Thursday, August 26, 2010

Swallowed Up

For years I was annoyed at pundits who believed in hunches more than numbers. Or who'd jump on a single poll, misinterpreting noise as "momentum." And if you scratched most so-called experts they turned out to be partisan boosters. Which is why a site like FiveThirtyEight was so helpful--one of the few places to give clear-eyed projections backed by solid analysis.

Nate Silver, who created the site, explains how it worked in the last election:

The FiveThirtyEight forecasting model, which was based on a rigorous analysis of polling in past presidential elections, gave Mr. Obama [...] a 98.7 percent chance [of winning] by Election Day. And yet, on the weekend before the election on the television program, “The McLaughlin Group,” three of the five pundits suggested that the election was “too close to call,” and one other said she expected a narrow McCain victory.

[....] In baseball (which I covered prior to politics), “intangibles” like clubhouse chemistry are sometimes treated as being more important than batting average, or E.R.A. But you wouldn’t find very many sportswriters who would claim, in a game in which the Yankees were trailing Boston 7-2 in the 9th inning, that it was “too close to call,” no matter how shaky the Red Sox bullpen looked, or how confident Mark Teixeira seemed at the plate. That’s the equivalent of what those pundits were doing on The McLaughlin Group.

Instead, there seems to be something about politics that can make the rational parts of the brain turn off. FiveThirtyEight was designed to be the antidote to that. [....T]he blog is devoted to the rational analysis of politics, and sometimes other data-rich subjects. In Congressional and presidential elections — for which there is a lot of high-quality data available — this will sometimes take the form of quite explicit forecasts, [....] In other cases, it simply means trying to prioritize objective information over subjective information in dealing with issues in the news.

This is what makes his site worthwhile.  So I have to admit I'm a little nervous that it's now become part of The New York Times.  Nate Silver is a partisan himself, but he keeps those politics out of his analysis.  I hope there's no institutional pressure to align his work with Times' editorial policy.

5 Comments:

Anonymous Anonymous said...

In the overall predictions department he's good, but Nate Silver regularly editorializes like everyone else, and even sometimes uses statistics that he favors to fight for his side.

2:15 AM, August 26, 2010  
Anonymous Denver Guy said...

I like 538, but recently, in preparing rankings of national pollsters, Nate Silver seems to have allowed his partisan thoughts to interfere with a pure analysis. He has introduced into his rankings a score for how transparent a pollster is (ie, whether they share raw data and explanations of their methodology). While that information may be interesting in deciding whether to trust a pollster, the quality of their results should overwhelmingly be decided by their accuracy, imho.

The result of this technique has been to lower Rassmussen a few pegs in the overall rankings. Rassmussen clearly has a Republican lean, but their ultimate results have been very good for several election cycles. Still, Rassmussen is ranked in the top 15 I think, and in the end, Silver's rankings are mostly for viewers of his web site. The media usually sites the most recent polls without much explanation of who the pollster was.

8:27 AM, August 26, 2010  
Anonymous Lawrence King said...

His Senate election forecast is amazingly detailed. I will be very interested in how accurate it is in November.

He brags that he predicted all the Senate winners in 2008. Yes, that is impressive. However, since there are usually less than half a dozen Senate seats that seem seriously up-for-grabs this close to an election, it's not statistically that astounding to pick the winners. What he does not state is how accurate he was overall in predicting the vote percentages.

There's been a lot of hype about this election. Scott Brown won in January 2010, and his victory was declared by most of the Right to be a harbinger of November 2010. Rational voices (this blog among them) pointed out that it was still almost a year before the election. But since then, the polls have been in some very non-traditional points. The approval rating for the Democratic Party has plummeted, and yet the approval rating for the Republican Party has remained as low as it was a couple years ago -- indeed, it's still lower today than the Democrats have been.

Based on voter turnout (e.g., the Florida primary this week) I would definitely bet on big Republican gains. But what happens then? The GOP didn't put together any kind of Contract With America this time (either due to disorganization, or to internal disagreements, or because the leaders are content to win votes as the Opposition Party and have no actual goals). Right now, the Tea Party, the social conservatives, the Republican establishment, and the foreign policy hawks are united -- by their disdain for the Obama administration. But can that actually be turned into a party with a goal? If not, won't the Republican approval ratings stay in the gutter?

On the political level, it's not an unreasonable prediction that a 2010 Republican takeover of the House would actually help Obama's chances in 2012. On the other hand, if the econmy stays bad, and the GOP doesn't nominate someone crazy, I certainly wouldn't bet in Obama's favor for that election.

In other words, 538 works because people change their minds slowly enough so that ten weeks of change isn't much. But this doesn't really help us predict anything about politics in 2011... which will be an interesting year. Peaceful pollsters, pls refudiate.

11:12 AM, August 26, 2010  
Blogger LAGuy said...

FIveThirtyEight in its analysis (as opposed to its editorials) understands polls much better than the general punditocracy. They also understand certain effects--for example, the bump after a convention usually fades before the election. They also understand math beats hunches.

I think it's true that, though the Republicans are looking pretty good right now, there are internal tensions (particularly between party regulars and the tea party). On the other hand, they're not as important as the general and fairly overwhelming opposition to Obama (not unlike the Dems opposition to Bush). Some suggest Repubs will do even better after Labor day when people return to the their parties and independents look at the unemployment figures and drift rightward. I guess we'll see.

A word about the Contract With America. It was a collection of popular ideas tied together in a package, but there are many who question if it really made any difference. Were voters even that aware of it, or were they just tired with two years of Clinton? Certain beliefs pass into popular culture, relatively unexamined. (Another example is if Perot had not run in 1992 Bush would have won.) This is where sites like FiveThirtyEight can be useful--they can actually plow into the numbers and give you a reasonably rigorous interpretation, not just a feeling. Numbers can't tell you everything, but they can help a lot.

12:06 PM, August 26, 2010  
Anonymous Lawrence King said...

Certain beliefs pass into popular culture, relatively unexamined. (Another example is if Perot had not run in 1992 Bush would have won.) This is where sites like FiveThirtyEight can be useful--they can actually plow into the numbers and give you a reasonably rigorous interpretation, not just a feeling. Numbers can't tell you everything, but they can help a lot.

The most extreme -- and absurd -- example is the claim that Mayor Daley's rigging of the vote in Chicago in 1960 is what put JFK over the top. "Richard Nixon's victory is sitting at the bottom of Lake Michigan," Daley is said to have boasted. I have heard this claim many times in the past few decades.

FALSE. A fifteen-second check of the numbers will reveal that even if Illinois had voted for Nixon in 1960, Kennedy still would have had a majority of electoral votes.

5:47 PM, August 26, 2010  

Post a Comment

<< Home

web page hit counter