Use the data in OKUN.RAW to answer this question; see also Computer Exercise 11.11.
(i) Estimate the equation pcrgdpt = (0 + (1 (unemt + ut and test the errors for AR(1) serial correlation, without assuming [(unemt;. t = 1, 2, ...} is strictly exogenous. What do you conclude?
(ii) Regress the squared residuals, 2t, on kunem (this is the Breusch-Pagan test for heteroskedasticity in the simple regression case). What do you conclude?
(iii) Obtain the heteroskedasticity-robust standard error for the OLS estimate 1. Is it substantially different from the usual OLS standard error?