At one point in the early summer of 1980, Anderson was in a statistical tie with Jimmy Carter and Ronald Reagan, according to the Gallup poll. As voters began to plug into the campaign, his numbers fell. A similar phenomenon happened to Ralph Nader in 2000. In many polls during that cycle, Nader was polling in the high single digits. His numbers sank as election day approached, and he underperformed almost all of the final polling estimates.

The cases of Anderson and Nader are consistent with what Duverger would predict: as voters really start to engage in the process, they become aware that only one of two candidates can realistically win, and most choose not to “waste” their vote. Again, this is not an ironclad law, just a tendency. And to counteract it, you usually need a substantial, counteracting force to push voters in the opposite direction—like a TR in 1912 or a Perot in 1992.

There is nothing about Johnson or Stein that is nearly so interesting. After all, the two of them ran in 2012, and combined they accounted for just 1.35 percent of the vote. The only significant force this cycle is the widespread disdain that voters have for the major-party candidates. People really dislike Trump and Clinton, so it is possible that they stick with Johnson and Stein. This would be unprecedented, but—weirdly enough—be typical of the 2016 campaign, which seems insistent on breaking every available precedent.