Published online by Cambridge University Press: 26 February 2016
The rationality test is based on regressing the actual series of variable in period t on its forecast formed in period t − 1. The nonrejection of the rationality hypothesis requires expectations' unbiasedness and efficiency. We show that in a model with t − 1 dating, where expectations strongly and positively affect the economy, the rationality test suffers from low power. It provides a high probability of nonrejection of the rationality hypothesis against numerous alternatives for expectations formation. This result is attained because the realization of the economy is driven by the public's expectations, irrespective of how well they are formed, via the structural relationship between them. The parameters in this test are predetermined by the parameters of the structural model, supporting expectations' unbiasedness under reasonable assumptions. Thus, successfully passing the rationality test could be misleading in terms of interpretation of the quality of the expectations and could lead to questionable accuracy of many applications.
I wish to thank two anonymous referees and the associate editor for their very helpful comments and suggestions in substantially improving the paper. I also wish to thank Thomas Laubach, Guy Segal and Jonathan Benchimol for their helpful comments and suggestions.