Natural Disasters and Migration

Ariel Belasen Department of Economics and Finance

Southern Illinois University Edwardsville, Illinois

abelase@siue.edu

Solomon W. Polachek Department of Economics

State University of New York at Binghamton Binghamton, New York

polachek@binghamton.edu

December 12, 2011

Preliminary Draft

Comments Welcome

 

 

1

Introduction

Climatic and weather-related events have long shaped migration patterns on Earth. From

tectonic shifts to meteors to hurricanes and tsunamis, life on Earth has long been subject to

unexpected changes in its surroundings and had to adapt. While earlier inhabitants of Earth,

such as the dinosaurs, were unable to cope with such natural disasters, more recently humanity

has found ways to withstand the elements via construction of solid shelters. But even some of

the most well-built cities have fallen victim to natural disasters forcing their populations to either

flee or suffer the consequences. Sometimes such disasters are caused by human activity as in the

string of ghost towns along the Aral Sea following the rerouting of waterways. Other times

unforeseen natural disasters can cause similar shifts in population, such as the combination of

Hurricanes Katrina and Rita in New Orleans, Louisiana in 2005 in which 300,000 people sought

refuge in nearby states.

In this chapter we examine the implications of natural disasters on communities and

differentiate between them based on magnitude, level of development for the impacted region,

and a number of additional factors. We present a critique of past approaches in examining the

impact of a disaster on migration and provide alternatives to the literature. Section I highlights

some of the historical situations involving natural disasters. Sections II through V examine the

literature regarding the economic impact of natural disasters on migration. And sections VI

through X provide an alternative approach to examining this impact as well as a sample

application of this approach. Section XI provides some concluding remarks.

 

 

2

I. Natural Disasters

Climate experts have recently predicted with 99% certainty that the number of weather-

related disasters will be increasing over the next few decades, forcing some regions of the world

to become “increasingly marginal as places to live” (Borenstein 2012). The basis for this claim

is rooted in history. Time and again, natural disasters have destroyed thriving communities

around the world. One of the earliest recorded casualties of a natural disaster is the ancient city

of Ephesus which is located in modern-day Turkey. Ephesus is known for having been the site

of one of the Seven Wonders of the World, the Temple of Artemis, and for later becoming the

second-largest city in the Roman Empire. While Ephesus grew in importance and stature in the

ancient world, the geography of the town began to shift over the centuries as silt deposits from

the Kucuk Menderes River (or Cayster River) began building up in the seas. What had once been

a magnificent port city became fully land-locked by the Byzantine Period and was soon

abandoned to Arab conquerors. The majority of the estimated 400,000-500,000 people of

Ephesus fled to nearby Byzantine strongholds such as Constantinople and, due to the diminished

capabilities of the port-less port city, never returned (Ephesus.us 2010). Today, silt from the

Cayster River has pushed the coastline more than five miles from the original Ephesus harbor.

Fifteen hundred years later, another city was abandoned north of the Black Sea from

Ephesus. The city of Prypiat, Ukraine was founded in 1970 as the ninth Atomograd, or nuclear

cities, of the Soviet Union. Prypiat was home to 50,000 residents, most of whom worked at or

were related to workers of the nearby Chernobyl Nuclear Power Plant. Following the nuclear

disaster on April 27, 1986, the city was hastily abandoned over the course of two days and

declared an abandoned zone. Today the city has no permanent inhabitants and the majority of

 

 

3

buildings are in ruins. Despite clean-up efforts, as of August 2011, the radiation levels in Prypriat

are more than three times higher than in nearby Chernobyl and more than seven times the safe

levels of the closest major city, Kiev (Pripyat.com 2011).

Similar incidents to these have happened in the United States in the 1970s. Prior to the

Chernobyl disaster and the 2011 Japanese earthquakes, the worst nuclear disaster took place at

the Three Mile Island nuclear facility in Pennsylvania in 1979. While nearby townships were

evacuated, 98% of residents returned to their homes within three weeks of the evacuation order

(Cutter and Barnes 1982). On the other hand, the residents of Centralia, Pennsylvania and Love

Canal, New York just north of Three Mile Island were not as fortunate.

Centralia was settled in 1841 as a coal mining town and became famous for being a

hotbed of bloody labor disputes at the end of the 19th Century between mine owners and the

semi-organized Irish workers known as the Molly Maguires. Nearly a century later, the coal was

depleted and the mines were turned into a landfill. In 1962, unbeknownst to residents, coal ashes

ignited a fire in the Centralia landfill which led to an amassing of carbon monoxide underground.

Between 1979 and 1982 several incidents related to the fire, including the appearance of a 150

foot sinkhole, alerted residents to the fire. Research indicated that the fire would burn for

upwards of 250 years leading the US Congress to begin relocating residents of Centralia to other

towns in the county in 1984. Today, only a handful of residents remain in defiance of the

government eviction orders (O’Carroll 2010).

Similarly, Love Canal, a white collar suburban neighborhood of Niagara Falls, New York

experienced a human-related disaster in the 1970s that caused residents to leave en masse. In

1953, the Niagara Falls School Board purchased the land of Love Canal from Hooker Chemical

for $1 with direct knowledge that it had been a toxic dump site for the past 12 years. While it is

 

 

4

somewhat common for toxic dumps to be reclaimed, the problem at Love Canal was that the

waste was buried too shallow so that by 1978 leakages led to countless health problems,

especially birth defects. President Jimmy Carter declared a federal health emergency on August

7, 1978 and the government began the process of relocating families from Love Canal. The 1979

EPA Report on Love Canal reported that 56% of families had at least one child who suffered

from birth defects. The neighborhood was evacuated and most of the residential buildings were

destroyed. The leakage in Love Canal and the subsequent dioxin leak in Times Beach, Missouri

in 1983 which also led to a full-scale evacuation were instrumental in the passage of the

Superfund Act which former EPA administrator Eckardt Beck stated would “defuse future Love

Canals” (Beck 1979).

More recently, severe weather catastrophes over the last few years have led to large-scale

evacuations in such places as New Orleans, Lousiana (hurricanes), Sumatra, Indonesia

(earthquake and tsunami), and Joplin, Missouri (tornados). But none of them faced the problems

of Fukushima, Japan. On March 11, 2011 a series of small earthquakes shook the Pacific Rim.

That afternoon, a major 8.9 earthquake (the most powerful to ever hit the region) struck off the

coast of Tohoku, Japan. The earthquake was so strong that Honshu Island (the main island of

Japan) shifted geographically by 8 feet. As bad as the earthquake was, the tsunami that followed

was even worse. The flooding from the tsunami touched off a nuclear event that was as bad as

the Chernobyl event in magnitude, and just as with Prypriat, an entire region of the country was

evacuated. Nuclear experts estimated that the tens of thousands of homes that were evacuated

would be unlivable for several thousand years, rendering Tohoku, Japan uninhabitable (Chico

2011).

 

 

5

II. Scientific Literature

The literature reveals a number of approaches to examining the impact of natural

disasters on migration. Among others, researchers have used OLS, difference-in-difference

(DD), difference-in-difference-in-differences (DDD), logistical regressions, instrumental

variable regressions, and even simple anecdotal statistical analyses. Regardless of the approach

used, the main finding is that disasters will always lead to temporary migration, and more often

than not, can also lead to permanent migration especially in developing regions of the world.

In the United States, natural disasters trigger interstate migration which leads to a long

run labor supply shock. For example, a number of studies examined the migration of workers

out of Louisiana and into nearby states following Hurricanes Katrina and Rita in 2006. The bulk

of the migrants headed across the border to Texas. Richard L. Clayton and James R. Spletzer

(2006) used a simple analysis of descriptive statistics to examine migration from Louisiana to

Texas. The study found that prior to Hurricane Katrina most migrants came to Texas for the

opportunity to receive wage gains. However, once the hurricane struck, the new wave of

migrants that left for Texas created a massive spike in labor supply that depressed the wages of

all the Louisiana migrants. Molly F. McIntosh (2008) used a DD framework to examine Current

Population Survey (CPS) data and found that Houstonian natives also experienced wage and

employment declines due to the wave of immigration from New Orleans following Katrina.

Makiko Hori, Mark J. Schafer, and David J. Bowman (2009) used data from the Louisiana

Health and Population Survey to differentiate between the intrastate and interstate migration of

New Orleans residents following Katrina and through an analysis of the descriptive statistics,

they found substantial differences in the likelihood that individuals would return to an area

 

 

6

following a hurricane. Additionally, Dakshina G. De Silva, Robert P. McComb, Yong-Kyu

Moh, Anita Schiller, and Andres J. Vargas (2010) studied labor market differences between

Houston and Dallas using a DDD approach. Using Quarterly Census of Employment and Wages

(QCEW) data, they found that disasters lead to long term permanent trends in migration, which

in turn leads to much stronger wage effects than previously thought due to substantial changes to

local labor supply.

Lisa K. Zotarelli (2008) provided anecdotal evidence for the De Silva, et al. (2010)

findings by showing that black residents of New Orleans that had been displaced by Hurricane

Katrina were less likely to find a job upon returning to New Orleans than those that never left.

Zotarelli’s (2008) findings were based on a logit analysis of Gallup Survey data collected in

2006 that examined the probability of return. The study concludes that workers who left after

the hurricane were more likely to stay in the city they took refuge in during the aftermath of the

storm if they have a better opportunity there. Jeffrey A. Groen and Anne E. Polivka (2010)

followed up Zotarelli’s (2008) findings with another logit analysis, showing that the individuals

who returned home after Katrina were predominantly older people who had less to gain from

leaving the city to begin with. Using CPS data, Groen and Polivka (2010) found that in the end,

less educated, low income, workers stayed away leaving New Orleans with a higher percentage

of high educated, high income, workers. James R. Elliott and Jeremy Pais (2010) point out that

this withdrawal is typically a feature of urban areas but not rural areas. They used descriptive

statistics to analyze differences among population density and found that in rural areas the poor

generally have less of a chance to migrate out and wind up bearing the brunt of the storm.

In developing countries, the poor face similar constraints. Without the resources to leave

following a disaster, many of the poor are forced to wait until conditions improve. Sally E.

 

 

7

Findley (1994) examined the descriptive statistics for a major drought in Mali and found that

despite occurring over a 3 year period, the drought did not significantly lead to long term

migration. Examining descriptive statistics from the Sahel Institute, Findley (1994) found that

the bulk of the migration was temporary and circular, with most people simply trying to wait out

the drought. Clark Gray and Valerie Mueller (2011) employed a multivariate event history

analysis to study another lengthy drought and found that in the Ethiopian highlands mobility

increased as the drought conditions worsened. Similar to Findley (1994), the study utilized local

data and found that migration during droughts primarily remained regional only, such that

displaced populations tended to remain close to their original location. The migratory behavior

tended to be nomadic in the sense that the drought caused people to change their daily routine

into a constant search for water. Such nomadic migration appears to be unique to the developing

world primarily because resources are too limited for people to outright move out of developing

regions.

In addition, Alessane Drabo and Linguere Mously Mbaye (2011) found that natural

disasters in developing countries affect highly-educated individuals as well, and in recent years

they have contributed heavily to the brain drain. Using fixed effects analysis for a panel of 88

countries, they investigated the relationship between net migration rates and natural disasters of

three types – meteorological (events caused by storms), hydrological (events caused by floods,

drought and wildfire) and climatological (events caused by extreme temperature) – which are

instigated by climate change, in developing countries. They found that only higher-educated

people will have the means to leave, and subsequently, without higher-skilled individuals, these

countries have a harder time dealing with the disasters and are thus even more susceptible to

emigration of high skilled workers to developed countries the next time a disaster strikes.

 

 

8

Other studies have found that natural disasters can disrupt long term trends in developing

countries and bring about unexpected out-migration of the poor as well. Adriana Kugler and

Mutlu Yuksel (2008) examined Hurricane Mitch’s impact on immigrants to the United States

and found more of a long term trend. Using OLS and IV methods to analyze US Census and

Community Survey data, Kugler and Yuksel (2008) showed that as immigrants entered into

specific regions of the United States in the 1980s, the native population moved away, thus when

the hurricane created another wave of immigration in the 1990s, the increase in labor supply

really only impacted the earlier group of immigrants. Therefore, the increase in low-skilled

emigration to the United States depressed the wages of previous migrants to the United States

and did not have a significant impact on natives. Denise L. Stanley (2010) also examined the

impact of Hurricane Mitch on Latin America by analyzing the descriptive statistics before and

after the hurricane. Specifically, she used data from the Honduran Population Census to look at

the impact on farmers in Honduras. She found that before the hurricane, there was regional out-

migration in shrimp farming towns and in-migration into melon farming towns as impoverished

rural residents sought out more consistent sources of income. However, once the hurricane hit

destroying small farms, this localized migration was replaced with more permanent international

migration. Similarly, Gordon H. Hanson and Craig McIntosh (2010) used OLS to analyze

census data from the United States, Canada, the United Kingdom, and Spain. Hanson and

McIntosh (2010) showed that labor opportunities drive international migration as well, in that

higher population growth in Latin America has led to waves of migration to more prosperous

nations. They found that when labor opportunities are destroyed by natural disasters in Latin

America, the bulk of emigrants leave for the United States more so than elsewhere.

Additionally, Sergio O. Saldana-Zorilla and Krister Sandberg (2009) utilized a spatial model to

 

 

9

find that weather-related disasters accounted for 80% of economic losses in Mexico particularly

in agricultural regions, in turn, leading to permanent mass-migration north into the United States.

Timothy J. Halliday (2006) reported similar findings using an ordered response model to

examine household-level panel data in El Salvador. Halliday (2006) found that earthquakes and

agricultural shocks increased the rate of migration of poor farmers to the United States. In

general, it appears that the ease of the potential move plays a role in determining whether out-

migration will occur following a disaster. Regions that experience relatively high annual out-

migration, such as Latin America and the Caribbean, will see out-migration numbers increase

significantly following a hurricane. On the other hand, more nomadic areas with relatively low

annual out-migration will continue to remain that way even after a natural disaster.

III. Methodological Problems

A number of methodological flaws possibly mar how these studies evaluate the impact of

natural disasters on population shifts. These methodological flaws range from small sample size

to poor or no selection of controls (see, for example, Stephane Hallegatte, Auguste Boissonnade,

Marc-Etienne Schlumberger, and Robert Muri-Wood (2008)), to a poor identification of the

treatment (see, for example, De Silva, et al. (2010)), and in the case of the descriptive statistical

analyses to potential for spurious relationships (see, for example, Elliott and Pais (2010)).

One of the most common issues is a lack of a legitimate set of controls which in turn precludes

the study from having a true baseline comparison. For example, Stephane Hallegatte, Auguste

Boissonnade, Marc-Etienne Schlumberger, and Robert Muri-Wood (2008) examine damage

claim forms following hurricanes and found that the construction sector soared due to a surge in

demand for reconstruction which, in turn, raised prices. They then remove the demand surge

 

 

10

(which they assume to be homogenous across the state) to assess losses at the pre-hurricane price

level. However, they fail to account for both the role of the housing bubble in raising

construction prices as well as the migration of people seeking warmer weather moving from the

Northeast down to Florida1 which would have been a major part of their demand surge. Thus,

without a real baseline comparison, Hallegatte, et al. (2008) cannot properly measure the true

impact of hurricanes on local demand-induced migration. Similarly, De Silva, et al. (2010) relies

on an assumption that Houston and Dallas are comparable cities with the only major difference

that occurred in the last decade stemming from the Katrina migrants. However, Dallas is a

wealthier city than Houston, with per capita income and housing prices more than 25% higher

than in Houston. Hence, a comparison examining how much lower wages are in Houston than

Dallas begins with the fundamental flaw that the two cities are not exactly comparable. Thus, De

Silva et al. (2010) comes up with results that are inconsistent with the earlier findings in

McIntosh (2008) which examines Houston by itself following the hurricane.

IV. Summary of Past Findings: Meta-Analysis

Because of potential biases in any one study, it makes sense to examine all studies in

their entirety to determine which conclusions generalize across studies. We do so within the

context of a meta-analysis. In general, these previous studies have come to the conclusions that:

(1) natural disasters will lead to migration in the short term and possibly the long term as well,

with the bulk of that outmigration occurring in developing countries; and (2) people living in

rural areas (especially in developing countries) will have less mobility than people in urban areas

                                                            1 Olivier Deschenes and Enrico Moretti (2009) showed that cold weather can drive people to warmer places in order to improve their long term health and life expectancy.

 

 

11

following a disaster. To verify these key results in the literature we turn to a meta-analysis as

well as an event study which will be discussed in the following section.

Table 1 provides detail on the main findings in 52 studies which examined natural

disasters and provided some evidence of an impact of disasters on migration.2 Authors are given

an identification number and in our analysis are then further sorted into two groups: economists

and other social scientists. Furthermore, we note the specific techniques used by each of the

studies: 2SLS, DD, DDD, Event Study, FE, GMM, IV, Logit, LPM, OLS, OR, Poisson, Probit,

Simplex, Spatial Analysis, Tobit, WLS or whether it was simply an analysis of descriptive

statistics. Next we gathered data on location and the type of disaster(s) examined in the studies

along with estimates of the damage cost in billions of dollars.3 Finally we looked at the details

on migration brought upon by the disaster. Those disasters that led to migration were noted both

through a simple dummy variable whether or not the disaster led to outmigration, as well as by

noting the rate of migration (defined as the number of migrants divided by the total number of

people affected). We have converted this data into a series of categorical variables that can be

used in a probit analysis to examine the likelihood of international migration which tests the first

of the two conclusions mentioned above. The data also enable us to examine whether or not the

results differ when written by economists compared to other social scientists, and whether they

differ based on incorporating econometrics rather than simple descriptive statistics. We run a set

of regressions to test these hypotheses using each study reported in Table 1 as the unit of

observation. One set of regressions examines the impact of disaster size measured in terms of

death and damage on international outmigration. The other is based on characteristics of the

                                                            2 We especially thank Marlon Tracey, Michael House, and Chris Pathmon for invaluable help in compiling the table. 3 Note that we also collected data on the number of disaster-related deaths and the total number of people affected by the disaster, however due to space constraints that data was left out of Table 1.

 

 

12

study, whether published by an economist and whether econometric analysis was used. Table 2

lists the results of our first analysis, namely whether or not more destructive disasters triggers

higher levels of migration as well as whether or not developing countries tend to have more

outmigration than developed countries. In order to maximize degrees of freedom, using the

meta-data in Table 1, we ran four individual probit estimations to assess the likelihood that a

disaster would lead to outmigration into a different country.4

The coefficients can be interpreted as the change in likelihood of international migration.

Thus, as expected, developing countries are 60% more likely than developed countries to have

international outmigration following a disaster. The number of deaths due to a disaster is

positive, though statistically insignificant in determining whether there will be international

outmigration. As for the damage factors, the negative coefficients appear to be driven by a major

outlier in that the damage to the United States during Hurricane Katrina was the most expensive

in the series; however, it did not lead to any outmigration. Thus we expand the model by

including a dummy variable that isolates Hurricane Katrina. Running the estimations once more

while accounting for the studies that examined Katrina by including a separate Hurricane Katrina

dummy variable, we find that the factors that measure damage lose their statistical significance.

Table 3 reports these findings.

Next, we created a dummy variable that identified which of the studies had been

published in economics journals to see if there were any differences in results. Table 4 compares

two more probit analyses, one that identifies the impact of economics publications and one that

identifies the impact of whether or not econometric analysis was used or if the study simply used

                                                            4 Unfortunately, due to missing data we could not run the meta-analysis using migration rate. However, the high correlation of 0.722 between the number of migrants and the dummy variable for international outmigration reveals that the results should be consistent between the two.

 

 

13

descriptive statistics. The results indicate that there is no statistical difference when comparing

results for papers published in economics journals and in other social science journals.

Furthermore, it does not appear that the results were significantly understated in studies that only

used descriptive statistics as their main analytical tool.

V. Summary of Past Findings: Event Study

Rather than examining each study as the unit of observation, we now in this section use

the natural disaster itself as the unit of observation. These 52 studies in Table 1 entail 23

individual natural disasters which we summarize in Table 5. We divide them into four groups:

droughts (D); hurricanes (H); water-related disasters including floods and tsunamis (W); and

land-related disasters including earthquakes, sinkholes, and tornadoes. In addition, we classify

locations which underwent multiple disasters (M = 1), and we take account of whether the

country is a developing nation (DEV = 1). Finally, we denote events resulting in migration for

which the number of migrants exceeds 10,000 in a given year (Migration = 1), and whether the

outmigration was international (INT = 1).

Once again we will use a probit estimation to assess the conclusions of the literature.

This time we focus on the second conclusion that rural areas are less likely than urban areas to

experience outmigration following a disaster. In addition, we examine type of disaster and we

distinguish between domestic and international migration. The results of these regressions can

be found in Table 6. Our analysis estimates the likelihood of the incidence of mass migration

(defined as more than 10,000 people leaving their country within the span of one year) and the

incidence of international migration. In this particular estimation, we examine whether or not the

disaster struck a rural region to see what, if any, difference there will be in the resulting level of

 

 

14

outmigration. Model 1 lists the results of this estimation indicating less out-migration from rural

areas. Model 2 adds type of disaster (hurricane, draught, and water-related disasters relative to

land-related events). None, with the exception of hurricanes, are statistically significant.

Due to the presence of Hurricane Katrina as an outlier in the data set, the hurricane

dummy variable yields a negative coefficient in Model 2. By including a dummy for Hurricane

Katrina in Model 3, the hurricane coefficient becomes positive and statistically significant.

Aside from that, once again the type of disaster does not appear to play a significant role in

determining the likelihood of a mass migration episode occurring, however again, multiple

disasters are statistically significant and the occurrence of multiple disasters at one event site

increases the likelihood of mass migration by as much as 51%. In support of the major

conclusions from the literature, we find that mass migration is far less likely to occur when

disasters impact rural areas (about 75% less than urban areas).

Finally, we examine how disasters relate to international migration rather than overall

migration level. To do so, we run two additional regression models, one using the international

migration dummy variable as the dependent variable and the second which features an

interaction term between international and the high-migration dummy variable as the dependent

factor. These results are summarized under Models 4 and 5 respectively. As can be seen, while

rural regions are less likely to feature high level migration, statistically speaking, individuals in

those regions are about as likely, on average, to migrate abroad as are their urban counterparts.

VI. Improving the Estimation

 

 

15

The common thread throughout most of the previous comparative studies is a failure to

account for overlaying trends in the data via a legitimate control set. In addition, DD studies, in

particular, have identification problems with regard to the treatment group versus the control

group as Joshua A. Angrist and Alan B. Krueger (1999) demonstrate; and other studies overlook

potential issues within the controls themselves.5 This results in untrustworthy coefficient

estimates due to biases in the estimates themselves as well as in their standard errors. We

propose to examine the mean difference between a treatment set and control set based upon a

number of individual events. Using a mean effect rather than a unique seemingly-exogenous

shock minimizes the possible bias resulting from potential unobservable heterogeneity between

the control and experimental groups. To that end, the Generalized Difference-in-Difference

(GDD) model improves upon the DD model by incorporating a multitude of events through

which the average exogenous effect can be felt (Belasen and Polachek (2008)).

The GDD model calls for a rotating pool of control groups and treatment groups such that

likelihood of identification problems resulting from nonrandom sampling is eliminated. Each

individual in that group can wind up as a member of the control group or the treatment group

depending on the impact of the event. So essentially, DD is a special case of GDD in which only

one event occurs at one given period of time. Of course, by generalizing the study via GDD, the

probability that unobserved heterogeneity has biased the results is minimized. Additionally, the

identification problems will also be minimized because the control and treatment groups will be

made up of the same individuals albeit at different points in time.

The GDD model is similar to the DD model in the sense that a variable is chosen which

takes the value of zero or one if a specific exogenous event occurs. We denote this as the                                                             5 See Marianne Bertrand, Esther Duflo, and Sendhil Mullainathan (2002) and Jeffrey D. Kubik and John R. Moran (2003) for a more complete discussion of endogeneity issues.

 

 

16

treatment which either occurs or does not occur. Thus our variable of interest, Y, is different

depending on the outcome of event A in time t. Let Y0 represent the value of Y if the event does

not impact individual c, and let Y1 be the value if the event does occur for individual c.

Following Angrist and Krueger’s (1999) specification,6 the conditional means take the following

form with c representing the locale of interest:

[ ] cttt AcYE γβ +=,|0 (1)

[ ] [ ] ][,|,| 01 δEAcYEAcYE tttt += (2)

δ is the value assigned to the exogenous shock when the two equations are differenced.

However, herein lies the difference between the two models. While the DD model assumes that

this incident is isolated for one specific locale c, the GDD model makes no such assumption,

such that c is a (1 x k) vector of which an event A can occur for any of the k number of locales

within that vector. At that specific period of time, t, in which A occurs, those locales affected by

A will take a value of one and all other locales will take a value of zero. Furthermore, while

event A is independent and unpredictable, there will be j such unique events such that A is a (j x

1) vector. Thus, at any moment of time, a specific locale within c may take a value of one or

zero depending on the outcome of A.

[ ] ktjtkt AcYE γβ +=,|0 (3)

[ ] [ ] [ ]jkEAcYEAcYE jtktjtkt ,|,|,| 01 δ+= (4)

                                                            6 See Angrist and Krueger (1999) equations (18) and (19).

 

 

17

As with the DD model, the equations are differenced across time and across locales to

differentiate the locales that were affected by event A from those who were not, however now

the GDD approach differs in that all of the locales that were unaffected will be used jointly to

provide a more-balanced control group:

[ ] [ ] *][)||( 01 δEAYEAYE jttjtt =−Δ (5)

The new estimate for the exogenous shock, δ*, is the difference between the mean value

of Y across time and between two sets of locales conditional on one set of locales affected by a

unique event in a given period of time and the other consisting of the mean value of those locales

unaffected by the event, given that several such events occur in the set of time t = 1, … , T.

Therefore, δ* effectively becomes the time and event averaged exogenous shock resulting from

the vector of events A.

VII. Analyzing the Impact of Hurricanes on Migration

In a previous study, Belasen and Polachek (2008), we found that hurricanes impact the

labor market. Using a GDD model to analyze data from the Quarterly Census of Employment

and Wages, we compared counties in Florida that were hit by hurricanes to counties that were

not hit. The GDD model enabled us to isolate the average impact of a hurricane on a county by

examining a series of 19 hurricanes that directly struck the State of Florida between 1988 and

2005. We found that counties that are directly hit by hurricanes will experience reductions in the

growth rate of employment and concomitantly will have increases in the growth rate of wages.

 

 

18

This impact diminishes further away from the center of the storm, in that if a hurricane directly

impacts a bordering county, the impact is lessened. For these counties, the change in the growth

rate of employment is not statistically significant, and the growth rate of wages will fall. Hence,

we hypothesize that at least a portion of the labor market effect of hurricanes is in fact due to an

unforeseen surge in migration out of afflicted counties. Thus when a hurricane hits a county, it

should reduce the population growth rate in a county as potential migrants choose to settle

elsewhere. To that end, we have taken a set of hurricanes and examined their impact on

population shifts in those same counties.

VIII. Methodology

In order to assess the impact of hurricanes on migration we will adapt the model used in

Belasen and Polachek (2008, 2009) in such a way to capture the population growth rate in place

of the growth rates of wages or employment. Following Belasen and Polachek (2008) equation

(5), the following equation isolates the impact of highly destructive hurricanes (i.e. hurricanes

with maximum wind speeds over 100 miles per hour) on the differences in population growth

rate for each county across time.7

itNit

Dititit uHHWP +++Δ+=Δ − 32110 %% αααα (6)

P represents the population in county i in time t, and ∆P represents the net migration into

and out of that county. W is the corresponding average wage for that county in the previous time

period. The two H variables represent the impact of hurricanes on counties, both as direct hits,

                                                            7 Note that this model was also run using weak hurricanes, but the results were not statistically significant.

 

 

19

D, and as indirect hits, N. Following Belasen and Polachek (2008, 2009), direct hits occur when

a county falls within a specific locus of destruction, whereas indirect hits represent strikes to

neighboring counties and hence are used to capture spillovers. In order to assess the exogenous

impact of the hurricanes on population growth, the GDD approach can be used to isolate the

average impact of a hurricane on a county that was hit relative to the average county that was

unaffected by the hurricane. Hence, the following model will be regressed:

Nit

Dittittit HHWWPP 32111 %(%)%(% γγγ ++Δ−Δ=Δ−Δ −− (7)

IX. Data

The hurricane data used in this study come from individual reports from the National

Hurricane Center of the National Oceanic and Atmospheric Administration (NOAA). NOAA

reports that most hurricanes that strike the United States strike the Gulf States and the

Southeastern States. Florida, as a member of both groups of states, was hit by 19 hurricanes in

the 18 year period of interest for this study. We focus on the seven most-destructive hurricanes

to Florida in that sample: Hurricanes Andrew, Opal, Charley, Ivan, Jeanne, Dennis, and Wilma.8

Wage data comes from the Bureau of Labor Statistics’ (BLS) Quarterly Census of Employment

and Wages (QCEW). The BLS surveys employers to gauge wages and employment by county.

The annual growth rate of wages for the average county in the sample was 6.36%, with a

relatively wide range running from a high annual rate of 36.75% down to a low contractionary

rate of -29.43%.

                                                            8 See Belasen and Polachek (2009) for a complete description of each of the hurricanes in the sample. Note that while Hurricanes Katrina and Rita were the most destructive hurricanes overall in the timeframe, they were not very strong when they swept through the Florida Keys.

 

 

20

The population data used in this study comes from the U.S. Census’s Population

Estimates by County (PEC). The Census estimates population levels for April 1 of each year

using the Census Survey and the American Community Survey (ACS). In order to accurately

gauge population changes, a cohort component method is used which breaks down population

changes into two main effects: a net birth rate effect as well as a migration effect. Migration is

further broken down into domestic (as measured by Internal Revenue Service (IRS) tax returns)

and international (measured by ACS reports). Finally, with respect to disaster-related migration,

the PEC updates the IRS data with specific data from the Federal Emergency Management

Administration (FEMA) on funding disbursements. Overall the average county experienced an

annual growth rate of 2.52% with a range between -4.70% to 17.79% growth.

Unfortunately, since the population data is only available on an annual basis, our results

will likely underreport the true impact of a disaster on migration by underemphasizing the short-

term migration. Furthermore, since hurricanes need warm surface water to form, hurricane

season runs from June 1 through November 30 of each year; and since the population data is

estimated for April 1, it will be necessary to lag the results such that we will be examining the

impact of hurricanes in time t-1 on a county’s population growth rate in time t. Therefore, while

we cannot examine short-term migration brought on by a hurricane, we will be able to identify

the impact of a hurricane on longer-term migration rates that persists for at least six months.

X. Empirical Results

 

 

21

A summary of the results of the regression analysis for equation (7) can be found in Table

7 below. As one would expect, the evidence of higher wage growth coincides with an increase in

the population growth rate for an individual county in the following period. For each one

percent increase in the wage growth rate, the population growth rate increased by just under 6%.

Additionally, hurricanes play a very significant role in the longer-term population growth rates

for a county that is directly hit by reducing growth rates by 74.8% on average. That indicates

there is legitimate evidence that a hurricane will impact net migration into a county. This is

contrasted with the counties bordering a county that was directly hit by a hurricane. Neighboring

counties will experience a 40.5% increase in their population growth rates – which may be

driven by people fleeing the directly impacted counties.

XI. Conclusions

Since the beginning of civilization, natural disasters and environmental degradation have

forced populations to relocate. The scientific literature points to two major conclusions

regarding the impact of disasters on migration. First, natural disasters will definitely lead to

short-term migration, and possibly long-term, as well. Second, populations in urban areas are

more likely to leave than their counterparts in rural areas. This latter conclusion stems from the

fact that urban dwellers will typically have a higher level of human capital and thus will have an

easier time of adjusting in a new location. This is particularly the case when populations migrate

from developing to developed countries.

We carried out a meta-analysis as well as an event study using 52 individual studies to

test these conclusions. Additionally, we provided an alternative approach to assessing the true

impact of a natural disaster on migration by utilizing the Generalized Difference-in-Difference

 

 

22

(GDD) model. Using GDD we were able to better isolate the exogenous implication of a series

of hurricanes on county-level populations in Florida. We found that population growth rates in

hurricane-battered counties can fall as much as 75% relative to unaffected counties.

But still there are numerous issues yet to be studied. Among the questions are: Who

among a population moves? Is it the young, the old, or the wealthy? Do families move? Do

they move together? What happens to them once they move? Does their economic position

rise? Do they become poorer? Do their children benefit? What about human capital transfers?

Do language barriers pose a problem? What about other capital lossess, such as farm

equipment? Do those emigrants stay forever, or is there return migration? Who returns and

why? Obviously the same questions regarding specific effects apply to return migrants. Next,

there are questions about who does not move. Who are the non-migrants? Are they the old, the

young, families, etc? What happens to their wealth, their health, and what happens to other

aspects of their well-being?

Given all these questions, we still need to zero in on aspects of natural disasters. While

there is some evidence on type and magnitude of disasters, we still need more precision,

particularly in terms of timing the short-run implications. We suggested some techniques such as

the GDD, but there are others that need be developed. So whereas there is clear evidence that

natural disasters do, in fact, provide stimulus for populations to move in particular when the

likelihood for improvement in quality of life is high, we nonetheless require more accuracy in

order to pinpoint when, where and the effects that occur when populations shift as a result of

natural disasters.

 

 

23

Table 1: Meta-Data

ID Author Technique

Natural Hazard

Damage Cost in Billions

Country/ region Origin

Destination of Migrants

Migration Rate

1 McIntosh

DD, WLS, Probit

Hurricane 125 USA Local

2 Clayton & Spletzer DS Hurricane 125 USA Local

3 Hori & Bowman DS Hurricane 125 USA Local

4 DeSilva, McComb et al DDD Hurricane 125 USA Local

5 Zotarelli Logit Hurricane 125 USA Local 20%

6 Groen & Polivka Logit Hurricane 125 USA Local 30%

7 Elliot & Pais OLS Hurricane 26.5 USA Local

8 Findley DS Drought Mali Both 30%

9 Gray & Mueller Event Drought Ethiopia Local 45%

10 Drabo & Mbaye FE Multiple Developing International

11 Kugler & Yuksel IV Hurricane 6.01

Central America

International

12 Hanson & McIntosh OLS Multiple 68.41 LAC International 4%

13 Saldana-Zo & Sandberg spatial Multiple 7.74 Mexico Local

14 Halliday OR Earthquake 1.8485 El Salvador International 26%

15 Hallegate & Boissonnade et al Simplex Hurricane 210.59 USA

16 Gray & Frankenberg et al Logit Tsunami 4.4516 Indonesia Local 19%

17 Ezra & kiros Event Drought Ethiopia Local 22%

18 Reuveny & Moore

Tobit, OLS, Robust

Multiple Developing International

19 Landry, Bin et al Logit Hurricane 125 USA Local

20 Barrios, Bertinelli et al FE Drought 3.18 SSA Local

21 Smith & McCarty DS Hurricane 26.5 USA Local 4%

22 Gray Event Drought 0 Ecuador Both 53%

23 Fussell & Elliot DD Hurricane 125 USA Local 35%

24 Vu, Vanlandingham,, et al Logit Hurricane 125 USA Local 35%

25 Mueller & Osgood FE Drought 2.95 Brazil

26 Pugatch & Yang IV Drought/Flood 3.14 Mexico International

27 Smith & McCarty Logit Hurricane 53.06 USA Local 25%

28 Afifi & Warner OLS Multiple World International

29 Kick et al OLS Hurricane 7 USA Local 22%

30 Tse LPM, FE Multiple 6.73 Indonesia Local 9%

 

 

24

31 Yang LPM, FE Earthquakes 1.8485 El Salvador Both 38%

32 Grote, Engel et al Logit Tsunami 1.32 Sri Lanka 24%

33 Paul DS Tornado Bangladesh None 0%

34 Edwards & Gray DS Drought Australia

35 Ezra DS Drought/Famine Ethiopia Both 34%

36 Deng Probit Multiple China Local 22%

37 Paxson & Rouse LPM Hurricane and flood 158.23 USA Local

38 Alexeev, Good et al FE Weather-related Disasters World International

39 Alexeev, Good et al FE Non-weather related World International

40 Henry, Schoumaker, et al Event Drought Burkina Faso Local

41 Henry, Schoumaker, et al Event Drought Burkina Faso International

42 Henry, Boyle et al Poisson Drought Burkina Faso Local 2%

43 Myers, slack et al OLS, GLS Hurricane 125 USA Local

44 Kirchberger

(Ordered) Probit, OLS Earthquake 8.95 Indonesia Both

45 Naude GMM Drought, disasters 10.91 SSA International

46 Marchiori, Maystadt et al FE, 2SLS Multiple 6.68 SSA Both

47 Geest DS Drought 0.00001 Ghana Both

48 Strobl & Valfort LPM, OLS Multiple 0.044

Senegal and Uganda Local

49 Badiani & safir FE-linear, FE-logit Rainfall shocks 11.5 India Local 17%

50 Munshi FE-OLS, IV

Drought conditions Mexico International 11%

51 Gray & Bilsborrow Event Drought Ecuador Local 18%

52 Gray & Bilsborrow Event Drought Ecuador International 18%

 

 

25

Table 2: Probit Estimation Results for the Likelihood of a Disaster Leading to International Outmigration Factor Coefficient Standard Error n R2 High Damage (dummy) -0.4477*** 0.1151 52 0.1884 ln (Damage Cost) -0.0653*** 0.0182 33 0.1983 Number of Deaths (1000s) 0.0007 0.0004 41 0.0557 Origin=Developing 0.6000*** 0.0844 52 0.3290 Notes: *Significant at the 10% level; **Significant at the 5% level; ***Significant at the 1% level

Table 3: Probit Estimation Results for the Likelihood of International Outmigration When Omitting Hurricane Katrina

Factor Coefficient Standard Error n R2 High Damage (dummy) 0.0678 0.2201 52 0.3024 ln (Damage Cost) -0.0129 0.0244 33 0.4141 Number of Deaths (1000s) 0.0002 0.0005 41 0.3020 Origin=Developing 0.6000*** 0.0853 52 0.3290 Notes: *Significant at the 10% level; **Significant at the 5% level; ***Significant at the 1% level

Table 4: Probit Estimation Results for the Likelihood of International Outmigration Factor Coefficient Standard Error n R2 Economics Journal (dummy) 0.0415 0.1417 52 0.0017 Econometric Analysis (dummy) 0.0341 0.1902 52 0.0006 Notes: *Significant at the 10% level; **Significant at the 5% level; ***Significant at the 1% level

 

 

26

Table 5: Event Data Origin Years Rural INT D H W L M DEV Migration Sources

Bangladesh 1970s-1980s 0 0 1 0 1 0 1 1 1 Baechler (1999); Hafiz and Islam (1993)

China 1980s-1990s 1 0 0 0 1 0 1 1 1 Baechler (1999); Brown et al., (1994); and Smil (1995)

Ecuador 1970s-1990s 1 0 1 0 0 0 1 1 0 Bilsborrow (2001); Pichon and Bilsborrow (1999); and UN (2001)

North Korea 1995-2000 1 1 1 0 0 0 0 1 0 Chu-Whan (1999); Lee (2001); and Yoon (1998)

Somalia 1980s-1990s 1 1 1 0 0 0 0 1 1 Cooper (2001); Kane (1995a,1995b); and Kibreab (1994)

Guatemala 1950s-1980s 1 1 0 0 1 0 0 1 0

Bilsborrow and DeLargy (1990); Sader, Reining, Sever, and Soza (1997); and UN (2001)

Dominican Republic 1940s-1980s 1 0 0 0 0 1 0 1 0

Bilsborrow (2001); UN (2001); and Zweifler, Gold, and Thomas (1994)

Canada 1931-1939 1 0 1 0 0 0 0 0 0 IISD/EARG (1997); Lockeretz (1978); and Rosenzweig and Hillel (1993)

Mexico 1970s-2000s 0 1 1 0 0 0 0 1 1 Arizpe (1981); Liverman (2001); NHI (1997); and Schwartz and Notini (1994)

Kenya 1960s-1990s 1 0 1 0 0 0 0 1 0 Dietz (1986); Gould (1994); and IOM (1996)

Uzbekistan 1970s-2000s 1 0 0 0 0 1 0 1 0

Shestakov and Streletsky (1998); Small, van der Meer, and Upshur (2001); and UN (2001)

Caspian Sea Region 1990s 1 1 0 0 1 0 0 1 0

Chuykov (1995); Shestakov and Streletsky (1998)

Russia 1980s-1990s 1 0 0 0 0 1 0 1 1 Kane (1995b); Specter (1994)

Burkina Faso 1960s-2000s 1 0 1 0 0 0 0 1 0

Binama (1996); Cordell, Gregory, and Piche (1996); Henry, Schoumaker, Beauchemin, and Dabire (2003)

India 1978-1983 1 0 1 0 0 0 0 1 0 Jacobson (1989)

Zimbabwe 1980s-2010s 1 0 1 0 0 0 0 1 0 Lonergan (1998); Scoones (1992)

Thailand 1980s-1990s 1 0 0 0 0 1 0 1 0

Bilsborrow (2001); Cropper, Griffiths, and Mani (1997); Panayotou and Sungsuwan (1994)

Russia 1990s 1 1 0 0 0 1 0 1 1 Shestakov and Streletsky (1998)

Tanzania 1950s-1990s 1 0 0 0 0 1 0 1 0 Charnley (1997); Mwakipesile (1976); Odgaard (1986)

USA 2005 0 0 0 1 1 0 1 0 0

McIntosh (2008); Clayton and Spletzer (2006); Hori et al. (2009); DeSilva et al. (2010); Zotarelli (2008); Groen and Polivka (2008)

Ethiopia 1983-1985 1 0 1 0 0 0 0 1 0 Findley (1994); Gray and Mueller (2011)

Honduras 1994 1 1 0 1 0 0 0 1 1 Kugler and Yuksel (2008); Stanley (2010)

Mexico 1990s 1 1 1 0 0 1 1 1 1 Saldana-Zorilla and Sandberg (2009)

 

 

27

Table 6: Probit Estimation Results for the Impact of Disasters on Mass Migration Coefficient: Model 1 Model 2 Model 3 Model 4 (INT) Model 5 (>10K) Rural

Coefficient: -0.3667 -0.6238** -0.7551* -0.1182 -0.4564 RSE: (0.3043) (0.2476) (0.3731) (0.5133) (0.3681)

Hurricanes Coefficient: -0.4379** 1.2737*** 0.7455*** 0.7709*** RSE: (0.1924) (0.4177) (0.1858) (0.1785)

Drought Coefficient: -0.1963 -0.2469 0.1000 -0.0600 RSE: (0.2679) (0.2818) (0.2609) (0.2258)

Flood Coefficient: -0.2510 -0.2377 0.3273 -0.4055 RSE: (0.2664) (0.2717) (0.3728) (0.2554)

Multiple Coefficient: 0.4454 0.5092* -0.2727 0.1545 RSE: (0.3215) (0.2854) (0.4227) (0.3210)

Katrina Dummy Coefficient: -0.9177*** -1.1727** -1.2054*** RSE: (0.3895) (0.5280) (0.3757)

R2 0.0672 .3570 .3977 .1723 .3221

F 1.45 5.44 1.41 0.56 1.27 n 23 23 23 23 23 Notes: *Significant at the 10% level; **Significant at the 5% level; ***Significant at the 1% level

Table 7: GDD Estimation Results for the Population Growth Rate in the Average County in Florida Coefficient: Wage Growth Rate

Coefficient: 0.0581*** RSE: (0.0122)

Direct Impact of Hurricanes Coefficient: -0.7476* RSE: (0.4148)

Neighboring Impact of Hurricanes Coefficient: 0.4050** RSE: (0.1557)

R2 .0977

F 7.80 n, groups 1135, 67 Notes: *Significant at the 10% level; **Significant at the 5% level; ***Significant at the 1% level

 

 

28

References Afifi, Tamer and Koko Warner. 2008. “The Impact of Environmental Degradation on Migration

Flows across Countries,” UNU-EHS Working Paper 5/2008. Alexeev, Alexander, David H. Good, and Rafael Reuveny. 2011. “Weather Related Disasters and

International Migration,” working paper. Angrist, Joshua A. and Alan B. Krueger. 1999. “Empirical Strategies in Labor Economics,” in

Orley C. Ashenfelter and David A. Card, editors, Handbook of Labor Economics, 3A, Amsterdam: Elsevier, 1277-1366.

Arizpe, L. 1981. “The rural exodus in Mexico and Mexican migration to the United States,”

International Migration Review, 15(4): 626-649. Badiani, Reena and Abla Safir. 2008. “Coping with Aggregate Shocks: Temporary Migration

and Other Labor Responses to Climatic Shocks in Rural India,” working paper. Baechler, Gunther. 1999. Violence through environmental discrimination. Dordrecht,

Netherlands: Kluwe. Barrios, Salvador and Luisito Bertinelli. 2006. “Climatic Change and Rural-Urban Migration:

The Case of Sub-Saharan Africa,” Journal of Urban Economics, 60(3): 357-371. Beck, Eckardt. 1979. “The Love Canal Tragedy,” EPA Journal, January 1979. Belasen, Ariel, and Solomon W. Polachek. 2008. “How Hurricanes Affect Wages and

Employment in Local Labor Markets,” American Economic Review, 98(2): 49-53. _____. 2009. “How Disasters Affect Local Labor Markets: The Effects of Hurricanes in

Florida,” Journal of Human Resources, 41(1): 251-276. Bertrand, Marianne, Esther Duflo, and Sendhil Mullainathan. 2002. “How Much Should We

Trust Differences-in-Differences Estimates?,” The Quarterly Journal of Economics, 119(1): 249-275.

Bilsborrow, R. E. 2001. “Migration, population change, and the rural environment,” Population, environmental change, and security paper series, 2. Ann Arbor: University of Michigan.

Bilsborrow, R. E., and DeLargy, P. F. 1990. “Land use, migration, and natural resource deterioration: the experience of Guatemala and the Sudan,” Population and Development Review, 16(2): 125-147.

Binama, E. 1996. The relationship between environmentally-induced population displacements and environmental impacts. Paper presented to the International Organization for

 

 

29

Migration conference on environmentally-induced population displacements and environmental impacts, Geneva.

Borenstein, Seth. 2012. “Future Holds More Extreme Weather,” The Associated Press, November 1, 2011. http://hosted.ap.org/dynamic/stories/U/US_SCI_CLIMATE_ EXTREMES?SITE=ORBEN&SECTION=HOME&TEMPLATE=DEFAULT

Brown, L., Kane, H., and Roodman, D. N. 1994. Vital signs 1994. New York: W.W. Norton Charnley, S. 1997. “Environmentally-displaced peoples and the cascade effect: lessons from

Tanzania,” Human Ecology, 25(4): 593-618. Chico, Harlan. 2011. “Japan Quake: With Two Natural Disasters and a Nuclear Emergency,

Recovery Begins,” Washington Post, http://www.washingtonpost.com/wp-dyn/content /article/2011/03/12/AR2011031201452.html?hpid%3Dtopnews, March 12, 2011.

Chu-Whan, S. 1999. North Korean escapees are treaty-defined refugees. International

Conference of NGOs, Seoul. Chuykov, Y. S. 1995. “Ecological problems of the Northern Caspian region and Caspian Sea,”

International Conference on the Caspian Sea Problems, 1995 Proceedings, Astrakhan. Clayton, Richard L. and James R. Spletzer. 2006. “Worker Mobility before and after Hurricane

Katrina,” Montly Labor Review, 129(8): 11-21. Cooper, M. H. 2001. Global refugee crisis. Global issues. Washington, DC: CQ Press. Cordell, D., Gregory, J., and Piche, V. 1996. Hoe and wage: A social history of a circular

migration system in West Africa. Boulder, CO: Westview. Cropper, M. L., Griffiths, C., and Mani, M. 1997. “Roads, population pressures, and

deforestation in Thailand, 1976-89,” Policy Paper 1726. Washington, DC: World Bank. Cutter, Susan and Kent Barnes. 1982. “Evacuation Behavior and Three Mile Island,” Disasters,

6(2): 116-124. De Silva, Dakshina G., Robert P. McComb, Young-Kyu Moh, Anita Schiller, and Andres J.

Vargas. 2010. “The Effect of Migration on Wages: Evidence from a Natural Experiment,” American Economic Review, 100(2): 321-326.

Deng, Wei, Baochen Shi, Xiaoli He, Zhihua Zheng, Jun Xu, Biao Li, Jian Yang, Lunjiang Ling,

Chengping Dai, and Boqin Qiang. 2004. “Evolution and Migration History of the Chinese Population Inferred from Chinese Y-Chromosome Evidence,” Journal of Human Genetics, 49(7): 339-348.

 

 

30

Deschenes, Olivier and Enrico Moretti. 2009. “Extreme Weather Events, Mortality, and Migration,” The Review of Economics and Statistics, 91(4): 659-681.

Dietz, T. 1986. “Migration to and from dry areas in Kenya,” Tijdschrift voor Economische en

Sociale Geografie, 77(1): 18-26. Drabo, Alassane and Linguere M. Mbaye. 2011. “Climate Change, Natural Disasters and

Migration: An Empirical Analysis in Developing Countries,” Institute for the Study of Labor, Discussion Paper No. 5927.

Edwards, Ben, Matthew Gray, and Boyd Hunter. 2009. “A Sunburnt Country: The Economic and

Financial Impact of Drought on Rural and Regional Families in Australia in an Era of Climatic Change,” Australian Journal of Labor, 12(1): 108-131.

Elliott, James R. and Jeremy Pais. 2010. “When Nature Pushes Back: Environmental Impact and

the Spatial Redistribution of Socially Vulnerable Populations,” Social Science Quarterly, 91(5): 1187-1202.

Ephesus.us. 2010. “Ephesus! History, Information, and Pictures of Ephesus Ancient City,”

http://www.ephesus.us. Ezra, Markos. 2001. “Demographic Responses to Environmental Stress in the Drought- and

Famine-Prone Areas of Northern Ethiopia,” International Journal of Population Geography, 7(4): 259-279.

Ezra, Markos and Gebre-Egziabher Kiros. 2001. “Rural Out-migration in the Drought Prone

Areas of Ethiopia: A Multi-Level Analysis,” International Migration Review, 35(3): 749- 771.

Findley, Sally E. 1994. “Does Drought Increase Migration? A Study of Migration from Rural

Mali during the 1983-1985 Drought,” International Migration Review, 28(3): 539-553. Fussell, Elizabeth and James R. Elliott. 2009. “Introduction: Social Organization of

Demographic Responses to Disaster: Studying Population-Environment Interactions in the Case of Hurricane Katrina,” Organization Environment, 22(4): 379-394.

Gould, W. 1994. “Population growth, environmental stability and migration in Western Kenya,”

In B. Zaba, & J. Clarke ed., Environment and population change, 247-268. Liege, Belgium: Ordina.

Gray, Clark L. 2010. “Gender, Natural Capital, and Migration in the Southern Ecuadorian

Andes,” Environment and Planning A, 42(3): 678-696. Gray, Clark L. and Valerie Mueller. 2011. “Drought and Population Mobility in Rural Ethiopia,”

World Development, forthcoming.

 

 

31

Groen, Jeffrey A. and Anne E. Polivka. 2010. “Going Home After Hurricane Katrina:

Determinants of Return Migration and Changes in Affected Areas,” Demography, 47(4): 821-844.

Grote, Ulrike, Stefanie Engel, and Benjamin Schravin. 2006. “Migration Due to the Tsunami in

Sri Lanka: Analyzing Vulnerability and Migration at the Household Level,” ZEF discussion paper on development policy 106.

Hafiz, F. A., & Islam, N. 1993. Environmental degradation and intra/interstate conflicts in

Bangladesh. Environmental and conflicts project e Occasional Papers, No. 6. Zurich: Center for Security Studies and Conflict Research.

Hallegatte, Stephane, Auguste Boissonnade, Marc-Etienne Sclumberger, and Robert Muir-

Wood. 2008. “Demand Surge and Worker Migration in Disaster Aftermaths: Application to Florida in 2004 and 2005,” Cente International de Recherche sur l’Environnement et le Developpement working paper.

Halliday, Timothy J. 2006. “Migration, Risk, and Liquidity Constraints in El Salvador,”

Economic Development and Cultural Change, 54(4): 893-925. Hanson, Gordon H. and Craig McIntosh. 2010. “Birth Rates and Border Crossings: Latin

American Migration to the US, Canada, Spain, and the UK,” NBER Working Paper No. 16471.

Henry, Sabine, Paul Boyle, and Eric F. Lambin. 2003. “Modelling Inter-Provincial Migration in

Burkina Faso, West Africa: The Role of Socio-Demographic and Environmental Factors,” Applied Geography, 23(2-3): 115-136.

Henry, Sabine, Bruno Schoumaker, and Cris Beauchemin. 2004. “The Impact of Rainfall on the First Out-Migration: A Multi-Level Event-History Analysis in Burkina Faso,” Population & Environment, 25(5): 423-460.

Hori, Makiko, Mark J. Schafer, and David J. Bowman. 2009. “Displacement Dynamics in

Southern Louisiana After Hurricanes Katrina and Rita,” Population Research and Policy Review, 28(1): 45-65.

IISD/EARG. 1997. Agriculture and climate change: A prairie perspective. Toronto:

International Institute for Sustainable Development. IOM. 1996. Environmentally-induced population displacements and environmental impacts

resulting from mass migrations. Geneva: International Organization for Migration Jacobson, J. L. 1989. “Environmental refugees: nature’s warning system,” Populi, 16(1): 29-35.

 

 

32

Kane, H. 1995a. “Leaving home,” In L. R. Brown, et al. ed., State of the world 1995, 132-149. New York: W.W. Norton.

_____. 1995b. The hour of departure: Forces that create refugees and migrants. Paper 125.

Washington, DC: Worldwatch. Kibreab, G. 1994. Migration, environment and refugeehood. In B. Zaba, & J. Clarke ed.,

Environment and population change, 115-130. Liege, Belgium: Ordina. Kick, Edward L., James C. Fraser, Gregory M. Fulkerson, Laura A. McKinney, and Daniel H.

De Vries. 2011. “Repetitive Flood Victims and Acceptance of FEMA Mitigation Offers: An Analysis with Community-System Policy Implications,” Disasters, 35(3): 510-539.

Kirchberger, Martina. 2010. “Distributional Effects of Natural Disasters: Who Are the Winners

and Losers?,” working paper. Kubik, Jeffrey D. and John R. Moran. 2003. “Can Policy Changes be Treated as Natural

Experiments? Evidence from Cigarette Excise Taxes,” The Journal of Law and Economics, 46: 1-26.

Kugler, Adriana and Mutlu Yuksel. 2008. “Effects of Low-Skilled Immigration on US Natives: Evidence from Hurricane Mitch,” Institute for the Study of Labor, Discussion Paper No. 3670.

Landry, Craig E. Okmyung Bin, Paul Hindsley, John C. Whitehead, and Kenneth Wilson. 2007.

“Going Home: Evacuation-Migration Decisions of Hurricane Katrina Survivors,” Southern Economic Journal, 74(2): 326-343.

Lee, S. W. 2001. Environmental matters: Conflict refugees and international relations. Tokyo:

World Human Development Institute Press. Liverman, D. M. 2001. “Vulnerability to drought and climate change in Mexico,” In J. X.

Kasperson, & R. E. Kasperson ed., Global environmental risk, 201-216. New York: UN University Press and Earthscan.

Lockeretz, W. 1978. “The lessons of the Dust Bowl,” American Scientist, 66(5): 560-569. Lonergan, S. 1998. “The role of environmental degradation in population displacement.

Environmental change and security project report,” research report 1. Toronto: University of Toronto.

Marchiori, L., J-F. Maystadt, and I. Schumaker. 2011. “The Impact of Weather Anomalies on

Migration in Sub-Saharan Africa,” Institut de Recherches Economiques et Sociales de l’Universite Catholique de Louvain discussion paper 2011-34.

 

 

33

McIntosh, Molly F. 2008. “Measuring the Labour Market Impacts of Hurricane Katrina Migration: Evidence from Houston, Texas,” American Economic Review, 98(2): 54-57.

Mueller, Valerie A. and Daniel E. Osgood. 2009. “Long-term Impacts of Drought on Labor

Markets in Developing Countries: Evidence from Brazil,” Journal of Development Studies, 45(10): 1651-1662.

Munshi, Kaivan. 2001. “The Identification of Network Effects: Mexican Migrants in the US

Labor Market,” working paper. Mwakipesile, J. S. 1976. “Peasants and migrants: a case study from Usangu Plains.” MA thesis,

Sociology. Tanzania: University of Dar es Salaam.

Myers, Candice A., Tim Slack, and Joachim Singelmann. 2008. “Social Vulnerability and Migration in the Wake of Disaster: The Case of Hurricanes Katrina and Rita,” Population & Environment, 29(6): 271-291.

Naude, Wim. 2008. “Conflict, Disasters, and No Jobs: Reasons for International Migration from

Sub-Saharan Africa,” UNU-Wider working paper RP2008/85. NHI. (1997). Breaking the cycle: Desertification and migration on the US-Mexico border. San

Francisco: National Heritage Institute. O’Carroll, Eoin. 2010. “Centralia , PA: How an Underground Coal Fire Erased a Town,”

Christian Science Monitor, http://www.csmonitor.com/Environment/Bright-Green/ 2010/0205/Centralia-Pa.-How-an-underground-coal-fire-erased-a-town, February 5, 2010.

Odgaard, R. 1986. “Tea - does it do the peasant women in Rungwe any good?” In J. Boesen, K.

J. Havnevik, J. Koponen, & R. Odgaard ed., Tanzania - Crisis and struggle for survival, 207-224. Uppsala: Africana.

Panayotou, T., and Sungsuwan, S. 1994. “An econometric analysis of the causes of tropical

deforestation: the case of northeast Thailand,” In K. Brown, & D. W. Pearce ed., The causes of tropical deforestation, 192-210. London: UCL.

Paul, Bimal Kanti. 2005. “Evidence Against Disaster-Induced Migration: The 2004 Tornado in

North-Central Bangladesh,” Disasters, 29(4): 370-385. Paxson, Christina and Cecilia Elena Rouse. 2008. “Returning to New Orleans after Hurricane

Katrina,” American Economic Review, 98(2): 38-42. Pichon, F. and Bilsborrow, R. E. 1999. “Land use systems, deforestation, and demographic

factors in the humid tropics: farm-level evidence from Ecuador.” In R. E. Bilsborrow and D. Hogan ed., Population and deforestation in the humid tropics, 175-207. Liege,

 

 

34

Belgium: International Union for the Study of Population. Pripyat.com. 2011. “Website of the City of Pripyat – Chernobyl Disaster News, Articles,

Media,” http://www.pripyat.com/en. Pugatch, Todd and Dean Yang. 2011. “The Impact of Mexican Immigration on U.S. Labor

Markets: Evidence from Migration Flows Driven by Rainfall Shocks,” working paper. Reuveny, Rafael. 2007. “Climate Change-Induced Migration and Violent Conflict,” Political

Geography, 26: 656-673. Reuveny, Rafael and Will H. Moore. 2009. “Does Environmental Degredation Influence

Migration? Emigration to Developed Countries in the Late 1980s and 1990s,” Social Science Quarterly, 90(3): 461-479.

Rosenzweig, C., and Hillel, D. 1993. “Plant and environment interactions, the Dust Bowl of the

1930s,” Journal of Environmental Quality, 22(1): 9-22. Sader, S. A., Reining, C., Sever, T., and Soza, C. 1997. “Human migration and agricultural

expansion: an impending threat to the Maya Biosphere Reserve,” Journal of Forestry, 95(12): 27-32.

Saldana-Zorrilla, Sergio O. and Krister Sandberg. 2009. “Spatial Econometric Model of Natural

Disaster Impacts on Human Migration in Vulnerable Regions of Mexico,” Disasters, 33(4): 591-607.

Scoones, I. 1992. “Coping with drought: responses of herders and livestock in contrasting

Savanna environments in Southern Zimbabwe,” Human Ecology, 20(3): 293-314. Schwartz, M. L., and Notini, J. 1994. Report on desertification and migration: Case studies and

evaluation. San Francisco: Natural Heritage Institute. Shestakov, A., and Streletsky, V. 1998. Mapping of risk areas of environmentally-induced

migration in the Commonwealth of Independent States. Geneva: International Organization for Migration.

Small, I., van der Meer, J., and Upshur, R. E. 2001. “Acting on an environmental health disaster:

the case of the Aral Sea,” Environmental Health Perspectives, 109(6): 547-549. Smil, V. 1995. China’s environmental refugees: causes, dimensions and risks of an emerging

problem. In K. R. Spillmann, & G. Baechler ed., Environmental crisis: Regional conflicts and ways of cooperation, 75-91. Bern, Switzerland: Swiss Peace Foundation

Smith, Stanley K. and Christopher McCarty. 1996. “Demographic Effects of Natural Disasters:

A Case Study of Hurricane Andrew,” Demography, 33(2): 265-275.

 

 

35

Specter, M. 1994. Far North in Russia, the mines’ fatal blight, New York Times March 28,

1994. Stanley, Denise L. 2010. “Outmigration, Human Development and Trade: A Central American

Case Study,” Journal of Human Development and Capabilities, 11(2): 315-337. Strobl, Eric and Marie-Anne Valfort. 2011. “The Effect of Weather Induced Internal Migration

on Local Labor Markets: Evidence from Sub-Saharan Africa,” working paper. Tse, Chun Wing. 2011. “Do Disasters Lead to Forced Migration: Global Analysis and Case

Study of Indonesia,” SSRN working paper 1908958. UN. 2001. World population monitoring: Population, environment and development. New York:

United Nations. Van der Geest, Keer. 2011. “North-South Migration in Ghana: What Role for the

Environment?,” International Migration, 49 (s1): e69-e94. Vu, Lung, Mark J. VanLandingham, Mai Do, and Carl L. Bankston III. 2009. “Evacuation and

Return of Vietnamese New Orleanians Affected by Hurricane Katrina,” Organization Environment, 22(4): 426-432. 

Yang, Dean. 2003. “Risk, Migration, and Rural Financial Markets: Evidence from Earthquakes

in El Salvador,” working paper. Yoon, Y. S. 1998. On the situation of North Korean defectors in China. Field survey report

conducted in China in 1998. Seoul: Korea Political Development. Zottarelli, Lisa K. 2008. “Post-Hurricane Katrina Employment Recovery: The interaction of

Race and Place,” Social Science Quarterly, 89(3): 592-607. Zweifler, M. O., Gold, M., and Thomas, R. 1994. “Land-use evolution in hill regions of the

Dominican Republic,” Professional Geographer, 46(1): 39-53.