Testing Near-Rationality Using Detail Survey Data

This paper considers the evidence of “near-rationality†in household inflation expectations, as described by Akerlof, Dickens, and Perry (2000), hereafter ADP. According to ADP, the economic incentive to anticipate inflation varies from agent to agent, and as inflation falls, some agents stop trying to accurately predict inflation (“hyper-rational†) and either underweight it or, in the extreme, ignore it altogether (“nearly rational†). A key implication of the ADP model is that a particular rate of inflation minimizes unemployment in the long-run. In this paper, we bring the idea described by ADP to detailed survey data on household inflation expectations for the U.S. and Sweden, two countries where the existence of ADP-type near-rationality has been identified in earlier research. We find that the survey data do not, in general, support the specific form of near-rationality suggested by ADP. However, we also show that inflation expectations are not distributed across households in a smooth and continuous way. Rather, households appear to hold inflation expectations that tend to discrete, and largely fixed “focal points,†suggesting that a substantial share of both Swedish and U.S. households do not form precise inflation predictions, but instead gauge inflation prospects in rather qualitative terms. We further document that in Sweden, the combination of a low inflation environment, coupled with an inflation target, has been accompanied by a disproportionately high proportion of households reporting the expectation of no inflation, consistent with one type of “nearly rational†behavior posited by ADP. However, a similarly low inflation trend in the U.S., which does not have an explicit inflation target, reveals no such rise in the proportion of households expecting no inflation. This observation suggests that how the central bank communicates its inflation objective may influence inflation expectations independently of the inflation trend they actually pursue.