Political Corruption, Punishment, and Institutions in American States

Transcription

1 Political Corruption, Punishment, and Institutions in American States Linjing Wang Abstract Institutions have dominated explanations of political corruption. For instance, political and judicial institutions play an important role in explaining corruption in the American states. This empirical study focuses on the effects of judicial institutions upon corruption convictions in 1997, 2002, and The results indicate that the strength of law enforcement is positively associated with corruption convictions in the states when controlling the number of public employees; and that surprisingly, severity of punishment is positively associated with corruption convictions and takes away the explaining effect of the strength of law enforcement. The coefficient is statistically significant and consistent in all years. Contrary to the prediction deduced from the Becker s (1968) theory that severe punishment can compensate a lower probability, this empirical finding suggests states with severe punishment have more corruption convictions. Index Terms Corruption, institutions, punishment, United States I I. INTRODUCTION nstitutions have emerged as major explanations of political corruption in recent empirical research [1] [2] [3]. In the United States, heterogeneity in state institutions also accounts for variance in corruption across the states [4] [5] [6]. The size of the government, strength of law enforcement, socioeconomic variables such as higher education percentage, urban population, and state political institutions such as initiatives, progressive governors are significantly associated with corruption in states. This article follows institutional explanation to explore the underlying causes of different corruption conviction rates across states. Among all the institutional factors, government anti-corruption efforts have relied more on law enforcement since the 1970s. Modern theoretic work on law enforcement dates back to Becker s [7] paper on crime and punishment, where he argued that an increased probability of conviction absorbs public and private resources and that it can be compensated for through severe punishment in forms of imprisonment and fines. Game-theoretic researchers [8][9]simulated the strategic relations among law enforcers, criminals, and policy makers based on the simple model developed by Becker and Stigler [10], and found that crime is unaffected by the magnitude of the penalty. Friehe s [11] simulation of corruption even produced a counterintuitive result that harsher anti-corruption measures can increase crime incentives. My study found that states with severe punishment had higher rates of corruption conviction, ceteris paribus. The rest of the paper is structured as follows. In section 2, anti-corruption efforts at the federal level are described. Section 3 details the heterogeneity of state institutions in previous literature. Section 4 presents the theory and model of corruption conviction. In section 5, empirical data and results are discussed. Section 6 offers discussion and concluding remarks. II. FEDERAL ANTI-CORRUPTION EFFORTS Anechiarico and Jacobs [12] traced the history of efforts to control corrupt behavior by public officials in the United States. They categorized four phases of corruption control during , , , and 1970-present as anti-patronage, progressive (professionalism), scientific management, and panoptic vision. The panoptic vision, currently predominating in large governments in the United States, assumes that officials will succumb to corrupt opportunities and advocates comprehensive surveillance, investigation, and target-hardening strategies (pp. 466). It is no longer primarily attributed to incompetence, absenteeism, laziness or partisan influence, but to inadequate rules, threats, and controls. Law enforcement is hence heavily emphasized. Watergate scandals and the fiscal crisis of the mid-1970s are the key factors that triggered anti-corruption efforts and added fiscal accountability to the purview of administrators. The panoptic vision has led to the expansion of anti-corruption institutions and strategies and to enhancement of the authority of anti-corruption units and personnel [12] (pp. 468). Measures adopted include corruption vulnerability assessment 1, auditing, and internal surveillance. New anti-corruption laws have been established and new governmental institutions have been created since the late 1970s. For instance, the Ethics in Government Act was passed in 1978, and the Public Integrity Section (PIN) was created in 1976 to consolidate the Justice Department s oversight responsibilities for the prosecution of criminal abuses of the public trust by government officials. As the vast majority of federal corruption prosecutions are handled by the local United 1 Corruption vulnerability assessment measures the potential for corruption inherent in an agency s operations and organization. For example, the New York City Comptroller s Office requires agency heads to conduct a review of the adequacy of internal controls annually. Based on these results, specific agencies and operations were selected for intensive audits, which are more costly and time consuming. (see [12], pp ) DOI: _ _ GSTF

2 States Attorney s Office for the geographic district where the crime occurred, the PIN facilitates prosecution as a result of recusals, sensitive and multi-district cases, federal agency referral, or special request within the Justice Department. PIN reports annually to the Congress. In accordance with the Report to Congress on the Activities and Operations of the Public Integrity Section for 2009, an Election Crimes Branch was created within PIN in 1980 to supervise the Department s nationwide response to election crimes, such as voter fraud and campaign financing. Hence, the categories of corruption expanded beyond bribe, graft, and fraud, and now include election crimes and conflicts of interests. With each act of corruption exposed, the corruption- hunting apparatus lobbies for both greater resources and a broader definition of its mission. The inevitable consequence is that more corruption is uncovered [12] (pp.468). III. HETEROGENEITY OF STATE INSTITUTIONS The legal environment in the US is often assumed to be homogeneous, but the pattern of the corruption convictions shows variance among the states. New York and California had total over 200 cases while Vermont had only 1 case in the years 1997, 2002 and The difference is partly attributed to the size of government. Dividing the cases number each year and standardizing them per 100,000 employees, the pattern of corruption convictions changes dramatically. For example, Alaska, Louisiana, and Alabama are the most corrupted states, while New York and California only ranked No.17 and No. 30. Four main explanations of corruption in the American states are found in the literature. First, historical and cultural explanations of corruption rely on the socioeconomic and demographic characteristics of states. For example, urban political machines facilitated the rise of immigrants to power in the United States [13]. The unemployment rate is also negatively related to corruption [5]. Second, the size of the state and the state government plays an important role: the larger the government, more expenditure, more personnel, and the greater the potential for state corruption. Third, political institutions produce mixed results. Meier and Hilbrook [4] found that great voter turnout and party competition reduced corruption while the percentage of elected officials being black led to more corruption convictions due to racial targeting (pp. 150). No statistically significant effects of campaign report filings, state centralization, initiative requirements, or referendum requirements were found, except that single purpose special districts (water districts, public corporations) seem related to corruption. Fourth, the state institutional effects of law enforcement played an important role. The seminal articles on crime, punishment, and law enforcers [7] [10] have revived the interest in the importance of law enforcement and the substituting effects of punishment for the probability of captures. However, empirical studies provide mixed results of the effect of judicial resources, including US Attorney, federal judges, cases backlogged, police expenditures, state s attorneys, FBI, other justice department employment, and the severity of punishment[4][5]. IV. THEORY AND MODEL OF CORRUPTION CONVICTION 1. Corruption conviction reflects the probability of conviction as well as the corruptness of the government. Equation (1) shows the rate of corruption conviction is related to three variables: the probability of capture and conviction (P 1 ), the percentage of public employees being corrupt (P 2 ), and the size of the public employee body (E). These three variables are all positively related to the corruption convictions of states. That is, when holding other two variables constant, larger state, more corrupt public employees and higher conviction rates will yield more corruption cases. C = f ( P 1, P 2, E ) (1) Where, C The rate of corruption convictions P 1 The probability of capture and conviction P 2 The percentage of public employees being corrupt E Total number of public employees And 2. The operationalization of the variables P 1 and P 2 can be found in the explanations of corruption discussed in section III. The strength of law enforcement, determined by the available social resources allocated, sets the probability of capture and conviction P 1. Since most societies cannot tolerate corruption, corrupt officials are prosecuted and thus P 1 >0. However, similar to other criminal prosecution, where an increased probability of conviction absorbs public and private resources [7] (pp184), corruption prosecution and conviction takes social resources, such as policemen, judges, juries, etc. Hence, it is not practical to expect the capture and conviction of all corrupt officials, thus P 1 <1. P 1= f (law enforcement strength) (2) =f (judicial personnel, severity of punishment, and reward to enforcer) (3) As (2) and (3) indicate, the probability of capture and conviction is determined by the available social resources as well as set by the judicial system. First, states with larger gross domestic product can hire more and better skilled law enforcers and offer them higher salaries. Better judicial resources lead to higher probability of capture and detection P 1. Second, Becker[7] noted the tendency during the eighteenth and nineteenth centuries in Anglo-Saxon countries, and even today in many Communist and underdeveloped countries, to punish those convicted of criminal offences rather severely, at the same time that the probability of capture and conviction was set at rather low values (pp.184). Low probability saves social resources yet is not sufficient to deter corruption alone, severe punishment is used to compensate for the low probability of capture. When applying this theory, if we observe variances in the severity of punishment in the American states, we should be able to deduce conversely the GSTF

3 states relative probabilities of capture. A state with relatively severe punishment should be associated with low probability of capture (P 1 ) and also with greater deterrence which leads to low percentages of public employees being corrupt (P 2 ). Therefore, when we assume the same number of public employees working in all states, states with severe punishment should have much less corruption convictions. Simply put, severe punishment should be negatively associated with corruption convictions, according to the theory. 3. P 2 measures the percentage of public employees being corrupt. The value of P 2 is between 1 and 0; that is, it is false to say that all public officials are corrupt or that none of them is corrupt. When corrupt opportunities are present, individual public officials act differently: some would still remain clean and some couldn t resist temptation. P 2= f (socioeconomic conditions, government salary level, state political institutions) (4) The percentage of public employees being corrupt (P 2 ) also varies with socioeconomic conditions, their salary level, and the competitiveness of state political institutions, shown in (4). Socioeconomic conditions, such as inequality in distribution of wealth [14] (pp.418) and lower government salaries [5] (pp. 113), make bribe-taking a more attractive way to accumulate wealth. Higher salaries lessen the need for corruption [4](pp.146). Hence, P 2 decreases as inequality improves and salary level rises. However, when the state political environment is relatively competitive and more uncertainty arises over their jobs, public officials are thus less likely to be corrupt. Term limits produces two way effects on corruption. On the one hand, as the grassroots movement Term Limits 2 advocates, term limits force out career politicians who are more concerned with their own gain than the interests of the American people We want a government of the people, by the people, and for the people, not a tyrannical ruling class who care more about deals to benefit themselves, than their constituents. Thus term limits can check the corrupt behavior of public officials. On the other hand, as Alt and Lassen [6] (pp349) argue, if the incumbent was too sure of continuing in office, corruption would be high. The converse can also be true. If the incumbent is almost certain not to have a next period nothing is lost in terms of re-election possibilities by increasing rents, or not combating corruption. In states with term limits law, public officials in their last terms fit into both scenarios. In this sense, corruption levels should be high in these states. A. Variables and Data V. EMPIRICAL ANALYSIS The dependent variable CORRUPTION is the aggregated raw data for each state from the district data reported by US Attorneys, which are in the Public Integrity Section annual reports to Congress. The reported convictions include cases of 2 See federal, state, local public employees as well as private citizens 3 involved in the cases. In contrast to previous studies, which use perception score or standardized case number as dependent variable, I use raw data on corruption cases in each state and apply a simple OLS model. This is because I control for the numbers of public employees in the states. The Census Bureau conducted surveys in 1997, 2002, and These surveys calculate the total state and local government employees and their total income in March of respective years. These are reflected in the STATELOC and PUBPAY variables. INEQUALITY is the Gini coefficient of states, with value between 0 and 1. Larger values reflect more inequality in that state. The inequality levels of the states vary from year to year. The data on 1997, 2002 and 2007 state inequality levels are used in the three year data sets respectively. URBANPOP is the percentage of urban population in the states in EDUCATION measures the percentage of people 25 years and over who have completed a Bachelor's degree from American Community Survey 5-year estimates. For the LAWENFORCE variable, using the number of law enforcers, such as US Attorneys, judges, FBI budget or personnel would violate the mean independence assumption of OLS model which means when we try to analyze the effect of law enforcement upon corruption, the direction of causality in reality could be reverse, i.e. uncovered severer corruption calls for strengthening the law enforcement. Simply put, more corruption cases in the past result in more judicial resources allocated to that area in the present and future. To solve this problem, an ideal instrumental variable is the number of judicial districts in each state. Since there is one US Attorney and one US district court in each judicial district, which was initially established in the Judiciary Act of 1789 or established when the states joined the Union (current breakdown of the districts in states are at 28U.S.C ), the number of districts in a state is uncorrelated with the severity of corruption situation today and thus exogenous to the dependent variable. At the same time, it is highly representative of the strength of the law enforcement. 4 LEGLIMIT is a dummy variable which measures the legislative term limit in each state. The values for this variable changes from 1997 to 2002 to 2007, as the laws for term limits in states starting to become effective during this period. The changing nature of this variable provides more explanatory power than the gubernatorial term limit used in previous study, if term limits really affect corruption. 5 SENTENCE variable measures the mean maximum sentence by state. In order to test the effect of severity of sentence upon corruption, lagged coding is used specifically, 1996 sentencing records are used for the 1997 data set and Data on private individual convictions by state are not available to the author. As these private individual cases involve corrupt public officials, the impact of this inclusion should be small. 4 For the 2011 data, the number of US Attorney s offices and total key personnel in a state are correlated with the number of judicial districts in that state by and source: 5 This variable has more variance than gubernatorial term limits, which, however, produces similar insignificant results in Model GSTF

4 records are for the 2002 and 2007 data sets 6. TIS is a dummy variable capturing states adopting Federal Truth-in-Sentence standard by year-end This standard requires that violent offenders 7 serve not less than 85% of their prison sentence before becoming eligible for release. TIS standard makes punishment more substantial to criminals. These two measures did not directly apply to corruption cases but are good proxies for the purpose of comparing the severity of punishment across the states. Southern states are commonly regarded as having harsher punishment even for relatively minor crimes. The coding of SOUTH dummy is adopted from V. O.Key [15]. B. Models for empirical analysis OLS models are used for the analysis: Model 1: CORRUPTION=α 1 STATELOC+α 2 LAWENFORCE+ α 3 INEQUALITY+α 4 PUBPAY + ε Model 2: CORRUPTION=β 1 STATELOC+β 2 LAWENFORCE+ β 3 INEQUALITY+β 4 PUBPAY+β 5 LEGLIMIT +β 6 EDUCATION +β 7 URBANPOP + ε Model 3: CORRUPTION=γ 1 STATELOC+γ 2 LAWENFORCE+ γ 3 INEQUALITY+γ 4 PUBPAY +γ 5 SENTENCE +γ 6 TIS +γ 7 SOUTH+ ε To control for the potential heteroskedastisity, robust standard errors are used in all models when calculating t-values[6][16](pp ). C. Results 1) Table 1 presents the results of Model 1 and Model 2 for the year 1999, 2002, and The results confirm the theoretical prediction that institutional characteristics of a state do matter. The numbers of state and local public employees (STATELOC) are positively associated with the number of corruption convictions via positively affecting E in equation (1). The strength of law enforcement (LAWENFORCE) is also positively associated with the number of convictions since the stronger the law enforcement the higher probability of capture and conviction P 1. However, the magnitude of its effect increased in 2002 but decreased significantly in TABLE I. THE OLS MODELS OF CORRUPTION CONVICTIONS 6 Each year, certain states didn t report the sentencing data. These missing data causes fewer observations in respective data sets. The same applies to the variable TIS. 7 Violent offenders are defined as the Federal Bureau of Investigation's Uniform Crime Reporting Program offenses of murder and nonnegligent manslaughter, rape, robbery, and aggravated assault. The result of truth-in-sentencing practices has been to reduce discretionary release of offenders by parole boards in favor of mandatory release according to statutory provisions. Similarly to the results of gubernatorial term limit in previous studies, the state legislative term limit (LEGLIMIT) shows mixed effects on corruption convictions and not statistically significant. This could be due to the fact that state legislators have incentives to maintain their reputations for the post-legislator employment, such as seeking higher level government positions or consultant jobs in the private sector. Socioeconomic conditions have strong effects on corruption. Where more inequality exists, P 2 rises, and thus more corruption occurs. However, education and urban population have no significant relationship with corruption convictions as expected. URBANPOP has the correct sign in all three models while the EDUCATION variable has mixed effects. When these two variables enter in Model 2, both the coefficient and significance level of the INEQUALITY variable declined. This means that EDUCATION and URBANPOP take away some explanation power from INEQUALITY. At the same time, the results suggest that inequality level of a state reflects the impacts of educations and the percentage of urban population upon corruption. Moreover, the significance sign of INEQUALITY means that it captures more socioeconomic impact on corruption than EDUCATION and URBANPOP do. Hence, it is a better variable. In model 3, EDUCATION and URBANPOP are dropped and only INEQUALITY is included to measure the effects of socioeconomic conditions. The payment level (PUBPAY) is not significant in any model. The coefficient suggests that contrary to the theory s predictions, higher salary levels are positively associated with higher corruption convictions. This does not confirm the theoretic prediction that higher salaries impose a cost on violations to trust (Becker and Stigler 1974, pp.12). The discrepancy may be a result of the magnitude of the salaries, which has not reached a commanding level. 2) Model 3 presents a very different picture about the GSTF

5 effects of judicial institutions, shown in Table 2. The R-square values are larger than both model 1 and 2 for all years. In this model, STATELOC remains significant and similar magnitude. However, the LAWENFORCE and INEQUALITY variables gradually lose their explanation power, in terms of significance level and coefficient. Southern states (SOUTH) make a difference in the 1997 dataset but loses their impact later. The Truth-in-Sentence standard (TIS) gradually showed its impact. When a state adopted the Truth-in-Sentence standard, on average about 7 more convictions are observed. TABLE II. THE OLS MODEL OF CORRUPTION CONVICTIONS (MODEL 3) The prominent explanatory variable is SENTENCE, which plays an increasingly significant role as time goes on. More surprisingly, the coefficient is positive. 8 As previously mentioned, this variable is lagged to prevent reverse causality, i.e. harsher sentence results from more convictions. The coefficients indicate that when controlling for other variables, a state with 10 more months of sentence in 1996 had 0.8 more corruption conviction cases in 1997 on average; 10 more months of sentence level in1999 resulted in 1.45 more cases in 2002 and 1.52 cases in As the means of CORRUPTION in the three years are 16.7, 16.9, and 19.7 respectively, merely increasing the sentence level by 10 more months (about 10% of the mean sentencing) leads to the conviction rate rising by 4.8%, 8.6%, and 7.7% in 1997, 2002, and In other words, severe punishment results in more corruption convictions! This is contradicts the theories predictions: severe punishment is to compensate for low probability of conviction in order to deter the criminals and thus is expected to be associated with low 8 Goel and Nelson(1998) also found this positive and statistically significant coefficient for the data. They dismissed it as spurious correlation with the dependent variable (pp.115). conviction rates. VI. DISCUSSION As corruption conviction indicates the true level of corruption, the findings shows a counterintuitive relationship that harsher punishments results in more corruption. How could a state with severe sentence fail deterring corruption? This can also be observed in country level cases. Taking to the extremes, death penalty and life sentence is rare in relatively clean countries such as the United States and Scandinavian countries, but relatively commonly used in corrupt countries such as China. What is the explanation? Let s go back to the formula (1) in section IV. Convictions are positively associated with the probability of conviction P 1 as well as the percentage of public employees being corrupt P 2. Harsher punishment is designed to compensate a low probability of conviction, which is often unfortunately set by the available resources in a society. Less social resources mean smaller budget for law enforcement, reduced size of police and surveillance forces poorly equipped, insufficient prison and jail facilities, early discharge of prisoners, etc. Meanwhile, the information is not confidential and potential corrupt public officials can gauge the probability of convictions being small. Thus, an increased P 2 is an inverse result of the harsher punishment. More potential corrupt public official are hence endeavored to take the chance even when the punishment is harsher. 9 A simple example about shoplift will help to illustrate the logic. Shoplift is similar to corruption in the sense that potential criminals are allowed to gather information and gauge the probability of getting caught before taking action. Punishment for supermarket shoplifts is often very heavy, even with store surveillance, security systems and personnel adding on the deterrence. For example, on the door of a supermarket in the United States says [The supermarket name] will prosecute shoplifters to the fullest extent of the law. On the contrary, no roadside ice cream store has similar announce on the door. The difference is simply that even without monitoring system the latter has 100% probability of conviction no pay, no ice cream. The potential shoplifters make decision upon their estimation of the probability of conviction. When presented with corruption opportunities, decisions of taking bribe or not, for example, are affected by the public employees subjective calculation of being caught and convicted. The degree of how harsh a punishment is in a state affects the subjective probability of being caught, inversely. As a result, harsher punishment encourages potential criminals taking chance to violate law. VII. CONCLUDING REMARKS This empirical study confirms that institutional factors play important role in corruption convictions in the American 9 One might argue that there are people who never become corrupt under any circumstances due to cultural, religion, or other reasons. As corruption exists in all countries only with various degree of seriousness, this paper focuses on those who succumb to corruption under circumstances GSTF

6 states. Reflecting the emphasis on law enforcement as measures to control corruption, however, the results fail to confirm the prediction that severe punishment deters crime and compensates for the low probability of capture and conviction. When controlling for other variables, states with severe punishment convicted more corrupt officials in 1997, 2002, and This requires more empirical analysis to untangle the underlying mechanism between the magnitude of punishment and the rate of corruption conviction. The explanation of this anomaly probably relies on the mixed effect of punishment on corruption convictions through the probability of capture (P 1 ) and the percentages of public employees being corrupt (P 2 ). On the one hand, severe punishment deter public employees becoming corrupt; on the other hand, severe punishment is designed to compensate low probability of conviction, which encourages P 2, that is, small chance of being convicted encourages public employees to take chance and commit corruption. Therefore, punishment is negatively associated with the probability of capture and positively with the percentages of public employees being corrupt. Punishment demonstrates a greater encouraging effect than the deterrence effect to potential corrupt public employees. As a result, what we observe is that states with more punishment have more corruption convictions. [9] L. Andreozzi. "Rewarding Policemen Increases Crime. Another Surprising Result from the Inspection Game." Public Choice, vol. 121, no.1, pp , [10] G. Becker and G. Stigler. Law Enforcement, Malfeasance, and Compensation of Enforcers. The Journal of Legal Studies, vol. 3, no.1, pp.1-18, [11] T. Friehe, Correlated Payoffs in the Inspection Game: Some Theory and An Application to Corruption. Public Choice, vol. 137, pp , [12] F. Anechiarico and J. B. Jacobs. "Visions of Corruption Control and the Evolution of American Public-Administration." Public Administration Review, vol.54, no.5, pp , [13] F. I. Greenstein,"The Changing Pattern of Urban Party Politics." Annals of the American Academy of Political and Social Science, vol. 353, pp.1-13, [14] Nye, Joseph S. Corruption and Political Development: A Cost-Benefit Analysis. American Political Science Review, vol.61, pp , [15] V. O. Key, Southern Politics in State and Nation. Vintage Press, [16] W. H. Greene, Econometric Analysis. 6th ed. Prentice Hall, 2007 ACKNOWLEDGMENT I am grateful to William McLauchlan, Glenn R. Parker, Bert A Rockman, Samuel Standaert, Harry Z.G. Tzimitras, and Eric Waltenburg for comments and suggestions. Insightful comments of the reviewer are also appreciated. All remaining errors are my own. This article was originally prepared for presentation at the Annual International Conference on Political Science, Sociology and International Relations, Singapore, November Linjing Wang was born in Heilongjiang, China in She earned a B.A. in Economics at University of International Business and Economics (2000), Beijing, China, an M.A. in Political Science (2005) and an M.S. in Economics (2010) at Purdue University, West Lafayette, Indiana, USA. She is currently pursuing her Ph.D. in Political Science at Purdue University She became an Assistant Professor in 2007 and is now an Associate Professor and director of Bilingual Education Department at the college of Economics and Management of Inner Mongolia Agricultural University in Hohhot, Inner Mongolia, China. Her current research interests include political economy, institutions, and American politics. REFERENCES [1] M. Manion, Corruption by Design: Building Clean Government in Mainland China and Hong Kong. Harvard University Press, [2] R. Karklins, The System Made Me Do It: Corruption in Post- Communist Societies. M.E. Sharp, [3] J. Lambsdorff, Institutional Economics Corruption Reform: Theory, Evidence and Policy. Cambridge University Press, [4] K. Meier and T. Hilbrook, I Seen My Opportunities and I Took 'Em: Political Corruption in the American States. The Journal of Politics, vol.54, pp , [5] R. Goel and M. Nelson. Corruption and Government Size: A Disaggregated Analysis. Public Choice, vol. 92(1), pp , [6] J. Alt and D. Lassen. The Political Economy of Institutions and Corruption in American States. Journal of Theoretical Politics, vol. 15, no.3, pp , [7] G. Becker, "Crime and Punishment: An Economic Approach." The Journal of Political Economy, vol.76, pp , [8] G. Tsebelis. The Abuse of Probability in Political Analysis: The Robinson Crusoe Fallacy. American Political Science Review, vol.83, pp , GSTF