unitsystem

MISSISSIPPI WEB SITE

STEVEN A. McCALEB
103 ALVERADO DRIVE
LONG BEACH, MISSISSIPPI 39560
PHONE & FAX: (228)-868-8428

E-MAIL: mccaleb4thdist@aol.com

WEB SITE: www.mississippiwebsite.com

No Pretense to Honesty: County Government Corruption in Mississippi*

GÖKHAN R. KARAHAN†, LAURA RAZZOLINI‡ AND WILLIAM F. SHUGHART II‡
†Department of Economics and Finance, Nicholls State University, Thibodaux, LA 70310 USA
‡Department of Economics, University of Mississippi, P. O. Box 1848, University, MS 38677-1848 USA
May 2003
Please do not quote without permission
ABSTRACT
This paper explores the determinants of government corruption exploiting a unique dataset generated by an FBI investigation of county purchasing activities that ultimately led to the conviction of 55 of Mississippi’s 410 county supervisors, one county road foreman, two state highway commissioners and 13 vendors on bribery, extortion and other felony charges. Evidence is reported that corruption occurs more frequently in rural, less densely populated counties where voter-taxpayers have fewer years of schooling. Corruption is also more likely in counties where supervisors are paid more, ceteris paribus, casting doubt on the proposition that efficiency wages purchase honest public officials.
JEL Keywords: corruption; rent seeking; county governments
* We are grateful to James Crockett for generously granting us early access to extracts from his since-published book manuscript. We benefited from discussions with Robert Tollison and from the comments of Hilary Shughart, attendees at the 2002 meetings of the Southern Economic Association and the 2003 meetings of the Public Choice Society. Among the latter group, especially helpful suggestions were offered by Lee Coppock, Arthur Denzau, Fred McChesney, Michael Munger and Paul Pecorino. As is customary, however, we take full responsibility for any errors herein.
No Pretense to Honesty: County Government Corruption in Mississippi
Only if one looks at the fine structure of political and economic systems can one go beyond a showing that corruption is harmful to an understanding of the way it operates in different contexts…. We need more systematic knowledge of just how corruption and self-dealing affect the operation of government programs and private markets. (Rose-Ackerman, 1999: 4)
I. Introduction
The State of Mississippi has two different county government systems. The underlying distinction between them turns on the extent to which legislative and executive powers are combined at the county supervisor level. Each of Mississippi’s 82 counties is managed by five supervisors, who are elected to concurrent four-year terms and are responsible for administering all county governmental functions except the public schools.1 Owing to this division of public policymaking labor, as a practical matter Mississippi’s boards of supervisors spend most of their time and budgetary resources building and maintaining county roads and bridges.
Under the so-called unit system, many of the routine activities of county government, including purchasing, personnel policies, and inventory control and storage, are centralized. In carrying out their responsibilities with respect to transportation infrastructure, supervisors serve primarily in a policymaking capacity, collectively establishing priorities by simple majority rule and then delegating authority for executing their decisions to a hired professional road manager who supervises the day-to-day activities of county road crews. Under the so-called beat system, on the other hand, every county is partitioned geographically into five separate districts or “beats”. After dividing the county’s road budget either into equal shares or on the basis of existing road mileages, each supervisor is independently responsible for setting and executing policies within his or her own beat. The more decentralized beat system grants individual
1 Mississippi’s county public school systems fall under the separate administrative jurisdiction of elected county school boards. -1-
2
supervisors more direct control over the allocation of county funds and county resources such as labor and capital equipment.
In August 1988, after months of debate about whether to force each county to switch from the beat system to the unit system or to allow voters to petition to have the beat-unit system issue placed on the November ballot, the Mississippi legislature passed a bill letting each county chose its system of governance.2 In addition, however, the bill mandated that all of the state’s 82 counties adopt a centralized system for administering purchasing, inventory control and storage, and personnel policies. In other words, the legislature’s action meant that Mississippi’s voters would vote only on the issue of centralized versus decentralized road districts. Forty-six counties voted for the unit system and 36 counties voted for the beat system.3
The impetus for these events was a political scandal that erupted when 57 of the state’s 410 county supervisors were charged with corruption in connection with county purchasing activities. Caught on tape by federal agents posing as sales representatives of a bogus company seeking to do business with county governments in Operation Pretense, an FBI “sting” that ran from March 1984 to sometime in late 1987,4 the charges ranged from bribery, extortion, mail fraud, bid-rigging, and accepting kickbacks from suppliers to “busting” county invoices, a practice whereby supervisors authorized payment for supplies never delivered and then split the overcharge with vendors. Fifty-five supervisors in 26 Mississippi counties, one county road foreman, two state highway commissioners and 13 vendors eventually were convicted of engaging in corrupt practices. Only one of the seven supervisors who went to trial was found
2 This special bill was necessary because Mississippi’s constitution does not allow for voter initiative or referendum measures.
3 See Karahan, Razzolini and Shughart (2002) for an analysis of the November 1988 election and its sequel, in November 1992, in which two unit-system counties, Jones and Tate, voted to revert to the beat system. According to Crockett (2003: 297), Lincoln County switched back to the beat system in early 2000.
4 The first indictments were handed down on Friday the 13th of February 1987 (Crockett, 2003: 3).
3
innocent; one supervisor was found mentally incompetent to stand trial and another died while under indictment and awaiting the disposition of his case. A guilty verdict was likewise handed down by the jury empanelled to hear the case of the sole vendor who contested the charges filed against him. The remaining defendants, including all of the supervisors, pled guilty to various felonies in federal or state court; all were ordered to pay fines, make restitution of ill-gotten gains, or both, and received prison sentences of up to 10 years.5
The purpose of this paper is to attempt to identify the factors that led to the filing of corruption charges against one-eighth of Mississippi’s county supervisors. Using a unique county-level dataset, we find that governments tend to be more corrupt in poor, rural, less well-educated counties. We also find that a county’s supervisors were more likely to engage in corruption when corruption existed in neighboring counties. Finally, the evidence suggests that, other things being the same, corruption was more prevalent in counties where supervisors were paid more.
The paper is organized as follows. Section II summarizes our data and reports descriptive statistics on corrupt and non-corrupt counties in the State of Mississippi. Empirical models and estimation results are presented in Section III. Section IV concludes the paper.
II. County Government Organization and County Government Corruption
Proponents of the unit system of county government, led by then-Governor Ray Mabus, who had served as the State Auditor of Mississippi while Operation Pretense was underway and based his successful campaign for the governor’s mansion on his office’s close cooperation with the FBI’s undercover investigation, highlighted the cost savings that would follow from reducing wasteful
5 See Crockett (2003: 301–7) for a list containing the identities of the individuals charged under Operation Pretense and the disposition of their cases. The charges against many of the indicted vendors and public officials subsequently were reduced – and in a few instances dismissed – in return for cooperation in the FBI’s investigation.
4
duplication of capital and other resources due to multiple road districts. The reformers also pointed to the unit system’s advantage in reducing the incentives for corruption that existed under the beat system which gave individual supervisors a “license to steal” owing to their personal and unmonitored control over county funds and resources: “each supervisor is a one-man business.... He can sign the purchase order, the receiving slip and the order for payment.”6 In expressing support for the unit system, U.S. Attorney Robert Whitwell argued that, “The state needs honest professionals running county government, not untrained road foremen and good ol’ boys.... The county unit bill will help eliminate supervisors who abuse the system and line their pockets at the expense of taxpayers by removing supervisors from direct involvement in road building and maintenance.”7
On the other hand, opponents of the unit system, spearheaded by the Mississippi Association of Supervisors (MAS), maintained that such efficiencies were largely based upon the untested assumptions that a centralized county government would exercise greater control over unethical and illegal behavior, allocate the county’s resources public-spiritedly toward projects whose benefits were high relative to their costs, and, thus, promote more efficient provision of local public goods.8 The MAS contended that, by preventing direct communication between elected supervisors and their constituents, the unit system would in fact do just the opposite, namely worsen road repair and maintenance services.
6 Crockett (2003: 4), quoting an editorial published in the Star-Herald (Kosciusko, Miss.), 19 February 1987.
7 Clarion-Ledger (Jackson, Miss.), 26 October 1988, p. 1B.
8 To our knowledge, only one empirical study comparing beat and unit system counties has ever been conducted. Published in August 1987, the study examined the experience of Neshoba County, whose supervisors voted unanimously on July 2, 1984, to switch from the beat to the unit system. The transition was completed on September 1, 1986. A comparison of total county expenditures for personnel, utilities and contractual services, and materials for fiscal year 1985–86 (Mississippi’s fiscal year runs from July 1 to June 30) with annualized data for the first three months of fiscal year 1986–87 showed that the unit system produced savings on the order of 38%, or $533 per county road mile (Crockett, 2003: 182–83).
5
Figure 1 shows the Mississippi counties operating under the beat system as of November 1992, following the votes in 22 counties that had petitioned to revisit the decision taken in November 1988; two of those 22 counties reverted to the beat system. An empirical analysis of those two votes suggests that less populous counties, counties with smaller farms, and counties spending relatively more on roads and road repair and maintenance were more likely to choose the beat system over the unit system (Karahan, Razzolini and Shughart, 2002). Interestingly, the same study also found that the choice of county governance system in either 1988 or 1992 was not significantly related to evidence of corruption: other things being the same, voters in counties where one or more supervisors had been caught in Operation Pretense’s dragnet were no more likely to favor the unit system than their counterparts in counties where indictments had not been handed down.9
Figure 2 shows the 26 Mississippi counties in which supervisors were convicted on charges brought to light by Operation Pretense. The investigation was launched in March 1984, some two years after the Reverend John Burgess, a Pentecostal minister who owned a building supply business in Carthage, Mississippi, a town in centrally located Leake County, first contacted the FBI to inform them that he had learned, after investing in Polk Concrete, a pipe manufacturing company just south of Jackson, that the company’s sales representatives could not do business with some counties unless they kick backed 10 percent of the purchase price to supervisors (Crockett, 2003: 4).10 Burgess subsequently agreed to open an FBI front operation,
9 At the time of the FBI’s investigation, virtually all of Mississippi’s county governments operated under the beat system. The exceptions are Neshoba County, described in footnote 8 above, and Coahoma County, where, according to Crockett (2003: 231), “the unit system has been in operation more than 55 years….”
10 Bribes equal to ten percent of the value of the benefits conferred seem to be something of an iron law of government corruption. In her summary of the relevant literature, Rose-Ackerman (1999) reports “commissions” of that magnitude being paid for import licenses in the Philippines (p. 11), for reductions in taxes owed by property owners in New York City (p. 19), and for the award of international construction contracts in Paraguay (p. 28). Also see Atkinson, Couch and Shughart (1992).
6
Figure 1. Systems of County Governance in Mississippi circa November 1992
Notes: Beat system counties indicated by square dots. Lincoln County reverted to the beat system in early 2000.
7
Figure 2. Corrupt and Non-Corrupt Mississippi Counties
Note: Square dots indicate counties where one or more supervisors were convicted of corruption on charges brought under Operation Pretense.
8
named Mid-State Pipe Company, next to his business in Carthage, and to allow federal agents to pose as sales representatives of Mid-State, providing cover for a broad investigation of county government purchasing practices (ibid.: 5).
There is reason to believe that the sample of indicted supervisors does not include all of the county officials who engaged in illegal activities during the 1980s. Avoiding the possibility that defendants could claim entrapment required “predication” – evidence that a supervisor was predisposed to corruption. In the early stages of Operation Pretense, Jerry King, one of the federal undercover agents, met with Ray Davis, a vendor who began cooperating with the FBI in June 1986. At that meeting, “King took the list of members of the Mississippi Association of Supervisors and went over it with Davis, as they told each other which supervisors they had paid kickbacks” (Crockett, 2003: 9). In an interview aired in May 1987 on 60 Minutes, Davis said that “he had made kickbacks to twenty-five to thirty-five county supervisors all over the state” (ibid.: 278; emphasis added). These statements suggest that the FBI’s agents cast their net widely and pursued most, if not all, of the supervisors for whom they had documented evidence of a propensity to corruption.11 On the other hand, at the end of his exhaustive study of Operation Pretense, Crockett (2003: 298) concludes that
No doubt the percentage of supervisors convicted would have been much greater had the undercover investigation covered all eighty-two counties rather than twenty-six. Because of the strain on the undercover agents and security concerns, the investigation was halted by the federal authorities when they had secured enough evidence to make an “Impact Statement” concerning the corrupt system. Local authorities were expected to pick up the ball and effect the necessary reforms.
11 Additional anecdotal evidence that corruption was widespread and, hence, that our dataset represents a sample of the universe of corrupt Mississippi supervisors, even if, given the legal requirement of “predication”, it necessarily is not a random one, was supplied in a personal communication with the authors by a 2003 candidate for the post of supervisor in Lafayette County. The candidate claimed that the FBI had prepared Operation Pretense indictments against two of that county’s supervisors, but that the indictments were quashed as a result of local political pressure. We are unable to determine whether such scenarios played out elsewhere in Mississippi.
9
With that caveat in mind, Table 1 classifies the 26 counties in which Operation Pretense convictions were obtained according to the number of supervisors found guilty at trial or entering guilty pleas in federal or state court. The table shows the mean fines assessed, restitutions ordered and jail sentences imposed per guilty official.
Table 1. Fines, Restitutions and Jail Sentences
Number of Corrupt Supervisors
Number of Counties
Average Fine per Supervisor
Average Restitution per Supervisor
Average Sentence per Supervisor
(months)
1
10
$3,600
$2,399
37
2
7
6,143
2,135
41
3
7
4,833
1,198
43
4
1
2,375
926
152
5
1
1,700
0
114
Two points about these data are worth emphasizing. First, Table 1 shows the penalties actually imposed on the 54 Mississippi county supervisors who were convicted or pled guilty to corruption.12 Except for six of the seven individuals who contested the charges against them and subsequently were found guilty at trial, all of these penalties resulted from plea bargains with U.S. Attorneys. As such, the fines, restitutions and jail sentences ordered by the courts emerged from deals in which the government’s lawyers accepted reductions in the original number of felony counts in return for guilty pleas that avoided the costs of litigation. Penalties were reduced most in plea bargains with supervisors who agreed to cooperate with the FBI’s investigation and the most cooperative supervisors often were those against whom the most serious charges initially had been filed. Across all 26 counties, plea bargaining produced dramatic decreases in
12 As mentioned earlier, one of the 57 indicted supervisors was found mentally incompetent to stand trial (Perry County), one supervisor (Claiborne County) was found not guilty of all charges and another (Newton County) pled guilty but died while awaiting the final disposition of his case. The indictment against the deceased supervisor subsequently was dismissed. See Crockett (2003: 17, 66 & 196). The number of supervisors indicted in each of the 26 Mississippi counties is listed in the appendix.
10
the penalties for corruption, lessening the mean fine per guilty supervisor from $527,259 to $4,463 and shortening prison sentences considerably. Deals with prosecutors resulted in much smaller reductions in restitution amounts – from $2,252 to $1,532, on the average.13
Second, the court-ordered restitutions tend to be small in dollar value. That is because most of the convicted supervisors took advantage of state procurement laws which allowed them to “make purchases of up to $500 without competitive bids.” While “purchases between $500 and $2,500 required two bids or quotes, ... the supervisors did not have to advertise for bids. Advertising for competitive bids was required for purchases costing more than $2,500” (Crockett, 2003: 10). These legal requirements frequently were evaded by splitting invoices for large purchases into two or more small purchase orders, although that practice was itself illegal. Hence, computed as they normally were at 10 percent of the invoice, the kickbacks supervisors illegally accepted on any one transaction rarely exceeded $200.
III. Empirical Models of Corruption
Micro-level studies of governmental corruption are few in number. Most prior work in this area involves cross-country comparisons and focuses on the consequences – as opposed to the causes – of rent-seeking, corruption, or both (Tullock, 1967; Krueger, 1974; Bhagwati, 1982; Tollison, 1982; Elliot, 1997). The papers by Mauro (1997) and Ades and Di Tella (1999) are the most recent attempts explicitly to model corruption. Mauro studied its negative effects on private
13 Restitutions were determined by the amounts supervisors feloniously had accepted in kickbacks on transactions with vendors. The monies at issue were ordered paid to their respective county governments. Under Mississippi law, it is not illegal for public officials to accept gratuities, provided that the gifts are declared and turned over to the public treasury (Crockett, 2003: 82). Rose-Ackerman (1999, 55) observes that, by itself, “the right of recovery” of ill-gotten gains “is a weak deterrent to corrupt payoffs because the recovery is not multiplied by a factor that reflects the probability of detection.”
11
investment and rates of economic growth.14 Ades and Di Tella, by contrast, attempted to explain the causes of corruption. Per capita GDP, schooling, lack of political rights, factors related to the levels of competition and regulation were analyzed in a number of multivariate models. Some interesting results reported by Ades and Di Tella are that rents encourage corruption and that education and per capita GDP do not have significant impacts on corruption.15
A. Data, Models and Estimation Methodology
Our dataset combines information from Crockett (2003), the Mississippi Statistical Abstract (Office of External Affairs, various years), and the Census Bureau’s County and City Data Book (U.S. Department of Commerce, various years). The first of these sources identifies the county supervisors who were indicted in the course of Operation Pretense and reports on the disposition of their cases.16
In order to explore the determinants of corruption amongst Mississippi’s county supervisors, we estimated the following cumulative logistic regression model:
NUMBER OF CORRUPT SUPERVISORS = f(SUPERVISOR’S SALARY, BORDER, EDUCATION, POPULATION DENSITY, RATIO).
14 Bardhan (1997) contains a survey of the literature addressing the impact of corruption on economic development. The growth-retarding effects of corruption may operate by enlarging the shadow economy (Schneider and Enste, 2000: 90–91).
15 These results are based on data for the 1990s. During the 1980s, by contrast, Ades and Di Tella report that corruption is negatively and significantly related to the level of income per capita and educational attainment in a country.
16 Prior to the publication of Crockett’s (2003) book, we had constructed a list of indicted supervisors by consulting stories published in the Jackson, Miss., Clarion-Ledger, 15 February 1987, p. 1A; 9 November 1988, p. 1A; and 13 September 1987, p. 1A. A Freedom of Information Act request submitted by the authors to the Jackson Office of the U.S. Attorney generated an incomplete list of the fines and restitutions ordered in Operation Pretense-related cases; it did, however, identify the 26 counties in which felony convictions were obtained.
12
The observations were recorded at the county level and cover the years between 1984 and 1990. The ordered dependent variable assumes a value of 5 if all of a county’s supervisors were convicted of corruption, a value of 4 if four of five supervisors were convicted, and so on. The dependent variable is set equal to 0 if none of a county’s supervisors were caught in Operation Pretense’s net. The independent variables are defined as follows.
SUPERVISOR’S SALARY: In the State of Mississippi, the salaries of supervisors are fixed by law and based on a county’s total assessed value. Given the assessed value, the Mississippi legislature determines the maximum income the members of each county’s board of supervisors are authorized to earn. That figure is known as a county’s “code income” and all five of a county’s supervisors are paid the same amount. Unfortunately, we have not been able to obtain information on supervisors’ actual salaries, which may be less than code income in some counties.17
Be that as it may, the economic theory of crime and punishment (Becker, 1968; Becker and Stigler, 1974) suggests that public officials are more likely to be corrupt when the income earned from public service is low. Put differently, high pay deters corruption in an efficiency wage sense because officials fear the loss of a secure income stream if malfeasance is detected and they are forced to resign. Goel and Nelson (1998) report evidence consistent with the efficiency-wage story.18 On the other hand, Van Rijckeghen and Weder (1997) reject the
17 See Mississippi Code of 1972 § 25-3-13 for authorized county supervisor salaries. For the purposes of this study, we have used the 1989 grand assessed value (the earliest date available to us) of each county to determine supervisors’ salaries based on the numbers suggested by the 1985 amendment to Mississippi Code § 25-3-13 (House Bill. No 185, approved March 25, 1985, effective October 1, 1985). Paul Smith of the Mississippi State Tax Commission provided us with the 1989 grand assessed values of Mississippi’s 82 counties. The models were also run using the figures suggested by the 1988 amendment to the Mississippi Code. The results are identical and available upon request.
18 In particular, using 1983–87 data from the U.S. states, they find that the conviction rates of public employees for abuses of office are significantly lower where the average wage of state and local government employees is higher relative to state personal income. On the other hand, Goel and Nelson also find a positive and significant relationship between corruption and government size, as measured by real total state and local government
13
hypothesis that higher relative wages lead to lower levels of public corruption, as measured by a corruption index.19 The implication is that, if corrupt activities are highly remunerative, a high official salary may not be enough to deter corruption. As Rose-Ackerman (1999, 78) puts it,
high pay may simply increase the bribe an official demands in order to overcome the risks of losing what is now a desirable job. Officials may go from being “lean and mean” to being “fat and mean.” The incidence of bribery may fall as fewer officials solicit or accept payoffs, but the size of each bribe increases.
In short, the data will have to tell us which effect dominated in Mississippi at the time of Operation Pretense. Other things being the same, was corruption more or less likely in counties where the salaries of supervisors were higher?
BORDER is an indicator variable assuming a value of 1 if a county borders a corrupt county; 0 if not. This variable is constructed to account for spatial autocorrelation and to test whether corrupt behavior in one county transfers to (or is imitated by) neighboring counties. Figure 2 preliminarily suggests that this is so: of the 26 corrupt counties in Mississippi, 24 bordered one another; of the 56 non-corrupt counties, 36 bordered corrupt counties.
POPULATION DENSITY: A negative relationship between the population density of a county (observed here in 1985) and public corruption is hypothesized. That a priori algebraic sign is suggested by the logic of collective action (Olson, 1965), which implies that it is more costly for voter-taxpayers to monitor their elected public officials and control corrupt activities in rural counties where the population is more dispersed geographically.
expenditures per capita. Since one would expect the salaries of public employees to be higher in big-government states, as a test of the Becker-Stigler model these empirical results are ambiguous at best. For additional applications of the Becker-Stigler model to governmental corruption, see Basu, Bhattacharya and Mishra (1992), Besley and McLaren (1993) and Mookherjee and Png (1995).
19 This result is obtained on time-series data from 25 countries. In the same study, using a cross-sectional methodology, Van Rijckeghen and Weder find a negative relationship, ceteris paribus, between relative pay (the ratio of civil service to manufacturing wages) and the corruption index.
14
EDUCATION: The level of education in a county is measured by the percentage of the population holding bachelor’s degree or higher in 1990. Consistent with previous studies, as well as on the theory that a less educated citizenry is both less likely to be informed about their county supervisors and to vote, we posit a negative relationship between education and corruption.
RATIO: This variable is constructed by dividing per capita county taxes by per capita direct general county expenditures as of 1987. It is reasonable to assume that corruption is lower in counties where this ratio is higher. Taxpayers would have more incentive to monitor the behavior of their elected supervisors in counties where local taxes are higher relative to local spending.
Table 2 reports descriptive statistics. It appears on simple pair-wise comparisons that average supervisor salaries are slightly higher in counties where one or more supervisors were convicted on corruption charges. This is consistent with the results obtained by Van Rijckeghem and Weder (1997). Corrupt Mississippi counties also appear to have lower education levels and higher population density. RATIO seems to be higher in the corrupt counties. Whether these relationships hold in a multivariate context is the question we answer in the following section.
Table 2. Descriptive Statistics
Non-Corrupt (56 Counties)
SALARY ($)
EDUCATION (%)
POPULATION DENSITY
RATIO
Mean
18,730
7.08
51.32
0.1242
Standard Deviation
2,252
3.05
44.40
0.0373
Minimum
15,500
3.25
6.19
0.0592
Maximum
24,600
17.55
286.85
0.2644
Corrupt (26 Counties)
SALARY ($)
EDUCATION (%)
POPULATION DENSITY
RATIO
Mean
19,719
6.59
53.41
0.1319
Standard Deviation
2,313
2.27
54.46
0.0409
Minimum
16,300
3.61
13.69
0.0684
Maximum
24,600
12.47
271.37
0.2642
15
B. Empirical Results
Estimates of two empirical models are presented in Tables 3a and 3b. The first set of results is based on the cumulative (ordered) logit model described above; the second set of results replaces the five intercept terms denoting the number of county supervisors convicted of corruption (0, 1, 2, 3, 4, or 5) with a binary dependent variable taking a value of 1 if any supervisors were convicted in a county and set equal to 0 otherwise.20 The two regressions tell essentially the same story.21
Table 3a. The Likelihood of Observing Corruption (Cumulative Logit)
Variable
Coefficient Estimate
p-value
Intercept-1
–16.2165
0.0001
Intercept-2
–17.0835
< .0001
Intercept-3
–17.8548
< .0001
Intercept-4
–19.5591
< .0001
Intercept-5
–20.2981
< .0001
SUPERVISOR’S SALARY
0.0009
0.0007
BORDER
2.5612
0.0032
EDUCATION
–54.4450
0.0108
POPULATION DENSITY
–0.0230
0.0456
RATIO
8.854
0.2618
Wald chi-square for overall significance: 0.0052
Percentage of correctly classified responses: 0.78
20 The cumulative logit model allows for different intercepts. In Table 3a, for example, Intercept-5, indicating counties where five supervisors were convicted of corruption, is the predicted log-odds ratio of being in the five- corrupted-supervisors category rather than being in categories where four, three, two or one supervisors were convicted, when the values of the other independent variables are constrained to be zero (Allison, 1999: 134–42). The standard logit model in Table 3b assumes a binary response where the dependent variable can take on two values: 1 for those counties having one or more corrupt supervisors and 0 otherwise.
21 It is possible that there may be over- or under-dispersion in our data on account of omitted variables, outliers, or the small numbers of observations in some sub-categories. Correcting for this using the Pearson chi-square statistics (SAS Institute, 1995: 81–86) resulted in the following p-values for the cumulative logit model: SUPERVISOR’S SALARY < 0.0001, BORDER = 0.0005, EDUCATION = 0.0025, POPULATION DENSITY = 0.0178, and RATIO = 0.1835. The corresponding p-values for the binary logit model were: SUPERVISOR’S SALARY = 0. 0006, BORDER = 0.0024, EDUCATION = 0.0050, POPULATION DENSITY = 0.0534, and RATIO = 0.1984.
16
Table 3b. The Likelihood of Observing Corruption (Binary Logit)
Variable
Coefficient Estimate
p-value
Intercept
–19.9166
< .0004
SUPERVISOR’S SALARY
0.0011
0.0002
BORDER
2.9046
0.0011
EDUCATION
–72.3613
0.0025
POPULATION DENSITY
–0.0227
0.0373
RATIO
11.3518
0.1621
Wald chi-square for overall significance: 0.0076
Percentage of correctly classified responses: 0.84
In particular, the likelihood of observing supervisor corruption is higher in counties where supervisors’ salaries are higher. This finding is consistent with the hypothesis that high pay for public officials by itself is not enough to deter corruption when the extra income available from corrupt activities is also high. Of course, it is also possible that our model fails to include relevant personal characteristics of the supervisors who engaged in corruption.22
As expected, corruption is more likely in counties with lower education levels and lower population densities. Moreover, it appears that corruption is contagious: BORDER is positive and significant, indicating that the existence of corruption in one county raises the likelihood of observing corruption in neighboring counties. RATIO is not significant.23
In addition to providing support for the conjecture of Van Rijckeghem and Weder (1997), our finding of a positive relationship between corruption and pay is also consistent with the following interpretation. As noted earlier, the Mississippi Code determines the maximum
22 One possibility is a supervisor’s length of service in office. This could go either way. Trudie Westmoreland of Perry County, the first Mississippi supervisor tried and convicted of Operation Pretense-related charges, argued in court that her inexperience – she was serving her first term – caused her erroneously to accept kickbacks from suppliers. On the other hand, her fellow Perry County supervisor, Junie Mixon, who died while awaiting sentencing on corruption charges, had been in office for 19 years. See Crockett (2003: 15ff.).
23 Neither is the size of government measured in terms of county taxes or county spending entered separately either as totals or as per capita amounts.
17
allowable salaries of county supervisors. A given county’s supervisors are nevertheless free to pay themselves less than the allowed maximum, an action they might take if they fear voter opposition to increasing their own incomes.24 In such cases, supervisors may have incentive to capture the difference under the table by accepting or extorting bribes from vendors.25 To test this possibility, we replaced SUPERVISOR’S SALARY in our two regression models by a new independent variable, defined as the difference between the supervisor’s salary and the per capita income in a county.26 Without affecting the signs or significances of the other included explanatory variables (DENSITY is the sole exception), the estimated coefficient on the new variable is positive and significant at less than one percent, providing some support for this conjecture.27
Yet another possibility is that SUPERVISOR’S SALARY is simply a proxy for a county’s assessed value, which is itself a proxy for available rents. After all, “code income” in a county is determined exactly that way.28 Replacing SUPERVISOR’S SALARY with GRAND ASSESSED VALUE, we find that the estimated coefficient on the latter variable is positive and significant at the ten percent level. EDUCATION remains negative and significant at ten percent and BORDER is positive and significant at five percent. RATIO and POPULATION DENSITY
24 Mississippi law requires a simple majority of a county’s supervisors to vote affirmatively to accept a legislatively authorized pay increase and to record the results of such voting on the board’s minutes.
25 “If public pay is very low, corruption is a survival strategy…. In these cases, officials are likely to take second jobs or accept payoffs as salary supplements” (Rose-Ackerman, 1999: 72).
26 The mean per capita income in the 26 counties with one or more corrupt supervisors is $8,485, with a standard deviation of $964.57; the mean per capita income in the non-corrupt counties is $8,365 (standard deviation = $1,215.50).
27 An anecdote recounted by a candidate for Lafayette County supervisor (see footnote 11) reinforces this interpretation. According to him, it was common knowledge before Operation Pretense began making headlines that supervisors took money under the table to supplement the low pay they earned from performing what was then a part-time job. Voter indifference to revelations of government corruption is illustrated by a story from Pontotoc County. On the evening prior to the date on which one of that jurisdiction’s five convicted supervisors was scheduled to report to federal prison, the members of his church threw a going-away party in his honor.
28 The simple correlation coefficient between the 1989 grand assessed county value and code salary is +0.76, which is significantly different from zero (p < 0.0001).
18
are insignificant in this specification. Thus, we cannot disentangle empirically two alternative interpretations of our regression results. One is that high pay for public officials fails to deter corruption, at least in the State of Mississippi. The other is that corruption is more likely, ceteris paribus, in counties where more rents are up for grabs.
We have run other models to see whether our results would change if we were to account for the possibility that a Newton County supervisor had not died while under indictment and a Perry County supervisor had not been declared mentally incompetent to stand trial. The results reported in Table 3 are unaffected by adding these two individuals to the sample of convicted public officials.
C. Competition and Corruption
Across the 26 Mississippi counties in which supervisors were convicted of the unlawful practices brought to light by Operation Pretense, corruption involved as few as one and as many as five supervisors, with an overall mean of 2.08 individuals. The available empirical evidence affords an opportunity to test whether the level of governmental corruption is a function of the number of public officials who participated in breaking the law.
In extending the theory of oligopoly to corruption in the public sector, Shleifer and Vishny (1993) show that competition among a set of officials who have authority to grant similar favors tends to reduce the bribe any one of them can demand.29 This result follows from observing that, if one officeholder insists on a supranormal kickback, rational favor-seekers will respond by shifting their business to alternative officials who are willing to accept lower fees for service. On the other hand, when a single officer acts as gatekeeper, or two or more public
29 Rose-Ackerman (1978) apparently was the first scholar to suggest that corruption can be reduced by introducing competition at the level of the official receiving bribes.
19
employees having such authority are able successfully to collude, bribes approach monopoly profit-maximizing levels.
Because collusive agreements are cheaper to organize and to enforce the fewer the parties involved (Stigler, 1974), theory predicts less wealth extraction from vendors in counties where more of the supervisors engaged in the corrupt practices uncovered by Operation Pretense. Accordingly, we divided our sample of 26 Mississippi counties into the following two categories: 17 counties with a below-average level of corruption (one or two convicted supervisors) and nine counties where the level of corruption observed was above the average (three, four or five convicted supervisors). We then calculated the arithmetic means of the monetary penalties (fines and restitutions) imposed per guilty supervisor in the two subsamples. The results are reported in Table 4.
Table 4. Average Fines and Restitutions per Supervisor Grouped by Numbers of Guilty Officials
Number of Corrupt Supervisors per County

Monetary
Penalties
One or Two
(17 Counties)

Three or More
(9 Counties)
Average Fine
$5,083 - $3,983

Average Restitution
2,245
962

Note: The mean number of guilty supervisors per county is 54/26 = 2.08.
The available data suggest that corrupt behavior was less costly to vendors in counties where more of the supervisors were on the take. The restitutions ordered in the nine jurisdictions where convictions were obtained against three or more supervisors were substantially smaller, on the average, than those ordered in counties where only one or two supervisors were found guilty of corruption. That result is consistent with the hypothesis that kickbacks were demanded less frequently from vendors in the more corrupt counties or, alternatively, represented smaller percentages of the dollar value of the unlawful transactions uncovered by Operation Pretense.
20
Supervisors in counties where three or more convictions were obtained also paid smaller fines, on the average, reinforcing the conclusion that corruption in those jurisdictions was less egregious.
Additional insight into corruption is gained by comparing supervisors’ salaries in these two groups of counties. Throughout the second half of the 1980s, average supervisors’ salaries were higher in the 17 counties where Operation Pretense produced the largest monetary recoveries per guilty supervisor. In particular, the mean annual income of supervisors in that sample of counties was $20,294 in 1985.30 During the same year, by contrast, supervisors earned salaries averaging $18,633 per year in the nine counties where three or more of the officials were found to be corrupt and fines and restitutions were lower,31 a difference of $1,661 or 8.9%.32 To the extent that court-ordered restitutions are indicators of the level of bribery, we have some evidence that high pay reduces the number of officials who engage in unlawful behavior: among the 26 counties where at least one supervisor was corrupt, fewer supervisors were convicted where average salaries were higher. On the other hand, a high-paid official who becomes corrupt tends to demand larger bribes or demands them more often. “Officials … go from being ‘lean and mean’ to being ‘fat and mean’”, as Rose-Ackerman (1999: 78) suggests.
Based as they are on small numbers, these results are far from definitive. The evidence at hand nevertheless indicates that competition between corrupt public officials tends to reduce the level of bribery and, moreover, that officials who are paid more extract bigger bribes. But to reiterate our earlier findings, Mississippi’s county supervisors were more likely to have engaged in corruption where their salaries were higher, all else equal.
30 Supervisors’ salaries in these 17 counties ranged from a low of $17,700 (Perry and Claiborne counties), to a high of $24,600 (Harrison, Jackson and Rankin counties).
31 The salary range in these nine counties was $16,300 (Greene County) to $22,100 (Hancock County).
32 Supervisors’ salaries in the two groups of counties differed by $1,787, or 8%, in 1988.
21
IV. Summary and Conclusions
This paper models public corruption at one of the lowest levels of governance. Previous studies have examined corrupt behavior largely on an international scale. Thanks to Operation Pretense, our dataset, by contrast, contains well-documented cases of corruption among approximately one-eighth of Mississippi’s 410 elected county supervisors.
At the micro level, we find evidence that governments tend to be more corrupt in rural, less densely populated counties where voter-taxpayers are not well-educated. We do not find that high pay for public officials leads to less corruption, ceteris paribus. As a matter of fact, corruption is found to be more likely in counties where supervisors were compensated more generously. Some rethinking of the efficiency wage story for public officials thus seems to be in order, especially so in light of the evidence reported herein that, among the sample of Mississippi counties in which any governmental corruption was observed, fewer supervisors engaged in corruption where their salaries were higher, but those who did were penalized more heavily, suggesting that bribes were larger or demanded more frequently. On the other hand, competition among county supervisors for bribes is shown here to reduce the level, if not necessarily the prevalence, of thievery.
22
Appendix. Number of Supervisors Indicted by County
County
Number of Corrupt Supervisors
Pontotoc 5
Hancock 4
Attala 3
Copiah 3
Greene 3
Leake 3
Neshoba 3
Newton 3
Wayne 3
Clarke 2
Jasper 2
Lamar 2
Lauderdale 2
Monroe 2
Perry 2
Rankin 2
Claiborne 1
Covington 1
Harrison 1
Jackson 1
Lincoln 1
Marion 1
Panola 1
Scott 1
Smith 1
Winston 1

Totala
54
a Operation Pretense indictments were handed down against 57 of Mississippi’s 410 county supervisors. The total shown here excludes one supervisor who was found mentally incompetent to stand trial (Perry County), one supervisor (Claiborne County) who was found not guilty of all charges, and a third (Newton County) who pled guilty but died while awaiting the final disposition of his case.
23
References
Ades, A. and R. Di Tella (1999). “Rents, Competition, and Corruption”, American Economic Review 89: 982–93.
Atkinson, K. E., J. F. Couch and W. F. Shughart II (1992). “Ethics Laws and the Outside Earnings of Politicians: The Case of Alabama’s ‘Educator-Legislators’”, Public Choice 73: 135–45.
Allison, P. (1999). Logistic Regression Using the SAS® System: Theory and Application. Cary, NC: SAS Institute, Inc.
Bardhan, P. (1997). “Corruption and Development: A Review of Issues”, Journal of Economic Literature 35: 1320–46.
Basu, K., S. Bhattacharya and A. Mishra (1992), “Notes on Bribery and the Control of Corruption”, Journal of Public Economics 48: 349–59.
Becker, G. S. (1968). “Crime and Punishment: An Economic Approach”, Journal of Political Economy 76: 169–217.
Becker, G. S. and G. J. Stigler (1974). “Law Enforcement, Malfeasance, and Compensation of Enforcers”, Journal of Legal Studies 3: 1–18.
Besley, T. and J. McLaren (1993), “Taxes and Bribery: The Role of Wage Incentives”, Economic Journal 103: 119–41.
Bhagwati, J. N. (1982). “Directly Unproductive, Profit-seeking (DUP) Activities”, Journal of Political Economy 90: 988–1002.
Crockett, J. R. (2003). Operation Pretense: The FBI’s Sting on County Corruption in Mississippi. Jackson: University Press of Mississippi.
Elliott, K. A. (Ed.) (1997). Corruption and the Global Economy. Washington, DC: Institute for International Economics.
24
Goel, R. K. and M. A. Nelson (1998). “Corruption and Government Size: A Disaggregated Analysis”, Public Choice 97: 107–20.
Karahan, G. R., L. Razzolini and W. F. Shughart II (2002). “Centralized versus Decentralized Decision-Making in a County Government Setting”, Economics of Governance 3: 101–15.
Krueger, A. O. (1974). “The Political Economy of the Rent-Seeking Society”, American Economic Review 64: 291–303.
Mauro, P. (1997). “The Effects of Corruption on Growth, Investment, and Government Expenditure”, in K. A. Elliot (Ed.), Corruption and the Global Economy, 83–107. Washington, DC: Institute for International Economics.
Mookherjee, D. and I. P. L. Png (1995), “Corruptible Law Enforcers: How Should They be Compensated?”, Economic Journal 105: 145–59.
Office of External Affairs, Division of Research (various years). Mississippi Statistical Abstract. Starkville, MS: Mississippi State University.
Olson, M. (1965). The Logic of Collective Action: Public Goods and the Theory of Groups. Cambridge: Harvard University Press.
Rose-Ackerman, S. (1978). Corruption: A Study in Political Economy. New York: Academic Press.
Rose-Ackerman, S. (1999). Corruption and Government: Causes, Consequences, and Reform. Cambridge: Cambridge University Press.
SAS Institute, Inc. (1995). Logistic Regression Examples Using the SAS System, ver. 6, 1st ed. Cary, NC: SAS Institute, Inc.
25
Schneider, F. and D. H. Enste (2000). “Shadow Economies: Size, Causes, and Consequences”, Journal of Economic Literature 38: 77–114.
Shleifer, A. and R. W. Vishny (1993). “Corruption”, Quarterly Journal of Economics 108: 599–617.
Stigler, G. J. (1974). “A Theory of Oligopoly”, Journal of Political Economy 72: 44–61.
Tollison, R. D. (1982). “Rent Seeking: A Survey”, Kyklos 35: 575–602.
Tullock, G. (1967). “The Welfare Costs of Tariffs, Monopolies, and Theft”, Western Economic Journal 9: 379–92.
Van Rijckeghem, C. and B. Weder (1997). “Corruption and the Role of Temptation: Do Low Wages in Civil Service Cause Corruption?”, IMF Working Paper WP/97/73.
U.S. Department of Commerce, Bureau of Census (various years). County and City Data Book. Washington, DC: USGPO.