Econometrics

A unified framework for causal inference in econometrics. Covers the potential outcomes framework, selection bias, ATE vs ATT vs LATE, randomized controlled trials, regression adjustment, propensity score matching, IPW, doubly robust estimation, and regression discontinuity design — with finance examples and a comparison of when to use each method.

The linear probability model (LPM) applies OLS to a binary dependent variable, interpreting coefficients as percentage-point changes in probability. Learn its advantages, limitations, when to use it vs. logit or probit, and why robust standard errors are mandatory.

Learn how logit and probit models handle binary dependent variables using maximum likelihood estimation. Covers odds ratios, marginal effects, and model comparison with worked loan default examples.

Learn how instrumental variables and two-stage least squares (2SLS) solve endogeneity in econometric models. Covers instrument validity, weak instruments, testing, and the LATE interpretation.

Learn panel data analysis: fixed effects, random effects, the Hausman test, pooled OLS, first differencing, and two-way fixed effects. Covers within transformation, GLS estimation, clustered standard errors, and practical decision rules for applied finance research.

Learn how difference-in-differences isolates causal effects from natural experiments by comparing treatment and control groups before and after a policy change.

Learn about cointegration, unit root tests and error correction models for analyzing long-run equilibrium relationships in finance.

Learn how dummy variables encode qualitative data in regression models, including intercept shifts, multiple categories, interaction terms, the Chow test for structural breaks, and the dummy variable trap.

Learn what omitted variable bias is, how to sign the direction of bias, and how multicollinearity, measurement error, and proxy variables affect regression estimates. Includes the OVB formula, VIF thresholds, and a comparison of solutions.

Learn how serial correlation affects regression analysis — detection with Durbin-Watson and Breusch-Godfrey tests, correction with Newey-West HAC standard errors and Cochrane-Orcutt/Prais-Winsten FGLS.