Small Area Modeling Research for the Occupational Employment Survey

Timothy Li


The Occupational Employment Survey (OES) is a yearly mail survey designed to produce estimates of employment and wages for more than 700 occupations. The OES sample contains approximately 400,000 establishments per year, accumulating a rolling sample of 1.2 million establishments every three years. The OES sample is stratified by geographic area, economic activity, and employment size class—with geographic area defined by State, Metropolitan Statistical Area (MSA) and balance of State area, and economic activity defined by three-digit Standard Industrial Classification (SIC) codes. While the sample is designed to produce reliable design-based estimates for large geographic areas, our research investigates synthetic and composite estimators for smaller geographic areas. We examine some standard small area models, as well as examine new methods which draw upon some of the unique features of the OES survey. We also examine a new method of variance estimation, which modifies the current Jackknife Random Groups approach.