Research on Improving the Job Opening and Labor Turnover Survey's Outlier Detection Procedures Using Winsorization Treatment

摘要

The U.S. Bureau of Labor Statistics embarked on the Job Openings and Labor Turnover Survey (JOLTS) in 1999. The JOLTS collects total employment, job openings, hires, quits, layoffs and discharges, other separations, and total separations data. The JOLTS initially used the interquartile method to detect outliers for each characteristic. The list of schedules with potential outliers was reviewed by analysts who determined which schedules were to be treated as atypicals because the value of one or more characteristics was not considered to be representative of the population. There are several statistical issues associated with the initial method. In this paper, we discuss a new procedure to addresses these issues. The new approach uses the Winsorization method.