U.S. Bank Capital Research (2000-2025)

This project was completed as part of an academic Summer Research Scholarship focused on constructing a clean longitudinal dataset for all U.S. commercial banks from 2000 to 2025. I built Python ETL pipelines to ingest and structure data from FDIC, EDGAR, and CRSP, covering 176k+ observations across 11,000+ institutions. The work included designing identifier-linkage logic across FDIC CERT, CIK, GVKEY, and PERMNO, implementing quality assurance scripts to validate capital ratios, accounting identities, and time-series consistency, and engineering financial variables such as ROE, NIM, funding costs, and regulatory indicators. I also adopted and extended existing research code to estimate dynamic power-law distributions of U.S. bank assets, supporting analysis of cross-sectional mean reversion and idiosyncratic volatility across bank size groups. Due to research confidentiality, the full repository is private, but the system architecture and methodology reflect production-grade financial data platforms used in industry.