Coordination and the Relative Cost of Distinguishing Nearby States

Rational Inattention and Sequential Information Sampling

“…But this result depends on the fact that the mutual-information cost function does not make it costly to have abrupt changes in signal probabilities as the state of the world changes continuously. Morris and Yang (2016) develop the complementary result, showing that even in the case of an endogenous information structure, if signal probabilities must vary continuously with the state, there is again a unique equilibrium.…”

“…With this alternative neighborhood structure, Lemma 6 no longer requires that the response frequencies be identical for any two states. Moreover, because the cost function The continuity of choice probabilities across points at which there are discrete changes in payoffs is also an important issue for the global games literature (Morris and Yang (2016)). However, this literature typically assumes a continuum of states, and many of the perceptual experiments that we have just referred to are naturally modeled with a continuum of states as well.…”

“…In this subsection, we continue our discussion of the neighborhood-based cost function proposed in the previous subsection, and consider the limit as the number of states of the world, |X|, becomes infinite. This example is motivated by the work of Yang (2015) and Morris and Yang (2016), who study global games (e.g. Morris and Shin (2001)) with endogenous information acquisition.…”

“…proposal, this proposal yields a highly parsimonious theory of rational inattention with only one free parameter, indexing the degree of overall scarcity of attentional resources, but one that we believe may be preferable to Sims' proposal for many applications, such as the problem of endogenous information in global games discussed by Morris and Yang (2016).…”

“…Shannon's mutual information has the first two of these features, but not the third; we will provide an example, motivated by perceptual experiments, in which rationally inattentive DMs' behavior is consistent with the neighborhood-based cost function, but not the mutual-information cost function. We also argue that this distinc-tion has consequences in economic applications, focusing on the global games example of Morris and Yang (2016). We provide an example neighborhood-based cost function, the expected Fisher information, which, like mutual information, has only a single parameter, but unlike mutual information, captures the notion that similar states are difficult to distinguish.…”