The Rise of Non-White Voters: Why The Racial Composition Of National 2016 Election Polls May Be Missing The Mark

Exit polling indicates that the non-white share of the electorate has increased by 2-4% in every presidential election since 1992. Picture courtesy of Jacquelyn Martin/A.P.

The Cook Political Report's Political Analyst David Wasserman recently tweeted the message below, regarding the likely racial make-up of the 2016 presidential election:

New @CookPolitical: White % of electorate may fall from 72% to 70%, Latinos & Asians/Others each up 1%, AA % stable: http://t.co/lMsalsgl5L

— Dave Wasserman (@Redistrict) June 19, 2015

Wasserman's tweet revives a point made very shortly following the 2012 election, when I posited that based on demographic shifts since 1992, white voters could expect to make up anywhere between 68-70% of the 2016 electorate.  Why? Because the white share of the presidential vote has dropped between two and four points every cycle since 1992.

Well, the prognosticators at The Cook Political Report have spoken. And given their level of expertise in these matters, I'll happily give them the benefit of the doubt and go with their estimate - the 2016 electorate should be roughly 70% white, and 30% non-white.

Based on Wasserman's analysis, it might be a bit surprising to learn that the racial composition of some pollsters' surveys looks little like his assumption of the 2016 electorate. Democratic firm Public Policy Polling's most recent national survey found likely voters identifying as 74% white, and 26% non-white. If PPP's past success boils down to, what their director Tom Jensen called in 2013 "... a well informed but still not entirely empirical hunch," you have to wonder what less-than-empirical hunch led them to peg the 2016 electorate at 74% white, 26% nonwhite. These figures represent an even LESS racially diverse electorate than the one that showed up in the 2012 presidential election. And as I've already noted, the electorate has become MORE racially diverse in every presidential election since 1992. In other words, for the 2016 electorate to resemble PPP's racial composition, a major reversal of precedent is required.

But pretending Jensen's hunch is correct, the GOP's failure to overtake Hillary with an electorate as white as it was in 2008 would obviously be bad news for Republicans. And perhaps this is why PPP has consistently been one of the most GOP friendly national pollsters, at least with regards to their November 2016 horse race polling (behind Fox News). For example, take Jeb Bush. While he trails Hillary Clinton an average 51-42% per Huffington Post's Pollster, and by six points per Real Clear Politics, PPP finds Clinton up just 45-41%. Had their racial composition more closely resembled Wasserman's prediction, the contest falls outside of the margin of error, and Jeb trails Hillary 46-40%.

The Rasmussen Poll is nearly as guilty as PPP, with their likely voter population being less racially diverse than the one we saw in 2012 (73% white, 27% non-white). Note the favorable sample does little to help Jeb Bush, who still trails Hillary Clinton 45-36%. Had Rasmussen's racial composition matched Wasserman's proposal, Clinton's lead grows to 46-35%.

Quinnipiac University's most recent national registered voter sample is identical to the racial make-up of the 2012 presidential election (72% white, 28% non-white), making it only slightly less diverse than the one being predicted by Wasserman. Unfortunately, Quinnipiac does not provide the crosstabs on how racial demographic groups plan to vote in 2016. But given what we know, it's safe to assume Hillary's margin over her GOP opponents would grow, albeit ever so slightly, had the sample's racial composition been 70% white, 30% non-white.

Now to be fair, not all pollsters are using less diverse racial compositions than suggested by The Cook Political Report. For example, the latest poll from CNN uses a dramatically MORE racially diverse sample than Wasserman predicts. Why? Two reasons: 1) The survey population is of all adults, rather than registered or likely voters, and thus naturally less white than the smaller pool of 'voters,' and 2) CNN weights their sample of adults to national Census figures on race. The official 2010 Census put the share of national adults that are white at 64%, with 36% identifying as non-white. Some back of the envelope extrapolation verifies that this is indeed the racial breakdown of adults in the CNN survey, making it six points more non-white than The Cook Political Report sees it.

Table only contains polls that 1) make publicly available the racial composition of the survey sample, and 2) ask respondents who they support in the November 2016 presidential election.

Not surprisingly, thanks to the notable lack of white voters in the sample,  Hillary Clinton has massive leads over all of her competitors in the CNN poll. Looking specifically at Jeb Bush, and presuming the CNN survey was re-weighted to match Wasserman's 2016 prediction, we see Clinton's lead falling from 54-41% to 52-42%. Still impressive, but a bit less so.

Table only contains polls that 1) make publicly available the racial composition of the survey sample, 2) ask respondents who they support in the November 2016 presidential election, and 3) provide racial demographic breakdowns of how those respondents would vote.

As you might expect, re-weighting the CNN, PPP, and Rasmussen polls to Wasserman's prediction on 2016 racial turnout, narrows the previously wide disparity in topline results. Before, Hillary's lead over Jeb ranged between 4-13 points. After the re-weighting, the range narrows to 6-11 points.

For the record, there was at least one pollster over the last several months to weight their poll in line with Wasserman's 70/30% white/non-white prediction - the March McClatchy/Marist poll finding Clinton leading Bush among registered voters 49-42%.

For now, the oversight by those pollsters weighting their demographics to pre-2012-like racial identification seems pretty inconsequential to the topline. But as the 2016 election takes shape, and partisan preferences harden to the levels they did in 2012, it makes a difference. Pay attention to whether the racial composition of upcoming surveys fall more in line with David Wasserman's thinking.



Side note: Unfortunately, as you can see in the final table above, very few pollsters publicly provide not only their racial crosstabs, but also ensure those tabs match their registered or likely voter samples used in general election polling. So a big thanks to CNN, PPP, and Rasmussen for making things easy. NBC/WSJ, ABC/WaPo, YouGov, Fox News, and Quinnipiac - get with it!