Corruption and Competition*

Franklin Allen† Jun “QJ” Qian Lin Shen

Imperial College London Shanghai Advanced Institute of Finance Department

f.allen@imperial.ac.uk and Finance The Wharton School

University of Pennsylvania Shanghai Jiao Tong University University of Pennsylvania

allenf@wharton.upenn.edu jqian@saif.sjtu.edu.cn shenlin@wharton.upenn.edu

May 17, 2016

Abstract

An interesting aspect of corruption is that its damaging effects on economic performance seem to

differ significantly across countries. In large and regionally diverse countries such as China and

India, rampant corruption has not slowed down economic growth; in many African and South

American countries it seems that corrupt officials severely retard economic development. We

examine corruption associated with the provision of government services and goods. Local officials

can charge a fee to cover the cost of provision. Due to an agency problem, corruption occurs, and

local officials set higher than social optimal fees as bribes. Central government can mitigate the

agency problem by paying for performance financed by tax revenues. However, budget-constrained

central governments have limited power in controlling corruption with such payment schemes. One

possibility is to use the law to try to rule out corruption. However, such attempts often fail. We

argue that a different approach is to combat corruption by introducing competition between officials.

With multiple officials providing the same service or good, the fee is determined competitively, and

the pernicious effects of corruption are minimized. Moreover, the cost of implementing the optimal

payment scheme by the central government is also minimized. This theory is consistent with some

countries growing at fast rates despite corruption while others are severely damaged by it.

Keywords: Corruption, institutions, competition, taxes, user fee.

JEL Classifications: O0; H0; P5.

* We appreciate helpful comments from Viral Acharya, Tim Baldentius, Pierluigi Balduzzi, Hanming Fang, Frank

Gevurtz, Ning Gong, Ed Kane, Andrei Shleifer, Phil Strahan, Shyam Sunder, Yong Wang, Luigi Zingales and

seminar/session participants at Journal of Law Finance and Accounting Conference (Boston), Boston College, China

International Conference in Finance (Chengdu), and the Symposium on “Rethinking Corruption” at the McGeorge

School of Law, University of the Pacific. Jason Xu provided excellent research assistance. Data on country-level

corruption indexes provided by Martha Dye of Transparency International and financial support from Boston College

and the Wharton Financial Institutions Center are gratefully acknowledged. The authors are responsible for all

remaining errors.

† Corresponding author: Franklin Allen, Imperial College Business School, Tanaka Building, South Kensington Campus,

London SW7 2AZ, UK. E-mail: f.allen@imperial.ac.uk.

1

Corruption and Competition

May 17, 2016

Abstract

An interesting aspect of corruption is that its damaging effects on economic performance seem to

differ significantly across countries. In large and regionally diverse countries such as China and

India, rampant corruption has not slowed down economic growth; in many African and South

American countries it seems that corrupt officials severely retard economic development. We

examine corruption associated with the provision of government services and goods. Local officials

can charge a fee to cover the cost of provision. Due to an agency problem, corruption occurs, and

local officials set higher than social optimal fees as bribes. Central government can mitigate the

agency problem by paying for performance financed by tax revenues. However, budget-constrained

central governments have limited power in controlling corruption with such payment schemes. One

possibility is to use the law to try to rule out corruption. However, such attempts often fail. We

argue that a different approach is to combat corruption by introducing competition between officials.

With multiple officials providing the same service or good, the fee is determined competitively, and

the pernicious effects of corruption are minimized. Moreover, the cost of implementing the optimal

payment scheme by the central government is also minimized. This theory is consistent with some

countries growing at fast rates despite corruption while others are severely damaged by it.

Keywords: Corruption, institutions, competition, taxes, user fee.

JEL Classifications: O0; H0; P5.

2

I. Introduction

Corruption has many damaging effects in any economy, including the distortion of

incentives and misallocation of resources, and it occurs in various forms in almost all industries

(e.g., Shleifer and Vishny 1993, Khwaja and Mian 2005). In its worst form, corruption is essentially

outright theft and robbery done on a massive scale by high-ranking government officials. Corruption is

also a pervasive and persistent problem, and takes on many different forms.1 It occurs in many

countries, including some of the most developed ones, and seems very difficult to eradicate.

Despite ample evidence at the micro-level on the prevalence, scale and damaging effects of

corruption, researchers generally do not find a negative and significant relation between corruption,

measured by a number of indexes provided by international organizations (most notably the

Corruption Perception Index, or the CPI, published by the Transparency International), and

economic growth in cross-country studies (e.g., Mauro 1995; Svensson 2005). One reason for this

interesting and perhaps puzzling fact is that these indexes are imperfect measures for the multiple

facets of corruption (e.g., Olken 2009; Gutmann et al. 2015). Another reason is that the negative

effects of corruption on growth differ significantly across countries (e.g., Olken and Pande 2012).

For example, in countries like China and India, rampant corruption has not slowed down growth

(e.g., Allen et al. 2005, 2012); but in many African and South American countries, its effects seem

far more damaging (e.g., Fisman and Svensson 2007).

Why does corruption occur and why is it so difficult to eradicate? How to explain the cross-

country variations on the impact of corruption? In this paper, we address these questions by

focusing on corruption associated with the provision of government services and goods to the public

such as banking licenses, initial public offering approval for firms that want to go public, and many

1 Extrapolating from firm and household survey data, the World Bank estimates that total bribes paid to officials around

the world are about US$1 trillion per year, or 3% of world GDP (Rose-Ackerman 2004). For descriptions on the

severity of corruption in individual countries, see reports from IMF, World Bank and Transparency International.

3

other services such as provision of driver’s licenses. We first argue that corruption arises due to an

agency problem. Officials can charge fees to cover the cost of provision. However, the fees charged

cannot be perfectly monitored by the central government, therefore officials have incentive to take

advantage of their market power and charge higher than social optimal user fees. In this regard, we

define bribes as fees paid in excess of the cost of provision and corruption as officials receiving

bribes over and above the costs of supplying the socially optimal level of services and goods. The

central government can reduce corruption by paying bonuses to officials who increase the provision

of services and goods to firms and the public. However, the central government might not have

sufficient tax revenues to implement such costly salary schemes. Even if the government is not

budget-constrained, it would optimally tolerate moderate levels of corruption to save the cost of

such conditional salary payments. This is related to the idea of capitulation wages (e.g. Besley and

McLaren 1993, Gorodnichenko and Peter 2007), which basically means wages are so low that only

corrupt citizens would accept employment as government officials, and this is sometimes more

efficient than paying incentive compatible wages to achieve the social optimum. In the long run,

one way to curtail corruption is to raise more revenues directly through taxes and raise the

sensitivity of payment to performance for officials.

With insufficient tax revenues in the short run, however, can anything be done to minimize

the damaging effects of corruption? Using the legal system to make corruption illegal is one way to

try to prevent corruption. However, the rampant corruption in many countries shows the limitations

of this kind of approach. We argue that if services and goods are provided in a competitive

environment, so that multiple officials compete to provide the same service or good (e.g., issuing

licenses for banks), then the pernicious effects of corruption can be significantly reduced. In this

case, the bribe charged by an official is similar to a fee that is competitively determined (e.g.,

through Bertrand competition) and substitutes for taxes. As a result, the level of services and goods

4

provision can approach the first best (socially optimal) level. On the other hand, when a

government official is the monopolist in the provision of public goods or services, a much higher

user fee will be charged, leading to a lower level of provision of the goods or services and slower

economic growth.

To ensure that there is competition among officials, economic agents demanding the good or

services must be able to move from one provider to another (or credibly signal that they can do so)

and have the information regarding the user fees in other regions to exploit the differences in user

fees. In a more general and realistic setting with transaction costs (for switching), we show that an

imperfect Bertrand competition yields an output level that is below the first best level (with perfect

competition) but higher than that of monopoly, and that the cost of implementing the optimal salary

scheme is lower. Thus, the central government is less likely to be constrained by insufficient tax

revenues. An implication of this result is that economic agents can alleviate the damaging effects of

corruption through reducing their own transaction costs. For example, small firms are easier to

relocate than large firms; firms operate in industries that require small scales of investment in fixed

assets can also relocate at low costs. Therefore, the size and choice of assets makes it more difficult

for corrupt officials in any region to extract rents from firms.

The main policy implication from our analysis is that setting up the provision of government

services and goods in a competitive environment reduces the damaging effects of corruption and

promotes growth. In practice, anti-corruption legislations and strong punishment on corruption

have been proved to be inefficient in combating corruption in some countries/organizations (e.g.,

Olken and Pande 2012, Ogus 2004). One of the reasons is that corruption also occurs in law

enforcement and monitoring system, and corrupt officials can get away with it by bribing the

inspectors (e.g., Damania 2004). What we are proposing, introducing competition in the

organizational structure, is more effective in controlling corruption in the sense that it is incentive

5

compatible and therefore is immune from corruption in the enforcement system. One way to

implement this would be to decentralize the provision of services and goods to local government

officials or agencies. A potential problem with the delegation of the provision to local officials or

agencies is quality control (e.g., banking licenses to incompetent firms), and thus monitoring may

be necessary to ensure that quality is not compromised under decentralization. In large and

diversified countries, a competitive environment can also be created by encouraging inter-regional

economic activities and lowering the costs (and bureaucracy) of relocation across regions, while in

small and homogeneous countries this can be done by encouraging international trade and lowering

the costs of relocation across countries. Another policy implication to introduce competition and

reduce friction is through facilitating information revelation, specifically the fee charged by other

officials. For example, the central government can reward officials whose regions have shown the

fastest economic growth and attracted the largest amount of economic activities (and investment)

from non-local sources as in China. These policies encourage inter-regional competition among

officials and induce more government services and goods to be supplied. It may in circumstances

be better not to heavily penalize corruption to facilitate the provision of price information.

Our results are broadly consistent with evidence on corruption and growth and provide new

predictions. First, corruption is more likely to occur in countries whose governments have low

taxing ability and tax income – most of these countries are poor and developing countries. Second,

in large and regionally diverse countries such as China and India, there is a limited amount of bribes

officials can extort because of regional competition (e.g., Burgess et al 2012). In some African and

South American countries with homogeneity across regions (in part due to historical reasons from

colonial eras), corrupt officials effectively have monopoly power and this allows them to extract

much larger amounts of rents, which in turn significantly reduces the level of economic activities.

Third, consistent with the implications of our model, there is cross-country evidence on the negative

6

relation between “openness” of an economy (e.g., exports/GDP) and corruption, and a positive

relation between corruption and excessive regulation of entry of new firms (e.g., Djankov et al.

2002; Svensson 2005).

Fourth, our model shows that corruption has less damaging effects on industries and firms

that can be easily relocated. Consistent with this prediction, in both China and India, the sectors

that have witnessed the fastest growth are small- and medium-sized firms from services and labor-

and technology-intensive industries (e.g., Allen et al. 2005, 2012), while in Russia, an economy

dominated by large corporations in heavy manufacturing industries, firm-growth in a large number

of industries has lagged behind other countries. Further empirical studies can systematically

examine the impact of corruption on different industries across countries or different regions within

the same country.2 Finally, in China, one of the key yardsticks in measuring the performance of

local officials is economic growth and the amount of investment from non-local sources (e.g., Li

1998; Li and Zhou 2005); in other countries the most corrupt government officials hold the highest

positions in the government, making it impossible to implement incentive-aligning policies for

lower level officials.

Our paper contributes to the related strands of literature on political economy, institutions

and economic growth. Proponents of institutional development argue that a country’s institutions

restraining the government and powerful elites, determine the country’s long-run economic growth

(e.g., Rajan and Zingales 2003a, b; Acemoglu and Johnson 2005). Unlike this line of research and

most of the literature on corruption that focuses on its negative impact on growth, we show that,

because of competition among officials, lack of corruption is not a necessary condition for

economic growth. We also propose different degrees of competition (among officials) as the new

rationale to explain cross-country (and regional) differences of corruption and growth. In addition,

2 With survey data on firms from Uganda, Svensson (2003) finds that firms with less sunk costs (of capital stock) pay

smaller amount of bribes, consistent with our model prediction.

7

our model shows reducing corruption can be expensive and therefore depends on the budget of the

central government. As a result, during early phrases of economic development, policies aiming at

eliminating corruption by raising salaries of officials are not likely to succeed, while policies to

increase the degree of competition among officials in providing the same services and goods can be

much more effective in reducing the damaging effects of corruption.

The rest of the paper is organized as follows. In Section II, we review related work and

present empirical evidence on corruption, governments’ taxing ability and economic growth.

Section III presents a model of corruption based on standard models of industrial organization and

market power. Section IV discusses model implications and limitations, and Section V concludes.

II. Related Work and Facts about Corruption

The ideas that corruption does not necessarily slow down economic growth, and conceivably

certain forms of “efficient corruption” can actually promote growth, can be traced back to at least

Leff (1964) and Huntington (1968). These authors argue that corruption is more likely to occur

during periods of rapid economic and social development and modernization. They also argue that

through bribing corrupt officials economic agents can bypass bureaucracy and bad, rigid laws and

regulations and expedite activities that lead to faster economic growth. Subsequently, researchers

formalized these ideas by modeling competition among economic agents (e.g., banks applying for

licenses) in submitting bribes to officials (e.g., Lui 1985; Bliss and Di Tella 1997; Ades and Di

Tella 1999). However, this type of competition may not be enough to reduce the damaging effects

of corruption, because corrupt officials may favor more distortions and secrecy than granting the

license to the most efficient agent as argued by Shleifer and Vishny (1993).3 In our model, we take

the competition among agents as given and focus on competition among officials in providing the

3 Most of the literature focuses on the negative impact of corruption on economic development. See, for example,

Svensson (2005) for a review of the literature.

8

same service or good, and show that this latter form of competition is more important in terms of

reducing the damaging effects of corruption.

While Shleifer and Vishny (1993) focus on reasons why corruption can have a significantly

negative impact on economic development, we focus on how competition among officials can

minimize these damaging effects. Like their paper, we show that the ‘industrial organization’ of the

provision of government goods and services is an important determinant of the effects of corruption.

Unlike their paper, we argue that the fundamental reason for the occurrence of corruption is an

agency problem, while the government’s inability to collect sufficient taxes to compensate the

officials adequately leads to the prevalence of corruption. Competition among officials can yield the

same outcome as the first best situation when the government has sufficient tax income. We also

argue that competition among officials is a robust mechanism in dealing with corruption, even if the

central government cannot monitor the process through which the services or goods are provided by

local officials (e.g., officials can lie about the quantity of services sold or the amount of money they

have collected).

There is a strand of literature examining how the central government can control corrupt

officials at local levels providing goods and services. Following Becker and Stigler (1974), most of

the literature regards the relation between central and local government officials in the framework

of a principal-agent relationship, and study ways to motivate the ‘agent’ (local officials) to carry out

their tasks congruent with the goals of the ‘principal’ (central government) that is to promote

economic growth. We also have the principal-agent problem in our model. In addition, we focus on

what policies and performance evaluation measures the central government can come up with to

motivate competition among officials. We argue that these policies and salary contract paid by

central government to local officials are both of first order importance, and they substitute each

9

other in the sense that when competition among officials is high the optimal payment made by the

central government is low.

Our work is also closely related to the literature on the structure of the bureaucracy. Tiebout

(1961) argues that decentralization enhances efficiency because of the informational advantage and

tailored regulation by local officials. Albornoz and Cabrales (2013) argued and provided empirical

evidence that decentralization reduces (aggravates) corruption for sufficiently high (low) levels of

political competition. Our model predicts that increasing competition at the same level always

reduce corruption. Djankov et al (2010) and Banerjee et al (2011) showed that transparency helps

curtail corruptions by allowing the public to monitor government officials. Our model also features

the positive effect of transparency in controlling corruption, however through facilitating the

comparison of user fees and thus mobility across regions.

Finally, there is also a literature on institutions and long-run economic growth. For example,

Rajan and Zingales (2003a, b) and Acemoglu and Johnson (2005) argue that institutions restraining

the government and the powerful elites promote economic growth. In this regard, the evidence in

China and India provides important counterexamples to this literature, in that despite powerful elites

and corruption these countries have achieved impressive growth records. In this regard, our model

provides a new rationale on why lack of corruption is not a necessary condition for economic

growth among emerging economies.

Insert Tables 1 and 2 here.

Table 1 lists the largest twenty economies in the world as of 2014 using both (unadjusted)

simple exchange rates and those that are based on purchasing power parity (PPP); the table also lists

the top twenty economies that have the highest growth rates (GDP and per capita GDP) during the

period 1990-2014. As shown in Table 3-B, some of the most corrupt countries (measured by the

CPI and highlighted in bold) are also among the largest and fast growing economies, led by China

10

and India. These facts reinforce our research goal that understanding how corruption affects

economic growth differently across countries is an important question for the world economy and

development. Table 2 presents the results of Bribe Payers’ Index (BPI) in 2011 and 2008,

published by the Transparency International. The BPI is based on the responses of 3,016 in 2011

(2,742 in 2008) business executives worldwide to questions about the propensity of foreign firms

that do the most business in their country to pay bribes or to make undocumented extra payments.

The answers are converted to a score between 0 (bribes most frequent) to 10 (bribes never occur),

and the ranking of 28 in 2011 (22 in 2008) of the leading countries, including all G20 countries,

listed shows the average score of a country. Consistent with the rankings of CPI, firms from the

three largest emerging countries (using PPP), China, India, and Russia, have the lowest BPI

rankings or the highest propensity of paying bribes.

Tables 3-A and 3-B compare government’s taxing ability, measured by total taxes collected

over GDP, economic growth and income level, measured by both GDP growth and per capita GDP,

across three subgroups of countries ranked by corruption. As mentioned above, the corruption

measure we adopted is the CPIs of Transparency International, which provides the most

comprehensive (both in terms of number of countries covered and time horizon; using other

corruption measures produced very similar results; see, e.g., Svensson 2005). We divided all

countries into two groups according to their average CPIs from 1995 to 2014. The more (less)

corrupt countries have CPIs above (below) the sample median. Consistent with prior research,4

countries in the more (less) corrupt group are on average poor (rich) and have low (high) taxing

abilities proxied by a smaller (larger) fraction of total taxes collected by the government of GDP.

These results motivate our model assumptions that the reason corruption alleviates is because the

4 There is a strand of literature documenting the scale of shadow economies around the world (e.g., Friedman et al. 2000;

Schneider and Enste 2000, 2002). Researchers find that the size of the shadow economy is larger (relative to GDP) in

poor countries, especially where the tax rates are high, and that there is a positive correlation between size of shadow

economy and corruption.

11

government can raise enough tax revenues to pay the government officials for providing goods and

services at socially desirable levels.

Tables 3-A and 3-B also show that while the average GDP growth rate (during the 1981-

2014 period) for the more corrupt group is higher than that of the less corrupt group, this difference

is small; perhaps not surprisingly, the standard deviation of growth rates is also higher in the most

corrupt group than for the least corrupt group. In most of the regression analysis (e.g., Mauro 1995;

Svensson 2005), researchers generally do not find a statistically significant relation between

corruption and GDP growth after controlling for other standard growth factors (such as human

capital, initial growth conditions, etc). While this puzzling fact may be explained by measurement

errors of corruption, we interpret it as evidence showing that the forms and damaging effects of

corruption vary significantly across countries. The main purpose of our model is to provide a

rationale for this interpretation and draw policy implications.

Table 3-B furthers our interpretation of the lack of significant relation between corruption

and growth. Panel A compares government’s taxing ability and economic growth (measured by

both GDP growth and per capita GDP) across the largest more corrupt countries in the world, while

Panel B lists the largest less corrupt countries in the world. There are significant variations in terms

of economic growth among the countries with lowest scores on corruption in Panel A. China

appears to be the most significant ‘outlier’ in the group: While its corruption index/score (based on

Transparency International’s CPI index, ranges from 1 to 10 with 1 being the most corrupt) had

improved from 2.2 in 1995 to 3.6 in 2014, its annual GDP growth rates (using PPP measures) of

almost 13% during the period 1981-2014 is the highest in the world; in PPP terms it currently has

the largest economy in the world; and per capita GDP increased from US$343 in 1981 by 37 fold to

almost $12,880 in 2014. Similar (but less spectacular) cases are other large and regionally

diversified countries such as India and Indonesia, in that despite rampant corruption, these countries

12

have done quite well in terms of economic growth. At the other extreme, many countries in Africa

and South America with similar scores on corruption have basically stagnated and there is nearly no

growth in sight. Finally, a large and regionally diversified country with rampant government

corruption, Russia, has done poorly in terms of economic growth during the same time period. Our

model is consistent with these facts about corruption and economic growth across countries.

III. A Simple Model of Government Goods and Services Provision and Corruption

We develop the simplest possible model for conveying our main message that competition

among officials reduces the damaging effects of corruption. We consider a central government’s

problem of providing a service (or good) to the public when there is an agency problem. We

analyze how this can be done with different structures of supplying the good and the efficiency of

different structures. We first examine the first best case without an agency problem. Next, we

analyze different cases where the government uses tax revenue as a resource of payment to local

official(s) to motivate the provision of the government service.

III.1 Elements of the Model and First Best Solution

We assume that the government service is homogeneous, in that there is no difference in

quality provided by different officials in different regions (of the same country or different

countries). We also assume that the demand for this good among economic agents from different

regions can be characterized by a downward sloping demand function. For simplicity, we assume

this function to be linear, and takes the form

P(Q) = W – aQ,

where Q is the equilibrium level of the good supplied in the economy, W indicates highest possible

consumer surplus (agent obtaining the good with price 0), and the parameter a ( > 0 ), indicates the

13

sensitivity of quantity demanded when price changes.5 We further assume that the provision of the

government service is costly, and the cost is linear in the quantity supplied, i.e.

cQQC ,

where the constant c is the marginal cost for producing and supplying each additional unit. Local

government officials can charge fees to cover the cost of provision. The First Best case is achieved

when the fees charged by local government officials are observable.

With our model setup, it is straightforward to derive the First Best level of this government

service, Q*. This is achieved by increasing the supply of the service until the marginal cost of

supplying the last unit equals the marginal benefit of the buyer of the final unit, or

MC(Q)= c = P(Q). (1)

Solving for Q in (1) above gives

a

cWQ FB

,

and the associated user fee charged is

cPFB .

From standard demand theory we know that at this output level there is no deadweight loss and

social welfare is maximized. The First Best case is summarized in the following result.

Proposition 1 Without an agency problem, the government official(s) will charge PFB to supply QFB

units of the service in the economy.

5 Our assumptions on the demand function for the government service can be restrictive for two reasons. First, the

consumption or use of some services by one agent can have a positive (or negative) externality on other agents. This is

not captured by the above demand function. Second, one agent may need to purchase multiple government services and

goods in order to undertake her own activity, in which case the joint determination of the provision of all the services

and goods becomes important.

14

III.2 Agency Problem: Monopoly vs. Perfect Bertrand Competition

Now suppose that there is an agency problem, and the local officials can potentially charge

higher than socially optimal fees to maximize their own payoffs. We assume that the monitoring of

officials’ fundraising and the provision of the good is imprecise, so that the supervisor of the

officials providing the service cannot differentiate the reasons why the total quantity supplied in the

economy is below QFB: This can be induced by officials charging excessive fees and thus reducing

demand, or because the officials lie about the amount that can be raised in the economy to cover the

production/supply costs (this can be due to the fact that there is some uncertainly in the economy

about how much funds can be raised for different purposes). In other words, the central government

cannot observe or derive the fee charged from the demand function. The lack of effective

monitoring implies the officials providing the service have the discretion to restrict the quantity that

is supplied so as to maximize their own benefits. This is our definition of corruption, with the user

fees charged in excess of PFB by officials corresponding to bribes. Although the central government

cannot monitor the local official(s), it may be able to mitigate the problem through a salary contract

as a function of the quantity of government good (or service) provided. For simplicity, we restrict

attention to the linear contract6

T(Q)=tQ+s.

We believe this linear payment contract captures the basic idea of pay for performance with the

simplest form. In practice, government officials may not directly receive pecuniary rewards for

good performance; however, performance is usually rewarded indirectly via promotion. For

example, in China the likelihood of promotion of local officials increases in their performance

including local economic growth and the amount of investment from non-local sources (e.g., Li

6 Holmstrom and Milgrom (1987) showed that linear contracts are more robust to agent manipulation.

15

1998; Li and Zhou 2005). As long as the reward scheme increases in performance, the general result

that the central government can mitigate corruption by adopting such a reward scheme goes through.

Next, we examine different structures of the government services market, focusing on the

supply side while taking as given the demand for the service. We first examine the case where the

provision of the service is carried out by one official so that he has monopoly power. This case can

be associated with the scenario that in one region there is only one government branch/office that

the service can be obtained. In countries with multiple regions where the same service is supplied,

this case corresponds to the situation in which it is impossible for agents from one region to move to

another, so that the official in each region is a monopolist in providing the service.

If the central government is budget constrained and cannot allocate any resource to pay the

local officials, given the monopoly and assuming that the demand function is relatively inelastic

(a >> 0), the official has an incentive to cut back the quantity supplied, raise the fee (assuming he

cannot price discriminate between buyers of the service so he can only set one fee for all units) and

maximize his total fees from providing the good. To do this, the monopolist official supplies the

government service until the marginal cost of supplying the last unit equates that of his marginal

revenue of selling the unit, which is below marginal benefit of the buyer of the final unit:

aQW

dQQQdP

QMRcQMC 2

. (2)

Solving for Q in (2) yields

a

cWQM

2

,

and the corresponding excess user fee for each unit of the government good charged by the

monopolist official is given by:

2

cWPQPf FBMM

. (3)

16

Again, from standard industrial organization and market structure models, we know that the output

level under monopoly, QM, is strictly below the First Best level, QFB, derived in (1). The

comparison of output levels under different structures are illustrated in Figure 1, which also depicts

the profits earned by the monopolist official as well as the deadweight loss generated by the less-

than-optimal level of provision of the government good.

Insert Figure 1 here.

The profits earned by the official, net of the costs of supplying QM to the agents in the

economy, is indicated by the shaded area in Figure 1, while the deadweight loss is indicated by the

dark triangle. The size of the loss, indicated by the shaded area of the triangle, is proportional to the

product of the differences in output levels (QFB – QM) and the profit margin (P(QM) –c)):

a

cWcQPQQDWLoss MMFBM

8

)(

2

1 2 , (4)

where the last equality is derived by plugging in the expressions for Q* and QM.

If the central government has enough tax revenues to pay the local official to increase the

provision of government good, and the payment contract {t, s} is

0

)(stQ

QTif

ifM

M

QQ

QQ

The objective of the local official now is to maximize the profit from fees plus compensation:

)()(max QTcQQQPQ

, (5)

Solving for Q in (5) yields

MQa

tcWtsQ ,

2max),( . As long as t>0, the monopolistic official

is willing to increase the quantity of the service supplied, and therefore the government can

successfully mitigate the agency problem by offering payment with high sensitivity to quantity.

However, it is also costly for the central government to adopt such a payment scheme. To model the

17

cost, we assume that the total social welfare created consists two parts: that from supplying the

specific government good (service) and that from other government activities. A central government

can make use of its tax revenues of amount g to create social welfare v(g) through other government

activities, where 0)( gv and 0)( gv for all 0g . The cost of payment scheme can be

interpreted as the opportunity cost of reducing welfare through other government activities. And a

negative second order derivative of v(g) indicates that the marginal cost of implementing the

payment scheme is lower for “wealthier” central government. The average welfare created per unit

of the government good (service) is ))((2

1cQPcW , thus the total welfare created from

supplying the government good (service) is QcQPcW ))((2

1 . If the central government

allocates T(Q) for the payment scheme, the welfare created from other activities is then v(g-T(Q)).

Therefore the optimal contract to offer for a government with total budget g can be calculated as

follow,

)())((2

1max

,QTgvQcQPcW

st (6)

subject to

MQa

tcWQ ,

2max (7)

0)( MQT (8)

The solution (see proof in appendix A.1) to the central government’s problem {tM, sM} is such that

4

1

42

2

M

M

t

cW

a

tgv (9)

MM ta

cWs

2

. (10)

The corresponding quantity supplied and excess user fee charged are

18

),(2

ˆ FBMM

M QQa

tcWQ

(11)

),0(2

ˆ MM

M ftcW

f

. (12)

respectively. The central government needs to allocate

a

tQT

MMM

2)ˆ(

2

(13)

to pay the monopolistic local official, and the total social welfare loss compared to the first best

case is

),0()2

()(8

)(22

MMM

M

T DWLossa

tgvgv

a

tcWDWLoss

. (14)

Note that even if the government has enough resources, the quantity supplied under the payment

scheme is strictly less than the first-best level. In other words, even if it is able to eliminate

corruption completely the central government chooses not to. This is because the marginal cost of

inducing the official to increase the supply of government good increases. It is simply too costly to

eradicate corruption, and the central government is better off to live with moderate corruption and

save the tax revenue for other purposes. This result may be qualitatively sensitive to the structure of

payment scheme, but the general intuition of increasing marginal cost of combating corruption

should hold as long as the limited liability constraint (8) is imposed. Proposition 2 (see proof in

appendix A.2) below summarizes the results for the monopolist case.

Proposition 2 With a monopolist official providing the government good, the central government

optimally offers payment contract },{ MM st as in equation (9) and (10), and the excess user fee

charged for each unit is set at ),0(ˆ MM ff as in equation (12), resulting in a total of

),(ˆ FBMM QQQ units supplied in the economy as in equation (11). The total payment to the

19

monopolistic official )ˆ( MM QT as in equation (13) increases when: i) W increases; ii) c decreases;

iii) a decreases and iv) g increases. The deadweight loss of the monopolist

supplier ),0( MM

T DWLossDWLoss as in equation (14) increases when: i) W increases; ii) c

decreases; iii) a decreases; and iv) g decreases.

The comparative statics on the total payment of the incentive scheme )ˆ( MM QT are intuitive.

A higher W and (or) a lower c indicates higher consumer surplus from consuming the government

good, and hence more incentive for the central government to reduce corruption and increase the

quantity supplied. Therefore, the central government would allocate more resource to the incentive

scheme to control corruption when W increases and (or) c decreases. A lower a, the slope of the

demand curve implies that consumers are more reactive to price changes. If the corrupt government

official reduces the bribe it charges, the quantity demanded would increase a lot, and the social

welfare would be restored a lot. In other words, the incentive scheme is more valuable and the

central government is willing to allocate more resources to implement it. Lastly, for a government

with more tax revenues, the same incentive scheme is relatively cheaper due to the concavity of v(g),

and therefore the government can afford to pay for a larger incentive scheme.

The comparative statics on the size of the deadweight loss M

TDWLoss are generally the same

with those on the total payment )ˆ( MM QT with the exception of tax revenues g. The general

intuition is that when it is more efficient to implement the incentive scheme which leads to higher

)ˆ( MM QT , there is also more room for profits for the corrupt official and hence larger M

TDWLoss .

Specifically, a higher W and (or) a lower c indicates higher consumer surplus and therefore more

profits for the corrupt official leading to a larger deadweight loss. A lower a implying higher

elasticity of demand to price, which gives the corrupt official more bargaining power in abusing its

20

monopoly power and turn it into profits. Thus, M

TDWLoss decreases in a. Finally, although a

“wealthier” government spend more tax revenue on the incentive scheme, the benefit of reducing

corruption overweighs the cost. To see why this is the case, consider two governments with

different budgets. Suppose the deadweight loss generated under optimal incentive scheme for the

resource-constrained government is D1. If the resourceful government were to implement the same

incentive scheme, the resulting deadweight loss D2 must be smaller. Yet the resourceful government

would optimally choose a larger incentive scheme leading to deadweight loss D3. Since D3 is

generated under optimal choice, it must be smaller than D2, therefore smaller than D1.

Next suppose that the government sets up two suppliers for provision of the service (results

would be very similar if there are more than two suppliers), and all consumers are perfectly

informed of the fees charged by both suppliers. This can occur if in any region, where the good is

provided, there are two government branches/offices from which agents can purchase the service.

This can also occur if agents can move from one region to another (to exploit differences in fees)

without any costs. In either case there is competition among officials. Since in most cases the

officials set and announce the user fees to all potential buyers, we assume that there is Bertrand

competition for the same service.7

Given that there are no costs in resetting prices and that the pricing information becomes

immediately known to all agents in the economy, similar to standard one-period Bertrand

competition models, there is a unique Nash equilibrium in this case, where each supplier will set the

user fee to be the marginal cost of supplying the last unit. But this is exactly the same condition in

the First Best, indicated in (1). Hence the total units supplied in the economy is the First Best level,

Q* again. Since each official is earning zero profits under Bertrand competition, it does not matter

7 Alternatively, government officials may choose the quantity to sell and the same (per unit) fee for the service is

determined in the market place. In this case we have Cournot competition and the output level is higher than the

monopoly case but lower than the Bertrand competition case.

21

how they split the total supply between them; without loss of generality we assume each official

provides exactly half of Q*. Hence Bertrand competition generates outcomes that are identical to

the First Best solution in Figure 1, and the equilibrium fee can be found by equating the marginal

cost function to the demand curve. The key for the above results is that, if the fee set by one of the

officials is higher than marginal cost of producing the last unit, by undercutting the fee by the

other official can capture the entire market (recall that all agents can move to another region without

costs) and maximize profits, in the process of doing so also driving the demand for the official with

the higher user fee to 0. In this case, competition also completely eliminates the agency problem,

therefore the central government doesn’t need to pay the local officials to align their incentives with

the social optimum.

Proposition 3 With two or more officials providing the same government service, the First Best is

restored, and the unique equilibrium excess user fee for each unit is set at 0Bf , resulting in a

total of QB =QFB units supplied.

Bertrand competition not only yields the First Best output level for the provision of the government

good, it also requires almost no monitoring of the officials. This is because the market mechanism

of competition is self-enforcing due to each official’s incentive to undercut the user fee set by other

officials and capture the entire market demand for the same good. However, all the analyses in this

paper rely on the assumption that the officials cannot collude on reducing supplies to achieve higher

profits.

III.3 Agency Problem: Imperfect Bertrand Competition with Transaction Costs

22

In the previous subsection we considered two extreme cases of how the government good is

provided. More generally, the industrial organization structure in the provision of the service will

often lie between monopoly and perfect Bertrand competition. In this subsection, we consider

imperfect Bertrand competition, in that the agents from each region can move to another region at a

constant cost K (we assume that 2

cWK

)8. The cost K can also be interpreted as the cost of

acquiring information about the fee charged by the non-local government official. Since now agents

cannot move from region to region to explore differences in user fees for free, each of the two

officials has some monopoly power in setting the user fee. In particular, each official knows that, in

order to capture the entire market of demand for the same government good, he must undercut the

other’s user fee by at least K, as any smaller cut would not attract any agents to move from the high-

fee region to his region. On the other hand, if there is room to cut the user fee by K and still have a

positive profit margin, then an official will indeed cut the fee by K.

Given the above intuition, it is then straightforward to show that without the compensation

scheme offered by the central government, the equilibrium pricing strategy for each local official is

to set the fee K above the marginal cost of providing the last unit:

MC(Q) + K = P(Q), (15)

which yields a lower total quantity supplied (with each of the two officials supplying half of the

total),

a

KcWQB

K

,

than that under perfect Bertrand competition (QB), and a higher excess user fee

Kf B

K ,

8 If 2

cWK

, it is too costly for an agent to move to another region, and we get back to the monopoly case.

23

than that under perfect Bertrand competition (fB). Note that it is implicitly assumed that each local

official can charge different user fees for local and nonlocal agents. Specifically, the excess fee

charged for nonlocal (local) agents is 0 ( Kf B

K ), so that all agents, local or nonlocal, are all

indifferent between going to any of the two officials for the public good. If we assume the cost of

moving to another region K is a deadweight loss, then all agents going to the local official is

socially optimal. The equilibrium fee and quantity (relative to the perfect Bertrand competition) are

illustrated in Figure 2. The current structure of government good provision will also result in a

deadweight loss (shaded area), however smaller than the monopoly case:

a

KKQQDWLoss B

K

FBB

K22

1 2

. (16)

Insert Figure 2 here.

Now we explore the central government’s optimal payment scheme under imperfect

Bertrand competition. Since the severity of the agency problem lies in between the monopoly and

the perfect competition, intuitively, the payment contract should also lie in between the two cases.

Below is the formal characterization of the central government’s problem:

))(())((2

1max

,QTgvcQPcWQ

st (17)

subject to

B

KQa

tcWQ ,

2max (18)

0)( B

KQT (19)

When KK , the imperfect Bertrand competition outcomes are already close to the First Best case.

The bonus required to incentivize the local government officials to cut their profit from corruption

further is so costly that the central government would rather leave the moderate level of corruption

as it is. To see it clearly, the quantity supplied with the incentive scheme is

24

a

KcW

a

tcWQB

K ,2

maxˆ . Hence a effective payment scheme needs to have pay-for-

performance a

KcWt

2

2 which decreases in K . When the friction is relatively small, t needs to

be high to induce the corrupt officials to reduce their bribe charges further and therefore the total

payment of the incentive scheme would be large. Therefore, the equilibrium excess fee charged

B

K

B

K ff ˆ , the quantity supplied B

K

B

K QQ ˆ , the total payment to local officials 0)ˆ( B

K

B

K QT and the

deadweight loss B

K

B

TK DWLossDWLoss , . Proposition 4 (see proof in Appendix A.3) below

summarizes the results for the imperfect Bertrand competition case with low transaction cost.

Proposition 4 With imperfect Bertrand competition and transaction cost ],0( KK , the central

government optimally choose to offer zero payment contract.

When )(21

)()(

gv

gvcWKK

, the optimal salary contract is B

K

B

K st , is such that

)22(22

2

cWKt

tcW

a

cWKttgv

B

K

B

K

B

K

B

K

(20)

B

K

B

K

B

K Qts . (21)

The equilibrium excess user fee charged is

)ˆ,0(2

ˆ MB

KB

K ftcW

f

, (22)

resulting in the quantity of public good supplied

),ˆ(2

ˆ FBMB

KB

K QQa

tcWQ

. (23)

The central government needs to allocate

25

)ˆ(,02

2)ˆ( M

B

K

B

KB

K QTa

cWKttQT

(24)

in total to pay the local officials, and the total social welfare loss compared to the first best case is

M

T

B

K

B

K

B

KB

TK DWLossa

cWKttgvgv

a

tcWDWLoss ,0

2

2)(

8

)( 2

,

. (25)

Note that under the optimal incentive scheme, the imperfect Bertrand competition case is always

closer to the first best than the monopoly case, i.e. MB

K ff ˆˆ , MB

K QQ ˆˆ and

M

T

B

TK DWLossDWLoss , . Proposition 5 (see proof in Appendix A.3 and A.4) below summarizes the

results for the imperfect Bertrand competition case with high transaction cost.

Proposition 5 With imperfect Bertrand competition and transaction cost )2

,(cW

KK

, the

central government optimally offers payment contract },{ B

K

B

K st as in equation (20) and (21), and the

excess user fee charged for each unit is set at )ˆ,0(ˆ MB

K ff as in equation (22), resulting in a total

of ),ˆ(ˆ FBMB

K QQQ units supplied in the economy as in equation (23). The total payment to the

monopolistic official )ˆ( B

K

B

K QT as in equation (24) increases when: i) a decreases; ii) g increases;

and iii) K increases. The deadweight loss of the monopolist supplier ),0(,

M

T

B

TK DWLossDWLoss as

in equation (25) increases when: i) a decreases; ii) g decreases; and iii) K increases.

The comparative statics on the total payment of the incentive scheme )ˆ( MM QT are intuitive.

Similar to the monopoly case, a lower a implies a more elastic demand and also a more valuable

incentive scheme (a small reduction in bribe induces a large increase in demand), therefore the

central government is willing to allocate more resources to implement it. For a government with

26

more tax revenues, the same incentive scheme is relatively cheaper due to the concavity of v(g), and

therefore the government can afford to pay for a larger incentive scheme. Lastly, the total payment

decreases in friction K. The intuition is that when the friction is relatively small, it is very costly to

induce the corrupt officials to reduce their bribe charges further and therefore the government

would not spend much on the incentive scheme. Since the objective welfare function and the

constraints are all continuous in K, this intuition also explains that when KK , the optimal

incentive scheme payment reduces to zero.

The comparative statics on the size of the deadweight loss M

TDWLoss are also intuitive. A

lower a implying higher elasticity of demand to price, which gives the corrupt official more

monopolistic profits. Thus, M

TDWLoss decreases in a. A government with sufficient tax revenues

(high g), has more power and flexibility in controlling corruption, and therefore the resulting

deadweight loss is small. Finally, the deadweight loss increases in the transaction cost K, because

the corrupt officials have more monopoly power if the friction K increases. In addition, as K

increases, the imperfect Bertrand competition case approaches the monopoly case explained in

Proposition 2 (and if 2

cWK

these two cases are identical).

Note that competition helps alleviate the agency problem in two ways. First, as the friction

in competition K decreases, the power that the officials have in extracting surplus from the agents

decreases directly, as evidenced by a

KcWQB

K

decreasing in K. Secondly, as the friction in

competition K diminishes, the payment that the central government makes decreases holding

everything else equal. From another perspective, the competitiveness in the supply side of the

service and the compensation contract can be thought of as substitutes in reducing corruption.

Lowering competition friction, on one hand, reduces corruption, while on the other hand, making

27

incentive scheme less effective and more costly to implement. In terms of policy implication, if a

central government has sufficient tax revenues, it has more tools and flexibility in controlling

corruption. However, if a central government has limited resources to allocate on salary payment,

the central government can instead work on reducing the frictions in competition among local

officials.

III.4 Multiple Government Goods

In this section, we compare different forms of government organizations when there are

multiple government goods supplied. In particular, if the government goods (service) supplied are

substitutes or complements, corruption on one of the goods might have externalities on the others.

Therefore, corrupt officials might have the incentive to jointly make the decision on bribe charges

and split the profits. In this situation, ensuring coordination among various officials is perhaps as

important as creating a competitive environment to reduce user fees set by each official. We show

that such organized corruption generates higher profits for corrupt officials than the case where they

charge bribes independently. We also show that the central government can reduce welfare loss by

altering organizational structure with the interdependency of different government goods.

Specifically, the central government can effectively combat corruption by introducing competition

when the two goods are substitutes and introducing cooperation when the two goods are

complements. Nonetheless, the results that increasing regional competition and introducing

incentive scheme are effective in reducing corruption still hold.

For simplicity, we assume there are two symmetric government goods with demand function

2

1

2

1

P

P

Q

Q

ab

ba

W

W.

28

The elasticity of demand can be shown to be22 ba

a

dP

dQ

i

i

and

22 ba

b

dP

dQ

i

i

. We assume

that ba to ensure that the price change of one good has more effect on the demand of such good

than the price change of the other good. Since 22 ba

b

dP

dQ

i

i

, the two goods are complements if b

< 0; they are substitutes if b > 0; and they are independent if b = 0, in which case each good can be

analyzed separately as in the main body. The cost of supplying one unit of good 1 and the cost of

supplying one unit of good 2 are both Wc ,0 . First, we consider the case where two cooperative

corrupt officials set 1P and 2P to maximize their total profits. The maximization problem is as follow,

cbQaQWQcbQaQWQQQ

122211, 21

max .

The solution to the maximization problem, i.e. the quantity supplied under monopolistic organized

corruption, is )(2

21ba

cWQQQ CCC

. And the resulting bribes charged, i.e. the price paid in

excess of the cost of supplying one unit of government good, is 2

21

cWfff CCC

. While in

the first best case, 021 FBFBFB fff and ba

cWQQQ FBFBFB

21 . Since the total welfare

created in the first best case changes in b, to analyze the effect of b on welfare changes, we

introduce a new measure, the percentage deadweight lossFBWelfare

DWLossDWL % . Following similar

calculation as in equation 4, it can be shown that the deadweight loss as a percentage of the total

welfare

%25

)(2

1)(

2

12

1

2

1

%

21

222111

cWQcWQ

fQQfQQ

DWLFBFB

CCFBCCFB

C .

29

Note that although the absolute deadweight loss increases in the complementarity between two

goods, the percentage deadweight loss remains constant.

To demonstrate the externality of corruption on good to the other, we compare it to the case

where there are two non-cooperative corrupt officials each supplying one of the two goods. The

maximization problem of official i is then

cbQaQWQ iiQi

i-max .

The best response of one official to the other can be solved from the first order condition, and it can

be shown to bea

bQcWQQ i

i

BR

i2

)(

. By solving the fixed point in the best response mapping,

we get a unique equilibrium in which ba

cWQQQ NCNCNC

221 ,

ba

cWafff NCNCNC

2

)(21

and 2

2

)2(%

ba

aDWLNC

. Note that the percentage deadweight loss increases in the

complementarity between two goods. In addition, when the two goods are complements (b < 0),

CNC DWLDWL %% ; when the two goods are substitutes (b > 0), CNC DWLDWL %% ; when the

two goods are independent, CNC DWLDWL %% . The intuition is as follow. If the two goods are

complements, raising bribe charge on one good reduces the demand for the other good. In the

cooperative case, the corrupt officials endogenize this effect and therefore charges lower bribes than

in the non-cooperative case, where they don’t endogenize this effect. Similarly, if the two goods are

substitutes, raising bribe charge on one good increases the demand for the other good, and therefore

the cooperative officials charge higher bribes than non-cooperative officials. Although the social

welfare loss is affected by the interdependency of the two goods, the profit made by the two

officials in the cooperative case

)(22

2

ba

cWfQ CC

is always higher (strictly higher when 0b )

30

than that in the non-cooperative case

2

2

)2(

22

ba

cWafQ NCNC

, which implies that the corrupt

officials always have the incentive to engage in this type of organized corruption. Therefore, when

the two goods are complements, the central government can effectively reduce corruption by

encouraging cooperation between the two officials, which is also incentive compatible. While when

the two goods are substitutes, the central government would prefer local officials to make decisions

competitively. However, since the corrupt officials always have the incentive to engage in

cooperative corruption, the central government might need to introduce incentive scheme such as

relative performance to increase competition among local government officials.

Regardless, our results in the one good case that both competition and pay-for-performance

scheme are effective in combating corruption still hold in the multiple goods case. The

corresponding competition under the two goods context is that there exists at least two regions each

offering both goods, and consumers in the two regions can pay some transaction cost K to “buy” the

goods in non-local regions. Therefore, by nature the goods offered by different regions are

substitutes. As analyzed before, by introducing competition among different regions, the central

government can effectively reduce welfare loss.

IV. Discussion

The premise of our model is that the reason corruption occurs is an agency problem, and the

government can use a compensation contract to mitigate the problem and reduce corruption. Our

model implies that another and perhaps more effective method to reduce the pernicious effects of

corruption is setting up the provision of government services so that there is competition.

Corruption, in this case, is a legitimate way of collecting user fees to cover the cost of provision of

the service. In the ideal case of perfect Bertrand competition (without transaction costs), corrupt

31

officials (by charging a fee equal to the cost of provision) do not have any distortionary effects, and

the central government does not need to pay bonuses to the officials. In more general cases with

imperfect Bertrand competition while the outcome is not First Best, the competition among officials

still yields a better outcome than having a monopolist official providing the service, which is the

worst case.

An important reason for the damaging effects of corruption is the secrecy of corrupt

officials’ activities (including fees charged and bribes received), leading to distortions of incentives

and resource allocations as stressed by Shleifer and Vishny (1993). Having a competitive

environment for the provision of government goods and services lessens this problem, because each

official providing the services has the incentive to disclose his fee structure for granting the services

so as to attract more clients. In Asian countries such as China, there are usually publicly known

‘market’ prices for government services or goods, and with such transparency bribes can ‘grease the

wheels’ and improve the process of providing these services and goods.

The results of our model are broadly consistent with evidence on corruption and growth.

First, consistent with existing research, corruption is more likely to occur in countries whose

governments have low taxing ability and tax income, most of which are developing countries.

There is also evidence that corruption is more prevalent in developing countries or regions with

higher tax rates or higher federal transfers, all of which proxy for lower government taxable income,

while corruption is less prevalent where government activities are more decentralized (e.g., Fisman

and Gatti 2002a, b; Fisman and Wei 2004). Second, in large and regionally diverse countries such

as China, India and Indonesia, there is a limited amount of bribes officials can extort because of

regional competition. The bribes are fees in excess of the cost of provision and are very limited due

to competition, and incentives are efficient. In some African and South American countries with

homogeneity across regions (in part due to historical reasons from colonial eras), corrupt officials

32

effectively have monopoly power. As a result, we get the bad equilibrium with under supply of the

government service and overcharging by officials and extracting much larger amounts of rents,

which in turn significantly reduces the level of economic activities. In fact, consistent with the

implications of our model, there is cross-country evidence on the negative relation between

“openness” of an economy (e.g., exports/GDP) and corruption, and a positive relation between

corruption and excessive regulation of entry of new firms (e.g., Djankov et al. 2002; Svensson

2005). In addition, a recent empirical paper, Hanousek et al. (2015), presented evidence that greater

variance in perception of corruption is associated with more efficiency. The argument is that a high

variance indicates different local sub-environments, among which the competitive ones are

relatively efficient and improves the welfare of the whole environment.

The main policy implication from our analysis is that setting up the provision of government

goods and services in a competitive environment reduces the damaging effects of corruption and

promotes growth. A natural follow-up question would be how to introduce more competitiveness to

the government services market. In large and diversified countries, this can be achieved by

encouraging inter-regional economic activities and lowering the costs (and bureaucracy) of

relocation across regions and entry barriers for firms and agents. In small and homogeneous

countries this can be done by encouraging international trade and lowering the costs of relocation

across countries and entry barriers in different countries. These policies are more effective in

promoting economic growth in emerging countries than those intending to eliminate corruption,

because governments in emerging countries are more budget constrained and do not have the

resources to buy cleanness. This problem cannot be mitigated by the alternative of raising the

payments to officials until the country’s economic conditions have been significantly improved (in

the long run) and the government has the ability to raise sufficient tax revenues to cover the

payments to government officials. In Appendix B.1, we generalized our model to multiple

33

government goods case, and showed that the above policy implications are robust to this variation.

Moreover, the central government can combat corruption by encouraging competition for supplying

substitutes and encouraging cooperation for supplying complements.

Our model also implies that corrupt officials have an incentive to increase their monopoly

power, since the total amount of bribes received by a monopolist official is much higher than a

counterpart in a competitive environment. They can regain market power by, for example, blocking

information about similar goods and services provided in other regions/countries, or by increasing

the transaction costs for agents to relocate. Solving problems of this sort must involve the central

government. For example, the central government can reward officials whose regions have shown

the fastest economic growth and attracted the largest amount of economic activities (and investment)

from non-local sources. These policies encourage inter-regional competition among officials and

induce more government services and goods supplied and higher levels of growth. In China, one of

the key yardsticks in measuring the performance of local officials is economic growth and the

amount of investment from non-local sources (e.g., Li 1998; Li and Zhou 2005); in other countries

the most corrupt government officials hold the highest positions in the government, making it

infeasible to implement incentive-aligning policies for lower level officials.

The situation depicted in Proposition 4 with imperfect Bertrand competition is perhaps a

more general case. With transaction costs, each of the officials providing the services or goods

captures some rents by setting prices above marginal costs of supplying them by exactly the size of

the transaction costs (of agents). But the distortion is less than that of a monopolist official. An

implication of this result is that economic agents can alleviate the damaging effects of corruption

through reducing their own transaction costs. For example, small firms are easier to relocate than

large firms; firms operate in industries that require small scales of investment in fixed assets can

also relocate at low costs.

34

Consistent with this prediction, in both China and India, the sectors that have witnessed the

fastest growth are small- and medium-sized mobile firms from services and labor- and technology-

intensive industries (e.g., Allen et al. 2005, 2012). On the other hand, as Table 4 indicates, Russia’s

economy is dominated by large corporations in natural resources (“other” industry in the last

column) and heavy manufacturing industries. Firms in these industries face much higher relocation

costs and hence the damaging effects of corruption may be one of the reasons that economic

development in post-Soviet Union era (of Russia) has been lagging behind that of many smaller

former socialist countries in Eastern Europe. Further empirical studies can systematically examine

the impact of corruption on different industries across countries or different regions within the same

country.

We close this section by pointing out some limitations of our simple model. First, the

objective of the central government is not necessarily to maximize social welfare. The mapping of

our model to the reality depends on the benevolence of the central government. In the extreme case

that the central government is totally corrupt and aims to collect all bribe payments, introducing

competition only impedes the central government’s objective. Second, our static model doesn’t

allow collusion by competitive local officials. If we extend the model to infinite horizon repeated

game, a possible Nash Equilibrium is that the two local officials collude to charge a monopoly price.

However, in reality, there is often rotation of local officials after their tenures, hence the real

situation is closer to a finite repeated game or static game. As long as the game is finite, the unique

equilibrium is just the repetition of the static equilibrium as in our model. Third, our simple model

implies that the central government should decentralize and delegate the provision of goods and

services to local government officials. A potential problem with this delegation is quality control

(e.g., driving tests before issuing a driver’s license; see, e.g., Bertrand et al. (2006) for details). As

shown in the literature on franchising, designing uniform and enforceable contracts across the board

35

and monitoring the local officials effectively are necessary conditions to ensure that the quality of

the goods and services provided is not compromised under decentralization.

V. Summary and Concluding Remarks

This paper proposes an explanation for the puzzling fact that rampant government

corruption is not associated with slower economic growth at the country level. We focus on

corruption associated with the provision of government goods and services, including those to the

public such as driver’s license and those to firms such as a banking license for financial firms. We

first show that corruption occurs because of an agency problem, and the government cannot

effectively control corruption without sufficient revenues to pay the officials. The bribes received

by officials can be regarded as user fees in excess to the cost of provision of the service. We then

argue that the structure of the provision of government services and goods partially determines the

negative effects of corruption and the cost of curbing it: when multiple officials undertaking the

same project, the fee is determined competitively such that the pernicious effects of corruption are

minimized, and the payments to the officials are also minimized; when each official is a monopolist

in the provision of the services and goods and the damaging effects are much higher, as are the

payments to the officials.

The main policy implication from our analysis is that there are two substitute ways to

control corruption: setting up the provision of government services and goods in a competitive

environment and incentivizing officials with performance-sensitive payment contracts financed by

tax revenues. In large and diversified countries competition can be introduced by encouraging

inter-regional economic activities and lowering the entry barriers for firms and agents from different

regions. In small and homogeneous countries this can be done by encouraging international trade

and lowering the costs of relocation across countries. Economic agents can make the competition

36

mechanism more effective by reducing their own transaction costs when moving from region to

region. An implication for this result is that corruption has a smaller negative impact on small firms

and firms that have smaller amount of fixed assets. Finally, the central government can also control

corruption of this type by rewarding officials whose regions have shown the fastest economic

growth and attracted the largest amount of economic activities from local and non-local sources.

These policies encourage inter-regional competition among officials, and are more effective in

promoting growth than policies aiming at eliminating corruption considered in our paper in

emerging countries that do not have sufficient tax revenue for the appropriate compensation of local

officials.

37

References

1. Acemoglu, Daron, and Simon Johnson, 2005. “Unbundling Institutions,” Journal of Political Economy,

113 (5), 949-995.

2. Ades, Alberto, and Rafael Di Tella, 1999. “Rents, Competition and Corruption,” American Economic

Review 89(4), 982-993.

3. Albornoz, F., & Cabrales, A., 2013. “Decentralization, political competition and corruption,” Journal of

Development Economics, 105, 103-111.

4. ____, Jun Qian, and Meijun Qian, 2005. “Law, Finance, and Economic Growth in China,” Journal of

Financial Economics 77 (1), 57-116.

5. Allen, Franklin, Rajesh Chakrabarti, Sankar De, Jun Qian, and Meijun Qian, 2012. “Financing Firms in

India,” Journal of Financial Intermediation 21, 409-445.

6. Banerjee, A., Kumar, S., Pande, R., & Su, F, 2010. “Do informed voters make better choices?

Experimental evidence from urban India”. Unpublished manuscript. http://www. povertyactionlab.

org/node/2764.

7. Beck, Thorsten, Asli Demirgüç-Kunt and Ross Levine, 2005. “SMEs, Growth and Poverty: Cross-

country Evidence,” Journal of Economic Growth 10 (3), 199-229.

8. Becker, Gary and George Stigler, 1974. “Law Enforcement Malfeasance, and the Compensation of

Enforcers,” Journal of Legal Studies 3, 1-19.

9. Bertrand, Marianne, Simeon Djankov, Rema Hanna, and Sendhil Mullainathan, 2006. “Does Corruption

Produce Unsafe Drivers? Working paper, MIT.

10. Bliss, Christopher, and Rafael Di Tella, 1997. “Does Competition Kill Corruption?” Journal of Political

Economy 105(5), 1001-1023.

11. Burgess, Robin, Matthew Hansen, Benjamin A. Olken, Peter Potapov, and Stefanie Sieber, 2012. “The

Political Economy of Deforestation in the Tropics,” The Quarterly Journal of Economics 127(4), 1707-

1754.

12. Damania, Richard, Per G. Fredriksson, and Muthukumara Mani, 2004. “The Persistence of Corruption

and Regulatory Compliance Failures: Theory and Evidence.” Public Choice 121.3-4, 363-390.

13. Djankov, Simeon, Rafael La Porta, Florencio Lopez-de-Silanes, and Andrei Shleifer, 2002. “The

Regulation of Entry,” Quarterly Journal of Economics 117 (1), 1-37.

14. ____, ____, ____, and ____, 2003. “Courts,” Quarterly Journal of Economics 118 (2), 453-517.

15. ____, ____, ____, and ____, 2008. “The Law and Economics of Self-Dealing,” Journal of Financial

Economics 88(3), 430-465.

16. ____, ____, ____, and ____, 2010. "Disclosure by Politicians,” American Economic Journal: Applied

Economics 2(2): 179-209.

38

17. Demirgüç-Kunt, Asli, and Ross Levine, 2002. Financial Structure and Economic Growth: Cross-country

Comparisons of Banks, Markets, and Development, MIT Press, Cambridge, Massachusetts.

18. Fisman, Raymond, and Roberta Gatti, 2002a. “Decentralization and Corruption: Evidence from U.S.

Federal Transfer Programs,” Public Choice 113, 25-35.

19. ____, and ____, 2002b. “Decentralization and Corruption: Evidence across Countries,” Journal of Public

Economics 83, 325-345.

20. ____, and Jakob Svensson, 2007. “Are Corruption and Taxation really Halmful to Growth? Firm Level

Evidence,” Journal of Development Economics 83, 63-75.

21. ____, and Shang-jin Wei, 2004. “Tax Rates and Tax Evasion: Evidence from ‘Missing Imports’ in

China,” Journal of Political Economy 112, 471-496.

22. Friedman, Eric, Simon Johnson, Daniel Kaufmann and Pablo Zoido-Lobaton, 2000. “Dodging the

Grabbing Hand: The Determinants of Unofficial Activity in 69 Countries,” Journal of Public Economics

76, 459-493.

23. Gorodnichenko, Y., & Peter, K. S., 2007. “Public sector pay and corruption: Measuring bribery from

micro data,” Journal of Public Economics 91(5), 963-991.

24. Gutmann, Jerg, Fabio Padovano, and Stefan Voigt, 2015. “Perception vs. Experience: Explaining

Differences in Corruption Measures Using Microdata,” Working Paper.

25. Hanousek, Jan, Anastasiya Shamshur, and Jiri Tresl, 2015. “Is bread gained by deceit sweet to a man?

Corruption and firm efficiency,” Working Paper.

26. Holmstrom, Bengt, and Paul Milgrom, 1987. “Aggregation and linearity in the provision of intertemporal

incentives,” Econometrica: Journal of the Econometric Society, 303-328.

27. Huntington, Samuel, 1968. Political Order in Changing Societies. New Haven: Yale University Press.

28. La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Shleifer, and Robert Vishny, 1997. “Trust in Large

Organizations,” American Economic Review (proceedings issue), 87 (2), 333-338.

29. ____, ____, ____, and ____, 1998. “Law and Finance,” Journal of Political Economy, 106, 1113-55.

30. ____, ____, ____, and ____, 1999. “The Quality of Government,” Journal of Law, Economics, and

Organization, 15 (1), 222-279.

31. Leff, Nathaniel, 1964. “Economic Development through Bureaucratic Corruption,” American Behavioral

Scientist 82, 337-341.

32. Li, David D., 1998. “Changing Incentives of the Chinese Bureaucracy,” American Economic Review 88,

393-397.

33. Li, Hongbin, and Li-An Zhou, 2005. “Political turnover and economic performance: the incentive role of

personnel control in China,” Journal of Public Economics, 89(9), 1743-1762.

34. Lui, Francis, 1985. “An Equilibrium Queuing Model of Bribery,” Journal of Political Economy 93 (4),

760-781.

39

35. Mauro, Paolo, 1995. “Corruption and Growth,” Quarterly Journal of Economics 110, 681-712.

36. Ogus, Anthony, 2004. "Corruption and Regulatory Structures." Law & Policy 26.3‐4, 329-346.

37. Olken, B. A., 2006. “Corruption and the costs of redistribution: Micro evidence from Indonesia,” Journal

of public economics, 90(4), 853-870.

38. Olken, B. A., 2009. “Corruption perceptions vs. corruption reality,” Journal of Public economics, 93(7),

950-964.

39. Olken, B. A., & Pande, R, 2012. “Corruption in Developing Countries,” Annual Review of

Economics, 4(1), 479-509.

40. Rajan, Raghuram, and Luigi Zingales, 2003a. “The Great Reversals: The Politics of Financial

Development in the Twentieth Century,” Journal of Financial Economics 69, 5-50.

41. _____, and _____, 2003b. Saving Capitalism from Capitalists: Unleashing the Power of Financial

Markets to Create Wealth and Spread Opportunity, Random House, New York.

42. Rose-Ackerman, Susan, 2004. “Governance and Corruption,” in Global Crises, Global Solutions, B.

Lomborg, ed. Cambridge: Cambridge University Press, Chapter 6.

43. Schneider, F. and D. Enste, 2000, “Shadow Economies: Sizes, Causes and Consequences,” Journal of

Economic Literature 38 (1), 77-114.

44. ____, 2002. “Size and Measurement of the Informal Economy in 110 Countries around the World,”

World Bank.

45. Spagnolo, Giancarlo, 1999, “Social Relations and Cooperation in Organizations,” Journal of Economic

Behavior and Organization 38, 1-25.

46. Shleifer, Andrei, and Robert Vishny, 1993. “Corruption,” Quarterly Journal of Economics 108, 599-617.

47. Svensson, Jacob, 2003. “Who Must Pay Bribes and How Much? Evidence from a Cross Section of

Firms,” Quarterly Journal of Economics 118, 207-230.

48. ____, 2005. “Eight Questions about Corruption,” Journal of Economic Perspectives 19(3), 19-42.

49. Tiebout, C. M., 1961. “An economic theory of fiscal decentralization,” Public Finances: Needs, Sources,

and Utilization (pp. 79-96). Princeton University Press.

40

Appendix A Proofs

A.1 Central Government’s Maximization Problem: Monopoly

Denote the optimal contract as {tM, sM}, the induced quantity supplied as MQ̂ and the

induced excess user fee charged as FBMM QQPf )ˆ(ˆ .

First, assume tM>0, which implies that MM

M Qa

tcWQ

2ˆ . Note that the total social

welfare decreases in T(Q)=T(QM)+t(Q-QM), thus decreases in T(QM). Therefore, the limited

liability condition (equation 8) must bind, i.e. sM =tMQM. Plug in s=tQM and a

tcWQ

2

into the

expression for the total welfare. Denote the total welfare as L(t). The maximization problem can

then be simplified as follow,

a

tgv

a

t

a

cWt

a

cWtL

t 284

)(

8

)(3)(max

222

The first order derivative of the above simplified problem is

a

tgv

a

t

a

tcWtL

24)(

2

Note that 0)(4

1)0( cW

aL and 0

2

)()(

2

a

cWgv

a

cWcWL , which implies the

optimal ),0( cWt M and verifies the assumption made at the beginning of the proof. By

equating the first order derivative to zero, we can solve for the optimal sensitive to pay tM, which

solves 4

1

42

2

M

M

t

cW

a

tgv .

Since ),0( cWt M , the induced quantity supplied is ),(2

ˆ FBMM

M QQa

tcWQ

41

and ),0(2

ˆ MM

M ftcW

f

. The total payment made by the central government is

a

tQT

MM

2)ˆ(

2

. The welfare in the first best case is )(2

)( 2

gva

cWWelfare FB

, and the resulting

deadweight loss with the optimal incentive scheme is

02

)(8

)()(

22

a

tgvgv

a

tcWtLWelfareDWLoss

MMMFBM

T . Since tM>0, L(tM)>L(0)

and MFBMFBM

T DWLossLWelfaretLWelfareDWLoss )0()( .

A.2 Comparative Statics in Proposition 2

Given that the optimal sensitive to pay tM is such that 4

1

42

2

M

M

t

cW

a

tgv . By implicit

function theorem, 0

24)(

22

2

3

4

2

2

a

tgvatcWa

ta

tgv

tda

d

MM

MM

M . Similarly, it can be verified

that 0MtdW

d, 0Mt

dc

d and 0Mt

dg

d.

Given that the deadweight loss is )2

()(8

)(22

a

tgvgv

a

tcWDWLoss

MMM

T

. The

signs of the partial derivatives can be verified to be 0

M

TMDWLoss

t, 0

M

TDWLossa

,

0

M

TDWLossW

, 0

M

TDWLossc

and 0

M

TDWLossg

. Intuitively, since Mt minimizes

deadweight loss, the first order condition ensures that 0

M

TMDWLoss

t.

42

Given that a

tQT

MM

2)ˆ(

2

, it can be verified that 0)ˆ(

a

tQT

t

MM

M,

02

)ˆ(T2

2

a

tQ

a

MM and 0)ˆ(T)ˆ(T)ˆ(T

MMM Qg

Qc

QW

.

The total derivatives can then be calculated as follow,

M

T

MM

TM

M

T

M

T DWLossx

tdx

dDWLoss

tDWLoss

xDWLoss

dx

d

and

MM

M

MM tdx

dQT

tQT

xQT

dx

d)ˆ()ˆ()ˆ(

.

Plugging in the signs, it can be verified that 0W

M

TDWLossd

d, 0

cM

TDWLossd

d,

0M

TDWLossda

dand 0M

TDWLossdg

d. Similarly, it is straightforward to verify that

0)ˆ( MQTdW

d, 0)ˆ( MQTdc

d and 0)ˆ( MQT

dg

d. As for the sign of )ˆ( MQT

da

d, it takes a few lines

to calculate:

0

28)(2

)(

24)(

22

2)ˆ(

2

3

2

2

3

4

2

2

22

2

a

tgvatcWa

cWt

a

tgvatcWa

ta

tgv

a

t

a

tQT

da

d

MM

M

MM

MM

MMM

.

A.3 Central Government’s Maximization Problem: Imperfect Bertrand Competition

Denote the optimal contract as B

K

B

K st , , the induced quantity supplied as B

KQ̂ and the induced

excess user fee charged as FBB

K

B

K QQPf )ˆ(ˆ .

43

First, assume KcWt B

K 2 , which implies that B

K

B

KB

K Qa

tcWQ

2ˆ . Note that the

total social welfare decreases in )()()( B

K

B

K QQtQTQT , thus decreases in )( B

KQT . Therefore the

limited liability condition (equation 19) must bind, i.e. B

K

B

K

B

K Qts . Plug in B

KtQs and

a

tcWQ

2

into the expression for the total welfare. Denote the total welfare as H(t). The

maximization problem can then be simplified as follow,

a

cWKttgv

a

t

a

cWt

a

cWtH

t 2

2

84

)(

8

)(3)(max

22

The first order derivative of the above simplified problem is

a

cWKttgv

a

cWKt

a

tcWtH

2

2

2

22

4)(

The first part in the derivative is the marginal benefit, i.e. the marginal welfare restored by

increasing the sensitivity of pay for performance. And the second part in the derivative is the

marginal opportunity cost, i.e. the marginal burden on other government activities of increasing the

sensitivity of pay for performance. The optimal incentive scheme can be found by equating

marginal cost to marginal benefit, mathematically by equating the first order derivative to zero. The

optimal sensitive to pay B

Kt is such that

)22(22

2

cWKt

tcW

a

cWKttgv

B

K

B

K

B

K

B

K

.

The last step in solving for the optimal incentive scheme is to verify the assumption that

KcWt B

K 2 , which is equivalent to verifying 0)2( KcWH . Plug KcWt 2 into

the first order derivative:

)(21

)()(0

2

2

2)2(

gv

gvcWKKgv

a

KcW

a

KKcWH

.

44

If KK , the friction in the competition is so small that the equilibrium is close to the first best,

and t needs to be really high to induce further reduction in the bribes charged by corrupt officials.

To see it clearly, consider a K close to zero. In order to reduce corruption through incentive scheme,

the sensitivity of pay for performance t needs to be close to W-c. Since

0))(

(2

)(

a

KcWgv

a

KcWcWH ,

i.e. the marginal benefit goes to zero, yet the marginal cost is strictly positive. Therefore, it is not

cost-efficient to offer such huge incentive package, and the optimal incentive scheme to offer is any

KcWt 2 and B

KtQs , which is outcome-equivalent to offering no incentive scheme. Note

that since 2

cWK

, the parameter space such that an effective incentive scheme is offered, i.e.

)2

,(cW

KK

, is non-empty.

Recall that the optimal contract in the monopoly case satisfies4

1

42

2

M

M

t

cW

a

tgv .

Calculate the first derivative of the total welfare at tM:

a

cWKttgv

cWKt

a

tgvt

atH

MMM

MMM

2

2

22

1)(

2

Since 2

cWK

and 0Mt , it must be true that

0

2

2

2

2

a

cWKttgv

a

tgv

MMM

02

cW

Ktt MM ,

therefore 0)( MtH and MB

K tt . Note that

02

2)(

a

cWKgv

a

KtcWH ,

45

therefore the optimal cWt B

K .

Since ),( cWtt MB

K , the induced quantity supplied and the bribe charged are

),ˆ(2

ˆ FBMB

KB

K QQa

tcWQ

)ˆ,0(2

ˆ MB

KB

K ftcW

f

.

The total payment made by the central government is

a

cWKttQT

B

K

B

KB

K2

2)ˆ(

,

and the resulting deadweight loss is

02

2)(

8

)()(

2

,

a

cWKttgvgv

a

tcWtHWelfareDWLoss

B

K

B

K

B

KB

K

FBB

TK .

Note that tM is just a special case of B

Kt when 2

cWK

, i.e. )ˆ()ˆ(

2

B

cW

M QTQT . To prove that

)ˆ()ˆ( MB

K QTQT , it suffices to show that 0)ˆ(

dK

QdT B

K . By implicit function theorem, we can solve

for dK

dt B

K from equation

)22(22

2

cWKt

tcW

a

cWKttgv

B

K

B

K

B

K

B

K

. Specifically,

1

Bt

TA

BK

TA

dK

dt B

K where 0)22)((2 cWKtTgvA and 0)(4 TgvB . Finally,

1

)()ˆ(

Bt

TA

K

T

t

T

K

TB

dK

dt

t

T

K

T

dK

QdT B

K . Since 0

a

t

K

T, 0

2

22

a

cWKt

t

T and

02

2

a

KcW

t

T

K

T, it must be true that 0

)ˆ(

dK

QdT B

K and therefore )ˆ()ˆ( MB

K QTQT for all

46

2

cWK

.

Since B

Kt maximizes the function H(t), it must be true that )()( MB

K tHtH and

.2

)(8

)(

2

2)(

8

)()(

22

2

,

M

T

MM

MMMMFBB

TK

DWLossa

tgvgv

a

tcW

a

cWKttgvgv

a

tcWtHWelfareDWLoss

A.2 Comparative Statics in Proposition 4

Given that the optimal sensitive to pay tM is such that

)22(22

2

cWKt

tcW

a

cWKttgv

B

K

B

K

B

K

B

K

. By implicit function theorem, it is

straightforward to show that 0B

Ktda

d, 0B

KtdW

d, 0B

Ktdc

d, 0B

Ktdg

dand 0B

KtdK

d.

Given that

a

cWKttgvgv

a

tcWDWLoss

B

K

B

K

B

KB

TK2

2)(

8

)( 2

, , the signs of the

partial derivatives can be proved to be 0,

B

TK

K

DWLosst B

, 0,

B

TKDWLossa

,

0,

B

TKDWLossW

, 0,

B

TKDWLossc

, 0,

B

TKDWLossg

and 0,

B

TKDWLossK

. Intuitively,

since Mt minimizes deadweight loss, the first order condition ensures that 0,

B

TK

K

DWLosst B

.

Given that

a

cWKttQT

B

K

B

KB

K2

2)ˆ(

, it can be verified that 0)ˆ(

B

KB

K

QTt

,

0)ˆ(T

B

KQa

, 0)ˆ(T

B

KQW

, 0)ˆ(T

B

KQc

, 0)ˆ(T

B

KQg

and 0)ˆ(T

B

KQK

.

The total derivatives can then be calculated as follow,

47

B

TK

B

K

B

TKB

K

B

TK

B

TK DWLossx

tdx

dDWLoss

tDWLoss

xDWLoss

dx

d,,,,

and

B

K

B

KB

K

B

K

B

K tdx

dQT

tQT

xQT

dx

d)ˆ()ˆ()ˆ(

.

Plugging in the signs, it can then be verified that 0, B

TKDWLossda

d, 0

W, B

TKDWLossd

d,

0c

, B

TKDWLossd

d, 0, B

TKDWLossdg

d and 0, B

TKDWLossdK

d. Similarly, it can be verified that

0)ˆ( B

KQTda

d, 0)ˆ( B

KQTdg

d and 0)ˆ( B

KQTdK

d.

48

Table 1 Top 20 Economies in the World: GDP and Growth

GDP in 2014 (simple

exchange rates) GDP in 2014

(PPP*)

GDP growth:

1990-2014

(constant prices)

Per capita GDP

growth: 1990-2014**

(const. prices)

Rank Country

/Region

US$

billion

Country

/Region

Int’l

$ billion Country

/Region

Annual

growth

Country

/Region

Annual

growth

1 U.S. 17,419 China 17,617 China 9.8% China 9.0%

2 China 10,380 U.S. 17,419 Sudan 8.5% Sudan 6.7%

3 Japan 4,616 India 7,376 Vietnam 6.9% Vietnam 5.5%

4 Germany 3,860 Japan 4,751 Uganda 6.7% Sri Lanka 4.8%

5 U.K. 2,945 Germany 3,722 Nigeria 6.6% Korea 4.7%

6 France 2,847 Russia 3,565 India 6.5% India 4.7%

7 Brazil 2,353 Brazil 3,264 Ethiopia 6.4% Taiwan 4.4%

8 Italy 2,148 Indonesia 2,676 Angola 6.3% Nigeria 3.8%

9 India 2,050 France 2,581 Malaysia 6.0% Bangladesh 3.7%

10 Russia 1,857 U.K. 2,549 Sri Lanka 5.9% Chile 3.7%

11 Canada 1,789 Mexico 2,141 Ghana 5.6% Malaysia 3.7%

12 Australia 1,444 Italy 2,128 Korea 5.5% Ethiopia 3.6%

13 Korea 1,417 Korea 1,779 Bangladesh 5.5% Indonesia 3.6%

14 Spain 1,407 Saudi Arabia 1,606 Taiwan 5.1% Thailand 3.6%

15 Mexico 1,283 Canada 1,592 Indonesia 5.1% Uganda 3.3%

16 Indonesia 889 Spain 1,566 Tanzania 5.1% Poland 3.2%

17 Turkey 806 Turkey 1,508 Thailand 4.5% Ghana 3.0%

18 Saudi Arabia 752 Iran 1,334 Pakistan 4.4% Angola 3.0%

19 Nigeria 574 Australia 1,095 Saudi Arabia 4.4% Peru 2.8%

20 Poland 547 Taiwan 1,075 Iran 4.3% Iran 2.7%

Notes: * The PPP conversion factor is obtained from the World Bank Development Indicator (Table 5.6, World Bank. For details on

how to calculate the indicator, see “Handbook of the International Program.” United Nations, New York, 1992).

**: Countries with population less than 20 million or GDP less than USD20billion are excluded from this ranking.

49

Table 2 Transparency International – Bribe Payers’ Index (BPI)

This table shows the results of BPI, published by the Transparency International. The BPI is based on the responses of

business executives worldwide to questions about the extent to which foreign companies engage in bribery when doing

business in their country. The answers are converted to a score between 0 (bribes most frequent) to 10 (bribes never

occur), and the ranking of the leading economies, including all G20 countries for year 2006 and 2011, listed below

shows the average score of a country.

2011 2006

Country/ Region

Rank Score No. of respondents Rank Score No. of respondents

Netherlands 1 8.8 273 8 7.3 1,821

Switzerland 1 8.8 244 1 7.8 1,744

Belgium 3 8.7 221 9 7.2 1,329

Germany 4 8.6 576 7 7.3 3,873

Japan 4 8.6 319 11 7.1 3,279

Australia 6 8.5 168 3 7.6 1,447

Canada 6 8.5 209 5 7.5 1,870

Singapore 8 8.3 256 12 6.8 1,297

United Kingdom 8 8.3 414 6 7.4 3,442

United States 10 8.1 651 9 7.2 5,401

France 11 8.0 435 15 6.5 3,085

Spain 11 8.0 326 13 6.6 2,111

South Korea 13 7.9 152 21 5.8 1,930

Brazil 14 7.7 163 23 5.7 1,317

Hong Kong 15 7.6 208 18 6.0 1,556

Italy 15 7.6 397 20 5.9 2,525

Malaysia 15 7.6 148 25 5.6 1,319

South Africa 15 7.6 191 24 5.6 1,488

Taiwan 19 7.5 193 26 5.4 1,731

India 19 7.5 168 30 4.6 2,145

Turkey 19 7.5 139 27 5.2 1,755

Saudi Arabia 22 7.4 138 22 5.8 1,302

Argentina 23 7.3 115 N/A N/A N/A

United Arab Emirates 23 7.3 156 14 6.6 1,928

Indonesia 25 7.1 153 N/A N/A N/A

Mexico 26 7.0 121 17 6.5 1,765

China 27 6.5 608 29 4.9 3,448

Russia 28 6.1 172 28 5.2 2,203

No. of Countries 28 30

No. of Respondents 3,016 11,232

Average Score 7.8 6.4

Std. of Score 0.7 0.9

Min. Score 6.1 4.6

Max. Score 8.8 7.8

50

Table 3-A Corruption, Taxing Ability and Economic Growth

The Corruption Perception Index for a country is the average value (0 to 10; lower score means more corrupt) of

Transparency International from 1995 to 2014. GDP data, starting from 1980 to 2014, are all based on PPP (purchasing

power parity) and from IMF World Economic Outlook Database. Tax/GDP variable, starting from 1990 to 2014, is the

percentage of tax revenue over GDP and from World Bank.

Statistics

Corruption

Perception

Index

Tax/GDP

GDP Growth

Rate per year

(1981-2014)

Per Capita

GDP

(1981)

GDP per

capita

(2014)

More Corrupt Countries

(Above Median)

Mean 2.68 13% 6.65% 1,939 7,749

Stdev. 0.51 5% 2.50% 1,831 6,964

Less Corrupt Countries

(Below Median)

Mean 5.88 19% 5.97% 9,898 30,689

Stdev. 1.83 9% 1.67% 13,442 22,970

51

Table 3-B Corruption, Taxing Ability, and Economic Growth: An International Comparison

The Corruption Perception Index for a country is the average value (0 to 10; lower score means more corrupt) of Transparency International from 1995 to 2014.

GDP data, starting from 1980 to 2014, are all based on PPP (purchasing power parity) and from IMF World Economic Outlook Database. Tax/GDP variable,

starting from 1990 to 2014, is the percentage of tax revenue over GDP and from World Bank.

Country

Corruption

Perception Index Taxes/GDP

GDP Growth

Rate (1981-14)

Stdev. of GDP

Growth

Per Capita

GDP at 1981

Per Capita

GDP at 2014

GDP at 1981

(PPP billion)

GDP at 2014

(PPP billion)

Panel A More Corrupt Countries (Below Median)

Afghanistan 1.42 7% 10% 6% n/a 1,937 n/a 61

Myanmar 1.64 4% 12% 4% n/a 4,706 n/a 242

Sudan 1.68 6% 10% 12% 472 4,267 9 159

Equatorial Guinea 1.90 13% 19% 30% n/a 32,266 0 25

Nigeria 1.95 3% 8% 7% 1,454 6,031 102 1,049

Angola 1.98 26% 8% 9% 1,794 7,203 16 176

Bangladesh 1.99 7% 8% 1% 560 3,373 47 534

Congo Dem. Rep. 2.01 5% 4% 6% 656 704 20 56

Burundi 2.06 15% 5% 5% 449 911 2 8

Guinea 2.07 11% 6% 2% n/a 1,313 2 15

Cambodia 2.07 9% 10% 3% n/a 3,263 n/a 50

Tajikistan 2.11 9% 6% 9% n/a 2,688 n/a 22

Paraguay 2.13 11% 6% 5% 3,062 8,449 10 58

Kyrgyzstan 2.18 15% 4% 7% n/a 3,361 n/a 19

Azerbaijan 2.19 14% 8% 14% n/a 17,618 n/a 165

Cameroon 2.19 9% 6% 6% 1,457 2,981 13 67

Congo Rep. 2.19 7% 7% 6% 2,082 6,559 3 28

Kenya 2.20 17% 6% 3% 1,074 3,084 18 132

Central African Rep. 2.24 9% 3% 9% 587 607 1 3

Papua New Guinea 2.28 21% 7% 5% 750 2,399 2 18

Venezuela 2.29 14% 5% 7% 8,208 17,695 126 539

Yemen 2.29 10% 6% 4% n/a 3,774 n/a 104

Laos 2.33 13% 9% 4% 645 4,987 2 34

Cote d'Ivoire 2.35 14% 5% 4% 1,981 3,131 16 71

Pakistan 2.37 11% 8% 3% 980 4,736 83 882

Indonesia 2.42 14% 8% 4% 1,608 10,641 242 2,676

Ukraine 2.42 16% 1% 10% n/a 8,668 n/a 371

Sierra Leone 2.45 9% 5% 10% 1,117 2,027 3 13

Russia 2.45 15% 4% 7% n/a 24,805 n/a 3,565

Honduras 2.47 15% 6% 3% 1,588 4,729 6 39

Uganda 2.48 11% 9% 4% 415 2,023 5 77

Ecuador 2.55 16% 6% 4% 3,533 11,244 30 180

Zimbabwe 2.55 22% 1% 9% n/a 2,046 n/a 27

Kazakhstan 2.56 10% 6% 7% n/a 24,020 n/a 418

Nepal 2.58 9% 7% 3% 484 2,376 7 67

Iran 2.58 7% 7% 7% 4,756 17,114 194 1,334

52

Country

Corruption

Perception Index Taxes/GDP

GDP Growth

Rate (1981-14)

Stdev. of GDP

Growth

Per Capita

GDP at 1981

Per Capita

GDP at 2014

GDP at 1981

(PPP billion)

GDP at 2014

(PPP billion)

Panel A (Cont.) More Corrupt Countries (Below Median)

Togo 2.64 15% 5% 5% 953 1,450 2 10

Nicaragua 2.66 13% 5% 4% n/a 4,736 7 29

Maldives 2.68 13% 10% 6% 1,439 14,383 0 5

Bolivia 2.73 15% 6% 3% 2,186 6,221 12 70

Tanzania 2.75 12% 7% 2% 692 2,667 13 127

Niger 2.77 11% 5% 6% 574 1,048 3 18

Gambia 2.78 10% 6% 6% 614 1,599 0 3

Ethiopia 2.79 9% 8% 7% 334 1,589 12 145

Mozambique 2.81 19% 8% 6% 208 1,174 3 31

Philippines 2.85 14% 6% 4% 2,061 6,962 102 692

Lebanon 2.87 15% 6% 17% 6,724 17,986 17 81

Madagascar 2.87 10% 4% 4% 886 1,437 8 34

Guatemala 2.88 10% 6% 2% 3,252 7,503 21 119

Albania 2.89 14% 6% 8% 2,354 11,377 6 32

Moldova 2.91 17% 3% 9% n/a 4,979 n/a 18

Mali 2.95 15% 7% 4% 569 1,729 3 27

Armenia 2.96 16% 8% 8% n/a 7,374 n/a 24

Dominican Rep. 3.04 14% 7% 4% 2,791 13,012 16 138

Zambia 3.05 16% 6% 5% 1,777 4,064 10 61

Algeria 3.05 38% 6% 3% 5,025 14,259 97 552

Belarus 3.06 19% 6% 7% n/a 18,161 n/a 172

India 3.09 9% 9% 3% 643 5,855 448 7,376

Kiribati 3.10 16% 4% 4% 998 1,713 0 0

Liberia 3.10 0% 6% 10% n/a 882 n/a 4

Benin 3.11 16% 6% 3% 710 1,870 3 20

Argentina 3.14 11% 5% 6% 6,411 22,583 181 948

Egypt 3.16 16% 7% 3% 2,320 10,877 97 943

Mongolia 3.21 15% 8% 6% n/a 11,882 3 35

Vanuatu 3.26 18% 6% 4% 1,050 2,608 0 1

Senegal 3.34 19% 6% 3% 913 2,311 5 34

Sao Tome & Principe 3.35 14% 5% 4% 1,452 3,153 0 1

China 3.36 10% 13% 3% 343 12,880 343 17,617

Mexico 3.37 10% 5% 4% 6,351 17,881 462 2,141

Bosnia & Herzegovina 3.37 21% 6% 4% n/a 9,833 n/a 38

Thailand 3.37 16% 8% 5% 1,758 14,354 84 986

Burkina Faso 3.37 13% 8% 3% 363 1,682 3 29

Romania 3.41 14% 4% 6% 5,220 19,712 118 393

Sri Lanka 3.42 15% 8% 3% 1,171 10,372 17 217

Colombia 3.43 13% 6% 3% 3,041 13,430 88 640

Georgia 3.44 14% 8% 4% n/a 7,653 n/a 34

Grenada 3.45 20% 6% 5% 2,017 11,979 0 1

Panama 3.45 12% 8% 5% 3,605 19,455 7 76

53

Country

Corruption

Perception Index Taxes/GDP

GDP Growth

Rate (1981-14)

Stdev. of GDP

Growth

Per Capita

GDP at 1981

Per Capita

GDP at 2014

GDP at 1981

(PPP billion)

GDP at 2014

(PPP billion)

Panel B Less Corrupt Countries (Above Median)

Macedonia 3.51 18% 4% 4% n/a 13,349 n/a 28

Suriname 3.52 20% 4% 4% n/a 16,623 2 9

Jamaica 3.56 26% 4% 4% 3,263 8,609 7 24

Belize 3.57 20% 8% 6% 1,491 8,248 0 3

Serbia 3.59 21% 5% 4% n/a 13,329 n/a 95

Morocco 3.59 23% 7% 5% 1,605 7,606 32 252

Brazil 3.76 14% 5% 3% 4,927 16,096 598 3,264

Lesotho 3.77 45% 7% 3% 494 2,764 1 5

El Salvador 3.79 13% 5% 3% 2,057 8,021 10 51

Bulgaria 3.79 19% 4% 6% 5,338 17,860 47 129

Ghana 3.79 15% 8% 3% 883 4,129 9 108

Peru 3.80 14% 6% 6% 3,531 11,817 63 371

Turkey 3.91 20% 7% 5% 3,956 19,610 171 1,508

Rwanda 3.92 11% 7% 11% 473 1,698 2 19

Trinidad and Tobago 3.92 25% 5% 6% n/a 32,139 9 43

Croatia 3.94 21% 4% 5% n/a 20,889 n/a 88

Fiji 4.00 22% 5% 5% 2,312 8,236 2 7

Latvia 4.22 9% 5% 7% n/a 23,707 n/a 48

Slovakia 4.25 14% 6% 4% n/a 28,175 n/a 153

Greece 4.34 20% 4% 4% 9,602 25,859 93 284

Samoa 4.43 0% 5% 4% n/a 5,180 0 1

Italy 4.55 22% 4% 3% 11,607 35,486 656 2,128

Kuwait 4.55 1% 7% 19% 28,390 71,020 41 284

Tunisia 4.58 20% 7% 3% 2,394 11,300 16 124

Czech Rep. 4.64 15% 4% 4% n/a 29,925 n/a 315

Poland 4.65 17% 5% 3% 4,607 25,105 165 954

Seychelles 4.71 28% 7% 5% 4,756 25,607 0 2

South Africa 4.77 25% 5% 3% 5,297 13,046 155 705

Namibia 4.77 26% 7% 3% n/a 10,765 n/a 24

Lithuania 4.84 54% 7% 6% n/a 27,051 n/a 80

Jordan 4.84 19% 7% 6% 4,557 11,927 11 80

Korea 4.84 14% 10% 5% 2,520 35,277 98 1,779

Mauritius 4.93 19% 8% 3% 2,050 18,553 2 23

Malaysia 4.98 16% 9% 5% 3,664 24,654 52 746

Hungary 5.04 22% 4% 4% 7,075 24,942 76 246

Costa Rica 5.04 14% 7% 3% n/a 14,864 8 71

Bahrain 5.31 1% 7% 3% 18,168 51,714 7 62

Oman 5.32 5% 8% 6% 9,738 39,681 14 162

Cape Verde 5.40 20% 8% 4% 899 6,324 0 3

Dominica 5.53 22% 6% 4% n/a 10,800 0 1

Bhutan 5.74 8% 11% 5% 574 7,641 0 6

Malta 5.87 44% 4% 2% n/a 33,216 n/a 14

54

Country

Corruption

Perception Index Taxes/GDP

GDP Growth

Rate (1981-14)

Stdev. of GDP

Growth

Per Capita

GDP at 1981

Per Capita

GDP at 2014

GDP at 1981

(PPP billion)

GDP at 2014

(PPP billion)

Panel B (Cont.) Less Corrupt Countries (Above Median)

Botswana 5.98 24% 10% 6% 1,897 16,036 2 34

Slovenia 6.06 19% 5% 4% n/a 29,658 n/a 61

Uruguay 6.06 18% 5% 4% 4,628 20,556 14 70

Cyprus 6.08 40% 7% 4% 7,008 30,769 4 27

Estonia 6.21 17% 6% 7% n/a 26,999 n/a 36

St. Vincent and the

Grenadines 6.27 23% 6% 4% 1,757 10,778 0 1

United Arab Emirates 6.30 0% 7% 8% 81,045 64,479 89 600

Spain 6.31 13% 5% 3% 8,570 33,711 323 1,566

Portugal 6.33 20% 5% 4% 6,747 26,975 66 280

Qatar 6.47 20% 9% 10% 71,815 143,427 19 321

Israel 6.57 25% 7% 3% 7,870 32,691 31 268

Belgium 6.90 25% 5% 2% 11,958 42,973 118 481

France 6.93 21% 5% 2% 11,830 40,375 639 2,581

Saint Lucia 7.02 23% 6% 6% 2,386 11,594 0 2

Japan 7.12 10% 5% 4% 9,655 37,390 1,135 4,751

Chile 7.15 18% 7% 5% 3,925 22,971 45 409

Bahamas 7.15 14% 5% 4% 9,292 25,049 2 9

Barbados 7.30 26% 4% 3% 5,972 16,183 1 5

United States 7.47 10% 5% 3% 13,966 54,597 3,211 17,419

Ireland 7.65 25% 7% 4% 8,406 49,195 29 227

Austria 7.76 19% 5% 2% 12,173 46,420 92 395

Hong Kong SAR 7.87 12% 8% 5% 7,952 54,722 41 398

Germany 7.91 11% 4% 3% 12,261 45,888 949 3,722

United Kingdom 8.24 26% 5% 2% 9,436 39,511 532 2,549

Luxembourg 8.48 25% 7% 4% 16,865 92,049 6 51

Australia 8.62 23% 6% 2% 11,696 46,433 176 1,095

Norway 8.72 27% 5% 2% 16,574 66,937 68 345

Canada 8.75 14% 5% 3% 12,840 44,843 318 1,592

Netherlands 8.78 21% 5% 2% 12,657 47,355 180 799

Switzerland 8.79 10% 4% 2% 18,821 58,087 119 473

Iceland 8.95 24% 5% 4% 12,022 43,637 3 14

Singapore 9.10 14% 10% 5% 10,202 82,762 26 453

Sweden 9.18 21% 5% 3% 11,827 45,986 98 448

New Zealand 9.39 30% 5% 3% 10,247 35,152 32 159

Finland 9.40 21% 5% 4% 10,637 40,347 51 221

Denmark 9.465 32% 4% 3% 12,518 44,343 64 250

55

Table 4 Industry Compositions in the Most Corrupt Countries

Source of data in the table is World Development Indicators 1990-2010 (World Bank). The industrial origin of value added is determined by the International

Standard Industrial Classification (ISIC) revision 3. ‘Agriculture’ corresponds to ISIC divisions 1–5 and includes forestry and fishing; ‘Industry’ covers mining,

manufacturing (also reported separately), construction, electricity, water, and gas (ISIC divisions 10–45); and ‘Services’ correspond to ISIC divisions 50–99; this

sector is derived as a residual (from GDP less agriculture and industry) and may not properly reflect the sum of services output, including banking and financial

services. Manufacturing corresponds to industries belonging to ISIC divisions 15–37.

Country

Agriculture

(% of GDP)

Services

(% of GDP)

Industry

(% of GDP)

Food,

beverages &

tobacco (% of

manufacturing)

Textiles &

clothing (% of

manufacturing)

Machinery &

transport

equipment (% of

manufacturing)

Chemicals

(% of

manufacturing)

Other (% of

manufacturing)

1990 2010 1990 2010 1990 2010 1990 2010 1990 2010 1990 2010 1990 2010 1990 2010

Algeria 11 9 40 40 48 51 13 46 17 4 – 19 10 11 60 20

Argentina 8 8 56 61 36 31 20 – 10 – 10 – 12 – 49 –

China 27 10 32 44 41 46 15 – 15 – 16 – 13 – 42 –

Congo 31 23 40 42 29 34 – – – – – – – – – –

Egypt 19 14 52 48 29 38 19 16 15 6 6 4 14 15 46 59

Ethiopia 52 45 38 45 10 10 62 – 21 – 1 – 2 – 14 –

India 29 18 44 55 26 27 12 10 15 9 17 19 14 15 42 48

Indonesia 19 14 41 42 39 44 27 23 15 10 9 22 9 14 40 30

Kenya 30 28 51 51 19 21 38 31 10 4 5 3 9 6 38 57

Nigeria 32 24 23 54 45 22 – – – – – – – – – –

Russia 17 4 35 61 48 35 – 17 – 2 – 10 – 10 – 61

Venezuela 5 6 34 42 61 52 17 – 5 – 3 – 9 – 66 –

Germany 2 1 61 69 38 30 – 8 – 1 – 40 – 11 – 40

Japan 2 1 60 71 38 28 9 13 5 2 25 27 10 12 52 47

U.S. 2 1 70 79 28 20 12 – 5 – 23 – 12 – 48 –

56

Figure 1. Comparing Perfect Bertrand Competition and Monopoly

Figure 2. Imperfect Bertrand Competition with Transaction Costs

Price (user fee)

P(Q)=W-aQ

Q QM QFB 0

PFB

PFB + fM

MC(Q)=c

MR(Q) DWLossM

Price (user fee)

P(Q)=W-aQ

Q B

KQ QFB 0

PFB

PFB + K

MC(Q)=c

MR(Q)

B

KDWLoss