Two Answers, One Belief What Household Inflation Expectations Measure

The New York Fed Survey of Consumer Expectations asks households for both a point forecast of inflation and a probability distribution over the same horizon. The two responses differ by 3.46 percentage points on average. The analysis shows that much of this gap reflects the artifact of survey design rather than differences in beliefs. The reported standard deviation is mechanically generated by which bins a household ticks for about 42% of respondent-months, and restricting the sample to densities inside the interior bins cuts the average divergence by 89%. What remains is a reporting error rather than a preference, which is two-sided, and its persistence decays to half in about three months. Between 65 and 67 percent of the crosssectional dispersion in reported point forecasts is measurement error rather than disagreement, and the implied degree of information rigidity is 0.54 using the density means against 0.38 using the point forecasts.