”The bankruptcy system is supposed to work for everyone, but in many cases it works only for the powerful” – House Judiciary Committee Chairman Jerrold Nadler, July 28th, 2021.
Since at least Frank (1931), researchers have recognized that judicial outcomes are subject to the biases of the ruling judge. To alleviate concerns of fairness, courts in both the U.S. and abroad claim to assign judges to individual court cases randomly as shown by Shayo and Zussman (2011) and Abrams Bertrand, and Mullainathan (2012). From a policy perspective this randomization is valuable as it promotes public confidence in the judicial process by limiting forum shopping and the individual influence of any single judge.
This randomization is also important from the perspective of empirical social scientists, who take advantage of the random assignment process as a kind of “natural experiment laboratory.” Scientists run experiments with treated and control subjects to isolate the effect of a single variable (like providing a vaccine to a patient). But social scientists do not often get to run these same types of experiments to isolate the effects of social policies. Instead, we must find “natural experiments” that allow us to isolate the effect of a specific policy intervention. If judges are both (i) randomly assigned and (ii) disagree over decisions, then we can study the causal effects of court outcomes just like a lab scientist. This is a common strategy in modern social science, especially economics. For instance, from our own analysis, we count over 20 papers published in the top economics journals (American Economic Review, Journal of Political Economy, and Quarterly Journal of Economics) since 2015 that exploit the random assignment of cases to judges.
For instance, bankruptcy court outcomes are analyzed by finance economists to understand the effects of bankruptcy on people and businesses. Bankruptcy is the legal process to discharge unsecured debt. Businesses generally file either Chapter 11 bankruptcy (which allows the firm to develop a repayment plan for their debts and stay in business) or Chapter 7 bankruptcy (where all the assets of the firm are sold off to pay the debts and the business is dissolved). When companies file for Chapter 11, sometimes the judge assigned to the case will decide the firm should actually be liquidated and converts the case to a Chapter 7.
In order for financial economists to explore this setting to understand the effects of different bankruptcy regimes, it is necessary that (i) some judges are more likely than others to convert Chapter 11 cases to Chapter 7, and (ii) judges are assigned randomly and not correlated with the characteristics of the case. Past research has already provided plenty of evidence that judges differ in how they treat bankruptcy cases: for instance, Bris et al. (2006) show that judge fixed effects account for 10% of conversion decisions. Similarly, Bernstein et al. (2019) estimates the conversion rate separately for each bankruptcy judge and finds that a one standard deviation in the conversion rate increases the likelihood of conversion by 7.5 percentage points. Researchers have also provided evidence that judges are randomly assigned. For instance, Bernstein et al. (2019) considers a range of debtor characteristics, and shows case assignments do not appear correlated with any of these characteristics. In addition, Iverson et al. (2017) contacted all U.S. Bankruptcy Courts regarding the assignment process; of the 81 courts that responded, only one court (the Eastern District of Wisconsin) reports assigning cases to judges non-randomly. Despite this evidence, legal scholars (Levitin, 2021), policy makers (Merle and Bernstein, 2019), and the public (Randles 2020) have become increasingly concerned that debtors and creditors are choosing their preferred judge in bankruptcy judge. Missing from these discussions is any large-scale empirical evidence of non-random assignment.
In a new paper, we revisit the claim of randomized judicial assignment by analyzing investments in distressed firms made prior to a bankruptcy filing. Specifically, we examine whether the investments of active creditors predict the assignment of judges to U.S. Chapter 11 corporate bankruptcy cases. Past research finds active investors routinely influence a wide range of post-bankruptcy outcomes such as emergence and the structure of repayments (Hotchkiss and Mooradian (1997), Ayotte et al. (2012)); since bankruptcy judges have significant authority over these outcomes (Chang and Schoar (2006), Bris et al. (2006), and Bernstein et al. (2019)), we argue investors have similar incentives to influence the assignment of cases. By focusing on investments made before the assignment of a bankruptcy judge, our technique is not suspect to standard critiques that predictability is merely an outcome of ex-post data mining; instead, in order for investors to systemically invest in firms that are later assigned a preferred judge, it must be possible to infer future judicial assignments.
Our analysis focuses on the investment decisions of hedge funds investing in private debt markets. Private debt investments have expanded dramatically as investments in private credit approached $600 billion globally by the end of 2016 and fund raising in private credit has grown 2.5 times the annual growth rate of private equity since 2010. Within this sector, distressed debt represents the largest investment strategy with 45% of all committed capital, and 43% of large corporate bankruptcies have one or more private debt funds acting as creditors (Ivashina et al. (2016)). As hedge funds are major investors in distressed firm debt (Aragon and Strahan (2012)), hedge funds routinely influence a wide range of bankruptcy outcomes including emergence and debt restructurings (Jiang et al. (2012) and Lim (2015)). The prevalence of these investors allows us to explore a new channel of activism in the distressed debt market not yet studied by the hedge fund or bankruptcy literature: activist influence in judicial assignment process prior to filing.
To begin our analysis, we first aggregate the judge conversion decisions for each judge over the prior three-year period, which provides us a time-varying measure of a judge’s propensity to convert a given case. We therefore evaluate whether filings involving a hedge fund creditor are consistently assigned a judge with a conversion rate different from filings without a hedge fund creditor. By focusing on the judge’s past conversions, rather than the outcome of the current case, hedge funds must be influencing the assignment process itself and not the decisions of the judge following the assignment.
To identify non-random assignment, we exploit the fact that opposing regimes (reorganization vs. liquidation) lead to different repayment outcomes among creditors: secured creditors have a well-known liquidation bias (Bergstrom et al. (2002), Ayotte and Morrison (2009), and Vig (2013)), while unsecured creditors recover more under the repayment plan in reorganization (Bris et al. (2006) and Antill (2021)). This distinction leads us to our empirical specification: we test whether unsecured hedge fund creditors are assigned a judge less likely to convert the case to a liquidation relative to a similar debtor with a secured hedge fund creditor.
To begin our analysis, we collect data on the universe of U.S. Chapter 11 bankruptcy cases during 2010-2020 from court dockets. Second, we collect information on debtor characteristics including (i) industry, (ii) size, (iii) access to public equity markets, and (iv) location. Third, for each filing, we also collect information on the bankruptcy outcomes including the (i) assigned judge, (ii) filing date and district, and (iii) conversion decision. Finally, we collect information on hedge fund debt investments in distressed firms, including debt terms, to determine whether a bankrupt firm had a hedge fund creditor at the time of filing.
Relative to other cases in the same year and court district, we estimate being assigned a judge with a 10 percentage point higher past conversion rate increases the likelihood a given case is converted to liquidation by 2.2 percentage points, equivalent to 22% of the mean conversion rate. To identify hedge fund creditors, we match cases to information on private debt agreements in the Preqin database. In total, we analyze nearly 20,000 case filings including nearly 600 cases with hedge funds acting as creditors at the time of bankruptcy filing.
In our baseline findings, we estimate that relative to a hedge fund acting as a secured creditor in the same court district and year, unsecured hedge funds are assigned a judge with a 3.3 percentage point lower mean conversion rate. As we estimate a mean judge conversion rate of 10%, we estimate a 33% reduction relative to the mean. The difference is statistically significant at the one-percent level, holds after controlling for debtor characteristics, and is robust to excluding small- and medium-size debtors from the analysis. In addition, we find that unsecured hedge fund claimants are assigned a preferable judge more commonly when the hedge fund invested shortly before the bankruptcy filing, suggesting a portion of hedge funds choose to invest explicitly to influence the filing.
In order for creditor investments to predict future judicial assignment, creditors must be able to convince the debtor to file when optimal. As equity holders and management have the same financial preferences for reorganization over liquidation as unsecured creditors (White (1989) and Eckbo et al. (2016)), we argue it is only unsecured creditors that should be able to influence the time of filing. In line with this argument we find no evidence that filings involving a secured hedge fund are assigned a different judge than otherwise similar cases. Furthermore, among the unsecured creditors, we show the effects are greatest when the hedge fund is directly or indirectly connected to the board of directors of the debtor at the time of filing, providing further support for the role of communication between debtor and creditor. Last, we confirm our results continue to hold when excluding involuntary bankruptcies that are filed by the creditor.
There are three separate concerns with our analysis. First, it is possible our results are simply the result of noise. If this is the case, we should find a judge’s future conversion rate (after controlling for the past conversion rate) is also correlated with hedge fund investments. However, we find no evidence of any correlation, suggesting hedge funds are explicitly influencing judicial assignment based on information regarding past judicial outcomes. Second, it is possible the assignment process is non-random for certain districts and this is public knowledge; our results may then be driven by this subset of districts. However, focusing on the subset of districts that explicitly state random assignment within their district (according to Iverson et al. (2017)), we continue to find hedge fund investments predict assignment. Third, cases may be assigned at the office-level rather than the district-level; in this instance, our results are no longer evidence of non-random assignment. To test this hypothesis, we include district-office-year fixed effects in our analysis and continue to find a relationship between hedge fund investments and assignment.
We next extend our analysis to an alternate bankruptcy outcome measure: the unsecured creditor recovery rate according to the confirmed plan. While we observe this measure for only a subsample of the full dataset, this measure allows us to examine variation within filings that are ultimately reorganized. As before, we estimate each judge’s unsecured creditor recovery rate for previously assigned cases and continue to find (i) the past recovery rates of a given judge predict future recovery rates, (ii) unsecured hedge funds are far more likely to be assigned a judge with previous high unsecured recovery rate, and (iii) the coefficient is similar for the subsample of districts that explicitly state random assignment.
The results above provide convincing evidence hedge funds can predict judicial assignment; we next demonstrate judicial assignment is also predictable to the econometrician. Assuming overseeing a large corporate case is time-consuming (Iverson et al. (2017)), we argue courts may be less inclined to assign multiple large cases to the same judge within a narrow window. In line with this theory, we provide new evidence that large bankruptcy filings are negatively serially correlated; being assigned a large bankruptcy in the previous week decreases the likelihood of being assigned a large bankruptcy filing this week. In contrast, small filings are not predictable based on recent judicial assignments. In addition, we find evidence that unsecured hedge fund creditors appear to exploit these patterns: while the filing dates of cases with unsecured hedge fund creditors can be partially explained by the recent case filings, similar cases of secured hedge fund creditors do not exhibit these same patterns. Overall, the results suggest econometricians can predict judicial assignment, just like hedge funds.
So, given these findings, how can we improve the bankruptcy process? We believe there are two potential policies that can alleviate these issues. The first, and simplest, is for policy makers to develop a truly randomized process. However, the obvious downside of this proposal is that judges may occasionally be assigned multiple large filings in a row, impacting judicial outcomes (Iverson (2017) and Muller (2022)). Alternatively, and following the suggestions of Iverson et al. (2020), policy makers can instead increase the number of bankruptcy judges. In this scenario, creditors will lose their predictability powers even if assignment is not fully randomized. Policy makers intent on a fairer judicial system should consider both proposals.
Niklas Hüther is an Assistant Professor of Finance at Indiana University
Kristoph Kleiner is an Assistant Professor of Finance at Indiana University
This post is adapted from their paper, “Are Judges Randomly Assigned to Chapter 11 Bankruptcies? Not According to Hedge Funds” available on SSRN.
The views expressed in this post are those of the authors and do not represent the views of the Global Financial Markets Center or Duke Law.