Google+ Followers

Thursday, June 13, 2013

There is no Evidence that Unsecured Recoveries in Chapter 11 Are Diminishing (Part 1: Overview)

The American Bankruptcy Institute formed a 22-member commission in 2011, having as its stated purpose "to study and propose reforms to Chapter 11 and related statutory provisions that will better balance the goals of effectuating the effective reorganization of business debtors—with the attendant preservation and expansion of jobs—and the maximization and realization of asset values for all creditors and stakeholders".  Obviously, the premise of the commission and its mission statement is that there is something wrong with the recent operation of the Bankruptcy Code.  In a June 2011 article in the ABI Journal, the co-chairs contended that  "The ubiquitous and expansive use of secured credit is a primary challenge to the traditional reorganization regime. How can reorganization be financed when there is little or no apparent equity in the assets? In the case of overleveraged balance sheets, how are employee, tort claimant and trade creditor interests to be protected, or at least recognized, much less equity interests? An examination of priorities, adequate protection, surcharge and other factors may be in order."   The co-chairs also claimed that the ABI should undertake the task of fixing the perceived problems because “it is not a political organization, but rather is unbiased, and thus has no interest other than intellectual honesty and an open process.”  (That claim, of course, is only true if you think the only bias in the world is political, as opposed to, say, economic.) 
On March 14, the commission held a public session at which Joshua Gotbaum, head of the PBGC, offered a written, oral and visual presentation that argued that “[o]ver the decades, the balance has tilted away from debtors, shareholders, and unsecured creditors, and toward secured creditors.”  As proof, he offered a visual aid that purported to show that the “magnitude of the shift is breathtaking.”  The visual purported to show that, “in the early 1990s, a general unsecured claim in a large public company bankruptcy paid on average 77 cents on the dollar.  By 2010, that had declined to 45 cents on the dollar.”

The visual aid stated that it was based on an article entitled “The Decline of Unsecured Creditor and Shareholder Recoveries in LargePublic Company Bankruptcies” by one Andrew A. Wood, a 2011 graduate of UCLA Law School, who appears to have written the article while a third-year student under the tutelage of Professor Lynn LoPucki of that school.  The article appears in volume 85 of the American Bankruptcy Law Journal (2011). 

Wood’s article compares (1) a set of 42 chapter 11 cases that had confirmed plans of reorganization in the 2009-10 timeframe to (2) a larger set of chapter 11 cases that had confirmed plans of reorganization in the 1991-96 timeframe, which was analyzed in similar fashion by LoPucki, in an article entitled “The Myth of the Residual Owner” in the Washington University Law Quarterly (2003).

I am familiar with the article and, while I want to be diplomatic -- as it was written by someone who was only a student at the time -- it is not very good at all.  It is rife with major errors and oversights, dozens of which I was able to spot within only a few hours of investigation of the dockets of the bankruptcy cases sampled in the article.  It contains hypotheses that are totally unsupported by the verifiable data.

But, before I turn to a detailed analysis of the Wood article, I don’t believe that Gotbaum has even accurately presented the numbers in it.  Gotbaum said the 1991-96 cases provided a recovery of 77% for “general unsecured creditors”, but the Wood article has conflicting statements on that.  On page 436, Wood writes “General Unsecureds [capitalized terms his] recovered 77 cents in 1991-96 but only 53 cents in 2009-10.”  On page 430, however, Wood says “on average, general unsecured creditors in the 1991-96 cases recovered 60 cents on the dollar.”  I have no idea how Gotbaum decided to select 77% over 60%, or how Wood derived or reconciled two inconsistent statements.[1]  Further, Wood’s only citation for the 77% figure is the last two pages of LoPucki’s 2003 article, but “77%” does not appear on those pages or in any other location in LoPucki’s 2003 article (which I searched using both my own eyes and Adobe Acrobat Reader’s “Find” function) and Wood does not show how he calculated 77%.

Worse, I also found no explanation for the 45% figure Gotbaum employed.   Wood consistently uses the figure of 53% for general unsecured recoveries over his 2009-10 sample.  You can see the 53% figure in the quote from page 436 in the preceding paragraph.  On page 430, as well, he states “On average, general unsecured creditors in the 2009-10 cases recovered only 53 cents on the dollar”.  Finally, on page 437, a graphic representation of his analysis continues to use the "53 cents" figure.  I cannot find anywhere in Wood’s article where the 45% figure crops up at all, and I cannot tell if Gotbaum is comparing apples to apples as his visual implies. 

In sum, the “bottom line” numbers Gotbaum presented to the commission appear very shaky and unreliable, even before I get into the flaws in the Wood article.

Overview of the Main Problems with Wood’s Thesis.

Wood’s article contends that one of three hypotheses explains the purported decline in unsecured creditor recoveries: (1) increased prevalence of 363 sales; (2) increase in secured debt’ (3) decline in valuations attributable to the “Great Recession”.  He decides “The most plausible explanation is that the amount of secured debt held by companies has rapidly increased over the last decade.  This is partly due to the increase in second-lien financing.” 

As I shall explain in my next post, his conclusion is wrong for one simple reason: even though the bankruptcy court dockets for each of the cases listed in Wood’s article are publicly available on PACER (and in many cases on claims’ agent’s websites) and Wood claims to have checked many of the disclosure statements for other purposes, Wood did not investigate the capital structures of any of those debtors for the presence of second-lien debt.  Had he done so, as I show in my next post, he would have observed that (1) second-lien debt showed up in less than one-quarter of the sample and less than one-third of the cases with recoveries below the “77%” figure cited by Gotbaum;  (2) most of the cases that contain second liens are in fact “split-lien” deals that are relatively conservative forms of financing; (3) second-lien recoveries on average were less than general unsecured recoveries in his sample; and (4) second lien recoveries meaningfully exceeded general unsecured creditors in only two cases – less than 5% -- of  Wood’s total sample. Wood clearly had no basis to support his thesis about second lien financings.  

Beyond his failure to investigate the facts of his sample pertinent to his thesis, Wood’s data and analysis are badly flawed and skewed in several other very significant ways.

First, his article relies entirely on a database known as the “UCLA – LoPucki Bankruptcy Research Database” which purports to hold data regarding every publicly traded company that filed for bankruptcy since 1980 (Wood article p. 431).  Unfortunately, that database is essentially useless on the subject of creditor recoveries.

Wood states that the recovery percentages in the UCLA database come from the “Bankruptcy Data Source database” and were generally used “without further investigation”.[1] However, where there was not “sufficient information”, he “attempted to clarify by reading the Disclosure Statement on PACER…. If the PACER docket did not provide the needed information, [he] omitted the case from the study. ”  He also claimed that when “PACER revealed … that the plan confirmed was not the plan analyzed in the BDS database” he “adjusted the data accordingly”.  Last, he claimed that, in certain cases, specifically mentioning Abitibi-Bowater, the disclosure statement was “insufficient to support independent calculation [so he] omitted such cases from the study.”  Despite his research, he reported that he was only able to establish sufficient data for “50% of the cases … having plans confirmed in 2009 and 2010.” [2]

For the cases he did select, Wood ascribes a recovery percentage to unsecured creditors of that debtor.  From those recovery figures, and other work he reports doing, (a) Wood appears to have derived his sample’s recovery percentage, whether it be “45%” or “53%” and (b) someone, whether LoPucki or Wood, must have derived the “77%” figure attributed to the 1991-96 sample. 

I will show in my third post that the recovery figures given by Wood are astonishingly inaccurate.  First, the database itself provides much less recovery data than Wood suggests. All it contains is a column identified as “BX” and titled as “DistribUnsecPct” that seems intended to say what unsecured creditors received in a given case.  However, for the vast majority of cases, the column is empty.  Notably, the absence of recovery data in the “database” is pervasive: it is missing not just from the recent era of cases that Wood focuses on, but from the earlier era of cases that LoPucki covered in his 2003 article.  In the version of the database that I downloaded, which is dated March 6, 2013, there were 971 cases recorded, but only 119 – just 12% -- had data in the “DistribUnsecPct” column. The remaining 88% had no data on unsecured recoveries.

Even when the “DistribUnsecPct” column does contain a recovery percentage for a given case, there is no explanation of how it was calculated.  But unsecured creditors are often classified differently and different classes may have different recoveries. There is no way of knowing by looking at the database what class or classes of unsecured creditors are covered by the figure.   Thus, anyone who accepts the figures in the database “without further investigation” is, I am sorry to say, ignorant of chapter 11’s complexities. 

Last, even when the database gives recovery information, it is sometimes just inaccurate -- for example, it specifies unsecured recoveries in Six Flags at 13.1% when they actually received over 100%.   So the database by itself contains little useful information on the subject of unsecured recoveries and is completely unreliable.

Instead of relying on the database for recovery percentages, I simply went to the underlying source documents – typically the recovery estimate table in the court-approved disclosure statement -- in each case.  I found a high number – more than 1/4 -- of the recovery estimates listed in Wood’s article for “General Unsecured Creditors” were materially wrong. 

I was able to check the cases listed by Wood because, being quite recent, the cases were all available on PACER.  But, the disclosure statements or other documents relevant to recovery information in the 1991-96 cases cited in the LoPucki study are generally not available on PACER and cannot be checked nearly as easily. Yet, the extent of error that I shall show in Wood’s tabulations of 2009-10 cases has to throw his (or LoPoucki’s) calculations of 1991-96 recoveries into question:  how reliable can their calculations of older cases be when their calculations of the more recent era is often wrong? 

There is no basis to believe they got one era right and the other era wrong.  Rather, because of the pervasiveness of the errors in Wood’s analysis of 2009-10 cases, not only should the recovery percentages for that era that Wood derived be disregarded, but likewise the “77%” recovery figure attributed to the earlier cases has to be regarded as unreliable. Neither figure can be used as a basis for comparison when each is derived from such a poorly maintained database and the author(s) have failed to show how they derived either figure.

After reviewing the database errors, I will go on to show in my fourth post that the difference in recoveries between the two samples is partly a misleading artifact that arises from the methodological error of comparing two mismatched samples.  First, the 1991-96 time period was a 6-year bull market with an IRR of approximately 18%.  The 2009-10 time period was a 2-year bull market with an IRR also of approximately 18%.  Because Wood took simple averages of both samples, and valuations increased for a much longer time over the earlier sample, simple arithmetic should make the earlier sample’s average a higher number.  There were no legal or structural or policy factors at work. 

Second, as one can tell just by looking at the tables in the LoPucki study and the Wood article, there is a meaningful drop in the number of cases with debt falling into one of the “subordinated debt” columns from the former to the latter.   But one type of subordinated debt that neither table sets forth is “structurally subordinated” debt, i.e., debt at a holding company level that is subordinated in effect to the claims of debt at the operating subsidiaries.  It appears that, instead of maintaining a separate column for structurally subordinated debt, both authors drop it into the “general unsecured” category, or sometimes into the “senior unsecured” category, or bury it in a range of recoveries.  This has the effect of understating recoveries of general unsecured creditors, compared to the figure that would result were structurally subordinated debt to be placed in its own category as contractually subordinated debt is.  I provide further detail of this analysis in my fifth post.

Last, because the Wood and LoPucki studies focus just on disclosure statement estimates of unsecureds’ recoveries, they ignore the several other ways in which unsecured creditors are taken care of in chapter 11 cases.  Using a small subset of Wood’s sample, my fifth post on this topic shows that the size of the classes of general unsecured creditors in the 2009-10 cases that were presented in the disclosure statements is but a small fraction of the actual amount of general unsecured debt existing when the cases commenced.  This means that many such unsecured debts were satisfied outside the plan of reorganization, for example through “first-day orders” or through other provisions of the plan, such as cure payments upon assumption, that are not captured in the recovery estimate for the remaining unsecureds.   In order to properly comprehend unsecureds’ recoveries in any given case, or the change over time in such recoveries, it is necessary to find out the amounts paid to them in all contexts, and not merely the one quantified in the disclosure statement, which proves to be a small piece of the puzzle.

I have also posted a sixth essay identifying three further questionable aspects of the Wood/LoPucki analysis: disregard of post-petition interest; calculating average recoveries on a case-weighted basis as opposed to claim weighted basis; and relying exclusively on disclosure-statement estimates of recoveries instead of considering post-emergence appreciation, which was in some cases significant. .

In my next post, I will explore in detail the lack of evidence for the thesis that the supposed increased prevalence of second lien debt has diminished unsecured creditors recoveries.

I updated this on June 29 to provide the links to the later posts and to include a link to the sixth essay.

[1]           All quotes in this paragraph are from page 432 of Wood’s article.
[2]           Among the cases omitted by Wood are such mega-cases with full recoveries for unsecureds as General Growth  Properties.  As GGP was a solvent estate, its unsecureds obviously recovered at least 100%, and I cannot fathom how Wood failed to include it.

[1]           Wood gives a definition of the capitalized term “General Unsecured” that does not shed any light on the inconsistency: “’General Unsecured’ is a variable that describes how many cents on the dollar general unsecured creditors recover” and quoting LoPucki tells the reader that “’all unsecured creditor classes [are assumed] to be general unsecureds unless words suggesting different priority are used’”.  Neither of those explains how one sentence assigns a 77% recovery and another a 60% recovery to what appear to be the same group.