Princeton Election Consortium: Clinton had a 99 percent probability of winning [With Comment by John]

On the eve of the 2012 election, I viewed the Obama-Romney as a toss-up that the challenger would, slightly more likely than not, win. After all, the polls showed the president up by only around 1 point nationally and late deciders are thought generally to break against the incumbent.

When I expressed this view to members of the smart set — both just before and just after the election — I received quizzical looks or worse. Didn’t I read the New York Times? Didn’t I know that Nate Silver had put the probability of an Obama victory at around 90 percent? What was I, some kind of science denier?

One of the fringe benefits of the 2016 election is the end, I hope, of such hubris. This time around, Silver put the likelihood of a Clinton victory at around 70 percent.

And Silver was “conservative.” Most of the other big-data election modelers thought the likelihood of a President Hillary Clinton was significantly higher than 70 percent. Indeed, the Princeton Election Consortium (PEC) said, and argued at great length, that the probability of a Clinton win was between 98 and 99 percent.

For a good laugh, read the PEC’s analysis.

All in all, Silver and his FiveThirtyEight crew didn’t have such a bad election. Clinton will end up winning the popular vote, as he predicted, and Silver was one of the few analysts I’m aware of who talked up the chance that the electoral vote would go one way (in favor of Trump) and the popular vote the other.

Moreover, Silver tried pretty early on to persuade liberals that Trump had a realistic chance of winning (he received some abuse for his trouble). He also explained more than once that events with a one-third chance of occurring actually occur pretty often. (Is anyone surprised when the roll of a die produces something other than a 3, 4, 5, or 6?)

At the end of the day, though, Silver failed accurately to pick the outcome of the election. He is not the master of the election universe, after all (nor, unlike some of his admirers, did he claim to be, as far as I know).

I never viewed Silver as a master of the universe. I did conclude that his sophisticated analysis of poll data is more reliable than my casual analysis coupled with a failing intuition about the electorate. I still think it is (my intuition was that Trump would lose). Thus, I will continue to look to Silver in these matters, along with Steve’s friend Henry Olsen who also saw the real possibility of a Trump victory.

But the main point is that predicting the outcome of presidential races is more difficult than the smart set thought. The new generation of data-crunching gurus, all or nearly all of whom seem to be leftists, aren’t infallible and probably aren’t all that brilliant.

Disagreeing with their predictions doesn’t make one a science denier. It may be that folks with a good sense of the electorate — a cohort that no longer includes me — are as likely to call elections correctly as the new gurus, provided they pay attention to the data.

Good to know.

JOHN adds: Nate Silver is a mediocrity with a passable knowledge of statistics. But number crunching is the easy part: the hard part is knowing whether you have good numbers in the first place. Silver wrongly attacked something I wrote years ago, based on an embarrassingly foolish crunching of numbers that he didn’t understand.

There were, of course, a few pundits who, unlike Silver, predicted that Trump would win. Like me.

Responses