Retrospective Error in SOEP Calendar Data: The Case of Unemployment

This paper compares current and one-year retrospective data on unemployment in the German SOEP. Thirteen percent of all unemployment spells remain unreported one year later, and another 7 percent are misreported. Individuals with weak labour force attachment, e.g., women with children or individuals close to retirement, have the largest propensity to underreport unemployment retrospectively. The data are consistent with evidence on retrospective bias found by cognitive psychologists and survey methodologists.