This package contains methodology documentation, source code, data schemas, sample datasets, and interactive demos for the OBP analyses. All data sources are publicly available. All code is anonymized. Nothing in this package identifies specific attorneys, firms, or cases.
Contents: 8 documentation files, 4 Python scripts, 1 interactive demo, 1 sample dataset.
Cross-references FJC IDB prior filing history against case disposition across 4.9 million Ch.13 cases to identify 391,951 cases where prior filers received discharge with no evidence of eligibility verification.
Full description of the FJC cross-reference approach, data sources, filtering logic, and limitations.
View methodology.mdNational summary: 391,951 cases (27.4%), policy impact, Advisory Committee citation.
View findings.mdThe 391,951 finding broken down by all 94 federal judicial districts. Generated from actual FJC IDB data. Includes per-district prior filer rates, discharge rates, and estimated violations.
View district_breakdown.mdStandalone Python implementation of the 1328(f) screening algorithm. Takes PACER CSV exports and outputs violations. Also includes FJC IDB national screening mode. Zero dependencies beyond Python 3.8 stdlib.
View screener_logic.pySelf-contained HTML page that runs the screening algorithm in-browser against the sample dataset. Open in any browser - no server needed. Inspect the JavaScript to verify the logic matches the Python implementation.
Open interactive_demo.htmlStatistical tools for comparing attorney performance against district baselines. Two approaches: targeted comparison (specific attorney vs. baseline) and blind outlier detection (scan all attorneys, find anomalies).
Docket entry classification, response matching, rate calculation, and controls.
View methodology.mdCompares any attorney's metrics against their district's baseline. Computes Z-scores on 6 dimensions (dismissal, discharge, OSC, MTD, fee density, stay relief). Anonymized - no firm/attorney names in code.
View district_comparison.pyPoint-and-shoot anomaly scanner. Give it a court's data, it finds the statistical outliers. No target specified. Ranks all attorneys by composite Z-score. Includes anonymization mode.
View blind_outlier.pyPython script for calculating opposition rates from downloaded PACER dockets. Anonymized.
View opposition_rate.pyHow free, public court RSS feeds are used for real-time case monitoring at $0 cost. Covers architecture, event classification, attorney cross-reference, and replication instructions.
View rss_methodology.mdAnonymized results table showing response rates by motion type.
View sample_output.mdComplete documentation of every FJC field used in the analysis. Field names, types, code meanings, and how each maps to derived metrics. Includes the local analysis database schema.
View fjc_schema.md70-row anonymized dataset for testing the screener. Sequential case IDs (SAMPLE_001-070), generic debtor labels (Debtor_001-040), synthetic dates. Run screener_logic.py against this to verify the algorithm works.
Download sample_dataset.csvStep-by-step instructions for independently reproducing both analyses using public data.
View VALIDATION.mdAll analyses use publicly available data:
The GitHub repository with all source code is available on request: github.com/openbankruptcyproject/obp-methodology (private, collaborator access required).
You May Also Find Helpful