# District bankruptcy timelines, v1 (FJC-derived)

Generated 2026-09-20 by `scripts/build_district_timelines.py`. Data file: `district_timelines_v1.csv` (947 rows: 475 Chapter 7 cells, 472 Chapter 13 cells, 25 suppressed).

## What it is

For every U.S. bankruptcy district and filing fiscal year, the 25th percentile, median, and 75th percentile number of days from the petition date to:

- **Chapter 7:** the debtor's standard discharge (FJC code `d1fdsp = 'A'`), filing FY2019 through FY2023.
- **Chapter 13:** case closing, for cases whose first-listed debtor received a discharge (`d1fdsp IN ('A','B')`, standard or hardship), filing FY2015 through FY2019. Only cohorts old enough to have almost entirely resolved are included (pending share 1.5% for FY2019; FY2020 is 5.5% and is left out).

Filing fiscal year is the federal fiscal year of the petition date (October 1 through September 30; FY2023 = 2022-10-01 to 2023-09-30).

## Source and snapshot

Federal Judicial Center Integrated Database, Bankruptcy Petition NewSTATS Snapshots (BPNS). One row per case: the latest quarterly snapshot row for each `casekey` (the FJC file carries one row per case per reporting period; a naive count overstates by about 3.5x). Snapshot in use: **filings through 2025-12-31, dispositions through 2026-01-15, spine built 2026-06-12** (`daylight/aggregates/fjc_spine.db`, `cases_latest`).

Universe: voluntary petitions only (`casetyp='v'`); intra-district transfer records excluded (`origin!='i'`); chapter is the chapter at filing (`orgflchp`). Dates in the FJC file are MM/DD/YYYY text and are reformatted to ISO before the day arithmetic. Rows where the disposition or closing date falls on or before the filing date (a source defect, about 2% of Chapter 7 discharges) are excluded from the quantiles and reported in `n_excluded_nonpositive`.

## The denominator convention, pinned

This dataset publishes **counts and durations, not rates.** Each cell carries:

- `n_cohort_filed`: every case in the universe filed in that district and fiscal year, any disposition or none.
- `n_pending_no_disposition`: cohort cases with a blank disposition code at the snapshot.
- `n_measured`: the cases the quantiles are computed on. Chapter 7: `d1fdsp='A'` with a positive day count. Chapter 13: `d1fdsp IN ('A','B')` with a positive day count to `closedt`.
- Chapter 13 only: `n_plan_completed_closed_ch13` and `median_days_plan_completed` restrict further to cases that closed under Chapter 13 (`clchpt='13'`), which excludes cases converted to Chapter 7 and then discharged.

No column divides one count by another. If you compute `n_measured / n_cohort_filed` yourself, you are computing a discharge share over *filings*, which is not the convention behind the site's published Chapter 13 discharge and dismissal percentages (those are over a terminated-case denominator whose exact code set is still being reconciled). Do not present such a ratio as the site's figure.

## Survivor bias: read this before quoting a median

**A timeline measures the cases that finished the way the column says.** The Chapter 7 median is the median among cases that reached a standard discharge. Cases dismissed, converted, closed without discharge, or still open are not in it. The Chapter 13 median is the median among discharged cases only; in a typical cohort more Chapter 13 cases are dismissed than discharged, and those dismissals, which usually happen earlier, are not in the number. A district with fast discharges and many dismissals will look faster than a new filer should expect. These are descriptive durations for completed cases, not a forecast of how long any particular case will take.

Cells with fewer than 30 measured cases keep their counts but carry no quantiles (`suppressed=1`).

## National reference values

| chapter | filing FY | n_measured | p25 | median | p75 |
|---|---|---:|---:|---:|---:|
| 13 | 2015 | 131,586 | 1173 | 1673 | 1958 |
| 13 | 2016 | 127,720 | 1184 | 1707 | 1958 |
| 13 | 2017 | 128,180 | 1209 | 1722 | 1959 |
| 13 | 2018 | 126,758 | 1216 | 1734 | 1963 |
| 13 | 2019 | 126,718 | 1217 | 1740 | 1965 |
| 7 | 2019 | 440,193 | 97 | 103 | 112 |
| 7 | 2020 | 380,183 | 97 | 103 | 114 |
| 7 | 2021 | 287,869 | 95 | 100 | 109 |
| 7 | 2022 | 208,574 | 94 | 99 | 109 |
| 7 | 2023 | 225,096 | 95 | 101 | 109 |

## Cross-check against an independent source

Three districts were recomputed from `bankruptcy_master.db` (PACER-sourced rows only, `source='pacer_api'`; CourtListener-sourced termination dates are never used for timelines because their ordering-error rate is two orders of magnitude higher). Chapter 7: days from `date_filed` to `date_discharged` where `disposition='Standard Discharge'`. Chapter 13: days from `date_filed` to `date_termed` where a discharge date exists. Master-DB cells with fewer than 30 cases show no median. The two sources are independent (FJC statistical extract vs. PACER docket reports) and cover different case samples, so agreement is a consistency signal, not a proof.

**Chapter 7** (FJC: filing to `d1fdspdt`, code A; master: filing to `date_discharged`, Standard Discharge):

| filing FY | district | FJC n | FJC median | master n | master median | delta (FJC minus master) |
|---|---|---:|---:|---:|---:|---:|
| 2019 | ksbk | 3545 | 97 | 605 | 98 | -1 |
| 2020 | ksbk | 2775 | 97 | 534 | 98 | -1 |
| 2021 | ksbk | 2148 | 98 | 468 | 98 | 0 |
| 2022 | ksbk | 1481 | 97 | 345 | 98 | -1 |
| 2023 | ksbk | 1583 | 98 | 429 | 98 | 0 |
| 2019 | mowbk | 4253 | 101 | 586 | 100 | 1 |
| 2020 | mowbk | 3426 | 105 | 500 | 103 | 2 |
| 2021 | mowbk | 2563 | 105 | 399 | 103 | 2 |
| 2022 | mowbk | 1756 | 101 | 239 | 99 | 2 |
| 2023 | mowbk | 1681 | 102 | 261 | 99 | 3 |
| 2019 | flsbk | 9054 | 100 | 878 | 100 | 0 |
| 2020 | flsbk | 8566 | 100 | 330 | 101 | -1 |
| 2021 | flsbk | 7675 | 100 | 261 | 100 | 0 |
| 2022 | flsbk | 4816 | 100 | 156 | 100 | 0 |
| 2023 | flsbk | 4586 | 101 | 157 | 101 | 0 |

**Chapter 13.** Two comparisons, because the two sources measure different things. Column pair 1: the published measure (FJC filing to `closedt`, all discharged cases including those converted to Chapter 7) against master filing to `date_termed`. Column pair 2: FJC plan completions only (`clchpt='13'`) against master filing to `date_discharged`. The PACER-sourced master rows for Chapter 13 are almost entirely plan completions (converted cases carry chapter 7 there), and PACER's termination date trails the discharge date by roughly two to three months, so pair 1 is expected to disagree and pair 2 is the like-for-like test.

| filing FY | district | FJC n (all disch.) | FJC median to close | master n | master median to termed | delta 1 | FJC plan-completed n | FJC plan-completed median | master median to discharge | delta 2 |
|---|---|---:|---:|---:|---:|---:|---:|---:|---:|---:|
| 2015 | ksbk | 2073 | 1659 | 678 | 1810 | -151 | 1839 | 1747 | 1748 | -1 |
| 2016 | ksbk | 1849 | 1669 | 617 | 1909 | -240 | 1665 | 1770 | 1837 | -67 |
| 2017 | ksbk | 2041 | 1639 | 703 | 1768 | -129 | 1874 | 1689 | 1686 | 3 |
| 2018 | ksbk | 2208 | 1623 | 840 | 1697 | -74 | 2008 | 1680 | 1625 | 55 |
| 2019 | ksbk | 2141 | 1658 | 888 | 1738 | -80 | 1951 | 1702 | 1658 | 44 |
| 2015 | mowbk | 1287 | 1740 | 382 | 1816 | -76 | 1188 | 1774 | 1742 | 32 |
| 2016 | mowbk | 1266 | 1720 | 410 | 1799 | -79 | 1147 | 1780 | 1718 | 62 |
| 2017 | mowbk | 1358 | 1725 | 451 | 1818 | -93 | 1255 | 1770 | 1735 | 35 |
| 2018 | mowbk | 1363 | 1754 | 483 | 1834 | -80 | 1285 | 1777 | 1751 | 26 |
| 2019 | mowbk | 1526 | 1781 | 544 | 1821 | -40 | 1436 | 1803 | 1737 | 66 |
| 2015 | flsbk | 4290 | 1719 | 221 | 1909 | -190 | 3997 | 1852 | 1822 | 30 |
| 2016 | flsbk | 3772 | 1770 | 203 | 1915 | -145 | 3576 | 1855 | 1818 | 37 |
| 2017 | flsbk | 3546 | 1840 | 317 | 1912 | -72 | 3388 | 1883 | 1812 | 71 |
| 2018 | flsbk | 3429 | 1880 | 353 | 1919 | -39 | 3235 | 1912 | 1827 | 85 |
| 2019 | flsbk | 3572 | 1694 | 491 | 1903 | -209 | 3339 | 1837 | 1816 | 21 |

Summary: Chapter 7 absolute delta ranges 0 to 3 days across 15 cells. Chapter 13 pair 1 ranges 39 to 240 days (15 cells); pair 2 ranges 1 to 85 days (15 cells). The residual in pair 2 is not explained here; the master-DB sample is a fraction of each cohort and was pulled for reasons unrelated to timing, so it is not a random sample of the district.

## Columns

`chapter, filing_fy, district, district_label, fjc_district_code, measure, n_cohort_filed, n_pending_no_disposition, n_measured, n_excluded_nonpositive, p25_days, median_days, p75_days, suppressed, n_plan_completed_closed_ch13, median_days_plan_completed`. `district` is the CourtListener/PACER court id with a trailing `k` (for example `mowbk` = W.D. Missouri); `_national` rows aggregate all 94 districts.

## What this cannot tell you

No attorney, judge, trustee, county, plan-payment, or fee field exists in the FJC file. The pro se flag is one the FJC itself marks as unchecked and is not used here. Termination reasons beyond the disposition code are not available. The FJC file is a snapshot that is overwritten each release; this dataset reflects the state of each case at the snapshot above.

## License

CC0 1.0 Universal (public domain dedication): https://creativecommons.org/publicdomain/zero/1.0/. No attribution required. Suggested citation if you want one: Open Bankruptcy Project, "District bankruptcy timelines v1 (FJC-derived)," https://1328f.org/data/district_timelines_v1.csv, 2026-09-20.