Midterm Demographics Didn't Sink the Democrats

Midterm Demographics Didn't Sink the Democrats

By Sean Trende - November 19, 2014

A congealing conventional wisdom surrounding the 2014 elections is that Democrats had a long night because of an unfavorable Senate map and because Democratic constituencies failed to show up. One storyline growing out of this is that once Democrats can enjoy a “presidential electorate” rather than a “midterm electorate,” their fortunes will turn, and Democrats will run well.

This isn’t entirely correct.  The major factors driving the different results between 2012 and 2014 were not demographic.  The major difference was that in 2012 Barack Obama was a moderately popular president.  In 2014, he is an unpopular president.  If this does not change between now and 2016, demographic shifts alone will not save the Democratic nominee.

We can illustrate this best by borrowing a page from Harry Enten, and seeing what would have happened if the 2014 electorate had instead more closely resembled the 2012 electorate. That is to say, let’s keep whites voting 60-38 for Republicans, Hispanics voting 62-36 for Democrats, and so forth, as they all did in 2014, but alter their shares of the electorate to resemble 2012 (72 percent white, 10 percent Hispanic, and so forth) rather than 2014 (75 percent white, 8 percent Hispanic, and so forth). This allows us to isolate the effects of demographic change between 2012 and 2014.

The results are underwhelming: If the 2014 electorate had resembled the 2012 electorate in terms of race, the Republican vote share would shrink by just 1.97 percentage points.  In other words, in a 2012 electorate, Republicans would have won the popular vote for the House by 4.5 points, rather than 6.5 points.  That’s not nothing, as they say, but it still only explains a relatively small share of the difference between the 2012 and 2014 results. Put differently, if Obama had put up the same vote shares among racial groups in 2012 as Democrats ultimately did in 2014, he’d have lost.

Perhaps the difference is not so much differences in the racial makeup of the electorate, but rather differences in the age makeup of the electorate? The 2014 electorate was, in fact, quite a bit older than the 2012 electorate.  This isn’t necessarily surprising, given that the elderly population is actually set to grow substantially in the next decade. Regardless, if we reduce the 65+ share of the electorate from 2014’s 22 percent to 2012’s 16 percent, increase the 18-24 year old share from 7 percent (2014) to 11 percent (2012), and adjust everything in between accordingly, the Republican advantage contracts by ... 1.94 points.

Now you might look at this and say, “Well, that’s a total of four points!”  The problem with this approach is that there is a substantial double count going on.  Democrats do better among young voters in large part because that demographic is less white; younger whites don’t vote that differently from older whites.  So this isn’t a cumulative exercise.

To get around this, we can look at the age-race crosstabs.  That is, the exit polls tell us how not just 18-29-year-olds voted and African-Americans voted, but also how 18-29-year-old African-Americans voted (and so forth). If those groups had turned out in a way as to re-create the 2012 electorate, the Republican margin constricts by a bit more than if we looked at race alone or age alone, but the change still only amounts to about two points.

In other words, even if Democrats had managed to re-create 2012-style age or racial demographics in 2014, they still would have had a rough year.

Maybe a better way to see the differences is to look beyond demographic splits and instead compare the ideologies of the two electorates.  But if the 2014 electorate had consisted of the same proportions of liberals, moderates, and conservatives as the 2012 electorate, the Republican margin would have contracted by just three points.

By now you should see a pretty distinct trend.  When Harry Enten compared the demographics of 2012 to 2010, he found that demographics helped Republicans by about three points (net) in 2010.  Looking at 2012 and 2014, I find that demographics helped Republicans by about two points (net) in 2014.  It seems reasonable to say that demographics were responsible for two-to-three points’ worth of difference between presidential and midterm election results.

Now we do see some substantial changes in outcomes if we compare party identification. It still isn’t enough to turn the popular vote over to the Democrats, but more importantly, this commits the exact same fallacy that poll unskewerers were committing on the left in 2004 and on the right in 2012: assuming that an individual’s party ID is constant.

But, in fact, Democrats went from a five-point advantage among adults in November 2012 to a four-point disadvantage in October 2014 (their largest disadvantage since late 2005) among the total adult population. In other words, when the electorate shifted from a six-point Democratic edge in the electorate in 2012 to a one-point Republican edge in 2014, it was just mimicking changes representative of what was occurring in the overall population, rather than reflecting issues with turnout.

Now, some may press further, note that the Electoral College is where the action is, and ask whether perhaps we might see more substantial demographic shifts in certain swing states.  If Joni Ernst, Cory Gardner and Thom Tillis would have lost in a 2012 electorate, perhaps Republicans should be more worried.

In reality, this is a hopeless exercise. Candidates matter. Republicans won’t be running against Bruce Braley in Iowa in 2016, but their nominee won’t likely have the baggage of an unpopular legislature in North Carolina.  More to the point, if individual 2014 Senate/governor races are indicative of how the Electoral College will shake out in 2016, then Democrats should be very worried about places like Illinois and Maryland. Of course, they shouldn’t worry, because we all know that these races aren’t necessarily representative of the campaign dynamics in two years, irrespective of demographic results.

Nevertheless, we see shifts at the state level similar to those we see at the federal level when we journey from 2012 electorates to 2014 electorates.  If we estimate the Republican share of the non-white vote in states that don’t have full demographic crosstabs, we find, for example, that Cory Gardner probably lost non-white voters by about 28 points (while winning whites by 10).  Applying the demographic splits from the 2012 election in Colorado only moves the results about 0.4 points to the left, resulting in another Gardner win.

We see similar results in other states. In a 2012 electorate, Rick Scott’s victory margin would have been a half-point narrower. Bruce Rauner’s margin would have been 3.7 points narrower. Rick Snyder’s margin would have been 1.5 points narrower. Tillis would have lost, but the swing is still small: 3.8 points.  Mark Warner’s lead would have expanded by just 0.7 points. Scott Walker’s margin would shrink by 1.38 points.

States like Iowa and New Hampshire don’t have enough racial diversity to really make anything of those crosstabs, but replicating the 2012 electorate in terms of ideology moves things 0.26 points away from Ernst and 0.75 points away from Scott Brown.  Replicating the 2012 age breakdowns in those states moves things just .04 points away from Ernst and .35 points away from Brown.  Incidentally, if we look at age and ideology in North Carolina, rather than race (which was problematic, since we estimated for the non-white/non-black population), the swings shrink back to around two points: Tillis wins by a half-point when we look at age and loses by a half-point with respect to ideology.

No matter how you slice it, demographic changes in the midterm electorate account for a relatively small portion of the Democrats’ problems in 2014.  The real difference between 2012 and 2014 isn’t changes in the demographic makeup of the electorate.  It is changes in the way that demographic groups voted. This, in turn, has everything to do with the president’s job approval rating.

On Election Day 2012, the president had a 49.9 percent job approval rating and a 47.4 percent disapproval rating.  In 2014, by contrast, the president had a 42 percent job approval rating, and a 53.3 disapproval rating.  Notably, this isn’t ascribable to likely voter screens; the highest the president has been in polls of adults since June was 45 percent.

This low job approval interacted with state partisanship more heavily than it did with state demographics.  After all, Obama was popular enough in 2012 in places such as Montana and North Dakota to enable Democratic Senate victories. If Obama’s job approval had been 54 percent in the overall 2014 electorate, rather than 44 percent (as exit polls indicated), the Republican purple state wins would not have occurred, and some of the red states would have elected blue senators.

Of course, this should be intuitive.  After all, several models based on “fundamentals” had suggested that Democrats were in trouble in these elections.  Yet none of these models, to my understanding, include a variable distinguishing between Democratic performance in midterm elections and in presidential elections. If this were a substantial factor, the models would skew badly toward Democrats in midterms and toward Republicans in presidential elections.

We might even go a step further and inquire if perhaps what is frequently being called the “presidential” electorate is really simply the “Obama” electorate, something political scientists like John Sides have questioned, and something that Democratic strategists are increasingly fretting about.

Regardless, the Democrats’ problem in 2014 was not simply the map, nor was it mostly a demographic/turnout issue. It was an unpopular Democratic president.  If Hillary Clinton (or whoever) does not perform better among these groups in 2016 (and she/he might!), the best turnout machine in the world will not save her or him.

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.

How Democrats Can Get Their Mojo Back
Caitlin Huey-Burns · December 8, 2014
A Reply to Rush Limbaugh
Mona Charen · December 9, 2014
Against the Wave: Three Winning House Democrats
David Byler · December 14, 2014
Obama, Congressional Dems Show Cracks in Unity
Jim Kuhnhenn · December 7, 2014

Sean Trende

Author Archive

Follow Real Clear Politics

Latest On Twitter