Classical Assumptions

The five assumptions behind OLS properties: A1 mean-zero errors, A2 exogeneity E[u∣xk]=0E[u\mid x_k]=0, A3 no autocorrelation, A4 homoskedasticity, A5 normality of the errors.

When to use

Map each to what it buys: A1 and A2 give unbiasedness, A1 to A4 give Gauss-Markov Theorem (BLUE), A1 to A5 give normal β^\hat\beta for exact t and F inference. Then ask which assumption a problem breaks: Omitted Variable Bias breaks A2, Heteroskedasticity breaks A4.

See Lecture 4 and Lecture 13.

Appears in