The Effect of Bureaucracy on Corruption: Evidence from the Regions of the Russian Federation

Ryan Gillette
Middlebury College Department of Economics 2008

Abstract: What is the relationship between bureaucracy and corruption? Using firm-level data from the 2005 BEEPS survey along with information from the Russian State Statistics department, this paper investigates the connection between bureaucracy and bribery across 14 of the regions of Russia. Results reveal a robust inverse relationship between size of bureaucracy and corruption. This could be due to competition between bureaucrats, increased oversight, or better staffing.

*Special thanks to professors Pyle, Prasch, Holmes, Maluccio, Warin, and all my family and friends for their support and guidance along the way. 1

A bureaucrat is the most despicable of men, though he is needed as vultures are needed, but one hardly admires vultures whom bureaucrats so strangely resemble. -Marcus Tullius Cicero

Bureaucracy and corruption have long been considered to be intrinsically connected. Bureaucrats are stereotyped as greedy, corrupt drains on the economy, siphoning away public funds and extracting bribes for their own personal gain. While bureaucrats are often in a unique position to abuse their governmental powers, the connection between bureaucracy and corruption is not as simple as it initially appears. The presence of corruption can undermine the legitimacy of a government, and inhibit the growth of efficient and equitable markets. If left unchecked, it can incite crime and destabilize formal legal institutions. While virtually every country experiences some level of corruption, it is especially prevalent in post-communist countries. As these countries underwent transitions towards a market based economy, the resulting liberation of prices and privatization of firms proved to be extremely detrimental to governments and economies. Uncertainty and confusion associated with the transitions exacerbated problems with rent seeking and clouded the chain of authority. As a result of the lack of clear leadership, government officials were better able to take advantage of their power to extract bribes from newly privatized firms, which needed support to adapt to the rapidly changing free market. This phenomenon was especially prevalent in Russia. The communist legacy coupled with the troubles of transition resulted in serious problems with bribes and corruption throughout the government and across the country. In 1998, the INDEM foundation estimated that the monetary amount paid out in bribes in Russia exceeded combined expenditures on science, education, health care, culture and art. Another study


suggested US$500 million was paid by small businesses in corruption payments every month towards the end of the twentieth century (Levin & Satarov, 1999). In some instances, these bribes proved to be nothing more than a trifling hurdle, but in other cases they became a serious impediment. For example, to enroll a child in kindergarten, parents had to typically pay US$20 and supply the director with chocolates or flowers; to secure a job at a public school, however, applicants often had to forego a whole year’s salary in bribes (Leonhardt, 1998). In addition to extensive corruption, Russia also has an expansive bureaucracy. The country’s vast size and considerable population have historically necessitated a large government with a dense bureaucracy that has carried over into modern times. Combined with the high prevalence of corruption, Russia provides an opportunity to study the connection between bureaucracy and corruption in a controlled manner. The current Russian Federation consists of 84 regions, each with diverse resources, populations, and regional governments. However, their differences are bounded by their shared cultural and political history. Furthermore, the overarching communist legacy enables easier control of political factors, as the central planners dictated development and growth of local governments in each region. After the fall of the Soviet Union, these regions gained unprecedented levels of independence, enabling them to grow in unique directions. This divergent development allows for a thorough examination of the effect of size of government on corruption. Recent data includes detailed information about bureaucratic concentration and bribes paid to government officials in several of the regions of Russia. While it is generally believed that bureaucrats are associated with higher instances of bribes and corruption, this may not be the case. In fact, there is extensive literature which suggests an opposite 3

effect of bureaucracy on corruption. Through various mechanisms, such as competition, heightened oversight, and increased efficiency of public services, it is possible that increases in the size of bureaucracy may lead to a decrease in the frequency and size of bribes. The purpose of this paper is to investigate the effect of bureaucracy on corruption by examining the relationship between the concentration of bureaucrats in each region and the frequency and size of bribes. Using this information, it becomes apparent that having more bureaucrats in a region is typically associated with lower levels of reported corruption.

Literature Review:

Definition of Corruption: An especially difficult aspect of corruption is the complexity in creating a precise definition of the problem. One of the most widely accepted definitions is provided by Treisman (2000), who states that corruption is defined as “the misuse of public office for private gain.” While this definition provides a reasonable generalization of the topic, it fails to cover many of the deeper intrinsic qualities of corruption. For example, the definition makes no distinction between types of corruption. Prasch (2007) describes different forms of fraud, and how “simple fraud” can occur at the employee level, but a more damaging “control fraud” can occur at the highest levels of an organization, in which a single corrupt executive can undermine an entire corporation.


Governmental corruption functions in a similar manner. Corruption can occur at different points of the government hierarchy and have varying effects. High level corruption is corruption that occurs in the upper echelons of a government, reaching up to those officials with the most power and influence. This is typically characterized by activity such as state capture, in which firms exert enough influence over law makers that laws are actually amended to give advantages to certain parties, or large government funds are transferred away from one sector of the economy to another where it is easier to extract bribes. Low level corruption, on the other hand, occurs at a more local level among inferior officers. Examples of this include bribes to hasten the procurement of certain government permits or patents, payments to government agencies to prevent inspections and thus avoid compliance, or kickbacks to regional decision makers to secure local government contracts and other public services. This sort of corruption is much more widespread, as it often goes undetected and unnoticed due to the relative insignificance of any one instance of petty bribery. In aggregate, however, the broad reaching implications of low level corruption make it at least as great a threat to a country as high level corruption.1


There is also the issue of economic and political rent seeking. Rent seeking occurs when officials abuse their power to profit from a good, without increasing or contributing to productivity. This was especially prevalent under command economies, as officials who had access to resources could sell them on the black market for significant profits. This phenomenon continues to be a problem, even in liberal capitalist economies. Rent seeking can be considered a form of corruption, but it falls under a different category than the other two types. The primary difference is that rent seeking is not limited to government officials. Anyone with access to below-market-price goods can engage in rent seeking. It can also be done in a semilegal fashion, as exhibited through the powerful agriculture lobbies in the United States, who transfer some of the burden of costs to the government by petitioning for subsidies. As a result, rent seeking is very hard to measure and define, and will not be a major focus of this paper.


Effects of Corruption: The consequences of corruption can be significant. Several studies have found a robust relationship between perceived corruption and the level of investment. Anderson & Gray (2006) show in a multinational study of countries across Europe and Central Asia that as perceptions of corruption increase, investment tends to decline, as firms fear their investments will be seized by corrupt bureaucrats. This not only discourages foreign direct investment, but also inhibits the formation of new small businesses. Entrepreneurs are less likely to invest capital in a new company if they fear they will be unable to pay the necessary bribes to get the company off the ground. As a result, corruption can function like a tax, creating a similar barrier to entry without providing the social benefit of increased public revenue. Other works examine how corruption encourages inefficiencies in markets. Contrary to a model proposed by Mauro (1998), in which it is suggested that bribes might encourage government employees to work harder, and speed the most efficient firms through the permit queue, Leite & Weidmann (1999) find that corruption actually decreases the efficiency of markets. The ability to extract bribes instead provides bureaucrats with an incentive to devise a slow and inefficient system for the processing of various documents, forcing firms to pay bribes to receive them in a timely manner. Mauro (1998) also shows that corruption can lead to an inefficient allocation of government funds. Resources can be siphoned away from beneficial public programs to areas in which it is easier to disguise costs and seize funds. For example, a corrupt politician may use money earmarked for public school textbooks to purchase a new fighter jet or finance a large government building, since it is easier to disguise costs in these fields. Dreher, Kotogiannis & McCorriston (2007) attempt to model this by 6

measuring cement consumption across countries, proposing that more corrupt countries will have more large scale construction projects to hide fund seizure. Results reveal a positive relationship between cement consumption and corruption, confirming the theory that corruption may lead to a misallocation of funds. These studies suggest that there is a link between higher levels of corruption and lower GDP growth. However, as Schleifer & Vishny (1993) show, once controlling for level of investment, there is no relationship between corruption and GDP growth. Combined with findings from Anderson & Gray (2006) which reveal no clear link between corruption and GDP growth, these studies conclude that any effect corruption may have on GDP growth is channeled primarily through lost investment spending. Another study by Slinko, Yakolev & Zhuravskaya (2004) shows that while regions with high instances of state capture exhibit lower rates of development for small businesses, overall growth is not significantly affected, suggesting the effects of corruptive protectionism to be more redistributive than draining. Furthermore, while it is possible that corruption could be causing a decline in GDP growth, it is also plausible that countries that already exhibit low GDP growth could have higher levels of corruption due to some third factor, such as poor political or legal institutions, which influences both. Alternatively, causation could run in both directions, with each factor being affected by the other. There is also literature showing that many of the detrimental effects of corruption may be caused by perceptions of corruption rather than actual corruption experiences. As Treisman (2006) explains, many commonly used corruption indices determine values by polling experts and scholars instead of the actual population, resulting in discord between real and predicted values of corruption. In a survey of countries from Burkina Faso to 7

Togo, experts predicted 54% of the population would encounter corruption during the year, while reported experiences revealed that only 13% actually experienced acts of corruption (Razafindrakoto & Roubaud, 2005). Furthermore, Mocan (2004) shows that, controlling for endogeneity and institutional quality, perceived corruption is determined more by institutional quality than by actual corruption experience. This is especially troubling, since many investors, both foreign and local, rely on corruption perception indices to evaluate investment risks in countries. It is thus possible that many of the apparent detrimental effects of corruption are caused by unfounded assumptions about levels of corruption within a country.

Causes of Corruption: The causes of actual corruption are varied. According to Ades & Di Tella (1999), one of the most prominent sources is an abundance of natural resources. Natural resources create significant rent seeking opportunities, attracting corruption and inhibiting proper development of markets. This is especially problematic with high value resources such as oil and diamonds. In some African countries, competition over valuable gem deposits has even lead to civil war, demonstrating the incredible corruptive power of natural resources (Lujala et al, 2005). Other studies hypothesize that the concentration of population contributes to corruption. Rose-Ackerman (1996) and Anderson & Gray (2006) both find that corruption thrives in cities and other densely populated regions. Mocan (2004) suggests this could be due to more impersonal relationships between bureaucrats and citizens in urban areas, allowing larger abuses of their governmental powers and greater ease in asking for bribes. It is also possible that corrupt officials are drawn to metropolitan areas 8

due to larger public budgets from which they can seize funds. However, it is also possible that regions with lower populations will experience higher corruption, due to decreased political competition and a greater ability to retain power despite suspicions of scandal. Thus, the quality of elections and democracy seems to play a role in corruption, as areas with poor democratic institutions may not be able to remove corrupt officials from office. Treisman (2000) also proposes that education plays a role in corruption. Along with Dreher et al. (2007) and Dininio & Orttung (2005), they find that highly educated societies typically exhibit lower levels of corruption. This could be due to the fact that more educated regions will tend to have higher literacy rates, which yields a more literate populace that is better informed and well read about issues such as corruption, and thus more capable of combating it. As a result, the degree to which the press is free to report news also has an influence on the level of corruption. Furthermore, as Alt & Lassen (2003) show, societies with higher incomes often enjoy lower levels of corruption, so education could be affecting corruption through the income channel as well. Bureaucratic wage is theorized to have an effect on corruption by Di Tella & Schargrodsky (2003). They hypothesize that by paying bureaucrats higher wages, they will have less incentive to attempt to extract bribes due to higher opportunity costs should they be caught. As Larina (2001) describes, in many cases it is not even possible to earn a living wage on bureaucratic pay. As such, bribes are considered a benefit of the job to boost income to an equilibrium level. Furthermore, when the difference between business pay and government pay is substantial, bribe size is hypothesized to increase to balance out the pay differential. Perhaps the most interesting factor is the role of bureaucratic concentration on corruption. The literature on this topic is varied and comes to diverse conclusions. 9

Dininio and Orttung (2005) undertook a study similar to this one of bureaucracy in regions of Russia, and found that an increase in the number of bureaucrats per region resulted in an increase in perceived corruption. However, their study relies heavily on measures of corruption provided by Transparency International and INDEM. As discussed earlier, these indices are often poor measures of actual corruption experience, and in some cases are not correlated at all. Thus, while they find a positive relationship between bureaucracy and corruption, their results must be regarded with caution due to the nature of their data. Furthermore, Schleifer & Vishny (1993) model the market for bribes, and find that with the presence of several corrupt officials, competition between these bureaucrats will drive down the price of bribes. When attempting to obtain permits and licenses, firms will seek out the bureaucrat who charges the lowest bribe. Barring any collusion among officials, other bureaucrats will be forced to match this price, or risk losing market share in the arena for bribes. Increasing the pool of bureaucrats could thus stimulate competition, dampening the level of corruption in a region. Rose-Ackerman (1996) suggests that there are two types of markets for bribes. In some cases, the market is roughly competitive, and approximately the same bribe price is charged to all firms. This typically occurs in the procurement of routine government services, such as general permits and documents. In the absence of competition, however, bureaucrats may be able to price discriminate, charging higher bribe prices to those most wiling to pay. Those firms that stand to gain the most from the bribe will be willing to pay the highest price. For example, a firm with a high volume of exports would be much more willing to pay large bribes to customs officials than small firms that trade mostly domestically. 10

Brown et al. (2007) find a negative relationship between size of bureaucracy and frequency of bribes in Russia, but proffer a different explanation. Instead of competition, they suggest that more bureaucrats could lead to better staffed public agencies. As a result, firms will not be forced to resort to bribery to receive government services in a timely fashion. Mauro (1998) and Rose-Ackerman (1996) both discuss how heightened oversight can lead to lower levels of corruption. By increasing the chances of getting caught taking or giving a bribe, some bureaucrats and firms will find it too risky to engage in corrupt practices. As a result, increasing the number of supervising bureaucrats could help curtail corruption. However, as Treisman (2000) finds, countries with more tiers of government tend to have higher levels of perceived corruption. Instead of regulating their underlings, these supervising bureaucrats become just another obstacle, generating even more bribes. Waller, Verdier & Gardner (2002) reach an interesting mixed conclusion. By examining the number of tiers of government, they find that a single bribe setting monopolist will set a lower bribe price per investment project relative to a chain of successive monopolists. However, the total volume of bribes obtained will be higher, due to the increased amount of investment. Yang (2005) also finds that an increase in the perceived probability of meeting an honest official can drive down bribe frequency. The wide diversity of opinions regarding the role of bureaucracy in determining corruption demonstrates the importance of the issue. Combining the findings of all the various studies suggests that bureaucracy plays an uncertain role in determining corruption. The presence of more bureaucrats in a region has the potential to increase the volume of bribes due to additional bureaucratic hurdles. At the same time, however, competition between these bureaucrats may drive down the price of any one bribe. The 11

added risks of heightened oversight and greater efficiency of government agencies could also affect the amount of corruption in an area. There are many other potential factors that could influence corruption. Studies have found connections between colonial history, cultural factors, legal origins, religion, and other sociopolitical indicators to affect levels of corruption. For the scope of this paper, however, these variables are insignificant since this study examines regions of Russia, which share a common political and cultural background. As a result, strong entho-cultural unity throughout Russia will effectively control for many of these aspects.

Data & Methodology: The primary source of data for this paper is the 2005 Business Environment and Enterprise Performance Survey (BEEPS) jointly provided by the World Bank and European Bank for Reconstruction and Development. The survey was first carried out in 1999 in 25 different transition countries, and was repeated in 2002, 2004 and 2005. The 2005 survey was conducted through a series of face-to-face interviews with over 9000 firms in 27 different transition countries. The questionnaires were designed to illuminate issues in transition, ranging from success of privatization to levels of competitiveness to sources of corruption. Firms were selected randomly, but the sample was designed to ensure certain sampling thresholds. At least 10% of the firms were required to be “small” (2-49 employees), and 10% “large” (250+ employees). Firms larger than 10,000 employees were removed from the sample, as were new firms that had been established after 2002. A minimum of 10% of the firms had to have foreign ownership and another 10% state control; additionally, 10% of firms had to be in a “small city/countryside” location 12

(population under 50,000). The sectoral composition of the sample was selected to reflect relative contribution to GDP. Of all the firms that were contacted in Russia, 31% completed an interview, 41% refused or were unavailable, and the remaining 28% were not eligible or unnecessary due to fulfillment of sampling quotas. Refusals were typically attributed to lack of time, disinterest, or need for approval from those in charge. Usually, the owner or director of the firm was interviewed. However, in some cases a manager or financial officer was surveyed. There is no reason to believe any of these factors would affect a firm’s answers regarding bribery and corruption. To ensure honesty of answers, firms were assured of their anonymity throughout the survey. Each interview was carried out privately and personally in the local language by trained individuals, and firms were coded with a serial number so that no names were attached to their responses. Before sensitive questions regarding bribes were asked, interviewers were instructed to read to participants the following assurances:
We are interested in your opinions in a personal capacity, we do not imply in any way that your company makes unofficial payments/gifts, and we recognize that your company neither approves nor condones the use of unofficial payments/gifts. The responses you give will be aggregated and presented in purely statistical terms; any comments you give me cannot be attributed to either you or your company.

Sensitive questions were phrased in a manner to encourage honesty through depersonalization; two of the critical questions to this paper are structured as such: “It is common for firms in my line of business to have to pay some ‘additional payments/gifts’ to get things done” and “on average, what percentage of total annual sales do firms like yours typically pay in unofficial payments/gifts to public officials?” As a result, instances of no response are low, but it is unclear how truthfully firms chose to respond. However, since there is little incentive to inflate the actual bribe values, any bias would probably


downplay the true level of corruption. If these real values could be measured, they would likely further strengthen the robustness of this study. The most valuable component of the 2005 survey is that it contains a locational variable for firms in Russia, allowing for distinction between different cities and regions within the country. This makes it possible to build a database of firm-reported corruption and bribery within different regions of Russia. The 2005 BEEPS survey covered 601 enterprises in Russia, but unfortunately the firms were only distributed among 14 of Russia’s 84 regions. Figure 2 in Appendix 1 shows the breakdown of firms by region. The BEEPS data provide valuable measures of corruption and bribery based on actual experiences rather that perceptions. There are two central questions from this survey that serve as the focal point for this paper. The first is a question regarding the frequency of “additional payments/gifts to get things done with regard to customs, taxes, licenses, regulations, services, etc.” This is measured on a scale of 1 (never) to 6 (always). Another question asks what percentage of annual sales is used to make “unofficial payments/gifts to public officials.” This, combined with another question at the end of the survey asking for an estimate of total sales makes it theoretically possible to determine the total monetary size of bribes paid per year. Unfortunately, nearly half the firms declined to report sales estimates, significantly affecting the usefulness of this information in larger regressions. There are several other important variables from the 2005 BEEPS survey that are included in regressions. One is another question pertaining to corruption, but at a higher level. It asks on a yes-no basis if firms attempt to affect legislation for the benefit of their company, providing a potential proxy for state capture. Lobbying, however, would be


included under this variable, so it is difficult to distinguish illicit imposition of power from simple legal influence. Another series of variables inquires about the frequency with which various departments are bribed, again on a 1-6 scale. This allows for partial control of the “type” of official being bribed. Since it is unreasonable to assume all bureaucrats have the same governmental powers and hence the same ability to extract bribes, this variable allows for a closer comparison of similar officials. The types of departments range from tax collection to the court system to fire inspection. The 2005 BEEPS survey also included a panel component, allowing for comparison between the 2005 and 2002 surveys. While there were over a thousand firms surveyed in both 2002 and 2005, only 41 of those were in Russia, preventing any sort of useful analysis of trends. A preemptory glance at all the data from both surveys, however, shows some interesting patterns, illustrated in Table 1 below. The frequency of bribes is similar in both data sets, both in terms of median and mean. The percentage of sales paid in bribes, however, is quite different. The 2002 data set has a much wider variance than the 2005 survey, with a median of .5% and a maximum of 20% of sales. The 2005 data set, however, has a median of only .1% and a maximum of 10%. These differences are significant at the 1% level.



Std. Deviation


2002 1.43 0.5 2.49 20 2005 1.06 0.1 1.66 10 Table 1: % of sales paid in bribes in 2002 and 2005

When extrapolating monetary size of bribes based on the firms that chose to report annual sales data, however, we find that more money was paid out in 2005 than in


2002. In 2002, an average of $11,728 was paid in bribes, while in 2005 over $18,000 was paid (see Table 2 below). This shows that while larger percentages were paid in 2002, more money was actually lost to bribes in 2005. Since the reported frequency of bribes is essentially identical in both data sets, this means the increase is likely due to growth in reported sales from 2002 to 2005. It also suggests that bureaucrats have some knowledge of firm profit margins, and are able to charge higher bribe prices as companies become more successful. This difference, however, is only significant at the 88% level, perhaps due to the fact that nearly 40% of the firms in 2005 declined to report sales data. As a result, the monetary size of bribes was excluded as a dependent variable in regressions due to the low response rate and lack of significance of the overall regressions.



Std. Deviation


2002 $11,728 $488 $11,728 $600,000 2005 $18,500 $500 $114,490 $1,851,800 Table 2: Monetary size of bribes paid in 2002 and 2005

The BEEPS data also contains questions pertaining to various firm characteristics, such as size and origin. These questions are used as controls in some regressions, but it is possible the survey does not accurately represent the different types of firms in a region. For the regions with a high volume of survey respondents, such as Moscow, it is safe to assume there is an accurate representation of the different sizes and types of firms within the region. For many of the smaller regions, such as Kemerovo, it is possible that only the largest or most prominent firms were interviewed, potentially biasing the true effects of firm characteristics on corruption. The Russian Federation is particularly discreet regarding its bureaucracy. As a result, it is difficult to find constant, detailed data on the size of bureaucracy at the 16

regional level. The most recent available data was published in the 2001 Statisticheskii Builleten’ created by the Russian State Statistics Department Goskomstat. This summary of Russian bureaucracy includes the number of bureaucrats per region, both at the federal and regional levels. It also divides bureaucrats by branch of government, separating executive from legislative from judicial bureaucrats. Throughout this study, bureaucracy was measured as both the absolute number of government employees per region and the number of bureaucrats per capita. When running regressions, coefficients followed the same general trends regardless of the measure, but the results were often far less statistically significant and the adjusted Rsquared values were lower when using the per capita measure of bureaucracy. As suggested in Dininio & Orttung (2005), using the number of bureaucrats is a more accurate indicator of the absolute size of the state than using a per capita measure. Due to the drastically different sizes of Russian regions, both in terms of population and territory, measuring bureaucracy on a per capita basis places too heavy an emphasis on population as a determinant of the size of bureaucracy. More expansive regions will require larger numbers of bureaucrats in assorted local government positions, despite the fact that they may have drastically smaller populations. For example, Moscow has the largest number of bureaucrats, but the smallest number of bureaucrats per capita. This does not accurately reflect the fact that in Moscow all the bureaucrats are in a much closer proximity to the population relative to a remote, expansive region such as Krasnoyarsk. As a result, using a per capita measure of bureaucracy is not as useful a tool for this study as the absolute number of bureaucrats. Nonetheless, once controlling for the size of the region, results were similar, but less significant. Regressions using a


measure of bureaucrats per area suffered from similar problems, and were typically highly insignificant. The data also include information regarding government salaries by region, which will help control for wage differentials. Several other control variables, such as regional population, regional domestic product, geographic size, and education are taken from other surveys provided by Goskomstat. Presence of natural resources would be included in regressions, but none of the 14 regions covered in the BEEPS survey exhibit significant amounts of any valuable natural resource. A series of political variables was taken from the Социальной Атлас Российских Регионов (Social Atlas of the Russian Regions) provided by the independent institute Social Politics. This group evaluated the quality of different political factors in the regions of Russia, such as fairness of elections, freedom of press, and political transparency. Unfortunately, their methodology was largely based on aggregating expert opinions, calling into question the validity of their results. As such, these variables will be used with caution, since this paper has already discussed the flaws of perception-based measurements. One of the biggest concerns when studying corruption is endogeneity. There are many reasons why the concentration of bureaucrats in any region may not be exogenous to the amount of corruption. It is possible that the presence of significant corruption profits for government employees would draw other officials to the region, increasing the concentration of bureaucrats. Another possibility is that the poor business environment created by high levels of corruption could encourage firms to move to less corrupt regions, thus decreasing tax revenues in these corrupt areas and lowering the regional budget for employing bureaucrats. 18

The nature of this study helps alleviate some of these problems. As Hill and Gaddy (2003) discuss in their book, the central planners played a critical role in the development of regional governments, especially in the later years of the Soviet Union. This means that the initial distribution of bureaucrats at the time of the collapse of the USSR was almost entirely determined by economic factors and not affected by other elements. As a result, there has been a much shorter period of time, largely characterized by significant changes in economic and political structure, for bureaucrats to discover opportunities for corruption profits and seek entry into these high corruption regions. The desire of local bureaucrats to prevent entry of other corrupt officials combined with the short time horizon for transferring means that any movement trends are likely small and still in progress, and thus not a significant impediment to the results of this paper. Other questions in the 2005 BEEPS data help show that corruption does not affect the performance of a business enough to make a firm move to a different region. When asked if corruption was an impediment to business success, most firms responded that corruption was no more than a “minor obstacle.” This suggests that corruption does not drastically affect firms in a way that would encourage them to relocate. Also, a question regarding percentage of actual sales reported for tax purposes shows only a small, weak correlation between reported sales and corruption, demonstrating that, all else equal, more and less corrupt regions should expect to collect comparable tax revenues. This means the regional governments should be equally well funded and equally capable of hiring and retaining bureaucrats. The host of controls introduced helps eliminate many of the other pathways by which corruption could influence the size of the bureaucracy. Controlling for regional population is an important factor, as it directly affects both corruption and bureaucracy. 19

More highly populated regions likely require more bureaucrats simply due to greater demand for government services, while at the same time population density is shown to be associated with higher levels of corruption. Regional domestic product is likely to have a similar effect. For the regressions in this paper, both of these factors were controlled for using the average of data from the years 2001 and 2005. Since the bureaucracy data are from 2001, and the BEEPS survey is from 2005, using the mean values of data from the two years helps avoid bias. Regressions were checked using the individual year data to ensure accuracy of results. For other variables this is either not possible or not relevant, so the most appropriate available data were used. Detailed descriptions of all the variables included in this study can be found in Figure 1 of the first appendix.2 Multicollinearity also proved to be a significant issue with many of the regressions. Figure 3 in Appendix 1 lists correlation coefficients between several of the independent variables. This may have limited the statistical significance of the regressions, and inhibited the ability to accurately determine the individual sources of variation in corruption. Removing highly correlated variables did not significantly alter the magnitude of coefficients, but lowered R-squared values. Thus, they were all typically included in regressions due to the extensive literature noting their importance.

An instrumental variable model was also run in addition to the regressions presented as a validity check. Results are robust and similar to those found in the body of the paper, suggesting endogeneity is not a significant impediment to this analysis. See Appendix 2 for more details.



Results: To investigate the role of bureaucracy on corruption, a series of regressions were run using a variety of models (see Table 3 below). The primary dependent variables used in the regressions were bribesize and bribefreq, which measure the size and frequency of bribes. The first set of regressions examines the effects of bureaucracy on bribe frequency. Since the indicator of bribe frequency from the BEEPS data is measured on a scale of 1-6, binary variables representing different ranges of bribe frequency were created, and probit models were run to examine changes in the probability of paying bribes.3 The first regression uses the binary variable anybribe, which is equal to “1” if a firm pays any bribe. This basic model confirms that bureaucracy is closely tied to corruption, demonstrating that the likelihood of paying a bribe declines as the number of bureaucrats increases. Although statistically significant, the magnitude of the drop not large: adding an additional 1000 bureaucrats to a region only leads to a 1% drop in the probability of paying a bribe. Furthermore, this initial model is consistent with many of the patterns of corruption found in other literature. It shows that the probability of paying a bribe increases in more populated regions along with areas exhibiting a higher regional domestic product. The coefficient on diffwage shows the wage differential between the average employee and the average bureaucrat is associated with a lower existence of bribes. Past theories have proposed the opposite relationship, but it is possible that bureaucrats make up the wage gap by simply extracting larger bribes as opposed to more bribes. The coefficients on unistudents and democracy have expected signs, but are not

An alternative model was run using an ordered probit model. Overall results were similar to probit analysis. This model is described in more detail in the second appendix.



Table 3: Probit regressions on bribe presence and frequency4 anybribe oftenbribe

I bureaucrats burpc popavg territory rdpavg diffwage unistudents mediafreedom democracy privatized private foreign subsidary small medium mine construction manuf transport trade













































































Standard deviations presented in parentheses below. * denotes significance at 10%, ** at 5%, *** at 1%








real_estate hotel_restau central northwest siberian southern urals distmosc













Pseudo R2 N

0.0355 548

0.0431 548

0.049 548

0.0545 548

0.0377 548

significant. Mediafreedom has a counterintuitive sign, but is also insignificant. Since both mediafreedom and democracy are subjective measures, however, results must be interpreted with caution. Due to the lack of significance and concerns about multicollinearity, these variables were dropped in subsequent models. Models II through V use the variable oftenbribe, which is equal to “1” if firms pay bribes often, and “0” otherwise.5 Model II reveals a strengthened negative relationship between bureaucrats and bribe frequency with a coefficient over double in size. Furthermore, this regression introduces firm level characteristics to control for firm type


All regressions were checked using each of the dependent variable measures. Trends were the same no matter the measure, and variations in coefficients can be reliably attributed to changes in included independent variables as opposed to the specific measure of the dependent variable. Other measures are included to demonstrate consistency of results.


and size. Small firms appear to enjoy a hearty decline in bribe frequency relative to large sized firms, suggesting bureaucrats prey more heavily on larger firms, which likely have a greater ability to pay. Private firms with no state history also show a statistically significant increase in bribe frequency relative to the omitted category, otherorigin. This could be due to the fact that privatized firms often already have established relationships with government employees, reducing the need for impersonal bribes. Once controlling for sector, however, privatized firms also exhibit an increase in bribe frequency.6 Sector dummies were also introduced to examine trends in different areas of the economy. While these dummies tended to be insignificant in regressions, the dummy for construction firms was routinely positive. When excluding other firm level characteristics, it even proved to be significant at the 10% level. This suggests it is easier for officials to extract bribes from construction firms, perhaps due to the high volume of permits and inspections required, along with large, soft budgets in which it is easy to hide unofficial extra payments. Because of this, firms that specialize in construction tend to pay more bribes on average than firms that work in other sectors. Model IV introduces controls for geographic features in an attempt to capture any differences that may arise from a region’s location within Russia. One method of accomplishing this was to study the next level of regional organization. In 2000, President Putin divided the country’s oblasts and regions into seven federal districts, each with its own presidential envoy to oversee activity in the region. These envoys are appointed by the president, and are charged with ensuring regional compliance with federal laws and regulations. As a result, it is possible that the degree to which corruption


Several regressions were also run using interaction terms between bureaucracy and assorted firm characteristics including size, sector and origin; these coefficients proved to be highly insignificant.


is detected and punished may vary depending on the district and the envoy, allowing for variation in the extent to which bureaucrats attempt to extract bribes. Figure 4 in Appendix 1 shows the distribution of observations among the districts in the data set. Regressions including these dummies revealed that on average, firms located in the Northwest federal district are 37% less likely to pay frequent bribes relative to the other federal districts. All the Northwest firms in this data set are situated in or around St. Petersburg7, so it is difficult to assess whether the effect is caused by characteristics of this particular city, or some aspect of the Federal supervision. St. Petersburg is often considered the most “European” of Russian cities, but it is unclear if this notion transcends into the realm of politics. Citizens appear to be more politically aware, however, as evinced by large public unrest over a 2003 campaign for governor that was deemed unfair and dishonest (BBC, 2003). This could perhaps translate into a higher intolerance of fraud and corruption. It is also the birth city of Vladimir Putin, whose diligent pursuit of law and order may resonate especially strongly in his home town. As Francken et al (2005) show, the remoteness of a region or distance from central authority can also affect corruption. If regions are more isolated from the central government, it may be easier for bureaucrats to extract bribes without fear of repercussion. To control for this, another variable was introduced measuring the linear distance of each region’s administrative center from Russia’s capital, Moscow. This variable, however, proved to be largely insignificant, but often negative in probit regressions.


The administrative center of Leningradskaya Oblast is the city of St. Petersburg, but due to the immense size, the city itself is an independent oblast. The same is true of Moscow and Moscovskaya Oblast.


Model V uses bureaucrats per capita as a measure of bureaucracy to illustrate the consistency of results. When using the variable burpc, the statistical significance of many coefficients declines, along with pseudo R-squared values. General patterns, however, remain similar to other regressions. Another set of probits was run to examine the effects of bureaucracy on bribes to specific departments. These results are presented below in Table 4. Firms were asked how often, on a scale of 1 (never) to 6 (always) they paid bribes to individual departments and for certain services. In this study, all of the dependent variables are binary, taking on a value of “1” if a firm ever paid a bribe to that department (similar in definition to anybribe). This measures the existence of bribes in different departments and agencies. Results reveal that the number of bureaucrats has a negative, statistically significant effect on the probability of paying a bribe in several processes, including obtaining permits, getting connected to basic infrastructure, dealing with the health and fire inspectors, and resolving legal affairs in courts. While the data do not break down bureaucrats by agency, the trends suggest that as the number of government officials in a region increases, the presence of bribes in specific departments tends to decline. Some of these services, such as obtaining permits and getting connected to local infrastructure, are likely rendered by bureaucrats who serve essentially identical functions. Since it is easy to turn to another bureaucrat to get a basic permit, it is possible competition between officials helps reduce corruption. Similarly, in departments in which bureaucrats offer unique services, such as the procurement of government contracts, there is no apparent connection between bureaucracy and the presence of bribes. This could be due to the fact that there is no substitute for a government contract. As a result, firms have much less 26

Table 4: Probit regressions on existence of bribes in different departments

Dependent Variables:
permits bureaucrats popavg territory rdpavg diffwage unistudents mine construction manuf transport trade real_estate hotel_restau getconnected govcontracts healthinspector fireinspector environinspect taxcollector customsimports courts






















































































































firm level controls fed district controls Pseudo R2 N

yes yes
0.0755 542

yes yes
0.0803 550

yes yes
0.1722 480

yes yes
0.0772 535

yes yes
0.0652 549

yes yes
0.0792 549

yes yes
0.0488 541

yes yes
0.122 486

yes yes
0.0988 487


bargaining power when negotiating for bribes, as it is more difficult to turn to another official to receive the same service. The court system is a curious intermediary, since firms are typically stuck with one judge. However, there are many aspects of the legal process in which bureaucrats play a role. In the Russian court system, claimants are required to pay legal fees before any hearings or trials. It is possible to ask for postponement of payment, but there is a complicated procedure associated with the process. Additionally, claimants can request to have a panel of three judges instead of just one, requiring an equally lengthy application (Hendley, 1998). This provides judicial bureaucrats the opportunity to demand bribes to hasten or accept these requests, causing them to be subjected to mechanisms such as competition in the same fashion as other bureaucrats. At the same time, however, an overburdened court system could be streamlined by the addition of more employees. It is possible that regions with more bureaucrats have better staffed courthouses that are more capable of efficiently handling cases. As a result, firms do not need to make extra payments to receive a court ruling in a timely fashion, suggesting better staffing plays an important role. Alternatively, a higher number of court affiliated bureaucrats could simply imply a better functioning legal system, reducing the need for bribery and increasing the consequences if caught.8 It is also possible that declines in bribe frequency are a result of better staffed agencies. As Brown et al (2007) suggest, if a department is understaffed, it may function inefficiently. This could cause queuing problems and create delays, encouraging firms to pay extra to expedite the process. With a well staffed bureaucracy, agencies may perform


A set of regressions was run using the number of “judicial” bureaucrats in a region. The coefficient on this variable was negative, but only significant at the exclusion of the other types of bureaucrats. As a result, it is unclear if lower instances of bribery in the legal system are associated with more judicial bureaucrats, or simply more bureaucrats in general.


better and faster, reducing the need for these efficiency payments. Examining sector dummies also shows which industries are most affected by which departments. As discussed earlier, construction firms seem have the greatest likelihood of offering a bribe. Results reveal that businesses in the construction sector are roughly 25% more likely to pay bribes for permits, fire inspection, and court services, and over 55% percent more likely to pay bribes to obtain government contracts relative to the omitted sector, “other industries.” This seems logical, since these services are all essential to the construction industry. Similarly, the mining sector exhibits significant increases in the probability of paying a bribe to the health inspector and customs officials, two of the major departments that mining operations interact with on a regular basis. In addition to the differences between departments, it is possible there is a difference between types of bureaucrats. The government of the Russian Federation is structured similarly to the American government, separating power into three branches – legislative, executive, and judicial. Additionally, there are three levels of governmental hierarchy – federal, regional, and local. Thus far, the measure of bureaucracy has included all bureaucrats in each region, but there is reason to believe bureaucrats in certain branches and hierarchies may have differential effects on corruption. For example, there is an agency of the Kremlin ominously entitled The Control Directorate. Founded in 1991 as an arm of the Presidential Executive Office, its purpose is essentially to oversee implementation and adherence to federal law and to punish violators.9As such, the Directorate is also charged with curtailing and controlling corruption within regions. This suggests that larger quantities of federal level executive bureaucrats may translate into a greater presence of the Control Directorate, discouraging bribery and corruption in

For a more complete description, check the Kremlin website


the area. The data on bureaucracy breaks down bureaucrats by branch, and executive bureaucrats by hierarchy (see Figure 5 in Appendix 1 for breakdown). Using a probit model with oftenbribe as the dependant variable, Regression I in Table 5 below shows that increases in the number of executive branch bureaucrats leads to a statistically significant decline in the probability of paying frequent bribes. Regression II further breaks down executive branch bureaucrats by hierarchy, with interesting results. Increases in regional and local level bureaucrats leads to decreases in corruption, but for federal bureaucrats, the coefficient is insignificant but positive. The coefficients for regional and local bureaucrats are not surprising, as they mirror earlier findings, but the federal coefficient is perplexing. It is possible that the number of federal level bureaucrats has no effect on the size or effectiveness of departments such as the control directorate, negating any causal link to corruption. Furthermore, since federal level bureaucrats are likely involved with larger governmental affairs, it is probable they are not in a position to extract bribes from local firms, causing them to have no impact on bribery. This suggests that the decline in corruption associated with higher levels of bureaucracy is not a result of better monitoring: the increased prevalence of overseeing officials appears to be unconnected to changes in bribe frequency. The strong coefficients for regional and local bureaucrats, however, reinforce the mechanisms of competition and better staffing: as regions become more saturated with low-level government employees, corruption tends to decline. Greater competition among these officials could drive down bribe prices, or better staffing could reduce the need for firms to pay bribes. Alternatively, since federal level bureaucrats are typically appointed and thus easily transferable, it could be that federal level bureaucrats are sent to regions already exhibiting higher instances of bribery in an attempt to better combat 30

corruption, leading to an endogeneity problem. However, since bureaucracy data is already effectively lagged by four years relative to corruption data, an accurate analysis of the effects of federal level bureaucrats would require a more in-depth study with more recent data.

Table 5: Probit models breaking down bureaucratic hierarchy (dependent variable oftenbribe) I executive federal regional local legislative judicial otherorgans popavg territory rdpavg diffwage unistudents II




















(0.0062) yes yes no 0.0536 510

sector controls firm level controls fed district controls10 Pseudo R2 N

yes yes no 0.057 510


Dummies for the different federal districts were omitted due to perfect co-linearity of a few of the districts with the disaggregated bureaucratic data.


The high levels of correlation between the independent variables complicated the analysis of bribe size (see Table 6 below). Multicollinearity made it difficult to obtain significant coefficients on several of the independent variables. Since
Table 6: Tobit regressions using bribesize as the dependent variable I bureaucrats popavg territory diffwage rdpavg unistudents small medium privatized private subsidary foreign _cons II III





















(0.4629) no no 0.0075 556

(1.1277) yes yes 0.021 556

sector controls fed district controls Pseudo R2 N

no no 0.0041 556

the variable for bribe size as a percentage of sales is bounded by 0 and 100, Tobit models were run to examine trends in corruption. Although Model I in Table 6 showed a 32

statistically significant decline in bribe size as the concentration of bureaucrats increased, this significance vanished with the inclusion of more variables in Models II and III. The difference between average regional per capita income and bureaucratic wage proved to be an important factor in determining bribe size. Model II shows that as the difference increases, the size of bribes tends to grow as well, suggesting that bureaucrats augment their salaries to closer match this outside opportunity. However, when controlling for regional domestic product, even this significance vanishes. This could be due to the fact that bureaucratic wage, per capita income, and regional domestic product are all very highly correlated, with correlation coefficients upwards of .98. This means it is possible the increase in bribes could be attributed to the size of the regional economy and greater potential for extortion, reducing the relevance of the wage differential. Model III introduces firm, sector, and regional level controls. Coefficients on the individual sectors and federal districts were highly insignificant and hence omitted from tables. Firm level characteristics reveal medium sized firms pay larger bribes as a percentage of sales relative to large firms, with statistical significance just shy of 10%. Furthermore, firms owned by foreign partners suffer a huge penalty in bribe size, paying over 2.5% more of annual sales in bribes than other types of firms. This could be because foreign firms do not have the same connections as local firms, and bureaucrats may choose to exploit the ignorance of foreign owners about local customs and traditions.11

In addition to investigating low level corruption, several regressions were run to examine the effects of bureaucracy on high level corruption. Results proved to be largely inconclusive due to several factors. There were only two variables pertaining to high level corruption in the data set, asking whether firms had attempted to influence national or regional law. Less than 10% firms interviewed responded yes, with no indication of frequency of success. Regressions on these variables yielded no significant coefficients, and the overall regressions were often insignificant themselves. Only the dummy for small firms exhibited a significant, negative coefficient – a rather obvious result, as high level corruption is much more expensive and outside the scope of a small firm’s budget. Furthermore, it is unclear what, if any, effect size of bureaucracy should have on high level corruption. If anything, lower levels of bureaucracy should encourage a greater frequency of high level corruption, as each bureaucrat would have more power and potential influence over regional law, making high level corruption easier and perhaps cheaper. Still,



Conclusion: Contrary to common assumptions about bureaucracy, this paper finds that larger bureaucracies tend to be associated with less corruption than smaller bureaucracies. This may be because in areas with a larger pool of bureaucrats to choose from, it is easier to find bureaucrats who will perform their duties without demanding additional payments. Furthermore, bribes could be a choice by firms due to inefficient, understaffed bureaucracies. As a result, regions with more government officials may be able to provide public services more effectively, reducing the need for bribes to hasten the process. Improved oversight may play a role as well, but this paper found no apparent connection between increases in supervising officials and lower levels of corruption. Since Russia is considering halving the number of regions and significantly trimming the bureaucracy, it is important that the implications of such an action be considered. Reducing the size of bureaucracy has the potential to increase levels of corruption, perhaps due to easier collusion among bureaucrats, fewer alternatives for services, or reduced efficiency of government agencies. Thus, while bureaucrats have long been chastised for their corrupt practices, an ironic solution to the problem may lie in simply increasing the quantity of these loathed officials.

the connection is tenuous at best, and would warrant a more in-depth study than is possible with the available data.


Appendix 1: Variables and Statistics

Figure 1: Descriptions of variables

Variable Name


Mean (Std Dev)
1.06 (1.66)


size of total bribes as % of annual sales


frequency of bribes on a scale of 1 (never) to 6 (always) a binary variable, equal to 1 if a firm ever pays a bribe (bribefreq>1) a binary variable, equal to 1 if a firm often pays bribes (bribefreq>3) hundreds of bureaucrats in a region

3.18 (1.62) .790 (.407)



.392 (.488) 298.3 (121.7)



number of bureaucrats per captia in a region

6.56 (1.57) 90.2 (252.4 5184 (3263) 4754 (2461) 4969 (2859) 1292612 (1621599) 461522 (543729) 877.1 (1082)


size of a region in thousands of sqaure km


regional size of population in 2005, measured in thousands of inhabitants regional size of population in 2001, measured in thousands of inhabitants average regional population from 2001 to 2005 measured in thousands regional domestic product in 2005, measured in millions of rubles regional domestic product in 2001, measured in millions of rubles average regional domestic product from 2001 - 2005, billions of rubles








per capita monthly income in rubles, 2001

9899 (7070) 7389 (366)


average monthly bureaucratic salary in 2001. Across Russia, all bureaucrats receive a base salary of 4050 rubles/month, but receive "bonuses" that vary by region. This variable measures base salary plus any bonuses. difference between monthly per capita income and monthly bureaucratic salary in rubles, 2001 number of university students per 10,000 people


2509 (6904) 512.5 (210.4)



freedom of media on a scale of 1 (poor) to 5 (excellent) the number of employees in the firm in 2005

3.74 (1.01)


154.4 ( 589.8) 371209 (4341767) 465.3 (503.6) .416


number of employees in 2005 squared


linear distance (in miles) of administrative center of region to Moscow dummy equal to 1 if firm uses an external certified auditor to check books dummies for size of company based on number of employees. 2-49 is small, 50-249 medium, 250 and over is large. dummies representing origin of company. Privatized companies are former state owned enterprises. Private companies have no state history. Subsidaries are offshoots of former state owned enterprises. Foreign firms are joint venture private firms with a foreign partner


small medium large privatized private subsidary foreign otherorigin

.66 .22 .11 .11 .73 .01 .02 .10


getconnected permits govcontracts healthinspector fireinspector taxcollector customsimports courts mine construction manuf transport trade real_estate hotel_restaurant central northwest siberian southern volga urals

a series of variables indicating the frequency with which firms pay bribes for various services/departments, ranging from getting connected to basic infrastructure to avoiding inspections to paying off courts. Measured as binary variables, equal to 1 if a firm has ever paid a bribe for this service

dummies representing self-identified Sector firm works in. Ommited sector is "other"

dummies representing the administrative district in which each firm resides. Districts are the next level of organization above oblast.


Figure 2: Firm breakdown by region

Oblast Kemerovo Krasnoyarsk Kursk Leningradskaya o. Moscow12 Moscovskaya o. Nizhegorod Novosibirsk Rostov Samara St. Petersburg Sverdlovsk Udmurtia Voronezh Total

# Firms 7 7 25 32 157 8 58 42 61 45 52 50 10 47 601

Figure 3: Correlation coefficients between several highly correlated variables bureaucrats bureaucrats burpc popavg rdpavg diffwage unistudents 1.000 -0.702 0.971 0.899 0.858 0.785 burpc 1.000 -0.816 -0.776 -0.788 -0.717 popavg rdpavg diffwage unistudents

1.000 0.959 0.932 0.842

1.000 0.989 0.848

1.000 0.829



Since so many of the observations are concentrated in Moscow, regressions were double-checked removing Moscow firms. Results were typically unaffected by the exclusion of Moscow observations.


Figure 4: Distribution of observations by federal district Federal District Central # Observations 237

Regions Kursk Moscow Moscovskaya Voronezh Leningradskaya St. Petersburg Kemerovo Krasnoyarsk Novosibirsk Rostov Sverdlovsk Nizhegorod Udmurtia Samara





Southern Urals Volga

61 50 113

Figure 5: Percent of Executive branch bureaucrats in each tier of government by region

Oblast Kemerovo Krasnoyarsk Kursk Leningradskaya Moscow Moscovskaya Nizhegorod Novosibirsk Rostov Samara St. Petersburg Sverdlovsk Udmurtia Voronezh

Federal 36.7 38.2 31.9 32.6 38.7 33.9 38.2 45.0 39.9 37.2 50.9 38.6 39.8 38.2

Regional 10.3 10.0 32.9 4.7 50.2 4.4 6.0 13.2 4.9 9.5 42.5 21.8 12.6 10.9

Local 53.0 51.8 35.2 62.7 11.1 61.7 55.8 41.8 55.2 53.3 6.6 39.6 47.6 50.9


Appendix 2: Alternative Models Ordered Probit: In addition to the probit models run on bribe frequency, an ordered probit model was estimated as well. The ordered probit allows for changes in the effects of independent variables depending on the level of the dependent variable. In the context of this paper, it means bureaucracy may play a different role when moving from “never paying bribes” to “seldom paying bribes”, versus from “usually paying bribes” to “always paying bribes”. Furthermore, it allows for differences in categorical sizes that may be lost through a binary probit model. The results of the regression, presented below in Table 7, represent the overall net effect of each independent variable on bribe frequency. Bureaucracy continues to have a statistically significant negative impact on the frequency of bribes, matching results from the probit models. Table 8 shows the marginal effects of bureaucracy at each level of bribe frequency. The positive, significant coefficients on bureaucracy at low levels of bribe frequency show that higher numbers of bureaucrats increases the probability of paying no or very few bribes. The negative coefficients at high levels of bribe frequency show that increases in the size of bureaucracy drive down the probability of paying frequent bribes. This demonstrates that the findings of this paper hold under the ordered probit, as increases in bureaucracy lead to a higher probability of paying low or no bribes.


Table 7 (below): Ordered probit model, using bribefreq as dependent variable

bureaucrats popavg territory rdpavg diffwage unistudents privatized private foreign subsidary small medium mine construction manuf transport trade real_estate hotel_restaur central northwest


siberian southern volga distmosc Pseudo R2 N







(0.0018) 0.0233 548




Table 8 (below): Marginal effects of bureaucracy at each level of bribe frequency



Level of bribe frequency 1 (never)

Marginal effect of bureaucray 0.0035




2 (seldom)




3 (sometimes)



4 (frequently)




5 (usually)



6 (always)






Instrumental Variable Probit analysis: Endogeneity is always a concern when studying corruption and bureaucracy. As this paper discussed earlier, there are several ways by which the level of bureaucracy could be affected by the amount of corruption. In some cases, the higher potential for corruption in a region could attract bureaucrats seeking to profit from bribes. At the same time, corruption that is too high could force firms out of the region, reducing regional tax levels and budgets for employing bureaucrats. Since it is unclear which, if either, of these effects is at work, it is difficult to ascertain if endogeneity is really a problem, as it is even possible the two effects cancel each other out. Nevertheless, to confirm results, an instrumental variable probit analysis was run as well. The available data included information on the number of federal employees by region back to 1995. In 2001, the number of federal employees proved to be an effective predictor of the overall size of bureaucracy, exhibiting a correlation coefficient of .97. The number of federal employees in 1995 also effectively predicts size of bureaucracy in 2001, with a slightly lower correlation coefficient of .90. This suggests that the sizes of regional bureaucracies change at similar rates, and the ratio of federal employees to the overall pool of bureaucrats remains relatively constant. At the same time, it is difficult to explain how the number of federal level bureaucrats in 1995 would affect the amount of corruption in 2001, suggesting that the number of federal bureaucrats in 1995 could serve as an effective instrument. Results of the regression in Table 9 below show that even after instrumenting for potential endogeneity, bureaucracy has a negative effect on bribe frequency. These results hold using both the maximum likelihood estimator and Newey’s two-step


estimator. Instrumental variable analysis on bribe size reveals no significant coefficients, but this was often the result even without instrumentation.
Table 9: Instrumental variable probit model using oftenbribe as dependent variable, instrumenting for bureaucracy using federal employees from 1995 (continued) central northwest siberian southern volga distmosc constant N -2.1335 (1.7591) -3.2111 (2.2498) 1.6525 (1.5527) -0.6027 (0.6529) -2.0247 (1.6149) -0.0026 (0.0023) 4.3187 (3.3242) 548

bureaucrats popavg territory rdpavg diffwage unistudents privatized private foreign subsidary small medium mine construction manuf transport trade real_estate hotel_restaur

-0.0191 (0.0069)*** 0.0009 (0.0003)*** 0.0004 (0.0005) -0.0024 (0.0025) 0.0003 (0.0003) 0.0020 (0.0017) 0.4489 (0.2468)* 0.3236 (0.2040) 0.4896 (0.3948) 0.2493 (0.4387) -0.3905 (0.199)** -0.1585 (0.2059) -0.1929 (0.5168) 0.2482 (0.2375) -0.0843 (0.2215) 0.1235 (0.2925) 0.0257 (0.2150) 0.0497 (0.2600) 0.0227 (0.3423)


References: Acemoglu, Daron, James Robinson & Simon Johnson. 2001. “The Colonial Origins of Comparative Development: An Empirical Investigation.” American Economic Review. December 91, pp. 1369-401 Ades, Albero and Rafael Di Tella. 1999. “Rents, Competition and Corruption.” American Economic Review, 89:4, pp. 982-93 Ahlin, Christian, & Pinaki Bose. 2007. “Bribery, Inefficiency, and Bureaucratic Delay.” Journal of Development Economics. v84, pp. 465-486 Alt, James & David Dreyer Lassen. 2003. “The Political Economy of Institutions and Corruption in American States.” Journal of Theoretical Politics. 15:3, pp. 341-365 Anderson, James & Cheryl Gray. 2006. “Anticorruption in Transition 3: Who is Succeeding and Why?” World Bank. Bardhan, Pranab & Dilip Mookherjee. 2005. “Decentralization, Corruption And Government Accountability: An Overview.” Handbook of Economic Corruption. BBC News Network. 2003. “Putin Ally Leads Petersburg Poll.” Available Brown, David, John Earle & Scott Gehlbach. 2007. “Helping Hand or Grabbing Hand? State Bureaucracy and Privatization Effectiveness.” Di Tella, Rafael & Ernesto Schargrodsky. 2003. “The Role of Wages and Auditing during a Crackdown on Corruption in the City of Buenos Aires.” Journal of Law and Economics. 46:1, pp. 269-92 Dininio, Phyllis & Robert Orttung. 2005. “Explaining Patterns of Corruption in the Russian Regions.” World Politics. Vol. 57, pp. 500-529 Dreher, Axel, Christos Kotsogiannis & Steve McCorriston. 2007. “Corruption Around the World: Evidence from a Structural Model.” Journal of Comparative Economics. v35, pp. 443-466 Francken, Nathalie, Bart Minten & Johan Swinnen. 2005. “Listen to the Radio! Media and Corruption: Evidence from Madagascar.” LICOS discussion paper No. 155/2005 Granovetter, Mark. 2004. “The Social Construction of Corruption.” Standford University Gorodnichenko, Yuriy, & Klara Peter. 2007. “Public Sector Pay and Corruption: Measuring Bribery from Micro Data.” Journal of Public Economics. v91, pp. 963-991


Hendley, Kathryn. 1998. “Remaking an Institution: The Transition in Russia from State Arbitrazh to Arbitrazh Courts.” The American Journal of Comparative Law. 46:1, pp. 93127 Hill, Fiona & Clifford Gaddy. 2003. Siberian Curse: How Communist Planners Left Russia Out in the Cold. Washington DC: Brookings Institute Press. Hoff, Karla, Shale Horowitz & Branko Milanovic. 2005. “Political Alteration, Regardless of Ideology, Diminishes Influence Buying.” Carnegie Endowment for International Peace. Hotelling, Harold. 1929. “Stability in Competition.” The Economic Journal. 39:153, pp. 41-57 La Porta, Rafael, Florencio Lopez-de-Silanes, Andrei Schleifer & Robert Vishny. 1998. “Law and Finance.” Journal of Political Economy. 106, pp. 1113-155 Larina, Ekaterina. 2001. “Bureaucracy Faces the Knife – and Many Questions.” The Russia Journal. 7:13 Leite, Carlos & Jens Weidmann. 1999. “Does Mother Nature Corrupt? Natural Resources, Corruption and Economic Growth.” IMF Working Paper. Leonhardt, M. 1998. “Report on Fieldwork in Derbent, Dagestan, November 1997”, unpublished MS. Levin, Mark, & Gregory Satarov. 2000. “Corruption and Institutions in Russia.” European Journal of Political Economy. v16, pp. 113-132. Lujala, Paivi, Nils Gleditsch & Elisabeth Gilmore. 2005. “A Diamond Curse?” Journal of Conflict Resolution. 49:4, pp. 538-562 Mauro, Paolo. 1998. “Corruption: Causes, Consequences and Agenda for Further Research.” Finance and Development Mocan, Naci. 2004. “What Determines Corruption? International Evidence from Micro Data.” NBER Working Paper #10460 Perloff, Jeffrey & Steven Salop. 1986. “Firm-Specific Information, Product Differentiation, and Industry Equilibrium.” Oxford Economic Papers. v38, pp. 184-202 Prasch, Robert. 2007. “The Economics of Fraud.” Middlebury College Razafindrakoto, M & F. Roubaud. 2005. “How far can we trust expert opinions of corruption? An experiment based on surveys in francophone Africa.” Global Corruption Report 2005, pp. 292-5


Rose-Ackerman, Susan. 1996. “The Political Economy of Corruption.” Institute For International Economics. Schleifer, Andrei & Robert Vishny. 1993. “Corruption.” The Quarterly Journal of Economics. 108:3, pp. 599-617 Slinko, I, E. Yakovlev & . Zhuravskaya. 2005. “Laws for Sale: Evidence from Russia.” American Law and Economics Review. 7:1, pp. 284-318 Svenson, Jakob. 2003. “Who Must Pay Bribes and How Much?” Quarterly Journal of Economics. 118:1, pp. 207-30 Svenson, Jakob. 2005. “Eight Questions about Corruption.” Journal of Economic Perspectives. 19:3, pp. 19-42 Treisman, Daniel. 2000. “Decentralization and the Quality of Government.” IMF. Treisman, Daniel. 2006. “What have we learned about the causes of corruption from ten years of crossnational empirical research?” Annual Review of Political Science. Vol. 10, pp. 211-244 Waller, C., T. Verdier & R. Gardner. 2002. “Corruption: Top Down or Bottom Up?” Economic Inquiry. 40:4, pp. 688-703 Yang, David. 2005. “Corruption by Monopoly: Bribery in Chinese Enterprise Licensing as a Repeated Bargaining Game.” China Economic Review. v16, pp. 171-188