How Long is a Spell of Unemployment?: Illusions and Biases in the Use of CPS Data

Abstract

Most data used to study the durations of unemployment spells come from the Current Population Survey, which is a point-in-time survey and gives an incomplete picture of the underlying duration distribution. We introduce a new sample of completed unemployment spells obtained from panel data and apply CPS sampling and reporting techniques to replicate the type of data used by other researchers. Predicted duration distributions derived from this CPS-like data are then compared to the actual distribution. We conclude that the best inferences that can be made about unemployment durations using CPS-like data are seriously biased.