Point in Time Equity Fundamentals: Restatement-Safe Sources

Point in time equity fundamentals is a property of a database, not a field inside it. A point-in-time table keeps every historical snapshot of a company's reported figures keyed to the period they describe, so a query for Q2 as seen on any given date returns exactly what the filing said then — not the number that survived three later revisions.

The contrast matters because most convenient sources are revision-shaped rather than history-shaped. A daily bars feed such as Tiingo Financial Markets API exists to price portfolios, and its corporate-action-adjusted end-of-day series for 80,000+ monitored assets back to 1962 is deliberately rewritten every time a split or dividend lands. That is correct for returns and fatal for factor construction: rerun a value screen against a fully adjusted, fully restated file and you have quietly handed the model information that did not exist at decision time.

Nasdaq Data Link (formerly Quandl) is the catalog's answer. It aggregates hundreds of products from Nasdaq and third-party providers — Sharadar equity fundamentals, Mergent, Zacks, CFTC positioning and QuoteMedia among them — and its tables products expose row-level entity records: one row per ticker per period, with multi-decade histories. Individual tables run from thousands of rows up to billions for full equity fundamentals histories.

Look-ahead bias enters whenever the data available to a strategy differs from the data that existed at the simulated moment. Three mechanisms drive most of it, and each has a concrete fix in these catalogs.

First, restatements. When a company reissues prior-period statements, a last-version-wins table silently rewrites history; a factor computed from the revised figure could not have been computed at the time. Row-level tables on Nasdaq Data Link keep the record keyed to its reporting period, so the as-first-reported value stays addressable.

Second, alignment drift. Sharadar-style per-ticker-per-period granularity lets you join on fiscal period rather than calendar guesswork, which is what keeps a December fiscal-year-end company from contributing January data into a March-quarter bucket.

Third, adjustment contamination. Corporate-action-adjusted prices are the right input for returns and the wrong input for valuation ratios. Tiingo adjusts end-of-day OHLCV for splits and dividends across 80,000+ monitored assets — precisely why those series should never be divided by a stale book value from another vintage. Keep the fundamental table frozen and let the price engine do its own adjusting.

Depth varies more than vendors advertise. On the price side, Tiingo reaches furthest in this industry with end-of-day prices back to 1962 and fundamentals stretching 20+ years; Alpha Vantage's intraday archive runs to January 2000 while its indicator library applies to 1-minute through monthly bars; Massive carries US stock tape from 2003, options from 2014, currencies from 2009, futures from 2017 and indices from 2023; LSEG Data & Analytics holds tick history from 1996 in an archive exceeding 80 petabytes alongside 53 million economic indicators and instruments.

For statement-level fundamentals, the honest comparison is between Tiingo's 20+ year window across 80,000+ monitored assets and the multi-decade Sharadar and Mergent tables distributed through Nasdaq Data Link, whose individual tables scale to billions of rows. Neither figure is a rounding error difference: a 20-year window covers roughly two full credit cycles, while multi-decade coverage reaches into pre-GICS regime shifts that change how value and quality factors behave.

Macro context stretches deepest of all, and it is free. BIS Statistics Exchange and Derivatives Data publishes total credit covering 1,133 series x 344 quarters reaching back to 1940-Q2 across 40+ economies, quarterly, under unrestricted use with citation — long enough to condition factor exposure on leverage regimes that predate every commercial fundamentals table above.

A defensible build uses one vendor per layer rather than forcing a price feed to impersonate a research database.

The financial exchanges data data guide maps all 19 datasets in this industry — 11 primary plus 8 secondary — including the price-history and macro sources around this core. Pair this page with the tick-by-tick stock data cluster if microstructure matters to your strategies, and check the split dividend adjusted price history cluster before trusting any adjusted series in a ratio denominator. Every figure above traces to those catalog records as of August 2026.