Despite numerous criticisms, state lotteries continue to receive popular support. The popularity of lotteries is easy to comprehend when considered from the perspective of a self-interested individual. Individuals who desire to gamble view lotteries as a legal and readily available means to do so.

Those who do not participate, including those who otherwise oppose lotteries for moral or societal reasons, perceive that a portion of their tax burden is shifted to the participants. Hersch and McDougall (1989) find that higher-income individuals vote in favor of lotteries, not in order to participate, but in the expectation that other income groups will assume a portion of their taxes. Nonparticipants expect to benefit. Participants bear any attributable costs voluntarily in exchange for the privilege of gambling. From the perspective of self-interest, both participants and nonparticipants seem to have reasons to support lotteries.

The perception that nonparticipants benefit does not take into account an essential factor: the effect of lotteries on state legislatures. Lotteries provide state legislatures with a revenue source without the political risk of passing an explicit tax. Does this inflow of "free money" alter the expenditure decisions of state legislatures? Applying the principal-agent model, this paper examines how state lotteries increase agency costs that are borne by all citizens of the state. The result raises an additional criticism of lotteries that has implications for nonparticipants whose support for lotteries is based on expected tax benefits.

I

Criticism of State Lotteries

CRITICISM OF STATE LOTTERIES can be categorized as either (1) morally and ethically based or (2) tax based. The first line of criticism focuses on the moral implications of government-sponsored gambling. Some regard gambling, in any form, as an immoral act that should not be tolerated or legalized (Glaston and Wasserman 1996). Adherents to this position not only would not participate in the lottery but would prohibit others from participating as well. As with other vices, some view gambling as the voluntary choice of individuals that should not be prohibited but also should not be sponsored by government. Though a permissible choice of individuals, it is unethical for government to sponsor, encourage, and exploit the vices of its citizens in order to raise revenue. State governments raise tax revenue from alcohol and tobacco, but do not manufacture and encourage their use (Glaston and Wasserman 1996). The advertising and promotion of lotteries does not provide accurate information and, therefore, is deceptive (Stearns and Borna 1995). Lotteries and their advertising undermine the core principles of society by equating success with luck rather than with work (Glaston and Wasserman 1996). The enactment of lotteries correlates with an increase in crime (Mikesell and Pirog-Good 1990). A certain percentage of the population are dormant gambling addicts who previously abstained from gambling because it was not readily available or because the available avenues were illegal. At the margin, legalized gambling in the form of a state lottery will increase the number of active gambling addicts (Riverbank 1998; Madhusudhan 1996).

A second line of criticism faults the tax mechanism of lotteries. Lotteries constitute a regressive tax that falls disproportionately on those least able to pay (Price and Novak 2000; Borg and Mason 1988). Lottery participants are more likely to be less educated and to be members of minorities (Stranahan and Borg 1998; Borg and Mason 1988). Lotteries do not result in a promised increase in funding for education (Spindler 1995; Borg and Mason 1988). Lotteries slightly reduce proceeds from other forms of legalized gambling (Gulley and Scott 1989).

Despite criticism, lotteries receive majority support. Both moral and tax-based arguments against lotteries rest on the perspective of society as a whole, not that of the self-interested individual. A nonparticipant who believes lotteries harm society expects to benefit as an individual from shifting the tax burden to participants. The nonparticipant weighs the cost to society against perceived personal benefit. Lotteries receive support from nonparticipants for whom the perceived personal gain outweighs the cost to society as a whole.

II

Principal-Agent Model

THIS PAPER EVALUATES THE BENEFIT to nonparticipants, using the principal agent model to examine the impact of lotteries on state expenditures. In the generalized principal-agent model, the agent represents the principal with the designated purpose of serving the best interests of the principal. However, the agent's actions inevitably are influenced by his or her own best interests. Since the principal and agent are different individuals, their best interests and objectives will never be identical. The objectives of the principal and agent diverge, resulting in agency costs. The actions of the agent may be brought more closely in line with the objectives of the principal through monitoring or aligning the incentives so that the best interests of the agent are linked to the objectives of the principal.

Applied to state government, state legislators are the agent for the citizens of the state, the principal. Legislators represent the voters with the purpose of acting in the best interests of the citizens. At least to some extent, the objectives of the citizens and legislators diverge. (1) Voters want a capable representative. The legislator wants to be that representative. Legislators have an incentive to take actions that enhance their personal reelection prospects, even if those actions are not necessarily best for the state.

The citizens receive the benefit of services provided by state expenditures and pay the cost in the form of taxes. The question is not do the citizens want more services, but how much personal consumption are they willing to forego to pay for these services. The optimal state budget occurs at the point at which the marginal benefit to the citizens equals the marginal cost to the taxpayers. As agent for the citizens, state legislators should balance the benefit of expenditures against the cost to the taxpayers. However, reelection chances are increased by providing greater services for their constituents.

The agency conflict is most likely to emerge with expenditures that are not popular on a statewide basis but serve a limited constituency. Assume a park in one district and a bridge in another district are not included in an optimal state budget. However, securing funding for these projects enhances the reelection prospects of each district's legislator. Each legislator has an incentive to vote for the other's project in exchange for support for his or her own. The result is spending above the level that is optimal for the principals, in order to benefit the agent. This incentive to overspend is constrained by taxpayer monitoring of their explicit taxes.

A lottery provides the legislature with the ability to expend additional funds without raising direct taxes. Enacting a lottery does not free the legislature from its responsibility to weigh the benefits and costs of its expenditures, but it does lessen monitoring since it does not increase explicit taxes. If state legislators view lottery proceeds as "free money," they are less likely to weigh the marginal benefits of these expenditures against the marginal costs. This yields a testable implication. Applying the agency framework to lotteries implies that implementation of a state lottery will be associated with higher than normal growth in state expenditures.

III

Data and Methodology

FOR EACH STATE THAT IMPLEMENTED the lottery, total expenditures were obtained for the five-year time period beginning two years preceding implementation of the lottery and ending two years following implementation. Total expenditures for all states combined were obtained for the same five-year time period. Data were obtained from the Government Finances (U.S. Department of Commerce, U.S. Bureau of Census) for the years 1962-1996. Data included 36 states that implemented the lottery between 1964 and 1994. (2)

Percentage changes in total expenditures are computed for each state on an unadjusted basis and after subtracting the percentage change in combined state expenditures.

Unadjusted Percentage Change in State [Expenditures.sub.t] = (State [Expenditures.sub.t]/State [Expenditures.sub.t-1]) - 1

Adjusted Percentage Change in State [Expenditures.sub.t] = (State [Expenditures.sub.t]/State [Expenditures.sub.t-1])-(Combined State [Expenditures.sub.t]/Combined State [Expenditures.sub.t-1]),

where t = -1 ... 2, with 0 denoting the year in which the lottery was implemented.

Subtracting the percentage change in combined state expenditures adjusts for differences in nationwide economic conditions and inflation rates that occurred over the 35-year time period. Since spending generally increases, unadjusted percentage changes are expected to be positive and significant in all four years. Of interest is whether the increase in state spending, after adjusting for the nationwide increase in state spending, is higher than normal during the year in which the lottery is implemented. An increase in state spending that exceeds the national average is consistent with the hypothesis that implementation of a lottery increases agency costs in state government.

Some states may increase expenditures above the national average in every year because the state's growth rate exceeds that of the country. To account for differences in spending trends among the states, the percentage change in Adjusted State Expenditures in the implementation year (year 0) is compared to the percentage change in Adjusted State Expenditures in the three other sample years.

DIFFERENCE = Adjusted State [Expenditures.sub.0] - Adjusted State [Expenditures.sub.t],

where t=-1, 1, 2.

If implementation of a lottery results in above normal increases in state spending, Adjusted State Expenditures in the implementation year should be larger than in other years.

A logit regression determines if implementation of the lottery is a significant determinant of above normal increases in state spending. The logit regression is specified as follows:

ABOVE AVERAGE INCREASE IN STATE [SPENDING.sub.t] = a + [b.sub.1] COMBINED [EXPENDITURES.sub.t] + [b.sub.2] [LOTTERY.sub.t]

t = -1 ... 2, where

ABOVE AVERAGE INCREASE IN STATE SPENDING = Dummy variable assigned a value of 1 if the percentage increase in state spending exceeded the national average during that year. Dummy variable assigned a value of 0 if the percentage increase in state spending was below the national average.

COMBINED EXPENDITURES = Percentage change in expenditures for all states combined.

LOTTERY = Dummy variable assigned a value of 1 if the state implemented the lottery during that year, and 0 otherwise.

The sample time period consists of the four years surrounding the implementation of the lottery. Each of the 27 largest states in the sample is included for the year preceding implementation of the lottery, the year of implementation, and the two subsequent years for a total of 108 observations. The dependent variable is assigned a value of 1 if the percentage change in state expenditures exceeds the national average and a value of 0 if the change in state spending is below the national average. The lottery dummy variable is assigned a value of 1 during the implementation year, and 0 in the three other sample years. If the coefficient for the LOTTERY dummy variable is positive and significant, the increase in state spending is more likely to exceed the national average during the year in which a lottery is implemented than during other years.

IV

Empirical Results

TABLE 1 SUMMARIZES THE mean and median percentage changes in state spending during the four-year time period surrounding implementation of the lottery. As expected, the states report significant increases in expenditures on an unadjusted basis for all four years, regardless of whether the lottery was implemented or not. After adjusting for the percentage change in national expenditures, the median increase in state expenditures exceeded the national average by 1.67 percent in the year in which the lottery was implemented. Increases in state expenditures also exceeded the national average in the year preceding implementation of the lottery and were below the national average in the two years following.

Milgrom and Roberts (1992) and Shapiro (1973) state that agency costs, particularly those attributable to overspending as the result of special interest lobbying, increase with the size of the governmental budget. Costs associated with monitoring and organizing opposition impede taxpayer efforts to control excessive spending (Shapiro 1973). Costs of monitoring and controlling the actions of the agent increase with greater dispersion of the principals) As a result, agency costs are likely to be higher in large states and lower in small states. Results for the full sample are not statistically significant. However, when the nine smallest states (lowest quartile of state expenditures) are excluded, the results show a statistically significant median increase in Adjusted State Expenditures of 1.68 percent during the year the lottery was implemented. The mean increase of 1.59 percent is also statistically significant.

For both the full sample and subsample of 27 larger states, all mean and median differences in Adjusted State Expenditures between the implementation year and the three other years are positive. After considering differing growth rates of states, increases in state expenditures are above normal during the year in which the lottery is implemented.

During the year in which the lottery is implemented, increases in state expenditures are above normal after adjusting for national spending trends and trends in spending for individual states. The finding is consistent with the hypothesis that a state lottery increases agency costs, resulting in higher than normal state spending. On average, implementation of a state lottery results in above normal increases in state expenditures of 1.68 percent.

Table 2 reports the results of the logit regression that evaluates the determinants of above average increases in state expenditures. The dependent variable is assigned a value of 1 if the increase in state spending in that year exceeded the national average, and 0 otherwise. The coefficient associated with the lottery dummy variable is positive and significant at the 10 percent level. The result indicates that increases in state expenditures are more likely to exceed the national average in the year in which a lottery is implemented than in other years.

The magnitude of the percentage increases and the statistical significance reported in Tables 1 and 2 appear relatively small. However, the economic impact is more significant. A 1.68 percent above normal increase in state spending equates to $9.4 billion for the 37 states that have a lottery (based on 1998 expenditures). Total lottery net proceeds for 1998 were $12.2 billion, suggesting 77 percent of lottery net proceeds are used for above normal spending increases. Nonparticipants whose support for the lottery is based on the expectation of a reduced tax burden will find that the expected benefit is substantially dissipated by above normal spending increases.

V

Conclusion

SOME NONPARTICIPANTS OPPOSE LOTTERIES on moral or societal grounds but favor lotteries for the perceived personal benefit of a reduced tax burden. The principal-agent model and the empirical results of this paper suggest that implementation of a lottery is associated with higher than normal increases in state spending. Nonparticipants will find that the perceived tax benefit is largely diminished by the above normal spending increases.

If nonparticipants do not receive the tax benefit they expect, the question arises of why popular support for lotteries has not diminished. Public attitudes toward tobacco and alcohol may provide a useful analogy. Like lotteries, tobacco and alcohol are regarded as vices, and as a legally permissible choice of individuals. When the perception existed that costs were borne by participants with little effect on nonparticipants, public use of tobacco and driving after consumption of alcohol were tolerated by nonparticipants. When it was found that nonparticipants incurred significant costs through secondhand smoke and accidents caused by inebriated drivers, laws governing the use of tobacco and alcohol became more restrictive. The analogy suggests that support for lotteries will wane if evidence suggests that nonparticipants incur significant costs. This paper contributes to that debate with the argument that perceived tax benefits to nonparticipants are illusory.

References

Ang, James S., Rebel A. Cole, and James Wuh Lin. (2000). "Agency Costs and Ownership Structure." Journal of Finance 55: 81-106.

Borg, Mary O., and Paul M. Mason. (1988). "The Budgetary Incidence of a Lottery to Support Education." National Tax Journal 41: 75-85.

De Alessi, Louis. (1973). "Private Property and Dispersion of Ownership in Large Corporations." Journal of Finance 28: 839-851.

Fama, Eugene F., and Michael C. Jensen. (1983). "Separation of Ownership and Control." Journal of Law and Economics 26: 301-325.

Glaston, William A., and David Wasserman. (1996). "Gambling Away Our Moral Capital." Public Interest 123: 58-71.

Gulley, David O., and Frank A. Scott, Jr. (1989). "Lottery Effects on Pari-Mutuel Tax Revenues." National Tax Journal 42: 89-93.

Hersch, Philip L., and Gerald S. McDougall. (1989). "Do People Put Their Money Where Their Votes Are? The Case of Lottery Tickets." Southern Economic Journal 56: 32-38.

Jensen, Michael C., and William H. Meckling. (1976). "Theory of the Firm: Managerial Behavior, Agency Costs and Ownership Structure." Journal of Financial Economics 3: 305-360.

Madhusudhan, Ranjana G. (1996). "Betting on Casino Revenues: Lessons from State Experiences." National Tax Journal 49: 401-412.

Milgrom, Paul, and John Roberts. (1992). Economics, Organizations, and Management. Prentice-Hall.

Mikesell, John, and Maureen A. Pirog-Good. (1990). "State Lotteries and Crime: The Regressive Revenue Producer is Linked with a Crime Rate Higher by 3%." American Journal of Economics and Sociology 49: 7-19.

Price, Donald I., and E. Shawn Novak. (2000). "The Income Redistribution Effects of Texas State Lottery Games." Public Finance Review 28: 82-92.

Riverbank, William C. (1998). "The Tax Incidence of Casino Gambling in Mississippi." Public Finance Review 26: 583-598.

Shapiro, David L. (1973). "Can Public Investment Have a Positive Return?" Journal of Political Economy 81: 401-413.

Spindler, Charles J. (1995). "The Lottery and Education: Robbing Peter to Pay Paul." Public Budgeting and Finance 15: 54-62.

Stearns, James M., and Shaheen Borna. (1995). "The Ethics of Lottery Advertising." Journal of Business Ethics" 14: 43-51.

Stranahan, Harriet, and Mary O. Borg. (1998). "Horizontal Equity Implications of the Lottery Tax." National Tax Journal 51: 71-82.

Notes

(1.) At the extreme, agency costs include accepting bribes, supporting legislation that favors large campaign contributors, supporting self-serving legislation, and developing lucrative pension plans for state legislators.

(2.) Complete data were not available for New Mexico and the District of Columbia.

(3.) Although not directly addressing monitoring by voters, the corporate governance literature states that greater dispersion of stockholders (principals) reduces monitoring and increases agency costs (Fama and Jensen 1983; Jensen and Meckling 1976). Ang, Cole, and Lin (2000) find that agency costs increase with the number of nonmanager shareholders. De Alessi (1973) states: "More diffused ownership not only implies greater transaction costs, including collecting and disseminating information regarding the efficiency of managers' decisions, but also a smaller return to each stockholder from seeking to police inefficiencies. Among other results, the smaller would be the owners' incentive to police

the managers' contractual obligations and the greater would be the latter's opportunity for utility maximizing behavior."

Richard B. Whitaker, Professor Whitaker is an Associate Professor of Finance at Eastern Illinois University. His research interests include agency costs and the costs of financial distress. He has published empirical studies on the effect of financial distress and agency costs on firm performance.

Table 1
Percentage Changes in State Expenditures Upon
Implementation of a State Lottery

Table 1 reports the mean and median percentage changes in
state expenditures for the four years around the implementation
of a state lottery. UNADJUSTED is computed as the percentage
change in state expenditures from the previous year. ADJUSTED
represents Adjusted State Expenditures, which are computed as
the percentage change in state expenditures minus the
percentage change in expenditures for all states combined.
DIFFERENCE is computed as the Adjusted State Expenditures in
the implementation year (year 0) minus the Adjusted State
Expenditures in each of years -l, 1, and 2. Subscripts denote
the measurement year, with 0 being the year the lottery went
into effect. Statistical significance of means is determined
using the test of two means. Statistical significance of
medians is determined using the sign rank test

Percentage Change in State Expenditures

                         FULL SAMPLE           LARGE STATES
                           N = 36                N = 27
                      MEAN       MEDIAN       MEAN      MEDIAN

UNADJUSTED (-1)     10.77% **   10.53% **   11.29% **   10.75% **
UNADJUSTED (0)      10.74% **   10.46% **   11.22% **   11.43% **
UNADJUSTED (1)       9.28% **    9.53% **    9.91% **    9.76% **
UNADJUSTED (2)      10.89% **   10.80% **    9.94% **    9.19% **
ADJUSTED (1)         0.97%       0.26%       1.55%       1.23%
ADJUSTED (0)         1.14%       1.67%       1.59% *     1.68% *
ADJUSTED (1)        -0.06%      -1.11%       0.80%      -0.64%
ADJUSTED (2)        -0.17%      -0.83%      -0.83%      -1.23%
DIFFERENCE (-1)      0.17%       1.93%       0.04%       2.41%
DIFFERENCE (1)       1.20%       1.09%       0.79%       0.80%
DIFFERENCE (2)       1.31%       2.06%       2.42% *     2.52% *

** Significant at the 1% level.

* Significant at the 5% level.

Table 2
Determinants of Above Average Increases in
State Expenditures

Table 2 reports the results of a logit regression that evaluates
the determinants of above average increases in state expenditures.
The dependent variable is assigned a value of 1 if the
percentage change in state expenditures exceeds the national
average, and 0 otherwise. The independent variables are the
percentage change in total state expenditures, and a lottery
dummy variable assigned a value of 1 if the state implemented
a lottery in that year, and 0 otherwise. [X.sup.2] are reported in
parenthesis. N= 108

                              ESTIMATE

INTERCEPT                      1.3993
                              (5.326) **
COMBINED EXPENDITURES        -15.187
                              (6.5644) ***
LOTTERY                        0.785
                              (2.689) *

*** Significant at the 1% level.

** Significant at the 5% level.

* Significant at the 10% level.