4.6 The Social Costs of Court Backlog
4.6.2 Estimating Changes to the Interest Burden of US Corporations
To translate these numbers into an aggregate cost, I draw on the approach ofDrehmann et al.(2015) to calculate the US-wide interest payments of non-financial corporations. They calculate the macroeconomic debt service burden as follows, where I omit sub- scripts for brevity:
Interest Burden= Interest Rate
(1−(1 +Interest Rate)−M aturity)×Debt, (4.3)
where Debt refers to the total stock of non-financial corporate debt; Interest Rate to the average interest rate on the existing stock of debt; and M aturity to its average remaining maturity. For the baseline estimates, I ignore the effect of an exogenous drop in caseload on maturities. This is highly conservative, because incorporating the increase in maturities substantially increases the macroeconomic savings associated with more efficient bankruptcy courts. In the online appendix, I provide back-of-the- envelope calculations with a change in maturities that suggest the findings presented here are an extreme lower bound of the social costs of overburdened judicial systems. Because I am interested in how much the cost of debt service change with judicial
efficiency, and ignore the change to maturities, this yields the following equation: ∆Interest Burden= ∆Interest Rate×Debt. (4.4) Note thatInterest Rateconsists of an interest ratespreadand an underlying reference rate. Because I am only interested in the change to spreads, I directly translate it into changes in interest rates. Keeping reference rates constant, a decrease in spreads and increase in maturities can be directly translated into a reduction in aggregate corporate debt servicing costs.
Because the BAPCPA setting allows me to calculate the elasticity of a change in spreads to a change in caseload, I could directly calculate its aggregate effects by re- placing∆Interest Ratetin equation4.4with the estimated elasticity:
∆Interest Burden= −90.144/10,000
1092.878 ×Caseload Drop×Debt, (4.5) where−90.144 /10,000 is the estimated coefficient for the change in spreads (in basis points) and 1092.878 the estimated coefficient for the caseload drop. Caseload Drop
specifies the size of the drop in the weighted caseload per judge one is interested in. However, the aggregate effect may depend crucially on how firms differ in their ex- posure to such changes and how much debt they have. For example, firms that would be highly exposed to an increase in judicial efficiency may be large and highly levered, or small and hold no debt at all. I thus allow for firm-level variation in the caseload drop and outstanding debt, which yields:
∆Interest Burdeni =
X
i
−
90.144/10,000
1092.878 ×Caseload Dropi×Debti
. (4.6)
I will refer to equation 4.6 as “direct estimation”, compared to the formula-based es- timation in equation4.3. The advantage of calculating effects directly is that I can use firm-level variation in outstanding debt and exposure without having to assume ho- mogeneity in a firm’s debt structure (bonds versus loans) or average financing terms. A disadvantage is that I cannot easily gauge the effect of a change in maturities, be- cause calculating the outstanding maturity of a firm’s outstanding debt is difficult (see also Drehmann et al., 2015). Another disadvantage is that the large, publicly listed borrowers in my estimation sample are not representative; this introduces a strong downward bias, because smaller firms are more likely to file for bankruptcy, and thus more affected by changes to the efficiency of bankruptcy courts (see the results section 4.5.2). The firms in my sample also hold only a fraction of total non-financial corporate
debt in the US, and may pay different spreads on bonds and loans.
I address these challenges two-fold. First, I will scale up the outcomes of equation 4.6using the share of debt of firms in my sample as a fraction of total debt to arrive at more representative values. Note that to maximize the external validity of my calcu- lation, I use data on all firms in the matched Dealscan-Compustat data set for which I can identify their bankruptcy district.35 Second, I compare the firm-level estimates to changing the inputs of the aggregate debt service ratio in equation 4.3, which I refer to as “formula-based”. This has the advantage that it allows for different interest rates and maturities for loans and bonds (which are assumed to be equal across firms), and also makes calculating changes to maturities manageable. According to data from the US Financial Accounts, non-financial corporates had a total of $5,070.1 billion in debt at the end of 2004 (around 58% in bonds), and $7,680 by the end of 2016 (around 66% in bonds).
4.6.3
BAPCPA and the Interest Burden of Corporate Borrowers
Equipped with point estimates for the effect of a drop in caseload on the spreads of corporate borrowers, and the methodology outlined above, we can make a back-of- the-envelope calculation for the macroeconomic savings of BAPCPA. To quantify the effect, I use the total drop in the caseload per judge between 2006-2007 and 2004-2005, equal to around 493 hours (see section4.2). Because firms were more exposed to the caseload drop than others, I weight the firm estimates by the district’s exposure, i.e. the non-business share in the weighted caseload of the bankruptcy district. This yields the following equation:
∆Interest Burden=X i − 90.144/10,000 1092.878 ×493×Exposurei ×Debti , (4.7)
where Debti is the sum of long-term debt and debt in current liabilities from Com-
pustat, averaged over 2004 and 2005. As outlined above, I assume that the district differences in Exposure map linearly into improvements in caseload (1092.878 from table4.2) and spreads (−90.144from column (7) in table4.14). The assumption is thus that lower caseload allows firms to roll over pre-existing debt at better financing con- ditions – which is consistent with my findings on borrowers with exogenous financing needs in section4.5.2.
35Note that the results presented here yield highly similar results if I instead use only the firms in
Table 4.7: THESOCIAL COSTS OF CONGESTEDBANKRUPTCYCOURTS
This table presents the inputs required to calculate the social costs of excessive bank-
ruptcy court caseload. The Implied elasticity is calculated as the ratio of Exposure on
spreads, divided by Exposure on the caseload drop. The Estimated drop in spreads is cal-
culated as the Drop in caseload per judge (average) times the implied elasticity. The Judge
multiplier is the ratio of savings in interest burden to required judgeships. Net gains
are the difference between the savings in interest butden and required judgeships.
Direct Estimation Formula-Based
In-sample Total Total
debt corporate debt corporate debt
(1) (2) (3)
Panel A: Point Estimates
Exposureon spreads -90
Exposureon caseload drop 1,092.878
Implied elasticity (bps/caseload hour) -0.082
Panel B: Estimated Effect of BAPCPA
Total debt (2004, in $ billion) 2,686 5,070.1 5,070.1
Drop in caseload per judge (average) 493 493 493
Estimated drop in spreads -40.599 -40.599 -40.599
Savings in interest burden (in $ billion) 8.109 15.308 9.536
Panel C: Estimated Effect of Bankruptcy Judgeship Act of 2017
Total debt (latest, in $ billion) 5,303 7,680 7,680
Drop in caseload per judge (average) 154 154 154
Savings in interest burden (in $ billion) 0.476 0.689 0.414
Required judgeships (in $ billion) 0.004 0.004 0.004
Judge multiplier 119 172 104
Net gains 0.472 0.685 0.41
Panel D: Estimated Effect of New Judges for Highly Congested Courts
Total debt (latest, in $ billion) 5,303 7,680 7,680
Drop in caseload per judge (average) 310 310 310
Savings in interest burden (in $ billion) 0.681 0.986 0.592
Required judgeships (in $ billion) 0.008 0.008 0.008
Judge multiplier 85 123 74
I start with the in-sample calculation for the publicly listed firms in the matched Dealscan-Compustat sample in column (1) of table4.7. The borrowers in the sample had a total of $2,686 billion in outstanding debt prior to BAPCPA in 2004. This means they account for a substantial fraction of 53% of the total outstanding debt of non- financial corporates of around $5,070 in that year. Equation4.6implies that the drop in caseload per judge following BAPCPA saved the sample firms $8.109 billion in interest payments per year - a substantial magnitude, even if one abstracts from changes to maturities.
Next, I turn to the aggregate debt service burden of allnon-financial corporations in column (2). Because the borrowers in my sample make up a significant part of total borrowing, I their geographical distribution is likely a solid approximation for the geographical distribution of total debt. With that assumption in mind, I can calculate the macroeconomic savings due to BAPCPA by scaling up the in-sample estimate of $8.109 billion by the factor of total to in-sample debt ($5$2,,070686 billion). This implies total interest burden savings of $15.308 billion as a result of the caseload drop following BAPCPA.
How do these firm-level estimates compare to those from changing the inputs in the formula in equation4.3? Answering this question requires a number of additional inputs, are assumed to be equal across firms: the fraction of total debt accounted for by bonds and loans, as well as the average maturity and interest rates of the outstanding debt in each category. I describe how to calculate these in more detail in the online appendix and compare them to data reported by the BIS. Accounting for the drop in spreads due to BAPCPA in the aggregate implies total savings of $9.536 billion. This is slightly lower than the estimate of $15.307 billion when taking heterogeneity in firm- level exposure into account. It does, however, suggest that changes to the aggregate interest burden yield similar results to those based on firm-level data.
How does these savings in interest burden compare to the costs of BAPCPA? The United States Government Accountability Office estimated implementation costs of around $72.4 million – but these were almost exclusively related to the legal changes for consumer cases, and in fact do not incorporate the 28 new temporary judgeships BAPCPA created (USGAO,2008). As I will discuss in more detail below, bankruptcy judgeships cost approximately $1 million a year, adding $28 million in annual ex- penses. These numbers pale in comparison with the estimated savings from the exo- genous drop in caseload for my sample firms alone, which clocked in at around $8.1