Panel Data

Repeated observations on banks or funds can separate changes within an entity from persistent differences between entities. This distinction matters when estimating the association between capital, risk, or active share and performance. This lesson compares pooled OLS, fixed effects, and random effects. It states their exogeneity assumptions, explains the Hausman comparison, and identifies the additional problem created by a lagged outcome in a short panel.

• formulate pooled OLS, FE, and RE;
• derive the within transformation and state what FE cannot estimate;
• state the RE orthogonality assumption and GLS quasi-demeaning idea;
• use Hausman as a check of RE vs FE;
• recognise when short-$T$ dynamic panels call for Arellano–Bond-type methods.

1. One model, three treatments
2. Hausman test
3. Arellano–Bond as a frontier pointer