Selected tools, data sources, and resources for empirical finance research.
Documentation for statistical software, data management, econometrics, graphics, and reproducible workflows.
Software for applied econometrics, panel data, causal inference, and reproducible analysis.
Panel and longitudinal data — fixed-effects, random-effects, dynamic-panel, IV, and DiD features.
reghdfe — linear models with multiple high-dimensional fixed effects.
ivreghdfe — instrumental-variable estimation with absorbed fixed effects.
coefplot — coefficient and confidence-interval graphics from stored estimates.
Packages for data cleaning, econometrics, prediction, and geospatial analysis.
pandas — tabular data and data cleaning.
NumPy — numerical arrays and scientific computing.
statsmodels — statistical models and hypothesis tests.
scikit-learn — preprocessing, model selection, and machine learning.
GeoPandas — spatial joins, vector data, and map-based analysis.
Packages for data manipulation, econometrics, and statistical graphics.
tidyverse — consistent data-science workflow.
fixest — fixed-effects, IV, and DiD estimation.
data.table — memory-efficient data work.
ggplot2 — layered statistical graphics.
SQL / PostgreSQL — structured-data queries, joins, and management.
LaTeX / Overleaf — manuscripts, tables, equations, and bibliographies.
MATLAB — numerical modeling and matrix-oriented computation.
Common sources for finance and applied economic research. Several require an institutional subscription.
Access to licensed financial, accounting, ownership, governance, and market databases, with tools for querying and linking sources.
Historical security prices, returns, identifiers, and indexes. Common uses include return measurement, event studies, and sample construction.
Standardized company financial statements and market information for accounting measures, firm characteristics, and panel construction.
Executive compensation and leadership records for covered public firms. Common variables include pay, incentives, tenure, and governance measures.
Trades and quotes across U.S. equity markets for research on liquidity, spreads, price discovery, and market microstructure.
Institutional voting records and proxy-voting information used in shareholder voting, governance, and stewardship research.
Federal campaign-finance filings, committee records, and transaction-level contributions for research on political contributions and connections.
Demographic, geographic, business, and economic tables for regional controls, community characteristics, and population measures.
Employment, earnings, hires, separations, and job creation by place, industry, and demographic group.
Licensed business and establishment records, including location and industry fields, for establishment-level and geographic matching.
Private-company, investor, funding-round, and acquisition records for entrepreneurial-finance and venture-capital research. Coverage and licensing should be documented.
Licensed market data, news, analytics, screening, and company information for market analysis, valuation, and classroom demonstrations.
Public options, equities, and futures statistics, including volatility, volume, and market-activity measures.
Macroeconomic and financial time series from the Federal Reserve Bank of St. Louis, including rates, inflation, employment, output, and credit measures.
Free public access to company filings and filing APIs for financial statements, disclosures, governance documents, events, and text analysis.
Data-use note. These descriptions do not imply that licensed data may be redistributed. Researchers should preserve source provenance, access dates, variable definitions, and applicable license restrictions.
Selected empirical designs and references. Credibility depends on the identifying assumptions, not the software command.
Specify the estimation window, event window, benchmark model, and inference procedure. WRDS overview · MacKinlay (1997)
Compare changes across treated and comparison groups. Diagnose treatment timing, parallel-trends assumptions, anticipation, and heterogeneous effects. Callaway & Sant’Anna · did documentation
Build event-specific cohorts and comparison samples before stacking them. State cohort weights, comparison-group rules, and clustering choices. Cengiz et al. · Baker, Larcker & Wang
Use an instrument that shifts treatment while satisfying relevance and exclusion restrictions. Report first-stage strength and interpret the estimand carefully. Stata IV overview · ivreghdfe
Match the fixed effects and standard errors to the research design and source of identifying variation. Stata panel data · fixest
Check support, singleton handling, collinearity, and the appropriate level of clustering. reghdfe · Correia
Separate training and evaluation samples, use economically meaningful loss functions, and compare predictive performance with tests suited to the forecast setting. Diebold & Mariano · Clark & West
Use cross-validation and out-of-sample evaluation. Compare flexible models with simple benchmarks; guard against leakage and overfitting. Gu, Kelly & Xiu · Cross-validation
Document coordinate systems, geocoding quality, spatial joins, boundary vintages, and distance definitions. Treat unmatched and ambiguous locations explicitly. GeoPandas · TIGER/Line
A written workflow helps audit data construction and reproduce results.
Literature search. Google Scholar, SSRN, and backward and forward citation searches.
Research design. Record the question, sample, estimand, timing, assumptions, and planned diagnostics before coding.
Data construction. Separate raw, intermediate, and analysis files; keep data dictionaries and a decision log.
Reproduction check. Run from clean inputs, validate outputs, version code, and archive final dependencies.
Use Google Scholar and SSRN for discovery; manage sources and PDFs with Zotero. Save the search terms and access dates used for a review.
Overleaf supports LaTeX collaboration. Git and GitHub track code and text.
The AEA Data Editor gives guidance on data and code availability. Gentzkow and Shapiro’s practitioner’s guide covers project organization.
Maintain a dated log of data pulls, transformations, modeling decisions, errors, robustness checks, and open questions. Record why a choice changed, not only what changed.
WRDS Video Support — CRSP, Compustat, ExecuComp, TAQ, event studies, linking, and related workflows.
Kenneth French Data Library — public factor returns and research portfolios used in empirical asset pricing.
NBER Data — public datasets and links for economic research.