FLIP NC TUrnout Model Details

We estimate our statistical model of voting behavior by using voter and district characteristics to predict the outcomes of the 2016 election for Governor and the 2018 elections for NC House and Senate at the precinct level.

Given our focus on the NCGA, we use the 2016 Governor race because it was the strongest predictor (among the three major statewide races) of outcomes in contested legislative races in 2016. We give equal weight to predicting 2016 and 2018 elections and, within 2018, give equal weight to predicting the House and Senate outcomes.

Our model includes variables that characterize the composition of the electorate by voters’ party, race, ethnicity, and sex. We allow the impact of these variables to vary with the urbanicity and college education level of the associated NC house district. The main conclusions of our analysis are not sensitive to the set of specific variables included in the model.