SEC Financial Statement Data Sets

Each quarterly ZIP holds four tab-delimited TXT tables — SUB (submissions), TAG (tags), NUM (numbers) and PRE (presentation of statements). Because SUB carries each filing’s accession number and filed date, quarterly vintages can be assembled without look-ahead, which is what fundamental factor construction and backtest universes need from a fundamentals source. The numbers are extracted from registrants’ XBRL-tagged filings and appear as filed; the Commission adds its own filing identifiers and fields such as the Standard Industrial Classification.

Downloads follow a fixed pattern — https://www.sec.gov/files/dera/data/financial-statement-data-sets/2026q1.zip — with no registration. Automated clients must declare a descriptive User-Agent, or the SEC returns an “Undeclared Automated Tool” error, and the FAQ states a ceiling of 10 requests per second. The SEC publishes its own CC0-licensed loading code at sec-gov/python-for-dera-financial-datasets. Publication lags: as of 17 August 2026 the newest posting is 2026 Q1, and 2026q2.zip returns 404.

Three things to know before relying on it. In December 2024 the SEC reposted the series reprocessed to cover only submissions and numeric data from the primary financial statements as rendered by the Commission, replacing the earlier non-dimensional and co-registrant compilation and adding a segments field to NUM — cached pre-2024 downloads of a given quarter are therefore not comparable to current ones. Early coverage is thin because the XBRL mandate phased in: 2009 Q1 is only 13.22 KB, unusable as a broad universe. And the SEC disclaims accuracy, since the data derives from registrant submissions where extraction errors are possible, and states the data sets are not a substitute for the filings themselves. There is no licence file to point at; the reuse permission is prose in the agency’s FAQ, which says “All Government-created content on sec.gov and EDGAR public filing content are free to access and reuse.”