Fixed Effects, giving every firm its own line
Fixed effects gives each firm its own intercept and removes any time-invariant trait, staying consistent even when that trait correlates with the regressors, but cannot estimate constant variables.
A 5.5 minute lesson on fixed effects, the within estimator. Built for ECON3006 and FIN301 students at Western Sydney University, worked in Stata.
The problem is one line drawn through three firms. A pooled fit ignores who each firm is and lets persistent firm traits leak into the slope. Fixed effects gives each firm its own intercept by subtracting its own average over time, which removes any time-invariant trait entirely, so the estimate stays consistent even when that trait is correlated with the regressors.
The catch is the flip side. Because demeaning sweeps out everything constant over time, fixed effects cannot estimate the effect of a time-invariant variable like gender or industry, and it still needs strict exogeneity to handle time-varying confounders. Add time effects for common shocks and cluster the standard errors by firm. Pair the video with the Atlas concept page for the within transformation, the Stata commands, a quiz, and citations.