As more and more data gets released about the 2016 election, the decisive factor for Trump occupying the White House remains elusive (was it more Putin or more Comey?). However, we can begin to eliminate some reasons people have given. Several articles in the immediate aftermath placed the blame on Libertarian Gary Johnson and Green Jill Stein (see here, here, here, here, among many others). On the surface, it does look like they harmed Hillary Clinton. Both Johnson and Stein ran in 2012 and received 1.35% of the vote (.99% for Johnson, .36% for Stein). In 2016, they each tripled their support (3.27% for Johnson, 1.06% for Stein). Trump got a smaller share (45.94%) than Mitt Romney (47.15%), but Hillary dropped by even more (48.03% compared to 51.01% for Obama). Thanks to the Cooperative Congressional Election Study (CCES), we can examine the minor party vote with more rigor. No matter how you cut it, Johnson, Stein, McMullin, and the rest did not make a difference. They ended up masking more Trump supporters than ones for Hillary.
Trump gained disproportionate support in the final weeks of October. The CCES interviewed people both before and after the election. Taking those who said they are registered to vote and did surveys before and after election day (44,667 total interviews), Trump gained seven points while Hillary gained just three. Because of the wisdom of several old, rich, white slaveowners 230 years ago, that was enough for Trump to “win”.
Trump kept a greater share of his voters than Hillary did her. More than triple the share of initial Hillary supporters went for Trump than vice versa (2.6% vs. 0.8%). Half of initial Johnson and Stein supporters went for someone else. Johnson voters tilted more towards Trump while Stein voters went for Hillary. Those undecided went for Trump by a 50.0% to 34.2% margin.
Using Random Forest regression, we developed a Trump/Clinton model based on the demographics and responses in the poll. We trained the data using a random selection of 15% of the poll respondents and applied it to the remaining 85% (click here for data and R code). This gives us a probability of voting for Trump and Clinton for each participant. The model works well in classifying voters.
Assigning the candidate with the higher probability as the predicted vote, the model gets 95.8% of Clinton’s voters and 94.2% of Trump’s voters correct (the dots above the diagonal are mostly blue and the ones below red). Among Johnson’s voters, 33.7% have a higher probability of voting for Clinton and 66.3% for Trump (note the higher density of observation below the diagonal line).
Among Stein’s voters, 81.5% have a higher probability of voting for Clinton and 18.5% for Trump. This level of support for Trump among Stein voters should not surprise given the hostility her prominent supporters had for Clinton (see here, here, here) and her ties to prominent Trump surrogates:
Moreover, 12% of Bernie Sanders primary voters went for Trump, as did 11% of Obama 2012 voters and 5% of those who consider themselves ‘liberal’.
Evan McMullin (81.8% predicted Trump, 18.2% predicted Clinton) and those supporting other candidates (63.8% predicted Trump, 36.2% predicted Clinton) also tilt heavily towards Trump.
Allocating the minor party vote to Hillary and Trump results in Trump expanding his margins in the key swing states of Wisconsin, Michigan, Pennsylvania, and Florida.
Whatever Hillary would gain from Stein voters, she loses far more with the other candidates removed. But what if we just get rid of Stein? If we allocated just Stein voters and not anyone else, Hillary still loses. Even though Stein got more votes than what Trump won by in the decisive states of Wisconsin, Michigan, and Pennsylvania, there is no data that justifies 100% of Stein voters going to Hillary. If Hillary received 81.5% of Stein voters, she only flips Michigan.
Given the attitudes of voters that we can examine more deeply, we can’t justify “blaming” Gary Johnson, Jill Stein, Evan McMullin, and other people either on the ballot or written in for Trump. The Orange Idiot would have received 270+ Electoral Voters in both a two-party race and one without Jill Stein.