Autopsy-Turvy

Strategists teaching the wrong lessons

So I bet a lot of you thought the official party autopsy of the 2024 presidential election, released on Thursday, was probably concealing some awful ideological truth the DNC didn’t want you to know, and that’s why Ken Martin wanted to keep it secret, but it turns out it was merely incompetent, assigned to a writer who either didn’t know how to do it or didn’t have time, and produced a document that was woefully incomplete both in form (a lot of the chapters promised in the table of contents don’t seem to have been written at all) and in content—as Michelle Goldberg wrote for The Times,

What’s most striking is its utter lack of substance. The words “Israel” and “Gaza” don’t appear once in its 192 pages. It offers little insight into why the Democratic Party lost large numbers of Black and Latino men, or its failure to speak to disconnected, irregular voters. Much of it is a string of platitudes, like this: “It’s imperative that Democrats meet the moment — by identifying and preparing the leaders and organizers who will deliver positive change for America.” I wondered if it was written by A.I....

and so poorly conceptualized that Martin felt the DNC couldn’t take responsibility for the thing and festooned it with red marks like an angry professor reacting to a bad term paper.

Leaving the rest of us largely mired in the same old debate as to whether we ought to pull our wagon out of the rut and over toward the middle of the road or out to the left shoulder for the next phase of the journey, as if that humble metaphor provided an adequate characterization of all the human possibilities, only with even less hard information than usual.

---

No, there is no population of Goldilockses at a vast persuadable center, judiciously weighing lists of policies (”this porridge is much too left!”) in search of the most perfectly moderate bowlful who will give you their votes as soon as you give them a program of exactly the right degree of tepidity; and while there is a discernible left, it’s never demonstrated how it can muster votes out of a welter of competing priorities (unlike the right, where antitax maniacs and antiabortion fanatics cheerfully coexist as if they had anything in common but their enemies)—it still enrages me that hardly anybody was able to even recognize Joe Biden’s anti-inequality, anti-racist, pro-child, green, industrial planning program as of the left, other than Senator Manchin, who hated it and personally worked to gut it exactly for that reason, of course.

A much more useful kind of autopsy came out Friday, at the Substack of the “data journalist” G. Elliott Morris, who became famous as an undergraduate in 2018 for predicting the Democratic wave in that year’s midterms, and is now using Substack to rebuild something like the old 538, which expired when ABC fired the whole staff last year: Morris comes right out to propose “The Real Reasons Democrats Lost in 2024“, and they aren’t anything like what you might expect.

Namely, Morris doesn’t discuss anything the candidates did or the way they were chosen, or the running of the campaign, not that he thinks any of these weren’t relevant, but he doesn’t need them for the explanation of what happened; it’s fully accounted for by just two variables encapsulating the way voters felt about the incumbent party and its economic management: Gallup’s number for presidential approval as measured in June 2024 and the two-year average of the University of Michigan Consumer Sentiment Index between June 2022 and June 2024. These are used to create a model that accurately estimates the share of the two-party vote (third parties are omitted) that the incumbent party’s presidential candidate will receive in November.

The chart below shows the incumbent party’s actual share of the presidential vote in elections from 1956 (at the far right, with about 57.5% for Eisenhower) to 2024 (with the solid blue dot at 49.3% for Harris). The dashed line shows what the percentage would be if the model was perfect, from under 45% for Republicans in 2008 (McCain beat the prediction with a good 46%), to 60% for the 1956 Republicans, (in which Eisenhower overperformed). The blue line under Harris’s dot shows the full range of her prediction, from 46.6% to 49.9%—she overperformed too, it turns out, finishing within the predicted range but near its top, losing by 1.5 points instead of the expected 4.

Other especially interesting years are 1960, where it looks as if the Kennedy campaign really came close to defeating the model, and 2000 and 2016, where it looks as if W. Bush and Trump respectively did (can’t say, of course, since I don’t have the prediction ranges), and the chart may offer some credence to rumors of cheating of one kind and another in all three elections.

Anyway, the Harris result is especially remarkable, given what we know about the parameters: consumer sentiment had been at record lows for two years. even as the rate of inflation increase finally began to slow, and Biden’s Gallup approval was terrible that June, already at 37% before the awful debate (which didn’t lower it that much further—indeed it rose a good deal after he gave up running). It seems as if the result was basically decided a good month before Harris and Walz became candidates in July, and there wasn’t much of anything they could have done about it. And nothing especially an autopsy can recommend, except get rid of inflation before the midterms, which is what did Republican Ford and Democrat Carter in (Paul Volcker, made Fed chair in 1979, succeeded in whipping inflation by the most brutal means, but not until after Reagan was inaugurated, so Carter didn’t benefit; there’s the two-year rule in operation). And get rid of Bidens early too, I guess, though I really hate that, because I still think he was the most effectively left president since Roosevelt, though I don’t imagine he thinks of himself that way. Neither did Roosevelt!

Worse, by attributing the loss to strategic failures, the autopsy invites the party to learn the wrong lessons. If you decide Harris lost because she didn’t run negative enough ads against Trump, you’ll spend 2028 running more negative ads. If you decide she lost because the campaign didn’t have a clear definition of the candidate, you’ll spend 2028 obsessing over the candidate definition. Neither will help if the next Democratic nominee inherits another period of high inflation or low presidential approval. And neither will be necessary if they inherit a recovering economy and a popular incumbent.

The deeper problem with the autopsy is that it imagines a voter who doesn’t exist. The kind of voter the report’s recommendations would persuade — someone weighing Harris’s issue positions against Trump’s, watching campaign ads carefully, updating their beliefs in response to messaging frames — is essentially a Washington consultant, not your grandma who can’t afford to pay her bills because gas is up 50% and electricity subsidies just ended. One of the problems with autopsies is that voter psychology takes a lot of work to understand well, but the people who have that skillset largely aren’t the type of person the DNC is hiring to audit their choices.

Which brings us back to the point for which I was citing Oliver Willis above: the ways Democratic strategists think about voters aren’t insightful—not that we need to start thinking the way Republican strategists do, that’s even worse, as I believe we will finally understand after the next two years of inflation driven by terrible policy and a president already more unpopular than Biden except for that one awful summer. I mean things I’ve been talking about for quite a while now: of the need to get ahead of the one-dimensional left-to-right model of ideological identity; of the need to start getting serious about charisma; in particular of the need to stop imagining this imaginary population of persuadable “centrist” voters we can gently woo to our side with the judicious moderation of our policy proposals and focus instead on the real voters, the unlikely ones, the ones whose reasons for switching from Obama to Trump may not be entirely rational, who don’t spend all that much time thinking about politics or policy, not because they’re dumb but because they have a lot of other stuff to think about, and who need to be surprised! In a way politicians haven’t often surprised them, but maybe Harris and Walz did, for at least a moment, and others perhaps relatively left and right, Ocasio-Cortez and Talarico, and so on.

I’ve been handicapped in these efforts not just because I’m not a political scientist, but also because those who are haven’t generally been collecting the kind of data I need, and the thing that’s been striking me the most in Morris’s newsletter is that he has: that he has real numbers I can use on the behavior of nonvoters and unlikely voters and undecideds, not the caricatures interviewed by The New York Times on their Cletus safaris and their not totally honest formulations, but the actual quantifiable behavior in the voting booth and survey studies. I hope to be writing more about these matters in the near future.