The current issue and full text archive of this journal is available at www.emeraldinsight.com/0144-3585.

htm

JES 37,1

Endogenous corruption in economic development

Keith Blackburn
Economic Studies, Centre for Growth and Business Cycles Research, University of Manchester, Manchester, UK

4
Received 28 July 2008 Accepted 12 September 2008

Niloy Bose
Department of Economics, University of Wisconsin-Milwaukee, Milwaukee, Wisconsin, USA, and

M. Emranul Haque
Economic Studies, Centre for Growth and Business Cycles Research, University of Manchester, Manchester, UK
Abstract
Purpose The purpose of this paper is to present an analysis of the joint determination of bureaucratic corruption and economic development. Design/methodology/approach The analysis is based on a simple model of growth in which bureaucrats are employed as agents of the government to collect taxes from households. Findings Corruption is reected in bribery and tax evasion as bureaucrats conspire with households in providing false information to the government. Costly concealment of this activity leads to a loss of resources available for productive investments. The incentive for a bureaucrat to accept a bribe depends on economy-wide outcomes, which, in turn, depend on the number of other bureaucrats who accept bribes. The paper establishes the existence of multiple development regimes, together with the possibility of both history- and frequency-dependent equilibria. The predictions of the analysis accord strongly with recent empirical evidence. Originality/value The paper provides insights into the issue, and in doing so, makes further inroads to the macroeconomics of misgovernance. Keywords Corruption, Bribery, Economic development Paper type Research paper

1. Introduction Public sector corruption may be broadly dened as the illegal, or unauthorised, proteering by public ofcials who exploit their positions in public ofce to make personal gains. To many observers, this type of behaviour is an inevitable aspect of state intervention, which typically entails some transfer of responsibilities from the government to a bureaucracy in a principal-agent type relationship. Such delegation of authority endows the bureaucracy with administrative discretion that may be used to capture economic rents through side payments or bribes. These rents may be signicant and the incentive to seize them may be tempered only mildly by imperfect mechanisms of
Journal of Economic Studies Vol. 37 No. 1, 2010 pp. 4-25 q Emerald Group Publishing Limited 0144-3585 DOI 10.1108/01443581011012234

JEL classication D73, H26, O11 The authors are grateful for the nancial support of the ESRC (Grant No. L138251030). The usual disclaimer applies.

prevention based on costly and imprecise monitoring, together with inadequate and inappropriate penalties. A considerable amount of research has been devoted towards understanding in detail the causes and consequences of bureaucratic corruption. Most of this research has been partial equilibrium in nature, focusing on the microeconomic aspects of incentives, information and enforcement in motivating or deterring corrupt practices which inuence efciency and welfare (Banerjee, 1997; Carillo, 2000; Klitgaard, 1988, 1990, 1991; Mookherjee and Png, 1995; Rose-Ackerman, 1975, 1978, 1999; Shleifer and Vishny, 1993). Much less research has been directed towards analysing the joint determination of corruption activities and economic outcomes within the context of fully specied dynamic general equilibrium models. This is particularly notable given that the macroeconomic consequences of corruption have become an increasing concern to both economists and policy makers who have shared a deepening belief that a fundamental requirement for economic development is good quality governance[1]. In this paper, we present an analysis of corruption and growth that lends general support to this presumption, subject to some important qualications. By its nature, corruption is a clandestine activity which takes place away from the glare of publicity and which is difcult to measure empirically. Since the early 1980s, however, a number of organisations most notably, Business International Corporation (BIC), Political Risk Services Incorporated and Transparency International (TI) have provided measures of corruption that have gradually gained the condence of researchers. Known as corruption perception indices, the datasets are constructed on the basis of questionnaire surveys sent to networks of correspondents around the world, whose responses are used to compose a ranking of countries in terms of the extent to which corruption is perceived to exist. Whilst differing in various aspects (e.g. coverage, methodology and availability) and whilst being susceptible to the usual caveat associated with survey data (i.e. the possibility of bias), the indices are all highly correlated with each other and all highly correlated with key economic variables properties which suggest that they are, in fact, measuring the same phenomenon and that spuriousness is not a major issue. For these reasons, the indices have become widely accepted as providing reliable (possibly the best) estimates of corrupt activity[2]. Their publication has given rise to a urry of empirical investigations into the relationship between corruption, development and other phenomena. Two major ndings from these investigations are as follows. First, there is overwhelming evidence that corruption adversely affects growth and development. According to Mauro (1995), the principal mechanism through which this occurs is a change in private investment: an improvement in the corruption index by one standard deviation is estimated to increase investment by as much as 3 per cent of output. Other authors report correlations between growth and corruption that are consistently negative and signicant (Gyimah-Brempong, 2002; Keefer and Knack, 1997; Knack and Keefer, 1995; Li et al., 2000; Sachs and Warner, 1997)[3]. Second, there is also strong evidence that the relationship between corruption and growth is two-way causal: bureaucratic rent-seeking not only inuences, but is also inuenced by, the level of development. In Treisman (2000), it is found that corruption is generally much higher in rich countries than in poor countries, with as much as 50-73 per cent of the variations in corruption indices being explained by variations in per capita income levels. These ndings, supported in other studies (Ades and di Tella, 1999;

Endogenous corruption

JES 37,1

Fisman and Gatti, 2002; Montinola and Jackman, 1999; Paldam, 2002; Rauch and Evans, 2000), suggest that cross-country differences in the incidence of corruption owe much to cross-country differences in the level of prosperity. By the way of illustrating the relationship between corruption and development, we present some summary statistics in Table I, constructed on the basis of the World Banks income classication of countries, together with the corruption indices of BIC, International Country Risk Guide (ICRG) and TI. This table provides information about the corruption ratings among different groups of countries for various periods over the past 20 years, including four consecutive recent periods. A compelling feature of the data is the much higher corruption rating of poor countries than rich countries, irrespective of which index is used and which period is looked at. This is indicative of the negative correlation between corruption and development that has been reported in many empirical studies. Another notable feature one that has received very little publicity is the much greater diversity in corruption levels among middle-income countries, for which the range of each index is signicantly larger than the range for either low- or high-income countries in all of the periods. A comparison of the variances of the indices across different income groups gives the same impression: the variance for the middle-income group is consistently higher than the variance for either the low- or high-income groups, in spite of the denser and larger sample of the middle-income group. Further evidence of this can be seen in Figure 1 which plots the rolling standard deviation of corruption as one moves through the cross-country distribution of income[4]: as is clearly seen, corruption displays much greater variability within the middle range of the distribution than in either of the tails.
Index Total rangea Year Number of countries Totalb Low income Middle income Lower middle income Upper middle income High income Range of index Low income Middle income Lower middle income Upper middle income High income Variance of index Low income Middle income Lower middle income Upper middle income High income BIC 1.0-10.0 1980-1983 59 5 37 21 16 17 1.0-4.0 1.5-10.0 1.5-8.8 3.3-10.0 7.5-10.0 2.00 4.07 4.41 3.44 0.33 ICRG 1.0-6.0 1991-1997 113 33 59 43 16 21 1.4-4.0 1.0-5.0 1.0-5.0 1.1-5.0 4.4-6.0 0.55 0.79 0.67 1.14 0.34 TI 0.0-10.0 2002 2003 94 24 50 29 21 20 1.2-3.9 1.7-6.5 1.7-5.7 2.5-6.5 6.3-9.7 0.36 1.47 0.93 1.50 1.09 121 36 65 38 27 20 1.3-3.3 1.6-7.4 1.6-4.9 2.1-7.4 6.9-9.7 0.23 1.39 0.69 1.59 0.88

2001 85 20 45 23 22 20 1.0-3.4 2.0-7.5 2.0-5.3 2.8-7.5 6.6-9.9 0.33 1.39 0.96 1.22 0.95

2004 133 43 70 40 30 20 1.5-3.6 1.9-7.4 1.9-5.3 2.3-7.4 6.9-9.7 0.24 1.56 0.69 1.84 0.80

Table I. Corruption across countries

Notes: aGreater levels of corruption are indicated by lower values of the indices; bto facilitate comparison between the indices, oil exporting countries have been excluded from the sample. In particular instances, certain other countries have also been excluded due to questions about the reliability of the data

2.10

2.10

1.90

1.90

Endogenous corruption

1.70 Standard deviation of corruption

1.70

1.50

1.50

1.30

1.30

1.10

1.10

0.90

0.90

0.70

0.70

0.50 5.43

0.50 5.87 6.28 6.72 7.13 7.43 7.70 8.00 8.24 8.45 8.97 9.43 9.72 9.93 10.11 10.27 Log (GDP per capita)

Figure 1. Rolling standard deviation of corruption index

These observations suggest that there is more to the relationship between corruption and development than has been revealed so far in empirical studies. Whichever way one looks at it, the picture that emerges is one of considerable diversity in the incidence of corruption among middle-income countries, compared to the uniformly high levels of corruption among low-income countries and the uniformly low levels of corruption among high-income countries. The foregoing empirical observations pose three important challenges for theoretical research. The rst challenge is to provide a rigorous, formal account of why corruption can be damaging to economic development. Of the few such analyses that exist, Ehrlich and Lui (1999) and Sarte (2000) are credited with making two of the rst contributions[5]. The former develop a model in which corruption has the effect of diverting resources away from growth-promoting activities (investments in human capital) towards power-seeking activities (investments in political capital), while the latter presents a framework in which corruption causes resources to be diverted away from the formal (more efcient) sectors of the economy towards the informal (less efcient) sectors. Whilst revealing and insightful, these analyses do not delve too deeply into other important questions, such as what might give rise to corruption to begin with and what might cause corruption to persist or decline over time. The second challenge is to explain simultaneously why corruption may, itself, be affected by development. Doing this would allow one to account for the joint, endogenous determination of corruption and development in a relationship that is fundamentally two-way causal that is, a relationship where the quality of governance both inuences and is inuenced by the forces of growth. Some progress in this direction has been made by Blackburn et al. (2006) and Blackburn and Forgues-Puccio (2007) who model more fully the dynamic general equilibrium interactions between rent-seeking behaviour and economic activity. It is shown how these interactions can lead to threshold effects and multiple development regimes, and how the limiting outcome of an economy may depend crucially on initial conditions. The third challenge

JES 37,1

is to make even further in-roads by explaining also why corruption may differ markedly between countries at similar levels of development (as observed for middle-income countries). A seemingly natural way to address this is to appeal to differences in the institutions (Kunicova and Rose-Ackerman, 2005; Persson et al., 2003), regulations (Ali and Isee, 2003; Lederman et al., 2005; Wei, 2000) and cultures (Chang and Golden, 2007; Suphachalasai, 2005) that form the social fabric of societies. Yet, whilst having its merits, such an approach runs the risk of being almost tautological and of emphasising factors that often have remote and rather tenuous connections to the development process. Ideally, what one would like is an explanation that is grounded more rmly on economic fundamentals and that is fully unied with an account of the mutual dependence between development and corruption. As far as we know, this has yet to be tackled. The objective of this paper is to take on the above challenges. We do so by developing a simple model of growth in which public agents (bureaucrats) are delegated the responsibility for collecting taxes from private individuals (households) on behalf of the political elite (the government). Bureaucrats have the opportunity to engage in corrupt practices which are difcult to monitor by the government. Specically, bureaucrats may exploit their powers of public ofce to collude with households in bribery and tax evasion: a bribe to a bureaucrat holds the promise that the income of a household will be reported falsely and exempt from any tax. Such behaviour impacts on growth through its effects on aggregate savings and capital accumulation. A key implication of our analysis is that the incentive for a bureaucrat to engage in corruption depends on economy-wide outcomes which, in turn, depend on the behaviour of all other bureaucrats. As a consequence, bureaucratic decision-making entails strategic interactions that are capable of producing multiple, frequency-dependent equilibria associated with different (low and high) incidences of corruption. In general, such non-uniqueness is explained by appealing to the notion that, for one reason or another, individuals are more likely to be corrupt when others are corrupt and vice versa. For example, the more corrupt people there are, the less might be the probability that each one of them will be caught, the less might by the penalty that each one of them may incur and the less might be the moral costs, or social stigma, that each one of them feels. These ideas have been incorporated into several partial equilibrium models of corruption. Typical of these are the models of Andvig and Moene (1990) and Cadot (1987) in which multiple equilibria arise because a bureaucrats expected punishment for being corrupt is a decreasing function of the number of other corrupt bureaucrats[6]. In a slightly different vein, Tirole (1996) establishes multiple equilibria that are history-dependent due to group reputation effects, whereby good or bad actions in the past motivates good or bad actions in the present. Our own account of the phenomena stands in sharp contrast to these analyses and centres around the surplus that accrues to households and bureaucrats as a result of their illegal proteering. Ceteris paribus, the greater is the level of corruption the higher are the taxes that households must pay if the government is to balance its budget. In order to evade these higher taxes, households are willing to cede more in bribes which reinforces the rent-seeking incentives of bureaucrats. The upshot is that a bureaucrats expected gain from being corrupt depends positively on the number of other bureaucrats who are corrupt hence the possibility of contagious behaviour and, with this, multiple equilibria. We emphasise that this is only a possibility since there

are circumstances in our model where such behaviour does not arise and there exists a unique equilibrium. Signicantly, these circumstances relate to the level of economic development, as determined by the process of capital accumulation. This is another distinguishing feature of our analysis. Up to now, the question of how an economy may move from one equilibrium to another has been addressed largely on the basis of comparative static exercises (i.e. studying the effects of exogenous changes in parameter values). In our case, the selection of an equilibrium is partly endogenous, being linked to an economys position along its development path. The precise effect of corruption in our model is to reduce the amount of resources available for productive investments as bureaucrats seek other (less conspicuous, but costly) ways of disposing of their illegal income. In this way, our analysis allows for the endogenous interaction between corruption and development: on one hand, the selection of an equilibrium with a particular incidence of corruption is governed, in part, by aggregate economic activity; on the other hand, growth in economic activity through capital accumulation is determined by the equilibrium level of corruption. Together, with the above considerations, this interdependence means that our model is able to generate both frequency-dependent and history-dependent outcomes. According to our results, an economy may nd itself in one of three distinct types of development regime: the rst, a low development regime, is characterised by a unique equilibrium associated with a high incidence of corruption; the second, a high development regime, is also characterised by a unique equilibrium but one that entails a low incidence of corruption; the third, an intermediate development regime, is characterised by multiple equilibria with varying incidences of corruption. The existence of multiple equilibria means that different levels of corruption may be displayed by countries at similar stages of development. Consequently, and in accordance with the empirical evidence, our analysis is able to explain not only why there is more corruption in poor countries than in rich countries, but also why there is more diversity in corruption among middle-income countries. It is also able to account for persistence in both corruption and income inequalities across countries: transition from a low development (high corruption) regime to a high development (low corruption) regime is not inevitable in our model, and it is possible for an economy to become trapped in a vicious circle of widespread poverty and wholesale misgovernance. This accords with the casual observation that many of the most poor and corrupt countries in the past are among the most poor and corrupt countries today. The remainder of paper is organised as follows. In Section 2, we describe the economic environment in which agents make decisions. In Section 3, we study the incentives of agents to engage in corruption. In Section 4, we reveal the adverse effects of corruption on growth. In Section 5, we establish the reverse causation from development to corruption. In Section 6, we offer some concluding remarks. 2. The environment Time is discrete and indexed by t 0; . . . ; 1. There is a constant population of two-period-lived agents belonging to overlapping generations of non-altruistic families. Agents of each generation are divided into two groups of citizens private individuals (or households), of whom there is a xed measure of mass m, and public servants (or bureaucrats), of whom there is a xed measure of mass n , m[7]. Households are differentiated according to differences in their labour endowments,

Endogenous corruption

JES 37,1

10

which determine their relative incomes and their relative propensities to be taxed. Specically, we suppose that a fraction, m [ 0; 1, of households are endowed with l . 1 units of labour and are liable to pay tax, while the remaining fraction, 1 2 m, are endowed with only one unit of labour and are exempt from paying tax. Taxes are lump-sum and are collected by bureaucrats on behalf of the government which requires funding for public expenditures. For simplicity, we assume that each bureaucrat has one unit of labour endowment (which exempts him from taxation) and that each bureaucrat has jurisdiction over the same number, mm=n, of taxable households. Corruption arises from the incentive of a bureaucrat to conspire with a household in concealing information (the households income) from the government. In doing this, the bureaucrat expects to gain from his acceptance of a bribe and the household expects to gain from its evasion of tax. We assume that a fraction, h [ 0; 1, of bureaucrats are corruptible in this way, while the remaining fraction, 1 2 h, are non-corruptible, with the identity of each bureaucrat being unobservable by the government[8]. All agents are risk neutral, working (and saving) only when young and consuming only when old. Production of output is undertaken by rms, of which there is a continuum of unit mass. Firms hire labour from households and rent capital from all agents. All markets are perfectly competitive. 2.1 The government We envisage the government as providing public services which contribute to the efciency of output production (Barro, 1990). Expenditure on these services, gt is assumed to be a xed proportion, u [ 0; 1, of output. The government also incurs expenditures on bureaucrats salaries which are determined as follows. Any bureaucrat (whether corruptible or non-corruptible) can work for a rm, supplying one unit of labour to receive a non-taxable income equal to the wage paid to households. Any bureaucrat who is willing to accept a salary less than this wage must be expecting to receive compensation through bribery and is therefore immediately identied as being corrupt. As in other analyses (Acemoglu and Verdier, 1998), we assume that a bureaucrat who is discovered to be corrupt is subject to the maximum ne of having all of his income conscated (i.e. he is dismissed without pay). Given this, then no corruptible bureaucrat would ever reveal himself in the way described above. As such, the government can minimise its labour costs, while ensuring complete bureaucratic participation, by setting the salaries of all bureaucrats equal to the wage paid by rms to households[9]. The government nances its expenditures each period by running a continuously balanced budget. Its revenues consist of the taxes collected by bureaucrats from high-income households, plus any nes imposed on bureaucrats who are caught engaging in corruption. We denote by tt , the lump-sum tax levied on each high-income household. Since the government knows how much tax revenue is due in the absence of corruption (since it knows the number of taxable households and since it is responsible for setting taxes), any shortfall of revenue below this amount reveals that corruption is occurring. Under such circumstances, the government investigates the behaviour of bureaucrats using an imprecise monitoring technology. This technology implies that a bureaucrat who is corrupt faces a probability, p [ 0; 1, of avoiding detection, and a probability, 1 2 p, of being found out. The tax-evading household with whom the bureaucrat conspires faces the same probabilities of remaining anonymous and being exposed. In the event of the latter, the household is forced to pay its full tax liability[10].

2.2 Households Each young household of generation t is paid a wage, wt, from supplying inelastically its labour endowment to a rm. Depending on this endowment, a household is either a low-income earner and exempt from paying tax, or a high-income earner and liable to pay tax. In each case the household saves its entire net income as (non-consumable) capital which is rented to rms at the market rate of interest, rt 1, in order to nance old-age consumption. For a household with one unit of labour endowment, wt is equal to total labour income which is not subject to tax, so that wt is also equal to net income. Obviously, this type of household has no incentive to engage in tax evasion and its (expected) lifetime income, or utility, is simply rt 1wt[11]. For a household with l units of labour endowment, total labour income is lwt from which the government requires payment of the lump-sum tax, tt . This type of household may conspire with a corruptible bureaucrat in bribery and tax evasion. In the absence of such corruption, the household expects to earn a net income of lwt 2 tt [12]. In the presence of corruption, its expected net income depends on the amount of bribe paid to the bureaucrat and the probability of being caught. Let bt denote the bribe. In return for this, the bureaucrat agrees to dissemble the identity of the household by declaring that it is a low-income type and is therefore not liable to pay tax. With probability p, the household and bureaucrat succeed in their conspiracy and the households net income is lwt 2 bt . With probability 1 2 p, their collusion is exposed and the household is forced to pay its tax, implying a net income of lwt 2 bt 2 tt . Given these outcomes, we may write the expected lifetime utility of a high-income household as: ( if bt 0; r t1 lwt 2 tt ut 1 r t1 lwt 2 bt 2 1 2 ptt if bt . 0:

Endogenous corruption

11

2.3 Bureaucrats Each young bureaucrat of generation t is paid the salary wt from supplying inelastically his unit labour endowment to the government. Each bureaucrat is responsible for collecting taxes from mm=n households to which he reveals himself as being either corruptible or non-corruptible. Like all households, all bureaucrats save their entire income to nance old-age consumption. By denition, a non-corruptible bureaucrat is never corrupt. The income of such a bureaucrat is always wt, implying a lifetime utility of rt 1wt. By contrast, a corruptible bureaucrat may or may not be corrupt. If the latter, then he expects to receive an income of wt, as above. If the former, then his expected income depends on the bribes that he receives, the chances of being caught, the resources spent on trying to avoid detection and the penalties incurred if he is exposed. In general, corrupt individuals may try to remain inconspicuous by hiding their illegal income, by investing this income differently from legal income and by altering their patterns of expenditure. For the purposes of the present analysis, we assume that a corrupt bureaucrat must dispose of all side payments immediately if he is to stand any chance of not being caught: if he holds on to these payments, or invests them himself, then he is certain of being found out, in which case he ends up with nothing. The concealment of bribes is not without cost, however.

JES 37,1

12

Specically, we imagine that such subterfuge requires the assistance of some other group of individuals who specialise in the laundering of illegal income. These individuals are able to channel this income into areas where it is difcult to trace by the government, such as the informal sector of the economy and overseas bank accounts. In return for this service, a bureaucrat must pay a commission, equal to some proportion, d [ 0; 1, of the total bribe income that he seeks to conceal, mm=nbt . We also make the common assumption that black money (i.e. money that is unaccounted for) earns a lower rate of return than white money (i.e. money that is accounted for). This assumption captures the idea that hidden investments in the economys informal sector are unable to access the same support infrastructure that investments in the formal sector can. As a consequence, the productivity of the former is lower than that of the latter[13]. We model this by specifying the black market rate of return to be a fraction, r [ 0; 1, of the ofcial market rate of return, rt 1. Given this description of events, we may write the initial net income of a corrupt bureaucrat as wt mm=nbt with probability p and wt mm=nbt 2 f t with probability 1 2 p, where ft is the ne incurred if the bureaucrat is caught. In the case of the former, wt is invested in the formal sector to earn a return of rt 1, while mm=nbt is invested in the informal sector to earn a net return of rr t1 2 d. In the case of the latter, f t wt mm=nbt the bureaucrat being ned the full amount of his legal and illegal earnings. It follows that the expected lifetime utility of a corruptible bureaucrat is: 8 if bt 0; < r t1 wt mm vt 2 : p r t1 wt rr t1 2 d n bt if bt . 0: 2.4 Firms The representative rm produces output, yt, according to the following technology:
b g yt Al a t kt g t ;

A . 0; a; b; g [ 0; 1; b g , 1 where lt denotes labour and kt denotes capital[14]. The rm hires labour at the competitively determined wage rate wt and rents capital at the competitively determined rental rate rt. Prot maximisation implies:
a21 b g wt Aal t kt g t ; b21 g r t A bl a gt : t kt

4 5

3. From development to corruption Corruption occurs if a high-income household and a corruptible bureaucrat nd it mutually advantageous (or non-disadvantangeous) to conspire with each other in concealing information from the government. Under such circumstances, there is bribery and tax evasion. In the analysis that follows we study the individual incentives of private and public agents to behave in this way. As we shall see, these incentives depend on aggregate economic activity. Thus, we establish the rst direction of causation between corruption and development that runs from the latter to the former.

A high-income household is willing to pay a bribe if its expected utility from doing so is no less than its expected utility from not doing so. The maximum bribe that such a household is willing to concede is determined by strict equality of this condition. From equation (1), this maximum bribe payment is deduced as: bt p tt : 6

Endogenous corruption

Intuitively, the household is prepared to bribe a bureaucrat by no more than what it expects to save in taxes. Similarly, a corruptible bureaucrat is willing to accept a bribe if he expects to be no worse off from doing this than from not doing this. From equation (2), this requires that: bt $ n1 2 pr t1 wt : mmprr t1 2 d 7

13

Accordingly, the bureaucrat demands a higher bribe payment the more he expects to lose in legal income if he is caught and the less he expects to gain in illegal income if he is not caught. For corruption to take place, both equations (6) and (7) must be satised simultaneously. This yields the condition: ptt $ n1 2 pr t1 wt : mmprr t1 2 d 8

The key feature of this condition, alluded to above, is that it depends on the economy-wide variables tt , rt 1 and wt. We summarise the implications as follows. Proposition 1. A corruptible bureaucrat is more likely to be corrupt the higher is the level of taxes, the higher is the rate of interest and the lower is the level of wages. Proof. The left-hand-side of equation (8) is increasing in rt, whilst the right-handA side of equation (8) is decreasing in rt 1 and increasing in wt. As we shall see, each of the above variables can be expressed as a function of the existing stock of capital, kt. One therefore deduces that capital accumulation is an important factor in determining the incidence of corruption. We explore this in detail later when we also expose some further implications of the condition in equation (8). In particular, we shall reveal how the incentive for an individual bureaucrat to be corrupt may depend crucially on the number of other bureaucrats who are corrupt, inviting the possibility of multiple, frequency-dependent equilibria. Again, this is due to the fact that individual incentives are inuenced by aggregate outcomes. It is therefore appropriate that we rst turn our attention to the determination of these outcomes. 4. From corruption to development We proceed to study the behaviour of the economy under two opposite scenarios one in which no corruptible bureaucrat is corrupt and the other in which all corruptible bureaucrats are corrupt. Our ultimate objective in each case is to deduce the path of capital accumulation along which the economy evolves over time. This is obtained from the equilibrium condition that the total demand for capital is equal to the total supply of savings (except for any savings that are hidden). By showing how this path is affected by rent-seeking, we establish the second direction of causation between corruption and development that runs from the former to the latter. In conducting our

JES 37,1

14

analysis we make use of some of our earlier results and assumptions. Specically, we recall that gt uyt and observe that, in equilibrium, l t l lm 1 2 mm[15]. From f2 1 21 ak f , where equations (3)-(5), we then have g t Fukf t , wt Fl t and r t Fbkt a g 1=12g and f b=1 2 g. F Al u Consider, then, the case in which no corruptible bureaucrat is corrupt. The government obtains the maximum tax revenue of mmtt which is used to nance its expenditures on public services, gt, and bureaucrats salaries, nwt. The tax imposed on each high-income household is determined from the governments budget constraint as: ^t t g t nwt Ful an f ^kf kt ; t t : l mm mm 9

A household that wished to evade this tax would be willing to pay a bribe up to a ^ t pt ^t . Total savings in the economy comprise the savings of all maximum of b ^t and low-income households, 1 2 mmwt of all high-income households, mmlwt 2 t of all bureaucrats, nwt. Collecting these terms together, and exploiting (9), we may derive the following expression for capital accumulation: ^ t1 lwt 2 g t Fa 2 ukf ; K ^ kt ; k t 10

where we assume that a . u[16]. For future reference, we also note that, given equation ^ f21 , or: ^t1 Fbk (10), the equilibrium rate of interest is determined as r l 1
f f 2 1 ^ kt ^t1 FbFa 2 uf21 kt ;R r

11

Now consider the case in which all corruptible bureaucrats are corrupt. The total population of such bureaucrats is hn and the total population of bribe-paying households is hmm[17]. Within these groups, there is a fraction, p, of agents who evade detection by the government, while the remaining fraction, 1 2 p, are caught. From each bureaucrat in the former cohort the government receives zero tax revenue, whilst from each bureaucrat in the latter cohort the government receives mm=n t t in tax revenue plus wt mm=nbt in nes. The populations of non-corruptible bureaucrats and non-bribe-paying high-income households are 1 2 hn and 1 2 hmm, respectively. From each non-corrupt bureaucrat the government receives mm=ntt in tax revenues. As before, total government expenditure is equal to expenditures on public services, gt, plus expenditures on bureaucrats salaries, nwt. It follows from the governments budget constraint that the tax imposed on each high-income household is:

t ~t

~t g t 1 2 1 2 phnwt 2 1 2 phmmb 1 2 phmm

12

F{ul 1 2 1 2 phan} f f kt ; t ~kt ; 1 2 p 2 h l m m ~ t pt where b ~t is the maximum bribe that a tax-evading household would be willing to pay. Total savings of households comprise the savings of low-income households, 1 2 mmwt , of high-income households that do not pay bribes, 1 2 hmmlwt 2 t ~t , ~ t 2 1 2 pt and of high-income households that do pay bribes, hmmlwt 2 b ~t . Total savings of bureaucrats consist of the savings of non-corruptible bureaucrats,

1 2 hnwt , and of corruptible bureaucrats, hnpwt . Like before, each of these terms corresponds to the amount of income invested in the formal sector of the economy. We assume that any other income that is, the income from bribes which corrupt bureaucrats conceal and launder through the informal sector does not contribute to the amount of savings available for capital formation[18]. Together, with equation (12), the terms may be combined to express the capital accumulation process as:
f ~ t Fa 2 u 2 p 2 hmmt ~ t1 lwt 2 g t 2 phmmb ~ kt ; k ~kt ; K

Endogenous corruption

13

15

where we assume that [ ] . 0. As above, we also note for future reference that, given ~ f21 , or: ~t1 Fbk equation (13), the equilibrium interest rate is computed as r t 1
f f 2 1 ~ kt : ~t1 FbFa 2 u 2 p 2 hmmt r ;R ~f21 kt

14

We are now in a position to establish our rst set of key results. In doing so we recall that f , 1 and appeal to the fact that p 2 . 1 2 p by virtue of a similar condition that we impose later in order to ensure that our analysis is non-trivial. Proposition 2. For any given existing stock of capital, kt, the equilibrium levels of current taxes, future capital and future interest rates in the absence and presence of ^ t1 . k ~ t1 and r ^t1 , r ~t1 . ^t , k ^t , t corruption satisfy t Proof. The rst result follows from equations (9) and (12), the second from equations (10) and (13), and the third from equations (11) and (14). A Thus, taxes are lower, capital accumulation is higher and interest rates are lower in the absence than in the presence of corruption. Intuitively, the prospect of lost tax revenues under corruption means that the government must raise taxes on high-income households in order to satisfy its budget constraint. Each of these households is ^ t as a way of evading its higher tax ~t . b therefore willing to pay a larger bribe b liability. The diversion of bribe income towards hidden, unproductive activities reduces ~ t ), leading to higher aggregate savings and capital accumulation (by the amount phmmb interest rates by virtue of diminishing returns to capital in output production. In Figure 2, we illustrate the two capital accumulation paths described by equations (10) and (13). Associated with each one is a unique steady state level of capital either ^ * Fa 2 u1=12f in the case of the former or k ~ * Fa 2 u 2 p 2 hmmt k ~1=12f in the case of the latter. In the absence of corruption the economy evolves along the ^* , ^ and converges to the high steady state equilibrium, k high development path, K whilst in the presence of corruption the economy evolves along the low development ~* . ~ and converges to the low steady state equilibrium, k path, K 5. Corruption and development in general equilibrium Our analysis has so far revealed how rent-seeking by public ofcials can both inuence, and be inuenced by, aggregate economic activity. It remains for us to bring these two aspects together so as to provide a complete account of the joint, endogenous determination of corruption and development. We do so by returning to the incentive condition in equation (8) which identies the circumstances under which an individual bureaucrat will choose whether or not to be corrupt. Proceeding as above, we study this choice under each of the two opposite scenarios in which corruption is wholly absent or entirely widespread.

JES 37,1

kt+1 () K

16
K ()
~

Figure 2. Capital accumulation

45 k*
~ c k1

k2c

* k

kt

As we have seen, the equilibrium levels of taxes and interest rates differ between ^ t 1 ^t and r non-corrupt and fully-corrupt environments: in the case of the former we have t ~t1 in equations (10) in equations (9) and (11), whilst in the case of the latter we have t ~t and r and (14). Accordingly, equation (8) actually describes a pair of conditions that are context-specic. Recalling the expression for wages, wt, we may write these conditions as: ^ kt $ R ^ dp 2 l mmt ^; ;V 2 ^ 2 F1 2 pan rp l mmt 15

~ kt $ R

rp 2 l mmt ~2

dp 2 l mmt ~ ~: ;V F1 2 pan

16

In words, equation (15) is the condition for an individual corruptible bureaucrat to be corrupt, given that no other such bureaucrat is corrupt, whilst equation (16) is the condition for an individual corruptible bureaucrat to be corrupt, given that all other such ^ . 0 and V ~ . 0 if there is to be bureaucrats are corrupt. Evidently, it is necessary that V any possibility that a bureaucrat will choose not to be corrupt under each scenario. A sufcient condition for this, alluded to earlier, is rp 2 . 1 2 p which we assume to be satised. ^ and R ~ are decreasing monotonically in kt. Note also that Note that both R ^ , R ^ .V ~ for any given kt ~ (since t ^, t V ~ from our previous results), while R (again from our previous results). Given these observations, we may dene two critical c ^ c ^ ^ ^ levels of capital, kc 1 and k2 , such that the following hold: Rk1 V, with R . V for c c c ^ ^ ~ ~ ~ ~ all kt , k1 and R , V for all kt . k1 ; and Rk2 V, with R . V for all kt , kc 2 ~ , V ~ for all kt . kc . Evidently, kc , kc . We are now in a position to establish and R 2 1 2 some further key results. Proposition 3. For kt , kc 1 there exists a unique equilibrium in which all corruptible bureaucrats are corrupt. c ~ ~ Proof. Suppose that kt , kc 1 , implying kt , k2 as well. Then R . V and ^ ^ R . V, in which case it pays each corruptible bureaucrat to be corrupt, irrespective

of whether other such bureaucrats are corrupt or non-corrupt. No bureaucrat has an incentive to deviate from corrupt behaviour, whilst each bureaucrat has an incentive to deviate from non-corrupt behaviour. The only equilibrium is one in which all bureaucrats are corrupt. A This result demonstrates that low levels of development are associated with high (maximum) levels of corruption. Proposition 4. For kt . kc 2 , there exists a unique equilibrium in which no corruptible bureaucrat is corrupt. c ^ ^ Proof. Suppose that kt . kc 2 , implying kt . k1 as well. Then R , V and ~ , V ~ , in which case it pays each corruptible bureaucrat to be non-corrupt, R irrespective of whether other such bureaucrats are non-corrupt or corrupt. No bureaucrat has an incentive to deviate from non-corrupt behaviour, whilst each bureaucrat has an incentive to deviate from corrupt behaviour. The only equilibrium is one in which all bureaucrats are non-corrupt. A This result demonstrates that high levels of development are associated with low (zero) levels of corruption. c Proposition 5. For kt [ kc 1 ; k2 there are multiple equilibria in which all corruptible bureaucrats are either corrupt or non-corrupt. c ^ ^ ~ ~ Proof. Suppose that kt [ kc 1 ; k2 . Then R . V, but R , V, in which case it pays each corruptible bureaucrat to be either corrupt or non-corrupt, depending on whether other such bureaucrats are corrupt or non-corrupt. No bureaucrat has an incentive to deviate from corrupt behaviour if all others are corrupt, and no bureaucrat has an incentive to deviate from non-corrupt behaviour if all others are non-corrupt. There are two equilibria, one in which all bureaucrats are corrupt and the other in which all bureaucrats are non-corrupt. A This result demonstrates that intermediate levels of development may be associated with either low or high levels of corruption. We illustrate the above results in Figure 3, from which we are led to distinguish between three types of development regime for the economy. The rst a low development regime is one in which the incidence of corruption is always at its maximum for any given level of capital below the lower threshold level, kc 1 . The second

Endogenous corruption

17

R () () R
c k1 ~

k2c

kt

Figure 3. Equilibrium corruption

JES 37,1

18

a high development regime is one in which the incidence of corruption is always at its minimum for any given level of capital above the upper threshold level, kc 2 . And the third an intermediate development regime is one in which the incidence of corruption may be either at its maximum or at its minimum for any given level of capital between the two thresholds. The intuition is as follows. Each corruptible bureaucrat chooses to be corrupt or non-corrupt according to whether the condition in equation (8) is satised or violated. This condition depends on taxes, interest rates and wages, all of which depend on the existing aggregate stock of capital (measuring the level of development) and the rst two of which depend also on the aggregate incidence of corruption (reecting the behaviour of all other bureaucrats). At sufciently low or sufciently high levels of development, a bureaucrats incentive to behave in one way or another is unaffected by how other bureaucrats are behaving: what matters most is the level of development, itself. For capital stocks below kc 1 , interest rates are always high enough to ensure that the condition in equation (8) is satised[19]. As such, a corruptible bureaucrat will always be corrupt, irrespective of what others around him may be doing. Since this is true for all such bureaucrats, then the only equilibrium from which there is no incentive to deviate is one in which corruption is the unique choice of strategy. Conversely, for capital stocks above kc 2 , interest rates are always low enough such that the condition in equation (8) is violated. In this case a corruptible bureaucrat will never be corrupt, regardless of what others may be up to. Being true for all such bureaucrats, this means that the only equilibrium from which defection will not occur is one in which non-corruption is the singular choice of action. In contrast to these scenarios, a bureaucrats incentive to transgress at intermediate stages of development depends critically on the exploits of others. For any given stock of capital between kc 1 and kc 2 , the condition in equation (8) is satised if corruption is widespread but is violated if corruption is absent. A corruptible bureaucrat will now be corrupt or non-corrupt according to whether other such bureaucrats are corrupt or non-corrupt. Consequently, there are two candidate equilibria that are frequency-dependent and that are equally likely to arise. As mentioned previously, our account of multiple equilibria is quite different from other accounts that currently exist. In our case, multiplicity arises because, ceteris paribus, the joint surplus of a household and a bureaucrat from colluding with each other is higher (lower) when corruption in total is higher (lower). This follows from our earlier result that, for any given level of capital, a higher (lower) incidence of corruption is associated with a higher (lower) level of taxes as the government strives to maintain budget balance. Higher (lower) taxes means that households are willing to pay larger (smaller) bribes which, in turn, means that each bureaucrat has a stronger (weaker) incentive to engage in rent-seeking. In this way, both good and bad behaviour can be contagious as a bureaucrats compliance in corruption may depend critically on the compliance of others. In fact, the diversity of outcomes at intermediate levels of development is greater than what we have suggested so far. Each of the equilibria that has been constructed is a pure strategy equilibrium in which all corruptible bureaucrats are either corrupt or non-corrupt. But there also exists a mixed strategy equilibrium in which bureaucratic behaviour is heterogenous that is, an equilibrium in which a fraction, 1 [ 0; 1, of corruptible bureaucrats are corrupt, while the remaining fraction, 1 2 1, are non-corrupt. We establish this in an Appendix by c demonstrating that, for each kt [ kc 1 ; k2 , there exists an 1 such that the condition in

equation (8) holds with equality. It is therefore possible for a middle-income country to be in one of three equilibria where the incidence of corruption is high, low or somewhere in between. To many observers, it is not surprising that the relationship between corruption and development may sometimes be a little fragile. Indeed, there is a widely-held view that, at least in the rst instance, development may do little to reduce (and may even foster) corruption as the process of modernisation (including economic, political and social reforms) brings with it new incentives and new opportunities for public agents to engage in corrupt practices. For example, it is often alleged that this has been true in countries undergoing transition from controlled to more market-oriented economies (Bardhan, 1997; Basu and Li, 1998). In addition to the above, our analysis is able to explain why corruption and poverty may co-exist as persistent, rather than transient, phenomena. Recall Figure 2 which shows the capital accumulation paths for a non-corrupt and fully corrupt economy, as given in equations (10) and (13). Imagine the two threshold levels of capital being c ~ ~ located at some points such that kc 1 . k* and k2 , k* . The economy is on the low c ^ , for kt . kc and ~ , for kt , k , the high development path, K development path, K 1 2 c c either of the paths for kt [ k1 ; k2 . What transpires from this scenario is a poverty trap ~ * . In other words, if the economy is poor and corrupt to begin with equilibrium at k (i.e. if its initial capital stock is below kc 1 ), then it will be destined to remain poor and corrupt unless fundamental changes take place so as to dictate otherwise. For example, exogenous shifts in the stock of capital may cause a switch in development regime by pushing the economy above kc 1 . Alternatively, changes in the values of structural parameters may produce a similar turn of events by altering the transition function and ~ * . kc . In both cases a switch in regime is more likely to the threshold, itself, such that k 1 occur the closer is an economy to kc 1 to begin with. Accordingly, should circumstances change in these ways, then it is those countries at the upper end of the distribution below kc 1 that are most likely to feel the effects, while those in the lower tail remain as they are. Even for the former, however, there is no guarantee that the result would be low corruption and high growth, nor any assurance that the upper threshold, kc 2 , would also be breached: of the good and bad equilibria that exist between the two thresholds, the economy may nd itself stuck in the latter as corruption emerges as the equilibrium strategy. These observations suggest that the divisions between poor and rich, corrupt and non-corrupt, economies are unlikely to vanish quickly or easily, if at all. 6. Conclusions Public sector corruption is pervasive throughout the world. In one form or another, and to a lesser or greater degree, it has existed, and continues to exist, in all societies. Over the past few years, there has been a growing concern among the academic community and international organisations about the causes and consequences of corrupt behaviour within government bureaucracies. This has been motivated by a strengthening conviction that good quality governance is essential for sustained economic development and that corruption in the public sector is a major impediment to growth and prosperity. Recent innovations at the empirical level have allowed this conviction to be tested, and there is now a large body of evidence to support it. By contrast, there remains relatively little by way of formal theoretical analysis that would lend rigour and precision to the arguments involved. Our objective in this paper has been to provide such an analysis.

Endogenous corruption

19

JES 37,1

20

We have dened public sector corruption in the usual way as the abuse of authority by bureaucratic ofcials who exploit their powers of discretion, delegated to them by the government, to further their own interests by engaging in rent-seeking activities. We have also addressed the archetypal form of public sector corruption, whereby a bureaucrat is bribed by a private individual to conspire in the concealment of valuable information from the government. Of course, to the extent that bribery entails a transfer of resources between agents, there need not be any net social costs associated with such behaviour. As with any type of illegal or unauthorised activity, however, there are costs to both individuals and society of deception and secrecy, on the one hand, and detection and prosecution, on the other. In our case, corruption results in a loss of resources available for investment such that capital accumulation is depressed. It has been suggested elsewhere that corruption may also result in a misallocation of resources towards inefcient investments with similar consequences. Either way, the costs of corruption are potentially signicant, especially since it takes only small changes in the growth rate to produce substantial cumulative gains or losses in output and welfare. Our analysis respects the notion that bureaucratic corruption not only inuences, but is also inuenced by, economic development. This two-way causality is reected in the existence of threshold effects and multiple equilibria, which allow us to explain why the incidence of corruption may vary markedly across countries, even if countries share essentially the same structural characteristics. At any point in time, an economy may be located in a low development regime, a high development regime or an intermediate development regime. Cross-country variations in the level of corruption may occur both across and within these regimes. For example, two otherwise identical economies may end up with very different levels of corruption if one of them is in the low regime and the other is in the high regime, or if both of them lie in the intermediate regime. The predictions that follow from this accord well with the empirical observations of a high incidence of corruption among low-income countries, a low incidence of corruption among high-income countries and a diverse incidence of corruption among middle-income countries. The results are also consistent with the idea of persistence in corruption since transition from one regime to another is not inevitable but requires the crossing of a threshold that may be prohibitive. Of course, there are many other factors besides economic considerations that may help to explain why corruption levels differ across countries. The recent empirical literature suggests a number of intriguing possibilities. Yet even after controlling for these factors, economic development remains highly signicant and is undoubtedly a major determinant. The relationship between corruption and development is an issue on which much has been written but about which there is still much to learn. Our intention in this paper has been to provide further insights into the issue and, in doing so, to make further in-roads to the macoreconomics of misgovernance.
Notes 1. The concept of governance is broader than that of corruption, though the two are closely connected: just as bad governance fosters corruption, so corruption undermines good governance. For an appreciation of the importance of corruption to international policy makers, see the World Bank and IMF web sites: www.worldbank.org/publicsector/

anticorrupt and www.imf.org/external/np/exp/facts/gov. For some excellent reviews of the litera ture on corruption, see Bardhan (1997), Jain (2001), Rose-Ackerman (1998, 1999) and Tanzi (1998). 2. For more detailed discussions, see Ades and di Tella (1997), Jain (1998), Tanzi and Davoodi (1997) and Treisman (2000). The only other (objective) measure of corruption that exists is the number of public ofcials who have actually been convicted for corrupt offences. This has been used in several studies, but it is not without its problems. Higher rates of conviction do not necessarily mean that there is higher corruption, but may merely reect improved methods of detection and prosecution. More fudamentally, numbers of convictions provide information only on corrupt practices that are exposed, which arguably account for only a small fraction of total corrupt activity. 3. By contrast, there is very little evidence to support the so-called speed money hypothesis, according to which corruption (bribery, in particular) is a means of improving efciency by helping to circumvent cumbersone regulations (red tape) in the bureaucratic process (Huntington, 1968; Leff, 1964; Leys, 1970). 4. We compute this rolling standard deviation based on a window of 20 countries which we slide along the income distribution, beginning with the poorest countries. 5. In a purely static context, Acemoglu and Verdier (1998, 2000) conduct a general equilibrium analysis of how corruption may form part of an optimal allocation in which market failure is traded off against government failure. 6. The incidence of crime has been explained in a similar way. In Sah (1991), for example, an individual is more (less) likely to engage in criminal activity if there are many (few) others engaged in such activity because the chances that he will be caught are lower (higher). 7. We assume that agents are differentiated at birth according to their abilities and skills. A population of m agents lack the skills necessary to become bureaucrats, while a population of n agents possess these skills. The latter are induced to become bureaucrats by an allocation of talent condition established below. Thus, as in other analyses (Blackburn et al., 2006; Blackburn and Forgues-Puccio, 2007; Sarte, 2000), we abstract from issues relating to occupational choice. In doing so, we are able to simplify the analysis by not having to consider possible changes in the size of the bureaucracy and possible changes in the level of corruption that may result from this. 8. This assumption may be thought of as capturing differences in the propensities of bureaucrats to engage in corruption, whether due to differences in prociencies at being corrupt or differences in moral attitudes towards being corrupt (Acemoglu and Verdier, 2000; Besley and McLaren, 1993; Blackburn et al., 2006; Tirole, 1996). The main purpose of the assumption is to allow us to determine the wages of bureaucrats in a relatively straightforward way that does not demand additional assumptions about how public sector pay is determined. In fact, all we need for this purpose is that there be at least one bureaucrat who is non-corruptible all other bureaucrats may well be potential transgressors. 9. This has the usual interpretation of an allocation of talent condition. The government cannot force any of the n potential bureaucrats to actually take up public ofce, but it induces all of them to do so by paying what they would earn elsewhere. 10. As indicated earlier, it is possible to obtain multiple equilibria by allowing p to depend (positively) on the number of corrupt individuals. We deliberately avoid doing this in order to highlight a quite different mechanism for generating such phenomena. We also assume that monitoring is costless for the government, though the model could be extended by relaxing this assumption (and perhaps allowing p to be a function of monitoring expenditures) without altering its main implications. To a large extent, our results would be strengthened in the sense that there would be an additional loss of resources from corruption.

Endogenous corruption

21

JES 37,1

Likewise, our results would not change substantially if one were to assume that, in addition to paying its tax liability, a household is ned or punished in some other way if it is caught trying to evade taxes. Again, we choose not to include this for simplicity. 11. As we shall see, rt 1 is a function of currently observable variables and is therefore known to agents at time t.

22

12. Throughout our analysis, we assume appropriate restrictions on parameter values to ensure that lwt 2 tt . wt . This means that after-tax incomes are always positive and that high-income households are always better off than low-income households. 13. The assumption is also consistent with the implications of various government schemes that have been implemented to bring black money back into circulation. One such scheme (exemplied by the 1981 Special Bearer Bond Act in India) has been for governments to issue bearer bonds in xed denominations, the proceeds from which on reaching maturity are allowed to be introduced into regular books of account without questions being asked about the money that was originally exchanged for the bonds. The rate of interest on such bonds is substantially lower than the market rate. Another approach, based on the principle of voluntary disclosure, has been to allow individuals to report previously undisclosed income which is taxed at a high rate but which grants an individual immunity from prosecution. Because of the high tax rate, the return on this income is again lower than the market rate. 14. The parameter restriction b g , 1 ensures the existence of a steady state level of capital associated with a strictly concave capital accumulation path. 15. This latter expression denes equilibrium in the labour market, where the total supply of labour is equal to the labour supply of high-income households, lmm, plus the labour supply of low-income households, 1 2 mm. 16. Since au is the share of labour (government expenditure) in national income, this assumption is justied empirically. 17. This follows from the fact that each bureaucrat colludes with mm=n households. 18. This assumption may be relaxed either partially or wholly without altering our results. In the rst instance, one could allow for some, but not all, illegal income to contribute to productive investments. In the second instance, one would still recognise the fact that investments in the informal sector are less productive than investments in the formal sector. Either way, corruption would impact on capital accumulation. Our stronger assumption serves merely to save on notation and to simplify exposition. 19. Observe that the dependence of taxes and wages on capital is irrelevant since both variables are proportional to kf t which therefore cancels out in equation (8). References Acemoglu, D. and Verdier, T. (1998), Property rights, corruption and the allocation of talent: a general equilibrium approach, Economic Journal, Vol. 108, pp. 1381-403. Acemoglu, D. and Verdier, T. (2000), The choice between market failures and corruption, American Economic Review, Vol. 90, pp. 194-211. Ades, A. and di Tella, R. (1997), The new economics of corruption: a survey and some new results, Political Studies, Vol. 45, pp. 496-515 (special issue). Ades, A. and di Tella, R. (1999), Rents, competition and corruption, American Economic Review, Vol. 89, pp. 982-93. Ali, M.A. and Isee, H.S. (2003), Determinants of economic corruption: a cross-country comparison, Cato Journal, Vol. 22, pp. 449-66.

Andvig, J.C. and Moene, K.O. (1990), How corruption may corrupt, Journal of Economic Behaviour and Organisation, Vol. 13, pp. 63-76. Banerjee, A.V. (1997), A theory of misgovernance, Quarterly Journal of Economics, Vol. 112, pp. 1289-332. Bardhan, P. (1997), Corruption and development: a review of issues, Journal of Economic Literature, Vol. 35, pp. 1320-46. Barro, R.J. (1990), Government spending in a simple model of endogenous growth, Journal of Political Economy, Vol. 98, pp. S103-25. Basu, S. and Li, D.D. (1998), Corruption in transition, Working Paper No. 161, University of Michigan Business School, Ann Arbor, MI. Besley, T. and McLaren, J. (1993), Taxes and bribery: the role of wage incentives, Economic Journal, pp. 119-41. Blackburn, K. and Forgues-Puccio, G. (2007), Distribution and development in a model of misgovernance, European Economic Review, Vol. 51, pp. 1534-63. Blackburn, K., Bose, N. and Haque, M.E. (2006), The incidence and persistence of corruption in economic development, Journal of Economic Dynamics and Control, Vol. 30, pp. 2447-67. Cadot, O. (1987), Corruption as a gamble, Journal of Public Economics, Vol. 33, pp. 223-44. Carillo, J.D. (2000), Corruption in hierarchies, Annales dEconomie et de Statistique, Vol. 10, pp. 37-61. Chang, E. and Golden, M. (2007), Electoral systems, district magnitude and corruption, British Journal of Political Science, Vol. 37, pp. 115-37. Ehrlich, I. and Lui, F.T. (1999), Bureaucratic corruption and endogenous economic growth, Journal of Political Economy, Vol. 107, pp. 270-93. Fisman, R. and Gatti, R. (2002), Decentralisation and corruption: evidence across countries, Journal of Public Economics, Vol. 83, pp. 325-45. Gyimah-Brempong, K. (2002), Corruption, economic growth and income inequality in Africa, Economics of Governance, Vol. 3, pp. 183-209. Huntington, S.P. (1968), Political Order in Changing Societies, Yale University Press, New Haven, CT. Jain, A.K. (Ed.) (1998), The Economics of Corruption, Kluwer Academic, Boston, MA. Jain, A.K. (2001), Corruption: a review, Journal of Economic Surveys, Vol. 15, pp. 71-116. Keefer, P. and Knack, S. (1997), Why dont poor countries catch up? A cross-national test of an institutional explanation, Economic Inquiry, Vol. 35, pp. 590-602. Klitgaard, R. (1988), Controlling Corruption, University of California Press, Berkeley, CA. Klitgaard, R. (1990), Tropical Gangsters, Basic Books, New York, NY. Klitgaard, R. (1991), Adjusting to Reality: Beyond State Versus Market in Economic Development, ICS Press and International Centre for Economic Growth, San Francisco, CA. Knack, S. and Keefer, P. (1995), Institutions and economic performance: cross-country tests using alternative institutional measures, Economics and Politics, Vol. 7, pp. 207-27. Kunicova, J. and Rose-Ackerman, S. (2005), Electoral rules and constitutional structures as constraints on corruption, British Journal of Political Science, Vol. 35, pp. 573-606. Lederman, D., Loayza, N.V. and Soares, R.R. (2005), Accountability and corruption: political institutions matter, Economics and Politics, Vol. 17, pp. 1-35. Leff, N.H. (1964), Economic development through bureaucratic corruption, in Jain, A.K. (Ed.), The Economics of Corruption, Kluwer Academic, Boston, MA.

Endogenous corruption

23

JES 37,1

Leys, C. (1970), What is the problem about corruption?, in Heiden-heimer, A.J. (Ed.), Political Corruption: Readings in Comparative Analysis, Holt Reinehart, New York, NY. Li, H., Xu, L.C. and Zou, H. (2000), Corruption, income distribution and growth, Economics and Politics, Vol. 12, pp. 155-82. Mauro, P. (1995), Corruption and growth, Quarterly Journal of Economics, Vol. 110, pp. 681-712. Montinola, G.R. and Jackman, R.W. (1999), Sources of corruption: a cross-country study, British Journal of Political Studies, Vol. 32, pp. 147-70. Mookherjee, D. and Png, I.P.L. (1995), Corruptible law enforcers: how should they be compensated?, Economic Journal, Vol. 105, pp. 145-59. Paldam, M. (2002), The big pattern of corruption, economics, culture and seesaw dynamics, European Journal of Political Economy, Vol. 18, pp. 215-40. Persson, T., Tabellini, G. and Trebbi, F. (2003), Electoral rules and corruption, Journal of the European Economic Association, Vol. 1, pp. 958-89. Rauch, J.E. and Evans, P.B. (2000), Bureaucratic structure and bureaucratic performance in less developed countries, Journal of Public Economics, Vol. 76, pp. 49-71. Rose-Ackerman, S. (1975), The economics of corruption, Journal of Public Economics, Vol. 4, pp. 187-203. Rose-Ackerman, S. (1978), Corruption: A Study in Political Economy, Academic Press, New York, NY. Rose-Ackerman, S. (1998), Corruption and development, paper presented at Annual World Bank Conference on Development Economics 1997, Washington, DC, pp. 35-57. Rose-Ackerman, S. (1999), Corruption and Government: Causes, Consequences and Reform, Cambridge University Press, Cambridge. Sachs, J.D. and Warner, A.M. (1997), Sources of slow growth in African economies, Journal of African Economics, Vol. 6, pp. 335-76. Sah, R.K. (1991), Social osmosis and patterns of crime, Journal of Political Economy, Vol. 99, pp. 1272-95. Sarte, P.-D. (2000), Informality and rent-seeking bureaucracies in a model of long-run growth, Journal of Monetary Economics, Vol. 46, pp. 173-97. Shleifer, A. and Vishny, R. (1993), Corruption, Quarterly Journal of Economics, Vol. 108, pp. 599-617. Suphachalasai, S. (2005), Bureuacratic corruption and mass media, Environmental Economy and Policy Research working paper, University of Cambridge, Cambridge. Tanzi, V. (1998), Corruption around the world: causes, scope and cures, IMF Working Paper No. 98/63. Tanzi, V. and Davoodi, H. (1997), Corruption, public investment and growth, IMF Working Paper No. WP/97/139. Tirole, J. (1996), A theory of collective reputation (with applications to the persistence of corruption and to rm quality), Review of Economic Studies, Vol. 63, pp. 1-22. Treisman, D. (2000), The causes of corruption: a cross-national study, Journal of Public Economics, Vol. 76, pp. 399-457. Wei, S.J. (2000), Local corruption and global capital ows, Brookings Papers on Economic Activity, Vol. 2, pp. 303-52.

24

Further reading Rose-Ackerman, S. (1996), Democracy and the grand corruption, International Social Science Journal, Vol. 158, pp. 365-80. Appendix We establish the existence of a mixed strategy equilibrium in the intermediate development c regime. Suppose that, for kt [ kc 1 ; k2 , there is a fraction, 1 [ 0; 11 2 1, of corruptible bureaucrats who are corrupt (non-corrupt), with a corresponding fraction, 1h1 2 1h, of high-income households who are bribe payers (non-bribe payers). Proceeding in the usual way, we may derive expressions for taxes:

Endogenous corruption

25

t t

t gt 1 2 1 2 p1hnwt 2 1 2 p1hmmdb 1 2 p1hmm

A1

F{ul 1 2 1 2 p1han} f f k ;t 1kt ; {1 2 1 2 1 2 pdp1h}l mm t capital accumulation:

f  t1 lwt 2 gt 2 1hmm1 2 db  t Fa 2 u 2 p1hmm1 2 dt k 1kt ;

A2

and interest rates:

ff21  kt ; 1; t1 FbFa 2 u 2 p1hmm1 2 dt ;R r 1f21 kt

A3

 t pt where b t . The condition for a corruptible bureaucrat to be corrupt is:  kt ; 1 $ R

dp 2 l mmt 1

rdp 2 l mmt 1  1: ;V 2 F1 2 pan

A4

^t , t It is straightforward to verify that, for a given kt and a given 1 [ 0; 1; t ~t (hence t , t ^t , b ^ t1 . k ^ .V ~ t ), k ~ t1 , r t , b  t 1 . k ~ . It is also straightforward  .V ^t1 , r ~t1 and V  t 1 , r b ^t ; k ^ t1 ; r ^ if 1 0, while t t b  t1 k  V ^t1 and V t1 r ^t ; b to verify that t ~t ; t t t t ~ ~   ~   is increasing in ~  bt bt ; kt1 kt1 ; rt1 rt1 and V V if 1 1. Finally, we note that R  is decreasing in 1. In terms of Figure 1, these properties imply that, for any given 1, while V ^ and R ~ , while the line V  always lies between the curves R  1 [ 0; 1, the curve R ^ and V ~ . It follows that, within the region kc ; kc , there is a single always lies between the lines V 1 2  and V  . This means that, for any given kt [ kc ; kc there exists an intersection between R 1 2   1 [ 0; 1 such that R V , implying that each corruptible bureaucrat is indifferent between being corrupt and non-corrupt. This 1 is the fraction of corrupt corruptible bureaucrats that supports a mixed strategy equilibrium.

Corresponding author Niloy Bose can be contacted at: nbose@uwm.edu

To purchase reprints of this article please e-mail: reprints@emeraldinsight.com Or visit our web site for further details: www.emeraldinsight.com/reprints