Michael Gerson’s Latest

by John Sides on November 5, 2012 · 15 comments

in Political Science News

Gerson’s latest column contains a strange critique of quantitative approaches to elections and political science more generally.  Talking about election forecasting and polling aggregation, Gerson writes:

The main problem with this approach to politics is not that it is pseudo-scientific but that it is trivial. An election is not a mathematical equation; it is a nation making a decision. People are weighing the priorities of their society and the quality of their leaders. Those views, at any given moment, can be roughly measured. But spreadsheets don’t add up to a political community. In a democracy, the convictions of the public ultimately depend on persuasion, which resists quantification.
Put another way: The most interesting and important thing about politics is not the measurement of opinion but the formation of opinion. Public opinion is the product — the outcome — of politics; it is not the substance of politics. If political punditry has any value in a democracy, it is in clarifying large policy issues and ethical debates, not in “scientific” assessments of public views.

There are two false dichotomies here.  One is the measurement of opinion vs. the formation of opinion.  In political science, we do the former so we can do the latter.  Indeed, although forecasting models and polling aggregators take centerstage in an election year, far more political science research actually focuses on the formation of opinion.  In fact, far more political science research about elections is about how voters make decisions—you know, in books like How Voters Decide—and not about averaging polls.  And it’s about how voters may or may not be persuaded by the arguments of candidates—you know, in books like The Message Matters.  (And, yes, persuasion can be quantified.  It’s attitude change in the face of new information.  You can measure attitude change and information with quantitative data!  I don’t know why Gerson thinks otherwise.)

The second false dichotomy is between empirical inquiry and normative argument.  Gerson says that political science, like the rest of social science, cares only about empirical inquiry—it has “precision envy”—and not about values.  Gerson wants pundits to avoid this mistake:

Over the past decade, there has been a revolt among political scientists against a mathematical methodology that excludes substantive political debates about justice and equality. A similar revolution is increasingly needed in political commentary. The problem with the current fashion for polls and statistics is that it changes what it purports to study. Instead of making political analysis more “objective,” it has driven the entire political class — pundits, reporters, campaigns, the public — toward an obsessive emphasis on data and technique. Quantification has also resulted in miniaturization. In politics, unlike physics, you can only measure what matters least.

That’s funny, because when I read political science, I see a lot of interest in topics like justice and equality.  Here’s a book that’s primarily quantitative, about the formation of opinion, and about justice: Jon Hurwitz and Mark Peffley’s Justice in America: The Separate Realities of Blacks and Whites.  Here’s Larry’s Bartel’s book about (in)equality.  And Martin Gilen’s. And Nathan Kelly’s.  Shoot, the American Political Science Association commission a whole report about “Inequality and American Democracy.”  All of these are empirical.  All engage normative arguments.

I’m not suggesting that every work of political science qualifies, including many of my own.  But I’m not cherry-picking either.  It is easy to find political science research that marries objective analysis—even mathematics!—with “substantive political debates.”

Gerson purportedly “crack(s) open most political science journals” and “find(s) a profusion of numbers and formulas more suited to the study of physics.” He should read a little more, because he doesn’t really know what he’s missing.

[UPDATE: See also Greg Weeks.]


Andrew Gelman November 5, 2012 at 5:52 pm


I don’t know if this will make you feel any better, but . . . before seeing this blog post of yours, I’d never heard of Michael Gerson before. Washington Post political coverage really seems to have declined since the days of Art Buchwald!

P.S. I do agree with Gerson, however, that reporting probabilities such as “86.3 percent” is pretty silly. Also I agree with his implied statement that more students should consider going into statistics.

Joel November 5, 2012 at 9:30 pm

I like the idea that 86.3 percent is silly, but 80% and 85% aren’t silly. They’re good American numbers. Those decimals are kinda European, and it doesn’t matter if they’re math.

Andrew Gelman November 5, 2012 at 10:12 pm


I think 86.3% is silly for the same reason I think it would be silly to describe someone as 5 feet 4.352031 inches tall. As a statistician, I don’t like distracting people with numbers that are basically noise.

Craig November 7, 2012 at 7:51 pm

Public health professionals routinely report height and weight measurements in decimals. I’m sure they’ll all stop now that they know some guy on a blog thinks they’re silly for being precise.

Andrew Gelman May 30, 2013 at 6:59 am


If public health professionals reports a probability such as 86.3% (or, for that matter, a height as 5 feet 4.352031 inches), I will criticize them too. Just cos someone’s a public health professional, that doesn’t make them perfect. And just cos I am a guy on a blog, that doesn’t make me wrong.

Jerry Sena November 6, 2012 at 4:35 am

As a non statistician, I find the decimal reassuring, perhaps because it implies that a carefully crafted mathematical model underlies the figure. I realize my opinion is subjective, but so too is your judgment that the decimal is “silly.”

Andrew Gelman May 30, 2013 at 7:01 am


That’s one of the problems with hyper-precision: it can make people over-confident in a number. There’s no reason why the model underlying the “86.3%” number is carefully crafted.

Alexi Maschas November 5, 2012 at 7:30 pm

While I think the gist of your response to Gerson is on the mark, I do think it’s slightly disingenuous to hand wave away the serious schism between theorists and the methods-driven mainstream of American political science. Gerson makes the point poorly, but contemporary political science does lean heavily towards methods and models at the expense of theory.

John Sides November 5, 2012 at 7:48 pm

Alexi: Fair point. I don’t mean to portray political science as a big happy family without disagreements about substantive focus, methods, etc. And I personally would like more engagement between scholars who do empirical research and those who do normative theory. My hope in writing the post is that Gerson would see that there is political science research that engages both.

klhssh November 5, 2012 at 8:41 pm

Maybe we should just have a policy not to use any numbers in any journal, maybe that would make things better. I know we are all supposed to eschew numbers and just talk through it all with our postmodern arguments, but just because someone thinks that there is too much modeling doesn’t mean a particular model is invalid or doesn’t provide proper justification for claims. In fact, every single numbers-based argument in every journal could be completely valid, and informative even if polisci has a theory problem. I don’t see how models detracts from the discussion; polisci doesn’t need less models it needs more theory. And anyone who makes the claim that the number of quantitative arguments are somehow detracting from the field need to provide evidence of the claim, or this is no more an argument about your personal preferences.

This is why the entire piece is so silly: the claim of who is likely to win the election is a positive one which can be easily quantified (the statistics really ain’t that hard). And what are we supposed to do? Evaluate Silver’s numbers based on the fact that there’s a theory problem in polisci? That we should not even be asking the question because it has a air of weather-report-like triviality. You have *GOT* to be kidding me.

Jim Fearon November 5, 2012 at 11:47 pm

Yes, there’s lots of Pol Sci on the kinds of things Gerson says he wants more of in this op-ed. But what’s really remarkable is the way he is blaming Silver et al for making the punditocracy overly obsessed with polls and the horse race. Rather, the reason Silver and company have been successful is just that they are doing better what the pundits have traditionally been obsessed with and claimed to be doing all these years: horse race analysis.

It would be great to have more informed political commentary on substantive rather than horse race stuff. But blaming poll averaging and political science research that points out how the things that pundits get all excited about don’t seem to explain much about the horse race is ridiculous. Instead, the problem seems to be that the horse race is what people are mainly interested in, if they are interested at all.

Cornelieus Fitzedward Pope II November 6, 2012 at 12:33 am

It is the task of polisci normative theory to explain why people are consumed with the horse race and not what it considers its substantive questions, and it is the task of polisci modeling to measure and predict the schism between it and normative theory.

Pat November 6, 2012 at 11:06 am

I suspect that Gershon’s real problem has to do with a simpler problem. It’s pretty easy to write about the horse race non-quantitatively. You merely express your goals for the outcome and say that they will happen. Unfortunately, with people like Silver and Wang around, it makes you sound like a fool. Then the editors expect you to write articles about the substance of the race, and you have to think harder. Gershon’s probably getting a little old for that crap.

Jacob H. November 6, 2012 at 12:01 pm

I think there’s a strong argument that Gerson is not making that, forecasters and hobbyists aside, academic political science is overly focused on quantitative approaches right now. I enjoy quantitative methodologies very much (this blog highlights why), but there are other epistemological values that are presently being neglected. I’m kind of shocked, for example, that the famous politics department next door to me no longer requires any foreign language of its graduate students.

LFC November 9, 2012 at 2:56 pm

I agree with Jim Fearon. The problem is that people are so intensely interested in the horse race. Silver, so I gather (I don’t read him myself), does that much better than the pundits, and the obsession w the horse race is not Silver’s fault. But that obsession does detract from a focus on substance.

Comments on this entry are closed.

Previous post:

Next post: