Election Modeler Split Ticket Gives Harris 53% Chance Of Winning
S E E D E D C O N T E N T
The Case For Kamala Harris
If you’re of the thought that Kamala Harris will win, you’d be excused for being bewildered by the prediction market swings over the last few weeks. The bearishness on her comes from three main angles: the previous cycles’ under-estimation of Trump, the recent tightening in national surveys, and the early vote.
But the fundamentals look quite good for Harris, at least compared to polling. The most significant of these is Washington’s top-two, nonpartisan “blanket primary”, which actually points to a result around D +3–4 nationally . Moreover, the non-urban part of the state, which cratered for Democrats in 2016 and showed surprisingly weak results in 2020, now shows a swing to the left in 2024, as Democrats gain with secular and non-college white voters.
In 2016, 2018, and 2020, the results in Washington portended the Democratic struggles in the Midwest, and they warned of a race that was nowhere near as blue as surveys implied. This time around, however, they suggest that the party may be in better shape there than the polling might suggest. While they don’t indicate a blue wave, they do point to the party holding the demographically-similar Blue Wall states of Michigan, Pennsylvania, and Wisconsin.
We’ve also seen the early vote analysis people have peddled, and we’ve extensively detailed our skepticism regarding them . We’ve already given examples of why applying 2020 splits by vote method (mail, early in-person, and election day) is an incredibly poor method of analysis, and we’ve also highlighted that election day is set to be a lot bluer than it was in 2020 and likely even 2022, as Democrats are simply returning their ballots later .
Most of the GOP strength in early voting is simply because a non-trivial chunk of their 2020 election day voters are now voting early, as Trump and the GOP embrace early voting. In North Carolina, 40% of the registered Republicans who voted on election day in 2020 have now voted early, while just 20% of Democrats have. In Nevada’s Clark County, this number is 33% for Republicans, while it’s just 15% for Democrats. And in Georgia, which has no party registration but is heavily polarized by race, roughly 33% of white voters who voted on election day in 2020 have voted early; however, for Black voters, this number is just 20% .
There are two ways to reconcile this reality. The first is to suggest that Democrats are simply going to vote later, and that election day will be substantially bluer (and less white) than it was in 2020. This would track with the return patterns we are seeing and the statistics cited above. You can see an example of this in a chart assembled for us by Daniel Sokul , who has tracked Harris County’s returns in Texas. Democrats are returning their ballots far later than they were in 2020, and as the election nears, their voting rate has dramatically picked up (a stark reversal from 2020).
If delayed Democratic turnout does not materialize, the only possibility is that Republicans are going to enjoy a landslide turnout advantage, with Democrats essentially boycotting the election en masse . Given the incredibly poor track record that unqualified early voting analysis tends to have, we lean towards the first theory — in a presidential cycle, both parties will likely show up, and any turnout advantage the Republicans have will likely be marginal at best in most states, though it reduces the turnout the GOP will need on Tuesday.
Finally, the majority of skepticism and pushback on Harris that we receive comes from people burned from previous polling misses. To this, we’d point out that polls rarely continue to miss in the same direction, because pollsters adjust for what they missed the last time around. We have no evidence to suggest that polls are set to overestimate Trump once more.
To begin with, we have strong evidence that the response environments this time are simply different. In 2016, we now know that a primary driver of polls missing on Trump was due to not weighting on certain factors, such as education, that had become dividing lines in American politics. In 2020, the cause of the miss was different — polls were simply far more likely to reach Democratic-leaning voters, who were staying home amidst COVID and were thus more likely to respond.
Looking at the response environments, we can see more evidence of the underlying shift in reality. Take a look at YouGov’s data and you can see that Democrats have become significantly less likely to respond to polls over the course of the cycle, which is extremely different from how polling worked in 2016 and 2020 (and even immediately post- Dobbs in 2022). For the first presidential cycle in a while, pollsters are actually getting a pretty good balance of Democrats and Republicans in their raw samples, even before stratifying and weighting their electorates.
There are real reputational and psychological costs to repeatedly making these types of mistakes, and we can observe that pollsters have taken a lot of steps to ensure they get enough Trump supporters this time around (as they did in 2022). Whether through weighting by recalled vote (which explicitly tries to get a set number of self-identified Trump supporters in a survey), through more GOP-friendly likely voter screens , or other methodological choices , it is clear that something has changed.
As a last point, there are a shade too many tied polls in Pennsylvania (at least, compared to what would mathematically be possible, even after accounting for how variance shrinks when you weight surveys on common factors). If this is true, then the aggregate will likely miss by more than it has in the past . We have no way to know which direction this error cuts in, but it is probably worth noting that registered voter polls in Pennsylvania consistently find Harris up, even as likely voter (LV) polls find her down, and that there is a curious discrepancy between national and statewide LV polls .
Then, there’s the whole matter of outliers and that Selzer poll. And while it’s just one poll, it’s perhaps one of the only ones worth its weight in gold, as Selzer’s outliers have consistently been more correct than the results of the industry as a whole, over the course of a full decade.
So, About That Selzer Poll.
Do you get deja vu?
Four years ago, the race in Iowa was looking extremely tight. FiveThirtyEight’s polling average actually had Joe Biden practically tied with Trump in the Hawkeye State. At the time, this hardly seemed unusual; polls consistently said Biden was poised to make a big recovery with the Midwestern working class whites who had abandoned Hillary Clinton, and Iowa was the epicenter of that shift. The Selzer poll from September had found both candidates tied, 47–47, reinforcing this belief.
Then, Selzer dropped a bomb in the final week, with a survey that showed Trump ahead 48–41. This was a finding that suggested Trump was actually well-positioned to replicate his 2016 performance in Iowa, with big implications for the rest of the Midwest. It was Trump’s best poll in months, and arguably his best one of the entire year. And at the end of the day, Selzer turned out to be right, just as she was in the 2008 Democratic caucuses, in the 2014 Senate election, and the 2016 election — her outlier polls, once again, continued to consistently catch something that no other pollster was seeing.
This was, however, a huge deviation from basically every other poll of the cycle, and came with a number of caveats. Selzer is one of the most respected pollsters in the country, but no pollster is free from the laws of the universe that govern statistics. Sampling error is always a possibility, and her polls, even with their phenomenal track record, are not infrequently off by a point or two. And, of course, it was just one poll — the data in aggregate said something different.
All of these caveats apply to the astonishing poll that Selzer released this evening, where Kamala Harris leads Donald Trump, 47–44, in Iowa . It is inarguably one of the best results for Harris of the cycle. Although Iowa polling has been sparse (neither candidate is treating it as a swing state, given Trump’s 8-point margin there in 2020), there is scant evidence anywhere of the surge in support Harris would need to flip Iowa.
Nevertheless, Selzer’s track record is close to immaculate, and consequently her polls are given substantial weight in our model. The model is ultimately skeptical of this result, which goes against almost the entirety of the available data. Just today, for example, Emerson College found Trump ahead 53-43 in Iowa. Still, this poll alone moves Iowa to Leans Republican in our forecast, with Trump being a 76% favorite to keep the state in his column — a marked decline from the 88% chance the forecast gave him just yesterday.
If this poll is anywhere near correct, the implications are vast. It is difficult to imagine a universe in which Kamala Harris achieves the level of white support necessary to be competitive in Iowa, but does not win Michigan, Wisconsin, Pennsylvania, and Nebraska’s second congressional district – and with it, the election.
While no other poll has shown quite this monumental of a shift, if you squint, there are perhaps hints of something similar happening in polls of similar states. Harris has polled exceptionally well in Nebraska’s second congressional district, and some polls of Nebraska statewide show a shift toward her as well. There was also a recent poll of Kansas that only had Trump up 48-43, a seeming outlier, but one perhaps worth taking a second look at in the wake of this poll.
Does this poll imply a Harris landslide? That’s one interpretation we’re skeptical of — even setting aside the outlier nature of this poll, it is worth noting that even a perfectly accurate Iowa poll cannot say much about states like Georgia or Arizona, where the whites vote differently from the Midwest. Barack Obama, for example, won Iowa easily en route to losing both of those states, and the idea of Harris possibly replicating something closer to the Obama coalition seems far-fetched. (But, then again, so does the idea of Harris being competitive in Iowa at all).
We have criticized pollster “herding,” a dynamic in which polls crowd around the expected result, instead of being willing to publish outliers or other surprising results. One thing is clear: Ann Selzer is not herding. It is possible that this becomes the first Selzer poll to seriously miss, through no fault of its own — again, sampling error can be unavoidable. On the other hand, perhaps Selzer is once again correct, and pollsters have been seriously underestimating Harris, an outcome that would remind people, at the very least, that polling error is unpredictable.
In three days, we will have our answer.
Tags
Who is online
417 visitors
I wasnt familiar with this site until a few days ago, but apparently they have a very good reputation, and you can see from this article that they go into great detail, some of it off the beaten path. For example concluding that the non-partisan primary in Washington state months ago may indicate Harris strength in the blue wall states of the upper midwest. They also mention polling in Iowa that may foretell Harris strength elsewhere.
Split Ticket gives Harris a 53% chance of winning the electoral college, sort of a mirror image of the 54% chance Nate Silver gives Trump.
The Iowa poll, sponsored by The Des Moines Register and conducted by Ann Selzer, is encouraging. The switch to Harris is being driven by older women.
If Harris wins Iowa we may be looking at a minor Harris landslide.
That's what it portends. Remains to be seen, though.
If Iowa really is a toss up it may mean that some people that were thought to be gone may be returning to their senses.
That would be a very welcome development.
Hillary Clinton’s odds of winning the presidency rose from 78% last week to 91% Monday before Election Day, according to CNN’s Political Prediction Market.
With hours to go before Americans vote, Democrat Hillary Clinton has about a 90 percent chance of defeating Republican Donald Trump in the race for the White House, according to the final Reuters/Ipsos States of the Nation project.
Nate Silver’s FiveThirtyEight website this week put Clinton ’s chances of victory at the highest they’ve ever been. On Monday, the probability of the Democrat winning had reached 96.4 percent and was at 86.3 percent Wednesday.
The difference bewteen 2016 and today is that in 2016 the biggest asshole in America hadnt been infesting everyones lives for 8 years. He was just getting started.
That's a valid point, I posted the 2016 polls just to show how far off they were.
Do you think that's the only difference?
Nate Silver this morning
and tonight
... to have a prominent, high-quality pollster like this (Selzer) at a time when most other pollsters are herding toward the consensus suggests the possibility that other pollsters could be lowballing Harris .
Silver Bulletin 2024 presidential election forecast