We present a unified framework to solve the endogeneity problem in time-series regression models with serially a correlated disturbance term with ARMA($p$,$q$) dynamics. Our focus is on the case in which lagged regressors are used as instrumental variables. The control function approach provides us with the solution to the problem. Besides, it provides us with an easy method for testing for endogeneity. Our Monte Carlo experiments confirm that the proposed two-step estimation procedure and the proposed test of endogeneity perform well for a sample size as small as 250.