Nate Silver: Smart, but not a social scientist

As a political scientist, it's often frustrating to see Nate Silver being treated as an expert in the quantitative study of politics. Silver is obviously very bright, but he is a blogger who can run regressions and make charts, not a trained social scientist. As a result, while I like his energy and his quantitative approach, I typically find his analyses to be ad hoc and lacking any grounding in previous research.

Here's a recent example. A few days ago, he wrote a post that compared the voting records of Senator Arlen Specter and Rep. Joe Sestak, who might challenge Specter in the 2010 Democratic Senate primary. To do so, Silver used the DW-NOMINATE scores developed by Keith Poole and Howard Rosenthal. Specifically, he compared the post-switch DW-NOMINATE score he projected (rather arbitrarily) for Specter with Sestak's DW-NOMINATE score in the 110th Congress. The problem is that DW-NOMINATE scores are not comparable between chambers, which Silver would know if he had used them for research purposes. (Poole has developed other algorithms which generate scores that can be compared across chambers.) Even if you accept Silver's questionable projection methodology, the whole exercise -- which was cited by Josh Marshall -- is simply invalid.

If you like Silver's work and are interested in analytical blogging about politics, I would also recommend the work of top political scientist bloggers like Charles Franklin, Nolan McCarty, Andrew Gelman, and the folks at The Monkey Cage.

Comments

[David]

Despite this, Silver's comprehension of statistics remains far beyond the average commentator, and he's certainly much brighter as well. The punditocracy could certainly do worse than elevate to Nate's analytical abilities.

[tofias]

@David,

Are junk stats better than no stats?

I'm all for the quantitative analysis of politics. And I'm all for Nate Silver. When I consume his baseball analysis, I don't care that the work isn't peer reviewed and he doesn't show me what's going on in the black box.

Silver can't approach political analysis the same and and get a free pass just because he's smarter than most pundits.

Well he can, but he just risks coming off as a hack like John Lott. He's passing infotainment off as a science.

[Jinchi]

I don't care that the work isn't peer reviewed ...

A peer reviewed blog?

I think you and Brendan are missing the point here. Nate writes commentary on a blog discussing political trends and he does a fairly comprehensive analysis to back up his posts. I don't know that he's ever claimed to be a social scientist.

More importantly, he discusses polling at a much more sophisticated level than virtually all of the news organizations do on a routine basis. What bothers me isn't someone like Nate putting forth his understanding of social dynamics, but rather the incessant articles with titles like "Is Hillary Unstoppable" or the obsession over which candidate had "momentum" during the campaign.

[TB]

Certainly Silver's analyses could be better, and I'm sure they would be if he spent several years in a quality Ph.D. program. But they are still pretty decent, and quite transparent, which is more than we can say for the work of many if not even most political scientists churning out research today. The question is, why can't we get more political scientists doing work like Silver's (in substantive/topical terms) such that it might be more present and useful in political debates and decisions? Joe Sestak and other PA Democrats, for example, could really use some quality predictions right now about what Specter's voting may look like in the next year and a half and how he might compare to Sestak in that regard. A peer-reviewed article on the subject that comes out in 2012 is not going to be particularly helpful.

[Theda Skocpol]

I agree with the overall point but, frankly, one does not have to be a quantitative political scientist to know that Silver's comparison of Sestak and Specter is flawed. It was apples and oranges, because Representatives who reflect voters in their smallish districts have much room to move when they run for the Senate in a large, diverse state. So even if the nominate scores were comparable, they would not tell us much about how these two would compare in state-wide representation. And of course, Specter is leaving plenty of room for a competitor in his primary to the moderate left.

I do admire Nate Silver. But he got this one all wrong.

[beowulf888]

I have trouble with political scientists passing themselves off as scientists. There's little science (i.e. a cohesive and testable theoretical framework that meets Popper's criteria of falsifiability).

[beowulf888]

CORRECTION: I deleted a segment of my text by mistake. My post should have read: "I have trouble with political scientists passing themselves off as scientists. There's little science in political science (i.e. a cohesive and testable theoretical framework that meets Popper's criteria of falsifiability).

[paul]

So what happens if you use the numbers that are comparable across chambers?

[Woodrow L. Goode]

I find this post amusing, coming, as it does, from someone who made what little reputation he has posting media criticism, even though he has absolutely no credentials in journalism or mass media.

Beowulf888's point is also well taken. The commonality between a political scientist and a creation scientist is that they're both misapplying the word to establish credibility. Mr. Nyhan, if you could show me something from your field that qualifies as science-- as opposed to, say, reverse engineering of data-- I'd be interested to hear it.

[Lisa]

The thing Nate gets right is that he writes about concepts in a way which is accessible.

Too many ivory tower academics can't do that, and therefore, their brilliance goes unread by the average political junkie. It's no good if you have all the know-how, but lack the means to communicate it.

Rather than bemoan Silver's site, why don't you find someone who can hook up your knowledge with pithy writing? Maybe you could fill a niche?

Also, I must agree with what some of the others here say - political science is science in name only. It's almost akin to voodoo. Nate happened to have a good grasp of 2008's dynamics. Maybe he'll keep it up, maybe he won't, but no matter, his writing is insightful and INTERESTING.

Also, he has said several times upfront, yes, he's partisan.