KEY POINTS FROM THIS ARTICLE
— A persistent finding in swing state polls is that Democrats are doing better in Senate races than Joe Biden is doing in the presidential race.
— At the topline, 2016 and 2020 produced hardly any split presidential and Senate results, suggesting that perhaps the presidential and Senate polling should converge.
— However, even in those years, there still was variation from state to state between the presidential and Senate margins.
— Focusing on the Senate races in the presidential swing states distracts from the races that will truly decide the Senate majority: red state seats with Democratic incumbents, Montana and Ohio.
What’s with the swing state Senate polls?
If you follow Senate elections at all, and if you’re reading this we’re sure that you do, you may very well be aware of a fact that we and others often cite.
In both the 2016 and 2020 elections, the party that won each Senate race was the same as the party that won that state for president, with just one exception: In 2020, Sen. Susan Collins (R-ME) won reelection despite President Biden winning her state for president. Our J. Miles Coleman tracked the history of split Senate/presidential results in the post-World War II era; such split results used to be common but have been rare in the past two presidential election cycles. In another Crystal Ball article, Miles documented the decline of Senate/presidential ticket-splitting over the last six presidential cycles.
The presidential and Senate results have become increasingly correlated. Gary Jacobson, one of the nation’s leading experts on congressional election trends, found that “the correlation between the major party share of votes won by candidates for president and Senate was 0.95 in 2020, slightly higher than the previous record set in 2016 (0.94).”
The big picture trend is that presidential and Senate election results are more connected now than they’ve been in the past. So it has been a notable curiosity that polls are currently showing a disconnect between presidential and Senate results in the key swing states.
The presidency will likely be decided by how many of the following six states Biden can hang onto, all of which he carried in 2020 but by less than his 4.5-point edge in the national popular vote: Arizona, Georgia, Michigan, Nevada, Pennsylvania, and Wisconsin. Of these states, 5 of the 6 have Senate races (all but Georgia). Democratic Senate candidates generally lead in those states while Joe Biden does not (more on the specifics below in Table 3). What follows is a look at what happened in 2016 and 2020, in which there was more difference in the presidential and Senate results by state than those big-picture (but true) observations above might suggest, as well as some other points about how we are thinking about the potential for differences between presidential and Senate outcomes this year.
1. The recent history is actually more nuanced
The recent history cited at the top of this article about a lack of split-ticket results in the 2016 and 2020 cycles does obscure that the results for Senate and presidential elections were not identical in those elections. Tables 1 and 2 compare the presidential and Senate results to one another in the 2016 (Table 1) and 2020 (Table 2) cycles. We included nearly every race from both years (see the note below the tables for the particulars). The key column for the purposes of this discussion is the one on the far right labeled “Difference.” A positive number shows when the Senate margin in a state was more Democratic than the presidential margin, and a negative number shows when the Senate margin was more Republican than the presidential margin.
Table 1: Presidential vs. Senate margins, 2016
Table 2: Presidential vs. Senate margins, 2020
Notes: 2016 California and 2020 Arkansas Senate races excluded because those general elections featured only one of the two major parties. 2016 Alaska is also excluded because Sen. Lisa Murkowski’s (R-AK) main challengers ended up being third party candidates. 2016 and 2020 Louisiana and 2020 Georgia Senate races show first-round voting (as opposed to runoffs, when held) and, when applicable, shows the combined Democratic, Republican, and other votes in all-party primaries. 2020 Alaska considers Al Gross as a Democrat even though he ran as an independent. GA S denotes the special Senate election in Georgia in 2020.
Source: Dave Leip’s Atlas of US Presidential Elections
In 2016, on average across the 32 races included, there was an 8.7-point difference in margin between the presidential and Senate outcome in a given state. However, note that in some of that cycle’s most-watched races, like New Hampshire, Nevada, and Pennsylvania, the topline differences were either small or even nonexistent. Of the 11 races where there was more than a 10-point marginal difference between the presidential and the Senate races, only 1 of them—Sen. Roy Blunt’s (R-MO) reelection against a spirited challenge from then-Missouri Secretary of State Jason Kander (D)—ended up being among the cycle’s closest races.
Four years later, in 2020, the variation in presidential and Senate margins was still present, but on average it was smaller—just 4.4 points, or roughly half that of 2016 (34 races are on this list, including both of the Georgia Senate races from that year). One of the cycle’s biggest overperformances came in the one split result over the past two cycles, Susan Collins’s reelection in Maine, although there is a slight caveat there—Maine has a ranked-choice voting system that would have triggered had Collins not won an outright majority (she got 51%). The margin probably would have been closer had that happened, and voter behavior might have been different had the race been decided traditionally, too (third party voting, at 6.6% of the total electorate, was the third-highest of any Senate race that year). However, that’s splitting hairs—Collins did generate (and would’ve generated in a traditional race) a great deal of crossover voting that year.
We could probably make many more observations about 2016 and 2020 crossover voting, and maybe we will in a future issue. The point, though, is that while there isn’t a ton of crossover voting, there is still some, and that could make the difference in producing one or more split outcomes this year.
2. Look at vote share, not margin, in polls
Let’s look at the polls specifically. Table 3 shows the RealClearPolitics average for the presidential and Senate elections in the five key presidential swing states (FiveThirtyEight, another polling aggregator, does not yet have Senate averages). The comparisons are not perfect—some polls didn’t ask specifically about both races, for instance. But the basic finding of Democrats doing better at the Senate level than the presidential level has shown up often enough that we think it’s a real trend right now.
Table 3: Presidential vs. Senate races in polling averages
Source: RealClearPolitics averages of polls; cited presidential polls are for the two-party Trump vs. Biden results as opposed to the full ballot
Note that in 4 of the 5 states listed, there is a fairly hefty difference in margin between the presidential and Senate races—Biden is behind in all the swing states (although the deficits are small in Pennsylvania and basically nonexistent in Michigan and Wisconsin) while the Democrats are up in all the Senate races, with only the Michigan races showing just a modest difference between presidential and Senate margins. Michigan is an open-seat race where the two leading candidates, Rep. Elissa Slotkin (D) and former Rep. Mike Rogers (R), come from the same basic House district and might not be all that well-known outside of the Lansing area.
This could give us a hint as to what might be going on here: Democratic Senate incumbents may just be better-known and have better brands than the president, while Republican Senate challengers may just be less-known and have less of a brand than Donald Trump. If one focuses just on vote share, the Democrats are closer to Biden, generally, than the Republicans are to Trump. That will probably correct itself, at least to some extent, over the course of the campaign.
Democratic incumbents are on the ballot in Nevada, Pennsylvania, and Wisconsin, and while the two leading candidates in Arizona, Rep. Ruben Gallego (D, AZ-3) and 2022 gubernatorial nominee Kari Lake (R), are not incumbents (nor have they formally won their party nominations yet), Lake did just run for governor in 2022 and has some favorability problems. Gallego’s share of the vote compared to Biden’s is better, but not overwhelmingly so, while Lake is running more clearly behind Trump. Lake should be able to capture plenty of those voters eventually, so the real difference between the presidential and Senate races is likelier to be smaller than what polls show now (although we still believe Gallego could win even if Biden loses the state).
3. Realignment sometimes lags down the ballot
Presidential polling has sometimes shown major weaknesses for Biden among typically core Democratic constituencies, like young voters and voters of color. We can see Biden doing worse among these groups than he did in 2020, although we think some individual poll findings (Trump outright leading among 18-29 year-olds or Latinos, for instance) won’t actually come to pass.
That said, what if there is a gigantic shift among these kinds of voters? It may be that Biden ends up being weaker than a “generic Democrat” and he hemorrhages voters who otherwise vote Democratic down the ballot. This sort of thing wouldn’t be unprecedented—there are all sorts of historical examples of voters moving first at the presidential level and then later at the congressional level. For decades, the South was redder at the presidential level than it was down the ballot, with the last vestiges of down-ballot Democratic strength not being erased in some places until 2010 or even later. West Virginia became Republican-leaning for president in 2000, but Democrats still dominated other races for another decade, easily reelecting incumbent Democratic senators in both 2008 and 2012 while the state was voting comfortably Republican for president. The last statewide Democrat of prominence just formally disappeared last week (when retiring Sen. Joe Manchin changed his party registration from Democratic to independent).
Even in 2016, in states like Pennsylvania and Wisconsin, there were considerable differences across the state in the presidential and Senate races, as Republican Senate candidates outran Trump in suburban Milwaukee and Philadelphia but Democratic Senate candidates did better in outstate areas (see these maps from my aforementioned colleague Miles). This netted out to the Republican Senate incumbents running a bit ahead of Trump statewide, but the similar toplines obscured the amount of differences at the county and sub-county levels. In 2020, that same basic dynamic was present in several key Senate races, though with somewhat less granular variation between the presidential and Senate races.
We know from polling that there are going to be plenty of cross-pressured voters in this election—some of them might express this ambivalence in how they vote. There also is some evidence that voters sometimes will split their tickets against the party that they believe will win the White House (we’ve mentioned that in the Crystal Ball before). Currently, Trump is seen as a favorite both in polling (as Debra Leiter and Michael Lewis-Beck observed in the Crystal Ball earlier this week) and also in betting markets. Maybe that could contribute to ticket-splitting too. Just because the trend has been toward less Senate ticket-splitting doesn’t necessarily mean that trend will continue this year.
4. Third party presidential voters may decide
This election is likelier to be more similar to 2016 than 2020 in that there should be a higher share of third party presidential votes cast. So in at least most places, the major party vote share in the Senate races should be higher than the major party share for president—that was true in nearly every race in 2016 (but not so in 2020, when the overall national third party presidential vote was just about 2% as opposed to 6% in 2016). While true crossover voters (one major party for president, the other for Senate) will appear, there also will be some crossover voters who vote third party for president but major party for Senate. That dynamic could contribute to an overall split result here or there, and the vital swing voters in the Senate races could be disproportionately part of the third party presidential pool.
5. The swing state races aren’t deciding the majority
Notice that we have focused here on the potential for ticket-splitting in the presidential battlegrounds that have Senate races. That’s in part because these are the states that tend to have more recent public polling—beggars can’t be choosers, after all. But let’s remember, crucially, that the battleground Senate races are almost certainly not going to be the ones that actually decide the Senate majority: Democrats could win and Republicans could lose all five of the Senate races included in Table 3, and Republicans could still end up winning the Senate.
Just to reiterate the basic math, Democrats have a 51-49 Senate majority now (that includes the independents who caucus with them). West Virginia is effectively already lost for Democrats with Manchin’s retirement, unless he uses his new independent status to run for reelection (but that seems like more of a consideration for a late run for governor based on recent reporting, and Manchin would be an underdog in the context of any 2024 statewide bid in West Virginia). So that reduces the Democratic margin to 50-50, and they don’t have any clear offensive targets. In addition to holding all of the swing state seats, Democrats also need to defend Sens. Sherrod Brown (D-OH) and Jon Tester (D-MT) in states that are going to vote for Trump by, respectively, 5-10 (or more) and 15-20 (or more) points. We didn’t include these races in Table 3 because there’s little recent nonpartisan polling in either race. Brown and Tester have both waded carefully in the aftermath of Trump’s conviction in a New York trial last week, which makes sense given the potential for the conviction to further nationalize the electorate at a time when Brown and Tester both need a lot of crossover support to win.
Scan Tables 1 and 2: You can find examples of the level of marginal Senate overperformance that Tester and Brown will need, although it often (not always) involved incumbents running in sleepy races—those Montana and Ohio races will be anything but. That even now such crossover is possible is part of the reason we still view both incumbents as being in Toss-up races, but the hill to climb is steep and arguably getting steeper.
Overall, it will be important to continue to monitor the differences between the presidential and the Senate polling. We suspect that actual margins in the key states will be closer than polls currently show, but it’s not unimaginable that we’ll get some split presidential-Senate results this year. And Democrats will almost certainly need to produce at least two such results—in Montana and Ohio—to salvage even a 50-50 split in the Senate.