decis econ 22: business profitability versus social...

MANAGERIAL AND DECISION ECONOMICS

Manage. Decis. Econ. 22: 143–162 (2001)

Business Profitability versus SocialProfitability: Evaluating Industries with

Externalities, The Case of CasinosEarl L. Grinolsa,* and David B. Mustardb,*

a Department of Economics, Uni6ersity of Illinois, USAb Department of Economics, Terry College of Business, Uni6ersity of Georgia, USA

Casino gambling is a social issue, because in addition to the direct benefits to those who ownand use casinos, positive and negative externalities are reaped and borne by those who do notgamble. To correctly assess the total economic impact of casinos, one must distinguishbetween business profitability and social profitability. This paper provides the most compre-hensive framework for addressing the theoretical cost–benefit issues of casinos by groundingcost–benefit analysis on household utility. It also discusses the current state of knowledgeabout the estimates of both the positive and negative externalities generated by casinos.Lastly, it corrects many prevalent errors in the debate over the economics of casinogambling. Copyright © 2001 John Wiley & Sons, Ltd.

DOI: 10.1002/mde.1004

INTRODUCTION

Between 1990 and 1998, commercial casino rev-enues increased from $8.7 billion to over $22.2billion, or 156%.1 The number of counties withcasinos rose from 26 to almost 200 in the sametime. Including Class III American Indian casi-nos,2 casino revenues totaled $29.5 billion in 1998,representing expenditures of $153 per adult aged20 or over.

The rapid expansion of casinos to new parts ofthe country generated extensive debates about theimpact of casinos on a range of social, economic,and political issues.3 These concerns were suffi-ciently pronounced to cause Congress to establishthe National Gambling Impact Study Commis-sion (NGISC) in 1996 to conduct an exhaustivestudy of the impact of casinos.4 At the conclusionof its investigation, the commission recommendeda national moratorium on the expansion of gam-

bling and more study of gambling’s effects, costsand benefits, before making further decisionsabout it.

The literature on the costs and benefits ofcasino gambling is fraught with inadequacy andconfusion. Even studies that purport to evaluatethe economic impact of casinos commonly exhibita great deal of misunderstanding about whatshould be included among benefits and costs, andprovide little or no guidance about how the costsand benefits relate to one another or should becomputed. Many studies pay a great deal of atten-tion, for example, to estimating the number ofdirect and indirect jobs that casinos create and totallying the taxes casinos pay, but do not explainthe social value of an additional job or calculatethe lost taxes of competing non-casino busi-nesses.5 In general, the costs and benefits dis-cussed are casually listed, vary by study, and arecommonly presented with little or no justificationof how they were selected or why other potentialcosts and benefits were excluded.

A recent paper, Eadington (1999), is instructive.It identified three principal benefits of casinos: (1)

* Correspondence to: aDepartment of Economics, Universityof Illinois, 1206 S. 6th Street, Champaign, Illinois 61820.E-mail: [email protected]; bDepartment of Economics, TerryCollege of Business, University of Georgia, 528 Brooks Hall,Athens, Georgia 30602. E-mail: [email protected]

Copyright © 2001 John Wiley & Sons, Ltd.

144 E.L. GRINOLS AND D.B. MUSTARD

gain in utility (for those gambling in moderationfor entertainment), (2) ancillary economic benefitssuch as ‘job creation, investment stimulation,tourism development, economic development orredevelopment, urban or waterfront revitalization,or the improvement of the economic status ofdeserving or underprivileged groups’, and (3) ad-ditional revenues to the public sector. He lists twoprincipal costs: (1) ‘moral disapproval’ and (2)‘fears of adverse social impacts’, such as patholog-ical gambling, crime, or political corruption. Thenet increase in profits to business, unless this ismeant to be part of ancillary economic benefits, isabsent from the list of benefits.6 Although Ead-ington lists gain in utility (clearly internal to theindividual or household) as a benefit, he writesthat ‘many of the costs identified are internal tothe individual or the household, as opposed toexternal—borne by society—and are thereforedifficult to place into a cost/benefit framework’.This view of costs (including the references tomoral disapproval and fears of consequences in-stead of the actual consequences) suggests that theauthor believes costs are more subtle and possiblyless tangible than benefits. However, because theprocess to determine how items are included isnot explained, there is little theoretical guidanceabout how the identified cost–benefit componentsrelate to one another in an overall assessment ofthe impact of casinos or how competing costs andbenefits are reconciled. We will show how cost–benefit components based on utility can be placedinto the evaluation framework.

To bring uniformity and more theory to bearon the cost–benefit treatment of casinos, thispaper demonstrates the construction of an ex-haustive and utility-grounded framework to iden-tify costs and benefits. It outlines an explicittaxonomy for costs and benefits based on theprinciple of real resource use, and reviews theavailable studies that contain original researchestimating one or more cost–benefit components.Although the primary purpose of this paper is torectify theoretical cost–benefit reasoning as it ap-plies to casinos, the methodology applies moregenerally to the evaluation of projects in otherindustries. We also review existing empirical esti-mates of the costs and benefits of casinos ar-ranged according to the theoretically groundedprinciples. Unfortunately, there has been rela-tively little research on many of the most impor-tant social cost–benefit components, while much

of research has examined less significant issues orissues that are not even part of a properly definedanalysis of social costs and benefits. Some re-search that purports to evaluate costs or benefitsactually examines local and not total social costsor benefits. Another concern is that much of theresearch has been conducted by organizationswith a vested interest in the outcome of the re-search, institutes with industry ties, or state agen-cies. Relatively little research is in peer-reviewedjournals. A review of the empirical literature thatestimates correctly defined components of socialcosts and benefits indicates that the costs of casi-nos are at least 1.9 times greater than benefits.

The remainder of the paper is arranged asfollows. The next section constructs a theoreticalcost–benefit measure based on economic funda-mentals. The third and fourth sections examinethe social benefits and costs of casino gambling,respectively. The fifth section concludes by sum-marizing our contributions and outlining the im-plications of this work for future research.

THEORY

Linking Cost–Benefit to Utility

In this section, we lay out the foundations ofcost–benefit analysis for casino gambling. Toavoid the mistakes that have plagued cost–benefitanalyses, especially confusion about what can beincluded on each side of the cost–benefit ledgerand how each item should be computed, we startfrom the most fundamental cost–benefit conceptpossible—individual utility. The framework weemploy can be as comprehensive and general asdesired, although our objective is to provide justenough detail to include all of the major elementscommonly considered relevant to the economiceffects of gambling and enough explanation toindicate what would change in a more detailedapplication of the framework.

Our starting point is the change in the individu-al’s utility, u1−u0, where superscripts distinguishutility in two situations. In one, casinos are wide-spread geographically (alternative 1) and in theother, casinos are less widely spread (alternative0). We assume that u(x) is a continuous utilityfunction representing locally non-satiable prefer-ences defined on consumption x�Rn. A positiveelement of x denotes consumption of a good orservice, while a negative component stands for the

Copyright © 2001 John Wiley & Sons, Ltd. Manage. Decis. Econ. 22: 143–162 (2001)

145BUSINESS PROFITABILITY VERSUS SOCIAL PROFITABILITY

provision of a good or service.7 For example, theprovision of 10 hours of labor by the individualwould appear as −10 in the labor component ofx. We define the expenditure function e(d, p, u) asthe minimum expenditure needed to achieve util-ity u when the consumer buys and sells at prices p,and d is the distance to the nearest casino. It isstrictly monotonic in u for any choice of fixed dand p. The sign of e(d1, p1, u1)−e(d1, p1, u0) is,therefore, identical to the sign of u1−u0. In otherwords, for fixed distance and prices d and p, e(d,p, u(x)) is a utility function whose natural moneymetric records utility in dollars.8

We compare the social welfare between the twosituations. We presume for simplicity that gam-bling is a standardized good; casinos offer gam-bling on essentially the same terms as casinos inother locations.9 The primary advantage to theconsumer of more casinos, therefore, is closerproximity to the nearest one. Let di

1 be the dis-tance to the nearest casino for consumer i in thepost casino alternative 1. Our measure of socialprofitability is the change in welfare for allconsumers

DW=%i

wi [ei(di1, pi

1, ui1)−ei(di

1, pi1, ui

0)],

where �i wi=m, and m is the number of con-sumer households. Equation (1) allows for differ-ent weights for dollar gains to differenthouseholds, a topic to which we will return below.However, in applying (1) to produce a workingmeasure of social profitability, we explicitly ad-dress many issues left unspoken in some studiesand that are a source of confusion in others. Theinitial model provides the simplest framework foranalyzing the impact of casinos. We list our as-sumptions at the outset for clarity.

– We assume that a dollar of utility to onehousehold is equal to a dollar of utility toanother.10 With respect to Equation (1) thisimplies that wi=1 for all households. It alsomeans that firm profits do not need to beassigned artificial premia or discounts basedon which individuals or households happen toown them.

– Firm profits are equally important to socialwelfare regardless of which firm generatesthem. For example, casino profits are valuedthe same as the profits of a non-casino firm.

– To allow for regional tax differences, con-sumers and firms may face different prices. Inthe limit, each firm and household could havea different, personalized set of prices. House-hold i faces price vector pi, firm j faces pricevector pj, and endowments are traded at pricespV.

– We allow for the possibility that consumersmay be constrained in their labor supply deci-sions, resulting in unemployment. People havea reservation wage but cannot always find ajob at that wage, and lowering their wage willnot increase the chances of their getting a job.

– Firms and economy endowments are owned byhouseholds. Household i owns share uij of firmj, �i uij=1 and endowment Vi�R+

n , where�i Vi=V, the economy endowment vector.

– The government spends tax revenues to pur-chase goods and services, and private house-holds receive utility from these expenditures.To implement this assumption, we employ theartificial device of having the government re-turn tax dollars to households in a lump-sumfashion. Households then spend the transfersas part of their income and experience utilitygains based on their purchases.

– In addition to direct benefits and costs, casinosmay generate positive or negative externalities.Positive externalities add value to the economynot reaped by the agent creating them, whilenegative externalities remove value not paid bythe causing agent, following the usual defini-tion. For example, if a casino’s presence re-duces crime in an area, leading to less need forpolice presence, this frees real resources to therest of the community and represents a posi-tive externality. If the reverse is true, and thecasino increases the need for police, real re-sources are removed from final consumptionx, and this is a negative externality. The thirdand fourth sections discuss the nature of bene-fits and costs in more detail. The net resourcesgained or lost to the system are denoted by g.If g\0 negative externalities outweigh positiveexternalities, which decrease the resourcesavailable for consumption x, and therebylower social welfare. Social cost accounting inreal terms requires

x+g=y+V+z,

where x �i xi is aggregate consumption, andy �j yj is aggregate production. For each



firm j, yj is the associated production vector11;z is the economy trade vector.12

The above remarks provide the simplest frame-work that is sufficiently inclusive to discuss aneconomy’s social costs and benefits of gambling.

Application

Consider now the following carefully chosen iden-tity, a telescoping sum where each term cancelspart of the preceding term.

%i

[ei(di1, pi

1, ui1)−ei(di

1, pi1, ui

0)]=

%i

[ei(di1, pi

1, ui1)−pi

1 ·xi1] (2.1)

(Consumption Constraints in Situation 1)

+%i

[pi1 ·xi

1−pi0 ·xi

0] (2.2)

(Income Effects)

+%i

[pi0 ·xi

0−ei(di0, pi

0, ui0)] (2.3)

(Consumption Constraints in Situation 0)

+%i

[ei(di0, pi

0, ui0)−ei(di

1, pi0, ui

0)] (2.4)

(Distance Benefits)

+%i

[ei(di1, pi

0, ui0)−ei(di

1, pi1, ui

0)] (2.5)

(Consumer Surplus)

Expression (2.1) measures the welfare impact ofconstraints on the consumer’s choice that preventhim from being at his optimal bundle given theprices he faces. The primary example of this kindof constraint is unemployment. ei(di

1, pi1, ui

1) bydefinition is the least costly way of achieving theutility achieved in situation 1. Consumption bun-dle xi

1 satisfies u1=u(xi1) and also achieves utility

u1. Because choice of xi1 was constrained (in the

case of unemployment, by the consumer’s abilityto supply labor), it will lead to a greater expendi-ture than ei(di

1, pi1, ui

1). Therefore, the differencein expression (2.1) is the amount of money theindividual would be willing to pay to remove theconstraint. The same argument applies to expres-sion (2.3) in situation 0.

Expression (2.4) measures the value to the con-sumer of having the nearest casino distance di

1

away compared to distance di0. For example, in

the initial situation the consumer needed ei(di0, pi

0,ui

0) to reach initial utility. When the nearest casinois closer, distance di

1Bdi0, the income needed to

maintain original utility, ei(di1, pi

0, ui0), is smaller

(presuming the individual gambles). The differ-ence in expression (2.4), therefore, is the amountthe consumer would be willing to pay to have thenearest casino closer.

Expression (2.5) is the conventional measure ofconsumer surplus. The only difference betweenthe two terms in the expression is the price vector.If prices pi

1 are better for the household thanprices pi

0 (lower for goods purchased and/orhigher for goods sold, such as labor), then expres-sion (2.5) is positive and measures the amount ofmoney the consumer would be willing to give upto have the better set of prices.

Now examine expression (2.2). Use the house-hold budget identity

pi ·xi=%j

uijPj+pV ·Vi+Ti−Ei (3)

to transform the income effects in (2.2) where Pj

is the profit of firm j, pV ·Vi is earning from thehousehold’s endowment, Ti is the household’sshare of taxes, and Ei is the household’s share ofthe cost of gambling-induced externality expendi-tures. Summing (3) over households and differ-encing between the initial and final situations13

yields

%i

[pi1 ·xi

1−pi0 ·xi

0]=%j

DPj (DProfits)

+DpV ·V (Endowment Capital Gains)

+DT (DTaxes)−DE (DExternality Costs) (4)

Substituting (4) into (2); writing the distance ef-fects in differential form and rearranging gives thetaxonomy of cost–benefit elements that we seek:

%i

[ei(di1, pi

1, ui1)−e(di

1, pi1, ui

0)] DW=

%j

DPj+%i

& di1

di0

(ei

(di

ddi+DT−DE

+Consumption Constraints+DpV ·V

+%i

[ei(di1, pi

0, ui0)−ei(di

1, pi1, ui

0)], (5)

where ‘Consumption Constraints’ is the sum(2.1)+ (2.3).

The seven components in Equation (5) are anexhaustive, exact tabulation of the cost–benefit



elements for evaluating the economic effects ofcasinos. Moreover (5) shows precisely how eachterm should be computed theoretically. For exam-ple, the effect of casino gambling on firm profitsshould be summed over all firms, not just casinos.The increased profits of the casinos should benetted against lost profits of other firms thatcompete for consumer spending. Comparablestatements apply to the computation of employ-ment benefits and costs, taxes, and social costs.

There is one obvious simplification we canmake to (5). Because gambling industry revenue(casinos, lotteries, racetracks and other forms ofgambling) is relatively small,14 it will have a negli-gible effect on creating capital gains or losses onendowments. It is unlikely that the cost of capital,for example, will differ because of the presence orabsence of casinos in the economy. A similarstatement applies to consumer surplus effects thatdepend on gambling to influence overall prices.15

Therefore, for the remainder of the paper weassume that firm and household prices are invari-ant to the amount of gambling (pi

0=pi1, pj

0=pj

1, pV0 =pV

1 ), which means that the last two termsin Equation (5) related to capital gains on endow-ments and consumer surplus gains and losses dropout.

Conceptual Corrections

Equation (5) allows us to address some commonerrors and misconceptions of cost–benefit analy-sis applied to gambling.

The first error is the tendency to identify busi-ness profitability, �j Pj, and its improvement,�j DPj, with social profitability. The two are dif-ferent. Business profitability is clearly importantto social profitability and contributes to it, but thetwo are not synonymous. Failure to account forall of the components of social profitability isperhaps the most common mistake. Casino profitsare visible and prominent. Other costs and bene-fits may be less so.

The second error is to evaluate the economicimpact of gambling with respect to the taxes andprofits of a subset of firms—typically the profitsof firms in one state or region and sometimes theprofits of local gambling firms only. Equation (5)sums profits over all firms, not just casinos orfirms in one location. Ignoring firms that loseprofits due to the expansion of gambling is equiv-alent to selecting weights for them in Equation (1)

that are zero. Because households own these otherfirms, this violates the assumption that house-holds are treated equally.

The third is to consider only the taxes of asubset of households or regions. It is not uncom-mon, for example, for studies to focus only oncosts within the state, even though casinos thatborder another state have ramifications for citi-zens of the neighboring jurisdiction. Equation (5)sums taxes over all households and regions.

Evaluations that consider only the costs orbenefits of a subset of households or regions areinaccurate and incomplete. For example, thecost–benefit measure in (5) does not treat a job ina given location as more valued than a job inanother location. Many economic impact studiesperform regional net export multiplier analyses ofthe effects of casinos. They erroneously report thenumber of jobs in a given location as a benefit.According to (5) the value of employment in onelocation (part of the determination of firm profits)must be netted against the value of employment inanother location. There is no net gain to theeconomy from shifting a job from one location toanother unless it increases profits to theeconomy.16

The last common error is that much empiricalwork purports to show casinos decrease unem-ployment, but fails to prove what employmentwould ha6e been in the absence of casinos. Mostcasinos were introduced after 1991, when thecountry was recovering from the recession of1990–1991. The period from 1991 to 2000 alsocoincided with the longest economic expansion inAmerican history. As the country emerged fromthe recession, unemployment declined in areaswith and without casinos. If casinos temporarilyreduced unemployment faster than it would havefallen otherwise, this transitory effect could cor-rectly be counted as a benefit of casinos. How-ever, we know of no study that has made thiscase. On the contrary, the failure to account forthe decline in unemployment that would haveoccurred anyway leads to a classic post hoc, ergopropter hoc fallacy of logic. For a more detailedexample, see Appendix A, which discusses TheEvans Group (1996). Although it argued thatcasinos reduced unemployment, it did not reportthat areas without casinos with comparable start-ing unemployment rates experienced comparable,and in many cases, larger reductions in the unem-ployment rate.



BOUNDING BENEFITS

This section reviews the studies that estimate thebenefits from casinos based on the theoreticallycorrect cost–benefit computation in Equation (5).We discuss in order the net increase in firms’profits plus taxes paid due to the presence ofcasinos, the consumer distance benefits of nearercasinos, employment benefits and total benefitsfrom the expanded gambling opportunities.

Profits and Taxes

This benefit is calculated by determining thecasino profits and taxes minus the reduction inprofits and taxes of other businesses caused bycasinos. Although casino profits and taxes arehighly visible, they are invalid measures of socialbenefits because they do not adjust for the entireeconomy for the lost profits and taxes of compet-ing businesses. This point is not special to casinos.Any business—be it Wal-Mart or a drugstorechain, that attracts consumer sales, employs laborand other inputs, and displaces competing busi-nesses—should be evaluated on the same basis.

Because many casinos do not have to reporttheir profits or pay taxes (for example, casinosowned by American Indian tribes), there are nodata on industry profits. However, we can esti-mate revenues from annually published informa-tion. We provide a brief overview of casinogambling in the US before estimating the benefits.

Table 1 reports total and per capita gamblingrevenue.17 For comparison, we provide data onthe tobacco industry.18 Many studies estimate po-tential casino revenues using the amount of gam-bling per person in areas where casino gambling isa prominent activity. For example, the City ofChicago Gaming Commission funded a study(Deloitte and Touche, 1992) that reported thatadults within 35 miles of Atlantic City lost $198per adult annually to casinos. Adjusted for eight

years of price changes, this figure is approxi-mately $230. In its study, the Mirage Hotel (1993)estimated that annual per capita gambling rev-enues for persons residing within a 50-mile radiusof its proposed Chicago suburb gambling facilitywould be $200.19 In Iowa, in 1995, a Christiansenand Cummings Associates study for the stateRacing and Gaming Commission found that theaverage adult lost $172 to the casinos (this figureis lower than $230 because casinos are still not inclose proximity to all parts of Iowa). These dataare comparable to revenue for other areas.

In addition to averages we are interested in theconcentration of gambling among users. Manystudies examined gambling markets in differentlocations and at different times. Taken together,they provide a general estimate of how frequentlyresidents gamble. In a market with readily avail-able gambling opportunities including casinos, ap-proximately 30% of the population does notgamble, meaning that they will not have gambledin the past year.20 Another 50–60% could betermed light bettors, who gamble less than onceper week. This group includes those who enjoy anight out at the casino once in a while, but do notfrequent casinos. About 5–15% could be termedheavy bettors who gamble twice per week ormore. The last 2–5% of the population consists ofproblem and pathological (P&P) gamblers, whosuffer from compulsive gambling disorders, whichare expressed when the opportunity to gamble ispresent and sufficient time has elapsed for theproblem to become evident. This group might bein the casino daily, for long periods of time, andat unusual hours. Two-thirds to 80% of gamblingrevenues come from the 10% of the populationthat gambles most heavily.21 Expressed in reverse,90% of the population may provide as little as20% of casino revenues. Consequently, the greatmajority of adults are indifferent, or nearly indif-ferent, to the availability of casino gambling.Although the average American adult loses

Table 1. The Casino Market

Casino gambling revenues All gamblingAll gambling revenues Total revenues ($)per adult ($) revenues ($)per adult ($)

29.5 billionUS 1998 282153 54.4 billion:230 359 44.4 billion 69.3 billion‘Saturated market’

Tobacco industry 39 billion



approximately $153 per year and might lose closerto $230 per year were gambling more widespread,these revenues come from a few who gamble a lot,instead of many who gamble a little.

We now return to our original question—whatis the social value of this amount of casino gam-bling? According to Equation (5) we need theprofits and taxes attributable to casinos, minusthe reduction in profits and taxes of other busi-ness due to casinos. To these we must add theconsumer distance benefits of casinos (which weaddress in the next section). Because profits are afunction of market structure and the presence offree entry and exit, if casinos were deregulated,market contestability and free entry of casinoswould drive economic profits to zero. In thatevent, from the perspective of profits, a largercasino sector and smaller remainder of the econ-omy would represent a net wash because eco-nomic profits in the economy would be no greaterwith casinos than without. The sole contributionof casinos to social welfare in that case would bethe direct consumer benefits.

However, in the current legal environment,casinos in many locations are effectively regionalmonopolies sustained by government licensing re-strictions.22 We, therefore, make the followingadjustment to allow for the higher monopolyprofits of some casinos. In 1998, profits beforetaxes23 of all non-financial corporate business inthe United States were 13.8% of sales.24 Assumingthat casinos average 30% profit rates before taxes(more than double the normal business rate ofprofit) implies that social benefits in the form ofprofits and taxes from shifting $153 of revenuefrom other businesses to casinos is (0.30–0.138)153=$25 rounded up to the nearest dollar. In thenext section, we add to this consumer distancebenefits of casinos to produce an upper bound ontotal casino social benefits.

Consumer Distance Benefits

Equation (5) also idenifies di1

di0 ((e/(d)ddi as a

direct social benefit of casinos, where di is con-sumer i ’s distance to the nearest casino. Distancebenefits have been little studied, even though theyconstitute a primary direct benefit of casinos. Toour knowledge, only Grinols (1999) estimatedthese benefits and compared them with the othercomponents of (5). Assuming that utility depends

on goods x, the number of casino visits V, theamount gambled (spent) per visit g, and the dis-tance traveled to the casino, u=u(g, V, I(g, d))where I(g, d) is an enjoyment factor or visit ‘in-tensity’ factor that rises with g and falls with dand g is consumption of other non-casino, goods.The envelope theorem and consumer optimizationconditions show that di

1

di0 ((e/(d)ddi5di

1

di0 Vdg.

This inequality allows inferences about welfare tobe made from data that relate to the number ofvisits and amount gambled per visit to the dis-tance from the casino. Grinols (1999) estimatedthat the upper bound for direct conumer benefitsof casinos was $50 per adult (again, rounding upto the nearest round figure to produce an upperbound on casino benefits) when no allowance ismade for the significant portion of revenues fromproblem and pathological gamblers. If the rev-enues of non-P&P gamblers only are used tocalculate consumer distance benefits, then thebenefit figure falls to under $34.25 This numbercan be interpreted as the answer to the question,‘How much would you be willing to pay each yearto have the opportunity to gamble in a casinonearby compared with the alternative where casi-nos are 1000 miles away?’

Employment Benefits

Although the topic of employment benefits is oneof the most studied issues about casino gam-bling,26 it also contains a widespread and centralmisunderstanding—that the benefits of new busi-nesses are measured by the jobs they create in agiven location. While it may be legitimate to askwhat effect a new business will have on employ-ment, what taxes it will pay, and from where itsrevenues will come, these answers do not assessthe social benefits and costs of the business. In-creasing jobs in one location at the expense of lostjobs in another is not a social benefit. Businessprofitability is not social profitability. Social costbenefit is grounded on consumer utility and re-sults in a list of relevant factors different fromtracking income and employment effects.27

Total Social Benefits

Based on the previous sections, if casinos werefully deregulated and allowed to spread freelynationwide, economic profits would be driven tozero. The net increase in profits and taxes from



expanding the casino sector at the expense of therest of the economy, therefore, would be zero.The consumer distance benefits of casinos wouldbe less than $50 per adult, or if the revenues ofP&P gamblers are subtracted, $34 per adult.

If casinos are regulated and granted regionalmonopoly status in some jurisdictions, the eco-nomic profits of casinos will remain positive, butthe distance benefits will drop. Assuming averagepre-tax profits equal to 30% of sales (more thandouble the rate for non-financial corporate busi-ness in the US) implies that the net profit and taxbenefits of casinos are less than $25 per adult.However, if there is not free entry, distance bene-fits will average less than $50 per adult (less than$34 adjusting for P&P gamblers) because someareas will not have casinos close to consumers.We are, therefore, left with three upper bounds.The preferred number, $34 per adult, is the mostcorrect upper bound because it represents the fullsocial value of casinos under circumstances inwhich all of the benefits would be captured byconsumers if the industry were deregulated toallow free entry. Fifty-nine dollars combines thefull estimate of consumer distance benefits ad-justed for P&P gamblers with a generous profitfigure. It is too high because the consumer benefitis overstated, and in addition, because it fails torecognize that distance benefits would declinewith regional monopolies present that do not putcasinos close to all consumers. Finally, $75 peradult adds consumer benefits to profits withoutmaking any adjustments. We emphasize that thesenumbers are upper bounds on the estimatedbenefits.

COUNTING COSTS

Researchers estimate the social costs of casinosusing two methods. The first is through the studyof problem and pathological gamblers. The sec-ond is through statistical analyses of cost-creatingactivities such as crime, suicide, and bankruptcy.The former approach ties the cost activities togamblers, but overlooks social costs that do notderive from problem and pathological gamblers.The latter approach, determining the effect ofcasinos on social costs such as crime by examiningaggregate statistics, is direct and more inclusivebecause it looks at more than just the crimescommitted by P&P gamblers.

The remainder of this section consists of twoparts. The first derives a detailed taxonomy ofcost classifications tied to the theoretical analysisin the second section. When discussing these clas-sifications, we cite cost studies of both types listedabove. The second part of this section is a moredetailed review of all the studies that focusedspecifically on problem and pathological gam-blers. We calculate costs per pathological andproblem gamblers, and estimate the costs for theentire nation. These sections constitute the mostcomprehensive compilation of the social costs ofgambling available to date.

Cost Taxonomy

The underlying principle, based on Equation (5),is that each social cost uses physical resources g inways that do not directly enter utility or thatreduce economic efficiency. We arrange socialcosts into nine disjoint groups and discuss eachone briefly.

1. Crime: Of all the social costs, the link betweencasinos and crime has received the most atten-tion.28 Crime costs are real resources used forthe apprehension, adjudication, incarceration,and rehabilitation of criminals, or the policecosts that result from the need for increasedpolice presence in areas of greater gamblingactivity. One significant problem that hasplagued the majority of the casino-crime liter-ature is analogous to the problem present incalculating the profit and tax benefits of casi-nos: To estimate social costs, one should notcount new crime around the casino only, butalso consider whether casinos reduced crime inother locales (for example, this could happenif casinos move crime from other locations).Counting only local crime as a cost is similarto counting only local profits as a benefit.

The most comprehensive analysis of thecasino-crime link is Grinols et al. (2000),which evaluated county-level data for sevenoffenses in every US county over 20 years, andcontrolled for about 50 variables. It concludedthat on average, 8–10% of crime in casinocounties in 1996 could be attributed to thepresence of casino gambling in the county,resulting in costs of $63 per adult annually inthese counties. Furthermore, counties thatborder casinos also experience increasedcrime rates, which suggests that casinos truly



increase crime, not merely shift it from onelocation to another. Estimates of the cost ofnon-Index crimes would add to total crimecosts. For example, insurance fraud is not anFBI Index I crime. Estimates of the fraudcommitted by gamblers is $1.3 billion peryear,29 or $6.61 per adult annually.30

Studies of problem and pathological gam-blers have found similar effects. Maryland De-partment of Health and Mental Hygiene (1990)reported that 62% of gamblers in treatmentcommitted illegal acts as a result of their gam-bling, 80% committed civil offenses and 23%were charged with criminal offenses. Lesieur(1998b) surveyed nearly 400 members of Gam-blers Anonymous, 57% of whom admittedstealing to finance their gambling. On averagethese 400 people stole $135000, and their totaltheft was over $30 million. Lesieur (1992) re-ported on illegal activities and civil fraud en-gaged in by pathological gamblers to gambleor to pay gambling debts in five samples fromhospital inpatients, Veterans Administrationand Gamblers Anonymous groups, male pris-oners, female prisoners, and a female Gam-blers Anonymous sample that includes thewhite-collar crime and other crimes listed initem 1.31

2. Business and Employment Costs: These costsinclude lost productivity on the job, lost timeand unemployment: sick days off for gambling,extended lunch hours, leaving early to gamble,and returning late after gambling. Problem andpathological gamblers often impose costs ontheir employers (in addition to theft or em-bezzlement discussed in the section on abuseddollars below) in the form of an unreliablepresence on the job and reduced productivitywhen present. Between 21 and 36% of problemgamblers in treatment reported losing a jobbecause of their gambling (Lesieur, 1998b).Firing an employee imposes costs on theworker in terms of lost output during theperiod of unemployment and on the employerin terms of additional costs of hiring andtraining new employees. These costs are higherthe greater the firm-specific human capital.

3. Bankruptcy: Lawsuits and legal costs, and billcollection costs, bill collector harassment areamong the consequences of bankruptcy.Pathological gamblers often follow a pre-dictable path of exhausting personal resources,

selling insurance policies, selling possessions,and ‘borrowing’ from family and friends. Theirsearch for funds may lead them to acquiremultiple credit cards that they use to the limit.Debts will be paid off, of course, when theindividual wins big in his next gamblingepisode. Bankruptcy entails costs to creditorsattempting to collect and costs to the legalsystem in court time and resources. SMR Re-search Corporation (1997, p. 118) indicatedthat these costs may be quite large, ‘We set outthis year to interview many of the leading USexperts on gambling, gambling addiction, andthe financial impacts of gambling. Their stud-ies have suggested, fairly consistently, thatmore than 20% of compulsive gamblers hasfiled for bankruptcy as a result of their gam-bling losses’.

4. Suicide: Lesieur (1992) concluded that problemand pathological gamblers have higher suiciderates than the general public.32 Dozens of sto-ries have been reported of gamblers killingthemselves after losing at the casino, some-times on the premises.33 Consistent with this,Phillips et al. (1997) found that deaths in LasVegas were 2.5 times more likely to be a resultof suicide than deaths in other comparablysized metropolitan areas. Visitors to AtlanticCity and Reno were 1.75 and 1.5 times morelikely to die in suicides than tourists to othernon-gambling areas, and in Atlantic City thesuicide rates did not become elevated untilafter casinos were introduced in 1978.McCleary et al. (1998), funded by the Ameri-can Gaming Association, contested Phillips’findings. While we recognize the impact ofcasino gambling on suicide, the literature hasnot provided sufficiently reliable social costestimates, and, therefore, we do not accountfor such costs in the table below.

5. Illness: Among the forms of sickness associatedwith gambling or affected by it are depression,stress-related illness, chronic or severeheadaches, anxiety, moodiness, irritability, in-testinal disorders, asthma, cognitive distor-tions, and cardio-vascular disorders. Manysickness costs are borne by the gambler, butthey also appear as resource costs when thegambler seeks treatment. Gambler-borne costs,even when not absorbing resources, however,are tangible costs to the extent that the gam-bler would be willing to pay to eliminate theproblem.



6. Social Service Costs: This category of costsincludes therapy/treatment costs, unemploy-ment and other social service costs (includeswelfare and food stamps).

7. Government Direct Regulatory Costs: Socialservice and government direct regulatory costsare paid primarily through the government.The gambling industry has been regulated be-cause it has historically been subject to fraudand abuse. Social service costs transfer re-sources from one segment of society to an-other, consuming resources in the process. Ifsocial costs include the financial burden placedon the non-gambling society that would not bepresent in the absence of gambling, then thesecosts should be included for a complete assess-ment of the effects of gambling. Regulatorycosts differ by state and depend on the type ofcasinos (i.e. riverboat, Indian reservation,etc.), and extent of the responsibilities of theregulatory agencies.

8. Family Costs: Families of problem and patho-logical gamblers bear gambling-related costsof divorce, separation, spousal abuse, andchild neglect. Although these costs are non-pecuniary, they are, nevertheless, tangible andreal. They can be quantified in terms of theamount of money an individual would be will-ing to pay to remove the problem. In practice,such costs are rarely measured. When socialservices become necessary, as when gamblingleads to divorce proceedings, they representresources lost to other uses in society and canbe measured by the cost of the servicesprovided.

9. Abused Dollars: The final category representslost gambling money acquired from family,friends, or employers under false pretenses.Two examples are stealing that is never re-ported because the thief is a relative, andmoney ‘loaned’ under duress that is neverrepaid. Abused dollars represent costs to thenon-gambling population. To the extent thatabused dollars represent purchases of gam-bling services that are inefficiently sub-optimalfrom the gambler’s perspective or create mar-ket inefficiencies, a significant portion repre-sents social costs to society as a whole evenallowing for gains by the gambler or gamblingsector.34

Social Cost Estimates Tied Directly to P&PGamblers

Table 2 reports the results of all eight studies thatcontain original research that ties social costsdirectly to pathological gamblers.35 The first tworows show the location studied and the author(s),respectively. The first column shows the categoryof costs, as outlined in the previous section. Thestudies are listed in order of date of publication.With the exception of the pathbreaking paper byPolitzer et al. (1981), the studies were publishedbetween 1994 and 1999. The column totals rangefrom a low of $1,195 (Gerstein et al., 1999) to ahigh of $30,235 (Politzer et al., 1981). The Execu-tive Office of the Governor (1994) is the highestpost-1994 estimate. Because all studies omit someof the costs, these totals will understate the actualtotals.

A large share of the differences in the totals isexplained by differences in the number of costcomponents the studies estimated. The ExecutiveOffice of the Governor (1994) estimated onlycrime costs in Florida, while Thompson andQuinn (1999) estimated ten components. Thestudy with the lowest total cost (Gerstein et al.,1999) estimated only four categories. No studyestimated all the components.36 By far, crime andabused dollars are the largest cost estimates. Ger-stein et al. (1999) is the only study that completelyomits crime costs, and only the Executive Officeof the Governor (1994) and Gerstein et al. (1999)omit estimates of abused dollars. One importantcommon characteristic of all but one of thesestudies is that they are not published in peer-reviewed journals. The Executive Office of theGovernor (1994), Ryan et al. (1999), Thompsonand Quinn (1999) and South Dakota LegislativeResearch Council (1998) were either published byor prepared for state agencies. Thompson et al.(1996) was published by the Wisconsin PolicyResearch Institute. Politzer et al. (1981) was pre-sented at the Fifth National Conference on Gam-bling and Risk Taking, Gerstein et al. (1999) waspresented to the NGISC. The paper by Thompsonet al. (1998) was presented at the Twelfth Na-tional Conference on Problem Gambling, andlater published in Gaming Research and Re6iewJournal.

We used many strategies to ensure that the finalestimates of costs per pathological gambler were




Tab

le2.

Ann

ual

Soc

ial

Cos

tspe

rP

atho

logi

cal

Gam

bler

SDL

AU

SSC

Row

MD

FL

WI

CT

Exe

c.SD

Leg

.P

olit

zer

Tho

mps

onav

erag

esT

hom

pson

Rya

nG

erst

ein

Tho

mps

onet

al.

etal

.O

ffic

eet

al.

for

stud

ies

and

Qui

nnet

al.

etal

.R

esea

rch

(199

9)(1

981)

(199

6)C

ounc

il(1

999)

1994

–19

99(1

998)

ofG

ov(1

999)

($)

($)

($)

($)

1998

–19

99($

)($

)($

)(1

994)

($)

($)

Cri

me

1000

5311

625

771

44A

ppre

hens

ion

and

incr

ease

dpo

lice

cost

s12

34A

djud

icat

ion

(cri

min

alan

dci

vil

994

2764

947

667

617

88ju

stic

eco

sts)

382

690

451

3065

Inca

rcer

atio

nan

dsu

perv

isio

nco

sts

2828

1522

175

888

9

1126

5B

usin

ess

and

empl

oym

ent

cost

sL

ost

prod

ucti

vity

onjo

b10

8210

8259

3632

0L

ost

tim

ean

dun

empl

oym

ent

2156

2913

2717

3436

118

316

515

Ban

krup

tcy

Suic

ide

700

700

Illn

ess

Soci

alse

rvic

eco

sts

The

rapy

/tre

atm

ent

cost

s11

475

396

3083

189

437

971

549

6014

531

844

2U

nem

ploy

men

tan

dot

her

soc.

svc.

606

(inc

l.w

elfa

rean

dfo

odst

amps

)

Gov

ernm

ent

dire

ctre

gula

tory

cost

s

Fam

ilyco

sts

111

111

Div

orce

,se

para

tion

240

3175

2436

3834

3802

1435

495

19A

buse

ddo

llars

1358

6


lower bounds.37 First, in calculating the averageannual cost per pathological gambler by category(shown in the last column of Table 2 on the right)we omitted Politzer et al. (1981).38 This study hadthe highest cost estimates, but was conducted at adifferent time and in a different gambling environ-ment from the other studies. Second, costs forsuicide and government regulation are omitted,because none of these studies estimated them.Third, we did not price adjust the estimates, butrather took the values as given by the authors.Last, many studies combined their estimates forpathological and problem gamblers. We treatedthe numbers as if the costs we report apply only topathological gamblers. Because costs due topathological gamblers are higher than costs due toproblem gamblers, the estimates further under-estimate the costs connected to pathologicalgamblers.

Table 2 shows that the total average social costof eight studies is $13,586 per pathological gam-bler per year. If 1.5% of 196.65 million US adultswere pathological gamblers, this would imply an-nual social costs of $40.1 billion or $204 peradult. If pathological gamblers are 1% of the

population, the estimate reduces to $136 peradult.

Table 3 replicates Table 2 for problem gam-blers. Only Gerstein et al. (1999) and SouthDakota Legislative Research Council (1998) esti-mated any separate costs per problem gambler.These studies estimated only three of the manycost categories. The average annual cost per prob-lem gambler by cost category is shown in the lastcolumn. For the same reasons discussed in analyz-ing the results for pathological gamblers, theTable 3 total cost estimate of $912 due to problemgamblers understates the actual cost.

Table 4 applies the information in Tables 2 and3 to produce annual national social costs peradult. To test the robustness of these cost esti-mates, we use the 95% confidence bounds on thenumbers of problem and pathological gamblersset by Shaffer et al. (1997).39 This confidenceinterval sets the fraction of pathological gamblersbetween 0.9 and 1.38% of the adult popula-tion, and the fraction of problem gamblersbetween 1.95 and 3.65% of the adult popula-tion. Based on these lower and upper bounds,annual national social costs from problem and

Table 3. Annual Social Costs per Problem Gambler

US SDS. Dakota, Row averages:Gerstein et al.

studies 1994–1999(1999) 1998–1999($) ($) ($)

CrimeApprehension and increased police costsAdjudication (criminal and civil justice costs)Incarceration and supervision costs

Business and employment costsLost productivity on job

200Lost time and unemployment 200

Bankruptcy

Suicide

Illness

Social service costs360Therapy/treatment costs 360

Unemployment and other soc. svc. 352155 549(incl. welfare and food stamps)

Government direct regulatory costs

Family costsDivorce, separation

Abused dollars

912



Table 4. National and per Adult Social Costs

pathological gambling range from 27.5 billion toover $43 billion. On a per adult basis, the num-bers range from a low of $140 to a high of $221.Because Shaffer et al. (1997) estimated these con-fidence bounds based on samples of the nationbefore the time of publication including areaswith different degrees of casino gambling theyclearly understate the fractions of the entire USpopulation that would be identified as pathologi-cal or problem gamblers if casinos were expandedfully. The costs of Table 4, therefore, also under-state the associated costs of full gambling expan-sion.

IMPLICATIONS FOR FUTURE RESEARCH

This paper makes many contributions to the dis-cussion of social costs and benefits of casinogaming, and has numerous implications for futureresearch in this area. First, we provide the firsttheoretical justification of what should be in-cluded as costs and benefits. This justification isbased on individual utility and distinguishes busi-ness and social profitability for industries withexternalities. The lack of a clear theoretical basishas impaired the entire research agenda on thisissue. Much research has examined relatively mi-nor issues or issues that are not even part of aproperly defined cost–benefit analysis. Con-versely, there are relatively few estimates of someof the key components of social costs and bene-fits. Consequently, a well-grounded theoreticalframework of costs and benefits will make futureresearch more productive.

Second, using this theoretically grounded cost–benefit analysis we corrected several commonconceptual mistakes prevalent in the casino andgambling literature. One example of a commonerror is the focus on local rather than total socialcosts or benefits. On the benefits side, increases inlocal profits and taxes are often weighted heavilywhile losses in profits and taxes from geographi-cally distant areas are weighted less or not at all.Similarly, on the cost side, local crime is oftenweighted heavily while there is little discussionabout whether crime was simply moved fromother areas. Another error is the frequent use ofthe net export-multiplier modeling of jobs, aninappropriate method to determine social costsand benefits. Clearly, identifying these errors willreduce them in future research.

Third, we used the theory to construct a cleartaxonomy of benefits and costs as applied to thecasino industry. To estimate these costs and bene-fits we reviewed the available studies that dooriginal research on this topic. This literatureshows that the extreme upper bound on annualtotal social benefits is $75 per adult. The lowerbound for social costs, based on the estimates ofcosts associated with prevalence of problem andpathological gamblers, was between $140–$221per adult. Consequently, the available researchindicates that when using the highest estimates ofbenefits and the lowest estimates of costs, casinogambling fails a cost–benefit test by a ratio of 1.9to one or greater.40 Standard Pigouvian correctivetheory for an industry with externalities is that itshould be taxed by an amount equal to the coststhat it imposes on society. Relative to the



revenues for a representative casino of about $230per adult each year from nearby residents, Pigou-vian corrective taxes would represent between 61and 96% of casino revenues.

Fourth, we showed that the available researchindicates there is a lack of quality research on boththe benefit and cost sides of the debate, and thatthere is an important need for better research.There is a need for more uniformity in the mannerin which costs and benefits are treated. Peer-review-quality studies not funded by the casinoindustry or by pro- or anti-gambling groups areespecially needed to refine and improve the cost–benefit numbers that are currently available. Tofurther refine the cost–benefit analysis of casinogaming the following questions must be addressed.

What is the Effect of Casinos on the Number andGambling Patterns of Problem and Pathological

Gamblers?

Because the social costs of the casino industry aregenerated primarily by problem and pathologicalgamblers, it is essential to know how casinos affectproblem and pathological gamblers. There isabundant evidence that increased gambling oppor-tunities increase problem and pathological gam-bling. For example, the NGISC reported that thepresence of a casino within 50 miles roughlydoubled the prevalence of problem and pathologi-cal gambling.41 Other indicators include thetremendous increase in the numbers of gamblersseeking help when casinos enter a market, theincrease in gamblers anonymous groups whengambling enters a state, and the evidence fromsurvey data on the number of problem and patho-logical gamblers before and after casino expan-sion.

Casinos may also affect the amount of gamblingby problem and pathological gamblers. An aver-age adult is expected to lose $200–300 each year incasinos if they are nearby, while a typical patho-logical gambler often loses 10–20 times thisamount. Therefore, a small number of pathologi-cal gamblers accounts for a significant portion ofcasino revenues. A related issue is to determine theshare of casino revenues that derive from problemand pathological gamblers. Does this share differby type of gambling? For example, lotteries receivea smaller portion of their revenues from P&Pgamblers because lottery play attracts a largerportion of the population.

How much does an Additional Active Problem oran Additional Active Pathological Gambler Cost

Society?

This question is best addressed by studying prob-lem and pathological gamblers directly. However,estimates derived from this sample may be biasedbecause only a small fraction of P&P gamblersseek formal treatment. If those who seek helpimpose the greatest costs on society, our costestimates of P&P gamblers would be overstated.

What is the Life Cycle of a Problem andPathological Gambler?

For example, when casino gambling becomesavailable for the first time, what is the behavioraltime profile for individuals who enter and leavethe states of problem and pathological gambling?Do individuals begin with a period of increasinggambling dependence, move through a period ofproblem gambling, progress to pathological gam-bling, seek treatment (or withdraw unilaterallyfrom the problem), and abstain thereafter? Or arethere relapses and continued problems if treat-ment is not sought. This information could beused to predict how many currently active prob-lem and pathological gamblers to expect for givenpopulation as a function of the availability ofcasino gambling.

What Effect do Different Types of Treatmenthave on Problem and Pathological Gamblers?

Such information would help people to know howto efficiently allocate funding resources for treat-ment interventions.

How can Casino Gambling be Offered toMinimize its Social Costs?

Quinn (2001) discusses many possible ways ofoffering casino gaming to reduce social costs. Toevaluate the effectiveness of these interventionsand their impact on casino benefits one wouldneed to estimate the elasticity of both P&P andnon-P&P gamblers to such actions.

What are the Net Profit and Tax Benefits ofIncreasing Casino Gambling?

Rather than estimating a true social benefit, manystudies estimate only the gross increases in profitsor only weight the increased benefits to local firmswhile ignoring lost profits to other firms.



What are the Distance Benefits of IncreasingCasino Gambling?

To date only one study examines this importantquestion. Testing the robustness of this result willprovide more insight into this understudied area.

Focusing future research questions andmethodologies on a clearly formulated theoreticalfoundation will allow us to make our estimates ofboth the costs and benefits of casino gaming moreprecise.

APPENDIX A

A Study of the Economic Impact of the GamingIndustry Through 2005, by The Evans Group: A

Partial Critique

International Game Technology (& IGT), a man-ufacturer of computerized casino gaming productsand video gaming machines, and operator of pro-prietary gaming systems, commissioned TheEvans Group, an econometric consulting firm, toproduce a study of the impact of the gamblingindustry in 1996. The 9 September 1996 pressrelease for the resulting report entitled A Study ofthe Economic Impact of the Gaming Industrythrough 2005 issued by & IGT reported,

States and localities that permit casino gaminghave improved their overall economic perfor-

mance . . . The study . . . reports that where6ercasino gaming has been implemented, employmenthas risen, unemployment fallen, and additional taxrevenues have been generated. (Emphasis added.)

The Evans study describes impacts for individualstates. We will briefly examine the findings relatedto Illinois, a state with which the authors arefamiliar. On page 4-3 the report states:

Based on these data, it would appear that theopening of a casino reduced the unemployment ratein that county in both the year it was opened andin the following year. The average employment inthese eight counties...implies a total of 37 000 ex-tra jobs. These multiplier figures are much higherthan ordinarily obtained, and employment in thesecounties might have risen for other reasons aswell. Nonetheless, the figures do indicate thatcasino gaming has been a boon to these counties,especially those that are more rural. (Emphasisadded.)

Most casinos opened after 1991. The period1991–1996 covered by the study, therefore, coin-cided with the nationwide economic expansioncoming out of the recession of 1990–1991. Em-ployment was rising and unemployment wasfalling in many counties, with or without theintroduction of casinos. The authors, therefore,were right to feel uneasy. Their caution that ‘em-ployment in these counties might have risen forother reasons’ shows they knew that simple be-fore-and-after comparisons finding declining un-employment and increasing employment proved

Figure 1. The Evans Group study, reproduced figure 4-1.



Figure 2. Casino counties are indistinguishable from non-casino counties.

nothing about the effects of casinos in a countryrecovering from recession. Figure 1 reproducesfigure 4-1, provided in the original study. Theauthors explain that the observed drop in casinocounty unemployment rates exceeded the stateaverage by 0.3 and 0.2 percentage points on aver-age in the first and second year after introduction.The authors’ conclusions are noted above. Therest of the story is provided below.

The study gives the impression that countiesthat opened casinos experienced better economicperformance than those that did not. However,Illinois contains 102 counties. We can select othercounties that had the same unemployment rate(within 0.1 percentage point) as the casino countyin the inital period and compare their peformancedirectly. This is done in Figure 2. As shown there,the unemployment rate dropped in all countieswith similar initial unemployment. Some countiesdid better than casino counties, some counties didworse. From left to right, bottom row first, thecasino counties are numbers 6, 1, 3, 2, 2, 3, 7, 3.Nineteen counties performed better than theircasino cousin, while 19 performed worse.

A statistical test confirms that the drop inunemployment of casino counties is statisticallyinsignificantly different from the drop experiencedby the comparable non-casino counties shown inFigure 2. Let DU denote the change in countyunemployment rate minus the change in stateunemployment rate for the same period, and let

Casino identify counties that introduced casinos inthe initial period (Casino=1 if a county intro-duced a casino, 0 otherwise). Then running thefollowing regression,

DU=a+b Casino+o

reveals that coefficient b is 2.75 (consistent withthe 0.2 and 0.3 percentage point differences re-ported by The Evans Group), but with a standarderror of 0.856 implying a p-value of 0.4. Coeffi-cient b is, therefore, statistically indistinguishablefrom 0 at conventional levels.

NOTES

1. Gambling revenue is the net amount of money thatthe gambling operator extracts from patrons. Itequals the ‘handle’ (gross amount wagered—which may reflect the same chips being bet manytimes before it is ultimately retained or lost) lesspayouts, prizes, or winnings returned to players.For example, if players wager $1 000 000 on out-comes of a roulette wheel over the course of anevening, and $880 000 is returned to them as win-nings (some roulette slots are reserved for thehouse), then operator revenue is $120 000.

2. According to the Indian Gaming Regulatory Act of1988, Class I gambling consists of ‘social gamessolely for prizes of minimal value’. Included inClass I gambling are traditional Indian games iden-tified with tribal ceremonies and celebrations. ClassII gambling includes bingo and ‘games similar tobingo’. Class III gambling includes ‘all forms of



gaming that are not Class I gaming or Class IIgaming’, such as blackjack, slot machines, roulette,and other casino-style games.

3. Kindt (1994), Grinols (1996), Grinols and Omorov(1996) and Henriksson (1996) discussed a numberof these.

4. Public Law 104-169 of the 104th Congress estab-lished the NGISC. For more information about itsmission, composition and findings see http://www.ngisc.gov/.

5. We show below that both concepts are necessary toa proper cost–benefit assessment of casinos.

6. We will show below that it should be present.7. We follow throughout the paper standard general

equilibrium accounting conventions for describinginputs and outputs in consumption and production.

8. That is, $100 of utility is defined to be the utilitythat can be achieved by optimally spending $100 atprices p, with nearest casino d miles away.

9. For example, the returns to playing roulette, slotmachines, or a blackjack game are approximatelythe same regardless where offered. The frameworkcould be modified to allow for different qualities ofgambling. In this case the model would deal withmultiple, imperfectly substitutable goods.

10. The transfer of wealth in gambling is generallyfrom relatively poor to relatively wealthy. There-fore, if a dollar generates more utility for rich thanpoor, our assumption understates the social bene-fits. If a dollar generates more utility for the poorthan the rich, our assumption understates the socialcosts of casinos.

11. A positive element of yj denotes output of a goodor service, and a negative component denotes theuse of an input.

12. Although it is not central to our objective in thispaper, we include z to be consistent with the gen-eral framework we develop. Excluding z does notaffect the central arguments of this paper. Compo-nents of z are economy excess demands for tradedgoods. A zero denotes a non-traded good, while apositive entry denotes imports.

13. Use the fact that �i uij=1.14. In 1998, gambling revenues were approximately

0.5% of GDP and casino revenues were approxi-mately 0.25%.

15. It is conceivable, of course, in certain circumstancesthat the introduction of casinos could change pricesenough to matter to local residents. For example, ifcasinos increased employment and the local popu-lation, the demand for local housing would in-crease, thus raising housing prices and creatingcapital gains for residents. In such cases, however,the reduction in demand for residential propertyand capital losses in the areas from which the newresidents came would have to be taken into ac-count. Over time, if new housing responded to theincreased demand, the prices of the existing stockof housing would decrease. Because gamblingdoesn’t create new people, but only moves themfrom one place to another, a reasonable first ap-proximation is that the net effect of gambling on

capital gains and consumer surplus considerationswould be small.

16. We presume that the jobs being compared in twolocations are comparable. Blair et al. (1998) arguedthat ‘employees in gaming industry occupations areless satisfied with their jobs than those in otherindustries’. If jobs are different in two locations,then the jobs would appear in the formula asdifferent because workers would demand compen-sating wage differentials, and this would affectprofitability. If compensating wage differentials donot arise, but workers face non-market constraintsthat cause them to work hours that are not optimalgiven the wages paid, these costs would appear inthe unemployment terms of Equation (5).

17. For industry revenue data, see International Gamingand Wagering Business (1999, p. 24).

18. The value of tobacco grown each year is $39 bil-lion. Encarta Encyclopedia, http://encarta.msn.com/find/Concise.asp?ti=02A43000cs12.

19. The proposal was for West Dundee, Illinois. Thestudy reported, ‘Both Christiansen/Cummings andMirage Resorts estimate local gaming demand byapplying gaming win per capita factors to the popu-lation residing within concentric circles of a gamingvenue. The factors decline as distance increases.The $200 win per capita applicable to the 0–50 milesegment was developed jointly by representatives ofMirage Resorts and Dr Cummings to apply to thelocal population in the New Orleans environs in a1992 evaluation of the New Orleans gamingmarket.’

20. See, for example, GLS Research (1994) ClarkCounty (Las Vegas, Nevada) Residents Study1993–1994. Even in Las Vegas, one-third of thepopulation does not gamble.

21. For example, a study of wagers in Minnesota(Smith and Craig, 1992; Tice, 1995) found that 1%of gamblers accounted for 50% of wagers, and that10% accounted for 80%. An Illinois study (GazelR, Thompson WN. 1996. Casino gamblers in Illi-nois: who are they? Manuscript, 1–25 (plus datasupplied by the authors)) found that 10% of bettorsaccounted for 66% of wagers. Heavy gambling isnot the same as problem and pathological gamblingeven though the revenues of P&P gamblers figuredisproportionately among the revenues of thehighest-gambling segment of the population. Whencompared with the population at large, the amountgambled by P&P gamblers implies that the share ofcasino revenues from problem and pathologicalgamblers can be as much as 1/4 to 1/2 of casinorevenues (see Grinols and Omorov, 1996). Lesieur(1998b) reported that 48.7% of casino revenues inNova Scotia came from problem gamblers, andthat 55% of revenues for casino cards and dicegames came from problem gamblers in Washing-ton. In other locations he found that percentagesranged between 26.7 and 41.4%. In Montana, 37%of the revenues of video gambling machines wasestimated to come from problem and pathologicalgamblers (Polzin et al., 1998). The Productivity



Commission (1999) reported that problem gamblersaccount for 2.1% of the adult population but one-third of all gambling revenues in Australia. Volberget al. (2001) also examined the distribution of rev-enue from different types of gamblers.

22. This description applies in Illinois and many othermidwestern states. In Minnesota, for example, onlyAmerican Indians operate casinos. In locationssuch as Atlantic City or the Gulf Coast of Missis-sippi, regulations allow entry to all as long ascertain operating requirements are met. In theselocations competitition drives economic profits tozero.

23. Non-Indian casinos paid over $2 billion in taxes tothe various states on gaming revenues in 1997. CT’stwo Indian casinos paid $236 million to the statethat year. In comparison, states generated revenuesof approximately $10 billion from net proceeds oflotteries in 1997, or $51.15 per adult.

24. See Economic Report of the President, 1999, TableB-15, column 8.

25. How should we treat demand derived from addic-tion? If addiction is not rational then its deriveddemand should be treated differently. We, there-fore, report both figures above. In the lower figure,we assumed that 32% of casino revenues are fromP&P gambling.

26. A survey of this literature and list of references canbe found in Adam Rose and Associates (1998) andthe NGISC (1999), appendix 5 on EconomicDevelopment.

27. Leven et al. (1998) provide an example of how thefocus on job creation may mislead the unwary oruntrained. They wrote,

‘This study seeks to take an objective look at theeconomic impact of the gaming industry on theMissouri economy. Where do the gaming rev-enues come from? How are they redistributed inthe economy? By how much do state and localgovernments benefit? What is the net bottom-lineeconomic impact? . . . [Gaming] does add spend-ing, income, and jobs to the Missouri economy.It should be addressed in this context.’

While the authors do not claim that the answers totheir questions constitute a cost–benefit evaluation,their plea that gambling adds ‘spending, income,and jobs to the Missouri economy’ and that ‘itshould be addressed in this context’ could easily bemisinterpreted to mean that a calculation of in-come, jobs, and employment is synonymous with acost–benefit evaluation. In their summary (p. 75)they wrote:

‘The focus of this study has been the determina-tion of whether net new output (and jobs andemployment) have been created state-wide inMissouri as a consequence of casino gamingoperations, and if so how much . . . The ‘bottomline’ is that significant additions to the Missourieconomy have been achieved. As of 1997, almost

18 000 net new jobs, $500 million in added per-sonal income, and over $750 million of addedoutput have benefited the state’s economy.’

Who would argue with such figures? Or be awarethat regardless of their accuracy, casinos in Mis-souri might fail to pass a cost–benefit test and thusbe harmful to state welfare?

28. See Grinols et al. (2000) for a complete review ofthis literature. Each of the following crimes hasbeen alleged in the literature to be associated withgambling. Index I Violent Crime (Aggravated As-sault, Robbery, Rape, Murder), Property Crime(Larceny, Burglary, Auto Theft), and non-Index Icrime such as Embezzlement and Employee Theft,Loan Fraud, Insurance Fraud, Forgery (includingcheck forgery), Tax Evasion, Tax Fraud, ConGames (Swindles, Hustling Cards, Dice or OtherGames), Bookmaking, Working in an Illegal Game,Pimping, Prostitution, Selling Drugs, and FencingStolen Goods.

29. Lesieur (1992, p. 45) and Executive Office of theGovernor (1994, p. 67).

30. National population data by age cohort are on theUS Census Bureau website http://www.census.gov/population/estimates/nation/intfile2-1.txt. As of 25August 2000, the US had a population of275 130 000. Moreover, 196 649 000 were aged 20or older.

31. See Table 2.32. See also Frank et al. (1991).33. Representative of such cases is the following ac-

count, ‘A Florida man who lost about $50 000while gambling here (Atlantic City) during the pasttwo days died Tuesday after he jumped seven floorsfrom a Trump Plaza Hotel and Casino roof ontoColumbia Place, officials said’. Brian Hickey, StaffWriter, 18 August 1999, South Jersey PublishingCo.

34. The minimum social costs of this category are thevalue of the resources spent by those trying to stealand cover up their offenses and the value of theresources spent by potential victims to decreasetheir likelihood of being victimized. There may beanother component of cost in addition, however.Social costs can be higher if the original owners ofthe property value it more than the offenders do.For example, if the owners valued their property at$1000 and the offenders who stole it sold it tosomeone who valued it at $300, there would be anadditional social loss of $700. Furthermore, if thethief is a pathological gambler and spends thewrongly acquired $300 gambling, his expendituresmay reflect addiction rather than rational choice.In that case there would be social cost equal tosome or all of the $300 because of his sub-optimalallocation of resources to the gambling sector. Last,although there is some debate about whether tocount stolen dollars as costs to all of society (whichincludes the thief) because ‘the thief gets themoney’, it is clear that the non-gambling portion ofsociety will be made worse off by such actions, andlosses to the rest of society are important in the



policy debate because they suggest that all of theabused dollars represent social costs to the non-gambling sector.

35. Westphal et al. (1999) is not used in Table 2, butsupplements Ryan et al. (1999). The South DakotaResearch Council study was completed in 1998, butaddenda were added in 1999. See also Finance andAdministrative Cabinet, Commonwealth of Ken-tucky (1999), Florida Department of Law Enforce-ment (1994), Florida Sheriffs Association (1994),Iowa Racing and Gaming Commission (1995),Lesieur (1998a).

36. As an alternative way of showing that the differencesin the totals are driven largely by the number of costcategories estimated, we compared the totals after‘filling the gaps’ in each study using the average costfor a given category from those studies that didestimate those particular costs. When doing so, thevariance in the totals decreased substantially. Thelowest totals were for South Dakota LegislativeResearch Council (1998–1999) and Thompson andQuinn (1999), $7396 and $8047, respectively. Thelargest were $25 742 by the Executive Office of theGovernor (1994) and $18 203 by Thompson et al.(1998).

37. In addition to our use of the numbers, some studies,such as Thompson et al. (1998) intentionally formedtheir original estimates conservatively to understatecosts.

38. Including the nominal value of this study wouldincrease the cost estimate for three of the four costsit estimates. Using the values adjusted for 19 yearsof price level changes would have significantly in-creased the estimates of all four costs.

39. See Table 5, p. 34.40. Our highest estimate of benefits was $75; our lowest

estimate of costs $140. Applying the per adult costsof $221 from Table 4 to the estimate of benefitsadjusted for P&P gamblers of $34 implies that casinogambling fails a cost–benefit test by a ratio of 6.5:1.

41. NGISC (1999, p. 4-4).

REFERENCES

Adam Rose and Associates. 1998. The regional economicimpacts of casino gambling: Assessment of the literatureand establishment of a reasearch agenda. Report pre-pared for the National Gambling Impact Study Com-mission (August).

Blair BF, Schwer RK, Waddoups JC. 1998. Gamblingas an economic development strategy: the neglectedissue of job satisfaction and nonpecuniary income. TheRe6iew of Regional Studies 28(1): 47–62.

Deloitte and Touche. 1992. Economic and other impactsof a proposed gaming, entertainment and hotel facility.Prepared for the City of Chicago Gaming Commis-sion.

Eadington WR. 1999. The economics of casino gam-bling. Journal of Economic Perspecti6es 13(3): 173–192.

The Evans Group. 1996. A study of the economic impactof the gambling industry through 2005. Study commis-sioned by International Game Technology, Septem-ber, iii+105.

Executive Office of the Governor. 1994. Casinos inFlorida. Office of Planning and Budgeting: The Capi-tol, Tallahassee, FL.

Finance and Administrative Cabinet, Commonwealth ofKentucky. 1999. Assessment of the economic andsocial impact of expanded gaming in the common-wealth and neighboring region. December, iii+182.

Florida Department of Law Enforcement. 1994. TheQuestion of Casinos in Florida: Increased Crime: Is itWorth the Gamble? State of Florida: Tallahassee, FL.

Florida Sheriffs Association. 1994. Casinos and crime: isit worth the gamble? A summary report and positionpaper. Florida Sheriffs Association: Tallahassee, FL.

Frank ML, Lester D, Wexler A. 1991. Suicidal behavioramong members of gamblers anonymous. Journal ofGambling Studies 7: 249–254.

Gerstein D, Murphy S, Toce M, Volberg R, HarwoodH, Tucker A, Christiansen E, Cummings W, SinclairS. 1999. Gambling impact and behavior study: reportto the National Gambling Impact Study Commission.National Gambling Impact Study Commission, (1April).

GLS Research. 1994. 1993–1994 Clark County ResidentsStudy. Las Vegas Convention and Visitor’s Authority:Las Vegas, NV.

Grinols EL. 1996. Incentives explain gambling’s growth.Forum for Applied Research and Public Policy 11(2):119–124.

Grinols EL. 1999. Distance effects in consumption.Re6iew of Regional Studies 29(1): 63–76.

Grinols EL, Mustard DB, Dilley CH. 2000. Casinos,Crime, and Community Costs. University of Illinois:University of Georgia; 1–35.

Grinols EL, Omorov J. 1996. Development or dreamfielddelusions?: assessing casino gambling’s costs and ben-efits. The Journal of Law and Commerce 16(1): 49–87.

Henriksson LE. 1996. Hardly a quick fix: casino gam-bling in Canada. Canadian Public Policy XXII(2):116–128.

International Gaming and Wagering Business. 1999.August.

Iowa Racing and Gaming Commission. 1995. DesMoines Register. 13 October. Christiansen/CummingsAssociates: Arlington, MA.

Kindt JW. 1994. Increased crime and legalized gamblingoperations: the impact on the socio-economics ofbusiness and government. Criminal Law Bulletin. 43:538–539.

Lesieur H. 1992. Compulsive gambling. Society 29(4):43–50.

Lesieur H. 1998a. Testimony before the National Gam-bling Impact Study Commission, Atlantic City, NewJersey, 22 January 1998, Institute of ProblemGambling.

Lesieur H. 1998b. Costs and treatment of pathologicalgambling. The Annals of the American Academy ofPolitical and Social Science (Gambling: socioeco-



nomic impacts and public policy, Frey JH, specialeditor) 556: 153–171.

Leven C, Phares D, Louishomme C. 1998. The eco-nomic impact of gaming in Missouri. Report to CivicProgress, St. Louis, April.

Maryland Department of Health and Mental Hygiene,Alcohol and Drug Abuse Administration. 1990. TaskForce on Gambling Addiction in Maryland. MarylandDepartment of Health and Mental Hygiene, Alcoholand Drug Abuse Administration: Baltimore, MD.

McCleary R, Kenneth C, Feng W, Merrill V, Napoli-tano C, Moles M, Graffeo B. 1998. Suicide andgambling: An analysis of suicide rates in US countiesand metropolitan areas. Report to the AmericanGambling Association. Manuscripts.

Mirage Hotel. 1993. Response of the developers of theproposed Fox River resort to the West Dundee river-boat project task force.

National Gambling Impact Study Commission. 1999.National Gambling Impact Study Commission FinalReport www.ngisc.gov [accessed 18 June 1999].

Phillips DP, Welty WR, Smith MM. 1997. Elevatedsuicide levels associated with legalized gambling. Sui-cide and Life-Threatening Beha6ior 27: 373–378.

Politzer RM, Morrow JS, Leavey SB. 1981. Report onthe societal cost of pathological gambling and thecost–benefit/effectiveness of treatment. The JohnsHopkins Compulsive Gambling Counseling Center,Presented at Fifth National Conference on Gamblingand Risk-Taking.

Polzin PE, Baldridge J, Doyle D, Sylvester JT, VolbergR, Moore WL. 1998. Final Report. Presented to theMontana Gambling Study Commission, 30 September1998, Bureau of Business And Economic Research(University of Montana-Missoula).

Productivity Commission. 1999. Australia’s gamblingindustries. Final Report, Summary (26 November,Report No. 10).

Quinn FL. 2001. First do no harm: What could be doneby casinos to limit pathological gambling. Manage-rial and Decision Economics 22: 133–142.

Ryan TP, Speyrer JF, with Beal ST, Burckel DV,Cunningham BR, Kurth MM, Scott LC, Wall JL,

Westphal JR. 1999. Gambling in Louisiana: a benefit/cost analysis. Prepared for the Louisiana GamingControl Board, Louisiana State University MedicalCenter (April).

SMR Research Corporation. 1997. The PersonalBankruptcy Crisis, 1997: Demographics, Causes, Im-plications & Solutions. SMR: Hackettstown, NJ.

Shaffer HJ, Hall MN, Vander Bilt J. 1997. Estimatingthe Pre6alence of Disordered Gambling Beha6ior in theUnited States and Canada: A Meta-Analysis. HarvardMedical School: Boston.

Smith F, Craig WJ. 1992. Whos in for how much?CURA Reporter. Center for Urban and RegionalAffairs 22: 11–14, 16 (plus data supplied to theauthors).

South Dakota Legislative Research Council. 1998. Eco-nomic and fiscal impacts of the South Dakota gamingindustry. 1–146+Attachment: Letter of 7 January1999 to Mr. Terry Anderson, Director, South Da-kota Legislative Research Council, Table South Da-kota Total Estimated Incremental Social Costs.

Thompson WN, Gazel R, Rickman D. 1996. The socialcosts of gambling in Wisconsin. Wisconsin PolicyResearch Institute Report 9(6): 1–44.

Thompson WN, Gazel R, Rickman D. 1998. Socialcosts of gambling: A comparative study of nutmegand cheese state gamblers. Twelfth National Confer-ence on Problem Gambling, NV 1998. Gaming Re-search and Re6iew Journal 5(1): 1–15.

Thompson WN, Quinn FL. 1999. An economic analy-sis of machine gambling in South Carolina. TheEducation Foundation of the South Carolina PolicyCouncil, May 1999.

Tice DL. 1995. Why the press hates gaming. CasinoExecuti6e August: 57–58.

Volberg R, Gerstein DR, Christiansen EM. BaldridgeJ. 2001. Assessing self-reported expenditures on gam-bling. Managerial and Decision Economics 22: 77–96.

Westphal JR, Johnson LJ, Stevens L. 1999. Estimatingthe social costs of gambling disorders in Louisianafor 1998. Study Performed Under Subcontract FromUniversity Of New Orleans, UNO c313-204-124 (23March).


decis econ 22: business profitability versus social...

Documents