|
 |
The blogosphere is back! In the early 2010s when I started writing, there were tons of interesting debates carried on between blogs — one person would write a post, and someone else would link to it and respond to it on their own blog, and they’d go back and forth for a few rounds. It reminded me a little of the “Republic of Letters”, the network of intellectual exchange that existed in Europe and the Americas in the 1600s and 1700s. In the mid to late 2010s, this epistolary exchange was superseded by Twitter fights; instead of slow, measured responses, intellectuals would “dunk on” each other with 280-character denunciations. Something important was lost.
But I’m happy to report that with the degradation of X and similar platforms, and the rise of Substack and other new blogging utilities, we’re starting to see some of the old debate style return. A good example is the recent debate over whether moderate Democrats are more electable. So in traditional blogosphere fashion, I’ll try to give a rundown of the debate itself, and then see if I can add anything new.
Run moderate candidates, or turn out the base?
This argument has been simmering for a long time (perhaps for centuries), but recently a new crop of analysts has started to bring new statistical methods to bear on the question. Since the late 2010s, political scientists have begun to borrow a concept called “wins above replacement” from sports. Sports WAR is fairly complicated, since it involves isolating a single player’s contribution from the contributions of the rest of the team. But elections are an individual sport; when political scientists talk about WAR, they really just mean doing some regressions on election results to figure out what individual characteristics cause candidates to outperform.¹
A data analysis company called Split Ticket, headed by Lakshya Jain, is probably the most famous for using this methodology. They’ve consistently found that moderate candidates do better than strongly ideological ones on both the Democratic and the Republican sides. For example, here’s a chart where Split Ticket breaks out recent performance by Congressional caucus:
 |
This seems pretty solid, but it’s subject to a subtle caveat. Split Ticket tries to control for every important feature of a Congressional district that might influence the outcome of elections. But there might be interactions between those characteristics and a candidate’s level of moderation. For example, Blue Dog Democrats might outperform in their own districts, but if you plunked them down in the districts that elected AOC or Ilhan Omar, they might do even worse than those progressives. It’s hard to say. But this is sort of a second-order problem; these results still suggest that Democrats should try running more moderate candidates.
Some political scientists have disagreed with this conclusion. For example, Bonica et al. (2025) use an alternative methodology to assess the benefits of moderation. They measure candidate’s ideology based on a combination of how they vote in Congress and which donors donate to them.² They find that moderation is beneficial, but the benefit is smaller than what Split Ticket finds. Bonica and his co-authors argue that this means that base turnout is ultimately more important, and that politicians should therefore be unafraid to embrace strong ideology in order to fire up the base. Here are some excerpts from a thread Bonica wrote:
 |
 |
This conclusion has a major problem, and hopefully you can already see what it is. Suppose a researcher comes to you and tells you that people in hospitals are more likely to die of disease than people outside of hospitals. Should this make you avoid hospitals when you’re sick? No, of course not. People go to hospitals because they’re sick, so of course those people are more likely to die of disease!
Similarly, Bonica’s observation about Democrats’ national election performance interprets correlation as causation, when there’s an obvious reason not to interpret it that way. Obama won in 2008 running a campaign that was more lefty than usual, and won. But maybe he was able to run on a more lefty platform precisely because the electorate was in a more lefty mood than normal! After all, voters in 2008 were very mad about the Iraq War and the financial crisis. Obama harnessed that anger to win, and the anger also caused high turnout.
But in 2010, the electorate was in a far more conservative mood, and brought in the Tea Party Congress. Democrats ran to the center that year, and still lost big. But they might have run to the center precisely because the electorate was in a conservative mood! And had they not run to the center, their performance in a conservative-leaning year might have been even worse!
Kamala Harris tacked to the center in 2024 and still lost a fairly close contest. But suppose Kamala Harris had come out as a progressive fire-breather in 2024, the way she tried to in the 2020 primary. Suppose she had railed against systemic racism, called for an end to military aid to Israel, proposed cutting police budgets, and offered a full-throated defense of Biden’s tolerant immigration policies. Do we really believe that this strategy would have kept the election as close as it was, or even won it by turning out the base? That’s what Bonica and his co-authors would have us believe. And though we can’t prove them wrong, it doesn’t really pass the smell test, does it?
Of course the WAR analyses for congressional races also don’t separate correlation