COMPARISON OF DIRECT, MIXED MODEL, AND BAYESIAN METROPOLITAN STATISTICAL AREA ESTIMATES FOR THE INSURANCE COMPONENT OF THE MEDICAL EXPENDITURE PANEL SURVEY (MEPS)

The Medical Expenditure Panel Survey is a family of sample surveys. The Insurance Component of MEPS provides national and state level estimates of insurance offered and provided by employers in the U.S. In recent years the demand for reliable data at the state level and below, regarding healthcare insurance has greatly increased. Previous research has been conducted to produce direct design-based estimates using the MEPS Insurance Component design structure. However, the number of Metropolitan Statistical Areas (MSAs) for which direct estimates can be produced with acceptable reliability is limited. In this paper, we evaluate mixed models and Bayesian models, incorporating a time covariate, to produce MSA level estimates for smaller MSAs in ten states. We examine estimates of MSE and RSE of two types of estimates based on direct and indirect estimation techniques.