How Much Did Demographics Matter in Va. Race?

How Much Did Demographics Matter in Va. Race?

By Sean Trende - November 12, 2013

The morning after last week's election, I wrote, regarding Virginia's surge in non-white voters:

"There was a bounce-back from 2009 lows, as expected, but the demographic shifts were probably about more than a bounce-back. To use racial crosstabs as an example, the 2012 electorate was 70 percent white, while the 2009 electorate was 78 percent white. The 2013 electorate was 72 percent white. Most of that difference came from increasing the African-American share of the electorate vis-à-vis 2009. This is probably the most encouraging data point for the Democrats for the night.”

Other analysts, like Washington Monthly’s Ed Kilgore and Democratic demographic godfather Ruy Teixeira, have made similar arguments. The latter argues:

“In 2009, Virginia voters were 78 percent white and 22 percent minority. In 2013, they were just 72 percent white and 28 percent minority -- not far off the 70/30 split in the 2012 presidential election. There you have the key to [Terry] McAuliffe’s victory: Despite performing much better among white voters than the hapless Creigh Deeds, [Bob] McDonnell’s Democratic opponent, McAuliffe would nevertheless have lost this election if the white/minority voter distribution had mirrored that of 2009. It was the increase in the minority vote that put him over the top.”

I almost hate to be contrarian here, because we all are on the same basic page -- this is an important data point. If Democrats can mirror the Obama coalition without Obama on top of the ticket, that’s a very good sign for them in midterm elections going forward, and a very good sign in presidential elections.

With that said, I think there are three important caveats to keep in mind before we get too far ahead of ourselves:

1. This shift in turnout is probably in part about Creigh Deeds and Ken Cuccinelli.

A lot of people forget this, but Deeds had real problems with the African-American community in 2009. Former Gov. Doug Wilder refused to endorse him, and the cash-strapped candidate failed to mount a get-out-the-vote effort. Because of this, African-American participation four years ago was probably depressed somewhat. If our baseline is artificially low, then our measurement of the swing will be artificially high.

At the same time, it is possible that Cuccinelli’s polarizing persona turned off some moderate white voters. If this is the case, however, the effects were probably marginal: White turnout was up 4 percent, while the white population grew by about 5 percent. Regardless, I’d guess that turnout in 2013 was about halfway between the “real” floor in 2009 and the 2012 numbers. That’s still significant.

2. The exit polls might overstate the surge in African-American votes.

There are two important quirks in the data to keep in mind here. The first has to do with the nature of the exit polls in general. They found a 72 percent white electorate in 2004, a 70 percent white electorate in 2008 and a 70 percent white electorate in 2012. If these numbers are correct, then the size of the white electorate isn’t that daunting for Republicans, since George W. Bush was able to win Virginia handily in 2004. It also suggests that the 78 percent number that the exit polls found in 2009 was a real outlier.

But the 2 percent swing from 2004 to 2008 in a state that is 20 percent African-American looks pretty anomalous. By contrast, the Current Population Survey, the census bureau’s estimate of voter participation, found that the electorate was 80.6 percent white in 2004, 74.4 percent white in 2008 and 73.5 percent white in 2012.

The CPS estimates have their own problems -- in particular more people report voting than actually did so, and there are some good reasons to believe that the over-reporting issue isn’t uniform across demographic groups. More importantly, we don’t have CPS estimates for elections in odd years, so we can’t do an apples-to-apples comparison. Regardless, let’s stipulate that there’s at least a chance that exit poll findings here are off.

The second quirk is related to the first, but is more specific to the 2013 exit poll data. We know that non-whites tend to be younger, more liberal, and more Democratic than whites. Yet while the increase in younger, liberal Democratic voters was there, it was less pronounced than we’d have expected given the reported shift in the electorate.

Remember, exit polls have a lot of the same response bias problems that traditional polling companies are encountering, without the benefit of a full sample of “all adults” to weight against. The exits are eventually weighted to actual results in a sort of bizarre game of 15-dimensional chess. It’s tough to get right, and it is one more reason that it is important to pay more than lip service to the notion that exit polls have error margins.

The youth vote was up 3 percent from 2009, the Democratic vote was up 4 percent from 2009, and the liberal vote was up 2 percent from 2009. Again this backdrop, the increase of 6 percent in the nonwhite vote looks anomalous.

If you want to place a further question mark on the exit poll findings: We only have crosstabs for the white and African-American populations, presumably because the sample sizes for Hispanics, Asians, and “others” are too small for reportable results.

If you just look at the share of the electorate that was reported as white or black, and then look at the shares that Cuccinelli, McAuliffe and Libertarian candidate Robert Sarvis received, it works out to a two-point McAuliffe win. This makes sense, until you realize that we haven’t factored in the share of the electorate that is neither white nor African-American.

Basically, unless you think that the remainder of the non-white vote only went for McAuliffe by seven points (any case for this would probably have to involve Sarvis, who is half-Chinese, doing extremely well among Asian voters), the demographic data reported in the exits pertaining to race are off. By quite a bit, in fact; if McAuliffe actually won these voters by a more traditional 70-30 split, then these “back out” to a five-point McAuliffe win.

Even this doesn’t mean that the white share of the electorate was understated; it may be that Cuccinelli did better among whites than reported, rather than whites constituting a larger share of the electorate that reported. If you’re looking for evidence of this possibility, turn to the African-American majority city of Petersburg, where total votes cast increased from 6,400 in 2009 to 8,300 in 2013. My bet is on a combination of both, but it’s really hard to know for certain here. It’s probably enough just to remember that there’s an asterisk on these reported data (as tends to be the case with exit polls).

3. Shifts within groups mattered more than demographic shifts.

Finally, I do have to take issue with Ruy’s assertion that “[i]t was the increase in the minority vote that put him over the top.” In a sense it is literally true; had the minority vote not increased from 2009, Cuccinelli would be governor-elect.

But in a close race, just about everything puts the winning candidate over the top. What we need to ask is what was the most important factor. In one sense, the 60,000 votes George W. Bush received from Florida residents who were gay put him over the top in the state in 2000, but if you were looking to explain his win, I think most analysts would to start with white evangelicals.

Regardless, Bush needed all groups to swing toward him that year. Likewise, McAuliffe needed the demographic swings that he helped create, but he also needed improved Democratic performance within those groups. Without both, he’d have lost. Now we need to sort out which was more important.

We know that McAuliffe won by about 2.5 points, and that McDonnell won by 17.3 points in 2009, for a total net swing of 19.8 points. Remember that number.

If we take the McDonnell/Deeds vote share among the various age cohorts (e.g. 54/44 McDonnell among 18-29-year-olds, 56/44 McDonnell among 30-44-year-olds), and substitute those numbers for the McAuliffe/Cuccinelli numbers in the 2013 breakdown (e.g., the 2013 electorate was 13 percent 18-29-year-olds, 23 percent 30-44-year-olds), we can see how much the numbers swing. We can then estimate how much of the 19.8 point swing was due to that particular demographic change.

When we replace the McAuliffe/Cuccinelli numbers with the Deeds/McDonnell numbers, the results swing 19 points toward Republicans from the 2013 baseline. So we can estimate that, at least with respect to age cohorts, demographic change was responsible for 0.8 points worth of change, while changes within age cohorts (e.g., McDonnell won those over 65 years old by 20 points, while Cuccinelli won by five) were responsible for 19 points worth of the shift. Put differently, if a Republican performs as well as McDonnell in the different age cohorts, she’d win by 16.4 points, even with the new demographic breakdown.

What about with race? It’s a tougher call, because as we noted above there are some questions about the racial crosstabs. But the swing is 11 points, representing a little more than half of the actual swing. Put differently, even in this electorate, a Republican who ran as well as McDonnell ran with whites and blacks would win by 8.5 points.

Of course, it’s not an either/or proposition: There were more young voters and more minority voters. But there’s also a substantial crossover here; by bringing out more minority voters, McAuliffe was necessarily making the electorate younger. In other words, it was harder for Cuccinelli to do as well as McDonnell among young voters because those voters were less white, and it was harder for him to do as well among white voters because they skewed younger this time. It’s really difficult to sort out.

So let’s turn to some factors that are byproducts of all the demographic changes together: Things like partisanship, ideology, and presidential approval. These things have their own problems -- unlike demographics, they aren’t fixed in -- but it at least gives us a different sort of look at the electorate.

So what about partisanship? If Cuccinelli had done as well as McDonnell with Democratic, Republicans and Independents, he would have performed 11 points better. Again, we see that changes in the electorate matter a little less than half as much as changes in voting patterns within the electorate.

What about ideology? If Cuccinelli had performed as well as McDonnell with conservatives, moderates and liberals, he’d have done about 12 points better. If Cuccinelli had performed as well as McDonnell with those who approved and disapproved of the president, he’d have done a whopping 23 points better.

Incidentally, it isn't just the true that if Cuccinelli had performed as well as McDonnell with these groups, he'd have won. If you go through this exercise using the Romney/Obama numbers from 2012, you find that if Cuccinelli had done as well as Romney with these groups, he'd have won.

There are four key takeaways here. First, note that despite all of these demographic changes in the Democrats’ favor, a greater share of the Virginia electorate disapproved of the president than in 2009; we should note that this “coalition of the ascendant” blamed the parties equally for the government shutdown and it disapproves of Obamacare.

Second, Democrats really should be pleased with these changes. No matter how you slice it, the McAuliffe campaign’s ability to turn out previously marginal voters helped the Democrats up and down the ticket. If Democrats can do this in 2014 in other races, their candidates will benefit and otherwise-close losses will become wins

Third, we should be careful about reading too much into the demographic shifts. If I had to guess, I’d say that about half of the shift from McDonnell’s big win to McAuliffe’s narrow win was due to changing demographics, while the other half was due to the differences in the way the candidates appealed to these demographics.

This is the fourth point, and it is important: Even in this emerging coalition, Republicans can win if they run acceptable candidates. A candidate with McDonnell’s broader support would have put this race away easily. We should also keep in mind that if 30,000 voters had changed their minds, an electorate that seemingly came close to approximating the 2012 electorate would have elected an underfunded candidate with little establishment backing who made some of the more outlandish attacks on himself believable by embracing positions on social issues that really did go beyond what a traditional successful Republican in Virginia would embrace. Demographics matter, but they aren’t everything. 

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.

Sean Trende

Author Archive

Follow Real Clear Politics

Latest On Twitter