Our new study offers some of the first causal evidence on a question that has motivated nine proposed congressional bills over several decades: should firms be allowed to file for bankruptcy outside their home state? Drawing on a newly assembled dataset of US corporate bankruptcies linked to confidential US Census microdata, we find that forum shopping to Delaware is associated with fewer liquidations, shorter cases, improved creditor recovery, and a meaningful increase in post-bankruptcy employment. The findings offer a useful empirical reference point for ongoing reform discussions, including Senator Warren’s 2023 Bankruptcy Venue Reform Act.
The policy debate and the data challenge
Bankruptcy forum shopping is common: 22% of US Chapter 11 cases are filed outside the debtor’s headquarters state, and roughly half of those land in Delaware. Critics, most prominently Lynn LoPucki, have argued that managers, lawyers and secured lenders may favour Delaware because its judges deliver outcomes that benefit insiders at the expense of workers and other stakeholders. Proponents counter that Delaware’s specialised judges and deep precedent yield faster, more predictable, and more efficient proceedings.
Until recently, the empirical literature has had difficulty adjudicating this debate. Existing studies have generally relied on small, hand-collected samples of large public firms, leaving most of the corporate bankruptcy universe unobserved. We address this gap by assembling a panel of 155,059 corporate Chapter 11 cases, roughly 82% of those filed since 2005, by merging the Federal Judicial Center’s Integrated Database with case-level records from PACER, and then linking the merged data to the US Census Bureau’s Longitudinal Business Database. The result is an establishment-level panel that tracks employment, closure, and payroll for both public and private bankrupt firms, before and after filing.
A novel research design
We address the inherent selection problem, namely that firms choosing Delaware differ systematically from those that do not, with an instrumental-variables strategy. Among firms headquartered in Delaware-adjacent states (Maryland, New Jersey and Pennsylvania), physical proximity to the Wilmington bankruptcy court strongly predicts the decision to file there. A one-standard-deviation reduction in distance (62 miles) is associated with a 5–10 percentage-point increase in the Delaware filing rate. Within-state variation in distance is plausibly unrelated to firm quality, particularly once we absorb state-by-year fixed effects, county-level economic controls, industry effects, and binned controls for assets, liabilities and number of creditors drawn from the bankruptcy petition.
Two placebo tests support the exclusion restriction. First, distance to Delaware does not predict pre-bankruptcy employment trends among the bankrupt firms themselves. Second, in a placebo sample of never-bankrupt establishments matched by ZIP code, distance to Delaware is uncorrelated with employment dynamics across all horizons. Neither result would be expected if proximity were proxying for unobserved growth prospects.
What the estimates suggest
The 2SLS estimates are economically meaningful. Filing in Delaware appears to reduce the rate of conversion from Chapter 11 to Chapter 7 liquidation by roughly 31 percentage points. Establishment closure three years after filing falls by approximately 30 percentage points. Case duration shortens substantially, and the probability that unsecured creditors receive a strictly positive recovery rises. Notably for the policy debate, post-bankruptcy employment increases by roughly 6.6 employees per establishment relative to pre-filing average employment of 26.65, about a 25% increase.
The OLS coefficients tell a complementary story. They are close to zero, suggesting negative selection: firms appear more likely to forum shop when their prospects are weakest. Simple correlations may therefore understate the underlying treatment effect.
The Delaware-judge mechanism
Why might Delaware deliver better outcomes? We construct a ‘Judge DE Gap‘ measure capturing how closely each judge’s leave-one-out conversion rate, recovery rate, case duration and experience resemble the Delaware average. Exploiting the well-established random assignment of judges within courts, we find that Delaware-like judges tend to produce faster cases and fewer liquidations across the country, not just in Delaware. Filing in Delaware mechanically increases the probability of drawing such a judge. When judge characteristics are added as controls, the 2SLS effect of forum shopping shrinks by half and loses statistical significance. Analysis of the 2005 Bankruptcy Abuse Prevention and Consumer Protection Act reform reinforces the mechanism: as home-court caseloads fall and home-court efficiency improves, forum-shopping rates decline.
Implications for venue reform
Senator Warren’s concern that ‘affected communities—like workers, creditors, and consumers—lose‘ when firms forum shop is, on this evidence, in some tension with the data. Workers in the sample appear to benefit through higher post-bankruptcy employment; unsecured creditors through improved recovery; and the bankruptcy system as a whole resolves cases more quickly. One caveat is that our findings are specific to Delaware and silent on forum shopping to other courts, where local precedents on issues such as the priority of severance pay differ materially.
For policymakers, one takeaway is that restrictions targeting Delaware specifically could push marginal firms into less efficient home courts, with potentially adverse consequences for jobs and creditor recoveries. An alternative response would be to invest in judicial expertise and reduce caseload congestion elsewhere. Addressing the underlying quality gap that appears to drive forum shopping in the first place.
The authors’ paper can be found here.
Samuel Antill is an Assistant Professor of Business Administration, Harvard Business School.
Aymeric Bellon is an Assistant Professor of Finance, University of North Carolina (UNC) at Chapel Hill.
Share:
With the support of
