“Travels with MRP: Election Forecasting in the U.S., U.K., Germany and Alabama,” Douglas Rivers, Stanford University
QUANTITATIVE RESEARCH METHODS WORKSHOP
The combination of multilevel regression and post-stratification (MRP) offers a promising alternative to conventional weighting methods for correcting samples for selection bias. There are, however, many practical issues in encountered in real-world applications:
* Models are needed for both turnout and vote choice.
* Vote choice is not dichotomous, even in two candidate elections (some voters are undecided and there are write-ins).
* Selection bias is often not ignorable when conditioning on just demographics.
* Post-stratification targets are usually not available for the cross-classification of the full set of demographic and political variables.
I will describe modeling choices that my colleagues and I have made in five recent elections across the world and how they performed in practice.
Douglas Rivers is Professor of Political Science at Stanford University, Senior Fellow at the Hoover Institution, and Chief Scientist at YouGov.
This workshop series is being sponsored by the ISPS Center for the Study of American Politics and The Whitney and Betty MacMillan Center for International and Area Studies at Yale with support from the Edward J. and Dorothy Clarke Kempf Fund.