In this paper it is demonstrated by simulation that, contrary to a widely held belief, pure seasonal mean shifts - i.e., seasonal structural breaks which affect only the deterministic seasonal cycle - really do matter for Dickey-Fuller long-run unit root tests.

The main purpose of this work is to provide a panoramic and almost exhaustive view on the "rational expectations hypothesis" literature published until 1992. Several issues less well known are emphasized. Many results and comments are integrated and presented in a somewhat innovative fashion.