KEY POINTS FROM THIS ARTICLE
— A Des Moines Register/Mediacom poll showing Kamala Harris trailing by just 4 points in Iowa got a lot of attention over the weekend, but the final DMR poll in that state is sometimes more accurate and more Republican than ones released around this time in the election calendar.
— Polling in states that are somewhat competitive for president but outside the group of 7 that is likely to decide this election generally reflect what happened in 2020, although Donald Trump is usually performing slightly better than the actual 2020 results in those states, much like the polling in the key battlegrounds that we covered last week.
— Polls in states that are more clearly blue or red sometimes differ more from the 2020 results, but it’s not uncommon for the dominant party in a given state to be understated in pre-election polling.
Polls in the non-battlegrounds
Of all the polls that came out over the weekend, the one that seemed to get the most eyeballs from poll junkies was one that did not survey one of the 7 key battleground states in this election.
Instead, the poll of interest came in Iowa, a one-time swing state that shifted hard to Donald Trump, and the poll itself was a Des Moines Register/Mediacom poll conducted by J. Ann Selzer, who has a well-earned reputation for accurate final polls in her state.
Selzer found Trump leading Kamala Harris 47%-43% in a poll released Sunday, a lead just half that of Trump’s 8-point victory there in 2020. This also was a huge shift from Selzer’s previous poll, taken in mid-June when Joe Biden was still a candidate, showing Trump up 18 points in Iowa. So while no one expects Iowa to be a prominent part of this year’s presidential battleground—it remains Safe Republican in our ratings, despite this poll—this was an encouraging finding for Democrats in a state that shares some demographic similarities with the broader Midwest (particularly Wisconsin).
In our overview last week of 2016-2020 polling error in the key swing states, we did note that Wisconsin’s polls may be too rosy for Democrats now, and it’s also possible that this Iowa poll is a little too rosy for them as well.
It has not been uncommon in recent years for Democrats to do better in a Selzer Iowa poll further away from the election than they actually perform (or that the final Selzer poll shows when it typically is released right before the general election). Here’s a brief history of the poll over a number of races in the last 6 cycles:
2012: A late September poll showed Barack Obama up 4 points, and the final poll showed him up 5; both came very close to his eventual victory of a little under 6 points.
2014: A late September poll showed Joni Ernst (R) leading Bruce Braley (D) by 6 points in an open-seat Senate race. But an early October poll showed Ernst leading by only 1 point. The final poll showed Ernst back up 7, and she won by 8 points.
2016: In early October, the poll found Trump leading Hillary Clinton by 4 points. The final poll showed Trump by 7, and he won by a little under 10 points.
2018: In mid-September, the poll found Fred Hubbell (D) leading Kim Reynolds (R) by 2 points for governor. The final poll showed the same, but Reynolds outperformed the poll and won by a little under 3 points.
2020: A mid-September survey showed Biden and Trump tied. The final poll had Trump leading by 7 points, extremely close to his 8-point victory. That same poll in the Senate race that year had Sen. Joni Ernst down by 3 points in mid-September. The final poll had Ernst up by 4, and she won by about 6.5 points.
2022: A mid-October poll showed Sen. Chuck Grassley (R) leading his Democratic challenger by just 3 points. The final poll had him up by 12, nailing the final margin exactly. Both the October and final polls showed Reynolds leading in her reelection bid by 17 points, very similar to her final 18.5-point margin.
So there are a number of instances where an earlier Selzer poll had the Democrats doing better than they would ultimately perform, only for the final poll to hit very close to the final margin, with the only real miss coming in the 2018 gubernatorial race (the history here, particularly the final poll in these races, is highly impressive).
Our suspicion is that Trump is going to do better than just winning by 4 points in Iowa, and that the final Selzer poll may very well indicate that he is doing better than he is doing now. But when we are gathering all of the information we can over the final weekend before the election, this poll is one of the indicators we’ll be looking at, both because it has a good track record in Iowa and because it may have some broader implications—namely, we think it correctly suggested that Trump was going to do better in the Midwest than many other polls in many other states suggested in both of his elections.
This closely-watched poll of a state that is not at the center of the 2024 Electoral College battlefield leads to an obvious question: What are the polls saying about other states outside the “Magnificent Seven” battlefields of Arizona, Georgia, Michigan, Nevada, North Carolina, Pennsylvania, and Wisconsin?
Tables 1 and 2 show the RacetotheWH polling averages for states that were somewhat competitive in 2020 (Table 1, states decided by more than 3 points but less than a dozen points) and not competitive at all (Table 2, states decided by a dozen points or more). We only included states that had at least 2 polls conducted after Kamala Harris entered the presidential race by 2 different pollsters; as you can see, that low bar still excludes many states, particularly in Table 2, where there are only a handful of states listed. Positive numbers indicate Democratic leads/overperformances, negative ones indicate Republican leads/overperformances.
Table 1: Current polls vs. 2020 results, somewhat competitive presidential states
Notes: We used RacetotheWH for this average instead of the blended RealClearPolitics/FiveThirtyEight average we used for the key battlegrounds piece last week because RacetotheWH included averages for more states (also, we have to be honest, using one average is easier than using two—we decided to include this frank bit of honesty for those diligent enough to read the notes under tables). *NE-2 polls are just for the single electoral vote in Nebraska’s 2nd Congressional District.
Source: RacetotheWH as of Tuesday morning for poll averages, Dave Leip’s Atlas of Presidential Elections for actual results; NE-2 results are from The Downballot, formerly Daily Kos Elections.
The polls are very reflective of the 2020 results in the states listed in Table 1. One clear trend since Harris entered the race is that Democratic margins in some of the lighter blue states (Minnesota, New Hampshire, and Virginia) have regenerated. When Biden was in the race, these states seemed like they could potentially be up for grabs, but they’re all showing Harris with decent-sized margins.
The pair of Florida and Texas—each of which have Senate races that Democrats are watching as they search for anywhere to play offense on this difficult Senate map—are each polling almost exactly at their 2020 actual margin, with Trump doing very slightly better than his 2020 result in each.
That is generally true across these states—to the extent the average is different than the 2020 margin, it’s Trump who is usually doing better. That includes the single electoral vote in Nebraska’s 2nd Congressional District, where Harris has a lead that is very comparable to but not quite as large as Biden’s actual 2020 margin. This single electoral vote could be important if this election is very close (the single electoral vote in ME-2, won by Trump in 2016 and 2020, did not have enough polling to be included in this average). Overall, the polling in Table 1 looks similar to the averages we put together last week for the 7 key swing states: very reflective of 2020 with Trump usually showing slightly better margins than the actual 2020 results.
Note that Iowa is not included in Table 1: The Selzer poll is the only recent survey there. Ohio, thanks to the state’s highly competitive Senate election, has been polled to some extent recently, and the poll average looks almost identical to the 2020 results. This is a change from 2020, when polls on balance actually showed a very close presidential race in Ohio: Trump led in the final RealClearPolitics and FiveThirtyEight averages by only a point, leading to a 7-point misfire that was identical in size to the notably large aggregated 2016/2020 polling miss in Wisconsin we documented last week (Trump only had a similarly small lead in Iowa in the final 2020 averages too, with the Selzer survey standing out as the only poll there that reflected the actual result). Ohio, while much bigger than Iowa, has moved off the list of presidential battlegrounds in a manner similar to how Iowa has. The polling in Ohio being consistent with 2020 also suggests the recent Selzer Iowa poll is perhaps a little too good for Harris compared to what the final result there will end up being.
The presence of competitive down-ballot races, such as Montana and Maryland’s Senate races, helps explain why we have some presidential polling in those otherwise uncompetitive states. The big states of California and New York, which have prominent in-state pollsters, also have been polled at least twice since the Harris switch. Here we see a little more difference from 2020, particularly in the blue states.
Table 2: Current polls vs. 2020 results, uncompetitive presidential states
Source: RacetotheWH as of Tuesday morning for poll averages, Dave Leip’s Atlas of Presidential Elections for actual results.
Missouri and Montana, states that should vote for Trump by at least 15 points, look very similar to the 2020 results, albeit slightly better for Democrats. The other 3, all states that Biden won by more than 20 points, show a smaller average Democratic margin than the 2020 results. This is especially true of New York, although Harris is still up about 15 points there in the polling average.
The dynamic here reminds us of something that poll-watcher Dan Guild (@dcg1114 on X/Twitter and a past Crystal Ball contributor) has previously found and reiterated to us in an exchange of messages on Monday: Polls, particularly when the election is still somewhat far away, are likelier to understate the majority party in a state they are likely to win than they are to understate the final margin in a more competitive state. In other words, it’s not uncommon for a red state to look less red than it will ultimately be, or a blue state to look less blue than it ultimately will be, in polling released roughly 1-2 months before the election. This history adds a little extra context to the recent Iowa poll, too (Iowa is not a very red state, but it’s not purple either).
However, it’s also worth noting that polls in California and New York at this time four years ago, if anything, were overstating the ultimate Democratic margin in those states, and we have also heard from sources that Harris’s polling performance in key House districts in both states is lagging Biden’s 2020 showing. This of course doesn’t matter for the Electoral College—Harris is going to easily win both of these states—but the margin itself may matter in the many districts across both states that Biden carried but that have Republican House incumbents. So this is something to monitor, even though it may also be a mirage: Guild told us that while polling from this time of the election did overstate Biden’s eventual 2020 margin in California and New York, polls from the 2000-2016 presidential elections at this time of the calendar were understating Democrats in each state.
It’s also possible that if the Democratic presidential margin is smaller in California and New York, that could have the effect of reducing the pro-Republican bias in the Electoral College if Harris holds up better in the key swing states than she does in the landslide states. Together, California and New York should cast around 15% of the total presidential vote, so Democratic slippage in both states, if that actually happens in a significant way, could have an impact on the overall national popular vote.
Generally speaking, we would advise healthy skepticism in clearly red or blue states that show something wildly different from 2020. Polls are a great tool, but the results themselves from 2016 and 2020 are too, particularly (and as we argued last week) because Trump is on the ballot again and the political environment does not seem to be wildly different from 2016 or 2020.