The Demographics of Victory [Fiddling with Numbers]

When we ran the various possible House models last week, one of the most interesting (yet not at all surprising) results was a strong and statistically significant correlation between the gain for the opposition party and the party of the president. That is, all other things being equal (such as the president's approval rating and the generic ballot), House Republicans tended to do better against Democratic presidents than House Democrats did against Republican presidents. This was not something immediately obvious, but makes sense when you consider that a tie in the generic ballot predicts a six-seat gain for Republicans. The reason for this is primarily because of fundamental differences between presidential elections and midterm elections, which boil down to this graph:

Source: McDonald, Michael P. "American Voter Turnout in Historical Perspective." In Oxford Handbook of American Elections and Political Behavior, ed. Jan Leighley. Cambridge, UK: Oxford University Press, 2010.

The main story that should be drawn from this graph is that, at least since the Civil War, presidential elections have always seen higher turnout--usually by almost twenty points. Today, not even a majority of the voting eligible population (VEP) goes to vote in years that aren't leap years; they haven't since the 19th century. A lot of theories have been proposed to explain why the gap between presidential elections and midterm elections occurs: presidential candidates generate a lot more enthusiasm compared to House candidates, for example.

The significance for midterm elections, however, is simpler: we focus on who leaves the electorate from 2012 to 2014. If turnout decreases by a third (as it did from 2008 to 2010), it doesn't mean that 33% fewer men voted, 33% fewer women voted, 33% fewer whites voted, 33% fewer African-Americans voted, etc. It means that 33% fewer people voted. The drop in turnout is disproportionately concentrated among specific demographics. Here's the breakdown (from CNN exit polls from 2008, 2010, and 2012):

• Race. In 2008, 74% of voters identified as white; similarly, in 2012 72% of voters did so. Because the trend appears to be a "darkening" of the electorate, the expected proportion of voters who were white would be between 74% and 72%, but the percentage actually increased by three points to 77% in 2010. Conversely, black and Hispanic voters as a percentage of the electorate declined in 2010 compared to either 2008 or 2012.
• Age. In 2008, voters aged 18-29 represented 18% of the electorate, and in 2012, they represented 19%. In 2010 they constituted 12% of the vote. At the same time, senior citizens, who represented 16% of the electorate in 2008 and 2012, represented 21% of the vote in 2010.
• Education. The electorate in 2010 was more educated, with 53% who had completed at least an associate's degree, compared to 44% and 47% in 2008 and 2012, respectively.
• Party identification. 32% of voters in 2008 and 2012 identified as Republicans, compared to 35% in 2010. At the same time, about 39% of voters in 2008 and 2012 identified as Republicans, while in 2010 it was about 35% (in other words, the Democratic ID lead over Republicans shrank by 7 points).
• Political ideology. 2010's electorate was more conservative than either 2008 or 2012, with 42% identifying as conservative in 2010 compared to 34% in 2008 and 35% in 2012.
• Religiosity. In 2008, 40% of voters attended religious services at least weekly, and in 2012 about 42% did. In 2010, that number was 48%.

We'll summarize it in the following hyper-general sentence: "The average voter in a midterm election is whiter, older, more educated, more conservative, more religious, and more likely to be a Republican than the average voter in a presidential election." Demographically speaking, this pretty much describes the Republican base:

• Race. In 2008, Sen. John McCain won whites 55-43; Gov. Mitt Romney improved on that margin in 2012, winning whites 59-39.
• Age. There's a large generational gap at work in American elections. In 2008, Sen. Obama won voters between 18 and 29 by a margin of 66-32, while Sen. McCain won voters 65 and older by a much smaller margin of 53-45. The gap was more pronounced in the closer 2012 election, in which President Obama won the 18-29 group 60-37 and Gov. Romney won the 65 and older cohort by 56-44.
• Education. Except for those with postgraduate degrees (who trend heavily Democratic), higher levels of education slightly correlate with a greater lean toward Republicans. Gov. Romney won college graduates by a narrow 51-47 margin in 2012.
• Party ID. Republicans vote more Republican than Democrats do. Moving on now.
• Political ideology. Political sorting over the last forty years or so has made it so that, with the exception of Susan Collins (R-ME), the last remaining centrist Republican, the most conservative Democrat in the Senate is still more liberal than the most liberal Republican. The House is almost finished this sorting process as well. In other words, political parties have become less umbrella groups for ideologically diverse populations and more ideological in themselves: to generalize, Democrats are the liberal party, Republicans the conservative one.
• Religiosity. Evangelicals and born-again Christians broke 74-24 for Sen. McCain and 78-21 for Gov. Romney. Even among non-evangelicals, those who attended religious services at least once a week voted 55-43 for Sen. McCain and 59-39 for Gov. Romney, while those who attended religious services less often voted 60-39 for Sen. Obama in 2008 and 57-40 in 2012.

Sum total, these small shifts incrementally add up to a large advantage to Republicans in 2010 compared to in 2008 and 2012--simply because of the demographics of midterm elections. We can demonstrate this quite well by imagining what President Obama's 2012 victory would have looked like had the electorate been like 2010 in racial makeup. By multiplying President Obama and Gov. Romney's respective proportions of each racial group by the 2010 racial makeup, we calculate that President Obama's 51.1-47.2 popular vote victory becomes a 48.6-49.5 popular vote loss. Fiddling with the demographic numbers in this same way works in President Obama's narrowest victories as well: for example, in Florida, adjusting the racial makeup of Florida to its 2010 composition turns President Obama's 50.0-49.1 victory into a 47.1-51.2 loss, and in Ohio the president's 3-point lead decreases to a dead heat.

The point is this: despite President Obama's victories among minorities in 2012 being touted as the beginning of the end for the GOP, Democrats still have a lot of work to do if they want to make use of those victories. Democratic victories in 2008 and 2012 were not just the result of high minority appeal but also the ability to turn out minority voters. Even the slightest depressions in minority vote--and the resulting increase in white share of the electorate--can tip the scales. The lesson for Democrats is that probably the most efficient way to increase their chances of victories in November is to do what President Obama did so effectively in 2008 and 2012: create an electorate that looks more a presidential year and less like a midterm one. More like 2012 and less like 2010, in short.