Caesars at Woodbury: Problem Gambling ? No Problem







Summary: This is a critique of Attachment IX.A.2.a_A2  in  section  07 – IX.A. Assessment of Local Support and Mitigation Local Impact  of the  application by Caesars Entertainment to the New York State Gaming Facilities Location Board.    Being an epidemiologist and physician familiar with problem gambling as a public health problem,  I found  Attachment IX.A.2.a_A2  extremely biased in downplaying,  to near zero.  the possible health impacts  of a new Woodbury Casino.   The report asserts  as follows:

  • no socio-economic costs of pathological gambling and problem gambling warrant $ consideration, as none can be quantified so that all parties are close to agreement.
  • making casinos more convenient hardly  increases the prevalence of pathological gambling and problem gambling in the surrounding population in the long run.
  • the only population within a 50 mile radius of Woodbury that is  at theoretical  risk of having even a temporary surge in prevalence of pathological gambling and problem gambling  is that of Orange, Dutchess and Putnam Counties.
  • efforts by Caesars elsewhere to address “problem gambling” have been highly successful and will minimize “problem gambling” in southern New York State.

The  report greatly understates the possibility of harm to residents of the region due to a casino in Woodbury to residents of the region.  This essay addresses the first three  points in the above order.  The fourth  I have discussed in an e-mail to the NYS Gaming Commission last April.


In reading the Caesars Entertainment Inc application for a Woodbury casino I focused as a physician versed in public health  on the 40- page report   Study of Addiction and Public Health Implications of a Proposed Casino and Resort in Woodbury New York by Bo J. Bernhard Ph.D., Khalil Philander Ph.D., and Brett Abarbanel Ph.D.

The authors are all experienced consultants for gambling-related  enterprises. Two are senior members of the International Gaming Institute (IGI) at University of Nevada at Las Vegas. This is a highly polished presentation by experts who know the field but hide large tracts of it from view.   It  dismisses or never mentions four crucial facets of the ecology of pathological gambling and problem gambling. The report basically concludes that

  • socio-economic costs of pathological gambling and problem gambling don’t warrant consideration, as none can be quantified so that all parties are close to agreement.
  • making casinos more convenient does not much increase the prevalence of pathological gambling and problem gambling in the surrounding population in the medium  run of 2 to 4 years.
  • the only population within a 50 mile radius of Woodbury that is now under-served by racinos or casinos (and hence at theoretical  risk of having even a temporary surge in prevalence of pathological gambling and problem gambling) is that of Orange, Dutchess and Putnam Counties.
  • efforts by Caesars elsewhere to address “problem gambling” have been highly successful and can be relied on to minimize “problem gambling” in southern NY

The report nowhere mentions a statistic often cited by opponents of predatory gambling but never addressed head-on by casino advocates and never refuted: 40-50% of revenue at the average casino comes from pathological and problem gamblers, who comprise perhaps 12-15% of its customers, maybe 4% of all adults. [] For the casino lobby to refute the statistic (if it is refutable) they would have to acknowledge that they can spot pathological and problem gamblers among their “visitors” while those persons are still active customers. This means before the person has loudly threatened suicide within an employee’s hearing or left town suddenly or thrown an ugly scene on the “gaming floor” or been arraigned or jumped.

Casinos will not acknowledge they have any  ability to spot problem or pathological gambling signs and symptoms that are not florid and end-stage. Why not? To move in even gently on such persons would risk offending them so they would go elsewhere or sending them to premature recovery before they have been “played to extinction.” [ ] .  To recognize the problem gamblers before they are end-stage yet not do anything for them  would reveal how insincere are the “preventive measures.” .

The casino cartel does not deny that its net revenues follow the Pareto principle: most come from a small proportion of gamblers. What casino promoters won’t say is what proportion on the average of that small proportion are pathological gamblers or problem gamblers. The promoters just do not want to know who among their customers is a problem gambler or pathological gambler until the gambler hits bottom or worse.   Promoters and detractors alike recognize that not everyone who loses a lot of money over time at casinos is a problem gambler. Anti-casino activists hold that most are; the American Gaming Association counters that most are affluent people having fun with their disposable income.

Assuming the central statistic is close to truth, casinos are not motivated to sincerely counter problem gambling and pathological gambling.  A successful effort to do so would lower their revenues by 40-50%.  Nor is government motivated; lower casino gross gaming revenues   would reduce  government’s share  by a like amount.

The report prepared for Caesars (in this no different from all the literature on problem gambling) also does not recognize that “unchanging prevalence” of problem and pathological gambling requires the formation of replacement problem gamblers and pathological gamblers to fill the shoes of those who have recovered, died, moved far away or are no longer free-living. What might appear a steady state is built on creating new problem gamblers. The more effective  the casino is at encouraging current problem gamblers and pathological gamblers into lasting recovery before they have fiscally and emotionally wiped out themselves and and ten people around them, the faster it must generate replacement problem and pathological gamblers to keep up its high profit margins.

I will now cover the first three bullet points above in more detail. The fourth bullet point I wrote about in the above-mentioned letter to the Gaming Commission.

Bernhard and colleagues devote the first few pages to the case that that socio-economic costs of legalized gambling cannot be quantified for econometrics and therefore will not be considered in their analysis. They imply that all attempts at estimating costs are inflated by intangible dimensions of social cost, such as emotional suffering. Such dimensions were in fact entered into some estimates years ago. The Australian Productivity Commission report of 1999 [ ] briefly reviewed literature on social costs. The report notes that estimates for social costs per year for “problem gamblers” [some estimates with no distinction between “pathological and problem ] were as high as $52,000. The report also mentions (see box 9.2, p 353) an estimate of $13,600 U.S. per year in 1981 by the respected researcher Rachel Volberg. The next page cites an estimate from the University of Manitoba in 1996 of $56,000 (Canadian) per year per pathological gambler. The APC Report is not referred to by Bernhard and colleagues.

Bernhard and colleagues overlook the most comprehensive and  most recent work on the econometrics of legalized gambling, Earl Grinols’s Gambling in America: Costs and Benefits (Cambridge University Press, 2004). Grinols is Distinguished Professor of Economics at Baylor. His book carefully and clearly quantifies selected and aggregated socioeconomic costs of legalized gambling. Bernhard et al. do refer to an earlier (2001) article by Grinols and Mustard, . [“Social Profitability versus Business Profitability: Evaluating Industries with Externalities, the Case of Casinos.” Managerial and Decision Economics 22: 143-162.] . Some of this article was reworked into the 2004 book.   Prof. Grinols’s choices on what dimensions to include and what values to assign to them may be disputed, but they all have an empiric basis. not a “what would you pay to avoid this?” basis. Bernhard and colleagues pit against the 2001 article opinions of Prof. Douglas Walker, an economist at the College of Charleston who has been a consultant to the American Gaming Association. See Dr Walker’s 2007 white paper for the AGA The Bernhard et al report writes

“Among other issues, Walker found that much of the data cited examined disparate (earlier) historical periods, when casino development was quite different than it is today.

Furthermore, many of the works cited were not subject to double-blind peer-review, and hence their reliability and trustworthiness were called into question. [emphases added]

Unfortunately, this leads the top researchers in this field [emphasis added] to question whether the current literature has generated any social cost estimates that policymakers can trust. In our view, there simply are not any “sound dollar estimates” of the costs of problem gamblers that can provide a strong foundation for policy decisions.

At this stage, then, we agree with the most careful analysts, and conclude that more harm than good would result from a pecuniary measure of social cost that would necessarily rely upon discredited research.”  [end of quotation]

Bernhard and colleagues thus imply that Grinols’s work has been discredited, being based on data too old and factoring in valuations outside the ken of econometrics. The work may be scoffed at by them, who speak ex cathedra as “the most careful analysts” and “the top researchers.”  Thus discounting it, however, is pure bias. It is not based on data that are too old or on valuing intangibles. The Grinols and Mustard article, and the 2004 book by Grinols left out of their calculations the highest-cost estimate (Politzer et al, 1981) because it had been done long before the other eight tabulated in Grinols’s 2004 book. These eight came from seven different states and one national sample. None was done by a group with an anti-gambling agenda. They were dated 1994-2003, not so long before Dr. Walker was writing in 2007. Grinols acknowledges they were not peer-reviewed (though what Bernhard et al mean by “double-blind peer-review” is mumbo-jumbo).

Some of the studies in Grinols’s 2004 book were done for state agencies such as the Louisiana Gaming Control Board (Ryan et al 1999) or the Executive Office of the Governor of Florida (1994). The study with the lowest estimated socio-economic cost per pathological gambler was done by National Opinion Research Center at the behest of the National Gambling Impacts Study, commissioned by the U.S. Congress. One highly-respected researcher was an author on four of the other studies. This was William N. Thompson, Professor of Business Administration at  UNLV since 1988, author of (among many other books) Gambling in America: An Encyclopedia (2001). In the 2004 book Grinols developed averages of estimates for costs of pathological gambling from the eight studies, controlled for co-morbidity to the extent permitted by the original data.  His figure, $10, 330 (dollars of 2003) per year was between the highest and the lowest estimates of the eight.

That mythical creature, the rational person, would actually find Grinols’s case quite reasonable. It is crucial to know that Grinols did not include in his cost estimates several quantities entered into the higher-cost figures like $52,000. Bernhard and colleagues want their readers to think otherwise. They want readers to think that Grinols included valuations for emotional impacts.

  • There is no line in Grinols’s table (pp. 172-173) for the ill-gotten gains of indictable crimes, such as embezzlement or robbery. Though these can be objectively measured, they are treated by an economist as transfer of wealth.
  • There is no line for the emotional and psychological havoc wrought by pathological gamblers on families, friends and associates. Divorce, separation, psychological abandonment, loss of ability to trust – no value can be arrived at for these losses.  They are not in Grinols’s estimates.
  • Suicide in Grinols’s view cannot be quantified to where a panel of economists can reach approximate consensus (though in wrongful death suits re particular cases a jury or judge may do so). Therefore suicide, the worst possible outcome of gambling addiction,  has no place in Grinol’s figures, nor does murder (much less common an event in this population). For that matter, actual suicide was not accounted for in earlier higher-cost figures. No one ventured to evaluate suicide econometrically for cost-benefit analysis.

It is crucial also to understand, however, that Grinols does include “abused dollars.” Some observers take exception to this, saying abused dollars should be treated as a transfer of wealth. “Abused dollars” [sometimes called “bailout money”] are monies taken from one party who has at least an equal right to them by another, but without a criminal prosecution.   They can be personal “loans” never repaid; siphoning from a joint checking account or a spouse’s IRA; diversion of insurance or mortgage payments; draining a grandchild’s college trust account; or selling home furnishings. A gambler with job and salary who stints on food or lunch money or clothing for the kids   abuses what people might say is his or her own money. Isn’t it just as much theirs?   .

Two of the eight studies Grinols cites in the 2004 book have no figure for abused dollars. In the other six the proportion of total quantifiable socioeconomic cost per pathological gambler per year ranges from 11% to 60% with an unweighted average of 39%. In Grinols’s 2004 total estimate abused dollars come to a 28% share.  Thus, if a panel of experts declared that abused dollars should not be reckoned as a cost of gambling, Grinols’s overall estimate for pathological gambling would be lowered by 28% .  This yet-“harder” figure is still thousands of dollars per year above the zero valuation that Bernhard and colleagues proffer.

Abused dollars are not directly borne by the government, as are costs of welfare or of running the justice system. Yet they are borne by society.  When a wife whose gambler husband left discovers that he hadn’t paid the mortgage for three months and the marshalls are at the door, that is a quantifiable socioeconomic cost. It has the value of those unpaid installments.

Bernhard and colleagues have on their own decided who are the “top” researchers and whose work has been “discredited” without naming names or giving valid reasons. That arrogance discredits their own report.

Second bullet point: prevalence studies of pathological gambling and problem gambling and the convenience effect.  The report cleaves to the “adaptation hypothesis” promulgated by members of the Division on Addictions at the Cambridge Health Alliance. This says that when a new gambling locus arrives,  the prevalence of pathological gambling and problem gambling nay rise locally for a time, then drifts down to status quo ante. At a supra-regional level of observation the hypothesis is proclaimed by members of the Division on Addictions at the Cambridge Health Alliance when they write that the prevalence of problem gambling nationwide has not gone up noticeably between the mid-70s (very few casinos nationally) and 2005 (many casinos across the nation). I cannot refute this dictum but have written elsewhere [ why it should not be accepted as fact if based just on the three national surveys cited.

The authors adduce an article by DW Black et al. Prevalence of problem gambling in Iowa: Revisiting Shaffer’s adaptation hypothesis. Ann Clin Psych Nov 2012, 24(4): 279-284) to buttress the adaptation hypothesis re the State of Iowa. This study, however, was according to the authors “not strictly comparable in sampling methods” to earlier studies by Volberg and colleagues with which they compare its findings.   The prevalence of pathological gambling in the 2012 report was 1.4%, less than the 1.9% observed in the 1995 Iowa replication study. The prevalence of problem gambling in the 2012 report was 2.2%, below the 3.5% observed in the 1995 study. The overall level of gambling participation was lower in the 2012 report (13.5% vs. 22.9%). Since many casinos became active between the mid-1990s and the study years (2006-2008), while prevalences of specified gambling behaviors seemed to have declined, Frank Fahrenkopf (then head of AGA) in a published letter to the New York Times called the report evidence that more casinos do not lead to more gambling problems.

Mr. Fahrenkopf’s conclusion, dutifully printed by the New York Times after a Clyde Haberman column about problem gambling, has no sound  basis. The figures are based on “lifetime” prevalence, not past-year activity. If someone has ever been in any of the categories, he or she is there lifelong even if in recovery. An apparent 25-33% drop in lifetime prevalence of any of the above categories over a ten-year period can best be explained, therefore, by reasoning that the sampling frames of the two studies are different, with the later study tapping a frame in which gambling at all levels is less than would be found in a true random sample of Iowa adults. The authors acknowledge this possibility, saying “the study was not designed as a survey to be generalized to the state population” [as had been the Iowa studies reported in 1989 and 1995]. The only alternative explanation (assuming truthful self-reports) is that a huge proportion of the gamblers in each category as of the early 1990s have died, moved out of state, are incarcerated or have no telephone.

It is not clear why the authors decided the results might be valid for the state population and therefore published the article.This study is flimsy support for the adaptation hypothesis and its corollary of stable prevalence.

Credible evidence for lack of recent rise in prevalence of problem gambling nationally is cited by Bernhard et al. This is the recently–reported study by Welte et al, [Gambling and Problem Gambling in the United States: Change between 1999 and 2013 J. Gambling Studies published on line June 2014] . The authors found no significant increase 1999-2010 in the prevalence of persons with SOGS (South Oaks Gambling Survey ) score of 3 or greater. This study, however, did confirm the association found in an earlier study by the same researchers [J Gambling Studies 2004 vol 20] between residence near a casino and prevalence of problem gambling. This is an ecological correlation, not necessarily a causal one-way relationship. It is consistent with a distance or convenience effect.

Never forget: even if prevalence of pathological gambling is rock-stable over a period of years in a defined population, that  requires the formation of new pathological gamblers. The goal in prevention of problem gambling should not be to keep a steady prevalence but to drive the prevalence down year by year to zero. Stable prevalence is not a win.

Bernhard et al make use of an important study from Quebec [ Jacques and Ladouceur 2006 “A Prospective Study of the Impact of Opening a Casino on Gambling Behaviors: 2 and 4-year Follow up.” Can J. Psychiatr vol 51 no 12 764-773]. This followed over time a cohort of randomly-selected residents near a new casino and also followed a control cohort from a city without a nearby casino. This research did support the “adaptation hypothesis” of initial rise followed by return to near pre-casino levels. The study is unusual in having observations on incidence (new cases per time interval), though there is a puzzling anomaly: incidence over four years of current probable pathological gambling was zero. . Attrition in follow-up from the Hull cohort (exposed) was high  (44% pre to post-test) enough that the possibility of differential loss (higher for those who developed gambling problems) could not be ruled out. This is a universal problem in longitudinal studies.   Funding was received in part from Loto-Quebec

Bernhard et al smugly wrap up thus their discussion of studies of prevalence over time: “At the very least, this literature suggests that these problem gambling impacts are more complex than previously assumed, and the notion that problem gambling rates inexorably rise as exposure increases has been debunked (page 42). [emphases added]”

Third bullet point: Geographic region at risk for new problem gambling:   Based on the Quebec findings,  Bernhard et al project (mid-range estimate) that four years after the opening of the proposed Woodbury casino there will be an additional 344 problem gamblers in the Counties of Dutchess, Orange and Putnam. They look, however, only at those three counties, as if the casino site had not been chosen for its nearness to Westchester County and New York City and northern New Jersey.   The total populations of these three counties in 2010 made up less than 15% of the total NYS population within the 50 mile radius of Woodbury. The authors must assume that the large piece of NYC within the 50 mile ring already has so many casino and racino opportunities (e.g. Yonkers, Aqueduct, Monticello, Atlantic City, Poconos, Connecticut) that no one in it will it develop a gambling problem. This is saying that those customers from NYC and environs who are not already problem gamblers, when they switch from their previous favorite place to Woodbury will be already immunized against becoming problem gamblers. This is not credible. It is also not credible that no one within the 50 mile radius but outside the three counties who is not a casino-racino goer in late 2014 will become a problem gambler if Caesars opens at Woodbury. The authors should have estimated, with upper and lower bounds at their discretion, how many new problem gamblers would develop inside the entire 50 mile radius; they should not have carved out a sector of the three “counties of interest.

The report emphasizes that efforts to address problem gambling, such as age-limited advertising, signage, employee education and self-exclusion are an important component of preventing problem gambling. An executive of Caesars Entertainment  presented corporate policy on these topics at the April 9th forum on Problem Gambling. I commented in an unsolicited letter to the Gaming Commission that problem gambling was not defined in that presentation; in fact, it was deliberately blurred to identify a problem gambler as a gambler who causes problems on the gambling floor.  None of the measures described by the speaker can be effective as long as casinos depend for half their revenues on the losses of problem gamblers who are still “visitors.”

This report is arrogant and deliberately deceptive.  Sophistic is the right word.  It is sad indeed that decisions may be made based upon it.

The opinions in this piece are those of the writer, Stephen Q; Shafer MD MA MPH,  and do not necessarily reflect those of any or all other members of Coalition Against Gambling in New York.  Permission is granted to reproduce in whole or part as long as the permalink is cited.