An Optimization Approach to Reconciling Sample Allocations

David S. Piccone and Matthew Dey

Abstract

As part of a larger project, a research team at the Bureau of Labor Statistics (BLS) created an alternative sample design for the Occupational Employment Statistics (OES) survey. There are three sample allocations for the new sample design, each geared towards improving the estimator in different ways. There is an efficient allocation that aims to lower the sampling error of the OES estimates, and two minimum allocations that set a lower sample size threshold for area and industry domains. Each of the three sample allocations are stratified designs, however they use different strata definitions. This paper describes how we reconcile the the three allocations using an optimization approach.