The Ephemeric

So let's get to it. The White House, House of Representatives and Senate are all up for grabs. Despite what you may have heard in the media, there is a favourite in all three contests, although it remains very close. The topline prediction is this: The Ephemeric predicts Kamala Harris to win the Presidency, the Democrats to take control of the House of Representatives, and Republicans to take control of the Senate.

Presidential Election Verdict: Kamala Harris Elected President

Predicted Electoral Map: Kamala Harris (D) - 286, Donald Trump (R) - 252.

The above map is based on data from a variety of aggregators, including Split Ticket, Fivethirtyeight, and analysts including Sabato and Cook Political, and shows the expected electoral map. The rest is pretty self explanatory: dark blue represents safe Democrat wins, light blue leans Democrat, grey is toss up. Meanwhile on the other side, light red to dark red represents lean to likely Republican.

A few notes about this map. I have attempted to remove "toss ups" to the extent reasonable. I think analysts in general have become a bit too lazy about calling close states toss ups, rather than making an actual prediction. For example, Cook recently referred to a Wisconsin Senate race, with a polling average margin of 6%, as a toss up, which is just silly. For me, a toss up is a state that is literally too close to reasonably call, and I am drawing that line at under a 0.5% margin, and generally only where there is high quality polling indicating a victory for both candidates. A candidate leading by 1% in a state, yeah it's close but I can reasonably say that it's leaning towards one candidate based on the available data, so long as that lead is consistent in multiple high quality polls over an extended period of time. A 0.1% lead either way, on the other hand, is completely meaningless. I want to be clear though, just because a state isn't shown above as a toss up doesn't mean it isn't competitive, or realistically winnable by either side, just that there is a reasonably identifiable favourite as it stands.

There are seven realistically competitive battleground states in the 2024 Presidential race: WI, MI, PA, NV, AZ, NC and GA. Beyond these seven, we could theoretically get an upset, like Democrats winning Florida or Texas, but it's unlikely. I would be quite confident in saying that all the other states will vote as indicated on the above map.

While this race remains close and competitive, Kamala Harris is favoured. There has been a lot of hype in the media about how this race is a toss up, or too close to call, and that really just isn't the case. Kamala Harris leads in the polls, both nationally and in sufficient battleground states to win, and has done pretty much constantly and consistently since she entered the race.

In fact, for all the talk about how unpredictable this election is, or the constant media oscillations between who's on top, this has been about as boring and consistent an election as we've seen in recent years. That may come as a surprise in a race where one candidate was convicted of multiple felonies, and the other quit halfway through the race to be replaced by his Vice President, but despite this the polling in recent weeks has been very steady. There's been a lot of chatter of different candidates having momentum, or surges in the polls, but as you can see, there's been basically zero net movement in the last several weeks. 

The high quality polling currently has Kamala around 3% ahead nationally. As anyone who follows American politics will know, the national popular vote is not what wins you the election, it's the electoral college. This gives us the so-called electoral college gap, ie the national margin that a candidate needs to have in order to be roughly tied in sufficient states to win the election. In 2020, Biden won the electoral college somewhat comfortably with a national popular vote margin of around 4.5%. It is generally expected that the Democratic Party candidate will need to be winning by around 2% nationally in order to win the electoral college, although there is some evidence that this gap may be narrower in 2024. The long and short of it is that a 3% national margin should be sufficient for Kamala to win, and in fact she can probably afford for the race to be a little bit narrower than that. If her lead ends up being less than 2%, however, then it starts getting a bit dicey. 

But while we can infer, to an extent, Kamala's chances based on the national margin, ultimately it's victory in the individual key states that will win this election. There's no sense beating around the bush, the key to this election is the rust belt battleground, WI, MI and PA. If Kamala wins those three states, she wins the election regardless of what happens elsewhere. 

The good news for Democrats is that Kamala has led in all three of these states pretty much constantly since entering the race, and is favoured to win all three. In fact, you need to look pretty hard to find high quality polling showing Kamala trailing in any of these three states. The data is consistent enough for me to say that Kamala is favoured in both MI and PA, albeit narrowly, with only WI remaining a true toss up. 

The Pennsylvania position requires a little more elaboration. Throughout this election, the primary focus of media coverage has been on Pennsylvania. It's obvious why, it's a massive state with a lot of electoral votes. It is reasonable to say that whoever wins this state is favoured to win the election as a whole. It is also the state that, in most analysts' estimation, is the closest in the polls. The last few days have seen a number of polls ranging from Trump +1 to Kamala + 2, with most showing a tie, averaging out at around +0.5% Kamala. So why don't I consider this a true toss up? A few reasons. Local and district-level polling has tended to be stronger for Kamala than the big national pollsters, with most showing Kamala either matching or improving upon Democrats' numbers in 2020. This is supported by the early numbers (keeping in mind my caveat, above, about reading too much into EV data) which is showing a higher than expected turnout in Democratic strongholds. 

But more to the point, it's difficult to see where Trump gets the votes he needs to win PA. Even the polls which have been more bullish for him still only show him with around 46% of the vote. There aren't enough votes in the rurals to get him up to the 49% or so he needs to plausibly win. He would need to get some serious movement towards him among either college educated suburban voters, or non-white voters in urban districts. Right now, the pollsters showing a toss up are betting on this happening, but there isn't a lot of tangible data to support it, and certainly we're not seeing that in the early vote or local polling. Conversely, there is some suggestion that pollsters may simply be "hedging" and herding towards a tie to avoid getting egg on their face - it is worth noting that most of these polls are making a point of providing data favourable to both sides, eg a poll showing Trump ahead head-to-head, but losing in the full field of candidates, or a poll showing Trump narrowly ahead with likely voters, but comfortably behind among registered voters. There does seem to be an element of pollsters not wanting to stick their necks out too far for fear of being wrong, and with that in mind it leaves me inclined to trust the local polling and fundamentals to a greater extent. Trump absolutely has a path to victory in PA, but it requires just about everything to go right for him. You would rather be Harris right now.

In my opinion, more people should be talking about Wisconsin. Of these three states, Wisconsin is the most purple, it is the one that has been closest in each of the last five electoral cycles, and the only state to regularly vote for Republicans in statewide elections in recent years. Crucially, while Pennsylvania has been trending left across recent years, Wisconsin has not, so it's much easier to believe that Wisconsin would suddenly shift to the right of 2020 and 2022 than Pennsylvania. Let's put it another way: I can't see a situation where Kamala wins Wisconsin, and loses Pennsylvania, but the reverse is very plausible. In other words, Wisconsin is the state to watch. If Kamala fails to win MI and PA, I would be surprised. Wisconsin, on the other hand, is a true toss up, and pretty much a must-win state for Trump.

Yet, this demonstrates just how much of an uphill battle Trump faces. If Kamala does win MI and PA (which seems more likely than not) then she only needs to pick up one out of WI, AZ, NC and GA. Trump would need to run the board on all of them. 

Of these, Kamala leads the polls in Wisconsin. NC and GA are essentially tied, with some polls showing Kamala ahead, others showing Trump ahead, but with the margins negligible. Statistically speaking, you would expect Kamala to win one of these tied races. Arizona is a race where Trump leads, narrowly but consistently enough that you would have to call him a favourite (JD note: since writing, there have been a glut of polls in AZ showing Kamala ahead, so this may end up being more of a true toss-up than initially thought).

All in all, these four states are a coin flip. You would expect Kamala to win at least one or two of these just by probability. She could win more, she could win fewer. The bad news for Trump is he needs to win all of them. It's very possible that he does win all of them, but certainly the less likely outcome.

It's also worth noting the down-ballot races in these states (more on these later). Democrats are currently running with significant leads against Republicans in statewide races in WI, AZ and NC. The latter two in particular feature especially problematic Republicans with major scandals or popularity issues. The suggestion is that this could potentially have a knock-on effect on the Presidential election. While this is plausible, history tells us that the top of the ticket race tends to be the greater driver to the polls. In other words, Trump is more likely to lift up unpopular candidates, than be dragged down by them. In practice, the result will likely be somewhere in the middle, but closer to Trump's numbers. This may seem minor, but with such close margins it could be material.

Then there are the often overlooked primary and special election results from this past year. Throughout this election cycle the election results have been indicative of a far more left-leaning year than the polls, with primary and special election results very strongly bullish towards the Democrats. Democrats have consistently overperformed their polls, including in some key bellwether contests and open primaries where they were able to run against Republicans. Trump, meanwhile, generally underperformed his primary polling to the tune of 15-20%. He even managed to lose a few states, making him one of the only incumbent or quasi-incumbent candidates to do so in the modern era.

In addition, the fundamentals clearly favour Kamala, with a strong economy, indicators trending the right way for the incumbent party, and a Presidential election year which, historically, has resulted in greater turnout from Democrat-leaning demographics. When you get this close to the election, the polls matter more, but in a tied race, these factors may just prove definitive.

Lastly, there is the effect of Trump himself. Trump likes to brand himself as a winner, but the undeniable fact is that Democrats have won pretty much every election since 2016, with the Republican nominee's divisive brand of politics receiving increasingly diminishing returns with each cycle. 

So Kamala is the favourite, but that isn't really how the race has been portrayed in the media. In fact, the average person following the election might be forgiven for thinking that Trump leads in the polls, or is favoured, as that is often how this election has been covered. 

Why the disparity between the data and the coverage? Well there's a few reasons. The first is that the "conventional wisdom" seems to hold that Trump will over-perform his polls, as he did in 2016 and 2020. To be blunt, this view is not a particularly sharp one. Polling errors are rarely consistent from one election to the next, even with the same candidates involved - in fact, no political party has ever overperformed polls in three consecutive presidential elections. Pollsters generally adjust their methodology to account for such errors, and in many cases are more likely to overshoot than repeat the same error. 

More to the point, the "polling errors" in 2016 and 2020 are a little overhyped. The 2016 polling miss resulted from the high number of undecided voters breaking for Trump, which analysts either missed or ignored - there are fewer undecided voters in 2024 (practically zero, compared to double digits in 2016), and so a much lesser risk of this happening again. The 2020 polling miss was nowhere near as substantial as claimed, and also not consistent from state to state. For example, Pennsylvania actually had zero error in 2020, while Georgia polls were actually biased towards Trump, things which never seem to be mentioned when the media discusses polling errors.

The simple fact is that there is no particular reason to think the polls will underestimate Trump. The polls in 2024 are not really comparable to either 2016 or 2020. If anything, they most closely resemble the 2022 polls, which underestimated Democrats significantly. 

The most notable feature of 2022 polling was the extent to which polling averages were influenced by, frankly, junk polls. This was driven by well-known propaganda pollsters like Rasmussen and Trafalgar, but also by a new crop of unknown polling organisations which seemed to pop out of the ether with no track record. These pollsters deluged the averages and created the impression of a right wing wave election which never materialised. By contrast, the high quality pollsters indicated a close 2022 election. As a reminder, this blog was one of the few forecasts to correctly identify this phenomenon and predict a better than expected election night for Democrats. It's important because we're seeing the exact same thing in 2024. To be clear, polling averages are still invaluable, and the best indicator we have as to an electoral outcome, but some critical scrutiny is required.

It's also an unfortunate fact that there are simply far fewer high quality polls than there used to be, and this lack of clarity is contributing to the uncertainty among forecasters. By comparison, September and October 2020 saw the release of some 20+ Pennsylvania polls from high quality pollsters, this year we have seen basically nothing until the final days, and even then only from a small number of pollsters. Why this may be the case is a topic for a separate, much longer discussion.

All of this is to say: the polls are close. There's no particular reason to think Trump will over-perform them. There may in fact be greater reason to think that Democrats will over-perform. Your best bet for predicting the election is, as always, to follow the average, while filtering for high quality, reputable pollsters.

Ultimately, you would have to favour Kamala based on the data, and Trump faces the more uphill battle to reach 270. Make no mistake, the polls are super close, and Trump could still win this, but it would be a surprise. He is certainly not the favourite, as some are suggesting. If forced to make a firm prediction, I would say Kamala to win WI and one of GA/NC. WI has polled narrowly, but consistently to Kamala's favour. NC is close enough that I think the down-ballot races could give Kamala the edge. NV is a total toss up, but the fact that this state has been trending Republican in recent years gives me the sense that it will go for Trump. At a push, I would make Kamala to be a 60/40 favourite right now, which is pretty close to where the highest rated forecasters currently have the race.

Caveat, estimating the result in the House is far from precise. The balance ultimately comes down to individual races being won or lost, which depends on a myriad of factors beyond simply the national vote. The national vote merely serves as a useful estimate for a forecast that would ultimately require detailed analysis of hundreds of individual races. So consider this an estimate with a wide margin for error. Nevertheless, the odds remain in the Democrats' favour.