What are the best datasets for financial research?

For financial research, the best datasets depend on whether you need market prices, fundamentals, macro indicators, or alternative data, and whether you require free/open or commercial/academic sources. Below is a practical shortlist of widely used, high‑quality datasets in 2026. forage

Free / open datasets (strong starting points)

Macroeconomic and economic indicators

• FRED (Federal Reserve Economic Data)

  • ~845,000 time series on GDP, inflation, employment, interest rates, yield curves, money supply, housing, etc., with vintage data for revised series. forage
  • Free API; standard baseline for US macro research and many global series via partner sources. forage

• World Bank Open Data, IMF Data, OECD Stats, Eurostat

  • Country‑level development, fiscal, trade, debt, and macro indicators; essential for cross‑country work. forage

Company fundamentals and filings

• SEC EDGAR

  • All US public company filings (10‑K, 10‑Q, 8‑K, etc.) as filed; canonical source for fundamentals, footnotes, and event dates. forage
  • Free, but you’ll need to parse filings yourself or use a wrapper library.

• Tiingo, Polygon.io, Financial Modeling Prep (free tiers)

  • Provide income statements, balance sheets, cash flows, and some price data for US equities with rate limits; useful for prototyping. forage

Market prices (equities, crypto, some futures/FX)

• HF Data Library

  • Free 1‑minute OHLCV for 1,391 US stocks/ETFs (2002–present), positioned as a free TAQ/CRSP‑like alternative for academic work. hfdatalibrary

• Polygon.io, Tiingo, Alpha Vantage, Yahoo Finance (unofficial)

  • Offer historical and (in some cases) delayed prices for equities, ETFs, FX, and crypto; free tiers have rate limits and coverage gaps. deepsdata

• Crypto exchange archives (e.g., Binance, Coinbase)

  • Direct trade/orderbook dumps from exchanges; preferred for crypto microstructure research over scraped aggregators. deepsdata

Alternative / event data (free or partially free)

• GDELT
  • Global news/events dataset at massive scale; useful for sentiment, geopolitical risk, and event‑driven strategies. forage
• Specialized free packs (e.g., fffinstill’s CSVs)
  • Curated datasets on insider trades, congressional trading, valuation multiples, risk metrics, etc., often with ready‑to‑load CSVs and code examples. fffinstill

Commercial / academic databases (gold standard for serious research)

These are the core stacks cited in top finance journals and used by quant teams: worldmetrics

• CRSP + Compustat (via WRDS)

  • CRSP: survivorship‑bias‑free US equity prices and returns.
  • Compustat: standardized fundamentals for global firms.
  • Standard combo for asset pricing, corporate finance, and accounting research. libguides.usc

• I/B/E/S

  • Analyst estimates and revisions; widely used for expectation‑based studies and earnings surprise research. libguides.usc

• Refinitiv (Datastream, Tick History), FactSet, S&P Global (Capital IQ, Compustat alternatives), Bloomberg

  • Broad coverage of prices, fundamentals, ownership, M&A, bonds, derivatives, and macro; strong for institutional research but expensive. worldmetrics

• Nasdaq Data Link (formerly Quandl)

  • Mix of free and premium datasets; convenient API delivery for many alternative and reference series. worldmetrics

How to choose

• Academic / publishable work: If you have access, use CRSP + Compustat (+ I/B/E/S) via WRDS; supplement with FRED/SEC EDGAR for macro and filings. forage
• Independent quant / ML research on a budget: Combine FRED, SEC EDGAR, HF Data Library (for intraday), and a paid API like Polygon.io or Tiingo for broader price/fundamental coverage. forage
• Macro / cross‑country studies: Start with FRED, World Bank, IMF, OECD, and Eurostat; add GDELT if you need news/event proxies. forage
• Crypto‑focused research: Use exchange‑provided archives plus a market data API (Polygon, Tiingo) for convenience and normalization. deepsdata

If you tell me your focus (asset class, region, frequency, and whether you need tick‑level or daily), I can suggest a concrete dataset stack and example download paths.