Friday, December 13, 2013

2012 Election Model: House Races

In the last post I modeled the change in turn out from 2008 to 2012 (at the county level) to see if there was any statistical significance with electorate demographics and economic factors for the presidential race. In this post, I performed the same regression analysis, but instead on the House races. House races, like the Presidency, are on the ballot for every U.S. citizens. On the other hand, governor and Senate races are harder to evaluate for trends because not every seat is up for election during presidential election cycles. Of course, all house seats were up for election during 2010, but this analysis will only look at the voter difference between 2012 and 2008 House races because the turnout is more comparable and much higher than mid-term elections. In fact, the turn-out between 2008 and 2012, though smaller, was very similar demographically – The ideological breakdown favored Democrats in 2008 by +7 and in 2012 by +6; Minorities made up 26% of electorate in 2008 and 28% of the electorate in 2012; and women made up 53% of the electorate in 2008 and 2012. In both cases the economy was also in bad shape. Here is the breakdown:

Age – Just as in the Presidential race, there was plenty of statistical significance, but the results were inconclusive. For males the Democrats had strong statistical significance in 5 age segments, but weak significance in 8 whereas the Republicans has strong significance in 6 age segments and weak significance in 5. On the female side the Democrats were strong in 6 and weak in 5 and the Republicans were strong in 4 and weak in 4 others. Overall the trend was a bit more favorable to Republicans as one would suspect since they won a majority of the vote.

Race – The results were surprising here. Democrats showed not only weak significance among Whites, but with African Americans! Not surprisingly, Democrats were strong with Hispanics. Republicans, on the other hand, had strong significance with Whites and weak significance among Asians and Hispanics.

Income and Employment – Republicans had strong statistical significance among the wealthy and the Democrats among the poor, but there was no correlation along employment status for either group. In other words, the economy had no impact on voters’ decisions.

Gender – Surprising, there was no gender gap that is evident in the Presidential race. There is absolutely no evidence that women overwhelmingly support Democrats and men overwhelmingly support Republicans.

Education – Republicans had strong significance with all demographic groups with at least a high school diploma. Republicans had weak significance with those who did not have a high school diploma. However, surprisingly, the Democrats had significance with any groups.

Food Stamps – People who did not collect food stamps overwhelmingly voted against Democrats, but other than that there was no other statistical significance.

Marriage – No surprise here, married folks broke heavily for Republicans, but single family household showed no statistical significance towards Democrats or Republicans.

So how did Republicans hold the House and win a majority of the vote while Obama defeated Romney by 3.5 points? There are several key differences in voting trends for the Presidency and the House. First, African-Americans may be more likely to abstain from voting in House races. Secondly, women are more likely to vote for a Republican. Thirdly, people on food stamps are also more likely to abstain from voting in House elections.


  1. There's a lot of important information in this analysis, Patrick. Good job!

  2. Thanks CW. Probably the most surprising finding is there was no gender gap for House races.

  3. Hmmmm... why do you think that is?

    1. Not too sure. My gut tells me that there would be a gender gap in most Senate and Governor races. I think House races do not get the same amount of money and media attention and therefore candidates are not portrayed as "hating" women. It is an interesting result.