![IconGems-Print](images/IconGems-Print.png)
282 – New Developments in Small Area Estimation Research at the U.S. Census Bureau
Multilevel Regression and Poststratification (MRP) for Small Area Estimation: An Application to Estimate Health Insurance Coverage Using Geocoded American Community Survey
Xingyou Zhang
Economic Research Service, U.S. Department of Agriculture
Sam Szelepka
U.S. Census Bureau
Blandine Bawawana
U.S. Census Bureau
Alfred O. Gottschalck
U.S. Census Bureau
Knowledge of the geographic distributions of population socioeconomic and health outcomes is critical for public social and health policy deliberation, formulation, delivery and program planning and evaluation. More granular or local population socioeconomic and health data are often needed but usually do not exist. A variety of small area estimation techniques have been developed to address this significant data gap. Sociodemographic and health surveys have become routinely geocoded in federal statistical agencies, which means that we could have both individual characteristics of survey respondents from the survey itself but also their geographic context that might have great influence on their individual social, economic and health behaviors. Thus, we are developing and validating an innovative multilevel regression and poststratification (MRP) approach that applies multilevel regression models to geocoded surveys; takes account for both individual characteristics and area level factors at multiple geographic levels; predicts individual-level social, economic and health outcomes in a multilevel modeling framework; and estimates the geographic distributions of population socioeconomic and health outcomes. We applied this innovative multilevel approach for small area estimation using geocoded American Community Survey (ACS) data. We demonstrate that MRP provides a flexible statistical linkage and modeling platform that makes full use of geocoded ACS data and available geodemographic data to generate small area estimates of percentages of the population without health insurance coverage. We will also compare our model-based health insurance coverages with those based on the current SAHIE model and direct ACS survey estimates.