Doctoral Dissertation Research: Estimating Congressional District-Level Opinions from National Surveys using a Bayesian Hierarchical Logistic Regression Model

Columbia University, New York NY

Investigators

Abstract

This dissertation research project constructs a hierarchical logistic regression model for the mean of a binary response variable conditional on poststratification cell. The object is to produce congressional district-level opinions from national surveys. This approach combines the modeling approach often used in small-area estimation with the population information used in poststratification. The NSF dissertation research grant will be used to purchase data from the US Census Bureau to obtain reliable estimates of the joint population distributions of sex, ethnicity, age, and education at the congressional district-level. In order to poststratify on all the variables listed above, along with the congressional district, we need the joint population distribution of the demographic variables within each congressional district. Currently, the U.S. Census only provides without charge the joint distributions of sex, ethnicity, and age. The model has produced reliable measures that the state-level.