The demand for casino gaming with special reference to a smoking ban.

Thalheimer, Richard ; Ali, Mukhtar M.

I. INTRODUCTION

There is a sparse literature on the determinants of the demand for casino-style gaming. This study adds to that literature. A major focus of this analysis is on the effect on gaming demand of the introduction of a smoking ban into an existing gaming market, in this case the State of Delaware. A comprehensive statewide ban on smoking in public places was implemented in Delaware on November 27, 2002. The ban was titled the "Delaware Clean Indoor Air Act" (the Act). Gaming facilities open to the public were one of the locations specifically cited by the Act as subject to the ban.

The subjects of this analysis are the three Delaware racinos over the period September 1996 through December 2004. A racino is a pari-mutuel racetrack that has the statutory authority to offer electronic gaming devices (slot machines) to its customers. In 1990, there was only one racino in the United States, Mountaineer Park in West Virginia. By 2005, there were 13 states that permitted slot machine wagering at pari-mutuel racetracks. (1)

In some states, the electronic gaming devices at a racino are operated by the entity that owns the racetrack. In other states, the devices are operated under the auspices of the state lottery with the racetrack acting as a lottery agent. In such cases, the electronic gaming devices or slot machines are alternatively referred to as video lottery terminals (VLTs). Such is the case in the State of Delaware. (2) To show the importance of VLT gaming in Delaware, consider that in fiscal 2004 total state lottery revenue was $641 million. Video lottery revenue was $532 million or 83% of total state lottery revenue. The remaining 17% of total lottery revenue came from traditional on-line and instant lottery games (Delaware Lottery, 2004).

A system of three demand equations was developed to examine the effect of a set of determinants on slot machine wagering (handle) at the three Delaware racinos. The system is described by the seemingly unrelated regressions (SUR) model and was jointly estimated with the well-known seemingly unrelated regression estimation technique. VLT handle was found to be positively related to the number of VLTs, per capita personal income, and population. VLT handle at a racino was found to be negatively related to the number of VLTs at the other Delaware racinos. This negative relationship of the number of VLTs at other racinos to a particular racino's wagering demand indicates that they are substitutes. Gaming demand was found to decrease with the implementation of the Delaware statewide smoking ban.

The study proceeds as follows. Section II is a review of the literature on the demand for casino wagering. Also included is a review of the literature on the effect of a smoking ban on revenues of gaming and non-gaming establishments. Section III develops a three-equation system demand model, one equation each for the three Delaware racinos. Section IV reports the estimated models and presents empirical results on the factors affecting the demand for casino wagering. Section V summarizes the findings and draws further conclusions.

II. PREVIOUS STUDIES

Only two studies have examined the determinants of the demand for casino gaming. Thalheimer (1998) investigated the determinants of the demand for slot machine wagering at the first racino in the United States, Mountaineer Park, a thoroughbred racetrack in West Virginia. During the study period, 1990-1991, Mountaineer Park was permitted to offer a limited number of electronic gaming devices to its customers under the auspices of the state lottery on an experimental basis. VLT wagering demand was found to be positively related to the number of VLTs. The price of VLT wagering did not vary over the study period, and hence, its relationship to the demand for wagering could not be ascertained. Thalheimer and Ali (2003) investigated the determinants of slot machine wagering at 24 riverboat casinos and three racinos in the Midwestern states of Illinois, Iowa, and Missouri. The study period was from 1991 through 1998. Slot machine handle at a riverboat or racino was found to be positively related to the number of slot machines at the facility. Handle was found to be negatively related to the price of wagering at the facility and to the availability of competing gaming facilities.

Unlike the limited number of studies on the demand for casino wagering, there have been a number of studies that have examined the determinants of casino wagering revenue. Casino wagering revenue is total handle less the amount paid back as winnings to customers. Revenue is also referred to as win since it is the amount "won" by the casino after payback of winnings to the customer. Revenue and wagering (handle) are intimately related. As the price of casino wagering is win percent, defined as revenue from wagering (win) divided by total wagering, revenue is the product of wagering and the price of wagering. Prior studies of the determinants of casino revenue include those of Cargill and Eadington (1978), Nichols (1998a, 1998b), Hunsaker (2001), Levitzky, Assane, and Robinson (2000), Mandel, Alamar, and Glantz (2005a, 2005b), and Pakko (2005, 2006).

The studies of gaming revenue by Mandel, Alamar, and Glantz (2005a, 2005b) and Pakko (2005, 2006) included a smoking ban variable as a revenue determinant. In addition to the smoking ban variable, Mandel, Alamar, and Glantz (2005a) included the following gaming revenue determinants: quadratic trend, number of VLTs, income, and a seasonal dummy variable for winter. The introduction of the Delaware smoking ban was found to have had a positive but insignificant impact on aggregate Delaware VLT revenue and revenue per machine. In an erratum to this study, Mandel, Alamar, and Glantz (2005b) acknowledged that it contained a data error and subsequently found evidence of heteroskedasticity in the equation for total revenue. Correcting for the data error and adjusting for heteroskedasticity, they found that the introduction of the smoking ban had a negative but insignificant impact on total revenue.

Pakko (2006) reestimated the aggregate VLT revenue model used in Mandel, Alamar, and Glantz (2005a, 2005b) making adjustments to methodology and found the effect of the smoking ban to be negative and significant. In another article, Pakko (2005b) estimated a revenue model for each of the three Delaware racinos with a modified version of the model used in Mandel, Alamar, and Glantz (2005a, 2005b). In addition to the smoking ban variable, gaming revenue determinants included the following: quadratic trend, number of VLTs, index of economic activity, monthly dummy variables, and a dummy variable for a month in which a snowstorm occurred. The estimated impacts of the Delaware smoking ban on VLT revenue at each of the three Delaware racinos were found to be negative and significant.

There is an inherent difficulty in evaluating the results of the prior studies that have examined the impact of a smoking ban on casino revenue rather than on casino handle. The main reason for this is that revenue is affected through two sources: wagering and the price of wagering. Thus, factors, including the smoking ban, that affect wagering and those that affect the price of wagering must be the factors that affect revenue. To assess the impact of a smoking ban on revenue, one must include all these factors, both those that affect wagering and those that affect the price of wagering, in the revenue equation. In addition, racino revenue may also be affected by competition from other nearby gaming opportunities such as, in this case, casinos in Atlantic City, New Jersey, a racino in Charles Town, West Virginia, state lotteries in Delaware and surrounding states, and competing parimutuel facilities. Exclusion of any one of these relevant factors in the previous casino revenue models that included a smoking ban as a revenue determinant may result in a biased estimate of the smoking ban's effect on revenue.

While the effect of a smoking ban on the demand for casino gaming is the focus of this study, there are a number of past studies that examined the impact of a smoking ban on the revenues of restaurants and bars. Notable among these are Dunham and Marlow (2000a, 2003). Dunham and Marlow (2000a) examined the effects of smoking laws on revenues using a 1996 national survey of 1,300 restaurants and bars. Respondent owners were asked if revenue would be affected by the implementation of a virtual smoking ban. Six percent of restaurant owners expected revenue to rise, 39% expected revenues to fall, 51% expected no change, and 4% did not know. For bar owners, 2% expected revenues to rise, 83% expected revenues to fall, 13% expected no change, and 2% did not know. While a slight majority of restaurant owners expected no change, an overwhelming majority of bar owners expected a negative effect from a virtual smoking ban. The differential effects on revenue of an assumed virtual smoking ban are observed not only on the type of business, in this case restaurants or bars, but also between businesses, for example, restaurants and bars.

Following Dunham and Marlow (2000a), Dunham and Marlow (2003) analyzed the effect of a smoking ban on profits using a survey of restaurant and bar owners in Wisconsin in 2001. The study results confirmed the findings of Dunham and Marlow (2000a) that the effects of a smoking ban vary widely across establishments. Their analysis also suggests that restaurants that cater relatively more to smoking customers are expected to experience a greater loss when faced with a smoking ban.

In addition to the survey-based studies of the effects of a smoking ban on establishment revenues and profits, Dunham and Marlow (2000b, 2004) examined the effects of a smoking ban on nonsmoking seating allocations of restaurants and bars. Dunham and Marlow (2000b) reanalyzed the data used in Dunham and Marlow (2000a), while Dunham and Marlow (2004) reanalyzed the data used in Dunham and Marlow (2003). An important finding of these studies is that the firm's choice of the mix of smoking/nonsmoking seating is intimately related to the smoking preferences of customers, and this choice of seating allocation is consistent with the profit-maximizing behavior of a firm. As expected, given the smoking preference of the customer, converting seating for smoking use to nonsmoking use rewards nonsmokers, resulting in a gain in profits, while it penalizes smokers, resulting in a loss in profit.

A profit-maximizing firm chooses the allocation of seating, where at the margin the gain in profits matches the loss in profits. Thus, the optimal share of nonsmoking seating should be inversely related to the percentage of customers who smoke. Any deviation from this optimum allocation will result in a loss in profit for the firm. As the policy of a complete smoking ban reallocates all the smoking seats to nonsmoking use, it is expected that such a policy will result in a loss to a firm that is operating with the optimum or near-optimum allocation of seats as long as there are customers in the population who smoke. Furthermore, the loss in profit will be positively related to the percentage of customers who smoke. However, if the firm is operating with no seats allocated to nonsmoking use, it is not clear whether a comprehensive smoking ban policy will result in a gain or loss in profit to the firm. This is because both the allocation of (i) no seats to nonsmoking use and (ii) all seats to nonsmoking use are suboptimal, resulting in a reduction in profit from that which would result with optimal allocation. The magnitudes of these reductions cannot be predicted without knowledge of the extent of the suboptimality of these two allocations. However, as the optimal share of nonsmoking seats decreases with an increase in percentage of customers who smoke, it is expected that there will be a loss in profit with the policy of a comprehensive smoking ban if the percentage of customers who smoke is large.

The above analysis suggests that the policy of a comprehensive smoking ban will have differential effects on firms depending on each firm's pre-ban share of nonsmoking seats and the percentage of its customers who smoke. Furthermore, the loss in profit will be positively related to the percentage of customers who smoke. This observation is consistent with the findings of Dunham and Marlow (2000a, 2003), where it was found that revenue or profit loss is more likely among bar owners than among restaurant owners. It may also be concluded that revenue or profit loss will decrease as the percentage of smokers in the population decreases. For our study, each of the three establishments (racinos) sell the same product, that is, slot machine wagering, and are located in a relatively similar environment with regard to the percentage of smokers in the market area populations. To the extent that all three establishments did not allocate any seats for nonsmoking use before the policy of complete smoking ban was introduced, the policy of a comprehensive smoking ban is expected to affect the profitability of these establishments similarly. Unfortunately, the direction (positive or negative) of these effects cannot be predicted.

While Dunham and Marlow (2000a, 2000b, 2003, 2004) analyzed the impact of a smoking ban on individual establishments, a number of studies aggregated all restaurant, bar, and/or hotel business establishments into one "community-wide" impact. Almost all these studies (Glantz and Charlesworth, 1999; Glantz and Smith, 1994, 1997; Goldstein and Sobel, 1998; Huang et al., 2004; Hyland, Cummings, and Nauenberg, 1999; Sciacca and Ratliff, 1998), which used historical taxable sales, concluded that businesses, as an aggregate, do not suffer reduced sales as the result of a smoking ban. In light of the findings by Dunham and Marlow (2000a, 2003) of differential effects (both positive and negative) of a smoking ban on individual establishments, the finding of an insignificant aggregate effect is not a surprise. However, as our study deals with three individual establishments each selling the same product, that is, slot machine wagering, findings from these earlier studies are not relevant to evaluating the effect of a comprehensive smoking ban on these establishments.

III. DETERMINANTS OF THE DEMAND FOR VLT WAGERING

In this study, the demand for VLT wagering at each of the three Delaware racinos is determined. Demand determinants include the price and product characteristics of the subject facility, the price and product characteristics of competing product facilities, market area per capita income, population, and other external market environment conditions.

In addition to expanding the literature on the effects of traditional demand variables on wagering demand, a major focus of this study is on the effect of a smoking ban on slot machine wagering demand. While prior studies of the effects of a smoking ban on casino gaming addressed the relationship of a smoking ban to racino revenue, this study addresses the impact of a smoking ban on casino wagering demand (handle).

The study period for this analysis was chosen as the 100-month period from September 1996 through December 2004. The study period was chosen such that it began 1 mo following the start-up of the last racino to come on-line in Delaware, Harrington Raceway. As a result, there were an equal number of observations for each racino in the data set. The study period includes 25 mo following the introduction of the Delaware statewide smoking ban.

The demand variable is defined as total VLT wagering (handle). The price of VLT wagering is long-run average win, the amount retained by the racino after payback of winnings to customers, as a percent of total wagers (handle). The realized win percent for a month provides an estimate of this long-run win percent. Although there were small variations in the estimated win percent over the sample period, these were largely random variations and not likely to be associated with racino operator decisions to change the price. Thus, win percent was not included in the demand equations. Since an increase in the number of VLTs is expected to increase access to the gaming product and thus reduce the associated cost of waiting for an opportunity to play, the number of VLTs serves as a measure of some aspect of price (reward) of wagering. We have included this variable in the demand equations. Of course, the demand for VLT wagering is expected to be positively related to the number of VLTs.

Each of the three Delaware racinos faced competition from the other two Delaware racinos over the sample period. The demand for wagering at any one of the three Delaware racinos is expected to be affected by product characteristics of the other two measured, in this case, by the number of VLTs at the competing racinos. To the extent that the Delaware racinos are substitutes for one another, the demand for wagering at any one racino is expected to be negatively related to the number of VLTs at the other Delaware racinos. Only those substitute racino products that were found to be significant were included in the demand equations.

The Delaware racinos faced external market competition from casinos in Atlantic City, New Jersey, and a racino in Charles Town, West Virginia, over the study period. The win percent at each of these locations was unchanged over the study period. Thus, win percent at these locations was not included in the demand equations. The number of slot machines at each of these locations over the study period was not found to have a significant effect on handle at the Delaware racinos, and so these competition variables were dropped from the demand equations as well. (3)

The Delaware racinos faced pari-mutuel wagering competition from racetracks and/ or off-track betting facilities in the states of Maryland, New Jersey, New York, Pennsylvania, Virginia, and West Virginia. Pari-mutuel wagering, either simulcast or live and simulcast, was held at these facilities year-round. There was no change in takeout rate for the live races over the study period. (4) For this reason, the effect of pari-mutuel competition on Delaware racino handle could not be determined. In a prior study, Thalheimer and Ali (2003) found that competition from parimutuel wagering facilities did not have a significant effect on casino wagering.

Finally, even though the Delaware racinos were under the auspices of the Delaware state lottery, the Delaware racinos faced competition from traditional (non-VLT) Delaware lottery games. The Delaware racinos also faced competition from neighboring-state lotteries in Maryland, New Jersey, and Pennsylvania. The weighted average lottery takeout rate (100% less the percent of sales paid out in prizes), using total lottery sales in each state as the weighting factor, was computed for the Delaware state lottery and the three surrounding state lotteries. (5) The lottery takeout rate was not found to have a significant effect on handle at the Delaware racinos, and so this competition variable was dropped from the demand equations.

For the reasons given above, the external market competitors were not included in the demand equations.

VLT wagering demand for each racino is expected to be positively related to its market area per capita personal income and population. Following Thalheimer and Ali (2003), market area population for each of the three racinos was defined as the population in those counties that were located within a 100-mile radius of the racino. Following the analysis in Section II of this report, neither the direction nor the magnitude of the effect of the change in the external environment caused by government restrictions on smoking in public places, that is, the Delaware smoking ban, can be predicted. However, it is expected that the effect of a smoking ban will not vary across the three racinos. A dummy variable to represent this change in environment was included in the equations. In addition to the aforementioned variables, 11 monthly seasonal dummy variables were also included to capture seasonal variations, if any, in wagering.

Equations (1)-(3) show the specification of the demand for VLT wagering at each of the three Delaware racinos.

(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(2) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

(3) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]

where RHAND_DEL, RHAND_DOV, and RHAND_HAR are annual real VLT (slot machine) handles for the three Delaware racinos; FEB, MAR, APR, MAY, JUN, JUL, AUG, SEP, OCT, NOV, and DEC are binary (0, 1) seasonal dummy variables; VLTS_DEL, VLTS_DOV, are VLTS HAR are the number of VLTs at each of the racinos; VLTS_[DEL.sup.2], VLTS_[DOV.sup.2], and VLTS_[HAR.sup.2] are the square of the number of VLTs at each of the racinos; RPCI_DEL, RPCI_DOV, and RPCI_HAR are real per capita income for the three racinos; POP_DEL, POP_DOV, and POP_HAR are market area populations for the three racinos; SMOKEBAN is the fraction of the month over which the Delaware smoking ban was in place; and u is the overall error term. More detailed variable definitions, construction, and data sources are given in the Appendix. Summary statistics are given in Table 1.

In specifying the functional form of the slot machine handle equations, note that handle is by definition positive. A simplistic functional form that guarantees this positive condition is one that relates the logarithm of handle to its determinants. Following Thalheimer and Ali (2003), this functional form was chosen for the analysis. It is hypothesized that the change in VLT handle increases at a decreasing rate with respect to the number of slot machines. For this reason, the number of slot machines was specified in quadratic form.

IV. MODEL ESTIMATION AND ANALYSIS

The system of three demand equations, (1)-(3), was estimated using the SUR method (Zellner, 1962). The SUR method allows for possible correlations among the errors of these equations and therefore improves the estimation efficiency over that of estimating equation-by-equation using the ordinary least squares method. The estimated equations are given in Table 2.

The models fit the data well with [R.sup.2] greater than 0.78 in each case. Twenty-one of the 33 seasonal variables were significant at the 5% level or better. The coefficients of the linear component of own number of VLTs were positive as expected, significant at the 7% level or better in all three equations. The coefficients of the quadratic component of own number of VLTs were negative as expected and significant at the 5% level or better in all three equations.

The coefficients of the cross product (number of VLTs) variables in each equation were negative and significant at the 5% level or better. This shows substitutability among these products. The coefficients of per capita personal income were positive in each equation, as expected, significant at the 5% level or better in two equations and negative but insignificant in the other. The coefficients of population were positive, as expected, in all three equations and significant at the 5% level or better in one. Finally, the coefficients of the smoking ban variable were negative and significant at the 5% level or better in each of the three equations.

The introduction of restrictions on smoking in public places in Delaware, specifically at gaming facilities open to the public, was found to have had a large impact on the demand for slot machine wagering at the three Delaware racinos. Specifically, the impact of the smoking ban was found to have resulted in reductions in slot machine handle of 15.7%, 17.8%, and 12.7%, respectively, at Delaware Park, Dover Downs, and Harrington Raceway. (6) A standard chi-square test for equality of reductions across the three racinos showed that the hypothesis of equality cannot be rejected at any reasonable level of significance. (7) Thus, the reductions in wagering, as expected, do not vary significantly across the three racinos. The weighted average impact of the smoking ban over the three racinos was computed to be 15.9%. (8)

V. SUMMARY AND CONCLUSIONS

The estimated equations of the demand for VLT wagering at the three Delaware racinos fit the data well. The number of slot machines at a racino facility and the market area income were found to be significant demand determinants. The number of slot machines at competing in-state racino facilities was also found to be a significant demand determinant. Study results indicate that the Delaware racinos are substitutes for one another.

A major focus of this study was on the impact of a smoking ban on the demand for gaming. The introduction of restrictions on smoking in public places in Delaware, specifically at gaming facilities open to the public, was found to have had a large negative impact on the demand for VLT wagering at the three Delaware racinos. Specifically, the impact of the smoking ban was found to have reduced VLT handle for all three racinos but the impacts were not found to vary significantly across them. The weighted average reduction in VLT handle due to the smoking ban was found to be 15.9%. To the extent that the smoking ban was not fully enforced, its impact on VLT handle may have been underestimated. The findings of a significant and negative impact of a smoking ban on VLT handle (demand) are similar to the findings in Pakko (2005, 2006) with respect to the impact on VLT revenue and are in contrast to the insignificant negative impact on VLT revenue found in Mandel, Alamar, and Glantz (2005b). They are also in contrast to earlier findings of an insignificant impact of a smoking ban on taxable retail sales for various other businesses such as restaurants, bars, and hotels.

Following Dunham and Marlow (2000a, 2000b, 2003, 2004), it is expected that the effects of a smoking ban on businesses will decrease over time as the smoking prevalence of the population continues to decrease and customers are less likely to be affected by smoking ban laws. To illustrate the historical trend in smoking ban incidence, consider that the smoking prevalence of U.S. adults 18 years of age and older decreased from 42.4% in 1965 to 20.9% in 2004 (U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, 2006). If this trend continues, it is likely that, with or without smoking ban laws, the effects of smoking bans on business revenues and profits will decrease in the long run. This speed of this potential adjustment process was not a subject of this analysis.

ABBREVIATIONS

SUR: Seemingly Unrelated Regressions

VLT: Video Lottery Terminals

APPENDIX

VARIABLE DEFINITION, CONSTRUCTION, AND DATA SOURCES

RHAND_DEL, RHAND_DOV, RHAND_HAR

The gross annual wager on VLTs at a racino is termed handle. Handle was converted to its real dollar equivalent using the consumer price index. Data source for monthly VLT handle: Delaware Lottery.

FEB, MAR, APR, MA Y, JUN, JUL, A UG, SEP, OCT, NOV, and DEC

Binary (0, 1) variables for seasonality in the data. VLTS_DEL, VLTS_DOV, and VLTS_HAR

The number of VLTs at a racino. Data source for monthly number of VLTs: Delaware Lottery.

POP_DEL, POP_DOV, and POP_HAR

Population for the three racino market areas. Following Thalheimer and Ali (2003), market area population for each of the three racinos was obtained as the population in those counties, which were located within a 100-mile radius of the racino. Annual county population estimates for 1996 through 2004 were obtained from the U.S. Department of Commerce, Economics and Statistics Administration, Bureau of Economic Analysis, Regional Information System.

RPCI_DEL, RPCI_DO V, and RPCI_HAR

Per capita income of the racino market area. Per capita income was converted to its real dollar equivalent (RPCI) using the consumer price index. Annual county per capita income estimates for 1996 through 2004 were obtained from the U.S. Department of Commerce, Economics and Statistics Administration, Bureau of Economic Analysis, Regional Information System.

SMOKEBAN

The fraction of a month over which the Delaware smoking ban was in place. SMOKEBAN takes the value 0 prior to November 2002. The smoking ban was in effect beginning November 27, 2002, and so SMOKEBAN was assigned a value of 0.13 (4 d/30 d) for that month. For all months subsequent to November 2002, SMOKEBAN takes the value 1.

RICHARD THALHEIMER and MUKHTAR M. ALI*

* We would like to thank the two anonymous referees for their insightful comments, which have helped us improve our presentation and analysis. As is customary, the authors assume full responsibility for the analysis and the remaining errors.

REFERENCES

Cargill, T. F., and W. R. Eadington. "Nevada's Gaming Revenues: Time Characteristics and Forecasting." Management Science, 24, 1978, 1221-30.

Delaware Lottery. "The Delaware Lottery 2004 Annual Report For the Fiscal Year Ending June 30, 2004." Dover, 2004.

Dunham, J., and M. L. Marlow. "Smoking Laws and their Differential Effects on Restaurants, Bars, and Taverns." Contemporary Economic Polio,, 18, 2000a, 326-33.

--. "The Effects of Smoking Laws on Seating Allocations of Restaurants, Bars, and Taverns." Economic Inquiry, 38, 2000b, 151-7.

--. "The Economic Incidence of Smoking Laws." Applied Economics, 35, 2003, 1935-42.

--. "The Private Market for Accommodation: Determinants of Smoking Policies in Restaurants and Bars." Eastern Economic Journal, 30, 2004, 377-91.

Glantz, S. A., and A. Charlesworth. "Tourism and Hotel Revenues Before and After Passage of Smoke-Free Restaurant Ordinances." Journal of the American Medical Association, 281, 1999, 1911-8.

Glantz, S. A., and L. R. A. Smith. "The Effects of Ordinances Requiring Smoke-Free Restaurants on Restaurant Sales." American Journal of Public Health, 84, 1994, 1081-5.

--. "The Effects of Ordinances Requiring Smoke-Free Restaurants and Bars on Revenues: A Follow-Up." American Journal of Public Health, 87, 1997, 1687-93.

Goldstein, A. O., and R. A. Sobel. "Environmental Tobacco Smoke Regulations Have Not Hurt Restaurant Sales in North Carolina." North Carolina Medical Journal, 59, 1998, 284-7.

Huang, P., A. K. De, M. E. McCusker, and E. I. S. Officer. "Impact of a Smoking Ban on Restaurant and Bar Revenues--E1 Paso, Texas, 2002." Morbidity and Mortality Weekly Report, 53, 2004, 150-2.

Hunsaker, J. "The Impact of Riverboat Casinos on the Demand for Gambling at Casino Resorts: A Theoretical and Empirical Investigation." Managerial and Decision Economics, 22, 2001, 97-111.

Hyland, A., K. M. Cummings, and E. Nauenberg, E. "Analysis of Taxable Sales Receipts: Was New York City's Smoke-Free Air Act Bad for Restaurant Business?" Journal of Public Health Management Practice, 5, 1999, 14-21.

Levitzky, I., D. Assane, and W. Robinson. "Determinants of Gaming Revenue: Extent of Changing Attitudes in the Gaming Industry." Applied Economics Letters, 7, 2000, 155-8.

Mandel, L. L., B. C. Alamar, and S. A. Glantz. "Smoke-Free Law Did Not Affect Revenue from Gaming in Delaware." Tobacco Control, 14, 2005a, 10-2.

--. "Erratum to Mandel, L. L., Alamar, B. C., and Glantz, S. A. 'Smoke-free Law Did Not Affect Revenue from Gaming in Delaware' Tobacco Control 14 (2005), 10-12." Tobacco Control On-Line (Electronic Letter), May 23, 2005b.

Nichols, M. W. "Deregulation and Cross-Border Substitution in Iowa's Riverboat Gaming Industry." Journal of Gambling Studies, 14, 1998a, 151-72.

--. "The Impact of Deregulation on Casino Win in Atlantic City." Review of Industrial Organization, 13, 1998b, 713-26.

Pakko, M. R. "No Smoking at the Slot Machines: The Effect of Smoke-Free Laws on Gaming Revenues," Working Paper No. 2005-054B, Federal Reserve Bank of St. Louis, June (revised October), 2005, 1-8.

--. "Smoke Free Law Did Affect Revenue from Gaming in Delaware." Tobacco Control, 15, 2006, 68-72.

Sciacca, J. P., and M. I. Ratliff. "Prohibiting Smoking in Restaurants: Effects on Restaurant Sales." American Journal of Health Promotion, 12, 1998, 176-84.

Thalheimer, R. "Parimutuel Wagering And Video Gaming: A Racetrack Portfolio." Applied Economics, 30, 1998, 531-43.

Thalheimer, R., and M. M. Ali. "The Demand for Casino Gaming." Applied Economics, 35, 2003, 907 18.

U.S. Department of Health and Human Services, Centers for Disease Control and Prevention. "Smoking Prevalence Among U.S. Adults." Tobacco Information and Prevention Source (TIPS), October 2006.

Zellner, A. "An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias." Journal of the American Statistical Association, 57, 1962, 348-68.

Thalheimer: President, Thalheimer Research Associates, Inc., 107 West Short Street, Lexington, KY 40507. Phone 1-859-255-3073, Fax 1-859-254-8103, E-mail rthal@gte.net

Ali: Professor, Department of Economics, University of Kentucky, Lexington, KY 40506. Phone 1-859-2577636, Fax 1-859-323-1920, E-mail mmalil@uky.edu

(1.) The 13 states which permitted racino gaming in 2005 were Alabama, Arkansas, Delaware, Florida, Iowa, Louisiana, Maine, New Mexico, Oklahoma, New York, Pennsylvania, Rhode Island, and West Virginia.

(2.) In this study, the terms VLTs and slot machines are used interchangeably.

(3.) Data for the Charles Town racino slot machine win percent and number of machines were available from the West Virginia Lottery. Data for the Atlantic City casino slot machine win percent and number of machines were available from the New Jersey Casino Control Commission.

(4.) Data were available from the Maryland, New Jersey, Virginia, and West Virginia Racing Commissions, the Pennsylvania State Horse Racing Commission, and the New York State Racing and Wagering Board.

(5.) Data were available from the Delaware, Maryland, New Jersey, and Pennsylvania lotteries.

(6.) VLT handle impacts are computed using coefficients in Equations (1)-(3) as follows: Delaware Park, [exp(0t17) - 1]100; Dover Downs, [exp([[beta].sub.17]) - 1]100; Harrington, [exp(y17) - 1]100.

(7.) The Wald chi-square test with 2 degrees &freedom was 2.59 with p value of 0.2737.

(8.) The weighted average smoking ban impact is computed as each individual racino's smoking ban impact weighted by that racino's share of total Delaware 2004 VLT handle. Handle shares are 50.6%, 32.1%, and 17.3% for Delaware Park, Dover Downs, and Harrington Raceway, respectively.

TABLE 1
Summary Statistics

Variable Mean (100 Standard
 observations) Deviation

Common variables
 JAN 0.0800 0.2727
 FEB 0.0800 0.2727
 MAR 0.0800 0.2727
 APR 0.0800 0.2727
 MAY 0.0800 0.2727
 JUN 0.0800 0.2727
 JUL 0.0800 0.2727
 AUG 0.0800 0.2727
 SEP 0.0900 0.2876
 OCT 0.0900 0.2876
 NOV 0.0900 0.2876
 DEC 0.0900 0.2876
 SMOKEBAN 0.2513 0.4346
Delaware Park (DEL)
 1n(RHAND_DEL) 18.72 0.22
 VLTS_DEL 1,699 511
 [VLTS.sup.2]_DEL 3,145,140 1,681,579
 RPCI_DEL 18,850 763.8
 POP_DEL 19,100,000 384,046
Dover Downs (DOV)
 1n(RHAND_DOV) 18.18 0.33
 VLTS_DOV 1,705 5143
 [VLTS.sup.2]_DOV 3,170,159 1,679,492
 RPCI_DOV 18,218 772.3
 POP_DOV 14,500,000 270,944
Harrington Raceway (HAR)
 1n(RHAND_HAR) 17.53 0.28
 VLTS_HAR 1,016 345
 [VLTS.sup.2]_HAR 1,149,736 686,713
 RPCI_HAR 17,770 782
 POP_HAR 13,100,000 241,586

Variable Minimum Maximum

Common variables
 JAN 0 1
 FEB 0 1
 MAR 0 1
 APR 0 1
 MAY 0 1
 JUN 0 1
 JUL 0 1
 AUG 0 1
 SEP 0 1
 OCT 0 1
 NOV 0 1
 DEC 0 1
 SMOKEBAN 0 1
Delaware Park (DEL)
 1n(RHAND_DEL) 18.18 19.12
 VLTS_DEL 1,000 2,500
 [VLTS.sup.2]_DEL 1,000,000 6,250,000
 RPCI_DEL 17,106 19,786
 POP_DEL 18,500,000 19,700,000
Dover Downs (DOV)
 1n(RHAND_DOV) 16.94 18.68
 VLTS_DOV 527 2,500
 [VLTS.sup.2]_DOV 277,729 6,250,000
 RPCI_DOV 16,520 19,206
 POP_DOV 14,000,000 14,900,000
Harrington Raceway (HAR)
 1n(RHAND_HAR) 16.91 18.05
 VLTS_HAR 498 1,435
 [VLTS.sup.2]_HAR 248,004 2,059,225
 RPCI_HAR 16,116 18,902
 POP_HAR 12,700,000 13,400,000

TABLE 2 SUR: Dependent variable--RHAND

 Delaware Park

 Standard P > [absolute
 Coefficient Error Z value of Z]

CONSTANT 14.90 2.058 7.24 0.0000
FEB 0.0782 0.0519 1.51 0.1320
MAR 0.1961 0.0529 3.70 0.0000
APR 0.1507 0.0529 2.85 0.0040
MAY 0.1713 0.0531 3.23 0.0010
JUN 0.1478 0.0531 2.78 0.0050
JUL 0.1578 0.0530 2.98 0.0030
AUG 0.2240 0.0531 4.22 0.0000
SEP 0.0934 0.0514 1.82 0.0690
OCT 0.0966 0.0524 1.84 0.0650
NOV 0.0693 0.0525 1.32 0.1870
DEC -0.0282 0.0533 -0.53 0.5970
VLTS_DEL 9.082E-04 1.064E-04 8.53 0.0000
VLTS_[DEL.sup.22] -1.620E-07 1.280E-08 -5.25 0.0000
VLTS_DOV -2.194E-04 8.370E-05 -2.62 0.0090
VLTS_[DOV.sup.2]
VLTS_HAR
VLTS_HAR2
RPCI 1.553E-04 4.740E-05 3.28 0.0010
POP 8.730E-09 1.280E-07 0.07 0.9450
SMOKEBAN -0.1711 0.0482 -3.55 0.0000
Equation statistics
 Observations 100
 Parameters 17
 RMSE 0.1036
R.sup.2] 0.7818
 [chi square] 470.93
P 0.0000
 1.553E-04

 Dover Downs

 Standard P > [absolute
 Coefficient Error Z value of Z]

CONSTANT 3.13 3.558 0.88 0.3790
FEB 0.0963 0.0630 1.53 0.1260
MAR 0.2456 0.0640 3.84 0.0000
APR 0.1957 0.0640 3.06 0.0020
MAY 0.2130 0.0645 3.30 0.0010
JUN 0.2319 0.0648 3.58 0.0000
JUL 0.3149 0.0650 4.85 0.0000
AUG 0.3792 0.0650 5.83 0.0000
SEP 0.2532 0.0637 3.98 0.0000
OCT 0.2164 0.0648 3.34 0.0010
NOV 0.1778 0.0649 2.74 0.0060
DEC 0.0642 0.0663 0.97 0.3330
VLTS_DEL
VLTS_[DEL.sup.22]
VLTS_DOV 3.304E-04 1.781E-04 1.85 0.0640
VLTS_[DOV.sup.2] -1.620E-07 4.400E-08 -3.68 0.0000
VLTS_HAR -2.605E-04 9.020E-05 -2.88 0.0040
VLTS_HAR2
RPCI 5.016E-04 7.830E-05 6.41 0.0000
POP 4.120E-07 3.040E-07 1.35 0.1750
SMOKEBAN 0.1960 0.0647 -3.03 0.0020
Equation statistics
 Observations 100
 Parameters 17
 RMSE 0.1257
R.sup.2] 0.8515
 [chi square] 599.23
P 0.0000

 Harrington Raceway

 Standard P > [absolute
 Coefficient Error Z value of Z]

CONSTANT 8.99 2.962 3.04 0.002
FEB 0.106 0.0559 1.90 0.058
MAR 0.2316 0.0568 4.07 0.000
APR 0.1311 0.0570 2.30 0.021
MAY 0.1481 0.0573 2.59 0.010
JUN 0.1766 0.0573 3.08 0.002
JUL 0.0975 0.0574 1.70 0.089
AUG 0.2418 0.0574 4.21 0.000
SEP 0.1301 0.0558 2.33 0.020
OCT 0.1002 0.0569 1.76 0.079
NOV 0.0741 0.0571 1.30 0.194
DEC -0.0786 0.0581 -1.35 0.176
VLTS_DEL
VLTS_[DEL.sup.22]
VLTS_DOV -1.723E-04 7.900E-05 -2.180 0.029
VLTS_[DOV.sup.2]
VLTS_HAR 1.215E-03 2.503E-04 4.850 0.000
VLTS_HAR2 -2.610E-07 1.210E-07 -2.160 0.031
RPCI -6.550E-05 7.390E-05 -0.890 0.376
POP 6.870E-07 2.880E-07 2.380 0.017
SMOKEBAN -0.1354 0.0543 -2.490 0.013
Equation statistics
 Observations 100
 Parameters 17
 RMSE 0.1120
R.sup.2] 0.8344
 [chi square] 658.47
P 0.0000