C. M. Kelly (1985) claims that long-run solutions from econometric models may be seriously misleading when expectations variables are erroneously replaced by observed outcomes. It is shown that his results derive uniquely from an invalid exogeneity assumption. All inferences are therefore potentially invalid, illustrated by a case where the long-run is correct while the short-run is biased. Using an encompassing framework, error-variance rankings and related tests distinguishing expectational...