2016 Election Results Broken Down by Race and Gender

The electorate in this model is 73% White, 12% African-American, 10% Hispanic, 3% Asian, and 2% Other. 52% Female and 48% Male. These numbers were chosen on the basis of historic trends (in terms of how much the exit polls undersample or oversample certain parts of the electorate), and current data on racial turnout available in a few states. The American Community Survey will provided definitive numbers on the composition of the electorate sometime in mid-2017, but until then, this is the best estimate I believe we have.

I created scatter plot regressions of the various racial groupings based on exit polls to determine how non-White racial groups change their partisan leaning based on the demographics and partisan lean of a state. I utilized exit polls from Latino Decisions and Asian American decisions to give a sense of how the sub-categories for Hispanics and Asians voted (such as the more traditionally GOP-leaning Cuban-Americans and Vietnamese-Americans) voted in relation to the mean (though I did not define the total national support based on those results). For Native Americans, I did a large number of ecological inferences for Native American reservations (which have a different partisan behavior compared to the Native Americans living outside reservations) to determine the particular partisan lean of Native Americans in states with reservations. Combining all of these methods together, I believe I have a pretty strong model for what the vote resembled. Scatter plots were created to determine the gender gaps within states.

Obviously with these estimates, there is going to be some inherent error here I cannot control for, but I believe this is about as accurate as I can make it with the data available.

Detailed results here.