The Casino Industry and the Corruption of U.S. Public Offic

The Casino Industry and the Corruption of U.S. Public Officials in the U.S.

Douglas M. Walker and Peter T. Calcagno

College of Charleston

April 1, 2011

2

What is corruption?

Criminal acts perpetrated by politicians or government employees, with the goal of illegitimate personal gain

Dept

of Justice reports that election crimes are good indicators of corruption

Vote fraud, campaign finance crimes, political shakedowns

Bribes, favors for friends/family, racketeering, conspiracy, extortion, etc.

Glaeser

and Saks (2006) argue that federal corruption convictions are better indicators state crime convictions

If the state is corrupt, it may affect the number of charges filed by the state

Federal conviction #s range from 0 to over 100 per year in a particular state

3

Anecdotal evidence of a link

Las Vegas has a history of organized crime and corruption

Atlantic City, NJ has a long-established bad reputation

Over half of the last 9 mayors have been arrested on corruption charges

Louisiana governor Edwards convicted in 2000 of 17 corruption charges related to riverboat casino licenses

Illinois governor Blagojevich being sued by Illinois casino operators

FBI wiretaps indicate a shake-down of horse racetrack owners for support of a bill that requires casinos to share profits with the racetracks

Observers in Kentucky suggest that corruption has started there, even before casinos are being built

In AL, 11 people (including a few senators) were arrested for buying/selling votes related to casino legislation (2010)

4

Theoretical explanations for a link

Casinos are big, cash businesses

See Table 1, Figure 1

Gambling industry PAC and individual contributions to politicians are large

See Table 2

Likely to be large illegal “contributions” too?

Industry requires explicit government approval to exist

Huge rents are created by this restriction and regulation of the industry

Laws and regulations that apply only to this industry

Requires interaction between law makers and industry

Politicians can be influenced

State legalization, local permits, regulatory framework, etc., are opportunities for corruption

5

StateYr. LegalOrder LegalizedCasinoOpening# CasinosOperating in 20072007 Revenues(millions $)2007 Taxes Paid(millions $)Colorado19904Oct. 199145819115 Illinois19904Sept. 199191983834 Indiana19936Dec. 1995112625842 Iowa19893Apr. 1991171363315 Louisiana19915Oct. 1993182566559 Michigan19967July 199931335366 Mississippi19904Aug. 1992292891350 Missouri19936May 1994121592417 Nevada193111931270*128491034 New Jersey197621978114921475 Pennsylvania20048Oct. 200761090473 S. Dakota19893Nov. 1989369815 Totals---467341325795 

Table 1: US Casino states, 2007

Source: AGA (2008)

6

ElectionCycleInd.RankTotalContrib.FromIndividualsFromPACsSoft MoneyContrib.ToDemocratsToRepublicans% toDems% to Repubs200833$16.8m$14.4m$1.9mN/A$10.2m$6.0m6337200632$11.6m$3.7m$7.9mN/A$6.2m$5.4m5347200435$11.2m$4.2m$7.0mN/A$6.5m$4.7m5842200226$15.0m$2.1m$4.4m$8.6m$7.6m$7.5m5050200036$12.9m$2.7m$2.3m$7.9m$7.2m$5.6m5644199838$6.4m$1.5m$1.1m$3.9m$2.5m$3.8m4060199640$7.1m$2.0m$1.1m$4.1m$3.8m$3.3m5347199449$3.1m$1.2m$0.4m$1.5m$1.7m$1.4m5446199269$1.5m$0.8m$0.2m$0.5m$1.0m$0.5m6535199075$0.5m$0.2m$0.2mN/A$0.3m$0.1m7129Total39$85.7m$32.9m$26.4m$26.4m$47.0m$38.6m5545

Table 2: Industry contributions to US politicians

Source: Center for Responsive Politics (http://www.opensecrets.org)

7

8

A perfect recipe for corruption?

Lots of casino revenues, tax revenues

Large contributions to politicians from individuals & PACs related to casinos/gambling

Government approval needed for industry to exist

Every facet of the industry is regulated

Corruptible politicians

But there’s no empirical evidence of a link (at least, in the literature)…

9

Corruption literature

Numerous studies

Rose-Ackerman (1978),

Glaeser

& Saks (2006),

Fisman

&

Gatti

(2002), Alt & Lassen (2008), Lee &

Chelius

(1989), etc.

International studies (cross-sectional, at the national level)

Corruption is worse when firms are shielded from foreign competition

Negative relationship between corruption and

political freedom

decentralization of power

economic growth

“Law and order” and “democratic accountability” reduce corruption

US studies (at the state-level)

Higher education and income reduce corruption

Income inequality increases corruption

Racial “dissimilarity” (diversity?) increases corruption

No studies that examine a specific industry as a potential cause of corruption

10

Our data

We are interested in determining if there is a link between

commercial

casinos and corruption

Corruption is measured by federal corruption convictions of state/local government employees

By year and state: 1985-2005, 11 casino states

Covers most legalization (1989-96), except PA, NV, NJ

Data source: US

Dept

of Justice

Casino activity is measured by casino revenues

By year and state, 1985-2005, 11 casino states

Data source: State gaming regulatory agencies

231 observations (11 states, 21 years)

In most states there are several pre-casino observations

[For all states, 21 years, we have 1050 observations. But the model isn’t determinate using all states.]

11

Granger causality analysis

Standard Granger causality for time series data (Granger,

Econometrica

1969):

X

t

=



X

t

-j

+



Y

t

-j

+



Y “causes” X if inclusion of Y improves prediction of X

Y

t

=



Y

t

-j

+



X

t

-j

+



X “causes” Y if inclusion of X improves prediction of Y

Not exactly the same meaning as the common concept of “causality”

We cannot analyze casino legalization – an event – using this empirical analysis

We maybe should focus on this instead…

12

Granger causality for panel data

Walker & Jackson (1998) adapt Granger causality to panel data; we follow that method

We have two series:

Corrupt

and

Revenue

Step 1

Detrend

the panel data series of state- and time-specific information

State dummies, time trend, state-trend interaction terms

Use the residuals from each equation

Test the residuals for

stationarity

; if hypothesis of unit root is rejected, move on

13

Empirical analysis, continued

Step 2

Determine the autoregressive process that generates each series of

detrended

residuals

We want the fewest lag periods on which the variable must be regressed so that the residuals are white noise

Iterative process, adding additional lag periods until correlograms and Box-Pierce

Q

-statistics indicate residuals from these regressions are white noise

If these residuals are white noise, then if adding the other variable’s residuals improves prediction of the first series, then Granger causality exists

Models indicate that

Corrupt

requires

2

lag periods;

Revenue

requires

3

lag periods

Sample size must be adjusted to account for these

14

Empirical analysis, continued

Step 3

Run the regressions on each

detrended

series:

(1)

Corr

t

=

a

1

+

a

2

Corr

(t-1)

+

a

3

Corr

(t-2)

+

a

4

Rev

(t-1)

+

a

5

Rev

(t-2)

+

a

6

Rev

(t-3)

+

ε

(2)

Rev

t

=

g

1

+

g

2

Rev

(t-1)

+

g

3

Rev

(t-2)

+

g

4

Rev

(t-3)

+

g

5

Corr

(t-1)

+

g

6

Corr

(t-2)

+

ε

Granger causality tests (

F

-tests):

“

Revenue

does not cause

C

orruption

”

in equation (1) above, test coefficient restriction:

a

4

=

a

5

=

a

6

= 0

“

Corruption

does not cause

Revenue

”

in equation (2) above, test coefficient restriction:

g

5

=

g

6

= 0

15

Possible results

4 possible results

Casino revenues Granger cause corruption convictions

Casino revenues are used to bribe politicians to expand the industry or loosen regulation

Corruption convictions Granger cause casino revenues

Casinos are legalized in relatively corrupt states

Independence

No causal relationship either way

“No result” is still an interesting result (fortunately)

Feedback, or simultaneous determination

Each variable is contributing to the other

e.g.

, quid pro quo

between casinos and politicians

Timing here must be examined

16

Actual results

Table 4. Granger causality test results

HypothesisF-statisticProbabilitya4 = a5 = a6 = 0(Revenue does not cause Corrupt)1.180.319g5 = g6 = 0(Corrupt does not cause Revenue)5.030.008

17

Discussion

The first econometric evidence of a link

Analysis does not indicate

why

there is a link

A reasonable story can be told whatever the result

Suggests that casinos may be more likely to be legalized in relatively corrupt states

Anecdotal evidence: IL, LA, NJ, MS; next AL?

No evidence that casinos use revenues to corrupt politicians after casinos open

Casino licenses are expensive; unlikely to be put at risk…

Consistent with a public choice perspective, that politicians have the most power to extract rents when they are formulating casino legislation

Corruption is most likely to occur at this point

18

Discussion, cont.

Analyze legalization point instead of revenue stream?

Hazard model

Casino states only or all states?

Timing issues need to be addressed:

corruption ‘event’ and convictions

casino legalization and casino revenues

[see the following scenario]

Suggestions for robustness checks?

Granger causality provides evidence for subsequent analysis

Corruption leading to revenues/legalization

19

Figure 3. Possible timing of corruption convictions and casino revenues

Year 1

Casinos proposed

Some politicians solicit bribes to support casinos

Year 2

Casino bill

passesCorruption charges filed and trials begin; some trials end in conviction

Year 3

Casino building beginsAdditional trials end in conviction

Year 4

Casinos open for business (revenues begin)