ABSTRACT
Autocorrelation (also called serial correlation) is the violation of the assumption that E(ε i ε j ) = 0. When the error in one period is related to the error in another period then OLS is no longer BLUE. Moreover, the R 2 may be overestimated, standard errors underestimated and t-statistics overestimated. If the regressors include a lagged dependent variable then OLS estimates are biased.
