Large public-company bankruptcies

Download the data and code

• bankruptcies_monthly.csv558 months · 15 KBThe monthly series, with counts by chapter and the twelve-month mean.
  Download CSV
• bankruptcy_cases.csv81 cases · 13 KBEvery case from January 2023, with assets, chapter and a link to the SEC filing.
  Download cases
• update_bankruptcies.py19 KBPython script that rebuilds the series and, with --harvest, finds new filings on EDGAR for review. Needs pandas.
  Download Python
• update_bankruptcies.m11 KBMATLAB script that rebuilds the series from the case list and draws the chart.
  Download MATLAB
• not_cases.csv1 KBFilings that were read and not counted, with the reason. The harvest skips them.
  Download

The counts to 2022 are written into both scripts. To update the series yourself, run the Python script with --harvest once a month has ended, read the filings it lists, add the genuine petitions to bankruptcy_cases.csv, and run it again. Python and MATLAB give identical output.

Read the code

Loading the script…

How the series is built

The series counts Chapter 7 and Chapter 11 petitions by US public companies with total assets of at least $100 million in 1980 dollars, the definition used by the LoPucki Bankruptcy Research Database. Each case is counted in the month its petition was filed. The database ended with its December 2022 release, so the series joins two sources at January 2023.

From January 2023 each case passes through five steps:

1. Search. SEC full-text search finds every 8-K whose items include 1.03. Two search phrases are used, so that a filing missed by one is caught by the other. A company must file the 8-K within four business days of its petition.
2. Size. The filing is kept if the company’s largest total assets reported in its XBRL financial statements over the fifteen months before the event exceed $100 million in 1980 dollars, scaled by the annual consumer price index.
3. Read. The Item 1.03 text must show that the company itself filed a Chapter 7 or Chapter 11 petition. A filing by a subsidiary alone, a later report on the plan or the company’s emergence, and a receivership do not count.
4. Date. The case is dated by the petition date stated in the filing, not the 8-K’s report date, which often refers to an earlier event such as a restructuring agreement.
5. Count once. A case reported in several 8-Ks, or by several registrants in one jointly administered case, is counted once.

bankruptcy_cases.csv lists every case from 2023 with a link to its filing. not_cases.csv lists the filings that were read and not counted, with the reason.

Checking the new source against the database

Both sources exist for 2012 to 2022, so the SEC-filing method can be tested against the database over eleven years. It finds 266 cases against the database’s 276. The monthly counts have a correlation of 0.88, rising to 0.96 for five-month averages and for annual totals.

The SEC count misses companies that deregister before they file and companies that report the petition under another 8-K item, about one database case in seven over the overlap. It also finds some cases that the database lists under a different company, or not at all. The misses are spread evenly over time, so the counts track each other closely.

Before you use it

• Recent months can change. If a missed filing turns up later, the case is added and that month goes up by one. Counts before 2023 are fixed.
• What the SEC count cannot see. From 2023 the series misses companies that deregister before filing, those that report the petition only under Item 7.01 or 8.01, those with no XBRL balance sheet in the fifteen months before the petition, and foreign companies that file 6-Ks instead of 8-Ks.
• When a month is complete. A month enters the series about a week after it ends, once the four-day 8-K deadline has passed for every petition in it.
• Small numbers. Since 2021 there have usually been zero to three large cases a month, so one case moves the monthly count a lot. The twelve-month mean, cases_ma12, is the steadier measure.
• Chapters. cases_ch11 and cases_ch7 add up to cases, except for three database cases with no order for relief.
• The database itself. The case-level LoPucki database is licensed by its authors. Only monthly counts derived from it are published here.

Why it matters

Large bankruptcies are a direct record of corporate distress. Credit spreads and default probabilities mix the risk of default with the price investors put on it; the count of filings does not. It rises in every recession on the chart: 97 large cases in 2001, 91 in 2009 and 56 in 2020, against 16 in 2025.

Continuing the LoPucki series from SEC filings, with the database’s own definition, keeps that record going and lets results built on it be tested on data that did not exist when they were found.

How to cite

If you use the series, please cite the LoPucki database for 1980 to 2022. A reference for this page, which covers the continuation from 2023, is included in case it helps.

@misc{LoPuckiBRD2023,
  author       = {LoPucki, Lynn M.},
  title        = {Florida--UCLA--LoPucki Bankruptcy Research Database},
  howpublished = {University of Florida Levin College of Law and UCLA School of Law, final release January 2023},
  year         = {2023}
}

@misc{PybisBankruptcies,
  author       = {Pybis, Sam},
  title        = {Large {US} public-company bankruptcies, monthly, 1980 to September 2026},
  year         = {2026},
  howpublished = {\url{https://sampybis.co.uk/updated-data/bankruptcies.html}},
  note         = {Updated 20 October 2026}
}