Site hosted by Build your free website today!


An Assessment of the Indirect Effect of Casino Revenues on Federal Allocations For Education









The Mississippi Legislature adopted casino gaming in 1990 for the purpose of curing financial ills that have traditionally plagued the Magnolia state.  Local policy makers were given the opportunity to tax the casino industry at 3.2 percent, with an additional .8 percent if these local government stakeholders deemed it necessary to extract additional supplemental revenue from the casino industry.  One program designated as a beneficiary of this revenue-generating source was education.  This paper borrows research techniques from the lottery literature in an effort to measure the impact of casino gaming dollars indirect effect federal on per pupil spending in Mississippi.  The conclusions reached by this research suggests that the thirteen school districts receiving casino proceeds for education are witnessing an increase in federal per pupil expenditures due to outdated funding formulas for education.


The revitalization of Mississippi occurred in 1990 with the passage of the Mississippi Gaming Control Act, which authorized casino gambling in local communities that chose to adopt this revenue-generating device.  During the 1980s, Mississippi was faced with severe budgetary hardships.  Mississippi was operating on a budget of around $2 billion, which was not sufficient enough to cover all the expenses that the state was incurring, and they were forced to slash governmental programs.  The one program that received the largest cut was education. Because of Mississippi’s traditional lag in per pupil expenditures compared to other states over time, policy makers viewed additional supplemental revenue from casinos as a possible panacea for alleviating this legendary funding problem in Mississippi (Oliver, 1995).  This research fills a gap in the existing literature on casinos because no other quantitative empirical study deals specifically with the impact of casino dollars on federal allocations for education.  This research will address the following question: Do federal revenue patterns differ in school districts with casino tax revenue from similar school districts without casino revenue? 

Literature Review

State Supported Gaming In America

Public administrators and political functionaries, in Mississippi and other American states, experienced a most intense and challenging decade during the 1990s.  State governments witnessed a tremendous increase in demands on their governmental services, and an unprecedented number of un-funded mandates from the federal government, along with a tax- payer revolt (Ryen, 1992).  As the demand for social intervention programs increased, and the amount of available resources for funding these programs decreased, governmental officials used their ingenuity in generating revenue to offset the cost of running their government, its policies, and programs.  “Games of chance,” in one variation or another were the mechanism chosen by many state governments as their “economic savior” (Rivenbark and Rounsaville, 1995: p.3). 

One of the primary arguments used to rally support for legalized gambling is per pupil expenditures.  Numerous political and appointed bureaucratic functionaries stipulate that legalized gambling, whether lotteries, casinos, or other types of gambling, generate enough revenue to significantly enhance per pupil expenditures.  In theory, these government officials are asking the populace to invest in the future of their community and country by using gambling dollars to educate the younger generations.  The basic premise of their argument is that the education system in America is lagging behind most countries, and the only way America is going to compete in the global economy is by investing in the education of it’s children.  They paint a desperate picture that some sort of gaming device must be adopted in order to allow our children a chance for survival in the newly emerging global economy.  Many proponents of legalized gambling believe that the education system in America has traditionally been under-funded, and they view legalized gambling as a means to end this disparity (Alexander and Salmon, 1995).                     

            Legalized gambling is the mechanism used by policy makers to offset the cost of education for taxpayers.  The type of legalized gambling that is most popular is the lottery, but the literature suggests that lotteries have not turned out to be the panacea that policy makers had originally intended.  However, Florida and Georgia do represent specific cases where the lottery has displayed some success in funding education (Miller and Pierce, 1997). 

State Operated Lotteries

            Lotteries have proven to be appealing mechanisms for producing revenue because they are considered a voluntary tax: individuals pay the tax because they want to, instead of having to pay the tax because the government demands it.  The voluntary aspects of lotteries are extremely appealing to governors and legislators because resources for social intervention programs are generated without unpopular tax increases (Rubin, 1993).  Theoretically, legalized gambling intends to raise revenues without increasing the tax burdens of the lower class (Mikesell, 1989).     

            The most popular gambling device today is by far the lottery (Mikesell and Zorn, 1986).  The allure of lotteries and other forms of gambling as a source of revenue enhancement for state and local governments ascribes amply to the continued emergence of legalized gambling over the past two decades.  Currently, thirty-seven states and the District of Columbia operate lotteries, while other states debate their legalization.  Legalized gambling accounts for one of the fastest growing industries in the United States.  From 1982 to 1990, expenditures on legalized gaming increased at almost two times the rate of income; and by 1992, revenues from state sanctioned gambling operations averaged approximately $30 billion a year (Gross, 1998). 

            While lotteries are touted by many as a means of increasing funds for needy state programs, opponents contend that lotteries are not the panacea that policy makers and voters raved about.   Miller and Pierce (1997) examined the financial aspects of education lottery’s short-term and long-term effects.  They found that states that adopted lotteries increased spending on education per capita during the early years of the lottery, but as time passed, these same states witnessed an overall decrease in spending for education.  In turn, through pooled time series analysis, the authors were able to determine that states without lotteries actually increased their spending on education over time.  Four major problems permeate the literature on why lotteries, as a source for generating revenue are a “fiscal hoax. (p.34).”  They are: 1) lottery proceeds decline over time; 2) lottery dollars are actually shifted to other programs (fungibility); 3) lottery revenues are used to finance a tax cut; and finally 4) states with lotteries receive less federal funding for education compared to states without lotteries. 

            The second major problem with lotteries funding education is the idea of fungibility.  Spindler (1995) reinforces the notion of fungibility in reference to lottery dollars for education.  Spindler examines the lotteries of New York, New Hampshire, Ohio, Michigan, California, and Montana to determine their impact on educational revenue enhancement of public education expenditures.  Through ARIMA time-series modeling, the author successfully supports the notion that lottery revenues are fungible.  He attributes this fungibility to the “politics of the budgetary process” because education expenditures are highly visible to the public, and are plagued with fiscal and political restraints (p. 60).  Spindler contends that in states where lottery revenues are earmarked for education, revenues actually substitute for general fund expenditures.  Hence, Spindler concludes by postulating that state lotteries “are robbing Peter to pay Paul” (p.61).  

            Fields (1996) supports Spindler’s notion, and contends that the failure of Florida’s lottery in meeting everyone’s expectations of success expounds on the limitations of this revenue enhancing mechanism.  He points out that even though Florida’s educational system has received billions of dollars from lottery proceeds, the state legislature has taken non-lottery monies previously designated for education for the funding of other state commitments.  Public education’s share of the state budget in Florida has decreased more than 5 percent over the past decade since the lottery began in 1986 (National Education Association, 1997).  Even though revenues from lottery sales were intended to enhance the state’s educational system, the legislature was not legally bound to boost education with these profits. As a result, the earmarking of revenues from lotteries to replace regular, budgeted educational funds, instead of enhancing education, depicts Florida’s education policy. 

A third major problem with lotteries occurs when the proceeds are used to finance a tax cut.  Lotteries have proven to be appealing mechanisms for producing revenue because they are considered a voluntary tax: individuals pay the tax because they want to, instead of have to, pay the tax.  The voluntary aspects of lotteries are extremely appealing to governors and legislators because resources for social intervention programs are generated without unpopular tax increases, and in some cases tax cuts occur because a surplus of revenue exists from the lottery (Rubin, 1993). 

This is quite appealing to governors and legislators in their reelection bids for office.  Rodgers and Stuart (1995) stipulate that “the revival of lotteries,” despite immoral concerns and “negative distributional effects,” has occurred because of the belief that lotteries, instead of other tax instruments, raise additional revenue by generating smaller efficiency losses than other taxes; therefore, lotteries are less painful to voters (p. 244).  In turn, political leaders will endorse tax cuts and replace the lost revenue with lottery dollars.  Tax cuts are highly favorable political platforms used by incumbents for being reelected.  Unfortunately, many times social intervention programs, such as education, will be the first to suffer so politically ambitious individuals can maintain their tenure in politics (Jones and Amalfitano, 1994). 

As previously suggested by the literature, lotteries are failing to provide the anticipated revenue policymakers had hoped for.  Lottery revenues have attributed to a negative indirect effect on federal spending for education.  French and Stanley (2002) conducted empirical research measuring the impact of lottery proceeds on federal and state spending in southern states.  Through the use of pooled time series cross-sectional regression analysis, they found that per pupil expenditures in southern states with lotteries varied little in the amount of federal dollars appropriated for education, compared to southern states without lotteries.  The inference drawn from their research is that southern states utilizing lotteries dollars as a supplemental source of revenue for per pupil expenditures, witnessed no real increases in federal or state spending for education.  The scholars later employed this model to all fifty states, with similar results.  States with lotteries experienced no increase in federal per pupil expenditures for education, compared to states without lotteries (Stanley and French, 2002).  The authors hypothesized that states with lotteries should receive less federal funding for education because supplemental state funds from the lottery would increase, ultimately replacing federal dollars.  Therefore, federal dollars could be reallocated to states opting not to support the lottery as a supplemental source of revenue for education.  Stanley and French found that richer states tend to support state lotteries more so than less affluent states (southern states).  With disparities in per pupil expenditures for education quite substantial between affluent and less affluent states, one would anticipate more federal education dollars going to states without lotteries (Alexander and Salmon, 1995).  However, the authors found that federal allocations for education to states with and without lotteries have remained virtually the same.  The authors concluded in both articles that outdated federal funding formulas, used in allotting per pupil expenditures to the American states may serve as one explanation as to why funding disparities remain in the American states. 

The purpose of this research project is to employ the argument previously discussed by Stanley and French in reference to federal allocations for education to local school district utilizing casino revenues as a supplemental source of revenue.  This project will examine federal per pupil funding disparities between casino school districts in Mississippi with matching noncasino school districts.  This project will focus on the possibility that local school districts are experiencing the same educational funding disparities from the federal government as the American states. 



Do federal revenue patterns differ in school districts with casino tax revenue from similar school districts without casino revenue? 




H1: The federal government tends to allocate the same amount of federal dollars to school districts with casinos, compared to matching noncasino school districts.


Data & Methodology

Conceptual Definitions

Federal Spending Per Pupil by School District (Dependent Variable) – the amount

of spending per pupil by the federal government.


Casino Tax Revenue - the amount of revenue casino school districts in Mississippi

receive from the gaming tax placed on casinos.


Per Pupil Assessment Value Average Per Pupil Assessment Value based on Average Daily Student Attendance (measured in $100 thousand)[1]


Number of Students - the number of students in each Mississippi school district.


Millage Rates – the percentage of taxable income levied on real and personal property in each Mississippi school district.


Casino Presence – Dummy variable coded 0 = casino school districts; 1 = Non-casino school districts.



The model tested in this research project for empirical results using pooled time series analysis are: Total spending on per pupil expenditures for education. The time frame used in this analysis is eleven years: 1989 – 90 to 1999 – 2000 school year. 

Federal Spending Per Pupil by School District – Mississippi State Superintendent’s Report on Education


Casino Tax Revenue – Mississippi Department of Education, total casino spending on education


Control Variables

Per Pupil Assessment Values – Mississippi Report Card on Education, Mississippi

Department of Education


Number of Students – Mississippi Statistical Abstracts, Mississippi State University


Millage Rates - Mississippi State Superintendent’s Report on Education


Casino PresenceMississippi Report Card on Education, Mississippi Department of Education



The regression equation and formal model tested in this research project is as follows:


[Insert Table One]


Pooled time series cross-sectional data analysis is chosen as the advanced measuring device for testing the previously stated hypotheses (Sayers, 1989).  One of the most promising advantages of using pooled time series cross sectional analysis is its ability in offering explanations of the past, while simultaneously predicting the future behavior of exogenous variables in relation to endogenous variables.  Pooled time series cross sectional analysis allows the researcher to focus on more than one case in predicting social phenomenon, whereas simple time series analysis strictly deals with a specific case at different time points, causing data management complications.

Furthermore, accurate findings on the effects of casinos over a long period of time may fail to represent reality when multivariate regression models of data analysis are used to explain revenue development.  Spindler, writing on lotteries, noted that multivariate regression for data analysis in revenue enhancement over time possesses problems of high levels of multicollinearity between variables and collinearity in the time series (Spindler, 1995).  Spindler opted for an ARIMA model (simple time series analysis) of time series to correct for this problem.  ARIMA time-series methods of data analysis place an overwhelming emphasis on the burden of controlling for autocorrelation and heteroskedasticity to ensure data dependability.  Autocorrelation and heteroskedasticity do pose threats to data analysis, however, according to Beck and Katz (1996) they are more of a “nuisance” than a real threat when the N is larger than the T in the pooled time series regression model (p. 3).  ARIMA models of time-series analysis focus more on controlling autocorrelation and heteroskedasticity than discovering and explaining social phenomenon (McDowall et. al, 1980).

Despite the numerous advantages of pooled time series analysis using N (number of cases) at T (time points) for predicting the future of a particular social intervention program, a number of methodological disadvantages limit the usage of this data measuring device. The basic assumptions underlying traditional Ordinary Least Squares (OLS) regressions are violated in a pooled model, and such departures may exhibit severe consequences for the reliability of the estimators (Stimson, 1985). For instance, the following assumptions are usually made in regards to the error term in pooled time series regression.

1) The error term has a mean of zero,

2) The error term has a constant variance over all observations,

3) The error terms corresponding to different points in time are not correlated (Ostrom, 1978).


The accuracy of the regression model is inevitably measured by the error term.  Hence, if the standard error is small, then all of the sample estimates based on the sample size tend to be similar and are considered representative of the population parameters.  The exact opposite is true if the error term is large, then the statistics fail to represent the population parameters.  Of the previously mentioned assumptions, the error term corresponding to different points in time failing to correlate is the most important assumption violation.  When this violation occurs autocorrelation is present, creating estimators that negate true representation of social phenomena.  Autocorrelation violates an assumption of the regression model that the residuals are independent of one another. A lagged regression model relates a current endogenous variable to past values of the exogenous and endogenous variables, reducing the risk of autocorrelation.

            All the financial indicators in this data set were lagged one year (spending on education, casino proceeds, and per pupil assessment values). These predictors were lagged one year because the budgetary cycle used by the state and local governments are determined one year before the actual proceeds are allotted to specific programs.  Therefore, the availability of the data for the research is at least one year behind current figures.  Furthermore, the data has been adjusted for inflationary factors (Mississippi Department of Education, 2000). 

            By July 15 of each year, the school board must submit a formal request to the levying authority (county board of supervisors or city board alderman) for the ad valorem tax needs for the ensuing fiscal year.  The levying authority will determine the necessary millage rate to generate the ad valorem tax requested by the schools.  That millage rate will be in effect October 1 of that year because the fiscal school year for school districts in Mississippi begins on October 1 and runs through September 30. When considering the current school year (2000-2001), the millage is generally referred to as the 2000-2001 millage rates that are applied to the 1999-2000 assessment rolls.  Therefore, since millage rates are determined and set in the same year, this financial variable was not lagged.  Again, these efforts were made to prevent autocorrelation in the data set (Beck and Katz, 1996).

            A second major methodological problem with pooled time series cross-sectional data analysis is heteroskedasticity.  In pooled data, some units, for a variety of reasons, are inherently more various than others at all times.  Such differential variability is usually of modest concern in un-pooled data because it affects only a single case at a time.  In pooled data, however, it is likely to inflict a larger amount of harm to data sets.  For instance, basic size differences between units are one such endemic source of heterogeneity.  For example, Miller and Pierce (1997), when studying the effect of lotteries on education across the American states, found that the error terms for California and New York are more likely to be greater than those for New Hampshire and North Dakota.  This is simply because California and New York budgetary data used for the pooled time series analysis was larger than New Hampshire and North Dakota.  To account for the differences among states, intercepts for the cross-sectional unit are often employed. 

The standardization of the data set, according to Beck and Katz (1995), assists in alleviating heteroskedasticity problems in the data set.  Controlling for panel heteroskedasticity, Stimson contends (1985), and later Beck and Katz (1995 and 1996), argue that adding a dummy variable will assist in the prevention of this problem.  Stimson (1985) also stipulates that adding dummy variables to “shallow pools (small data sets)” will also control for autocorrelation.  Despite these concerns, this data set is not considered shallow because it is measuring 26 school districts over a period of eleven years (n = 286). 

In creating the pooled time series cross-sectional regression model, the dependent variable (total spending on education) was lagged one year to control for autocorrelation.  In using the Lagrange multiplier, the residuals are regressed from an OLS estimation of the equation on the first lag of those residuals, along with all the independent variables used in the OLS estimation.  The estimated coefficient on the lagged residual term yields an estimate of the remaining serial correlation of the errors.  If serial correlation remains present in the model, instrumental variables are employed as well as dummy variables.  Placing dummy variables in the equation offers explanations of the unexplained variance controlling for autocorrelation.  Therefore, the Durbin-Watson statistic was consulted to check for autocorrelation in the regression model. 

            The variance inflationary factor (VIF) checks for multicollinearity among the variables (a situation in the data set where two or more variables are highly correlated) in the regression equation.  Instead of, however, accepting the validity of this statistic on the assumption that SPSS is right, measures were taken to test for this statistical problem.  All the variables in the equation were regressed against one another to ensure that, according to Fox (1991), no variables indicated a VIF of 5.6 or more.  Fox further contends that it is more appropriate to view the tolerance levels instead of the VIF reports.  According to Fox (1991), if the tolerance variable reports levels of .9 and 5.6 (VIF) or less, multicollinearity is not a problem in the data set.  All the variables in this data set reported VIF statistics below 2.2 indicating that multicollinearity was not a problem in the data set.  Furthermore, White’s test for suggested that heteroskedasticity was not a problem.  In reference to autocorrelation, the Durbin-Watson M suggested that this methodological concern was not a problem in the data set either (Durbin, 1970).[2]

Twenty-six school districts were used as the units of analysis in this project (thirteen-casino school districts compared to matching non-casino school districts).  The comparison groups were chosen based on similar assessment values, student populations and per pupil spending for education.  Once the comparison groups were established a comparative means test was conducted on the two groups to determine there the mean differences.  The means test was conducted in 1994, the year prior to the allocation of casino revenue to school districts.  The means test reported a .8 percent difference between the two groups.  Since the results were less than 1 percent the year prior to the initial flow of casino revenue began matriculating to school districts the school districts were deemed similar enough to continue the study.  Following is a graph demonstrating the similarities between casino school districts and matching non-casino school districts before casino gaming revenues began finding there way to school districts in Mississippi.

According to the Mississippi Gaming Commission (2000), 1993 was the first year that casinos began contributing revenues to state and local governments.  However, according to the Mississippi Department of Education (2000) the first casino dollars used to fund education did not come until 1995. The comparison groups for the quantitative aspect of this project were chosen based on number of students, spending on education per pupil, and per pupil assessment values in 1994 (year before casino tax revenues were allotted to local school districts); (Mississippi Department of Education, 2000).  The comparison groups were chosen premised on previous studies conducted by the Mississippi Department of Education.[3]  These studies utilized a process for chosen comparison groups based on approximation ranges in number students, spending on education per pupil, and per pupil assessment values by each school district.  The range categories used in selecting the comparison groups were as follows: 1000 – 15,000 for number of students, $2,500 to $5,000 for per pupil expenditures, and $10,000 – $50,000 for the assessment value variable.  The casino school districts are in bold.

[Insert Table Two]


[Insert Table Three]

Even though the adjusted R2 shows that only four percent of the variance in the equation is being explained (.043), the significance value of the casino variable indicates that a significant relationships exists between casino tax revenue and federal spending on education.  The statistical information provided in table three suggests that for every unit increase in casino revenue an increase of .320 will occur in federal per pupil expenditures for education.  The negative relationship recorded by the dummy variables impact on the regression model, despites its insignificant p>. value, suggests that noncasino school districts in Mississippi are witnessing a slight decrease in federal allocations for education.  Therefore, the null hypothesis, the federal government tends to allocate the same amount of federal dollars to school districts with casinos, compared to matching noncasino school districts, is rejected.

What have we learned from these statistical tests?

            According to funding statistics provided by the Mississippi Department of Education for the fiscal year 1998-1999, the state provided 54.47 percent of education revenues while the federal government assisted with 13.87 percent, and the local government share was 31.66 percent.  The State of Mississippi currently uses what is called a “Minimum Foundation Program” as it’s funding formula for allotting state revenue to local school districts.  The formula is driven by the average daily attendance of students, along with calculating the number of teachers in each school district, to determine the amount of state funding each school district will receive.  The teachers considered in this calculation are those positions established as necessary for conducting the bare minimum requirements of teaching.  Also, school districts in Mississippi hire other teachers that are not covered under the Minimum Foundation Program.  Those excluded from the calculation are special education teachers, substitute teachers, and other teachers hired on a temporary basis. This formula has existed in Mississippi since 1953, and is currently in the process of being replaced.  Mississippi is slowly incorporating what is known as the “Mississippi Adequate Education Program” as its funding formula. 

This formula is also a type of Combination Foundation and Guaranteed Tax Based program, but accounts for more variables in determining the amount of revenue each local school district will receive from the state.  This funding formula is different than the Minimum Foundation Formula in several ways.  The most important difference between the two is that the Mississippi Adequate Education Program is premised on what is called cost-based education.  The Mississippi Adequate Education Program relies on data that assesses the initial cost of each school district in Mississippi and, from this analysis, determines the amount of revenue for education each district will receive from the state (Mississippi Department of Education, 2000).                    

            Currently under the Minimum Foundation Formula, the wealth of a school district is considered in the overall funding calculations used by the state.  One particular program, the Uniform Millage Assistance Funding Program, assists those districts in Mississippi that are considered poor in the amount of local revenues they spend on education.  However, this program is going to be repealed in 2002-2003 when the Mississippi Adequate Education Program (MAEP) is fully implemented.  As previously mentioned, the MAEP, which premises it’s allotments off educational costs, will take funding equity issues into consideration when dispersing state revenues among school districts.  In other words, for those local school districts that cannot place adequate revenues into their specific educational programs, the MAEP will provide the additional revenue to cover the expense. However, under MAEP, local school districts are required to maintain a certain tax rate.  For instance, the MAEP does require a minimum ad valorem tax contribution by the school district.  That contribution is the dollar value of 28 mills, or 27 percent of the cost of MAEP, whichever is less (Mississippi Department of Education, 2000).  

Federal funds appropriated to states and localities are determined by a formula similar to the funding formulas used by individual states.  In determining the amount of federal spending for education in each state, federal administrators use the following formula:  income of taxpayers + population of each state / number of students.  The formula is the number of dollars of revenue raised for each student from each $100 of income received by each member of the population.  In 1996, the national effort for elementary and secondary education was 23.5 percent, a slight decrease of 3.2 percent from 1994.  Federal spending per pupil was 20.6 in 1996, 10.7 points below it’s benchmark of 31.3 in 1966.  However, according to the Department of Education, federal education spending per student has been relatively stable since 1970, except for a drop in the early 1980s.  After remaining relatively stable during the 1980s, elementary and secondary public education revenue, as a percentage of Gross Domestic Product (GDP), rose between 1988 and 1992.  Higher education spending as a percentage of GDP has remained about 1 percent since 1970 (Department of Education, 1999).  Therefore, the federal government plays a role in the amount of revenue spent on education in local school districts. 


This research project suggests that school districts without casinos are witnessing less federal dollars for education, compared to matching casino school districts.  Despite the virtuous intentions of federal public educational administrators, educational expenditures from the federal government vary from year to year, again leaving states with the problem of adequately funding education.  With America’s emergence into a global economy, it is perilous for the United States to establish an education system second to none (Grissmer, Flanagan, and Williamson, 1997).  In order to establish this type of school system, states and local school districts must generate tax dollars to fund social intervention programs like education.


Alexander, Kern; Salmon, Richard G. 1995.  Public School Finance.  Boston:  Allyn and Bacon Publishing.


Beck, Nathaniel; Katz, Jonathan N.  1996.  “Nuisance Vs. Substance: Specifying and Estimating Time Series Cross Section Models.”  Political Analysis.  4: 1-37.


________.  1995.  “What To Do (and not to do) With Times Series Cross Section Models.”  American Political Science Review.  89: 634-47.


Brown, Neil M.; Kubasek, Nancy K. 1997.  “Should We Encourage Expansion of the Casino Gaming Industry?”  Review of Business.  Vol. 18, # 3, p. 9.


Durbin, John 1970.  “Testing for Serial Correlation in Least-Squares Regression When Some of the Regressors Are Lagged Dependent Variables,” Econometrica 38:  410-412.


Fields, Greg.  1996.  “Lottery No Education Panacea, Florida Finds.”  Knight-Rider.  p. 1-2.


Fox, John. 1991.  Regression Diagnostics:  Quantitative Applications In The Social Sciences.  California:  Sage Publication.


Franchkiewicz, Vic. 1993, The States Ante Up:  An Analysis of Casino Gaming Statutes.  Loyola Law Review.  Vol. 38, p. 1123-1157.


Gross, Meir.  1998.  “Legal Gambling as a Strategy for Economic Development.”  Economic Development Quarterly.  12: 203-211.


Jones, Paul; Holmes, Elizabeth; Garner, Larry; Perkins, William.  1999.  Tables Of Fortune:  Lost Hope… Lost Lives, Gambling In America.  Mississippi:  Thistle, LLC.


Jones, Thomas; Amalfitano, John. 1995. Public School Finance and State Lotteries.  Lancaster, PA.:  Technomic.


McClendon, McKee J. 1995.  Multiple Regression and Causal Analysis.  Illinois:  F.E. Peascock Publishers, Inc.


McDowall, David; McCleary, Richard; Meiginger, Errol; and Hay, Richard.  1980.  Interrupted Time Series Analysis.  California: Sage Publication.


Mikesell, John L.; Zorn, Kurt C.  1986.  “State Lotteries as Fiscal Savior or Fiscal Fraud:  A Look at the Evidence.”  Public Administration Review.  (July/August), p. 311-320.


Mikesell, John L 1989.  “A Note on the Changing Incidence of State Lottery Finance.”  Social Science Quarterly.  70: 513-521.


Miller, Donald E.; Pierce, Patrick A. 1997.  “Lotteries for Education:  Windfall or Hoax?”  State and Local Government Review.  29: 34-42.


Mississippi Department of Education.  2000.  Official Homepage. (


National Education Association.  1997.  “In Florida, The Bet’s OFF: If Your State Lawmakers Think A Lottery Alone Can Fund Schools, Pack’em Off To The Sunshine State.”  NEA Today.  16: 10-11.


Ostrom, Charles W. Jr.  1978.  Time Series Analysis: Regression Techniques.  California: Sage Publication.


Rodgers, William M.; Stuart, Charles.  1995.  “The Efficiency of a Lottery as A Source of Public Revenue.”  Public Finance Quarterly.  23:242-255.


Rubin, Irene 1993.  The Politics of Public Budgeting.  New Jersey:  Prentice-Hall Publishers.


Ryen, Dag 1992.  “The State Agenda for the Coming Years”.  The Journal of State Government. (April/June), p. 53-57.


Sayers, Lois W. 1989.  Pooled Time Series Analysis.  California: Sage Publication.


Spindler, Charles J. 1995.  “The Lottery and Education: Robbing Peter to Pay Paul?”  Public Budgeting and Finance.  3: 54-62.


Stanley, Rodney E.; French, P. Edward 2002.  “Can Students Truly Benefit From State Lotteries?  A Look AT Lottery Expenditures Toward Education In Southern States.”  Forthcoming publication by the Social Science Journal.


________ 2002.  “Lotteries, the American States and the Federal Government: A Formula For Perpetual Success or Inevitable Destruction in Education Policy?”  Forthcoming publication by Journal of Public Budgeting, Financial Management and Accounting.


________ 2002.  “The Lottery, Southern States and the Federal Government:  A Formula For Perpetual Success or Inevitable Destruction in Education Policy?”  Forthcoming publication by Public Administration Quarterly.


Stimson, James A.  1985.  “Regression in Space and Time: A Statistical Essay.”  American Journal of Political Science.  4: 915-947.


Von Herrman, Denise; Ingram, Robert; Smith William C.  2000.  “Gaming In the Mississippi Economy.”  Published by the University of Southern Mississippi.


Appendix One

Results of Statistical Tests for Heteroskedasticity & Autocorrelation


                                                       b                           st.e                     beta                           t                               p.

(Constant)                             45.722                     55.709                                                     .817                         .414

Population                            -1.575                      .001                         -.061                        -.858                        .392

Casino Revenue                   9.783                       .005                         .018                         .197                         .844

# of Students                        2.572                       .015                         .003                         .017                         .986

Millage Rate                          -.247                        .938                         -.018                        -.264                        .792

Assessment Value               -1.254                      .001                         -.084                        -1.189                      .236

Dummy                                  -.379                        .694                         -.073                        -.545                        .586



White’s Test For Heteroskedasticity

                Calculated Value: 5.004

                Critical Value: 15.067


[Table One]


Pooled Time Series CROSS-SECTIONAL Regression Equation


Y (FEDSPEDU)  = a +  (B1) CASINO t-1  +  (B2) PER PUPIL ASSESSMENT t-1  + (B3) NUMSTUD t-1  + (B4) MILLAGE t-1  + (B5) CASINO PRESENCE + E


Formal Model: 1989-1990 to 1999-2000


Per Pupil Assessment Values                                                            

            Number of Students                                                    Federal Per Pupil S.

            Millage Rates                                                            (Dependent Variable)

            Casino Tax Revenue                                                          

            Casino Presence





[Table Two]





School District

Number of Students

Spending On Education $

$ Per Pupil Assessment Values

Bay St. Louis




Benton County








Carroll County




Clarksdale City




Coahoma County




Greenville City




Gulfport City




Hancock County




Harrison County




Hattiesburg Municipal




Jackson County




Lee County




Leland City




Moss Point Municipal




Natchez-Adams County




Ocean Springs




Oxford Municipal




Pascagoula City




Rankin County




Tunica County




Tupelo City




Vicksburg City




Webster County




Western Line




Yazoo City







[Table Three]





                                         b                    st.e               beta                   t           p.



(Constant)                    -287448.7        309069.95                               -.930       .353

Population                    383.773           2129.414         .014                 .180        .857

Casino Revenue            .320                 .103                 .202                 3.112      .002

Number of Students      15.597             25.359             .046                 .615        .539

Millage             6356.190         5650.712         .069                 1.125      .262

Assessment Value         5.099               5.779               .055                 .882        .378

Dummy Variable           -4732.341        141388.17       -.002                -.033       .973

(0 = Casino Districts)

(1 = Non-Casino Districts)


R                                  .247

R2                                           .061

Adjusted R2                       .043

Df                                6

F                                  3.301

Sig. Of   F                    .004

N =                              286


Note: letters mean the following:

b- slope of the regression line

st.e – standard error

B – Beta

T – t-test

p>. Significance Value


Note: See Appendix for tests checking for Heteroskedasticity and Autocorrelation in the pooled time series cross-sectional regression models.



[1] However, due to the absence of a law stipulating the timing that school districts must re-assess land, the results of these statistical tests may be skewed.  Timing means that after 1992 school districts must re-assess 25 percent of their land every four years according to Mississippi law. Prior to the passage of this law, school districts were not required to re-assess 25 percent of their land every four years. 

[2] See appendix one for Durbin-Watson M and White’s tests for autocorrelation and heteroskedasticity.

[3] The variables per pupil assessment value, per pupil expenditures, and number of students are considered by education finance scholars as appropriate indicators on which to base comparison groups when studying the financial impact that a social intervention program (such as casinos) is having on education (Dr. Gary Johnson, January 8, 2001, 2:08 p.m.).  Charles Shivers, Director of Financial Accountability, Mississippi Department of Education notes that the Mississippi Department of Education, when conducting comparative studies, relies primarily on the project being studied.  In other words, if a study is concerned with financial matters then those indicators expressing financial data are most appropriate.   Mr. Shivers adds that the Mississippi Department of Education has used the following indicators in the past to determine comparative school districts in various studies: average daily attendance, 1st month enrollment, property per pupil assessment values, whether the districts have 16th section trust lands, whether they are municipal or county districts, or rural or urban, per pupil spending, and total federal spending.  Mr. Shivers endorses the indicators (population, per pupil assessment value and spending per pupil) chosen by this study for generating the comparative school districts that were studied (Charles L. Shivers, CPA, Tuesday, January 9, 2001, 3:28 p.m.).