Business Associations, Bureaucratic and Political Corruption: An Empirical Analysis of Lobby Group Membership. Eugene Kiselev.

Business Associations, Bureaucratic and Political Corruption: An Empirical Analysis of Lobby Group Membership Eugene Kiselev Brandeis University International Business School October 2, 2012 Abstract This paper examines how lobbying eorts by business associations aect the bribe payments rms make to bureaucrats in order to get around unfavorable regulations (bureaucratic corruption), and the payments they make to legislators to aect regulatory change (political corruption). Utilizing a rm-level panel dataset of Eastern European enterprises, the empirical results indicate that rms who join lobby groups do not reduce their bribes to bureaucrats, and rms that are more impacted by bureaucratic corruption are no more likely to join a lobby group than their counterparts. On the other hand joining a lobby group increases a rm's probability of bribing legislators and other policymakers. These results indicate that lobby group membership is not a substitute for bribe payments to bureaucrats, but rather introduces the possibility of state capture by giving rms access to policymakers that they wouldn't otherwise have. Eugene Kiselev is currently a PhD student at Brandeis University International Business School. e-mail: ekiselev@brandeis.edu. 1

1 Introduction Since the fall of the Soviet Union, rms in many post-soviet states have been in an ongoing battle against corruption and government interference while operating in an environment of excessive regulation and legal uncertainty. Still the business environment in these countries continues to mature, and a potentially promising development has been the emergence business associations. These groups provide a number of services to rms, including dispute resolution mechanisms, domestic and international product market information, help with licensing, and perhaps most importantly the lobbying of local and federal governments to achieve outcomes favorable to association members. Recent literature has explored whether lobbying eorts by business associations can eventually replace bribe payments to bureaucrats. Researchers argue that at the countrylevel, lobbying should eventually overtake corruption as a mechanism for achieving outcomes in the private sector as economies grow and political and legal institutions develop. A similar argument has been put forward at the rm-level, suggesting that as enterprises grow they will forego bribing rule enforcers (bureaucrats) in favor of lobbying rule makers; the logic being that the need to bribe rule enforcers to get around complex and unfavorable regulations will disappear once lobbying eorts succeed in making these regulations more rm-friendly or in removing them entirely. While this outcome appears plausible, as yet there have been no empirical studies attempting to establish a strong casual link between bureaucratic corruption and lobbying by associations. In addition, while recent studies have examined the relative impact of lobbying and bureaucratic corruption on rm outcomes such as growth in sales and political inuence, researchers have yet to empirically investigate political corruption and whether business associations play a role in state capture through private payments directed at policy-makers. This paper investigates the relationship between lobbying eorts by business associations and rms' involvement in both bureaucratic and political corruption. With regard to bureaucratic corruption, I challenge the notion that a causal relationship exists. I rst investigate whether bribe payments to bureaucrats have any bearing on business association membership and whether this relationship is aected by the size of the rm. 2

I further analyze this relationship in the context of membership dynamics to determine whether bribes to bureaucrats aect a rm's decision to join or leave an association. As lobbying is not the sole function of a business association the analysis distinguishes between the many services that associations provide in order to determine whether dierent types of rms derive dierent benets from membership. Finally I ask the question whether association membership leads directly to political corruption by investigating whether rms use these organizations to gain access to and bribe rule-makers and other high-level ocials. Utilizing the Business Environment and Enterprise Performance Survey (BEEPS) conducted by the World Bank in 2002 and 2005 I rst examine the rm-level determinants of association membership and corruption. The empirical analysis focuses on former Soviet republics as well as additional Eastern European states and Turkey. Vital to the analysis is that the BEEPS data allows me to create a panel of nearly 1500 rms that were interviewed in 2002 and then again in 2005. I nd that neither the incidence nor magnitude of rms' bribe payments to bureaucrats aect their probability of being a business association member or their probability of joining or leaving one between 2002 and 2005. Likewise, joining an assocation does not aect the incidence or magnitude of bureaucratic corruption as evidenced by rm's bribe payments. Finally I nd no indication that this relationship depends on the size of the rm. The results indicate that neither lobbying nor any other benet of association membership is a substitute for bribe payments to bureaucrats that many rms are forced to make. Turning the focus to political corruption I nd that business association members are more likely to make "unocial" payments to legislators than non-members. In other words, not only do association members continue paying bribes to bureaucrats, but they are also more inclined to inuence legislation through bribe payments to policy-makers. Members involved in political corruption are also less likely to leave an association than their more scrupulous counterparts. However a rm's baseline level of exposure to political corruption does not aect their probability of joining, suggesting that corrupt rms do not self-select into business associations but that associations provide certain services that change rms' preferences for political corruption. 3

To determine whether business association lobbying eorts in particular are associated with political corruption activities, I analyze rms' characteristics vs. the value ratings they assign to dierent association services. The ndings indicate that rms engaged in political corruption are signicantly more likely to list lobbying as the most important service, whereas other characteristics such as rm size, whether the rm is an exporter, or measure of regulatory burden are associated with other membership benets such as assistance with government regulations or domestic and international product market information. I conclude that business associations serve a multitude of functions, many of which are unrelated to political corruption, but for some rms lobbying through their association is particularly important as a means to access and bribe policy-makers. The remaining question is whether in fact joining a business association increases the likelihood of a rm bribing policy-makers. To that end I examine the change in rms' political corruption activities between 2002 and 2005, controlling for their baseline score. I nd that rms become signicantly more corrupt after joining an assocation relative to those that remained non-members. This indicates conclusively that there is indeed a corrupting inuence on the rm of lobbying eorts by business associations. Taken together the analysis concludes that business associations serve many functions and that rms utilize them for various reasons depending on the constraints they face and the characteristics they have. In particular the results indicate that some rms use business association lobbying eorts as an opportunity to engage in political corruption by gaining access to policymakers to aect favorable outcomes through private payments. The paper proceeds as follows: Section 2 provides background on business associations and lobby groups in Europe and elsewhere, and discusses relevant theoretical and empirical literature. Section 3 presents the data used in the analysis, while section 4 discusses the empirical strategy and presents the results. Section 5 concludes. 4

2 Business Associations and Lobby Groups: Theory and Evidence Many post-soviet countries share a number of economic and regulatory problems that require collective action on the part of private actors to facilitate change. These include complicated and frequently changing legal codes, regulatory obstacles that increase the costs of starting and operating businesses, few reliable dispute resolution mechanisms, and of course bureaucratic corruption. The emergence of business associations in the late 1990's and early 2000's can be attributed to the desire of businesses to improve many of these conditions. Country-specic case studies have shown that associations targeting specic obstacles can have signicant success in promoting eciency and precipitating regulatory change, including expediting business licensing and registration, providing rms with legal services to combat predatory government practices, and even providing governments with the necessary impetus to undertake large-scale endeavours such as macroeconomic reform (Sullivan, 2007; Duvanova, 2006; Doner & Schneider, 2000). Empirical analyses utilizing surveys of business associations and association members have provided additional support for the positive inuence of these organizations. Analyses of Russian associations show that membership oers benets such as dispute resolution, consulting and other services, and that relative to non-members, members are quicker to adopt new technologies, are more likely to import/export, and have higher levels of investment (Pyle, 2005; Pyle, 2006). The lobbying component of business associations cannot be overlooked, and the Olsonian assessment of lobby groups as rent-seeking organizations is a constant theme in the literature. Much of the theoretical research assumes lobbying is simply a transfer of resources from one group to another, so that the distinction between lobbying and political corruption is nebulous at best. 1 Deliberate attempts to separate out the effects of lobbying and corruption in a theoretical setting are relatively uncommon, but studies that have done so draw the distinction on the basis of information provision to 1 See Grossman & Helpman (1994), Coate & Morris (1999), Yalcin & Damania (2005). Surveys of the theoretical literature on lobbying include Grossman & Helpman (2002) and Austen-Smith (1997) 5

politicians vs. direct contributions (Bennedsen & Feldmann, 2001) or in the transfer cost dierences of the two activities (Begovic, 2005). A unique recharacterization of the distinction between lobbying and corruption is presented by Harstad and Svensson, who dene lobbying as any activity (including a transfer of resources) that aims to change existing policy, whereas corruption (in the form of bribery) is a tool to get around rules and regulations (Harstad & Svensson, 2011). They model this relationship in the context of rms paying bribes to avoid regulatory hurdles. Bureaucrats cannot commit to not asking for higher bribes in the future, and as bribes are rising in rms' investment levels they nd that continuing to bribe rule enforcers becomes relatively less desirable than joining a lobby group to change existing regulations. The result at the micro-level is that small rms bribe whereas large rms lobby. A small empirical literature has sprung up alongside the Harstad and Svensson model. Utilizing the 2000 World Business Environment Survey, Bennedsen and Feldmann nd that large rms have more political inuence relative to smaller ones. In addition they perceive corruption to be less of a problem and pay bribes to get things done less frequently. They also nd benets to political inuence in the form of higher sales and more government subsidies (Bennedsen & Feldmann, 2009). Campos and Giovannoni make use of the 2002 Business Environment and Enterprise Performance Survey, and while nding no direct correlation between bureaucratic corruption and lobby membership their analysis does conclude that large rms are associated with lobbying while small ones are more likely to bribe. In addition they nd that lobbying increases both political inuence and sales growth whereas corruption does not (Campos & Giovannoni, 2008). However their study suers from using association membership and lobbying interchangeable, as they do not examine other association functions. They also do not distinguish between bureaucratic and political corruption, and in doing so misrepresent their bribe payments variable in the survey. The most recent study to examine the lobbying and corruption relationship is by Sukiassyan and Nugent. They employ both the 2002 and 2005 BEEPS surveys, as well as the panel of rms surveyed in both. Exploring primarily rm outcomes they nd that association membership increases not only sales growth but also growth in exports and R&D. Most importantly they discover that lobbying has the smallest aect 6

on rm outcomes, with other association benets such as product market information and accreditation standards being more signicant. Like previous studies they nd that bribe payments do not aect rm performance (Sukiassyan & Nugent, 2011). The latter two studies are discouraging in how they approach potential endogeneity arising from either association membership or bribe payments being endogenous to rm outcomes such as political inuence and growth. While both make extensive use of instrumental variables and do in fact list the ones used, they make little or no attempt to establish the reasoning behind excluding these variables from the second-stage regressions. It is natural to become skeptical of these results as they are founded entirely on an econometric technique rather than logic or intuition. In the hope of avoiding these same concerns this study approaches endogeneity from a dierent direction, making use of the panel structure of the data to inform the analysis and obtain intuitive results regarding causality. In addition I extend the previous analyses in multiple directions, both by examining bribe payments in the context of changing membership status and by exploring the link between association membership and political corruption. 3 Data The econometric analyses utilize the 2002 and 2005 Business Environment and Enterprise Performance Surveys (BEEPS) conducted by the World Bank and EBRD. Local organizations administered the surveys in each country through face-to-face interviews with rm representatives. I limit the data to only those rms who were surveyed in both years, creating a two-period panel of 1446 rms for a total of 2892 observations. Table 1 presents the countries included in the panel and the associated number of observations, while table 2 describes the outcome and explanatory variables relevant to the analysis. Two key sets of variables concern association membership and corruption. Association membership status is available for all rms for both survey years, and a combination of rms' initial and nal status allows me to segment all rms into groups with regard to membership: in-in, in-out, out-in, out-out. From this I derive two additional variables, join_association (out-in vs. out-out) and leave_association (in-out vs. in-in), allowing me to analyze the characteristics of rms that join or leave associations between sur- 7

vey years. These variables capturing the dynamics of membership enable the empirical models to tackle some endogeneity concerns by examining the initial 2002 characteristics of rms that eventually join or leave an association, and exploring the change in these characteristics by 2005. Bureaucratic corruption data is encompassed in two variables, pay_bribe and bribe_pct_sales. The former takes a value of 1 if a rm has paid a bribe to get things done with regard to customs, licenses, or other services that bureaucrats provide, and 0 otherwise. The latter variable is the percent of sales that the rm has paid in all interactions with the bureaucracy. Net business association membership was little changed between the two survey years, with 44.5% of rms reporting membership in 2002 and 45.9% in 2005. Between the surveys 135 rms left their association, constituting 21% of all rms in business associations in 2002. However an additional 154 rms joined during this same period thus leading to an overall increase in membership. Conversely, the incidence of bribe payments to bureaucrats to get things done went down substantially, with 44% of rms paying in 2002 and only 36% in 2005. Of those rms that reported making bribe payments, the average amount paid also fell during this period from 3.6 to 3.1 percent of sales. While data on bribe payments deals only with bureaucratic corruption, the surveys do provide additional questions to gauge the rms' involvement in higher-level corruption such as unocial payments to legislators to aect votes, payments to judges to aect court decisions, and illegal contributions to political parties. It is this set of variables that the analysis will utilize when examining whether business associations encourage higherlevel corruption by opening the door to a previously inaccessible level of government. For this group of variables the question is asked in the following way: In many countries, rms are said to give unocial, private payments or other benets to public ocials to gain advantages in the drafting of laws, decrees, regulations, and other binding government decisions. To what extent have the following practices had a direct impact on your business: (a) Private payments or other bene- ts to Parliamentarians to aect their votes (b) Private payments or other benets to Government ocials to aect the content of government decrees (c) Private payments or other benets to judges to aect the decisions of court cases (d) Illegal 8

contributions to political parties and/or election campaigns to aect the decisions of elected ocials. The responses are ordinal on a scale of 0-4 from no impact to decisive impact. Though the survey makes a point of not directly ask rms their corruption activities with regard to high-level ocials, this is clearly the intention of the question. On average nearly 80% of respondents claimed that these practices had no direct impact on their business, suggesting that most rms do not engage in political corruption. Finally, the majority of studies utilizing the BEEPS data have used association membership and lobbying interchangeably. However associations can serve multiple functions and provide many benets to members. The BEEPS survey asks rms to rate the six association services described in Table 2 on a scale of 0 to 4 from being of no value to the rm to being of critical value. 2 Examining the mean values of each category I nd that rms identify lobbying as the second least important function, with only dispute resolution being less valuable. For both survey years rms cite product market information, information on government regulations, and help with certication and accreditation as the most valuable services provided to them. 4 Empirical Framework and Results 4.1 Corruption and Association Membership I rst determine whether rms do indeed move from bureaucratic corruption to association membership and whether this relationship depends on rm size by analyzing whether bribe paying rms are more likely to be members of business associations and vice versa. I also explore what rm-level characteristics aect corruption and association membership and whether they aect the two dierently. To that end I estimate probit models sharing most of the same explanatory variables but with dierent dependent variables. They are of the following form: 2 The categories are no value, minor value, moderate value, major value, and critical value 9

P (Y i,t = 1) = φ(x i,t β + v i + w t ) (1) When modeling association membership Y i,t takes the value of 1 if rm i is an association member in year t and 0 otherwise. When modeling corruption, it takes the value of 1 when a rm has paid a bribe to a bureaucrat. x i,t is a set of explanatory variables including new_product, regulations_pct_time, impact_bribe_legislators, log_employees, log_age, pct_foreign, exporter and wdi_gdpc. 3 v i is a vector of time-invariant control variables including the industry and country in which the rm operates, while w t is the year eect. In the association membership model I include pay_bribe as an explanatory variable, as well as an interaction term between pay_bribe and log_employees to determine whether the relationship is a function of rm size. In the corruption model I include rm_association. I also analyze an additional model with pay_bribe in 2005 as the dependent variable using join_association as the predictor of interest and include the 2002 value of pay_bribe and the other explanatory variables mentioned above as controls. This model will determine whether rms who have joined an association between 2002 and 2005 are subsequently less likely to pay bribes to bureaucrats relative to those rms that remained non-members. It is a way to look at the short-run dynamics of association membership and changes in corruption. The results for this and the previous model are presented in tables 3 and 4. I nd no statistically signicant relationship between association membership and bribe payments, i.e association members are not less likely to pay bribes to bureaucrats and bribe payers are not aected by membership status. The interaction term in model 1 of table 3 is also insignicant, indicating that this relationship is unaected by rm size. The results in table 4 indicate that rms do not change their behavior with regard to bureaucratic corruption even after joining an association, as evidenced by the fact that join_association comes up non-signicant as a predictor of bribe-paying. 3 In this and all future regression models using impact_bribe_legislators as an explanatory variable the regressions were also performed using impact_bribe_ocials as well as political_bribes, which is the rst principal component of these two individual political corruption variables. The results are both qualitatively and quantitatively similar. 10

The results indicate that either most business association members are primarily interested in services other than lobbying and these other functions are unrelated to bureaucratic corruption, or if we accept that lobbying plays a major role in association membership, that there is still an inherent dierence between bureaucrats and policymakers and bribing the former is not necessarily related to lobbying the latter. This is especially true if we consider lobbying by associations to be a public good, where lobbying eorts to reduce regulatory burdens aect all rms in an industry and not simply association members. Non-members may therefore choose to save themselves membership fees and still benet. The results also demonstrate that rms that recently introduced new products, rms burdened by government regulations, and those that are government owned are more likely to both be members of associations and pay bribes to bureaucrats. 4 In addition rms that bribe legislators are more likely to be association members but are also more likely to pay bribes to bureaucrats. This may suggest that certain rms are simply more willing to engage in corruption at any level, whether it is bribing rule-enforcers to sidestep regulations or rule-makers to precipitate regulatory change. Lastly I nd that association membership is positively correlated with older, larger exporting rms and rms with more foreign ownership. 4.2 Joining and Leaving Associations Making use of the panel structure of the data I now examine rm characteristics associated with joining / leaving an association between 2002 and 2005. I again employ the probit model dened in equation 1 above but now use join_association and leave_association, dened in the data section and in Table 2, as outcome variables and regress them on lagged values of the explanatory variables. The new model is as follows: 4 While the coecient on log_employees is negative, it is only signicant in a model that excludes the government ownership variable. These two have a correlation of 0.35, indicating (as we would expect) that the government generally invests in larger rms. When both variables are included as covariates it appears that having connections to the government is driving the results, and that larger rms are not any less likely to pay bribes to bureaucrats than smaller ones. 11

P (Y i = 1) = φ(x i,2002 β + v i ) (2) Since both join_association and leave_association as well as the xed eects take the same value in both 2002 and 2005, the above regression model utilizes exclusively the 2002 data and examines the rm characteristics that aect the likelihood of joining (leaving) an association over the following three years. Because of the short panel this approach is preferred over a xed eects regression or a rst-dierence model that uses association membership status as the dependent variable. Since the majority of variation in the data is cross-sectional rather than across time, the within estimator of a xed eects model is likely to be inconsistent and parameter estimates from xed eects models will have large standard errors. Conversely this approach does not suer from those drawbacks. I present these results in Table 5. The results of this analysis are in line with the previous results and demonstrate that rm exposure to bureaucratic corruption in 2002 does not aect the subsequent decision to join or leave an association. Note that since the sample available for this analysis is signicantly smaller, many of the explanatory variables have lost signicance as compared to the previous model. It is also plausible that the determinants of entry and exit are not the same as characterstics of an average member rm. Nevertheless I do nd that large and foreign owned rms are more likely to join, and foreign rms are less likely to leave than their domestic counterparts. Most importantly the decision to join an association is unaected by the impact on the rm of bribing legislators. However, the level of involvement in political corruption of existing association members is signicantly negatively associated with their propensity to leave in the following three years. That rms beneting from political corruption are less likely to leave associations indicates that the opportunity for political corruption is an advantage that association membership may oer. This is in accordance with the idea of selective incentives proposed by Olson, wherein exit becomes costly if the association provides a crucial service to its members (Olson, 1965). In addition we might expect that rms wanting to join 12

an association to bribe legislators would have had little success doing so on their own, so that the coecient on impact_bribe_legislators would be negative and signicant in the join_association model. That this is not the case may suggest that the desire to bribe high-level ocials does not play into the decision to join an association but rather becomes a factor after membership. I will test this hypothesis more directly in section 4.4. 4.3 Membership Benets and Services Provided Though lobbying is frequently used interchangeably with association membership there are in fact many other services that associations oer. As mentioned earlier the rms in the BEEPS data actually value lobbying to a far lesser degree than product market and regulatory information or accreditation and licensing services. But do all rms seek the same benets, or are certain rm characteristics associated with one particular association function or another? This issue is explored empirically using the six questions regarding the value of dierent association services (described in table 2 and in the data section) as outcome variables in an ordered probit framework of the following form: P (Y i,t = j) = φ(µ j x i,t β v i w t ) φ(µ j 1 x i,t β v i w t ) (3) The probability of observing an ordinal response category j is a function of the underlying latent continuous variable Y i, itself a function of the regressors, being within empirically determined threshold values µ j and µ j 1. x i,t is again the vector of explanatory variables used in the previous analyses, while v i is again the industry and country eects and w t is the year eect. The results are presented in Table 6. As expected all rms do not value every association service equally. Exporters are concerned with international product market information but not much else, whereas rms burdened by regulations are interested in obtaining information about regulatory hurdles as well as assistance in resolving disputes that likely result from non-compliance. Large rms and foreign-owned rms are more likely to derive multiple benets from membership, including lobbying services and information on regulations. Large rms 13

also benet from accreditation and certication services while foreign-owned rms value dispute resolution assistance, which is not surprising considering the inherent risks of investing in companies that operate in developing countries and particularly in the republics of the former Soviet Union. Of particular interest is that impact_bribe_legislators is positive and signicance only as a predictor of the lobbying and dispute resolution scores. That is, rms engaging in political corruption are more likely to cite lobbying and dispute resolution services as the most important reasons for being association members. Of the six services provided by associations these are clearly the two areas where political corruption can play a signicant role, either in bribing policymakers to aect favorable regulatory change or other public ocials to gain an upper hand in disputes with rms and regulators. The above result implies a close relationship between lobbying and political corruption within the framework of business association membership. 4.4 Do Associations Facilitate Corrupt Behavior? In section 4.2, when analyzing the factors inuencing rms' decisions to join or leave business associations, the results indicated that the likelihood of joining an association was unaected by the baseline political corruption level. However rms were less likely to leave an association if they were more engaged in political corruption in 2002, hinting at the corrupting inuence of association membership. This section will directly measure the aect of association membership on political corruption by analyzing the changing impact of bribing legislators on those rms that joined relative to non-members. As such the outcome variable of interest is the change in the impact score between 2002 and 2005. It is a discrete variable taking values between -4 and 4, where a value of 4 indicates that a rm was not impacted by bribes to legislators in 2002 but is decisively impacted in 2005 while a value of -4 indicates the opposite. I estimate an ordered probit regression model using the change in impact of political corruption as an outcome variable. The main explanatory variable is join_association and the model controls for the initial impact score as well as a number of additional 2002 rm characteristics, including log_employees, log_age, pct_foreign, exporter, wdi_gdpc, 14

as well as industry and country eects. Table 7 presents these results. The analysis indicates that rms who joined an association between 2002 and 2005 are signifcantly more likely to see an increase in their political corruption score than those who did not. Also, the higher the initial political corruption score, the less likely a rm is to increase their corruption activities. This is an obvious result as rms who are already decisively impacted by payments to legislators can at best maintain the same level of political corruption and are more likely to see it fall. That I observe signicance while controlling for the initial corruption score and other observables demonstrates that the change in political corruption activities is in fact the result of joining an association. But is it specically lobbying that makes bribing legislators so appealing, or do business associations exert a corrupting inuence on all members, regardless of whether they value lobbying or not? To get at this question I segment the join_association variable into two components: the rst, which I call join_to_lobby, takes a value of 1 if a rm joins an association between 2002 and 2005 and when interviewed in 2005 lists association lobbying services as at least of moderate value. The comparison group for which join_to_lobby takes a value of 0 is again all rms who were not members in 2002 and remained non-members. The second variable, called join_information, takes a value of 1 if the rm joined an association but listed lobbying as being of no value. It captures those rms whose primary interest in joining business associations is information provision regarding regulations, product markets, etc. Like the previous variable it takes a value of 0 if the rm did not join an association between the two survey periods. The analysis performed is identical to the previous regression, except I replace join_association with join_to_lobby and join_information (in two separate model specications) to determine their individual eects on the change in political corruption. I present these results in table 8. The results indicate that only those rms who join business associations and use them to lobby the government see an increase in their political corruption score. The nonlobbyists represented by join_information are not signicantly impacted. In addition the parameter estimate of join_to_lobby is also larger than join_association in the previous model meaning the eect is even stronger for this subset of rms. Combining 15

this result with the previous analysis strongly suggests that bribing policymakers is an activity many rms only begin to engage in after joining, and more to the point lobbying through a business association. 5 Conclusion Utilizing a panel dataset of rms based on the 2002-2005 BEEPS I analyze the relationship between rm-level bureaucratic corruption and business association membership. The results indicate that there is no statistically signicant relationship between association membership and bribe payments. Furthermore rms who have recently joined an association are no less likely to engage in bureaucratic corruption than their non-member counterparts. While the results generally support a big rms lobby, small rms bribe conclusion, the analysis makes clear that this is in no way causal. Simply put, lobbying eorts by business associations do not substitute for or reduce the scale of bribery and bureaucratic corruption. When looking at the dynamics of association membership in the context of a rm's decision to join or leave, I nd that the decision to join an association is unrelated to the impact on the rm of bribing legislators and other policy-makers; however for those rms that were already members in 2002 the impact of bribing legislators is signicantly negatively associated with the propensity to leave. The implication is that members that come to appreciate the benets of political corruption associated with membership are not likely to give it up, while non-members do not quite appreciate this opportunity until they join. Lobbying is an important but not singular function of business associations. To that end I examine other services provided by associations and nd that specic rm characteristics correspond to the choice of services they are seeking. For example foreign rms value dispute resolution assistance, which is not surprising given a large number of highly publicized conicts between foreign companies and their local partners. The analysis results demonstrate that rms with a high impact of bribing legislators score are signicantly more likely to be interested in lobbying and dispute resolution services. 16

It is clear therefore that corrupt rms see associations primarily as tools for bribing policymakers or other public ocials to either aect favorable regulatory changes or to gain an upper hand in disputes with rms and regulators. Finally, when analyzing the change in rms' exposure to political corruption between those rms that joined associations and those that did not, the results demonstrate that political corruption scores increased signicantly more among association members, again suggesting that bribing policymakers is an activity many rms only begin to appreciate and engage in after becoming association members. Business associations provide a broad range of services and rms do in fact take advantage of many of them. As previous studies have shown, these associations can improve the business climate, encourage investment and help rms grow. However it does appear as though the opportunity for political corruption that associations oer through their lobbying eorts is simply too tempting for many members to resist. These results suggest that the emergence of lobby groups in former Soviet republics and other countries will not result in a decrease of low-level bureaucratic corruption, and may instead facilitate state capture. 17

References [1] D. Austen-Smith. Interest groups: Money, information and inuence. Perspectives on public choice: A Handbook, page 2960321, 1997. [2] B. Begovic. Corruption, lobbying and state capture. CLDS Working Paper, 2005. [3] M. Bennedsen, S. Feldmann, and D. Lassen. Lobbying and bribes-a survey-based analysis of the demand for inuence and corruption. CESifo Working Paper Series No. 3496, 2011. [4] M. Bennedsen and S.E. Feldmann. Informational lobbying and political contributions. Journal of Public Economics, 90(4):631656, 2006. [5] N.F. Campos and F. Giovannoni. Lobbying, corruption and political inuence. Public Choice, 131(1):121, 2007. [6] N.F. Campos and F. Giovannoni. Lobbying, corruption and other banes. William Davidson Institute Working Paper No. 930, 2008. [7] S. Coate and S. Morris. Policy persistence. The American economic review, 89(5):13271336, 1999. [8] R. Damania and E. Yalc yn. Corruption and political competition. Economics Discussion Papers, 2008. [9] R.F. Doner and B.R. Schneider. Business associations and economic development: why some associations contribute more than others. Business and Politics, 2(3):261 288, 2000. [10] D. Duvanova. Business associations in post-communist russia: Explaining sectoral variation. In Midwest Political Science Association Conference, Chicago, 2006. [11] G.M. Grossman and E. Helpman. Protection for sale. American economic review, 84(4):833850, 1994. [12] G.M. Grossman and E. Helpman. Special interest politics. The MIT Press, 2002. 18

[13] B. Harstad and J. Svensson. Bribes, lobbying, and development. American Political Science Review, 105(01):4663, 2011. [14] M. Olson. The logic of collective action: Public goods and the theory of groups, volume 124. Harvard Univ Pr, 1965. [15] M. Olson. The rise and decline of nations: Economic growth, stagation, and social rigidities. Yale Univ Pr, 1982. [16] W. Pyle. Contractual disputes and the channels for interrm communication. Journal of Law, Economics, and Organization, 21(2):547575, 2005. [17] W. Pyle. Collective action and post-communist enterprise: The economic logic of russia's business associations. Europe-Asia Studies, 58(4):491521, 2006. [18] G. Sukiassyan and J.B. Nugent. Lobbying or information provision. Eastern European Economics, 49(2):3063, 2011. [19] J.D. Sullivan, K.E. Bettcher, and A. Shkolnikov. Business associations, business climate, and economic growth: evidence from transition economies. In annual conference of the International Society for New Institutional Economics, Boulder, Colorado, 2006. 19

List of Tables 1 Countries in Panel.............................. 21 2 Variable Descriptions............................. 22 3 Determinants of Association Membership & Corruption.......... 23 4 Corruption after Joining an Association................... 24 5 Joining and Leaving Associations...................... 25 6 Services Provided by Business Associations................. 26 7 Association Membership and Changes in Political Corruption Activities. 27 8 Lobbying and Political Corruption...................... 28 20

Table 1: Countries in Panel Country N Albania 130 Armenia 98 Azerbaijan 136 Belarus 92 Bulgaria 178 Croatia 122 Czech Republic 72 Estonia 138 Georgia 116 Hungary 118 Kazakhstan 120 Kyrgyzstan 80 Latvia 108 Lithuania 112 Macedonia 68 Moldova 64 Poland 156 Romania 128 Russia 82 Serbia 86 Slovakia 58 Slovenia 150 Tajikistan 36 Turkey 94 Ukraine 294 Uzbekistan 56 Total 2892 21

Table 2: Variable Descriptions Variable Name Type Description Association Membership: rm_association Binary Value of 1 if rm is in a business association and 0 otherwise join_association Binary Value of 1 if rm joined an association between 2002 and 2005 leave_association Binary Value of 1 if rm left an association between 2002 and 2005 Association Services (0-4; no value, minor, moderate, major, critical value): value_lobbying Ordinal Lobbying government value_dispute_res Ordinal Dispute resolution between rms or rms and government value_domestic_mkt Ordinal Domestic product market information value_int_mkt Ordinal International product market information value_info_reg Ordinal Information on government regulations Corruption Variables: pay_bribe Binary Value of 1 if rm has paid a bribe to bureaucrats to get things done bribe_pct_sales Continuous Total bribe amount paid as a percentage of sales impact_bribe_ocials Ordinal Impact on rm of private payments to ocials to aect content of decrees (no impact to decisive impact) impact_bribe_legislators Ordinal Impact on rm of private payments to legislators to aect votes (no impact to decisive impact) political_bribes Continuous First principal component of impact_bribe_ocials and impact_bribe_legislators Firm Characteristics: log_employees Continuous Log of the number of employees of the rm log_age Continuous Log of the age of the rm pct_foreign Continuous Percentage of rm owned by the foreign private sector exporter Binary Value of 1 if the rm is an exporter pct_govt_owned Continuous Percentage of rm owned by municipal or federal government new_product Binary Value of 1 if rm has developed major new product line in last 3 years regulations_pct_time Continuous Percent of senior management's time spent dealing with government regulations year Discrete Year (2002 or 2005) industry Categorical Industry in which rm operates cname Categorical Country in which rm operates wdi_gdpc Continuous GDP per capita in country of operation (in US $) 22

Table 3: Determinants of Association Membership & Corruption (1) (2) rm_association pay_bribe pay_bribe -0.193 (0.146) log_employees 0.217-0.0108 (0.0305) (0.0243) pay_bribe*log_employees 0.0377 (0.0419) rm_association -0.0889 (0.0708) new_product 0.230 0.257 (0.0694) (0.0651) regulations_pct_time 0.0108 0.0130 (0.00320) (0.00302) pct_govt_owned -0.00485-0.00602 (0.00117) (0.00112) impact_bribe_legislators 0.0883 0.274 (0.0395) (0.0380) log_age 0.129-0.135 (0.0496) (0.0463) pct_foreign 0.00457-0.000732 (0.00122) (0.00107) exporter 0.328 0.153 (0.0943) (0.0877) wdi_gdpc -0.00000635-0.0000410 (0.0000768) (0.0000731) year_2005 0.0828-0.0703 (0.148) (0.138) constant -0.545 0.320 (0.617) (0.502) Industry FE: Yes Yes Country FE: Yes Yes N 2189 2189 R 2 0.301 0.140 Standard errors in parentheses p <.10, p <.05, p <.01 23

Table 4: Corruption after Joining an Association 2005.pay_bribe join_association -0.0564 (0.169) 2002.pay_bribe 0.804 (0.130) 2002.log_employees 0.0469 (0.0585) 2002.new_product 0.275 (0.133) 2002.regulations_pct_time -0.0107 (0.00652) 2002.pct_govt_owned -0.00613 (0.00239) 2002.impact_bribe_legislators 0.0692 (0.0822) 2002.log_age 0.0406 (0.102) 2002.pct_foreign -0.00277 (0.00286) 2002.exporter -0.161 (0.229) 2002.wdi_gdpc 0.000469 (0.000268) constant -2.085 (0.971) Industry FE: Yes Country FE: Yes N 560 R 2 0.208 Standard errors in parentheses p <.10, p <.05, p <.01 24

Table 5: Joining and Leaving Associations (1) (2) join_association leave_association pay_bribe 0.0842 0.195 (0.143) (0.175) new_product 0.195-0.0354 (0.148) (0.192) regulations_pct_time 0.00155-0.00530 (0.00634) (0.00682) pct_govt_owned 0.000723 0.00360 (0.00249) (0.00284) impact_bribe_legislators 0.0185-0.269 (0.0790) (0.108) log_age -0.0393-0.0376 (0.109) (0.103) log_employees 0.214-0.0768 (0.0572) (0.0613) pct_foreign 0.00770-0.00591 (0.00249) (0.00298) exporter 0.263-0.126 (0.215) (0.208) wdi_gdpc 0.000425-0.000582 (0.000209) (0.000198) _cons -2.262 2.198 (0.797) (1.240) Industry Dummies: Yes Yes Country Dummies: Yes Yes N 615 379 R 2 0.2244 0.2369 Standard errors in parentheses p <.10, p <.05, p <.01 25

Table 6: Services Provided by Business Associations (1) (2) (3) (4) (5) (6) value_lobbying value_dispute_res value_domestic_mkt value_int_mkt value_info_reg value_accredit new_product 0.184 0.120 0.169 0.309 0.0947 0.0987 (0.0869) (0.0837) (0.0756) (0.0768) (0.0775) (0.0794) regulations_pct_time 0.00211 0.00861 0.00302 0.00412 0.00849 0.00508 (0.00355) (0.00353) (0.00344) (0.00348) (0.00337) (0.00358) pct_govt_owned -0.00111-0.000373 0.00115-0.000179 0.000791-0.0000750 (0.00144) (0.00140) (0.00125) (0.00140) (0.00120) (0.00126) impact_bribe_legislators 0.148 0.179 0.0613 0.0230 0.0653 0.0866 (0.0529) (0.0450) (0.0412) (0.0458) (0.0471) (0.0462) log_age 0.101 0.0305 0.0129-0.0355 0.0301 0.0147 (0.0625) (0.0577) (0.0520) (0.0521) (0.0473) (0.0534) log_employees 0.101 0.0281 0.0316 0.0481 0.0540 0.0582 (0.0293) (0.0275) (0.0277) (0.0278) (0.0265) (0.0269) pct_foreign 0.00279 0.00216 0.00121 0.00208 0.00233 0.00117 (0.00112) (0.00108) (0.00114) (0.00107) (0.00107) (0.00109) exporter -0.0531 0.00489 0.0366 0.429-0.0338 0.118 (0.103) (0.0941) (0.0899) (0.0864) (0.0850) (0.0874) wdi_gdpc -0.000178-0.000112-0.0000777-0.0000526-0.0000519-0.000118 (0.0000955) (0.0000877) (0.0000827) (0.0000862) (0.0000868) (0.0000849) year_2005 0.403 0.181 0.279 0.189 0.193 0.367 (0.196) (0.178) (0.166) (0.175) (0.181) (0.176) Industry FE: YES YES YES YES YES YES Country FE: YES YES YES YES YES YES µ1 0.704 0.899-0.290-0.0745 0.444-0.103 (0.532) (0.624) (0.513) (0.534) (0.507) (0.540) µ2 1.317 1.627 0.290 0.518 1.042 0.393 (0.531) (0.625) (0.514) (0.536) (0.506) (0.539) µ3 2.086 2.371 1.079 1.126 1.814 1.111 (0.537) (0.628) (0.514) (0.537) (0.507) (0.539) µ4 3.007 3.333 2.210 2.221 2.839 2.070 (0.544) (0.622) (0.517) (0.546) (0.511) (0.538) N 928 931 992 971 982 961 R 2 0.0729 0.0643 0.0309 0.0422 0.0506 0.0307 Standard errors in parentheses p <.10, p <.05, p <.01 26

Table 7: Association Membership and Changes in Political Corruption Activities (1) _political_corruption join_association 0.407 (0.184) 2002.impact_bribe_legislators -1.846 (0.170) 2002.log_age 0.0447 (0.0935) 2002.log_employees 0.0145 (0.0496) 2002.pct_foreign -0.00221 (0.00247) 2002.exporter 0.106 (0.207) 2002.wdi_gdpc 0.000640 (0.000133) Industry Dummies: Yes Country Dummies Yes µ 1-5.002 (0.828) µ 2-3.172 (0.733) µ 3-1.637 (0.664) µ 4-0.160 (0.622) µ 5 3.507 (0.639) µ 6 3.801 (0.636) µ 7 4.204 (0.629) µ 8 5.222 (0.713) N 572 R 2 0.3881 Standard errors in parentheses p <.10, p <.05, p <.01 27

Table 8: Lobbying and Political Corruption (1) (2) change_political_corruption change_political_corruption join_to_lobby 0.707 (0.333) join_information 0.411 (0.266) 2002.impact_bribe_legislators -1.888-1.909 (0.179) (0.193) 2002.log_age 0.0699-0.0451 (0.110) (0.105) 2002.log_employees -0.00728 0.0554 (0.0596) (0.0547) 2002.pct_foreign -0.00222-0.00131 (0.00272) (0.00277) 2002.exporter 0.0965 0.187 (0.215) (0.234) 2002.wdi_gdpc 0.000747 0.000588 (0.000248) (0.000151) Industry Dummies: Yes Yes Country Dummies Yes Yes µ 1-4.858-5.872 (1.021) (0.908) µ 2-3.032-4.007 (0.941) (0.800) µ 3-1.374-2.183 (0.881) (0.705) µ 4 0.110-0.731 (0.841) (0.663) µ 5 3.934 3.104 (0.864) (0.677) µ 6 4.236 3.438 (0.859) (0.675) µ 7 4.749 3.845 (0.861) (0.664) µ 8 5.867 4.762 (0.916) (0.747) N 489 509 R 2 0.418 0.405 Standard errors in parentheses p <.10, p <.05, p <.01 28