|
Abstract:
|
The Current Population Survey (CPS) is a two-stage sample that seeks to estimate unemployment level at both state and national level. Primary Sampling Units (PSUs) are created out of counties/groups of county-equivalent entities, some of which are drawn with certainty. For the non-certainty PSUs, these are stratified, and one PSU is selected from each stratum for a 10-year period, with further sampling of households for monthly estimates. Our goal is to minimize within-strata variance over the 10-year period. A set of stratification algorithms will be benchmarked against a set of simulated populations, that will vary population distribution and labor force characteristics across time. A final evaluation will include a population that is derived from American Community Survey estimates to provide a realistic simulation example. Robustness of the stratification strategies will be evaluated based on total error in the simulations.
|
