A guide to 87 datasets for research in empirical finance and macro-finance, with notes on what each one is for and how to use it well. Free sources come first, starting with the two libraries behind most work on stock returns: Kenneth French’s factors and portfolios, and Amit Goyal’s equity premium predictors. They are followed by uncertainty measures, factor libraries, yield curves, real-time and survey data, academics’ own data pages, official statistics and the main subscription databases.
Kenneth R. French, Tuck School of Business, Dartmouth
Free
Frequency Daily, weekly (some), monthly, annual
Coverage US from July 1926; five-factor data from July 1963; developed and emerging markets from the 1990s
Why it’s useful
This is the reference library of empirical asset pricing, and the first place to go for almost any project on stock returns. Its factors are the standard benchmarks for risk adjustment: an alpha against the CAPM, the three-factor model, Carhart’s four-factor model (add momentum) or the five-factor model is computed from these series in almost every paper, so using them makes your results directly comparable with the literature.
The sorted portfolios are the standard test assets for asset pricing models (the 25 size–book-to-market portfolios, for example, in time-series GRS tests and Fama–MacBeth regressions). The industry portfolios open up lead–lag, sector-rotation and information-diffusion questions. The RF series is the usual one-month risk-free rate for turning returns into excess returns, and the NYSE breakpoints let you build your own portfolios in a way that matches the literature. All of it is free, long, and updated every month.
What’s in it
Fama–French three- and five-factor returns, plus momentum and short- and long-term reversal factors
Portfolios sorted on size, book-to-market, profitability, investment, E/P, CF/P, D/P, accruals, beta, net issuance and variance (one-, two- and three-way sorts)
Industry portfolios in eight classifications: 5, 10, 12, 17, 30, 38, 48 and 49 industries
NYSE breakpoints for size, book-to-market, profitability, investment and prior returns
Factors and portfolios for developed regions (Europe, Japan, Asia Pacific ex Japan, North America) and emerging markets
The one-month Treasury bill rate (RF) used by most of the literature
Amit Goyal, University of Lausanne and Swiss Finance Institute
Free
Frequency Monthly, quarterly, annual
Coverage Index, dividend and earnings series from 1871; most predictors from the 1920s or later; updated to 2025
Why it’s useful
This is the benchmark dataset for forecasting the equity premium, and the natural second stop after Kenneth French. Welch and Goyal (2008) showed that most predictors proposed in the literature fail to beat the historical average out of sample; their data, kept up to date by Amit Goyal, defines the set of predictors that any new one has to beat. Using it makes your in-sample and out-of-sample results comparable, line by line, with dozens of published papers.
The 2024 follow-up with Zafirov is the current state of play. Re-examining 29 variables proposed since 2008, it finds that more than a third are no longer significant even in sample, that half of the rest perform poorly out of sample, and that only a small number still work in both. If you are proposing a new predictor, such as uncertainty or an industry signal, this is the competition. The dataset is also the standard source for the price, dividend and earnings series needed for valuation ratios and return decompositions.
What’s in it
The Welch–Goyal (2008) predictors, updated to 2025: dividend–price ratio, dividend yield, earnings–price ratio, payout ratio, stock variance, book-to-market, net equity expansion, Treasury-bill rate, long-term yield and return, term spread, default yield and return spreads, inflation, and others
S&P 500 index level, dividends and earnings, CRSP value-weighted returns with and without dividends, and the risk-free rate
Goyal, Welch and Zafirov (2024): 29 newer variables from 26 papers published after 2008, alongside the original 17, with data to 2021 and to 2025 (MATLAB and CSV) and an Internet Appendix
A single Excel file combining both papers’ data up to 2025
Uncertainty, risk and sentiment
Measures of the state of the economy and of investors, used as predictors, regime indicators and controls.
Sydney Ludvigson (Jurado, Ludvigson and Ng, 2015; Ludvigson, Ma and Ng, 2021)
Free
Frequency Monthly
Coverage July 1960 onwards; updated about twice a year
Why it’s useful
Most papers proxy uncertainty with stock market volatility. These indexes instead measure how unpredictable a large panel of series is: the financial index summarises 147 financial series, from valuation ratios and spreads to the returns on industry, size, value and momentum portfolios. That makes them a cleaner measure of genuine uncertainty, and the macro–financial split lets you ask which kind matters.
Ludvigson, Ma and Ng show that shocks to financial uncertainty drive downturns rather than simply responding to them. In finance the indexes work as return predictors, as regime indicators for when other predictors work, as inputs to volatility forecasts, and in VARs of the macro-financial transmission of uncertainty.
What’s in it
Macroeconomic, financial and real uncertainty indexes at one-, three- and twelve-month horizons
Data vintages, so you can see the index as it was published at earlier dates
Scott Baker, Nicholas Bloom and Steven Davis (Quarterly Journal of Economics, 2016)
Free
Frequency Daily, monthly, quarterly (by index)
Coverage US news-based index from 1985 and a historical index from 1900; global and around 50 country indices
Why it’s useful
EPU captures uncertainty about policy: who will decide what, and when. Because it is built from newspaper text, it spikes around close elections, wars, debt-ceiling fights and fiscal battles, and Baker, Bloom and Davis show that it is associated with higher stock volatility and lower investment in policy-sensitive sectors.
In finance it is useful as an alternative or complementary uncertainty measure (it is only weakly related to financial uncertainty), for event studies around political episodes, and for international comparisons thanks to the country coverage. It is also a good robustness check whenever a result is attributed to “uncertainty”.
What’s in it
US EPU built from newspaper coverage, plus categorical indices (monetary, fiscal, trade, climate policy and more)
A global index and country indices, including the UK
Related trackers: World Uncertainty Index, Equity Market Volatility tracker, trade policy uncertainty
Dario Caldara and Matteo Iacoviello (American Economic Review, 2022)
Free
Frequency Daily and monthly
Coverage Recent index from 1985; historical index from 1900
Why it’s useful
A text-based measure of adverse geopolitical events and the risk of them, with spikes around the world wars, the Korean War, the Cuban Missile Crisis and 9/11. Higher geopolitical risk foreshadows lower investment and employment and is linked to larger downside risks.
Useful for studying safe-haven flows, oil and defence stocks, tail risk, and whether an uncertainty result is really about geopolitics. The split between threats and realised acts is especially handy for event-style analysis.
What’s in it
Aggregate geopolitical risk, split into threats and acts
The market’s own forward-looking estimate of volatility over the next month, often called the “fear gauge”. It is the most common real-time measure of financial-market stress, a natural regime indicator, and an input to the variance risk premium (implied minus realised variance), which is itself a well-known return predictor.
Bloom (2009) used the older VXO as his uncertainty measure, so the VIX family is the obvious comparison when you use broader uncertainty indexes such as Ludvigson, Ma and Ng’s.
What’s in it
Daily VIX (30-day implied volatility of the S&P 500) and related volatility indices
Malcolm Baker and Jeffrey Wurgler (Journal of Finance, 2006)
Free
Frequency Monthly
Coverage July 1965 to December 2025
Why it’s useful
The standard measure of market-wide investor sentiment. Baker and Wurgler show that when sentiment is high, stocks that are hard to value and hard to arbitrage (small, young, volatile, unprofitable or distressed firms) subsequently earn low returns, and vice versa. That makes it a key conditioning variable for anomalies and mispricing.
For aggregate return prediction it is the starting point for refinements such as the aligned sentiment index of Huang, Jiang, Tu and Zhou (2015). It is also a useful contrast with uncertainty: sentiment and uncertainty regimes are related but not the same, and Chu, He, Li and Tu (2022) show that forecasting regimes based on sentiment matter too.
What’s in it
A sentiment index built as the common component of several market-based proxies (such as IPO activity, turnover and the equity share of new issues)
Coverage Global stock, bond, FX and macro connectedness measures
Why it’s useful
A ready-made measure of how shocks spread across markets and countries, built on the Diebold–Yilmaz variance-decomposition framework. Useful for volatility-spillover questions, systemic-risk monitoring, and as a benchmark if you estimate your own spillover indices.
What’s in it
Total, directional and pairwise connectedness indices based on variance decompositions
Frequency Quarterly (monthly versions for some series)
Coverage Around 140 countries from the 1950s
Why it’s useful
The broadest cross-country uncertainty measure, built consistently for advanced and emerging economies alike. Use it when EPU is not available for a country, or for panel studies of how uncertainty affects growth, investment and asset prices around the world.
What’s in it
Country uncertainty indices built from the Economist Intelligence Unit’s country reports
Related indices on trade, pandemic and other uncertainty
Robert Engle and the Volatility Institute, NYU Stern
Free online
Frequency Daily
Coverage Thousands of assets and markets worldwide
Why it’s useful
Ready-made volatility and systemic-risk measures from the lab founded by Robert Engle, who won the Nobel prize for ARCH models. Useful as benchmark volatility forecasts, for SRISK as a measure of how undercapitalised financial firms would be in a crisis, and as a quick check on your own GARCH estimates.
What’s in it
GARCH-family volatility estimates and forecasts
Dynamic correlations, SRISK systemic-risk measures, long-run VaR, liquidity and climate-risk analysis
A single weekly read on US financial conditions, with positive values meaning tighter than average. It works as a financial-stress state variable, a regime indicator alongside uncertainty, and a control in macro-finance regressions. The St. Louis Fed Financial Stress Index in FRED is a close alternative.
What’s in it
The NFCI and its adjusted version (ANFCI), with risk, credit and leverage sub-indices
A weighted average of over 100 money-market, debt, equity and banking indicators
Neely, Rapach, Tu and Zhou (2014), replicated and updated by Sam Pybis
Updated here
Free
Frequency Monthly
Coverage S&P 500 from December 1950; 18 commodities, starting between 1961 and 1987
Why it’s useful
The fourteen moving-average, momentum and volume rules of Neely et al. (2014), reproduced exactly from the paper’s files and extended every month, with the same price rules for oil, metals and crops. The share of rules on a sell signal gives a trend uncertainty index that moves with the VIX, and the rules themselves are standard technical predictors of the equity premium.
What’s in it
Every buy or sell signal, by rule, asset and month
Trend uncertainty index, buy share and trend disagreement for each asset
Python and MATLAB code, and a check against the paper’s replication files
Factors and the cross-section of returns
Factor returns and anomaly libraries for testing whether a new signal is genuinely new.
Andrew Y. Chen and Tom Zimmermann (Critical Finance Review, 2022)
Free
Frequency Monthly (portfolios and signals)
Coverage Hundreds of published cross-sectional predictors; latest release October 2025
Why it’s useful
The “factor zoo”, replicated and downloadable. If you have a new cross-sectional predictor, this is how you check whether it is genuinely new: test it against the existing anomalies, see whether it is subsumed, and put your t-statistics in the context of everything already tried. The open code is also an excellent way to learn how published signals are actually constructed.
What’s in it
Portfolio returns for hundreds of published anomalies, replicated with open code
Firm-level signal data
Code in Python (translated from Stata in the 2025 release), with Python and R download packages
Theis Jensen, Bryan Kelly and Lasse Pedersen (Journal of Finance, 2023)
Free
Frequency Monthly
Coverage 153 factors in 93 countries
Why it’s useful
The best free source for international out-of-sample evidence. Jensen, Kelly and Pedersen show that most factors replicate, cluster into 13 themes, and work in a large new data set covering 93 countries. If a result only holds in the US, this lets you see whether it travels, and the theme clusters are a convenient way to control for the whole zoo with a handful of factors.
What’s in it
Factor returns for 153 characteristics across 93 countries, grouped into 13 themes
Regional and global portfolios, with documentation and code
Kewei Hou, Haitao Mo, Chen Xue and Lu Zhang (global-q.org)
Free
Frequency Daily, weekly, monthly, quarterly, annual
Coverage January 1967 onwards; updated annually
Why it’s useful
The main investment-based alternative to the Fama–French five-factor model. Reporting alphas against both is now common practice, because the q-factors explain many anomalies that the Fama–French factors do not. If an alpha survives both, a referee is much less likely to argue that it is a known factor in disguise.
What’s in it
The q5 model: market, size, investment, return on equity and expected growth factors, plus the T-bill rate
Well-known factors that are not in the Fama–French library, constructed by their original authors and available across many countries. Useful as additional controls (betting against beta and quality are frequent referee requests), for value and momentum outside equities, and as a check on how sensitive a value-related result is to the timing of prices.
What’s in it
Betting Against Beta (Frazzini and Pedersen, 2014)
Quality Minus Junk (Asness, Frazzini and Pedersen, 2019) factors and portfolios
The Devil in HML’s Details (Asness and Frazzini, 2013): a value factor using up-to-date prices
Value and Momentum Everywhere (Asness, Moskowitz and Pedersen, 2013) factors and portfolios across asset classes, plus momentum indices
Ľuboš Pástor, Chicago Booth (Pástor and Stambaugh, Journal of Political Economy, 2003)
Free
Frequency Monthly
Coverage 1962–2025 (updated December 2025)
Why it’s useful
The standard measure of market-wide liquidity risk. Pástor and Stambaugh show that stocks more exposed to aggregate liquidity shocks earn higher returns, and the series picks out liquidity crunches such as 1987, 1998 and 2008. Use it as a factor control, as a state variable, or to test whether a predictor is really about liquidity.
What’s in it
Aggregate liquidity levels, innovations and the traded liquidity factor
Decile portfolios sorted on historical liquidity betas
Data for Pástor, Stambaugh and Taylor’s green-returns paper
Robert F. Stambaugh, Wharton (with Yu Yuan and Jianan Liu)
Free
Frequency Daily and monthly
Coverage US mispricing factors 1963–2016; China factors updated separately
Why it’s useful
A behavioural alternative to the Fama–French and q-factor models, built from 11 well-known anomalies, plus the mispricing scores used to study sentiment and arbitrage limits. The China factors are the standard benchmark for research on Chinese stock returns, where size and value behave differently.
What’s in it
The four-factor mispricing model (market, size and two mispricing factors) and 11 anomaly long–short returns
Stock-level mispricing scores (Stambaugh, Yu and Yuan)
CH-3 and CH-4 size and value factors for China (Liu, Stambaugh and Yuan)
Interest rates and the yield curve
Zero-coupon curves, term premia and shadow rates: the inputs for bond-return, term-structure and monetary-policy work.
Federal Reserve Board (Gürkaynak, Sack and Wright, 2007)
Free
Frequency Daily
Coverage 1961 onwards; updated about weekly; a separate TIPS (real) curve
Why it’s useful
The standard daily zero-coupon Treasury curve, built from off-the-run notes and bonds. It is the usual source for bond returns, term spreads and forward rates in asset pricing, and the TIPS curve gives market-implied inflation compensation.
What’s in it
Fitted Svensson curve parameters and smoothed zero-coupon yields at maturities out to 30 years
A companion TIPS curve with real yields and inflation compensation
Yan Liu and Jing Cynthia Wu (Journal of Financial Economics, 2021)
Free
Frequency Daily and monthly
Coverage 1961–2025
Why it’s useful
A newer curve that fits observed bond prices more closely than parametric methods, especially at short maturities and in turbulent periods. It is increasingly used in place of older curves for bond-return predictability and term-structure estimation.
What’s in it
Non-parametric zero-coupon Treasury yields across the maturity spectrum
Splits Treasury yields into expected future short rates and a term premium, the compensation for holding duration risk. Use it to study what drives long rates, how monetary policy and quantitative easing work, and how bond risk premia relate to equity risk premia.
What’s in it
Treasury term premia, risk-neutral yields and fitted yields at maturities from one to ten years
Federal Reserve Bank of Atlanta (Wu and Xia, Journal of Money, Credit and Banking, 2016)
Free
Frequency Monthly
Coverage Covers the zero-lower-bound periods
Why it’s useful
When policy rates sit at zero, the fed funds rate stops measuring the stance of monetary policy. The shadow rate fills the gap, so it is the standard way to keep a monetary-policy variable in a VAR or predictive regression through 2009–2015 and 2020–2022.
What’s in it
A shadow policy rate that can go below zero when the actual rate is stuck at the lower bound
Forecasts, surveys and real-time data
What forecasters and households expected, and what the data looked like at the time: essential for honest out-of-sample work.
Macro data are revised for years after first release. This dataset records each vintage, so you can forecast using only what was actually known at the time. It is the reference source for real-time forecasting and for testing whether a macro predictor really helped investors.
What’s in it
Output, consumption, investment, prices, employment and money as they were published on each date
Coverage US from 1968: the oldest quarterly survey of US macro forecasts
Why it’s useful
A record of what professional forecasters expected. It is the benchmark that model forecasts should beat, a measure of expectations for testing rational-expectations models, and a direct measure of disagreement, which links to the disagreement explanation for why predictability changes over time.
What’s in it
Individual and median forecasts for output, inflation, unemployment and interest rates
Probability forecasts, and forecast dispersion as a measure of disagreement
A real-time, high-frequency measure of the business cycle. It is a natural state variable for return-predictability studies that want something faster and less revised than NBER dates.
What’s in it
A daily index of real business conditions built from mixed-frequency data, with zero as average conditions
Coverage US; monthly since 1978, quarterly from the 1950s
Why it’s useful
The standard measure of household sentiment and inflation expectations. In finance it is used as a sentiment proxy and as a household-expectations series, for example to compare how households and markets price inflation risk.
What’s in it
The Index of Consumer Sentiment and its components
Household inflation expectations one year and five years ahead
Macro and long-run data
Macroeconomic series, real-time vintages and long histories for the US and the UK.
Coverage Hundreds of thousands of US and international series; ALFRED keeps past vintages
Why it’s useful
The quickest route to almost any macroeconomic or interest-rate series, with consistent identifiers and an API. For forecasting work, ALFRED matters as much as FRED: macro data are revised after publication, so using today’s revised values in an out-of-sample exercise builds in look-ahead bias. ALFRED lets you use the data as investors actually saw them.
What’s in it
Interest rates and yields, inflation, output, employment, money, credit spreads, exchange rates and more
ALFRED: the archive of past releases (vintages) of revised series
An API, used by the Python, R and Stata readers listed below
Michael McCracken and Serena Ng (Journal of Business & Economic Statistics, 2016)
Free
Frequency Monthly (FRED-MD) and quarterly (FRED-QD)
Coverage FRED-MD: around 125 US macro series from 1959; FRED-QD: around 250 quarterly series; monthly vintages
Why it’s useful
The standard “big data” macro panel. It is designed for factor models, diffusion indexes and machine-learning forecasts, and it replicates the predictive content of the older Stock–Watson datasets while being updated every month. In finance, factors extracted from this kind of panel are a well-known source of return and risk-premium predictability. It is also the obvious “macro kitchen sink” to control for.
What’s in it
A balanced panel of macroeconomic and financial series covering output, labour, housing, prices, money, credit, interest rates and stock prices
Transformation codes for making each series stationary
The longest consistent monthly US stock market series, and the home of CAPE. Essential for long-horizon predictability, valuation and bubble questions, and for any sample that needs to start before CRSP. It is also one of the sources behind the early years of the Welch–Goyal data.
What’s in it
S&P composite prices, dividends and earnings
Interest rates and the CPI
The cyclically adjusted price–earnings ratio (CAPE)
Frequency Mostly annual, with some monthly and quarterly series
Coverage UK back to the 13th century in places; version 3.1 runs to 2016
Why it’s useful
The richest long-run source for the UK, and the natural starting point for UK return-predictability or macro-finance work over long samples. It lets you test whether a US finding holds in a market with a very different history, using one consistent spreadsheet.
What’s in it
GDP, prices, wages, interest rates, Bank Rate, gilt yields, equity prices, money and credit
Other research datasets on the same page, including the Bank’s balance sheet since 1696
The standard proxy for global demand in oil-market and commodity research. It separates demand-driven from supply-driven oil price shocks, which matters because the two have very different effects on stock returns.
What’s in it
A business-cycle index of global industrial commodity demand, built from dry-bulk shipping rates
Kothari and O’Doherty (2023), updated by Sam Pybis
Updated here
Free
Frequency Monthly
Coverage US from January 1951
Why it’s useful
Job openings as a share of employment: the labour-market predictor of Kothari and O’Doherty (2023), who find that a high ratio is followed by lower stock market returns. Their series stops with their sample. This one is extended every month from JOLTS and the Current Population Survey, joined to Barnichon’s help-wanted index before 2001.
What’s in it
The monthly ratio, and a version using only data public at each month end
The job openings and employment series behind it
Python and MATLAB code that rebuilds it from FRED
Academics’ data pages
Datasets built and maintained by researchers, many of them the standard references in their area.
Coverage US and global by industry and region; updated each January
Why it’s useful
The most widely used free source for valuation inputs: implied equity premia, country risk premia and industry costs of capital. In research it is a handy forward-looking equity-premium series to compare with return-based estimates, and the long annual return history is ideal for teaching.
What’s in it
Implied equity risk premium for the S&P 500, and country risk premiums
Historical returns on US stocks, bonds and bills since 1928
Industry betas, costs of capital, multiples, margins, capital structure and dividend policy
Coverage US IPOs mainly 1975–2025; international underpricing for 55 countries
Why it’s useful
The reference source on IPOs: hot and cold markets, underpricing, long-run IPO performance and the shrinking number of listed firms. IPO activity is also one of the inputs to the Baker–Wurgler sentiment index.
What’s in it
IPO counts, first-day returns (underpricing) and gross proceeds
Founding dates, underwriter rankings, VC-backed and dual-class IPOs
cay, the deviation of consumption from its long-run relation with wealth and labour income, is one of the best-known predictors of stock returns at business-cycle horizons. It also appears in the Welch–Goyal data, so it is a natural benchmark for quarterly predictors.
What’s in it
The consumption–wealth ratio cay, with earlier vintages for replication
Currency risk-premia data (Lettau, Maggiori and Weber)
Coverage 18 advanced economies from 1870 (release 6)
Why it’s useful
The standard long-run, cross-country dataset for macro-finance. It underpins research on credit booms and financial crises, and its asset-return data give a much longer and broader view of the equity premium than US data alone.
What’s in it
48 real and nominal series, including bank credit, mortgage lending and bank capital
Long-run total returns on equities, housing, government bonds and bills (“the rate of return on everything”)
Carmen Reinhart and co-authors (Harvard Business School, Behavioral Finance & Financial Stability)
Free
Frequency Annual
Coverage More than 70 countries from 1800
Why it’s useful
The reference chronology of financial crises, from This Time Is Different. Use it to define crisis periods, to study how returns and predictability behave around crises, and as a long-run complement to the Jordà–Schularick–Taylor data.
What’s in it
Dates of banking, currency, inflation and stock market crises and sovereign defaults, country by country
Tim Loughran and Bill McDonald, University of Notre Dame
Free for academic, non-commercial use
Frequency Updated periodically (latest March 2026)
Coverage US company filings
Why it’s useful
The standard toolkit for textual analysis in finance. General-purpose dictionaries misclassify finance words (“liability” or “tax” are not negative in a 10-K), and these word lists were built to fix that. Combine them with EDGAR filings to measure the tone, uncertainty or readability of disclosures.
What’s in it
The Loughran–McDonald Master Dictionary with finance-specific sentiment word lists
Parsed 10-K and 10-Q text from EDGAR, and textual-analysis code
Official statistics and international data
Central banks, statistical offices and international organisations: the primary sources behind most macro series.
Coverage UK interest rates, yield curves, money, lending and exchange rates
Why it’s useful
The primary source for UK monetary and financial series. It is the place to get the UK risk-free rate, gilt yields and term spreads when you rebuild Welch–Goyal-style predictors for the UK market.
What’s in it
Bank Rate and market interest rates
Nominal and real yield curves
Money and credit aggregates, effective exchange rates
Coverage UK national accounts, prices, labour market, trade
Why it’s useful
The official UK macro source, and the counterpart to the US agencies (BEA, BLS) behind FRED. Use it for UK inflation, output and labour data, and to extend the Bank of England’s long-run dataset beyond 2016.
What’s in it
GDP, CPI and CPIH inflation, employment and earnings, trade and public finances
European Central Bank (successor to the Statistical Data Warehouse)
Free
Frequency Daily to annual
Coverage Euro area monetary, financial and macro statistics
Why it’s useful
The source for euro-area financial conditions. It is useful for international versions of return-predictability tests, and its stress indicators work as European counterparts to US uncertainty measures.
What’s in it
Euro area yield curves, policy rates, bank lending, exchange rates and the CISS financial-stress indicator
The best cross-country source for credit and property cycles. Credit-to-GDP gaps and house prices are standard macro-financial state variables and early-warning indicators of crises.
What’s in it
Credit to the non-financial sector and credit-to-GDP gaps
Residential property prices, central bank policy rates, effective exchange rates
Coverage Global: International Financial Statistics, World Economic Outlook, balance of payments
Why it’s useful
The broadest country coverage for macro and financial series, including emerging and frontier markets that other sources miss. The WEO forecasts are also useful as a record of expectations.
What’s in it
International Financial Statistics (IFS), WEO forecasts, balance-of-payments and reserves data
Comparable series for advanced economies, including the composite leading indicators, which are handy business-cycle state variables outside the US where NBER dates do not apply.
What’s in it
Main Economic Indicators, composite leading indicators, interest rates, share prices, national accounts
Coverage US national accounts, personal income and spending, trade, industry accounts
Why it’s useful
The original publisher of US GDP and consumption, the series behind consumption-based asset pricing (consumption growth, cay) and most macro forecasting. Go here for the detailed tables, methodology and release schedule that FRED does not carry.
What’s in it
GDP and its components, personal consumption expenditures (PCE) and the PCE price index
Personal income and saving, international transactions, GDP by industry
Coverage US prices, employment, wages and productivity
Why it’s useful
The source of US inflation and labour-market data. Its release calendars matter for announcement studies: CPI and payrolls are the macro releases that move markets most, so getting their release times right is essential when studying news effects on returns and volatility.
What’s in it
CPI and PPI, payroll employment and the unemployment rate
Wages, productivity and consumer expenditure surveys
Coverage US and world energy prices, production and inventories
Why it’s useful
The main free source for energy-market data: oil prices and inventories for commodity, inflation and oil-shock research, and the weekly inventory releases that commodity futures react to.
What’s in it
Crude oil, natural gas and electricity prices
Weekly petroleum inventories, production and consumption
The source for household wealth (an input to cay), aggregate equity issuance and leverage. These are the building blocks for consumption-based and flow-based asset-pricing research.
What’s in it
Flows and balance sheets by sector: households, firms, governments and financial institutions
Household net worth, corporate equity outstanding and debt
Coverage Global stocks, indices, ETFs, FX and crypto
Why it’s useful
Fast, free price data for teaching, coursework and first looks at a question. For research, prefer CRSP or Datastream: free sources have survivorship bias (delisted firms disappear), and their adjusted prices can change without notice.
What’s in it
Historical prices and adjusted closes, dividends and splits
Coverage Financial, economic and alternative datasets from many publishers
Why it’s useful
A convenient single API for a mix of free and commercial series, including futures and alternative data. Useful for prototyping, though for core research series the original sources above are usually better documented.
What’s in it
A catalogue of free and premium datasets with a common API and Python/R packages
Coverage Long-running collection of US and international economic time series
Why it’s useful
One of the oldest economic time-series sites, and still useful for quickly finding and charting an obscure series. For research, trace each series back to its original publisher (often via FRED) for definitions and revisions.
What’s in it
Tens of thousands of economic time series with quick charts and downloads
Coverage All US public-company filings since the 1990s
Why it’s useful
The primary source for company disclosures and the raw material for textual analysis in finance: the tone of 10-Ks, risk-factor disclosures, 8-K event timing and 13F holdings. Pair it with the Loughran–McDonald word lists below.
Continues the Florida–UCLA–LoPucki Bankruptcy Research Database
Updated here
Free
Frequency Monthly
Coverage US from April 1980
Why it’s useful
The LoPucki database was the standard source on bankruptcies of large US public companies, but it stopped adding cases after 2022. This series continues its monthly count from SEC 8-K filings, using the database’s own size threshold, with every new case listed and read by hand before it is counted.
What’s in it
Monthly counts of Chapter 11 and Chapter 7 petitions
The list of cases since 2023, with links to each filing
Python and MATLAB code, including the search of SEC filings
Replication archives
Where published papers' data and code live: the fastest way to start from a known result.
Queen’s University; newer files on the ZBW Journal Data Archive
Free
Coverage Data for JAE papers from the late 1980s onwards
Why it’s useful
One of the oldest and most complete replication archives in econometrics. Replicating a published result is the best way to learn a method, and the JAE’s replication section publishes exactly that kind of paper.
What’s in it
Data and often code for every paper in the Journal of Applied Econometrics, including its replication section
Coverage NBER public-use datasets, including the US business-cycle chronology
Why it’s useful
The NBER chronology is the standard definition of recessions, used everywhere to split samples into good and bad times and to show that predictability is concentrated in recessions.
What’s in it
US business-cycle peak and trough dates
Public-use data and links from NBER research programmes
Coverage Data and code for Watson’s papers, many with James Stock
Why it’s useful
A goldmine for anyone forecasting with many predictors: the data and code behind the Stock–Watson diffusion-index, dynamic-factor and forecast-combination papers, which underpin modern macro forecasting.
What’s in it
Replication files for macro-forecasting, factor-model, VAR and time-series econometrics papers
The best place to learn applied causal methods (IV, regression discontinuity, difference-in-differences) on the original data, from the author of Mostly Harmless Econometrics.
What’s in it
Replication files for instrumental-variables, natural-experiment and labour-economics studies
The standard gateway to the commercial databases most finance papers rely on. Its main advantages are consistent identifiers across databases (the CRSP–Compustat link, for example), programmatic access for reproducible data steps, and a server that can handle intraday data.
What’s in it
One platform for CRSP, Compustat, I/B/E/S, OptionMetrics, TAQ and many more
Web queries, plus SAS, Python and R access and a cloud server for large jobs
Center for Research in Security Prices (Morningstar), via WRDS
Subscription (via WRDS)
Frequency Daily and monthly
Coverage NYSE from the 1920s; AMEX and NASDAQ added later
Why it’s useful
The definitive source for US stock returns. It is survivorship-bias free, it is what the Kenneth French factors and most published portfolios are built from, and you need it whenever you construct your own portfolios, signals or firm-level tests.
What’s in it
Prices, returns with and without dividends, volume, shares outstanding, delisting returns and corporate actions
Market indices, and the CRSP–Compustat Merged link
Frequency Annual and quarterly (some monthly files)
Coverage North America and Global
Why it’s useful
The accounting side of empirical finance: book equity, profitability, investment, accruals and every other fundamental-based signal. Its short-interest data underlie the aggregate short-interest index of Rapach, Ringgenberg and Zhou (2016), one of the strongest known predictors of the equity premium.
What’s in it
Balance-sheet, income-statement and cash-flow data
Frequency Monthly summaries and individual forecasts
Coverage US and international
Why it’s useful
The standard source for expectations: earnings surprises, revisions, and forecast dispersion as a measure of disagreement among informed investors, which links directly to the disagreement explanation for why predictability changes across regimes.
What’s in it
Analyst earnings forecasts, recommendations and price targets
The source for forward-looking risk measures: implied volatility, skew and the variance risk premium at index and firm level. These are natural complements, and competitors, to uncertainty indexes in both return and volatility forecasting.
Needed for anything intraday: realised volatility from high-frequency returns, liquidity and microstructure measures, and the reaction of prices and volatility to scheduled macroeconomic announcements within the trading day.
Coverage Global equities, indices, bonds, FX, commodities and macro data
Why it’s useful
The workhorse for international and UK data: long histories for non-US stock markets, sector indices and bond yields that are not in CRSP. If a project compares markets or focuses on the UK, this is usually the most practical source. Many business schools also have Bloomberg terminals, which cover similar ground with better real-time and fixed-income detail.
What’s in it
International stock and index prices with long histories
Government and corporate bond yields, exchange rates and economic series
Subscription (usually library or trading-room terminals)
Frequency Real time to annual
Coverage Global markets, fixed income, derivatives, news and economics
Why it’s useful
The deepest single source for fixed income, derivatives and real-time market data, and for exact announcement times through its economic calendar. For many bond and international questions it is the only practical source.
What’s in it
Prices, yields and curves for almost every traded instrument
Company fundamentals, analyst estimates, news and economic calendars
The quickest way to get macro and interest-rate series into a Stata do-file reproducibly.
What’s in it
import fred GS10 CPIAUCSL downloads series straight into Stata, including vintages
No datasets match. Try a broader search.
Moved or retired sources
Older data-link lists often point to these. Here is where they went.
Florida–UCLA–LoPucki Bankruptcy Research Database: Stopped adding cases after 2022 (final release January 2023). Its monthly count of large public-company bankruptcies is continued here from SEC filings.
EconData.uk: No longer online. ONS, Nomis and the Bank of England Database cover the same UK series.
Quandl: Now Nasdaq Data Link; many of the old free Quandl databases have been retired.
ECB Statistical Data Warehouse: Replaced by the ECB Data Portal.
CBRT Electronic Data Delivery System (EDDS): Now the CBRT’s EVDS data portal.
Penn World Table (old University of Pennsylvania site): Now maintained by the Groningen Growth and Development Centre.
Journal of Statistics Education: Renamed the Journal of Statistics and Data Science Education; the data archive is still online.
Details checked against each provider’s site in October 2026. Data sources change: if a link breaks or a series moves, please let me know.