The Democratic primary elections to pick a candidate for the 2020 US Presidential election commenced yesterday. The media is filled with all kinds of arguments, partly based on data, typically coming from polls.
I'm not going to predict who will win, or who to vote for. (Nate Silver is still the best in class for this, if that's your things.) I like to comment on some refrains that we will hear in the coming months.
Be careful of story time
Every pundit references a poll or two, and everyone spins stories around these numbers. Story time is the moment, after you've been lulled in with some poll data, when the pundit feeds you stories that are not supported by, or have nothing to do with, the data. Be careful of story time. Stay awake!
The "electability" tautology
Electability is not an objectively defined idea. It is an amoeba. The best pollsters can do is to write questions that collect indirect measurement of "it". "It" is ultimately an opinion.
Individual opinions on electability are worthless. It doesn't matter whether I think someone is "electable". Electability takes on meaning only as an aggregate opinion. If a candidate loses in the election, then I suppose s/he is not electable. How is electability different from being elected? It seems to me that you'd need to know the result of an election in order to confidently answer whether someone is electable.
So, electability is the average prediction of whether a candidate will win a future election. Is the prediction of electability not the same as the prediction of the election?
One who is electable will win, because one who will win is electable. Thus, if a candidate says s/he will win because of electability, it's the same as saying s/he will win because s/he will be the winner. It's a tautology.
Of course, one of the more popular poll questions for electability is: "which candidate would you vote for if the election pits A against B?" You might think you're expressing an individual opinion but not making a prediction about other people's opinion. But electability is defined across all respondents, and then generalized to likely voters. When a pollster reports that 52% of likely voters say they would vote for A in a head-to-head against B, this is really no different from a prediction that 52% of likely voters would vote for A.
As the 2016 election demonstrated, it's mighty hard to predict an election accurately. Even when the experts like Nate Silver are accurate, they hang their hats on predictions based on polling data including polls from the days leading up to the election. It's highly suspect that we could collectively predict who would win months before the election. If we can't predict the election, we can't predict electability.
Candidates: please don't tout electability. Voters: please don't vote on electability.
P.S. What about the fringe candidates? That's a good follow-up thought. There's a reason why the electability argument is only used when races are close. In a lop-sided election, the leading candidate would never ask for our votes based on electability. In this case, the election result is easily predictable. It is correct that the leading candidate is "electable". But it's also self-evident so that it's pointless for the presumptive winner to claim electability.
Also, here is Andrew Gelman and Julia Azari's take on the 2016 elections. They mention electability in the following section: "If nearly everyone is voting on party lines, then 'electability' is not such a concern. In the wake of the victory of Donald Trump and the loss of Hillary Clinton, it will be difficult for moderate leaders of either party to persuade primary election voters to set aside their hearts and choose the purportedly safe option."
Comments