Government restrictions and the demand for casino and parimutuel wagering

Similar documents
Proponents of expanding gambling in. tracks will save the horse industry by increasing i purses for horse races by

INITIATIVE FINANCIAL INFORMATION STATEMENT. Authorizes Miami-Dade and Broward County Voters to Approve Slot Machines in Parimutuel Facilities

Trends in Gambling Revenue to the States

THE ECONOMIC IMPACT OF THE IOWA RACE HORSE INDUSTRY ON THE IOWA ECONOMY

Danish gambling market statistics First quarter 2017

OVERVIEW OF PRESENTATION

Danish gambling market statistics Third quarter, 2017

New York Racing By the Numbers in 2005

Department of Legislative Services Maryland General Assembly 2007 Special Session FISCAL AND POLICY NOTE

For the First Time, a Smaller Jackpot

New York State Racing by the Numbers in 2008

Video Lottery/Tourism Promotion

STATE OF NEW YORK OFFICE OF THE STATE COMPTROLLER 110 STATE STREET ALBANY, NEW YORK September 2015

New York Racing By the Numbers in 2006

Danish gambling market statistics First quarter 2018

Pennsylvania Gaming Control Board Lawrence Downs Casino and Racing Resort Public Input Hearing

Availability of Legal Gaming in Canada As of July 2006 By: Rhys Stevens, Alberta Gaming Research Institute

Session of HOUSE BILL No By Committee on Commerce, Labor and Economic Development 2-12

Danish gambling market statistics Third quarter 2018

ARE THE NEW ENGLAND STATES IN A GAMBLING ARMS RACE?

Department of Legislative Services Maryland General Assembly 2012 Session

STATE OF NEW YORK OFFICE OF THE STATE COMPTROLLER 110 STATE STREET ALBANY, NEW YORK September 2015

Comments on Your Government

Department of Legislative Services 2012 Session

Casinos Newly Authorized in Five U.S. States: A Comparison of Legislative Objectives, Licensing Procedures and Actual Outcomes

New York Racinos: Casinos of Convenience

UCO. Department of Agricultural Economics. 0 t c (J. ~ 1991 WORKING PAPER SERIES. Agricultural fconomics Library

Back in the Black. States Gambling Revenues Rose in Introduction Gambling revenues to states rose modestly in fiscal 2010, HIGHLIGHTS

Revenue Estimating Conference Lottery Sales Executive Summary

Gamblers Favor Skewness, Not Risk: Further Evidence from United States Lottery Games

ARTICLE 14. CASINO SIMULCASTING

SENATE, No STATE OF NEW JERSEY. 217th LEGISLATURE INTRODUCED JANUARY 9, 2017

IC Chapter 7. Conduct of Gambling Games at Racetracks

The Economic Impact of Colonial Downs in Virginia

Indiana Gaming Tax Revenue. Prepared for the Gaming Study Committee Legislative Services Agency August 24, 2009

The New Jersey State Lottery has been in business

Revenue Estimating Conference Lottery Sales March 6, 2013 Executive Summary

William Crawford, Executive Director Ohio State Racing Commission

Impact of VLTs on Lottery Ticket Sales in New York

Effect of Proposed Lawrence Downs Casino and Racing Resort on Racing in Western Pennsylvania

For more than two decades, states saw lotteries and casinos as a bonanza of new dollars for education

Slot machines and slot machine components

Babson Capital/UNC Charlotte Economic Forecast. May 13, 2014

2010 Progress Report. a report to the legislature and our community

Revenue Estimating Conference Executive Summary for Lottery Ticket Sales and the EETF December 4, 2018

Working Draft: Gaming Revenue Recognition Implementation Issue. Financial Reporting Center Revenue Recognition

Revenue Estimating Conference Lottery Sales Executive Summary

Education Committee Economic Background and Issue Review

RTC TRANSIT OPERATING STATISTICS RTC RIDE RTC RAPID RTC INTERCITY SIERRA SPIRIT

SEMINOLE COMPACT , FS 1 Ratifies the 2015 Gaming Compact executed by the Governor- with required, specified amendments.

Yonkers Raceway may consider moving track as part of expansion

The 2019 Economic Outlook Forum The Outlook for MS

RTC TRANSIT OPERATING STATISTICS RTC RIDE RTC RAPID RTC INTERCITY SIERRA SPIRIT

Ontario Horse Racing Industry Performance Results [Fiscal Year Q2]

80th Oregon Legislative Assembly 2019 Regular Session PRELIMINARY

Original Article. Correlation analysis of a horse-betting portfolio: the international official horse show (CSIO) of Gijón

COPYRIGHTED MATERIAL. Industry Overview. Chapter 1. Gaming in the United States. Casino Gaming 2014, AICPA AAG-CAS 1.07

SECTION 1: NARRATIVE a. Current Law: Currently no video lottery terminals (VLT s) are allowed at licensed pari-mutuel facilities in Florida.

Utility Debt Securitization Authority 2013 T/TE Billed Revenues Tracking Report

ROYAL MONETARY AUTHORITY OF BHUTAN MONTHLY STATISTICAL BULLETIN

Southwest Ohio Regional Economy in Context. Richard Stock, PhD. Business Research Group

Ontario Horse Racing Industry Performance Results [Fiscal Year Q1]

ROYAL MONETARY AUTHORITY OF BHUTAN MONTHLY STATISTICAL BULLETIN

Revenue Estimating Conference Lottery Sales Executive Summary January 16, 2018

Maryland Thoroughbred and Harness Horse Racing Tracks

ROYAL MONETARY AUTHORITY OF BHUTAN MONTHLY STATISTICAL BULLETIN

Ontario Horse Racing Industry Performance Results [Fiscal Year Q3]

[First Reprint] SENATE, No STATE OF NEW JERSEY. 218th LEGISLATURE INTRODUCED SEPTEMBER 24, 2018

Sidoti & Company, LLC Fourteenth Annual Emerging Conference March 23 rd, 2010 New York City, NY

Key Issues Affecting the Future of Racing in New Jersey

REFERENCE 1. COM 2. FT&C 3. BILL NO. AND SPONSOR:

Revenue Estimating Conference Lottery Sales Executive Summary August 1, 2018

Enabling Legislation New York Thoroughbred Breeding and Development Fund

Ontario Horse Racing Industry Performance Results [Fiscal Year Q1]

IMPACT STUDY: West Virginia Greyhound-Racing Subsidies

Department of Legislative Services Maryland General Assembly 2007 Session FISCAL AND POLICY NOTE

Ticket Sales. Other Income. Unclaimed Prizes. Distribution to EETF from Lottery Receipts

THE FUTURE OF SALES TAX REVENUE

SUMMARY Revises provisions relating to pari-mutuel wagering. (BDR )

Economic Overview. Melissa K. Peralta Senior Economist April 27, 2017

Pennsylvania Gaming Trends

DOWNLOAD PDF TELEPHONE NUMBERS THAT PROVIDE THE WINNING NUMBERS FOR EACH LOTTERY.

REVENUE & RIDERSHIP REPORT SEPTEMBER 2018

Hearing on Keno Bingo & Slots IRS Auditorium, Internal Revenue Building, 1111 Constitution Avenue NW., Washington, DC. June 17, :00 a.m.

(a) Advertised Jackpot Prize- The estimated annuitized Jackpot Prize amount as determined by the Mega

Ch. 870 SUPER 6 LOTTO CHAPTER 870. SUPER 6 LOTTO

Agricultural Weather Assessments World Agricultural Outlook Board

Agricultural Weather Assessments World Agricultural Outlook Board

2017 Nebraska Profile

[First Reprint] ASSEMBLY, No STATE OF NEW JERSEY. 218th LEGISLATURE INTRODUCED DECEMBER 10, 2018

NEVADA SLOT MACHINES: HISTORICAL HOLD PERCENTAGE VARIATIONS ANNUAL AND MONTHLY HOLD PERCENTAGES, CENTER FOR GAMING RESEARCH, NOVEMBER 2017

S 0037 S T A T E O F R H O D E I S L A N D

11/2/2015. The Horse Racing Industry in Delaware. A Little History on the Sport. More. Mark Davis

Session of HOUSE BILL No By Committee on Federal and State Affairs 2-8

Proposed Changes to 205 CMR 6.00 Pari-mutuel rules for thoroughbred racing, harness racing, and greyhound racing.

Department of Legislative Services Maryland General Assembly 2005 Session FISCAL AND POLICY NOTE

HORSE RACING AND CALIFORNIA FAIRS

Livestock and Dairy Market Outlook

Tabcorp Holdings Limited. Citigroup Wagering Forum 9 November 2011

MISO Energy and Peak Demand Forecasting for System Planning

Transcription:

Applied Economics, 2008, 40, 773 791 Government restrictions and the demand for casino and parimutuel wagering Richard Thalheimer Thalheimer Research Associates, Lexington, KY 40507, USA E-mail: rthal@gte.net Growth in casino wagering in the United States from the mid-1980s forward has been extraordinary. Over the same period, however, parimutuel wagering has declined. Concern for this decline has prompted a number of states to allow parimutuel racetracks to offer casino-type gaming devices at their facilities. Such operations are commonly referred to as racinos. If the gaming devices at a racino are under the auspices of a state lottery, they are referred to as video lottery terminals (VLTs). Separate parimutuel and VLT wagering demand models were estimated for a racino facility. A number of restrictions were imposed on the VLT s at the beginning of the study period including: number of VLT s, type of game and machine, maximum bet per play and VLT location. The effect of these restrictions on VLT and parimutuel handles was a major focus of this article. Also of importance was an examination of the relationship between the VLT and parimutuel products. Relaxation of government restrictions on the VLT s was found to have resulted in a large increase in VLT wagering. On the other hand, parimutuel wagering was found to decrease with the relaxation of restrictions on the VLT s. The presence and growth of the VLT product was found to decrease parimutuel handle. The presence and growth of the parimutuel product was found to increase VLT handle. I. Introduction The casino gaming and parimutuel wagering industries have exhibited markedly different growth rates over the past several decades. The growth of casino gaming in the United States from the mid-1980s forward has been extraordinary. To illustrate, consider that over the two-decade period, 1982 to 2002, casino-and-related wagering 1 increased 636%. Over that same period, parimutuel wagering increased only 25%. Real growth in casino-and-related wagering was 301% while parimutuel wagering declined by 32% (Christiansen, 1989; Christiansen Capital Advisors, LLC, 2004). The increase in casino wagering has occurred largely due to the introduction of new gaming venues and locations augmenting the previously existing land-based casino gaming venues of Nevada and Atlantic City, New Jersey. The new gaming 1 Includes land-based casino gaming, riverboat casino gaming, Indian casino gaming, legal bookmaking, card rooms and charitable gaming. Applied Economics ISSN 0003 6846 print/issn 1466 4283 online ß 2008 Taylor & Francis 773 http://www.tandf.co.uk/journals DOI: 10.1080/00036840600749672

774 R. Thalheimer venues include riverboats, land-based casinos, Indian casinos, and slot machines or card rooms at parimutuel racetracks. Reasons for the decline in parimutuel horse race wagering are well-known and include increased competition from casino gaming (Thalheimer and Ali, 1995a; Ali and Thalheimer, 2002) and state lotteries (Simmons and Sharp, 1987; Vasche, 1990; Thalheimer and Ali, 1992, 1995b, 1995c; Ali and Thalheimer, 1997). Not only do growth rates between casino gaming and parimutuel wagering vary greatly, but there are marked differences in the size of each of these gaming industry sectors. For example, in 2002 casino-and-related handle was $812.2 billion and parimutuel handle was $17.9 billion (Christiansen Capital Advisors, LLC, 2004). Concern for the decline in handle and revenues for the parimutuel racing industry has prompted a number of states to allow parimutuel racetracks to install casino-type gaming devices at their facilities. Parimutuel racetracks that are permitted by state statute to offer such gaming devices are commonly referred to as racinos. In some states these gaming devices are part of the state lottery and are placed under the lottery s regulatory authority. In those instances, the gaming devices are referred to as video lottery terminals (VLT s). In other states where electronic gaming devices are permitted at parimutuel wagering facilities, the privately owned devices are placed under the regulatory authority of a state racing or gaming commission and are referred to as electronic gaming devices (EGD s) or slot machines. In either case VLT s or EGD s are equivalent casino-type gaming devices, differing only in regulatory accountability but not in substance. In almost all instances where racino gaming is permitted in a state, a percent of the gaming revenue is statutorily allocated to purses for owners of horses (greyhounds) racing at those racetracks in an attempt to slow down or reverse the decline in the racing industry due to competition from other gaming venues. The first racino in the United States was Mountaineer Park, a small thoroughbred racetrack in northern West Virginia. On 9 June 1990 a limited number of video gaming devices was installed at Mountaineer Park racetrack under provisions of a nonstatutory pilot project involving the West Virginia Lottery and the racetrack. Under terms of the agreement, Mountaineer Park, Inc. acted as agent for the Lottery. Following this initial experiment, full-scale, racino gaming authorized by state statute was permitted in the states of Louisiana (2002), 2 Rhode Island (1992), West Virginia (1994), 3 Delaware (1995), Iowa (1995), New Mexico (1998) and New York (2002). 4 Gaming devices at racetracks in West Virginia, Rhode Island, Delaware and New York are under the auspices of the state lottery while racinos in the remaining states are regulated by state gaming commissions. 5 To illustrate the importance of VLT gaming to total state lottery revenue consider that in fiscal year 2002 total West Virginia lottery revenues were $728.4 million. Racetrack video lottery revenues generated by all four West Virginia racinos accounted for $596.0 million or 81.8% of that total (West Virginia Lottery, 2003). In this study the demand for VLT and on-track parimutuel wagering at a racino location, Mountaineer Park Racetrack and Resort (Mountaineer Park), is determined. In March 1994, The West Virginia Video Lottery Act was passed giving statutory authority to the West Virginia Lottery to conduct VLT gaming at all four (two thoroughbred and two greyhound) of the West Virginia parimutuel racetracks subject to local referendum. Following passage of the required local referendum approving the use of VLT s at parimutuel racetracks, Mountaineer Park on 2 September 1994 ended its pilot project program which had been judged to be a success. In this study, the estimation period is from 1994 to 2002, a period which included both pre-pilot project and post-pilot project VLT gaming. Under provisions of the West Virginia Lottery Act, a number of restrictions were imposed on the VLT s. These included the number of VLT s, the type of game and machine, the maximum bet per play and VLT location. The effect of these government restrictions on VLT and parimutuel handles is a major focus of this article. Also of importance is an 2 Limited stakes poker machine gaming devices were permitted at racetracks and other locations such as truck stops, restaurants and lounges in 1992. Legislation resulting in casino style slot machines, restricted to racetracks, resulted in full-scale racinos beginning with Delta Downs racetrack in 2002. 3 Video lottery terminal gaming was offered as permitted by state statute in March 1994, and after local referendum beginning in September 1994. At that time, three of the four West Virginia parimutuel racetracks, Mountaineer Park, Wheeling Downs and Tri-State, became racinos. Charles Town installed VLT s in September 1997 after local referendum. 4 Although, New York passed enabling racino legislation in 2002, racinos did not come on-line in that state until 2004. 5 In 2004, racinos were authorized but not yet implemented in the states of Maine, Oklahoma and Pennsylvania. The gaming devices were not under the state lottery in any of these states.

Demand for casino and parimutuel wagering 775 examination of the relationship between the VLT and parimutuel products. The plan of this study is as follows. Section II reviews the existing empirical literature on the demand for racino gaming. Section III develops two econometric models of the demand for racino gaming, one for VLT wagering and one for parimutuel wagering. Section IV reports the estimated model and presents empirical results on the factors affecting the demand for racino wagering. In addition to reporting results for traditional demand variables, the effects of government restrictions over the estimation period are reported. Section V summarizes the findings and draws further conclusions. II. Previous Studies Prior econometric models of the determinants of the demand for parimutuel horse race wagering include those of Gruen (1976), Coate and Ross (1974), Suits (1979), Morgan and Vasche (1979, 1982), Simmons and Sharp (1987), DeGenarro (1989), Church and Bohara (1992), Thalheimer and Ali (1992, 1995a, 1995b, 1995c) and Ali and Thalheimer (1997, 2002). Unlike studies of the demand for parimutuel wagering, there have been few econometric studies of the demand for casino gaming. Cargill and Eadington (1978) developed a quarterly forecasting model of casino win (revenue). Nichols (1998b) used a Box Jenkins time series intervention model to investigate the effect of deregulation of gaming floor space on slot machine and total win at the Atlantic City, New Jersey casinos. The relaxing of floor space restrictions was found to have a significant positive effect on win. Nichols (1998a) developed regression models for win, win per admission and admissions for riverboats in Iowa. Deregulation was found to have a significant positive effect on casino win at Iowa riverboats. Hunsaker (2001) developed separate models of casino win in Atlantic City, New Jersey and the Las Vegas, Nevada Strip and their relationship to riverboat win in other states. An excellent review of various economic issues affecting the casino gaming industry can be found in Eadington (1999). None of these models focused on the economic behaviour of customers with regard to wagering demand, rather they were reduced form equations designed to either forecast or assess the impacts of various factors such as government regulations on gaming revenue (win). While previous studies examined casino win or revenue, the only study of the demand for casino wagering was that of Thalheimer and Ali (2003). The subjects of the analysis were all riverboats and racinos in the Midwestern states of Illinois, Iowa and Missouri. Specifically, an econometric model was developed relating the demand for slot machine wagering to its determinants. One important finding of that study was that government restrictions on bet size/total wager, and restrictions on customer access resulted in significant reductions in slot machine handle. Only one previous study examined the demands for parimutuel and VLT (slot machine) wagering at a racino location (Thalheimer, 1998). In that study, separate econometric demand models were specified for VLT gaming and on-track parimutuel horse race wagering. The subject of the analysis was Mountaineer Park. The period of analysis was daily from 1989 1991 during which a limited number of VLT s were placed at the racetrack as a pilot project under the auspices of the West Virginia State Lottery. It was found that the introduction of VLT gaming at a racetrack caused a significant decline in on-track parimutuel wagering. However, VLT revenue was found to be sufficient to offset the decline in on-track parimutuel revenue. III. Data Description and Model Specification The Mountaineer Park Racetrack and Resort racino in West Virginia is located near the Ohio and Pennsylvania borders. The market area for Mountaineer Park is defined as including those potential customers who reside in counties lying within a 100-mile radius of the racino. This market area definition was used in an earlier study of the demand for casino gaming (Thalheimer and Ali, 2003). Market area population averaged 8.47 million over the first 12 months of the study period, falling slightly to 8.39 million in 2002. In 2002, the market area population was distributed as follows: 4.72 million in Ohio, 3.23 million in Pennsylvania and 0.44 million in West Virginia. During the study period Mountaineer Park included a hotel, Mountaineer Lodge, with 101 rooms and a restaurant, located at a short distance from the racetrack. In addition to Mountaineer Lodge, a new hotel was opened near the end of the study period in 2002. Video lottery gaming and wagering on the simulcasts of races from off-site racetracks were available both at the racetrack and lodge. In the previous study for which Mountaineer Park was also the subject of analysis (Thalheimer, 1998)

776 R. Thalheimer the demand for racino gaming was estimated using daily data from 1989 to 1990. The period of analysis was immediately prior to and after the nonstatutory pilot project when a limited number of VLT s were installed at the racetrack under the auspices of the state lottery. In this study the period of analysis is July 1994 through December 2002. Data are weekly with 443 observations over the study period. The data cover a period that includes the end of the limited-vlt pilot project period and subsequently the period where the statutory provisions of the West Virginia Lottery Act of 1994 replaced the temporary provisions of the pilot project period. In what follows, two econometric models are developed, one each for the demand for VLT and on-track parimutuel wagering both occurring at a single racino location. Dependent variable Demand is measured as handle per unit of market area population for VLT wagering (PCHAND_VLT) and for on-track parimutuel wagering (PCHAND_PARI). Annualized total VLT handle increased from $131 million to $2.6 billion from the first to the last 12 months of the study period. On a weekly basis, per capita VLT handle increased from $0.30 to $5.92. Real per capita VLT handle (RPCHAND_VLT) increased 1.580%. Annualized total on-track parimutuel handle increased from $38.2 million to $39.4 million from the first to the last 12 months of the study period. On a weekly basis, per capita parimutuel handle increased from $0.087 to $0.090 while real per capita parimutuel handle (RPCHAND_PARI) declined by 13%. The relative importance of parimutuel handle to total VLT plus parimutuel handle decreased over the study period. On-track parimutuel handle fell from 22.7% to 1.5% of total handle. The functional form chosen for the analysis relates the logarithm of real per capita handle to its determinants. This specification was chosen to ensure that handle will be positive. The two dependent variables transformed into their logarithmic equivalents are LRPCHAND_VLT and LRPCHAND_PARI. Demand determinants The major determinants of the demand for a product are price and product characteristics of the product, price and product characteristics of competing products, market area income and other factors influencing the market environment such as government restrictions. Following is a discussion of the demand determinants for both parimutuel and VLT wagering. VLT and parimutuel demand determinants general market environment General market environment factors affecting the demands for both VLT and parimutuel wagering include,. trend. seasonality. holiday. market area per capita income. extreme weather conditions. competition A stochastic trend term was introduced to capture effects such as continuing capital improvements not captured by the other included variables. The trend variables for VLT and parimutuel wagering demands are the lagged dependent variables LRPCHAND_VLT(-1) and LRPCHAND_PARI (-1), respectively. Seasonality in the data is captured by the binary(0, 1) month variables: JAN, FEB, MAR, APR, MAY, JUN, JUL, AUG, SEP, OCT, NOV, DEC. The holiday effect is captured by the binary(0, 1) national holiday variable, HOLIDAY. Extreme weather conditions which were observed for the winter week of 9 January 1999 and which resulted in limited wagering availability, are captured by the binary(0, 1) variable, BADWEATH. The market area economic environment is captured by per capita income, PCI. The independent variable real per capita income is denoted by RPCI. It is expected that both VLT and parimutuel wagering demands will be positively related to real per capita income and will increase at a decreasing rate. For this reason, RPCI is transformed into its logarithmic equivalent, LRPCI. The major competitors for Mountaineer Park over the study period were Wheeling Downs greyhound racino in West Virginia, located approximately 55 miles away and the West Virginia State Lottery. Since both Wheeling Downs and the state lottery were in place over the entire study period their effects on the demand for wagering at Mountaineer could not be determined. The effects on handle of the new hotel which was opened in 2002, augmenting the existing Mountaineer Lodge, could not be determined, possibly due to the lack of a sufficient time period over which to determine its impact.

Demand for casino and parimutuel wagering 777 VLT and parimutuel demand determinants VLT-specific Each of the VLT-specific demand variables, i.e. number of VLT s, type of VLT, location of VLT s and maximum VLT bet, was subject to government restrictions at some point over the study period. Video lottery terminals-specific factors affecting the demands for both VLT and parimutuel wagering include,. number of VLT s. VLT type. VLT location. VLT maximum bet per play. VLT wagering days. VLT price The number of VLT s was increased greatly over the study period. In fiscal year 1994 at the beginning of the study period Mountaineer Park had 165 VLT s under the pilot program. These were replaced in September 1994 by 400 newer generation VLT s permitted with enactment of the West Virginia Lottery Act of 1994. Subsequent increases in the number of VLT s required permission from the West Virginia Lottery Commission. Upon request by Mountaineer Park, the number of VLT s was increased over the study period from 400 in 1994 to 800 in 1995, to 1000 in 1997, to 1200 in 1998, to 1345 in 1999, to 1905 in 2000, to 2500 in 2001 and to 3000 by the end of 2002. VLT demand is expected to be positively related to the number of VLT s and to increase at a decreasing rate (Thalheimer and Ali, 2003). For this reason, the VLT s variable (VLTS) was transformed into its logarithmic equivalent, LVLTS. Parimutuel wagering demand is expected to be negatively related to the number of VLT s (Thalheimer, 1998). Parimutuel demand was specified as a linear function of VLTS. In this study, the relationship of the type of VLT to VLT wagering demand is examined for the first time in the literature. The type of VLT offered to customers is constantly changing with new game innovations and new technologies. The VLT s available and permitted at various times over the study period varied by combinations of payment method and game option. There were two payment methods available over the period. Initially, during the pilot project period and with passage of the West Virginia Lottery Act, all VLT s had to pay out winnings by issuing a ticket, paper receipt when a player finished play. This receipt could then be cashed by the racino cashier. Machines with this type of payment option are referred to as ticket-out machines. 6 The second type of payment method, permitted later in the study period, is referred to as a coin-out or coindrop. Coin-out machines pay winnings as coins dropped into a container on the front of the machine. In addition to payment method, the type of VLT was also defined by different game options permitted at various times over the study period. Some machines offered multiple game options while others offered single game options. All machines offering multiple game options were video gaming devices. The single game machines could offer either a video game such as a multi-payline slot machine-type game or a classical mechanical spinning reel Las Vegas-style slot machine game. The naming convention for machine type used in the remainder of this study is: VLT_payment type_number of games_machine type(game options). Payment type can be ticket-out (T) or coin-out (C). The number of games on a machine can be one (SINGLE) or more (MULTI). Machine payment type can be video (VID) or mechanical spinning reel (REEL). A chronology of the introduction of various machine types now follows. The pilot project period ended in September 1994 with implementation of the West Virginia Lottery Act of 1994. The first-generation ticket-out, single game, video, VLT s permitted over the pilot project period are designated as VLT_T_SINGLE_VID (pilot: card, keno or slot). The West Virginia Lottery Act initially provided for 400 second-generation ticket-out multi-game video machines which were installed at the racino on 2 September 1994. These machines replaced all of the original 165 pilot project machines. They offered card and keno games and are designated as, VLT_T_MULTI_VID(card, keno). While offering both card and keno games on a single machine, the Lottery Act prohibited them from offering a video slot machine game option. In July 1996, a video slot machine game option was permitted for use on the VLT s and was available as an additional game option on the existing ticket-out machines, VLT_T_MULTI_VID(card, keno, slot). 7 By October 1998, all of the existing and new ticket-out VLT s had the video slot game option. In June 1999, the statutes were again changed to permit coin-out payment machines with various 6 In recent years, receipts from ticket-out machines can be used as input to other machines in lieu of receiving cash. 7 See MTR Gaming Group, Inc., United States Security and Exchange Commission, Form 10-K, Annual Report, 31 December 1996. The simulated slot machine game startup was obtained from earlier reports.

778 R. Thalheimer associated new game options. There were three game options associated with the newly permitted coin-out machines. In November 1999, the first coin-out machines were placed on the racino floor. These were classic Las Vegas-style single game mechanical spinning reel slot machines, VLT_C_SINGLE_ REEL(classic slot). These were followed in April 2000 by multi-game coin-out video machines with card and slot game options, VLT_C_MULTI_VID (card, slot). Finally, in August 2000, single game multi-payline video slot machines, VLT_C_SINGLE_VID(multi-line slot) were installed on the racino floor. These video machines (generally with low coin input denominations such as penny or nickel) have symbols on more than one line and offer many possibilities of connecting symbols both on a single line and between lines. The distribution of machine types in a given period was determined as the ratio of the number of each available machine type divided by the total of all existing machine types. Each machine type in a given period, thus assumed a value between 0 and 1 and the total of machine types in a given period summed to 1. Because the number of VLT s largely accounted for the VLT impact on parimutuel wagering, the distribution of machines by type was not included in the parimutuel demand model. 8 As was the case for machine types, the effect on demand of the location of VLT s has not been examined in previous studies. Video lottery terminals were located at both the racetrack and lodge. The placement of the VLT s at the lodge and racetrack is measured by the VLT lodge-to-track ratio (VLT_LODGETOTRACK). Prior to enactment of the West Virginia Lottery Act, there were no restrictions on location of the VLT s. The ratio of VLT s at the lodge relative to the racetrack was 0.4. The Lottery Act provided for restrictions on the location of the VLT s at the racino over the study period. The Lottery Act specified that a 1 : 1 ratio of machines at the lodge vs. those at the racetrack be maintained. The statutes were amended in June 1998 to permit a 2 : 1 lodge-to-track ratio. Finally, in 2000, the statute was amended to eliminate the relative placement requirements of the VLT s between the racetrack and lodge. The nonrestricted lodge-to-track ratio was 3.4 : 1 in the last year of the study period. The maximum bet per play on a VLT as permitted by statute was initially set at $2 which was raised to $5 on 21 April 2001 as a result of changes in the statute. This government restricted maximum bet is defined by the binary (0,1) variable, VLT_BETLIMIT2TO5. 9 The number of VLT days per week was not included in the estimating equations, since the VLT s were operated year-round 7 days per week, with the exception of Christmas Day, and so their effect on the demand for VLT and parimutuel wagering could not be determined. The price of VLT gaming is the win percent. This is the amount remaining after payout of winning bets relative to total handle. The win percent which averaged 8.5% over the study period was relatively constant throughout the period and so its effect on VLT or parimutuel wagering demand could not be determined. VLT and parimutuel demand determinants parimutuel-specific VLT-specific factors affecting the demands for both VLT and parimutuel wagering include,. parimutuel wagering days. number of on-track full-card simulcast programs per parimutuel wagering day. live race days per parimutuel wagering day. average daily purse. presence of stakes race. price of parimutuel wagering. The days per week over which parimutuel wagering was conducted (DAYSPARI) varied from five to seven over the study period. Wagering at a racetrack location on a full schedule (card) of races simulcast from other racetrack who are conducting live racing is referred to as full-card simulcasting. Wagering on simulcasts of live horse and dog races sent to Mountaineer Park from other in-state and outof-state parimutuel racetrack locations was available each week over the study period. It is expected that parimutuel wagering demand is positively related to DAYSPARI and increases at a decreasing rate. For this reason, DAYSPARI was transformed into its logarithmic equivalent, LDAYSPARI. The intensity of the full-card simulcast product was defined by the number of racetracks whose signals were received at Mountaineer Park over a week divided by the number of parimutuel wagering days in the week (FULLCARD). The average number of full-cards per parimutuel wagering day increased 8 The overall explanatory power of the included demand determinants was virtually unaffected by this action. The adjusted coefficient of determination was 0.875 with VLT types included and 0.865 when they were excluded. 9 See MTR Gaming Group, Inc., United States Security and Exchange Commission, Form 10-K, Annual Report 31 December 2002. The simulated slot machine game startup was obtained from earlier reports.

Demand for casino and parimutuel wagering 779 from 4 per day over the first year of the study period to 18 per day over the last year of the study period. It is expected that parimutuel wagering demand is positively related to FULLCARD and increases at a decreasing rate. For this reason, FULLCARD was transformed into its logarithmic equivalent, LFULLCARD. Live horse racing was conducted on all but 5 of the 443 weeks in the study period, but not every day of the wagering week. When offered, live racing was always conducted on a day when full-card simulcasting was offered. The number of live race days in a week was equal to or less than the number of simulcast days in a week. The availability of live horse racing on a parimutuel wagering day (LIVECARD) was computed as the number of live race days in a week divided by the number of parimutuel wagering days in that week. LIVECARD varies between zero and one. Live race purse per wagering day, LIVEPURSE, is a measure of product quality to the wagering customer. Live race purse per race day increased from $23 500 to $163 100 over the study period. The independent variable real live race purse is denoted by RLIVERPURSE. To show the importance of statutory payments to purses from VLT revenues consider that in fiscal year 1995, the first 12 months of the study period, payments from VLT revenues were 36% of total live race purses, the remaining 64% being funded by payments from parimutuel handle. By 2002, the last 12 months of the study period, payments to live race purses from VLT revenues were 79% of the total with payments to purses from parimutuel handle accounting for the remaining 21%. Stakes races are the highest quality of race offering the highest purse at a racetrack. In this study, the stakes race of greatest regional importance was the West Virginia Derby (STAKE_WVDERBY) conducted live at Mountaineer Park. Simulcasts of the three triple crown stakes races of greatest national interest, the Kentucky Derby (STAKE_KYDERBY), the Preakness Stakes (STAKE_PREAKNESS) and the Belmont Stakes (STAKE_BELMONT) were available to Mountaineer customers each year over the study period. In addition, the simulcast of the Breeders Cup (STAKE_BRCUP) another stakes event of national importance was also available to Mountaineer customers each year over the study period. The price of parimutuel wagering is the takeout rate, the statutory percent withheld from each wager for payment to government, racetrack and horsemen (purses, breeders awards). The takeout rate on live racing, which averaged 23% over different bet types, was not changed by statute over the study period and so its effect on VLT or parimutuel wagering demand could not be determined. Summary of model specification VLT The VLT demand model is specified as LRPCHAND VLT ¼ 0 þ 1 LRPCHAND VLTð 1Þþ 2 JAN þ 3 FEB þ 4 MAR þ 5 APR þ 6 MAY þ 7 JUN þ 8 JUL þ 9 AUG þ 10 SEP þ 11 OCT þ 12 NOV þ 13 HOLIDAY þ 14 LRPCI þ 15 LVLTS þ 16 VLT T MULTIðcard, kenoþ þ 17 VLT T MULTI VIDðcard, keno, slotþ þ 18 VLT C SINGLE REELðclassic slotþ þ 19 VLT C SINGLE VIDðmulti line slotþ þ 19 VLT C MULTI VIDðcard, slotþ þ 21 VLT LODGETOTRACK þ 22 VLT BETLIMIT2TO5 þ 23 LDAYSPARI þ 24 LFULLCARD þ 25 LIVECARD þ 26 RLIVEPURSE þ 27 STAKE WVDERBY þ 28 STAKE KYDERBY þ 29 STAKE PREAKNESS þ 30 STAKE BELMONT þ 31 STAKE BRCUP þ 32 BADWEATH þ u ð1þ where the variables and their functional forms are as previously discussed and u is the error term. Summary of model specification parimutuel The demand determinants for the parimutuel demand equation are identical to those in the VLT demand equation with the exception that the VLT machine type variables are not included. Additionally, the VLT variable, which is expected to be negatively related to parimutuel handle, is specified in linear form. The parimutuel demand model is specified as: LRPCHAND PARI ¼ 0 þ 1 LRPCHAND PARIð 1Þþ 2 JAN þ 3 FEB þ 4 MAR þ 5 APR þ 6 MAY þ 7 JUN þ 8 JUL þ 9 AUG þ 10 SEP þ 11 OCT þ 12 NOV þ 13 HOLIDAY þ 14 LRPCI þ 15 VLTS þ 16 VLT LODGETOTRACK þ 17 VLT BETLIMIT2TO5 þ 18 DAYSPARI

780 R. Thalheimer þ 19 LFULLCARD þ 20 LIVECARD þ 21 RLIVEPURSE þ 22 STAKE WVDERBY þ 23 STAKE KYDERBY þ 24 STAKE PREAKNESS þ 25 STAKE BELMONT þ 26 STAKE BRCUP þ 27 BADWEATH þ u ð2þ where the variables are as previously discussed and u is the error term. Data sources and summary statistics Data definitions and sources are given in appendix Table A1. Summary statistics are given in appendix Table A2. VLT handle was also found to be positively related to each of the machine type variables. The impact of each machine types is computed relative to the omitted pilot period machine type variable, VLT_T_SINGLE_VID(pilot: card, keno or slot). To compare the relative impacts on VLT handle for each of the machine types, their impact is computed for a 1% share of total available machines as follows: 11 VLT_C_SINGLE_REEL(classic slot) 2.28% VLT_T_MULTI_VIDEO(card, keno, slot) 1.73% VLT_C_SINGLE_VIDEO(multi-line slot) 1.33% VLT_C_MULTI_VIDEO(card, slot) 0.85% VLT_T_MULTI_VIDEO(card, keno) 0.80% IV. Model Estimation and Analysis The parimutuel and VLT wagering demand models were estimated using ordinary least squares. VLT demand model The estimated VLT handle demand model is given in Table 1. The estimated equation fits the data well with an adjusted R 2 of 0.991. The equation is statistically significant. The Breusch Godfrey LM (Lagrange multiplier) statistic indicates the absence of serial correlation. This suggests that the equation is likely free of specification errors such as those caused by omitted variables or incorrect functional form. Of the 32 variables, 25 were significant at the 5% level or lower. All of the VLT-specific variables and the stochastic trend variable were significant. Per capita personal income and the parimutuel variables, live race purse and each of the stakes race variables were not significant. Video lottery terminal handle was positively related to the stochastic trend. This indicates that variables not specifically included in the equation, such as capital improvements made over the period, had a net positive effect on VLT handle. Video lottery terminal handle was also found to increase at a decreasing rate with the number of VLT s. The long-run elasticity with respect to the number of VLT s was found to be 0.36. 10 The machine type with the greatest impact on VLT handle was the classical Las-Vegas, coin-out spinning reel slot machine. Another interesting finding was that the permitted addition of a video slot machine game option to the existing ticket-out, multi-game, video card and keno machines, i.e. VLT_T_MULTI_ VIDEO(card, keno, slot) vs. VLT_T_MULTI_ VIDEO(card, keno), more than double the impact of those machines on VLT handle. It should be mentioned that other factors included in the product characteristic mix for each machine-type are characteristics that may vary between and within machine types. Such characteristics would include, for example, win percent 12 and minimum bet (e.g. nickel, quarter, dollar). Available information did not permit measurement of the effect of each of these factors on VLT handle independent of the machine-types with which they were associated. The individual impacts of the various machine types at the Mountaineer Park racino are facility and location-specific and may not be able to be generalized with respect to other racino locations. What is apparent from this aspect of the analysis is that customers showed distinct preferences for machines with different product characteristic packages such as payment method, type of game, and number of game options. VLT handle was found to be positively related to the VLT lodge-to-track ratio. Video lottery terminal handle was also found to increase with an increase of the maximum bet limit restriction from $2 to $5. 10 Long-run elasticity computed as: 15 /(1 1 ), where 1 and 15 are given in Equation 1 with corresponding values in Table 1. 11 Long-run percent change in handle for each machine type is computed as: [exp( i /(1 1 )(0.01)) 1]100, i ¼ 16, 20, where i and 1 are given in Equation 1 with corresponding values in Table 1. 12 Although the win percent could vary by machine, the aggregate win percent remained fairly constant over the study period.

Demand for casino and parimutuel wagering 781 Table 1. VLT wagering demand model (dependent variable: LRPCHAND_VLT) Variable Coefficient SE t-statistic Prob. C 9.2670 4.8872 1.90 0.0586 LRPCHAND_VLT( 1) 0.3799 0.0354 10.74 0.0000 JAN 0.1262 0.0283 4.46 0.0000 FEB 0.1792 0.0288 6.23 0.0000 MAR 0.1718 0.0270 6.36 0.0000 APR 0.2033 0.0265 7.67 0.0000 MAY 0.1447 0.0315 4.60 0.0000 JUN 0.1417 0.0292 4.85 0.0000 JUL 0.1859 0.0268 6.94 0.0000 AUG 0.2030 0.0277 7.33 0.0000 SEP 0.1312 0.0246 5.33 0.0000 OCT 0.1488 0.0242 6.14 0.0000 NOV 0.0744 0.0244 3.05 0.0024 HOLIDAY 0.1463 0.0145 10.10 0.0000 LRPCI 0.5408 0.5109 1.06 0.2905 LVLTS 0.2241 0.0420 5.33 0.0000 VLT_T_MULTI_VID(card, keno) 0.4963 0.0675 7.36 0.0000 VLT_T_MULTI_VID(card, keno, slot) 1.0629 0.0947 11.23 0.0000 VLT_C_SINGLE_REEL(classic slot) 1.4005 0.1433 9.78 0.0000 VLT_C_SINGLE_VID(multi-line slot) 0.8194 0.2135 3.84 0.0001 VLT_C_MULTI_VID(card, slot) 0.5262 0.1841 2.86 0.0045 VLT_BETLIMIT2TO5 0.0645 0.0281 2.29 0.0224 VLT_LODGETOTRACK 0.0534 0.0192 2.78 0.0057 LDAYSPARI 0.3328 0.1206 2.76 0.0061 LFULLCARD 0.2750 0.0397 6.92 0.0000 LIVECARD 0.1957 0.0437 4.48 0.0000 RLIVEPURSE 2.80E-07 4.80E-07 0.58 0.5592 STAKE_WVDRBY 0.0040 0.0465 0.09 0.9312 STAKE_KYDRBY 0.0590 0.0398 1.48 0.1384 STAKE_PREAK 0.0566 0.0398 1.42 0.1553 STAKE_BELMONT 0.0047 0.0380 0.12 0.9020 STAKE_BRCUP 0.0542 0.0334 1.62 0.1056 BADWEATH 0.6632 0.0978 6.78 0.0000 Summary statistics: Observations 443 Observations adjusting endpoints 442 Adjusted R 2 0.990 F-statistic 1368 Prob(F-statistic) 0.00 Breusch Godfrey LM test statistic 0.64 Prob(LM test statistic) 0.42 With respect to the parimutuel-specific variables, VLT handle was found to increase at a decreasing rate with the number of parimutuel wagering days and the number of full-card simulcasts per wagering day. The long-run demand elasticities with respect to parimutuel wagering days and full-card simulcasts per wagering day were 0.54 13 and 0.44, 14 respectively. To show the importance of the parimutuel wagering product to VLT handle, consider that while the average number of parimutuel wagering days per week did not vary greatly over the study period, full-card simulcast per wagering day (FULLCARD) at Mountaineer Park did vary greatly, increasing from 4 per day over the first 12 months to 18 per day over the last 12 months. The impact on VLT handle from increasing the intensity of the full-card simulcast product over the study period was estimated to be 94.8%. 15 13 Long-run elasticity computed as: 23 /(1 1 ), where 1 and 23 are given in Equation 1 with corresponding values in Table 1. 14 Long-run elasticity computed as: 24 /(1 1 ), where 1 and 24 are given in Equation 1 with corresponding values in Table 1. 15 Long-run percent change in handle computed as: [exp(( 24 /(1 1 ))(ln(fullcard{18}) ln(fullcard{4})) 1]100, where 1 and 24 are given in Equation 1 with corresponding values in Table 1.

782 R. Thalheimer Table 2. Impact on VLT handle of lifting government restrictions on machine type, 1994 2002 Machine-type Share (1994) Share (2002) Individual impact(2) VLT_T_MULTI_VID(card, keno) 1.000 0.000 48.5% VLT_C_SINGLE_REEL(classic slot) 0.000 0.368 129.5% VLT_T_MULTI_VID(card, keno, slot) 0.000 0.246 52.5% VLT_C_SINGLE_VID(multi-line slot) 0.000 0.227 34.9% VLT_C_MULTI_VID(card, slot) 0.000 0.159 14.5% Video lottery terminal wagering was found to increase with the number of live race days. The presence of one live race day per week was found to increase VLT handle 4.6%. 16 The number of live race days per week varied from 0 to 7 over the study period. The maximum impact on parimutuel handle from offering live race wagering can be computed by comparing the absence of live racing to offering it each day of the parimutuel wagering week. This is estimated to be 37.1%. 17 From these results it can be seen that the presence of the horse racing and wagering resulted in a large increase in the demand for VLT wagering. 18 Each of the VLT-specific demand variables were subject to government restrictions at some time over the study period. The effect of lifting government restrictions on the VLT product over the study period is now examined. As a result of petition by the racino to relax the restriction on the number of VLT s, the number of VLT s was increased from the original 400 permitted by the West Virginia Video Lottery Act in 1994 to 3000 by the end of the study period. This increase in the number of VLT s was estimated to result in a 107.1% increase in VLT handle. 19 The effects on VLT handle of lifting restrictions on machine types with associated changes in their relative distribution over time are given in Table 2. 20 Upon passage and implementation of the West Virginia Lottery Act in 1994, all of the VLT s were ticket-out, multi-game, video machines with card and keno game options without slot game options, VLT_T_MULTI_VID(card, keno). Eventually, as permitted by changes to the Lottery statutes, a video slot machine game option was added to each of these machines, VLT_T_MULTI_VID(card, keno, slot) replacing all of the original no-slot machine game option machines. The decline in the relative share of the original machines from 1 to 0 accounts for the negative impact on handle shown in Table 2. None of the remaining machine types was permitted in 1994 and all were phased in at various times over the study period due to changes in statutory provisions of the West Virginia Lottery Act. The joint impact on VLT handle from permitting changes in VLT types relative to the initially permitted ticket-out, multi-game, video machines with card and Keno options is computed to be 178.3%. 21 The lifting of the statutory restriction on placement of VLT s between the lodge and racetrack resulted in the VLT lodge-to-track ratio being increased from the government imposed 1 : 1 in 1994 to the customer determined 3.4 : 1 after the restriction was lifted. This ability to change the relative VLT placement from racetrack to lodge resulted in a 23.0% increase in VLT handle. 22 Finally, the effect of changing the government restricted maximum bet per play from $2 to $5 is computed to be 11.0%. 23 16 Long-run impact of one live day computed as: [exp(( 25 /(1 1 ))LIVECARD{1/7}) 1]100. 17 Long-run percent change in handle computed as: [exp(( 25 /(1 1 ))(LIVECARD{1} LIVECARD{0}) 1]100, where 1 and 25 are given in Equation 1 with corresponding values in Table 1. 18 To test the robustness of the positive relationship between the parimutuel variables and VLT handle, Equation 2 was reestimated without LVLTS. The coefficients of LDAYSPARI, LFULLCARD and LIVECARD, all remained positive and significant. 19 Long-run percent change in handle computed as: [exp(( 15 /(1 1 ))(LVLTS{3000} LVLTS{400})) 1]100, where 1 and 15 are given in Equation 1 with corresponding values in Table 1. 20 Long-run impacts computed using: exp(( i /(1 1 ))(s i,2002 s i,1994 )) 1, i ¼ 16, 20, where i and 1 are given in Equation 1 with corresponding values in Table 1. S i,2002 and s i, 1994 are shares of each machine type as shown in Table 2. 21 Joint long-run percent change computed as: [exp(( i /1 1 ))(s i,2002 s i,1994 )) 1]100, i ¼ 16, 20, where 1 and i are given in Equation 1 with corresponding values in Table 1. S i,2002 and s i,1994 are shares of each machine type as shown in Table 2. 22 Long-run percent change in handle computed as: [exp(( 21 /(1 1 ))(VLT_LODGETOTRACK{3.4}) VLT_LODGE TOTRACK{1})) 1]100, where 1 and 21 are given in Equation 1 with corresponding values in Table 1. 23 Long-run percent change in handle computed as: [exp(( 22 /(1 1 ))(VLT_BETLIMIT2TO5{1} VLT_ BETLIMIT2TO5{0})) 1]100, where 22 and 1 are given in Equation 1 with corresponding values in Table 1.

Demand for casino and parimutuel wagering 783 Table 3. Parimutuel wagering demand model (dependent variable: LRPCHAND_PARI) Variable Coefficient SE t-statistic Prob. C 11.7675 2.5982 4.53 0.0000 LRPCHAND_PARI( 1) 0.0999 0.0309 3.23 0.0013 JAN 0.0417 0.0245 1.70 0.0897 FEB 0.2304 0.0256 9.00 0.0000 MAR 0.2356 0.0255 9.24 0.0000 APR 0.2467 0.0254 9.71 0.0000 MAY 0.1785 0.0325 5.50 0.0000 JUN 0.2152 0.0288 7.46 0.0000 JUL 0.3289 0.0278 11.85 0.0000 AUG 0.3025 0.0282 10.72 0.0000 SEP 0.1829 0.0246 7.42 0.0000 OCT 0.1481 0.0235 6.32 0.0000 NOV 0.0912 0.0240 3.80 0.0002 HOLIDAY 0.1346 0.0144 9.37 0.0000 LDAYSPARI 0.4867 0.1097 4.44 0.0000 LFULLCARD 0.1490 0.0310 4.81 0.0000 LIVECARD 0.7129 0.0445 16.02 0.0000 RLIVEPURSE 1.46E-06 4.68E-07 3.12 0.0019 STAKE_WVDRBY 0.2599 0.0458 5.67 0.0000 STAKE_KYDRBY 0.3942 0.0406 9.71 0.0000 STAKE_PREAK 1.86E-01 0.0403 4.60 0.0000 STAKE_BELMONT 1.49E-01 0.0378 3.95 0.0001 STAKE_BRCUP 0.2472 0.0334 7.41 0.0000 VLTS 1.74E-04 2.91E-05 5.97 0.0000 VLT_BETLIMIT2TO5 0.0236 0.0237 1.00 0.3196 VLT_LODGETOTRACK 6.27E-02 1.25E-02 5.00 0.0000 LRPCI 0.7781 0.2733 2.85 0.0046 BADWEATH 0.7661 0.0970 7.90 0.0000 Summary statistics: Observations 443 Observations adjusting endpoints 442 Adjusted R 2 0.850 F-statistic 93.5 Prob(F-statistic) 0.00 Breusch Godfrey LM test statistic 1.81 Prob(LM test statistic) 0.18 Parimutuel demand model The parimutuel wagering demand model is given in Table 3. The estimated equation fits the data well with an adjusted R 2 of 0.85. The equation is statistically significant. The Breusch Godfrey LM statistic indicates the absence of serial correlation. Of the 27 variables, all but JAN and VLT_BETLIMIT2TO5 were significant at the 5% level or lower. VLT_BETLIMIT2TO5 was significant at the 9% level. Parimutuel wagering is positively related to the stochastic trend variable. This indicates that variables not specifically included in the equation, such as capital improvements made over the period, had a net positive effect on parimutuel handle. Parimutuel wagering was found to increase at a decreasing rate with an increase in market area per capita personal income. The long-run elasticity of market area per capita personal income was 0.86. 24 Parimutuel wagering was found to increase at a decreasing rate with the number of pari-mutuel wagering days and the number of full-card simulcast racetrack programs imported to the racino per wagering day. The long-run demand elasticities with respect to the number of pari-mutuel wagering days and the number of full-cards simulcasts per day were 0.54 and 0.17, respectively. 25 24 Long-run elasticity computed as: 14 /(1 1 ), where 14 and 1 are given in Equation 2 with corresponding values in Table 3. 25 Long-run elasticities computed as: 18 /(1 1 ) and 19 /(1 1 ), where 18, 19 and 1 are given in Equation 2 with corresponding values in Table 3.

784 R. Thalheimer The number of pari-mutuel wagering days was relatively constant over the study period. The impact on parimutuel handle from increasing full-card simulcasts from 4 to 18 per day was estimated to be 28.6%. 26 Parimutuel wagering was found to increase with the number of live race days per week. The presence of one live race day per week was found to increase parimutuel handle by 12.0%. 27 The maximum impact on parimutuel handle from offering live race wagering every day of the week relative to its absence was estimated to be 120.8%. 28 On-track parimutuel wagering on all sources, live and full-card, was found to increase with an increase in average live race purse. The long-run live race purse elasticity at the mean was computed to be 0.076 indicating that total (full-card plus live) handle is very inelastic with respect to a change in live race purse. 29 The presence of a stakes race of regional or national importance was found to positively impact wagering demand in each case. The respective impacts on parimutuel handle, for the corresponding week during which the stakes races were conducted, were estimated to be, 33.5% for the West Virginia Derby, 54.9% for the Kentucky Derby, 22.9% for the Preakness Stakes, 18.1% for the Belmont Stakes and 31.6% for the Breeders Cup Stakes. 30 Relaxing restrictions on the VLT variables which were subject to government restrictions resulted in decreases in parimutuel handle. Parimutuel handle was found to decrease with an increase in the number of VLT s. The long-run demand elasticity with respect to the number of VLT s, valued at the study period mean, was 0.25. 31 As a result of petition by the racino to relax the restriction on the number of VLT s increasing them from 400 to 3000 over the study period, on-track parimutuel handle was estimated to have decreased by 38.6%. 32 Lifting restrictions on the relative placement of the VLT s between the Lodge and racetrack resulted in a decrease in on-track parimutuel handle of 15.4%. 33 Finally, lifting restrictions on the maximum bet per play from $2 to $5 resulted in a 2.6% decrease in parimutuel handle. 34 V. Summary and Conclusions A pilot program with a limited number of VLT s conditionally permitted at Mountaineer Park racetrack in West Virginia was begun in 1990 under the auspices of the State Lottery. Declaring the pilot project a success, the West Virginia Video Lottery Act was passed in 1994 marking permanent slot machine gaming at West Virginia parimutuel racetracks, subject to local referendum. Demand models were estimated for the two racino products, video gaming and parimutuel wagering, at Mountaineer Park. Data were weekly from July 1994 to December 2002. Over that period real VLT handle grew 1.580%. The demand models fit the data well. Real per capita VLT handle was found to increase at a decreasing rate with the number of VLT s. Each of the VLT-specific demand factors included in the VLT wagering demand equation was subject to government restrictions over the study period. Following is a summary of the effect on handle of relaxing these restrictions on the VLT factors over the study period: Availability of VLT types 178.3% Number of VLT s 107.1% VLT lodge-to-track placement ratio 23.0% Bet limit $2 to $5 11.0% 26 Long-run percent change in handle computed as: [exp(( 19 /(1 1 ))(ln(fullcard{18)) ln(fullcard{4})) 1]100, where 19 and 1 are given in Equation 2 with corresponding values in Table 3. 27 Long-run percent change in handle computed as: [exp(( 20 /(1 1 ))(LIVECARD{1/7})) 1]100, where 1 and 20 are given in Equation 2 with corresponding values in Table 3. 28 Long-run percent change in handle computed as: [exp(( 20/ (1 1 ))(LIVECARD{1} LIVECARD{0})) 1]100, where 20 and 1 are given in Equation 2 with corresponding values in Table 3. 29 Long-run elasticity computed as: 21 /(1 1 )(RLIVEPURSE{mean}), where RLIVEPURSE{mean} is the sample mean, 21 and 1 are given in Equation 2 with corresponding values in Table 3. 30 Long-run percent change computed as: [exp( i /(1 1 )) 1]100, where 1 and i, i ¼ 22, 26 are given in Equation 2 with corresponding values in Table 3. 31 Long-run elasticity computed as 15 /(1 1 )(VLTS{mean}), where VLTS{mean} is the sample mean, and 1 and 15 are given in Equation 2 with corresponding values in Table 3. 32 Long-run percent change in handle computed as: [exp(( 15 /(1 1 ))(VLTS{3000} VLTS{400})) 1]100, where 1 and 15 are given in Equation 2 with corresponding values in Table 3. 33 Long-run percent change in handle computed as: [exp(( 16 /(1 1 ))(VLT_LODGETOTRACK{3.4}) VLT_LODGE TOTRACK{1})) 1]100, where 1 and 16 are given in Equation 2 with corresponding values in Table 3. 34 VLT_BETLIMIT2TO5 was not significant at the 10% level. Long-run percent change in handle computed as: [exp(( 17 / (1 1 ))(VLT_BETLIMIT2TO5{1} VLT_ BETLIMIT2TO5{0})) 1]100, where 17 and 1 are given in Equation 2 with corresponding values in Table 3.