How do we estimate public opinion in the states?

How do we estimate public opinion in the states?

Description

We compare two approaches for estimating state-level public opinion: disaggregation by state of pooled national surveys and an updated simulation or imputation approach using poststratification and multilevel modeling. We present a systematic assessment of the predictive accuracy of each, and how this is affected by sample size and state population. We give practical advice to applied researchers, including when each method should be used and how complex an imputation model need be. To do so, we use an original dataset of over 100 surveys asking respondents their opinions of same-sex marriage. Both methods work well under optimal conditions, but imputation is superior when samples are smaller and performs equally well across sample sizes. These results have striking implications, most notably that imputation can accurately estimate state opinion using only a single large national survey. This greatly expands the scope of issues for which researchers can study sub-national opinion directly or as an influence on policymaking.

People

Papers