Corruption in Indonesia1 J. Vernon Henderson Ari Kuncoro Brown University Brown University and University of Indonesia July 30, 2004 Bribes by firms in Indonesia arise principally from regulations --licenses and levies --imposed by local government officials. Regulations generate direct revenues (fees) plus indirect revenues in the form of bribes. The expected value of the latter is capitalized into lower salaries needed by localities to compensate public officials. Localities in Indonesia are hampered by insufficient revenues from formal tax and transfer sources to pay competitive salaries plus fund “demanded” levels of public services, because local tax rates are capped by the center and inter-governmental transfers are limited. Thus the direct and indirect revenues from local regulations are critical to local finances. The paper models and estimates the key aspects of corruption -- the relationship between bribes, time spent with local officials, and different forms of regulation. It models how inter-jurisdictional competition for firms limits the extent of local regulation and how greater sources of tax or inter-governmental revenues reduce the need for regulation and corruption. The paper estimates a large reduction in regulation in better funded localities. The findings are directly relevant to Indonesia where corruption is high and the country is in the throes of major decentralization and local democratization efforts. Key words: corruption, inter-jurisdiction competition, fiscal decentralization. JEL codes: D2, D7, H2, H7, O1 Corruption in Indonesia is widespread and costly. Based on the detailed survey this paper utilizes, firms report spending on average over 10% of costs on bribes and over 10% of management time in “smoothing business operations” with local officials. But the extent of corruption varies enormously across local jurisdictions, with, for example, the average of bribes to costs ranging from .56% to 31% across localities in the survey. This paper focuses on two issues. First we examine the nature and costs of corruption, asking how money and time costs vary as the local regulatory environment a firm faces varies. Second, we argue and present evidence that local regulatory environments depend on local fiscal arrangements and local fiscal situations, as determined in part by central government policy. Localities with more restricted formal fiscal situations use regulations and the resulting corruption and bribes as a form of indirect taxation to help fund adequate compensation for local officials. This relationship is of particular interest given two key political issues in Indonesia today—ameliorating corruption and achieving full devolution of many governmental functions to localities. 1 This paper was written while Kuncoro was a Visiting Professor of Population Studies at Brown University, for which funding from the Mellon Foundation is gratefully acknowledged. We thank Gilles Duranton for helpful comments on an early draft of the paper. In analyzing, first, the nature and costs of corruption, we ask what types of public regulations invite bribes (Kaufmand and Wei, 1998) and how important is each type. For regulations such as required licenses and levies, how does the bribing process work? What are the magnitude of bribes needed either to avoid requirements or to supplement official fees, and what determines the amounts of time spent wooing local officials? How heterogeneous are responses across firms and how predictable is this heterogeneity of response? Answering these questions will help us understand the extent to which creation of regulations enhances the ability of officials to extract bribes. We also examine forms of corruption, involving defraud of the state of, say, tax revenue that individual firms might favor (Shleifer and Vishny, 1993). To examine corruption, we will utilize a data set collected in 2001-2002 by LPEM at the University of Indonesia covering 1808 firms in 64 (out of about 300) local government areas, which is unusual in two aspects. First is the detailed micro information on forms of regulations and interactions with local officials. Second is the high response rate – in terms of willingness to report bribes, willingness to report other corruption information, and candor about the magnitude of bribes paid. For example, 75% of all firms sampled report positive bribes, and we infer that about another 5% correctly report zero bribes. Given missing values on other responses, in all in our data, we work with a sample of valid bribe responses of over 70% of the original surveyed firms. As we will see, non-respondents on bribe payments are not a random group, with typically different behaviors in other corruption dimensions than respondents. In contrast, in Uganda which is a country viewed at least as equally corrupt (Bardhan, 1997), Svensson (2003) ends up analyzing bribes reported by just 48% of original surveyed firms from a general economic survey of firms. In Svensson’s survey mean bribes are only about 3% of profits (and presumably a much smaller fraction of costs), in comparison to mean bribes to costs of 10.5% in our survey. The magnitudes reported for Uganda are similar to what Indonesian firms report as corruption costs (“gifts given”) in the formal Industrial Economic Census. But the carefully crafted interviewing for our survey specifically focused on corruption, with various indirect checks on accuracy, brings out very different responses and response rates. In examining the second issue of how fiscal arrangements affect regulatory and corruption environments, the corruption most firms face involves interaction with local officials, who administer regulations and taxation. Thus we focus on government at the kabupaten level, which is the main local government institution, and is similar in geographic scope to north-east US counties. This focus is of particular interest, since several months before our survey started, Indonesia implemented full decentralization of expenditure functions, with some degree of local democracy, moving away from a 2 formerly more unitary, authoritarian system. 2 The regime shift is expected to help curtail corruption, by increasing the role of the local government and accentuating the forces of inter-jurisdictional and political competition. Corruption in Indonesia is a major on-going political issue with wide press coverage and public discussion and has been a focus of both World Bank (2003) and local academic (Kuncoro, 2003) study. It is recognized that democracy may have a limited impact on corruption (Rasmusen and Ramseyer, 1994) and that regime switches are difficult given the role of history, culture, and expectations (Tirole 1996, Sah 1988, Andvig and Moene 1990, and Bardhan 1997). However, as we will explain, the concern in this paper is that decentralization may worsen corruption overall and seems certain to worsen it in some kabupaten because, while expenditure functions are decentralized in the reforms, revenue functions are not. To understand the potential link between corruption and fiscal arrangements, for the moment think of the government of a kabupaten, elected or not, as being embodied in the regent (the Indonesian term for the head of the local government), who hires local officials to administer regulations, as well as provide services. While the regent has a local property tax base, de facto tax rates are capped at low levels; and fiscal transfers are modest. It is widely acknowledged that revenues from tax and transfer sources both before and after decentralization are insufficient to pay for even minimal mandated public service levels, so the regent needs to seek other forms of revenue. Local regulations such as licenses and “levies” provide indirect revenues in the form of bribes, as well as direct revenues. Bribes received by local officials to ameliorate the impact of regulations mean the regent can pay lower salaries to officials, freeing up money for other purposes– i.e., expected bribes received are capitalized into lower official salaries. The use of regulations and corruption to provide local revenues is not without consequence. Non- competitive levels of regulation and excessive bribe demands make a locality unattractive to firms, driving firms to other kabupaten. Loss of firms impinges on a kabupaten’s tax base, lowering tax and bribe revenues. It also may affect local economic growth, local wages, property values, and wealth of residents. In a context where the last 15 years in Indonesia has seen rapid and massive movement of manufacturing firms out of central cities into ex-urban areas (Henderson, Nasution and Kuncoro, 1996), rural kabupaten that are less corrupt may compete and develop more effectively. The need for kabupaten to rely on regulation and corruption depends on inter-governmental fiscal arrangements. If the regent has access to higher effective tax rates on the kabupaten tax base or enjoys greater 2 Local legislators are now elected. However the local government head, or regent, is still de facto appointed by the center. After 2001, she became subject to censor and impeachment by the local elected parliament; and in November 2003, new legislation was enacted to have her be directly elected as well in the future. Note regents themselves as well as central government officials are corrupt, perhaps inhibiting their ability to fight lower level corruption (Andvig and Moene, 1990). 3 inter-governmental transfers, her need to rely on corruption is reduced. The effect of the recent decentralization on local government finances in Indonesia is still unclear; kabupaten have much greater expenditure responsibilities but also greater inter-governmental transfers. However it seems clear to observers that both before and after decentralization kabupaten governments are under-funded and that there are definite fiscal losers from decentralization. As we will see, in Indonesia because of oddities in the national tax-transfer system, in our data on local fiscal budgets, there is wide variation in the extent of fiscal transfers from the center as a fraction of local GDP, as well as wide variation in regulated tax revenues. We will link quantitatively the extent of regulation and local fiscal situations. The paper starts with a conceptual framework, modeling aspects of corruption, inter-jurisdictional competition, and the effect of fiscal arrangements on corruption, reviewing the relevant literature as we go along. Next, it turns to an empirical assessment of regulation, corruption, and the interaction between firms and local officials. Finally, we review center-local fiscal arrangements in Indonesia and examine the effects of fiscal arrangements on corruption. 1. Modeling Corruption and Its Effects In this section we outline a simple model of “demand and supply” of bribes in a locality. Firms supply bribes in interaction with local governmental employees, under the “efficient grease” hypothesis to reduce the impact of regulations (e.g., Liu, 1985, and Becker and Maher, 1986, as reviewed by Bardhan, 1997). Given the response of firms and local government employees, local governments “demand” bribes through the imposition of regulations (e.g., Banerjee, 1994, Kaufman and Wei, 1998). The new ingredients we bring to this traditional modeling of corruption are twofold. The choice of the extent of regulation and related corruption by local governments is constrained by strategic competition across localities for firms. Second the choice to use regulation and corruption as a form of indirect taxation and compensation of local government employees is influenced by local fiscal situations and the access of local governments to other forms of public revenues. The model we outline will form the basis for the empirical study of corruption. 2.1 Modeling Firm Response to Regulation How do firms respond to regulations (e.g., licenses, “levies”), or what is often called “red tape” or just “harassment” (Kaufman and Wei, 1999), in interacting with local officials? Bribes are supplied to ameliorate the impact of regulations, such as waiting times for specific licenses needed to carry out different aspects of business. In thinking about the bribing process, evidence in Kaufman and Wei (1999), Svennson (2003) and our data suggests that bribes and firm time engaged with public officials are positively correlated. To reduce 4 the costs of regulation takes both time and money. In our data we will have six categories of the fraction of time spent by management with local officials smoothing operations. The mid-point values of time are 2.5%, 10%, 20%, 37.5%, 62.5% and 87.5%. The mean percent of bribes in production costs changes across firms in each category, taking average values respectively of 7.8%, 9.3%, 12.5% 16.8%, 14.4% and 19.4%, so the average rises by 2.5 fold moving from the lowest to highest category. Why this positive, rather than negative correlation? In part time and bribes are complements. Firms need to bribe officials to facilitate or at least not deliberately hold-up, say, their application for a specific license; but they also need to spend time to learn about the unofficial application requirements, to prod the application along even while bribes are paid, and to monitor the process. In addition based on common perceptions of the social forces involved, officials enjoy spending time with firms, whether it is because of power issues (“I can force firm i to devote time to me”) or, particularly in Indonesia, because the official does not want to be seen as a simple thief but rather wants to cultivate a “gift relationship” among “friends”. Also as a pure economic explanation, time is spent trying to assess what level of bribe is needed to get cooperation of specific officials.3 In modeling, to simplify, we utilize a “black-box” formulation, noting that all these elements underlie the formulation. To model bribes, we adapt Kaufman and Wei, so that later in analyzing the demand for regulations and bribes, we can capture the notion of inter-jurisdictional competition for firms. In doing so, we separate the specification of firm production technology from that of corruption technology. We start with corruption technology and the supply of bribes. In the model, “harassment” is for example the number of licenses, h, which the local government imposes. For a firm, there is a “black-box” bribing “technology” to reduce the effects of harassment, f(b,t), where effective harassment h%=h− f(b,t), and f , f >0, f , f <0. Bribes are b t bb tt 3Economic micro-foundations for the f(b,t) function below about time and bribes could involve a learning story. For any license and license grantor, there is a minimum payment, say θ, an official will accept, depending on the official’s “tastes”. That minimum is private information and the firm only knows the distribution. The firm gets, say, two tries (visits by the official to the factory), to bribe him. If he fails on both accounts, he gets no license that year and that imposes specific costs. Suppose θ, is uniformly distributed between 0 and 1. The firm can offer θ on the first visit 1 which has a probability θ of being accepted. If θ is too low, no license is granted on the first visit and the official 1 1 incurs a cost c, with probability (1−θ). On the second visit the factory can offer θ which has a probability 1 1 2 (θ −θ) of being accepted. If it is rejected, the firm then bears a cost c with probability (1−θ), where c >c. 2 1 2 2 2 1 Optimizing with respect to θ and θ certain regions of parameter space yield an interior solution where 1 2 θ = 2/3 c + c /3 and θ = 2/3 c + c /3. Given θ > θ in such a solution, time spent (number of visits) with 1 1 2 2 2 1 2 1 the local official rises with observed bribes; and the solution also has the feature that with greedy officials bribes may never be paid although heavy costs (c, c and 2 visits) are incurred. 1 2 5 b and time by firm management is t. So time and bribes may be spent to reduce waiting time to get a license, which is required to proceed with certain firm functions. For the firm, effective harassment imposes costs, c(h%) where c', c'' >0. The firm seeks to minimize harassment costs plus bribes paid plus the cost of time, or c(h− f(b,t))+b + w(t) (1) where w' >0, w''≥0. The firm chooses b and t such that c'(h− f(⋅)) f (⋅)=1, (2a) b c'(h− f(⋅)) f (⋅)= w' (2b) t If we differentiate (2a) (with a symmetric condition for (2b)), if w'' =0, which we will generally assume, we get c'f −c'' f f c''f db= bt t b dt+ b dh (3) c''f 2 −c'f c''f 2 −c'f b bb b bb In (3), while the coefficient of dh is unambiguously positive (c'' >0, f <0), the coefficient of bb dt can only be positive if f >0 and f is “large”, or if b and t are “strong complements”. bt bt For well behaved functions, eqs. (2) can be solved so b= b(h) (4a) t = t(h) (4b) Utilizing eq. (3) and the corresponding term from differentiating (2b), we can show that dt = c''c' (f f − f f )/D≥0 (5a) dh b bt t bb db = c''c' (f f − f f )/D≥0 (5b) dh t bt b tt where D= (c''f 2 − c'f ) (c''f 2 − c'f )− (c''f f − c'f )2 >0 from second order conditions. t tt b bb b t bt db/dh and dt/dh are assumed to be positive so that b and t are “normal” goods in response to increases harassment; a sufficient condition for this is that f >0. For later reference, to ensure well-behaved interior bt outcomes it is convenient but not essential to assume d2b/dh2, dt2/dh2 ≤0, so that increases in harassment increase time and bribes at a decreasing rate. In the empirical section we estimate a form of eq. (4), yielding the coefficients in (5). We also examine the complementarity effects in (3) directly. 6 If harassment increases both time and bribes enjoyed by local government employees, what limits harassment, apart from local social norms? One limit is firm exit, where in Bliss and Tella (1997), given heterogeneous firms, corruption forces less efficient firms out of business. Another limit which we take here is to emphasize inter-jurisdictional competition, and exit to other jurisdictions. 2.2 Local Government Demand for Regulation and Corruption There are two stages to modeling local government demand for regulation. First is to analyze firm location decisions and inter-jurisdictional competition for firms, so as to see how harassment affects the number of firms in a locality. The effect of harassment on the number of firms in a locality and hence on the local tax base is something local governments anticipate in policy decisions. The second stage formulates the nature of the local governments, their optimization problem and their fiscal situation. In the work to follow we assume there are two regions, i and j, with the model directly generalizing to n regions. At some points we will impose symmetry across regions, to simply and illustrate key points. 2.2.1 Harassment and Inter-Jurisdictional Competition for Firms For our firms, there is a fixed amount of labor, L and L , available in each region, split among the (identical) i j firms in the region. Each firm is run by an entrepreneur where nationally there are a fixed number, N , of entrepreneurs and firms, which are perfectly mobile so N +N =N. We first examine firm production i j technology, and then combine that with the corruption technology to look at location decisions. Firm Production Technology. If all entrepreneurs are identical, then per firm output in a region i is x(L /N ), where x'>0,x"<0. i i Assuming competitive labor markets, so wage, w=x', we define firm revenue from production activity as R(N /L)=x(L /N )−wL /N , where we can show that4 i i i i i i dw /dN =−x"(L /N )⋅(L /N2)>0 i i i i i i (6) dR /dN ≡ R' = x"(L /N )⋅(L2 /N3)<0 i i i i i i i Location Decisions. Combining production and corruption, in region i, per firm total profits are π = R(N /L)− c(h − f(b, t ))−b −w(t )−T. (7) i i i i i i i i i i Firms get net production revenue R(⋅), have harassment and bribe costs of c(h − f(b, t ))+b + w(t ) from i i i i i i (1), and pay a tax T set by the center and partially remitted to the local government. The regent anticipates i 4 dR= (−x''+w) L /N2dN −∂w /∂N (L /N ) dN. But given x'=w, we have the second equation in (6). i i i i i i i i 7 that firms move between regions so ψ=π −π =0. Given this, we can determine how a regent perceives i j harassment levels will affect the number of firms in her region, in a Nash context, where the regent is choosing harassment levels. Totally differentiating (7), we get R' dN /L −c' dh + (c'f −1) db + c' (f −w )dt = R ' dN /L −c' dh +c' (f −1) db i i i i i i b i i t t i j j j j j j b j i i i j + c' (f −w )dt , where the regent sees T as fixed. Utilizing eq. (2), imposing Nash perceptions at j t t j i j j equilibrium by the regent in i that dh ,db ,dt =0, and having the regent recognize the national constraint on j j j numbers of firms so that −dN =dN , we get j j dN c' i = − i <0 (8) dh (−R'/L −R' /L ) i hj i i j j where R' <0. In (8), as h rises in the numerator that increases costs and reduces profits to the representative i firm which induces exit to region j. However, from the denominator, exit is limited by diminishing net firm revenues in j that occur with entry and rising net revenues in i that occur with exit. Again for later reference, to have well-behaved interior solutions, it is convenient but not essential to assume d2N /dh2 ≤0, so that as i i harassment increases firms exit at an increasing rate. A sufficient condition in a symmetrical situation for this to hold is that f f − f2 ≥0. bb tt bt 2.2.2 The Local Government’s Demand for Harassment Levels For regions, we look first at the regent’s revenue sources and expenditure requirements. Based on taxes collected from firms in the region, the T in eq. (6), the center remits to the locality a portion of that tax,β, so i that the regent’s tax revenue is βTN . This βTN could be adjusted to include also general fiscal transfers i i i i dependent on local GDP or other measures of local economic activity that are influenced by the number of firms in the locality. Out of βTN , the regent must pay a salary, S, to the local government employee, or i i official, who works for her. The post salary residual is “profits” or net revenue kept by her or spent on improving public good quality. But it is thought that these capped tax revenues are insufficient to pay the salaries of the base number of local officials needed to provide even mandated public services. Corruption enters this process through its effect on the formal salary required for the local official. The local official hired by the regent must be paid utility W. The local official has a utility function, y+U(t(h)N), where y is income. The second term is utility enjoyed by the official from time spent with local firms, where we assume U'>0,U"≤0. While, in practice, some of firm time facilitating the easing of a regulation is spent prodding a “faceless bureaucracy” or learning procedures, rather than directly with the 8 official, for simplicity we have the official enjoy all smoothing time endured by at firm. Treating W is as fixed, the regent sees the official’s required salary as S =W −b(h)N −U(t(h)N ); or as the market utility i i i i i less bribes collected by the official less the utility received by the local official from time devoted by local firms. Note we have not added in extra direct revenue that the regent gets from the sale of licenses and imposition of levies; we simply normalize that revenue to zero.5 What is the objective function for the regent in region i? Given the equation for salary, S, the regent’s total net budget after paying the salary of the local official B is i B =βT N − (W −b(h)N −U(t (h)N )), (9) i i i i i i i i i In the period for our data where harassment is still based on the pre-decentralization, pre-local democracy situation, we model the regent as a Leviathan government that seeks to maximize this budget. After the introduction of limited democracy in Indonesia, following Panizza (1999) and Arzaghi and Henderson (2004), we could model the political process as having an objective function that weights the utility of voters against that of the government, with weights representing the relative bargaining power of each group. For example the objective function could be Ω =V(w,B)θB1−θ, where V(w,B)is the quasi-indirect utility function of i i i i i i the representative worker. His income equals the local wage rate w ; and, normalizing the cost of local public i goods to 1, B equals the level the level of local public services, g . θis the relative bargaining power of the i i representative voter in the political process, or the degree of democratization. But for the pre-decentralization period, without loss of generality as to concepts involved, we assume Ω is simply B in a Leviathan situation i i where θ=0. Optimizing with respect to h , we have i dB /dh =∂B /∂h +∂B /∂N ⋅∂N /∂h =0 i i i i i i i i (10) d2B /dh2 <0. i i In the first order condition, the ∂B /∂h term represents the benefits of increasing harassment— i i increased corruption and bribes for the local official which reduces his formal salary and increases the public budget (from (9), ∂B /∂h =N[db/dh+U'⋅dt/dh] which from (5) is positive). The term ∂B /∂N ⋅∂N /∂h i i i i i i represents the costs of increased harassment: the lost revenues because firms exit the kabupaten in response to 5 If p is some exogenously regulated harassment fee so total harassment revenues are pNh , that adds to the h i i i harassment determination terms reflecting lost harassment fees when a firm exits plus gained monies when h is increased. 9 increased harassment (from (9) ∂B /∂N ⋅∂N /∂h = [βT + t⋅U'+b] dN/dh, where dN/dh<0).6 The first i i i i order condition balances out these two forces. For the second order condition to hold, sufficient conditions are dN d2N db dt d2b d2t i <0, i ≤0, i , i >0, i , i ≤0, U"≤0. Note we are assuming that each kabupaten dh dh2 dh dh dh2 dh2 i i i i i i perceives taxes are capped at too low a rate, so dB /dT >0at the currentT . i i i Here we have modeled the choice of harassment and related corruption by the regent as helping ensure local officials receive market compensation, with the degree of harassment limited by inter- jurisdictional competition for firms. There are other limits on corruption, although so far in Indonesia these do not appear to be direct penalties for being caught (Mookerjee and Png, 1995 and Cadot, 1987). But in the empirical work, we do consider observed and unobserved heterogeneity of the characteristics of local officials, in particular their degree of education, which might affect either their sensibilities to media scrutiny and social pressure to limit corruption or their perceptions about the long-term effects of corruption on local economic growth and ultimately the resources available to them. With this framework we can examine the effect on corruption of changes in fiscal situations. In particular we want to show that if inter-governmental transfers or money rebated from tax collections in the locality go up, harassment, bribes and corruption are likely to decline. That is, we want to show that an increase in βT (because either or both βand T are increased) reducesh . We can do this by simple i i i differentiation for the symmetrical case where βT is increased in all kabupaten by the same amount, so i eq.(10) continues to apply everywhere with the same arguments;7 imposing symmetry simplifies the impact on reaction functions. Differentiating eq. (10) with respect to βT and h yields i i dh d2B /dh2 i =− i i <0. (11) d(βT) d(dB /dh)/d(βT) i i i i We know from the second order condition in (10) that d2B /dh2 <0. Given (9), i i d(dB /dh)/d(βT)=dN /dh <0. Based on this result for the symmetric case with a Leviathan i i i i i government, we argue that the forces are there for increased inter-governmental transfers to reduce 6 Under partial democracy the benefits and costs of increasing h are respectively [θVθ−1V B1−θ+(1−θ)VθB−θ]N[db/dh+U'⋅dt/dh] and g {θVθ−1B1−θV dw/dN+[θVθ−1B1−θV +(1−θ)VθB−θ][βT+ t⋅U'+b]} {dN/dh}. In the cost expression, the first term in w g the first curly brackets is the effect of changes in firms on wages. While the expressions are more complicated they have the same economic interpretation. 7 Note the T,T terms cancel out in the derivation leading to eq. (8). i j 10