Effects of Repeated Interviewing on Estimation of Labor Force Status

This paper studies the effects of repeated interviewing (termed interview group bias) on the accuracy of aggregate unemployment rates computed from the Current Population Survey. Studies of rotation group bias at the aggregate level have shown that reports of labor force status vary systematically with the number of times a household has been sampled. This paper presents estimates of the systematic variation using microdata and shows that previous studies have underestimated its magnitude. It is shown that rotation group bias is a special case of interview group bias and that the magnitude of interview group bias is substantial enough that unemployment rates across groups interviewed only once can be more than 50% higher than for groups interviewed four times.