Casinos in Maine: A Costly Choice

Douglas Muir

May 20, 2003

Abstract

On November 4, voters in Maine will decide whether they want to legalize casino gambling and slot-machine gambling, when questions on these subjects will go to a statewide referendum. To help prepare voters for this crucial decision, this paper attempts to shed light on the financial aspects of this decision: casino revenues, taxes and social costs. We find that a large gambling casino would impose social and governmental costs on the general public in Maine that are larger than gambling tax revenues. After the development of casino competition in the region, the costs to the general public are expected to exceed gambling tax revenues by $134 million per year, while the benefit to the Maine tribes is estimated to drop to less than $16 million per year.

Reference to this document: http://home.comcast.net/~impact-forum/Gaming_Economics.htm

Contents

Section 1. Introduction

Section 2. The Task Force

Section 3. This Work

Section 4. Dependence of Casino Visitation on Distance

Section 5. Dependence of Gambling Expenditures on Distance

Section 6. Effect of Variations in Per Capita Income

Section 7. Casino-Expansion Scenarios

Section 8. Social Costs

Section 9. Summary and Conclusions

References

Tables

Table 1. Patron Origins at Foxwoods and Mohegan Sun in 1995 and 1999

Table 2. Dependence of Casino Gambling and Problem Gambling on Socioeconomic Status

Table 3. Impact of Casino Gambling on Social Costs in Maine Counties

Table 4. Annual Profit and Loss from Casino Gambling in Maine

1. Introduction

On November 4, voters in Maine will decide whether they want casino gambling and slot-machine gambling in their state when questions on these subjects will go to a statewide referendum. Details of the referendum questions can be obtained from the Maine Department of the Secretary of State (2003).

A simple "yes" or "no" vote on one of the referendum questions will either enact or defeat a substantial change to the 1980 Maine Indian Claims Settlement Act, including the authorization for two Maine tribes to operate a gambling casino, probably in southern Maine. A second referendum question will decide the fate of a complicated proposal to permit slot machine gambling at the Bangor and Scarborough racetracks. We will address mainly the proposal of the Maine tribes.

The decision that voters will face on November 4 will not be easy. For those who tend to focus mainly on questions of social justice, the problem will be to weigh the financial benefits that this project would bring to Maine's native peoples against the misery that fast-paced casino gambling inflicts upon gambling addicts and people close to them. Those concerned with job generation in the state will have to weigh the benefits of new service jobs in the host city against jobs lost in other segments of the economy, both now and in the future. Those concerned mainly with reducing taxes will have to weigh the benefit of new gambling-tax revenues against the financial costs and new tax burdens that problem gamblers would impose on the general public. Voters will also have to weigh the attraction of having a glitzy new entertainment venue in the state against the threat of political corruption, which often flourishes when too much wealth is concentrated in too few hands.

To help prepare voters for this momentous decision, we will try to shed some light on the financial aspects of this decision: casino revenues, profits, taxes and social costs. Although the mathematics gets a little involved in places, the issues we address are completely down-to-earth. Voters should study this material with the same level of attention that they give their annual tax bill, because a wrong choice in November could cost them a lot of money.

2. The Task Force

On September 3, 2002, the Maine State Legislature convened the Task Force to Study the Impact of a Maine-Based Casino, in the hope of obtaining guidance on the costs and benefits of legalizing large scale casino gambling in the state. Unfortunately, as noted in the Recommendations section of the Final Report of the Task Force, published by the Office of Policy and Legal Procedure of the Maine Legislature (2002), "the Task Force generated more questions than answers."

For example, at the second meeting of the Task Force, James Klas, an economic consultant hired by the Maine tribes, emphasized the benefits of a casino in southern Maine. On the important subject of job creation, Mr. Klas estimated that a casino would create 4740 jobs. However, as reported by Grace Murphy in the Portland Press Herald (2002), University of Illinois economist Earl Grinols testified before the same Task Force that jobs created at the site of a casino are generally "cannibalized" from other businesses in the region, with no net gain of jobs.

The overall picture regarding revenues was likewise left up in the air. In the fifth year of operation, Mr. Klas projected that casino gross revenues (the amount of money lost by patrons in the casino) would reach $610 million per year and that direct tax revenues would reach $119 million per year, or about 20% of gross revenues. Although not part of Mr. Klas' written testimony, Grace Murphy reports that the two Maine tribes hope to receive a total of about $100 million per year, or about one-sixth of the gross revenues of the project.

Mr. Klas also testified that a casino in Maine would establish strong customer loyalty, but made no other reference to the possibility of future casinos being built in neighboring states. From his testimony and from information that we have developed independently, it is clear that the revenue projections presented by Mr. Klas rest on the assumption that a casino in Maine would enjoy a monopoly position in northern New England.

During the conduct of the Task Force, this assumption was called into question. Several speakers made reference to the fact that legislatures in Massachusetts and Rhode Island are already considering bills that would introduce casinos and/or expand slot machine gambling at racetracks, with the stated purpose of reducing the flow of gambling tax revenues out of their states and into Connecticut. It would not be surprising if a large new casino in southern Maine were to trigger the same defensive reaction in the adjoining states.

David Siegel, Executive Director of the Maine Innkeepers Association, in his testimony before the same meeting of the Task Force, raised the question of the long-term survivability of a Maine-based casino, in view of the casino-expansion "opportunities" open to New Hampshire and Massachusetts.

Earl Grinols, who has testified as an expert on casino economics before legislative task forces and committees in five different New England states in recent years, testified before the Maine Task Force that the addition of large new casino in Maine would make it difficult for political leaders in Massachusetts and New Hampshire to maintain their long record of resistance to the expansion of casino gambling in their states. Grinols testified that "Any study purporting to project how Maine will fare with casinos, failing to account for casinos that would be sited in neighboring states, is not worth the paper it is printed on."

Another important issue is the financial costs that pathological and problem gamblers impose on the non-gambling public. While Mr. Klas characterized these costs as "difficult to determine", Prof. Grinols presented the results of his own, very detailed studies of these costs. Grinols' research, performed in collaboration with University of Georgia economist David Mustard is described in a peer-reviewed article in Managerial and Decision Economics (2001). According to his research, the national average of the "social costs" associated with pathological and problem gambling are comparable in magnitude to casino gross revenues.

3. This Work

To help provide voters with a better understanding of the costs and benefits of establishing a casino industry in the state of Maine, we have critically reviewed the published information on casino impact, especially in areas where divergent views were presented to the Task Force. For example, we reviewed published information on the magnitude of the social costs associated with casino gambling. While, as Mr. Klas indicated, they are rather difficult to estimate, it is also clear that they are very large. We conclude that social costs can be expected to be around 90% of casino gross revenues.

Another important is the geographic distribution of these costs. In his Task Force testimony, James Klas testified that only 12% of casino visits and 10% of gambling revenues would come from residents of Maine. A reasonable assumption is that social costs would be distributed in about the same proportion as casino revenues. Mr. Klas' figures, if correct, would suggest that nearly all of the casino-related social costs would be exported to Maine's neighboring states.

A substantial portion of the present work is devoted to an effort, based on information from a wide variety of sources, to better understand the distribution of casino visits, gambling revenues, and social costs among the counties and states of the region. Our analysis indicates that 27% of casino visits and 21% of gambling revenues would come from within the state of Maine in the initial period of casino operation. While these new estimates are considerably higher than those provided by Mr. Klas, it is important to note that our methodology agrees very well with actual data on casino revenues and casino visitation rates in the New England region.

Stimulated by the testimony at the Task Force regarding regional competition, we have applied our methodology to a case where a casino is built in southern Maine and, following this, several "defensive" casinos are built elsewhere in the northern New England region. In this "post-competition" scenario, we conclude that the in-state fraction of gambling revenues and social costs would jump from 21% to 98%. In this case, all casino gambling venues would become purely local businesses.

4. Dependence of Casino Visitation on Distance

In this and the following Sections, we present the detailed evidence for our conclusions regarding the origins of casino revenues and the distribution of costs to specific counties and states in New England.

We begin by examining the willingness of New England gamblers to travel to casinos located at varying distances from their home. Probably the best information this subject comes from data collected at the existing two-casino complex in southeastern Connecticut. By simply counting the number of automobiles and tour buses arriving from various states in the parking lots at Foxwoods and Mohegan Sun, one can get very useful information on the place of origin of the clients of those facilities.

In Table 1, below, we list the results of Grant (2000). Focusing for the moment on the middle portion of the Table, column 2 is the result of a survey conducted at Foxwoods in October 1995, column 3 is the result of a survey conducted at Foxwoods in February 1999, column 4 is the result of a survey conducted at Mohegan Sun in February 1999.

In order to obtain an estimate of patron visitation at the combined Foxwoods/Mohegan Sun complex in February 1999, we performed a weighted average of columns 3 and 4, using the fiscal 1999 revenues of the two casinos as weights. According to Gamingfloor.com Ltd. (1999), around 60.4% of the total gambling revenue of Foxwoods and Mohegan Sun in that year was earned at Foxwoods with the remaining 39.6% at Mohegan Sun. The weighted average number of visits appears in column 5.

Table 1. Patron Origins^a at Foxwoods and Mohegan Sun in 1995 and 1999

State of Origin	Modelb Oct 95	Foxwoods Oct 95	Foxwoods Feb 99	Moh. Sun Feb 99	Averagec Feb 99	Modelb Feb 99
Massachusetts	35.86%	33.04%	35.96%	22.80%	30.75%	32.87%
Connecticut	34.42%	29.04%	28.32%	46.00%	35.32%	36.03%
Rhode Island	14.90%	14.30%	13.34%	7.80%	11.15%	12.94%
New York	11.58%	10.76%	11.96%	15.60%	13.40%	15.12%
Other	3.24%	12.85%	10.41%	7.70%	9.34%	3.04%

a. Because of round-off, columns may not sum to exactly 100%
b. See text for further discussion of the model calculations.
c. Weighted average of Foxwoods and Mohegan Sun visitation rates.

Several interesting patterns emerge from an examination of these data. For example, although Foxwoods and Mohegan Sun are only about 10 miles from each other, the data in columns three and four indicate that there is a consistent pattern of casino preference, strongly favoring Foxwoods in the case of the "easterly" states of Massachusetts and Rhode Island and strongly favoring Mohegan Sun in the case of the "westerly" states of Connecticut and New York. This provides clear evidence that gamblers do not normally drive past one casino in order to reach a similarly equipped casino further away.

One can also obtain from Table 1 data on the ratio of number of Massachusetts visits to the number of Connecticut visits, Rhode Island visits to Connecticut visits, and New York visits to Connecticut visits. Since the New England states are small, these state-to-state ratios give fairly detailed information on the willingness of New England residents to travel from various distances to visit a gambling venue. We have employed these data to construct a simple mathematical model of how casino visitation rates and per capita casino revenues depend on distance to the nearest casino.

To provide an explanation of the basis of this model requires the introduction of a certain amount of mathematics. Some readers may wish to skip over this material and go directly to Eq. (9), which summarizes the results.

In developing a model, we were guided by a recent detailed study, performed by Thalheimer and Ali (2003), of a large number of factors that influence racino and riverboat casino visits and revenues in the Midwest. For the i-th county in the market area of a given facility, they computed a quantity called the ACCESS parameter as follows,

ACCESS_i = exp(-d_i/32.038) (1)

where exp(t) = e^t. In this expression, d_i is the distance in miles between the casino and the population centroid of the given county. Thalheimer and Ali chose the constant term in this function so as to reproduce an observation, Illinois Gaming Board (1997), that 21% of riverboat casino visits were made by participants who lived at least 50 miles away. Thus ACCESS characterizes the steadily decreasing accessibility of casino gambling with increasing distance from the facility. By performing a population-weighted sum of ACCESS_i over all counties in the market area of a given facility, they obtain the overall accessibility of that facility.

Thalheimer and Ali performed a regression analysis in which the observation data were the annual revenues of 30 different riverboat casinos and racinos measured over the period from 1991-1998. The analysis revealed a systematic relationship between facility revenue and facility accessibility, as defined above. In numerical terms, a 10% increase in accessibility from the sample average was found to be accompanied, on average, by a 4% increase in revenues.

We have modeled the New England visitation rate data in Table 1 with an empirical function similar in form to that shown in Eq. (1). Let V(d) be the average number of casino visits made per year by a person living at a distance d from the nearest casino. We assume that V(d) has the functional form

V(d) = k exp(-(d/d₀)^x) (2)

In Eq. (2) x is an adjustable constant, which we expect to have a value near 1, and d₀ is a second adjustable constant. The quantity d₀ is the point where V(d) drops to 37% of its value at d = 0, so it can be interpreted as the maximum distance that typical patrons are willing to travel in order to gamble. The second parameter x controls the general shape of the curve V(d). Large values of x tend to raise the value of V(d) in the near-casino region between 0 and d₀, while lowering the value at greater distances. Therefore x determines how steeply the rate of casino visits drops off in the neighborhood of d₀.

We assume that the frequency of visits per person depends on distance only, and not, for example, on income, age, gender, or any other demographic factor. We return to this point in Section 6 below.

To get the total number of casino visits per year, one needs, in addition to V(d), information on the spatial distribution of population in the New England region. This was obtained at the county level from the US Census Bureau (2002), which provides the population and the latitude and longitude of the population centroid of each county at the time of the 2000 US census. Analogous information at the level of "minor civil divisions" (municipalities) was obtained for New London County, CT, and York County, ME, from the Missouri Census Data Center (2003) and the US Census Bureau US Gazetteer (2003).

Let Pop_i indicate the population of the i-th county or (in the case of New London County and York County) municipality. To obtain the total number of visits per year of patrons from a given county or state, one needs only to perform a sum over the index i of the product [Pop_i V(d_i)]. The distances d_i from the casino are computed as if the entire population the i-th county or municipality resided at its population centroid. Distance is taken to mean the simple "crow flight" distance, as determined from the latitude and longitude of the population centroid and the latitude and longitude of the casino.

In this work, we have varied these parameters d₀ and x in Eq. (2) to get a best fit to the casino visitation rates for the states listed in Table 1. The best fit to the data in Table 1 was found at d₀ = 29.323 miles and x = 0.82409. In performing this fit, we have considered only state-to-state visitation ratios, so the constant k was not determined. The result then is

V(d) = k exp(-(d/29.323)^0.82409) (3)

In order to illustrate the quality of the fit, V(d) in Eq. (3) was used to produce the "Model" results listed in column 1 of Table 1. The model results in this column should be compared with the corresponding survey data for Foxwoods visits in October 1995 listed in column 2. Similarly, column 6 lists the results of a model calculation of the two-casino complex, as configured in February 1999. For the purposes of this calculation, the two-casino complex is considered to have an effective location with spatial coordinates that are 1999 revenue weighted averages of those of the two casinos, meaning a location about 4 miles from Foxwoods and 6 miles from Mohegan Sun. The "Model" data in column 6 should be compared with the weighted-averaged survey data in column 5.

In spite of the fact that there are only two free parameters available to fit the six independent visitation ratios (three for 1995 and three for 1999), the quality of the fit is quite good. Note that all general features of the visitation data are reproduced by the fitted model, including the striking fact that "moving" the casino just 4 miles westward reverses the visitation ranking of Massachusetts and Connecticut and also reverses the ranking of Rhode Island and New York.

5. Dependence of Gambling Expenditures on Distance

We now turn to the related problem of estimating the average amount of casino revenue obtained per year from an individual residing at a distance d from the casino, R(d).

The following general considerations lead one to expect that the function R(d) will have a smaller rate of decline, with increasing distance from the casino, than does the visitation function V(d). Each gambler has a certain amount of money at his disposal for the purpose of gambling. Most sources of income, whether earned at a job, received as a welfare payment, or for that matter, obtained as the result of criminal activity, are recurring resources. If a particular gambler, for example, moves from one city to a different city at a greater distance from a casino, his frequency of visits would go down, but, because his visits are less frequent, he will have accumulated a larger amount of money with which to gamble when he does visit the casino.

To see how one might express this idea mathematically, consider two extreme models of gambling behavior. In one model, which might be called the "efficient saver" model, R₀(d), gamblers accumulate resources efficiently between visits and, in this way, are able to spend a constant amount of money per year at the casino, regardless of the frequency of visits. In this case, the revenue function will be a constant independent of d,

R₀(d)/R₀(0) = 1 = [V(d)/V(0)]⁰ (4)

In the other extreme, which might be called the "inefficient saver" model, R₁(d), gamblers spend a fixed amount of money per visit to the casino, regardless of the frequency of visits,

R₁(d)/R₁(0) = [V(d)/V(0)] = [V(d)/V(0)]¹ (5)

It seems clear that the behavior of real gamblers lies somewhere between the extremes of R₀(d) and R₁(d). A comparison of Eqs. (4) and (5) suggests that the revenue function and the visitation function may be related according to

R(d)/R(0) = [V(d)/V(0)]^b , (6)

where the exponent b is a constant having a magnitude between 0 and 1.

Substituting the expression for V(d) from Eq. (2) into Eq. (6) and collecting together the various constants, we obtain the final form of our revenue model,

R(d) = c [exp(-(d/d₀)^x)]^b/[exp(-(8/d₀)^x)]^b (7)

Here b and c are adjustable parameters, which will be determined from available revenue information. By including the extra constant expression in the denominator of Eq. (7), the constant c is made equal to the value of R(d) at a distance d = 8 miles from the casino. The reason for defining c in this particular way will become clear shortly.

One quantity that is fairly well known is the local rate of casino participation. This is the average number of dollars lost through gambling per person per year by people living close to the casino, for example, in the host city.

According to Deloitte and Touche (1992), the amount lost per adult residing in or near Atlantic City in 1991 was $198 per year. The cumulative inflation rate over the period from 1991 through 2001, according to the Columbia Journalism Review Internet inflation calculator, was 1.30. This gives a local rate of participation in 2001 dollars of $198 x 1.30 = $257 per adult per year. According to the Center for Responsive Politics (2000), adults comprise around 74% of the US total population, so this converts to $190 per person per year.

The Deloitte and Touche data was collected over a region within 35 miles of the center of Atlantic City. Taking into account the variation of population density within the region, we estimate that the mean distance from casino to residence in the study of Deloitte and Touche was approximately 8 miles. Thus the parameter c, which is equal to R(8), has an estimated value of $190 per person per year.

The next step is to determine the constant b. We now determine this quantity by requiring that our model produce a correct total revenue figure for Foxwoods/ Mohegan Sun complex in 2001.

Although the operators of Foxwoods and Mohegan Sun casinos are not required to publish their revenues from table games, only their slot machine revenues, it is widely reported, for example by Green (2002), the two casinos had total gambling revenues of close to $2 billion in 2001.

The fraction of that revenue which originated within the New York/New England region can be estimated from the data in Table 1. The survey data given in columns 2 through 5 on the line labeled "Other" includes contributions from all other states, including states outside the New England region. In contrast, the "Model" results in columns 1 and 6 on the line labeled "Other" include contributions only from other "within-region" states, namely, New Hampshire, Vermont and Maine. Subtraction of these smaller amounts from the survey data gives a rough estimate of the "out-of-region" visits, 9.6% in 1995 and 6.3% in 1999. >From these numbers we estimate that roughly 10% of the revenues of Foxwoods and Mohegan Sun are gained from persons residing outside of the region of New York and the New England states. We arrive in this way at figure of $1.8 billion for the total amount of gross revenues that Foxwoods and Mohegan Sun gained in 2001 from residents of the New York/New England region.

The market for Foxwoods and Mohegan Sun in 2001 was that part of the northeastern US lying north of the Atlantic City market area. For calculational purposes, we assume that counties with centroids closer to Atlantic City fall in the Atlantic City market area, and counties closer to Foxwoods/Mohegan Sun fall in the Foxwoods/Mohegan Sun market area, with no cross-over patronage between market areas.

As before, let Pop_i denote the population of the i-th county or municipality and d_i the distance from the Foxwoods/Mohegan Sun complex to the i-th population centroid. In our exponential model, the revenue gained from the residents of the i-th county or municipality is given by the revenue function R(d_i) in Eq. (7). Since R(d_i ) depends strongly on the parameter b, we can choose the value of b so as to force the summation of the product [Pop_i R(d_i )] over the market area to give the value of $1.8 billion. Doing so yields the desired value of the final adjustable parameter, namely, b = 0.40845.

Although there is not a one-to-one relationship between our visitation function V(d) and the ACCESS parameter of Thalheimer and Ali, it is nevertheless interesting to recall that in their study, a 10% increase in facility accessibility was accompanied by an average 4% increase in casino revenue (in the region that they studied). Using our exponential revenue model fitted to visitation and revenue data from Foxwoods and Mohegan Sun, if one were to compare two counties, located at different distances from the casino complex by an amount such that the two values of the function V(d) differ by 10%, then from Eq. (6) our exponential revenue model predicts a difference in per capita revenues of 1.10^0.40845 - 1 = 4.0%.

Inserting b = 0.40845 into Eq. (7), we arrive at the desired estimate R(d) of the average amount of casino revenue earned from the gambling losses of a person living d miles from the nearest casino,

R(d) = 218.56 [exp(-(d/29.323)^0.82409)]^0.40845 (8)

or, equivalently,

R(d) = 218.56 exp(-(d/86.911)^0.82409) dollars per person per year (9)

We note that the form of the final result is very similar to the distribution of casino visits, V(d) in Eq. (3). The biggest difference is that R(d) drops off with a longer characteristic distance than V(d). Consider a town located 107 miles from the nearest gambling casino. This, for example, is the distance from Skowhegan, ME, to Sanford, ME. At this distance from a casino, Eq. (9) indicates average casino gambling losses of $66.70 per year for every person living in that community.

6. Effect of Variations in Per Capita Income

The development of the revenue model in Eq. (9) neglected the dependence of gambling activity on a wide variety of demographic factors. Several of the factors that might conceivably have an impact on gambling behavior can safely be neglected in discussing the behavior of large populations, which is our intended application.

For example, while men and women exhibit different gambling behavior, the ratio of male to female residents does not change much in different parts of the region, so gender-related behavioral differences would tend to disappear in the population averaging process. Similarly, the fraction of adults in the population varies by only a few percent at the state level, so variations in the fraction of adults from state to state likewise cannot produce a major effect.

However, one demographic feature, per capita income, does remain evident when viewed at the county and state level. For example, according to the US Bureau of Economic Analysis (2003), the per capita disposable annual income in Connecticut in 2001 was $34,200 and that in Maine was $23,000. One might well ask if Eq. (9), which is based on data taken in southern New England can be applied without change in areas of New England having a lower per capita income.

To answer this question, one must ask if casino gambling behavior depends on the income level of gamblers, and, if so, in what way. This question was addressed in a recent paper by Welte et al., who conducted a nationwide telephone survey of the gambling habits of 2630 persons selected at random. Responses were grouped in several ways to look for trends.

In one grouping, respondents were separated into five groups of roughly equal size (quintiles) based on a determination of "socioeconomic status," which was determined by family income, level of education and occupational prestige. The lowest quintile had an average family annual income of $22,000, the middle quintile an income of $59,000 and the highest an income of $115,000. Table 2 below contains a partial listing of the results given in Tables 2 and 4 of Welte et al.

Table 2. Dependence of Casino Gambling and Problem Gambling on Socioeconomic Status

Grouping Based on Socioeconomic Status	Percentage who Gambled in a Casino in Past Year	Mean Casino Involvementa of those who Gambled in a Casino ($/yr)	Mean Casino Involvementa of those in Quintile ($/yr)	Percentage of P&P Gamblers of those in Quintile
Low Quintile	17.00%	$1,690.00	$287.30	5.30%
2nd Quintile	25.00%	$1,047.00	$261.75	4.00%
3rd Quintile	30.00%	$1,207.00	$362.10	5.10%
4th Quintile	29.00%	$1,308.00	$379.32	1.40%
High Quintile	33.00%	$888.00	$293.04	1.60%

a. The "casino involvement" of an individual is the average number of casino visits per year, times the absolute magnitude of the amount won or lost in the most recent visit.

Thus, as income level rises, there is a visible trend toward increasing frequency of casino visits, but this is countered by a downward trend in spending per visit. The net effect, column 3, is that average casino expenditures per person are nearly the same for all socioeconomic groups. Welte et al. refer to this and other related trends as showing the "democratization" of gambling.

It is interesting to note that the average rate of casino involvement, if expressed as a fraction of annual income, is actually around five times larger for the lowest quintile than for the highest quintile in this study. This rather surprising result may have a connection with the information presented in column 4 of the Table.

Column 4 lists the fraction of survey respondents who fit the criteria being a pathological or problem (P&P) gambler, using criteria developed by psychologists who treat gambling disorders. Ronald Pavalko (1999) provides the following summary of the behavior of these gamblers: "Problem gamblers have an intense preoccupation with gambling. Their lives are focused on gambling, to the exclusion of other interests. They gamble more often and with more money than they intend, and they have great difficulty controlling the amount of money they wager or the amount of time they spend gambling."

As reported by Welte et al., and shown in column 4 of the Table above, the lowest three socioeconomic groups show a statistically significant higher rate of P&P gambling behavior than those in the higher groups. This may help explain why persons in the lower socioeconomic groups spend a disproportionately large fraction of their income on gambling activities.

Thus the data of Welte et al. seem internally consistent. We conclude that it is reasonable to apply the same revenue function R(d), Eq. (9), throughout the states and counties of the New England region, regardless of variations in per capita income.

7. Casino-Expansion Scenarios

We recall that the model in Eq. (9), when combined with population data from the US Census Bureau, yields a total revenue for Foxwoods and Mohegan Sun of $1.8 billion per year, in agreement with reported revenues of the two-casino complex. Of that $1.8 billion, our model calculation indicates that $33.9 million of Foxwoods and Mohegan Sun revenue comes from residents of Maine. These figures provide a baseline of casino gambling activity to which we can compare various scenarios for casino expansion in northern New England. In discussing various possible future developments, we shall refer to this as the "Base Case."

The revenue model was next applied to assess the impact of adding a single large casino in the town of Sanford in central York County, Maine. Sanford is one of the sites under active consideration by the project proponents. "Scenario 1" assumes that this casino joins the existing casinos in Connecticut, but that the three casinos have no other significant competition in the region. Application of the methodology we have developed leads to a total estimated annual revenue of $605 million per year for a casino in Sanford. This result is quite consistent with the $610 million annual revenue projected in the fifth year by proponents of this project. This close agreement suggests that our approach may be similar to the commercially developed planning tools employed by casino developers.

The revenue in Scenario 1 coming from the residents of Maine is predicted to be $127.9 million. This corresponds to an in-state revenue fraction of 21.1%. This figure is significantly higher than the estimate of 10% most recently provided by project proponents. No credible change in our model or parameters would give a fraction as low as 10%.

In discussing casino revenues, it is essential to consider the impact of casino competition. As we mentioned earlier, casino competition is not a vague or remote possibility. Legislatures in Massachusetts and Rhode Island are actively considering bills that would introduce casinos and/or expand slot machine gambling at racetracks, with the stated purpose of reducing the flow of gambling tax revenues out of their states and into Connecticut. Levinthal (2002) has reported on recent pressures to add video slot machines at the Rockingham Racetrack in Salem, NH, to help that track compete with the casinos in Connecticut.

Even within the state of Maine, concerns over the horse racing industry's ability to compete with the proposed casino in southern Maine has prompted the placement of a referendum question on the November 2003 ballot to permit the addition of slot machines to the existing racetracks in Bangor and Scarborough, thereby converting those facilities into "racinos."

Given the likelihood that the construction of a new casino in southern Maine would initiate a wave of casino building in the region, we have examined a third scenario. This case, which might be called the "post-competition" case, includes, in addition to an initial casino in Sanford, additional ones in Bangor, ME, Rochester, NH (14 miles south of Sanford), and the Salem, NH - Methuen, MA area.

The revenue model of Eq. (9) predicts, for Scenario 2, annual revenues of $79.0 million for a casino or racino in Bangor, $95.2 million for one in Sanford, and $110.8 million for one in Rochester. The amount of casino revenues available in Scenario 2 for taxation by the state of Maine (the sum of revenues at Bangor and Sanford) would be $174.2 million per year, far lower than the $605 million predicted for the pre-competition Scenario 1. The biggest single reason for this decline would be the loss of access to the large Massachusetts and New Hampshire markets.

While driving down taxable revenues, casino competition would also drive up gambling expenditures and related social problems within the state of Maine, by moving casinos closer to the place of residence of the gamblers. The amount of money lost by Maine residents in the three mentioned facilities in the post-competition case is predicted to be $170.7 million per year. In this plausible scenario, nearly all of the revenue available for taxation by the state of Maine would be generated from the gambling activities of Maine residents.

8. Social Costs

Up to this point, we have spoken mainly of casino revenues and their geographic distribution. We now turn to the question of social costs.

It is clear that gambling takes a large amount of money from the pockets of gamblers, but a portion of that expenditure represents a legitimate purchase of entertainment by people who can afford it. This portion of casino revenues cannot be characterized as a "social cost."

The social issue arises from the substantial number of gamblers who gamble out of control, losing far more money than they can afford. According to work of H. J. Shaffer, et al. (1997, 1999), 1.14% of all adults can be categorized as pathological gamblers, and another 2.80% are problem gamblers, based on past-year gambling behavior. Given the opportunity to gamble frequently, anonymously and for high stakes, these pathological and problem (P&P) gamblers first gamble away their own personal resources, and then they turn to crime or abuse of a position of trust to obtain the funds needed to support their habit. This behavior does impose significant costs on the wider community, including non-gamblers.

The cost of P&P gambling to the general public appears in many different forms, including increased police, court and jail costs, business losses (lost productivity, absenteeism, unemployment), bankruptcy, welfare payments, cost of treatment and therapy, misuse of household monies, and costs related to divorce and suicide. Some of these "social costs" are paid in the form of actual tax increases, some in the form of increases in the cost of insurance, and some are paid involuntarily by the family, friends and employers of problem gamblers, as a kind of hidden tax.

Grinols and Mustard (2001), by applying Shaffer's prevalence data and the results of published studies of the social costs of P&P gambling performed in many parts of the country, have estimated that the cost of P&P gambling paid by the general public in the US is $180 per adult per year. This value reflects the average of published dollar costs, not adjusted for inflation. (Most of the studies upon which this average is based were published in the period from 1996-1999.) For purposes of comparison, is interesting to note that W. N. Thompson et al. (1999a) compiled social cost data for gamblers in Wisconsin and Connecticut. Their cost results are completely consistent with the national average obtained by Grinols and Mustard.

We next need to estimate the social costs that are specifically attributable to casino gambling activities. Since the social cost data were reported in the 1996-1999 time period, we will base this estimate on casino revenue data for 1998, summarized in Table 1 of Grinols and Mustard. As reported by there, in 1998 casino gambling revenues were $153 per adult per year and total gambling revenues were $282 per adult per year, so that casino gambling represented 54% of total gambling expenditures. Thus the ratio of casino gambling social costs to total gambling social costs lies somewhere in the range from 54% (assuming all forms of gambling produce the same social cost) to 100% (assuming casino gambling is much more damaging than other forms). Given the faster pace, the higher stakes and the increased anonymity of casino gambling, 100% is probably closer to the truth than 54%. To be on the safe side, we take an average of these extremes, or 77%. Thus a reasonable estimate of the social cost of casino gambling in 1998, averaged over the entire US, was 77% of $180, or $138 per adult per year. Finally, our estimate of the ratio of casino-induced social cost to casino gambling revenue is $138 divided by $153, or 0.90.

We next consider the geographic distribution of the social costs. We introduce the notation C(d) to indicate the average amount of social cost (dollars per person per year) experienced by members of the general public in a community located at a distance d from the nearest casino. This cost impact is imposed mainly on the family, friends and employers of problem gamblers and on local governments in the communities where these gamblers live and work. There is no reason to expect that the ratio of casino gambling social costs to casino gambling expenditures in communities in Maine would differ very much from the national average of 0.90. Our estimate for C(d), then, is just

C(d) = 0.9 R(d) (10)

Combining Eqs. (9) and (10), we have

C(d) = 196.7 exp(-(d/86.911)^0.82409) dollars per person per year (11)

The impact of casino-expansion Scenario 1 on the total social costs of a particular county in Maine is just the population of the county Pop_i times the difference between the per capita social costs before and after the construction of a casino in southern Maine,

Impact of Scenario 1 = Pop_i [C(d_scen1) - C(d_base)] (12)

Here d_base is the distance to the Foxwoods-Mohegan Sun complex and d_scen1 is the distance to the closest casino in Scenario 1, which for all locations in Maine would be Sanford.

Similarly, the cost impact in that same county in Scenario 2 is

Impact of Scenario 2 = Pop_i [C(d_scen2) - C(d_base)] (13)

where d_scen2 is the distance to the closest casino in Scenario 2, which, depending on the location in Maine, would be Bangor, Sanford or Rochester.

In Table 3, we list the results of our determination of the distribution of social cost impact of the two casino-expansion scenarios, using Eqs. (12) and (13), for all counties in Maine. For ease of comparison of the results in counties of similar size, the counties are ordered according to total population.

Table3. Impact of Casino Gambling on Social Costs in Maine Counties

County	County Seat	Population 2000	Cost Impact of Scenario 1 ($/yr)	Cost Impact of Scenario 2 ($/yr)
Cumberland	Portland	265 612	$25 249 000	$25 249 000
York	Alfred	186 742	$22 396 000	$22 611 000
Penobscot	Bangor	144 919	$4 343 000	$22 433 000
Kennebec	Augusta	117 114	$6 307 000	$9 070 000
Androscoggin	Auburn	103 793	$7 565 000	$7 565 000
Aroostook	Houlton	73 938	$821 000	$3 082 000
Oxford	South Paris	54 755	$3 746 000	$3 746 000
Hancock	Ellsworth	51 791	$1 652 000	$5 953 000
Somerset	Skowhegan	50 888	$2 122 000	$4 808 000
Knox	Rockland	39 618	$1 887 000	$3 250 000
Waldo	Belfast	36 280	$1 474 000	$4 221 000
Sagadahoc	Bath	35 214	$2 483 000	$2 483 000
Washington	Machias	33 941	$646 000	$2 693 000
Lincoln	Wiscasset	33 616	$1 988 000	$2 162 000
Franklin	Farmington	29 467	$1 462 000	$1 916 000
Piscataquis	Dover-Foxcroft	17 235	$515 000	$1 861 000
Maine	Total	1 274 923	$84 657 000	$123 103 000

To use Table 3 to find out the cost impact in a given county, first locate the county in the Table. For example, Franklin County can be found near the end of the Table. The county seat of Franklin County is Farmington, and the population of the county is 29,467. Construction of a casino in Sanford would cause an increase of casino-gambling-related social costs in Franklin County of $1,462,000 per year over the Base Case. The construction of casinos in both Sanford and Bangor would cause an increase of casino-gambling-related social costs of $1,916,000 per year over the Base Case. The cost impact of Scenario 2 is higher, because the population center of Franklin County is closer to Bangor (71 miles) than it is to Sanford (92 miles). It is clear from this Table that every county in Maine would feel a substantial impact in social costs if casino gambling were to be legalized in the state.

In addition to encountering increased social costs, by entering the casino business the state of Maine would face an increase in operating expenses due to the need for careful regulation of the casino industry. Thompson and Quinn (1999b) have estimated this cost to be about $30 million per year for the state of South Carolina at a time when that state had a slot machine gambling economy of $610 million per year. We can estimate from this that the cost of regulation would be about 5% of the casino gross revenues in the state.

It is also important to estimate the impact of the casino on normal community services, such as schools, police and fire protection, sewer and water services, etc. The casino proposed in southern Maine is expected to have a workforce of about 4700 people. Judging from the experience in Connecticut, most of these new employees would come from out of state and some from out of the country. In addition to the casino employees and their dependents, there would also be approximately 22,000 daily visitors, so the total "casino community" would be sizable. If we assume that each employee has 0.5 dependents and that casino visitors spend on average 6 hours in the host community, there would be an effective year-round population of around 11,500. Judging from the cost of operating town governments in southern Maine, the needs of these 11,500 people cannot be met for less than about $18 million per year, in excess of the modest property tax revenues that can be expected to be paid by the new employees. This community cost impact corresponds to about 3% of gross casino revenues.

There is also a question concerning the cost of improvements to highways, bridges and other items of physical infrastructure in the state. Proponents have stated that such improvements would be paid for out of the $119 million of gambling tax revenues. Without knowing the exact location of the casino, one can only make a rough estimate of the cost impact on the state budget. However, a useful reference is provided by an article by Beth Quimby in the Portland Press Herald, where it is reported that a new 8.5-mile bypass road project in Gorham, Maine, has an estimated cost of $25 million to $30 million. The figure includes the costs of all right-of-way acquisition, engineering, impact review and construction. From this figure, we estimate the cost to the state government of providing the necessary highway infrastructure for a large gambling facility to be around $6 million per year for a number of years after the opening of the casino, or about 1% of casino revenues.

These results can be combined with other cost factors to derive net profit and loss figures for casino gambling in the state of Maine in the base case and in the two casino expansion scenarios we have considered. The results are shown in Table 4.

Table 4. Annual Profit and Loss from Casino Gambling in Maine

Category of Income or Cost	Base Case Foxwoods and Mohegan Sun	Scenario 1 Base Case plus Sanford	Scenario 2 Base Case plus Bangor, Sanford, Rochester and Salem
Maine gross casino revenuesa	none	$605 000 000	$174 205 000
Casino gambling losses by Maine residentsa	$33 871 000	$127 934 000	$170 652 000
Cost of P&P gambling to Maine general publicb	$30 484 000	$115 141 000	$153 587 000
Regulatory costc	none	$30 250 000	$8 710 000
Community costd	none	$18 150 000	$5 226 000
Infrastructure coste	none	$6 050 000	$1 742 000
Total costs to Maine general public (b+c+d+e)	$30 484 000	$169 591 000	$169 265 000
Gambling tax revenuesf	none	$121 000 000	$34 841 000
Net loss to Maine general public (b+c+d+e-f)	$30 484 000	$48 591 000	$134 424 000

a. Amount and distribution of losses/revenues are calculated from Eq. (9).
b. 90% of in-state casino gambling losses. See Table 3 for breakdown of costs by county.
c. 5% of Maine gross revenues. See text.
d. 3% of Maine gross revenues. See text.
e. 1% of Maine gross revenues. See text.
f. 20% of Maine gross revenues. See text.

9. Summary and Conclusions

Casino proponents have pointed out that Maine is already experiencing the social costs of casino gambling in Connecticut and receives no revenue from this activity. This is true, as far as it goes, but it is not a reason for expanding gambling in the state of Maine.

As shown in column 2 of the Table, building a new casino in southern Maine would not improve the present situation at all. The negative financial impact of a first casino in southern Maine would consume all of the new state gambling tax revenues, and more.

If this casino then triggers a wave of casino building in the region, as experts predict, then the net loss (social and governmental costs, less gambling tax revenues) would reach $134 million per year. This corresponds to an average cost of $257 per year for each of the 523,000 households in Maine (US Census Bureau, American Community Survey).

These costs do not include the gambling losses of casino patrons. These are costs that would be paid by all members of the general public, including those who do not gamble, much like a tax. They take many different forms, including increased police, court and jail costs, business losses (lost productivity, absenteeism, unemployment), bankruptcy, welfare payments, cost of addiction treatment and therapy, misuse of household monies, and costs related to divorce and suicide.

After the development of casino competition, the benefit to the tribes (one-sixth of the gross revenues of a casino in southern Maine) would, according to our estimates, shrink to less than $16 million per year. If casino expansion is viewed as a way of providing financial assistance to the tribes, it would be a very inefficient means of doing so. In a mature casino economy, it would cost the general public $134 million per year to maintain this $16 million benefit.

The wording of the referendum question is "Do you want to allow a casino to be run by the Passamaquoddy Tribe and Penobscot Nation if part of the revenue is used for state education and municipal revenue sharing?" A better question would be "Do you favor establishing a casino industry that would have to be subsidized by the taxpayers of Maine?"

References

Columbia Journalism Review. Inflation from 1991 to 2001 was calculated to be 1.30, using the Columbia Journalism Review inflation calculator, which can been accessed on the Internet at www.cjr.org/resources/inflater.asp.

Center for Responsive Politics. 2000. The data can be accessed on the Internet at www.opensecrets.org/overview/DonorDemographics.asp?cycle=2000.

Deloitte and Touche. 1992. "Economic and other impacts of a proposed gaming, entertainment and hotel facility proposed for Chicago, Illinois," report prepared for the City of Chicago Gaming Commission.

Gamingfloor.com Ltd. Archives, North America. November 1999. This material can be accessed on the Internet at www.gamingfloor.com/archive/November_1999_NorthAm.htm.

Grant KM. 2000. "ANTE UP? An Examination of Class III Gaming in Massachusetts", Master of Science in Public Affairs Program University of Massachusetts Boston, November 2000. This material can be accessed on the Internet at www.mccormack.umb.edu/Publications/docs/Ante_Up.pdf.

Green R. 2002. "When Gambling is a Sure Thing", January 15, 2002. This material can be accessed on the Internet at www.winner.com/Articles/When_Gambling_is_a_Sure_Thing/When_Gambling_is_a_Sure_Thing363001.asp.

Grinols EL and Mustard DB. 2001. "Business Profitability versus Social Profitability: Evaluating Industries with Externalities, the Case of Casinos," Managerial and Decision Economics 22, 143-162 (2001). This material can be can be accessed on the Internet at www.terry.uga.edu/~dmustard/profitability.pdf.

Illinois Gaming Board. 1997. Annual report, p.9, survey results.

Levinthal D. 2002. Eagle Tribune, January 10, 2002. This material can be accessed on the Internet at www.eagletribune.com/news/stories/20020110/NH_002.htm.

Maine Department of the Secretary of State. 2003. "Current Statewide Citizen Initiatives and People's Veto." The text of the two referendum questions and the text of the proposed legislation can be accessed on the Internet at www.state.me.us/sos/cec/elec/pets02/pets02-1.htm.

Missouri Census Data Center. The data can be accessed on the Internet at mcdc2.missouri.edu/webrepts/sf3pros/ctindex.html#MCDs.

Murphy G. 2002. "State panel won't take a position on casino", article published in the November 19, 2002, edition of the Portland Press Herald. The article can be accessed on the Internet at news.mainetoday.com/indepth/casinos/021119casino.shtml.

Office of Policy and Legal Analysis of the Maine Legislature. 2002. "Final Report of the Task Force to Study the Impact of a Maine-Based Casino on the Economy, Transportation Infrastructure, State Revenue and Job Market," Section VII (Recommendations), Appendix I (James Klas), Appendix L (David Siegel) and Appendix R (Earl Grinols).

Pavalko RM. 1999. "Problem Gambling", National Forum: The Phi Kappa Phi Journal (Fall 1999): vol. 79, issue 4, pp. 28-32. This material can be accessed on the Internet at home.attbi.com/~rcg/2.html.

Quimby B. 2003. "Hurdles await casino group", article published in the March 23, 2003, edition of the Portland Press Herald. The article can be accessed on the Internet at www.pressherald.com/news/state/030323casino.shtml.

Shaffer HJ, Hall MN, and Vander Bilt J. 1997. "Estimating the prevalence of disordered gambling behavior in the United States and Canada: A meta-analysis." Boston, Massachusetts: President and Fellows of Harvard College. See a summary of their findings in the chapter on "Pathological and Problem Gambling" in the 1999 report of the National Gambling Impact Study Commission. The text of this chapter can be accessed on the Internet at www.casino-gambling-reports.com/GamblingStudy/ProblemGambling/.

Shaffer HJ, Hall MN, and Vander Bilt J. 1999. "Estimating the prevalence of disordered gambling behavior in the United States and Canada: A research synthesis." Am. J. Public Health, 1999, Vol. 89, pp. 1369-1376.

Thalheimer R and Ali MM. 2003. "The Demand for Casino Gaming," to be published in Applied Economics in 2003. A related paper can be accessed on the Internet at http://cbpa.louisville.edu/eip/Newsletters/research/casino_gaming01.htm.

Thompson WN, Gazel R, and Rickman D. 1999a. "Social Costs of Gambling: A Comparative Study of Nutmeg and Cheese State Gamblers," Gaming Research and Review Journal, Vol. 3, Issue 1, pp. 1-16.

Thompson WN and Quinn FL. 1999b. "An Economic Analysis of Machine Gambling in South Carolina," presentation at the South Carolina Policy Council Education Foundation May 18, 1999. This material can be accessed on the Internet at http://www.scfuture.clemson.edu/scdigest/gambling/papers/thompson_screpor.pdf.

US Bureau of Economic Analysis, Regional Accounts Data Annual State Personal Income. The data can be accessed on the Internet at www.bea.doc.gov/bea/regional/spi.

US Census Bureau, Census 2000 Centers of Population by County. 2002. The data can be accessed on the Internet at www.census.gov/geo/www/cenpop/county/ctyctrpg.html.

US Census Bureau, US Gazetteer. 2003. The data can be accessed on the Internet at www.census.gov/cgi-bin/gazetteer.

US Census Bureau, American Community Survey. 2003. The data can be accessed on the Internet at www.census.gov/acs/www/Products/Profiles/Single/2000/C2SS/Narrative/040/NP04000US23.htm.

Welte JW, Barnes GM, Wieczorek WF, Tidwell M-C, Parker J. 2002. "Gambling Participation in the U.S.--Results from a National Survey," Journal of Gambling Studies, Vol. 18, No. 4 (Winter 2002).