Abstract
Stuart Scott, George Stamas, Thomas Sullivan, and Paul Chester (1994) "Seasonal Adjustment Of
Hybrid Economic Time Series," Proceedings of the Section on
Survey Research Methods, American Statistical Association,
forthcoming.
Based on a sample of 380,000 employers, state industry
employment is estimated monthly by the U.S. Bureau of Labor
Statistics and seasonally adjusted with X-11-ARIMA. The
estimates are revised annually to reflect employment counts
available from administrative records of the Unemployment
Insurance (UI) programs of each state. Thus, the overall time
series is a hybrid of universe data up to the last benchmark
and sample data afterward. As pointed out by Berger and
Phillips (1993), there appear to be distortions in seasonal
adjustment due to differing seasonal patterns in these data
sources. The universe data dominate the historical data from
which the factors are derived, but these factors then get
applied to sample data. We compare seasonal factors from the
two data sources, and use revision statistics to evaluate
alternative approaches to seasonal adjustment . In addition,
twelve-month change estimates are examined. When such
estimates are across the last benchmark, the unadjusted
estimate contains seasonality.
Last Modified Date: July 19, 2008
|