2024 D&D Forecast Review
I’ve been forecasting elections since 2017, making 2024 my fourth cycle. Without a doubt, this has been my best cycle yet—and likely the best I’ll ever have.
Here’s how my forecast performed:
The forecast’s accuracy this cycle was really, really good though I’ll be the first to admit that luck played a significant role. To put it into perspective, I ran simulations of the 2024 Election based on the assumption that my forecast was perfectly accurate—meaning every candidate's projected chances to win were precisely correct. Even under this assumption, there was only a 4% chance my forecast would get every presidential state correct and a 6% chance that my forecast would perform as well as it did (or better) overall from an accuracy standpoint across all races.
I will deliver a comprehensive overview of each chamber, compare my forecast with other prominent predictions, and examine the races I missed, offering an analysis of why this happened.
President
My forecast slightly outperformed those from Nate Silver, FiveThirtyEight, and The Economist in key presidential states this year. This outcome isn’t surprising, as my model consistently had higher expectations for Trump in nearly every state heading into election night and Trump overperformed the expectations of all four forecasts. Of course, a single election (N=1) isn’t nearly enough to determine which forecasts are ultimately better or worse—I’d still bet that mine would underperform the other three in the long run. But this year, it was about a point better in key states.
The only notable adjustment I made was manually shifting Michigan’s fundamentals about 0.5% toward Trump to account for the “Dearborn effect,” where Muslim Arab and South Asian voters in Dearborn, Dearborn Heights, and Hamtramck moved away from Democrats en masse. Typically, I reserve this kind of manual adjustment for block-voting communities like the Hasidic Jewish villages of New Square and Kiryas Joel in NY-17 and NY-18, where 200% swings between elections are common and entirely predictable. This year, however, I made an exception for Dearborn due to its comparable dynamics—slightly less extreme block voting than Hasidic villages but still featuring a swing I expected to be around 50% and on a larger scale. This adjustment ultimately led me to favor Trump over Harris in Michigan, because the forecast so narrowly favored Harris without this adjustment. Using some results-based analysis, it’s clear the approach worked exceptionally well, as it accurately flipped the state prediction. The proccess was sound too, reflected in Dearborn’s dramatic shift from Biden+39 to Trump+6.
My final forecast gave Donald Trump a razor-thin 0.05% advantage in Michigan, a 0.6% edge in Wisconsin, and a 0.9% lead in Pennsylvania. He was also projected to win Nevada, Arizona, North Carolina, and Georgia by margins of a point or more. While Trump only exceeded my expectations by an average of 1.2% in these states—far less than his overperformance in 2016 and 2020—he was already slightly favored heading into Election Day this time so it was enough.
Michigan, Wisconsin, Pennsylvania, North Carolina, and Georgia were all some of Kamala Harris’s strongest states compared to Joe Biden’s performance in 2020 and my pre-election expectations. This explains why she actually came pretty close to winning, despite her overall national performance being quite bad. Her collapse was concentrated in areas with large Latino and Asian populations, resulting in heavy losses in states like New York, New Jersey, California, Texas, and Florida. However, despite their size, these states are largely irrelevant in the Electoral College calculus. Electoral College Bias was roughly even in 2024!
US Senate
I missed just one Senate race: Pennsylvania. The loss of incumbent Democrat Bob Casey to Republican challenger David McCormick was a genuine surprise. My forecast pegged the likelihood of Democrats losing at least one of the Wisconsin, Michigan, or Pennsylvania Senate races at about 50-50, but my forecast said Casey was the least likely of the three to lose. It’s remarkable that such a prominent figure in Pennsylvania politics was defeated. While we all knew it was possible heading into election night, it stands out as the most surprising result in an election cycle otherwise lacking major upsets.
Overall, my Senate forecast was as accurate as my presidential predictions. It performed especially well in states where Democratic Senate candidates polled far ahead of Harris, where fundamentals tempered the large polling leads built by Gallego and Brown, bringing Democratic margins closer to reality. Arizona, involved a bit of luck—Gallego overperformed Harris by as much as polls had predicted, but the GOP exceeded expectations by about 3% in both states.
US House
The analysis below covers all races rated as tossups by the Cook Political Report, along with NE-02, which was rated Lean D by CPR but ultimately predicted incorrectly. My US House forecast performed slightly worse than the other two but remained fairly comparable overall.
This is going to be really specific and nerdy, but I missed 10 races out of the 435, so let’s get into them.?
领英推荐
1. AK-AL: Prediction D+3, Result R+2
Peltola overperformed by a lot, but Trump’s stronger than expected coattails in the state pulled Begich over the finish line.?
2. CA-22: Prediction D+2, Result R+6
I have some regrets. David Valadao has a strikingly consistent pattern of performing significantly worse in midterm elections (2018 and 2022) than in presidential years (2016 and 2020). Typically, midterms are easier for candidates to win crossover support, but in CA-22, it seems that low-propensity voters with low trust in government are more inclined to stick with the familiar name—Valadao. His 2022 performance wasn’t particularly remarkable; yet, he still managed to run 5 points ahead of Trump in an ancestrally Democratic district where Democrats were actively competing. That kind of result is undeniably impressive. Looking ahead to 2026, the theory suggests Valadao will follow the same pattern: after an impressive showing in 2024, he might lose, as he did in 2018, and may not even run far ahead of the California GOP gubernatorial nominee. However, I hesitate to fully commit to this prediction. While my forecast currently assumes a uniform effect of midterm vs. presidential dynamics on split-ticket voting and incumbent overperformance across all districts, Valadao’s unique trend defies easy integration. After all, he’s the only candidate in recent memory exhibiting this exact pattern. So, while the trend suggests he’s headed for a loss in 2026, I can’t say for sure whether I believe it. Valadao has a habit of surprising us.
3. CA-27: Prediction R+0, Result D+3
Maybe Christy Smith was actually the problem? Or Mike Garcia’s voodoo magic randomly ran out in 2024, despite winning by 6% in 2022. Either way, the model was reasonably close to the final margin in this race, just on the wrong side of 50-50.
4. CA-45: Prediction R+0, Result D+0
The model was basically completely accurate from a margin standpoint, just wrong sides of even.
5. MI-08: Prediction R+2, Result D+7
Ouch, part two. There’s not much to add here—Paul Junge (R) and Kristen McDonald Rivet (D) were both unknown quantities heading into the race. In hindsight, Junge was clearly a very weak candidate, while McDonald Rivet was exceptionally strong, even in an ancestrally Democratic district. I had a gut feeling this was the case, but since my model doesn’t incorporate expert ratings and Junge was just as well-funded as KMDR, it was difficult to convey that intuition. McDonald Rivet not only lived up to but massively exceeded my already high expectations, which is truly impressive. Definitely my forecast’s worst prediction compared to expert ratings and other forecasts, just totally off.
6. NE-02: Prediction D+2, Result R+2
Don Bacon is one of three Republicans to win a district carried by Harris, alongside Brian Fitzpatrick in PA-01 and Mike Lawler in NY-17. Along with Young Kim and David Valadao—who easily won narrow Trump districts—they form the “five horsemen” of GOP overperformers this cycle. I’ll admit, I thought Bacon was in a tough spot heading into Election Night. While I recognized he was a strong candidate and likely to overperform significantly (mostly thanks to his incumbency in an ancestrally Republican district and a substantial spending advantage, but also a few points of personal appeal), the forecast still believed the district’s increasing Democratic lean might be enough to unseat him.
7. NY-17: Prediction D+2, Result R+6
I have some regrets. Mike Lawler was always poised to win; the Israel-Gaza situation significantly bolstered his margin against Mondaire Jones, as Jewish voters swung hard right at all levels of gov’t. Even without this factor, he would have been fine. Lawler is a rare political talent, and his brand will only strengthen if he remains in the district. Often at odds with the national GOP—a challenge for most Republicans—this divergence is his greatest strength, given that he’s running in a blue district. Jones was also a weak candidate, but it was hard to imagine anyone beating Lawler this cycle, especially with the favorable national environment for the GOP. Lawler also has a once-in-a-lifetime chance to flip the New York governor’s office red in 2026 if Hochul withstands a primary challenge.
8/9/10. CO-03, PA-07, and PA-08
The forecast got all three of these races wrong, but I’ll still take a small victory lap because my predictions were way closer than everyone else’s. One reason my House forecast performed pretty well this year overall was an adjustment I made for 2022 House performance based on the state-specific environment that year. For example, at first glance, Susan Wild in PA-07 might seem like an electoral powerhouse in 2022—winning a Biden+0.6 district by 2% during an in-party midterm. However, it’s important to consider the broader context: all House Democrats in Pennsylvania benefited from the Shapiro-Mastriano landslide. Wild’s apparent strength in 2022 had more to do with Shapiro than with her own appeal, and Shapiro wasn’t on the ballot in 2024. The same applies to Matt Cartwright, who is genuinely strong electorally (matching Wild’s performance despite being in a district 3–4 points more GOP-leaning). Still, his 2022 results were also bolstered by the favorable environment. Similarly, Yadira Caraveo in CO-08 barely eked out a win in a Biden+5 district during a blue wave year in Colorado, which doesn’t exactly scream “impressive.” These environmental adjustments helped my forecast better account for underlying electoral dynamics.