A Note on the Practice of Lagging Variables to Avoid Simultaneity

A common practice in applied econometrics consists of replacing a suspected endogenous variable with its lagged values. This note demonstrates that lagging an endogen¬ous variable does not enable one to escape simultaneity bias. The associated estimates are still inconsistent, and hypothesis testing is invalid. I show that the only time a strategy of replacing Xt with Xt-1 enables consistent estimation of structural parameters is when there is no simultaneity to begin with. The implication of this study is that researchers who employ this practice should be explicit about why lagging is an effective estimation strategy.