ECONOMIC CONSEQUENCES OF CATASTROPHES TRIGGERED BY NATURAL HAZARDSrs471cr7450/TR143... ·...
Transcript of ECONOMIC CONSEQUENCES OF CATASTROPHES TRIGGERED BY NATURAL HAZARDSrs471cr7450/TR143... ·...
Department of Civil and Environmental Engineering
Stanford University
ECONOMIC CONSEQUENCES OF CATASTROPHES TRIGGERED BY NATURAL HAZARDS
by
T.L. Murlidharan
and
Haresh Shah
Report No. 143
March 2003
The John A. Blume Earthquake Engineering Center was established to promote research and education in earthquake engineering. Through its activities our understanding of earthquakes and their effects on mankind’s facilities and structures is improving. The Center conducts research, provides instruction, publishes reports and articles, conducts seminar and conferences, and provides financial support for students. The Center is named for Dr. John A. Blume, a well-known consulting engineer and Stanford alumnus. Address: The John A. Blume Earthquake Engineering Center Department of Civil and Environmental Engineering Stanford University Stanford CA 94305-4020 (650) 723-4150 (650) 725-9755 (fax) earthquake @ce. Stanford.edu http://blume.stanford.edu
©2003 The John A. Blume Earthquake Engineering Center
ECONOMIC CONSEQUENCES OF CATASTROPHES
TRIGGERED BY NATURAL HAZARDS
A DISSERTATION SUBMITTED TO THE
DEPARTMENT OF CIVIL AND ENVIRONMENT ENGINEERING
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
T. L. Murlidharan
March 2003
Copyright by T. L. Murlidharan 2003 All Rights Reserved
i Abstract
Abstract ________________________________________________________________________
Several important questions related to catastrophes are addressed in this
dissertation. How closely are catastrophes and developmental process related? How is the
post event economic growth related to the losses from catastrophic events? How
important and how long lasting are the various effects likely to be? What is the record of
past catastrophes and what regularities can be inferred from them? Can theoretical
models explain some of these regularities? How will a regional economy behave after a
catastrophic event? How is the effect of a catastrophe propagated to an interacting
region? What pre-event conditions are crucial in explaining the fact that some economies
do better after an event? Can the models explain other post-event behaviors too?
The purpose of this dissertation is twofold. On the one hand, it seeks to detect empirical
regularities in the behavior of economies affected by catastrophes. On the other hand, it
develops various models to study the effect of catastrophes on an economy, which
explain some of the empirical regularities.
Appropriate economic models are developed to explain the observed phenomena.
Theoretical simulations start by perturbing the Ramsey’s model to study the effect of
sudden changes in capital and the post event changes in the productivity. Two extensions
of this model are examined. The first of these studies the effect of efficiency of post-
event reconstruction on subsequent behavior. The second extension studies the effect of a
catastrophe on interacting economies. The behavior of the models from numerical
simulations is corroborated with empirical regression results.
A cross-country study with data from countries from various income groups affected by
different types of natural hazards (earthquakes, floods, hurricanes, and droughts) is
presented. Results based on an econometric model imply that direct losses as a result of
catastrophes are negatively correlated with the post event growth. It is seen that only by
modeling the fact that after a catastrophe reconstructed capital takes time to become
productive can one explain the negative correlation of the direct loss with post-event
ii Abstract
growth. Evidence point to the fact that catastrophes increase the external debt, budget
deficit and inflation. However, these effects are only temporary. Two years after the
event the effect of the catastrophe on the economic growth is statistically insignificant.
A standard regional economic model was used to simulate post event economic behavior
for three historical events - the 1989 Loma Prieta earthquake, 1992 Hurricane Andrew,
and 1994 Northridge earthquake. The model was then used to study the effects of
scenario earthquakes in the Bay Area and the Silicon Valley. The gross regional product,
consumption, investment, and local government spending show declines during the first
two years and then recover depending on the external aid and the efficiency of the
reconstruction process.
One of the main messages of this dissertation is that catastrophes cause myriad problems
in the short term after an event. But efficient reconstruction policies should help the
affected communities to emerge as less vulnerable, more productive and hence
economically stronger regions in the long run. To achieve this efficiency catastrophe
management has to be intimately linked with development policies.
Acknowledgments
iii
Acknowledgments ________________________________________________________________________ The research reported in this dissertation was partially supported by the Shah Family
Fellowship and by the Stanford University Department of Civil Engineering. The author
would like to place on record his gratitude to the many individuals who helped bring the
project to fruition. Professor James Sweeney and Professor Edison Tse offered advice
and support consistently throughout the development of this work. Discussions with
Professor Charles Jones and Mr. Rishi Goyal were extremely useful.
Table of Contents iv
Table of Contents ________________________________________________________________________
ABSTRACT iv
ACKNOWLEDGMENTS vi
CONTENTS vii
LIST OF TABLES xii
LIST OF FIGURES xiv
CHAPTER 1. INTRODUCTION
1.1 CATASTROPHES AND DEVELOPMENT PROCESSES 2
1.2 THEORETICAL MODELS OF ECONOMIES AFFECTED BY CATASTROPHES 3
1.3 EMPIRICAL EVIDENCE 4
1.3.2 Data on Catastrophes 5 1.3.3 Theoretical Models and Evidence on Post-Event Economic Behavior 11
1.4 CATASTROPHES AND REGIONAL ECONOMIES 12
1.5 OUTLINE OF THE DISSERTATION 13
CHAPTER 2. CATASTROPHES AND DEVELOPMENT
2. INTRODUCTION 19
2.1 WHAT IS A CATASTROPHE? 21
2.2 DEVELOPMENT PROCESSES, VULNERABILITY AND CATASTROPHE 24
2.2.1 Macro-Level Determinants Of Vulnerability 29 2.2.1.1 Openness To World Economy 29 2.2.1.2 Development Induced Investment In Large-Scale Projects 30 2.2.1.3 Development And Population Growth 33 2.2.1.4 Development And Urbanization 34 2.2.1.4 Development And Urbanization 35 2.2.1.5 Development And Poverty 36 2.2.1.6 Vulnerability As A “Phase Of Development” 38 2.2.1.7 Development And Government 39 2.2.2 Micro-Level Determinants Of Vulnerability 42
Table of Contents v
2.2.2.1 Presence Of Uncertainties And Development Processes 43 2.2.3 Development, Households And Vulnerability 45 2.2.3.1 Health, Nutrition, And Education 47
2.3 HOW DO CATASTROPHES AFFECT DEVELOPMENT? 52
2.3.1 Macro – Level Effects 53 2.3.1.1 Effects On Development 53 2.3.1.2 Effect On Trade And Investment 54 2.3.2 Effects At A Household Level 57 2.3.2.1 Savings And Investment 57 2.3.2.2 Identifying Transitory Income 59 2.3.2.3 Risk Pooling And Consumption Smoothing 60 2.3.2.4 Effect On Human Capital Investments 61
2.4 METHODS OF COPING- RISK, INSURANCE, CREDIT AND SAVING 63
2.4.1 Households, Groups, Community, Villages 65 2.4.2 Insurance, Savings And Credit 66 2.4.3 Credit, Insurance And Long-Run Development And Growth 68
2.5 AID AND RECOVERY 70
2.5.1 Disaster Aid At The Macro Level 71 2.5.2 Disaster Aid At The Micro Level 73
2.6 POLICY ISSUES 74
2.7 SUMMARY 76
CHAPTER 3: SHORT TERM ANALYSIS USING THEORETICAL MODELS
3.1 INTRODUCTION 78
3.2 MODELING A CATASTROPHE 80 3.2.1 Impact On Consumption And Investment 83 3.2.2 Impact On Welfare 86 3.2.4 Numerical Experiments 87
3.3 MODEL INCLUDING THE EFFECTS OF EFFICIENCY OF POST-EVENT RECONSTRUCTION 98
3.3.1 Model 98 3.3.2 Numerical Experiments 101 3.4 MODEL FOR REGIONAL EFFECTS
3.4.1 Numerical Experiments 114
3.5 CONCLUSIONS 128
Table of Contents vi
CHAPTER 4 EMPIRICAL ANALYSIS 4. INTRODUCTION 131 4.1 PREVIOUS STUDIES 133 4.2.1 Change In Indicators Due To Catastrophes 138 4.3 GENERAL FRAMEWORK AND ECONOMETRIC MODEL 141 4.3.1 Approximation 142 4.3.2 Summary Statistics And Discussion Of The Sample 143 4.3.2.1 Economic Growth 143 4.4.2.2 Effect On Consumption, Investment, Government Expenditure, Net Exports And
Income 145 4.4 EFFECT ON THE ECONOMIC GROWTH 150 4.4.1 Primary Variables 150 4.4.1.1 Direct Physical Loss 150 4.4.1.2 Percentage Affected 151 4.4.1.3 Type Of Hazard 152 4.4.2 Control Variables 153 4.4.2.1 Pre-Existing Economic Conditions 153 4.4.2.2 Health 154 4.4.2.3 Poverty And Inequality 154 4.4.2.4 Government, Bureaucracy, And Institutions 156 4.4.2.5 Infrastructure 158 4.4.2.6 Education 159 4.4.2.7 Trade 160 4.5 INTRODUCTION TO ECONOMETRIC ISSUES 161 4.6 PROBLEMS WITH THE DATA 163 4.7 LIMITATIONS OF CROSS-COUNTRY REGRESSION STUDIES 164 4.8 RESULTS FROM REGRESSION ANALYSIS 165 4.8.1 Growth Rates – Short Term 165 4.8.2 Growth Rates – Average 168 4.9 EFFECT ON MAJOR ECONOMIC INDICATORS 171 4.9.1 Consumption 171 4.9.2 Investment 171 4.9.3 Government Expenditure 171
Table of Contents vii
4.9.4 Inflation, And Interest Rates 173 4.11 CONSUMPTION SMOOTHING AND SAVINGS BEHAVIOR 179 4.12 CONCLUSIONS, EXTENSIONS, AND LIMITATIONS 181
CHAPTER 5 REGIONAL IMPACT OF CATASTROPHES
5. INTRODUCTION
5.1 METHODOLOGIES USED TO STUDY REGIONAL IMPACTS 188
5.2 MODELING PROBLEMS 190
5.4 DESCRIPTION OF EVENTS 191
5.4.1 Loma Prieta Earthquake 191 5.4.2 Hurricane Andrew 192 5.4.3 Northridge Earthquake 193
5.5 A COMPARISON OF THE IMPACTS OF THE EVENTS 193
5.5.1 Effects On The Components Of Personal Income 198 5.5.2 Effects On The Components Of Net Earnings By Place Of Work 198 5.5.3 Dampening Out Effect 201
5.6 SIMULATION OF THE EFFECTS WITH THE REGIONAL MODEL 203
5.6.1 LOMA PRIETA Earthquake 204 5.6.2 Hurricane ANDREW 206 5.6.3 NORTHRIDGE EARTHQUAKE 214 5.7 SIMULATION OF IMPACT OF PROBABLE EARTHQUAKE
SCENARIOS IN THE BAY AREA 218
5.8 MODEL BEHAVIOR WHEN CRUCIAL PARAMETERS ARE VARIED 222
5.8.1 Transfer Payments Effects (Fig. 5.12) 222 5.8.2 Consumer Spending Effects (Fig. 5.13) 222 5.8.3 Government Spending Effects (Fig. 5.14) 224 5.8.4 Labor Supply Effects (Fig. 5.15) 224 5.8.5 Migration Effects (Fig. 5.16) 226 5.8.6 Production Or Fuel Costs (Fig. 5.17) 227 5.8.7 Business Taxes And Credits (Fig. 5.18) 227 5.8.8 Consumer Prices (Fig. 5.19) 228
Table of Contents viii
5.9 SUMMARY AND CONCLUSIONS 230 CHAPTER 6 CONCLUSIONS AND FUTURE WORK
6. INTRODUCTION 231 6.1 CONCLUSIONS 231 6.2 FUTURE WORK 234 REFERENCES 236 APPENDICES CDROM Appendix A - Loss and economic data
Appendix B - Determinants of vulnerability Appendix C - Mathematica© Code for Simulation of Perturbed Ramsey’s Model
Appendix D - Mathematica© Code for Simulation of Model Including Effects of Reconstruction
Appendix E - Mathematica© Code for Simulation of Interacting Regions Model Appendix F - Details of regression for discerning the effect on economic growth Appendix G - Details of regression for discerning the effect on consumption,
investment, government expenditure, and net exports using Penn World Tables
Appendix H - Details of regression for discerning the effect on real interest rate Appendix I - Details of regression for discerning the effect on other indicators
including inflation
List of Figures
viii
List of Tables ________________________________________________________________________
Table
4.1a Disasters in the Caribbean can have significant impact on GDP and growth
(World Disasters Report, 1997)
4.1b Description of variables and their data sources
4.2 Summary statistics for short-term growth
4.3 Summary statistics for average growth
4.4 Summary statistics for external debt
4.5 Summary statistics for budget deficit
4.6 Summary statistics for resource balance
4.7 Specifications and regression analysis describing the effect of catastrophes on
short-term economic growth
4.8 Specifications and regression analysis describing the effect of catastrophes on
average economic growth
4.9 Specifications and regression analysis describing the effect of catastrophes on
external debt
4.10 Specifications and regression analysis describing the effect of catastrophes on
resource balance
4.11 Specifications and regression analysis describing the effect of catastrophes on
budget deficit
4.12 Summary statistics for income, consumption, and savings in the five years
enveloping the disaster year
4.13 Summary statistics for percentage growth rates for income, consumption, and
savings in the five years enveloping the disaster year
4.14 Estimates of consumption changes using lagged changes in income
4.15 Estimates for income changes using lagged savings
4.16 Estimates for consumption changes using lagged savings
4.17 Catastrophic events, associated direct losses, and percent population affected
List of Figures
ix
5.1 Observations on the effects of the Loma Prieta earthquake, Northridge
earthquake, and Hurricane Andrew
5.2 Effect on county’s components of personal income
5.3 Growth rates of the components of net earnings
5.4 Effect on county’s components of net earnings by place of work
5.5 Main economic indicators before the events
5.6 Comparison of model predictions with observed values – Loma Prieta
Earthquake
5.7 Comparison of model predictions with observed values – Hurricane Andrew
5.8 Comparison of model predictions with observed values – Northridge
Earthquake
5.9 Earthquake scenarios and assumptions about regional capacity
List of Figures
x
List of Figures ________________________________________________________________________
Figure
1.1 Overall layout of the thesis
2.1 Determinants of Macro- and Micro- Vulnerability
3.1a Assumed changes in the capital share in the production function
3.1b Changes in initial consumption for various productivity levels
3.1c Evolution of consumption with various levels of productivity using Ramsey’s
model
3.1d Evolution of capital with various levels of productivity using Ramsey’s model
3.1e Evolution of output with various levels of productivity using Ramsey’s model
3.1f Growth of the economy with various levels of productivity using Ramsey’s
model
3.1g Phase space plot of consumption and capital with various levels of
productivity evolution using Ramsey’s model
3.2a Assumed changes in the capital share in the production function
3.2b Changes in initial consumption for various levels of productivity
3.2c Evolution of consumption with various levels of capital loss using Ramsey’s
model
3.2d Evolution of capital with various levels of capital loss using Ramsey’s model
3.2e Evolution of output with various levels of capital loss using Ramsey’s model
3.2f Growth of the economy with various levels of capital loss using Ramsey’s
model
3.2g Phase space plot of consumption and capital with various levels of capital loss
evolution using Ramsey’s model
3.2h Changes in the overall welfare for various levels of capital loss
3.3a Assumed changes in the conversion of maturing capital to productive capital
List of Figures
xi
3.3b Assumed changes in the capital share in the production function for extended
model and the evolution of external aid
3.3c Change in maturing capital with various rates of conversion from maturing
capital to productive capital using extended model
3.3d Change in productive capital with various rates of conversion from maturing
capital to productive capital using extended model
3.3e Change in consumption with various rates of conversion from maturing
capital to productive capital using extended model
3.3f 3D Phase space plot of consumption, maturing capital and productive capital
evolution using extended model
3.3g Change in output with various rates of conversion from maturing capital to
productive capital using extended model
3.3h Growth of the economy with various rates of conversion from maturing
capital to productive capital using extended model
3.4a Assumed changes in the conversion of maturing capital to productive capital
3.4b Assumed changes in the capital share in the production function for extended
model and the evolution of external aid
3.4c Change in maturing capital with various levels of loss using extended model
3.4d Change in productive capital with various levels of loss using extended model
3.4e Change in consumption various levels of loss using extended model
3.4f 3D Phase space plot of consumption, maturing capital and productive capital
evolution using extended model
3.4g Change in output with various levels of loss using extended model
3.4h Growth of the economy with various levels of loss using extended model
3.4i Initial changes in consumption with various levels of loss using extended
model
3.4j Overall welfare changes due to various levels of loss of loss using extended
models
3.5a Change in consumption in affected region with various levels of aid using
regional model
List of Figures
xii
3.5b Change in consumption in unaffected region with various levels of aid using
regional model
3.5c Change in capital in affected region with various levels of aid using regional
model
3.5d Change in capital in unaffected region with various levels of using aid using
regional model
3.5e Phase space plot of consumption and capital for the affected region with
various levels of aid using regional model
3.5f Phase space plot of consumption and capital for the unaffected region with
various levels of aid using regional model
3.5g Change in output in affected region with various levels of aid using regional
model
3.5h Change in growth of output in affected region with various levels of aid using
regional model
3.5i Change in output in unaffected region with various levels of using aid using
regional model
3.5j Change in growth of output in unaffected region with various levels of aid
using regional model
3.6a Change in consumption in affected region with various levels of loss using
regional model
3.6b Change in consumption in unaffected region with various levels of loss using
regional model
3.6c Change in capital in affected region with various levels of loss using regional
model
3.6d Change in capital in unaffected region with various levels of using loss using
regional model
3.6e Phase space plot of consumption and capital for the affected region with
various levels of loss using regional model
3.6f Phase space plot of consumption and capital for the unaffected region with
various levels of loss using regional model
List of Figures
xiii
3.6g Change in output in affected region with various levels of loss using regional
model
3.6h Change in growth of output in affected region with various levels of loss using
regional model
3.6i Change in output in unaffected region with various levels of using loss using
regional model
3.6j Change in growth of output in unaffected region with various levels of loss
using regional model
3.6k Change in overall welfare in affected region with various levels of loss using
regional model
3.6l Change in overall welfare in unaffected region with various levels of loss
using regional model
4.1 Comparison of pre- and post-event growth rates (short term)
4.2 Comparison of average pre- and post-event growth rates
4.3a Effect on short term external debt
4.3b Growth of external debt
4.3c Effect on average external debt
4.4 Effect on budget deficit
4.5 Effect on resource balance
5.1 Effect of Loma Prieta Earthquake on personal income of San Francisco-San
Jose CMSA
5.2a Effect on gross regional product (Loma Prieta)
5.2b Effect on consumption (Loma Prieta)
5.2c Effect on capital stock (Loma Prieta)
5.2d Effect on employment (Loma Prieta)
5.2e Effect on consumption deflator (Loma Prieta)
5.2f Effect on government spending (Loma Prieta)
5.3 Effect of Hurricane Andrew on personal income of Dade county
5.4a Effect on gross regional product (Andrew)
5.4b Effect on consumption (Andrew)
List of Figures
xiv
5.4c Effect on capital stock (Andrew)
5.4d Effect on employment (Andrew)
5.4e Effect on price index (Andrew)
5.4f Effect on government spending (Andrew)
5.5 Effect of Northridge Earthquake on personal income of Los Angeles County
5.6a Effect on gross regional product (Northridge)
5.6b Effect on consumption (Northridge)
5.6c Effect on capital stock (Northridge)
5.6d Effect on employment (Northridge)
5.6e Effect on government spending (Northridge)
5.6f Effect on consumption deflator (Northridge)
5.7 Effect of future scenario earthquakes on gross regional product of San
Francisco – San Jose CMSA
5.8 Effect of future scenario earthquakes on personal income of San Francisco –
San Jose CMSA without aid
5.9 Effect of future scenario earthquakes on personal income of San Francisco –
San Jose CMSA with reconstruction aid
5.10 Effect on gross regional product – probable scenarios with no external aid
5.11 Effect on gross regional product – probable scenarios with ant without aid
assuming 10% capital loss
5.12 Effect on gross regional product – probable scenarios with ant without aid
assuming 20% capital loss
5.13 Effect on gross regional product – probable scenarios with ant without aid
assuming 28% capital loss
5.14 Effect on gross regional product – probable scenarios with ant without aid
assuming 35% capital loss
5.15 Effect on gross regional product – transfer payments 10% of loss
5.16 Effect on gross regional product – 1% increase in consumption spending
5.17 Effect on gross regional product – 10% increase in government spending
List of Figures
xv
5.18 Effect on gross regional product – 10% decrease in occupational employment
5.19 Effect on gross regional product – 1% increase in migration of population
5.20 Effect on gross regional product – Increase in relative production costs by
10%
5.21 Effect on gross regional product – 10% decrease in business taxes or 10%
increase in tax credits
5.22 Effect on gross regional product – 10% increase in consumer prices
5.23 Effect on gross regional product – 1% decrease in wage rate
5.24 Effect on gross regional product – 1% decrease in housing prices
5.25 Simulation of personal income of Santa Clara County 1989-1997
5.26 Effect of scenario earthquake on gross regional product of Silicon Valley
5.27 Effect of scenario earthquake on personal income of Silicon Valley
Chapter One: Introduction
1
Chapter One
Introduction ________________________________________________________________________ 1. Introduction What are the effects of a catastrophe on the macro-economic processes? How closely are
catastrophes and developmental process related? What are the socioeconomic
determinants of vulnerability of countries to catastrophes? Do catastrophes actually retard
economic growth? How important and how long lasting are the various effects likely to
be? What trends do past data on catastrophes suggest and can theoretical models replicate
these trends? How will a regional economy behave after a catastrophic event? What
measures will best help the affected community to recover? This dissertation is an
attempt to answer these questions. The purpose of the dissertation is to detect empirical
regularities in the behavior of economies affected by catastrophes and to develop models
to study the effect of catastrophes on a typical economy, which explain some of the
empirical regularities.
A direct tangible consequence of a catastrophe is the huge economic loss, often in the
order of billions of dollars, suffered by the affected region. The vulnerability of a
community to natural hazards depends on various socioeconomic conditions. In addition,
a catastrophe disrupts and changes the complex web of interactions between ongoing
economic, social, and political processes. An intriguing question is whether the
development of the economy of the region is significantly altered by the occurrence of a
catastrophe.
1.1 Catastrophes and Development Processes Catastrophes are not caused by the extremes of nature alone. A catastrophe is
fundamentally a social phenomenon; it involves the intersection of the physical processes
of a hazard agent with the various on-going economic, social, and political processes. For
large segments of the world's underdeveloped population, occurrence of a natural hazard
may worsen an already deteriorating or fragile situation. In order to study the effect of a
Chapter One: Introduction
2
catastrophe on an economy the factors that describe socioeconomic conditions prior to
occurrence of the hazard event have to be identified. Socioeconomic conditions in a
region are mainly as a result of the developmental processes. The effect of a catastrophe
on the developmental process is complex, especially for developing regions.
Socioeconomic processes including development affect the vulnerability of a community
to natural hazards in subtle ways. The determinants of vulnerability are qualitatively
derived based on empirical observations connecting socio-economic indicators to the
observed losses both at macro - (community) and micro - (household) levels. Socio-
economic indicators that are used include per-capita income, population growth,
education, infrastructure, and quality of governance. This study attempts to bring together
the data and results in disparate fields of study such as growth and development
economics, and sociology of disasters to provide a firm foundation for the connections
between catastrophes and socio-economic processes. The complex ways in which the
occurrence of catastrophe affects the development process is then explained. Methods of
coping with the economic effects of catastrophes are examined. Some implications for
policy design are mentioned.
1.2 Theoretical Models of Economies Affected by Catastrophes
The evolution of an economy after a catastrophic event is investigated in order to analyze
the dynamic effects of a catastrophic event that destroys substantial capital stock.
Ramsey’s model and its extensions are used to address various aspects of the problem.
For these three models, a catastrophe, due to the occurrence of an earthquake or a
hurricane, is modeled by a discontinuous change in the capital stock. The models
simulate the behavior of a typical economy when perturbed by an unanticipated and large
change in the capital stock followed by an arbitrarily complex change in the affected
region’s productivity. The results indicate the initial impact on investment, consumption,
and production.
Chapter One: Introduction
3
The simulation results point to the importance of modeling the efficiency of the
reconstruction processes after an event. Unless the process whereby the maturing capital
is converted to productive capital is modeled, the fact that post-event growth rate is
negatively correlated with the magnitude of loss cannot be explained. Empirical evidence
presented in Chapter 4, based on data from 43 countries in which catastrophes have
occurred, strongly suggest that greater loss is associated with smaller post-event growth
rates. In addition, Chapter 4 presents evidence regarding an extensive set of pre-event
conditions that are important in post-event recovery. These factors are indicators for the
changes in productivity and the conversion factor that are assumed in simulation models.
Models also indicate the fall in consumption levels after the event. Empirical evidence
indicates losses are negatively correlated with post-event income and consumption
changes are positively related to changes in income. Greater changes in productivity are
reflected in the post-event changes in output.
1.3 Empirical Evidence
Catastrophes triggered by natural hazards such as earthquakes, floods, storms, volcanoes
and droughts are a major global problem. Between 1968 and 1992 major disasters have
affected an average of 113 million people and killed 140 thousand annually (IFRCRCS
1994). More major natural disasters occurred in 1998 than in any other year on record
(MunichRe, 2000). The World Meteorological Organization (WMO) confirmed 1998 as
by far the warmest year since records began. Most catastrophes occur in poorer countries
of the Third World: some 97 percent of deaths and 99 percent of people affected between
1971 and 1995 were in least developed countries (LDCs) (Twigg 1997). Physical
destruction, in absolute terms and the economic consequences of disaster can be very
great. Environmental refugees account for some 58 per cent of all refugees worldwide
(IFRCRCS 1999: 20). Elo (1994) estimated that in 1992 alone the world economy lost
more money from catastrophes triggered by natural events in the LDCs (US $62 billion)
than it spent on development aid (US $60 billion). Three million people per year are
made homeless by flooding (IFRCRCS 1999: 11). During the 1980s the total economic
Chapter One: Introduction
4
losses from natural disasters exceeded US $10 billion (at 1990 prices) and the average
cost from a single major disaster is now probably about $500 million at mid-1990s prices
(Smith 1996). Catastrophes (post 1970) that have resulted in causing more than the
average cost of $500 million from a single disaster are chosen for the present study.
1.3.2 Data on Catastrophes
Data regarding major catastrophes that occurred around the world between 1971-
1998 is obtained from Center for Research on the Epidemiology of Disasters (Sapir and
Misson, 1992). This includes data on the type of the event, the time when an event
occurred, the place of occurrence, the approximate estimated direct losses, and the
number of people affected. Data regarding economic indicators such as per capita
income, gross domestic capital formation and its growth, gross domestic savings, the
resource balance, and government consumption and its growth are obtained from the
World Development Indicators (World Bank, 2001). Data on institutions, bureaucracy,
education, life expectancy, health, infrastructure are obtained from the web download-
able databases maintained by Easterly and Levine (1997) and Barro and Lee (1995).
Table 1.1 lists the sources of data used for the present study. The complete data set is
provided in electronic form in Appendix A.
It should be noted here that the quality of data associated with catastrophes is not as good
as data on macro-economic indicators. Only in the recent past have efforts been made to
document data from disasters like the EM-DAT database from CRED (Sapir and Misson,
1992). Recognizing that most reporting sources have vested interests and figures may be
affected by sociopolitical considerations, CRED manages conflicts in information by
giving priority to data from governments of affected countries, followed by UNDHA, and
then the US Office of Foreign Disaster Assistance. Agreement between any two of these
sources takes precedence over the third.
Catastrophes are relatively rare events in any given country by definition. In order to
obtain a broad understanding of the effects of catastrophe in different countries, data has
Chapter One: Introduction
5
been compiled for catastrophes that have occurred in 43 countries. The World Bank
classifies countries according to the income per capita. The four main categories in which
the nations are classified are the high income (>$9266 GNP per capita), upper-middle
income ($2996 - $9265), lower-middle income ($755-$2995 GNP per capita), and low
income (<$755 GNP per capita). In the sample, there are thirteen countries belonging to
the high income group (Australia, Canada, Denmark, France, Greece, Italy, Japan, Korea,
Rep., Netherlands, Spain, Switzerland, United Kingdom, United States), five to the upper
middle income group (Argentina, Brazil, Chile, Mexico, South Africa), fifteen to lower
middle income group (Algeria, Colombia, Dominican Republic, Ecuador, El Salvador,
Guatemala, Indonesia, Iran, Islamic Rep., Jamaica, Mongolia, Peru, Philippines, Russian
Federation, Thailand, Turkey), and ten to lower income group (Bangladesh, Burkina
Faso, China, Honduras, India, Nepal, Nicaragua, Pakistan, Vietnam, Zimbabwe). A
distribution of per capita income of the countries in the sample at the time of occurrence
of a particular event is shown in Fig.1.1. The figure illustrates the fact that the sample
gives a sufficiently general representation of all income groups. Fig.1.2 gives the region-
wise classification of the loss ratios (defined as the total economic loss from all events in
a particular year for country as a proportion of its GDP). It is clear from the figure that
high loss ratios (greater than 1% of the GDP) are concentrated in developing regions of
the world whereas in the developed world majority of the loss ratios are below 1%.
Fig. 1.1 Distribution of per capita GDP in the event set
0%10%20%30%40%50%60%70%80%90%
100%
$100 $1,000 $10,000 $100,000
GDP per capita (USD current)
Chapter One: Introduction
6
Generality of the inferences depend on including different hazard types in the sample.
The present sample includes various types of natural hazards. It includes 24 earthquake
events, 62 floods, 57 hurricanes or cyclones or typhoons or storms, 20 droughts, and 6
other events such as bush fires, volcanoes, and landslides. It is important to find out
whether the type of hazard has a significant effect on the nature of macro-economic
changes.
The sample has concentrated on post-1970 catastrophic events. This is because no cross-
country study of the macro-economic effects for those events has been attempted. The
study by Albala-Bertrand (1993a, b) concentrates on a few events that occurred in the
1970s. Fig 1.3 is a plot showing the number of events in a year in the sample that caused
more than US $500 million (current) in losses. From the figure it is clear that there is an
increasing trend of such events worldwide. However, if the loss-GDP ratios are graphed
as shown in Fig. 1.4, there is no trend. The loss ratio is therefore an appropriate
normalized indicator of the catastrophe magnitude that can be used to study the
phenomenon over time. By normalizing the loss with its current year GDP, the issues
related to the present values of the past losses has been addressed. Comparison of a loss
of $500 million in 1970 with the same loss in 1990 would have otherwise been
problematic. A similar reasoning is applied to the proportion of a population of a country
affected by a catastrophe since Fig. 1.5 does not show a trend in time. These two
measures of catastrophe, namely economic loss ratio and the percentage of people
affected, are inter-related as Fig. 1.6 illustrates. Higher loss ratios are strongly associated
with more number of people being affected (as a proportion of the total population).
The data on socio-economic indicators are compiled based on the events. For example,
for the 1987 earthquake in Ecuador, all the relevant indicators such as the GDP per capita
are collected for ten years before the event date. The average of this data is used in the
study on the determinants of vulnerability in Chapter 2. Indicators are also compiled three
years around the event year i.e. three years before and after the event. This is used in the
study on the economic consequences of a catastrophe in Chapter 4. In the following two
Chapter One: Introduction
7
sections, a summary of the results, which will be discussed in detail in the subsequent
chapters of this dissertation, will be presented.
Fig. 1.3 Number of events causing more than US$ 0.5 billion economic loss are increasing
R2 = 0.53
0
2
4
6
8
10
12
14
16
1970 1975 1980 1985 1990 1995 2000
Year
Num
ber o
f eve
nts
Fig. 1.2 Regionwise distribution of the loss to GDP ratios
0.0%
0.1%
1.0%
10.0%
100.0%
Category of events that occurred in a year (1970-'98) in a nation
Loss
as
a pr
opor
tion
of G
DP
Chapter One: Introduction
8
Fig 1.4 Loss ratios do not exhibit any trend with time
-4.0-3.5
-3.0-2.5-2.0
-1.5-1.0-0.5
0.00.5
1970 1975 1980 1985 1990 1995 2000
Year
Loss
rat
io (l
og s
cale
)
Fig. 1.5 Population affected do not exhibit any trend with time
0.00
0.05
0.10
0.15
0.20
0.25
0.30
1970 1975 1980 1985 1990 1995 2000
Year
Popu
latio
n af
fect
ed (a
s a
ratio
of
tota
l pop
ulat
ion
log
scal
e)
Chapter One: Introduction
9
Fig. 1.6 Smaller economic loss is associated with smaller percentage of population affected
R2 = 0.30, N= 118
0.000%
0.000%
0.000%
0.001%
0.010%
0.100%
1.000%
10.000%
100.000%
0.0% 0.1% 1.0% 10.0% 100.0% 1000.0%
Economic loss as a percentage of GDP
Perc
enta
ge o
f pop
ulat
ion
affe
cted
(log
log
scal
e)
Chapter One: Introduction
10
1.3.3 Theoretical models and evidence on post-event economic behavior
Chapter 3 develops dynamic models that simulate the effect of catastrophes. The results
of model simulation suggest various hypothesis that are tested using cross-country study
data from 43 countries from all income groups affected by different types of natural
hazards (earthquakes, floods, hurricanes, and droughts). These included more than 155
events (in some countries, more than one event may have occurred in a year).
Based on an econometric model, detailed in Chapter 4, statistical regularities are inferred
that corroborated the theory-generated hypothesis. Important inferences from the
empirical and theoretical studies are summarized below.
The magnitude of economic loss (as a proportion of GDP) is:
• negatively correlated with the post-event annual percentage economic growth,
• negatively associated with the post-event income.
• associated with increase in inflation and the real interest rates.
• negatively associated with the post-event income and this results in changes in
consumption.
• associated with changes in ex-ante saving behavior at least temporarily after the
event.
A simple cause-effect relation cannot explain the interaction between the occurrence of a
catastrophic event and its impact. The empirical results enumerated above generally
imply that catastrophes retard economic growth and savings, and increase the real interest
rates, inflation, and government spending. However, these effects are only temporary,
since two to three years after the event the effect of the catastrophe on the economic
indicators is statistically insignificant. Nations and regions affected by catastrophe start
rebuilding immediately after the event. However, the recovery process may be complex.
The pre-event socioeconomic conditions to a large extent determine the magnitude of
Chapter One: Introduction
11
impact and the ‘coping’ strategy of the affected community. The strategies adopted for
coping, in-turn, determine the post-event socioeconomic conditions.
1.4 Catastrophes and Regional Economies Having discerned patterns of economic slow down immediately after the event followed
by growth, a regional economic model is used to explain the post event behavior of an
affected region, in Chapter 5. A standard regional economic model was used to simulate
three historical events. The three events were the 1989 Loma Prieta earthquake, 1992
Hurricane Andrew, and 1994 Northridge Earthquake. Actual observed personal incomes
of the affected counties, as reported by Bureau of Economic Analysis, were compared
with the model generated personal incomes for validation. The model performed well
with a mean absolute percentage error not exceeding 3%.
The model was then used to simulate the effects of hypothetical scenario earthquakes in
the Bay Area (comprising of eleven counties of the San Francisco – San Jose – Oakland
Combined Metropolitan Statistical Area) that might have occurred in the year 2000.
Various direct loss and job loss levels were studied. Simulation results indicate, that for a
$30 billion capital loss and 25,000 - job loss scenario, the Bay Area’s gross regional
product would be down by 14% without any reconstruction and aid (worst case scenario)
during the year of the event. With minimal aid and reconstruction assumptions, the gross
regional product will be lower by 7% and would have totally recovered by the year 2002.
Consumption, investment, and local government spending would show declines during
the first two years and then rapidly grow as the economy recovers. These simulation
results concurred with the simulation of the theoretical model of interacting regions. Two
policy alternatives were simulated – business credits for new investment after the event
and increased local governmental spending. It was concluded that business credits
resulted in assisting rapid recovery after a catastrophe.
One of the main messages of this dissertation is that catastrophes cause myriad problems
in the short run, including slower growth, lower levels of savings and increases in
consumption. But efficient reconstruction policies result in better production techniques
Chapter One: Introduction
12
for the affected communities. This results in the affected communities to emerge as less
vulnerable and economically stronger regions in the long run. Reconstruction policies
have an important role to play but these will again depend upon extant socioeconomic
factors such as preparedness. To achieve this efficiency catastrophe management has to
be intimately linked with development policies.
1.5 Outline of the Dissertation Fig. 1.7 provides an overview of the layout of the dissertation (chapter and section
number appear on top right corner of the boxes). The diagram brings out the
interdependencies of the chapters. The chapter following this introduction provides a
comprehensive review of literature on economic effects of catastrophes. There are few
studies that address the issues related to catastrophes triggered by natural hazards.
Chapter 2 maps statistically significant associations between socio-economic indicators
and the loss-GDP ratios in the chosen disaster data set based on empirical observations.
These empirical regularities are used to relate literature from development economics and
disasters and to determine the indicators of vulnerability of a nation to natural hazards.
Chapter 3 presents theoretical dynamic economic models that simulate the effect of
catastrophes. Chapter 4 investigates evidence from the consequences of major disasters
that have occurred around the globe and relates it to the theoretical models presented in
Chapter 3. Chapter 5 focuses on the effects of catastrophes at a regional county level. The
simulation results from a standard regional economic model are compared with the
theoretical model presented in Chapter 3. Chapter 6 concludes with pointers towards need
for future research.
Chapter One: Introduction
13
Table 1.1 Data Description and Sources
Variables Description Source
Primary Loss Current US dollars CRED Number of people affected
Includes persons dead, injured, homeless or otherwise affected
CRED
Type of disaster and its year of occurrence
Earthquake, floods, storms, hurricanes, cyclones, drought, forest fires, and avalanches
CRED
Control Growth of GDP Average annual growth of real GDP per
capita 2001 WDI, World Bank
Institutions Bureaucratic quality 0-6 index Dates : 1982, 1990
PRS; ICRG
Government Rule of law 0-6 index Dates : 1982, 1990
Easterly Levine (1997)
Institutions Freedom from Corruption 1-7 index Dates : 1982, 1990
Easterly Levine (1997)
Government Repression of Civil Liberties 1-7 index Dates : 1980, 1990
Gastil (1990), Gastil (1987).
Literacy Percent of population literate Dates : 1980
Banks (1984)
Illitercy Percentage of "no schooling" in population
Barro and Lee (1993)
Enrollment Gross enrollment ratio for higher education
Barro and Lee (1993)
Enrollment Gross enrollment ratio for secondary education
Barro and Lee (1993)
Enrollment Gross enrollment ratio for primary education
Barro and Lee (1993)
Life expectancy at age zero
Life expectancy at age zero Barro and Lee (1993)
Health Daily calorie intake. Could be used as a measure of poverty.
World Bank's BESD database
Health Daily protein intake (grams). World Bank's BESD database
Health services availability
Number of hospital beds per thousand inhabitants Dates : 1980, 1990
World Bank's BESD database
Roadways Paved Roads/Highways Dates : 1980, 1990
2001 WDI, World Bank
Railroad Mileage per square mile 2001 WDI,
Chapter One: Introduction
14
Variables Description Source
Dates : 1980, 1990 Average
World Bank
Income Inequality
Gini coefficient for income that can range from a low of 0 to a high of 100
Deininger and Squire (1996)
Poorest Bottom quintile in income distribution 1980, 1990; Average
Deininger and Squire (1996)
Richest top quintile in income distribution 1980, 1990; Average
Deininger and Squire (1996)
Gender Inequality
Female to male average schooling years, age 26+ 1980, 1990;
Barro and Lee (1993)
Balance of Payments
Balance of Payments as a percentage of GDP
2001 WDI, World Bank
Debt Debt as a percentage of GDP 2001 WDI World Bank
Trade Trade as a percentage of GDP 2001 WDI World Bank
Money Average annual growth rate of the money supply during the last five years minus the potential growth rate of real GDP
Derived from 2001 WDI, World Bank
Inflation Standard deviation of the annual inflation rate during the last five years
Derived from 2001 WDI, World Bank
Consumption Household, government, as a percentage of GDP and their growth
2001 WDI, World Bank
Genuine savings Savings as a percentage of GDP 2001 WDI, World Bank
Interest Rates Real, nominal, interest rate spreads 2001 WDI, World Bank
Government Size
(Government consumption/GDP) 2001 WDI, World Bank
Takings Transfers and subsidies as a percent of GDP
Gwartney and Lawson, 1997
International exchange
Difference between the official exchange rate and the black market rate
Gwartney and Lawson, 1997
International exchange
Actual size of the trade sector compared to the expected size
Gwartney and Lawson, 1997
Chapter One: Introduction
15
Fig. 1.7 Layout of the thesis and interdependencies among the chapters
2.2How do developmentprocesses determine
vulnerability?
3.2Model 1
Ramsey's growthmodel
3.3Model 2
Model includingefficiency of reconstruction
3Theoretical models
Economic growth, outputand consumption
4.2-6Effect of direct loss on
economic indicatorsincluding growth
6Summary
ConclusionsFuture work
4Empirical Data
Macro-economicIndicators
2.3How do catastrophesaffect development
processes?
3.3Model 3
Interacting Regions
5.3-7Case Studies
Loma Prieta, Hurricane AndrewNorthridge Earthquake
5-5.2Simulation
results froma regional model
3 and 5Catastrophes and
interaction betweenregional economies
2.4Methods of
CopingAid and Recovery
2.1Catastrophe
DefinitionPerspectives
1Introduction
Chapter One: Introduction
16
Chapter Two: Catastrophes and Development
17
Chapter Two
Catastrophes and Development
________________________________________________________________________
2. Introduction
One of the objectives of this chapter is to show the subtle ways in which various
socioeconomic processes including development affect the vulnerability of a community
to natural hazards. Reversing the causal arrow, the complex ways in which the
occurrence of catastrophe affects the development process is also explained.
About 25 percent of world’s population lives in areas at risk from natural hazard. But the
most vulnerable people are the poorest. 40 of the 50 fastest-growing cities are in
earthquake zones (IFRCRCS 1999: 18). It has been estimated that the richest billion
people on the planet have an average income about 150 times that of the poorest billion
people, who have little choice but to locate in unsafe settings, whether these be urban
shanties or fragile rural environments. The Intergovernmental Panel on Climate Change
(IPCC) says that 60 per cent of the world’s population will be living in potential malarial
zones by 2100. There could be an extra 50 to 80 million cases of malaria and 3.5 million
cases of river blindness (IFRCRCS 1999: 14).
What determines vulnerability of communities to natural hazards? In LDCs broad and
complex socioeconomic problems combine with insecure physical environments to create
a high degree of vulnerability. For vulnerable people in the LDCs, access to resources at
either a household or individual level is most critical factor in achieving a secure
livelihood or recovering effectively from disaster (Blaikie et al. 1994). In addition, risk
varies according to occupation, social class, ethnicity, caste, age, and gender making
vulnerability determination a complex question.
Chapter Two: Catastrophes and Development
18
In developed countries even major events rarely cost more than 0.1 percent of GNP but
according to Zupka (1988), the negative impact on poor countries can be 20-30 times
greater. In the Commonwealth of Independent States natural hazards have regularly been
responsible for taxing the economy 3-4 times more than in the USA (Porfiriev, 1992).
Some countries have been highly vulnerable. For example, the GNP of the five countries
of the Central American Common Market was reduced by 2.3 percent between 1960 and
1974 as a result of disasters triggered by natural events (Smith 1996). Similarly, small
island countries in the Caribbean and the Pacific Oceans that depend on a narrow range
of primary products (the Dominican Republic in 1979, Haiti, Saint Lucia and Saint
Vincent in 1980, Fiji in 1993) have suffered damage from Hurricanes equivalent to 15
percent of their GNP. Smith (1996) reports that whilst the number of disasters claiming
at least 100 deaths has more than doubled in 30-year period from 1963-1992, disasters
creating economic damage equivalent to 1 percent or more of GNP have risen well over
four-fold.
There has been marked fall in the fatalities for some hazards in many of the wealthier
countries. But the world trend is towards more disaster-related deaths and damages
driven mainly by increased vulnerability in the LDCs. As of 1999, half the world’s
population lives in coastal zones. Ten million are at constant risk of coastal flooding
(IFRCRCS 1999: 11). One of the chief reasons for disproportionately large numbers of
deaths in case of sudden onset hazards like earthquakes or flash floods is that the poor
live in most vulnerable environments – in structures that are either non-engineered or
semi-engineered which might be located in low lying and vulnerable areas. The United
Nations estimates that 80 per cent of the world will live in developing countries by 2025,
more than half of which will be "highly vulnerable" to floods and storms (IFRCRCS
1999: Chapter 2). Though technology exists for constructing even non-engineered
structures to withstand moderate levels of hazards, this technology is not adopted by the
poorest. Uncertainty of threshold whereby a non-engineered structure becomes unsafe
and almost negligible probability of occurrence of a severe hazard is partly responsible
for this behavior.
Chapter Two: Catastrophes and Development
19
Many other equally important reasons can be elicited by investigating the structures of
vulnerability generated by ongoing socioeconomic processes. Smith (1996) cites several
reasons why disaster impact is growing, even if frequency of geophysical events is
unchanged and despite the many positive steps being taken to reduce disasters. The
reasons Smith cites are population growth, land pressure, urbanization, inequality,
climate change, political change, economic growth, technological innovation, social
expectations, and global interdependence. Using data from historical catastrophes, these
factors are shown to be determinants of vulnerability in this chapter.
The organization of this chapter is as follows. The next section describes various
perspectives about catastrophe. The macro- and micro-level determinants of vulnerability
are described next. The connections between development processes and vulnerability to
catastrophes are presented using arguments based on empirical regularities observed from
data on disasters and indicators of socio-economic processes. How do catastrophes affect
development? This question is examined in the Section 2.3. Results from literature are
reviewed. Theoretical and empirical results, from the research presented in later chapters,
are discussed. The mechanisms used for coping with risk, such as insurance, credit, and
saving, are then discussed. Section 2.5 examines how external aid helps in recovery. This
review chapter concludes with a discussion of policy issues.
2.1 What is a Catastrophe?
In this section various perspectives of catastrophes are briefly reviewed. This
becomes imperative for determining the parameters and issues that would be of relevance
to the discussion, to understand the reasons of emphasis placed by disaster researchers on
seemingly different issues, to design well-balanced policies, and if possible to gain a
holistic picture of disaster research.
Gilbert (1998) lists three main paradigms for studying disasters. In the first paradigm
catastrophes is imputed to an external agent. The affected human population is a passive
“victim of the environment”. An extreme version of this paradigm sees catastrophes as
Chapter Two: Catastrophes and Development
20
“acts of God”. The second paradigm views disaster as the result of underlying community
logic, of an inward social process. Catastrophes result from the interaction of physical
hazards with ongoing vulnerable socioeconomic processes. The third paradigm views
disaster as an entrance into state of uncertainty. In this paradigm catastrophe is tightly
tied into the impossibility of defining real or supposed dangers, especially after the
upsetting of the mental frameworks we use to know and understand reality.
Russell Dynes (1998) indicates - “a disaster is a normatively defined occasion in a
community in which extraordinary efforts are taken to protect and benefit some social
resource whose existence is perceived as threatened.” Robert Stallings (1998) points out -
“disasters are fundamentally disruptions of routines.” For Anthony Oliver-Smith (1998)
disaster is “a process/event involving the combination of a potentially destructive agent
from the natural, modified and/or constructed environment and a population in a social
and economically produced condition of vulnerability, resulting in a perceived disruption
of the customary relative satisfactions of individual and social needs for physical
survival, social order and meaning.”
Uriel Rosenthal’s (1998) re-conceptualization of the notion of sudden onset disasters is
interesting. A dam collapse is usually thought of as sudden onset catastrophe. But is it
really so? Its vulnerability is determined by the quality of construction, the politics, path,
and the kinds of channelization for drainage, among many other factors. There are both
structural and non-structural aspects mentioned here, each with different places in social
time, and virtually all taking place long before the dam failed. Catastrophes therefore
have complex and interrelated origins as well as consequences. What determines the
structure of vulnerability is as important as what determines the vulnerability of the
structure. We need to look no further than ongoing socioeconomic process to understand
the structures of vulnerability.
For the purposes of this study, a catastrophe is a low probability high consequence (in
terms of either lost lives or direct physical damage) economy wide event that acts as a
strain on the affected region’s resources and socioeconomic processes. As a result, low-
Chapter Two: Catastrophes and Development
21
income countries may be forced to either borrow or dis-save huge amounts to recover.
Considerable time may be required to bring the community to pre-disaster conditions.
Potential losses from disasters are usually classified as: (i) direct or capital (ii) indirect or
income, and (iii) negative secondary or output effects. The financial value of damage to
and loss of capital assets – constructed facilities including buildings, infrastructure,
industrial plants, and inventories of goods including crops, account for direct losses.
Direct losses also include loss of lives and include measures of the total number of
affected people who are rendered homeless.
Direct losses are usually the most readily assessed after a catastrophe has struck. In
economic terms direct losses can be equated with stock losses. It is important to
distinguish between financial estimates of loss from economic loss in case of constructed
facilities. The financial loss would involve the replacement value of the lost asset,
independent of its condition or age. Thus the replacement of a collapsed bridge as a result
of an earthquake would involve the replacement cost of a new bridge, independently of
the age or condition of the collapsed bridge. On the other hand, if the destroyed bridge
were near the end of its useful economic life, the economic cost of its destruction might
be very small if replacement would soon have been necessary. Of course, the loss of a
new bridge would impose far greater economic losses, which would approximate the
financial losses incurred.
Indirect losses arise from interrupted production and services, measured by loss of output
and earnings. For example, damage to roads and ports can hold up exports, imports, and
distribution of basic necessities affecting health and education, as well as other
productive sectors. Depending on the magnitude of the direct loss, the impacts of
disasters may or may not affect the country’s GDP. In some cases effects can spread
beyond national borders. For example the 1985 Mexico earthquake destroyed the central
telephone exchange. Many Central American countries were affected as their
transmission lines ran through Mexico City. Such indirect losses can be equated with
flow losses.
Chapter Two: Catastrophes and Development
22
Secondary effects of disasters are felt through longer-term impacts upon economic
performance including the development processes. Secondary effects are not easy to
estimate, which is reflected by the fact that not many research studies have concentrated
on this question. In the following sections we examine the complex ways in which
development processes and catastrophes are interrelated. In Chapter 3, various models are
presented that simulate the changes in productivity that arises after a catastrophe has
struck a region. The capital loss and the consequent changes in productivity due to
reconstruction result in permanent changes in overall welfare of the affected region.
Chapter 3 explains how this welfare loss can be used to quantify the secondary losses.
Development processes change the structures of vulnerability of a population.
Development decisions without adequately addressing the question of sustainability lead
to creation of inefficient facilities and services that contribute increasingly to disaster
impact (Kreimer and Munasinghe, 1991). One of the unintended consequences of the
development-induced change is that when a natural hazard strikes a highly vulnerable
region it may result in a catastrophe. The following questions are now examined:
• What are the socioeconomic determinants of vulnerability? How do development
processes increase vulnerabilities of some people to natural hazards? (Section 2.2)
• How does the occurrence of a catastrophe act adversely for the development process in
the affected region? (Section 2.3)
2.2 Development Processes, Vulnerability and Catastrophes
Vulnerability of a population to natural hazards can be summarized at both the
economy-wide (macro-) and household (micro-) levels. The main purpose of this section
is to examine the various factors that determine macro- and micro-level vulnerability and
their interdependencies. Fig. 2.1 gives an overall picture of these interactions. The
interactions are explained in the following.
Chapter Two: Catastrophes and Development
23
Chapter Two: Catastrophes and Development
24
Fig 2.1 Determinants of Macro- and Micro- Vulnerability
OverallVulnerability
Vagaries ofinternational
markets
Openess toworld economy
Low levelsof disasterawareness
and preparedness
Inadequatelabor supplyafter event
Low levels ofheath
Susceptible todisease after event
Health,Nutrition, and
Education
Poverty andpopulation
growth
Sloppyresidential
construction
HouseholdVulnerability
Poor construction quality
UrbanizationForces poor
to occupy vulnerableregions
Lack ofdisaster
recovery facilities
Poorinfrastructure
facilities
Large scaledevelopment
projects
Lack ofinsurance, credit
Inadequate savings
Fragilesocial
networks
Phase ofdevelopment
Chapter Two: Catastrophes and Development
25
Change, in the widest sense of the term includes development processes and it is hard to
separate them. Development, viewed positively, brings with it: (i) increases in per capita
output; (ii) a shift of labor out of agricultural sector and into relative security of
manufacturing and services; (iii) the integration of regional markets, assisted by
improved transportation and communication networks; (iv) increased trade with the
outside world; and (v) improvements in governments services aimed at alleviating or
mitigating poverty.
One of the main indicators of economic development is per capita income and an
important observation from the compiled loss data is illustrated in Fig 2.2. It is clear from
Fig.2.2 that higher loss ratios (annual economic loss/GDP) are associated with countries
with low per capita income. In this figure and for all the subsequent figures in this
chapter, the economic indicators are an average of ten years before the occurrence of an
event. The power-law relationship between the loss ratio and the per capita income
illustrated in Fig. 2.2 has clearly many applications, which will not be addressed in this
dissertation. What is relevant for the present purposes is that increases in per capita
income are associated with lower loss-GDP ratios. Per capita income thus constitutes a
significant indicator for vulnerability of a nation to natural hazards. Fig 2.3 brings out a
similar the relationship between the percentage of people affected and the per capita
income. As a consequence of low per capita income many people in most third world
countries that are vulnerable either lack preparedness measures, or the level of protection
is inadequate, or their livelihood level lacks resilience to economy wide catastrophes. It is
often the case that they are unable to provide themselves with self-protection, and the
state is unable or unwilling to offer much relevant social protection against economy
wide catastrophes. In developed industrialized countries, preparedness levels may be high
and in general livelihoods are more secure and insurance makes them more resilient.
Per capita income is certainly one of the indicators that determine vulnerability, but it
should be noted that development, in reality, is a very uneven affair, with some people
benefiting, often at the expense of others. Distributional issues and equity are major
Chapter Two: Catastrophes and Development
26
problems that many regions have yet to deal with satisfactorily. It is therefore not
surprising to note that development may increase vulnerability of some to natural
hazards. A highway construction project certainly brings about changes to the community
it serves. The highway presumably develops trade between the various regions it passes
through. But it also increases the probability that skilled labor from less developed
regions migrates to more developed regions. If the regions are rich in natural resources,
then the exposure of these natural assets to exploitation is increased. Exploitation is one
of the more common unintended consequences of the highway project. Another example
is the construction of high-rise buildings in earthquake zones without using earthquake-
resistant techniques or building on flood plains. More generally, development projects
result in changing the vulnerabilities of population to natural hazards. Besides per capita
income, other indicators of the development processes that are determinants of
vulnerability as discussed in the following sections.
Analysis of various political economies and the way they structure societies such that
similar hazards lead to very different impacts on one society compared to another is
required to unravel the exact nature of the phenomena. In United States, the vulnerability
of people to hurricanes is much less than in Bangladesh (or the countries of the
Caribbean) because of generally higher levels of income (which enable recovery more
easily), and the high degree of preparedness. The socioeconomic framework of self-
protection and social protection has reduced vulnerability to natural hazards of many. But
there exist sizable groups even in the wealthiest parts of the world that are still
vulnerable. Class, gender, race and ethnicity are likely to be very significant indicators of
the variable impact of hazards. For instance in the US not everybody enjoys social
protection (preparedness and mitigation measures) against hurricanes or earthquakes.
Hurricane Andrew affected the Black community to the significantly greater level than
others (Morrow 1997). Black and non-Cuban Hispanic households, across income levels,
were much more likely than Anglo and Cuban households to report insufficient
settlements and this was in part due to differential access to policies with larger
corporations.
Chapter Two: Catastrophes and Development
27
Fig. 2.2 Greater per capita income is associated with smaller annual economic loss as a proportion of GDP
Loss/GDP = 2.3614*(GDP/capita)-0.7111
R2 = 0.4143
0.01%
0.10%
1.00%
10.00%
100.00%
1000.00%
100 1,000 10,000 100,000
Per capita income (current US$)
Ann
ual e
cono
mic
loss
as
a %
of
GD
P
Fig. 2.3 Greater per capita income is associated with smaller population affected
R2 = 0.20, N= 117
0.0000%
0.0001%
0.0010%
0.0100%
0.1000%
1.0000%
10.0000%
100.0000%
100 1,000 10,000 100,000
Per capita income (current US$)
Perc
enta
ge o
f pop
ulat
ion
affe
cted
(log
log
scal
e)
Chapter Two: Catastrophes and Development
28
Generally speaking, vulnerability to a catastrophe is the result of people’s positions
within various political, social and economic fields, and the manner in which various
institutions in these fields respond to hazard in terms of awareness, emergency and crisis
management, and reconstruction. The reasons why catastrophes happen can be inferred
only by taking an objective and holistic view of the determinants of vulnerabilities and
damage levels in economies under stress. In the following sub-sections the determinants
of vulnerability will be enumerated first at a macro-level and then at a micro- or
household level. Links will be established between these macro- and micro parameters.
Data used in the subsequent regressions are presented in electronic form in Appendix B.
2.2.1 Macro-level determinants of Vulnerability
The determinants of vulnerability to natural hazards at a macro-level are: (i)
openness to the world economy, (ii) large-scale investment in development projects, (iii)
stability of the monetary system, (iv) urbanization, (v) population growth, (vi) poverty,
and the phase of development.
2.2.1.1 Openness to world economy
Increasing globalisation of the economy as a consequence of development implies
that regions, nations, and sub-national regions are directly connected to the rest of the
world. Whilst this globalisation brings with it unprecedented access to global trade and
resources, the risks from a natural hazard have also gained new transmission channels.
The 1995 Hanshin-Awaji earthquake forced the closure of the major Kobe port. As a
result many firms in US that rely on imports for manufacturing suffered due to delay in
shipping. For the affected region, though, globalisation provides a means to recover from
the disaster.
Chapter Two: Catastrophes and Development
29
The pattern of financial relationship between the industrialized North and the Third
World has altered with decolonization. The Third World has traditionally depended on
agricultural and mineral exports the prices of which are falling. Simultaneously, prices of
imported energy and technology have increased. Most of the Third World countries have
little opportunity to process and market what they produce and are dependent on imports
from industrialized nations which are often highly priced or tied to aid packages. This has
created circumstances in which many Third World nations are faced with great difficulty
in maintaining their balance of payments. Therefore, a viewpoint often expressed is that
the functioning of world economy is against the LDCs thereby reinforcing hazard
vulnerability. Current account balance is an indicator of the level to which a country is
dependent on imports. Fig 2.4 illustrates a clear negative relationship between
dependence on imports and the level of economic losses from catastrophes.
Countries faced with severe debts usually resort to national policies favoring export
production. As a result land degradation may result from destruction of forest, soil,
wetlands, and water sources. In order to service debt, new lands are cleared for ranching
or commercial cropping. Coastal areas are drained, mangrove forests cut, in order to
accommodate the expansion of tourist hotels and other foreign installations that hold out
the hope of hard currency earnings. Population growth and urbanization increase demand
for energy and in many countries dams (often large-scale) are built to produce electricity.
These dams flood vast areas of forest and other lands, forcibly displacing the inhabitants
to more vulnerable areas. The result of this debt-induced activity is an increase in the
vulnerability of the exposure. A severe hurricane on a coastal tourist resort in the
Caribbean results in huge property losses.
2.2.1.2 Development induced investment in large-scale projects
Economic growth in the developed countries has increased the exposure to
catastrophic property damage. Due to shortage of prime land in urban areas, extremely
vulnerable sites are chosen for the development of real estate such as coastal Florida and
resorts in Hawaii. Intensive capital development has increased the probability that a
Chapter Two: Catastrophes and Development
30
hazard like a hurricane will encounter an increasing amount of constructed facilities
unless steps are taken to reduce risks within cities and on industrial sites.
Chapter Two: Catastrophes and Development
31
Fig. 2.4 Larger negative external balance on goods and services is associated with larger losses
R2 = 0.2201
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
(35.0) (30.0) (25.0) (20.0) (15.0) (10.0) (5.0) - 5.0 10.0
Econ
omic
loss
as
a %
of G
DP
Pre-event external balance on goods and services % of GDP
Fig. 2.5 More government repudiation of contracts are associated with larger losses R2 = 0.3848
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 2.0 4.0 6.0 8.0 10.0 12.0
Econ
omic
loss
as
a %
of G
DP
Government repudiation of contracts
Chapter Two: Catastrophes and Development
32
For the LDCs, development planners often introduce technology at the so-called “leading
edge” of whatever version of rapid, systemic change they define as “development”. This
may be irrigation technology in the form a large dam that displaces thousands of families
in what economists call “the short run”. It might take the form of low-income housing or
the development of an industrial complex. Such development initiatives, though well
intentioned and useful, can have a series of unintended, unforeseen consequences, the
most detrimental of which is an increase in vulnerability of the poorest.
In many developing countries, as the economy grows, production and exchange become
increasingly complex and the institutional structure of the economy changes accordingly.
Traditional institutions, like extended households that are important mechanisms for
surviving shocks, may no longer be optimal. Since the risk-sharing functions may be
performed by other institutional arrangements that vary in their abilities both to fit local
circumstances and to perform their respective tasks, transition to a developed society
leaves many groups marginalized thus increasing their vulnerability. In developing
countries for instance, insurance is not a well-developed risk-sharing institution.
Financial inter-mediation is usually poorly developed in the LDCs. Financial inter-
mediation is very important for economic development because individuals live in risky
environments, which makes savings, insurance and consumption credit yield direct
benefits in coping with risk. Another reason is that the development of credit and
insurance should enhance an economy’s investment efficiency and, possibly growth.
Failures of inter-mediation are intimately linked with misallocation of capital and
inefficiencies. Individuals who have the most productive investment opportunities may
be denied access to funds.
The interaction of production risks with information asymmetries may increase
opportunistic behavior and hence reduce production. In developing countries, institutions
such as courts, various kinds of bonding mechanisms, social norms, and the structure of
incentives in future contracts are not strictly implemented and thus they are not able to
Chapter Two: Catastrophes and Development
33
limit the practice of opportunistic behavior by different agents. This leads to further
inefficiencies in the economy. In fact, data from past record of catastrophe suggest that
higher rate of government repudiation of contracts introduces more risks in the economic
environment and hence is associated with higher losses (Fig 2.5). Constant threat of
occurrence of a natural hazard may increase the uncertainties and limit investment
decisions, thus dampening economic growth.
2.2.1.3 Development and Population growth
Economic development and population growth affect the relationship of people
with their environment in several complex ways, most of them negative. In the opinion of
Dando (1980, pp. 105-9), the pressure to produce ever more food is creating “an agro-
environment conducive for eco-catastrophes.” The debate about environmental
degradation is frequently linked to the Malthusian notion of a regional “carrying
capacity”, this being defined as the number of people and animals a specified area can
maintain over a period of time. When populations exceed this limit, a cycle of over-
exploitation of the land is set in motion, which ultimately degrades the natural resource
base to such an extent that human and animal survival is unsustainable.
Continued population growth outstrips the ability of governments to invest in education
and other aspects of social development including disaster preparedness measures.
Higher population growth has detrimental consequences on disaster susceptibility. This is
clearly indicated in Fig 2.6. Higher population growth is positively associated with higher
loss-GDP ratios. Population growth also creates further competition for land resources
and in urban areas this increases the vulnerability to natural hazards. In the very poorest
countries, the human use of natural resources has created a problem of food security and
fragile livelihoods. Only quarter of the people in Africa have access to safe drinking
water. As a result even a hazard of mild magnitude can have catastrophic consequences.
Chapter Two: Catastrophes and Development
34
Fig. 2.6 Higher population growth is associated with larger losses
R2 = 0.3359
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Econ
omic
loss
as
a %
of G
DP
Average population grow th
Fig. 2.7 Larger urban population growth is associated with larger losses
R2 = 0.3471
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
0 1 2 3 4 5 6 7 8
Econ
omic
loss
as
a %
of G
DP
Urban population grow th (annual %)
Chapter Two: Catastrophes and Development
35
2.2.1.4 Development and Urbanization
As a result of various processes associated with development, people from the
countryside move into cities seeking better job opportunities. This urbanization process
results in land pressure as migrants from outside move into already overcrowded cities.
The new arrivals are forced to occupy disaster susceptible regions. Catastrophes being
rare events by definition, the migrants rarely make their decision to migrate to vulnerable
regions based on natural hazard probabilities. Slum residents often incur greater risks
from natural hazards (especially landslide and fires) as a result of having to live in very
closely spaced unsafe shanties that are located in low-lying areas. Current projections
indicate that within the coming decade there will be twenty cities with populations
between 10 to 25 million. Of these, fourteen are in the Third World, eleven in hazardous
zones (Blaikie et al. 1994).
Moreover, these cities are expanding rapidly, with the obvious risk of sloppy construction
standards (Tyler 1990). Maskrey (1994) cites the example of Peru, which has become
more hazard-prone with the post-colonial shift of population from mountain communities
to high-risk urban centers. This is especially the case in the capital, Lima, which is
located in a seismic zone where houses of Spanish design with heavy roofs are crowded
with low-income families. Record from past catastrophic events indicates that higher
urban population growth is associated with larger losses as a proportion of GDP (Fig.
2.7).
Lipton (1977:18-19) argues that rural poverty and vulnerability to famine are often a
function of government policies which are biased in favor of the interests of urban elite,
and which therefore discriminate against the interests of the agricultural sector in general,
and of the rural poor in particular.
Chapter Two: Catastrophes and Development
36
2.2.1.5 Development and poverty
People become disaster victims because they are vulnerable. Because people have
different degrees of vulnerabilities, they suffer differently. Inefficient development
policies are associated with increasing inequalities between communities and households,
resulting in rising vulnerabilities for some groups of people. The more the inequalities
that exist in a community more pronounced are the effects of a hazard. Growing poverty
creates greater vulnerability to natural hazards for the population for several reasons.
Farmers may be dispossessed of land and compelled to grow cash crops rather than
subsistence food. Urbanites may be forced to live in most dangerous built up areas. A
catastrophe simply reinforces the growing gap between rich and poor. Even in “normal”
times the poorest sections of society are pressured to over-use the land, and when disaster
strikes, the conventional responses may merely accelerate the continued
underdevelopment and marginalisation. Vulnerability is also the result of third world
impoverishment perpetuated by technological dependency and unequal trading
arrangements between rich and poor nations (Susman et al., 1983).
It is important to note that vulnerability and poverty are not synonymous, although they
are often closely related. Vulnerability is a combination of characteristics of a person or
group, expressed in relation to hazard exposure, which derives from the social and
economic condition of the individual, family, or community concerned. High levels of
vulnerability imply a catastrophic outcome in hazard events. Vulnerability is a complex
combination of both the qualities of the hazards involved and the characteristics of the
people. Poverty on the other hand describes people’s lack or need. Vulnerability is a
relative and specific term, always implying a vulnerability to a particular hazard. A
person may be vulnerable to loss of property or life from floods but not to drought.
Poverty may or may not be a relative term, but there are no different types of poverty for
any one individual or family depending on the causes.
Chapter Two: Catastrophes and Development
37
As many as 850 million people live in areas suffering severe environment degradation
(Smith, 1996). In many LDCs more than 80 per cent of the population is dependent on
agriculture but many are denied an equal access to land resources. Poverty forces many
people to adopt unsustainable land use practices. Countries with a legacy of deforestation,
soil erosion and over-cultivation find their environment more vulnerable to natural
hazard, especially floods and droughts. In fact the record of historical catastrophes
suggests a statistically significant relation between deforestation and the loss levels as
shown in Fig. 2.8
Fig. 2.8 Greater net forest depletion is associated with larger lossesR2 = 0.1595
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0
Econ
omic
loss
as
a %
of G
DP
Pre-event Genuine savings: net forest depletion (% of GDP)
Chapter Two: Catastrophes and Development
38
2.2.1.6 Vulnerability as a “phase of development”
Poor rural households may experience increased vulnerability during the
transition from a peasant or semi-subsistence society to a market economy. Traditional
mechanisms of coping against natural hazards are disrupted due to the capitalist
penetration of subsistence economies. Most development schemes do not substitute
traditional coping mechanisms since the incentives for preparing against natural hazards
may be missing. Unless the traditional coping mechanisms are substituted with alternate
mechanisms, large segments of the population will be made more vulnerable to natural
hazards. Sen (1981, p.173) has written that: “The phase of economic development after
the emergence of a large class of wage laborers but before the development of social
security arrangements is potentially a deeply vulnerable one.” As Clay (1986, p.180) puts
it: “freeing the hidden hand where the basic needs of the majority of people is not assured
is potentially a recipe for disaster.” When entire communities are made vulnerable during
“the phase of the pure exchange system transition” and are further destabilized by
occurrence of a natural hazard coupled with adverse processes of development, such as
changes in modes of production, the result can be catastrophic.
In many rural areas in developing countries transition from a village economy to a market
economy brings with it hastily adopted building techniques. People invest in residential
buildings that are unsafe because they are not built according to standard safety
guidelines. For example, as far as seismic safety is concerned, a poor household living in
light roof building may be much safer than a middle-class household who live in a semi-
engineered building with heavy roofs. Most of the deaths in the 1993 Latur (Maharashtra,
India) and the 2001 Bhuj (Gujarat, India) earthquake resulted from building collapse and
damage.
Chapter Two: Catastrophes and Development
39
2.2.1.7 Development and Government
The basic functions of the state are to provide law and order and to protect the
property rights. Such services are generally exchanged in return for tax payments. As the
single group in society to which all belong and from which exit may be least possible, the
state may solve a variety of coordination problems and overcome externalities plaguing
other institutions. Usually it is the government that has to provide assistance to disaster
areas. In the LDCs, the government may be plagued with inefficiencies and corruption.
Development aid is mostly used for betterment of those in power, making the poor more
vulnerable. Many catastrophes occur because governments fail or have limited capacity
to provide basic transportation, communication, health and other infrastructure needs of
the poorest. Poor infrastructure with almost no maintenance, lack of welfare programs,
which results in inadequate housing and health provision combined with low nutritional
status results in increasing the vulnerability of poor in the community. The evidence from
past record of catastrophes clearly establishes the link between infrastructure and the
levels of losses experienced. Fig. 2.9 illustrates the fact that better infrastructure as
indicated by the availability of electricity is strongly associated with lower loss-GDP
ratios.
Fig. 2.9 Larger electric power consumption per capita associated with
smaller lossesR2 = 0.4922
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
Econ
omic
loss
as
a %
of G
DP
Electric pow er consumption (kw h per capita)
Chapter Two: Catastrophes and Development
40
Market and non-market (including governmental) institutions, which work quite
adequately, though not perfectly, in normal times can easily turn even a moderate
aggregate shock into a catastrophe. According to Sen (1981) markets and institutions
determine the factors, which bear on the likelihood that an exogenous shock will turn into
mass entitlement failure and hence a catastrophe. The factors include quality and
distribution of endowments, the structure of prices, and the pattern of transfers. The
factors that can transform a shock into a catastrophe appear to be intrinsic features of
quite normal economies – rather than peculiar features of highly distorted or badly
damaged economies. They are always present, but normally hidden from view. And they
do surface in any number of ways after a catastrophe has occurred. The devastating
earthquake that struck the most densely populated and industrialized area of Turkey on
August 17, 1999 can be seen as a recent example. Corruption was rampant in the building
industry leading to substandard construction. The earthquake brought forth these inherent
societal deficiencies (corruption) to full public view by causing severe damage to
buildings built using sub-standard techniques. Bureaucracies (Fig 2.10) and increasing
governmental intervention with mostly untrained civil servants result in inefficient
organizational structures, creating vulnerable societies
Fig. 2.10 Inefficient bureaucracies are associated with larger lossesR2 = 0.2931
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 1 2 3 4 5 6 7
Econ
omic
loss
as
a %
of G
DP
Bureaucratic quality (1-poor, 6-best)
Chapter Two: Catastrophes and Development
41
Weak social infrastructure as indicated by poor law enforcement (Fig. 2.11), weak
unprepared government, corrupt bureaucracies (Fig. 2.12), and a relatively closed
political regime, all enhance vulnerability to hazards.
Fig. 2.11 Better rule of law is associated with smaller lossesR2 = 0.271
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 1 2 3 4 5 6 7
Econ
omic
loss
as
a %
of G
DP
Rule of law (1- Poor, 6-Good)
Fig. 2.12 Higher prevalence of corruption is associated with larger losses R2 = 0.1743
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 1 2 3 4 5 6 7
Econ
omic
loss
as
a %
of G
DP
Corruption (1-High, 6-Low )
Chapter Two: Catastrophes and Development
42
2.2.2 Micro-level determinants of Vulnerability
But what are the root causes of vulnerability to natural hazards? How is the
severity of the catastrophe determined by the initial values of the micro (household) level
parameters affecting vulnerability? These are some of the questions that we will address
in this section.
According to Cannon (1994), vulnerability may be divided into three aspects: the first is
the degree of resilience of particular livelihood system of an individual or group, and
their capacity for resisting the impact of a hazard. This reflects economic resilience,
including the capacity for recoverability (another measure of economic strength and
responsiveness to hazards). This can be called “livelihood resilience”, and has some
affinity with Sen’s concept of entitlement (Sen, 1981). Sen introduces the concept of
entitlement failure as a primary reason for the occurrence of famines. Markets and
institutions determine the factors, which bear on the entitlement failure and hence famine.
The factors are – distribution of endowments, income opportunities, the structure of
prices, and the pattern of transfers. Even a small entitlement shock to poor people can
induce large changes in their survival prospects.
The second component is the degree of self-protection that includes “health”. The health
of individuals, the operation of various social measures for hazard protection including
preventive medicine and quality of constructed facilities including residential structures.
The third component is the degree of preparedness of an individual or group. This is
determined by the protection available (for a given hazard), something that depends on
people acting on their own behalf and on social factors. The notion of precautionary
savings is of relevance here as will be explained in a later section. Communities
repeatedly affected by hazards use various schemes like precautionary savings, social
networks, loans, and credits to help themselves insulate from these adverse income
Chapter Two: Catastrophes and Development
43
fluctuations. The presence of such schemes and their efficiency in smoothing
consumption greatly reduces the vulnerability to natural hazards.
An additional component is the type of hazard. An earthquake, for example, may result in
a catastrophe if it strikes a community that is otherwise well prepared for a hurricane.
One of the reasons for this might be that buildings designed solely to withstand
hurricanes might be vulnerable to earthquakes.
2.2.2.1 Presence of uncertainties and development processes
Risk and uncertainty problems are very important in LDCs since the sources of
risk and their magnitudes are sufficiently varied and large and the relevant probability
distributions of alternative outcomes are unknown. Higher uncertainty in the growth of
income (as measured by the standard deviation of economic growth) leads to more
vulnerability of the affected society. Fig 2.13 associates higher uncertainty in growth
rates to the larger loss-GDP ratios. The return period, intensity, and magnitude of natural
hazards can at best be estimated in probabilistic terms. Moreover, given a hazard of
particular intensity, the potential direct loss in the affected region can be estimated only
probabilistically. Development processes continually change the vulnerabilities of the
regions and the exact way in which vulnerability of a region evolves is complex. This in
turn increases the inherent risk.
Chapter Two: Catastrophes and Development
44
Fig. 2.13 Greater uncertainty in economic growth is associated with larger losses R2 = 0.1324
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
1.0 10.0
Econ
omic
loss
as
a %
of G
DP
Standard deviation of annual economic grow th
Fig. 2.14 Larger volatility in inflation is associated with larger losses
R2 = 0.115
0.01%
0.10%
1.00%
10.00%
100.00%
1000.00%
(0.5) - 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
Econ
omic
loss
as
a %
of G
DP
Log of Pre-event standard deviation of inf lation
Chapter Two: Catastrophes and Development
45
Development processes in the LDCs may be retarded because of increased risk aversion
at the household level. The degree of risk aversion may be larger in LDCs than
elsewhere, in part because incomes are nearer to the minimal subsistence level and in part
because these risks are more related to other problems. For example, an observed
production shortfall can be explained with several reasons. It may be more difficult with
the available information to distinguish among the following alternative explanations: (1)
the direct effects of bad weather on output, (2) the indirect effects of bad weather on
output via the effects on health and the effective labor supply, (3) producer mistakes in
resource allocation, and/or (4) shirking of the workers. Among a cross-section of
countries it is useful to compare some indicator of the overall risk with the loss level. One
such indicator is the volatility in inflation. Higher volatility in inflation implies higher
volatility in future expectations. This in-turn implies more risky environments for
investors as well as households. Fig 2.14 clearly brings out the association between
higher inflation volatility and higher losses. This in-turn implies that inflation volatility is
a key determinant of vulnerability. But it is also sometimes argued that the poor are less
likely to be risk averse, since they have little to lose even if they fail. Development is
retarded because of the limited number of choices the poor are faced with. This denial of
accessibility to the poor is partly responsible for making them vulnerable to hazards.
In the following section we establish broad relationships between development processes
on the household level and its connection to vulnerability. The determinants of
vulnerability at a household level are (i) health, food, and nutrition, (ii) education and
consequent disaster awareness, (iii) endowments, (iv) infrastructure including sanitation
and availability of drinking water, (v) preparedness measures, and (vi) quality of
residential structures.
2.2.3 Development, Households and Vulnerability
It is well recognized that in many developing economies, the household and
family are key economic decision-makers and intermediaries, whereas, as development
Chapter Two: Catastrophes and Development
46
progresses, the market or the state takes over some of these roles. The capacity to self-
organize in times of crisis is extremely important for absorbing the impact of a
catastrophe. In the very early stages of economic development, where income levels
hover around the subsistence level, the risk-reduction element of the basic economizing
function may be the dominant one. It is this element, therefore, which may provide the
basic rationale for some of the most important institutions, such as the family, the tribe,
or the kin group. In such contexts, institutions, such as the need to offer hospitality to
everyone in the village and to allow every individual to obtain much knowledge about
each other, may be very efficient. This is especially true where production techniques are
simple and most exchanges are personal and repeating. Even poor societies, which are
well organized and cohesive, can withstand, or recover from, a catastrophe better than
those where there is little or no organization and people are divided. Similarly, groups
who share strong ideologies or belief systems, or who have strong experiences of
successfully cooperating to achieve common social goals, even when struck by a
catastrophe, may be better able to help each other and limit some kinds of suffering than
groups without such shared belief. Opportunistic behavior is rare because the
aforementioned institutions make for its detection.
Some of the most important choices households make revolve around the human capital
of children and adults. The fact that human capital investments are associated with higher
standards of living and welfare has been repeatedly demonstrated both in aggregate data
and in studies that have used individual or household level micro data. The vulnerability
of households in LDCs to natural hazards is due to low level of human capital
investments in health, education and disaster awareness programs. Low levels of disaster
awareness may lead to scant attention to building standards resulting in sloppy
construction vulnerable even to moderate intensity hazards. The cyclone that struck an
impoverished eastern Indian state of Orissa brought to full view the abysmal levels of
human capital investments that existed in the region prior to the event.
Chapter Two: Catastrophes and Development
47
2.2.3.1 Health, nutrition, and education
What are the factors that determine a household’s ‘health’ and hence its
vulnerability to natural hazards? Studies have focused on the effects of wages, food
prices, health programs, and family planning, on child health, schooling and fertility
outcomes. In areas in which the government carried out agricultural intensification
activities, Rosenzweig (1982, 1990) finds that market returns to primary schooling and
school enrollments are found to be higher and fertility lower for farm households in those
areas. Higher enrollments (Fig 2.15) are important for easier communication of disaster
awareness programs and lower fertility reduces pressures on vulnerability created by
unchecked population growth. Human capital investments also depend on infrastructure,
such as water and sanitation quality; measures related to price and quality of health,
education and family planning facilities; and prices of other health or education inputs
such as foods.
Based on a few studies conducted on the impact of health infrastructure on health
outcomes that are relevant to our concerns, we can conclude the positive effects of
development processes that try to increase the access to health facilities of the poorest.
This improved access to health facilities results in contributing towards reducing the
vulnerability to natural hazards, as Fig. 2.16 illustrates. Rosenzweig and Wolpin (1982)
and Hossian (1989) show a negative relationship between clinics per capita and child
mortality in India and Bangladesh respectively, also using density measures. Thomas,
Lavy and Strauss (1992) show a positive relationship between doctors and child height in
Cote d’Ivoire, and Deolalikar (1992) finds a positive relation between health expenditures
per capita and child weight among low-income households in Indonesia. The data on past
catastrophes reveals that higher child mortality is associated higher loss-GDP ratios (Fig.
2.17).
Health of the members of a household determines its vulnerability to epidemics such as
typhoid or malaria after a flood as well its ability to recover after an event. These factors
are also dependent on availability of adequate sanitation infrastructure at the community
Chapter Two: Catastrophes and Development
48
level. Most LDCs lack the basic facilities of sanitation and health rendering the
households vulnerable to adverse health consequences after a catastrophe. Data on past
catastrophes reveals that there is positive association between access to health
infrastructure and the loss-GDP ratios (Fig. 2.18).
Fig. 2.15 Higher secondary school enrollment is associated with smaller lossesR2 = 0.4001
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 0.5 1.0 1.5 2.0 2.5
Econ
omic
loss
as
a %
of G
DP
Pre-event secondary school enrollment
Fig. 2.16 Availability of physicians is associated with a decrease in the losses R2 = 0.3035
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
(2.5) (2.0) (1.5) (1.0) (0.5) - 0.5 1.0
Econ
omic
loss
as
a %
of G
DP
Pre-event physicians (per 1,000 people, log)
Chapter Two: Catastrophes and Development
49
Fig. 2.17 Higher infant mortality rate is associated with greater losses
R2 = 0.4048
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 0.5 1.0 1.5 2.0 2.5
Econ
omic
loss
as
a %
of G
DP
Pre-event mortality rate, infant (per 1,000 live births)
Fig. 2.18 More number of hospital beds is associated with smaller losses
R2 = 0.3896
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
1 10 100 1,000 10,000
Econ
omic
loss
as
a %
of G
DP
# of hospital beds per thousand
Chapter Two: Catastrophes and Development
50
2.2.3.2 Labor and food intakes
The determination of returns of labor plays a central role in models of development since
labor is by far the most abundant resource in low-income countries. Most of the income
for households in LDCs results from labor-intensive employment. Current nutrient
intakes play a crucial role in enhancing productivity for most of these jobs; for instance
calorie intake increases maximum oxygen uptake (which is related to maximum work
capacity; Spurr, 1983). On the other hand many jobs, in the service-oriented industries of
the developed countries do not require maximum physical effort. It seems likely that the
impact of health on income depends on the nature of work. A laborer, for example, may
suffer a larger decline in income because of physical injury than would a more sedentary
worker. Therefore a catastrophe that seriously undermines the physical ability of a labor-
intensive household may have a drastic effect on its earning capacity. In other words the
vulnerability of households is critically dependent on its food intakes. Studies indicate
among the poor households there is a positive correlation between expenditure and
calorie intakes. As income (or expenditure) rises, households switch to higher valued
foods, not necessarily with higher nutrient content (Behrman and Deolalikar (1987),
Strauss and Thomas (1990), Subramanian and Deaton (1992)). From these studies it is
clear that at low levels of expenditure, the calorie intakes and expenditure are positively
correlated but when per capita calories reach about 2000 per day, the curves flatten out.
When a catastrophe strikes a poor household living on subsistence diet, loss of access to
food results in effects much more severe than for relatively richer households who
consume more than 2000 calories per day. In fact the observations connecting losses to
calorie or protein intakes seems to support this hypothesis. Fig 2.19 and Fig 2.20 relate
the calorie or protein intake respectively to the loss-GDP ratios. More food intakes are
positively associated with smaller loss-GDP ratios.
An earthquake that damages key infrastructure facilities of an urban conglomerate
may leave many sedentary workers unemployed. It is a common observation that there is
a spurt in construction activity after an earthquake. Though it is not obvious that energy
or other nutrient intakes should be correlated with either productivity or labor supply,
there is some evidence that the body adapts to changes over some range in energy intakes
Chapter Two: Catastrophes and Development
51
in such a way as to keep functioning intact. Therefore only at extremely low levels of
calorie intakes, a common feature in chronically poor nations productivity or labor supply
suffers. In times of a catastrophe, it is the poor households that cannot supply the required
labor because of their low calorie intakes. This may result in a slow recovery process in
LDCs.
Fig. 2.19 Larger daily calorie intake is associated with smaller losses
R2 = 0.296
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
1,000 10,000
Econ
omic
loss
as
a %
of G
DP
Daily calorie intake
Fig. 2.20 Greater daily protein intake is associated with smaller lossesR2 = 0.304
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
- 20 40 60 80 100 120
Econ
omic
loss
as
a %
of G
DP
daily protein intake (grams)
Chapter Two: Catastrophes and Development
52
The connections between the macro- and micro-level determinants of vulnerability of a
group to natural hazards are shown in Fig 2.1. To explain one of these connections,
consider how the fact that an economically backward community will have majority of
households that have low levels of investments in health, nutrition, and education. At
household level this in turn implies low levels of disaster awareness and preparedness,
vulnerable health, and inadequate labor supply when it is most needed, i.e. immediately
after the event. All these factors contribute towards increasing the vulnerability of the
household. Other factors contributing towards macro- and micro- vulnerability are shown
in Fig. 2.1, which is a summary of the discussion above. These complex relationships that
determine vulnerability explain the fact that a hazard of similar intensity can cause
different levels of damage to two different communities. Determinants of vulnerability
are important in choosing the environmental and control variables used for examining
empirical evidence regarding post-event economic behavior, as will be discussed in
Chapter 4.
2.3 How do catastrophes affect development?
Having examined the complex ways through which development changes the
vulnerability of a community to natural hazards, this section focuses on the effect of the
occurrence of a catastrophe on socioeconomic processes. Catastrophes disrupt
socioeconomic processes of the affected communities and consequently it behooves us to
relate the adverse effects to development process. As has been previously mentioned, the
main purpose of this dissertation is to investigate the economic consequences of
catastrophes. By reviewing the literature that examines the effects of economy-wide
shocks on various socioeconomic processes, the work reported in the following chapters
can be placed in the right context.
In a historical context, Jones (1987) conjectures that the contrasting paths of development
between East and West were caused by a different incidence of disasters. According to
Jones (1987), with respect to changes over time the aspects of natural environment that
seem most to influence economic history are the very sharpest category of changes
Chapter Two: Catastrophes and Development
53
including catastrophes, immediate adjustments to which were hard to make. On the other
hand, incremental changes of the kind documented in climatic history seem to possess
little independent explanatory power. Economies adjusted to them. Climatic and other
incentives were greater in the Orient than in the Occident. This, in turn, gave more
incentive for the peasants in the East to invest in larger families and less incentive in
physical capital. Consequently, development in the Occident led to rapid industrialization
while the Orient relied on agriculture for growth and industrialization process lagged
behind. Though such an extreme view of the effects of different incidence of disasters is
debatable, it nevertheless shows us the importance of natural hazards in explaining at
least some aspects of the variations in growth rates observed globally.
2.3.1 Macro – Level Effects
Occurrence of a catastrophe triggered by a natural event has potential
consequences for ongoing socioeconomic processes of the affected society at a macro-
level including: (i) development processes including growth, (ii) balance of payments
deficits, (iii) budget deficits, (iv) poverty and inequality, (v) trade and investment and (vi)
sudden movement of population.
2.3.1.1 Effects on Development
One important consequence of a catastrophe especially for developing countries is
the disruption to well-laid development plans when investment resources committed to
long-term programs are reassigned to emergency disaster operations. Long-term
development goals might be undershot or foregone altogether as the original development
programs lose their resources. To quote a recent example, Hurricane Mitch was the most
devastating hurricane of the 20th century. The statistics are staggering: 9,346 known
dead, 9694 missing, 2 million affected, and over $4 billion dollars (US) in damage.
Hurricane Mitch, it is claimed, has decimated the economies of Honduras, Nicaragua,
Chapter Two: Catastrophes and Development
54
Costa Rica, El Salvador, Guatemala, and other Central American countries, setting them
back several decades.
Mary Anderson and Peter Woodrow (1989: 107) cite four examples of development
projects that were interrupted by a catastrophe. In two of these examples (Burkina Faso
and Kordofan, Sudan), basic developmental and environmental work was interrupted by
drought, diverting the NGOs effort into emergency feeding programs. In Joyabaj,
Guatemala, several volunteer agencies involved in developmental activities had to join
together to respond to the earthquake of 1976. In Santo Domingo in the Bicol Region of
Philippines, a village development project of the International Institute of Rural
Reconstruction was pre-empted by a sudden volcano eruption in the area of the intended
work.
Results based on simulation of theoretical models in Chapter 3 indicate that there are
overall welfare losses after a catastrophic event. The post-event consumption is lower
than the pre-event levels as the affected region invests in rebuilding. Chapter 4 presents
empirical evidence showing that post–event growth rates are negatively correlated with a
measure of direct losses. Growth being an indicator of development, the studies in
Chapters 3 and 4 corroborate with the anecdotal evidence presented in the above
paragraph.
2.3.1.2 Effect on Trade and Investment
Increasing globalization in the world economy has resulted in developing various
channels through which a shock is transmitted worldwide. Peek and Rosengren (1997)
investigate the extent to which the sharp decline in Japanese stock prices was transmitted
to the United States via U.S. branches of Japanese parent banks and identify a supply
shock to US bank lending that is independent of U.S. loan demand. They conclude that
binding risk-based capital requirements associated with the Japanese stock market decline
resulted in a decrease in lending by Japanese banks in the US that was both economically
and statistically significant.
Chapter Two: Catastrophes and Development
55
Japan’s capital Tokyo is a central player in worldwide economic activity. The Tokyo
region accounts for roughly 30% of the nominal GNP of Japan and the stock exchange
ranks third in the value of the world trading volume handled each day. Most large
Japanese companies base their headquarters in greater Tokyo and several key industries
are heavily concentrated in the area, including banking, insurance, transportation, oil
refining, printing and publishing, and telecommunications
Eight of the world’s 10 largest banks have their headquarters in Tokyo. The ramification
of an earthquake striking at the political and economic center of Japan would be
tremendous. An earthquake in the Sagami Trough of similar magnitude that occurred in
the Tokyo metropolitan area in 1923 would be truly catastrophic. For the Tokyo
metropolitan area, including the Tokyo, Chiba, Kanagawa, Saitama, and Shinuoka
prefectures, Risk Management Solutions, Inc. (1995) estimated a total economic losses
ranging from $2.0 to $2.7 trillion with property loss due to shaking and fire alone ranging
from $1.0 to 1.2 trillion. The earthquake would cause 40,000 to 60,000 deaths. Hadfield
(1992) reports a study made by Tokai bank, which projected that such an earthquake in
the Tokyo region could cause a crash in the US stock and bond markets, a decrease in the
flow of Japanese funds to foreign countries, and a resulting international financial crisis.
In a region known to have a high strike rate for natural hazards there is an undermining of
business confidence and discouragement of investment. Investors require unattainable
high rates of return from their projects in order to compensate the risks of operating in
disaster-prone areas. Also, regression analysis (Chapter 4) point to the fact that
catastrophes cause increases in inflation and the real interest rates. This further
discourages investment. As a result the region’s economic development may be
dampened.
If as a consequence of a catastrophe, an export crop is destroyed by storms and flooding,
the balance of payments deficits may become unmanageable. The effect can be very
important in small economies where cash crops are the main source of foreign exchange.
Chapter Two: Catastrophes and Development
56
It is a factor taken into account by IMF when awarding emergency funding to countries
suffering from the effects of disasters.
If governments are forced to overspend on disaster recovery and reconstruction there is a
growing public sector deficit and debt. Empirical data presented in Chapter 4 indicate an
increase in budget deficit after an event. Development projects are assigned lower
priorities after a catastrophe. Also, since the poor are unable to bid themselves out of
disaster-prone situations as higher income households do, it results in increase of level of
poverty. Cross-country regression analysis presented in Chapter 4 indicates that there is
an increase in the rate of debt growth after an event.
Catastrophes may result in unexpected movements of population especially from
devastated rural areas to unaffected towns in search for a means of livelihood or
employment. This makes the development of already crowded towns more difficult. For
example, after the recent volcanic eruption in Boma, about 400,000 people migrated to
Rwanda and safer regions in Congo. Regional model simulation of the effects of
migration indicates that recovery process slows down (Fig. 5.33).
In economic terms, secondary losses such as these can be counted among the negative
externalities of disasters. There can also be positive effects of disasters, which provide
unexpected opportunities to upgrade plant and machinery or renew aging infrastructure.
In most cases, however, benefits are unlikely to outweigh the costs of the losses. In
Chapter 3, welfare changes after a catastrophe is quantitatively examined. Results
indicate that a greater loss in capital results in greater overall welfare losses.
Analyses of the effects of economy-wide shocks on human capital outcomes suggest a
mixed picture. It is to be expected that the effects of aggregate shocks will vary by
country. This may result from differences in levels in socioeconomic development,
market structure and sectoral mixes and also levels of publicly and privately provided
safety nets. The impacts are also likely to vary depending on the particular outcomes.
Palloni and Hill (1992) find in Latin America that macro shocks do affect infant and child
Chapter Two: Catastrophes and Development
57
mortality from respiratory tuberculosis and diarrhea. Hill and Palloni (1992), Palloni, Hill
and Aguirre (1993) suggest a Malthusian response of age at marriage or marital fertility
to changes in aggregate income.
These studies of economy-wide shocks have not examined specific differences in country
characteristics that may explain differential responses. In Chapter 4 we present evidence
and study the consequences of catastrophes for macro economic factors. Literature has
also not differentiated between aggregate-level shocks and more local shocks, let alone
the possibility that household’s ability to adjust to shocks may be associated with their
characteristics. For example, are the poor more vulnerable to the impact of adverse
economic shocks? It is necessary to turn to micro-level evidence to answer these
questions.
2.3.2 Effects at a Household Level
In this section we briefly review the research that focuses on both household
responses ex-post to adverse shocks and ex-ante to perceived-risks. Much of this recent
literature has addressed the question of whether households are able (both by themselves
and using community-level mechanisms) to smooth their consumption perfectly against
all risks. Other studies include household saving behavior in presence of shocks and
modeling effects on human capital related outcomes.
2.3.2.1 Savings and Investment
Studies such as Deaton (1992a,b) and Paxon (1992), attempt to test the permanent
income hypothesis. According to the permanent income hypothesis, people base
consumption on what they consider their "normal" income. In doing this, they attempt to
maintain a fairly constant standard of living even though their incomes may vary
considerably from month to month or from year to year. As a result, increases and
decreases in income which people see as temporary have little effect on their
Chapter Two: Catastrophes and Development
58
consumption spending. The idea behind the permanent-income hypothesis is that
consumption depends on what people expect to earn over a considerable period of time.
People smooth out fluctuations in income so that they save during periods of unusually
high income and dis-save during periods of unusually low income. Deaton (1992a,b) and
Paxon (1992) find little support for the strong form although Paxon finds a weaker
version does rather well for Thai farm households. Using estimated impacts of regional
time-series shocks in rainfall on current household income to identify transitory income,
she finds that large fractions of transitory income are saved and so household
expenditures are little affected. Nevertheless, a strict form of the hypothesis that all
transitory income is saved is rejected. Evidence presented in Chapter 4 indicates that
unanticipated change in income due to occurrence of a catastrophe results in changes in
consumption. Evidence is presented to show that catastrophes change ex-ante saving
behavior at least for two years after the event.
In addition to strong assumptions regarding credit markets, permanent income models
treat both permanent and transitory income as exogenous. As pointed out by Besley
(1995), this is very restrictive when there exist many potential income sources and more
so in dynamic models. As an alternative, dynamic models have been used that allow for
endogenous income or credit market imperfections. Rosenzweig and Wolpin(1993) and
Fafchamps (1993) specify dynamic programming models of farmer behavior response to
shocks. Rosenzweig and Wolpin model bullock investment and dis-investment decisions,
incorporating the tradeoff in livestock sales between the potential need for current cash
and foregoing future output due to the loss of livestock for traction.
For those studies that model the effects of shocks through income, a key issue is how
income is measured. The concept of permanent income assumes that all income is
exogenous. Including endogenous components in the income measure will result in the
volatility of transitory income being systematically understated. Rejecting risk pooling in
tests such as the Townsend-type test (Townsend 1994) may then be more difficult.
Chapter Two: Catastrophes and Development
59
2.3.2.2 Identifying Transitory Income
Measuring exogenous swings in income is extremely difficult to achieve in
practice. For studies of risk pooling labor income along with asset sales, transfers and
remittances from temporary migration and farm profits net of the value of family labor
are likely to be endogenous. For example, Morduch (1994) and Rosenzweig and
Binswanger (1993) find that poor Indian farmers are likely to adjust their farm
investments in ways that lower expected profits, but decrease profit variation. Fafchamps
finds that weeding labor is adjusted to rainfall received earlier, at planting time.
In view of the difficulty associated with measuring purely exogenous swings in income,
instrumental variables and household fixed effects techniques have been used to free the
model of unobserved error. Wolpin (1982) uses regional, time-series data on rainfall to
construct long-run moments to instrument current income (which measures permanent
income with error) in a savings equation. Paxson (1992) extends Wolpin’s (1982) study
by using deviations of rainfall from its long run mean (and functions thereof) to construct
a measure of transitory component of income for use in her savings equation. Measure of
permanent income and the (expected) variance of income are likewise constructed.
Rosenzweig and Stark also use rainfall, plus interactions with a household’s dry and
irrigated land owning, to instrument for the household-level mean and variance of farm
profits in explaining the variance of food consumption.
Rosenzweig (1988) uses household fixed effects to model the impacts of household full
income surprises on net transfers into the household and on household net indebtedness.
Likewise, many of the full income pooling tests implicitly use household fixed effects by
transforming the estimating equation so that consumption growth is the dependent
variable (Deaton 1992b).
Chapter Two: Catastrophes and Development
60
2.3.2.3 Risk Pooling and Consumption Smoothing
Given a measure of transitory shock, ex-ante and ex-post strategies used to
smooth consumption have been modeled. To the extent that households are successful in
smoothing consumption, it is less likely that idiosyncratic shocks will affect human
capital investment decisions. A set of studies have tried to estimate the impacts of
unanticipated income changes on various dimensions of ex-post methods of consumption
smoothing, including transfers, credit transactions, asset sales and labor force
participation.
In Chapter 3, effect of sudden decreases in capital due to catastrophes (economy-wide
shock) on consumption is studied using three models. Results, using Ramsey’s model
indicate an instantaneous drop in consumption. An extended model that simulating the
conversion of maturing capital to productive capital (these terms are explained in Chapter
3) indicates that consumption drops but is not instantaneous as predicted by Ramsey’s
model. After the changes in productivity have stabilized, consumption settles to a level
below its pre-event level, if we assume a permanent increase in the capital share in the
production function of the affected region.
A different set of evidence has been provided in the studies prompted by Townsend’s
(1994) work on testing for complete risk pooling against idiosyncratic shocks. The
intuition for Townsend’s test is that if households are able to perfectly smooth their
consumption (or at least smooth against idiosyncratic risk), then conditional on a
household fixed effect and aggregate village consumption (or alternatively a village-
specific time effect), consumption should be unrelated to household income. With panel
data, household fixed effects can be used, so that consumption growth is regressed on a
village-specific time effect and the lagged level or changes in household income.
Many of the empirical tests to date, most using ICRISAT’s India data are consistent with
Paxon’s (1992) tests of permanent income hypothesis. Full pooling is rejected (as is the
strong form of permanent income), as household income is found to affect changes in
Chapter Two: Catastrophes and Development
61
consumption over and above village-time specific fixed effects. However the estimated
income effects on consumption are small. These studies suggest that using household
expenditure (per capita) as a measure of long-run income human capital studies may be
reasonable.
A central hypothesis in the risk pooling literature is that wealthier households are better
able to pool their risk. This may arise because wealthier households have more access to
credit markets, because they have more assets to sell in case of need, because they may
have more diversified income sources (such as non-farm employment), or because
relatives living apart may be better able to afford help during times of distress. For
instance, Townsend (1994) and Morduch (1993) stratify their analyses on land owned
and find that perfect risk pooling is likely to be rejected among the landless or small
farmers. Rosensweig and Stark (1989) find that inherited assets help mitigate the impact
of farm profit variability on the variability of food consumption. It should be noted that
these tests are for risk pooling against idiosyncratic risks. Catastrophes, on the other
hand, are economy-wide events. The next section presents some empirical results of
studying economy-wide shocks.
2.3.2.4 Effect on human capital investments
Studies of responses of human capital investments to shocks have examined
expenditures on human capital inputs, such as food (Rosenzweig and Stark 1989,
Morduch (1993, 1994), Rosenzweig and Binswanger (1993), individual nutrient intakes
(Behrman and Deolalikar 1990), child growth (Foster, 1995); schooling attendance
(Jacoby and Skoufias, 1992), and infant mortality (Ravallion, 1987, 1990, 1997;
Razzaque, Alam, Wai and Foster 1990). A different set of studies has decomposed effects
of child mortality on fertility into expected (hoarding) and shock (replacement) effects
(Olsen and Wolpin 1983).
Ravallion and Razzaque et al. estimate the impact of the 1974 Bangladesh famine on
subsequent child mortality. Ravallion shows that in the Matlab area, time-series mortality
Chapter Two: Catastrophes and Development
62
rates closely track rice prices. Rice prices rose by 50 percent over a very short (3-month)
period at a time when Bangladesh did not have any public safety net programs, such as
the food-for-work program adopted in later years. Household, family and village
mechanisms were, in many cases, overwhelmed. Using vital event data linked to census
records from the Matlab area, Razzaque et al. demonstrate that higher mortality was not
uniformly distributed: smaller increases were registered among wealthier households.
Foster (1995) examines the impact of a major flood in rural Bangladesh on growth in
child weight during the subsequent three months. He derives an Euler equation that
represents changes in household utility associated with changes in child’s weight.
Changes in child weight are expressed as a function of changes in rice prices, changes in
the rate of change of rice prices, changes in the incidence of illness (instrumented by
lagged illness incidence), child age and gender and the elapsed time between the initial
weighing and the follow up. Interest rates are captured by the amount of borrowing by the
household and in aggregate by the village. Foster reports that a higher price of rice is
associated with significantly lower growth for children in landless households, but the
effect is not significant for children of better off households; this is consistent with the
hypothesis of differential ability to smooth. However this ability to smooth may be put to
test in the case of economy wide catastrophes. It is in such situations that the health of the
affected community is seriously undermined.
Is education affected by the occurrence of a catastrophe? Esther Dufflo (1993), based on
evidence from Indonesia, concludes that increased investment in education infrastructure
results in increases in percent of primary educated people in the population. Using
Dufflo”s result, albeit negatively, if the catastrophe results in major destruction of
education related infrastructure, or considerable loss of life, then this may result in a
negative impact on the community’s long term literacy rate.
Hannan Jacoby and Emmanuel Skoufias (1992) show that investment in children’s
education in India is responsive to adverse shocks. Jacoby and Skoufias (1992) use four
years of the ICRISAT data, divided into two cropping seasons, in each year, to examine
Chapter Two: Catastrophes and Development
63
whether changes in the time allocated to school attendance responds to changes in
measured full income, controlling for changes in the local child wage and for village
season effects. Changes in income do have a significant impact on school attendance, net
of the opportunity cost of children’s time, which the authors interpret as indicating a lack
of perfect consumption smoothing. One implication of this study is that if there is a
significant change in income after a catastrophe it may have an adverse affect on school
attendance.
Jacoby and Skoufias then use Paxon’s method to decompose full income into permanent
and transitory components, conditional on village-season-year effects. They find that
transitory income is only significant for landless households, while anticipated effects are
not significant for either type. This evidence is consistent with landless households using
their children as assets to borrow against bad times.
In sum, there are rather few studies that have attempted to measure effects of resource
shocks on human capital outcomes and how existing assets may condition those effects.
This is certainly a question, which is raised directly by the issue of whether economic
adjustment hurt the poor disproportionately as claimed by Cornia, Jolly and Stewart
(1987). The few studies reviewed here suggest that there are some impacts, certainly in
the case of major events such as a flood or famine, and the effects seem to hit the poor
hardest, consistent with intuition. Whether smaller changes have negative impacts,
though, or whether households are able to adjust in ways not detrimental to human capital
investment is still unclear. Furthermore, the aggregate time-series evidence indicates
differences among countries, which may be partly a function of the existence and quality
of social safety nets as well as of the level of market and human capital development and
thus households’ ability to adjust.
2.4 Methods of Coping- Risk, insurance, credit and saving
In LDCs disaster is often accepted as a “normal” part of life. In this situation
group coping strategies in the forms such as extended households are important. Nomadic
Chapter Two: Catastrophes and Development
64
herdsmen in semi-arid areas have tended to accumulate cattle during years with good
pasture as an insurance against drought (Smith 1996). In developed countries technology
and engineering design have provided a high degree of reliability for most urban services
against natural hazards. But a severe earthquake can easily disrupt road networks, electric
power lines or water systems. This can have damaging consequences because, when such
systems fail, there is frequently no alternative source of supply.
Claude Gilbert (1998) and Hewitt (1998) point out that disasters have often been
considered as the affairs of the public authorities rather than the affairs of citizens. In
doing this, the perception of the population as a whole is merely passive and bound to be
directed and commanded in cases of disasters. But nearly all studies carried out on deep
crisis situations (Gilbert 1998) show that human communities do participate in the
management of disasters. It is well known that in case of earthquakes, such as the one
that happened in Mexico City in 1985, the assistance to the victims comes first of all
from other survivors, with the means used by the authorities contributing to emergency
only to very little extent. In short, what is interesting in the empirical studies of disasters
is the surprising capacity of the reaction and self-organization of people outside any usual
public or institutional structure. Given the capacity of the citizens to react formidably in
disaster situations, there has been little effort to link studies in risk and disaster-coping
behavior in development economics literature with disaster studies. The disaster literature
has concentrated on investigations on functioning, good or bad, of public powers and
official emergency systems.
In the following section we present literature from development economics that presents
evidence regarding economic behavior in adversely affected low-income settings. People
and in particular disaster victims rely on various social coping systems: households,
groups, community, villages, government and non-government agencies, insurance,
credit, and international institutions.
Chapter Two: Catastrophes and Development
65
2.4.1 Households, groups, community, villages
Risk and consumption smoothing problems condition structure of rural
households. Indeed, the structure of households in the low-income settings appears to
differ distinctly from that of high-income industrialized countries characterized by more
organized markets, governmental social insurance schemes, more predictable income
sources, and technological change, where the dominant household form is the nuclear
family.
In the post-disaster recovery period each family’s social and economic position and
connections within the larger community are critical factors influencing outcomes for
their household (Drabek et al. 1975; Bolin 1982). Comparative disaster studies have
revealed three principal modes of family recovery: first, the autonomous use of personal
resources (such as savings or insurance); second, reliance on informal kinship support
systems; and third, the utilization of institutional resources (such as government
assistance) (Morrow, 1997). While victims often make use of all three, the extent to
which one dominates is primarily determined by the larger political end economic setting
(Bates and Peacock 1989). In modern industrialized settings, kinship assistance is less apt
to be the primary source of help, but still is important (Bolin 1982; Nigg and Perry 1988).
Morrow (1997) reports those families in Hurricane Andrew’s path provided a unique
opportunity to study the role of kin networks in disaster preparation and response. Only
14 per cent of households in the survey reported in Morrow (1997) received assistance
from relatives when preparing their homes and, among those reporting having kin in the
area, the rate was only 16 per cent. Using logistic regression models Morrow shows that
minority (Black and Hispanics) families are more apt to have been helped by relatives for
pre-disaster preparations as compared to Anglos households. Overall, kin networks
appear to be under-utilized during hurricane preparation: While nearly 75 per cent of the
respondents had relatives living nearby, less than 20 percent reported assisting or being
assisted by them.
Chapter Two: Catastrophes and Development
66
During the storm about 27 per cent of the total sample reported having relatives stay in
their home during the storm and the rate was significantly higher for Hispanics. After the
storm 24 per cent of those with relatives in the area reported receiving major assistance
with such things as supplies, debris removal, and repairs. Similarly, 30 per cent of those
with family in the area reported assisting relatives after the storm. 44 per cent among the
sub-sample from South Dade received help from relatives in the area. These findings
suggest that under severe conditions family networks become an important source of help
in the aftermath of a disaster, even in an industrialized country.
2.4.2 Insurance, Savings and Credit
The functions of savings, credit and insurance are intimately connected with one
another in most developing economies. However, at first sight, there ought to be a
division of roles between transactions that transfer resources across time, as with savings
and credit, and those that transfer resources across states of the world, as with insurance.
Moreover, decisions about how to allocate resources over time and states ought to be
separable in the sense that the consumer’s decision about how much to borrow and save
is independent of uncertainty about future events. This viewpoint is appropriate only
under the most restrictive of circumstances, when we are in a competitive economy with
a complete set of Arrow-Debreu securities and no externalities. An insurance scheme
may be approximated in the actual economy by various risk-sharing opportunities and
markets. Some possible sources of insurance include stocks in securities markets,
borrowing and lending in credit markets, unemployment insurance, contracts between
employer and employee, crop insurance for farmers, and insurance among family
members or close communities. The separation of the functions of insurance and
saving/borrowing no longer holds when markets are incomplete. It is limitations on
insurance possibilities that make it essential to treat savings, credit and insurance in a
unified way.
Chapter Two: Catastrophes and Development
67
Two features of developing economies are particularly germane to the link between
savings and insurance. First, the absence of markets for trading in risks is particularly
noticeable. Many types of insurance possibilities taken for granted in developed
countries, are simply not traded. This is especially striking given the relative importance
of risk in the lives of many inhabitants of LDCs, such as the risk of suffering certain
infectious diseases. Second, a large fraction of population is typically dependent on
agricultural income for their livelihood. The latter may be subject to drastic weather
shocks and commodity price fluctuations. Rosenzweig and Binswanger (1993) bring out
the relative importance of risk in an agricultural economy by the finding that the
coefficient of variation of income from south India is 137. For white males aged 25-29
surveyed in Longitudinal Survey of Youth in the US in 1971, the number is just 39.
When insurance markets are incomplete, saving and credit transactions assume a special
role by allowing households to smooth their consumption streams in the face of random
fluctuations. In a purely autarkic model, the savings decision is made in isolation. While
in a developed country we might naturally think of saving using demand deposits that
earn interest, this is not necessarily a good model for LDCs (Besley, 1995). There is
evidence that because of high transactions cost, low levels of literacy and numeracy,
mistrust of financial institutions, individuals will often accumulate savings in forms other
than demand deposits such as assets. The national accounts statistics of India, for
example, report that in 1987-88 households accounted for more than 80 percent and less
than 52 percent of household savings was in the form of financial assets, the rest being
direct saving in physical assets. Bevan, Collier, and Gunning (1989), discusses the
importance of accumulation in non-financial assets after the Kenyan coffee boom in the
late 1970s. The role of these non-financial assets in smoothing consumption against
economy-wide shocks needs to be examined.
Loans are less available when the local economy is subject to a common shock, such as a
late monsoon (Rosenzweig, 1988). Thus weather-induced profit variability may be far
less insurable than idiosyncratic or household specific profit variability necessitating ex
ante risk reduction through altering of portfolio of investments that differ in their
Chapter Two: Catastrophes and Development
68
sensitivity to weather outcomes. Investments would then be predominantly responsive to
weather risk.
Informal credit institutions are very important in LDCs and there is a huge diversity of
them in Asia as reported by Ghate (1992). Among the main sources of informal credit
are: loans from friends, relatives and community members; rotating savings and credit
association; moneylenders and informal banks; tied credit and pawning.
Udry’s (1990, 1994) studies of Northern Nigeria are based largely on loans from friends,
relatives and community members. Udry focuses on the risk-sharing function of loans,
with repayments being indexed to the borrowers’ and lenders’ economic circumstances.
Educational loans between family members may also be important with repayment
appearing as urban-to-rural remittances. The main enforcement mechanisms for such
loans tend to be informal social sanctions.
2.4.3 Credit, insurance and long-run development and growth
The relation between credit, insurance and long-run development and growth
works via the role of intermediaries in providing finance for industrialization and,
consequently, much lending of this sort occurs in an environment that is, arguably, not
very special to developing countries. For any given aggregate level of savings, the quality
of financial inter-mediation is a crucial determinant of the efficiency of investment
choices, i.e., in ensuring that savings find their way into the most productive
opportunities. Insurance may also be important, especially in relation to incentives to
adopt new, riskier technologies.
An important theme in the relationship between credit markets and long-run development
countries today, and many now developed countries historically, was a lack of institutions
for funds to flow to where capital could be most productively used. The evolution of
financial institutions can be understood in large part as trying to overcome this, leading to
a more efficient allocation of capital throughout the economy. This view is based on
Chapter Two: Catastrophes and Development
69
realizing gains from trade from differences in technologies. This discussion also seems
relevant to modern day developing countries where, as we remarked above, market
segmentation is significant. Improvements in infrastructure and communications, more
generally, also play a central role in providing market integration. Temporary disruption
of these facilities after an earthquake may result in breaking up the integration that
directly affects the development of the affected region.
Many of the inefficiencies that arise because insurance possibilities are lacking may take
the form of a failure to adopt new technologies and appropriate investments. In a LDC
context Eswaran and Kotwal (1989) have discussed how access to credit may affect
technology adoption decisions. Townsend’s (1992) study of northern Thai villages
attributes the non-adoption of new rice varieties to the absence of insurance possibilities.
Insurance is a key loss-sharing strategy in the developed countries. Like disaster aid, it is
a redistributive method but in this case people at risk join forces with a large financial
organization to spread the costs more widely. Most insurance takes place when an
individual perceives a hazard and purchases a policy from a commercial company, which
guarantees that any specified losses will be reimbursed. Hence, the policyholder spreads
the possibly crippling cash burden from one catastrophe over a number of years through
the repayment of an annual premium.
For the insurance company, risk spreading starts with the underwriting of property, such
as buildings or crops, against natural hazards. Policy underwriters try to ensure that the
property they insure is spread over diverse geographical areas so that only a small
fraction of the total value at risk could be destroyed by a single event. By this means,
payments to those policyholders suffering loss are spread over all policyholders.
Assuming the premiums are set at an appropriate rate, the money received from
policyholders can be used to compensate those suffering loss.
Chapter Two: Catastrophes and Development
70
2.5 Aid and recovery
Disaster aid is the inevitable outcome of humanitarian concern following a
catastrophic event, usually involving the loss of life. Though necessary for immediate
post-event relief aid can never fully alleviate the profound economic and social
disparities around the world, which are responsible for so much hazard vulnerability. Aid
is not a good long-term solution for disaster reduction since victims come to rely on such
external support. They might be indirectly encouraged to settle in high-risk areas driven
by the expectation of being compensated after a catastrophic event. Data from past
catastrophes indicates that more of government’s allocation of its expenditures to some
form of aid is associated with lager catastrophic losses as a percent of GDP (Fig. 2.21).
Fig. 2.21 Larger aid is associated with larger losses
R2 = 0.2069
0.0%
0.1%
1.0%
10.0%
100.0%
1000.0%
(2.0) (1.5) (1.0) (0.5) - 0.5 1.0 1.5 2.0 2.5
Econ
omic
loss
as
a %
of G
DP
Aid (% of central government expenditures)
Chapter Two: Catastrophes and Development
71
2.5.1 Disaster aid at the Macro Level
At a macro level, disaster aid flows to victims via governments and charitable
non-governmental organizations (NGOs), such as the International Federation of Red
Cross and Red Crescent Societies (IFRCRCS), Oxfam, Save the Children Fund and the
religious agencies. As many governments in the developed countries are less willing to
take welfare responsibility for the poor and the vulnerable, and governments in the LDCs
are less able to do so, the role of NGOs in disaster relief increases. Since the 1970s there
has been a substantial increase in the proportion of aid channeled through the NGOs. For
example, the European Community, which is the world’s largest single provider of
disaster aid, raised the proportion of its funding through NGOs from zero in 1976 to 40
percent by the mid-1980s (IFRCRCS, 1993, 1994). Other sources of International aid
include UN’s Disaster Relief Organization (UNDRO), Department of Humanitarian
Affairs (DHA), and USA’s Office of Foreign Disaster Assistance (OFDA).
Despite all these efforts to organize disaster relief, the results are often disappointing.
International aid flowing from developed to developing countries can be unreliable and
may well not reflect the true need. For example, severe earthquakes and tropical
cyclones, which invariably result in a high ratio of casualties to survivors, usually
generate a large donor response irrespective of need. Alternatively, droughts and floods
tend to produce comparatively low responses, despite the large numbers of survivors who
will be adversely affected and in need of support. Not only are some disasters apparently
more fashionable than others are but aid is often highly political and may even be used as
a weapon by the powerful donor nations.
The political relations of the affected country with the donor countries often affect the
flow of development and disaster aid. Development aid, in particular, may be tied to trade
agreements rather than targeted at the countries in most need. In Europe a great deal of
disaster aid is raised for former colonies in Africa and Asia whilst the USA most actively
supports friendly Latin American countries within its sphere of influence. Very abrupt
Chapter Two: Catastrophes and Development
72
changes in policy may occur. For example, the 1988 earthquake in Armenia, formerly
USSR, generated the largest initial donation (5 million pounds sterling) from the British
government following any natural disaster, despite the fact that the government had never
given any disaster aid to the former Soviet Union. This happened in 1976 when
Guatemala rejected earthquake assistance from Britain because the two countries were
then engaged in a territorial dispute over Belize in Central America.
Examples of badly managed aid relief abound. After the hurricane Gilbert aid consisted
of simply dumping surplus commodities such as fur coats, high heel shoes and heavy
winter clothing for victims in the tropics. Consideration should be given to food
requirements and the development needs of the recipient country since over-generous
donations can lower market prices and disrupt the local agricultural economy in some
developing countries. Excessive food aid may induce the government of the affected
region to lower its priority for self-sufficiency in agricultural base.
There is little hard evidence that disaster aid results in net benefit. Chang (1984) applied
an economic model of disaster recovery to coastal Alabama, USA, following a damaging
hurricane in 1979, and concluded that outside assistance from federal agencies and
insurance companies was not sufficient to replace lost assets. The inflows of capital into
an area immediately after a disaster and the improvement of some facilities do not
amount to an overall net gain accruing from a catastrophe.
Getting assistance after the catastrophe may be arduous because of the following reasons:
Getting to the correct location to file an application may be difficult because public
transportation is not available because of massive environmental destruction. Most street-
signs as well as prominent landmarks will also be damaged making the location of
assistance centers problematic.
Successfully negotiating the aid process – getting all of the assistance for which a
household is qualified – typically takes a great deal of time, energy, and skill in dealing
with bureaucracies. Poor people lack these assets. In case of Hurricane Andrew, it usually
Chapter Two: Catastrophes and Development
73
took several trips to the FEMA Disaster Assistance Centers (DACs) or other centers to
complete various, rarely integrated application processes.
The destruction of entire communities places extraordinary demands on insurance
companies, government agencies, material suppliers, and skilled labor. Many
homeowners lack the necessary resources for repairing their homes, usually because they
are uninsured, underinsured, or their insurance companies folded or paid too little.
2.5.2 Disaster aid at the Micro Level
There are a host of reasons why there may be economic links across households,
and these linkages manifest themselves in a variety of ways, including transfers of money
goods and time. First, households may be altruistic. Second, transfers may be motivated
by exchange through an implicit contract or strategic behavior (Berheim, Scheifler and
Summers 1985; Cox 1987). A particular form of exchange, insurance against income
shocks, may be a key motive for inter-household links in a developing country context
(Rosenzweig 1988a, Rosenzweig and Stark 1989). If so, then this suggests the pool of
potential donors and recipients may be very large.
Peacock, Killian, and Bates (1987) found that households residing in peripheral isolated
rural villages recovered more slowly than households residing in more economically and
politically complex central cities. Households within larger communities had better
access to external resources and were able to take advantage of post-disaster funding and
resource opportunities (Bates and Peacock 1993). The implication is that a community’s
position within a regional stratification system and exchange network has important
consequences for disaster-related recovery processes.
For example the 1987 drought in Pakistan’s Thar Region found many villagers living
with their relatives in other villages in their district. But risk sharing at the provincial
level is less likely because at that level distances may be too great to enact transactions
Chapter Two: Catastrophes and Development
74
and kinship may be weaker. Risk sharing has implications for the timing of transfers
(Rosenzweig 1988). The direction of net transfers should depend on which party has
faced positive or negative shocks and is unrelated to the life cycle. This hypothesis is
very hard to test without a (long) time series of data on both sides of the contract (and
possibly on all households in the pool.
Households may choose to mitigate risk through spatial diversification in living
arrangements (Rosenzweig 1988). Rosenzweig and Stark (1987) argue that in south India
where women, rather than men, marry and then to move to new households, families will
seek to locate daughters in different places if risk is a concern. The authors find almost
complete diversification, and that the extent of diversification is greatest for the least
wealthy, who are, presumably, those at the greatest risk in the face of a weather shock.
Furthermore, daughters tend to marry kin, who it is argued, are more likely to be
concerned about the origin family after marriage. Rosenzweig (1988) reports that
conditional on wealth, variability of agricultural profits positively affects the number of
co-resident daughters-in-law, which he interprets as a measure of intergenerational and
spatial extension. Though the effect is not precisely estimated, it is interesting to note the
various ways in which risk is managed in developing countries.
2.6 Policy issues
This section discusses some specific issues of policy towards orientating the
development process taking catastrophes into account. There are broadly two normative
criteria that can be applied to motivate policy. The first is based on concerns about
equity. Poor people are most likely to be excluded from trade in formal financial markets.
Some reasons for this is that poor people lack reliable forms of collateral, are less likely
to be literate, numerate, may face higher transactions costs and lack the influence needed
to gain subsidized loans. This suggests that interventions that genuinely broaden the
scope of financial inter-mediation may have a major impact on the poor in terms of
raising their self-reliance in times adverse situations that may result from a catastrophe.
Chapter Two: Catastrophes and Development
75
While this is a useful measure the question of whether intervention in financial markets
alone is an appropriate policy response is debatable. This raises the large subject of what
are appropriate policy interventions are required to reduce vulnerability against natural
hazards.
Recent discussions of disaster-relief policy have emphasized the synergy with longer-
term development goals (Anderson and Woodrow 1989). Catastrophes triggered by
natural events can have a significant effect on development, at all levels from that
individual households and local communities through to national level. Moreover,
development and its consequences are key factors in determining whether a natural
hazard is transformed into a catastrophe. Management of anticipated catastrophic events
should be integrated into a region’s development plans and no longer left as viewed in
terms of immediate post-event crisis management strategies of relief teams.
For disaster reduction and development to occur together, a reliance on local knowledge
rather than imported technology is required. In rural areas, for example, successful
development means “bottom up” strategies that start at community level, such as the
establishment of cooperatives to provide seed banks, crop insurance and credit for tools
and other assets lost in a disaster. Such measures would help to stabilize the rural base
and halt the migration to unsafe urban environments. Conversely, technical aid,
particularly beyond the emergency relief phases, is perceived as increasing vulnerability
by making a short-term problem semi-permanent through additional dependency.
It is difficult to disentangle disaster and development problems in the developing
countries and to make reliable assessments of disaster aid as an adjustment to an
environmental hazard. But, wherever possible, disaster aid should be minimized. Given
the fact that some emergency response will always be necessary in certain cases, attention
should be given to optimizing this form of relief. More training of local aid workers
would help, especially if continuity could be maintained by re-training core staff from
event to event. Aid needs to be carefully targeted in order to improve the situation of the
most vulnerable people.
Chapter Two: Catastrophes and Development
76
2.7 Summary
Hazards of similar intensities result in only a few people being affected in the
developed countries but result in thousands being affected in the developing regions of
the world. Why does this happen? In other words, what transforms a hazard into a
catastrophe? The answer to this complex question can be summarized in one word –
vulnerability. Hazards are converted into catastrophes by vulnerabilities in the existing
socioeconomic fabric of a community or a nation. What are the factors that determine
vulnerability?
Using literature on development and growth economics, and sociology of
disasters as pointers, the determinants of vulnerability (measured by the loss-GDP ratio)
to natural hazards were identified. Data on catastrophes demonstrates statistically
significant associations between the various socioeconomic indicators and the loss-GDP
ratios. These associations map the complex relationships between ongoing
socioeconomic and development processes and the determinants of vulnerability to
catastrophes.
Corrupt and inefficient governments and bureaucracies, poor physical
infrastructure facilities, excessive dependence on imports, poor health infrastructure,
large uncertainties in the macroeconomic environment, and low levels of literacy are all
factors contributing towards the vulnerability. It is not surprising to note that these factors
also determine the per capita income of a nation, and it has been shown (Fig. 2.2) that
they are inter-related. However, physical and human capital losses also depend on hazard
intensity.
It is apparent that reducing disasters is possible not only by modifying the hazard, but
also by reducing vulnerability. However, most of the efforts of those concerned with
disasters are focused either on reducing the impact of the hazard itself (sometimes in
expensive and inappropriate ways), or on reducing one aspect of vulnerability – social
Chapter Two: Catastrophes and Development
77
protection through certain forms of technological preparedness. The major determinants
that make people vulnerable (i.e. the social, economic and political factors, which
determine the level of resilience of people’s livelihoods, and their ability to withstand and
prepare for hazards) are rarely tackled. Building storm shelters on the cyclone-prone
regions of eastern India is a solution, which can manage the needs of a community
immediately after a cyclone. It does not address the wider problems associated with
growing vulnerability of certain sections of the population during “normal” times.
Mitigation of hazards is normally associated with attempts to reduce the intensity of a
hazard or to make some other modification, which is supposed to lessen its impact. It is
often a hazard-centered rather than a people centered-approach. Large-scale engineering
works to counter river floods and expensive satellite early warning systems for tropical
cyclones are two examples of the “technocratic” solutions to the problem. Other
approaches rely on the state and through local groups or NGO activities. But state
intervention is often unreliable. Though well intentioned, the state is usually a party to the
very same economic and social processes that lead people to be unable to protect
themselves in the first place.
More generally, national development plans and budgeting exercises should take
potential losses from disasters into account and make assessments of the likely
investments necessary to recover from future natural hazards, which may transform into
catastrophes.
Chapter Three: Theoretical Models
78
Chapter Three
Short-Run Analysis of the Economic Consequences of Catastrophes _____________________________________________________________
3.1 Introduction
This chapter investigates the dynamic effects of a catastrophic event that destroys
substantial capital stock of an economy. Three models are presented to address various
aspects of the problem. For the three models, a catastrophe, due to the occurrence of an
earthquake or a hurricane, is modeled by a discontinuous change in the capital stock.
Post-event reconstruction and behavior is modeled using perturbations in productivity
and the inflow of investment exogenously.
A major strength of the framework on which the models are built is that they are founded
in standard micro-economic principles. The study examines how an economy initially in
a steady state responds to unanticipated and large change in the capital stock followed by
an arbitrarily complex change in the affected region’s productivity. Laplace transforms
are used to study the perturbations of the resulting system of ordinary differential
equations.
The results presented below indicate the initial impact on investment, consumption, and
production due to occurrence of a catastrophe and the resultant changes in productivity.
For the purposes of the study we examine the dynamic effects two or three years after the
event. Hence the term ‘short run’ is used in the title of this chapter.
The first model is based on Ramsey’s growth model. Perturbations of the model to
simulate the effect of a catastrophe reveal an initial drop in consumption consequent to
the loss in capital. After an event there is a discontinuous drop in the output of region,
followed by a rapid growth rate. The growth rate is changed by the magnitude of change
in the productivity of the affected that occurs after the event. Immediately after the event,
the affected region’s productivity decreases due to non-availability of infrastructure,
Chapter Three: Theoretical Models
79
factory shutdowns, and increase in prices of various commodities. However, when new
capital is constructed, the productivity increases, because old and least maintained capital
stock is replaced by more productive capital. It may be noted that it is the old and least
maintained capital stock that is most vulnerable to natural hazards. These changes in
productivity are modeled by appropriate perturbations of the production functions.
The second model is an extension of the Ramsey’s growth model, which includes two
different types of capital – the maturing capital and the productive capital. The maturing
capital is present in the economy due to various ongoing construction processes. The
productive capital is the capital that has been commissioned and is being used to produce
goods or services. When a catastrophe strikes a region, the model simulates the impact of
efficiency of the reconstruction process. The results suggest that the output of the
economy drops discontinuously and starts to grow only after a period of time when its
productivity increases. This result is different form the previous model, wherein
economy grows, without delay, after the event. The model is able to simulate a more
realistic behavior of the economy as will be clear from the econometric evidence
presented in a Chapter 4. The reasoning behind it is that in real-life, reconstruction of
damaged capital takes time. It is impossible to fully utilize the capital that is being
reconstructed immediately, as a result of unavoidable start-up problems.
The third model examines the consequences of a catastrophe on a region that is
interacting with another region. This model is again an extension of Ramsey’s model.
The model simulates how resultant changes in productivity of the affected region are
propagated to the other region. The model simulates behavior of county level economies
interacting with state-level economies.
The chapter is organized as follows. Section 3.2 examines the first model. Effect on
initial consumption and investment are derived in Section 3.2.3. Results and discussion
based on a numerical simulation are presented in Section 3.2.4. The next section (3.3)
presents the second model. Results based on a numerical simulation are examined to
bring out salient features of modeling the efficiency of the post-event reconstruction
Chapter Three: Theoretical Models
80
process in Section 3.3.2. Section 3.4 presents the third model. Section 3.4.1 presents
numerical simulation results of the consequences of a catastrophe on interacting
economic regions. Section 3.5 summarizes the chapter’s main points.
3.2 Modeling a catastrophe
Assume that an economy consists of a large fixed number of identical and
infinitely lived economic agents. The common utility functional is assumed to be
additively separable in time with a constant pure rate of time preference, ρ:
∫∞
−=0
)( dtcueU tρ (3.2.1)
where c(t) is consumption of a single item at time t and u is the instantaneous utility
function. We assume the following form for the utility:
γ
γ
+=
+
1)(
1ccu (3.2.2)
γ measures the coefficient of risk aversion. The representative agent will chose a
consumption path, c(t), capital accumulation, k(t) such that he maximizes his utility:
∫∞
−=0
)(0 )(max)( dtcuekV t
tc
ρ (3.2.3)
s.t.
kckfk β−−= )(& (3.2.4) *
0 kk = (3.2.5)
Eq. 3.2.4 represents the fact that the change in capital stock, k& , occurs due to
output, )(kf , of the economy and decreases due to diversion of output towards
consumption, c, and depreciation of existing capital stock, k, at rate β. In other words,
investment, which measures the change in capital, is output less consumption and
depreciation. At t=0, the capital stock is assumed to be in equilibrium, k*. The following
form of the production function is assumed – the Cobb-Douglas form: σAkkf =)( (3.2.6)
Chapter Three: Theoretical Models
81
It is easily recognized that A is a measure of productivity of the economy and σ is a
measure of the capital share in the output. The resulting equilibrium solves the system of
differential equations:
))((/ kfcc ′−+= βργ& (3.2.7)
kckfk β−−= )(& (3.2.8)
At equilibrium, 0== kc && . This implies that the equilibrium values of c and k, denoted by
c* and k*, respectively, are given by:
11
* −
+
=σ
σβρ
Ak (3.2.9)
*** )( kkfc β−= (3.2.10)
A catastrophe is modeled by a discontinuous change in capital, for one period after the
occurrence of an event at time τ. This is accompanied by a change in productivity
associated with post-event reconstruction due to the inflow of aid. To model these
changes due to occurrence of a catastrophe the following perturbation equations are used:
))(1( 1 thεσσ +→ (3.2.11a)
))(1( 2 thAA ε+→ (3.2.11b)
)(3 thεββ +→ (3.2.11c)
Eq. (3.2.11a) simulates the changes in capital share,σ, in output after the occurrence of an
event using a perturbation ε, and a time varying function h1(t). In the rest of the in this
chapter, ε, will denote the perturbation. Eq.(3.2.11b) models the change in total factor
productivity, A, due to construction of capital. The discontinuous reduction in capital
stock after an event is modeled by assuming a Dirac Delta function for h3(t) in Eq.
3.2.11c.
Applying the perturbation in Eq.3.2.11, to Eq.3.2.7 and 3.2.8, the perturbed system of
differential equations can be written as follows: 1))(1(
1231))(1))((1()([/ −+++−++= thkththAthcc εσεεσεβργ& (3.2.12)
]))(())(1([ 3))(1(
21 kthckthAk th εβε εσ +−−+= +& (3.2.13)
)()( 33 τδε −= tath (3.2.14)
Chapter Three: Theoretical Models
82
δ(t-τ) is the Dirac Delta function. To make the dependence on ε explicit the equations are
written as: 1))(1(
1231))(1))((1()([/),(),( −+++−++= th
t kththAthtctc εσεεσεβργεε (3.2.15)
)],())((),(),())(1([),( 3))(1(
21 εεβεεεε εσ tkthtctkthAtk th
t +−−+= + (3.2.16)
*),0( kk =ε (3.2.17a)
∞<∞→ )(lim tkt (3.2.17b)
Eq. 3.2.17a states the fact that the system’s initial condition coincides with its
equilibrium. Since the economy is initially at the ε=0 steady state, the occurrence of a
catastrophic event essentially implies that ε has changed. The impact of this change in ε
on the critical variables at future times will be studied herein. Differentiating Eqs.3.2.15
and 3.2.16 with respect to ε and evaluating at the steady state, the following equations
result:
{ }])()()()()()([/
)0,())(1(/)0,(*2
1211*
3*
2**
kLogthththkAthc
tkkActc t
σσγ
σσγσ
εσ
ε
++−
+−−=−
−
(3.2.18)
σσσβ
βσσεεε
)*()(2)}*()(11)*()(3{*
}1)*(){0,()0,()0,(
kAthkLogthkAthk
kAtktcttk
+−+−−
+−−+−= (3.2.19)
In matrix form:
{ }
++−−++−
+
−′−
′′−=
−
−
σσ
σ
ε
ε
ε
ε
σβσσγ
βγ
)()()}()()()({])()()()()()([/
)0,()0,(
)(1)(/0
)0,()0,(
*2
*1
1*3
*
*2121
1*3
*
*
**
kAthkLogthkAthkkLogthththkAthc
tktc
kfkfc
tktc
&&
(3.2.20)
Using Laplace transform
{ }
++−−+++−+
+
=
−
−
σσε
σε
ε
ε
ε
ε
σβσσγ
)()()}()()()({)0,0(])()()()()()([/)0,0(
)()(
)()(
*2
*1
1*3
*
*2121
1*3
*
kAsHkLogsHkAsHkkkLogsHsHsHkAsHcc
sKsC
sKsC
s J (3.2.21)
where Hi(s) is the Laplace transform of hi(t). It should be noted here that cε(0,0) is the
change in c at t = 0 induced by ε. cε(0,0) is an unknown at this point, but the initial value
Chapter Three: Theoretical Models
83
of k is fixed at k*. This fact yields an initial condition kε(0,0)= k* = 0. Rewriting Eq.
3.2.21, the following equation results:
{ }
++−−++−+
−=
−
−−
σσ
σε
ε
ε
σβσσγ
)()()}()()()({])()()()()()([/)0,0(
)()()(
*2
*1
1*3
*
*2121
1*3
*1
kAsHkLogsHkAsHkkLogsHsHsHkAsHcc
ssKsC
JI
(3.2.22)
Eq. 2.22 is used to obtain the solution for Cε(s) and Kε(s) in terms of cε(0,0). To ensure
the uniqueness of cε(0,0) it is assumed that the eigenvalues of the linearized system, J,
are distinct, real and have opposite signs. That such is the case for Ramsey’s growth
model is explained in Barro and Sala-i-Martin (1995). Let µ and ξ be the two eigenvalues
of the Jacobian J and furthermore let µ be the positive eigenvalue and ξ be the negative.
Then, the positive eigenvalue is given by:
( ) 2/)(/42 ssss kfc ′′−+= γρρµ (3.2.23)
since at steady state,
βρ −′= )( sskf
The eigenvalues will be used in studying the impact on consumption and investment. 3.2.1 Impact on Consumption and Investment
We assume that the initial change in capital due to a perturbation ε, kε(t,0), is
bounded, which implies that a catastrophe causes a finite and bounded change in the
capital. In this case Kε(s) must be finite for all s>0, since Kε(s) is the Laplace transform of
kε(t,0). This again implies that when s = µ, Kε(µ) must be finite for any positive
eigenvalue, µ. However, when Eq. 3.2.22 is evaluated at s = µ, a singularity exists since
by definition of µ being an eigenvalue, µI – J, is a singular matrix. The matrix (sI – J)-1
can be written as
−
−−− 1121
1222
))((1
JsJJJs
ss ξµ (3.2.24)
In particular, the denominator is zero when s = µ. Therefore the only way for Kε(µ) to be
finite is that the following should be satisfied:
Chapter Three: Theoretical Models
84
{ }
=
++−−++−+
−
−
−
−
00
)()()}()()()({])()()()()()([/)0,0(
*2
*1
1*3
*
*2121
1*3
*
1121
1222
σσ
σε
σβσσγ
kAsHkLogsHkAsHkkLogsHsHsHkAsHcc
JsJJJs
(3.2.25)
This implies two conditions for cε(0,0); however, since µ is an eigenvalue, these
conditions are not independent thus giving a unique value of cε(0,0). Therefore:
{ }])()()()()()([/
/])()()}()()()({)[()0,0(*2
1211*
3*
12*
2*
11*
3*
11
kLogsHsHsHkAsHc
JkAsHkLogsHkAsHkJc
σσγ
σβµσ
σσε
++−−
++−−−−=−
−
(3.2.26)
Substituting the values for J11 (=0) and J21 (=-1), the following equation results:
{ }])()()()()()([/
])()()}()()()({[)0,0(*2
1211*
3*
*2
*1
1*3
*
kLogsHsHsHkAsHc
kAsHkLogsHkAsHkc
σσγ
σβµσ
σσε
++−−
++−−=−
−
(3.2.27)
Simplifying the above expression and substituting H3(µ) = e-τµ,
******2
***1
**1
]/[)](/)()[(
)]())()(1(/)[()()0,0(
kkcekfckfH
kLogkkLogHckfHc
βµγγµµ
µσµγµτµ
ε
−+−′++
++′=−
(3.2.28)
The Eq.3.2.28 gives the initial impact on consumption due to a sudden change in the
capital stock accompanied by changes in productivity. The expression tells us that
consumption decrease exponentially after the event (the third term), which is
accompanied by the normal depreciation of capital (the fourth term). This is in part offset
by increases in productivity during reconstruction represented by the first two terms. The
term γ/*c is the measure of coefficient of absolute constant risk aversion. The term
)( *kf ′ is the marginal product of capital at steady state. This measures the price of
capital at steady state. µµ)(H 1 measures the after-event change in productivity, that is,
discount the change in productivity at rate µ and multiply the result by µ. The expression
for µ, Eq.3.2.23, implies that it is greater than the pure rate of time preference, ρ, for
realistic values of crucial parameters. Since µ>ρ, µH1(µ) puts more weight on changes in
productivity immediately after the event relative to distant future changes than does
ρH(ρ). This implies that productivity changes decay rapidly relative to the utility
discount rate as the economy evolves away from the event date. This in turn implies that
Chapter Three: Theoretical Models
85
the initial effects of a catastrophe on the consumption, lasts for short spans after the
event. Numerical experiments, reported in a later section, indicate that the impact on
consumption lasts for a time when the changes in productivity have stabilized.
The initial change in capital kε(0,0), which is the impact on investment, can be
determined as shown below. Since
{ }( )( )
++−−+
++−+−
−−=
−
−
σσ
σε
ε
µµσµβµ
σµµµσµγ
ξµ
)()()}()()(/)({
)()()()()()([/)0,0(
))((1)(
*2
*1
1*3
*
*2121
1*3
*
kAHkLogHkAHks
kLogHHHkAHcc
sssK
(3.2.29)
Eq.3.2.29 can be written as:
{ }( )( )σσ
σε
ε
σβ
σσµγ
µξξµ
)()()}()()(/)({
)()()()()()([/)0,0(
)())((
*2
*1
1*3
*
*2121
1*3
*
2
kAsHkLogsHkAssHks
kLogsHsHsHkAHcc
sKss
++−−+
++−+−
=++−
−
− (3.2.30)
Taking limit of ∞→s and using the relations:
)0())0,0()((limit 2
s εεε ksksKs &=−∞→
(3.2.31)
)0()(limits εε kssK =
∞→ (3.2.32)
the following relation for initial investment can be obtained:
)0()())0()()()0(()0,0()0( 2*
1**
3* hkfhkLogkfhkcI +′+−−+−= βεε (3.2.33)
Since the assumed perturbations occur after the catastrophe (that occurs at time τ), the
functions, h1(0), h2(0), and h3(0) are zero. Eq.3.2.33 implies that initial investment occurs
to compensate for the changes in consumption.
The system of equations (Eq. 3.2.20) becomes a non-autonomous linear initial value
problem, which can be solved to yield a solution for kε(t,0) and cε(t,0). The procedure of
solving the initial value problem (IVP) is shown below.
( )
−−−
−−=
−−
=− −
skfcs
ssssadjs
1)(/
))((1
)det()( *''*
1 γρξµAI
AIAI (3.2.34)
Taking the inverse Laplace transform, the following results:
( )
−−−
−−=−= −−−
skfcs
ssLsLe
sst
1)(/
))((1 ''*
111 γρξµ
AIA (3.2.35)
Chapter Three: Theoretical Models
86
[ ] ( ) [ ]
( ) [ ] [ ]
++−−
−
−−
−−+−−=
tttt
tttt
t
eeee
eekfeee
ξµξµ
ξµξµ
ξξµµξµ
ξµγξρρµ
ξµ
)(11
)()()()(
1 *''
A (3.2.36)
The solution can be written as follows:
{ }[ ]{ } ds
kAthkLogthkAthkkLogthththkAthc
ee
kc
etktc
tst
t
∫
++−−++−
+
=
−
−
0*
1*
11*
3*
*2121
1*3
*
)()()()()()()()()()()()(/
)0()0(
)()(
σσ
σ
ε
ε
ε
ε
σβσσγAA
A
(3.2.37)
Eq. 3.2.37 gives the evolution of the consumption and capital stock when the economic
system 3.2.3-4 is perturbed by a catastrophic event. Numerical experiments on the
solution 3.2.37 are presented in Section 3.2.4.
3.2.2 Impact on Welfare
The overall welfare function can be written as:
∫∞
−=0
)),(()( dttcueU t εε ρ (3.2.38)
U(ε) represents the present value of overall welfare associated with different ε. One
measure of the impact of the catastrophe is dU/dε, which the change in the overall
welfare due to perturbations induced by catastrophe. This is calculated in the
neighborhood of the ε=0 paths. The change in U due to an infinitesimal change in ε is
∫∞
−=0
* )0,( dttcecddU t
ερ
ε (3.2.39)
( ) [ ]))()()()(())((
))(()(''
21**
3*
*1*
ρρσβρξρµργε
γ
HHkLogkfHkkfcddU
+++−−−
=+
(3.2.40)
The above expression (Eq. 3.2.40) can be used as a measure of the secondary effects of a
catastrophe. Since f’’(k)<0, ρ<µ, and γ<0, the first term of Eq.3.2.40 is positive. The
welfare change is negatively affected by the capital loss: - k*(H3(ρ)+β). This loss in
welfare can be compensated by appropriate reconstruction measures that boost the
Chapter Three: Theoretical Models
87
productivity. The changes in productivity are given by: f(k*)(σLog(k*)H1(ρ)+H2(ρ)) that
depends crucially on the steady state production level, f(k*). Steady state production level
means pre-event per capita output. H1(ρ) is the discounted (at rate ρ) change in the capital
share, σ, after the event. H2(ρ)is the discounted (at rate ρ) change in the technology, A,
after the event. Depending on the nature of these changes in productivity, the net overall
change in welfare may be controlled.
The following expressions (based on Taylor series expansion around the steady state)
could be used to determine the unit impulse response functions due to a catastrophe:
)0,()()1,( ** tckkftc εβ +−≅ (3.2.41)
)0,()1,( * tkktk ε+≅ (3.2.42)
3.2.4 Numerical Experiments
The Eq. 3.2.20 is solved numerically. The Mathematica© program is presented in
electronic form in Appendix C. The following Cobb-Douglas form of production function
is chosen:
σσρ kkf /)( = (3.2.43)
It is assumed that σ =0.25, and ρ = 0.04. ρ measures the time rate of preference or the
real interest rate, hence 4% is a reasonable value. σ measures the share of capital per unit
labor of 25% in the output, which is a reasonable assumption. It is also assumed that γ =
-0.5 and that time at which the catastrophe occurs, τ = 0.5. Productivity changes due to
the catastrophe are handled by the two functions. For changes in the capital share in the
production function, σ, the form of the function is
h1(t) = [(a4+b1)/(2-τ)t - (2a4+b1τ)/(last-τ)][UnitStep(t-τ) -UnitStep(t-τ)]
+ b1UnitStep(t-2) (3.2.44)
where, a4 = 0.05. The model is simulated for various values of b1. The plot for various
assumed functions simulating the changes in the capital share in the production function
is shown in Fig. 3.1a (b1 = 0.0, 0.05, 0.1, 0.15, 0.2). This type of change in productivity
may occur when the reconstructed capital replaces pre-event old and ill-maintained
capital stock. The reconstructed capital incorporates latest technology and hence causes
Chapter Three: Theoretical Models
88
Note: In all subsequent figures in this chapter time is measured in years, and the catastrophic
event occurs 0.5 units of time (years) after the system starts to evolve from its steady state. A bar
denotes the occurrence of an event.
Fig. 3.1a Changes in capital share in the production function
-10%
-5%
0%
5%
10%
15%
20%
25%
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5Time (in years)
Cap
ital s
hare
0% Change in capital share5%10%15%20%
Time of occurrence of the event
Fig. 3.1b Effect of loss intensity on initial consumption
-0.014-0.012-0.01
-0.008-0.006-0.004-0.002
00.0020.004
0 0.05 0.1 0.15 0.2 0.25
loss ratio
Initi
al c
onsu
mpt
ion
chan
ge
Chapter Three: Theoretical Models
89
permanent shifts in the capital share in the production function. The total factor
productivity (TFP) change is modeled using:
[ ])3())5.0(()sin()( 12 −−+−= tUnitSteptUnitSteptath τ (3.2.45)
The plot for the TFP change is shown in Fig. 3.2b. Capital reduction at time τ due to the
occurrence of a catastrophe is modeled using:
)()( 33 τδ −= tath (3.2.46)
It is assumed that 10 percent of the capital stock is destroyed or a3 = 0.1. The results are
shown in Fig. 3.1b to 1f.
Fig. 3.1b plots the initial consumption change required for a stable solution (Eq.3.2.28)
for various values of b1. The curve implies that as the after event capital share in the
production function increases, smaller initial consumption levels can ensure stability.
This implies that an economy with a lower pre-event consumption, for a given percentage
of capital loss, requires higher post-event capital share in its production function to
achieve stability. Conversely, a higher pre-event consumption level demands lesser
increase in post-event capital share for stability.
Fig. 3.1c shows the discontinuous drop in consumption when the event occurs. The
unanticipated change in income (output, Fig. 3.1e) due to sudden loss of capital causes a
drop in consumption and the empirical evidence presented in Chapter 4 supports this.
There is a drop of around 40% in consumption due to a 22% percent loss in capital stock.
It takes approximately 3 units of time to reach to its post-event stable consumption level,
which is 25%, less than its pre-event consumption level. From the graph (Fig. 3.1c) it is
clear that increase in post-event share of capital in production function increases the post-
event level of consumption.
Fig. 3.1d plots the evolution of capital after 22% destruction due to a catastrophic event.
It takes approximately 3 units of time to reach to its pre-event level. We see that the loss
of capital is compensated for the decrease in consumption level. The evolution of the
output of the economy is shown in Fig3.1e. The output clearly portrays the changes that
occur in the capital share. Increase in the TFP cause the output to rise at t=1. The output
traces this increase till it lasts (t=3). Greater the rise in TFP, greater is the output.
Chapter Three: Theoretical Models
90
Fig 3.1c Evolution of consumption
0%
1%
2%
3%
4%
5%
6%
7%
8%
9%
0 0.5 1 1.5 2 2.5 3 3.5
Time
Con
sum
ptio
n
0% Change in capital share5%10%15%20%
Fig 3.1d Changes in capital assets
0.10
0.12
0.14
0.16
0.18
0.20
0.22
0 0.5 1 1.5 2 2.5 3 3.5
Time
Prod
uctiv
e ca
pita
l
0% Change in capital share5%10%15%20%
Chapter Three: Theoretical Models
91
Fig 3.1e Evolution of output
0.6
0.61
0.62
0.63
0.64
0.65
0.66
0.67
0.68
0.69
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
Out
put
0% Change in capital share5%10%15%20%
3.1f Changes in economic growth over time
-2.0%-1.5%-1.0%
-0.5%0.0%0.5%1.0%1.5%2.0%
2.5%3.0%3.5%
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
Econ
omic
gro
wth
0% Change in capital share5%10%15%20%
Chapter Three: Theoretical Models
92
Growth (Fig. 3.1f) increases immediately after the loss in capital, consistent with the
standard growth model prediction that lesser the output, higher is the growth. The growth
then drops till t=1, where the rise in total TFP boosts growth rate. The growth rate then
continues to fall till t=3, when the changes in TFP cease. The kink at t=2, is the point
where the changes in capital share of production stabilize. After all the changes in the
production function has stabilized, at t=3, greater increases in productivity result in
greater post event growth rates.
After studying the impact of changes in productivity on the post-event behavior of an
economy, the impact of changes in the capital loss is studied. It is assumed that
a2=b1=0.1 (Fig.3.2a and 3.2b) but a3 varies from 0.0 to 0.2 simulating various levels of
capital loss. Greater the loss, lower the level of consumption (Fig. 3.2c) and capital (Fig.
3.2d). It takes around 3 units of time for the economy to attain stable solution (which is
the middle path in the five paths shown). Fig. 3.2e shows the evolution of output. Greater
the losses, lower are the output. The output grows after the event. At t=1 there is a
noticeable kink due to increase in the TFP. The output grows till it stabilizes around t=3,
when all the changes in productivity have stabilized.
Growth is plotted in Fig. 3.2f. The results show that greater the loss higher the growth
rate. This does not change even after all the changes in productivity have stabilized.
Empirical evidence from post event growth rates indicate that they are negatively related
to losses, i.e. greater the loss lower is the post event growth rate. The model is not able to
explain this empirical observation. The second model presented in the next section helps
us to better understand this empirical observation.
Fig. 3.2g gives a phase space portrait of the evolution of consumption and capital after a
catastrophe. Fig. 3.2h plots the changes in the welfare due to the occurrence of a
catastrophe. The plot indicates that a greater loss in capital results in greater loss in
welfare. This change in welfare due to occurrence of a catastrophe can be used as a
measure of the secondary impact.
Chapter Three: Theoretical Models
93
Fig. 3.2a Changes in capital share in the production function
-6%
-4%
-2%
0%
2%
4%
6%
8%
10%
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
Cap
ital s
hare
Fig 3.2b Changes in total factor productivity with time
0
0.02
0.04
0.06
0.08
0.1
0.12
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Chapter Three: Theoretical Models
94
Fig 3.2c Evolution of consumption
0.00
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0 0.5 1 1.5 2 2.5 3 3.5
Time
Con
sum
ptio
n
Loss ratio = 0%5%10%15%20%
Fig 3.2d Changes in capital assets
0.10
0.12
0.14
0.16
0.18
0.20
0.22
0.24
0 0.5 1 1.5 2 2.5 3 3.5
Time
Prod
uctiv
e ca
pita
l
Loss ratio = 0%5%10%15%20%
Chapter Three: Theoretical Models
95
Fig 3.2e Evolution of output
0.6
0.62
0.64
0.66
0.68
0.7
0.72
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
Out
put
Loss ratio = 0%5%10%15%20%
3.2f Changes in economic growth over time
-3%
-2%
-1%
0%
1%
2%
3%
4%
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
Econ
omic
gro
wth
Loss ratio = 0%5%10%15%20%
Chapter Three: Theoretical Models
96
Fig. 3.2g How does consumption vary with changes in capital?
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24Capital
Con
sum
ptio
n
Loss ratio = 0%5%10%15%20%
Fig3.2h Changes in welfare
-0.5-0.48-0.46
-0.44-0.42-0.4
-0.38-0.36
-0.34-0.32-0.3
0 0.2 0.4 0.6 0.8 1
Loss ratios
Pres
ent v
alue
of u
tility
Chapter Three: Theoretical Models
97
3.3 Model including the effects of efficiency of post-event reconstruction
In an economy, construction activity of some kind is almost always going on.
Capital in an economy can therefore be classified into two categories – the maturing
capital and the productive capital. The capital that is being constructed does not become
immediately productive. For example, it takes time to build a bridge. It is crucial to
model the effects of speed with which the maturing capital becomes productive capital.
The model described in the previous section does not model the effects of post event
construction. Typically, after a catastrophic earthquake, productive capital including
buildings and infrastructure will be damaged or destroyed. Reconstruction activities start
some time after an event. With the reconstruction efforts gaining momentum, the
conversion of maturing capital to productive capital may temporarily exceed its normal
values for a period of time depending on the inflow of investment in the affected region.
The model described below tries to simulate this behavior of an economy.
3.3.1 Model
The representative agent will choose a consumption path, c(t), capital
accumulation, k2(t) such that he maximizes his utility:
∫∞
−=0
)(0 )(max)( dtcuekV t
tc
ρ (3.3.1)
such that:
121 )( kckfk α−−=& (3.3.2)
212 kkk βα −=& (3.3.3)
where k1 is the maturing capital and k2 is the productive capital. Eq.3.3.2 states that the
maturing capital grows from investment (f(k2) ) but is partly offset by the consumption (c)
and partly by conversion into productive capital (αk1). Eq. 3.3.3 states that the productive
capital grows depending on a portion of the maturing capital (αk1) but the growth is
negatively affected by the depreciation of the productive capital (βk2). To study the
consequences of a catastrophe on an economy described in Eqs. 3.3.1-3, it is assumed
Chapter Three: Theoretical Models
98
that the economy initially is at equilibrium with the steady state values of capital given by
k*1 and k*
2 and consumption given by c*.
The following form of the production function is assumed – the Cobb-Douglas form:
σ2)( skkf = (3.3.4)
The current value Hamiltonian for Eqs.3.3.1-3 is given by
( ) ( )212121 )()( kkkckfcuH βαλαλ −+−−+= (3.3.5)
λi(t) (i=1,2) can be interpreted in simple economic terms as shadow prices. λ1(t) is the
change in the maximum attainable value of the objective function discounted to t, when
the economy has to acquire a single unit of maturing capital at time t. Similarly, λ2(t) is
the change in the maximum attainable value of the objective function discounted to t,
when the economy has to acquire a single unit of productive capital at time t.
The resulting equilibrium solves the system of differential equations:
211 )( αλλαρλ −+=& (3.3.6)
2122 )()(' λβρλλ ++−= kf& (3.3.7)
0)('0 1 =−⇒= λcuH c (3.3.8)
Assuming the utility function as in Eq.2.2 the following relationship is obtained: γλ /1
1=c (3.3.9)
A catastrophe is modeled by a discontinuous change in capital, for one period after the
occurrence of an event at time τ. This is accompanied by a change in productivity
associated with post-event reconstruction. To model these changes due to occurrence of a
catastrophe the following perturbation equations are used:
))(1( 1 thεαα +→ (3.3.10a)
)(2 thaidexternal ε→ (3.3.10b)
)(3 thεββ +→ (3.3.10c)
))(1( 4 thεσσ +→ (3.3.10d)
Eq. 3.3.10a simulates the changes in the rate at which maturing capital is converted to
productive capital. Eq.3.3.10b models the flow of external aid. The discontinuous
Chapter Three: Theoretical Models
99
reduction in capital stock after an event is modeled by assuming a Dirac Delta function
for h3(t) in Eq. 3.3.10c. Eq.3.3.10d models the productivity due to a change to the capital
share in the production function.
The perturbed system of differential equations can be written as follows:
)(),())((),(),(),( 211/1
1))(1(
214 thtkthttsktk th εεεαελεε γεσ ++−−= +& (3.3.11)
),())((),())((),( 23112 εεβεεαε tkthtkthtk +−+=& (3.3.12)
),())((),()))(1((),( 21111 ελεαελεαρελ tthttht +−++=& (3.3.13)
),())((),(),())(1(),( 2211))(1(
2424 ελτεδβρελεεσελ εσ ttttkthst th −−−++−= −+& (3.3.14)
Differentiating the above system with respect to ε, and evaluating at ε=0, the following
system of equations result:
+−−
−+−
+
+−+
−−−
=
−
*1
*24
*2
*23
*2
*11
*231
*1
21*14
*2
2
1
2
1
*2
*12
1/1*1
*2
2
1
2
1
))(1)(()(')())((
)()()()()()(
)0,()0,()0,()0,(
)()(')("0)(00
000)(/1)('
)0,()0,()0,()0,(
λσλλλα
αασ
λλ
βρλααρ
βαλγα
λλ
ε
ε
ε
εγ
kLogthkfthth
kththkththkthkf
tttktk
kfkf
kf
tttktk
ss&
&
&
&
(3.3.15)
Performing the Laplace transform:
+−+−+
−++−+
−=
ΛΛΚΚ
−
*1
*24
*2
*232
*2
*111
*231
*12
21*14
*21
1
2
1
2
1
))(1)(()(')()0,0())(()0,0(
)()()0,0()()()()()0,0(
)(
)()()()(
λσλλλλαλ
αασ
ε
ε
ε
ε
ε
ε
ε
ε
kLogthkfthth
kththkkththkthkfk
s
ssss
JI
&
&
&
&
(3.3.16)
Let the positive eigenvalue of J be µ. Then the conditions that k1(t), k2(t) < ∞ as t → ∞,
k1ε(0,0) = k2ε(0,0) = 0, and the singularity of the matrix (sI-J) at s = µ imply that:
Chapter Three: Theoretical Models
100
=
+−+−+
−+−
−
0000
))(1)(()(')()0,0())(()0,0(
)()()()()()()(
)(
*1
*24
*2
*232
*2
*111
*231
*1
21*14
*2
*2
λσµλµλλλµαλ
µµαµµαµσ
ε
ε
kLogHkfHH
kHHkHHkHkLogkf
sadj JI (3.3.17)
Eqs.3.3.17 can be solved to obtain the initial conditions λ1ε(0,0) and λ2ε(0,0). The
Eqs.3.3.15 can then be solved using standard procedures (Appendix ). The solutions so
derived will be used to explain the relevance of conversion of maturing capital into
productive capital in the post-event dynamics.
3.3.2 Numerical Experiments
The following Cobb-Douglas from of production function is chosen:
σskkf =)( (3.3.18)
It is assumed that σ =0.25, and ρ = 0.04. It is also assumed that γ = -0.5 and that time at
which the catastrophe occurs, τ = 0.5. Productivity changes due to the catastrophe are
handled by the function in Eq. 3.3.10d for changes in the capital share in the production
function, σ. The flow of external aid is modeled using (Fig 3b):
[ ])3())5.0(()sin()( 22 −−+−= tUnitSteptUnitSteptath τ (3.3.19)
A similar function is used to model the changes in rate at which maturing capital is
converted to productive factor:
[ ])3())5.0(()sin()( 11 −−+−= tUnitSteptUnitSteptath τ (3.3.20)
The Mathematica© program for simulating these equations is presented in electronic
form in Appendix D. The plot for various assumed functions simulating the changes in
the capital share in the production function is shown in Fig. 3.3a (a1 = 0.0, 0.05, 0.1, 0.15,
0.2). This change in conversion from maturing capital to productive capital occurs based
on the efficiency of the reconstruction process. More efficient the reconstruction, higher
the rate at which reconstructed capital becomes productive. Capital reduction at time τ
due to the occurrence of a catastrophe is modeled using Eq.3.3.10c, where it is assumed
that 10 percent of the capital stock is destroyed. Changes in capital share are modeled
using Eq.3.3.10d (Fig.3b). The results are shown in Figs. 3.3c to 3g.
Chapter Three: Theoretical Models
101
Fig. 3.3a Changes in total factor productivity
0%
5%
10%
15%
20%
25%
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
Cap
ital s
hare
0% change in TFP5%10%15%20%
Fig. 3.3b Changes in aid and capital share in the production function
-6%
-5%
-4%-3%
-2%
-1%
0%1%
2%
3%
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
Cap
ital s
hare
aid
capital share
Chapter Three: Theoretical Models
102
Fig. 3.3c plots evolution of maturing capital after the event. The maturing capital grows
slower after the catastrophe, until there is a change in the conversion factor α at t= 0.5.
Greater the change in conversion factor lesser is the level of maturing capital. Fig. 3.3d
plots the evolution of the productive capital. The productive capital is higher if the
conversion factor is higher as can be seen in Fig. 3.3d.
Fig. 3.3e shows the drop in consumption after the event. But unlike Figs. 3.1c or 3.2c,
where there is a discontinuous drop in consumption, Fig. 3.3e shows that the
consumption level drops and remains constant after the event until there is a change in
the conversion factor. Lower the conversion factor lower is the consumption. It takes
approximately 3 units of time to reach to its post-event stable consumption level, which is
7% point less than its pre-event consumption level.
Fig. 3.3d plots the evolution of productive capital after 15% destruction due to a
catastrophic event. It takes approximately 3 units of time to reach its stable path. There is
a permanent increase in capital by 7%. Thus, increase in capital is compensated by the
decrease in consumption level. The evolution of the output of the economy is shown in
Fig. 3.3f. The output clearly portrays the changes that occur in the capital. Greater the
rise in conversion factor, lesser is the output beyond t=3, when changes in productivity
stabilize.
Growth (Fig. 3.3g) increases immediately after the loss in capital, consistent with the
standard growth model prediction that lesser the output, higher is the growth. The growth
increases till t=1, where the rise in conversion factor boosts growth rate. Greater the
conversion factor higher is the growth rate. When the change in conversion factor stops
(t=2), the growth rate falls. It then stabilizes at t=3. After all the changes in the
production function has stabilized, at t=3, greater increases in conversion factor result in
greater post event growth rates.
Chapter Three: Theoretical Models
103
Fig 3.3c Changes in maturing capital
0.00
0.05
0.10
0.15
0.20
0.25
0 0.5 1 1.5 2 2.5 3 3.5
Time
Mat
urin
g ca
pita
l
Conversion rate 0%5%10%15%20%
Fig 3.3d Changes in productive capital
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0 0.5 1 1.5 2 2.5 3 3.5
Time
Prod
uctiv
e ca
pita
l
Conversion rate 0%5%10%15%20%
Chapter Three: Theoretical Models
104
Fig 3.3e Evolution of consumption
-0.70
-0.60
-0.50
-0.40
-0.30
-0.20
-0.10
0.00
0 1 2 3 4 5 6
Time
Con
sum
ptio
n
Conversion rate 0%5%10%15%20%
Fig 3.3f Evolution of output
1.09
1.1
1.11
1.12
1.13
1.14
1.15
1.16
1.17
1.18
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time
Out
put
Conversion rate 0%5%10%15%20%
Chapter Three: Theoretical Models
105
3.3g Changes in economic growth over time
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
3.5%
4.0%
4.5%
0 1 2 3 4 5 6
Time
Econ
omic
gro
wth
Conversion rate 0%5%10%15%20%
Chapter Three: Theoretical Models
106
After studying the impact of changes in productivity on the post-event behavior of an
economy, the impact of changes in the capital loss is studied. It is assumed that a1=0.1
(Fig. 3.4a). Changes in the capital share and the external aid are shown in Fig. 3.4b. a3
varies from 0.0 to 0.2 simulating various levels of capital loss. Greater the loss, lower is
the consumption (Fig. 3.4e) and capital (Fig. 3.4c and d). It takes around 3 units of time
for the economy to attain stable solution (which is the middle path in the five paths
shown). Fig. 3.4f shows the evolution of output. Greater the losses, lower is the output.
The output grows after the event. The output grows till it stabilizes around t=3, when all
the changes in productivity have stabilized.
Growth is plotted in Fig. 3.4g. The results show that greater the loss, higher the growth
rate, immediately after the event. However this changes soon after the changes in the
conversion factor. After t= 1.2, greater loss result in lesser growth rates. This behavior
continues till all the changes in productivity have stabilized and the conversion rate has
returned to its normal level. This concurs with the empirical evidence of post event
growth rates. The evidence indicates that losses are negatively associated with post event
growth rates, i.e. greater the loss, lower is the post event growth rate. Unlike the previous
model, the second model able to explain this empirical observation. The results point to
the importance of modeling two types of capital in an economy – the maturing and the
productive capital and the changes in the conversion process that typically follow after a
catastrophic event.
Fig. 3.4h plots the initial changes in consumption required for stable post event behavior.
The curve implies that the greater the loss, smaller should be the initial consumption
levels to ensure stability. Fig. 3.4h plots the changes the welfare due to the occurrence of
a catastrophe. The plot indicates that a greater loss in capital results in greater loss in
welfare. This change in welfare due to occurrence of a catastrophe can be used as a
measure of the secondary impact and is consistent with the results of the previous model.
Chapter Three: Theoretical Models
107
Fig. 3.4a Changes in capital share in the production function
0.0%
0.5%
1.0%
1.5%
2.0%
2.5%
3.0%
0 1 2 3 4 5
Time
Cap
ital s
hare
Fig. 3.4b Changes in aid and total factor productivity
-6%
-4%
-2%
0%
2%
4%
6%
8%
10%
12%
0 1 2 3 4 5
Time
Cap
ital s
hare
Chapter Three: Theoretical Models
108
Fig 3.4c Changes in maturing capital
-0.60
-0.40
-0.20
0.00
0.20
0.40
0.60
0.80
0 0.5 1 1.5 2 2.5 3 3.5
Time
Mat
urin
g ca
pita
lLoss ratio = 0.05%10%15%20%
Fig 3.4d Changes in productive capital
-0.40
-0.30
-0.20
-0.10
0.00
0.10
0.20
0.30
0.40
0.50
0 0.5 1 1.5 2 2.5 3 3.5
Time
Prod
uctiv
e ca
pita
l
Loss ratio = 0.0%5%10%15%20%
Fig 3.4e Evolution of consumption
-1.50
-1.00
-0.50
0.00
0.50
1.00
0 0.5 1 1.5 2 2.5 3 3.5
Time
Con
sum
ptio
n
Loss ratio = 0.0%5%10%15%20%
Chapter Three: Theoretical Models
109
Fig 3.4f Evolution of output
1.04
1.06
1.08
1.1
1.12
1.14
1.16
1.18
0 0.5 1 1.5 2 2.5 3
Time
Out
put
Loss ratio = 0.0%5%10%15%20%
3.4g Changes in economic growth over time
-2%
-1%
0%
1%
2%
3%
4%
5%
0 0.5 1 1.5 2 2.5 3
Time
Econ
omic
gro
wth
Loss ratio = 0.0%5%10%15%20%
Fig. 3.4h Effect of loss intensity on initial consumption and welfare
-0.35
-0.3
-0.25
-0.2
-0.15
-0.1
-0.05
0
0.05
0.1
0 0.05 0.1 0.15 0.2 0.25
loss ratio
WelfareInitialConsumption
Chapter Three: Theoretical Models
110
3.4 Interacting Regions
The model in this section simulates the behavior of two interacting regions when
a catastrophe strikes one of the regions. Region 1 is affected by a catastrophe. Region 1
interacts with Region 2 by exporting and importing goods. After a catastrophe strikes
Region 1, its productive capacity decreases. Region 2 tries to mitigate the situation in
Region 1 by diverting some of its output for relief and reconstruction to Region 1. After a
period of time, due to the construction of new capital in Region 1, its productive capacity
may increase. As a result, Region 1 is able to export more to Region 2. This situation
describes some of the dynamics when the Northridge Earthquake struck Los Angeles
County. As will be clear from the data and simulation presented in a later chapter, the
economy of Los Angeles did better after the Northridge Earthquake. In cases where there
is no appreciable change in the productive capacity, the economy may reach its pre-event
output levels. The model presented herein explains why and under what condition
economies revive and often do better after a catastrophe.
In the following the subscript 1 will denote variables belonging to Region 1 and subscript
2 denotes variables belonging to Region 2. The representative agents in both the regions
will choose consumption paths c1(t) and c2(t) such that they maximize their utility:
∫∞
− +=0
21)(),(0 ))()((max)(21
dtcucuekV t
tctc
ρ (3.4.1)
subject to the following conditions:
111121111 )()( kckfakfak β−−+=& (3.4.2)
221221212 )()( kckfakfak β−−+=& (3.4.3)
The current value Hamiltonian is:
))()(())()(()()(
221221212
11112111121
kckfakfakckfakfacucuHβλβλ
−−++−−+++=
(3.4.4)
λ1(t) is the change in the maximum attainable value of the objective function discounted
to t, when Region 1’s economy acquires a single unit of its capital at time t. Similarly,
Chapter Three: Theoretical Models
111
λ2(t) is the change in the maximum attainable value of the objective function discounted
to t, when Region 2’s economy acquires a single unit of its capital at time t.
Differentiating the Hamiltonian with respect to c1 and c2, the following first order
conditions result:
0)('00)('0 2211 21=−⇒==−⇒= λλ cuHandcuH cc (3.4.5)
Differentiating the relations in Eq.3.4.5 with respect to time we have:
222111 )(")(" ccuandccu &&&& == λλ (3.4.6)
The co-state equations are:
)('))(( 12121'
11111 kfakfa λβλρλλ −−−=& (3.4.7)
))('()( 22222'
12122 βλλρλλ −−−= kfakfa& (3.4.8)
Using the relations in Eq.3.4.6 in Eqs. 3.4.7 and 3.4.8, the following equations result:
)(')]([ 12111
21
'11
11 kfa
cckfacc −−−+= γ
γ
γβρ
γ& (3.4.9)
)(')]([ 21212
12
'22
22 kfa
cckfacc −−−+= γ
γ
γβρ
γ& (3.4.10)
The following perturbations are introduced to model the behavior of the economies after
a catastrophic event in Region 1.
))(1( 1 thεσσ +→ (3.4.11a)
))(1( 21212 thaa ε+→ (3.4.11b)
))(1( 22121 thaa ε−→ (3.4.11c)
)(3 thεββ +→ (3.4.11d)
Eq.3.4.11d models the fact that when a catastrophic event occurs in Region 1, there is a
change in its capital. This is followed by an increase in the amount that Region 1 imports
from Region 2, which is modeled by Eqs.3.4.11b-c. Reconstruction in Region 1 changes
the capital share coefficient of Region 1’s production function, which is modeled in Eq.
3.4.11a. Substituting the relations in Eqs.3.4.2-3 and Eqs.3.4.9-10, the following
perturbed system of equations result:
Chapter Three: Theoretical Models
112
),())((),()),(())(1(),(),( 1212112))(1(
11111 εεβεεεεε εσ tkthtctkfthatskatk th +−−++= +&
(3.4.12)
),(),()),(())(1(),(),( 222122))(1(
12121 εβεεεεε εσ tktctkfthatskatk th −−−+= +& (3.4.13)
1))(1(11211
1
2
1))(1(11113
11
1
1
),())(1(),(),(
]),())(1()([),(),(
−+−
−+
++
+−++=
th
th
tkthsatctc
tkthsathtctc
εσγ
γ
εσ
εεσεγε
εεσεβργε
ε&
(3.4.14)
122121
2
1
12222
22
),())(1(),(),(
]),())(1([),(),(
−−
−
++
−−+=
σγ
γ
σ
εσεεγε
εεσβργε
ε
tksthatctc
tkthsatctc& (3.4.15)
Differentiating the above system (Eqs.3.4.12-15) with respect to the perturbation ε
around the steady state, the following equations result:
dJ +
=
)0,()0,()0,()0,(
)0,()0,()0,()0,(
2
1
2
1
2
1
2
1
tktktctc
tktktctc
ε
ε
ε
ε
ε
ε
ε
ε
&
&&
&
(3.4.16)
where the elements of the matrix J and vector d are given below:
)])()1()(([1*2
*1
2111*111
γγβργ c
caakfJ −+′−+= (3.4.17a)
γ))(( *2
*1*
12112 cckfaJ ′−= (3.4.17b)
])([)( 1*221
*111
*1
13 *1
*2 −+
′′−= γ
γ cccacakfJ (3.4.17c)
014 =J (3.4.17d)
γ))(( *2
*1*
21221 cckfaJ ′−= (3.4.18a)
)])()1()(([1*2
*1
1222*222
γγβργ c
caakfJ −+′−+= (3.4.18b)
023 =J (3.4.18c)
])([)( 1*112
*222
*2
24 *2
*1 −+
′′−= γ
γ cccacakfJ (3.4.18d)
Chapter Three: Theoretical Models
113
131 −=J (3.4.19a)
032 =J (3.4.19b)
β−′= )( *1133 1
kfaJ (3.4.19c)
)( *1234 2
kfaJ ′= (3.4.19d)
041 =J (3.4.20a)
142 −=J (3.4.20b)
)( *2143 2
kfaJ ′= (3.4.20c)
β−′= )( *2244 2
kfaJ (3.4.20d)
The components of the vector d (Eq.3.4.16) are shown below.
))())((1()()( 1*221
*111
*1
*1
1
*1
1 *1
*2 −++
′−= γσ
γγ cccacakLogkfthcd (3.4.21a)
))()(()( 1*
112*2221
*2
2 *2
*1 −+
′= γ
γ cccacath
kfd (3.4.21b)
)()()()()()( *11
*111
*132
*2123 kLogthkfakththkfad σ+−= (3.4.21c)
)()()()()( 2*222
*11
*1214 thkfakLogthkfad −= σ (3.4.21d)
Performing the Laplace transform:
++++
−=
ΚΚ
−
42
31
22
11
1
2
1
2
1
)0,0()0,0()0,0()0,0(
)(
)()()()(
dkdkdcdc
s
sssCsC
ε
ε
ε
ε
ε
ε
ε
ε
JI
&
&
&
&
(3.4.22)
3.4.1 Numerical Experiments
The following Cobb-Douglas from of production function is chosen: σskkf =)( (3.4.23)
where it is assumed that σ =0.25, ρ = 0.04. It is also assumed that γ = -0.5 and that time at
which the catastrophe occurs, τ = 0.5. Productivity changes due to the catastrophe are
Chapter Three: Theoretical Models
114
handled by the function in Eq. 3.2.44 for changes in the capital share in the production
function, σ. The flow of external aid is modeled using:
[ ])3())5.0(()sin()( 22 −−+−= tUnitSteptUnitSteptath τ (3.4.24)
A similar function is used to model the changes in rate at which maturing capital is
converted to productive factor:
[ ])3())5.0(()sin()( 11 −−+−= tUnitSteptUnitSteptath τ (3.4.25)
Capital reduction at time τ due to the occurrence of a catastrophe is modeled using
Eq.2.46, where it is assumed that 5 percent of the capital stock is destroyed. The
simulation is done for various values of a1 = 0.0, 0.05, 0.1, 0,15, 0.2. The Mathematica©
program for simulating these equations is presented in electronic form in Appendix E.
The results are shown in Figs. 3.5a to 5j. Fig. 3.5a shows the discontinuous drop in consumption for Region 1 when the event
occurs. There is a drop of around 5% in consumption due to a 5% percent loss in capital
stock. It takes approximately 3 units of time to reach to its post-event stable consumption
level, which is 3% less than its pre-event consumption level. Fig.5b plots the evolution of
consumption in Region 2 after a catastrophic event has occurred in Region 1.
Consumption levels keep rising in Region 2 until it starts giving some of its output to
Region 1 for reconstruction. The consumption in Region 2 falls for the period when it
diverts its output (t=2) after which consumption rises again to reach its pre-event level at
t=3.
Fig. 3.5c plots the evolution of capital in Region 1 after a catastrophic event. It takes
approximately 3 units of time to reach to a stable level. The post event stable level of
capital is higher than the pre-event level. The increase in capital is compensated for by
the decrease in consumption level. The capital in Region 2 (Fig. 3.5d) reaches it pre-
event level at t= 4 after some changes. The phase portraits of the evolution of capital and
consumption in Regions 1 and 2 are shown in Figs. 3.5g and 5f, respectively. The
evolution of the output of the economy of Regions 1 and 2 are shown in Fig.5g and 5i,
respectively. The output in Region 1 stabilizes at t=3 above its prevent level. Growth rate
Chapter Three: Theoretical Models
115
(Fig. 3.5h) of Region 1 rises sharply after the event and the drops only to be raised by
increase of flow of funds from Region 2. Greater the inflow, greater is the growth.
Fig 3.5a Evolution of consumption in the disaster region
-5.00%
-4.00%
-3.00%
-2.00%
-1.00%
0.00%
1.00%
0 1 2 3 4 5 6 7 8
Time
Con
sum
ptio
n in
dis
aste
r reg
ion
Increase in TFP of affected region: 0%5%10%15%20%
Fig 3.5b Evolution of consumption in the adjacent region
-6.00%
-5.00%
-4.00%
-3.00%
-2.00%
-1.00%
0.00%
1.00%
2.00%
3.00%
0 1 2 3 4 5 6 7 8
Time
Con
sum
ptio
n in
adj
acen
t reg
ion Increase in TFP of affected region: 0%
5%10%15%20%
Chapter Three: Theoretical Models
116
Fig 3.5c Changes in capital in the affected region
-0.06
-0.04
-0.02
0.00
0.02
0.04
0.06
0.08
0.10
0 1 2 3 4 5 6 7 8
Time
Cap
ital i
n th
e af
fect
ed re
gion
Increase in TFP of affected region: 0%5%10%15%20%
Fig 3.5d Changes in capital of the unaffected region
-0.20
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
0 1 2 3 4 5 6 7 8
Time
Cap
ital i
n th
e un
affe
ctec
d re
gion
Increase in TFP of affected region: 0%5%10%15%20%
Chapter Three: Theoretical Models
117
Fig. 3.5e How does consumption vary with changes in capital of the affected region?
-0.05
-0.04
-0.03
-0.02
-0.01
0
0.01
-0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1
Capital in disaster region
Con
sum
ptio
n in
dis
aste
r reg
ion
Increase in TFP of affected region: 0%5%10%15%20%
Fig. 3.5f How does consumption vary with changes in capital of the unaffected region?
-0.06
-0.05
-0.04
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2
Capital in unaffected region
Con
sum
ptio
n in
una
ffec
ted
regi
on
Increase in TFP of affected region: 0%5%10%15%20%
Chapter Three: Theoretical Models
118
Fig 3.5g Output in the disaster region
0.78
0.79
0.80
0.81
0.82
0.83
0.84
0.85
0 1 2 3 4 5 6 7 8
Time
Out
put i
n di
sast
er re
gion
Increase in TFP of affected region: 0%5%10%15%20%
Fig 3.5h Economic growth in the disaster region
-1.00%
-0.50%
0.00%
0.50%
1.00%
1.50%
2.00%
2.50%
3.00%
3.50%
0 1 2 3 4 5 6 7 8
Time
Gro
wth
in d
isas
ter r
egio
n
Increase in TFP of affected region: 0%5%10%15%20%
Chapter Three: Theoretical Models
119
Fig 3.5i Output in the adjacent region
1.23
1.241.24
1.251.25
1.261.26
1.271.27
1.28
0 1 2 3 4 5 6 7 8
Time
Out
put i
n th
e ad
jace
nt re
gion Increase in TFP of affected region: 0%
5%10%15%20%
Fig 3.5j Economic growth in the adjacent region
-3.00%
-2.50%
-2.00%
-1.50%
-1.00%
-0.50%
0.00%
0.50%
1.00%
0 1 2 3 4 5 6 7 8
Time
Gro
wth
in th
e ad
jace
nt re
gion
Increase in TFP of affected region: 0%5%10%15%20%
Chapter Three: Theoretical Models
120
Growth of Region 2’s economy (Fig. 3.5j) increases after the event but drops after
Region 2 starts diverting its funds to Region 1. There is a sharp change in growth rate
once Region 2 stops diverting its funds to Region 1 at t=2. The growth then falls down
and reaches a lower value than its pre-event level after t=3.
After studying the impact of changes in productivity on the post-event behavior of an
economy, the impact of changes in the capital loss is studied. It is assumed that there are
no changes to productivity and that the aid from region 2 is constant. Various levels of
capital loss are modeled by varying a3 from 0.0 to 0.2. The levels of consumption in
Region 1 (Fig. 3.6a) and capital (Fig. 3.6c) are lower greater the loss. It takes around 3
units of time for the economy to attain stable solution (which is the middle path in the
five paths shown). The levels of consumption in Region 2 (Fig. 3.6b) and capital (Fig.
3.6d) are higher, greater the loss in Region 1. It takes around 3 units of time for the
economy to attain stable solution (which is the middle path in the five paths shown). The
phase portraits of the evolution of capital and consumption in Regions 1 and 2 are shown
in Figs. 3.6e and 6f, respectively. The evolution of the output of the economy of Regions
1 and 2 are shown in Fig.6g and 6i, respectively. The output in Region 1 stabilizes at t=3
above its pre-event level. Growth rate (Fig. 3.6h) of Region 1 rises sharply after the event
and the drops only to be raised by increase of flow of funds from Region 2. Greater the
loss, greater is the growth. This does not change even after all the changes in productivity
have stabilized. Growth of Region 2’s economy (Fig. 3.6j) increases after the event but
drops after Region 2 starts diverting its funds to Region 1. There is a sharp change in
growth rate once Region 2 stops diverting its funds to Region 1 at t=2. The growth then
falls down and reaches a lower value than its pre-event level after t=3. Greater the loss in
Region 1 greater is the post-event growth rate. This is due to the fact that unlike model 2,
the effects of the efficiency of the reconstruction process has not been modeled for the
affected Region 1.
Fig. 3.6k plots the changes the welfare due to the occurrence of a catastrophe for Regions
1 and 2 respectively. The plot indicates that a greater loss in capital results in greater loss
in welfare for both the regions.
Chapter Three: Theoretical Models
121
Fig 3.6a Evolution of consumption in the disaster region
-0.10
-0.08
-0.06
-0.04
-0.02
0.00
0.02
0.04
0 1 2 3 4 5 6 7 8
Time
Con
sum
ptio
n in
dis
aste
r reg
ion
Loss ratio = 0.05%10%15%20%
Fig 3.6b Evolution of consumption in the adjacent region
-0.05
-0.04
-0.03
-0.02
-0.01
0.00
0.01
0.02
0.03
0 1 2 3 4 5 6 7 8
Time
Con
sum
ptio
n in
adj
acen
t reg
ion Loss ratio = 0.0
5%10%15%20%
Chapter Three: Theoretical Models
122
Fig 3.6c Changes in capital in the affected region
-0.15
-0.10
-0.05
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0 1 2 3 4 5 6 7 8
Time
Cap
ital i
n th
e af
fect
ed re
gion
Loss ratio = 0.05%10%15%20%
Fig 3.6d Changes in capital of the unaffected region
-6.0%
-5.0%
-4.0%
-3.0%
-2.0%
-1.0%
0.0%
1.0%
2.0%
3.0%
4.0%
0 1 2 3 4 5 6 7 8
Time
Cap
ital i
n th
e un
affe
ctec
d re
gion
Loss ratio = 0.05%10%15%20%
Chapter Three: Theoretical Models
123
Fig. 3.5e How does consumption vary with changes in capital of the affected region?
-0.1
-0.08
-0.06
-0.04
-0.02
0
0.02
0.04
-0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.25 0.3
Capital in disaster region
Con
sum
ptio
n in
dis
aste
r reg
ion
Loss ratio = 0.05%10%15%20%
Fig. 3.5f How does consumption vary with changes in capital of the unaffected region?
-0.05
-0.04
-0.03
-0.02
-0.01
0
0.01
0.02
0.03
-0.06 -0.04 -0.02 0 0.02 0.04
Capital in unaffected region
Con
sum
ptio
n in
una
ffec
ted
regi
on
Loss ratio = 0.05%10%15%20%
Chapter Three: Theoretical Models
124
Fig 3.6g Output in the disaster region
0.70
0.75
0.80
0.85
0.90
0.95
0 1 2 3 4 5 6 7 8
Time
Out
put i
n di
sast
er re
gion
Loss ratio = 0.05%10%15%20%
Fig 3.6h Economic growth in the disaster region
-4.0%-3.0%-2.0%-1.0%0.0%1.0%2.0%3.0%4.0%5.0%6.0%7.0%
0 1 2 3 4 5 6 7 8
Time
Gro
wth
in d
isas
ter r
egio
n
Loss ratio = 0.05%10%15%20%
Chapter Three: Theoretical Models
125
Fig 3.6i Output in the adjacent region
1.25
1.25
1.25
1.25
1.25
1.26
1.26
1.26
0 1 2 3 4 5 6 7 8
Time
Out
put i
n th
e ad
jace
nt re
gion
Loss ratio = 0.05%10%15%20%
Fig 3.6j Economic growth in the adjacent region
-1.4%-1.2%-1.0%-0.8%-0.6%-0.4%-0.2%0.0%0.2%0.4%0.6%0.8%
0 1 2 3 4 5 6 7 8
Time
Gro
wth
in th
e ad
jace
nt re
gion
Loss ratio = 0.05%10%15%20%
Chapter Three: Theoretical Models
126
Fig3.6k Changes in welfare
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
00 0.05 0.1 0.15 0.2 0.25
Loss ratios
Pres
ent v
alue
of u
tility
Welfare in disaster regionWelfare in adjacent region
Chapter Three: Theoretical Models
127
3.5 Conclusions
In the preceding sections three models simulating the behavior of an economy
after the occurrence of a catastrophic event were studied. Changes in capital due to a
catastrophe and the subsequent changes in productivity are modeled by perturbing a
dynamic model of the economy. The simulation results point to the importance of
modeling the efficiency of the reconstruction processes after an event. Classical models
of growth (as demonstrated in the first model) suggest that lower levels of capital result
in higher growth rates. This would imply that if a catastrophe strikes a region and
destroys substantial capital stock it would grow at a faster rate than a region that loses
smaller amounts of capital. However, empirical evidence, based on data from 43
countries in which catastrophes have occurred, strongly suggest that greater loss is
associated with smaller post-event growth rates.
The main contribution of this chapter is to show that, unless the process whereby
the maturing capital is converted to productive capital is modeled, the fact that post-event
growth rate is negatively correlated with the magnitude of loss cannot be explained. Pre-
event conditions are important in post-event recovery. Pre-event conditions determine the
efficiency of the processes whereby newly reconstructed capital becomes fully
productive. In particular, pre-event conditions determine the changes in productivity.
The third model simulates the interaction between two regions one of which is struck
by a catastrophe. Model results suggest that greater the loss more is the effect felt on the
unaffected region. This is supported by examining the evidence from three catastrophic
events – 1989 Loma Prieta Earthquake, 1992 Hurricane Andrew, and 1994 Northridge
Earthquake. Higher loss ratio in Dade County made its effect felt at the state level,
whereas the lower loss ratios in Los Angeles and the Loma Prieta affected counties
resulted in localized effects. Evidence regarding the personal income suggests that
growth rate after the events are typically higher. This is also borne out in the model
behavior results. Though the models do not go into simulating sector specific effects such
Chapter Three: Theoretical Models
128
as impact on housing prices or government expenditures, by modeling consumption a
general equilibrium approach has been adopted that allows for qualitatively verifying the
behavior.
The models consistently demonstrate the importance of investment in the
disaster region. Coupled with efficient reconstruction, the models show that investment in
the disaster region can make the region more productive than before the event. Disasters
offer an opportunity to rebuild and convert vulnerable communities into robust ones.
Chapter Four: Empirical Studies
129
Chapter Four
An Empirical Study of the Macro-economic Effects of Catastrophes Triggered by Natural Events
4. INTRODUCTION
In this chapter we re-examine our understanding of the effects of catastrophes on
the economy based on empirical evidence. Questions addressed include the change or
absence thereof, in economic growth, consumption, saving, inflation, and real interest
rates. Data on these economic indicators are compiled for various countries for periods
immediately preceding and following the occurrence of a catastrophe. Data regarding
catastrophes such as the estimates of direct losses is also compiled. The regression
analysis employed suggests that catastrophes are negatively associated with all the
aforementioned economic indicators.
In order to study the effect of a catastrophe on an economy the factors that describe
socio-economic conditions prior to occurrence of the hazard event have to be identified.
The vulnerability of a society to natural hazards is the result of various on-going
economic, social, and political processes, as has been discussed in Chapter 2. For large
segments of the world's underdeveloped population, occurrence of a natural hazard may
worsen an already deteriorating or fragile situation. In such regions even a moderate
hazard, such as the 1985 Mexico earthquake, could trigger a catastrophe. Oliver-Smith
(1994) brings this out clearly in his analysis of the 1970 Peru Earthquake. He points out
that Peru's catastrophe was some 500 years in the making, rooted in the complex of
economic and political forces that structured development and the human-environment
relations. The earthquake and subsequent landslides was a trigger for a catastrophe
grounded in poverty, political oppression, and the subversion of previously sustainable
indigenous practices (Bolin and Stanford, 1998).
Socioeconomic conditions in a region are mainly as a result of the developmental
processes. The effect of a major catastrophe on the developmental process is complex,
Chapter Four: Empirical Studies
130
especially for developing regions. Globally, economies are evolving ‘complex’ systems.
This complexity in the economic systems is the result of the historical geography, the
political economy, the increased interdependencies among various sectors and regions of
an economy facilitated by the quantum leaps in the communication technology, and the
rapid globalisation of trade. In order to study the effect of a catastrophe on an economy
the factors that describe socio-economic conditions prior to occurrence of the hazard
event have to be identified. General statements regarding the economic consequences of a
catastrophe can be made only when these complexities are appropriately modeled.
An overview of some studies, which partly address these questions, is presented in the
next section. In section 4.2 connections between the occurrence of a catastrophe and
ongoing development processes of an affected region are made. Section 4.3 describes the
data used for the present study. The general framework and the particular econometric
model used to estimate the effect of catastrophes are presented in the next section.
Various factors that affect the growth rate are then presented. Section 4.5 presents a
discussion of various factors that may be important in determining the post event
economic indicators. Results of regression analysis are discussed in Section 4.6.
Chapter Four: Empirical Studies
131
4.1 PREVIOUS STUDIES
Studies on the effects of natural hazards on an economy have discussed direct and
indirect losses that result from such events (Development Technologies, 1992). Direct
losses are usually associated with direct physical damage and secondary effects, such as
damage caused by fire following an earthquake. Indirect damages relate to the effect on
flows of goods that will not be produced and services that will not be provided after a
catastrophe. They are measured in monetary terms. The impact of the catastrophe on
overall economic behavior, which has sometimes been termed as secondary effects, is
measured by changes in macro-economic variables. The work reported in this dissertation
focuses on secondary effects.
There are few studies on the macro-economic effects of catastrophes. They are based on
small data sets. Moreover, the conclusions are seemingly contradictory. Albala-Bertrand
(1993:163) argues "GDP normally does not fall after a disaster impact and if anything
tends to improve at least for a couple of post-disaster years." Albala-Bertrand's study
(1993) is based on a sample of catastrophes that occurred in the 1970's in mostly
developing countries. He uses three criteria for examining the effect of catastrophes on
economic growth, investment and sector outputs, public finance, and balance of
payments. The three criteria include: i) examining the change in the indicators according
to sign (positive meaning 'growth') and direction of change (up meaning 'acceleration'),
ii) the figures are averaged in per country terms for each period, and iii) comparison
between pre- and post-disaster averages. Limited by sample size, no other statistical
inferential procedures are used. The hypothesis he proposes is not validated since there
could be many factors that explain post-event economic behavior. For example, a country
might have experienced increased growth after an event because of reasons totally
unrelated to the occurrence of a catastrophe or due to efficient reconstruction policies.
However, this does not imply that a similar economy will sustain economic growth in the
absence of efficient reconstruction. Inferences from cross-country data are general only if
they are ‘normalized’ using control and environmental variables.
Chapter Four: Empirical Studies
132
The World Disasters Report (1997) expresses an apparently opposite viewpoint. The
report states, “Caribbean disasters can be costly, especially as a proportion of GDP. The
impact on national economies has been significant: hurricanes between 1980 and 1988
effectively reversed the growth rates.” This statement is again based on a simple
comparison of average growth for the affected countries between 1980-88 and 1989-91
(Table 4.1). All the five countries are small islands, which makes it difficult to generalize
the result.
Taken together these studies produce ambiguous conclusions regarding the effect of
catastrophes on ongoing economic processes.
Friesema et al. (1979) is an early study to analyze the effect of disasters on the long-term
growth patterns of four cities - Conway, Galveston, Topeka, and Yuba City. Their null
hypothesis is that disasters had no significant effect on employment, small business
activity (number of gas stations and restaurants), retail sales, and public finance. They
examine a time series of the indicators for a time period ten years before and after an
event. They conclude that local economic behavior patterns, barring slight disruptions,
were scarcely interrupted by the disaster events considered. They also mention that their
results are not surprising since in all the four cases the basic capital stock remained, and
the production process continued. This makes their sample unrepresentative of post
catastrophic economic behavior.
Table 4.1: Disasters in the Caribbean can have a significant impact on GDP and growth (World Disasters Report, 1997)
Country Average growth rate GDP 1980-88
Average growth rate GDP 1989-91
Dominica 4.9 4.3 Montserrat 3.7 -4.4 St.Kitts/Nevis 6.0 4.9 Antigua/Barbuda 6.8 2.2 Jamaica 5.0 0.8
Chapter Four: Empirical Studies
133
Wright et al. (1979) examine data for over 3100 counties in the US for effects of disasters
on growth trends of population and housing. Damage inflicted by the typical disaster in
their sample affected only a small proportion of structures, enterprises, and households of
typical counties. Based on regression studies they conclude that there are no significant
effects on growth trends in population and housing. However, these findings have been
questioned by the research of Yezer and Rabin (1987), who distinguish between
anticipated and unanticipated disasters. Their hypothesis is that “expected” disasters,
those occurring at a rate predicted by historical experience in a region, have no impact on
migration – such expectations have already been reflected in trend rate of migration. In
contrast, “unexpected disasters”, a spate in excess of those predicted by historical
experience, discourage migration. Empirical testing that explicitly distinguishes
“anticipated” from “unanticipated” supports the hypothesis.
The inferences from these studies cannot be generalized to effects of catastrophe in a
developing economy for several reasons. Firstly, the studies concentrate on regional
localized effects in a developed country. Secondly, the direct loss reported in the studies
is relatively small compared to the overall capital stock of the affected region. Finally,
they only examine changes in a subset of indicators that describe the social and economic
conditions of a region.
Chapter Four: Empirical Studies
134
4.2 CATASTROPHES AND ONGOING DEVELOPMENT PROCESSES
Losses from a catastrophe may be readily absorbed by a developed economy. To
cite an example, the Northridge earthquake occurred in a state with a Gross Regional
Product ranked 6th largest in the world. A US $30 billion direct loss due to the earthquake
manifested itself as a minor perturbation. This contrasts with the devastating Third World
disasters such as the 1976 Guatemala earthquake or the 1985 Mexico City earthquake. In
both cases, the catastrophes produced national crises with effects well beyond the
immediate physical impacts.
For a developing economy, like Bangladesh, direct losses from a catastrophe, which are
comparable to the Gross Domestic Product (GDP) might divert scarce resources from
development plans to reconstruction. Almost half of the 1988/89 Bangladesh's national
development budget was diverted to pay for ad-hoc relief and rehabilitation programs
(Brammer, 1990) after the 1988 flood. Development plans may include improving health
care, education, food supply, and institutions for crisis management. As Bates and
Peacock (1993) point out catastrophes "intervene in the development process as it
pertains to other important adaptive problems, and they redirect, deflect, retard, and on
rare occasions accelerate the development process."
The deep indebtedness of many Third World countries has made the cost of
reconstruction and the transition from rehabilitation to development unattainable. To see
how foreign debt burden can adversely affect the loss that a country suffers, take the case
of Jamaica struck by Hurricane Gilbert in 1988 (Blaikie, et al. 1994). Prior to the
Hurricane Gilbert, part of Jamaica's debt burden was in part due loans used to pay for
damages from previous hurricane. Jamaica introduced a structural adjustment program
that typically involved cuts in public spending. Services such as education, health, and
sanitation were reduced. Government programs to introduce preparedness or mitigation
measures were also cut as result of economic constraints. These decisions greatly reduced
the ability of the community to recover from the effects of a major hazard like Hurricane
Gilbert.
Chapter Four: Empirical Studies
135
Foreign debt also forced the government to intervene in the financial sector that resulted
in an increase of interest rates to over 20% and home mortgage rates ran between 14-
25%. Government forced rent control and import duty on construction materials. This
resulted in a rapid decline in new construction and other maintenance activity. The
quality of new construction also declined, since contractors tried to maximize profit by
using unsafe practices. This may have been partly responsible for the huge magnitude of
losses observed.
Delica (1993) brings out the relation between disasters and economic growth based on
her study of the natural hazards affecting Philippines. She argues that disasters have
practically negated the real economic growth achieved during the administration of
Carazon Aquino. From 1986 to 1991, damage to infrastructure, property, agriculture, and
industry from disasters were enormous, averaging about 2% of the GNP. Using simple
arithmetic, she argues that with an annual population growth of 2.3%, the economy needs
greater than 4.3% annual growth simply to maintain per capita income levels. But the
economy had only about 4% average annual growth, with the result vulnerability to
disasters has increased rather than decreased. This is because Philippines’ foreign debt
obligations have increased, from $26 billion in 1985 to $29 billion in 1992. The
government's spending on relief and rehabilitation has been tightly controlled and
increasingly dependent on external sources. Government's development strategy puts a
premium on export-orientation and attraction of foreign investment. This is at the
expense of ecological sustainability and environmental protection. Out of the 54% forest
cover required for a stable ecosystem only 20% remains as a result of deforestation. This
in turn increases the severity of floods and landslides.
Many poor countries try to solve their debt problems by adopting national policies
favoring raw material export. This typically results in land degradation since new land is
cleared for ranching and commercial cropping. Land degradation increases vulnerability,
which in turn increases the potential for catastrophic losses.
Chapter Four: Empirical Studies
136
Long-term development projects may be adversely affected by diversion of resources to
help an affected community rebuild. Twigg (1998) reports that the World Bank diverted
some $2 billion of existing loans between the 1987 and 1988 financial years to fund
reconstruction and rehabilitation after catastrophes triggered by natural events.
Catastrophes reveal the robustness or vulnerability of a country's socioeconomic
conditions. Various indicators can be used to quantitatively measure robustness or
vulnerability. The importance of these parameters, which are perhaps ignored in less
turbulent times, is revealed, tragically, only after a catastrophic event. A catastrophe can
unmask social and economic inequalities that come to the fore in the distribution of relief
aid. Catastrophes usually result in worsening the pre-event economic inequalities. It is
important to identify the factors that can be associated with vulnerability that explains the
wide variety of post event economic behavior. For example, we could examine the role of
infrastructure in changes in post-event economic growth. As has been pointed out by
Hewitt (1983), catastrophes are shaped and structured by economic, social, political and
cultural 'practices' and processes that existed prior to the occurrence of a physical event.
Indicators that describe some of these initial practices and processes need to be identified.
Whether there exists empirical evidence to support the hypothesis that ongoing
socioeconomic processes determine the post event economic behavior will be examined
in this chapter.
4.2.1 Change in Indicators Due to Catastrophes
In the past decade there has been an explosion of empirical studies of growth and
development. Efforts have been made to account for differences in growth rates between
various countries using indicators of education, health, infrastructure, institutions, and
political freedom. Results from these studies will be used to identify variables that can
make cross-country comparisons of changes in macro-economic indicators possible. The
parameters will act to control some of the variability across countries. Any effect due to
catastrophes on macro-economy can be detected only after the control variables
explained variability from other sources.
Chapter Four: Empirical Studies
137
As has been mentioned previously, there is an intimate relation between ongoing
development processes and the occurrence of a catastrophe. The various parameters,
which are associated with development such as education, infrastructure, and health, are
hypothesized as measures of a community's robustness (or pessimistically, vulnerability)
to a catastrophe. A combination of these parameters can be used to assess a community's
robustness. It is reasonable to expect that a robust community's development or growth
should not be adversely affected by occurrence of a catastrophe. The ongoing dynamics
of the developmental processes are capable of absorbing the effects of catastrophe.
Conversely, a society is weak if its development process is adversely affected by the
occurrence of catastrophe triggered by a natural event. The present work relies on
previous studies on the determinants of growth for choosing parameters that are
associated with the development process.
One important indicator of development of a country is its economic growth rate. The
following is a summary of some of the parameters that have been shown to be
determinants of growth. A percentage point in economic growth is associated with the
following:
• Increase of 1.2 years in average schooling of labor force
• An increase in secondary enrollment of 40 percentage points
• A reduction of 28 percentage points in the share of central bank in total credit
• An increase of 50 percentage points in financial depth (M2/GDP)
• An increase of 1.7% of GDP in public investment in transport and communication
• A fall in inflation of 26 percentage points
• A reduction in the government budget deficit of 4.3 percentage points of GDP
• An increase in (exports + imports)/GDP of 40 percentage points
• A fall in government consumption/GDP of 8 percentage points
• An increase in foreign direct investment/GDP of 1.25 percentage points.
(Barro 1991, Barro and Lee 1993, King and Levine 1993, Easterly and Rebelo 1993,
Fisher 1993, Easterly and Levine 1997, Easterly, Loayza, and Monteil 1997, Borensztein,
De Gregorio, and Lee 1994).
Chapter Four: Empirical Studies
138
These inferences are used to identify the variables that can be controlled when making a
cross-country comparison of post-event behavior of the economic growth.
Chapter Four: Empirical Studies
139
4.3 GENERAL FRAMEWORK AND ECONOMETRIC MODEL
The general framework to be used in empirical studies reported here will be
developed in this section. In Chapter 3, theoretical models simulated the occurrence of a
catastrophe as a perturbation of the ‘normal’ economic processes. A catastrophe was
modeled as a reduction of capital and subsequent changes in productivity of the affected
region. The economy was assumed to be initially in its steady state. Inspection of
Eq.3.3.15 reveals that growth of capital due to the catastrophe depends on the steady state
(the Jacobian term) and the perturbations. Growth of the economy in turn depends on the
changes in capital stock. Hence, the following relation is used to estimate the effect of a
catastrophe on the post-event growth rates:
growthwith hazard = f(damage, productivity-changes; y*) (4.1)
y* is the long-run steady-state level of per capita output and depends on the steady state
levels of capital stock, as shown in Eq. 3.2.9. y* depends on an array of choice and
environmental variables. The private sector’s choices include saving rates, labor supply,
and fertility rates, each of which depends on preferences and costs. The government’s
choices involve spending in various categories, tax rates, the extent of distortions of
markets and business decisions, maintenance of rule of law and property rights, and the
degree of political freedom. Also relevant for an open economy is the terms-of-trade,
typically given to a small country by external conditions. A cross-country empirical
analysis requires conditioning on the determinants of the steady states. Also, the pre-
event conditions to a large extent determine the post event productivity. These
determinants or the country specific factors, along with their relation to catastrophes, are
presented in Section 4.4. It is assumed that the country specific factors are invariant over
the period of interest – five years. Data for these factors are typically available as
constants over five- to ten-year periods.
Damage, in general, depends upon the intensity of the hazard and the vulnerability.
Vulnerability is the susceptibility of the exposed constructed facilities, economic and
Chapter Four: Empirical Studies
140
social structures of a region to be affected given a specified level of hazard. As discussed
in Chapter 2, vulnerability is intimately related to ongoing socio-economic processes.
Damage may be expressed as:
damage = h(hazard, vulnerability) (4.2)
It should be mentioned here the relation Eq.4.2 is expected to be highly non-linear. Even
for relatively simple structures such as single-family dwellings, the damage curves –
which relate the hazard intensity to the damage level (RMS, 1996) – are non-linear. Data
regarding the loss of capital and the changes in productivity are hard to come by. Hence
the loss of capital is modeled by the direct losses recorded after the event.
4.3.1 Approximation
The first step to estimate the model expressed in the relations (Eqs.4.1-2) above is
to use an approximate linear relation. Consequently, the relation in Eq.4.1 is
approximated by:
growthwith hazard = α1 + β1E + β2Damage + β3Hazard_type + ε1 …(4.3)
ε1 is an unobserved disturbance term. The indicators for Damage are the direct-loss to
GDP ratio and the percentage of population affected. E is a vector of time-invariant
country specific indicators of the economy that are considered as determinants of
economic growth. The vector E contains indicators from each of the following categories
of determinants of growth - Economic conditions, Individual Rights and Institutions,
Education, Health, Transport and Communications, Inequality across income and gender.
In particular the following indicators are used: Inflation variability, Average pre event
decade growth, SD of pre event decade growth, annual money growth, black market
premium, political rights, civil liberties, bureaucratic quality, government enterprises,
percent “no schooling” in population, daily protein or calorie intake, life expectancy at
age zero, radios per capita, and TVs per capita. Hazard-type is a dummy variable to
account for the type of hazard – earthquake, hurricane, or drought.
Chapter Four: Empirical Studies
141
It should be mentioned here that Damage as such would depend on factors in the vector
E. It is implicitly assumed that the indicators for damage are not correlated with the
factors in E. This may be a strong assumption if the measure of loss is in terms of
destroyed productive capital stock and E includes factors such as capital stock per
worker. Indicators chosen in E are such that they are only indirectly related to direct loss
term. Therefore, the assumption that E and damage are not significantly correlated is
reasonable. It is also assumed that the errors in measurement/estimation of damage are
not correlated with the error term ε1. The reduced form given in Eq.4.3 is estimated.
The results presented in Fig. 3.4h indicate that loss is negatively correlated with the post-
event growth rate. The hypothesis to be tested is that the coefficients β2 in Eq.4.3 are
statistically significant and negative.
Similar models for other economic indicators are estimated where the dependent variable
is chosen to be the post event budget deficit, external debt, resource balance, inflation,
interest rates, or consumer price index. Again, the hypotheses to be tested are that the
coefficients β’2 in Eq. 4.3 are statistically significant. 4.3.2 Summary Statistics and Discussion of the Sample
4.3.2.1 Economic growth
As a first step, the growth rates between two adjacent years are compared, that is,
the growth rate during the event year is compared with the growth rate immediately
preceding year. Both mean and median of the pre-event annual percentage growth are
greater than their post-event counterparts (Table 4.2). Presumably catastrophic events
also induce greater variance for the growth, as evidenced by comparing the pre- and post-
event variances in the growth (Table 4.2). Distribution of the pre- and post- event growth
rates are shown in Fig.4.1.
Chapter Four: Empirical Studies
142
Table 4.2 Summary statistics for short-term growth Pre event Post event
Mean 3.96 3.29Standard Error 0.30 0.33Median 3.80 3.18Standard Deviation 3.70 4.06Sample Variance 13.67 16.46Kurtosis 2.01 2.61Skewness -0.16 -0.61Range 23.36 26.60Minimum -9.10 -12.57Maximum 14.27 14.03Sum 605.25 503.99Count 153 153Confidence Level(95.0%) 0.59 0.65
Fig 4.1 Event year growth is clearly lower than the pre-event year growth
0%10%20%30%40%50%60%70%80%90%
100%
0 2 4 6 8 10
GDP annual growth (%)
Perc
entil
e 1yrBeforeEventYear
Fig 4.2 Average pre- and post-event growths
0%10%20%30%40%50%60%70%80%90%
100%
0 2 4 6 8 10
GDP annual growth (%)
Perc
entil
e Avg3yrsAfterAvg3yrsBefore
Chapter Four: Empirical Studies
143
It is apparent from Fig. 4.1 that the distribution for growth in the event year is shifted to
the left relative to growth one year before the event.
Table 4.3 summarizes the statistics for pre- and post- event average growth. Here again
average post-event growth rate is smaller than pre-event growth rate. But sample variance
of the average post-event growth is smaller than average pre-event growth, indicating
perhaps that the effects of the events are reducing.
Table 4.3 Summary statistics for average growth Average 3 years before
Average 3 years after
Mean 3.83 3.55Standard Error 0.27 0.24Median 3.48 3.61Standard Deviation 3.32 3.00Sample Variance 11.03 9.02Kurtosis 1.89 3.35Skewness 0.15 -0.44Range 22.96 22.66Minimum -9.42 -10.20Maximum 13.54 12.46Sum 586.68 542.90Count 153 153
It is apparent from Fig. 4.2 that the average post event growth is shifted to the left relative
to average pre event growth rates, though in this case the effect is not as pronounced for
growths less than 5%.
4.4.2.2 Effect on consumption, investment, government expenditure, net exports and income
The main components of the GDP are the consumption, investment, government
expenditure and net exports. Using the latest Penn World Table data (2002), the effect of
catastrophes on each of these macroeconomic indicators is investigated. As a first step,
each of these variables is graphed with the loss-GDP ratios. These graphs are shown in
Fig. 4.3 to 4.8.
Chapter Four: Empirical Studies
144
Fig. 4.3 Effect of catastrophes on consumption
40
50
60
70
80
90
100
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
eve
nt c
onsu
mpt
ion
(% o
f GD
P)
Fig. 4.4 Greater losses are associated with larger amount of government spending
05
1015202530354045
0.0% 0.1% 1.0% 10.0% 100.0% 1000.0%
Annual loss as a % of GDP
Post
-eve
nt in
vest
men
t as
a %
of
GD
P
Fig. 4.5 Greater losses are associated with larger amount of government spending
0
5
10
15
20
25
30
35
40
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
eve
nt g
over
nmen
t sp
endi
ng a
s a
% o
f GD
P
Chapter Four: Empirical Studies
145
Fig. 4.6 Greater losses are associated with higher openness
0
20
40
60
80
100
120
140
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
eve
nt o
penn
ess
as a
% o
f G
DP
Fig. 4.7 Larger losses are associated with smaller post event savings
-40
-30-20
-100
1020
3040
50
0.0% 0.1% 1.0% 10.0% 100.0% 1000.0%
Annual loss as a % of GDP
Post
eve
nt s
avin
gs a
s a
% o
f G
DP
Fig. 4.8 Greater losses are associated with lower post event GDP per capita
100
1,000
10,000
100,000
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
-eve
nt re
al G
DP
per
equi
vale
nt a
dult
Chapter Four: Empirical Studies
146
These (Fig. 4.3 to 4.8) depict important observed regularities between magnitude of
losses and post-event macroeconomic variables. For example, Fig. 4.5 depicts the
observation that higher losses are associated with higher post-event governmental
spending as a fraction of the GDP. Fig. 4.8 establishes a clear negative association
between loss magnitude and post-event GDP per capita. Regressions in the later sections
are performed to determine the robustness of these associations by accounting for country
specific factors.
Other variables are also examined. In particular the effect of losses on inflation and real
interest rates are presented in Fig. 4.9 and 4.10, respectively.
The next section discusses the primary and control variables that are used in the
estimation. An overview of linear regression analysis for the estimation of Eq.4.2 along
with the model adequacy checking is presented in Section 4.6. Following that,
econometric evidence associating changes in economic indicators with magnitude of loss,
the percentage affected, and the type of catastrophe is presented.
Fig. 4.9 Larger losses are associated with higher inflation
0.1
1.0
10.0
100.0
1000.0
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Log(Loss/GDP)
Post
Eve
nt In
flatio
n
Chapter Four: Empirical Studies
147
Fig. 4.10 Greater loss ratios are associated with higher post-event real interest rates
-10
-5
0
5
10
15
20
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
eve
nt re
al in
tere
st ra
te
Chapter Four: Empirical Studies
148
4.4 EFFECT ON THE ECONOMIC GROWTH
Relating the magnitude of a catastrophe with a change in the growth of an
economy is very complex since there are many factors that determine the economic
growth (Barro, 1997). Recent research in the determinants of cross-country economic
growth has revealed much regularity. Investment in physical capital, educational
attainment of the population, stable macro-economic policies, open trade regimes, better
developed financial markets are important factors exerting positive effect on growth
(Barro and Sala-i-Martin, 1995). There are several other factors that retard growth -
population growth, political instability, budget deficits, shocks resulting from terms of
trade changes, internal strife, and wars (Rodrik, 1998), policy distortions, government
consumption, and low bureaucratic quality (Commander, et al, 1997). In the following
sections we describe some factors that may explain the variety of observed changes in
ongoing economic processes after a catastrophe. This discussion is similar to the
discussion in Chapter 2 regarding the factors that determine the vulnerability to natural
hazards. The important difference here is that these factors are explained here as factors
that may contribute towards the recovery of a community after a catastrophic event.
4.4.1 Primary variables
Three primary variables are used as indicators of the catastrophe. They are: i) the direct
physical loss, ii) the percentage of population affected, and iii) the type of natural hazard.
4.4.1.1 Direct physical loss
One of the important variables that characterize a catastrophe is the resulting
direct loss. Direct damages include all damage to fixed assets (including property),
capital and inventories of finished and semi-finished goods, and business interruption
resulting from a catastrophe (HAZUS, 1997). Estimation of the macro-economic effects
involves a comparison of economic behavior with and without the change in a
Chapter Four: Empirical Studies
149
community's assets. The direct loss is one measure of the change in community assets
after a catastrophe.
Comparing direct loss across countries necessitates an approach based on purchasing
power parity (PPP). Converting the losses into a common currency, for e.g. the US dollar,
through the use of official exchange rates often misleads cross-country comparisons of
the losses. These nominal exchange rates do not reflect the relative purchasing power of
different currencies, and thus errors are introduced into the comparisons. Using PPP is
one way to obtain a correct measure of losses. In countries where the domestic prices are
low, the losses based on PPP will be higher than that obtained from official exchange
rates. For the purposes of this study we use ratio of loss (in current US dollars) to the
GDP (in current US dollars) as a measure of direct loss. Using a ratio makes comparison
of loss across countries valid, since PPP or exchange rates that appear both in numerator
and denominator of the ratio cancel out. As mentioned in the introduction, the loss to
GDP ratio does not exhibit any trend over the time period of the sample and hence is a
good indicator of catastrophes. This is important since the present study is based on
events during the last three decades. Comparison is only possible by using the annual
economic loss as a proportion of the total income (GDP).
4.4.1.2 Percentage affected
In a developing economy, where the majority are poor the number of people
affected is often a better indicator of the severity of a catastrophe than direct loss. The
number of people affected depends on the vulnerabilities of various groups that are
resident in the affected area. The vulnerability of groups in turn depends on the manner in
which assets and income are distributed between different social groups. Post event
recovery depends on the way resources are allocated and here too discrimination may
occur based on pre-existing conditions of inequality based on gender, ethnicity, and race.
It is these vulnerable sections of society that suffer most from catastrophes affecting their
lives, their settlements, and their livelihoods.
Chapter Four: Empirical Studies
150
Blaikie et al. (1994) point out that in many parts of the world each household's bundle of
property and assets and economic connections with others may be lost, enhanced,
disrupted, or reinforced in a number of ways due to hazards. The impact of the hazards
operate under the influence of rules and structures derived from existing social and
economic system, but are modified by the distinct characteristics of a particular hazard
and patterns of vulnerability.
4.4.1.3 Type of hazard
Different types of disaster have varying direct and therefore indirect and secondary
impacts. Given a vulnerable habitat, the damage pattern depends on type and intensity of
the physical event. For example, droughts ruin crops and forests but cause relatively little
damage to infrastructure. As a result productivity may remain the same after the event. In
the case of droughts, if the country has surplus of domestic food production, drought can
be managed. For example, one year after the 1982 Australian drought the country's
economy was back to 'normal'. But in countries with little surplus, the effects are more
tangible. Countries whose GDP is mainly represented by the rural economies are
especially vulnerable to droughts. Droughts cause major production losses. If the net farm
income falls during a drought in a farm based economy, it my cause a decline in the
overall output.
In contrast, earthquakes cause relatively little damage to standing crops, other than
localized losses resulting from landslides. But an earthquake can damage buildings and
underground infrastructure. A hurricane may cause extensive crop damage as well as
damage to structures. Reconstruction may result in changes to the productivity due to the
destruction and subsequent construction of new capital. Such changes in productivity
were modeled in Chapter 3. It is important to find out whether the type of disaster affects
the post-event growth rates.
Location and climate have large effects on income levels and income growth, through
their effects on transport costs, disease burdens, and agricultural productivity, among
Chapter Four: Empirical Studies
151
other channels. Major natural hazards that occur frequently in some parts of the globe
have definite effects on income levels and growth. Some countries may therefore be at a
geographical disadvantage due to being situated in hazard prone area.
4.4.2 Control variables
Previous studies do not explicitly spell out the explanatory variables that may be
related to the post-event economic growth rate. Furthermore, there is a lack of theoretical
analytical models describing the phenomena, which has been addressed in Chapter 3.
Theoretical models and simulations presented in Chapter 3 point to the importance for
modeling the post-event productivity changes. Changes in the productivity are reflected
in the post-event evolution of consumption, output, and growth. Based on a wealth of
studies conducted in the field of economic growth (mentioned in Section 4.2.1), variables
that may be important in determining the post-event productivity are discussed in the
following sections. These include indicators for describing pre-event economic
conditions, health, poverty and inequality, government, infrastructure, education, and
trade.
4.4.2.1 Pre-existing Economic Conditions
If a nation has a stable macro-economy with a steady growth, it would be
relatively easier to detect any fluctuations resulting from a catastrophe. Pre-event decade
mean and standard deviation of the annual percentage growth rates are included as
control variables, as indicators of past performance of a nation's macro-economy. Barro
(1995) finds that higher inflation variability goes along with a lower rate of economic
growth. Monetary institutions and policies that lead to substantial variations in the
general level of prices create uncertainty and undermine the efficacy of money. In the
event of a catastrophe, it is more likely in nations with high inflationary susceptibility
that the prices will go out of control. Inflationary pressures will have a negative effect on
the productivity. An indicator for standard deviation of the annual inflation rate during
the last five years is included as a control variable (Gwartney and Lawson, 1997).
Chapter Four: Empirical Studies
152
Another indicator of monetary stability that is included is the average annual growth rate
of the money supply during the last five years minus the potential growth rate of the GDP
(Gwartney and Lawson, 1997).
4.4.2.2 Health
Health problems are particularly highlighted in studies of floods on the West
Coast of South America brought about by El Nino in 1982-83. Blaikie et al. (1994),
quoting from a study of government health centers in north Peru, report that there was an
almost two-fold increase in number of deaths as result of disease and illness due to
epidemics following floods. People's basic health and nutritional status relates strongly to
their ability to survive disruptions of their livelihood systems. This status is important for
their resilience in the face of external shock. For most people living on a subsistence diet
and without proper access to health care, even a mild epidemic after a catastrophe may
prove fatal. The pre-event socioeconomic processes, to a large extent, determine the pre-
event health conditions of the community which in-turn determines the percentage of
people affected by a catastrophe. The post event reconstruction depends on an adequate
supply of labor immediately after the event. If the majority of population is affected by a
catastrophe for health reasons, there may be inadequate supply of labor resulting in
adverse changes in post-event productivity.
Various indicators are used to summarize the 'health' of a community. These include:
i) Life expectancy at age zero,
ii) Number of hospital beds per thousand, indicating the accessibility of health services
after a catastrophe, and
iii) The daily calorie and protein intakes.
4.4.2.3 Poverty and Inequality
The burden of poverty is spread unevenly - among the regions of the developing
world, among countries within those regions, and among localities within those countries
Chapter Four: Empirical Studies
153
(Meier, 1995, Ray, 1998). Alexander (1998) cites the example of Philippines and
compares it with Japan. Both the countries have similar risk profiles as far as occurrence
of types physical hazards are concerned. But Philippines has a GNP that is 2.75% of
Japanese and 49% of Philippines population lives below poverty line. This necessitates
Philippines to bear a heavier burden from losses it experiences from calamitous events.
Within regions and countries, the poor are often concentrated in vulnerable places: in
rural areas with high population densities, such as the Indo-Gangetic plain and the Island
of Java, Indonesia. Often the problems of poverty, population, and the environment are
intertwined: earlier patterns of development and pressure of rapidly expanding
populations mean that many of the poor are forced to live in highly vulnerable regions.
As Blaikie et al. (1994) point out, in Manila (Philippines) the inhabitants of squatter
settlements constitute 35% of the population vulnerable to coastal flooding, and Bogota
(Colombia) has 60% of population living on landslide prone steep slopes. Even in urban
areas, if there are no adequate measures to systematically maintain buildings, potential
losses may be high. For example in the 1985 Mexico earthquake, the decaying inner city
tenements were severely affected.
Rural-urban migration leads to the erosion of local knowledge and institutions required
for coping in the aftermath of a disaster. The loss of younger people, especially working
age males and those with skills which are marketable in the cities may alter the type of
building structures that can be constructed to something less safe than previously.
Obviously this results in greater number of people being affected by the catastrophe.
Certain groups within a community are more vulnerable. Women, children, elderly,
ethnic groups, and minorities suffer disproportionately as a result of catastrophe as has
been reported by Peacock et al. (1997) after Hurricane Andrew. Inequality is a crucial
factor in the ability of an affected community to recover after the occurrence of a
catastrophe. A more unequal society will result in a more unequal distribution of effects -
the poorest in the affected society bearing the brunt of the catastrophe. An inefficient
bureaucracy will allow the inequality to deepen by concentrating the relief in the already
Chapter Four: Empirical Studies
154
affluent people of the community. It has already been demonstrated by various macro-
economists (Barro 1995, Easterly, 1997) that higher the inequality slower is the economic
growth. Thus one of the effects of a catastrophe, given an inefficient bureaucracy, is to
indirectly retard growth by deepening inequalities. On the other hand the government can
view the occurrence of a catastrophe as an opportunity for initiating various programs to
boost economic growth. More efficient and modernized infrastructure may be constructed
replacing damaged structures increasing productivity, which acts as a catalyst for
economic growth of the affected region.
Indicators used to summarize the 'poverty' include the percentage of people living on less
than $1 a day (PPP 1981-95) (World Bank, 1997). The daily calorie intake is also an
indicator of poverty, though controversial. The decade average for Gini coefficient is
used as an indicator for inequality (Easterly and Levine, 1997). The ratio of the share of
the top twenty percent in the income distribution to the first quintile is also used as an
indicator of inequality (Easterly and Levine, 1997). Gender bias is represented using the
ratio of female to male average schooling years.
4.4.2.4 Government, Bureaucracy, and Institutions
Whether a poor country recovers quickly from a catastrophe depends, among
other factors, on its government. If the government has effectively implemented the
policies that make the country's development potential realizable, then a catastrophe will
be absorbed without much negative impact. But in many poor countries, the political
foundations for developmental efforts are not yet firm. Political instability,
undifferentiated and diffuse political structures, and inefficient governments are still too
prevalent (World Bank, 1997).
Commander et al. (1997) look at factors explaining the size of government and the
consequences of government for income growth and other measures of well-being, such
as infant mortality and life expectancy. They present partial evidence for the view that
governments use consumption to buffer external risk, particularly in low-income
Chapter Four: Empirical Studies
155
countries. With respect to the consequences for growth, they find a robust negative
association with government consumption and with an index of policy distortions and a
positive relationship with quality of bureaucracy. They also report that social sector
spending can exert a positive influence on infant mortality and life expectancy.
Primarily its bureaucracy (Knack and Keefer 1995 and Mauro 1993) gives an explicit
evaluation of the quality of government. This evaluation is put together from a set of
responses by foreign investors that focus on the extent of red tape involved in any
transaction, the regulatory environment and the degree of autonomy from political
pressure. These responses provide us with a composite index of the quality of
government bureaucracy or its capability. Mauro (1993) finds a strong relationship
between per capita income and average indices of red tape, inefficient judiciary, and
corruption. Clague, Keefer, Knack, and Olson (1996) likewise establish a relationship
between high per capita income and high quality institutions - freedom from
expropriation, freedom from contract repudiation, freedom from corruption, and rule of
law. After a catastrophe has occurred, it is the efficiency of the government bureaucracy,
which partly determines the efficiency of the processes that determine the post event
productivity. As Oliver-Smith (1994) points out, the assistance after the 1970 Peru
earthquake never reached the survivors because of the 'Byzantine bureaucratic design and
a bewildering division of responsibilities' of the principal agency in charge of relief and
reconstruction. Keefer and Knack (1997) find a strong association between per capita
income and trust between individuals in a society. Trust is important for post event
behavior.
Rodrik (1999) presents econometric evidence from countries that experienced the
sharpest drops in growth after 1975 were those with divided societies and with weak
institutions of conflict management. He contends that 'social conflicts and their
management - whether successful or not - played a key role in transmitting the external
shocks on to economic performance.' The strength of crisis management institutions
determines the recovery process of an affected community. Studies at community level
(e.g. Oliver-Smith 1990, and Bolin 1982) highlight the major impediments to the
Chapter Four: Empirical Studies
156
community recovery process even when they have received aid. Aid is not effective for
the following reasons: i) local disaster management staff are unprepared to deal with aid
recipients, ii) aid does not meet the needs of the poor, iii) outside donor programs exclude
local involvement, and iv) poorly coordinated and conflicting demands from national
government agencies. Many national governments have begun to initiate programs that
assist their local jurisdictions to prepare recovery and development plans (Kreimer and
Munasinghe, 1991).
Political will and respect for human rights are important factors for the successful
implementation of such plans. Without strong political will and freedom of expression in
a country, methodologies devised by a vulnerable community for coping after a
catastrophic event will not receive necessary impetus. As a result development processes
may suffer. On the other hand, if the most vulnerable in the society are empowered
through grass-roots organizations and non-governmental organizations there would be
greater equity in distribution of relief aid thus sustaining the development process. In
Central America, the importance of local organization has been demonstrated in sudden
onset disasters (earthquakes) and in terms of events with a longer preparatory phase
(hurricanes and flooding).
An indicator of government size is the general government consumption as a percentage
of GDP. Human rights index, bureaucratic quality, rule of law, freedom from corruption,
and repression of civil liberties are all indicators for individual rights and democracy
(Easterly and Levine1997).
4.4.2.5 Infrastructure
Typically, a catastrophe results in major disruption of infrastructure facilities.
Given the fact that infrastructure facilities are poorly maintained in developing countries,
the extent of damage is severe even for a moderate hazard. Moreover, the emphasis shifts
to re-building the community in the immediate aftermath of a catastrophe. In its report on
Infrastructure and Development (1994), the World Bank points out that 'when times are
Chapter Four: Empirical Studies
157
hard, capital spending on infrastructure is the first item to go and operations and
maintenance are often close behind. Despite the long-term economic costs of slashing
infrastructure spending, governments find it less politically costly than reducing public
employment or wages.' After a catastrophe, emphasis shifts to providing immediate relief
to the victims and as a result capital expenditures are cut with infrastructure capital
spending often taking the biggest reduction. The 1970 Peru earthquake completely
incapacitated the fragile infrastructure of roads, railways, airports and communications
(Oliver-Smith, 1994). The rails of one railway system that had been twisted beyond
repair have not been replaced 20 years after the event. Destruction of roadways or other
infrastructure may cause impediments to relief and rescue operations and lead to business
interruption. After the 1995 Kobe earthquake, the Kobe port was closed down for a long
time due to extensive damage. As a result ships started using other ports with a result that
there was a permanent damage to regional income. The degree to which key
infrastructure can function after a catastrophe determines the rate at which materials and
men can be moved to the affected region. This in turn affects the post-event productivity
of the region.
The indicators used for infrastructure are:
i) Number of television sets per capita, and
ii) Number of radios per capita
4.4.2.6 Education
Education is vital in creating awareness and facilitates communicating catastrophe
mitigation ideas. A society will be more aware of the risk of occurrence of natural hazard
event if specific measures are taken to incorporate an adequate knowledge of the
vulnerabilities of different zones or regions in the school curricula. An educated
community can devise and implement more effective self-management strategies in case
of a catastrophic event and rapidly recover thus restoring its pre-event productive
capacity or sometimes even exceeding it. But education is intimately related to ongoing
socioeconomic processes. Barro and Martin (1995) present evidence to show that
Chapter Four: Empirical Studies
158
average-years of male secondary and higher schooling and average years of female
secondary and higher schooling tend to be significantly related to subsequent growth.
Thus the level of education of a community is an important control variable used to
compare post-event economic behavior across various countries.
The percentage of "no schooling" in the population is used as an indicator for education.
4.4.2.7 Trade
Trade forms an important component of a country's output in the modern global
economy. Manufactures form a major portion of imports of a developing country while
exports are usually non-manufactures such as cash crops. If a catastrophe severely
destroys the main export merchandise then the country may be forced into a balance of
payments crisis thus dampening its post-event reconstruction efforts. Productivity after
the event may be forced to be below its pre-event levels. Also, prices, for agricultural and
mineral exports on which Third World has traditionally had to depend, are falling. On the
other hand, prices of imported energy and technology have increased. This worsens
balance of payments crises. Hurricane Gilbert deprived Jamaica of more than US $27
million in foreign exports in 1988-89. This may partly explain the decline in its growth
rate (Table 4.1). Foreign debt amounted to 60 per cent of GDP for Latin America during
1985 (Branford and Kucinski 1988).
Indicators which summarize a governments policy stance on trade over time is:
i) The Economic Freedom Index (Gwartney and Lawson, 1997), and
ii) The degree to which a country's exchange rate has been over-valued - as measured by
the black market premium on the exchange rate.
Chapter Four: Empirical Studies
159
4.5 INTRODUCTION TO ECONOMETRIC ISSUES
An overview of the main concepts involved in linear regression models is
presented in this section. An algebraic model represents the real–world system by a
system of equations. These equations may be behavioral, such as the consumption
function: C = C(Y), an equilibrium condition, such as the national income equilibrium
condition: Y = C+Z. The variables in this model are consumption C, national income Y,
and exogenous expenditure Z. All of these fundamental or basic equations of the model
may be termed structural equations. The model determines values of certain variables,
called endogenous variables, the jointly dependent variables of the model, which are
determined simultaneously by relations of the model. In the case of the model specified
in Eq.3.7, post event economic growth rate is the endogenous variable, which is to be
explained or predicted. The model also contains other variables, called exogenous
variables, which are determined outside the system but which influence it by affecting the
values of the endogenous variables. They affect the system but are not in turn affected by
it. Direct loss from a catastrophe is an example of exogenous variable. The model also
contains certain parameters (α and β), which are generally estimated using econometric
techniques and relevant data. Another important characteristic of an econometric model is
the fact that it is stochastic rather than deterministic. A stochastic model includes random
variables, whereas a deterministic model does not. Let the economic growth rate after a
catastrophic event be given by:
yt = α0 + β1yt-1 + β2D + β3E (4.4)
This function specifies that given the pre-event economic conditions yt-1, a vector D
describing the disaster including magnitude of direct loss, the number of people affected,
and a vector E describing the country specific characteristics, post-event economic
growth rate yt is determined exactly as given by Eq.4.4. This is clearly not reasonable.
Many factors other than the direct loss and the population affected determine the post
event growth rate, such as inflation variability, quality of the bureaucracy, health, and
education. Furthermore, the relationship may not be as simple as that given by the linear
relation as explained in Chapter 3 and the loss variables may be measured inaccurately. It
is therefore more reasonable to estimate yt at a given level of loss variables, as on average
Chapter Four: Empirical Studies
160
equal to the right hand side of Eq.4.4. In general, yt will fall within a certain confidence
interval, after controlling for country specific fixed effects, that is,
yt = α0 + β1yt-1 + β2D + β3E ± ∆, (4.5)
where, ∆ indicates the level above or below the average value such that with a high
degree of confidence, yt fall in the defined interval. The value of ∆ can be determined by
assuming that yt is itself a random variable with a particular density function. Because of
the central limit theorem the normal distribution is typically assumed. The term “on
average” generally refers to the mean or expected value, so the right hand side of Eq. 4.5
is the mean of yt. The ∆ can then be chosen, as illustrated, so that 90% of the distribution
is included in the confidence interval, where each of the tails of the distribution contains
5% of the distribution. In general, an econometric model uniquely specifies the
probability distribution of each endogenous variable, given the values taken by all
exogenous variables and given the values of all parameters of the model.
Algebraically, the stochastic nature of the relationship for the post event growth rate is
represented as
yt = α0 + β1yt-1 + β2D + β3E ± ε (4.6)
ε is an additive stochastic disturbance term that plays the role of chance mechanism. In
general, each equation of an econometric model, other than definitions, equilibrium
conditions, and identities, is assumed to contain an additive stochastic disturbance term.
If the stochastic disturbance term has a variance that is always identically zero, the model
reduces to a deterministic one. The other extreme case is where the model is purely
stochastic. The stochastic terms are unobservable random variables with certain assumed
properties (e.g. means, variances, and covariances). The values taken by these variables
of the model are not known with certainty; rather, they can be considered random
drawing from a probability distribution. Possible sources of such stochastic perturbations
could be relevant explanatory variables that have been omitted from the relationship
shown in the model or possibly the effects of measurement errors in the variables - in
particular errors related to reporting of direct losses. Other sources of such stochastic
disturbances could be mis-specified functional forms, such as assuming a linear
relationship when the true relationship is nonlinear, or errors of aggregation, which might
Chapter Four: Empirical Studies
161
be introduced into a macro equation when not all individuals possess the same underlying
micro relationship. It might be noted that sources of these perturbations can be quite
important in practice; thus the treatment of measurement error will be, in general, quite
different, depending on whether the measurement error is found in the dependent variable
or in one or more of the explanatory variables. In any case, the inclusion of such
stochastic disturbance terms in the model is basic to the use of tools of statistical
inference to estimate parameters of the model.
4.6 PROBLEMS WITH THE DATA
Data related to catastrophes are typically non-experimental data. There are several
problems encountered with these data, which are presented in the following.
The first is the degrees-of-freedom problem – that the available data simply do not
include enough observations to allow an adequate estimate of the model. In the use of
non-experimental data it is impossible to replicate the conditions that gave rise to them,
so additional data points cannot be generated. Data regarding catastrophic losses may be
available, but data on explanatory variables may be missing. This was particularly true in
panel data compiled for this study. In some cases the available was inadequate for
estimating a particular model but adequate for estimating an alternative model, which
will be clear when various model specifications are presented. By including more than
155 major events, the panel had adequate degrees of freedom in spite of the missing data.
Second is the multi-collinearity problem - the tendency of the data to bunch or move
together rather than being “spread out”. For example, for a complex model describing
economic growth, the variables exhibit the same trends over a cross-section of countries.
With experimental data it may be possible to vary the conditions of the experiment to
obtain an adequate spread. With non-experimental data such control does not exist, and
the real-world system may involve very small variation in the data, in particular a high
degree of interdependence among certain variables. This problem was circumvented to a
certain extent by including data from countries belonging to wide range economic
development as measured by per capita GDP.
Chapter Four: Empirical Studies
162
Third is the serial-correlation problem – the fact that when using data in two consequent
years (before and after the catastrophic event), underlying changes occur very slowly
over time. Thus conditions in time periods that are close together tend to be similar. To
the extent that the stochastic disturbance term represents conditions relevant to the model
but not accounted for in it explicitly, such as omitted variables, serial correlation itself in
a dependence of the stochastic disturbance term in one period on that in another period.
Various tests, to be described later, were performed to detect the serial-correlation and
suitably interpret the results of the specifications.
Fourth is the errors-in-measurement problem – that data are measured subject to various
inaccuracies and biases. In fact, data are sometimes revised because of later recognition
of these inaccuracies and biases. More fundamentally, potential inaccuracies result from a
lack of precision in conceptualization. For example, the GNP accounts are revised from
time to time on the basis of such changes in conceptualization (e.g. defining what is
included in consumption). Such changes in conceptualization necessitate refining the data
to make them comparable and consistent over time. Also, as has been mentioned
previously, the reporting of data related to catastrophes may be inaccurate and biased due
to several factors related to political, sociological, and anthropological factors.
To address these issues, the Extreme Bounds Analysis is used and this is described in the
following.
4.7 LIMITATIONS OF CROSS-COUNTRY REGRESSION STUDIES
There are substantial conceptual and statistical problems that plague cross-country
investigations (Levine and Renelt, 1992). Levine and Renelt (1992) point out that
statistically entries are sometimes measured inconsistently and inaccurately. Even putting
measurement difficulties aside, it is not clear whether we can include countries as diverse
as Bangladesh and Canada in the same regression. These countries operate in different
policy regimes and under different environments. A country may be at a particular stage
in a business cycle, or may be undergoing major policy changes, or experiencing political
Chapter Four: Empirical Studies
163
disturbances. All these factors affect economic activity and consequently economic
growth. Researchers (Barro, 1991; Easterly, 1997) have found that many individual
indicators of monetary, fiscal, trade, exchange-rate, and financial policies are
significantly correlated with long-run growth in cross country growth regressions. How
could one evaluate the “believability” of cross-country regressions? Extreme bounds
analysis (EBA) based Edward Leamer’s (1983, 1985) work can be used for testing the
results of regressions relating the direct loss to the post event indicators of the economy.
The EBA employs a linear, ordinary-least-squares regression framework. The variables in
the vector E are chosen from a set of indicators, which are known to affect the long-run
economic growth rate. The EBA involves varying the E variables to determine whether
that coefficient on the damage indicator, D, is consistently significant and of the same
sign when the right-hand-side variables change. If β2 is consistently significant and have
the same sign the results are termed as “robust”; otherwise the results are “fragile.” The
EBA is used to test the robustness of the empirical associations between the loss-GDP
ratio and various economic indicators. The results of these regressions are presented next.
4.8 RESULTS FROM REGRESSION ANALYSIS
Details of the regression results for growth changes and the effects of catastrophes
on consumption, savings, government expenditure, inflation and real interest rates are
presented in Appendices F to I. For these regressions various specifications are reported.
4.8.1 Growth rates – Short term
Based on the specifications (21 in all) for examining the effect of a catastrophe on
the short-term growth rate of an economy that the direct loss term enters statistically
significantly. Complete details of the specifications and the regressions are presented in
Appendix F, in electronic form. Table 4.3 presents one such specification. For this
particular specification, the coefficient for the loss term (b1) is –2.37 and highly
significant. Other specifications reveal that the coefficient (b1) ranges from –3.9 to –1.7
with a mean of –2.9. The coefficient remains highly significant (<0.001) in all
specifications. The coefficient for percentage of population affected is also statistically
Chapter Four: Empirical Studies
164
significant but is positive. Dummy for earthquakes indicate that they are associated
positively and significantly with post event growth rate, whereas droughts are negatively
associated with post event growth rate.
Summary|R| 0.543R2 0.295R2 adjusted 0.271Standard Error 3.486# Points 151PRESS 1963.96R2 for Prediction 0.214Durbin-Watson d 1.787First Order Autocorrelation 0.106Collinearity 0.334Coefficient of Variation 105.286
ANOVASource SS SS% MS F F Signif df
Regression 736.81 29 147.36 12.12 8.030e-10 5Residual 1762.30 71 12.154 145Total 2499.1 100 150
P value Std Error -95% 95% t Stat VIFb0 8.73 0.000 1.604 5.56 11.90 5.4b1 -2.37 0.000 0.500 -3.36 -1.38 -4.7 2b2 60.47 0.003 19.796 21.34 99.60 3.1 1b3 -3.10 0.000 0.511 -4.11 -2.09 -6.1 2b4 0.42 0.001 0.128 0.16 0.67 3.2 1b5 -1.32 0.006 0.475 -2.25 -0.38 -2.8 1
Table 4.3 Result of a typical regression used for testing the negative association of economic loss to post-event growth
Gr1yrAfter = b0 + b1*Log10(Loss/GDP) + b2*Log(1+TotoAff/Pop) + b3*Log10AvgGDP_per_capita + b4*COV_AnnualGrowth + b5*Log10Std_Dev_Inflation
Chapter Four: Empirical Studies
165
Some of the specifications control for immediate (one year preceding) pre event
economic conditions with indicators such as the pre event growth rate, the pre event gross
domestic fixed investment growth, and the pre event government size. The coefficients of
pre-event growth rate and the pre event gross domestic fixed investment growth enter
positively and significantly in explaining the post-event growth rate, as expected. One
point of the pre event growth rate explains 0.39 to 0.79 point of the post event growth
rate. One point growth in the pre event gross domestic fixed investment is associated with
0.22 points of post event growth rate. Greater share of the government expenditures in the
GDP appears negatively associated with the post-event growth rate.
Control variables such as measures of civil liberties, bureaucratic quality, black market
exchange rates, percentage of population without schooling are averages over longer
periods of time, typically five to ten years around the event. The inherent assumption is
that these variables change at a much slower rate than other macro-variables like the
annual percentage growth rate. As has been discussed in chapter two, these factors
nevertheless determine the vulnerability of a country to natural hazards, which in turn
determines the post event economic behavior. The regressions present econometric
evidence for associations between indicators of ongoing social, economic, and political
processes and the post event behavior.
Indicators of the monetary health of an economy, namely the inflation variability, the
monetary growth (average annual growth rate of the money supply during the last five
years minus the potential growth rate of real GDP) is negatively associated with growth.
Better bureaucracies are associated with higher post event growth. Indicators for
government enterprises (higher ranks imply lower role and presence of government
owned enterprises) are positively associated with post event growth. Greater civil
liberties and political rights are also positively and significantly associated with post
event growth. Better health (as indicated by the daily protein/calorie intakes) is also
positively associated with growth. Lack of education is negatively and significantly
Chapter Four: Empirical Studies
166
associated with post event growth. The signs and significance of the determinants of
growth appear as expected and in accordance with the discussion in Section 4.5.
4.8.2 Growth rates – Average
Table 4.4 presents the regression results for average growth rates. As has been
already mentioned, average growth rates refer to means of growth rates three years prior
to and after the event. It can be seen from Table 4.4 that the loss term appears negatively
and significantly in all the specifications. The coefficient ranges from –1.95 to –0.67 with
a mean of –1.28 over the specifications. This again implies that loss term is negatively
correlated with post event growth. Simulations based on Ramsey’s model indicate that
(Fig. 2f) loss is positively correlated with the post-event growth. Only if the effects of the
efficiency of the post event reconstruction are taken into account, as described in Section
3.3 using an extended Ramsey’s model, can the negative correlation between loss and the
post-event growth be explained. This brings out the importance of modeling the transient
processes immediately after the event.
Dummy for earthquakes indicate that they are associated positively and significantly with
post event growth rate. Droughts are negatively associated with post event growth rate.
Earthquakes typically result in capital being damaged or destroyed. Droughts do not
cause relatively damage to capital stock. After an earthquake, reconstruction activities
may have a positive effect on the region’s productivity. Such a change in productivity
was assumed in the theoretical models developed in Chapter 3. Numerical simulations
indicated that increases in productivity (Figs. 3.1f and 3.3h) result in increases in post
event growth rates. Empirical evidence that earthquake dummy is positively correlated
with the post-event growth rate lends support to the theoretical result that capital
regeneration after an earthquake increase the post event growth rate. This is further
reinforced by the fact that drought dummy is negatively correlated.
Chapter Four: Empirical Studies
167
Summary|R| 0.633R2 0.401R2 adjusted 0.376Standard Error 2.276# Points 151PRESS 862.82R2 for Prediction 0.307Durbin-Watson d 1.382First Order Autocorrelation 0.308Collinearity 0.311Coefficient of Variation 60.682
ANOVASource SS SS% MS F F Signif df
Regression 499.10 40 83.18 16.05 4.387e-14 6Residual 746.27 60 5.182 144Total 1245.4 100 150
P value Std Error -95% 95% t Stat VIFb0 3.54 0.217 2.854 -2.10 9.18 1.2b1 -1.68 0.000 0.327 -2.33 -1.04 -5.1 2b2 38.22 0.005 13.264 12.01 64.44 2.9 1b3 -2.72 0.000 0.336 -3.39 -2.06 -8.1 2b4 -0.73 0.021 0.311 -1.34 -0.11 -2.3 1b5 4.17 0.024 1.836 0.54 7.80 2.3 1b6 0.31 0.000 0.084 0.14 0.47 3.6 1
AvgAfter = b0 + b1*Log10(Loss/GDP) + b2*Log(1+TotoAff/Pop) + b3*Log10AvgGDP_per_capita + b4*Log10Std_Dev_Inflation +
b5*Log10AvgGrossCapitalFormation_%GDP + b6*COV_AnnualGrowth
Table 4.4 Result of a typical regression used for testing the negative association of economic loss to post-event average growth
Chapter Four: Empirical Studies
168
The coefficient of average pre-event per capita income enters negatively and significantly
in explaining the post-event growth rate, as expected. Greater is the pre-event per capita
income, smaller is the post-event growth.
Indicators of the monetary health of an economy, namely the inflation variability, the
standard deviation of growth over the past ten years is negatively associated with post
event growth. These indicators of the susceptibility of the economy to price volatility in
an economy enter negatively because price increases after the event result in lower
productivity.
Better bureaucracies are associated with higher post event growth. One possible reason is
that better bureaucracies will fuel the post event productivity and hence growth. Lack of
corruption enters positively and significantly in the post event growth rates. Indicators for
government enterprises (higher ranks imply lower role and presence of government
owned enterprises) are positively associated with post event growth. Greater civil
liberties and political rights are also positively and significantly associated with post
event growth. Better health (as indicated by the daily protein/calorie intakes) is also
positively associated with growth. Lack of education is negatively and significantly
associated with post event growth. The signs and significance of the determinants of
growth appear as expected and in accordance with the discussion in Section 4.5.
Chapter Four: Empirical Studies
169
4.9 EFFECT ON MAJOR ECONOMIC INDICATORS
After examining the data on economic growth, the effects of catastrophes on
major economic indicators such as the consumption, investment, government
consumption, inflation, and the real interest rates are examined. Details of these
regressions are presented in electronic form in Appendix G.
4.9.1 Consumption
It is clear from Table 4.5 that direct loss is positively and significantly associated
with consumption. The coefficient of the loss term has a minimum value of 2.0 and a
maximum of 3.2 with a mean of 2.6 over the specifications. This implies that a direct loss
of 10% of GDP is associated with 3.21 point increase in consumption. This response to
changes in income from catastrophes is further explained in the subsequent sections.
4.9.2 Investment
An examination of the specification (Table 4.6) reveals that catastrophes result in
lowering the amount of investments by concentrating on increases in consumption
expenditures.
4.9.3 Government expenditure
The loss term enters positively and significantly in all the specifications with a
mean of 1.14. A typical relation is as follows:
GovernmentConsumptionafter = -4.33
(0.002)
+ 0.86*GovernmentConsumptionbefore+1.4*Log(Loss/GDP)
(0.000) (0.011)
+ 2.395*Eq. + 0.559*GovtEnterp
(0.004) (0.006)
N=100; R2= 0.818; F=38; DW = 1.925
This implies that a direct loss of 10% of GDP is associated with 1.4 point increase in
government expenditures.
Chapter Four: Empirical Studies
170
Table 4.6 Effect of catastrophes on investmentsSummary
|R| 0.915R2 0.837R2 adjusted 0.835Standard Error 2.477# Points 149PRESS 951.04R2 for Prediction 0.827Durbin-Watson d 1.508First Order Autocorrelation 0.243Collinearity 0.814Coefficient of Variation 12.814
ANOVASource SS SS% MS F F Signif df
Regression 4592.7 84 2296.4 374.14 3.469e-58 2Residual 896.11 16 6.138 146Total 5488.8 100 148
AvgAfter = b0 + b1*Log10(Loss/GDP) + b2*AvgBeforeP value Std Error -95% 95% t Stat VIF
b0 1.998 0.0056 0.71 0.60 3.40 2.81b1 -0.896 0.0020 0.28 -1.46 -0.33 -3.15 1.23b2 0.791 0.0000 0.03 0.72 0.86 23.16 1.23
Table 4.5 Effect of catastrophes on change in consumption Summary
|R| 0.24R2 0.06R2 adjusted 0.05Standard Error 0.04# Points 150.00PRESS 0.20R2 for Prediction 0.03Durbin-Watson d 1.87First Order Autocorrelation 0.05Collinearity 1.00Coefficient of Variation 3.61
ANOVASource SS SS% MS F F Signif df
Regression 0.01234 6 0.01234 9.396 0.00259 1Residual 0.194 94 0.00131 148Total 0.207 100 149
ChangeInRealConsumption = b0 + b1*Log10(Loss/GDP)P value Std Error -95% 95% t Stat
b0 1.029 0.000 0.009 1.012 1.046 121b1 0.011 0.003 0.004 0.004 0.019 3.065
Chapter Four: Empirical Studies
171
4.9.4 Inflation, and Interest rates
The regression associating the loss term and the inflation is as follows:
Log(Inflation)after = 0.18 + 0.89 *Log(Inflation)before+ 0.07*Log(Loss/GDP) (0.121) (0.000) (0.057)
N=114; R2= 0.686; F=204; DW = 1.68 This implies that a direct loss of 10% of GDP is associated with 0.07 point increase in log(inflation)
Fig. 4.11 clearly illustrates the fact that the post-event real interest rates are higher
than the pre-event counterparts. Details on the regressions for examining the effect
on interest rates is presented in electronic form in Appendix H. The impact of
catastrophes on real interest rates is presented in Table 4.7. It is clear from the
table that catastrophes, as measured by the number of people affected, are
positively associated with post event real interest rates. Pre-event levels of
inflation and per capita income are positively associated with the post event real
interest rates. Higher the pre event gross investment in fixed capital, lower is the
post event interest rate. And if household spend more money, smaller will be the
real interest rate. An important conclusion of this section is that catastrophes and
financial markets may not be totally uncorrelated after all. More research is
required to unravel the connections between the catastrophes and financial
markets.
Summarizing the results of this section, greater loss-to-GDP ratios are positively
and significantly associated with increases in consumption, government
expenditure, and inflation, and real interest rates and are negatively and
significantly associated with investments. The effect of catastrophes on a host of
other economic indicators is presented in electronic form in Appendix I. These are
not discussed explicitly herein, but they indicate negative effects of a catastrophe.
Chapter Four: Empirical Studies
172
Fig. 4.11 There is perceptible increase in real interest rates after a catastrophic loss
0%10%20%30%40%50%60%70%80%90%
100%
-5.0 0.0 5.0 10.0 15.0 20.0
Real Interest rates
AvgBeforeAvgAfter
Table 4.7 Effect of catastrophes on the real interest ratesSummary
|R| 0.581R2 0.337R2 adjusted 0.306Standard Error 3.424# Points 112PRESS 1626.91R2 for Prediction 0.132Durbin-Watson d 1.311First Order Autocorrelation 0.343Collinearity 0.585Coefficient of Variation 62.590
ANOVASource SS SS% MS F F Signif df
Regression 632.53 34 126.51 10.79 2.097e-08 5Residual 1242.5 66 11.72 106Total 1875.0 100 111
RealIntRateEventYear = b0 + b1*Log10(1+TotAff/Pop) + b2*LogInflationCPI + b3*Log10AvgGDP_per_capita + b4*AvgGrossCapitalFormation_%GDP +
b5*Household_final_consumption_expenditure_(annual_%_growth)P value Std Error -95% 95% t Stat VIF
b0 -0.706 0.925 7.437 -15.45 14.04 -0.09b1 9.130 0.003 3.051 3.082 15.18 2.99 1.2b2 3.526 0.000 0.756 2.028 5.025 4.67 1.2b3 1.245 0.031 0.568 0.119 2.372 2.19 1.5b4 -7.396 0.042 3.587 -14.51 -0.285 -2.06 1.2b5 -0.316 0.004 0.106 -0.527 -0.105 -2.97 1.2
Chapter Four: Empirical Studies
173
4.10 Catastrophes and consumption smoothing
The main purpose of this section is to test the validity or otherwise of the
permanent income hypothesis (PIH) during the years surrounding the occurrence of
catastrophe. Do catastrophes cause predictable shifts in consumption? If the nations are
not able to smooth consumption in these adverse circumstances, then policies need to be
devised which will help mitigate the effects of a catastrophe. If there are predictable
shifts in consumption after a catastrophe, then this information could be used to design
policies to smooth consumption after a catastrophe.
Whenever a catastrophe occurs it will cause rational agents to change the way in which
past incomes affect forecasts of future incomes. Consumption depends on expected future
incomes. Flavin’s (1981) result shows that changes in consumption are predictable by
lagged changes in income (the excess sensitivity hypothesis). The first part of the test will
establish whether the catastrophe results in predictable changes in income. Knowing
about the process generating income (in this case the date of occurrence of a catastrophe)
we could generate forecast for consumption change based on lagged values of income
and consumption. There is excess sensitivity if consumption responds to any previously
predictable component of income change (Deaton, 1992: 164). For the purposes of this section cross-sectional data was used. These include differences
in consumption and income three years preceding and following the event. The data was
not pooled since the main intention is to find out whether lagged changes in income can
predict consumption changes with and without the occurrence of a catastrophic event.
Table 4.8 shows the means and standard deviations of income, saving, and consumption
for five years with the catastrophic event year as the third year. The one noticeable
characteristic is the enormous variation in data. This is to be expected since we have
pooled data over a wide spectrum of countries. Another noticeable feature is that the
standard deviation of income and consumption become largest at t+1 and then drop to
their smallest value at t+2, t being the time of occurrence of the catastrophe. For saving
too the standard deviation becomes large at t+1 and drops at t+2. One plausible inference
is that the occurrence of a catastrophe, which results in direct losses comparable to the
Chapter Four: Empirical Studies
174
GDP (typically greater than 1% of the GDP), may induce greater fluctuations in income,
savings, and consumption. Another plausible inference is that effects of catastrophe
attenuate two years after the event as evidenced by the relatively sharp drop in the
standard deviations for income, consumption and savings.
The data in Table 4.9 illustrate the fact that mean growth rates for both income and
consumption fall during the disaster year (t0 - t-1). The variability of the growth rates
increases for the income whereas for consumption the variability remains almost
constant.
Table 4.8 Summary statistics for income, consumption, and savings in the five years enveloping the disaster year Income Consumption Savings Mean s.d. Mean s..d. mean s.d. t-2 5168.5 7711.9 4039.0 5933.1 1187.7 2072.2 t-1 5198.8 7709.0 4100.0 5993.2 1157.0 1990.3 t 5267.8 7823.1 4170.5 6100.3 1153.8 1979.6 t+1 5399.8 8061.3 4228.8 6176.8 1185.2 2070.0 t+2 4707.9 6224.0 3751.0 5108.4 973.8 1238.2
Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of occurrence of the catastrophe and the previous year. Table 4.9 Summary statistics for percentage growth rates for income, consumption, and savings in the five years enveloping the disaster year
Income Consumption
Mean s.d. Mean s.d. (t-1- t-2) 0.64 1.60 0.69 2.24 (t0 - t-1) 0.38 2.34 0.45 2.12 (t+1- t0) 0.87 1.81 0.87 2.11 (t+2- t+1) 0.72 1.74 0.74 2.13
According to the permanent income hypothesis (PIH), changes in aggregate consumption
cannot be predicted by lags in income. Hall (1978) first proposed tests for the PIH by
adopting the technique of regressing changes in consumption using lagged income,
conditional on lagged consumption. For the PIH to hold variables lagged t-1 or earlier,
Chapter Four: Empirical Studies
175
and in particular lags of income should not help predict consumption in period t. The
expression for changes in consumption (Deaton, 1992:83) is:
∆ct = r/(1+r) Σ∞k=0 (1+r)
-k (Et+1 − Et)yt+k
This implies that a change in consumption ought to be the amount warranted by
innovation in expectations about future labor income. Knowledge of the process
generating income will enable us to check this prediction too (Deaton, 1992:84).
Presumably innovations in income occur after a catastrophic event. Do these innovations
predict the change in consumption? What differences do we observe if we compare
consumption change after a catastrophe with consumption change without any
catastrophic event?
In an important paper, Flavin (1981) tested the null hypothesis the truth of the PIH as
expressed in Eq. 4.7, together with an auto regressive specification for the process
governing labor income. Flavin’s ‘excess sensitivity’ hypothesis allows consumption to
respond to current and lagged changes in income by more or less than is required by the
PIH. The measurement of excess sensitivity is the measurement of the extent to which
consumption responds to previously predictable changes in income.
By using Deaton’s (1992:94) specification of regressing the change on consumption on
the lagged change in income we could avoid the unit root problems that may affect the
income generating process:
∆ct = α + β∆yt-1
However time-averaging problems may induce spurious correlation for adjacent
observations of a series that has been first differenced. This implies that Eq. 4.8 may
yield inconsistent estimates because ∆yt-1 is spuriously correlated with ∆ct. To avoid such
problems Deaton (1992) suggests that variables lagged two periods may be used as
instruments. Instrumentation by variables lagged by variables enables us to account for
transitory consumption that may result from a catastrophe.
Chapter Four: Empirical Studies
176
The results are shown in Table 4.10. The first row presents the results of regressing on
changes in consumption on lagged changes in income for the period t-1 (the period before
the event occurs). The second row presents a similar regression using the twice-lagged
changes in income and consumption as instruments. The third and fourth rows repeat the
same for the period when the catastrophe strikes. The rest of the table presents more
regressions for two following periods. From Table 4.10 it is clear that lagged changes in
income does not enter significantly in explaining the changes in consumption for the
periods: t-1, t+1, and t+2. This implies that there is evidence that the PIH is valid for the
above periods. For years immediately after the event (i.e. results for ∆ct), the results
indicate the change in consumption is the amount warranted by innovations in income.
Instrumentation of lagged changes in income with twice lagged changes in income and
consumption still leaves a significant (albeit reduced) coefficient for ∆yt-1. There is
therefore evidence of excess sensitivity once possible timing and transitory consumption
problems have been taken into account. These results can be taken to mean that
innovations in income due to the occurrence of the catastrophe result in predictable
changes in consumption.
Table 4.10 Estimates of consumption changes using lagged changes in income
Constant t Lagged ∆y t R2 F N ∆ct-1 (OLS) 57.66 (2.620) 0.29 (3.16) 0.16 9.95 48 ∆ct-1 (IV) 57.41 (2.458) 0.30 (1.49) - - 48 ∆ct (OLS) 54.00 (2.527) 0.51 (6.40) 0.46 40.89 48 ∆ct (IV) -8.43 (-0.191) 1.23 (3.75) - - 48 ∆ct+1 (OLS) 24.90 (0.821) 0.15 (1.64) 0.03 2.69 48 ∆ct+1 (IV) 50.63 (1.323) 0.18 (1.11) - - 48 ∆ct+2 (OLS) 70.52 (2.829) 0.19 (1.75) 0.04 3.06 48 ∆ct+2 (IV) 64.71 (2.001) 0.26 (0.90) - - 48
Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World
Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of
occurrence of the catastrophe and three years prior and following the event. Place and occurrence of
catastrophe are from Center for Research on Epidemiology of Disasters (Sapir and Misson, 1992 CRED).
The instruments in the IV are ∆yt and ∆ct lagged twice. t-values are shown in brackets.
In this section PIH and excess sensitivity were examined. There is evidence of excess
sensitivity once possible timing and transitory consumption problems have been taken
Chapter Four: Empirical Studies
177
into account. These results can be taken to mean that innovations in income due to the
occurrence of the catastrophe result in predictable changes in consumption.
4.11 Consumption smoothing and savings behavior
A catastrophic loss in a country's income will lead to changes in consumption
only if the savings are not able to offset the income fluctuations. Incomes of LDCs are
both low and uncertain. Losses of incomes of LDCs due to catastrophes may seriously
undermine the ability to smooth consumption. When insurance markets are incomplete
(this is true for most LDCs), saving and credit transactions assume a special role by
allowing households to smooth their consumption streams in the face of random income
fluctuations.
If income can be treated as stationary then PIH implies that savings have a mean of zero.
Assets are built up in advance of expected declines in income, and are run down when
current income is lower than its expected future level. Under the PIH saving acts as a
sufficient statistic for the agent’s future income expectations. Saving behavior contains
information about what nations expect to happen to their incomes. Forecasts of income
conditional on saving help us deal with the fact that representative agents may possess
private more information about future income than does an observer. This helps us to
infer whether the catastrophic events considered are truly unanticipated. If the events
were anticipated, the changes in consumption could well be explained by expected values
of income, which in-turn would have been predicted by the lagged savings. This has
important consequences for policies designed for preparedness against catastrophes
triggered by natural events. If surprises in income due to occurrence of a catastrophe are
unanticipated then consumption will not be smooth even if we use the agent’s private
information. Nations that expect catastrophes (of the type and magnitude considered in
this study) to occur should devise policies for precautionary savings to smooth
consumption.
As pointed out by Campbell (1987), past savings is a predictor of how income will
change the next period. It is possible that countries anticipate the occurrence of a
Chapter Four: Empirical Studies
178
catastrophe triggered by natural hazards. Though natural hazards occur with certain
regularity, their magnitude and point of occurrence remains uncertain. But a good
preparedness program in place would help nations to smooth their income. By regressing
the change income with lagged savings and comparing the no-disaster year with the
disaster year we could infer about the efficiency of the precautionary savings of the
countries to income shortfalls from a catastrophe triggered by a natural hazard. From
Table 4.11 it is clear that before the catastrophe occurs, lagged savings do not explain the
income change. This situation, however, changes one year after the event. The coefficient
for lagged savings becomes positive and significant in explaining income changes. The
value of the coefficient again drops two years after the catastrophe. This means that the
catastrophe changes the ex ante saving behavior at least for two years after the event.
Consumption change regressions (Table 4.12) show that it is positively related to lagged
values of savings. Before the catastrophic event the coefficient on savings has a lower
significance than two years after the event. One plausible inference is that the changes in
consumption due a catastrophe are only weakly anticipated for the collection of events
that have been considered. Since the events cover a large range of loss/GDP ratios (Table
4.13) it is plausible that the data set dampens out effects of unanticipated losses for
LDCs. Three years after the event the significance and magnitude fall to their pre-
disaster levels and lagged savings are not able to explain consumption changes in
accordance with the PIH.
If we use income lagged twice as an instrument in the regression on consumption change
on lagged saving we essentially get the same results.
Table 4.11 Estimates for income changes using lagged savings
Constant t Lagged Savings
t R2 F N
∆yt-1 37.32 (0.94) 0.0061 (0.36) 0.003 0.13 46 ∆yt 26.21 (0.67) 0.0520 (3.07) 0.17 9.45 46 ∆yt+1 -10.69 (-0.38) 0.1249 (10.22) 0.70 104.4 46 ∆yt+2 -22.71 (-0.64) 0.1031 (4.52) 0.31 20.43 45
Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World
Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of
Chapter Four: Empirical Studies
179
occurrence of the catastrophe and two years prior and following the event. Place and occurrence of
catastrophe are from Center for Research on Epidemiology of Disasters (Sapir and Misson, 1992 CRED). t-
values are shown in brackets.
Table 4.12 Estimates for consumption changes using lagged savings Constant t Lagged
Savings t R2 F N
∆ct-1(OLS) 20.74 (0.66) 0.034 (2.57) 0.11 6.62 48
∆ct-1 (IV) 26.34 (0.82) 0.029 (2.11) - - 48
∆ct(OLS) 9.42 (0.45) 0.053 (5.83) 0.42 33.95 48
∆ct(IV) 6.67 (0.31) 0.055 (5.78) - - 48
∆ct+1(OLS) 20.02 (0.88) 0.067 (6.76) 0.49 45.71 48
∆ct+1(IV) 16.83 (0.73) 0.069 (6.71) - - 48
∆ct+2(OLS) 21.21 (0.66) 0.020 (1.51) 0.03 2.27 48
∆ct+2(IV) 11.30 (0.35) 0.028 (2.01) - - 48 Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World
Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of
occurrence of the catastrophe and three years prior and following the event. Place and occurrence of
catastrophe are from Center for Research on Epidemiology of Disasters (Sapir and Misson, 1992 CRED).
The instrument in the IV is income lagged twice. t-values are shown in brackets.
In this section the efficiency of the precautionary savings of the countries to income
shortfalls from a catastrophe triggered by a natural hazard is examined. Results from
regressions of income change on lagged savings and comparison of the no-disaster year
with the disaster year are used for arriving at conclusions. Before the catastrophe occurs,
lagged savings do not explain the income change. But one year following the event,
lagged savings anticipate income changes. Evidence is presented to show that
catastrophes change ex ante saving behavior at least for two years after the event.
4.12 Conclusions, Extensions, and Limitations
The problem of finding empirical regularities in the ongoing socioeconomic
processes after the occurrence of a catastrophe was addressed in this chapter.
Connections between these statistical regularities and the results of the theoretical model
simulations presented in Chapter 3 were made. The results of the regression analysis
indicate that by studying disasters much can be learned about the way large-scale socio-
Chapter Four: Empirical Studies
180
economic systems affect and are affected by the occurrence of catastrophes. By making a
cross-country study with countries from all income groups affected by different types of
natural hazards, the results are expected to be sufficiently general. Previous empirical
results from the literature on the determinants of economic growth and on economic
development helped in identifying the explanatory control variables.
The main results of this study can be summarized as follows:
Summarizing the regressions on growth the following statistical regularities are
discerned:
• The models indicate very significant negative coefficient for the direct loss variable in
regressions for short-term growth. The coefficient for the loss variable in the long-term
growth has a lower significance, but remains negative. The magnitude of the coefficient
in the average growth rate regression is less than the short-term regression. This implies
that the associations between the loss term and the economic growth rate become harder
to detect with the passage of time. These results corroborate the results obtained by
simulating the model presented in Section 3.3.
• The pre-event economic growth rate is positive and very significantly associated with the
post-event growth rate, in both the short-term and average regressions. This implies that,
other variables being constant, an economy with a sufficient growth rate can absorb the
effect of a catastrophe. Growth itself is an indicator of the robustness of ongoing
developmental processes. This brings out the importance of having a robust
developmental process in place in absorbing the effect of a catastrophe. The coefficient
for pre-event general government consumption is significant and negative. This agrees
with the known fact that heavy consumption by the government sector retards growth.
• The coefficient for the percentage of people affected is positive and significant in short-
term growth regressions. Though this seems odd, it is should be noted that a catastrophe
affects many people only in developing countries. The amount of aid is to a certain extent
decided by the figures regarding people affected. It is probably this external aid
associated with the percentage affected that spurs growth. As the models described in
Section 3.3 and 3.4, greater inflow of aid results in greater growth.
Chapter Four: Empirical Studies
181
• The coefficient for daily protein/calorie intake appears positive in the short-term growth
regressions associating a healthier community with a more robust developmental process
• If the institutions of crisis management can be proxied by a combination of the size of the
government and the efficiency of the bureaucracy, then their coefficients are positively
and significantly associated with short- and long term (average) post event growth. This
brings out the importance governmental bureaucracy in mitigating the effects of a
catastrophe.
• The coefficient for inflation variability, which is a measure of the monetary robustness of
an economy, is associated negatively and significantly with the post event short- and
long-term growth. This once again ascertains the importance of the ongoing economic
processes in explaining the post-event economic behavior.
• Other factors including civil liberties, percentage of no schooling, economic freedom
index, freedom from corruption, and land-area had the expected signs.
The main results of examining the effects of catastrophes on consumption, investment,
government expenditure, net exports, inflation, interest rates gave the following results:
Large economic losses as a proportion of the GDP are associated with:
1. greater post-event consumption,
2. greater post-event government expenditure,
3. smaller post-event investments,
4. higher inflation, and
5. an increase in real interest rates.
Innovations in income due to the occurrence of the catastrophe result in predictable
changes in consumption. The efficiency of the precautionary savings of the countries to
income shortfalls from a catastrophe triggered by a natural hazard is examined. Results
from regressions of income change on lagged savings and comparison of the no-disaster
year with the disaster year are used for arriving at conclusions. Before the catastrophe
occurs, lagged savings do not explain the income change. But one year following the
event, lagged savings anticipate income changes. Evidence is presented to show that
catastrophes change ex ante saving behavior at least for two years after the event.
Chapter Four: Empirical Studies
182
There are limitations of the study, which are discussed in the following. The first is
regarding the heterogeneity and panel data that arise naturally in cross-country studies.
Omitted heterogeneity induces correlations between explanatory variables and the error
term in a way that has the same consequences as simultaneity bias. The factors that
appear on the right hand side of the specification (Eq.4.6) such as pre event growth may
have no general claim to exogeneity. The combination of genuine simultaneity and
heterogeneity has the further effect of ruling out the use of lags to remove the former.
These considerations would typically require further examination of the effect of
catastrophe on the economic indicators using alternative specifications based on first
differences. Another important limitation is the lack of appropriate instruments, which
are correlated with direct loss term but un-correlated with error term. These instruments
can be used to check whether the coefficients on the loss terms remain robust when they
are instrumented. If data on sectoral distribution of losses is available, this can be used to
instrument the direct loss variable. In other words, this requires details regarding losses in
the agriculture, industry, and service sectors. But such data is hard to obtain. It would be
ideal to develop a system of structural equations to explain the connections between all
the macro-economic variables affected by catastrophes. Lack of underlying theoretical
models forces us to use reduced form equations. These result in inference of statistical
regularities as opposed to full-fledged causal models. Increase of representation in the
sample of higher loss-GDP ratio events is required for the sake of generality.
Chapter Four: Empirical Studies
129
Chapter Four
An Empirical Study of the Macro-economic Effects of Catastrophes Triggered by Natural Events
4. INTRODUCTION
In this chapter we re-examine our understanding of the effects of catastrophes on
the economy based on empirical evidence. Questions addressed include the change or
absence thereof, in economic growth, consumption, saving, inflation, and real interest
rates. Data on these economic indicators are compiled for various countries for periods
immediately preceding and following the occurrence of a catastrophe. Data regarding
catastrophes such as the estimates of direct losses is also compiled. The regression
analysis employed suggests that catastrophes are negatively associated with all the
aforementioned economic indicators.
In order to study the effect of a catastrophe on an economy the factors that describe
socio-economic conditions prior to occurrence of the hazard event have to be identified.
The vulnerability of a society to natural hazards is the result of various on-going
economic, social, and political processes, as has been discussed in Chapter 2. For large
segments of the world's underdeveloped population, occurrence of a natural hazard may
worsen an already deteriorating or fragile situation. In such regions even a moderate
hazard, such as the 1985 Mexico earthquake, could trigger a catastrophe. Oliver-Smith
(1994) brings this out clearly in his analysis of the 1970 Peru Earthquake. He points out
that Peru's catastrophe was some 500 years in the making, rooted in the complex of
economic and political forces that structured development and the human-environment
relations. The earthquake and subsequent landslides was a trigger for a catastrophe
grounded in poverty, political oppression, and the subversion of previously sustainable
indigenous practices (Bolin and Stanford, 1998).
Socioeconomic conditions in a region are mainly as a result of the developmental
processes. The effect of a major catastrophe on the developmental process is complex,
Chapter Four: Empirical Studies
130
especially for developing regions. Globally, economies are evolving ‘complex’ systems.
This complexity in the economic systems is the result of the historical geography, the
political economy, the increased interdependencies among various sectors and regions of
an economy facilitated by the quantum leaps in the communication technology, and the
rapid globalisation of trade. In order to study the effect of a catastrophe on an economy
the factors that describe socio-economic conditions prior to occurrence of the hazard
event have to be identified. General statements regarding the economic consequences of a
catastrophe can be made only when these complexities are appropriately modeled.
An overview of some studies, which partly address these questions, is presented in the
next section. In section 4.2 connections between the occurrence of a catastrophe and
ongoing development processes of an affected region are made. Section 4.3 describes the
data used for the present study. The general framework and the particular econometric
model used to estimate the effect of catastrophes are presented in the next section.
Various factors that affect the growth rate are then presented. Section 4.5 presents a
discussion of various factors that may be important in determining the post event
economic indicators. Results of regression analysis are discussed in Section 4.6.
Chapter Four: Empirical Studies
131
4.1 PREVIOUS STUDIES
Studies on the effects of natural hazards on an economy have discussed direct and
indirect losses that result from such events (Development Technologies, 1992). Direct
losses are usually associated with direct physical damage and secondary effects, such as
damage caused by fire following an earthquake. Indirect damages relate to the effect on
flows of goods that will not be produced and services that will not be provided after a
catastrophe. They are measured in monetary terms. The impact of the catastrophe on
overall economic behavior, which has sometimes been termed as secondary effects, is
measured by changes in macro-economic variables. The work reported in this dissertation
focuses on secondary effects.
There are few studies on the macro-economic effects of catastrophes. They are based on
small data sets. Moreover, the conclusions are seemingly contradictory. Albala-Bertrand
(1993:163) argues "GDP normally does not fall after a disaster impact and if anything
tends to improve at least for a couple of post-disaster years." Albala-Bertrand's study
(1993) is based on a sample of catastrophes that occurred in the 1970's in mostly
developing countries. He uses three criteria for examining the effect of catastrophes on
economic growth, investment and sector outputs, public finance, and balance of
payments. The three criteria include: i) examining the change in the indicators according
to sign (positive meaning 'growth') and direction of change (up meaning 'acceleration'),
ii) the figures are averaged in per country terms for each period, and iii) comparison
between pre- and post-disaster averages. Limited by sample size, no other statistical
inferential procedures are used. The hypothesis he proposes is not validated since there
could be many factors that explain post-event economic behavior. For example, a country
might have experienced increased growth after an event because of reasons totally
unrelated to the occurrence of a catastrophe or due to efficient reconstruction policies.
However, this does not imply that a similar economy will sustain economic growth in the
absence of efficient reconstruction. Inferences from cross-country data are general only if
they are ‘normalized’ using control and environmental variables.
Chapter Four: Empirical Studies
132
The World Disasters Report (1997) expresses an apparently opposite viewpoint. The
report states, “Caribbean disasters can be costly, especially as a proportion of GDP. The
impact on national economies has been significant: hurricanes between 1980 and 1988
effectively reversed the growth rates.” This statement is again based on a simple
comparison of average growth for the affected countries between 1980-88 and 1989-91
(Table 4.1). All the five countries are small islands, which makes it difficult to generalize
the result.
Taken together these studies produce ambiguous conclusions regarding the effect of
catastrophes on ongoing economic processes.
Friesema et al. (1979) is an early study to analyze the effect of disasters on the long-term
growth patterns of four cities - Conway, Galveston, Topeka, and Yuba City. Their null
hypothesis is that disasters had no significant effect on employment, small business
activity (number of gas stations and restaurants), retail sales, and public finance. They
examine a time series of the indicators for a time period ten years before and after an
event. They conclude that local economic behavior patterns, barring slight disruptions,
were scarcely interrupted by the disaster events considered. They also mention that their
results are not surprising since in all the four cases the basic capital stock remained, and
the production process continued. This makes their sample unrepresentative of post
catastrophic economic behavior.
Table 4.1: Disasters in the Caribbean can have a significant impact on GDP and growth (World Disasters Report, 1997)
Country Average growth rate GDP 1980-88
Average growth rate GDP 1989-91
Dominica 4.9 4.3 Montserrat 3.7 -4.4 St.Kitts/Nevis 6.0 4.9 Antigua/Barbuda 6.8 2.2 Jamaica 5.0 0.8
Chapter Four: Empirical Studies
133
Wright et al. (1979) examine data for over 3100 counties in the US for effects of disasters
on growth trends of population and housing. Damage inflicted by the typical disaster in
their sample affected only a small proportion of structures, enterprises, and households of
typical counties. Based on regression studies they conclude that there are no significant
effects on growth trends in population and housing. However, these findings have been
questioned by the research of Yezer and Rabin (1987), who distinguish between
anticipated and unanticipated disasters. Their hypothesis is that “expected” disasters,
those occurring at a rate predicted by historical experience in a region, have no impact on
migration – such expectations have already been reflected in trend rate of migration. In
contrast, “unexpected disasters”, a spate in excess of those predicted by historical
experience, discourage migration. Empirical testing that explicitly distinguishes
“anticipated” from “unanticipated” supports the hypothesis.
The inferences from these studies cannot be generalized to effects of catastrophe in a
developing economy for several reasons. Firstly, the studies concentrate on regional
localized effects in a developed country. Secondly, the direct loss reported in the studies
is relatively small compared to the overall capital stock of the affected region. Finally,
they only examine changes in a subset of indicators that describe the social and economic
conditions of a region.
Chapter Four: Empirical Studies
134
4.2 CATASTROPHES AND ONGOING DEVELOPMENT PROCESSES
Losses from a catastrophe may be readily absorbed by a developed economy. To
cite an example, the Northridge earthquake occurred in a state with a Gross Regional
Product ranked 6th largest in the world. A US $30 billion direct loss due to the earthquake
manifested itself as a minor perturbation. This contrasts with the devastating Third World
disasters such as the 1976 Guatemala earthquake or the 1985 Mexico City earthquake. In
both cases, the catastrophes produced national crises with effects well beyond the
immediate physical impacts.
For a developing economy, like Bangladesh, direct losses from a catastrophe, which are
comparable to the Gross Domestic Product (GDP) might divert scarce resources from
development plans to reconstruction. Almost half of the 1988/89 Bangladesh's national
development budget was diverted to pay for ad-hoc relief and rehabilitation programs
(Brammer, 1990) after the 1988 flood. Development plans may include improving health
care, education, food supply, and institutions for crisis management. As Bates and
Peacock (1993) point out catastrophes "intervene in the development process as it
pertains to other important adaptive problems, and they redirect, deflect, retard, and on
rare occasions accelerate the development process."
The deep indebtedness of many Third World countries has made the cost of
reconstruction and the transition from rehabilitation to development unattainable. To see
how foreign debt burden can adversely affect the loss that a country suffers, take the case
of Jamaica struck by Hurricane Gilbert in 1988 (Blaikie, et al. 1994). Prior to the
Hurricane Gilbert, part of Jamaica's debt burden was in part due loans used to pay for
damages from previous hurricane. Jamaica introduced a structural adjustment program
that typically involved cuts in public spending. Services such as education, health, and
sanitation were reduced. Government programs to introduce preparedness or mitigation
measures were also cut as result of economic constraints. These decisions greatly reduced
the ability of the community to recover from the effects of a major hazard like Hurricane
Gilbert.
Chapter Four: Empirical Studies
135
Foreign debt also forced the government to intervene in the financial sector that resulted
in an increase of interest rates to over 20% and home mortgage rates ran between 14-
25%. Government forced rent control and import duty on construction materials. This
resulted in a rapid decline in new construction and other maintenance activity. The
quality of new construction also declined, since contractors tried to maximize profit by
using unsafe practices. This may have been partly responsible for the huge magnitude of
losses observed.
Delica (1993) brings out the relation between disasters and economic growth based on
her study of the natural hazards affecting Philippines. She argues that disasters have
practically negated the real economic growth achieved during the administration of
Carazon Aquino. From 1986 to 1991, damage to infrastructure, property, agriculture, and
industry from disasters were enormous, averaging about 2% of the GNP. Using simple
arithmetic, she argues that with an annual population growth of 2.3%, the economy needs
greater than 4.3% annual growth simply to maintain per capita income levels. But the
economy had only about 4% average annual growth, with the result vulnerability to
disasters has increased rather than decreased. This is because Philippines’ foreign debt
obligations have increased, from $26 billion in 1985 to $29 billion in 1992. The
government's spending on relief and rehabilitation has been tightly controlled and
increasingly dependent on external sources. Government's development strategy puts a
premium on export-orientation and attraction of foreign investment. This is at the
expense of ecological sustainability and environmental protection. Out of the 54% forest
cover required for a stable ecosystem only 20% remains as a result of deforestation. This
in turn increases the severity of floods and landslides.
Many poor countries try to solve their debt problems by adopting national policies
favoring raw material export. This typically results in land degradation since new land is
cleared for ranching and commercial cropping. Land degradation increases vulnerability,
which in turn increases the potential for catastrophic losses.
Chapter Four: Empirical Studies
136
Long-term development projects may be adversely affected by diversion of resources to
help an affected community rebuild. Twigg (1998) reports that the World Bank diverted
some $2 billion of existing loans between the 1987 and 1988 financial years to fund
reconstruction and rehabilitation after catastrophes triggered by natural events.
Catastrophes reveal the robustness or vulnerability of a country's socioeconomic
conditions. Various indicators can be used to quantitatively measure robustness or
vulnerability. The importance of these parameters, which are perhaps ignored in less
turbulent times, is revealed, tragically, only after a catastrophic event. A catastrophe can
unmask social and economic inequalities that come to the fore in the distribution of relief
aid. Catastrophes usually result in worsening the pre-event economic inequalities. It is
important to identify the factors that can be associated with vulnerability that explains the
wide variety of post event economic behavior. For example, we could examine the role of
infrastructure in changes in post-event economic growth. As has been pointed out by
Hewitt (1983), catastrophes are shaped and structured by economic, social, political and
cultural 'practices' and processes that existed prior to the occurrence of a physical event.
Indicators that describe some of these initial practices and processes need to be identified.
Whether there exists empirical evidence to support the hypothesis that ongoing
socioeconomic processes determine the post event economic behavior will be examined
in this chapter.
4.2.1 Change in Indicators Due to Catastrophes
In the past decade there has been an explosion of empirical studies of growth and
development. Efforts have been made to account for differences in growth rates between
various countries using indicators of education, health, infrastructure, institutions, and
political freedom. Results from these studies will be used to identify variables that can
make cross-country comparisons of changes in macro-economic indicators possible. The
parameters will act to control some of the variability across countries. Any effect due to
catastrophes on macro-economy can be detected only after the control variables
explained variability from other sources.
Chapter Four: Empirical Studies
137
As has been mentioned previously, there is an intimate relation between ongoing
development processes and the occurrence of a catastrophe. The various parameters,
which are associated with development such as education, infrastructure, and health, are
hypothesized as measures of a community's robustness (or pessimistically, vulnerability)
to a catastrophe. A combination of these parameters can be used to assess a community's
robustness. It is reasonable to expect that a robust community's development or growth
should not be adversely affected by occurrence of a catastrophe. The ongoing dynamics
of the developmental processes are capable of absorbing the effects of catastrophe.
Conversely, a society is weak if its development process is adversely affected by the
occurrence of catastrophe triggered by a natural event. The present work relies on
previous studies on the determinants of growth for choosing parameters that are
associated with the development process.
One important indicator of development of a country is its economic growth rate. The
following is a summary of some of the parameters that have been shown to be
determinants of growth. A percentage point in economic growth is associated with the
following:
• Increase of 1.2 years in average schooling of labor force
• An increase in secondary enrollment of 40 percentage points
• A reduction of 28 percentage points in the share of central bank in total credit
• An increase of 50 percentage points in financial depth (M2/GDP)
• An increase of 1.7% of GDP in public investment in transport and communication
• A fall in inflation of 26 percentage points
• A reduction in the government budget deficit of 4.3 percentage points of GDP
• An increase in (exports + imports)/GDP of 40 percentage points
• A fall in government consumption/GDP of 8 percentage points
• An increase in foreign direct investment/GDP of 1.25 percentage points.
(Barro 1991, Barro and Lee 1993, King and Levine 1993, Easterly and Rebelo 1993,
Fisher 1993, Easterly and Levine 1997, Easterly, Loayza, and Monteil 1997, Borensztein,
De Gregorio, and Lee 1994).
Chapter Four: Empirical Studies
138
These inferences are used to identify the variables that can be controlled when making a
cross-country comparison of post-event behavior of the economic growth.
Chapter Four: Empirical Studies
139
4.3 GENERAL FRAMEWORK AND ECONOMETRIC MODEL
The general framework to be used in empirical studies reported here will be
developed in this section. In Chapter 3, theoretical models simulated the occurrence of a
catastrophe as a perturbation of the ‘normal’ economic processes. A catastrophe was
modeled as a reduction of capital and subsequent changes in productivity of the affected
region. The economy was assumed to be initially in its steady state. Inspection of
Eq.3.3.15 reveals that growth of capital due to the catastrophe depends on the steady state
(the Jacobian term) and the perturbations. Growth of the economy in turn depends on the
changes in capital stock. Hence, the following relation is used to estimate the effect of a
catastrophe on the post-event growth rates:
growthwith hazard = f(damage, productivity-changes; y*) (4.1)
y* is the long-run steady-state level of per capita output and depends on the steady state
levels of capital stock, as shown in Eq. 3.2.9. y* depends on an array of choice and
environmental variables. The private sector’s choices include saving rates, labor supply,
and fertility rates, each of which depends on preferences and costs. The government’s
choices involve spending in various categories, tax rates, the extent of distortions of
markets and business decisions, maintenance of rule of law and property rights, and the
degree of political freedom. Also relevant for an open economy is the terms-of-trade,
typically given to a small country by external conditions. A cross-country empirical
analysis requires conditioning on the determinants of the steady states. Also, the pre-
event conditions to a large extent determine the post event productivity. These
determinants or the country specific factors, along with their relation to catastrophes, are
presented in Section 4.4. It is assumed that the country specific factors are invariant over
the period of interest – five years. Data for these factors are typically available as
constants over five- to ten-year periods.
Damage, in general, depends upon the intensity of the hazard and the vulnerability.
Vulnerability is the susceptibility of the exposed constructed facilities, economic and
Chapter Four: Empirical Studies
140
social structures of a region to be affected given a specified level of hazard. As discussed
in Chapter 2, vulnerability is intimately related to ongoing socio-economic processes.
Damage may be expressed as:
damage = h(hazard, vulnerability) (4.2)
It should be mentioned here the relation Eq.4.2 is expected to be highly non-linear. Even
for relatively simple structures such as single-family dwellings, the damage curves –
which relate the hazard intensity to the damage level (RMS, 1996) – are non-linear. Data
regarding the loss of capital and the changes in productivity are hard to come by. Hence
the loss of capital is modeled by the direct losses recorded after the event.
4.3.1 Approximation
The first step to estimate the model expressed in the relations (Eqs.4.1-2) above is
to use an approximate linear relation. Consequently, the relation in Eq.4.1 is
approximated by:
growthwith hazard = α1 + β1E + β2Damage + β3Hazard_type + ε1 …(4.3)
ε1 is an unobserved disturbance term. The indicators for Damage are the direct-loss to
GDP ratio and the percentage of population affected. E is a vector of time-invariant
country specific indicators of the economy that are considered as determinants of
economic growth. The vector E contains indicators from each of the following categories
of determinants of growth - Economic conditions, Individual Rights and Institutions,
Education, Health, Transport and Communications, Inequality across income and gender.
In particular the following indicators are used: Inflation variability, Average pre event
decade growth, SD of pre event decade growth, annual money growth, black market
premium, political rights, civil liberties, bureaucratic quality, government enterprises,
percent “no schooling” in population, daily protein or calorie intake, life expectancy at
age zero, radios per capita, and TVs per capita. Hazard-type is a dummy variable to
account for the type of hazard – earthquake, hurricane, or drought.
Chapter Four: Empirical Studies
141
It should be mentioned here that Damage as such would depend on factors in the vector
E. It is implicitly assumed that the indicators for damage are not correlated with the
factors in E. This may be a strong assumption if the measure of loss is in terms of
destroyed productive capital stock and E includes factors such as capital stock per
worker. Indicators chosen in E are such that they are only indirectly related to direct loss
term. Therefore, the assumption that E and damage are not significantly correlated is
reasonable. It is also assumed that the errors in measurement/estimation of damage are
not correlated with the error term ε1. The reduced form given in Eq.4.3 is estimated.
The results presented in Fig. 3.4h indicate that loss is negatively correlated with the post-
event growth rate. The hypothesis to be tested is that the coefficients β2 in Eq.4.3 are
statistically significant and negative.
Similar models for other economic indicators are estimated where the dependent variable
is chosen to be the post event budget deficit, external debt, resource balance, inflation,
interest rates, or consumer price index. Again, the hypotheses to be tested are that the
coefficients β’2 in Eq. 4.3 are statistically significant. 4.3.2 Summary Statistics and Discussion of the Sample
4.3.2.1 Economic growth
As a first step, the growth rates between two adjacent years are compared, that is,
the growth rate during the event year is compared with the growth rate immediately
preceding year. Both mean and median of the pre-event annual percentage growth are
greater than their post-event counterparts (Table 4.2). Presumably catastrophic events
also induce greater variance for the growth, as evidenced by comparing the pre- and post-
event variances in the growth (Table 4.2). Distribution of the pre- and post- event growth
rates are shown in Fig.4.1.
Chapter Four: Empirical Studies
142
Table 4.2 Summary statistics for short-term growth Pre event Post event
Mean 3.96 3.29Standard Error 0.30 0.33Median 3.80 3.18Standard Deviation 3.70 4.06Sample Variance 13.67 16.46Kurtosis 2.01 2.61Skewness -0.16 -0.61Range 23.36 26.60Minimum -9.10 -12.57Maximum 14.27 14.03Sum 605.25 503.99Count 153 153Confidence Level(95.0%) 0.59 0.65
Fig 4.1 Event year growth is clearly lower than the pre-event year growth
0%10%20%30%40%50%60%70%80%90%
100%
0 2 4 6 8 10
GDP annual growth (%)
Perc
entil
e 1yrBeforeEventYear
Fig 4.2 Average pre- and post-event growths
0%10%20%30%40%50%60%70%80%90%
100%
0 2 4 6 8 10
GDP annual growth (%)
Perc
entil
e Avg3yrsAfterAvg3yrsBefore
Chapter Four: Empirical Studies
143
It is apparent from Fig. 4.1 that the distribution for growth in the event year is shifted to
the left relative to growth one year before the event.
Table 4.3 summarizes the statistics for pre- and post- event average growth. Here again
average post-event growth rate is smaller than pre-event growth rate. But sample variance
of the average post-event growth is smaller than average pre-event growth, indicating
perhaps that the effects of the events are reducing.
Table 4.3 Summary statistics for average growth Average 3 years before
Average 3 years after
Mean 3.83 3.55Standard Error 0.27 0.24Median 3.48 3.61Standard Deviation 3.32 3.00Sample Variance 11.03 9.02Kurtosis 1.89 3.35Skewness 0.15 -0.44Range 22.96 22.66Minimum -9.42 -10.20Maximum 13.54 12.46Sum 586.68 542.90Count 153 153
It is apparent from Fig. 4.2 that the average post event growth is shifted to the left relative
to average pre event growth rates, though in this case the effect is not as pronounced for
growths less than 5%.
4.4.2.2 Effect on consumption, investment, government expenditure, net exports and income
The main components of the GDP are the consumption, investment, government
expenditure and net exports. Using the latest Penn World Table data (2002), the effect of
catastrophes on each of these macroeconomic indicators is investigated. As a first step,
each of these variables is graphed with the loss-GDP ratios. These graphs are shown in
Fig. 4.3 to 4.8.
Chapter Four: Empirical Studies
144
Fig. 4.3 Effect of catastrophes on consumption
40
50
60
70
80
90
100
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
eve
nt c
onsu
mpt
ion
(% o
f GD
P)
Fig. 4.4 Greater losses are associated with larger amount of government spending
05
1015202530354045
0.0% 0.1% 1.0% 10.0% 100.0% 1000.0%
Annual loss as a % of GDP
Post
-eve
nt in
vest
men
t as
a %
of
GD
P
Fig. 4.5 Greater losses are associated with larger amount of government spending
0
5
10
15
20
25
30
35
40
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
eve
nt g
over
nmen
t sp
endi
ng a
s a
% o
f GD
P
Chapter Four: Empirical Studies
145
Fig. 4.6 Greater losses are associated with higher openness
0
20
40
60
80
100
120
140
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
eve
nt o
penn
ess
as a
% o
f G
DP
Fig. 4.7 Larger losses are associated with smaller post event savings
-40
-30-20
-100
1020
3040
50
0.0% 0.1% 1.0% 10.0% 100.0% 1000.0%
Annual loss as a % of GDP
Post
eve
nt s
avin
gs a
s a
% o
f G
DP
Fig. 4.8 Greater losses are associated with lower post event GDP per capita
100
1,000
10,000
100,000
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
-eve
nt re
al G
DP
per
equi
vale
nt a
dult
Chapter Four: Empirical Studies
146
These (Fig. 4.3 to 4.8) depict important observed regularities between magnitude of
losses and post-event macroeconomic variables. For example, Fig. 4.5 depicts the
observation that higher losses are associated with higher post-event governmental
spending as a fraction of the GDP. Fig. 4.8 establishes a clear negative association
between loss magnitude and post-event GDP per capita. Regressions in the later sections
are performed to determine the robustness of these associations by accounting for country
specific factors.
Other variables are also examined. In particular the effect of losses on inflation and real
interest rates are presented in Fig. 4.9 and 4.10, respectively.
The next section discusses the primary and control variables that are used in the
estimation. An overview of linear regression analysis for the estimation of Eq.4.2 along
with the model adequacy checking is presented in Section 4.6. Following that,
econometric evidence associating changes in economic indicators with magnitude of loss,
the percentage affected, and the type of catastrophe is presented.
Fig. 4.9 Larger losses are associated with higher inflation
0.1
1.0
10.0
100.0
1000.0
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Log(Loss/GDP)
Post
Eve
nt In
flatio
n
Chapter Four: Empirical Studies
147
Fig. 4.10 Greater loss ratios are associated with higher post-event real interest rates
-10
-5
0
5
10
15
20
0.01% 0.10% 1.00% 10.00% 100.00% 1000.00%
Annual loss as a % of GDP
Post
eve
nt re
al in
tere
st ra
te
Chapter Four: Empirical Studies
148
4.4 EFFECT ON THE ECONOMIC GROWTH
Relating the magnitude of a catastrophe with a change in the growth of an
economy is very complex since there are many factors that determine the economic
growth (Barro, 1997). Recent research in the determinants of cross-country economic
growth has revealed much regularity. Investment in physical capital, educational
attainment of the population, stable macro-economic policies, open trade regimes, better
developed financial markets are important factors exerting positive effect on growth
(Barro and Sala-i-Martin, 1995). There are several other factors that retard growth -
population growth, political instability, budget deficits, shocks resulting from terms of
trade changes, internal strife, and wars (Rodrik, 1998), policy distortions, government
consumption, and low bureaucratic quality (Commander, et al, 1997). In the following
sections we describe some factors that may explain the variety of observed changes in
ongoing economic processes after a catastrophe. This discussion is similar to the
discussion in Chapter 2 regarding the factors that determine the vulnerability to natural
hazards. The important difference here is that these factors are explained here as factors
that may contribute towards the recovery of a community after a catastrophic event.
4.4.1 Primary variables
Three primary variables are used as indicators of the catastrophe. They are: i) the direct
physical loss, ii) the percentage of population affected, and iii) the type of natural hazard.
4.4.1.1 Direct physical loss
One of the important variables that characterize a catastrophe is the resulting
direct loss. Direct damages include all damage to fixed assets (including property),
capital and inventories of finished and semi-finished goods, and business interruption
resulting from a catastrophe (HAZUS, 1997). Estimation of the macro-economic effects
involves a comparison of economic behavior with and without the change in a
Chapter Four: Empirical Studies
149
community's assets. The direct loss is one measure of the change in community assets
after a catastrophe.
Comparing direct loss across countries necessitates an approach based on purchasing
power parity (PPP). Converting the losses into a common currency, for e.g. the US dollar,
through the use of official exchange rates often misleads cross-country comparisons of
the losses. These nominal exchange rates do not reflect the relative purchasing power of
different currencies, and thus errors are introduced into the comparisons. Using PPP is
one way to obtain a correct measure of losses. In countries where the domestic prices are
low, the losses based on PPP will be higher than that obtained from official exchange
rates. For the purposes of this study we use ratio of loss (in current US dollars) to the
GDP (in current US dollars) as a measure of direct loss. Using a ratio makes comparison
of loss across countries valid, since PPP or exchange rates that appear both in numerator
and denominator of the ratio cancel out. As mentioned in the introduction, the loss to
GDP ratio does not exhibit any trend over the time period of the sample and hence is a
good indicator of catastrophes. This is important since the present study is based on
events during the last three decades. Comparison is only possible by using the annual
economic loss as a proportion of the total income (GDP).
4.4.1.2 Percentage affected
In a developing economy, where the majority are poor the number of people
affected is often a better indicator of the severity of a catastrophe than direct loss. The
number of people affected depends on the vulnerabilities of various groups that are
resident in the affected area. The vulnerability of groups in turn depends on the manner in
which assets and income are distributed between different social groups. Post event
recovery depends on the way resources are allocated and here too discrimination may
occur based on pre-existing conditions of inequality based on gender, ethnicity, and race.
It is these vulnerable sections of society that suffer most from catastrophes affecting their
lives, their settlements, and their livelihoods.
Chapter Four: Empirical Studies
150
Blaikie et al. (1994) point out that in many parts of the world each household's bundle of
property and assets and economic connections with others may be lost, enhanced,
disrupted, or reinforced in a number of ways due to hazards. The impact of the hazards
operate under the influence of rules and structures derived from existing social and
economic system, but are modified by the distinct characteristics of a particular hazard
and patterns of vulnerability.
4.4.1.3 Type of hazard
Different types of disaster have varying direct and therefore indirect and secondary
impacts. Given a vulnerable habitat, the damage pattern depends on type and intensity of
the physical event. For example, droughts ruin crops and forests but cause relatively little
damage to infrastructure. As a result productivity may remain the same after the event. In
the case of droughts, if the country has surplus of domestic food production, drought can
be managed. For example, one year after the 1982 Australian drought the country's
economy was back to 'normal'. But in countries with little surplus, the effects are more
tangible. Countries whose GDP is mainly represented by the rural economies are
especially vulnerable to droughts. Droughts cause major production losses. If the net farm
income falls during a drought in a farm based economy, it my cause a decline in the
overall output.
In contrast, earthquakes cause relatively little damage to standing crops, other than
localized losses resulting from landslides. But an earthquake can damage buildings and
underground infrastructure. A hurricane may cause extensive crop damage as well as
damage to structures. Reconstruction may result in changes to the productivity due to the
destruction and subsequent construction of new capital. Such changes in productivity
were modeled in Chapter 3. It is important to find out whether the type of disaster affects
the post-event growth rates.
Location and climate have large effects on income levels and income growth, through
their effects on transport costs, disease burdens, and agricultural productivity, among
Chapter Four: Empirical Studies
151
other channels. Major natural hazards that occur frequently in some parts of the globe
have definite effects on income levels and growth. Some countries may therefore be at a
geographical disadvantage due to being situated in hazard prone area.
4.4.2 Control variables
Previous studies do not explicitly spell out the explanatory variables that may be
related to the post-event economic growth rate. Furthermore, there is a lack of theoretical
analytical models describing the phenomena, which has been addressed in Chapter 3.
Theoretical models and simulations presented in Chapter 3 point to the importance for
modeling the post-event productivity changes. Changes in the productivity are reflected
in the post-event evolution of consumption, output, and growth. Based on a wealth of
studies conducted in the field of economic growth (mentioned in Section 4.2.1), variables
that may be important in determining the post-event productivity are discussed in the
following sections. These include indicators for describing pre-event economic
conditions, health, poverty and inequality, government, infrastructure, education, and
trade.
4.4.2.1 Pre-existing Economic Conditions
If a nation has a stable macro-economy with a steady growth, it would be
relatively easier to detect any fluctuations resulting from a catastrophe. Pre-event decade
mean and standard deviation of the annual percentage growth rates are included as
control variables, as indicators of past performance of a nation's macro-economy. Barro
(1995) finds that higher inflation variability goes along with a lower rate of economic
growth. Monetary institutions and policies that lead to substantial variations in the
general level of prices create uncertainty and undermine the efficacy of money. In the
event of a catastrophe, it is more likely in nations with high inflationary susceptibility
that the prices will go out of control. Inflationary pressures will have a negative effect on
the productivity. An indicator for standard deviation of the annual inflation rate during
the last five years is included as a control variable (Gwartney and Lawson, 1997).
Chapter Four: Empirical Studies
152
Another indicator of monetary stability that is included is the average annual growth rate
of the money supply during the last five years minus the potential growth rate of the GDP
(Gwartney and Lawson, 1997).
4.4.2.2 Health
Health problems are particularly highlighted in studies of floods on the West
Coast of South America brought about by El Nino in 1982-83. Blaikie et al. (1994),
quoting from a study of government health centers in north Peru, report that there was an
almost two-fold increase in number of deaths as result of disease and illness due to
epidemics following floods. People's basic health and nutritional status relates strongly to
their ability to survive disruptions of their livelihood systems. This status is important for
their resilience in the face of external shock. For most people living on a subsistence diet
and without proper access to health care, even a mild epidemic after a catastrophe may
prove fatal. The pre-event socioeconomic processes, to a large extent, determine the pre-
event health conditions of the community which in-turn determines the percentage of
people affected by a catastrophe. The post event reconstruction depends on an adequate
supply of labor immediately after the event. If the majority of population is affected by a
catastrophe for health reasons, there may be inadequate supply of labor resulting in
adverse changes in post-event productivity.
Various indicators are used to summarize the 'health' of a community. These include:
i) Life expectancy at age zero,
ii) Number of hospital beds per thousand, indicating the accessibility of health services
after a catastrophe, and
iii) The daily calorie and protein intakes.
4.4.2.3 Poverty and Inequality
The burden of poverty is spread unevenly - among the regions of the developing
world, among countries within those regions, and among localities within those countries
Chapter Four: Empirical Studies
153
(Meier, 1995, Ray, 1998). Alexander (1998) cites the example of Philippines and
compares it with Japan. Both the countries have similar risk profiles as far as occurrence
of types physical hazards are concerned. But Philippines has a GNP that is 2.75% of
Japanese and 49% of Philippines population lives below poverty line. This necessitates
Philippines to bear a heavier burden from losses it experiences from calamitous events.
Within regions and countries, the poor are often concentrated in vulnerable places: in
rural areas with high population densities, such as the Indo-Gangetic plain and the Island
of Java, Indonesia. Often the problems of poverty, population, and the environment are
intertwined: earlier patterns of development and pressure of rapidly expanding
populations mean that many of the poor are forced to live in highly vulnerable regions.
As Blaikie et al. (1994) point out, in Manila (Philippines) the inhabitants of squatter
settlements constitute 35% of the population vulnerable to coastal flooding, and Bogota
(Colombia) has 60% of population living on landslide prone steep slopes. Even in urban
areas, if there are no adequate measures to systematically maintain buildings, potential
losses may be high. For example in the 1985 Mexico earthquake, the decaying inner city
tenements were severely affected.
Rural-urban migration leads to the erosion of local knowledge and institutions required
for coping in the aftermath of a disaster. The loss of younger people, especially working
age males and those with skills which are marketable in the cities may alter the type of
building structures that can be constructed to something less safe than previously.
Obviously this results in greater number of people being affected by the catastrophe.
Certain groups within a community are more vulnerable. Women, children, elderly,
ethnic groups, and minorities suffer disproportionately as a result of catastrophe as has
been reported by Peacock et al. (1997) after Hurricane Andrew. Inequality is a crucial
factor in the ability of an affected community to recover after the occurrence of a
catastrophe. A more unequal society will result in a more unequal distribution of effects -
the poorest in the affected society bearing the brunt of the catastrophe. An inefficient
bureaucracy will allow the inequality to deepen by concentrating the relief in the already
Chapter Four: Empirical Studies
154
affluent people of the community. It has already been demonstrated by various macro-
economists (Barro 1995, Easterly, 1997) that higher the inequality slower is the economic
growth. Thus one of the effects of a catastrophe, given an inefficient bureaucracy, is to
indirectly retard growth by deepening inequalities. On the other hand the government can
view the occurrence of a catastrophe as an opportunity for initiating various programs to
boost economic growth. More efficient and modernized infrastructure may be constructed
replacing damaged structures increasing productivity, which acts as a catalyst for
economic growth of the affected region.
Indicators used to summarize the 'poverty' include the percentage of people living on less
than $1 a day (PPP 1981-95) (World Bank, 1997). The daily calorie intake is also an
indicator of poverty, though controversial. The decade average for Gini coefficient is
used as an indicator for inequality (Easterly and Levine, 1997). The ratio of the share of
the top twenty percent in the income distribution to the first quintile is also used as an
indicator of inequality (Easterly and Levine, 1997). Gender bias is represented using the
ratio of female to male average schooling years.
4.4.2.4 Government, Bureaucracy, and Institutions
Whether a poor country recovers quickly from a catastrophe depends, among
other factors, on its government. If the government has effectively implemented the
policies that make the country's development potential realizable, then a catastrophe will
be absorbed without much negative impact. But in many poor countries, the political
foundations for developmental efforts are not yet firm. Political instability,
undifferentiated and diffuse political structures, and inefficient governments are still too
prevalent (World Bank, 1997).
Commander et al. (1997) look at factors explaining the size of government and the
consequences of government for income growth and other measures of well-being, such
as infant mortality and life expectancy. They present partial evidence for the view that
governments use consumption to buffer external risk, particularly in low-income
Chapter Four: Empirical Studies
155
countries. With respect to the consequences for growth, they find a robust negative
association with government consumption and with an index of policy distortions and a
positive relationship with quality of bureaucracy. They also report that social sector
spending can exert a positive influence on infant mortality and life expectancy.
Primarily its bureaucracy (Knack and Keefer 1995 and Mauro 1993) gives an explicit
evaluation of the quality of government. This evaluation is put together from a set of
responses by foreign investors that focus on the extent of red tape involved in any
transaction, the regulatory environment and the degree of autonomy from political
pressure. These responses provide us with a composite index of the quality of
government bureaucracy or its capability. Mauro (1993) finds a strong relationship
between per capita income and average indices of red tape, inefficient judiciary, and
corruption. Clague, Keefer, Knack, and Olson (1996) likewise establish a relationship
between high per capita income and high quality institutions - freedom from
expropriation, freedom from contract repudiation, freedom from corruption, and rule of
law. After a catastrophe has occurred, it is the efficiency of the government bureaucracy,
which partly determines the efficiency of the processes that determine the post event
productivity. As Oliver-Smith (1994) points out, the assistance after the 1970 Peru
earthquake never reached the survivors because of the 'Byzantine bureaucratic design and
a bewildering division of responsibilities' of the principal agency in charge of relief and
reconstruction. Keefer and Knack (1997) find a strong association between per capita
income and trust between individuals in a society. Trust is important for post event
behavior.
Rodrik (1999) presents econometric evidence from countries that experienced the
sharpest drops in growth after 1975 were those with divided societies and with weak
institutions of conflict management. He contends that 'social conflicts and their
management - whether successful or not - played a key role in transmitting the external
shocks on to economic performance.' The strength of crisis management institutions
determines the recovery process of an affected community. Studies at community level
(e.g. Oliver-Smith 1990, and Bolin 1982) highlight the major impediments to the
Chapter Four: Empirical Studies
156
community recovery process even when they have received aid. Aid is not effective for
the following reasons: i) local disaster management staff are unprepared to deal with aid
recipients, ii) aid does not meet the needs of the poor, iii) outside donor programs exclude
local involvement, and iv) poorly coordinated and conflicting demands from national
government agencies. Many national governments have begun to initiate programs that
assist their local jurisdictions to prepare recovery and development plans (Kreimer and
Munasinghe, 1991).
Political will and respect for human rights are important factors for the successful
implementation of such plans. Without strong political will and freedom of expression in
a country, methodologies devised by a vulnerable community for coping after a
catastrophic event will not receive necessary impetus. As a result development processes
may suffer. On the other hand, if the most vulnerable in the society are empowered
through grass-roots organizations and non-governmental organizations there would be
greater equity in distribution of relief aid thus sustaining the development process. In
Central America, the importance of local organization has been demonstrated in sudden
onset disasters (earthquakes) and in terms of events with a longer preparatory phase
(hurricanes and flooding).
An indicator of government size is the general government consumption as a percentage
of GDP. Human rights index, bureaucratic quality, rule of law, freedom from corruption,
and repression of civil liberties are all indicators for individual rights and democracy
(Easterly and Levine1997).
4.4.2.5 Infrastructure
Typically, a catastrophe results in major disruption of infrastructure facilities.
Given the fact that infrastructure facilities are poorly maintained in developing countries,
the extent of damage is severe even for a moderate hazard. Moreover, the emphasis shifts
to re-building the community in the immediate aftermath of a catastrophe. In its report on
Infrastructure and Development (1994), the World Bank points out that 'when times are
Chapter Four: Empirical Studies
157
hard, capital spending on infrastructure is the first item to go and operations and
maintenance are often close behind. Despite the long-term economic costs of slashing
infrastructure spending, governments find it less politically costly than reducing public
employment or wages.' After a catastrophe, emphasis shifts to providing immediate relief
to the victims and as a result capital expenditures are cut with infrastructure capital
spending often taking the biggest reduction. The 1970 Peru earthquake completely
incapacitated the fragile infrastructure of roads, railways, airports and communications
(Oliver-Smith, 1994). The rails of one railway system that had been twisted beyond
repair have not been replaced 20 years after the event. Destruction of roadways or other
infrastructure may cause impediments to relief and rescue operations and lead to business
interruption. After the 1995 Kobe earthquake, the Kobe port was closed down for a long
time due to extensive damage. As a result ships started using other ports with a result that
there was a permanent damage to regional income. The degree to which key
infrastructure can function after a catastrophe determines the rate at which materials and
men can be moved to the affected region. This in turn affects the post-event productivity
of the region.
The indicators used for infrastructure are:
i) Number of television sets per capita, and
ii) Number of radios per capita
4.4.2.6 Education
Education is vital in creating awareness and facilitates communicating catastrophe
mitigation ideas. A society will be more aware of the risk of occurrence of natural hazard
event if specific measures are taken to incorporate an adequate knowledge of the
vulnerabilities of different zones or regions in the school curricula. An educated
community can devise and implement more effective self-management strategies in case
of a catastrophic event and rapidly recover thus restoring its pre-event productive
capacity or sometimes even exceeding it. But education is intimately related to ongoing
socioeconomic processes. Barro and Martin (1995) present evidence to show that
Chapter Four: Empirical Studies
158
average-years of male secondary and higher schooling and average years of female
secondary and higher schooling tend to be significantly related to subsequent growth.
Thus the level of education of a community is an important control variable used to
compare post-event economic behavior across various countries.
The percentage of "no schooling" in the population is used as an indicator for education.
4.4.2.7 Trade
Trade forms an important component of a country's output in the modern global
economy. Manufactures form a major portion of imports of a developing country while
exports are usually non-manufactures such as cash crops. If a catastrophe severely
destroys the main export merchandise then the country may be forced into a balance of
payments crisis thus dampening its post-event reconstruction efforts. Productivity after
the event may be forced to be below its pre-event levels. Also, prices, for agricultural and
mineral exports on which Third World has traditionally had to depend, are falling. On the
other hand, prices of imported energy and technology have increased. This worsens
balance of payments crises. Hurricane Gilbert deprived Jamaica of more than US $27
million in foreign exports in 1988-89. This may partly explain the decline in its growth
rate (Table 4.1). Foreign debt amounted to 60 per cent of GDP for Latin America during
1985 (Branford and Kucinski 1988).
Indicators which summarize a governments policy stance on trade over time is:
i) The Economic Freedom Index (Gwartney and Lawson, 1997), and
ii) The degree to which a country's exchange rate has been over-valued - as measured by
the black market premium on the exchange rate.
Chapter Four: Empirical Studies
159
4.5 INTRODUCTION TO ECONOMETRIC ISSUES
An overview of the main concepts involved in linear regression models is
presented in this section. An algebraic model represents the real–world system by a
system of equations. These equations may be behavioral, such as the consumption
function: C = C(Y), an equilibrium condition, such as the national income equilibrium
condition: Y = C+Z. The variables in this model are consumption C, national income Y,
and exogenous expenditure Z. All of these fundamental or basic equations of the model
may be termed structural equations. The model determines values of certain variables,
called endogenous variables, the jointly dependent variables of the model, which are
determined simultaneously by relations of the model. In the case of the model specified
in Eq.3.7, post event economic growth rate is the endogenous variable, which is to be
explained or predicted. The model also contains other variables, called exogenous
variables, which are determined outside the system but which influence it by affecting the
values of the endogenous variables. They affect the system but are not in turn affected by
it. Direct loss from a catastrophe is an example of exogenous variable. The model also
contains certain parameters (α and β), which are generally estimated using econometric
techniques and relevant data. Another important characteristic of an econometric model is
the fact that it is stochastic rather than deterministic. A stochastic model includes random
variables, whereas a deterministic model does not. Let the economic growth rate after a
catastrophic event be given by:
yt = α0 + β1yt-1 + β2D + β3E (4.4)
This function specifies that given the pre-event economic conditions yt-1, a vector D
describing the disaster including magnitude of direct loss, the number of people affected,
and a vector E describing the country specific characteristics, post-event economic
growth rate yt is determined exactly as given by Eq.4.4. This is clearly not reasonable.
Many factors other than the direct loss and the population affected determine the post
event growth rate, such as inflation variability, quality of the bureaucracy, health, and
education. Furthermore, the relationship may not be as simple as that given by the linear
relation as explained in Chapter 3 and the loss variables may be measured inaccurately. It
is therefore more reasonable to estimate yt at a given level of loss variables, as on average
Chapter Four: Empirical Studies
160
equal to the right hand side of Eq.4.4. In general, yt will fall within a certain confidence
interval, after controlling for country specific fixed effects, that is,
yt = α0 + β1yt-1 + β2D + β3E ± ∆, (4.5)
where, ∆ indicates the level above or below the average value such that with a high
degree of confidence, yt fall in the defined interval. The value of ∆ can be determined by
assuming that yt is itself a random variable with a particular density function. Because of
the central limit theorem the normal distribution is typically assumed. The term “on
average” generally refers to the mean or expected value, so the right hand side of Eq. 4.5
is the mean of yt. The ∆ can then be chosen, as illustrated, so that 90% of the distribution
is included in the confidence interval, where each of the tails of the distribution contains
5% of the distribution. In general, an econometric model uniquely specifies the
probability distribution of each endogenous variable, given the values taken by all
exogenous variables and given the values of all parameters of the model.
Algebraically, the stochastic nature of the relationship for the post event growth rate is
represented as
yt = α0 + β1yt-1 + β2D + β3E ± ε (4.6)
ε is an additive stochastic disturbance term that plays the role of chance mechanism. In
general, each equation of an econometric model, other than definitions, equilibrium
conditions, and identities, is assumed to contain an additive stochastic disturbance term.
If the stochastic disturbance term has a variance that is always identically zero, the model
reduces to a deterministic one. The other extreme case is where the model is purely
stochastic. The stochastic terms are unobservable random variables with certain assumed
properties (e.g. means, variances, and covariances). The values taken by these variables
of the model are not known with certainty; rather, they can be considered random
drawing from a probability distribution. Possible sources of such stochastic perturbations
could be relevant explanatory variables that have been omitted from the relationship
shown in the model or possibly the effects of measurement errors in the variables - in
particular errors related to reporting of direct losses. Other sources of such stochastic
disturbances could be mis-specified functional forms, such as assuming a linear
relationship when the true relationship is nonlinear, or errors of aggregation, which might
Chapter Four: Empirical Studies
161
be introduced into a macro equation when not all individuals possess the same underlying
micro relationship. It might be noted that sources of these perturbations can be quite
important in practice; thus the treatment of measurement error will be, in general, quite
different, depending on whether the measurement error is found in the dependent variable
or in one or more of the explanatory variables. In any case, the inclusion of such
stochastic disturbance terms in the model is basic to the use of tools of statistical
inference to estimate parameters of the model.
4.6 PROBLEMS WITH THE DATA
Data related to catastrophes are typically non-experimental data. There are several
problems encountered with these data, which are presented in the following.
The first is the degrees-of-freedom problem – that the available data simply do not
include enough observations to allow an adequate estimate of the model. In the use of
non-experimental data it is impossible to replicate the conditions that gave rise to them,
so additional data points cannot be generated. Data regarding catastrophic losses may be
available, but data on explanatory variables may be missing. This was particularly true in
panel data compiled for this study. In some cases the available was inadequate for
estimating a particular model but adequate for estimating an alternative model, which
will be clear when various model specifications are presented. By including more than
155 major events, the panel had adequate degrees of freedom in spite of the missing data.
Second is the multi-collinearity problem - the tendency of the data to bunch or move
together rather than being “spread out”. For example, for a complex model describing
economic growth, the variables exhibit the same trends over a cross-section of countries.
With experimental data it may be possible to vary the conditions of the experiment to
obtain an adequate spread. With non-experimental data such control does not exist, and
the real-world system may involve very small variation in the data, in particular a high
degree of interdependence among certain variables. This problem was circumvented to a
certain extent by including data from countries belonging to wide range economic
development as measured by per capita GDP.
Chapter Four: Empirical Studies
162
Third is the serial-correlation problem – the fact that when using data in two consequent
years (before and after the catastrophic event), underlying changes occur very slowly
over time. Thus conditions in time periods that are close together tend to be similar. To
the extent that the stochastic disturbance term represents conditions relevant to the model
but not accounted for in it explicitly, such as omitted variables, serial correlation itself in
a dependence of the stochastic disturbance term in one period on that in another period.
Various tests, to be described later, were performed to detect the serial-correlation and
suitably interpret the results of the specifications.
Fourth is the errors-in-measurement problem – that data are measured subject to various
inaccuracies and biases. In fact, data are sometimes revised because of later recognition
of these inaccuracies and biases. More fundamentally, potential inaccuracies result from a
lack of precision in conceptualization. For example, the GNP accounts are revised from
time to time on the basis of such changes in conceptualization (e.g. defining what is
included in consumption). Such changes in conceptualization necessitate refining the data
to make them comparable and consistent over time. Also, as has been mentioned
previously, the reporting of data related to catastrophes may be inaccurate and biased due
to several factors related to political, sociological, and anthropological factors.
To address these issues, the Extreme Bounds Analysis is used and this is described in the
following.
4.7 LIMITATIONS OF CROSS-COUNTRY REGRESSION STUDIES
There are substantial conceptual and statistical problems that plague cross-country
investigations (Levine and Renelt, 1992). Levine and Renelt (1992) point out that
statistically entries are sometimes measured inconsistently and inaccurately. Even putting
measurement difficulties aside, it is not clear whether we can include countries as diverse
as Bangladesh and Canada in the same regression. These countries operate in different
policy regimes and under different environments. A country may be at a particular stage
in a business cycle, or may be undergoing major policy changes, or experiencing political
Chapter Four: Empirical Studies
163
disturbances. All these factors affect economic activity and consequently economic
growth. Researchers (Barro, 1991; Easterly, 1997) have found that many individual
indicators of monetary, fiscal, trade, exchange-rate, and financial policies are
significantly correlated with long-run growth in cross country growth regressions. How
could one evaluate the “believability” of cross-country regressions? Extreme bounds
analysis (EBA) based Edward Leamer’s (1983, 1985) work can be used for testing the
results of regressions relating the direct loss to the post event indicators of the economy.
The EBA employs a linear, ordinary-least-squares regression framework. The variables in
the vector E are chosen from a set of indicators, which are known to affect the long-run
economic growth rate. The EBA involves varying the E variables to determine whether
that coefficient on the damage indicator, D, is consistently significant and of the same
sign when the right-hand-side variables change. If β2 is consistently significant and have
the same sign the results are termed as “robust”; otherwise the results are “fragile.” The
EBA is used to test the robustness of the empirical associations between the loss-GDP
ratio and various economic indicators. The results of these regressions are presented next.
4.8 RESULTS FROM REGRESSION ANALYSIS
Details of the regression results for growth changes and the effects of catastrophes
on consumption, savings, government expenditure, inflation and real interest rates are
presented in Appendices F to I. For these regressions various specifications are reported.
4.8.1 Growth rates – Short term
Based on the specifications (21 in all) for examining the effect of a catastrophe on
the short-term growth rate of an economy that the direct loss term enters statistically
significantly. Complete details of the specifications and the regressions are presented in
Appendix F, in electronic form. Table 4.3 presents one such specification. For this
particular specification, the coefficient for the loss term (b1) is –2.37 and highly
significant. Other specifications reveal that the coefficient (b1) ranges from –3.9 to –1.7
with a mean of –2.9. The coefficient remains highly significant (<0.001) in all
specifications. The coefficient for percentage of population affected is also statistically
Chapter Four: Empirical Studies
164
significant but is positive. Dummy for earthquakes indicate that they are associated
positively and significantly with post event growth rate, whereas droughts are negatively
associated with post event growth rate.
Summary|R| 0.543R2 0.295R2 adjusted 0.271Standard Error 3.486# Points 151PRESS 1963.96R2 for Prediction 0.214Durbin-Watson d 1.787First Order Autocorrelation 0.106Collinearity 0.334Coefficient of Variation 105.286
ANOVASource SS SS% MS F F Signif df
Regression 736.81 29 147.36 12.12 8.030e-10 5Residual 1762.30 71 12.154 145Total 2499.1 100 150
P value Std Error -95% 95% t Stat VIFb0 8.73 0.000 1.604 5.56 11.90 5.4b1 -2.37 0.000 0.500 -3.36 -1.38 -4.7 2b2 60.47 0.003 19.796 21.34 99.60 3.1 1b3 -3.10 0.000 0.511 -4.11 -2.09 -6.1 2b4 0.42 0.001 0.128 0.16 0.67 3.2 1b5 -1.32 0.006 0.475 -2.25 -0.38 -2.8 1
Table 4.3 Result of a typical regression used for testing the negative association of economic loss to post-event growth
Gr1yrAfter = b0 + b1*Log10(Loss/GDP) + b2*Log(1+TotoAff/Pop) + b3*Log10AvgGDP_per_capita + b4*COV_AnnualGrowth + b5*Log10Std_Dev_Inflation
Chapter Four: Empirical Studies
165
Some of the specifications control for immediate (one year preceding) pre event
economic conditions with indicators such as the pre event growth rate, the pre event gross
domestic fixed investment growth, and the pre event government size. The coefficients of
pre-event growth rate and the pre event gross domestic fixed investment growth enter
positively and significantly in explaining the post-event growth rate, as expected. One
point of the pre event growth rate explains 0.39 to 0.79 point of the post event growth
rate. One point growth in the pre event gross domestic fixed investment is associated with
0.22 points of post event growth rate. Greater share of the government expenditures in the
GDP appears negatively associated with the post-event growth rate.
Control variables such as measures of civil liberties, bureaucratic quality, black market
exchange rates, percentage of population without schooling are averages over longer
periods of time, typically five to ten years around the event. The inherent assumption is
that these variables change at a much slower rate than other macro-variables like the
annual percentage growth rate. As has been discussed in chapter two, these factors
nevertheless determine the vulnerability of a country to natural hazards, which in turn
determines the post event economic behavior. The regressions present econometric
evidence for associations between indicators of ongoing social, economic, and political
processes and the post event behavior.
Indicators of the monetary health of an economy, namely the inflation variability, the
monetary growth (average annual growth rate of the money supply during the last five
years minus the potential growth rate of real GDP) is negatively associated with growth.
Better bureaucracies are associated with higher post event growth. Indicators for
government enterprises (higher ranks imply lower role and presence of government
owned enterprises) are positively associated with post event growth. Greater civil
liberties and political rights are also positively and significantly associated with post
event growth. Better health (as indicated by the daily protein/calorie intakes) is also
positively associated with growth. Lack of education is negatively and significantly
Chapter Four: Empirical Studies
166
associated with post event growth. The signs and significance of the determinants of
growth appear as expected and in accordance with the discussion in Section 4.5.
4.8.2 Growth rates – Average
Table 4.4 presents the regression results for average growth rates. As has been
already mentioned, average growth rates refer to means of growth rates three years prior
to and after the event. It can be seen from Table 4.4 that the loss term appears negatively
and significantly in all the specifications. The coefficient ranges from –1.95 to –0.67 with
a mean of –1.28 over the specifications. This again implies that loss term is negatively
correlated with post event growth. Simulations based on Ramsey’s model indicate that
(Fig. 2f) loss is positively correlated with the post-event growth. Only if the effects of the
efficiency of the post event reconstruction are taken into account, as described in Section
3.3 using an extended Ramsey’s model, can the negative correlation between loss and the
post-event growth be explained. This brings out the importance of modeling the transient
processes immediately after the event.
Dummy for earthquakes indicate that they are associated positively and significantly with
post event growth rate. Droughts are negatively associated with post event growth rate.
Earthquakes typically result in capital being damaged or destroyed. Droughts do not
cause relatively damage to capital stock. After an earthquake, reconstruction activities
may have a positive effect on the region’s productivity. Such a change in productivity
was assumed in the theoretical models developed in Chapter 3. Numerical simulations
indicated that increases in productivity (Figs. 3.1f and 3.3h) result in increases in post
event growth rates. Empirical evidence that earthquake dummy is positively correlated
with the post-event growth rate lends support to the theoretical result that capital
regeneration after an earthquake increase the post event growth rate. This is further
reinforced by the fact that drought dummy is negatively correlated.
Chapter Four: Empirical Studies
167
Summary|R| 0.633R2 0.401R2 adjusted 0.376Standard Error 2.276# Points 151PRESS 862.82R2 for Prediction 0.307Durbin-Watson d 1.382First Order Autocorrelation 0.308Collinearity 0.311Coefficient of Variation 60.682
ANOVASource SS SS% MS F F Signif df
Regression 499.10 40 83.18 16.05 4.387e-14 6Residual 746.27 60 5.182 144Total 1245.4 100 150
P value Std Error -95% 95% t Stat VIFb0 3.54 0.217 2.854 -2.10 9.18 1.2b1 -1.68 0.000 0.327 -2.33 -1.04 -5.1 2b2 38.22 0.005 13.264 12.01 64.44 2.9 1b3 -2.72 0.000 0.336 -3.39 -2.06 -8.1 2b4 -0.73 0.021 0.311 -1.34 -0.11 -2.3 1b5 4.17 0.024 1.836 0.54 7.80 2.3 1b6 0.31 0.000 0.084 0.14 0.47 3.6 1
AvgAfter = b0 + b1*Log10(Loss/GDP) + b2*Log(1+TotoAff/Pop) + b3*Log10AvgGDP_per_capita + b4*Log10Std_Dev_Inflation +
b5*Log10AvgGrossCapitalFormation_%GDP + b6*COV_AnnualGrowth
Table 4.4 Result of a typical regression used for testing the negative association of economic loss to post-event average growth
Chapter Four: Empirical Studies
168
The coefficient of average pre-event per capita income enters negatively and significantly
in explaining the post-event growth rate, as expected. Greater is the pre-event per capita
income, smaller is the post-event growth.
Indicators of the monetary health of an economy, namely the inflation variability, the
standard deviation of growth over the past ten years is negatively associated with post
event growth. These indicators of the susceptibility of the economy to price volatility in
an economy enter negatively because price increases after the event result in lower
productivity.
Better bureaucracies are associated with higher post event growth. One possible reason is
that better bureaucracies will fuel the post event productivity and hence growth. Lack of
corruption enters positively and significantly in the post event growth rates. Indicators for
government enterprises (higher ranks imply lower role and presence of government
owned enterprises) are positively associated with post event growth. Greater civil
liberties and political rights are also positively and significantly associated with post
event growth. Better health (as indicated by the daily protein/calorie intakes) is also
positively associated with growth. Lack of education is negatively and significantly
associated with post event growth. The signs and significance of the determinants of
growth appear as expected and in accordance with the discussion in Section 4.5.
Chapter Four: Empirical Studies
169
4.9 EFFECT ON MAJOR ECONOMIC INDICATORS
After examining the data on economic growth, the effects of catastrophes on
major economic indicators such as the consumption, investment, government
consumption, inflation, and the real interest rates are examined. Details of these
regressions are presented in electronic form in Appendix G.
4.9.1 Consumption
It is clear from Table 4.5 that direct loss is positively and significantly associated
with consumption. The coefficient of the loss term has a minimum value of 2.0 and a
maximum of 3.2 with a mean of 2.6 over the specifications. This implies that a direct loss
of 10% of GDP is associated with 3.21 point increase in consumption. This response to
changes in income from catastrophes is further explained in the subsequent sections.
4.9.2 Investment
An examination of the specification (Table 4.6) reveals that catastrophes result in
lowering the amount of investments by concentrating on increases in consumption
expenditures.
4.9.3 Government expenditure
The loss term enters positively and significantly in all the specifications with a
mean of 1.14. A typical relation is as follows:
GovernmentConsumptionafter = -4.33
(0.002)
+ 0.86*GovernmentConsumptionbefore+1.4*Log(Loss/GDP)
(0.000) (0.011)
+ 2.395*Eq. + 0.559*GovtEnterp
(0.004) (0.006)
N=100; R2= 0.818; F=38; DW = 1.925
This implies that a direct loss of 10% of GDP is associated with 1.4 point increase in
government expenditures.
Chapter Four: Empirical Studies
170
Table 4.6 Effect of catastrophes on investmentsSummary
|R| 0.915R2 0.837R2 adjusted 0.835Standard Error 2.477# Points 149PRESS 951.04R2 for Prediction 0.827Durbin-Watson d 1.508First Order Autocorrelation 0.243Collinearity 0.814Coefficient of Variation 12.814
ANOVASource SS SS% MS F F Signif df
Regression 4592.7 84 2296.4 374.14 3.469e-58 2Residual 896.11 16 6.138 146Total 5488.8 100 148
AvgAfter = b0 + b1*Log10(Loss/GDP) + b2*AvgBeforeP value Std Error -95% 95% t Stat VIF
b0 1.998 0.0056 0.71 0.60 3.40 2.81b1 -0.896 0.0020 0.28 -1.46 -0.33 -3.15 1.23b2 0.791 0.0000 0.03 0.72 0.86 23.16 1.23
Table 4.5 Effect of catastrophes on change in consumption Summary
|R| 0.24R2 0.06R2 adjusted 0.05Standard Error 0.04# Points 150.00PRESS 0.20R2 for Prediction 0.03Durbin-Watson d 1.87First Order Autocorrelation 0.05Collinearity 1.00Coefficient of Variation 3.61
ANOVASource SS SS% MS F F Signif df
Regression 0.01234 6 0.01234 9.396 0.00259 1Residual 0.194 94 0.00131 148Total 0.207 100 149
ChangeInRealConsumption = b0 + b1*Log10(Loss/GDP)P value Std Error -95% 95% t Stat
b0 1.029 0.000 0.009 1.012 1.046 121b1 0.011 0.003 0.004 0.004 0.019 3.065
Chapter Four: Empirical Studies
171
4.9.4 Inflation, and Interest rates
The regression associating the loss term and the inflation is as follows:
Log(Inflation)after = 0.18 + 0.89 *Log(Inflation)before+ 0.07*Log(Loss/GDP) (0.121) (0.000) (0.057)
N=114; R2= 0.686; F=204; DW = 1.68 This implies that a direct loss of 10% of GDP is associated with 0.07 point increase in log(inflation)
Fig. 4.11 clearly illustrates the fact that the post-event real interest rates are higher
than the pre-event counterparts. Details on the regressions for examining the effect
on interest rates is presented in electronic form in Appendix H. The impact of
catastrophes on real interest rates is presented in Table 4.7. It is clear from the
table that catastrophes, as measured by the number of people affected, are
positively associated with post event real interest rates. Pre-event levels of
inflation and per capita income are positively associated with the post event real
interest rates. Higher the pre event gross investment in fixed capital, lower is the
post event interest rate. And if household spend more money, smaller will be the
real interest rate. An important conclusion of this section is that catastrophes and
financial markets may not be totally uncorrelated after all. More research is
required to unravel the connections between the catastrophes and financial
markets.
Summarizing the results of this section, greater loss-to-GDP ratios are positively
and significantly associated with increases in consumption, government
expenditure, and inflation, and real interest rates and are negatively and
significantly associated with investments. The effect of catastrophes on a host of
other economic indicators is presented in electronic form in Appendix I. These are
not discussed explicitly herein, but they indicate negative effects of a catastrophe.
Chapter Four: Empirical Studies
172
Fig. 4.11 There is perceptible increase in real interest rates after a catastrophic loss
0%10%20%30%40%50%60%70%80%90%
100%
-5.0 0.0 5.0 10.0 15.0 20.0
Real Interest rates
AvgBeforeAvgAfter
Table 4.7 Effect of catastrophes on the real interest ratesSummary
|R| 0.581R2 0.337R2 adjusted 0.306Standard Error 3.424# Points 112PRESS 1626.91R2 for Prediction 0.132Durbin-Watson d 1.311First Order Autocorrelation 0.343Collinearity 0.585Coefficient of Variation 62.590
ANOVASource SS SS% MS F F Signif df
Regression 632.53 34 126.51 10.79 2.097e-08 5Residual 1242.5 66 11.72 106Total 1875.0 100 111
RealIntRateEventYear = b0 + b1*Log10(1+TotAff/Pop) + b2*LogInflationCPI + b3*Log10AvgGDP_per_capita + b4*AvgGrossCapitalFormation_%GDP +
b5*Household_final_consumption_expenditure_(annual_%_growth)P value Std Error -95% 95% t Stat VIF
b0 -0.706 0.925 7.437 -15.45 14.04 -0.09b1 9.130 0.003 3.051 3.082 15.18 2.99 1.2b2 3.526 0.000 0.756 2.028 5.025 4.67 1.2b3 1.245 0.031 0.568 0.119 2.372 2.19 1.5b4 -7.396 0.042 3.587 -14.51 -0.285 -2.06 1.2b5 -0.316 0.004 0.106 -0.527 -0.105 -2.97 1.2
Chapter Four: Empirical Studies
173
4.10 Catastrophes and consumption smoothing
The main purpose of this section is to test the validity or otherwise of the
permanent income hypothesis (PIH) during the years surrounding the occurrence of
catastrophe. Do catastrophes cause predictable shifts in consumption? If the nations are
not able to smooth consumption in these adverse circumstances, then policies need to be
devised which will help mitigate the effects of a catastrophe. If there are predictable
shifts in consumption after a catastrophe, then this information could be used to design
policies to smooth consumption after a catastrophe.
Whenever a catastrophe occurs it will cause rational agents to change the way in which
past incomes affect forecasts of future incomes. Consumption depends on expected future
incomes. Flavin’s (1981) result shows that changes in consumption are predictable by
lagged changes in income (the excess sensitivity hypothesis). The first part of the test will
establish whether the catastrophe results in predictable changes in income. Knowing
about the process generating income (in this case the date of occurrence of a catastrophe)
we could generate forecast for consumption change based on lagged values of income
and consumption. There is excess sensitivity if consumption responds to any previously
predictable component of income change (Deaton, 1992: 164). For the purposes of this section cross-sectional data was used. These include differences
in consumption and income three years preceding and following the event. The data was
not pooled since the main intention is to find out whether lagged changes in income can
predict consumption changes with and without the occurrence of a catastrophic event.
Table 4.8 shows the means and standard deviations of income, saving, and consumption
for five years with the catastrophic event year as the third year. The one noticeable
characteristic is the enormous variation in data. This is to be expected since we have
pooled data over a wide spectrum of countries. Another noticeable feature is that the
standard deviation of income and consumption become largest at t+1 and then drop to
their smallest value at t+2, t being the time of occurrence of the catastrophe. For saving
too the standard deviation becomes large at t+1 and drops at t+2. One plausible inference
is that the occurrence of a catastrophe, which results in direct losses comparable to the
Chapter Four: Empirical Studies
174
GDP (typically greater than 1% of the GDP), may induce greater fluctuations in income,
savings, and consumption. Another plausible inference is that effects of catastrophe
attenuate two years after the event as evidenced by the relatively sharp drop in the
standard deviations for income, consumption and savings.
The data in Table 4.9 illustrate the fact that mean growth rates for both income and
consumption fall during the disaster year (t0 - t-1). The variability of the growth rates
increases for the income whereas for consumption the variability remains almost
constant.
Table 4.8 Summary statistics for income, consumption, and savings in the five years enveloping the disaster year Income Consumption Savings Mean s.d. Mean s..d. mean s.d. t-2 5168.5 7711.9 4039.0 5933.1 1187.7 2072.2 t-1 5198.8 7709.0 4100.0 5993.2 1157.0 1990.3 t 5267.8 7823.1 4170.5 6100.3 1153.8 1979.6 t+1 5399.8 8061.3 4228.8 6176.8 1185.2 2070.0 t+2 4707.9 6224.0 3751.0 5108.4 973.8 1238.2
Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of occurrence of the catastrophe and the previous year. Table 4.9 Summary statistics for percentage growth rates for income, consumption, and savings in the five years enveloping the disaster year
Income Consumption
Mean s.d. Mean s.d. (t-1- t-2) 0.64 1.60 0.69 2.24 (t0 - t-1) 0.38 2.34 0.45 2.12 (t+1- t0) 0.87 1.81 0.87 2.11 (t+2- t+1) 0.72 1.74 0.74 2.13
According to the permanent income hypothesis (PIH), changes in aggregate consumption
cannot be predicted by lags in income. Hall (1978) first proposed tests for the PIH by
adopting the technique of regressing changes in consumption using lagged income,
conditional on lagged consumption. For the PIH to hold variables lagged t-1 or earlier,
Chapter Four: Empirical Studies
175
and in particular lags of income should not help predict consumption in period t. The
expression for changes in consumption (Deaton, 1992:83) is:
∆ct = r/(1+r) Σ∞k=0 (1+r)
-k (Et+1 − Et)yt+k
This implies that a change in consumption ought to be the amount warranted by
innovation in expectations about future labor income. Knowledge of the process
generating income will enable us to check this prediction too (Deaton, 1992:84).
Presumably innovations in income occur after a catastrophic event. Do these innovations
predict the change in consumption? What differences do we observe if we compare
consumption change after a catastrophe with consumption change without any
catastrophic event?
In an important paper, Flavin (1981) tested the null hypothesis the truth of the PIH as
expressed in Eq. 4.7, together with an auto regressive specification for the process
governing labor income. Flavin’s ‘excess sensitivity’ hypothesis allows consumption to
respond to current and lagged changes in income by more or less than is required by the
PIH. The measurement of excess sensitivity is the measurement of the extent to which
consumption responds to previously predictable changes in income.
By using Deaton’s (1992:94) specification of regressing the change on consumption on
the lagged change in income we could avoid the unit root problems that may affect the
income generating process:
∆ct = α + β∆yt-1
However time-averaging problems may induce spurious correlation for adjacent
observations of a series that has been first differenced. This implies that Eq. 4.8 may
yield inconsistent estimates because ∆yt-1 is spuriously correlated with ∆ct. To avoid such
problems Deaton (1992) suggests that variables lagged two periods may be used as
instruments. Instrumentation by variables lagged by variables enables us to account for
transitory consumption that may result from a catastrophe.
Chapter Four: Empirical Studies
176
The results are shown in Table 4.10. The first row presents the results of regressing on
changes in consumption on lagged changes in income for the period t-1 (the period before
the event occurs). The second row presents a similar regression using the twice-lagged
changes in income and consumption as instruments. The third and fourth rows repeat the
same for the period when the catastrophe strikes. The rest of the table presents more
regressions for two following periods. From Table 4.10 it is clear that lagged changes in
income does not enter significantly in explaining the changes in consumption for the
periods: t-1, t+1, and t+2. This implies that there is evidence that the PIH is valid for the
above periods. For years immediately after the event (i.e. results for ∆ct), the results
indicate the change in consumption is the amount warranted by innovations in income.
Instrumentation of lagged changes in income with twice lagged changes in income and
consumption still leaves a significant (albeit reduced) coefficient for ∆yt-1. There is
therefore evidence of excess sensitivity once possible timing and transitory consumption
problems have been taken into account. These results can be taken to mean that
innovations in income due to the occurrence of the catastrophe result in predictable
changes in consumption.
Table 4.10 Estimates of consumption changes using lagged changes in income
Constant t Lagged ∆y t R2 F N ∆ct-1 (OLS) 57.66 (2.620) 0.29 (3.16) 0.16 9.95 48 ∆ct-1 (IV) 57.41 (2.458) 0.30 (1.49) - - 48 ∆ct (OLS) 54.00 (2.527) 0.51 (6.40) 0.46 40.89 48 ∆ct (IV) -8.43 (-0.191) 1.23 (3.75) - - 48 ∆ct+1 (OLS) 24.90 (0.821) 0.15 (1.64) 0.03 2.69 48 ∆ct+1 (IV) 50.63 (1.323) 0.18 (1.11) - - 48 ∆ct+2 (OLS) 70.52 (2.829) 0.19 (1.75) 0.04 3.06 48 ∆ct+2 (IV) 64.71 (2.001) 0.26 (0.90) - - 48
Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World
Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of
occurrence of the catastrophe and three years prior and following the event. Place and occurrence of
catastrophe are from Center for Research on Epidemiology of Disasters (Sapir and Misson, 1992 CRED).
The instruments in the IV are ∆yt and ∆ct lagged twice. t-values are shown in brackets.
In this section PIH and excess sensitivity were examined. There is evidence of excess
sensitivity once possible timing and transitory consumption problems have been taken
Chapter Four: Empirical Studies
177
into account. These results can be taken to mean that innovations in income due to the
occurrence of the catastrophe result in predictable changes in consumption.
4.11 Consumption smoothing and savings behavior
A catastrophic loss in a country's income will lead to changes in consumption
only if the savings are not able to offset the income fluctuations. Incomes of LDCs are
both low and uncertain. Losses of incomes of LDCs due to catastrophes may seriously
undermine the ability to smooth consumption. When insurance markets are incomplete
(this is true for most LDCs), saving and credit transactions assume a special role by
allowing households to smooth their consumption streams in the face of random income
fluctuations.
If income can be treated as stationary then PIH implies that savings have a mean of zero.
Assets are built up in advance of expected declines in income, and are run down when
current income is lower than its expected future level. Under the PIH saving acts as a
sufficient statistic for the agent’s future income expectations. Saving behavior contains
information about what nations expect to happen to their incomes. Forecasts of income
conditional on saving help us deal with the fact that representative agents may possess
private more information about future income than does an observer. This helps us to
infer whether the catastrophic events considered are truly unanticipated. If the events
were anticipated, the changes in consumption could well be explained by expected values
of income, which in-turn would have been predicted by the lagged savings. This has
important consequences for policies designed for preparedness against catastrophes
triggered by natural events. If surprises in income due to occurrence of a catastrophe are
unanticipated then consumption will not be smooth even if we use the agent’s private
information. Nations that expect catastrophes (of the type and magnitude considered in
this study) to occur should devise policies for precautionary savings to smooth
consumption.
As pointed out by Campbell (1987), past savings is a predictor of how income will
change the next period. It is possible that countries anticipate the occurrence of a
Chapter Four: Empirical Studies
178
catastrophe triggered by natural hazards. Though natural hazards occur with certain
regularity, their magnitude and point of occurrence remains uncertain. But a good
preparedness program in place would help nations to smooth their income. By regressing
the change income with lagged savings and comparing the no-disaster year with the
disaster year we could infer about the efficiency of the precautionary savings of the
countries to income shortfalls from a catastrophe triggered by a natural hazard. From
Table 4.11 it is clear that before the catastrophe occurs, lagged savings do not explain the
income change. This situation, however, changes one year after the event. The coefficient
for lagged savings becomes positive and significant in explaining income changes. The
value of the coefficient again drops two years after the catastrophe. This means that the
catastrophe changes the ex ante saving behavior at least for two years after the event.
Consumption change regressions (Table 4.12) show that it is positively related to lagged
values of savings. Before the catastrophic event the coefficient on savings has a lower
significance than two years after the event. One plausible inference is that the changes in
consumption due a catastrophe are only weakly anticipated for the collection of events
that have been considered. Since the events cover a large range of loss/GDP ratios (Table
4.13) it is plausible that the data set dampens out effects of unanticipated losses for
LDCs. Three years after the event the significance and magnitude fall to their pre-
disaster levels and lagged savings are not able to explain consumption changes in
accordance with the PIH.
If we use income lagged twice as an instrument in the regression on consumption change
on lagged saving we essentially get the same results.
Table 4.11 Estimates for income changes using lagged savings
Constant t Lagged Savings
t R2 F N
∆yt-1 37.32 (0.94) 0.0061 (0.36) 0.003 0.13 46 ∆yt 26.21 (0.67) 0.0520 (3.07) 0.17 9.45 46 ∆yt+1 -10.69 (-0.38) 0.1249 (10.22) 0.70 104.4 46 ∆yt+2 -22.71 (-0.64) 0.1031 (4.52) 0.31 20.43 45
Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World
Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of
Chapter Four: Empirical Studies
179
occurrence of the catastrophe and two years prior and following the event. Place and occurrence of
catastrophe are from Center for Research on Epidemiology of Disasters (Sapir and Misson, 1992 CRED). t-
values are shown in brackets.
Table 4.12 Estimates for consumption changes using lagged savings Constant t Lagged
Savings t R2 F N
∆ct-1(OLS) 20.74 (0.66) 0.034 (2.57) 0.11 6.62 48
∆ct-1 (IV) 26.34 (0.82) 0.029 (2.11) - - 48
∆ct(OLS) 9.42 (0.45) 0.053 (5.83) 0.42 33.95 48
∆ct(IV) 6.67 (0.31) 0.055 (5.78) - - 48
∆ct+1(OLS) 20.02 (0.88) 0.067 (6.76) 0.49 45.71 48
∆ct+1(IV) 16.83 (0.73) 0.069 (6.71) - - 48
∆ct+2(OLS) 21.21 (0.66) 0.020 (1.51) 0.03 2.27 48
∆ct+2(IV) 11.30 (0.35) 0.028 (2.01) - - 48 Note – Data on consumption and income (GDP) in 1995 US dollars are from the World Bank – World
Development Indicators CD-ROM (1999). Consumption and income (GDP) data was chosen for the year of
occurrence of the catastrophe and three years prior and following the event. Place and occurrence of
catastrophe are from Center for Research on Epidemiology of Disasters (Sapir and Misson, 1992 CRED).
The instrument in the IV is income lagged twice. t-values are shown in brackets.
In this section the efficiency of the precautionary savings of the countries to income
shortfalls from a catastrophe triggered by a natural hazard is examined. Results from
regressions of income change on lagged savings and comparison of the no-disaster year
with the disaster year are used for arriving at conclusions. Before the catastrophe occurs,
lagged savings do not explain the income change. But one year following the event,
lagged savings anticipate income changes. Evidence is presented to show that
catastrophes change ex ante saving behavior at least for two years after the event.
4.12 Conclusions, Extensions, and Limitations
The problem of finding empirical regularities in the ongoing socioeconomic
processes after the occurrence of a catastrophe was addressed in this chapter.
Connections between these statistical regularities and the results of the theoretical model
simulations presented in Chapter 3 were made. The results of the regression analysis
indicate that by studying disasters much can be learned about the way large-scale socio-
Chapter Four: Empirical Studies
180
economic systems affect and are affected by the occurrence of catastrophes. By making a
cross-country study with countries from all income groups affected by different types of
natural hazards, the results are expected to be sufficiently general. Previous empirical
results from the literature on the determinants of economic growth and on economic
development helped in identifying the explanatory control variables.
The main results of this study can be summarized as follows:
Summarizing the regressions on growth the following statistical regularities are
discerned:
• The models indicate very significant negative coefficient for the direct loss variable in
regressions for short-term growth. The coefficient for the loss variable in the long-term
growth has a lower significance, but remains negative. The magnitude of the coefficient
in the average growth rate regression is less than the short-term regression. This implies
that the associations between the loss term and the economic growth rate become harder
to detect with the passage of time. These results corroborate the results obtained by
simulating the model presented in Section 3.3.
• The pre-event economic growth rate is positive and very significantly associated with the
post-event growth rate, in both the short-term and average regressions. This implies that,
other variables being constant, an economy with a sufficient growth rate can absorb the
effect of a catastrophe. Growth itself is an indicator of the robustness of ongoing
developmental processes. This brings out the importance of having a robust
developmental process in place in absorbing the effect of a catastrophe. The coefficient
for pre-event general government consumption is significant and negative. This agrees
with the known fact that heavy consumption by the government sector retards growth.
• The coefficient for the percentage of people affected is positive and significant in short-
term growth regressions. Though this seems odd, it is should be noted that a catastrophe
affects many people only in developing countries. The amount of aid is to a certain extent
decided by the figures regarding people affected. It is probably this external aid
associated with the percentage affected that spurs growth. As the models described in
Section 3.3 and 3.4, greater inflow of aid results in greater growth.
Chapter Four: Empirical Studies
181
• The coefficient for daily protein/calorie intake appears positive in the short-term growth
regressions associating a healthier community with a more robust developmental process
• If the institutions of crisis management can be proxied by a combination of the size of the
government and the efficiency of the bureaucracy, then their coefficients are positively
and significantly associated with short- and long term (average) post event growth. This
brings out the importance governmental bureaucracy in mitigating the effects of a
catastrophe.
• The coefficient for inflation variability, which is a measure of the monetary robustness of
an economy, is associated negatively and significantly with the post event short- and
long-term growth. This once again ascertains the importance of the ongoing economic
processes in explaining the post-event economic behavior.
• Other factors including civil liberties, percentage of no schooling, economic freedom
index, freedom from corruption, and land-area had the expected signs.
The main results of examining the effects of catastrophes on consumption, investment,
government expenditure, net exports, inflation, interest rates gave the following results:
Large economic losses as a proportion of the GDP are associated with:
1. greater post-event consumption,
2. greater post-event government expenditure,
3. smaller post-event investments,
4. higher inflation, and
5. an increase in real interest rates.
Innovations in income due to the occurrence of the catastrophe result in predictable
changes in consumption. The efficiency of the precautionary savings of the countries to
income shortfalls from a catastrophe triggered by a natural hazard is examined. Results
from regressions of income change on lagged savings and comparison of the no-disaster
year with the disaster year are used for arriving at conclusions. Before the catastrophe
occurs, lagged savings do not explain the income change. But one year following the
event, lagged savings anticipate income changes. Evidence is presented to show that
catastrophes change ex ante saving behavior at least for two years after the event.
Chapter Four: Empirical Studies
182
There are limitations of the study, which are discussed in the following. The first is
regarding the heterogeneity and panel data that arise naturally in cross-country studies.
Omitted heterogeneity induces correlations between explanatory variables and the error
term in a way that has the same consequences as simultaneity bias. The factors that
appear on the right hand side of the specification (Eq.4.6) such as pre event growth may
have no general claim to exogeneity. The combination of genuine simultaneity and
heterogeneity has the further effect of ruling out the use of lags to remove the former.
These considerations would typically require further examination of the effect of
catastrophe on the economic indicators using alternative specifications based on first
differences. Another important limitation is the lack of appropriate instruments, which
are correlated with direct loss term but un-correlated with error term. These instruments
can be used to check whether the coefficients on the loss terms remain robust when they
are instrumented. If data on sectoral distribution of losses is available, this can be used to
instrument the direct loss variable. In other words, this requires details regarding losses in
the agriculture, industry, and service sectors. But such data is hard to obtain. It would be
ideal to develop a system of structural equations to explain the connections between all
the macro-economic variables affected by catastrophes. Lack of underlying theoretical
models forces us to use reduced form equations. These result in inference of statistical
regularities as opposed to full-fledged causal models. Increase of representation in the
sample of higher loss-GDP ratio events is required for the sake of generality.
Chapter Five: Regional Impact of Catastrophes
183
Chapter Five
Regional Impact of Catastrophes 5. Introduction
The main purpose of this chapter is to study the regional impact of three
catastrophes – 1989 Loma Prieta earthquake, 1992 Hurricane Andrew, and 1994
Northridge earthquake. Results from simulating a standard regional economic model are
compared to the results obtained from the theoretical model presented in Section 3.5. The
regional economic model is used for studying the effect of probable earthquake scenarios
in the Bay Area.
An important question, in the field of disaster research, is what are the effects of
catastrophic events such as earthquakes and hurricanes on a regional economy. A
catastrophic event inflicts heavy damage to the capital assets of the affected community.
This damage has consequences that can be measured at three levels: direct damages,
indirect damages, and secondary effects (ECLAC, 1991). Direct losses are the damages
to fixed assets (including property), capital and inventories of finished and semi-finished
goods, raw materials and spare parts that occur as a direct consequence of the natural
phenomenon triggering a catastrophe. Indirect damages relate to the effect on flows of
goods that will whose supply and demand will be affected. They are measured in
monetary, rather than physical terms. Secondary effects refer to the impact on overall
economic performance as measured through important macro-economic variables. As
such, they cannot be added mathematically to the sum total of direct and indirect
damages. Relevant variables may include gross state product (GSP), net earnings of a
county, the employment and unemployment levels, inflation, and the state of public
finances.
Studies relating to the direct and indirect effects of catastrophes are extensive (FEMA,
1994), though much needs to be done. Methodologies have been developed to evaluate
the direct and indirect effects. In particular, HAZUS (1999) is a software that brings
Chapter Five: Regional Impact of Catastrophes
184
together state-of-the-art techniques from engineering and economics to estimate losses
from earthquakes in US. It is being extended to include other hazards as well.
Several methodologies have been used to study the indirect effects of catastrophes.
Roberts, Milliman and Ellson (1982) use a macro-econometric model for predicting the
effects of an earthquake. In a macro-econometric model of a region, a baseline forecast of
economic activity is generated. Shocking the exogenous variables of the model to yield a
post-disaster forecast then simulates the earthquake and the impacts of the disaster are
found by taking the difference between the baseline and post-disaster forecasts. However,
the post-disaster impacts are based on coefficients, which were derived under factor
supply conditions that are no longer relevant after a catastrophic event. Roberts, Milliman
and Ellson (1982) circumvented this problem by joining engineering process models with
an econometric modeling approach to produce an impact analysis that took into account
the changes in the stocks of capital caused by the earthquake. This technique is extremely
data-intensive.
To estimate direct and indirect losses from lifeline damage, Boisvert (1992) adopted a
national input-output model to develop improved estimates of the losses. Input-output
models are a description of the inter-industry flows in an economy. The critical
assumption is that the money value of goods and services delivered by an industry to
other producing sectors is a linear and homogenous function of the output level of the
purchasing sectors. It is also assumed that the system is in equilibrium at given prices,
constant returns to scale, and there is no substitution between inputs. Boisvert (1992)
argues that for every billion dollars of direct damages to lifelines direct business losses
were about $1.8 billion. Indirect business losses increase by three-quarters of a billion for
every billion dollars’ increase in direct business losses. In most cases, the combined
economic losses are less than one percent of the regional economies. Boisvert (1992)
concludes that from a national perspective, it is unlikely that such losses will seriously
disrupt markets.
Chapter Five: Regional Impact of Catastrophes
185
Brookshire and McKee (1992) introduce Computable General Equilibrium models (CGE)
for indirect loss measurement in a region. CGE models extend the framework of input-
output models to include multiple households and to include substitution (not fixed) input
possibilities in the production side and on the consumption side. The CGE models
determine welfare changes that occur as a result of exogenous changes and shocks to the
economic system being examined.
The following is the outline of this chapter. The next section gives an overview of the
methodologies for studying the regional economic impacts. Section 5.2 mentions some
difficulties in assessing the regional impact of catastrophes. Section 5.3 gives a
description of the events considered – 1989 Loma Prieta earthquake, 1992 Hurricane
Andrew, and 1994 Northridge earthquake. A comparative study of the impacts of these
events is presented in Section 5.4. A standard regional economic model is used to
simulate the economic effects of the three events and comparison of the results with
observed values personal income are made in Section 5.5. Using this validated model, the
behavior of the San Francisco Bay Area and Silicon Valley economies to scenario
earthquakes is studied in Section 5.6 and 5.7. Section 5.8 concludes this chapter.
5.1 Methodologies used to study regional impacts Cochrane (1992) using the income identity explains macro-economic effects of a
catastrophe:
Y = C + I + G + (EX - IM) (5.1)
Y is the gross regional product, C is the spending on consumption, I is the spending on
investment, G is the government spending, EX is the exports, and IM the imports in a
region. The occurrence of a catastrophic earthquake would cause a loss of income for
individuals who are laid-off or lose their jobs because of business interruptions and
failures. The physical destruction of private and public facilities however would create
new jobs in the construction industry as destroyed facilities are repaired and rebuilt.
Losses in income for some individuals would therefore be offset by gains in income for
others. Cochrane (1992) presents qualitative reasons for the effect of an earthquake on
each of the variables. His conclusions are:
Chapter Five: Regional Impact of Catastrophes
186
• The net effect on the aggregate income of a region is ambiguous.
• Investments in other areas may be more expensive and consequently may decrease.
• After reinvestment takes place, local incomes may be higher than before the event.
• The way in which individual losses and gains come together to determine aggregate
losses is a complex issue and unique to each set of local economic and physical
conditions, not to mention the severity of the event that occurs.
The estimation of macro-economic effects can also be based on a comparison between
the economic performance anticipated by public and private sector organizations and
academic or consultant analysts (if they are available), and a modified projected
performance estimated after the direct and indirect damages are assessed and valued.
(ECLAC, 1991). This involves a with-without rather than before-after analysis of
economic performance.
In assessing the overall macro-economic and social impact of a catastrophe consideration
must be given to:
i) the time frame in which the disaster occurs, such as its timing relative to agricultural
cycle and, more broadly, to the economic cycle or any short or medium trends
ii) dis-aggregation of the sectoral impacts, and
iii) structure of the economy
Analytical techniques used to evaluate disaster losses and reconstruction gains have
become increasingly sophisticated. As West and Lenze (1994) point out, the techniques
have advanced from descriptive case studies (Haas, Kates, and Bowden 1977) to formal
implementation of regression and time series techniques (Friesema et al. 1979; Chang
1983) to analysis based on regional econometric models (Ellson, Milliman, Poberts 1984;
Guimaraes, Hefner, and Woodward 1993), input-output models (Cochrane 1992a;
Boisvert 1992; Gordon and Richardson 1992), and computable general equilibrium
models (Brookshire and McKee 1992).
Chapter Five: Regional Impact of Catastrophes
187
5.2 Modeling Problems
Modeling the impact of a catastrophe on a regional economy presents many
problems. Estimates of the direct losses can only be given in probabilistic terms. This is
in marked contrast to most impact studies that start with some firm numerical input on
expenditure, employment, income, or tax rate changes. In addition, a catastrophe may
affect many sectors of the economy simultaneously unlike an event like the inauguration
of a theme park, which may affect only a specific sector(s). Extensive damages to
physical facilities including infrastructure and buildings may cause direct supply
interruptions to the affected region. Reconstruction immediately after the event increases
the demand, which may strain regional capacity. Another problem is modeling household
reactions to unanticipated destruction of homes, personal property, and neighborhoods. It
is difficult to identify conditions under which households will resort to energetic
rebuilding as opposed to migration from the impacted area. Changes of consumption
patterns after a catastrophe destroys most of the wealth are not well documented.
A regional economic model originally developed by Treyz (1993) will be used in
studying the regional impacts of catastrophes such as earthquakes and hurricanes. A
separate Appendix J details the model and the associated MATLAB program. The model
is validated by studying the effects of three recent catastrophes – 1989 Loma Prieta
earthquake, 1992 Hurricane Andrew, and 1994 Northridge earthquake. Studying these
historical events helps us to understand the complexity of behavior that a catastrophe can
induce. The model, thus calibrated, is then used for studying the effect of probable
earthquake scenarios in the Bay Area.
The model has five basic building blocks, namely, output, labor and capital, population
and labor supply, wages, prices, and profits, and market shares.
Comprehensive modeling of regional impact of a catastrophe requires data at
considerable levels of detail. There is a lack of data on direct exogenous and endogenous
variable impacts as well as changes in ‘normal’ linkages between the output, labor and
Chapter Five: Regional Impact of Catastrophes
188
capital, population and labor supply, wages, prices, and profits, and market shares
modules. Ex post analysis of major events are to a certain extent amenable to analysis,
since we have observed values of personal income to calibrate the model by changing the
parameters. Ex ante analysis of probable events requires careful study by changing the
parameters to establish reasonable bounds on the behavior of a regional economy.
5.4 Description of Events
5.4.1 Loma Prieta Earthquake
On October 17, 1989 at 5:04 p.m., an earthquake of a 7.1 magnitude struck the
San Francisco Bay Area and its environs. The earthquake caused over $6 billion in direct
property damage and disrupted transportation, communications and utilities. Brady and
Perkins (1991) observed that workers were affected by layoffs for a maximum period of
four months. The total number of workers affected by layoffs was 7100 (out of a total of
3 million jobs = 0.237 percent). This resulted in direct potential loss of wages and salaries
of about $54 million, resulting in a minimum potential loss in gross output (including
wages and salaries) of about $110 million during this period. The total economic
disruption resulted in an estimated maximum potential Gross Regional Product (GRP)
loss ranging from $735 million in one month to $2.9 billion over a maximum of two
months following the event. However, at least 80 percent of that loss was recovered
during the 1st and 2nd quarters of 1990. This implies maximum GRP lost ranges from
$181 million to $725 million only. The 1989 GRP for the Bay area was $174 billion. The
losses, when compared to the total size of the regional economy, can be viewed as
isolated.
San Francisco (SF) experienced the greatest loss in retail activity for the 4th quarter. The
damage and disruption to the Bay Bridge connecting SF to East Bay is a good indicator
of how a major transportation network disruption could affect economic activity. Data
analysis indicates the loss of approximately $73 million in taxable sales due to the closure
of the Bay Bridge for several weeks. The damage of the Cypress Freeway was minimal
because of alternative routes. However, the Bay Area Economic Forum and Metropolitan
Chapter Five: Regional Impact of Catastrophes
189
transportation Commission, Oakland, CA, have documented economic impacts of
approximately $20 million annually. This implies that major failure of infrastructure in a
future quake will result in severe regional impacts. Santa Cruz County experienced 85
percent increase in unemployment insurance claims.
Table 5.1 Loma Prieta Earthquake - Damage observed (Brady and Perkins 1991):
Homes Damaged 24347 Homes Destroyed 1119 Businesses damaged 4316 Businesses destroyed 382 Road Damage ($million) 833 Public utilities damage ($million) 43 PG&E losses ($million) 74
5.4.2 Hurricane Andrew
On August 24, Hurricane Andrew hit South Florida, resulting in the destruction of
85,000 dwelling units and buildings - nearly $23 billion (West and Lenze, 1994) physical
damages -- leaving hundreds of thousands of people homeless. Major areas of impact
were the residential suburbs south of downtown Miami. The following counties in South
Florida were affected - Broward, Collier, Dade, and Monroe. The Dade County bore the
brunt of the damages. Hurricane Andrew has been classified as one of the costliest
weather related catastrophe to have hit US in the recent years. The impact on measured
income was substantial. The hurricane reduced 1992 real income growth statewide in
Florida from 1.8 percent to 1.0 percent, turning stable per capita income into real per
capita income decline.
5.4.3 Northridge Earthquake
The Northridge earthquake of January 17, 1994 killed 57 people and injured an
estimated 10,000. The counties affected were Los Angeles, Ventura, and Orange. The
costs of repairing earthquake damage and providing relief to victims probably exceeded
Chapter Five: Regional Impact of Catastrophes
190
$30 billion, including $12- 15 billion in insured losses, making that event the most costly
disaster in U. S. history. The number of households and businesses that suffered losses in
the Northridge earthquake far exceeded the size of the victim population in other recent
major disasters in the U. S., including Hurricane Hugo in 1989 and Hurricane Andrew in
1992. The assistance effort launched after the earthquake was the largest ever undertaken
for an U.S. disaster. Applications to the Federal Emergency Management Agency for
various forms of housing assistance totaled well over half a million. In the year following
the earthquake, over 50,000 businesses applied to the U.S. Small Business
Administration for disaster loans, and over $1.3 billion in loans had been paid out.
5.5 A comparison of the impacts of the events
The gross state product of California in 1989 was $591 billion. The personal
income of the worst affected county i.e. San Francisco – San Jose was $ 149 billion. The
Loma Prieta earthquake occurred in one of the most technology advanced counties of
California. Fortunately the disruption was minimal since the highway and rail network
was not substantially affected. Businesses in downtown Santa Cruz were able to relocate
to other areas of Santa Cruz County (Brady and Perkins, 1991). The direct losses were
relatively minimal. The loss to county personal income ratio was around 4 per cent. The
loss to gross state product ratio was around 1.1 per cent (Table 5.2).
The gross state product of California in 1994 was $722 billion. The personal income of
the worst affected county i.e. Los Angeles was $205 billion. The Northridge earthquake
occurred in one of the most prosperous counties of California. Perhaps, for this reason the
direct economic losses were very high given the fact that the earthquake itself was not of
a severe magnitude. The loss to county personal income ratio was around 12 per cent.
The loss to gross state product ratio was around 4 per cent (Table 5.2).
Chapter Five: Regional Impact of Catastrophes
191
Table 5.2 Observations on the effects of the Loma Priets and Northridge earthquakes and Hurricane Andrew
Loma Prieta Earthquake Northridge Earthquake Hurricane AndrewPlace of occurrence (state) California California FloridaDate of occurrence 17-Oct-89 17-Jan-94 24-Aug-92Estimated direct losses (upper bound) in dollars 7,000,000,000 30,000,000,000 31,000,000,000 Estimated direct losses (lower bound) in dollars 6,000,000,000 25,000,000,000 26,500,000,000
Gross state product in the year of the event 590,961,519,000 722,223,500,000 270,820,652,000
Personal earnings of the worst affected county in the year of the event
San Francisco-San Jose CMSA Los Angeles Dade
148,574,598,000 204,872,592,000 34,303,921,000 Ratio of loss to gross state product 1.1% 3.8% 10.6%Percentage loss to county personal income (assuming 90% loss occurred in worst affected county) 3.9% 12.1% 75.4%
Chapter Five: Regional Impact of Catastrophes
192
We can contrast Northridge earthquake to the occurrence of the Hurricane Andrew that
occurred in the state of Florida. The gross state product of Florida, in 1992, was $271
billion, almost one-third of California. Yet, Hurricane Andrew resulted in approximately
the same amount of dollar damages as Northridge Earthquake. Dade, the southern Florida
County worst affected by Hurricane Andrew, had personal earnings of around $34
billion, almost a sixth of the net earnings of Los Angeles. The loss to county personal
income ratio was around 75 percent. The loss to gross state product ratio was around 10
percent (Table 5.2).
San Francisco – San Jose CMSA contribution to the California's gross product remained
almost unchanged if we compare the pre- and post - disaster trends. It was 25.31 percent
in 1988, 25.14 percent in 1989, and 25.16 percent in 1990 (Table 5.3). Los Angeles
contribution to the California's gross product remained almost unchanged if we compare
the pre- and post - disaster trends. It was 28.66 percent in 1993, 28.37 percent in 1994,
and 28.3 percent in 1995. The slight drop is the long-term trend, rather than the effect of
Northridge earthquake (Table 5.3). Dade's contribution to Florida's gross product showed
the effects of Hurricane Andrew. Dade's contribution was 13.81 percent in 1991, dropped
to 12.7 percent in 1992, but rebounded to 13.5 percent in 1993 (Table 5.3).
The difference in the loss-to-Gross Regional Product ratios is partly responsible for the
different ways the macro-economic variables such as net earnings are impacted. It can be
readily inferred that a high loss-to-Gross Regional Product ratio results in a greater
regional impact. Thus, Loma Prieta and Northridge earthquakes had relatively minimal
impacts on the regional economies, whereas Hurricane Andrew’s was felt statewide.
Chapter Five: Regional Impact of Catastrophes
193
Table 5.3 Effect on county's components of personal income
Loma Prieta Earthquake Hurricane Andrew Northridge Earthquake1988 1989 1990 1991 1992 1993 1993 1994 1995
Effect on county's contribution to GSP25.3% 25.1% 25.2% 13.8% 12.7% 13.5% 28.7% 28.4% 28.3%
(Net earnings by place of work)/Personal incomeState 70.4% 69.7% 69.3% 55.8% 57.1% 56.5% 67.1% 66.7% 65.8%Affected county 70.2% 69.8% 69.2% 61.5% 67.8% 63.1% 67.7% 67.2% 65.9%
(Dividends, interest, and rent)/Personal incomeState 18.2% 18.7% 18.7% 27.0% 24.3% 25.0% 17.2% 17.5% 18.6%Affected county 19.6% 19.9% 20.2% 22.4% 12.7% 19.0% 16.6% 16.8% 18.3%
(Transfer Payments)/Personal incomeState 12.7% 12.7% 13.1% 17.2% 18.7% 18.6% 15.7% 15.8% 15.6%Affected county 10.6% 10.8% 11.0% 16.8% 20.2% 18.6% 15.6% 16.0% 15.8%
Chapter Five: Regional Impact of Catastrophes
194
5.5.1 Effects on the components of personal income
The personal income of an area (BEA, 1997), is defined as the income that is
received by, or on behalf of, all the individuals who live in the area. It is calculated as the
sum of: the net earnings by place of work, the personal dividend income and the personal
rental income, and the transfer payments. To delineate the effects of events like the
Northridge earthquake from general economic trends, we compare the values of these
components at county level to the state level. If we compare the trends of the California
state to that of Los Angeles, they are almost indistinguishable (Table 5.3). For example,
California earned 67.11 percent of its personal income from net earnings by place of
work in 1993, 66.67 percent in 1994 and 65.82 percent in 1995. The trend in the affected
county, in this case Los Angeles, is almost same - 67.71 percent in 1993, 67.22 percent in
1994, 65.92 per cent in 1995. Notice, however, that the transfer payments have slightly
increased for Los Angeles in 1994 and 1995 (Table 5.3). This should be apparent since
federal aid is accounted for under transfer payments. Similar trends can also be observed
for the Loma Prieta earthquake.
In the case of Hurricane Andrew, the effects were more pronounced and different. There
was a drop of about 10 percentage points in Dade's dividend, interest and rent component
of the personal income though there was no appreciable change of the same component in
Florida's personal income (Table 5.3). Florida's net earnings by place of work component
of the personal income increased in 1992, but Dade's increase was higher. Details of this
will become apparent if we examine the various components that make up the net
earnings by place of work.
5.5.2 Effects on the components of net earnings by place of work
The net earnings by place of work have two components - farm and non-farm. In
this paper, we do not consider the effects of the catastrophic events on farm earnings
Chapter Five: Regional Impact of Catastrophes
195
since they form a small part of the San Francisco – San Jose CMSA, Los Angeles and
Dade's earnings.
The components of non-farm earnings include earnings from construction,
manufacturing, transportation, wholesale trade, retail trade, finance insurance and rental
income, services and government.
The methodology used herein to study the effects on each of the sectors is by tracking the
changes in the growth rates. For example, the growth rates for construction earnings for a
period two years prior (pre-event) to the occurrence of the catastrophe and three years
after (post-event) the occurrence of the catastrophe (including the year of occurrence) are
calculated (Table 5.4). The mean of the pre-event growth rates is compared to the post
event means.
From Table 5.4 it is clear that excluding the Loma Prieta earthquake, the events tended to
cause similar changes in direction of the growth rates for all sectors except finance,
insurance, and rental income. But the magnitude of change of growth rates was more
intense in the case of Hurricane Andrew in the following sectors – construction,
manufacturing, transportation, retail trade, governmental spending. In the case of Loma
Prieta the trends were opposite, except for the transportation sector. Partial explanation
for Bay Area’s opposite behavior is that U.S. economy was in a recession during the
1990-1991 period.
If we exclude the Loma Prieta case, all the sectors received a boost to their growth rates
except the government and finance, insurance, rental income of Los Angeles. For
example, the pre-event construction growth rate jumped from a negative 5.5 percent to a
positive 6.4 percent.
But to answer the question as to whether these changes in growth rates were propagated
to the state level we have to analyze taking into account the changes in the net earnings
by place of work of the state.
Chapter Five: Regional Impact of Catastrophes
196
Table 5.4 Growth rates of the components of net earnings
t-2 t-1
Event Year t0 t1 t2
Pre-event mean
Post-event mean Direction
ConstructionLos Angeles -9.5% -7.1% 12.7% 2.2% -1.8% -8.3% 4.4% increaseDade -0.4% -10.8% 3.8% 21.7% -6.1% -5.6% 6.5% increaseSF-San Jose 5.6% 8.4% 4.2% 2.2% -8.6% 7.0% -0.7% decreaseManufacturingLos Angeles -3.5% -3.9% -2.7% -3.1% 2.9% -3.7% -1.0% increaseDade 0.4% 1.1% 4.0% 0.8% 1.8% 0.8% 2.2% increaseSF-San Jose 5.3% 8.2% 6.9% 4.7% 4.7% 6.7% 5.4% decreaseTransportationLos Angeles 3.5% 4.2% 4.7% 4.5% 3.0% 3.9% 4.1% increaseDade 8.9% -0.1% 0.4% 13.6% 5.7% 4.4% 6.6% increaseSF-San Jose 5.0% 2.5% 4.7% 8.4% 3.9% 3.7% 5.7% increaseWholesaleLos Angeles 2.6% -5.1% 2.9% 4.7% 1.5% -1.2% 3.0% increaseDade 7.1% 1.4% 7.0% 3.2% 3.5% 4.2% 4.6% increaseSF-San Jose 5.9% 13.1% 8.0% 6.2% 1.3% 9.5% 5.2% decreaseRetailLos Angeles 1.0% -1.9% 1.8% 3.7% 2.7% -0.4% 2.7% increaseDade 2.9% -3.1% 3.8% 10.6% 3.0% -0.1% 5.8% increaseSF-San Jose 2.4% 7.0% 6.9% 4.8% 1.9% 4.7% 4.5% decreaseFinance,insurance, and real estateLos Angeles 10.7% 8.4% -3.7% 2.1% 4.8% 9.6% 1.1% decreaseDade 1.5% -2.7% 14.1% 6.9% 3.8% -0.6% 8.3% increaseSF-San Jose 9.1% 8.2% -1.7% 7.0% 3.6% 8.7% 3.0% decreaseServicesLos Angeles 4.5% 1.5% 1.6% 6.2% 6.8% 3.0% 4.9% increaseDade 6.6% 2.9% 7.9% 7.9% 4.8% 4.7% 6.9% increaseSF-San Jose 10.9% 14.6% 9.5% 11.2% 4.9% 12.8% 8.5% decreaseGovernmentLos Angeles 4.3% -0.4% 1.2% 2.3% 1.0% 1.9% 1.5% decreaseDade 9.6% 6.4% -1.6% 7.1% 5.6% 8.0% 3.7% decreaseSF-San Jose 7.8% 6.0% 6.8% 9.0% 5.7% 6.9% 7.2% increase
Chapter Five: Regional Impact of Catastrophes
197
5.5.3 Dampening out effect
The methodology adopted for this section relies on comparing the changes in the
components of earnings at the state and county levels for the years immediately before
and after the event (Table 5.5).
Hurricane Andrew affected the components that make up Dade’s net earnings by place of
work and these changes were reflected in Florida’s earnings. For example, the
construction component of earnings for Dade changes from 4.7% in 1992 to 5.3% in
1993, an increase of 12.5% while Florida’s increase was 3.8% (Table 5.5). An important
observation in the case of Hurricane Andrew is that all the components of net earnings of
Dade and Florida showed similar trends. One possible inference from this observation is
that the effect of Hurricane Andrew was propagated to the state level, though these
effects were considerably dampened at the state level.
The same cannot be said for the Northridge earthquake. Though the damage was of the
same magnitude as that of hurricane Andrew, the robustness of Los Angeles economy
was partly responsible for dampening out the effects at the county level itself. This can be
inferred from the behavior of each of the components of net earnings (Table 5.5). For
example, while the wholesale earnings of California dropped from 2.3% in 1994 to 1% in
1995, wholesale earnings in Los Angeles dropped by only 0.6 percentage points in the
same period (Table 5.4). In the case of the Loma Prieta earthquake, the direct losses
were relatively lower. The changes in the components were not propagated to the state
level as is clear from Table 5.5.
Chapter Five: Regional Impact of Catastrophes
198
Table 5.5 Effect on county's components of net earnings by place of work
Loma Prieta Earthquake Hurricane Andrew Northridge EarthquakeConstruction/Net earnings No dampening effect Dampened out Dampened out*State 1.0% -5.6% -6.6% 3.9% 5.6% -0.7%Affected county -2.2% -5.0% -2.4% 12.5% 11.6% -1.1%Manufacturing/Net earnings No dampening effect Dampened out Dampened outState -1.3% -4.4% -2.9% -4.9% -1.1% -2.0%Affected county 0.4% -2.7% -2.3% -6.8% -3.6% -6.2%Transportation/Net earningsNo dampening effect Dampened out Dampened outState 0.29% -0.28% -0.22% 2.19% 0.78% -0.90%Affected county -1.69% 0.76% -5.59% 5.04% 3.72% 1.17%Wholesale/Net earnings Dampened out Dampened out No Dampening observed#
State 1.22% -0.17% 0.13% -3.95% 2.34% 0.96%Affected county 1.34% -1.26% 0.60% -4.56% 1.92% 1.29%Retail/net earnings Dampened out Dampened out No Dampening observedState -0.1% -2.8% -1.8% -0.5% 1.1% -1.6%Affected county 0.4% -2.6% -2.4% 2.3% 0.9% 0.3%Fin.,ins.,etc./net earnings No dampening effect Dampened out No Dampening observedState -9.3% -1.7% 8.1% 3.6% -4.6% -0.2%Affected county -7.7% -0.5% 7.2% -1.1% -4.6% -1.2%Services/net earnings No dampening effect Dampened out No Dampening observedState 2.7% 4.4% 1.1% 0.6% 0.1% 1.8%Affected county 2.8% 3.4% 1.5% -0.2% 0.7% 2.8%Government/net earnings No dampening effect Dampened out No Dampening observedState 0.9% 2.0% -3.4% -2.1% -0.1% -2.2%Affected county 0.2% 1.3% -6.4% -0.8% 0.2% -0.8%
Notes:*Dampened out here means that the state trend follows the affected county's trend but to a lesser degree # No Dampening observed implies that the state trend is different from the affected county's trend Components of earnings = Component earnings/total earnings
Chapter Five: Regional Impact of Catastrophes
199
Hurricane Andrew and Northridge earthquake caused almost equal amounts of direct
losses. But the effects were both similar and different. The effects were similar to the
extent that almost all sectors of the economy received a boost to their growth rates,
probably due to flow of external aid and reconstruction. The effects were different in the
sense that Northridge earthquake occurred in a county with a robust economy as
compared to Hurricane Andrew. The effects of Loma Prieta and Northridge earthquakes
were localized, whereas effects of Hurricane Andrew were felt at the state level.
Data presented herein suggests that shock to a robust economy, such as California's from
a hazardous event of magnitude comparable to the Northridge earthquake, results in
localization of the effects. A higher magnitude of shock may propagate the effects to a
greater extent, but will be nevertheless localized. Hurricane Andrew was a greater shock
in terms of the loss-GSP ratio. But the propagation of the shock was felt at the state level,
though it dampened out rapidly both spatially and temporally.
5.6 Simulation of the effects with the regional model
In this section we describe the application of this model to simulate the impacts of three
events on the regional economies - Loma Prieta Earthquake, Hurricane Andrew, and
Northridge Earthquake. Appendix J lists a computer program in MATLAB© that
simulates this model. Personal income, population, and employment in the affected
regions before the occurrence of the catastrophe are shown in Table 5.6.
Table 5.6 Main economic indicators before the events
Loma Prieta Earthquake
Hurricane Andrew
Northridge Earthquake
Personal income (billions of dollars) 138.8 35.8 204.8Population (million number of persons) 6.1 2.0 9.1Earnings by place of work (billions) 105.3 26.7 163Dividends, interest, and rent (billions) 27.0 7.8 32.5Transfer payments (billions) 14.7 6.0 31.5Total full- and part-time employment (million)
3.8 1.1 4.9
Government and government employment
0.52 0.13 0.56
Average earnings per job (dollars) 27446 25208 32982Construction Employment (million) 0.19 0.04 0.16
Chapter Five: Regional Impact of Catastrophes
200
5.6.1 Loma Prieta Earthquake
The baseline prediction without an earthquake is shown as the curve labeled as
‘baseline’ in the Fig. 5.1. Assuming the earthquake caused a direct loss of $6.2 billion
and 7100 jobs were lost, the model forecasts the curve labeled ‘earthquake’ in the Fig.
5.1. It was also assumed that the regional capacity to satisfy regional demand fell by
2.5% (since the direct capital loss was around 3% of the pre-disaster level of capital).
Subsequently the regional capacity to satisfy regional demand is assumed to reach its pre-
disaster levels over a period of five years after the event. This curve assumes that no aid
was given to the affected area.
The model generates a forecast, which is labeled ‘aid’ in Fig. 5.1, based on two
assumptions. The first assumption is that $1.0 billion is used as transfer payments in the
first two periods after the earthquake. The second assumption is that the regional capacity
to satisfy regional demands falls by 2% (instead of 2.5% in no aid case) and recovers to
its pre-disaster values within a period of years as a result of reconstruction efforts. If we
compare the ‘aid’ curve with the actual observed values given by Bureau of Economic
Analysis (BEA) the mean absolute percentage error is 2.59 (Table 5.7).
Table 5.7 Comparison of model predictions with observed values – Loma Prieta
Earthquake Observed Personal Income ($ Billions)
With Eq. % error
Without Eq.% error
With aid % error
1989 149 -5.75 -1.56 -3.80 1990 161 -6.23 -2.35 -4.28 1991 166 -5.47 -2.11 -4.54 1992 176 -5.02 -2.24 -4.06 1993 181 -2.20 -0.01 -1.16 1994 188 0.37 1.83 1.46 1995 200 1.47 2.49 2.19 1996 216 1.10 1.62 1.39 1997 233 -0.67 -0.26 -0.43
Mean absolute % error
3.14 1.61 2.59
Chapter Five: Regional Impact of Catastrophes
201
Fig. 5.2a shows plots of percentage changes of the gross regional product with respect to
no-event scenarios. The curve labeled ‘without aid’ corresponds to difference in behavior
with respect to no-event scenario of the gross regional product when no external aid is
given. The curve labeled ‘with aid’ is a similar curve where aid in the form of transfer
payments is assumed. In subsequent discussions, the comparison with a baseline no-event
scenario is illustrated via similar plots. From Fig. 5.2a it is clear that without aid the GRP
would have been lower by 5.39 percent during 1989. Because of transfer payments and
improved regional capacity due to reconstruction, the GRP was lower by 3.78, a gain of
1.64 percent. With aid and increased regional capacity, the region recovers within 5
years, whereas without aid and slower increases in regional capacity, the region would
have recovered in 7 years. Consumption shows similar trends (Fig. 5.2b). The
consumption in no-aid case is lower by 4.18 percent whereas with aid it is lower by 2.2
percent. The theoretical model presented in Section 3.5 also shows similar trends in the
change output and consumption. Immediately following the event, consumption and
output fall. The theoretical models also predict that greater aid results in more rapid
recovery. Increases in regional capacity are modeled by changes in productivity.
The reasons for lower GRP than the no-event scenario can be readily discerned when one
plots the capital and employment as in Figs. 5.2c and 5.2d, respectively. Capital (Fig.
5.2c) is lower by about 4 percentage points three years after the event in the no-aid
scenario, whereas it is lower by 3.5 percent with aid. The capital steadily converges to the
no-event scenario. Employment (Fig. 5.2d) is lower by 5.37 percentage points in the no-
aid scenario and is somewhat ameliorated with aid (is lower by 3.76 points).
Prices are modeled as consumption deflator (Fig. 5.2e). Lower the deflator higher is the
real price. Without aid the deflator is lower by 0.08 to 0.14 percentage points during the
first three years. With aid, prices are nearer the no-event case. With aid the deflator is
lower by only 0.06 to 0.1 percentage points, implying that prices are lower than the case
of no-aid, but higher than no-event case.
Chapter Five: Regional Impact of Catastrophes
202
Fig. 5.1 Effect of Loma Prieta Earthquake on Personal Income of San Francisco - San Jose CMSA
140150160170180190200210220230240
1989 1990 1991 1992 1993 1994 1995 1996 1997
Pers
onal
inco
me
(bill
ions
, nom
inal
)
Baseline Earthquake Aid Actual(BEA)
Chapter Five: Regional Impact of Catastrophes
203
Fig. 5.2a Effect on Gross Regional Product (Loma Prieta)
-6
-5
-4
-3
-2
-1
0
1
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
year
without aid, Regional capacity = 0.975,0.98,0.985,0.99,0.995with aid, Regional capacity=0.98,0.985,0.99,0.995
Fig. 5.2b Effect on Consumption (Loma Prieta)
-4.5
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
year
without aid with aid
Fig. 5.2c Effect on Capital stock (Loma Prieta)
-4.5
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
year
Capital stock - without aid Capital stock - with aid
Chapter Five: Regional Impact of Catastrophes
204
Fig. 5.2d Effect on Employment (Loma Prieta)
-6
-5
-4
-3
-2
-1
0
1
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
year
without aid with aid
Fig. 5.2e Effect on Consumption Deflator (Loma Prieta)
-0.16
-0.14
-0.12
-0.1
-0.08
-0.06
-0.04
-0.02
0
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
year
without aid with aid
Fig. 5.2f Effect on Government spending (Loma Prieta)
-3
-2.5
-2
-1.5
-1
-0.5
0
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
year
without aid with aid
Chapter Five: Regional Impact of Catastrophes
205
5.6.2 Hurricane Andrew
Assuming the hurricane caused a direct loss of $26 billion and 10500 jobs were
lost, the model forecasts the curve labeled ‘hurricane’ in the Fig. 5.3. This curve assumes
that no aid was given to the affected area. It also assumes that the regional capacity to
satisfy regional demand was down by 15% (as compared to the no-event scenario) during
1992-1993. Hurricane Andrew destroyed about 65% of the capital stock. But the region
was operating at approximately 50% of its capacity during the pre-event period (West
and Lenze, 1994). It was therefore assumed that without aid the region has the capacity to
supply 85% of its local demand. This capacity is assumed to steadily rise to its full no-
event value within a period of 9 years, without aid. With aid, the model generates a
forecast, which is labeled ‘aid’ in Fig. 5.3. It is assumed that transfer payments of $0.268
billion in 1992 and $0.403 billion are made in the years 1992 and 1993 (West and Lenze,
1994). It is also assumed that reconstruction expenditures are as follows: $2.377 billion in
1992, $7.939 billion in 1993, $4.282 billion in 1994, $2.670 billion in 1995, and $0.045
billion in 1996 (West and Lenze, 1994). This is translated in jobs (thousands) as 3.846 in
1992, 28.472 in 1993, 27.904 in 1994, 22.288 in 1995, and 0.396 in 1996 (West and
Lenze, 1994). The ‘aid’ curve compares well with the actual observed values given by
Bureau of Economic Analysis (BEA): the mean absolute percentage error is 2.70 percent
(Table 5.8).
Table 5.8 Comparison of model predictions with observed values - Hurricane Andrew Personal
Income (BEA $
billions)
With Hurricane
% error
Without hurricane % error
With aid % error
1992 34.3 -6.12 16.26 -2.55 1993 39.2 -11.82 6.51 -2.00 1994 40.5 -9.37 7.33 1.61 1995 42.5 -7.39 6.87 1.85 1996 44.7 -3.06 8.55 2.35 1997 46.2 0.53 10.90 5.83
Mean Absolute Errors
6.38 9.40 2.70
From Fig. 5.4a it is clear that without aid the GRP would have been lower by 25 percent
during 1992. Because of reconstruction spending, transfer payments and improved
Chapter Five: Regional Impact of Catastrophes
206
regional capacity due to reconstruction, the GRP was lower by 20 percent in 1992 (a gain
of 5 points as compared to no-aid case) and jumped to 8.1 percent in 1993 (a gain of 11.5
points). Consumption shows similar trends (Fig 5.4b). The consumption in no-aid case is
lower by 18.9 percent whereas with aid it is lower by 15.9 percent.
Capital (Fig. 5.4c) is lower by about 8 percent in 1992 and falls steadily to 12.5 percent in
1995 after which it shows an upward trend in the no-aid scenario. With aid, however it is
6.8 percent lower in 1992 and 7.2 percent lower in 1993. The capital steadily converges
to the no-event scenario. Employment (Fig. 5.4d) is lower by 25 percentage points in the
no-aid scenario and aid helps in rapidly improving the situation. Employment is lower by
21 percent in 1992 but is only 8 percent in 1993.
Without aid the deflator is lower by 0.46 to 0.78 percentage points during the first three
years. With aid the deflator is lower by 0.4 percentage points during first three years and
converges to the no event case.
Fig 5.3 Effect of Hurricane Andrew on Personal Income of Dade county
30
35
40
45
50
55
1992 1993 1994 1995 1996 1997
Pers
onal
inco
me
(bill
ions
, nom
inal
)
Baseline Hurricane Aid Actual(BEA)
Chapter Five: Regional Impact of Catastrophes
207
Fig. 5.4a Effect on Gross Regional Product (Andrew)
-30
-25
-20
-15
-10
-5
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
year
GRP GRP(with aid)
Fig. 5.4b Effect on Consumption (Andrew)
-20-18-16-14-12-10-8-6-4-20
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
year
Consumption Cons.(with aid)
Fig. 5.4c Effect on Capital Stock (Andrew)
-14
-12
-10
-8
-6
-4
-2
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
year
Capital stock capital stock(with aid)
Chapter Five: Regional Impact of Catastrophes
208
Fig. 5.4d Effect on Employment (Andrew)
-30
-25
-20
-15
-10
-5
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
year
Employment Employment (with aid)
Fig. 5.4e Effect on Price Index (CP) (Andrew)
-0.9
-0.8
-0.7
-0.6
-0.5
-0.4
-0.3
-0.2
-0.1
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
year
Price index Price index(with aid)
Fig. 5.4f Effect on Government spending (CP) (Andrew)
-14
-12
-10
-8
-6
-4
-2
0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
year
Government spending Government spending(with aid)
Chapter Five: Regional Impact of Catastrophes
209
Fig. 5.4g Effect on Investment (Andrew)
-50
-40
-30
-20
-10
0
10
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
year
Without aid with aid
Chapter Five: Regional Impact of Catastrophes
210
5.6.3 Northridge Earthquake
Northridge earthquake caused a direct loss of $25 billion and 25,800 jobs were
lost in the Los Angeles County. The model forecasts the curve labeled ‘earthquake’ in the
Fig. 5.5. This curve assumes that no aid was given to the affected area. It also assumes
that the regional capacity to satisfy regional demand was down by 5 percent (as compared
to the no-event scenario) during 1992-1993. Hurricane Andrew destroyed about 7 percent
of the capital stock. Assuming the region had idle capacity of 2% of the capital stock it
can be assumed that without aid the regional capacity to satisfy regional demand was
lower by 5 percent. This capacity is assumed to steadily rise to its full no-event value
within a period of 6 years, without aid. With aid, the model generates a forecast, which is
labeled ‘aid’ in Fig. 5.5. It is assumed that transfer-payments of $0.5 billion each in 1994
and 1995. It is also assumed that reconstruction expenditures resulted in job (in
thousands) impacts as follows: 6 in 1994, 50 in 1995, 45 in 1996, 40 in 1997, and 20 in
1998. The ‘aid’ curve compares well with the actual observed values given by Bureau of
Economic Analysis (BEA): the mean absolute percentage error is 1.3 percent (Table 5.9).
Table 5.9 Comparison of model predictions with observed values – Northridge
Earthquake Percent Errors
Personal Income (BEA $
billions)
With Eq. Without Eq. With aid
1994 204 -7.52 0.87 -2.91 1995 214 -3.45 0.73 -0.77 1996 224 -2.72 0.39 -0.72 1997 234 -2.82 -0.50 -0.76
Mean
Absolute Errors
4.12 0.62 1.29
From Fig. 5.6a it is clear that without aid the GRP would have been lower by 11 percent
during 1994. Because of reconstruction spending, transfer payments and improved
regional capacity due to reconstruction, the GRP was lower by 5 percent in 1994 (a gain
Chapter Five: Regional Impact of Catastrophes
211
of 6 points as compared to no-aid case). Consumption shows similar trends (Fig 5.6b).
The consumption in no-aid case is lower by 8 percent whereas with aid it is lower by 4
percent.
Capital (Fig. 5.6c) is lower by about –1.8 percent in 1994 and falls steadily to –2.4
percent in 1996 after which it shows an upward trend in the no-aid scenario. With aid,
however it is 0.8 percent lower in 1994 and 1.0 percent lower in 1995-96. The capital
steadily converges to the no-event scenario. Employment (Fig. 5.6d) is lower by 10.7
percentage points in the no-aid scenario and aid helps in rapidly improving the situation.
Employment is lower by 5 percent in 1994 but is only 1.7 percent in 1995.
Without aid the deflator is lower by 0.15 to 0.17 percentage points during the first three
years. With aid the deflator is lower by 0.07 percentage points during first three years and
converges to the no event case.
Having discussed historical events, the next section presents simulation results of the
impacts of probable earthquake scenarios in the San Francisco – San Jose region (Bay
Area).
Fig. 5.5 Effect of Northridge Earthquake on Personal Income of Los Angeles County
185190195200205210215220225230235
1994 1995 1996 1997
Pers
onal
inco
me
(bill
ions
, nom
inal
)
Baseline Earthquake Actual(BEA) With Aid
Chapter Five: Regional Impact of Catastrophes
212
Fig. 5.6b Effect on Consumption (Northridge)
-9
-8
-7
-6
-5
-4
-3
-2
-1
0
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
year
without aid with aid
Fig. 5.6c Effect on Capital Stock (Northridge)
-2.5
-2
-1.5
-1
-0.5
0
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
year
without aid with aid
Fig. 5.6a Effect on Gross Regional Product (Northridge)
-12
-10
-8
-6
-4
-2
0
2
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
year
without aid with aid
Chapter Five: Regional Impact of Catastrophes
213
Fig. 5.6d Effect on Employment (Northridge)
-12
-10
-8
-6
-4
-2
0
2
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
year
without aid with aid
Fig. 5.6f Effect on Government Spending (Northridge)
-4
-3.5
-3
-2.5
-2
-1.5
-1
-0.5
0
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
year
without aid with aid
Fig. 5.6e Effect on Consumer price CP (Northridge)
-0.2
-0.18-0.16
-0.14-0.12
-0.1
-0.08-0.06
-0.04-0.02
0
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009
year
without aid with aid
Chapter Five: Regional Impact of Catastrophes
214
5.7 Simulation of impact of probable earthquake scenarios in the Bay Area
In this section results of simulation studies for various earthquake scenarios in the
San Francisco Bay Area are presented. The model that was used to study the Loma Prieta
earthquake is extended to include an scenario earthquake that occurs in the year 2000. A
catastrophe abruptly increases the difference between optimal and actual capital levels.
The influx of aid and insurance payments speeds up the adjustment process. But this
additional spending should be carefully modeled so as to avoid double counting by
including only the direct endogenous spending increase from reconstruction. Loss of
capital stock available to the production process reduces regional output, employment,
wage income, and proprietors’ income. Unless these losses are accounted for as
exogenous changes, the model will reflect only reconstruction gains from a catastrophe.
As restoration and reconstruction evolves, the regional output, employment, wage
income, and proprietors’ income return to their ‘normal values’ and this has to be suitably
modeled.
There are two possible ways in which this increased reconstruction investment can be
modeled. If the purpose of the model is to examine an event ex post and we have data
regarding the investment spending for reconstruction, then the job and output losses can
be restored in conjunction with the time path of capital reconstruction.
In the absence of data regarding reconstruction expenditure, we assume that only a
portion of the investment is used for building new structures and the rest is used for
reconstruction. The recovery of a region depends crucially on the pattern of the
reconstruction expenditure and the fractions apportioned to new buildings and restoration
of damaged structures. The uncertainty surrounding these fractions can be studied by
suitably changing some parameters. In the model proposed, suitable changes were made
to α to reflect the true dynamics of the recovery process.
One important effect of a catastrophe is to cause a change in the relationships between
local demands and demand for locally produced output. Regional patterns change
Chapter Five: Regional Impact of Catastrophes
215
abruptly as the regional share in satisfying local demand decreases abruptly. As a result
purchases made from outside the region increase. The decrease in local potential to
satisfy local needs may prolong the deleterious effect of a catastrophe if region has no in-
built excess capacity or receives no external aid or receives external aid that is not
suitably used for reconstruction.
The earthquake scenarios are generated based on assumptions of direct losses to capital
stock and the number of jobs lost. These scenarios are presented in Table 5.10. Four
possible scenarios are simulated. For each of these scenarios assumptions regarding
regional capacity are made. For example, we assume regional capacity is lower by 7%
(Table 9, row 4, col. 2) when 10% of the capital has been lost in the no aid case. When
the affected region gets external aid, its capacity is augmented, and the regional capacity
is lower by 3% only in the year 2000 (Table 9, row 4, col. 3). These scenarios are
simulated and the results are presented in Figs. 5.7 to 5.11. The results show the
importance to regional capacity in dampening out the effects of a catastrophe. For
example (Fig. 5.8), in the 10 percent capital loss scenario, a regional capacity increase of
4 percent (due to aid, rapid reconstruction, or inherent pre-event excess capacity) results
in absorbing 7 percent (14.4 – 7.1) of loss in the gross regional product during the year
2000. This motivates the study of the model by varying other crucial parameters that may
change after a catastrophe. Such studies will be discussed in the next section.
Table 5.10 Earthquake scenarios and assumptions about regional capacity
Loss Scenarios Regional capacities
10% of capital ($ 30 billion) and 25,000 jobs
20% of capital ($ 60 billion) and 50,000 jobs
28% of capital ($ 80 billion) and 75,000 jobs
35% of capital ($ 100 billion) and 100,000 jobs
Year No aid aid No aid aid No aid aid No aid aid 2000 7 3 17 10 25 18 32 25 2001 5 1 15 5 20 12 27 10 2002 3 - 10 3 15 7 22 15 2003 1 - 5 1 10 3 17 10 2004 - - 3 - 5 1 12 5 2005 - - 1 - 3 - 7 1 2006 - - - - 1 - 3 - 2007 - - - - - - 1 - Fig. No. 5.7 5.8 5.7 5.9 5.7 5.10 5.7 5.11
Chapter Five: Regional Impact of Catastrophes
216
Fig. 5.7 Effect on Gross Regional Product - Probable scenarios with no external aid
-60
-50
-40
-30
-20
-10
0
10
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year$30 billion and 25,000 jobs $60 billion and 50,000 jobs$80 billion and 75,000 jobs $100 billion and 100,000 jobs
Fig. 5.8 Effect on Gross Regional Product - Probable scenarios with and without aid (10% loss)
-16-14
-12-10
-8-6
-4-2
02
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year$30 billion and 25,000 jobs $30 billion and 25,000 jobs with aid
Chapter Five: Regional Impact of Catastrophes
217
Fig. 5.9 Effect on Gross Regional Product - Probable scenarios with and without aid (loss = 20% of capital)
-35
-30
-25
-20
-15
-10
-5
0
5
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year$60 billion and 50,000 jobs $60 billion and 50,000jobs (with aid)
Fig. 5.10 Effect on Gross Regional Product - Probable scenarios with and without aid (loss = 28% of capital)
-45-40-35-30-25-20-15-10-505
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year$80 billion and 75,000 jobs $80 billion and 75,000jobs (with aid)
Fig. 5.11 Effect on Gross Regional Product - Probable scenarios with and without aid (loss = 35 % of capital)
-60
-50
-40
-30
-20
-10
0
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year
$100 billion and 100,000 jobs $100 billion and 100,000jobs (with aid)
Chapter Five: Regional Impact of Catastrophes
218
5.8 Model behavior when crucial parameters are varied
By varying the many parameters of the model in a controlled fashion one can get
a better understanding of the underlying mechanism of the model. In this section the 2nd
damage scenario (loss $60 billion and 50,000 jobs) is used.
5.8.1 Transfer payments effects (Fig. 5.12)
In personal income, transfer payments are income payments to persons for which
no current services are performed. They are payments by government and business to
individuals and non-profit institutions. Federal aid after a disaster is put in this category.
The effect of making transfer payments of 10 percent of the capital loss during the years
2000, 2001, and 2002 is not very significant. Increases in the transfer payments improve
the gross regional product to a small degree.
5.8.2 Consumer spending effects (Fig. 5.13)
Immediately after a catastrophe, propensity to consume increases, at least
temporarily, in the affected region. Consumption spending as a proportion of the real
disposable income increases. People rely on savings, credit, or insurance to finance their
reconstruction spending. Increases in consumption spending result in an increase in the
personal income for the region. Whether this reconstruction gain offsets the losses from
reduction in regional output, employment, wage income, and proprietors’ income can be
answered only by examining data from a specific event.
Consumer spending preferences are increased by 10 percent. Within one year of the event
the economy does better than the baseline no earthquake scenario due to increased
spending.
Chapter Five: Regional Impact of Catastrophes
219
Fig. 5.12 Effect on gross regional product due to 10% increase in transfer payments
-25
-20
-15
-10
-5
0
5
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year$60 billion and 50,000 job loss transfer payments 10% of loss
Fig. 5.13 Effect on gross regional product due to 1% increase in consumer spending
-25
-20
-15
-10
-5
0
5
10
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year$60 billion and 50,000 jobs loss 1% increase in consumption spending
Fig. 5.14 Effect on gross regional product due to 10% increase in government spending
-25
-20
-15
-10
-5
0
5
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
yearGovt. spending increases by 10% $60 billion and 50,000 job loss
Chapter Five: Regional Impact of Catastrophes
220
5.8.3 Government spending effects (Fig. 5.14)
The government spending preference usually increases temporarily for the
affected region. This is modeled by increasing the value of government propensity to
spend, for some periods (usually two to three years) after the event. As in most of the
model parameters, almost no data is available for these temporary increases in spending
preferences. One way to understand the behavior of the model in absence of data is to use
reasonable bounds.
A 10 percent increase in governmental spending does not seem to affect the behavior of
the regional economy, though small improvements can be discerned.
5.8.4 Labor Supply effects (Fig. 5.15)
A decrease in employment as a result of catastrophe reduces the exports from the
local area. A decrease in employment decreases the change in the wage rate, which in
turn decreases the wage rate as compared to the no change in occupational-employment-
demand scenario. Decrease in the wage rate decreases the relative production costs and
relative sales price for regional industries. Decrease in wage rate reduces the labor and
proprietor’s income thus reducing the overall output, GRP.
When the occupational wage supply is decreased due to a catastrophe, the change in
wage rate relative to no-event scenario is smaller. This reduces the wage rate relative to
the no-event scenario and consequently reduces the labor and proprietors’ income.
Reduced wages also imply that the optimal capital stock reduces which may lead to lower
investment levels.
Chapter Five: Regional Impact of Catastrophes
221
Lower wages implies that the relative labor intensity average will be higher than the
baseline. As a result the labor-output ratio will increase prompting employment to
increase.
Increase in the changes in wages via a 10% increase in employment does not seem to
have much impact to the behavior.
Fig. 5.15 Effect on gross regional product due to 10% decrease in occupational employment
-25
-20
-15
-10
-5
0
5
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year10 % decrease in occupational employment $60 billion and 50,000 jobs loss
Fig. 5.16 Effect on gross regional product due to 1% increase in migration
-25
-20
-15
-10
-5
0
5
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
yearIncrease in migration 1% of population $60 billion and 50,000 jobs loss
Chapter Five: Regional Impact of Catastrophes
222
5.8.5 Migration effects (Fig. 5.16)
Immediately after a catastrophe, in-migration to the affected region decreases
partly because of lower relative employment opportunity (REO) in most sectors except
construction. The relative wage ratio (RWR) may also be smaller for the affected region
when compared to neighboring unaffected regions. Individuals are suddenly released
from liquidity constraint of selling their current homes and as jobs in national firms are
relocated outside the disaster region. This may induce out-migration. The effect of
reduced migration is a decrease in population when compared to a no-event scenario.
Smaller population implies lesser transfer payments, lesser dividends, interest, and rent,
and lesser governmental spending. But lesser (relative to baseline no-event scenario)
population results in lesser demand for housing, thus reducing relative housing price and
also a smaller consumer price deflator. A lower consumer price deflator reduces the wage
rate thus reducing the labor and proprietors’ income. Reduced wages also imply that the
optimal capital stock reduces which may lead to lower investment levels. Lower wages
implies that the relative labor intensity average will be higher than the baseline. As a
result the labor-output ratio will increase prompting employment to increase.
If it is assumed that one-percent of population migrates after the occurrence of an
earthquake, then there is a small decrease in gross regional product. This results from the
fact that as more people migrate, the population of the region decreases, which in turn
implies a lower income from dividends, interest, and rent, and government spending.
Chapter Five: Regional Impact of Catastrophes
223
5.8.6 Production or fuel costs (Fig. 5.17)
Damage to infrastructure, machinery, and buildings as a consequence of a
catastrophe may temporarily increase the production costs. Equipment and raw materials
may have to be brought from nearby unaffected regions. The effect of increase in
production costs is similar to a decrease in factor productivity, explained previously.
If the catastrophe causes extensive damage to capital stock, the factor productivity is
expected to drop, at least temporarily till the damaged machinery has been repaired and
the damaged buildings have been reconstructed. This drop in factor productivity may
result in increase of relative production costs. Increase in relative production costs
reduces the relative profitability for national industries thus lowering the attractiveness of
the region to new investment, at least temporarily. Drop in relative profitability reduces
the region’s share in satisfying the region’s demand, thus lowering the overall output.
A 10% market surge in prices causes the economy causes a further drop of 5 percentage
points in the GRP compared to no price-rise case. The market surge hurts the economy
since it takes a longer time to converge to the no-event scenario.
5.8.7 Business Taxes and credits (Fig. 5.18)
The effect of a decrease in business taxes and credits is to decrease the relative
production costs. The effect of decrease in production costs is similar to an increase in
factor productivity. An increase in factor productivity will encourage investors since the
relative profitability will increase. The region’s share in satisfying region’s demand will
increase, thus increasing the overall output. Thus decreasing business taxes in the
affected region may prove to be a good incentive to revive an economy. However, the
exact amount of credits that has to be given will depend on the severity of economic loss,
Chapter Five: Regional Impact of Catastrophes
224
the jobs lost, the extent to which public infrastructure in the affected region can support
new businesses.
Credits given to businesses immediately after a catastrophe, for example the SBA loans
prove very useful in bringing back the economy to the baseline. A 10 percent decrease in
business taxes or equivalently, 10 percent increase in business credits helps in even
surpass the baseline scenario within three years of the event.
5.8.8 Consumer Prices (Fig. 5.19)
It has been documented in literature (Brookshire, Thayer, Tshirhart, and Schulze,
1985) that houses sold for less in areas exposed to earthquake risk. Tobin and Montz
(1988, 1994) present empirical evidence that house values (median selling prices) fell
after the occurrence of floods. A fall in the housing price causes the relative price
changes to be smaller, thus reducing the wage rate. A fall in the wage rate reduces the
relative production costs and increases the relative profitability. The relative sales price
for regional industries drop. A fall in the wage rate causes a drop in labor and proprietors’
income and thus lowers the gross output.
Immediately preceding a hurricane-type event where there is sufficient warning and after
a catastrophic event costs of goods may increase temporarily anywhere from 10% to
30%. This decreases the consumer price deflator, which in turn increases the wage rate.
Increase in the wage rate raises the labor and proprietors’ income.
A 10 % increase in consumer prices has effects similar to a market surge in production
costs.
Chapter Five: Regional Impact of Catastrophes
225
Fig. 5.17 Effect on gross regional product due to 10% increase in relative production or fuel costs
-30
-25
-20
-15
-10
-5
0
5
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
yearIncrease in Rel. production costs by 10% $60 billion and 50,000 jobs loss
Fig. 5.18 Effect on gross regional product due to decrease in business taxes or tax credits
-25
-20
-15
-10
-5
0
5
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year10% decrease in business taxes or 10% increase in tax credits $60 billion and 50,000 jobs loss
Fig. 5.19 Effect of wage rate changes on gross regional product
-25
-20
-15
-10
-5
0
5
1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010
year1% decrease in wage rate $60 billion and 50,000 jobs loss 1% increase in wage rate
Chapter Five: Regional Impact of Catastrophes
226
5.9 Summary and Conclusions
Hurricane Andrew and Northridge earthquake caused almost equal amounts of direct
losses. But the effects were both similar and different. The effects were similar to the
extent that almost all sectors of the economy received a boost to their economic growth,
due to flow of external investment, reconstruction and changes in the underlying factors
of productivity. Replacement of old and damaged capital with new and efficient capital
enhances the productivity. This confirms the results of various numerical simulations
reported in Chapter 3. The effects were different in the sense that Northridge earthquake
occurred in a county with a robust economy as compared to Hurricane Andrew. The
effects of Loma Prieta and Northridge earthquakes were localized, whereas effects of
Hurricane Andrew were felt at the state level. Pre-event socioeconomic conditions
determine the post-event behavior – to repeat an observation from Chapter 2.
Data presented herein suggests that shock to a robust economy, such as California's
from a hazardous event of magnitude comparable to the Northridge earthquake, results in
localization of the effects. A higher magnitude of shock may propagate the effects to a
greater extent, but will be nevertheless localized. Hurricane Andrew was a greater shock
in terms of the loss-GSP ratio. But the propagation of the shock was felt at the state level,
though it dampened out rapidly both spatially and temporally.
A regional model was used to analyze the regional impacts of three major events
in the recent past – Loma Prieta earthquake, Hurricane Andrew, and Northridge
Earthquake. Model predictions for regional personal income matched with the actual
observed values within acceptable levels of error. The changes in consumption and
output as predicted from simulation of model used in this chapter concurred with the
trends predicted by the theoretical model presented in Section 3.3. The model was then
used to study the impacts of possible earthquake scenarios in the San Francisco – San
Jose CMSA.
Chapter Five: Regional Impact of Catastrophes
227
What are the factors that contribute towards dampening the propagation of
localized shocks caused by intense earthquakes or hurricanes to adjacent regions? Values
of the parameters of the model were varied exogenously to understand the working of the
model. These studies help one to design suitable policies for early recovery of a regional
economy. The main inferences from these numerical experiments were as follows:
1. The model predicts large gains if incentives are given to businesses in the affected
region.
2. Large increases in consumer prices and production and fuel costs after an event
delay recovery
3. Migration from the affected region delays its recovery
4. Governmental spending and transfer payments have a small positive effect on the
region’s recovery.
5. Increased consumer spending is an important factor driving the region’s economy
to rapid recovery.
The main conclusion that can be drawn from this study is that efficient recovery from a
catastrophic event is possible if the government introduces measures that check the short
term inflation of commodities, give incentives for the establishment and continuation of
investments, and encourage consumer spending (via tax rebates) in the disaster region.
Chapter Six: Future Work and Conclusions
228
Chapter Six
Conclusions and Future Work
________________________________________________________________________ 6. Introduction
The final chapter concludes with some final summary thoughts on the
contributions of this study. This chapter looks to the future through a discussion of some
natural extensions of the models and data described in the previous chapters. In particular
it proposes studies that link the results to financial and insurance markets. It proposes
refinements that might make it possible to model the post event economic behavior due to
man- made catastrophes.
6.1 Conclusions The research presented in this dissertation:
• connects socioeconomic indicators to determinants of vulnerability of a nation to
natural hazards,
• develops dynamic economic models for studying the economic behavior of
economies affected by catastrophes, and
• validates these models based on an empirical study at national and regional levels.
There were several contributions as a result of this study.
What are the determinants of the vulnerability of a region to natural hazards?
Corrupt and inefficient governments and bureaucracies, poor physical
infrastructure facilities, excessive dependence on imports, poor health infrastructure,
large uncertainties in the macroeconomic environment, and low levels of literacy are all
factors contributing towards the vulnerability. It is not surprising to note that these factors
also determine the per capita income of a nation, and a quantitative relationship exists
between the two. However, physical and human capital losses also depend on hazard
intensity.
Chapter Six: Future Work and Conclusions
229
What trends do past data on catastrophes suggest and can theoretical models replicate
these trends? Do catastrophes actually retard economic growth?
Data based on past catastrophes suggest a negative correlation of the loss with the post
event economic growth. A theoretical model was developed that explained this negative
correlation between the loss and the post event growth rate. This was achieved by
modeling the efficiency of post-event reconstruction. The observation that earthquakes
were associated positively with the post event growth rates were explained by the fact
that reconstruction of destroyed or damaged capital results in increases in productivity of
the region which in turn spurs the post-event economic growth. Empirical data also
suggest negative impacts on inflation, interest rates, and savings.
What quantity can be used as a measure of a catastrophe? How do we quantify the
secondary effects?
The annual economic loss as a percentage of GDP is extensively used in this study and
provided a robust indicator to study the impact of catastrophes. Catastrophe results in loss
of physical and human capital and this loss combined with the changes in the productivity
of the affected economy results in overall welfare losses. A measure of the secondary
effects of catastrophes based on these welfare losses was devised, which can be used to
assess to impact of a catastrophe on an economy.
How will a regional economy behave after a catastrophic event?
The theoretical models predicted the declines in income and consumption of the affected
region. Examination of three catastrophes – 1989 Loma Prieta earthquake, 1992
Hurricane Andrew, and 1994 Northridge earthquake, revealed similar results. Also, it was
shown that a low loss to output ratio results in localizing the effect of a catastrophe.
Hurricane Andrew’s effect was propagated to the state level, whereas Loma Prieta and
Northridge earthquakes were localized. The study revealed that a catastrophic earthquake
in San Francisco’s Silicon Valley causes the personal income and consumption of the
affected counties to drop. Depending on the external aid it receives, Silicon Valley could
Chapter Six: Future Work and Conclusions
230
fare better after the event. Since Silicon Valley’s interaction with the other regions is
high, economic effects of an earthquake will be felt in these regions. The overall welfare
losses to other regions are directly proportional to the direct losses caused by a
catastrophe.
What measures will best help the affected community to recover?
Theoretical model simulations reveal the importance of aid (in the form of investment)
the affected region receives from unaffected region. This is crucial for reconstructing lost
capital, thus, reviving the economy. The importance of incentives to business investments
in the affected regions is clearly demonstrated.
Why and under what conditions does the affected community fare better after an event?
How important and how long lasting are the various effects likely to be?
The affected region can easily recover within two years of the catastrophe, sometimes
even to better economic conditions as compared to the pre-event levels. This was clearly
demonstrated after the Northridge earthquake and Hurricane Andrew. The affected region
can fare better after an event if reconstruction results in permanent increase in the capital
share of the production function. This typically follows after productive capital in an
economy is destroyed and is replaced by new capital stock. Empirical evidence that
earthquakes are positively correlated with post-event growth rates clearly lends support to
the economic resurgence due to new capital influx in the affected region.
How closely are catastrophes and developmental process related?
Catastrophes reveal the most vulnerable sections of a socioeconomic fabric. Vulnerability
of a region to catastrophes is intimately related to the on going socioeconomic processes
including development. If the threat of occurrence of natural hazards is taken into account
while designing the development program of a region, then it may result in building of
robust engineering as well as social structures. Ex ante a catastrophe-threat induced
preparedness programs could result in many positive externalities. A well-planned
development strategy would not only result in raising the levels of livelihood of the
people concerned but also make them resilient towards the onslaughts natural hazards.
Chapter Six: Future Work and Conclusions
231
Ex post catastrophes can result in the building of a robust and less vulnerable region, if
appropriate measure are taken. The models and empirical data presented in the previous
chapters bring out the importance of pre-event conditions and efficiency of post-event
reconstruction in determining the evolution of an economy after an event. The occurrence
of a catastrophe gives the opportunity to invest, rebuild, and revitalize the economy of the
affected community. If this opportunity is seized, the affected community could emerge
better off than it was prior to the event.
6.2 Future work
Future work related to study of economic behavior after a catastrophe requires
two broad initiatives: relating the results of this study to financial markets and insurance
and to man-made catastrophes.
Since it has been shown in this dissertation that catastrophes have negative
impacts on such financial indicators as inflation, real interest rates, and more generally
economic growth, it will be interesting to examine the affect of catastrophes on financial
markets. Are catastrophes triggered by natural hazards and the financial markets really
not correlated? The results of such a study will be important in devising various financial
instruments including insurance and catastrophe bonds that will help in mitigating the
effects of catastrophes.
Man-made catastrophes are important especially after September 11, 2001
attacks. An obvious question to answer is how are the affects of man-made catastrophes
different from catastrophes triggered by natural hazards? Such a study will be useful in
identifying macroeconomic policies that need be enforced for efficient recovery.
References
232
References ________________________________________________________________________ Albala-Bertrand, J. M., (1993a) Natural Disaster Situations and Growth: A
Macroeconomic Model for Sudden Disaster Impacts, September 1993, World
Development, v21, n9, 0305-750X, 1417-34.
Albala-Bertrand, J. M., (1993b) Political economy of Large Natural Disasters - with
special reference to developing countries, Clarendon Press, Oxford, 1993
Anderson, Mary B. (1990), Analyzing the costs and benefits of natural disaster responses
in the context of development, The World Bank Policy Planning and Research
Staff, Environment Department.
Anderson, Mary B. and Woodrow, Peter J. (1989), Rising from the ashes: development
strategies in times of disaster, Boulder: Westview Press; Paris: UNESCO
Ayscue, J.K., (1996), Hurricane damage to residential structures: risk and mitigation,
Natural Hazards Research Working Paper #94, Institute of Behavioral Science,
University of Colorado, Web site:
http://www.colorado.edu/hazards/wp/wp94/wp94.html
Backus, David K., Kehoe Patrick J., and Kydland, Finn E. (1992) “International Real
Business Cycles,” Journal of Political Economy, 100, 745-775.
Barro, Robert J. (1990), Macroeconomics, 3rd ed. Imprint: New York : Wiley
Barro, Robert J. (1991) “Economic Growth in a Cross Section of Countries.” Quarterly
Journal of Economics 106:407 –43.
Barro, Robert J. (1997), Determinants of economic growth : a cross-country empirical
study Cambridge, Mass. : The MIT Press, c1997.
Barro, Robert J. and Xavier Sala-i-Martin. (1995), Economic growth, New York :
McGraw-Hill.
Bates, F.L., and Peacock, W.G., (1989) “Long term recovery”, International Journal of
Mass Emergencies and Disaster 7: 349-365.
Bates, F.L., and Peacock, W.G., (1993) Living Conditions, Disasters and Development -
An Approach to Cross-Cultural Comparisons, The University of Georgia Press,
Athens.
References
233
Behrman, J.R. and Deolalikar, A.B. (1990) “The intra-household demand for nutrients in
rural south India: Individual estimates, fixed effects and permanent income”,
Journal of Human Resources, 25(4):665-696.
Behrman, J.R. and Wolfe, B.L. (1984) “More evidence on nutrition demand; Income
seems overrated and women’s schooling under-emphasized”, Journal of
Development Economics, 14:105-128.
Behrman, Jere and Srinivasan, T.N., eds. (1995) “Handbook of Development
Economics.” Vol.3. Amsterdam: Elsevier, North Holland, 1995.
Behrman, Jere R. and Deolalikar, Anil B. (1987) “Will Developing Country Nutrition
Improve with Income?” Journal of Political Economy, June 1987, 95(3): 492-507.
Bernheim, D., Shleifer, A. and Summers, L. (1985) “The strategic bequest motive”,
Journal of Political Economy, 93(6):1045-1076.
Besley, Timothy. (1995) “Savings, Credit and Insurance.” Handbook of Development
Economics, 1995.
Bevan, D., Collier, P. and Gunning, J.W. (with Bigsten, A. and Horsnell, P.) (1989)
“Peasants and Government: An Economic Analysis.” Oxford: Clarendon Press.
Blaikie, P., Cannon, T., Davis, I., and Wisner, B. (1994) “At Risk - Natural Hazards,
Peoples Vulnerability, and Disasters,” Routledge, London.
Boarnet, M. G. (1995) "Business Losses, Transportation Damage, and the Northridge
Earthquake. Irvine, CA: Department of Urban and Regional Planning, School of
Social Ecology, University of California, Irvine.
Boisvert, R.N., Direct and Indirect Economic Losses from Lifeline Damage, FEMA,
Indirect Economic Consequences of a Catastrophic Earthquake, A study
Conducted by Development Technologies, Inc., National Earthquake Reduction
Program, July 1992.
Bolin, R. C. (1982) “Long Term Family Recovery from Disaster”, Boulder, University of
Colorado.
Bolin, R. with Stanford, L. (1998) “The Northridge Earthquake,” Routledge, London.
Brady, R.J. and Perkins, J.B., Macro-economic effects of the Loma Prieta earthquake,
Association of Bay Area Governments (ABAG), Nov. 1991.
References
234
Branford, S. and Kucinski, Bernardo, (1988), The debt squads : the US, the banks, and
Latin America / Imprint: London ; Atlantic Highlands, N.J. : Zed Books.
Brookshire, D.S., and McKee, M., (1992) Indirect Loss Measurement and Computable
General Equilibrium, FEMA, Indirect Economic Consequences of a Catastrophic
Earthquake, A study Conducted by Development Technologies, Inc., National
Earthquake Reduction Program, July 1992.
Brookshire, D.S., Chang S.E., Cochrane, H. Olson, R.A., Rose A., and Steenson, J.,
(1997) Direct and Indirect Economic Losses from Earthquake Damage,
Earthquake Spectra, 13(4), Nov. 1997, pp.683- 701.
Campbell, John Y. (1987) “Does saving anticipate declining labor income? An alternate
test of the permanent income hypothesis,” Econometrica, 55, 1249-73.
Cannon, T. (1994) “Vulnerability analysis and the explanation of “Natural” Disasters,” In
Varley, A. (ed.), 1994, Disasters, Development and Development, John Wiley,
Chichester.
Chang, S. (1984) “Do disaster areas benefit from disasters?” Growth and Change
Oct.:24-31.
Chiappori, P.-A. (1992) “Collective labor supply and welfare”, Journal of Political
Economy, 100(3):437-467.
Coate, Stephen and Ravallion, Martin. (1993) “Reciprocity Without Commitment:
Characterization and Performance of Informal Insurance Arrangements,” J.
Devel. Econ., Feb. 1993, 40(1), pp1-24.
Commander, Simon, Davoodi, Hamid R., and Lee Une J., (1997), The Causes of
Government and the Consequences for Growth and Well-Being, World Bank
Reading list, 2000/2001 World Development Report: Attacking Poverty.
Clague, Christopher, Philip Keefer, Stephen Knack and Mancur Olson (1996). “Property
and Contract Rights under Democracy and Dictatorship.” The Journal of
Economic Growth 1:2, 243-276 (June).
Cochrane, H., (1992) Is Northridge a model for future events? What are the economic
consequences with and without insurance and federal assistance? The Northridge
Earthquake: Analysing Economic Impacts and Recovery from Urban
References
235
Earthquakes: Issues for Policy makers, Proceedings of the Conference, Oct. 1996,
EERI and FEMA.
Cochrane, John H. (1991) “A Simple Test of Consumption Insurance”, Journal of
Political Economy, Vol. 99, Issue 5 (Oct. 1991), 957-976.
Coles, Jeffrey l. and Hammond, Peter J. (1995) “Walrasian Equilibrium Without
Survival: Existence, Efficiency and Remedial Policy,” in Kaushik Basu, Prasanta
Pattanaik, and Kotaro Suzumura, eds. 1995.
Cornia, G.A., Jolly, R. and Stewart, F., eds. (1987) “Adjustment with a human face,” Vol.
1, Oxford: Clarendon Press.
Cox, D. (1987) “Motives for private income transfers,” Journal of Political Economy,
95(3) 508-546.
Dacy, D.C. and Kunreuther, H., The Economics of Natural Disasters - Implications for
Federal Policy, The Free Press, 1969
Dando, W.A. (1980) “The geography of famine.” Edward Arnold, London.
Davis, L. (1971) “Capital Mobility and American growth”, In R.W. Fogel and S.L.
Engerman, (eds.), “The reinterpretation of American economic history.” New
York: Harper and Row.
Deaton, Angus. (1989) “Saving in Developing Countries: Theory and Review.” In
“Proceedings of the World Bank annual conference on development economics.”
Stanley Fischer and Dennis De Tray (Eds.), Washington, DC: World Bank, 1989,
pp. 611-96.
Deaton, Angus. (1992a) “Saving and Income smoothing in Cote d’Ivoire.” Journal of
African Economies, 1(1): 1-24.
Deaton, Angus. (1992b) “Household saving in LDCs: Credit markets, insurance and
welfare” Scandinavian Journal of Economics, 94(2): 253-273.
Deaton, Angus. (1992c) “Understanding Consumption,” Clarendon Press, Oxford.
Deolalikar, A.B. (1992) “Does the impact of government health expenditure on the
utilization of health services by children and on child health outcomes differ
across expenditure classes?,” mimeo, Department of Economics, University of
Washington. Quoted in Strauss and Thomas (1995).
References
236
Deininger, K., Squire, L. (1996). Measuring Income Inequality: A New Data Set. World
Bank Economic Review 10(3), 565-91
Development Technologies, Inc. (1992) Indirect Economic Consequences of a
catastrophic earthquake, National Earthquake Hazards Reduction Program,
NEHRP and FEMA.
Devereux, Stephen. (1993) “Theories of famine.” New York: Harvester Wheatsheaf,
1993.
Drabek, T.E., Key, W.H., Erickson, P.E. and Crowe, J.L. (1975) “The impact of disaster
on kin relationships,” Journal of Marriage and the Family, 481-494.
Dreze, Jean and Sen, Amartya, eds. “The Political economy of hunger” Vols. 1,2, and 3.
Oxford: Oxford U. Press, 1990.
Dreze, Jean and Sen, Amartya. (1989) “Hunger and public action.” Oxford: Oxford U.
Press, 1989.
Dufflo, Esther (1998) “Schooling and labor market consequences of school construction
in Indonesia: Evidence from an unusual policy experiment,” mimeo,
Massachusetts Institute of Technology, 1998.
Dynes, R. Russell. (1998) “Coming to terms with community disaster,” In “What is a
Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge,
London, 1998
Eguchi, R.T., Goltz, J.T, Taylor, C.E., Chang, S.E., Flores, P.J., Johnson,L.A., Seligson,
H.A., and Blais, N.C., The Northridge Earthquake as an Economic event: Direct
Capital Losses, Paper No.1, The Northridge Earthquake: Analysing Economic
Impacts and Recovery from Urban Earthquakes: Issues for Policy makers, Proc.
of the Conference, Oct. 1996, EERI and FEMA.
Elo, O. (1994) “Making disaster reduction a public policy.” STOP Disasters 17:14-15,
1994
EQE, Hurricanes Andrew and Iniki 1992 - An EQE Summary Report, 1992, Web site:
http://www.eqe.com/publications/iniki/
Easterly, William, and Rebelo, Sergio. (1993) “Fiscal Policy and Economic Growth: an
Empirical Investigation,” Journal of Monetary Economics December 1993, 32(3),
pp. 417-57.
References
237
Easterly, William and Levine, Ross. (1997) “Africa’s Growth Tragedy: Policies and
Ethnic Divisions,” Quarterly Journal of Economics, November 1997.
Eswaran, M. and Kotwal, A. (1989) “Credit as insurance in agrarian economies,” Journal
of Development Economics, 31(1): 37-53.
Fafchamps, M. (1993) “Sequential labor decisions under uncertainty: An estimable
household model of West African Farmers”, Econometrica, 61(5):1173-1198.
Fischer, Stanley, (1993) “The Role of Macroeconomic Factors in Growth,” Journal of
Monetary Economics, December, 32(3), pp. 485-511.
Flavin, Marjorie, (1981). “The adjustment of consumption to changing expectations
about future income,” Journal of Political Economy, 89, 974-1009.
Foster, Andrew D. (1995) “Prices, Credit Markets and Child Growth in Low-income
Rural Areas,” Economic Journal, May 1995, 105(430), pp. 551-70.
Friesema, H. Paul, [et al.]. (eds.) (1979), Aftermath : communities after natural disasters,
Beverly Hills : Sage Publications, c1979.
Gastil,R.D.(1975 –90,various issues) Freedom in the World. New York: Freedom House.
Gersovitz, Mark. (1983) “Savings and Nutrition at Low Incomes,” Journal of Political
Economy, October 1983, 91(5), pp.841-55.
Ghate, P. (1992) “Informal finance: Some findings from Asia.” Manila: Oxford
University Press for the Asian Development Bank.
Gilbert, Claude. (1998) “Studying disaster: Changes in the main conceptual tools,” In
“What is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.),
Routledge, London, 1998
Gordon, P. and H. W. Richardson, with B. Davis, C. Steins, and A. Vasishth. 1995. "The
Business Interruption Effects of the Northridge Earthquake." Los Angeles, CA:
Lusk Center Research Institute, School of Urban and Regional Planning,
University of Southern California.
Gordon, Peter and Richardson, H.W., Business Interruption Effects of a Major
Earthquake in the Newport/Inglewood Fault Zone, Submitted by the planning
Institute, School of Urban and Regional Planning, University of Southern
California, to the National Committee on Property Insurance, Boston,
Massachusetts, July, 1992
References
238
Gwartney, J.; Lawson,R.; and Block,W. (1997) Economic Freedom of the World,1975 –
1995.Vancouver:The Fraser Institute.
Haas, E.J., Kates, R.W., and Bowden, M.J., (1977) (eds.) Reconstruction following
Disaster, MIT Press, Cambridge, Massachusetts
Levine, R. and S. Zervos, 1998, "Stock Markets, Banks, and Economic Growth",
American Economic Review, vol 88, no 3, 537-558.
Hadfield, Peter. (1992) “Sixty Seconds that will change the World: The Coming Tokyo
Earthquake.” Boston: Charles E. Tuttle Company.
HAZUS (1997), Earthquake Loss Estimation Methodology, Technical Manuals (vols. I to
III), National Institute of Building Sciences, FEMA
Hewitt, Kenneth. (1998) “Excluded perspectives in the social construction of disaster,” In
“What is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.),
Routledge, London, 1998
Hill, K. and Palloni, A. (1992) “Demographic responses to economic shocks: The case of
Latin America”, mimeo, Department of Population Dynamics, School of Public
Health, Johns Hopkins University. Quoted by Strauss and Thomas (1995).
Hossain, S. (1989) “Effect of public programs on family size, child education and
health”, Journal of Development Economics, 30(1):145-158.
IFRCRCS (1993) World Disasters Report. Martinus Nijhoff, Dordrecht.
IFRCRCS (1994) World Disasters Report. Martinus Nijhoff, Dordrecht.
IFRCRCS (1999) World Disasters Report. Oxford University Press, Oxford.
INCEDE, 1995, The first 55 hours - Great Hanshin earthquake, January 17, 1995,
INCEDE newsletter Special Issue, January 1995, Institute of Industrial science,
The University of Tokyo.
Jacoby, Hannan and Skoufias, Emmanuel (1992) “Risk, Seasonality and School
Attendance: Evidence from Rural India.” Mimeo, University of Rochester, 1992.
Jones, E.L. (1987) “The European Miracle – Environments, economies and geopolitics in
the history of Europe and Asia.” Second Edition. Cambridge University Press,
Cambridge.
Kathleen J. T., Business Impacts Of The Northridge Earthquake, Disaster Research
Center, Department of Sociology and Criminal Justice, University of Delaware,
References
239
Newark, DE 19716 USA Paper prepared for Pan Pacific Hazards 96, Vancouver,
British Columbia, July 29-Aug. 2, 1996.
Kreimer, A. and Munasinghe, M. (eds.) (1991) “Managing natural Disasters and the
environment,” Environment Department, The World Bank, Washington, DC.
Knack, Stephen and Philip Keefer (1995). "Institutions and Economic Performance: Cross-
Country Tests Using Alternative Institutional Measures." Economics and Politics 7:3,
207-228 (November).
Knack, S. and Keefer, P., 1996. "Does social capital have an economic payoff?: a cross-
country investigation ", The Quarterly Journal of Economics, 112(4): 1251.
Kroll, C. A., J. D. Landis, Q. Shen, and S. Stryker. 1991. "Economic Impacts of the
Loma Prieta Earthquake: A Focus on Small Business." Berkeley, CA: University
of California--Berkeley, University of California Transportation Center and the
Center for Real Estate and Economics. Working Paper No. 91-187.
Leamer, E. Edward. (1983) Let ’s take the con out of econometrics.American Economic
Review 73,1:31 –43
Leamer, E. Edward. (1985) “Sensitivity Analyses Would Help.” American Economic
Review 75 (3):308 –13.
Lewis, Karen (1996) “What can explain the apparent lack of international consumption
risk sharing?” Journal of Political Economy, 104(2), 267-297.
Levine, Ross and David Renelt (1992). “A Sensitivity Analysis of Cross-Country Growth
Equations.” American Economic Review, September.
Levine, R. And S. Zervos, 1998, "Stock Markets, Banks, and Economic Growth",
American Economic Review, vol 88, no 3, 537-558.
Lipton, Michael and Ravallion, Martin. (1995) “Poverty and Policy” in Jere Behrman and
T.N. Srinivasan 1995, pp.2551-2657.
Lipton, Michael. (1977) “Why Poor People Stay Poor” Temple Smith, London, 1977
Quoted by Devereux (1993).
Mace, Barbara J. (1991) “Full Insurance in the Presence of Aggregate Uncertainty.”
Journal of Political Economy, Vol. 99, Issue 5 (Oct. 1991), 928-956.
References
240
Maskrey, A. (1994) “Disaster Mitigation as a Crisis of Paradigms: Restructuring after the
Alto Mayo Earthquake, Peru,” In Varley, A. (ed.), 1994, Disasters, Development
and Development, John Wiley, Chichester.
MATLAB compiler (2001): the language of technical computing: user's guide. Version 2.
Natick, MA: MathWorks, Inc., 2001.
Mauro, Paolo. (1995) “Corruption and Economic Growth,” Quarterly Journal of
Economics, August, 110(3), pp. 681-712.
McGregor, Pat. (1990) “A Model of Crisis in a Peasant Economy,” Oxford Economic
Papers, Oct. 1990, 4, pp.793-811.
Meier, Gerald M. (ed.) (1995) Leading issues in economic development, 6th ed., New
York, Oxford University Press, 1995.
Milliman, J.W. and Sanguinetty, J.A., Basic Concepts and Simple Analytic of Loss
Measurement for Major Earthquakes, in FEMA, Indirect Economic
Consequences of a Catastrophic Earthquake, A study Conducted by Development
Technologies, Inc., National Earthquake Reduction Program, July 1992.
Morduch, Jonathan (1994) “Poverty and Vulnerability” American Economic Review,
May 1994, 84(2), pp. 221-25.
Morduch, Jonathan. (1993) “Risk, Production and saving: Theory and Evidence from
Indian Households” Mimeo, Harvard U., 1993.
Morrow, B.H. (1997) “Stretching the Bonds: The Families of Andrew”, In Peacock,
W.G., Morrow, B.H., and Gladwin, H. “Hurricane Andrew – Ethnicity, gender
and the sociology of disasters”, Routledge, London, 141-169.
Munich Re (2000), Topics 2000, Natural Catastrophes: The current position.
Nigg, J.M. and Perry, R.W. (1988) “Influential first sources; Brief Statements with long-
term effects”, International Journal of Mass Emergencies and Disasters, 6(3), 311-
343.
NOAA, The costliest hurricanes in the United States - 1900-1996, Web site:
http://www.nhc.noaa.gov/pastcost.html/
Oliver-Smith, A. (1994) “Peru’s Five Hundred-Year Earthquake: Vulnerability in
Historical Context,” In “Disasters, Development and Environment,” A. Varley
(ed.), John Wiley, Chichester.
References
241
Oliver-Smith, Anthony. (1998) “Global changes and the definition of disaster.” In “What
is a Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge,
London, 1998
Olsen, R. and Wolpin, K. (1983) “The impact of exogenous mortality on fertility; A
waiting time regression with dynamic regressors”, Econometrica, 51(3):731-749.
Palloni, A. and Hill, K. (1992) “The effects of structural adjustment on mortality by age
and cause in Latin America”, mimeo, Center for Demography and Ecology,
University of Wisconsin. Quoted by Strauss and Thomas (1995).
Palloni, A. and Hill, K. and Pinto Aguirre, G. (1993) “Economic swings and
demographic changes in the history of Latin America”, mimeo, Center for
Demography and Ecology, University of Wisconsin. Quoted by Strauss and
Thomas (1995).
Paxson, Christina H. (1992) “Using Weather Variability to estimate the response of
savings to Transitory Income in Thailand.” American Economic Review, March
1992, 82(1), pp.15-33.
Peacock, W.G., Killian, C.D. and Bates, F.L. (1987) “The effects of disaster damage and
housing aid on household recovery following the 1976 Guatemalan earthquake,”
International Journal of Mass Emergencies and Disasters, 5, 63-88.
Peacock, W.G., Morrow, B.H., Gladwin, H. (eds.), (1997) “Hurricane Andrew -
Ethnicity, gender and the sociology of disasters,” Routledge, London.
Peek, Joe and Rosengren Eric S. (1997) “The International Transmission of Financial
Shocks: The Case of Japan” American Economic Review, Sep. 1997, pp. 495-
505.
The Penn World Table (PWT ver 6.1), Extension of - The Penn World Table (Mark 5):
An Expanded Set of International Comparisons, 1950- 1988," The Quarterly
Journal of Economics, May, 1991.
Pielke,Jr., R. A., and Landsea, C.W., Normalized hurricane damages in the United States:
1925-1995, Copyright 1998 American Meteorological Society, Web site:
http://www.dir.ucar.edu/esig/HP_roger/hurr_norm.html
Pielke,Jr., R. A., Trends in Extreme Weather Impacts, Web site:
http://ycymis.hq.nasa.gov/web_serv/files/southeast/
References
242
Porfiriev, B.N. (1992) “The environmental dimensions of national security: a test of
systems analysis methods.” Environmental Management 16: 735-742.
Quarantelli, E.L. (ed.) (1998) “What is a Disaster? Perspectives on the question,”
Routledge, London, 1998
Ravallion, M. (1987) “Markets and famines,” Oxford: Clarendon Press.
Ravallion, M. (1990) “Income effects on under-nutrition”, Economic Development and
Cultural Change, 38(3): 489-515.
Ravallion, Martin. (1997) “Famines and Economics”, Journal of Economic Literature,
Vol. XXXV (Sep. 1997), pp.1205-1242.
Razzaque, A., Alam, N., Wai, L. and Foster, A. (1990) “Sustained effects of the 1974-5
famine on infant and child mortality in a rural area of Bangladesh”, Population
Studies, 44(1): 145-154.
Ray, Debraj, (1998) Development economics, Princeton, N.J.: Princeton University
Press, c1998.
Red Cross, 1996, Hurricane Andrew results in one of the most comprehensive disaster
relief efforts in Red Cross history, Web site: http://www.redcross.org/hec/1980-
present/andrew.html/
RMS and FaAA, 1995, Japan - The great Hanshin Earthquake, Risk Management
Solutions, Inc., (RMS) and Failure Analysis Associates, Inc. (FaAA), 1995
RMS, Northridge earthquake January 17, 1994 - Event report, Risk Management
Software Inc., Menlo Park, California, 1994.
RMS, What if the 1906 Earthquake Strikes Again? A San Francisco Bay Area Scenario,
Topical issues series, Risk Management Solutions, Inc., CA, 1995.
Roberts, Blaine R., Milliman, Jerome W., and Ellson, Richard W. (1982) Earthquakes
and Earthquake Predictions: Simulating their Economic Effects, Technical
Report, University of South Carolina, Columbia, SC 29208
Rodrik, D. (1999) Where did all the growth go? External shocks, social conflict, and
growth collapses. Journal of Economic Growth 4(4), 385-412
Rosenthal, Uriel. (1998) “Future disasters, future definitions.” In “What is a Disaster?
Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge, London, 1998
References
243
Rosenzweig, Mark R. (1982) “Educational subsidy, agricultural development and fertility
change”, Quarterly Journal of Economics, 97(1) 67-88.
Rosenzweig, Mark R. (1988) “Risk, implicit contracts and the family in rural areas of low
income countries”, Economic Journal, 98(4): 1148-170.
Rosenzweig, Mark R. (1990) “Population growth and human capital investments: Theory
and evidence”, Journal of Political Economy, 98(5): S38-S70.
Rosenzweig, Mark R. and Binswanger, Hans P. (1993) “Wealth, Weather Risk and
Profitability of Agricultural Investments,” Economic Journal, Jan. 1993,
103(416), pp.56-78.
Rosenzweig, Mark R. and Evenson, R.E. (1977) “Fertility, schooling and the economic
contribution of children in rural India: An Econometric analysis”, Econometrica,
45(5):1065-1079.
Rosenzweig, Mark R. and Stark, O. (1989) “Consumption smoothing, migration and
marriage: Evidence from rural India”, Journal of Political Economy, 97(4): 905-
927.
Rosenzweig, Mark R. and Wolpin, Kenneth I. (1993) “Credit Market Constraints,
Consumption Smoothing and Accumulation of Durable Production Assets in
Low-Income Countries” Journal of Political Economy, April 1993, 101(2),
pp.223-44.
Sapir, D.G. and Misson, C. (1992) “The development of a database on disasters,”
Disasters 16: 74-80. (Center for Research on the Epidemiology of Disasters –
CRED)
Sen, Amartya. (1981) “Poverty and famines: An essay on entitlement and deprivation.”
Oxford: Oxford U. Press, 1981.
Smith, K. (1996) “Environmental Hazards - Assessing risk and reducing disaster,”
Second Ed., Routledge, London
Spurr, G.B. (1983) “Nutritional status and physical work capacity”, Yearbook of Physical
Anthropology, 26:1-35.
Stallings, A. Robert. (1998) “Disaster and the theory of social order,” In “What is a
Disaster? Perspectives on the question,” Quarantelli, E.L. (ed.), Routledge,
London, 1998
References
244
Strauss, J. and Thomas D.(1990) “The shape of the calorie expenditure curve”, Economic
Growth Center Discussion, Paper No. 595, Yale University.
Strauss, J. and Thomas, D. (1995) “Human Resources: Empirical modeling of household
and family decisions”, Chapter 34: In Behrman, Jere and Srinivasan, T.N., eds.
Handbook of Development Economics. Vol.3. Amsterdam: Elsevier, North
Holland, 1995.
Subramanian, S. and Deaton, A. (1992) “The demand for food and calories: Further
evidence from India”, mimeo, Woodrow Wilson School , Princeton university.
Susman, p., O”Keefe, P. and Wisner, B. (1983) “Global disasters: a radical
interpretation.” In Hewitt, K. (ed.) “Interpretations of Calamity,” pp. 263-283.
Allen and Unwin, Boston and London.
Thomas D., Lavy,V. and Strauss, J. (1992) “Public policy and anthropometric outcomes
in the Cote d”Ivoire”, Living Standards Measurement Study Working Paper no.
89, World Bank, Washington D.C.
Townsend, R. (1992) “Financial systems in Northern Thai Villages,” Typescript.
Townsend, Robert M. (1994) “Risk and Insurance in Village India.” Econometrica, May
1994, 62(3), pp. 539-91.
Treyz, George I., (1993) Regional economic modeling: a systematic approach to
economic forecasting and policy analysis, Boston : Kluwer Academic. George I.
Twigg, John. “Disasters, development and vulnerability.” In “Development at Risk?
Natural Disasters and the Third World,” The UK National Coordination
Committee for Natural Disaster Reduction (IDNDR), pp.2-7.
Tyler, C. (1990) “Earthquake Zones – Sites of the World’s Largest cities.” Geographical
Magazine 62, 3: 28-32.
Udry, Christopher R. (1990) “Credit markets in Northern Nigeria: Credit as insurance in
a rural economy”, World Bank Economic Review, 3:251-271.
Udry, Christopher R. (1994) “Risk and Insurance in a Rural Credit Market: An Empirical
Investigation in Northern Nigeria.” Review of Economic Studies, July 1994,
61(3), 495-526.
Udry, Christopher R. (1995),“Risk and saving in Northern Nigeria.” American Economic
Review, Dec. 1995, pp. 1287-1300.
References
245
Varley, A. (ed.), (1994), “Disasters, Development and Environment,” John Wiley,
Chichester.
West, C. and D. Lenze, 1994. Modeling the regional impact of natural disaster and
recovery: A general framework and an application of Hurricane Andrew.
International Regional Science Review, 17(2): 121-150.
Whitman, R.V., Anagnos, T, Kircher, C.A., Lagorio, H.J., Lawson, S., and Schneider, P.
Development of a National Loss Estimation Methodology, Earthquake Spectra,
Vol.13, No.4, Nov. 1997, pp. 643-661.
Wolfram, Stephen. (1999) The mathematica book, 4th ed. Champaign, IL : Wolfram
Media ; New York : Cambridge University Press.
Wolpin, K.I. (1982) “A new test of the permanent income hypothesis: The impact of
weather on the income and consumption of farm households in India”,
International Economic Review, 23(3): 583-594.
World development report, 1994: Infrastructure and Development. New York ; Oxford :
Oxford University Press, 1994.
World Development Indicators, 2001: World Bank.
World Disasters Report, (various years), International Federation of Red Cross and Red
Crescent Societies, Geneva.
Wright James D. [et al.]. (1979) After the clean-up: long range effects of natural
disasters, Beverly Hills: Sage Publications.
Yezer Anthony M. and Rubin Claire B. (1987) The local economic effects of natural
disasters, Imprint: [Boulder, Colo.]: Institute of Behavioral Science, University of
Colorado.
Zupka, D. (1988) “Economic impact of disasters.” UNDRO News Jan./Feb.: 19-22.