Monday, 24 March 2014

Nate Silver's 2014 election predictions: they do exactly what they say on the tin


Nate Silver (above), that famed statistical guru of the last two US Presidential elections, is back at the forecasting now - predicting that the Republicans have the edge in this year's race for control of the Senate. Some Democrats are holding their heads in their hands as if an all-seeing seer just pronounced on their doom.

Now Mr Silver is getting quite a lot of flak about this - again. Other crystal ball gazers basically prefer to look at the polls when they come out with actual numbers attached to each party's chances - and when, quite frankly, we're a bit closer to the November elections, and we know exactly who is running aginst whom. At the moment, Mr Silver's model is entirely model-driven, based on the past performance of such indicators as incumbency, general opinion polls, the economy, and the Democratic- or Republican-leaning nature of each state Senate contest. Competing experts, quite naturally and rightly, have their own approaches.

But you know what? Mr Silver is doing exactly what he's always said that he's doinng - looking at the odds, and the averages, on past performance. It's a use of data that he's run to fit past performance and, though he might be out a bit on either side (quite a bit, at this juncture), he's not saying that he's got some passkey to the truth. Have a listen to what he says in his 2012 book, The Signal and the Noise:

Pure objectivity is desirable but unattainable in this world. When we make a foreacst, we have a choice from among many different methods. Some of these might rely solely on quantitative varibales like polls... [while others] may consider qualitative factors as well. All of them, however, introduce decisions and assumptions that have to be made by the forecaster. Wherever there is humanjudgement there is the potential for bias. The way to become more objective is to recognise the influence that our assumptions play in our forecasts and to question themselves about them.

Now have a look at the page on his relaunched website: It's all couched in the language of might, maybe and possibly, not will, should and ought. The hidden wiring is all there, and it's still there, front and centre, in the conclusion: 'our forecast might be thought of as a Republican gain of six seats — plus or minus five. The balance has shifted slightly toward the GOP. But it wouldn’t take much for it to revert to the Democrats, nor for this year to develop into a Republican rout along'. This kind of probabilism shows just how difficult it is to watch time's arrow onto the target, but it also tells us exactly the odds. Right now, control of the United States' upper chamber depends on Kay Hagan's 50:50 run in North Carolina, and Mary Landrieu's 45:55 (against) chances of hanging on in Louisiana. They might do it, of course: many vulnerable Democrats clung on in 2010, when their fate looked sealed. But now we've got some numbers about their chances.

Now lots of sites that look at different types of data - including interviews with insiders - could tell you just about the same thing, despite Mr Silver's disdain for the partial nature of their mere 'opinions'. Control of the Senate rests on a knife edge: that's the headline. That's all we know right now right now.

But what Nate's website at FiveThirtyEight does is give us some benchmarks, which we can come back to in a month, two months and three months. Then we'll see how the race is shifting - at least as measured against Nate Silver's model.

Isn't that worth a few number-driven speculations about the underlying nature of electoral choice?

No comments:

Post a Comment