How to Build a Point-in-Time Fundamentals Database from SEC EDGAR (and When Not To)

Short summary

Point-in-time fundamentals prevent lookahead bias in backtests by tracking what data was reported and when it first appeared in SEC filings. This guide walks through a six-step methodology: pulling XBRL data from SEC EDGAR, filtering to 10-K filings, resolving tag synonyms, stamping first_filed dates, detecting restatements, and QA checks. It then honestly compares DIY builds (free but complex) against vendors like Tradevo Data ($49/mo), helping quants decide whether to engineer or buy.

• •PIT fundamentals solve lookahead bias by recording when data was first reported, not when it was restated
• •Six-step methodology covers companyfacts API queries, 10-K filtering, XBRL synonym resolution, restatement detection, and QA
• •Build vs. buy: DIY is free but fiddly with real pitfalls; $49/mo Tradevo Data handles 5,214 US companies and 18,539 labeled restatements

Generated with AI, which can make mistakes.