UNEMPLOYMENT BENEFITS AND LABOR-MARKET TRANSITIONS - A MULTINOMIAL LOGIT MODEL WITH ERRORS IN CLASSIFICATION

This paper utilizes validation data on survey response error in the Current Population Survey to generalize the standard multinomial logit model to allow for spurious events that result from classification error. Our basic approach could be used with other stochastic models of discrete events as well. We illustrate our algorithm by studying the effect of unemployment insurance (UI) on transitions from unemployment to employment and on labor force withdrawal. Our results confirm earlier work suggesting that UI lengthens unemployment spells, and show that correcting for classification error strengthens the apparent effect of UI on spell durations.