And their ignorance might keep them from voting.
By Nathan Collins
FiveThirtyEight founder Nate Silver. (Photo: Slaven Vlasic/Getty Images for AWXII)
People are not generally very good with probabilities. Usually, that’s not a big deal. But a new survey hints that our lack of numeracy might indirectly affect the presidential election: Around one in five Americans don’t understand what a politician’s “chance of winning” means, and the nature of their misunderstanding could keep them from voting in what looks to be an increasingly tight race.
The problem stems in part from a shift in how news outlets report polling data. Most of us grew up with polls that simply reported the fraction of people surveyed who said they’d vote for one candidate. But in the last decade there’s been a move toward reporting results in terms of the probability that one candidate or another will win, and those probabilities are almost always numerically higher than the popular vote estimates. For example, FiveThirtyEight’s estimates put Hillary Clinton’s share of the popular vote around 48 percent, but estimate her chance of winning the election at around 70 percent.
But do people actually understand those distinctions? Certainly, there’s plenty of reason to doubt so, Santa Fe Institute Professor Mirta Galesic writes in an email. A 2005 study, for example, found that around one-third of New Yorkers thought a 30 percent chance of rain meant 30 percent of the region would get rain or that it would rain for 30 percent of the day. (In fact, a 30 percent chance of rain means that forecasters have run tens of thousands of computer simulations of the weather on a given day, and recorded rain in 30 percent of them.)
Only 77 percent of those who misinterpreted the data said they’d vote, compared with 87 percent of those who understood what FiveThirtyEight’s numbers meant.
Galesic figured something similar might happen with probability estimates—in particular, people might misinterpret them as predicted vote shares—so she set up a quick survey this past Sunday using that day’s probabilities from FiveThirtyEight and the New York Times’ Upshot blog.
Of the 171 people who took part, only 81 percent correctly interpreted the numbers as the percent chance of winning based on statistical models of the election. Most of the remaining 19 percent misinterpreted the probabilities as estimates of the popular vote, and some thought the numbers referred to how many states Clinton or Donald Trump would win, or else gave other explanations.
And then there’s this: Only 77 percent of those who misinterpreted the data said they’d vote, compared with 87 percent of those who understood what FiveThirtyEight and the Upshot’s numbers meant. That’s consistent with the idea that people are less likely to vote if they think the outcome is already set. In other words, if someone incorrectly thinks that Clinton’s 79 percent chance of winning means she’s going to get 79 percent of the vote—which would be a landslide for the ages—they might not bother voting.
For now, it’s impossible to say for sure whether something like that is actually happening, Galesic writes, “but if there is a chance that people who think that elections are determined are also less likely to vote, this could influence the turnout and the election result.”
The solution, however, could be quite simple: “Much of this is, I think, just ignorance of what the numbers mean. If we were to explain the numbers … more carefully in media reports, the amount of misinterpretations would likely be lower,” Galesic writes.