Ongoing Challenges in CostBenefit Analysis of Casino Gambling

1
Ongoing Challenges in Cost-Benefit Analysis of
Casino Gambling

• Colloquium on the European and National
  Perspectives of the Regulation of Gambling
• Tilburg University, The Netherlands
• 23 November 2006
• Douglas M. Walker, Ph.D.

2
Gambling research in the U.S.

• Since casino gambling began expanding outside
  Nevada and New Jersey in the early 1990s,
  gambling research has increased.
• Politicians are interested in benefits relative
  to costs of legalization.
• Casino legalization is a state-level policy.
• State governments have been facing record-level
  budget deficits.
• Gambling research has been influential, as it is
  commonly referred to by politicians and the media.

3
The economics of gambling literature

• Some well-known studies include
• U.S. House of Representatives (hearing, 1994),
• National Gambling Impact Study Commission (NGISC
  1999)
• Australian Productivity Commission (APC 1999)
• National Opinion Research Center (NORC 1999)
• National Research Councils Pathological
  Gambling A Critical Review (1999)

4
Gambling literature, cont.

• Earlier studies these are mostly not
  peer-reviewed, and use arbitrary assumptions.
• Politzer et al. (1985)
• Maryland Task Force (1990)
• Casinos in Florida (1995)
• Goodman (1994a, 1995b)
• Grinols and Omorov (1996), Grinols (1994a, 1995a)
• Kindt (1994, 1995)
• Thompson, Gazel, and Rickman (1996, 1997, 1999)

5
Gambling literature, cont.

• More recent studies, but still flawed
• Thompson and Quinn (2000)
• Managerial and Decision Economics (2001), edited
  by Grinols and Mustard
• Grinols and Mustard
• Gazel, Rickman, and Thompson
• Kindt
• Comments on Kindts paper published, MDE 2004
• Grinols (2004)
• perhaps most influential

6
Gambling literature, cont.

• Thompson and Schwer (2003/2005)
• Study of Las Vegas that got publicity
• Methodology by Thompson et al. (1996)
• Hall Aitken (2006)
• Study of U.K.
• PolicyAnalytics (2006)
• Study of Indiana, with a focus on crime
• Relies on Grinols (2004)
• Grinols and Mustard (2006)
• Review of Economics and Statistics study on crime

7
Gambling literature, cont.

• Research in this area has been criticized by
• National Research Council (1999)
• Walker and Barnett (1999
• Shaffer et al. (2001)
• Federal Reserve Bank of Minneapolis (2003)
• Eadington (2004)
• Walker (2007)
• and elsewhere

8
Economic perspective

• Economic perspective on costs-benefits
• Represented by Collins and Lapsley (2003),
  Eadington (2003), and Walker and Barnett (1999).
• This perspective has a foundation in the
  economics literature.
• Detailed description in Walker (2007)

9
Economic perspective on benefits

• The benefits of legalized casino gambling
  include
• Tax revenues (transfers, but easy to avoid)
• Employment and wage effects
• Capital and labor inflow to a region
• Potential economic growth
• Walker and Jackson (1998, 2007)
• Consumer benefits (CS, variety benefits,
  competitive pressure on entertainment)
• This is the most commonly ignored benefit.
• Consumer sovereignty
• Some may argue that increased freedom of choice
  is, itself a benefit.

10
Benefits, cont.

• There have been relatively few academic studies
  of the benefits of gambling.
• However, these are arguably easier to measure
  than the costs.
• Politicians may care mostly about tax revenues,
  and so do not demand detailed studies of
  benefits.
• The most important benefits (CS, other consumer
  benefits) are most difficult to measure.

11
Economic perspective on costs

• There is debate over all aspects of the costs of
  gambling
• definition
• measurement
• Social cost is a decrease in aggregate societal
  wealth (Walker and Barnett, 1999).
• Wealth need not be in monetary terms it is
  utility.
• Transfers of wealth are not social costs.
• Private (internalized) costs are omitted.
• Most researchers agree that the costs are caused
  by pathological gamblers.
• There are different severities of the affliction,
  ignored here
• Compulsive, pathological, problem gamblers, etc.

12
Costs, cont.

• Even if the definition is conceptually valid,
  measurement is difficult, if not impossible.
• There is no agreement among researchers from
  different disciplines.
• Other social scientists do not like the exclusion
  of transfers and other private costs.

13
Why the focus on costs?

• Costs are typically the focus by media,
  politicians, and researchers.
• This may be a result of the existing illegal
  status of casinos
• Are the benefits of legalization/expansion worth
  the costs?
• Estimates of expected tax revenues, employment
  increases, and changes in average wages can be
  easily produced.
• The social/economic costs are more elusive, both
  in defining them and in measuring them.
• Costs are closely related to prevalence of
  pathological gambling.

14
Calculating social costs

• Most of cost studies use this formula
• estimated annual cost per pathological gambler
  X prevalence estimate () X population
  estimate
• estimated annual social cost of gambling

15
Variation in cost estimates

• Authors rarely define social cost before
  estimating it.
• Studies use ad hoc methodologies, resulting in
  cost estimates ranging from US9,000-50,000
  (7.000-39.000) per pathological gambler per
  year.
• Such a large range indicates that the
  methodologies are not the same.

16
Examples of problematic studies, discussed
hereafter

• Grinols (2004)
• Politzer et al. (1985)
• Thompson et al. (1996, 1997, 1999)
• Thompson and Quinn (2000)
• Thompson and Schwer (2003)
• Grinols and Mustard (2006)

17
Grinols (2004)

• Gambling in America demonstrates many of the
  problems in the literature.
• His discussion appears to be comprehensive and
  unbiased, but it is neither.
• He ignores work with which he disagrees.
• See Eadington (2004) for a discussion.
• He overstates the costs and discounts the
  potential benefits from gambling.

18
Grinols (2004), cont.

• Grinols estimates for the U.S.
• Annual cost to society of one pathological
  gambler 10,330.
• Annual cost per problem gambler 2,945.
• On a per adult basis, the average cost is
  estimated to be 219. (p. 175)
• After some analysis, it is clear that these
  estimates are meaningless.

19
Grinols (2004), cont.

• Cost estimates were derived by averaging over
  the available original studies for each
  category of social cost, adjusting to 2003
  dollars, and summing over cost types (p. 176).
• The studies used by Grinols to derive his cost
  estimates include
• 9 studies, none peer-reviewed.

20
Grinols (2004), cont.

• Grinols estimate is based (in part) on
• Politzer et al. (1981)
• Thompson et al. (1996)
• Thompson and Quinn (1999/2000)
• Thompson and Schwer (2003)
• These studies have been criticized by a number of
  researchers.

21
Grinols (2004), cont.

• Grinols ignores the controversy over how to
  define and estimate social costs.
• His presentation gives the impression that
  because the cited research is original it is
  legitimate.
• He does not explain differences in the studies or
  their potential flaws.
• Lets look at some aspects of the studies used by
  Grinols

22
Politzer et al. (1985)

• This concept is vague.
• What is an essential purpose?
• Is it defined the same for Bill Gates and Joe
  Sikzpak?
• Could be interpreted as the total bets placed
  (handle), much larger than actual losses.
• Any money borrowed to gambler is considered
  abused dollars.
• Use of this concept allows one to arrive at a
  very high social cost estimate.

• Define abused dollars
• the amount of money obtained legally and/or
  illegally by the pathological gambler which
  otherwise would have been used by the
  pathological gambler, his family, or his victims
  for other essential purposes. These abused
  dollars include earned income put at risk in
  gambling, borrowed, and/or illegally obtained
  dollars spent on basic needs and/or provided to
  the family which otherwise would have been
  covered by that fraction of earned income which
  was used for gambling, and borrowed and/or
  illegally obtained dollars for the partial
  payment of gambling related debts.

23
Thompson et al. (1997)

• Estimated social cost per compulsive gambler is
  9,469 (p. 87), but
• They do not define social cost, and include
  whatever negative effects they can measure with
  their survey on Gamblers Anonymous (GA) members.
• Using GA members in Las Vegas to is not a random
  sample of problem gamblers.
• Most of the costs are transfers.

24
Thompson et al. (1997), cont.

• Costs included
• Employment (lost work hours, unemployment
  compensation, lost productivity), 2941
• (Lost hours, productivity are internalized
  unemployment compensation is a transfer or
  fiscal externality.)
• Bad debts, 1487
• (Transfer of wealth)
• Civil court costs (bankruptcy, other), 848
• Criminal justice (theft, arrests, trials,
  probation, incarceration), 3498
• (Thefts are transfers)
• Therapy, 361
• Welfare, 334 (Transfer of wealth)
• TOTAL 9469
• (Revised to 2974 after eliminating transfers and
  internalized costs. Walker and Barnett 1999)

25
Thompson and Quinn (2000)

• Costs include money leaving South Carolina for
  purchase of video poker machines (46.5 million
  each year).
• Then any purchase represents a social cost?
• Money leaving the state is not a cost since the
  buyer values what is bought more than the money
  spent.
• A transaction at the grocery store is similar.

26
Thompson and Schwer (2003)

• They estimate the annual social cost of per
  pathological gambler in Las Vegas at 19,085.
• Accounting for estimated prevalence and
  population, the total annual cost estimate is
  between 301 and 470 million.
• Their cost figure is the result of numerous
  arbitrary assumptions.
• Well examine a component of this estimate later

27
Studies cited by Grinols (2004)

• Among the studies cited by Grinols, there is
• Disagreement about the types of costs to include
• Sometimes costs are excluded because of
  measurement difficulties.
• Disagreements on the estimated values of the
  individual cost categories
• The estimated values are often the result of
  arbitrary assumptions (see Walker 2003)

28
Adjudication costs (criminal and civil justice)

• The monetary estimates for these costs among the
  Grinols-cited studies are
• 3,619 (Maryland, Politzer et al. 1981)
• 733 (Wisconsin, Thompson et al. 1996)
• 568 (Conn., Thompson et al. 1998)
• 31 (S. Dakota, 1999)
• 420 (Louisiana, Ryan et al. 1999)
• 266 (S. Carolina, Thompson and Quinn 2000)
• 51 (Nevada, Thompson and Schwer 2003)
• Such a huge variation across states indicates
  these studies are not measuring the same thing
  the same way.

29
Thompson and Schwer (2003) cost estimate for
lost work time

• 50 of 89 (or 56) of Gamblers Anonymous survey
  respondents indicated they had missed work
  because of gambling.
• An average of 17.22 hours missed during each
  month due to gambling.
• The average loss is 9.67 hours/month allocated
  over the 89 respondents.
• This amounts to 116.1 hours per year, calculated
• (50 x 17.22)/89 x 12
• 116.1 hours is multiplied by 15/hr, the hourly
  rate based on Thompson et al.s (1996) use of an
  average annual pay rate of 23,610.
• This annual rate of 23,610 is based on the U.S.
  average salary (Thompson et al. 1996, p. 17).
• This results in an estimated cost of 1,742 for
  lost work time.

30
Grinols and Mustard (2006)

• This recent paper in Review of Economics and
  Statistics has numerous problems
• In calculating crime rates, they include crimes
  committed by visitors, but omit visitors from the
  measure of population at risk.
• Their estimate of costs of crime seems arbitrary.
  
• Other issues are discussed in Walker (2007)

31
General Problems in Estimating Social Costs of
Gambling

• Regardless of the definition of social costs
  chosen, there are measurement problems
• Counterfactual scenario
• Comorbidity
• Survey data and fungible budgets
• Government expenditures
• If a particular study does not address these, it
  should probably be ignored.

32
Counterfactual scenario

• For policy purposes, the economic and social
  effects of legalized casinos must be compared to
  the case in which casinos are not legal.
• This may be difficult to know in terms of
  employment, economic growth, etc.
• Unless the counterfactual is what is already
  happening
• In some stagnant economies, one could argue that
  no other industry would have come (e.g.,
  Mississippi Gulf Coast).

33
Counterfactual scenario, cont.

• In terms of problem/pathological gambling
• If casinos were not legal in the state, would
  people just go to other venues?
• If casinos were not available, would the
  pathological gamblers with coexisting disorders
  have more serious alcohol or drug problems?
• If yes, then it is possible that the gambling
  legalization would lead to lower social costs
  even if more people would become pathological
  gamblers.

34
Comorbidity

• This may be the biggest problem in cost-benefit
  analysis.
• Studies have found that many pathological
  gamblers have other disorders.
• Petry, Stinson, and Grant (2005, p. 569) find
• 74.2 have alcohol use disorders
• 38.1 have drug use disorders
• 41.3 have anxiety disorders
• 28.5 have obsessive-compulsive personality
  disorder
• How do you allocate the social costs to the
  different problems when many pathological
  gamblers have other disorders?
• Most of the published studies ignore this,
  resulting in overestimates of the social costs
  attributable to gambling.

35
Surveys and fungible budgets

• Diagnostic instruments and cost estimate surveys
  ask various questions about potential problem
  gamblers behavior.
• Blaszczynski et al. (2006) explain that without
  explicit instructions, respondents use different
  strategies in estimating their gambling losses.
• The result may be serious biases in gambling
  losses reported in the literature (p. 128).
• Walker (2007) discusses the inability to
  attribute specific expenditures to specific
  revenue sources.
• Budgets are fungible.

36
DSM-IV and SOGS criteria

• DSM-IV items
• 8. has committed illegal acts such as forgery,
  fraud, theft, or embezzlement to finance
  gambling.
• If a person cannot estimate gambling losses, can
  they correctly attribute their crimes to its
  cause?
• 10. relies on others to provide money to
  relieve a desperate financial situation caused by
  gambling.
• What if the person bought an expensive car, or is
  otherwise financially irresponsible?
• How do clinicians deal with this possibility?

37
DSM-IV and SOGS criteria, cont.

• SOGS items
• 14. Have you ever borrowed from someone and not
  paid them back as a result of your gambling?
• What if you dine out at fine restaurants too
  often? Will you attribute your financial problems
  to the proper cause?
• 16a-k. If you borrowed money to gamble or pay
  gambling debts, who or where did you borrow
  from? (many possible responses)
• How can a person attribute specific spending to
  specific sources of income, unless there is only
  one source of income?

38
Caveat

• Clinicians may argue that it does not matter
  whether it is possible to identify sources of
  spending. The screening instruments may serve
  their purpose anyway.
• Perhaps what matters is if the person thinks
  he/she has a problem.

39
Relevance of diagnostic tools to cost estimates

• The reason these types of questions are relevant
  to policy is that the questions about total
  losses and sources of money used to gamble are
  used in cost estimation studies.
• Abused dollars
• Bad debts
• Bailout costs
• Bankruptcy costs

40
Fiscal externalities

• Browning (1999) discusses how to handle costs of
  government policies.
• His application is to health care costs
  associated with smoking.
• Walker (2007) discusses this in the context of
  treatment of pathological gambling.

41
Fiscal externalities (cont.)

• The issue is how to classify costs of government
  for pathological gambling behavior.
• Is the monetary cost incurred by government a
  cost of pathological gambling, or a cost of our
  philosophy on government and policy?
• Suppose one country has very generous treatment
  reimbursement (150), while another countrys
  government pays only 50 of the costs.
• The social cost of the first country would be
  triple that of the second.

42
Fiscal externalities, cont.

• If government expenditures are social costs,
  then we can eliminate costs by eliminating
  spending.
• Obviously this isnt the correct way to think
  about social costs.

43
Critiques of C-B Analyses

• Reuter (1999) and Kleiman (1999) are
  enlightening.
• They argue that research effort would be better
  spent on the effects of policy changes.
• Applied to gambling, since gambling is already
  widely available, what can we do to minimize the
  costs/harms?
• This is similar to what public health
  perspectives advocate harm minimization.

44
Should we measure costs and benefits of casinos?

• There are motivations for this type of research
• Politicians need data to inform and defend their
  decisions.
• Policy entrepreneurs want to influence policy
  (Krugman 1996)
• Researchers benefit if there is a problem to
  investigate.
• But do cost-benefit studies provide good
  information? And are they important?
• Policymakers might make better decisions without
  such studies.

45
Should we measure costs and benefits of casinos?
(cont.)

• Peter Collins book, Gambling and the public
  interest (2003) has an excellent discussion on
  this issue.
• Policy discussions in the U.S. neglect the
  fundamental issues
• Property rights
• Freedom of choice/consumer sovereignty
• The role of government in a free society
• We instead focus on expected tax benefits
  compared to a murky understanding of social
  costs.

46
Conclusions

• Unfortunately, researchers have not developed
  valid or practical methods for estimating costs
  (and benefits) of casino gambling.
• Given that some researchers seem intent on
  influencing policy, as explained by Eadington
  (2004), perhaps we should focus more on the more
  basic issues of property rights, freedom of
  choice, and the proper role of government.
• At this stage in the development of research in
  this area, policymakers are advised not to put
  too much stock in existing studies.
• If one does rely on such studies, seek an
  external critique of the study to be aware of
  potential problems that may not be explicitly
  identified in the study.

47
Contact information

• Doug Walker
• Associate Professor of Economics
• Georgia College, CBX 14
• Milledgeville, GA 31061
• USA
• dougwalker1_at_gmail.com
• http//walker-research.gcsu.edu
• (This link has presentation slides and references
  from the presentation.)

• The Economics of Casino Gambling
• by Doug Walker
• Springer (Jan. 2007)
• ISBN 3-540-35102-7