Available at SSRN: https://ssrn.com/abstract=7202339
Abstract:
Critics have long charged that a casino drives the households nearest it into bankruptcy, and the claim now shapes local policy. Casinos have become a place-based development tool, sited deliberately in distressed communities: Virginia, for instance, recently authorized commercial casinos only in cities it designated as economically distressed. The evidence on whether casinos lead to more bankruptcies is dated and coarse, relying on data almost two decades old and measured at a county or state level, far too aggregated to isolate a casino's local effect (Anisfeld and Rosenthal-Kay 2025; Scavette 2023). We revisit the casino-bankruptcy link with spatial data fine enough to observe highly localized effects. We assemble a quarterly panel of consumer bankruptcy filings at the census-tract level for 2008:Q1 through 2026:Q1 and match it to the location and opening date of 1,130 U.S. casinos. Exposure is measured by the driving time from each tract's population centroid to the nearest casino. Using the partially pooled synthetic control estimator of Ben-Michael et al. (2022), we build each casino tract a counterfactual from distant but comparable tracts and trace the effect outward from the host tract. No specification we estimate shows a statistically significant increase in bankruptcy filings after a casino opens. At the host tract the estimates are positive but never significant; in the surrounding area they are predominantly negative. Where much earlier work reports a positive effect, the finer spatial scale that should make a casino's local influence clearest shows no increase at all.
Commentary: A Sophisticated Study, But a Much Less Certain Conclusion Than the Title Suggests
Gary Wagner and Douglas Walker's Casino Openings Do Not Bankrupt Their Neighbors deserves attention. It uses much finer geographic data and more sophisticated statistical techniques than most of the older research examining casinos and bankruptcy.
But those virtues should not cause policymakers—or consumer bankruptcy attorneys—to accept the title as something the study has actually proven.
What the paper establishes is considerably narrower:
Using the authors' statistical model, they did not find sufficiently reliable evidence of an increase in aggregate bankruptcy filings within the geographic areas and time periods they examined following certain brick-and-mortar casino openings.
That is interesting.
It is not the same as proving that casinos "do not bankrupt their neighbors."
In fact, several features of the study make me quite skeptical that bankruptcy lawyers should draw such a broad conclusion.
The Closest Communities Actually Point Toward More Bankruptcies
Start with a result that receives less emphasis than it deserves.
The place where one would presumably expect casino-related financial harm to appear most clearly is the census tract containing the casino itself.
In those host tracts, the estimated effect goes in the direction of more bankruptcies, not fewer.
The authors cannot say with sufficient statistical confidence that those increases are real rather than ordinary variation, so the results are technically inconclusive. As the authors expand the geographic area to include increasingly distant surrounding communities, the estimated effect turns negative.
For a fuller plain-English explanation of terms such as "statistical significance," "treatment groups," "donor pools," "covariates," "transformations," and "specifications," I have included a separate statistical glossary below.
But the practical point is simple.
The study did not prove that bankruptcies did not increase in the communities containing casinos. It lacked sufficiently precise evidence to say whether they did or did not.
That is quite different from the certainty conveyed by the title.
And although the paper begins with data concerning 1,130 casinos, once its exclusions are applied the narrowest—and arguably most important—analysis contains only 54 newly treated casino host tracts.
An uncertain result from 54 host areas should probably remain an uncertain result.
"And Who Is My Neighbor?"
There is also a more fundamental geographic question.
As was asked roughly 2,000 years ago in Luke 10:29:
"And who is my neighbor?"
That turns out to be a surprisingly pertinent question for this paper.
Wagner and Walker define the casino's potentially affected neighborhood in increasingly larger circles, ending with census tracts whose population centers are within a 60-minute drive of a newly opened casino. Comparison communities used to estimate what supposedly would have happened without the casino begin farther away, with some comparison groups starting at 75, 90 or 120 minutes.
But why should sixty minutes define the outer boundary of the casino's financially affected "neighbors"?
Casinos themselves do not necessarily define their customers so narrowly.
Casino feasibility and market studies routinely analyze customer bases well beyond an hour's drive. An Illinois Gaming Board feasibility study, for example, examined casino markets at 15, 30 and 60 minutes, while recognizing a regional customer base beyond the immediately surrounding neighborhood. A Kenosha, Wisconsin casino market evaluation described the casino's primary catchment area as adults for whom that casino would be the closest casino within a two-hour drive, with another assessment extending a secondary market to a 100-mile radius.
Perhaps more strikingly, an official Richmond casino market assessment modeled regional casino demand from households living within approximately a 2.5-hour drive of downtown Richmond.
Those are not academic estimates of gambling harm.
They are estimates of where casinos expect customers to come from.
That difference is important.
The Casino's "Neighbor" May Live Two Hours Away
If a casino advertises regionally and expects customers to travel 90 minutes, two hours, or even farther to gamble, why should a bankruptcy study presume that exposure substantially ends after sixty minutes?
A frequent gambler from ninety minutes away may be far more financially exposed to a casino than a person who lives five minutes away but never sets foot inside it.
Yet geography rather than actual gambling behavior determines exposure in this study.
That produces a potentially troubling problem with the comparison group.
Some census tracts the statistical model treats as sufficiently distant to help represent a world without meaningful casino exposure may contain substantial numbers of people whom the casino itself regards as potential customers.
In other words, some of the supposed "unexposed" comparison households may not be unexposed at all.
If both the casino area and the supposedly unaffected comparison area contain casino patrons who develop gambling debt, the difference in bankruptcy filings between the two groups shrinks.
The model could then conclude that the casino had little effect—not because casino gambling caused no financial distress, but because gambling exposure crossed the geographic boundary used to distinguish the affected group from the comparison group.
That possibility deserves much more attention.
Advertising Makes the Geographic Problem Even More Pronounced
The geographic assumption also seems divorced from how casinos actually seek business.
Casinos are not neighborhood convenience stores relying solely on people who happen to live nearby.
They operate loyalty programs, advertise regionally, offer entertainment and restaurants, promote hotel packages, market events and otherwise seek to induce customers to travel. Modern casino market studies expressly distinguish primary, secondary and tertiary markets extending well beyond the immediate community. A recent gaming feasibility study, for example, characterized 0–30 minutes as the primary market, 30–60 minutes as a secondary day-trip market, and 60–120 minutes as a broader regional market.
The economic proposition is obvious enough that casinos themselves build revenue projections around it:
People travel to gamble.
If the casino is willing to spend advertising dollars to persuade someone two hours away that he is close enough to be a customer, it seems peculiar for researchers to presume that he is too distant to be a relevant "neighbor" when looking for the resulting financial consequences.
That is particularly important for destination properties that combine gambling with concerts, restaurants, hotels and other amenities designed specifically to enlarge their geographic draw.
Geographic Proximity Is Not the Same as Gambling Exposure
This reveals a larger conceptual weakness.
The paper is fundamentally studying proximity to casinos, not exposure to gambling.
Those are not the same thing.
A household across the street from a casino whose members never gamble is treated as intensely exposed.
A household ninety minutes away where someone drives to that casino twice every week may be treated as comparatively unexposed.
That makes geography a decidedly imperfect substitute for the thing we actually care about.
The question relevant to bankruptcy policy is not really:
How many miles from the casino does the debtor live?
It is:
Did expanded gambling access cause the debtor to gamble more, accumulate debt, dissipate assets and eventually suffer financial failure?
The data in this study cannot answer that question.
Bankruptcy Is Usually the End of the Story, Not the Beginning
There is another mismatch between the study design and the experience of practicing bankruptcy lawyers: time.
A household rarely experiences the event that ultimately causes its insolvency and immediately files bankruptcy.
Financial failure accumulates.
A gambler can first use available cash.
Then credit cards.
Then cash advances.
Then savings.
Then retirement funds.
Then home equity.
Then loans from family.
Minimum payments may continue for months or years.
Collections eventually start.
Perhaps there is a lawsuit.
Perhaps a judgment.
Perhaps a garnishment.
Only then does the household finally call a bankruptcy lawyer.
By that point, determining when the financial collapse actually began can require going back years.
Life in the Sweatbox Makes the Timing Problem Difficult to Dismiss
Pamela Foohey, Robert M. Lawless, Katherine Porter and Deborah Thorne documented exactly this phenomenon in Life in the Sweatbox, 94 Notre Dame Law Review 219 (2018).
Using Consumer Bankruptcy Project data, they examined the lengthy period of serious financial distress that precedes bankruptcy—the "financial sweatbox." Their work describes "long strugglers" who remain in financial distress for more than two years before filing and documents the substantial personal and economic consequences that accompany prolonged efforts to avoid bankruptcy.
Available at: Life in the Sweatbox
That should matter enormously in evaluating this study.
Wagner and Walker use up to twenty post-opening quarters—approximately five years—to measure the effects of a casino opening.
Five years sounds substantial until one considers how consumer financial distress actually unfolds.
The Casino Opens in Year One. The Bankruptcy Could Come in Year Seven.
Consider a perfectly plausible chronology.
A casino opens.
During the first year, someone begins gambling regularly.
Over the next two years, $30,000 accumulates on credit cards.
The household then borrows from retirement accounts and refinances a vehicle.
For another two years, it manages minimum payments while moving balances around.
Eventually payments are missed.
Collections begin.
A creditor sues.
The family delays bankruptcy for another year trying to fix things.
Then—six or seven years after the casino opened—the debtor finally files Chapter 7 or Chapter 13.
A five-year statistical window risks classifying that eventual bankruptcy as outside the period in which the casino opening had an effect.
For consumer bankruptcy lawyers, that should be an obvious concern.
A bankruptcy petition is often a lagging indicator of financial distress.
It tells us when the debtor finally surrendered to insolvency—not necessarily when insolvency began.
This Timing Problem Also Cuts Against the Paper's Geographic Precision
The timing and geographic issues can compound one another.
Suppose a person living ninety minutes from a casino responds to regional casino advertising, becomes a regular patron, gradually accumulates debt for three years, struggles in the financial sweatbox for another three years and finally files bankruptcy in year six.
That person can fall outside the study's preferred concept of both:
where the casino's neighbors live; and
when casino-related bankruptcy should appear.
Yet from a bankruptcy lawyer's perspective, that may be precisely the person we would want the research to identify.
These Are Not Really Observed Census-Tract Bankruptcy Filings Either
There is another limitation behind the paper's claimed geographic precision.
The Federal Judicial Center data do not actually report the debtor's census tract.
They report ZIP codes.
The authors mathematically distribute bankruptcy filings among census tracts using HUD data estimating the proportion of residential addresses from each ZIP code located in each tract.
That may be a reasonable statistical method.
But it means the study does not actually know that Debtor Jones lived in Census Tract 123 and filed bankruptcy in a particular quarter.
The tract-level filings are partly estimated.
That is especially important because the central criticism of previous research is that county- and state-level geography was too imprecise.
This paper unquestionably improves upon that geography.
But "census-tract bankruptcy filings" sounds rather more exact than the underlying data actually are.
And the Population Denominator Remains Stuck in 2010
The bankruptcy rate also uses population from the 2010 Census, despite the study extending through 2026.
That seems particularly questionable in a study of casino openings.
Casinos are intentionally promoted as economic-development projects. If successful, they may alter employment, development, migration and population.
Yet a bankruptcy filing rate in 2024 or 2025 is being calculated using the tract's population in 2010.
For a paper built around fine geographic precision, using a population denominator potentially fifteen years out of date is not a trivial issue.
A Casino Could Increase Gambling Bankruptcies While Reducing Total Bankruptcies
There is also a basic aggregation problem.
Suppose a casino creates 500 jobs and thereby prevents thirty households from filing bankruptcy.
At the same time, twenty frequent gamblers suffer catastrophic losses and eventually file bankruptcy.
Total local bankruptcies decline by ten.
The statistical result would show fewer bankruptcies after the casino opened.
But it would be false to say that the casino did not bankrupt anyone.
Indeed, Wagner and Walker themselves suggest that casino-related increases in employment and income may offset losses suffered by people who gamble.
That is a fascinating possibility.
But it changes the interpretation of their findings considerably.
Perhaps the paper has found evidence that:
The aggregate economic benefits of casino development may offset gambling-related financial losses sufficiently that total community bankruptcy filings do not increase.
That would be an important economic-development conclusion.
It does not establish that gambling-related bankruptcies do not occur.
Bankruptcy Filings Are a Very Blunt Measure of Financial Harm
Nor is bankruptcy the only possible endpoint.
A gambler may exhaust retirement savings, lose a car, default on a mortgage, suffer collections, live with judgments, obtain payday loans, borrow from family, delay healthcare or simply remain insolvent without filing bankruptcy.
Wagner and Walker themselves acknowledge that BAPCPA increased the cost of bankruptcy and reduced nonbusiness filings. They suggest that in the post-BAPCPA era covered by their study, a particular financial shock may be less likely to push a marginal household into a formal bankruptcy filing.
That is not a minor concession.
If fewer financially distressed households respond to hardship by filing bankruptcy, then bankruptcy filings become a less sensitive measure of financial harm.
Finding no bankruptcy filing does not mean finding no financial distress.
Where Are the Actual Debtors?
The authors supplement their filing data with only four neighborhood characteristics: median household income, poverty, lack of health insurance and the percentage of married-couple families.
Those tell us about census tracts.
They tell us almost nothing about the bankruptcy filers themselves.
That makes the absence of debtor-level information from the Consumer Bankruptcy Project particularly noticeable.
The CBP has spent decades gathering demographic, household and financial information about actual consumer bankruptcy filers. It cannot simply be substituted for the massive FJC dataset, and its sample size and confidentiality restrictions may make precise matching to individual casino openings impossible.
But those limitations reveal what Wagner and Walker cannot tell us.
We do not know whether debtors exposed to casinos carried more credit-card debt, depleted savings, borrowed from retirement accounts, suffered gambling-related losses, spent longer struggling before filing or experienced other financial deterioration long before a bankruptcy petition appeared.
Those are questions worth studying.
The Sports-Gambling Evidence Should Increase Our Skepticism
The contrast with The Financial Consequences of Legalized Sports Gambling is important.
That research concludes that easier access to sports gambling—particularly online and mobile gambling—harms consumer financial health by increasing excessive debt.
Wagner and Walker themselves recognize evidence of increased bankruptcy following online betting and deterioration in savings, credit-card balances and overdrafts among financially constrained households.
They run additional tests attempting to remove online gambling as a competing explanation and continue to find no increase associated with physical casino openings.
That may indicate that brick-and-mortar casinos are genuinely less financially dangerous than online gambling.
Or it may demonstrate that actual household financial data are better at detecting gambling harm than geographic proximity and aggregate bankruptcy filings.
Either interpretation is more nuanced than declaring that casinos do not bankrupt their neighbors.
Then There Is Casinonomics
The paper states:
**Funder Statement—**This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
That statement may be entirely accurate.
But it answers only whether this particular study received specific funding.
Douglas Walker is associated with Casinonomics Consulting, a Charleston, South Carolina consulting firm focused on the socioeconomic consequences of gambling.
Its publicly identified clients include the American Gaming Association, Aristocrat Technologies, the Canadian Gaming Association, the iDevelopment & Economic Association, several gaming commissions and regulatory agencies, the Florida Legislature and the Pauma Band of Mission Indians.
That does not establish bias or invalidate the study.
Researchers should not be disqualified merely because they have performed professional consulting.
But research of this sort contains innumerable judgment calls—from the choice of outcome and time period, to geographic boundaries and comparison communities, to the interpretation of uncertain results and ultimately the title.
Bias need not be deliberate to matter.
Given those professional relationships, I would have preferred a direct conflict disclosure explaining them, rather than only the technically narrower statement that this particular research received no specific outside grant.
Greater disclosure would make the work more credible, not less.
So, Again: Who Is My Neighbor?
The question in Luke is not really a statistical one.
But it exposes a statistical assumption in this paper that deserves scrutiny.
When casinos themselves define regional customer markets extending one, two or even two-and-a-half hours away, limiting the search for bankruptcy consequences primarily to communities within sixty minutes seems arbitrary.
When those more distant communities can also supply the statistical comparison group, the problem becomes more serious.
And when actual bankruptcy filing may occur years after the gambling-related financial problems begin, both the geographic and temporal definitions of "neighbor" may be substantially too narrow.
A casino's economic reach cannot sensibly be broad when projecting customers, gaming revenue and economic benefits—but suddenly narrow when measuring possible financial harm.
What the Study Actually Shows
Wagner and Walker have produced a sophisticated and worthwhile study.
But the title substantially outruns the evidence.
What they have reasonably shown is this:
For the particular casino openings, geographic definitions, comparison communities, statistical assumptions and post-opening periods they studied, they could not reliably detect an increase in aggregate bankruptcy filings attributable to opening a brick-and-mortar casino.
They have not shown that casinos never cause bankruptcy.
They have not shown that gamblers do not accumulate harmful amounts of debt.
They have not shown that casino-related financial distress does not emerge years later.
They have not shown that people living more than sixty minutes away are unaffected.
And they have not shown that households suffering substantial gambling-related financial losses necessarily ever file bankruptcy at all.
Perhaps the appropriate response to Casino Openings Do Not Bankrupt Their Neighbors therefore remains the question asked long before econometrics existed:
"And who is my neighbor?"
Until that geographic question—and the equally important question of when financial distress finally becomes bankruptcy—is answered more convincingly, the findings here warrant interest.
They warrant considerably less reassurance.
To read a copy of the transcript, please see:
Blog comments