Final presidential election forecast, 2024

The last time we discussed my presidential election forecast models, I observed there was an ambiguity that had never happened in any other presidential election in the training set (1952-2016): a major party nominee withdrawing his candidacy after the primary elections were complete, in favor of a new nominee who had not won a single vote. My model made use of the share each nominee won of the primary votes in their respective parties. The question was this: should Harris inherit the primary vote total earned by Biden, or should she start off at zero? If the former, she was on her way to winning approximately a 13% in the national popular vote; if the latter, her margin in the national popular vote would be closer to 2.4%. With the Real Clear Politics (RCP) average of national polls now giving Trump a slight edge (0.1%), it looks like Harris did not end up inheriting the Biden primary vote after all.

I dispute the RCP aggregation methodology. They include polls of registered voters (RV), whereas they should be restricting to polls of likely voters (LV). I further restrict to polls conducted in the last two weeks (October 13 and later). Lastly, I compute the trimean rather than the overall mean, which gives more weight to polls closer to the center of the data. By doing so, I calculate that Trump has a 0.6% edge in the national popular vote.

Of course, the election is fought in the Electoral College (EC), not the popular vote. I assume that Trump comes in with a base of 219 electors (sorry Democrats, that includes Florida and Texas), and that Harris comes in with a base of 226 electors (sorry Republicans, that includes Minnesota, New Hampshire, and Virginia). 93 electors are in play from 7 battleground states: Arizona (11), Georgia (16), Michigan (15), Nevada (6), North Carolina (16), Pennsylvania (19), and Wisconsin (10). 270 are needed to win, so Trump needs 51 more votes, and Harris needs 44 more votes.

Next, I studied three poll aggregators: Nate Silver (now working independently), 538 (now part of ABC News), and RCP. I examined their poll numbers from the 7 battleground states, as well as computing their trimean in each state. Here are the results.

• Silver’s model is the most favorable to Harris, but even he is showing a Trump electoral victory, with Harris taking Michigan, Nevada, and Wisconsin. Trump wins the EC by 281-257. Margin of victory: 24.

• RCP’s model is the most favorable to Trump, with Trump winning all 7 battleground states. Trump wins the EC by 312-226. Margin of victory: 86.

• 538’s model is in the middle. It shows Harris winning only Michigan and Wisconsin, not Nevada. Trump wins the EC by 287-251. Margin of victory: 36.

• Not surprisingly, the trimean takes the middle position as well, with Harris winning Michigan and Wisconsin, and Trump winning the others, winning the EC by 287-251. Margin of victory: 36.

• But beware. The margins are razor-thin in some states: WI (0.15%), MI (0.27%), NV (0.35%), PA (0.45%), NC (1.22%). For completeness, here are the other margins: GA (1.85%), AZ (1.92%).

• Thus if Silver is correct about Harris winning MI, NV, and WI, Trump could only afford to lose AZ, but not any other battleground states (GA, NC, and PA). Those three states are all must-wins: not just one or two of them, but all three.

You can see my electoral maps here and here.

I understand that partisans on both sides will claim that certain other states are… “in play”. I heard that about New Jersey and New York from the Trump side in 2016. My response is, show me results.

After the election, we will check back to see which of the poll aggregators was closest to being correct. We will also check in with my ex-post model, that uses the results from the Ohio election as a further check on the original forecast model.

EDIT: Additional points.

1. RCP is reporting that Trump is polling nationally 7.3% better than 4 years ago. Apparently, this is not spread out equally in the 50 states. My guess is that Trump has improved the most in Democrat-leaning states, where his support has had more room to grow compared to Republican-leaning states. That has not translated to an obvious shift in the battlegrounds. I hope I am proven wrong and that MN, NH, and VA go for Trump.

2. The 16-year cycle I wrote about earlier may no longer be valid. We may be in a new era, perhaps an 8-year cycle, with alternating party control of the White House. We won’t know for sure until the middle of the next decade.

3. For anyone still on the fence, unhappy with Trump’s “morals”, please ask yourself if you prefer to live in a country governed by Harris’s “morals”. You can make a difference, and unlike in most of history where changing the course of events would require a war, all you need to do is to go vote. Vote as if your life depended on it.