谁必须支付贿赂以及支付金额为多少？来自企业横截面分析的证据

\WP5Z14 S6 POLICY RESEARCH WORKING PAPER 248 6 W ho M ust Pay Bribes Ugandan firms typically have to pay bribes when dealing and How M uch? with public officials whose actions directly affect the Evidence from a Cross-Section firms' business operations. How much? The rnore a firm of Firms can pay, the more it has to pay. Jakob Svensson The World Bank Development Research Group Macroeconomics and Growth H November 2000  | POLICY RESEARCH WORKING PAPER 2486 Summary findings Svensson exploits a unique data set on corruption business operations. And the amount paid in bribes is not containing information about estimated bribe payments a fixed sum for a set of public services but depends on by Ugandan firms. To guide the empirical analysis he the firm's ability to pay. develops a simple rent-extortion model, which yields Controlling for other potential explanations of the predictions on both the incidence of bribery and the relationship between "ability to pay" and equilibrium amount paid. Both predictions are consistent with the graft, Svensson shows that the more a firm can pay, the data. more it has to pay. Firms typically have to pay bribes when dealing with public officials whose actions directly affect the firms' This paper-a product of Macroeconomics and Growth, Development Research Group-is part of a larger effort in the group to study the causes and consequences of corruption. Copies of the paper are available free from the World Bank, 1818 H Street NW, Washington, DC 20433. Please contact Rina Bonfield, room MC3-354, telephone 202-473-1248, fax 202- 522-3518, email address abonfield@worldbank.org. Policy Research Working Papers are also posted on the Web at www.worldbank.org/research/workingpapers. The author may be contacted at jakob.svensson@iies.su.se. November 2000. (43 pages) The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the view of the World Bank, its Executive Directors, or the countries they represent. Produced by the Policy Research Dissemination Center  Who Must Pay Bribes and How Much? Evidence from a cross-section of firms* Jakob Svenssont *Not for citation without permission from the author. Views expressed here do not represent official opinions of The World Bank. tInstitute for International Economic Studies, Stockholm University, and Development Re- search Group, The World Bank. I am grateful for comments and suggestions by Ray Fisman, Roberta Gatti, Dani Kaufmann, Aart Kraay, Stefan Palmqvist, Torsten Persson, Ritva Reinikka, Susan Rose-Ackerman, Andrei Shleifer, David Str6mberg, and several seminar participants. Correspondence to: Jakob Svensson, Institute for International Economic Studies, Stockholm University, 106 91 Stockholm, Sweden. Email: jakob.svensson0iies.su.se.   1. Introduction Until recently it was considered impossible to systematically measure corruption at the firm level. As a result, questions such as "who must pay bribes" and "how much" were at best supported by anecdotal evidence. However, with appropriate survey methods and interview techniques firm managers are willing to discuss cor- ruption with remarkable candor. Thus, quantitative micro data on corruption can be collected. This paper exploits an unique data set containing such information for a cross-section of firms in Uganda. We find that the incidence of bribery is positively correlated with bureaucrats' control rights over firms' business opera- tions. Firms typically have to pay bribes when dealing with public officials whose actions directly affect the firms' business operations, and such dealings cannot be easily avoided when, for example, exporting, importing, or requiring public infras- tructure services. The amount paid, in turn, is a function of firm characteristics: a firm's current and expected future profits and to what extent the firm's capital stock is sunk. Thus, demanded bribes do not appear to be fixed sums for given public services, but depend on firms' abilities to pay. Controlling for other po- tential explanations of the relationship between "ability to pay" and equilibrium graft, we show that the more a firm could pay the more it has to pay. To guide the empirical analysis on the incidence and level of graft we develop a simple principal agent model. The model rests on three assumptions.' First, bureaucrats are assumed to be expected profit maximizers, subject to the con- straints that the firm might exit and that the bureaucrat might get caught and punished. Second, by exiting (which broadly should be interpreted as moving to another sector/region or using alternative production techniques so as to mini- mize contacts with the bureaucracy) firms can avoid paying bribes. Third, there are many bureaucrats, each being uncertain of remaining in a position to extract rents. Based on these three assumptions we develop a simple rent-extraction model. The model produces a set of structural equations on the relationship be- tween firm characteristics and the incidence and level of corruption that we take to the data. 'These assumptions are grounded in case-study evidence of corruption in Sub-Saharan Africa and elsewhere. Thomas (1999) argues that the lack of control over personnel decisions, the lack of performance-based evaluations and hiring, and the power to fire government post-holders instantly with minimal explanation, have given bureaucrats and office holders with hiring and firing power opportunity to demand a share of the income stream from those lower in the hierar- chy. Increased uncertainty of tenure has also created strong incentives for those in government posts to extract as much and as quickly as possible to protect against impending unemployment or transfer to a less lucrative position (see also Bayart, 1993). De Soto (1989), Johnson et al. (1998), Johnson et al. (2000), and Friedman et al. (2000) show that corruption (opportunity of rent extraction) drives firms to the unofficial economy. 2  Modern research on the economics of corruption began with Rose-Ackerman (1975, 1978). However, despite its practical importance for many developing countries, economic studies on corruption at the firm level are rather limited. Shleifer and Vishny (1993) analyze a bureaucracy selling a government-produced good (e.g., a permit), noting that if the officials do not coordinate to extract bribes they fail to internalize the effect of their demands for bribes on other officials' income, leading to very high corruption levels. Moreover, they argue that the need for secrecy makes corruption much more distortionary than taxation. Bliss and Di Tella (1997) study the relationship between corruption and competition. They show that if bureaucrats have the power to extract money from firms under their control, they will drive the most inefficient firms out of business, enhancing the profitability of remaining firms, which in turn makes it possible to demand larger bribes. Choi and Thum (1999) use a similar model to study the effects of repeated extortion.2 Our model builds on this body of work, although it differs in one key aspect: firms' ability to pay bribes or avoid them differs in observable ways, so public officials make different bribe demands across firms. The empirical literature on corruption has recently experienced a boom. Mauro (1995) analyzes the relationship between investment and corruption, showing that corruption has a substantially negative effect on private investment and growth. Wei (1997a,b) finds that corruption discourages foreign investment, and more so than direct taxation. Tanzi and Davoodi (1997) study the effect of corruption on public finance. They find that corruption adversely affects both the compo- sition and level of public expenditures/investment. Kaufmann and Wei (1998) examine the relationship between corruption and management time wasted with bureaucrats. Contrary to the "efficient grease" argument, they find that firms that face more "bribe demand" are also likely to spend more management time with bureaucrats. Ades and Di Tella (1997, 1999), Svensson (2000), and Treisman (2000) deal with the causes of corruption. Ades and Di Tella (1997, 1999) find that corruption is higher in countries with more active industrial policy, and in countries where firms enjoy higher natural or policy induced rents, while Svens- son (1998) shows that foreign aid is associated with higher corruption in polarized societies, but that democracies seem to be less subject to this adverse effect of aid. Treisman (2000) test several hypotheses on the determinants of corruption. He shows that countries with protestant traditions, histories of British rule, and long exposure to democracy are less corrupt. This paper extends the empirical literature in two ways. First, we use firm-level data on corruption. The existing body of empirical work is based on cross-country analyses that tell us little about the relationship between corruption and individ- ual firms. Second, our data set contains quantitative information on estimated 2For recent surveys of the literature on corruption, see Bardhan (1997) and Wei (1999). 3  bribe payments of Ugandan firms. The existing empirical literature exploits data on corruption derived from perception indices, typically constructed from foreign experts' assessments of overall corruption in a country. While making it possible to study broader macro determinants and consequences of corruption, it also raises concern about perception biases that we avoid by using quantitative information.3 This paper is organized as follows. In section 2 a simple model is presented. Section 3 discusses the implication of relaxing some of the simplifying assumptions in the model. Section 4 takes the model's prediction to the data and discusses the empirical findings. Section 5 and appendix A.2. deal with implications and extensions, and section 6 concludes. 2. A Model for Estimating the Incidence and Level of Graft Below we set out a simple model to guide the empirical work. The objective is to show that firm-specific features should have implications for both the incidence and level of graft. To derive explicit solutions to the graft level equation we make a number of simplifying assumptions. In section 3 and appendix A.2. we discuss the effects of relaxing them. Consider an economy consisting of a large number of firms and bureaucrats. Each firm is in the territory of one bureaucrat. We assume that the bureaucrats are expected profit maximizers. Thus, in each period he maximizes bribe payments subject to the constraints that the firm might exit (in which case no bribes are collected), and that he might get caught and punished. The bureaucrats have discretionary power within the given regulatory system in the sense that they can customize the nature and amount of harassment on firms to extract bribes. The extent to which this could be done depends on the bureaucrats' "control rights" over the firms' business operations. We consider only private firms so by control rights we mean the extent to which the bureaucrats can constrain the firms' business decisions and influence their cash flows. These indirect control rights steni from the existing regulatory system and the fact that bureaucrats have discretion in implementing, executing, and enforcing rules and benefits affecting firms, such as business regulations, licensing requirements, per- missions, taxes, exemptions, and public-goods provision. 3Kaufmann and Wei (1998) also use firm-level data (based on the Global Competitiveness Report index) to assess the validation of the "grease argument". However, the data on corruption and its correlates are based on answers to questions referring to the country in which the firm is active, rather than firm-specific experiences and characteristics. Moreover the data is primarily of a qualitative nature (indices). Ades and Di Tella (1999) utilize the same source but use country averages. 4  As in Shleifer and Vishny (1994) we could think of the degree of control rights as determining the threat point or the leverage in the "negotiation" between a public official and a firm. When bureaucrats have low control rights a firm may refuse to pay the demanded bribes without any major consequences on its business operations. However, when bureaucrats have high control rights, the firm must either pay the bribe or exit. Bureaucrats' degree of control rights differ across sector and location. As an example, bureaucrats' control over exporting firms is typically high since such firms need additional licenses (to export) and have to deal with more government agencies (such as customs), while bureaucrats' control rights over firms in the informal sector are low since such firms seldom deal with bureaucrats and (by assumption) operate outside the formal and regulated economy. To simplify, we assume there are two sectors, j = {s1, s2}, which differ with respect to bureau- cratic control. Specifically, firms in sector s, must pay if bribes are demanded, or exit, while firms in sector 82 have enough leverage to avoid paying bribes without any significant impact on their business operations. A bureaucrat dealing with a firm in sector s, will demand a bribe if the ex- pected gain of receiving the bribe is larger than the expected cost. That is, g - 6mg > 0 where g is the graft and 6 is the probability of getting caught. We assume that the punishment of getting caught (or personal cost of being fired under corruption ac- cusations) is proportional to the bribe payment, with m > 0 being the punishment coefficient. Thus, 6mg is expected punishment (or cost) of demanding bribes. As in Ades and Di Tella (1999), Erard and Feinstein (1994), and others, we allow for the existence of both honest and dishonest public officials. Thus we assume that the personal cost m differs across individuals. The distribution of m is assumed to be uniform over [0, mh] and is known to all players. Further, to capture the inherent uncertainty of tenure we assume that the bureaucrats at each time period face an exogenously given probability 1 - q of being fired.4 At time 0, the bureaucrats must choose what sector to work in. The wage rate is normalized to zero. Each sector (i.e., public agencies interacting with firms in that sector) employs 50 percent of the total number of public servants. A bureaucrat who is indifferent between working in sector s, or S2 will be randomly selected into a sector with openings. The equilibrium allocation of public servants is easy to characterize. All public servants with personal cost m < 6-1, i.e., bureaucrats that will always ask for a bribe, will choose to work with firms in sector sl, while all civil servants with 4A fired public servant is assumed to be replaced by a new bureaucrat with the same char- acteristics. 5  personal cost m > 6-1 will be randomly allocated to the remaining openings. The probability p that a randomly picked bureaucrat in sector s, will ask for bribes is hence p = 1 - (6fi)-' if 6' < 2 nd p = 1 otherwise. Bureaucrats more prone to demand bribes will choose to work in agencies that have discretionary power over firms.5 The probability that a randomly drawn firm i must pay bribes, denoted by p(i), can now be stated as p(i) = a(i E s,) * p (2.1) where o(i e s,) is the probability that firm i is active in sector s1. Given that a firm is matched with a corrupt bureaucrat in period t, we can solve explicitly for the probability of that firm being forced to pay bribes in each future period (as evaluated from period t), Pt+n. This probability function is given by Pt+n = q- qfp + p. There exists a large number of firms. Firms' objective is to maximize present discounted value of expected cash flows (i.e., profits net of bribes). Each firm i is endowed with capital k and an individual-specific skill factor qi (knowledge) of production in sector si. q1 is distributed according to a known distribution function G(.). Invested capital is partly sunk. Let a' be the share of invested capital that could be resold and reinvested; that is, a captures the sunk cost component of the firm's production technology. At time 0 each firm faces the choice of either investing in sector s1 or in sector 82. Due to indivisibilities of capital, the firm must decide to invest in only one sector. The firms produce goods xl and x2, which are traded on the world market. The world market prices 0 and 1, respectively, are exogenously given as the country is a price taker. The production technologies are given by xl = f(k0, l1; f) and xi = f(ki, li), where f,7 > 0 and 1 is labor. There is unlimited labor supply at the wage rate w (markup on the rural subsistence wage). We assume that the price of good 1 is uncertain; that is, Ot is a stochastic variable. Ot is assumed to be independently and identically distributed over time, with bounded support [W 6 ]. Time t profit in sector s1 can then be written as a function of the observable inputs k and 1, 7r(k, 1(w/Ot); i, Ot I s,) = Otf (k, l(w/0t); 2) - wl(w/9t), where firm-specific superscripts have been dropped for convenience and where the labor demand function, l(w/Ot), is implicitly defined by the first-order condition, 5This endogenous response to differences in control rights is consistent with recent empirical evidence on corruption in the public sector. For example, results from surveys of public officials in Albania, Georgia, and Latvia suggest that there may even be a market for "high rent" positions (World Bank, 1998a, see also Thomas, 1999). 6  Otfi(k, 1; 7) - w = 0. Thus, at each time period the firms adjust their labor force such that the marginal product of labor is equal to the real wage. Similarly, period t profits in sector 82 are 7rt(k, l(w) I s2) = f (k, I(w)) - wl(w). With no bribes, firm i:s value functions are, Vt(kij)=EtE n-1ir(k,Ij) forj=J{s,S2}, (2.2) n=1 where Et is the expectation operator conditional on information at time t. If a firm invests in sector sj and faces a corrupt bureaucrat, the firm must either pay the required bribe or exit the sector. The latter constitutes an optimal response if the expected loss of exiting (foregone net profits today and in the future) is lower than the expected gain (alternative return on reversible capital). That is, ir(k, Ot,. Is) - g(Ot) +t t F1/3n' [ir(k, Ot+n, 1 Si) -Pt+n(Si)9(Ot+n)] (2.3) nN=l -17r (a!k, -I S2) X where g(Ot) is graft in period t as a function of Ot. In (2.3), the first two terms are current net profit when facing a corrupt bureaucrat. The third expression is expected discounted future net profits. In each period t + n the firm makes expected profit Et7r(k, Ot+n, | SO) (evaluated at t), and with a probability Pt+n(Si) faces a corrupt official and must also pay bribes. The term on the right side of the exit constraint (2.3) is the discounted flow of profits the firm would make if it sold and reinvested its partly sunk capital in sector S2 the first period. Firms cannot borrow to pay bribes, so in each period the firms' realized cash flow must be non-negative; that is,6 7r(k, Ot, | sl) - g(Ot) > 0 (2.4) for all t. We can now determine the equilibrium graft. Assume (2.4) holds (a sufficient condition is stated below). The corrupt bureaucrat will demand bribe payments so as (2.3) just binds. Note that Et7r(k, 0t+n, -I Sl) is constant and independent of Ot for all n > 1. Thus the expected optimal bribe payment Etg* (Ot+n) is constant and independent of Ot for all n > 1. Consequently, at each time period t + n, the corrupt bureaucrat faces an "exit constraint" (2.3) that is identical apart from the first term, current profit 7r(k, Ot, | si). Rewriting (2.3) yields, ~~~~~~~~~~~~~00 g(Ot) = 7r(Ot, .1 sl)+Et o n3 [ir(Ot+n, I s1) -Pt+n(S1)9(0t+n)0_ ,3n-r(ak, - S2). n=1 n=1 (2.5) 6The results are not qualitatively affected if we allow the firms to borrow. 7  Equation (2.5) gives a mapping from the space of possible g(O) into itself: a given g(O) implies an expected future flow of net profits, which in turn implies a new g(9) from (2.5). The fixed point of this mapping is, g*i(o0) = r(k, Ot, .1 sl) + r(k, lis )-i)- 71rak(o S2) (26) where *(k,* sl) _ Et7r(k, 9t+,,-I sl) for all n > 1, and p' _ ,q(iO)+P ) and I,V = 00l-00(-P) 7 PI -(1-)(1-qj)- Equation (2.6) suggests that the amount of bribes a firm needs to pay depends on current profits (+), expected future profits (+), and the alternative return to capital (-), 7r(ak). Having a technology with low sunk cost component strength- ens the firm's "bargaining" position in that exiting becomes more profitable. As a result the public official will demand a lower bribe. Higher profits today or higher expected future profits have the reverse effect, the firm's bargaining posi- tion weakens and it is forced to pay higher bribes. Furthermore, equation (2.6) implies that g(Ot) is a negative function of p (and indirectly of p). That is, the lower the probability that bureaucrats will demand bribes, the higher the equilibrium graft when matched with a corrupt bureaucrat. Expected graft, p * g(Ot), however, is a positive function of p. The reason for this result is simple. Everything else being equal, a lower p (and p) implies increased expected future net profits. Higher future profits weaken the firm's bargaining position since exiting becomes relatively more costly. As a result, the corrupt bureaucrat can demand higher graft. In equilibrium, the increase in g(Ot) cannot outweigh the fall in p, since that would imply that a lower p would result in lower expected future net profits, and thus lower g(Ot) - a contradiction. From (2.6) it is straightforward to determine under what conditions the bor- rowing constraint (2.4) holds. Specifically, equation (2.4) holds if 1 -q) (0Ok,1a, 82) < (2.7) p(l -q)-T(k, li, *tsi)- 7To solve the fixed point problem note that for the exit constraint to bind in every period the difference d _ ir(k, 0, .1 si) - g(0) must be constant over time. Substituting d into (2.5), noting that -C Et E fl Lpt+,d + (1 - pt+.)7rt+n] = n=l -pq(lp) + 1 Ppd + (1-p)p(3 d1 q Et?rt+n and rearranging yields expression (2.6). 8  Thus, if p is sufficiently high, equilibrium graft is always less than gross profit. Equations (2.1) and (2.6) provide a structural model of the relationship be- tween graft and firm characteristics. The incidence of bribery is a function of where the firms choose to locate, the firms' main areas of activities, as well as the expected personal cost to the bureaucrat of being fired under corruption ac- cusations. Given that a firm faces a corrupt official and must pay bribes, the amount paid depends on firm characteristics: current profits, the extent to which the firm's capital is sunk, and expected future profits. Before proceeding to estimate equations (2.1) and (2.6), it is useful to consider relaxing some of the simplifying assumptions in the model. This is important not only to show to that the model's qualitative results are robust to alterations, but also to better understand the empirical findings presented below. 3. Extensions and implications In reality, a bureaucrat does not have full information about a firm from whom he wishes to extract bribes. The shock 0 and profits are not directly observed, neither is the sunk cost component. As illustrated in an example in appendix A.2., incomplete information will create informational rents that the firm can capture. Thus, the linear relationship between profits and grafts identified in equation (2.6) will only be an approximation. The qualitative results, however, remain when introducing asymmetric information. The likelihood of facing a corrupt bureaucrat in the future is uncertain in the model. This likelihood is a function of a number of exogenous parameters; probability of getting caught, distribution of the personal cost of getting fired under corruption accusations, and the inherent uncertainty of tenure, but also depends on the endogenous choice of bureaucrats to differences in control rights across sectors. With no uncertainty, a one-period model would suffice to study the problem, since future profits, tr(k, P, -I sl), would then not matter. In the model, each firm is in the territory of one bureaucrat. As in Bliss and Di Tella (1997) and Choi and Thum (1999), we thus abstract from coordination issues and competition among public officials. Allowing competition among bureaucrats would in some instances increase the firm's bargaining power and thus reduce the equilibrium graft, given 7r(k,Ot, -Is), tr(k,li, -Is), and 7r(ak, .1s2). Still the qualitative relationship between profits, alternative return, and corruption would remain.8 We have taken the technology choice as given, i.e., the sunk-cost component 8However, to the extent that officials impose costs rather than benefits, it is not clear why competition would reduce corruption (see discussion in Rose-Ackerman, 1999). 9  (asi) is exogenous. Allowing the firm to choose what capital goods to purchase complicates the picture (the technology choice in a model of repeated rent extor- tion is studied in detail in Choi and Thum, 1999). In our model, the extent to which capital investments are sunk or not influences the firm's bargaining posi- tion versus the bureaucrat. Low sunk costs imply that the cost of exiting becomes smaller, and from equation (2.6), lower grafts when matched with a corrupt offi- cial. Thus, the firm might find it profitable to choose a "technology" that yields higher per-period operation costs but indirectly reduces the amount of bribes the firm needs to pay. In appendix A.2. we endogenize the choice of ai and show that the choice of technology depends on the parameters of the model, and in particular on p. If the incidence of bribery is high, the relative return of adopting a technology with inefficiently low sunk cost component is also high. For the em- pirical work it should be noted that the "technology-effect" would tend to mask the negative relationship between the sunk-cost component and corruption, and thus work against us. Maybe the most restrictive assumption of the model is the assumption that profits are not influenced by the amount of bribes paid, and that there is no feedback from corruption to equilibrium profits (through entry and exit into the market).9 Our choice to abstract from these effects does not imply that we do not think they might be important. However, we believe our more restrictive set-up is a good first approximation for two reasons. First, most firms in the sample are small (median is 34 employee). Causal empiricism suggests that the regula- tory process is not captured by these types of firms in less developed countries in general and in Uganda in particular, but a small set of large, politically powerful enterprises. Second, the inherent uncertainty of tenure for those in government posts, documented by for example Thomas (1999), suggests that public officials heavily discount the future. Thus, dynarnic graft-schemes that intend to maxi- mize revenue by implicitly controlling entry and exit may simply not be credible, since the uncertainty of tenure creates strong incentives for those in government posts to extract as much and as quickly as possible to protect against impending unemployment or transfer to a less lucrative position. Finally, the feedback from corruption to profits has already been extensively studied in the literature (see Bliss and Di Tella, 1997). Therefore we abstract from it in order to focus on the novel issue of determining the differences in bribe demands across firms. Despite these arguments it is crucial to evaluate how the results would change 9For models on rent-seeking see Buchanan, 1980; Tollison, 1982; Tullock, 1967, Bhagwati, 1982; Krueger, 1974. On regulatory capture see Laffont and Tirole, 1994, and references given therein. Bliss and Di Tella, 1997; and Choi and Thum, 1999, develop extortion models in which the public official chooses graft, where corruption may cause exit (or restrict entry) which may affect profits of the remaining firms and thus their potential to pay bribes. 10  if these effects were allowed in the model. The rent-seeking and regulatory capture models would also predict a positive relationship between profits and corruption. In these types of models the association arises because bureaucrats and politicians compete for rents associated with bribes and kickbacks by selling government favors. Alternatively, regulations benefiting firms are "acquired" by industries through bribes. Thus, the relationship is driven by reverse causation. Note that the predicted association between the share of sunk investments and corruption would not arise from these models. Moreover, as discussed below, for the reverse causation argument to bias the results it must be the case that the size of the government favor is linked to the amount paid in bribes. A simple extension of the model (yielding the same empirical predictions) is that all bribe-paying firms receive preferential government treatment, for instance they obtain a valuable license. Our identifying assumption is that the price of this license is determined by the firm's ability to pay. The extortion model of Bliss and Di Tella (1997), where corruption may cause exit (or restrict entry) which may affect profits of the remaining firms and thus their potential to pay bribes, would also suggest a positive association between bribes and profits. The interpretation, however, would be slightly different: prof- itable firms are forced to pay higher bribes but one reason for why they are profitable in the first hand is that they could "afford" to pay bribes while other potential competitors which could not have been driven out of the market. As discussed in detail in section 4.3, empirically we try to separate the afore- mentioned effects by instrumenting for profits. 4. Estimating the Incidence and Level of Graft 4.1. Specification Equations (2.1) and (2.6) provide a structural framework to study the incidence and level of graft across firms. The incidence equation (2.1) states that the prob- ability that a randomly drawn firm i must pay bribes depends on sector/location specific factors and on the personal cost to the bureaucrat of getting caught. This personal cost is not observable, and we choose to capture it with the random variable v. Thus p = xwi +