G. Elliott Morris of 538 (the man who took Nate Silver’s place) has thoughts on the polls. I think you will find it reassuring:
Where does the 2024 election stand one year out?
We modeled six scenarios for the presidential race.
American voters cast 158 million ballots in the 2020 presidential election. Yet the winner was ultimately decided by about 43,000 voters across Arizona, Georgia and Wisconsin — the states that carried President Joe Biden over the 270 Electoral College votes he needed to win the presidency. The 2016 election, also closely contested, was similarly settled by about 78,000 votes in Michigan, Pennsylvania and Wisconsin. And if the 2024 election were held tomorrow, it would likely be very close too.
At least, that’s what the polls say. According to an average of national 2024 general election polls I’ve run using 538’s current polling average methodology, Biden and former President Donald Trump are currently neck-and-neck among likely voters, with Trump at 42.9 percent and Biden at 42.4 percent. Support for Robert F. Kennedy Jr. and Cornel West, both of whom have announced runs as independents, is hovering around 11 percent and 4 percent, respectively. Although there’s a long way to go before next November, these numbers suggest that 2024 at least starts out as a close election.
Today, 538 is also happy to release our live-updating average of polls of the national generic congressional ballot, which ask Americans which party they intend to support for Congress in 2024 without naming specific candidates. Since this is our first general election polling average of the 2024 cycle, we’ve also taken the opportunity to update our polling average methodology to adjust polls from partisan firms or sponsors (which tend to favor the party that releases them) and decrease the weight given to polls of all adults or registered voters in horse-race averages (we prefer likely voter polls for these averages). As of Monday at 3:30 p.m. Eastern, the average shows Democrats and Republicans locked in a dead heat for the House popular vote — Democrats technically lead by 0.4 percentage points, but that’s well within the margin of error for our average. That’s even closer than generic ballot polls were on Election Day 2022, when Republicans were leading our average by 1.2 points (they ended up winning by about 1.6 points, after accounting for uncontested seats).
When the polls are this close, it’s important to remember that which polls you include and how you average them could lead to different conclusions about who is currently “winning” or “leading” in a given race. The important takeaway is not who’s up or down, but that the election is close in the first place. We have much more confidence that an election, if it were held today, would turn on a knife’s edge than we do that Trump or House Democrats are leading.
Of course, the 2024 election will not be held today, so it’s incumbent on us to ask …
How useful are early polls?
The answer, frankly, is “not very.” For one thing, the two major parties have not officially selected their nominees. While Trump leads handily in both state and national polls of the Republican primary, there is still time for one of his challengers to turn their luck around — or for a major news event to significantly disrupt his campaign. Biden, too, has primary challengers, and there are persistent whispers that he still may decline to seek reelection due to his age. Neither party will officially pick its nominee until next summer; until that point, these polls are little more than hypothetical exercises.
The general election is also still a year away, leaving plenty of time for the polls to change. At this point in the campaign, most past presidential nominating contests have not had clear leaders, and it was anybody’s guess what issues would be most important to voters come the general election. That makes early polls of little use to campaigns and prognosticators. Research from political scientists Robert Erikson and Christopher Wlezien shows that, between 1952 and 2008, polls taken 300 days before the general election had no predictive value. In statistical terms, they found that polls have an R-squared value of roughly 0 in January of an election year. That’s basically the track record of an (untrained) monkey throwing darts at a dartboard. Plus, as of today, we’re currently 365 days out from the election — quite a bit further out than Erikson and Wlezien were even willing to look, given the variability of their data.
True, there is persuasive evidence that presidential vote preferences are more stable now than they were in the 20th century, and that could make early polls more predictive. But given how often early polls have misled us in the past, it would still be risky to place your faith in them this far in advance.
What about the Electoral College?
Another caveat still is that the national popular vote does not determine the winner of U.S. elections — either at the House or presidential level. It’s the states and congressional districts that do the deciding.
Since we don’t have very many state-level polls of the 2024 presidential election (and I wouldn’t really trust them if we did), the safest assumption at this point is that states will vote how they did in 2020, plus or minus some amount of (1) national shift toward or away from Biden and (2) state-level noise. We can only make guesses at these numbers right now, but I put together a bare-bones, no-polls model for the presidential election that can simulate a range of outcomes given various inputs for those two settings. Scroll down to the bottom of the article if you want to read my full methodology.
Let’s look at six possible scenarios for the 2024 election under this model:
–Scenario A is essentially the “2020 rerun” scenario: I take Biden’s state-level margins from the 2020 election and add the expected level of correlated noise.
–Scenario B is roughly what the national polls indicate today: a tied popular vote, plus noise again.
–Scenario C again uses Biden’s 2020 performance as the starting point, but it also assumes that the Electoral College’s bias toward Republicans will shrink between 2020 and 2024.* We got hints this may happen from the 2022 election, when the presence of abortion rights on the Michigan ballot and other state-level factors gave Democrats a relative boost in northern battleground states.
–Scenario D is another tied popular vote, but this time with the lower Electoral College bias as well.
–Scenario E uses the 2020 election as its starting point but increases the Electoral College’s bias toward Republicans by 1 point relative to the last election.
–Scenario F is the worst scenario for Democrats, simulating what would happen with a tied popular vote and a higher Electoral College bias.
Here is what the model says about how many electoral votes Democrats are likely to win in each scenario, and what odds of victory that translates to:
The 2024 election is likely to be close
Democrats’ range of electoral votes and chances of winning in six different sets of simulations of the 2024 presidential election
| ELECTORAL VOTES FOR DEMOCRATS | |||
|---|---|---|---|
| SCENARIO | 5TH PERCENTILE | 95TH PERCENTILE | CHANCE OF WINNING |
| 2020 repeat | 222 | 388 | 61% |
| Tied popular vote | 197 | 321 | 23 |
| 2020 with lower Electoral College bias | 226 | 388 | 71 |
| Tied popular vote with lower Electoral College bias | 199 | 335 | 34 |
| 2020 with higher Electoral College bias | 217 | 388 | 54 |
| Tied popular vote with higher Electoral College bias | 197 | 321 | 20 |
Based on a rough model of U.S. presidential elections that predicts future election results based on (1) historical year-to-year variance in state-level Democratic vote margins and (2) how correlated vote shifts are between states. The “percentile” columns show the upper and lower bound of our prediction for Biden’s electoral votes in each scenario.
As you can see, Biden is favored in the scenarios where he wins the popular vote by a 2020-esque margin, and Trump is favored in the event of a tied popular vote. But in each scenario, the underdog still has a respectable chance — at least 1-in-5 — of winning. Most simulations don’t point to an Electoral College landslide, either. The probability of a 365+ electoral vote win (on the magnitude of Barack Obama in 2008 or Bill Clinton in 1992) is less than 1-in-5 in all scenarios. In other words, neither party can assume an easy path to victory.
It’s worth re-emphasizing: It is still early days. The 2024 election will be decided on Nov. 5, 2024 — 365 days in the future. But all of the data we currently have points to another close race.