Does Indian casino gambling reduce state revenues? Evidence from Arizona.
Congress passed the Indian Gaming Regulatory Act (IGRA) in 1988 after a landmark Supreme Court decision affirming California Cabazon Indians' legal right to engage in gaming. IGRA requires that federally recognized Indian tribes codify specific gaming agreements, called compacts, with their respective state governments (Sokolow, 1990). Gaming compacts specify the type of gambling activities, the allocation of machines, and the distribution of profits. IGRA divides gaming into three classes, as shown in table 1. Depending upon the existence of other forms of gambling in the state where a tribe is located, IGRA allows tribes to offer high stakes bingo, non-banking table games, and slot machines (Class III) casino style gaming on their reservations.
According to International Gaming & Wagering Business Magazine, Indian gaming is one of America's fastest growing and most profitable industries. Annual revenues from the 120 Indian casinos operating in 24 states are conservatively estimated at $5 billion. According to Goodman (1995), casinos may divert funds from taxable to non-taxable activities. This paper assesses the validity of his concern. Specifically, does the introduction of Indian casinos cause a structural change in state tax collections?
II. ARIZONA REVENUE SOURCES
State revenue sources in Arizona range from various taxes to traffic violation fees. The most significant source of revenue is the Transaction Privilege, Use, and Severance Tax (TPT). Arizona does not have a traditional sales tax, which is imposed on the buyer. Instead, TPT is levied on the seller for the privilege of engaging in business (State of Arizona, Joint Legislative Budget Committee, 1995). TPT accounts for more than 52% of total state tax revenue. The Income Tax, the second largest revenue source, accounts for more than 42%. Severance taxes are imposed, in lieu of a TPT, on mining (2.5%) and timber (1.5%).
TPT rates range from 0.46875% to 5.5%, depending on the type of business, with most rates at 5%. An additional tax usually considered to be part of TPT is the rental occupancy tax levied at the rate of 3% on real property on a long-term lease. Table 2 presents data on Arizona tax collections for FY 1994 and FY1995. The approximately 133,000 TPT accounts generated more than $2.53 billion in gross revenues in the FY1995. As table 2 shows, retail, contracting, utilities, and restaurants and bars constitute approximately 79% of consolidated tax collections.
TABLE 1 IGRA Types of Gambling Permits Class Description Class I Social Games for prizes of minimal value and traditional forms of Indian gaming engaged in as part of tribal ceremonies or celebrations. Class II Bingo and games similar to it, pull tabs, tip jars and certain nonbanking card games. Class III All other forms of gaming including banking card games, slot machines, craps, pari-mutuel horse racing, dog racing and lotteries.
Reductions in tax collections largely indicate changes in the law that modified or eliminated previous taxes. For example, in table 2, the 21.1% reduction in commercial lease taxes is due to a reduction in the tax rate from 4% to 3%, and various exemptions that are part of an overall effort to phase out of commercial lease taxes (State of Arizona Tax Handbook, 1995). For the last 10 years, TPT has been an increasingly important revenue source for local and state government. Recent changes in the law reducing both corporate and state income taxes will significantly increase the proportion of state revenue derived from the TPT.
The total of transaction privilege, rental occupancy, and severance tax base (TPTB) is divided into two parts, distribution base and nonshared. The distribution base is the portion of the tax shared with counties and municipalities. This amount is divided among municipalities (25%), counties (40.5%), and the state general fund (34.5%). The non-shared portion (of transaction privilege taxes, rental occupancy, and severance taxes) is deposited directly to the state general fund. In FY 95, the State distributed $356.3 million to counties, and $219.9 million to cities. Maricopa County, which includes Phoenix, received over 60% of the total amount distributed to counties (Arizona Department of Revenue Annual Report, 1995).
TPT collections from Maricopa County constitute a significant portion of total state revenue. In FY 95, Maricopa County collected 25% of total state TPT revenues. If Indian gaming "crowds out" consumption expenditures, then Maricopa County's TPT Base (TPTB) could decrease. Due to strong economic growth, however, sectors most likely to lose revenues to gaming have not shown a decline in TPT collections.
III. ARIZONA TAXES AND INDIAN RESERVATIONS
Indian tribes, as sovereign nations, do not pay taxes to states. Arizona Indians are exempt from state and local taxes (including TPT) imposed on consumer goods sold on the reservation unless they are imposed by tribal governments. Although the State is prohibited from taxing Indian lands and Indian owned property within the boundaries of reservations, the state taxes property and business transactions of non-Indians who operate on reservations. Until 1973, the state taxed the income of Indians who lived and worked on the reservation. In 1983, as a result of decisions in numerous cases,(1) Arizona issued a tax ruling (No. 2-I-83) preventing the State from collecting income taxes from Indians who live and work on a reservation, and who are identifiable and affiliated members of the tribe for which that reservation was established.
[TABULAR DATA FOR TABLE 2 OMITTED]
Even before legalized Indian gaming in Arizona, the issue of state taxes was hotly contested by tribes. Members of the legislature argue that casinos have a strong negative effect on the State's economy and should be taxed. In defending their highly profitable gaming operations, Arizona tribes make two arguments: first, that their sovereignty, based upon centuries of treaties and laws, should be respected. Second, that after generations of poverty, gaming provides the resources to improve schools, hospitals, medical care for the elderly, and other essential programs (The Arizona Republic, 1995).
IV. THE ECONOMIC SIGNIFICANCE OF INDIAN GAMING
With 21 reservations, Arizona has one of the largest concentrations of Native Americans in the United States. These tribes operate their own governments and courts in accordance with the Indian Reorganization Act of 1934. Like other tribes, Arizona's Indians seek to improve tribal members' standard of living through local economic development. In the past, tribes relied heavily on government grants, agriculture, forestry, tourism, and some light manufacturing. Despite these efforts, [TABULAR DATA FOR TABLE 3 OMITTED] many Arizona Indians are unemployed and living in poverty (Anders, 1996).
Before IGRA, several Arizona reservations supplemented their economies with bingo. Expansion into other gaming activities was outlawed by the Johnson Act of 1956, which prohibited any "gaming devices" on Indian reservations. Since 1993, when compacts were first ratified by the state (four years behind other states), casino gaming has become the most important economic activity on the reservation. Thus far, 16 tribes have been able to secure compacts with the state (see table 3). Two of the 16 tribes are located in Maricopa County, AK-Chin Indian Community and Fort McDowell Mohave-Apache Indian Community; the Gila River Indian Community is in close proximity.
Currently, both AK-Chin Indian Community (575 members) and Fort McDowell Mohave-Apache Indian Community (849 members) are authorized to operate 475 gaming devices. All are fully utilized. The Gila River Indian Community, with 11,500 tribal members, is authorized to operate 900 gaming devices. Currently only 772 machines and only two out of three authorized sites are being used. A new casino is under construction.
Financial information on Indian casinos is not publicly available. The Arizona Republic estimates annual revenues for Fort McDowell at $200 million and for Ak-Chin, $100 million (Wilson, 1995). Given a larger allocation of slots machines and two casinos, a conservative estimate of Gila River's annual revenues is approximately $250 million. With combined annual revenues of over $550 million, the four tribal casinos near Phoenix are a lucrative source of income for these tribes.
Since the opening of the casinos in late June 1993, the number of gamblers has increased dramatically. The clientele includes both permanent and seasonal residents ("snow birds"). Many retired people (including snow birds) participate through special travel buses that run regular schedules. Ak-Chin offers buffets and entertainment to attract seniors. This observation agrees with Thompson, Gazel, and Rickman (1995), who find that people 65 and over make up approximate two-thirds of the clientele in Wisconsin Indian casinos. The other two casinos, however, have a larger number of younger, more hard core gamblers, who probably also frequent Las Vegas, Reno, or Laughlin, Nevada. Slot machines, the principal money makers, are in constant use on all three reservations, often with lines of players anxious to try their luck.
Casino gaming is an economic activity that competes with a number of businesses. Consumers that incorporate gaming into their expenditures may reduce their purchases of other goods and services. This is particularly significant for state governments, because the state collects sales taxes on non-reservation purchases, but not on reservation sales, including gaming, food, and drink. If gaming results in diversions from taxable to non-taxable sectors, state resources will be reduced. Leakage of tax revenues may lead to reductions in expenditures on education, social services, economic development, and environmental protection.
It is important to point out that the Arizona Department of Gaming collects an assessment of $500 per year on each slot or video poker machine in Indian casinos. Given their tax exempt status, Indian casinos may divert revenue from taxable sectors of the economy, leading to lower TPT collections. An alternative hypothesis is that casinos actually have a positive or neutral impact on TPT revenues. This would be true if Indian casinos attract gamblers who would have otherwise spent their money in other out-of-state locations, or that the distribution of the gaming proceeds to tribal members has a very high multiplier effect.
V. ECONOMETRIC MODEL AND STRUCTURAL STABILITY ANALYSIS
A standard econometric model used to forecast state tax revenues follows:
(I) TPT = f(EMPL, RETAIL)
or, as a log-linear regression equation with time series data:
(2) [LTPT.sub.t] = [[Beta].sub.0] + [[Beta].sub.1] [LEMPL.sub.t] + [[Beta].sub.2] [LRETAIL.sub.t] + [u.sub.t]
LTPT = logarithm of Transaction Privilege Taxes
LEMPL = logarithm of employment
LRETAIL = logarithm of retail sales
u = a classical disturbance term
and the t subscript indexes month t.
Increases in employment and retail sales should increase tax revenues. Thus, the expected elasticity measures will be positive and significant ([[Beta].sub.1] [greater than] 0, [[Beta].sub.2] [greater than] 0). Although the parameter estimates are of interest, the main issue is whether the regression coefficients have changed over time and if so, when? One of the implicit assumptions in a standard regression equation is that the coefficients do not shift over time. That is, estimation of equation (2) is contingent on the assumption that the [Beta]'s are time invariant.
The usual practice in assessing the constancy of regression coefficients over time is to use prior information concerning the true point of structural change in the nature of the regression relationship. The analysis here identifies an event or set of events that is hypothesized to cause structural change, estimates separate regressions, and examines whether the multiple sets of estimated coefficients are significantly different from each other using an F-test. This is the so-called Chow (1960) test.(2) Using the standard approach, one can estimate equation (2) before and after the introduction of Indian gaming.
A third approach is to estimate the following model over the full sample period:
(3) [LTPT.sub.t] = [[Beta].sub.0] + [[Beta].sub.1] [LEMPL.sub.t] + [[Beta].sub.2] [LRETAIL.sub.t] + [[Beta].sub.3] CASINO + [u.sub.t]
where CASINO is a dummy variable which is equal to 1 after the introduction of Indian casinos in June 1993 and 0 otherwise. If the displacement hypothesis is valid, [[Beta].sub.3] will be negative and statistically significant.
It is important to note that both of these approaches require prior information regarding the event that is alleged to cause the structural change. Brown, Durbin, and Evans (1975) develop a test for the structural stability of regression parameters that does not require prior information concerning the true point of structural change. An analysis of the cumulative sum of squared residuals from the regression determines where, if at all, a structural "break" or shift occurs. An attractive property of the Brown-Durbin-Evans "cusum" test is that it allows the data to identify when the true point of structural change occurs. These tests have been employed on time series data to analyze the demand for money (see Heller and Khan, 1979) and to examine whether the returns to R&D investment vary by firm size (see Link, 1981; Lichtenberg and Siegel, 1991). In the context here, the null hypothesis of this test is that the regression coefficients are constant over the time period.
The basic intuition behind the Brown-Durbin-Evans test follows: If the structure of equation (2) varies according to an index of time, a shift in the residuals will result, as compared to the constant coefficients model. The Brown-Durbin-Evans test uses the test statistic [S.sub.r], which is derived from the normalized cumulative sum of squared residuals from a recursive estimation model:
[Mathematical Expression Omitted]
where [w.sub.i] are the orthogonalized recursive residuals, k is the number of regressors, and N is the number of observations.
[S.sub.r] has a beta distribution with expected value,
[Mu] = r-k/N-k.
With constant coefficients, a graph of [S.sub.r] will coincide with its mean-value line, within a confidence interval
[+ or -] [C.sub.0] + r+k/N-K,
where [C.sub.0] is Pyke's modified Kolmogorov-Smirnov statistic ([C.sub.0]). Durbin (1969) presents a precise definition of [C.sub.0]. Values of [C.sub.0] for the sample here are calculated from distributions of this statistic contained in Ben-Horim and Levy (1981). The actual and expected values of the test statistic, [S.sub.r] and E([S.sub.r]), are calculated for each observation. The absolute value of the differences between [S.sub.r] and E([S.sub.r]) also are computed. If the regression coefficients do not vary over time, then this difference will fall within the specified confidence region. The point at which the value of [S.sub.r] - E([S.sub.r]) exceeds [C.sub.0] identifies the occurrence of a structural change.
VI. DATA AND EMPIRICAL RESULTS
TPT is the dependent variable in our model of revenue formation. The ideal regressor in this model would be Indian casino gaming revenues. The three Indian tribes refused to provide any financial information on their casino operations. The Arizona Department of Gaming receives audited financial reports from the tribes but cannot release this information to the public. Because casino revenues are not available, studying the structural stability of tax revenues requires using an indirect approach.
Monthly data on TPT, employment, and retail sales for Maricopa County for July 1990 through December 1996 are analyzed. Each data series is seasonally adjusted based on the XII procedure in SAS. Standard tests of stationarity for each time series also are conducted. The first column of table 4 presents the point estimates of equation (2). The fitted model is:
[LTPT.sub.t] = 5.71 + .57[LEMPL.sub.t] (7.74) (6.40)
+ 1.14[LRETAIL.sub.t] (t-stats in parentheses) (5.36)
As expected, the elasticity measures are both positive and highly statistically significant. The model fits well, with an [R.sup.2] of 0.921. The use of monthly observations raises concerns about autocorrelation, or the instance where the error terms are serially correlated. These concerns appear unwarranted because the Durbin-Watson statistic (1.86) is insignificantly different from two.
The second column of table 4 displays the first piece of evidence regarding structural stability of the parameter estimates. Note that the coefficient on the dummy variable (-0.027) is negative and statistically significant, which implies that (holding employment and retail sales constant) the introduction of casinos reduces TPT revenues. Columns three and four of table 4 present coefficient estimates separately [TABULAR DATA FOR TABLE 4 OMITTED] for months preceding and following casino gaming, respectively. The estimates are markedly different, which is confirmed by the highly significant F-statistic (12.01) associated with the Chow test. More importantly, the elasticity estimates are both significantly lower in the period after Indian gaming was introduced. A lower [R.sup.2] (0.538 vs. 0.969) indicates an erosion in the explanatory power of the regression model.
While these results are illuminating, it is important to note that they are based on a prior identification of the point of structural change. The Brown-Durbin-Evans procedure permits examining the existence of structural stability without imposing this condition. Values of [S.sub.r] are calculated for each observation based on the residuals. Figure 1 plots values of [S.sub.r] - E([S.sub.r]) as differences over time measured in monthly intervals. In addition to the plot of the difference between [S.sub.r] and E([S.sub.r]), the graph includes two horizontal lines representing the 95% and 99% confidence regions, respectively. If the test statistic exceeds the confidence bound, structural stability at the respective level of significance can be rejected. That is, when the difference between [S.sub.r] and E([S.sub.r] exceeds [C.sub.0], a point of structural change has been identified.
Figure 1 reveals that structural stability can be decisively rejected. The confidence bound is broken during July 1993 (month 25) at the 95% level and August 1993 (month 26) at the 99% level. Recall that the introduction of Indian gaming in Maricopa County began in June 1993. This is a powerful result because the timing of the event is not imposed on the data by splitting the sample into "before" and "after." Instead, the model is estimated over the sample period, and the test statistic for structural stability is constructed from the residuals. Thus, the data indicate that the "event," or structural change, occurred in July 1993.
Descriptive statistics on the residuals before and after month 25 provide evidence on the change in sign. Before month 25, the mean and median values for the residual are +0.0022 and +0.0023, respectively.(3) After month 25, the mean and median values for the residual are -0.0022 and -0.0026, respectively.(4) These results imply that, after casinos came on-line, forecasted revenues exceed actual revenues. Interpreting the residuals from the logarithmic regression as percentage errors, these results associate a 0.44% decline in tax revenues with the introduction of gaming. The difference in mean residual is slightly higher than the standard error of the estimate, implying that in percentage terms, the errors induced by the "event" are quantitatively significant.
Note that these empirical results relate to total TPT revenue. Table 5 provides sectoral evidence on destabilization. Specifically, the econometric model and the associated Brown-Durbin-Evans structural stability tests are estimated for each major sector ($10M in annual revenue). (The parameter estimates for each sector are not reported, but are available upon request.) Four major sectors appear to experience revenue displacement: retail, restaurants and bars, hotel/motel, and amusements. Furthermore, most of these structural changes appear to occur almost simultaneously.
VII. POLICY IMPLICATIONS AND FUTURE WORK
Indian gaming is a complex and controversial issue. In effect, the IGRA created gaming monopolies that have enabled impoverished tribes to generate a substantial windfall. Although competition exists in the form of Nevada casinos, horse and dog racing, and the state lottery, Indian casinos in Arizona have captured a large share of the local gambling market. One must be careful about making generalizations. Many Indian casinos are much less profitable than the ones near Phoenix. Moreover, several tribes such as the Navajo and Hopi have voted not to engage in gaming.
One possible drawback of Indian gaming is that some of the rise in casino revenues may have come at the expense of other business establishments that pay TPT taxes. To examine this proposition, one can analyze the impact of gaming activity on the Transaction Privilege Tax Base, an important source of Arizona state revenue. Results suggest that the introduction of Indian gaming may have shifted consumer spending from taxable to non-taxable sectors. In Maricopa County, the revenue displacements are not currently a matter of great concern. This is because TPT has grown in the past few years due to economic growth, and increases in population (both permanent residents and "snow birds") and tourism. However, these findings imply that TPT growth may be masking revenue leakages.
The major implication of this study is that because of lost tax revenues, the State of Arizona may be justified in attempting to renegotiate gaming compacts with tribes to share casino profits. Of course, the tribes are adamantly opposed to what they see as a violation of their tribal sovereignty. But when one also considers the additional demand for police and [TABULAR DATA FOR TABLE 5 OMITTED] fire protection, economic losses due to compulsive gamblers, and the increase in negative externalities arising from gambling, the case for sharing casino profits with the state looks even stronger.
Several caveats and extensions must be mentioned. One problem is inherent to an "event study" approach. By definition, this approach attributes any changes in the regression relationship to the event in question. Other factors, such as a delayed effects from a change in tax laws or an unexpected change in the economy, also could cause a structural change in the model. Another concern is that the empirical analysis here is based on one county in one state. This raises concerns regarding external validity. A comparative multi-state study would address this concern. Obviously, this is a major task given the nature of the industry. Under IGRA, Indian casinos have no reporting requirements, except to state and federal commissions that by law must keep their reports confidential. Economists and students of public policy face a serious problem when attempting to understand IGRA. In many cases, tribes operate under a veil of secrecy reinforced by years of distrust and, now incredible wealth. Given the economic importance of this sector, it is critical that researchers have access to more data in order to understand the policy implications of IGRA.
IGRA: Indian Gaming Regulatory Act
TPT: Transaction Privilege, Use, and Severance Tax
This is a revised version of a paper presented at the Western Economic Association International 71st Annual Conference, San Francisco, Calif., July 1, 1996, in a session organized by Reza Rahgozar. The authors acknowledge the valuable comments and suggestions of Bill Eadington, John Navin, Farrokh Hormozi, Jonathan Silberman, Daniel Swaine, and the reviewers.
1. McClanahan v. State Tax Commission of Arizona (U.S. 1973) 93 S. Ct. 1257; Warren Trading Post Co. v. Arizona State Tax Commission (1965) 85 S. Ct. 1242. 380 U.S. 685, 14 L. Ed.2d 165; Central Machinery Co. v. Arizona State Tax Commission (U.S. 1980) 100S. Ct. 2592; Washington v. Confederates Tribes of Colville, 447 U.S. 134. 100 S. Ct. 2069 (1980); Ramah Navajo School Board, Inc. et al., v. Bureau of Revenue of New Mexico 73 L. Ed.2d 1174, 102 S. 3394 (1982). The Supreme Court decision in McClanahan (March 27, 1973) prohibits the state from levying a tax on the income of those Indians residing on reservations and whose income derived from reservation sources (Subcommittee Report on Taxation and Services to Arizona Reservation Indians, Arizona State Indian Seminar, 1973).
2. These tests were popular in the 1960s and 1970s when applied to empirical analysis of the demand for money. Typically, authors estimated whether changes in monetary regimes or catastrophic events, such as WWI or WWII, caused fundamental changes in money demand equations.
3. The sum of the residuals is +0.040.
4. The sum of the residuals is -0.040.
Anders, G. C., "Native American Casino Gambling in Arizona: A Case Study of the Fort McDowell Reservation," Journal of Gambling Studies, 12:3, 1996, 253-267.
Arizona Department of Economic Security, Research Administration Unit, Maricopa County Labor Force and Employment from 1988 to 1995.
Arizona Department of Revenue, Annual Reports, 1994, 1995, 1996.
-----, Transaction Privilege Tax Base for Maricopa County from FY-91 to FY-96.
The Arizona Republic, "State Has no Right to Funds," editorial, January 25, 1995, B4.
Ben-Horim, M., and H. Levy, Statistics: Decisions and Applications in Business and Economics, Random House, New York, 1981.
Brown, R. L., James, D., and J. M. Evans, "Techniques for Testing the Constancy of Regression Relationships Over Time," Journal of the Royal Statistical Society (series B), December 1975, 149-163.
Chow, G. C., "Tests of Equality Between Sets of Coefficients in Two Linear Regression," Econometrica, June 1960, 591-605.
Durbin, J., "Tests for Serial Correlation in Regression Analysis based on the Periodogram of Least squares Residuals," Biometrika, March 1969, 1-15.
Goodman, R., The Luck Business, Free Press, New York, 1995.
Heller, R. H., and M. S. Khan, "The Demand for Money and the Term Structure of Interest Rates," Journal of Political Economy, January 1979, 109-129.
Lichtenberg, F. R., and D. Siegel, "The Impact of R&D Investment on Productivity-New Evidence Using Linked R&D-LRD Data," Economic Inquiry, April 1991, 203-229.
Link, A. N., Research and Development Activity in U.S. Manufacturing, Prager, New York, 1981.
Siegel, D., and G. Anders, "Displacement Effects of Riverboat Gambling in Missouri: A Pilot Study," Paper presented at the American Economic Association Meeting, New Orleans, Louis, January 5, 1997.
Sokolow, G., "The Future of Gambling in Indian Country," American Indian Law Review, 15:1, 1990, 151-183.
Thompson, W., R. Gazel, and D. Rickman, "The Economic Impact of Native American Gaming in Wisconsin," Wisconsin Policy Research Institute Report, 8:3, April 1995, 1-43.
Wilson, S., "This Indian Casino Willing to Give State a Piece of the Pie," The Arizona Republic, January 22, 1995, A2.
Anders, Gary C.: Professor, Department of Economics, School of Management, Arizona State University West, Phone 1-602-543-6214, Fax 1-602-543-6221, E-mail email@example.com
Siegel, Donald: Associate Professor, Department of Economics, School of Management, Arizona State University West, Phone 1-602-543-6217, Fax 1-602-543-6221, E-mail firstname.lastname@example.org
Yacoub, Munther: Graduate Student, School of Management Arizona State University West
|Printer friendly Cite/link Email Feedback|
|Author:||Anders, Gary C.; Siegel, Donald; Yacoub, Munther|
|Publication:||Contemporary Economic Policy|
|Date:||Jul 1, 1998|
|Previous Article:||Illicit drug use, absenteeism, and earnings at six U.S. worksites.|
|Next Article:||How well do we measure employer-provided health insurance coverage?|