Here’s Cowen’s post, and here’s our paper:
Moderation in the pursuit of moderation is no vice: the clear but limited advantages to being a moderate for Congressional elections
Andrew Gelman Jonathan N. Katz
September 18, 2007
It is sometimes believed that is is politically risky for a congressmember to go against his or her party. On the other hand, Downs’s familiar theory of electoral competition holds that political moderation is a vote-getter. We analyze recent Congressional elections and find that moderation is typically worth less about 2% of the vote. This suggests there is a motivation to be moderate, but not to the exclusion of other political concerns, especially in non-marginal districts. . . .
Perhaps I’m not understanding it fully, but on a read-through it seems like a weak (incomplete?) paper that supports the Median Voter theorem?
Blackthorne,
Yes, it’s an incomplete paper. That’s why it’s my Unpublished Papers page—we never finished it! I still think it has good stuff. I don’t really know why you call it “weak,” but if there are specific problems you see with it, we welcome comments.
I wouldn’t say that our paper supports or does not support the median voter theorem. As is generally the case with the application of mathematics to social science, the conditions of the theorem are only approximate. The point of the paper is that (a) moderate candidates do better, even in some settings where a naive view of the politics might have suggested otherwise, and (b) the benefits of moderation are limited, which gives us some insights into the tradeoffs involved in position-taking.
I don’t know how it shows that. This seems like a result which is consistent with the median voter theorem, where politicians successfully optimize such that they hit the median nearly exactly to trade off electability and their election goals (plus possibly the usual smidge of systematic error, like regression to the mean, of the sort which is hard to purge completely & yields small coefficients of the ‘everything is correlated’ variety).
The median voter theorem is a statement about optimization and control systems: politicians optimize for the level of ideological extremity which gets them elected. So any approach involving simple correlations (or regressions) is already in trouble, just like it would be if you were trying to investigate the ‘median thermostat theorem’. If you wanted to investigate it, the obvious approaches would be natural experiments like sudden shocks to ideologies of candidates and/or electorates (although of course quite difficult to establish exogeneity or that the shock isn’t shared by both).
This doesn’t seem to do that, and I don’t see how the election data here could show anything useful causally. Skimming hastily, this sounds like a result like this: “we investigated Dawe’s median thermostat theorem by using time-series from a thermometer we put near to the thermostat, and we found that at temperature extremes, the thermostat showed only a small ‘moderation’ effect where it was slightly more likely to turn on/off than our regression model using our thermometer’s data would predict. We thereby conclude that claims about thermostats controlling room temperatures are overstated, although they give a real but small benefit to room temperature.”
Gwern:
In the paper, we estimated the electoral benefit of moderation to be approximately 2% of the vote. I agree that moderation, as we defined it, is relative to the U.S. political system. The vast majority of U.S. legislators are winning by much more than 2% of the vote. For them, they could go much more extreme and still do fine. This is one reason that we get extreme median-voter-defying behavior such as the Oklahoma abortion law that passed on a 73-16 vote.
The point of being elected is to then be able to make votes different from the other guy. Passing unpopular laws doesn’t defy the median voter theorem, which is about elections and candidates, not single bills. How many of those subsequently lost elections solely because of their ‘extreme’ behavior in voting for that abortion law?
(And if they are elected on large margins because of primaries/one-party districts, that just moves the candidates’ optimization to the real election within the party, not the general election, and renders the analysis in the paper even more moot.)
Gwern:
Ultimately, it’s up to the voters, so in that sense the median voter theorem is trivially true. Setting aside gerrymandering, etc., the way to win the election is to get more votes than the other candidate. I’m not disputing this indisputable statement. Our paper addresses the question of how much of this voting can be attributed to candidates being closer or further from the center. In the linked post, Cowen points to a political party adopting a centrist position, which he attributes to the median voter theorem; there are also lots of cases of political parties adopting extreme positions.
What’s the Cowen angle, for those of us who can no longer bring ourselves to read him? I used to, until I realized how consistently he used banal, dispassionate, and yet still somehow self-satisfied language to support conservative talking points. More in sorrow than anger, of course.
Nah. it’s you not him. He has been pretty consistent in his views. You just don’t want to listen anymore to the points that you don’t agree with.
Absolutely. The act was cuter before the resurgence of fascism.
Dmitri:
Yeah, recall this post, where I wrote about a Harvard election denier:
Maybe one point: could it be that a candidate who appears moderate relative to the congress is not necessarily the moderate one within the district? As far as I understand incumbents would only need to appear moderate in comparison with the other candidate within the district.
I think you need to start thinking of Cowen as a David Brooks type. Often wrong, and just can’t admit he’s wrong. Bangs the drum for ChatGPT, and then doesn’t even admit his errors when the AI just fools him into posting fake Francis Bacon quotes on his blog. Between this and the UFO stuff, I’m not sure why people take him seriously. https://www.thenation.com/article/culture/internet-archive-publishers-lawsuit-chatbot/
Steve,
The fake quote thing is funny. I agree that it would make sense for Cowen to post something on the topic. We can learn a lot from our errors. I’ve learned a lot from engaging with the errors that I’ve made.
I find Cowen more interesting than Brooks. And I say that even though Cowen is a big fan of Brooks! Brooks doesn’t seem to know anything. Cowen knows a lot. So I enjoy reading Cowen, even if he has some enthusiasms that I don’t share, such as David Brooks, Malcolm Gladwell, Elon Musk, and space aliens.
Something to keep in mind about MR: seems likely that all those NYT and other paywall links are there for a reason. Media has to sell itself. At the very least, MR makes an effort to spread the media consumption love around, which, if it doesn’t generate a paycheck directly, at least keeps MR on the minds of the people who consider themselves important that it links to.
None of this takes away from the fact that both Tyler and Alex are highly intelligent, knowledgable and often insightful. It’s just that those characteristics don’t necessarily pay the bills. Also I feel like the comments section is very good, which is often why I read MR. After all, Tyler and Alex rarely offer more than a minor blurb. You seem to think the comments there suck, and sometimes they do, but there are also excellent discussions in the comments on a wide range of topics.
I checked the above link. I agree Tyler should have fessed up and posted an apology. Personally I feel like that’s the big danger of “AI” as it currently stands – that, in fact, it propagates *false* information, rather than true information, because it only knows what people feed it – and it’s not clear to me that it can detect bullshit in the way a human can detect bullshit. Just the same, Tyler’s primary claim about AI is that “it’s happening”. And *it is happening* and he’s far in front of most pundits on that front.
I didn’t read your paper, but I confess I’m skeptical about both Tyler’s claims and yours. On the Tyler side, I don’t see that we’re getting “median voter” legislation. Quite the contrary. On the Andrew side, any quantitative analysis of “moderate” candidates or positions seems hopelessly dependent on the definition of “moderate”, not to mention the issues of the moment.
You must have a stronger stomach than me. Yes, there are some smart people who comment on MR – but you have to wade through an enormous number of trolls and outright uncivil bullies and crude comments. I get enough of that from Trump’s speeches. Name calling and crude slurs do not make for “excellent discussion” and – for me – destroy the few valuable comments that are interspersed among the many that are not. MR makes me wish for a moderator.
I did a paper a long while ago that tracked some of this to asymmetric accounting: https://gsood.com/research/papers/inNout.pdf
With asymmetric util. functions, you can come up with optimal position taking also … but it won’t be ‘median voter’ in the commonly used sense