My take on the election
There was nothing Democrats could have done.
Could Harris have won?
No. The polls showed a very stable election, nothing Harris did had much of an effect and since Trump won handily, we can conclude that at no time during the race was Harris really ever ahead.
Why were the polls wrong?
In 2012 the idea that the polls were “skewed” in favor of Democrats became a popular idea on the right. As an illustration of this idea, I looked up the election day forecasts from electoral-vote.com (a poll aggregator I have followed since 2004) and applied a simple rule. Any state that Democrats won by 3 or fewer points I would give to the Republicans. By doing this I converted the solid Obama victory forecasted by the model to a narrow Republican victory (see Table 1) as Republican partisans were claiming. The skew-adjusted model was very wrong. (The bolded values in the table highlight the more accurate forecast). The unadjusted model forecasted a Democratic victory that was more conservative than the actual results. The polls seemed fine and the hypothesis of a systemic bias in favor of Democrats was rejected as nonsense.
Table 1. Election model and skew-adjusted model results compared to actuals
In 2016, the polls and the model forecasted a result very similar to that in 2012. As in 2012, addition of the skew changed the result to a narrow Republican victory. This time the model was wrong, while the same skew-adjustment rejected in 2012, forecasted the opposite result—and was right. There was much soul searching among pollsters and election forecasters as to how and why they got it wrong. In 2018, the polls seemed to work well so the problems were believed to have been solved and in 2020 a Democratic victory greater than Obama’s in 2012 was predicted. Applying the skew adjustment showed that Biden was still going to win, but by a narrow margin. Having been burned in 2016, I applied a skew adjustment to the polling data to reassure myself that this time, my side was going to win. And they did.
For 2024, the model forecast was worse than 2016 and the skew-adjusted result predicted a Trump blowout. I had read about how polls were being adjusted to force the results into a pre-Dobbs electorate composition based on the 2020 election. This electorate was presumably more Republican than a post-Dobbs electorate like that seen in 2022 would be. Based on this information and my belief that there was a ceiling for Trump support around 47%, I came to believe that the necessary skew adjustment was already built into the polling results, so adjusting for skew was not necessary. I believed the election would be close. As I told a friend on election day who had a bad feeling, “my head tells me Trump should win, but my gut says Harris will pull it out.” As it turned out the skew-adjusted result was the same as the actual. With the skew, Harris never had been ahead during the campaign. She hadn’t had a chance.
Analysts more knowledgeable than I have hypothesized a “Trump vote” consisting of unlikely voters whose preferences are not captured by polling. They come out to vote for Trump but otherwise stay home. The “skew adjustment” is just a fudge factor applied to elections in which Trump is a candidate but otherwise is invalid. Now that Trump has run his final election, it can be retired.
Could Democrats have won had Biden not chosen to run?
No. Table 2 summarizes the results of elections in which no incumbent is running since the beginning of the Democratic party in 1828. The differences are stark. In such elections, when a Democrat is president, the Democratic candidate has won two and lost ten. Conversely, when a Republican/Whig is president, their candidates have won five and lost five. Assuming a basal 50% probability of winning, the probability of one party losing 10 out of 12 elections by chance is just under 2%. In contrast, the Republican record of winning 5 of 10 is just what one would expect.
Table 2. Election results when no incumbent is running
Interestingly, Democratic incumbents have won 9 out of 12 times, while Whig/Republican incumbents have won 8 out of 16 times. Again, the Republican outcome is just what one would expect from random chance, while the Democratic one is not.
The historical record suggests that voters like Democratic incumbents but do not like non-incumbents running to extend Democratic rule. Before I conclude this, I note that in four of these lost elections (1868, 1952, 1968, 2024) the sitting Democratic president either withdrew or otherwise failed to get the nomination, resulting in another nominee being chosen who then went on to lose. We might categorize these elections as incumbent losses and add them to the Democratic incumbent record. This changes the Democratic incumbent record to 9 wins out of 16, which is consistent with chance.
This reduces the Democratic entries in Table 2 to eight, two of which are wins, which has a 15% probability of happening by chance. This still suggests a bias against Democrats, but before I conclude this, I figure controlling for the dispensation might be a good idea. I have been using Stephen Skowronek’s Political Time model as a framework for characterizing American electoral dynamics. This model holds that periodic reconstructive presidents create a new “political order” providing standards that define what policy is possible or desirable, which I call a dispensation. We are currently operating under the Reagan dispensation that began with the 1980 election. It was established by Reagan’s landslide victory in 1984 and the subsequent victory of his vice president in 1988. It was preceded by the FDR dispensation, which was established by four successive Democratic presidential victories after the first one in 1932.
Differences in electoral outcomes could reflect which party has the dispensation. The Table 2 entries that happened during either the Democratic Jackson (1828-1859) or FDR (1932-1979) dispensations are coded in red. In this case the democratic record is 2 wins in 5, which is consistent with chance. The Republican/Whig record is 0 for 3 (12% probability). Under the Lincoln (1860-1931) and Reagan (1980-present) dispensations the Democratic record has been 0/7 (0.8% probability) and the Republican record 5/7 (23% probability). It would seem that candidates running when a member of their party is already president are unlikely to win when the other party holds the dispensation. Otherwise, they face probabilities consistent with chance.
During Republican dispensations like now1 Democrats cannot win elections unless they are either running against a Republican incumbent perceived as a failure or when a successful Democratic president is running for a second term. No matter how successful an incumbent Democrat is, another Democrat cannot follow them. This means that if Democrats ever want to play a leading role in shaping the countries future, they need to focus on regaining the dispensation by making sure the next Democratic president practices the politics of reconstruction. I wrote about how FDR did this here.
Doing this necessarily requires replacing the Reagan dispensation, which I see as centered on maintaining a neoliberal economic environment that fosters shareholder primacy economic culture. This would require bold policy that would best be implemented in the context of an economic crisis or major war. To this end I plan to write about my understanding of how the American political economy works in a broad sense and, if my view is correct, how the Reagan dispensation might be overturned.
There was one way Harris could have won, which is if Biden had somehow become a Reconstructive president. Such presidents (Washington, Jefferson, Jackson, Lincoln, Roosevelt, Reagan) have always run for re-election and won handily. Biden did not.