Why I Use AACER's Filing Statistics

Every month, I try to post on the latest bankruptcy filing statistics using data provided by a private company, Automated Access to Court Electronic Records (AACER). Recently, Jason Kilborn, a law professor at Chicago's John Marshall Law School, posted a comment asking why AACER's filing statistics were lower than the ones provided by the Administrative Office of U.S. Courts (AO). For example, AACER reported 826,665 bankruptcy filings in the 2007 calendar year, about 2.9% lower than the 850,912 filings reported by the AO.

The differences seem to stem from the AO's including a old but reopened bankruptcy case in its count of bankruptcy filings where AACER's data only counts new filings. The AO data also appear to count transferred cases. Thus, for new filings, AACER's statistics are the ones to use. Besides, as I have mentioned previously, the AO's data come out woefully late, as much as three months after the fact. AACER is able to provide timely data.

Without access to the AO's data to audit and compare their data to AACER's data, I only can say that the difference "seems" to be explained by the way the AO counts reopenings and transferred cases. But why do I hold this belief?

First, anecdotally, a few persons connected with the bankruptcy courts have suggested to me that the official AO filing data count reopenings and transfers. Second and more convincingly, Several bankruptcy courts make available filing statistics for their judicial district and with one exception, these local judicial district statistics are virtually identical to AACER's count and are far off from the AO's count. Consider the following numbers published for the 2007 calendar year as total bankruptcy filings (all AO data are from Table F as released on April 15, 2008 here):

In all of these examples, save one, the AACER number falls within 0.1% of the local court count for total bankruptcy filings. Because the bankruptcy filing count involves thousands of cases, this small percentage difference is not particularly troubling or even surprising. The one district where the local court figure differs dramatically from the AACER figure is the Northern District of Iowa. Interestingly, the Northern District of Iowa count says expressly that it counts reopenings (but is silent on transfers). In contrast, the Northern District of California site expressly says it does not include reopenings and transfers, and its figure is virtually the same as AACER's figure. One district even lays it all out. The District of Rhode Island data sheet totals to 2,767 filings for 2007 (although the sheet curiously shows a total of 2,770 filings) but then separately lists the number of transfers and reopenings. The 56 transfers and reopenings makes the local court numbers then reconcile with the AO figures. Thus, it seems pretty clear that counting transfers and reopenings is why the AO's numbers always run a few percentage points higher than the numbers from AACER.

If you've read this far, you may share my obsession with bankruptcy court data. Because I heavily use the AACER data, it was important for me to reconcile its data with the AO's data, but why should should anyone else care? First, if you are a policymaker or someone else who uses the bankruptcy filing data as some measure of economic health for the United States, you should care that the AO data overstate the true number of new bankruptcy filings by a few percentage points. This is a good lesson in being careful with the data one uses. The AO's primary mission is to provide data for the federal court system, where information about reopened and transferred cases may be important to judicial administration and workload measurements. Second, if those of us who use the bankruptcy filing data for time series analysis, one might worry about how long the AO has counted reopened and transferred cases in their bankruptcy filing data. If the data measurement was changed during the period of the time series analysis, we would have a discontinuous time series.

Sources for local bankruptcy court data:
Northern District of California; Northern District of Iowa; Southern District of Iowa; District of New Jersey; District of Rhode Island; District of South Carolina

Comments

[Jason Kilborn]

Thanks for this, Bob. I read the whole thing, so I guess I'm obsessed, too. This really makes me nervous about the European and Asian consumer filing data I've been tracking, and I don't have access to fancy privately compiled data to check the "official" figures. Oh, well. After reading this post and another by Frank Pasquale on Concurring Opinions ( http://www.concurringopinions.com/archives/2008/05/funny_money.html#trackbacks ), I am beginning to lose faith in government generated statistics entirely.

[Adam Levitin]

What about conversions? Is a 13 converted to a 7 (or vice-versa) counted as one or two filings?

[Bob Lawless]

Adam, AACER does not count conversions as a new filing. I do not know of the AO's practice. Because the AO's data seems to reconcile to AACER's data with just accounting for transfers and reopenings, it does not seem that the AO counts conversions as a new filing either.