Abstract:
|
In the Small Area Income and Poverty Estimates program at the US Census Bureau, the parameters in the production models are assumed to be constant for all areas (states or counties) across the nation. In the county model, there are over 3100 counties which differ from each other is significant ways (population size, geographical size, demographic composition, etc) that may not be captured by the covariates used in the production model: Federal Tax data, government assistance programs participation rates (e.g. foodstamps) and population estimates. Preliminary analysis of splitting the counties into three groups based on population size suggested that the small area model parameters may vary across areas. Borrowing ideas from the geographically weighted regression (GWR) literature, we propose a method to evaluate the assumption that the small area model parameters are constant across all of the areas.
|