Understanding Public Opinion and Policymaking Using Multilevel Regression and Poststratification

Columbia University, New York NY

Investigators

Abstract

Understanding public opinion is one of the fundamental issues in political science and of key import in any democracy. National surveys are designed to produce good estimates of national public opinion, but not necessarily to give good estimates at the state or congressional district level or by finely-grained demographic sub-groups. Therefore usually knowledge of public opinion at these more local levels is lacking despite the fact that much important policymaking and politics is at these levels. This seriously limits the preference measures social scientists can employ, with real consequences for research into opinion formation, opinion dynamics, and government responsiveness. This project further develops and automates a technique known as multilevel regression and post-stratification (MRP) that allows for national surveys and publicly available Census data to be used to estimate public opinion sub-nationally. While the methodological and substantive potential of MRP has been established, it has not yet been advanced to its full potential either in usage or scope. This project enables several significant extensions. These include making the MRP method itself far more accessible to a broad range of researchers and poll analysts, by automating and routinizing the process, so that others can create estimates from their own data. Robustness checks and measures of uncertainty are developed and made part of the automated process. In addition, diagnostics are created to show if the method is working properly in a particular setting. The proposed research will promote the progress of science and the national welfare by: 1) Contributing to the growing literature on the role of public opinion in government policymaking and the degree to which majority will is represented, or, conversely, the degree to which non-majority groups prevail. 2) Developing statistical tools and methods for the study of policy responsiveness. 3) Involving undergraduate and graduate students in the conduct of our research. 4) Disseminating results to the academic community through conference presentations, journal articles, and other publications. 5) Disseminating results to the non-academic community through the mass media and through presentations to non-profit and civic organizations fostering democratic engagement and greater policy responsiveness from government agencies.