Paradox: Trump leads in the odds, but not the state map
Making sense of the 2024 betting odds
ElectionBettingOdds.com, the site I created, shows a seeming paradox right now:
It shows that Trump is more likely to win than Biden. But Biden leads in more states, with more total electoral votes.
How is that possible? Are the markets just inefficient? Is there an error in the site?
Here’s the real answer:
Trump has more paths to winning
Looking at the map, there are 6 states that are almost pure tossups:
Biden has a 50.5% chance in Nevada, a 52.5% chance in Pennsylvania, a 53.5% chance in Michigan, and a 54.5% chance in Wisconsin.
Trump has a 49.5% chance in Arizona, and a 57.5% chance in Georgia.
Roughly speaking, these are all coin-tosses.
Trump leads in non-tossup state electoral votes
The key to the “paradox” is that, if you add up all the non-tossups, Trump leads by 9 points. Trump has 235 electoral votes in non-tossup states, vs Biden’s 226. Plus, Trump is expected to win if there’s an electoral college tie.
As a result, there are a number of tossup combinations that Trump could win with, which don’t apply to Biden.
For example, Trump would become President by winning just Pennsylvania and Georgia, while losing all the other tossups. Or he could win with just Pennsylvania and Michigan. The same cannot be said about Biden.
Those are just a few of the unique combinations Trump could win with. If we assume that all non-tossups go the way people expect, then Trump has 39 unique winning combinations among the 6 swing states. Specifically:
In contrast, Biden has just 25 combinations of tossups he could win with:
If every tossup state were an independent coin flip, and all other states were guaranteed to go the way they lean, then I believe the above means that Trump would have a 61% chance of winning1 based on the state betting odds.
Note that while the betting gives Trump a 46.2% chance of winning, that rises to 51.3% if we assume he will actually be the candidate on the ballot on election day, and that third parties won’t win (the above analysis assumes those things, so that’s the comparable number.) Biden’s odds would also rise, if we assume the same for him; the betting would then have it as 51.3% Trump to 48.6% Biden.
Why do the odds give Trump 51%, if the combinations suggest 61%?
Every tossup state is not quite an independent coin flip.
Certain states tend to move together
In practice, the “rust belt”/“blue wall” states of WI, MI, and PA tend to move together.
Likewise, NV and AZ may move together.
As a result, one relatively plausible scenario is that Trump’s polling gains among hispanics come through, winning him GA, AZ and NV — but WI, MI, and PA go the way they went last election. In that case, Trump would lose by 1 electoral vote. (For that reason, there’s a particularly important single electoral vote in Nebraska CD-2, which is basically Omaha, Nebraska, which gets a separate vote from the rest of the state. Trump lost the district in 2020 by 6 points, but won it in 2016. It’s considered “lean Dem” but such a small area can be unpredictable.)
That scenario is relatively likely, compared to other combinations, due to how states tend to move together, and it may be a reason that Trump’s real odds are less than 61%.
The tossups aren’t the only ones in play
Another factor is that North Carolina only has 70.5% chance to go for Trump, whereas Biden’s weakest “leaning” states are Minnesota and New Hampshire, which are both at 83.5%.
It’s conceivable that Trump could win, say, PA and MI — which would have made him win in the above analysis — but a surprise in NC would snatch that victory away.
Conclusion — There is no contradiction between the “Presidential winner” and state odds
A full analysis would consider both the historical correlations between states, and also the “leaning” states as well as the tossups.
I suspect that analysis would bring Trump’s odds down a bit from the 61%, to be closer to the 51% chance the bettors actually give him.
People, in particular Nate Silver, are working on that full-fledged analysis. It is also certain that some of the bigger traders operating on places like Polymarket and Betfair have done their own analyses along these lines. Historically, Silver’s odds have been roughly as accurate as the bettors, but using an average of the betting odds and Silver’s predictions has been even more accurate.
39 Trump combinations / (39 Trump combinations +25 Biden combinations) = 61%