Weakly rational expectations (Q393281)

This paper presents a weakening of \textit{R. Aumann} and \textit{J. Drèze}'s context for the characterization of rational expectations in a game [``Rational expections in games'', Am. Econ. Rev. 98, No. 1, 72--86 (2008; \url{doi:10.1257/aer.98.1.72})]. While Aumann and Drèze assumed rational belief systems (in which there is common knowledge of rationality and common priors over the types space) to determine the interim expected values, the author only requires \textit{P. Barelli}'s action-consistent distributions [Games Econ. Behav. 67, No. 2, 363--375 (2009; Zbl 1180.91019)]. Rationality is now only necessary in the support of these distributions. The resulting belief system is deemed weakly rational. Theorem 1 in the paper claims that under such weakly rational belief system the same interim values are obtained as with Aumann and Drèze's rational belief systems. That is, the set of rational expectations under those two belief systems is the same. A corollary indicates, as in [Aumann and Drèze, loc. cit.], that in any zero-sum game, now under a weakly rational belief system, the value is the rational expectation.