Los conflictos olvidados.

makes it attractive to isolate the causal effect of judicial efficiency, but less attractive as an indication of what constitutes congestion in US bankruptcy courts. After all, the caseload drop following BAPCPA was a reflection of a substantial legal change, and specifying the excessive or undesirable component of the caseload per judge is clearly challenging. In this section, I propose two simple approaches and calculate the costs of making courts less busy.

First, I will use the Bankruptcy Judgeship Act of 2017 as a template for a min- imum desirable reduction in caseload. The Act passed the Senate on September 5, 2017 and added four new permanent judgeships for the districts Delaware (2), Flor- ida Middle (1), and Michigan East (1), following the recommendations of the Judi- cial Conference.36 _{Data from the Federal Judicial Caseload Statistics for Q1/2018 and} Q1/2017 allow me to calculate how the new judgeship affected the annual change in the weighted caseload per judge. For Delaware, the judge workload changed from around 626 to 476 hours (a drop of 150); for Florida Middle from 906 to 766 (a drop of 140); and for Michigan East from 948 to 777 (a drop of 171). The average drop was 154 hours per judge per year.

Second, I will estimate the effect of reducing the caseload per judge of districts above the 90th percentile in Q1/2017 by hiring an additional judge in each of these districts.37 _{This would reduce the annual workload per judge in these courts by an} average of 310 hours, ranging from a minimum of 102 caseload hours in the northern district of Illinois to a maximum of 731 hours in the northern district of Mississippi. Note that these courts do not overlap with the districts receiving new judgeships as part of the 2017 bill: the average weighted caseload per judge of Delaware, Florida Middle, and Michigan East in Q1/2018 was 827 hours per year, which lies between the 75th and 90th percentile of the caseload distribution.

In both scenarios, the majority of courts would be unaffected by hiring additional judges. I thus begin by using the firm-level calculation as in equation 4.6 to calcu- late the savings in interest rate burden, and then scale up these values using the frac-

36_{See https://www.congress.gov/bill/115th-congress/senate-bill/1107/text}

for the full text of the bill and http://www.uscourts.gov/news/2017/04/07/ judiciary-seeks-bankruptcy-judgeships-warns-crisis for the recommendations of the Judicial Conference. The Act also made permanent many previously temporary judgeships that were extended on a regular basis. Since these judges were already in place, this does not have an immediate effect on the court workload.

tion of total debt in Compustat in my sample to total non-financial corporate debt in Q4/2016.38 _{The in-sample firms accounted for approximately 69% of total debt in 2016} ($5_$7,_,303₆₈₀ billion≈0.69).

For the Bankruptcy Judgeship Act of 2017, this procedure yields estimated sav- ings in the interest burden of approximately $689 million for the entire stock of non- financial corporate debt ($476 million in-sample). If one were to hire an additional judge in each of districts with a caseload above the 90th percentile, the savings would amount to $986 million ($681 million in-sample). How do these numbers compare to those one would arrive at using the aggregate formula 4.3, ignoring the firm hetero- geneity? My estimates suggest a somewhat lower drop of $414 million for the Bank- ruptcy Judgeship Act and $592 million in the high-congestion scenario. Again, these differences arise because the aggregate debt service formula relies on assumptions about the maturity of outstanding debt on one hand, but also allows for different fin- ancing terms for loans and bonds. Taking the simple average of the four estimates on the total non-financial corporate debt suggest that the costs of overburdened courts are indeed “enormous”: at least $670 million per year.

How much would it cost to resolve excessive court inefficiency? For the Bank- ruptcy Judgeship Act of 2017, the Congressional Budget Office estimates annual salar- ies and benefits for bankruptcy judges of about $232,000.39 _{The Congressional Budget} Office also provides an estimate for judicial administrative costs for personnel, se- curity, and court operations of about $700,000 per judge per year. Hiring an addi- tional judge would thus cost approximately $932,000; I will round this estimate up to $1,000,000 for simplicity. Four judges in the Bankruptcy Judgeship Act scenario thus would cost $4,000,000, and the 8 new judgeships required by the 90th percentile scen- ario $8,000,000. Clearly, these costs are miniscule compared to the estimated benefits: from a fiscal policy perspective, they imply a “judge multiplier” of above 100 in almost all estimations. The average multiplier for the four total debt estimates is 118.

Taken together, I conclude that the social costs of court backlog are large, and that addressing these costs could be a potentially profitable avenue for government ex- penditures. Again, these findings rely on a number of simplifying assumptions. Des- pite this caveat, the range of values I present is likely an extreme lower bound for

38_{More precisely, I use the sum of debt in current liabilities and total long-term debt of firms in the}

Dealscan-Compustat sample, and average these values over 2015 through 2017 (as available).

39_{Bankruptcy judges are entitled to compensation equal to 92% of that of a district judge, which}

puts their listed annual salary at approximately $191,000. See https://www.law.cornell.edu/ uscode/text/28/153 for the background covering bankruptcy judge compensation and http: //www.uscourts.gov/judges-judgeships/judicial-compensationfor the time series of ju- dicial pay. District judges in the United States were entitled to $208,100 in compensation in 2018.

multiple reasons. First, they are solely based on the effect of judicial caseload on pub- licly listed firms with low bankruptcy risk. Second, they are based on some of the smallest point estimates I find. Third, they ignore knock-on effects of financing terms on foregone firm investments and employment, as well as the effect of judicial effi- ciency on the interest burden of households. Fourth, I do not consider costs arising from an inefficient resolution of bankruptcy cases due to congested courts (see e.g. Iverson,2016). Fifth, up to this point, I have not considered the effect of caseload on loan maturities. All of these factors bias my estimates downwards.