Learning Under Multiple Information Sets (JMP)

Abstract: This paper analyses how individuals form inflation expectations under information uncertainty. Using evidence from a period with multiple public inflation statistics and an increase in relative price dispersion (Argentina, 2007-2011), I show that individuals use their own inflation experience- based on the change in prices of the goods they purchase- to form inflation expectations. More specifically, the inflation expectations of lower-income households, who experienced higher average household-specific rates of inflation, were higher. I characterize inflation experience by using daily (online) data on prices and expenditure information for roughly 25,000 households to construct household-specific price indexes. To disentangle the effect of information uncertainty from price dispersion, I model expectations through a Bayesian learner who knows signals can be noisy but also biased. Results suggest that, even in situations with a unique inflation statistic, inflation experience may affect economic decisions and outcomes if there are doubts about the quality of public information.

Underreaction of Expectations After Large Depreciations (new draft soon!)

Information Frictions: The Biased-Channel (work in progress)

Abstract: How do inflation expectations react to large depreciations? I document an underreaction of consensus inflation expectations following large depreciations. Underreaction is increasing on the average size of the nominal exchange rate depreciation and, surprisingly, is more important in emerging rather than in advanced economies, both in the short and medium horizon. Results hold after controlling for other well know predictors of inflation forecast errors and after all new available information is considered. The evidence suggests that recent support for rational models of information rigidities might not necessarily be extrapolated to emerging economies and that deviations from full-information rational expectations may occur beyond the bounds of baseline models of sticky and noisy information.