A panel of banks may show that highly capitalised banks lend more, but that raw relationship can reflect durable differences in business model, regulation, or management quality. Panel estimators decide what to do with that persistent heterogeneity. Pooled OLS ignores it, fixed effects removes it, and random effects models it as a component of the error. The choice controls both what variation identifies the coefficient and whether it is consistent.
Learning objectives. By the end of this note you should be able to:
- state the pooled, fixed-effects, and random-effects specifications;
- derive the within transformation for fixed effects;
- explain why FE cannot estimate a coefficient on a time-invariant regressor;
- describe RE quasi-demeaning and its GLS rationale;
- choose an initial specification from the economic data-generating story.
In this lesson
- One model, three treatments of heterogeneity
- Random effects uses a stronger assumption
- Inference remains a separate choice
Before you start
The full lesson, with worked examples and exercise solutions, is for members.
Read the full lessonNew to SelfMadeQuant.com? See membership options or browse the Econometrics and Financial Time Series syllabus.