Small Area Estimation for the Medical Expenditure Panel Survey - Insurance Component (MEPS - IC)

• Activity Number: 386
• Type: Contributed
• Date/Time: Thursday, August 15, 2002 : 8:30 AM to 10:20 AM
• Sponsor: Section on Survey Research Methods*
• Author(s): Steven Riesz*+ and John Sommers
• Affiliation(s): U.S. Census Bureau and Agency for Healthcare Research and Quality
• Keywords: small area estimation ; health insurance ; model-assisted estimation ; logistic regression

Abstract

The MEPS-IC is an establishment survey that collects data about employer-sponsored health insurance. Because states are important in the health insurance market, the IC sample was designed to produce good state-level estimates. In addition, there is great demand for sub-state estimates, but the enhanced state-level samples cannot produce good estimates of sub-state groups, such as an industry within a state. To get good small area estimates, we extend the work of Walkup and Sommers (2001), who developed model-assisted estimates by predicting values for all of the establishments on the frame. We develop model-assisted estimates and composite estimates and evaluate them in terms of their bias and mean squared error. The variances of the estimates are estimated by the method of balanced half-samples. Both the model-assisted estimates and the composite estimates have smaller mean squared errors than the standard estimates. Adding weights to one of the models reduces the bias of one of the estimates.