Notes: Corruption and Development - Faculty Support...

Intro Definition Measurement Theories Macro Micro

Notes:Corruption and Development

Jorge M. [email protected]

March 2nd & 4th, 2010

Jorge Aguero Notes: Corruption and Development


Outline of today’s lecture

1 Introduction

2 What is corruption?

3 Measuring corruptionCross-country measuresCommon featuresMagnitude

4 TheoriesWagesIO approach

5 Corruption and Growth

6 Ugandan FirmsWho pays bribes?Is corruption harmful to growth?



First, some numbers

Mobutu Sese Seko, President of Zaire, looted some $5 billion(approx. Zaire’s external debt in 1997.)

Suharto (Indonesia) two times that figure and FerdinandMarcos (Philippines) seven times higher.

An IMF report found that in 2001 alone, $1 billion of oilrevenues vanished from Angolan state accounts. That’s $77per Angolan, where 3/4 of the population lives with less thana dollar a day.

The World Bank estimates that bribes account for around $1trillion in a given year.



2. What is corruption

Misuse of public office for private gain.

Corruption is an outcome.

It is a response to beneficial or harmful rules.

Individuals paying bribes to avoid penalties for harmfulconduct.Bribes allow you to get around bad policies or inefficientinstitutions.



Corruption and taxes and lobbying

Corruption could be seen as a tax or fee.

Like takes, bribes create a gap between actual and privatelyappropriated MPK.

But bribes do not go to Government’s accounts.

Unlike taxes, bribes are not enforceable in courts and havehigher transaction costs due to secrecy and uncertainty.



Corruption and lobbying

Like lobbying, bribes influence buying or campaigncontributions.

There are however three main differences1 Lobbying affects all firms in the industry. Bribes have

firm-specific effects.2 Lobbying has more permanent effects because there are cost

associate to re-enacting a law. A bribed official might ask foranothe bribe in the future.

3 In accepting or rejecting the lobbyist’s agenda the evaluation isbased on the costs and benefits for the governments. In thecase of bribes is depends on private costs and benefits.


Intro Definition Measurement Theories Macro Micro Cross-country measures Common features Magnitude

Cross-country measures I

Cross-country measures can be divided in two groups.

Soft evidence:

International Country Risk Guide: Based on risk assessment. Itmeasures the likelihood government officials will demandspecial payments and the the extent of these paymentsthroughout government tiers (e.g., Mauro 1995).Corruption Perception Index (Transparency International):Nationally-representative perception surveys.Control of Corruption (KKM, 2003).

All these indicators are highly correlated. ρ(CC,CPI)=.97 andρ(CC or CPI,ICRG)≈ .75.



Cross-country measures II

Hard evidence:

EBRD (World Bank): Firm estimates of share of annual sales“firms like yours” typically pay in unofficial payments to publicofficials (26 transition countries only.)

International Crime Victim Surveys (UN): Urban/large citieshousehold surveys about bribes paid to government officials. Itexisted since 1989.

Main problem: We lack hard evidence on corruption.



of theories argues that institutional quality (and thus corruption) is shaped by eco-nomic factors. In short, institutions develop in response to a county’s income level anddifferential needs (Lipset, 1960; Demsetz, 1967). A related view—the human capitaltheory—argues that growth in human capital and income cause institutional develop-ment (Lipset, 1960; Glaeser, La Porta, Lopez-de-Silanes and Shleifer, 2004). Forexample, education and human capital is needed for courts and other formal institu-tions to operate efficiently, and government abuses are more likely to go unnoticedand unchallenged when the electorate is not literate. These theories suggest looking atper capita income and education as causes of corruption.

The second set of institutional theories stress the role of institutions more directly.

Table 1The Most Corrupt Countries(the bottom 10 percent most corrupt countries from each data set)

Country CC Country CPI Country ICRG Country ICVS

Equatorial 1.9c,i,v Bangladesh 8.7v Zimbabwe 5.8v Albania 0.75Guinea Nigeria 8.6 China 5v Uganda 0.36

Haiti 1.7v Haiti 8.5v Gabon 5c,v Mozambique 0.31Iraq 1.4v Myanmar 8.4v Indonesia 5v Nigeria 0.30Congo, Dem. 1.4c,v Paraguay 8.4v Iraq 5v Lithuania 0.24

Rep. Angola 8.2v Lebanon 5v

Myanmar 1.4v Azerbaijan 8.2 Myanmar 5v

Afghanistan 1.4c,i,v Cameroon 8.2v Niger 5c,v

Nigeria 1.4 Georgia 8.2i Nigeria 5Laos 1.3c,i,v Tajikistan 8.2i,v Russia 5Paraguay 1.2v Indonesia 8.2v Sudan 5v

Turkmenistan 1.2c,i,v Kenya 8.1v Somalia 5c,v

Somalia 1.2c,v Cote 7.9v Congo, 5c,v

Korea. North 1.2c,v d’Ivoire Dem. Rep.Zimbabwe 1.2v Kyrgyzstan 7.9i,v Serbia and 5v

Indonesia 1.2v Libya 7.9v MontenegroAngola 1.1v Papua New 7.9v Haiti 4.8v

Bangladesh 1.1v Guinea Papua New 4.8v

Cameroon 1.1v GuineaNiger 1.1c,v

Sudan 1.1v

Azerbaijan 1.1Tajikistan 1.1i,v

Sample size 195 133 140 44

Notes: CC is the Control of Corruption Index for 2002 from Kaufmann, Kraay and Mastruzzi (2003). Theindex takes values between �2.5 to 2.5, with a higher score indicating higher corruption (rescaled). CPIis the Corruption Perception Index for 2003 from Transparency International. The index takes valuesbetween 0 to 10, with a higher score indicating higher corruption (rescaled). ICRG is the InternationalCountry Risk Guide’s corruption indicator for 2001 (average over 12 months). The index takes valuesbetween 0 to 6, with a higher score indicating higher corruption (rescaled). ICVS is the incidence ofbribes in 2000 (share of households responding they need or are expected to pay bribes in 2000) fromthe International Crime Victim Surveys.c indicates that the country is not included in the Corruption Perception Index ranking.i indicates that the country is not included in the ICRG ranking.v indicates that the country is not included in the ICVS survey.

Jakob Svensson 25



Common features of highly corrupted countries

Mostly developing or transition countries.

Had socialist governments.

Closed economies, except for Indonesia.



regression line, which shows that their perceived corruption is close to the expectedlevel given their per capita GDP.

The strong relationship between income and corruption is consistent with thetheories of corruption that argue that institutional quality is shaped by economicfactors. However, it is a weak test of these theories, since economic developmentnot only may create a demand for good government and institutional change, butmay also be a function of the quality of institutions. Moreover, the huge variationaround the regression line suggests that these theories are at best incomplete.

What can account for this variation? To explore this, I carried out a series ofregressions where the dependent variable is corruption, proxied with the threesubjective measures of corruption described earlier. The explanatory variables ineach regression include initial GDP per capita and initial human capital (bothmeasured in 1970) as control variables. I then add a series of country characteris-tics, one at a time, and test if the coefficient is significantly different from zero.These partial correlations, of course, do not identify causal effects. Even so, thecorrelations are interesting because they reveal something about common charac-teristics of corrupt countries, adjusting for initial income and human capital.

What are the results? Table 2 shows that corrupt countries have significantlylower levels of human capital stock, proxied by years of schooling of the totalpopulation aged over 25. This relationship holds independent of what measure ofcorruption is used.

Figure 1Corruption and Income

CC

200

2

Real GDP per capita 1995

309.81 35144.2

.11

4.39

..

Ethiopia

..

EritreaMadagascar

.

.

.

.

.

...

Burkina Faso

.

..

.

.

.

.

.

.

.

Mauritania

.

.

..

.

.

.

.

Equatorial Guinea

.

.

..

.

Haiti

.

.

.

.

Bhutan

.

..

.

....

.

..

..

.

.

.

.

.

.

.

.

.

.

.Indonesia

.

.

.

.

.

.

.

Turkmenistan

..

..

..

..

.

.

Paraguay

.

.

.

.

Kazakhstan

.

.

.

.

.

.

.

.

.

.

Venezuela

.

Russia

.

.

.

.

.

.

Gabon

.

.

.

.

.

.

.

Argentina

.. . .

....

.

.

.

..

..

.

.

.

New Zealand

.

.

Finland

.

.

.

.

Sweden

.

..

..

..

Macao

..

..

.

Singapore

.

.

.

Note: The graph depicts the regression line of corruption (CC 2002) on real GDP per capita (inlogarithms) 1995.

Eight Questions about Corruption 27



Institutional theories I

There are two main institutional views.

Institutions are shaped by economic factors

Institutions develop as response to the country’s income level.

Human capital: education and human capital is needed forcourts and formal institutions to operate efficienlty.

An illiterate electorate is more permissive to government abuse.

Income and education are the causes of corruption.



Institutional theories II

Direct effects

Institutions have a long-lasting effect and are inherited.

Think along the terms of AJR’s paper on settlers’ mortality.

More regulatory institutions (French system) leads to morecorruption (La Porta, Lopez, Shleifer and Vishny, 1998).

Religion: Protestant church arose, in part, as an opposition tostate-sponsored religion.

Politicians are less challenged in Catholic and Muslim countriescompare to Protestant countries (Landes, 1998).



Corrupt countries do have some significantly different policy characteristics.Table 3 shows the regression results from a measure of openness to exter-nal competition from imports (imports of goods and services as percent of GDP).Table 4 shows the regression based on the extent of regulation of entry of start-upfirms (time it takes to obtain legal status to operate a firm). Table 5 showsregression results based on freedom of the press (a subjective score from FreedomHouse). The findings are robust across data sets, although the openness proxy isinsignificant in some regressions. Corrupt countries are less open and regulateboth entry to the market and the press more. Replacing freedom of media with abroader measure of political freedom (like the broader Gastil index also producedby Freedom House) yields qualitatively similar results.9

Using the incidence of bribes from the International Crime Victim Survey as

9 I also carried out parallel regressions using a variety of other explanatory variables that provided lessrobust results. Tables showing these regression results are in an appendix attached to the on-line versionof this paper at the journal’s website, �http://www.e-jep.org�. A short summary of the results is that whensettler mortality is included in this sort of regression (which is used as a proxy variable for whether it wasattractive for Europeans to settle in a certain area), it cannot account for why some countries, givencurrent levels of physical and human capital, are more corrupt than others. Countries with a Frenchlegal system or a socialist legal system tend to have more corruption, although the connection is notstatistically significant in all data sets. The proportion of the population identified as Catholic ispositively correlated with several corruption indicators; however, correlations between the proportion ofthe population that is Muslim and measures of corruption are not statistically significant. The religiousand legal variables lose significance in a multiple regression with the policy variables as additionalcontrols.

Table 2Corruption and Country Characteristics: Human Capital

Dep. variable

Control ofCorruption

(2002)

CorruptionPerception

Index(2003)

ICRGCorruption

Score(1982–01)

ICRGCorruption

Score(2001)

IVSCIncidenceof Bribes(2000)

Real GDP per capita (log) �0.60*** �1.38*** �0.87*** �0.73*** �0.03**(.123) (.33) (.20) (.19) (.01)

Years of schooling (log) �0.62*** �1.53*** �0.53** �0.51*** �0.06*(.18) (.52) (.27) (.28) (.03)

Sample size 91 79 83 83 26

Notes: Control of Corruption Index for 2002 from Kaufmann, Kraay and Mastruzzi (2003). The index takesvalues between �2.5 to 2.5, with a higher score indicating higher corruption (rescaled). CorruptionPerception Index for 2003 from Transparency International. The index takes values between 0 to 10, witha higher score indicating higher corruption (rescaled). ICRG is the International Country Risk Guide’scorruption indicator for 2001 (average over 12 months). The index takes values between 0 to 6, with a higherscore indicating higher corruption (rescaled). ICVS is the incidence of bribes in 2000 (share of householdsresponding they need or are expected to pay bribes in 2000) from the International Crime Victim Surveys.Real GDP per capita in 1970 is from the Penn World Tables. Years of schooling of the total population agedover 25 in 1970 is from Barro and Lee (2000). Robust standard errors in parenthesis.*** statistically significant at 1 percent level.** statistically significant at 5 percent level.* statistically significant at 10 percent level.

28 Journal of Economic Perspectives



the dependent variable, rather than one of the subjective measures of corruption,drastically reduces the sample size, as shown in the final column of the tables.Somewhat surprisingly, only GDP per capita, the proxy for initial human capitalstock, and regulation of the press remain significantly correlated with corruption.

These associations suggest some general conclusions. First, corruption isclosely related to GDP per capita and to human capital. These correlations areconsistent with the economic and human capital theories of institutional develop-ment, but the correlations could also be driven by reverse causality or omitted

Table 3Corruption and Country Characteristics: Openness

Dep. variable


(2002)


Index(2003)

ICRGCorruption

Score(1982–01)

ICRGCorruption

Score(2001)


Real GDP per capita (log) �0.67*** �1.43*** �0.90*** �0.71*** �0.06***(.12) (.32) (.21) (.20) (.01)

Years of schooling (log) �0.51*** �1.36*** �0.47* �0.53*(.18) (.50) (.27) (.28)

Imports/GDP �0.01** �0.03*** �0.00 �0.01 �0.00(.00) (.01) (.00) (.00) (.00)

Sample size 89 77 83 81 44

Notes: For details on sources of data, see Table 2. Imports/GDP is imports of goods and services aspercentage of GDP (average from 1980–2000) from World Development Indicators (2004).*** statistically significant at 1 percent level.** statistically significant at 5 percent level.* statistically significant at 10 percent level.

Table 4Corruption and Country Characteristics: Regulation of Entry

Dep. variable


(2002)


Index(2003)

ICRGCorruption

Score(1982–01)

ICRGCorruption

Score(2001)



Years of schooling (log) �0.18 0.12 �0.17 �0.03(.31) (.61) (.40) (.34)

No. of business days toobtain legal status (log)

0.33*** 0.98*** 0.27*** 0.34*** 0.01(.09) (.20) (.10) (.10) (.01)

Sample size 61 60 61 61 35

Notes: For details on sources of data, see Table 2. Number of business days to obtain legal status is thetime it takes to obtain legal status to operate a firm, in business days (a week has five business days anda month has 22) from Djankov, La Porta, Lopez de Silanes and Shleifer (2002).*** statistically significant at 1 percent level.** statistically significant at 5 percent level.* statistically significant at 10 percent level.

Jakob Svensson 29



variables. Second, for a given level of income, the extent of corruption still variesgreatly. The cross-country evidence suggests that this variation can partly be ac-counted for by the degree of market and political competition.

What is the Magnitude of Corruption?

The rankings of countries as more or less corrupt are based on subjectivejudgments and as such cannot be used to quantify the magnitude of corruption.Thus, until recently, the magnitude of corruption had to be assessed using anec-dotal or case-study evidence.10 However, the past few years has seen a small butgrowing body of research on identifying and quantifying corrupt behavior.11

There is some firm-survey evidence on the magnitude of corruption. Svensson(2003) presents survey data from Ugandan firms. Although the survey was adjustedin several ways to encourage managers to report graft payments truthfully, somemisreporting surely remains in the sample. Nonetheless, the results provide agloomy picture of entrepreneurship in one of the fastest growing countries insub-Saharan Africa in the last 10–15 years. Over 80 percent of Ugandan firmsreported needing to pay bribes. Avoiding graft comes at a cost, since the 20 percent

10 As an example in this journal, see McMillan and Zoido (2004). They use recorded bribe transactionsof and by Peru’s former secret-police chief Montesinos and find that Montesinos paid television-channelowners 100 times in bribes what he paid judges and politicians. Using a revealed preference argument,they conclude that news media, consistent with the cross-country evidence discussed above, are thestrongest check on the government’s power.11 Again, the focus of this paper is on public corruption. There is a related literature on privatecorruption or collusion (for instance, McAfee, 1992; Porter and Zona, 1993; Duggan and Levitt, 2002).There is also a related literature on the value of political connectedness (for instance, Fisman, 2001;Khwaja and Mian, 2004).

Table 5Corruption and Country Characteristics: Freedom of Media

Dep. variable


(2002)


Index(2003)

ICRGCorruption

Score(1982–01)

ICRGCorruption

Score(2001)



Years of schooling (log) �0.65*** �0.97* �0.18 �0.22(.12) (.59) (.28) (.36)

Freedom of media index �0.05** �0.10* �0.06** �0.05* �0.01**(.02) (.06) (.03) (.03) (.00)

Sample size 91 79 83 83 44

Notes: For details on sources of data, see Table 2. Freedom of media index is the average score of the fourcriteria “Laws and regulations that influence media content,” “Political pressures and controls on mediacontent,” “Economic influences over media content,” “Repressive actions” for print and broadcastmedia, average over 1994–2001, from the Freedom House.

30 Journal of Economic Perspectives



Magnitude of corruption

Rankings cannot tell us much about the magnitude ofcorruption.

Here are some micro examples

Uganda (Svensson 2003, QJE): 80% of firms reported need topay bribes.

For graft-paying firms, graft is 8% of total costs.

Other studies include Olken’s project in Indonesia, Di Tellaand Schargrosky (2003) in Argentina.

Hsieh and Moretti (2005) paper on Iraq and UN Oil for FoodProgram.


Intro Definition Measurement Theories Macro Micro Wages IO approach

4. Theories

We will consider two main theories.

One sees the problem as what wages should be paid to agentsin order to avoid corruption.

The other sees corruption and Principal-Agent problem.



Bureaucrat’s wages

Becker and Stigler (1974) is a model of corruptible enforcers(e.g., auditors and police).

Let w be the wage and v outside wage.

If bribed:

With probability p, person is detected and receives outsidewage v.With probability 1− p is not detected and receives b+ w

Equilibrium wage (indifferent between taking bribes or not)

w = pv + (1− p)(b+ w)

This implies

w − v =1− p

pb

Compensation (w − v) increases with b and decreases with p.



Multiple equilibria

Consider wage w fixed.

But now the probability of detection (p) depends onproportion of corrupted people in the population (c).

Assume that p(c) = 1− c.

The agent is indifferent is

w = pv + (1− p)(b+ w)

In terms of c, take bribe if

w − v <c

1− cb

This generates multiple equilibria.

Small changes in wage or corruption crackdown can createpermanent changes.



Multiple equilibria

Theory Empirics Basic Model Multiple Equilibria IO Model

Multiple equilibria

c0 1

1c b

c−

w v−

I Implication: temporary wage increase or corruption crackdowncan have permanent e¤ects

Econ 270C Corruption Lecture



Multiple equilibria: Bardhan and Udry (1999)

Measure the proportion of corrupt agents along the x-axis.

The M and N -curve show the marginal benefits for corruptand non-corrupt officials.

With low corruption N > M and with high corruptionM > N .

N can even be negative at high levels of corruption (as shownin m).

There are three equilibrium levels: A, B and C.

A and C are stable. B is not.




B M curveA

N curve

C

0

m




This shows how two otherwise similar countries (same SESand moral attitudes) may have different levels of corruption.

It depends on how close these countries are to point B.

Like before, small changes can create permanent changes.



Does higher wages attract “better” politicians?

Ferraz and Finan (2009)

Reform in Brazil: salaries to politicians can vary due topopulation size.

These provides a natural experiment.

Income will increase due to “exogenous” variation.

Is the pool of applicants different when wages are higher?



Wages and population

1000

1200

1400

1600

1800

2000

Wag

es

5000 7500 10000 12500 15000Population

010

0020

0030

0040

0050

00W

ages

10000 30000 50000 70000 90000 110000 130000 150000Population

020

0040

0060

0080

00W

ages

100000 300000 500000 700000 900000Population

FIGURE 1: LEGISLATORS’ SALARIES BY POPULATION

Notes: Figure shows legislators’ salaries by population. Each figure presents the mean wage for a bin size of 200 inhabitants (hollow-circles) along with a locally weighted regression calculated within each population segment with a bandwidth of 0.5. The vertical lines denote the various cutoff points.



Ceteris paribus

Income per capita (log)

Private Sector Wages (logs)

Assistants per legislators

Total Expenditure 2000

Effective Number of Political Parties in 1996 Elections

Hours in session

44.

55

5.5

66.

5

0 20000 40000 60000 80000 100000Population

66.

57

7.5

88.

5

0 20000 40000 60000 80000 100000Population

02

46

0 20000 40000 60000 80000 100000Population

1415

1617

18

0 20000 40000 60000 80000 100000Population

24

68

0 20000 40000 60000 80000 100000Population

010

2030

40

0 20000 40000 60000 80000 100000Population

FIGURE 2: MUNICIPAL CHARACTERISTICS BY POPULATION

Notes: The figure shows municipal characteristics by population. Each figure presents the mean of the municipal characteristic for a bin size of 200 inhabitants (hollow-circles) along with a locally weighted regression calculated within each population segment with a bandwidth of 0.5. The vertical lines denote the various cutoff points.



“Better” pool of candidates

Table 7. The Effects of Politicians’ Wages on Candidate Selection

Dependent variable:

(1) (2) (3) (4) (5) (6) (7) (8)Panel A. Candidate characteristics

Log wages 0.584 0.602 0.062 0.068 0.062 0.062 0.026 0.028[0.352]* [0.318]* [0.037]* [0.035]* [0.028]** [0.027]** [0.017] [0.017]*

F-test 29.12 29.76 29.12 29.76 30.03 29.93 29.12 29.76(exc. instruments)

Panel B. Legislators' characteristics

Log wages 0.885 0.876 0.107 0.11 0.084 0.079 0.039 0.043[0.478]* [0.444]** [0.053]** [0.051]** [0.049]* [0.048] [0.031] [0.031]

F-test 29.12 29.76 29.12 29.76 30.03 29.93 29.12 29.76(exc. instruments)

Municipal characteristics No Yes No Yes No Yes No YesObservations 4887 4887 4889 4889 4890 4890 4889 4889

Log Years of schooling Share with at least a high school education

Share skilled occupations Share of female

Notes: The table reports the TSLS estimates for the effects of wages on the characteristics of those that ran and were elected as legislators in the 2004 elections. All regressions control for the number of hours the legislature functions per week and assistants per legislator. Municipal Characteristics include Log household income per capita, % urban population, Gini coefficient, % households with energy, % literate population, log average wage in private and public sector in municipality, and a linear spline in population. * indicates statistical significance at the 10% level, ** at the 5% level and *** at the 1% level. Robust standard errors are reported in brackets. The instruments used are the indicators for the cutoffs at 1{x>10,000}, 1{x>50,000}, 1{x>100,000}, 1{x>300,000} and 1{x>500,000}. The reported F-test refers to these excluded instruments.



Industrial Organization approach

Two types of corruption (Shleifer and Vishny, 1993):1 Corruption without theft: bribes are paid on top of official

fees. Here corruption decreases efficiency.2 Corruption with theft: bribes are paid instead of fees. The

impact on efficiency is unclear.



Principal-agent problem

The chief of a village (the principal) monitors villagers (theagents) and collect taxes from them.

As the village grows it is harder for him to keep track of allvillagers. So he hires “policemen” and tax collectors.

They are the “intermediaries”.

The problem: agents and intermediaries can engage in privatecontracts (bribes) so that intermediaries do not report to chief.

This is a problem with asymmetric information.

How do we solve the problem?

Just like in the typical problem, you will make a payment suchthat the policeman will tell you the truth.



Zero corruption?

Can we eradicate corruption at any cost?

The principal might not be well informed about theintermediary propensity for corruption.

Consider the case where a policeman can be corrupted by lowor high bribes.

To eliminate corruption you will have to make high paymentsto the former type.

It is much cheaper to make smaller payments.



5. Corruption and economic growth

Mauro (1995, QJE).

Question: Do more corrupt countries have less investment andslower growth?

Uses 1980-1983 Business International indices of corruption.

This is the risk-based assessment measure described above.

OLS estimates show a negative relationship betweencorruption and growth.

But is this causal?

He instruments using ethnolinguistic fragmentation.


Intro Definition Measurement Theories Macro Micro 687 CORRUPTION m D GROWTH

TABLE I BUREAUCRATIC INDEXEFFICIENCY

1.54.5 4.5-5.5 5.5-6.5 6.5-7.5 7.5-9 9-10

Egypt Algeria Angola Argentina Austria Australia Ghana Bangladesh Dominican Rep. Ivory Coast Chile Belgium Haiti Brazil Ecuador Kuwait France Canada Indonesia Colombia Greece Malaysia Germany Denmark Iran India Iraq Peru Ireland Finland Liberia Jamaica Italy South Africa Israel Japan Nigeria Kenya Korea Sri Lanka Jordan Hang Kong Pakistan Mexico Morocco Taiwan Zimbabwe Netherlands Thailand Philippines Nicaragua Uruguay New Zealand Zaire Saudi Arabia Panama Norway

Turkey Portugal Singapore Venezuela Spain Sweden

TrinidadlTobago Switzerland United Kingdom United States

BE is the bureaucratic efficiency index, which I compute as the simple 1980-1983average of three Business International indices: judiciary system, red tape, and corruption. A high value of the BE index means that the country's institutions are good.

that richer countries tend to have better institutions than poorer countries, and that fast-growers also tend to be among the countries with a higher bureaucratic efficiency index. Neverthe- less, there are a few of surprises. In 1980 BI reported Thailand to be the most corrupt country, yet its economic performance has been relatively good. Korea has been a fast grower, in spite of the fact that it was reported to have relatively inefficient institutions. l3

Figures 1-111 provide scatter plots of per capita GDP, the investment rate, and the per capita GDP growth rate versus the bureaucratic efficiency index for the 67 countries for which both Summers and Heston [I9881 and BI data are available in 1980- 1983. All these correlations are significant a t the 1percent level.

One of the most striking features of the data set is the strong association between bureaucratic efficiency and political stability.14 Table I1 arranges the countries in the data set in a matrix, grouping them by quintiles depending on their bureaucratic efficiency and

13. The BI indices refer to the period immediately following the assassination of President Park Chung-hee.

14. Corruption may be more deleterious and thus reported as a more serious problem in politically unstable countries. Shleifer and Vishny [I9931 argue that countries with weak (and, therefore, unstable) governments will experience a very deleterious type of corruption, in which an entrepreneur may have to bribe several public officials and still face the possibility that none of them really have the power to allow the project to proceed.



688 QUARTERLY JOURNAL OF ECONOMICS

Saudi Arabia

Kuwait

-

United States

-France*

** Austria +

ItalyTrinidad **

Spain - Mexico Ireland Singapore

3 **~outh~fr ica Iran

~ r a z i l * Clule

Thailand ' Jordan F'hilipplnes*** :* *e

Haiti Zaire eT Indonesia f l , . 7 Zimbabwe

I0 0 2 4 6 8 10

Bureaucratic Efficiency(BE) index

FIGUREI Per Capita Income and Bureaucratic Efficiency

BE index is 1980-1983 average of BI indices of corruption, red tape, and judiciary.

Per capita GDP at PPP in 1980 is from Summers and Heston L19881. 67 countries,r = 0.68.

political stability indices. Most countries lie near or on the diago-nal. The simple correlation coefficient between the bureaucratic efficiency index and the political stability index is 0.67, and the partial correlation coefficient controlling for per capita GDP in 1980 is 0.45, both significant at the 1percent level. Yet, several relatively stable countries are reported to have relatively ineffi-cient, corrupt bureaucracies. Conversely, several countries with relatively efficient, honest bureaucracies are relatively politically unstable. Based upon the 1980-1983 BI indices, Egypt, Greece, Indonesia, Saudi Arabia, and Turkey are at least two quintiles better on the grounds of political stability than on the grounds of bureaucratic efficiency. On the other hand, Angola, Chile, Iraq, Israel, Nicaragua, Peru, South Africa, and Zimbabwe score at least two quintiles better on bureaucratic efficiency than on political stability.15 For example, Indonesia under President Suharto was

15. A similar matrix appears in Coplin and O'Leary [19821. They classify 73 countries by political instability and restrictions of business. Their classification broadlyconfirms the one reported in Table 11.



CORRUPTION AND GROWTH 689

0 0 2 4 6 8 10 12

Bureaucratic Efficiency(BE) index

-Japan Singapore

Saudi Arabia Malaysia Fida;d

** - Ireland Switzerland Algeria

Jordan* * *

eGreece France* N o m y

Iran New Zealand Mei, : :* =. A".: "

.united States-Indonesia e*

Emt. *& Netherlands Thailand

Zimbabwe

Zaire Liberia ** United Kingdomf - Haiti Pakistan

*pemMorocco

-a Bangadesh

Angola

I I I I I

FIGUREI1 Investment and Bureaucratic Efficiency

BE index is 1980-1983 average of BI indices of corruption, red tape, and judiciary.

Average investment 1980-1985 from Summers and Heston [19881. 67 countries,r = 0.46.

relatively politically stable, although BI reports that companies were hindered by a corrupt, cumbersome bureaucracy. According to BI's consultants, Peru's fragile democracy and its problems with socialviolence and terrorism and SouthAfrica's racial tensions and active trade unions were in sharp contrast to their relatively efficient bureaucracies. Thus, even though bureaucratic efficiency and political stability are positively and significantly correlated, there is a wealth of information in the bureaucratic efficiency indicesthat can be used to analyze the determinants of investment and growth.

The fact that the indices reflect the subjective opinions of BI's correspondents presents both advantages and disadvantages. An advantage relates specifically to the political instability variables. Previous studies have used objective measures of political stability, such as the number of political assassinations or changes in government. Objective measures can often be misleading. For example, there have been over 50 changes of government in Italy since 1945,yet the country has been relatively politically stable. It


Intro Definition Measurement Theories Macro Micro 693 CORRUPTION AND GROWTH

TABLE I11 ETHNOLINGUISTIC 1960FRACTIONALIZATION,

Angola Canada Algeria Argentina Austria Dominican Bangladesh Ghana Belgium Australia Brazil Rep. India Malaysia Ecuador Finland Chile E m t Indonesia Pakistan Iraq France Colombia Germany Iran Peru Morocco Israel Denmark Haiti Ivory Coast Philippines New Zealand Kuwait Greece Hong Kong Kenya Thailand Singapore Mexico Jamaica Ireland Liberia Trinidad/ Spain Nicaragua Jordan Italy South Africa Tobago Sri Lanka Panama Netherlands Japan Zaire Switzerland Turkey Saudi Arabia Korea

Taiwan United Sweden Norway United Kingdom Venezuela Portugal

States Uruguay Zimbabwe

Theethnolinguistic fractionalization index for 1960is drawn from Taylor and Hudson [I9721

There is a negative and significant correlation between institutional efficiency and ethnolinguistic fractionalization, which makes the latter a good instrument.lg The ELF index has a simple correlation coefficient equal to -0.38 with the institutional efficiency index, -0.41 with the political stability index, -0.28 with the bureaucratic efficiency index, and -0.31 with the corruption index, all significant at the 1 percent level. A number of mecha- nisms may explain this relationship. Ethnic conflict may lead to political instability and, in extreme cases, to civil war. The presence of many different ethnolinguistic groups is also significantly associ- ated with worse corruption, as bureaucrats may favor members of their same group. Shleifer and Vishny [I9931 suggest that more homogeneous societies are likely to come closer to joint bribe maximization, which is a less deleterious type of corruption than noncolli~sive bribe-setting. Strictly speaking, the ELF index is a

has a positive and significant effect on productivity growth. They also argue that it is a predetermined proxy for political stability. However, they do not use the homogeneity index as an instrument for political stability. Hibbs [I9731 uses the index in a large system of simultaneous equations which is ultimately designed to explain mass political violence and other indicators of political instability.

19. Ethnolinguistic fractionalization is a valid instrument, while lags of the right-hand side variables such as beginning-of-period indicators of corruption and political instability would be unlikely to be valid instruments, because such institutional variables are highly autocorrelated.


Intro Definition Measurement Theories Macro MicroCORRUPTION AND GROWTH

TABLE VI INVESTMENT BUREAUCRATICON CORRUPTION, EFFICIENCY

Dependent variable: investment1GDP (1960-1985 Average)

Independent variable (1) (2) (3) (4) (5) (6) (7)

Constant 0.104 0.114 0.196 0.036 0.039 0.186 0.001 (3.03) (3.18) (4.65) (0.42) (0.40) (0.31) (0.01)

GDP in 1960 -0.008 -0.006 -0.004 -0.026 -0.021 -0.015 -0.017 (-1.31) (-0.81) (-0.60) (-1.57) (-1.41) (-2.50) (-2.73)

Secondary educa- 0.060 0.111 0.096 -0.078 0.017 0.082 0.115 tion in 1960 (0.97) (1.68) (1.40) (-0.56) (0.16) (1.60) (2.04)

Population -1.373 -0.620 -0.913 -2.754 -1.144 growth (-1.38) (-0.61) (-0.82) (-1.84) (-1.12)

Primary educa- 0.105 0.111 tion in 1960 (2.89) (3.36)

Government -0.166 -0.206 expenditure (-1.06) (- 1.39)

Revolutions and -0.009 -0.005 coups (-0.22) (-0.139)

Assassinations -0.164 -0.276 (-0.69) (-1.03)

PPI60 -0.058 -0.061 (-2.81) (-2.79)

PPIGODEV 0.043 0.035 (1.24) (1.04)

Africa 0.036 (1.92)

Latin America 0.017 (0.88)

High Bureaucratic efficiency dummy

Low Bureaucratic efficiency dummy

Political stability 0.013 0.014 index (1.64) (1.79)

Bureaucratic effi- 0.019 0.004 0.010 0.009 ciency index (4.04) (1.76) (2.19) (1.76)

Corruption index 0.013 0.034 (2.94) (1.56)

Estimation OLS OLS OLS 2SLS 2SLS OLS OLS method

R 2 0.51 0.47 0.44 (*) (*) 0.65 0.66

A high value of a BI index means the country hasgood institutions. One standard deviation equals 2.16 for the bureaucratic efficiency (BE) index, 2.51 for the corruption index, and 1.29 for the political stability index. The high (low) BE dummy takes the value one when the BE index is above 8.33 (below 5.801; there are 19 high BE and 19 low BE countries. There are 58 observations in the case of OLS and 57 in the case of 2SLS. White-corrected t-statistics are reported in parentheses. The Barro [19911 regressors used are per capita GDP, orimarv education. secondary education, the purchasinp-power parity value for the investment deflator (PPI6O) A d itsbeviation from the sample mean (PPI~ODEV)in-1960, the 1960-1985 average of the ratio of government consumotion ex~enditure (net of suendineon defense and education) to GDP. population mowth, the number of ~ - ~ - -

= ~ A . -revolutions and coups, the number of assassinations, and dummies for I,atinAmerica &d Sub-Saharan Africa where indicated. 2SLS indicates that the index of ethnolinguistic fractionalization in 1960, from Taylor and Hudson [19721, is used as an instrument. ('1 The R 2 is not an appropriate measure of goodness of fit with two-stage least squares.



Update on Mauro’s paper

country-specific fixed effects to control for time-invariant country characteristics,also yields insignificant results.13

This finding seems to lead to a puzzle. Most of the theoretical literature as well ascase study and micro evidence suggest that corruption severely retards development.However, to the extent we can measure corruption in a cross-country setting, it doesnot affect growth. The puzzle may arise from econometric problems involved inestimating the effects of corruption on growth using cross-country data. For example,the difficulties of measuring corruption may include omitted variables, like the extentof market regulation, and reverse causality, like whether modernization and rapidgrowth may increase corruption, as Huntington (1968) argued. Another plausibleexplanation for the mismatch between the micro and macro evidence is that corrup-tion takes many forms, and there is no reason to believe that all types of corruption areequally harmful for growth. Existing data, however, are by and large too coarse toexamine different types of corruption in a cross-section of countries.

Conclusion

In this paper, I posed eight questions about corruption. The answers are oftennot clear-cut, and there are many issues about corruption we simply know too littleabout. As the study of corruption evolves, three areas are of particular importance.

13 Using the two other subjective corruption indicators yields, in some specifications, a statisticallysignificant negative effect of corruption on growth. However, these indicators are measured at the endof the sample period, thus making it even more difficult to draw causal interpretations from corruptionto growth.

Table 6Growth and Corruption

Growth(1980–2000)

Growth(1980–2000)

Dep. variable Ordinary least squares Fixed effects

Real GDP per capita (log) �0.82* �6.50***(.47) (1.03)

Years of schooling (log) 1.86*** 6.63***(.66) (1.36)

Corruption �0.33 0.11(.24) (.24)

Countries 85 86Observations 85 335

Notes: For details on sources of data, see Table 2. Growth is growth in real GDP per capita overthe period 1980–2000 in specification (1) and growth in real GDP per capita over the periods1981–1985, 1986–1990, 1991–1995, 1996–2000 in specification (2). Real GDP per capita andyears of schooling are measured at the start of the sample period (in 1980 for specification (1)and in 1980, 85, 90, 95 for specification (2)). Corruption is the International Country RiskGuide’s corruption indicator, average for 1982–2000 in specification (1) and average over1982–1985, 1986–1990, 1991–1995, 1996–2000 in specification (2).

Eight Questions about Corruption 39


Intro Definition Measurement Theories Macro Micro Who pays bribes? Is corruption harmful to growth?

Two questions about corruption

1 Who pays bribes? (Svensson, 2003)

2 Is corruption harmful to growth? (Fisman and Svensson,2007)



Who pays bribes?TABLE I

PROBIT REGRESSIONS ON THE INCIDENCE OF CORRUPTION

Specification (1) (2) (3) (4) (5) (6) (7)

Constant 0.203 0.647 0.428 0.254 0.206 �0.090(.342) (.155) (.276) (.356) (.467) (.461)[.554] [.000] [.121] [.476] [.659] [.846]

Employment 8.4E-5 �7.9E-5 �8.2E-5 0.001 0.001 0.001 0.001(4.3E-4) (4.4E-4) (4.4E-4) (.001) (.001) (.001) (.001)[.848] [.857] [.852] (.280) (.278) (.477) (.380)

Infrastructure 0.192service (.094)

[.041]Trade 0.430

(.238)[.070]

Pay tax 0.374(.220)(.089)

Formal sector 0.140 0.141 0.213 0.200(.082) (.083) (.099) (.074)[.088] [.087] [.032] [.007]

Profit �2.6E-9 �4.0E-9 1.7E-8 2.4E-9(4.8E-8) (4.8E-8) (4.9E-8) (5.3E-8)[.957] [.935] [.730] [.964]

Capital stock �3.2E-7 �3.1E-7 �4.2E-7 �3.4E-7(2.5E-7) (2.6E-7) (2.5E-7) (2.8E-7)[.199] [.224] [.090] [.224]

Alternative �8.8E-7 �7.6E-7 2.4E-7 �6.3E-7return (1.1E-5) (1.1E-5) (1.1E-5) (1.1E-5)

[.934] [.884] [.983] [.956]Competition 0.003

(.018)[.884]

Sell to �0.337government (.272)

[.216]Exemption 0.515

(.216)[.017]

Industry — — — — — — 5.09[.885]

LR(z) 6.15 5.84 7.05 4.86[.104] [.119] [.070] [.183]

Observations 176 167 173 149 148 134 149

a. Dependent variable “incidence of graft” takes the value 1 if the firm reported positive bribe paymentsand 0 otherwise.

b. Standard errors are in parentheses, and p-values are in brackets.c. Industry is the likelihood-ratio test statistic for the H0 that the industry effects are equal.d. LR(z) is the likelihood-ratio test statistic for the H0 that the coefficients on the bargaining measures

(profit, capital stock, alternative return) are zero.



Who pays more?

2 displays the same regression once these outliers have beendropped. The fit of the regression improves, and the standarderrors of all bargaining measures are reduced.

Summarizing the basic findings on the magnitude of graft,the more a firm can pay; i.e., the higher are its current andexpected future profits, the more it must pay. The more profitable

TABLE IICORRUPTION REGRESSIONS

Specification (1) (2) (3) (4) (5)

Constant 17.1 14.2 38.8 �3.19(37.1) (35.9) (49.1) (46.2)[.646] [.694] [.432] [.945]

Profit peremployee 0.0040 0.0040 0.0042 0.0042 0.0038

(.0008) (.0008) (.0008) (.0008) (.0008)[.000] [.000] [.000] [.000] [.000]

Capital stock per 0.0041 0.0043 0.0040 0.0047 0.0041employee (.0024) (.0022) (.0024) (.0023) (.0027)

[.089] [.062] [.090] [.043] [.123]Alternative return �0.234 �0.239 �0.235 �0.253 �0.228per employee (.096) (.093) (.094) (.092) (.099)

[.017] [.012] [.014] [.007] [.024]Formal sector 9.83 9.61 8.20 12.2 7.13

(7.41) (7.22) (7.52) (8.31) (8.72)[.187] [.186] [.278] [.145] [.416]

Competition �1.30(1.75)[.460]

Sell to government �3.29(24.0)[.891]

Exemption 0.977(17.2)[.955]

Industry — — — — 8.41[.752]

LR(z)c 27.8 30.1 30.4 32.7 27.9[.000] [.000] [.000] [.000] [.000]

Observations 119 117 116 105 117

a. Dependent variable is graft in US$ per employee.b. Least-squares estimates with standard errors are in parentheses, and p-values are in brackets.c. Specification (1) includes two outliers.d. Industry is the likelihood-ratio test statistic for the H0 that the industry effects are equal.e. LR(z) is the likelihood-ratio test statistic for the H0 that the coefficients on the bargaining measures

(profit, capital stock, alternative return) are zero.

219WHO MUST PAY BRIBES AND HOW MUCH?



Is corruption harmful to growth?

industry fixed effects (with firms classified into five industrial sectors: commercial agriculture,agro-processing, other manufacturing, construction, and tourism. In these cases the variation usedto identify the effects come from the variation across industries (locations) in a given location(industry). However, including any one of these variables in the growth equation did notsignificantly affect the relationship between corruption and growth.

As a final test of the identifying assumption, we include two measure taken from Svensson(2003) of the control public officials maintain over the firms.12 The first variable is the percentageof senior management's time spent dealing with government regulations each month(REGULATION). The second variable is an index (INFRASERV) of the availability of five keypublic services (electricity, water, telephones, waste disposal, and paved roads). As stressed above,our identifying assumption would be invalid if some industries-locations have been systematicallydisfavored by the government in that they receive less and worse public services and governmentofficials systematically increase both the regulatory burden and demands for bribes. If this wouldbe the case, presumably firms in industry-locations with few public services and extensiveregulation would grow slower. In column (4), Table 3, we add the industry-location averages of

12 Svensson (2003) finds firms with more extensive dealing with the public sector are more likely to pay bribes.

Table 3Effect of bribery and taxation on growth: instrumental variable estimation

Dependent variable: GROWTH

(1) (2) (3) (4) (5)

Method IV IV IV IV IVBRIBE −3.320⁎⁎

(1.558)−3.291⁎⁎(1.641)

−3.635⁎⁎(1.671)

−4.173⁎⁎(2.100)

−3.485⁎⁎(1.628)

TAX −1.342⁎⁎(0.638)

−1.579⁎⁎(0.684)

−1.698⁎⁎(0.680)

−1.849⁎⁎(0.723)

−1.640⁎⁎(0.701)

LSALES95 0.008 (0.018) −0.009 (0.014) −0.019 (0.015) −0.022 (0.017) −0.019 (0.015)LAGE −0.063 (0.043) −0.049 (0.039) −0.052 (0.044) −0.060 (0.046) −0.051 (0.043)FOREIGN 0.261⁎⁎ (0.001) 0.216⁎⁎ (0.102) 0.211⁎⁎ (0.101) 0.2(0.102)TRADE 0.125⁎ (0.066) 0.133⁎⁎ (0.064) 0.125⁎ (0.066)INFRASERV 0.043 (0.039)REGULATION 0.012 (0.054)Cons 0.249 (0.340) 0.506⁎ (0.304) 0.671⁎⁎ (0.314) 0.569⁎ (0.307) 0.664⁎⁎ (0.308)F-test of instruments

(in BRIBE regression)24.05 {0.00} 24.14 {0.00} 23.65 {0.00} 19.66 {0.00} 13.33 {0.00}

F-test of instruments(in TAX regression)

18.04 {.000} 27.53 {.000} 23.61 {.000} 25.98 {.000} 14.11 {.000}

Hansen J-statistic 1.153 {.562}Observations 126 126 123 123 123

Standard errors in parentheses; all regressions use Huber–White correction for heteroskedasticity, allowing for clusteringby location-industry.The instruments are industry-location averages of BRIBE and TAX in specifications (1)–(4). In specification (5), industry-location averages of REGULATION and INFRASERV are added as additional instruments.F-test on instruments is the test statistic on the F-test of the joint significance of the instruments (BRIBE, TAX,REGULATION and INFRASERV) in the first-stage regressions, with p-values in braces. Hansen J-statistic is the teststatistic on the overidentification test of the instruments, with p-values in braces.⁎ Significant at the 10% level.⁎⁎ Significant at the 5% level.

72 R. Fisman, J. Svensson / Journal of Development Economics 83 (2007) 63–75



Is corruption harmful to growth?

REGULATION and INFRASERVas additional controls in the IV-specification. The coefficientson BRIBE and TAX become larger in absolute values, while both the extent of regulation andpublic service delivery enter insignificantly. In column (5), we instead add these two controlvariables as instruments. To the extent that REGULATION and INFRASERV have no direct effecton growth (as suggested in column 3) and since they influence to what extent firms are underbureaucratic control; i.e., the influence the public officials' bargaining power vis-à-vis the firms,they are valid instruments. Adding these additional instruments has the advantage that the model isnow overidentified and that the validity of the instruments can be tested. The instruments performwell. The F-statistic of their joint significance in the first-stage regression is 14.1 and is highlysignificant. The validity of the instruments (whether they are uncorrelated with the error process inEq. (5)) is tested, and the null hypothesis that the instruments satisfy the orthogonality conditionscannot be rejected. Importantly, the IV-estimates of BRIBE and TAX remain basically unchanged.These findings provide suggestive evidence in favor of our identifying assumption.

5. Conclusion

We have shown that there is a strong, robust, and negative relationship between bribery ratesand the short-run growth rates of Ugandan firms, and that the effect is much larger than theretarding effect of taxation. To our knowledge, this provides the first micro-level support for firm-based theories on the effects of corruption that have generated much attention in recent years.Much more work is still required in this area and while the results should be interpreted with caregiven the nature of the data and the problem of identifying causal effects, the evidence we havepresented and complementary, qualitative, information from firm managers, points in onedirection – corruption is a serious constraint in doing business.

The results of this paper also have significant policy implications. The donor community andother organizations have focused increasing attention on looking for ways to combat corruption indeveloping and transition countries. Our results suggest that such attention is justified by the data.Corruption significantly reduces firm growth, much more so than taxation.

Table 4Effect of bribery and taxation on growth, outliers excluded

Dependent variable: GROWTH

(1) (2)

Method OLS IVBRIBE −6.261⁎⁎ (2.973) −7.875⁎⁎ (3.728)TAX −0.314⁎ (0.171) −0.817⁎⁎ (0.388)LSALES95 −0.013 (0.009) −0.026 (0.015)LAGE −0.031 (0.026) −0.048 (0.028)FOREIGN 0.096⁎ (0.056) 0.136⁎ (0.071)TRADE 0.052 (0.044) 0.080⁎ (0.044)Cons 0.411⁎⁎ (0.161) 0.779⁎⁎ (0.284)R2 0.11Observations 114 119

Standard errors in parentheses; all regressions use Huber-White correction for heteroskedasticity, allowing for clusteringby location-industry.⁎ Significant at the 10% level.⁎⁎ Significant at the 5% level.

73R. Fisman, J. Svensson / Journal of Development Economics 83 (2007) 63–75



References

Main reference is Svensson (2005, JEP).

Fisman and Svensson (2007)“Are corruption and taxationreally harmful to growth? Firm level evidence”, Journal ofDevelopment Economics, 83(1):63-75.

Svensson (2003) “Who Must Pay Bribes and How Much?Evidence from a Cross Section of Firms”, Quarterly Journal ofEconomics, 118(1):207-30.

But also Olsen’s lecture notes.

Bardhan and Udry (1999) Development Microeconomics.

J.-J. Laffont chapter in Benerjee et al (2006) UnderstandingPoverty.


Notes: Corruption and Development - Faculty Support...

Documents

Transcript of Notes: Corruption and Development - Faculty Support...