State-Level Opinions from National Surveys: Poststratification Using Multilevel Logistic Regression

Editors:
  • DPDavid K. Park
    David K. Park
  • AGAndrew Gelman
    Andrew Gelman
  • JBJoseph Bafumi
    Joseph Bafumi

Abstract
This chapter focuses on the analysis of state-level opinions gathered from national surveys in the U.S. It criticizes the Erikson-Wright-McIver methodology for its insensitivity to short-term change in opinion and proposes the use of simulation combined with multilevel logistic regression to estimate state opinion. The evaluation of the proposed method reveals that it produces quite precise estimates of different examples of public opinion and suggests that it has opened up a new avenue for capturing dynamic properties in estimating state-level public opinion.