Variance Estimation by Replication for National CPS Seasonally Adjusted Series

Thomas D. Evans, Justin J. McIllece, and Stephen M. Miller

Abstract

There is much interest in month-to-month changes for Current Population Survey (CPS) seasonally adjusted labor force series at the national level. Much of this interest focuses on producing confidence intervals around the monthly change in the seasonally adjusted national unemployment rate. Unfortunately, variances for those series are not currently available due to the complexity of the X-11 seasonal adjustment process. Many studies are available on how to estimate those variances. A common approach is to utilize sampling error information and a linear approximation to X-11. Less work has been applied to an alternative approach of deriving variances from seasonally adjusted replicate series that also accounts for sampling errors. An adequate time series of consistent CPS replicate weights is now available which allows for an examination of new variance estimates for seasonally adjusted series by the replication approach. A description of our methodology and results are presented for the national employment, unemployment, and unemployment rate series.