Data Specifications
Time Coverage
2000–present
Geographic Coverage
United States
Granularity
Firm × Lender relationship level
Update Frequency
Annual
Methodology
We deploy Big Data and Large Language Model tools to digitise and standardise over 40 million archival loan documents containing detailed information on lending relationships between firms and their creditors in the United States.
The resulting dataset covers 1.8 million US firms, predominantly composed of small and medium-sized enterprises (SMEs) — a firm size segment that has been historically difficult to study due to data limitations.
The data includes detailed information on credit relationships, enabling research on the real effects of financial shocks on firm performance. Our analysis of 179 bank failures from 1990 to 2023 reveals that firms with a credit relationship to a failed bank are 6.7 percentage points (44.3%) more likely to fail themselves within five years, with surviving firms exhibiting 25% lower employment growth.
License & Terms
Commercial license. Contact us for pricing and academic access.