BAYESIAN SMOOTHING OF RATES IN SMALL GEOGRAPHIC AREAS

An enormous amount of socio-economic and public-health data come as ra tes (e.g., unemployment, per capita income, mortality rates, census un dercount) reported in small geographic areas. The U.S. Census Bureau r egularly publishes data series at the county level although the county is often a small area chosen for administrative convenience rather th an by design. The reported rates can be regarded as a noisy representa tion of the true geographic distribution of rates over the small areas . This article presents a Bayesian statistical method of smoothing raw rates. In order to illustrate the important features of the method, a data set on undercoverage in the 1980 U.S. Census will be used.