Subnational Public Opinion Estimation Using MrsP

Multilevel regression and poststratification (MrP) is a powerful tool for estimating local public opinion from national survey data. However, it has seen little use outside of the United States, owing to its rather stringent data requirements. In this paper, I prove that an extension of MrP using synthetic poststratification (MrsP) can overcome these limitations without introducing error. I then use this technique to disaggregate crossnational survey data from the World Values Survey and International Social Survey Programme. I find that these estimates are highly correlated with alternative measures of public opinion, and discuss how this technique could serve as a useful tool for election forensics.