We are only a few days past Election Day and with votes still being counted, it is going to take a while to get the full picture of what happened. But what is not in question is the fact that the view on who would ultimately win the Presidential race was wrong. We viewed Hillary Clinton’s election as a foregone conclusion; our assurances to friends and family that they did not have to worry about a Trump victory will stay with us for a while. With that in mind, we must take a hard look at the data and figure out if we missed something in the data itself, or find out if we just never have the right data to look at. Either way we have egg on our face, but if we don’t learn what happened, we are doomed to repeat it.
This data “autopsy” is going to have to be spread out, but our focus in this round are the Exit Polls and the national results.
As we attempt to put the pieces together, there are several demographic factors highlighted by exit poll data that may help us start to understand Trump’s shocking victory. According to the exit polls, Hillary Clinton beat Donald Trump by a 1-point margin (48% to 47%). As of the writing of this piece, Clinton is leading Trump by 0.3% and estimates have her ending up winning by 1-2%. Given this fact, we have no reason to feel that the exit polls are painting an inaccurate picture of what happened nationally.
Looking broadly at gender, we see a slightly higher proportion of men in 2016 (48%) over 2012 (47%). And obviously, the opposite is true among women as just slightly less turned out to vote in 2016 (52%) than 2012 (53%). Among the two cohorts, Clinton’s support among men was 4 points lower than Obama’s in 2012 (41% vs 45%). Seeing how Clinton was a historically significant female candidate and Trump has been singled out by evidence of harassment towards women, we expected Clinton to do better among women than Obama did. However, Clinton did 1 point worse than Obama among women overall (54% to 55%). At the same time, Trump did not see major success among women, managing to capture 42% of their votes overall, only 2 points lower than Romney (44%).
Moving to ethnicity, Trump did well with white voters, taking 58% of their vote, just about the same percentage as Romney in 2012 (59%). He promised to bring back industrial jobs, relied often on racially-charged rhetoric and messages of America’s downfall under the Obama administration. Arguably, his campaign was focused on appealing to white voters, many of whom felt disenfranchised and overlooked by eight years of Democratic leadership among other changing cultural attitudes. With this in mind, it is noteworthy that he actually performed worse among whites overall than Romney.
In terms of voter turnout, things shaped up just as we had predicted. 70% of voters who turned out to vote were white in the exit polls, down from 72% in 2012. Given these numbers, it does not appear that Trump’s victory was based on a surge in “angry” white voters. In fact, the percent support he received from white voters was on par with George W. Bush in 2004.
On the flip side, Clinton received the lowest level of support from white voters going back 12 years. 2 points is not dramatic, but given the fact that 7 in 10 voters were white, each percent can have a dramatic effect on the result.
Looking more closely at white voters by gender, Trump came in at one-point higher (63%) among white men than Romney (62%). Meanwhile, Clinton received 4 points lower than Obama in 2012 (31% vs 35%). The 4-point difference between Clinton and Obama among white men is certainly noteworthy but interestingly, only 1% of voters in this cohort moved to voting for the Republican.
Although he didn’t receive quite as much of the vote among white women when compared with Romney (56%), Trump still managed a majority (53%) of their votes. Clinton barely passed Obama with white women, receiving just one more percentage point in support (43% to 42%). Historically, white women have favored the Republican candidate. Indeed, since 2004, most white women have voted Republican; 55% in 2004, 53% in 2008 and 56% in 2012. The only exception to this trend was in 1992 (41%) and 1996 (43%), when less than 50% of white women voted for the GOP candidate (Perot clearly effected the numbers).
As has been written about previously, Obama received historically high levels of support among minority voters. In 2016, we expected Clinton to do as well among minorities and Trump to do even worse than Romney given the rhetoric coming from Trump and his supporters. However, this did not end up being the case.
Referencing the table above, we can see just how much Clinton under-performed when compared with Obama in both 2008 and 2012. Among minority voters overall, Clinton dropped by 5 to 6 points from Obama in 2012, depending on which group we look at. The most surprising drop was among Hispanic voters where Clinton saw a 6-point drop from Obama.
As surprising as Clinton’s drop may be, perhaps even more surprising is Trump’s surge among minority voters when compared with Romney. Trump manages to pull a few points higher than Romney with minority voters. While this increase is not necessarily dramatic in the big picture, it is a bit of a shock considering Trump’s alienation of minority voters and outlandish proposals when it comes to immigration and other major issues.
When it comes to race and gender, we can see perhaps the most dramatic drop for Clinton. In every breakout except for white women, among whom she only pulls a 1-point advantage over Obama, Clinton falls short of Obama’s numbers in every group. African American men and Hispanic women had especially dramatic drops for Clinton. As was the case with white voters, Trump did not benefit fully from this drop, although he did see growth among all groups except for white women, and Hispanic men.
Non-college graduates swung rather dramatically towards Trump (52%) when compared with Romney (47%). This difference is even more extreme when it comes to white non-college graduates. According to Pew Research, Trump out performed Romney among non-college whites by 6 points, winning two-thirds of their vote (67%) compared to 61% in 2012 and 58% in 2008. When broken up even further, Trump won 72% of the vote among white non-college men. We expected this to be a strong demographic for him. What was less expected is Trump’s success among white college educated women. In fact, there was only a 6-point difference between Clinton (51%) and Trump (46%) among this cohort.
Also of note is that in 2016, half (50%) of the vote was made up of college graduates as compared to 47% in 2012. This is one group in which Clinton has picked up momentum from Obama in 2012, capturing 52% of their votes compared to his 50%. On the other side, non-college graduates were a smaller makeup of the 2016 (50%) vote when compared with 2012 (53%). Here, Trump sees a significant surge in support compared to Romney. Even though non-college graduates made up a smaller proportion of the vote, Trump won their vote by 5 points (52%) when compared to Romney (47%).
Looking at age, we see a relatively consistent voter makeup from 2012 to 2016, apart from slightly less voters ages 30-44 and slightly more voters ages 45-64. Age is one breakout where Trump does the same or worse than Romney among all groups except those ages 45-64, where he gained 2 points from Romney (53% to 51%). However, Clinton under performs Obama in all age groups except she has a 1-point advantage among those who are 65 and older, 45% to 44%.
Obama famously mobilized and appealed to young voters, leading to record breaking turnout. This year, Clinton directed many funds and efforts toward young voters, whether through social media outreach or her appeal to young women. It turns out that while Clinton won with young voters over Trump (55% to 37%), she lost a significant 5 points from what Obama had among young people in 2012 (60%).
In May, we discussed two possible scenarios for a Trump victory. The first involved Trump excelling among white voters. More specifically, we came up with a hypothetical scenario in which we gave Trump 67% of the vote among white men and that he would match Mitt Romney’s share of the minority vote at 17%. In this scenario, he would have had to receive at least 60% of the white female vote. We factored in poor favorability ratings to predict what we thought would be Trumps’ failure to capture the vote among white women. In short, we thought that there simply were not enough of his target voters, whites and particularly white men to get him elected. We pointed to the recent historical decline of white voters as a percentage of the electorate, and ultimately declared this scenario unlikely.
In the second scenario, we said that if Clinton is unable to hold onto the share of votes that Barack Obama received in 2012, the election would be in Trump’s favor. We also assumed this scenario to be unlikely, but did not rule it out as impossible. Obama’s historic campaign shattered voter turnout especially among minorities and Democrats overall. We can imagine a lack of enthusiasm and even anger with the gridlock in Washington, as potential factors for Clinton’s inability to fully capture or exceed the Obama coalition of voters. However, for now these ideas are just speculation
Looking at 2016 exit polls, we can see that while Trump did not reach our projected levels of necessary support among while voters, he did do well. At the same time, Clinton under performed Obama in almost all ethnic cohorts. As it turns out, a hybrid of our two scenarios ultimately tipped the election toward Trump. An outcome we did not view as likely but so far appears to have been the case.
So far we know that the general percentages in the makeup of the electorate as a whole were about what was expected. And overall voter turnout is not pointing to a large surge in voters that carried the day for Trump or a huge drop off nationally. Given these factors it is not surprising that national polling was generally on target, predicting a Clinton win in the popular vote. However, it does not answer the questions of what happened. In our next phase, we are going to look at these numbers state by state to see if we find anything there to give us more insight into this question.
For a PDF version of this memo, please click here.
 Roper Center for Public Opinion Research. http://ropercenter.cornell.edu/polls/us-elections/how-groups-voted/
 New York Times Election Results. http://elections.nytimes.com/2008/results/president/national-exit-polls.html