In the aftermath of Donald Trump’s surprise election victory, a lot of Hillary Clinton voters bitterly blamed the result on those who had cast “protest” votes for Libertarian candidate Gary Johnson or Green candidate Jill Stein (or for other, even more marginal candidates)—an echo, they claimed, of Ralph Nader’s impact on the 2000 election, where only the tiniest (or, perhaps, the most illusory) of margins in Florida led to George W. Bush’s controversial victory over Albert Gore. In 2016, however, the complaint against third-party voters and candidates is ill-founded: If there had been only the two major-party candidates, there is no convincing reason to believe the outcome of the election would have been any different.
Sixteen years ago, the Florida vote was even more closely divided than it was this year: the final count (which may or may not have been perfectly accurate given ambiguities like the famous “hanging chads”, but was presumably not far off the mark) had Bush ahead of Gore by only 537 votes, out of something like eight million total votes cast in the state. Ralph Nader, running as the Green Party candidate, won almost 100,000 votes in Florida. Given a binary choice between Bush and Gore (who, among other advantages, had solid environmentalist credentials), it would seem overwhelmingly probable that a majority of Nader’s voters would have preferred Gore over Bush; even a 51-49 split of the Nader votes in favor of Gore would have made the Democrat a clear winner in Florida, and thus won him the national election.
Flipping off the States
In order test whether similar thinking is applicable to the 2016 presidential election, I gathered some final vote tallies and ran a fairly simple simulation. This is based on a final Electoral College result of 306 votes for Donald Trump and 232 votes for Hillary Clinton; in order for Clinton to have won in the Electoral College, 38 electoral votes would have to “flip” from Republican to Democratic.
A potentially “flippable” state is one that Donald Trump won by a narrow-enough margin that votes for alternate candidates could, theoretically, have changed the result. By this standard, Ohio, for example, is non-flippable: Trump won the state by 455,000 votes, and only 236,500 people voted for someone other than Trump or Clinton. I found only five flippable states: Arizona, with 11 electoral votes; Florida, with 29; Michigan, with 16; Pennsylvania, with 20; and Wisconsin, with 10. (North Carolina, with its 15 electoral votes, didn’t quite make the cut despite the fact that it had been “supposed” to vote Democratic; Jill Stein wasn’t on the ballot in the state, and even if all Gary Johnson’s voters had instead voted for Hillary Clinton, she would still have come up 50,000 votes short.) Altogether, these five states contributed 86 electoral votes to Donald Trump. (All my vote-counts come from the various state-by-state results published by the New York Times.)
Spoilers and Non-Spoilers
Any contention that third-party votes “should” have gone to a particular candidate must take into account the particular candidates running in the election. Take 1968 as an example: It is extremely unlikely that any significant number of George Wallace’s ten million voters would have instead voted for Hubert H. Humphrey had Wallace not run. Had Richard Nixon narrowly lost that election, he could have made a strong claim that George Wallace served as a “spoiler”; but had Humphrey lost narrowly, he would not have been able realistically to blame George Wallace for his defeat. (In fact, 1968 was something of an electoral blowout even though it was the last time a third-party candidate legitimately won electoral votes: Richard Nixon received 301 electoral votes, Wallace received 46 by winning the vote in five Southern states, and Humphrey won only 191 electoral votes.) By this standard, Ralph Nader was definitely a spoiler in 2000, as his candidacy almost certainly took more votes from Al Gore than from George Bush and thus determined the outcome of the national election.
The two “major” minor-party candidates in 2016 drew their support from different points on the ideological spectrum. Jill Stein, as the Green Party candidate, drew most of her support from environmentalists and other leftists; but Gary Johnson (like his running mate, William Weld) was a former Republican state governor, and explicitly sought support from “never Trump” Republicans as well as from disaffected Democrats and independents. The hypothetical allocation of the Libertarian vote is thus a matter of interest, both because Johnson drew around three times as many votes as Stein and because there is some question as to how Libertarian voters would have voted if they had no options other than Clinton or Trump.
In one of the flippable states, there is an additional wrinkle: Pennsylvania saw 20,896 votes cast for the “paleoconservative” Constitution Party, with Darrell Castle as its presidential candidate. Given its platform, it seems reasonable to allocate all these votes to Donald Trump in our hypothetical two-candidate election—it is impossible to imagine a Constitution Party voter casting a vote for Hillary Clinton, or indeed for any other Democrat.
Running the Numbers
My two-candidate-only simulation thus works as follows: All votes for the Green Party are allocated to Hillary Clinton, along with all unspecified “other” votes. All Constitution Party votes in Pennsylvania are allocated to Donald Trump. We can then calculate what percentage of Libertarian votes would have to be allocated to Hillary Clinton for the state to “flip” Democratic. (The relevant spreadsheet can be found here and the data are given in table form below.) In reality, it is unlikely that 100% of Green and “other” voters would have voted Democratic given a binary choice; some would have voted Republican, and others would have stayed home or expressed no preference in the presidential race. Thus the simulation is intended to be maximally “Hillary-friendly”, minimizing the percentage of Libertarian votes needed to flip a state.
This simulation also assumes that all Libertarian voters, presented with a binary choice between Donald Trump and Hillary Clinton, would have voted for one or the other; this means that for these voters to have any net positive impact on Clinton’s results, more than 50% of them would have to vote Democratic. As with Green and “other” voters, this assumption is not fully realistic: Some Libertarian voters, denied the possibility of casting a protest vote, would no doubt have stayed home or left the “President” portion of their ballot blank. However, given that all alternate-candidate voters clearly knew that their vote would have, at best, a declarative impact, it seems reasonable to assume that they were strongly motivated to cast a vote for somebody; and thus is hard to believe that in a binary election the Libertarian vote would have been split only between Hillary Clinton and “none of the above”. (Gary Johnson’s final vote tally was indeed far below what earlier polling indicated; as had been widely predicted, most of these voters “came home” to a major-party candidate.)
Here, then, are my state-by-state observations, in alphabetical order:
- In Arizona, Donald Trump won by 4.3%, or 84,500 votes. In order to win our hypothetical two-candidate-only race, Hillary Clinton would have needed to receive 90.5% of Gary Johnson’s votes as well as all the Green votes.
- In Florida, Clinton would have needed 57.4% of the Libertarian vote as well as all the Green and “other” votes in order to make up for her 1.3%, 120,000 vote defeat.
- In Michigan, Trump’s margin of victory was only 0.2%—a mere 12,000 votes out of 4.8 million votes cast. Hillary Clinton would thus have needed just 33.2% of the Libertarian vote along with the entire Green vote to win; Michigan is the most flippable state on our list.
- Pennsylvania gave Donald Trump a 1.1% victory, with a margin of 68,000 votes out of 6 million. Allocating all Green Party votes to Hillary Clinton and all Constitution Party votes to Donald Trump, Clinton would need 64.1% of the Libertarian vote to win in a two-party-only Pennsylvania election.
- Finally, Wisconsin was also a close race, with Trump winning by 27,000 votes, 0.9% of the state total. Had Hillary Clinton received all the Green Party votes as well as all 15,500 “other” votes, she would have needed 41% of the Libertarian votes to win. Wisconsin is thus the second-most flippable state on the list.
The last three states—Michigan, Pennsylvania, and Wisconsin—accounted for 46 electoral votes, one sixth of the total needed for an Electoral College victory and more than enough to have made Hillary Clinton President had she won them. Donald Trump won all three by a grand total of 107,000 votes, less than one percent of the three-state total vote.
Does the Flip Theory Fly or Flop?
All this means that if anything up to 57% of Gary Johnson’s Libertarian vote had gone to Hillary Clinton—along with all the Green Party and “other” votes, and with Constitution Party votes in Pennsylvania going Republican—she would have flipped Michigan and Wisconsin, with 26 electoral votes between them. This would have made the electoral vote 280 to 258—a much closer Electoral College result, but still a Trump victory. We need to turn up the Johnson-to-Clinton dial to 58% in order to flip Florida; its 29 electoral votes would then make the Electoral College vote 287 to 231 in Clinton’s favor. At 65% Johnson-to-Clinton, Pennsylvania flips Democratic as well, making Clinton’s Electoral College victory a more comfortable 307 to 231. Finally, if we assume that 100% of the Libertarian vote goes to Clinton in a two-way race, Arizona also goes Democratic and the Electoral College vote becomes a 318-220 near-blowout.
If we want to accuse alternate-candidate voters of being responsible for Donald Trump’s victory, we have to believe that at least 58% of those who voted Libertarian, as well as fully 100% of Green Party voters and “other” voters, would have voted Democratic in these five states given only a binary choice between the two major-party candidates. (More realistically, fewer than 100% of the Green Party and “other” votes would have been cast for Clinton, so 60% or more of the Libertarian vote in these states would have had to go Democratic to give Hillary Clinton an Electoral College victory.) While this is, of course, mathematically possible, it does not seem very probable: Libertarian principles, particularly in relation to economic issues, are more commonly associated with Republicans than with Democrats; and Donald Trump, despite being a major-party candidate, had a large degree of “protest candidate” allure and advocated a more Libertarian-style isolationist foreign policy.
It simply doesn’t seem reasonable to expect that the Libertarian vote, without a Libertarian to vote for, would have come out as a 60-40 landslide for Hillary Clinton; in fact, a Libertarian landslide against her is much more credible. Lacking convincing evidence that a large majority of Libertarian voters disliked Donald Trump more than they disliked Hillary Clinton and would have (or “should” have) voted Democratic, I believe that we must search for our culprits elsewhere.
|Arizona||Florida||Michigan||North Carolina||Pennsylvania||Wisconsin||Hyp. Flipped EV||EC Votes – Clinton||EC Votes – Trump|
|Trump margin %||4.3%||1.3%||0.2%||3.8%||1.1%||0.9%|
|Hyp. Trump margin w/o
|Johnson votes needed||67,956||118,147||57,506||152,662||91,437||43,599|
|Johnson % needed – all
|Johnson % needed – no