Since the election, various people have been pushing a narrative that Gary Johnson’s candidacy hurt Hillary and helped Trump get elected. This is consistent with the narrative that some were pushing before the election to the effect that Gary Johnson’s candidacy was bad for Hillary. This doesn’t seem to have been true then, and it isn’t true now. Gary Johnson’s candidacy helped Hillary; Democrats’ efforts to delegitimize Johnson hurt Hillary; and on Election Day, if Gary Johnson hadn’t been in the race, Trump would have won by more.
Jacob Levy has written about this before, and I linked to his analysis in a previous post called “Gary Johnson is slightly good for Hillary“. Later data has been consistent with this earlier story, so please read the earlier post as well. In addition, Jacob has also posted the following on Facebook and has given me permission to reprint it, so here goes.
1. The 9% of Democrats or 10% of liberals who voted for Trump are a larger group than Johnson and Stein voters put together.
2. Clinton has 47.7% of the popular vote. So let’s hold her constant by comparing the outcome to the day when that was her projected vote share at 538, October 3.
Oct 3 projection, 47.7 Clinton, 43.9 Trump, 7.0 Johnson
Outcome: 47.7 Clinton, 47.4 Trump, 3.3 Johnson
Comparing those two, Johnson lost 3.7. Trump gained 3.5. That looks an awful lot like the hypothesis “Johnson’s support was heavily soft-GOP, and it went home.” The reasonable guess about what would have happened if Johnson’s support had fallen further is a larger net gain to Trump (unfortunately).
3. The final 538 poll average the day before the election was:
Clinton 48.5, Trump 45.0, Johnson 5.0
Outcome: Clinton 47.7, Trump 47.4 , Johnson 3.3
What did the polls get wrong? Clinton -.8, Trump +2.4, Johnson -1.7. There’s no way to read that as “Johnson’s last-minute loss of a third of his support helped Clinton.” It can only be interpreted as “Johnson’s last-minute loss of a third of his support helped Trump.” It’s absurd to look at that and think “if only Johnson had fallen to 0, Clinton would have won.” You an explain a lot of the difference between the polls and the outcome by saying “at the last minute, undecideds and more a third of Johnson’s support broke for Trump.”
4. Exit polls: Johnson drew 2% of Republicans, 1% of Democrats, and 6% of independents. Independents split 48-42 for Trump. If we imagine Johnson away and send all those voters home (taking into account the different sizes of the three party ID groups) the final result is +.4% to Trump. Put another way, even at the end Johnson held Trump’s vote down by .4%. Not enough. But that was the direction of the effect.
5. A calculation has been going around claiming to “prove” that Johnson and Stein cost Clinton the election. It’s been picked up and propagated by outlets from CNN to Vanity Fair. Its method is to allocate 100% of Stein’s vote and 50% of Johnson’s to Clinton, and show that she would have won under those conditions. 50% is the nod toward recognizing that not all of Johnson’s voters were plausible Clinton voters.
What happens to the other 50% of Johnson’s voters, you ask? They don’t say, presumably because they know it would sound absurd. It stays with Johnson. The counterfactual here is “what if a bunch of third party voters defected to Clinton, and none defected to Trump?” It is, to use an unfortunate word, rigged. If instead you send Stein’s voters entirely to Clinton and split Johnson’s 50/50 between Clinton and Trump (which is still unrealistically optimistic, given 2-4), it’s not enough. Clinton still loses.