Wisconsin Could Be GOP Turning Point

Wisconsin Could Be GOP Turning Point

By Sean Trende - April 3, 2012

Shortly after Super Tuesday, I put together a model that projected Mitt Romney's share of the "non-Paul vote" (i.e., the Gingrich/Romney/Santorum vote) at the county level based on six variables: the percentage of Mormons, the percentage of African-Americans, the percentage of Latinos, the percentage of college-educated voters, the percentage of evangelicals, and whether a state was Midwestern and conducted a caucus (it also included a dummy variable for Georgia and Massachusetts, given "home field" advantages in those two for Newt Gingrich and Mitt Romney).

Initially, the model performed splendidly, accurately predicting Romney’s vote share in Kansas, Mississippi and Alabama, both at the county and statewide levels. The county-level errors seemed randomly distributed, suggesting that the demographic factors that had been driving the race continued to dominate.

But then there were two misses. In Illinois, Romney performed about six points better than expected, while in Louisiana he performed about six points worse. Neither misses were outside the standard error of the model, but it suggested that perhaps something had changed.

Upon closer inspection, however, these two states probably tell us more about the limits of predictive models than this one in particular. More than anything else, the misses are likely due to unique circumstances that the model could not account for.

Consider the following map of Louisiana. As a county gets redder, Rick Santorum over-performs the model’s projection by a greater margin. Bluer means Romney over-performed. Anything that is white means that the model pegged Romney’s share within five points.

As you can see, the big errors were clustered in south-central Louisiana. What’s unique about that area? Consider this map:

South-central Louisiana, also known as Acadiana, is dominated by French Catholics, and hence has a relatively low evangelical population. Usually, this would suggest a strong showing by Romney, as he has traditionally done well in counties with high percentages of mainline Protestants and Catholics. But Catholics in Louisiana have a unique cultural tradition, and it seems counties in Louisiana with more Cajuns are unusually disposed toward Santorum.

No other state in the nation has such a significant Cajun population, so this isn’t something likely to be repeated elsewhere. Outside of Acadiana, everything went largely according to plan, suggesting that there had been no demographic breakthrough for Team Santorum.

So, what about Illinois?

At first blush, this is a little more difficult to explain. I assumed that most of the misses would be concentrated in northeastern Illinois, in Chicago's collar counties (Lake, DuPage, etc.). If this had occurred, it would suggest a Romney breakthrough in the northern suburbs.

This did take place to some extent, but the errors were really concentrated more in Cook County than the collar counties. In addition, there was a band of counties across the center of the state where Romney over-performed. In particular, Sangamon (Springfield), Macon (Decatur), DeWitt and the surrounding counties were Romney Country to an unusual degree.

There are several plausible theses for what is going on in these places. In particular, one could hypothesize that Romney was beginning to run better in small Midwestern cities. But this wouldn’t explain his over-performance in Cook County, nor would it explain why he performed poorly in Effingham and about average in other small cities throughout the state.

One alternate explanation can be found in this map:

Remember, Santorum didn’t have delegates on the ballot in the 13th District, which takes in many of the counties in the middle of the state where Romney did better than expected. Unlike most states, Illinois has an actual direct election for delegates in addition to a “beauty contest” for the nominee. It’s plausible that voters saw the lack of Santorum delegates, and declined to cast a ballot in what they viewed as a fruitless exercise.

1 | 2 | Next Page››

Sean Trende is senior elections analyst for RealClearPolitics. He is a co-author of the 2014 Almanac of American Politics and author of The Lost Majority. He can be reached at Follow him on Twitter @SeanTrende.

Wisconsin Becomes 2016's Ohio
Salena Zito · November 9, 2014
Mitt Romney for Mayor
Carl M. Cannon · November 16, 2014

Latest On Twitter