Characteristics of a Model‐based Variance Measure for X‐11 Seasonal Adjustment

Stuart Scott and M. Sverchkov

Abstract

A model-based method due to Bell and Kramer (1999) and a moments-matching method due to Pfeffermann (1994) provide variance estimates for X-11 seasonally adjusted time series. An undesirable characteristic can arise when sampling error is large, relative to overall variability in the time series. Both empirical results and a detailed look at calculations for the model-based method help explain when this is likely to happen. The empirical results are based on a large-scale study of series from the Bureau of Labor Statistics' Current Employment Statistics program.