Beliefs About the Economy are Excessively Sensitive to Household-Level Shocks: Evidence from Linked Survey and Administrative Data

Linking the monthly Danish Consumer Expectations Survey to the Danish administrative registry, this paper shows that households’ inflation forecasts move strongly and negatively with their own recent and expected future income changes — even though those income changes are almost purely idiosyncratic and carry essentially no information about actual inflation. Two formal tests establish that the pattern is inconsistent with rational expectations, including limited-information versions, by a wide margin. The paper then traces the bias to selective recall cued by affect: the same events move backcasts of past inflation even more than forecasts, and so do family emergency-room visits. The magnitudes are the point. A one-log-point increase in recent household income is associated with essentially no change in realised inflation over the following twelve months (0.008, standard error 0.022) but with a 0.674 percentage point lower inflation forecast (standard error 0.140) — a gap an order of magnitude beyond the 0.02 percentage point bound the paper’s weaker test permits under rational expectations. The same pattern holds for survey-reported expected changes in household finances and for realised future income changes measured in tax data. The evidence on mechanism is that inflation backcasts covary more strongly with these events than forecasts do, and that controlling for backcasts significantly attenuates — in one case to statistical insignificance — the relationship between forecasts and the income measures, which the authors read as memory mediating the association. Family emergency-room visits in the survey month, a negative event entirely unrelated to prices and randomly timed relative to survey contact, also raise both backcasts and forecasts, again with a larger effect on backcasts. The claims are associational rather than experimental for the income results — the paper’s identification comes from the formal tests, from placebo income changes further in the past, and from a large battery of subsamples — while the emergency-room result leans on the random assignment of survey month.

Dmitry Taubinsky | Macro Paper Warehouse

Beliefs About the Economy are Excessively Sensitive to Household-Level Shocks: Evidence from Linked Survey and Administrative Data