How does the model handle the RFK Jr. dropout?

A brief update on the presidential race, and plans for the site for the near future.

2024 Presidential Forecast

Last Friday, perhaps one of the more predictable events in the 2024 election cycle took place, a rare occurrence in this age of uncertainty and what seems like constant changes: Robert F. Kennedy Jr., the son of former Attorney General and assassinated presidential hopeful Robert F. Kennedy, and the nephew of JFK, dropped out of the race. Or… did he?

Kennedy made it explicitly clear that he was only suspending his campaign, but not “terminating” it.¹ But although this is usually the case with candidates dropping out, his definition of “suspending” is a bit different. It’s more like a partial suspension because he only wants to (and only can) remove his name from the general election ballot in some states. Specifically, he indicated that he wanted to withdraw from the ballots of ten battleground states, as he was concerned he would hurt Donald Trump’s prospects if he remained an option for voters.

But wait, why is an independent candidate trying to so obviously tilt the scales towards one of the major two parties? Well, it doesn’t take much to see that this was always his goal. Kennedy appears to have run a long, unsuccessful, and frankly, wild campaign for the sole purpose of hurting Democrats. He first ran for the Democratic nomination in October 2023, hoping to either actually beat Biden or at least attack him enough. When he realized that was going nowhere, he switched strategies, running as an independent, hoping he could siphon enough of the Biden electorate to stop his re-election bid.

And for months, Republicans, Democrats, and the polls weren’t sure who was losing support to RFK. It seemed like one poll would have Biden losing on the net margin between a two-way and three-way race, and then the next would have the same for Trump. After the June 27th debate, however, and especially after Kamala Harris became the nominee, it’s become very clear that RFK hurts Trump. I, unfortunately, don’t have my polling averages set up so that we can easily compare between the head-to-head and full-ballot polling, so you’ll have to take my word for it when I say that RFK was a spoiler against Trump, not for him.

This really shouldn’t be surprising when you consider how Kennedy and his running mate, Nicole Shanahan postured themselves in the final weeks of their campaign. They went after the Democrats for switching nominees, using the same GOP claims of a “coup” to disparage Harris. They portrayed Trump’s survival of his assassination attempt as heroic and as if it proved he was presidential. Kennedy attacked COVID-19 lockdowns and mandates, a popular (and IMO, strong) argument from the GOP. Kennedy attacked Harris for “failing as Border Czar”, and the list goes on. Kennedy was fairly close to being a Republican candidate without being Trump.

But enough of the past. What’s changed in our polling averages and forecast as a result of this development? Well, nothing. I made the decision (it was kind of made for me too²) to lump all third-party/independent candidates as “Other(s)” across the board. So, when you look at our polling averages, the figure in yellow represents Cornel West, Jill Stein, Chase Oliver, (sometimes now) RFK Jr., and any other candidates. Sometimes surveys simply report an “Other” result, and that gets added too. So now that RFK has dropped out, I’m just going to keep adding polls as I did before. Some pollsters are looking at which states he will be removed from the ballot in, while others are just listing him as an option everywhere. It’s tricky to precisely deal with that myself, so I’m not changing anything in terms of the polling averages.

As far as actual movement in the model, candidates other than Harris & Trump are now expected to get just 5.3% of the national popular vote, a slight decrease from 5.9% the day RFK dropped out. However, they’ve been trending downward anyway since the model launch (a little over a month ago) where they were expected to receive 6.7%. In 2020, for comparison, they only received 1.9%. I believe that this number will continue to fall up until Election Day, but I’m not sure where it will end up.

The state of the race

But enough about RFK. The more important story is that of Kamala Harris rising in the polls and continuing to have the momentum, at least for now. She currently has around a 66% chance of winning, an all-time high in our model, is favored in states totaling 276 electoral votes, (although the model has her winning 287 on average), and is forecasted to win the popular vote by 3.5%, just a percentage point shy of Biden’s winning margin in 2020. She does seem to be getting a slight convention bounce, and I should note that our forecast DOES NOT adjust for this. Other models like Nate Silver’s (and I believe 538) will purposely limit and boost a candidate gets in the immediate weeks following their convention because those boosts usually don’t last in the long run. I don’t though, in part because these boosts seem to be getting smaller and smaller every four years, and in part because there’s just not much point. Sure, the model could be overconfident for Harris for a week or two, but if any bounce goes away within two weeks anyway, the model should be back to a realistic output in two weeks, well before the election. I’d care a lot more about a convention or debate adjustment if it was right before the election. For example, the first debate (in less than two weeks!!!) is well before Election Day, so I’m not accounting for it.

It’s safe to say (based on our model) that Harris is the favorite, and the race isn’t a toss-up. Most models have her at around a 60% chance, so ours is just a little more bullish, although Silver has Trump minimally favored, and I think his model has been over-adjusting for the convention because Harris still leads nationally and in almost all of Silver’s polling averages. But, if you ever looked back at FiveThirtyEight’s model in 2016, (then actually run by Silver) you’d find that Hillary Clinton had a 71% chance of winning. She didn’t. The 538 model wasn’t wrong at all and was well-calibrated compared to its competitors. It’s just that, well, a 29% chance outcome happening is not that surprising. I wanted to emphasize uncertainty interestingly with our forecast, and you’ll find a handy analogy right below the topline odds on the forecast page. Right now, for example, it says that a Trump win is about as likely as Lebron James making a three-pointer. Lebron makes plenty of threes, and for any given shot, the chance of it going in is about the chance that Trump has of winning right now.

As long as Harris can hold steady with her national/swing-state margins, her chance of winning will go slightly up, because the model accounts for less uncertainty and smaller polling errors the closer we are to Election Day. If she drops a little bit, her odds may remain somewhat stable, and if she grows her lead, she will rise even further. The upcoming debate on the 10th could be huge in determining whether or not she gets another bump. I’m now going to do a bit of looking at the swing states and how they compare to 2016 and 2020, and as most of the data is paywalled on the forecast, this section of the article will be too.

I hope to write more updates like this, and regret not being able to push out more lately. I’ve mainly been focused on getting more models out (our Big 8 Football and U.S. Senate forecasts were released, and a U.S. House polling tracker is in the works) and have a lot going on right now too with school quickly approaching. I’ll do my best to get at least two full articles a month if not more. Model updates for the presidential forecast will continue to run daily, while Senate/House updates will be more infrequent. Big 8 forecasts will update once each weekend.