The U.S. Census Bureau Small Area Income and Poverty Estimates program (SAIPE) currently uses an empirical Bayes estimation method similar to the Fay and Herriot (1979) model to produce biennial intercensal estimates of the poverty rates and counts of poor within counties. The dependent variable is formed from a three-year average of the March Current Population Survey (CPS) supplement, and the independent variables are formed from administrative data. The model includes two error terms. Problems with this estimation technique include a loss of data points due to the log transformation for counties whose CPS sample of poor is zero, and the requirement of using decennial census data to estimate the model error variance term. To address these problems, a hierarchical Bayes model based on a scaled binomial kernel has been developed (see Fisher and Asher (1999)). The scaling factor corrects for both the overdispersion of the variance and the complexity of the CPS sample design. This paper will discuss the effect of different scaling factor functions on the implementation and quality of this proposed model.