| Dear Readers: Today’s piece looking at trends in recent Senate race polling is by Jackson Hamilton, an excellent Crystal Ball summer intern in both 2023 and 2024. Jackson previously helped us with our coverage of last year’s Virginia state legislative elections.
We also want to thank Crystal Ball reader Greg Pasman, whose question to us about measuring Senate incumbents versus the 50% mark in polling prompted us to put together today’s analysis. — The Editors |
KEY POINTS FROM THIS ARTICLE
—Incumbent senators who polled below 50% in the summertime—or even right before the election—are not doomed to defeat.
—Of 70 races we studied from 2010-2022, the incumbent did not exceed 50% in either the mid-summer average or right before the election 48 times. But nearly 60% of those incumbents won anyway.
—Even incumbents stuck in the low 40s are not necessarily doomed.
—Incumbents polling at 47% or higher in the mid-summer average have consistently won over these 7 elections, which is good news for Sens. Bob Casey (D-PA) and Tammy Baldwin (D-WI) in this election cycle—both are currently over that mark in polling averages. Sen. Ted Cruz (R-TX) is also currently polling at that important 47% mark, and Sens. Sherrod Brown (D-OH) and Jacky Rosen (D-NV) are right around it.
Incumbent senators vs. 50%
It’s an argument we see in politics often: “[Insert incumbent] is below 50% in polling, so that person is in trouble.” The suggestion is that incumbents naturally poll higher than challengers because of name ID, so we should measure them not against their opponents, but against a 50% threshold, particularly before the campaigns really pick up in the fall and challengers become better known. Another, simpler form of this argument is basically that undecideds break heavily against the incumbent.
Back in 2010, Nate Silver largely disproved the accuracy of the 50% rule using a sample of polls going back to 1998. This piece looks at Senate races involving incumbents in competitive races since then, from 2010-2022, with the goal of determining what polling numbers can be reasonably said to typically result in the success or failure of an incumbent’s campaign for reelection.
It’s actually not even very common for incumbent senators to be polling at or over 50% in competitive races, whether at the start of August of the election year or even in the final polls right before the election. Those who did reach or surpass that 50% mark had a perfect record in our analysis, but incumbents who aren’t polling over 50% often ended up winning, too.
For our analysis, we’ll go year by year from the 2010 to 2022 cycles and offer some general takeaways along the way. We’ll also look ahead and consider what the polls say in this year’s competitive races featuring incumbent senators.
Race selection
We included 70 Senate elections from 2010 to 2022 in this analysis. In order to maximize the usefulness of the sample, each included election generally had to meet all of the following criteria: have an incumbent running for reelection; be considered to be a competitive election; and have sufficient polling to compute a midsummer polling average (we used poll averages as of Aug. 1, although in a few instances we included races where we couldn’t compute an Aug. 1 average—those are marked “N/A” in the tables) and a final polling average from right before the election.
We generally used RealClearPolitics’s historical polling averages for this exercise—that site has easily accessible archived polling info as well as averages we could use when available (or lists of polls from which we could make our own average). We also supplemented from FiveThirtyEight and Wikipedia’s polling archives, particularly for more recent elections. We made some judgment calls about which races to include or exclude; some of the averages included come from as few as just two polls or even just one, if we thought it was a quality poll, although often the listed averages include more polls than that (in Wisconsin 2022, for instance, the Aug. 1 average just consists of a single poll from the respected Marquette University Law School poll from mid-June).
Email us at [email protected] if you’re curious about any of the calculations. Any time an incumbent was at or over 50% in a polling average, we noted that in bold on the tables. In a few instances, we included races that went to runoffs—most notably the early 2021 Georgia election runoff duo that Democrats narrowly swept to flip control of the Senate. In those instances, the final poll average is the number right before the runoff as opposed to the initial election. The actual incumbent share from the election is from Dave Leip’s Atlas of U.S. Presidential Elections.
With that, we will move to the history, starting in 2010:
Table 1: 2010 Senate election polling in key races
The 2010 midterm elections featured a classic wave, in which Republicans flipped the House and made up major ground in the Senate. This Democratic vulnerability is clearly visible in the fact that 6 of the 7 competitive seats with an incumbent running listed here were held by Democrats. Interestingly, only 2 of the 6 Democratic incumbents in competitive races lost reelection, and those that did, Blanche Lincoln (D-AR) and Russ Feingold (D-WI), had the two lowest final polling averages (Republicans made the bulk of their gains in open seats in 2010). The rest of the Democratic incumbents managed to win reelection despite falling below the 50% mark in the final polling average.
Michael Bennet (D-CO) stands out as somewhat of an outlier, winning reelection despite being at only 43.3% of the vote in the mid-summer average and only having 46.3% in the final polling average. This comeback is more impressive because Bennet was an appointed incumbent running in what was then seen as a swing state (appointed incumbents sometimes lack a true power of incumbency). One of the key reasons behind Bennet’s victory may have been his Republican opponent’s steadfast opposition to abortion, even in cases of rape and incest.
Table 2: 2012 Senate election polling in key races
Going into the 2012 Senate elections, Democrats and the independents caucusing with them held 23 seats up for reelection. On paper, that degree of Democratic overexposure gave Republicans a great opportunity to flip the Senate if they could defeat multiple vulnerable Democratic incumbents, including a pair, Jon Tester (D-MT) and Claire McCaskill (D-MO), who represented states that President Obama was all but certain to lose in the concurrent presidential election. Instead, the strong overperformance of McCaskill, coming back from being at only 43% in the mid-summer polling average to win with almost 55% of the vote, helped Democrats expand their Senate majority. This overperformance can be largely attributed to McCaskill’s opponent, Todd Akin, who drew substantial criticism for responding to a question about exceptions from abortion bans in cases of rape by talking about “legitimate rape.” Akin had not even been nominated by Aug. 1, the time of the average we are using (the primary was Aug. 7, 2012), and he made his infamous comment not long after he won the nomination.
Nevada’s Senate race, won by appointed Republican Dean Heller, stands out for the fact that Heller actually underperformed both his mid-summer and final polling averages. But Heller still managed to win reelection with just under 46% of the vote as his own weakened challenger, Shelley Berkley, ran well behind Barack Obama’s 52.4% share of the vote.
Table 3: 2014 Senate election polling in key races
In 2014, the same class of senators who won during Obama’s landslide election in 2008 faced voters in a much more Republican environment. With Obama’s approval rating in the low 40s for the entire year, Republicans campaigned heavily against his Affordable Care Act, which had a notoriously rough rollout. The general environment may explain how the two Republican incumbents in competitive races both substantially overperformed their polling averages in states that typically vote for Republicans.
It is worth noting that Pat Roberts (R-KS) received an abysmal 42.6% in the final polling average before going on to win reelection by a double-digit margin anyway. The Aug. 1 polling average for this race was excluded from the table because Roberts’s Democratic opponent dropped out of the race on Sept. 3, leaving Greg Orman, an independent, as the leading challenger. Roberts, who worked to nationalize the race by connecting Orman to Democrats, ended up vastly outperforming his polls against Orman. Mitch McConnell (R-KY), a perennial target for Democrats due to his status as the Republican Senate leader, similarly failed to exceed 50% in his mid-summer polling average, but he managed to improve to 49% in polls by Election Day and ultimately won handily in the deep red state of Kentucky.
After the November election, Louisiana’s contest, where Sen. Mary Landrieu (D) led in the first round with 42% against, primarily, two Republicans, went to a runoff. Landrieu managed to substantially overperform her extremely poor final polling average of 37% in her runoff election but still fell well short of winning reelection. In the runoff, Landrieu actually did a good job of mobilizing Black voters, but her numbers with white voters were nowhere near their 2008 levels.
Table 4: 2016 Senate election polling in key races
As a result of the red wave that overwhelmed Democrats in the 2010 Senate elections, all of the competitive 2016 contests listed on Table 4 feature Republican incumbents. With this in mind, some notable performances stand out: John McCain (R-AZ), Richard Burr (R-NC), and Ron Johnson (R-WI) were all at 41% or 42% in the mid-summer polling average, but they each ultimately managed to substantially improve their polling by Election Day and win reelection. Rob Portman (R-OH) went on to a huge victory despite polling at only around 45% in early August. Trump’s surprising general election win likely helped all of these Republicans, although they also all ended up winning by larger margins than Trump in their states.
Table 5: 2018 Senate election polling in key races
Going into the 2018 Senate elections, the Democrats seemed extremely vulnerable despite the fact that it was a midterm with an unpopular Republican in the White House. Out of the 14 competitive Senate races with an incumbent, 12 of them were held by Democrats, including 10 in states that Donald Trump had won in 2016. Even though Republicans ultimately defeated 4 of those Democratic incumbents, Democratic incumbents still managed to win reelection in 6 Trump-won states, including West Virginia and Montana. When considering this strong Democratic performance in “red” states, it is even more impressive that the polling leading up to the 2018 Senate elections was surprisingly predictive of incumbent vote shares, with the final polling averages for many races being within 1-2 percentage points of the final outcome. This accuracy on incumbent performance extended to some of the most competitive races, including Texas, Nevada, and Montana.
Table 6: 2020 Senate election polling in key races
The steady unpopularity of Donald Trump combined with low early polling numbers for numerous Republican incumbents appeared to give Democrats the opportunity to pick up numerous Senate seats in 2020, including some in traditionally red states. Ultimately, though, these Republican incumbents generally improved at least somewhat from their poor polling start, and despite not a single candidate exceeding 50% in the final average, 7 of the 11 Republicans in competitive races won reelection.
Three Republican performances especially stand out: Susan Collins (R-ME) and Thom Tillis (R-NC) both came back from having low polling in the mid-summer polling average, with Collins’s win being especially remarkable given that Biden simultaneously won Maine with 53% of the vote. Steve Daines (R-MT) stands out for the size of his victory, going from 47% in the early average to 55% in the election despite running against Steve Bullock, who was then the incumbent Democratic governor of Montana.
The biggest reason for the overperformance of many Republican incumbents, with the exception of Collins, is likely simply the fact that they represented states that Trump won, and the decreasing prevalence of ticket-splitting allowed them to largely perform strongly enough to win.
Table 7: 2022 Senate election polling in key races
Going into the 2022 Senate elections, President Biden’s low approval ratings combined with the relatively high number of potentially vulnerable Democratic incumbents gave Republicans hope that they could flip the chamber red. However, these incumbents generally managed to overperform both their mid-summer and final polling averages en route to winning reelection. Ron Johnson (R-WI) stands out both as the only vulnerable incumbent to win reelection in a state that voted for the opposite party in the 2020 presidential election and for his remarkably narrow margin of victory over Mandela Barnes (D), especially given the fact that he exceeded 50% in the final polling average.
Overall, 2022 stands out for just how close all of the incumbents were to 50%, in both the mid-summer and final polling averages. Only one incumbent was below 46% in the mid-summer average, and only one was below 47% in the final average. This impressive polling could help explain how every single vulnerable incumbent in the 2022 Senate elections, both Democrat and Republican, won reelection.
Findings
When looking at this diverse set of data, encompassing political environments from a blue wave to a red wave, as a whole, the myth that an incumbent has to be at or above 50% of the vote in polling to win reelection simply does not hold up to reality.
Out of the 70 senate elections considered, there were 48 where the incumbent did not exceed 50% of the vote in either the Aug. 1 polling average or the final RCP polling average. In 28 of these 48 elections, the incumbent actually won reelection anyway. Based on this historical data, it is clear that receiving over 50% in polling is not by any means a necessary condition for an incumbent senator to win reelection. Conversely, because a majority of incumbents falling below that polling threshold even the day before the election ultimately win anyway, an incumbent being below 50% in polls taken months before the election is not even remotely reliable as a warning sign of potential vulnerability.
Beyond just disproving the accuracy of the “50% threshold” for incumbent polling, these data show that there is essentially no reliable threshold which can be used to reliably predict incumbent failure in advance of the election. Several incumbents ended up winning despite being stuck in the low 40s or even worse in the mid-summer average, such as Ron Johnson in 2016 and Susan Collins and Thom Tillis in 2020.
Despite the inefficacy of the 50% threshold in predicting the defeat of incumbents, it could actually serve a useful purpose in determining when an incumbent has such a large lead that they are in strong position to win. In the 22 races above where an incumbent was at 50% either in the early August or final averages, those incumbents always won, although not always easily. This record of recent electoral history suggests that, while falling beneath the 50% mark at any point does not indicate impending doom for incumbents, meeting or surpassing that threshold should be seen as a sign that an incumbent is in a good position to win reelection, barring a catastrophic event.
Table 8: 2024 Senate election polling in key races
Applying the findings from the historical data to the 2024 Senate elections, two races in particular stick out: The Democratic incumbents in Pennsylvania and Wisconsin, Bob Casey and Tammy Baldwin, both had over 47% of the vote in the Aug. 1 RCP polling average. This is notable because, in the competitive senate elections from 2010 to 2022, not a single incumbent who received at least 47% in the Aug. 1 polling average lost reelection While it is certainly still possible that the standing of Casey and Baldwin could weaken between now and Election Day, historical precedent suggests that they are in a strong position to win reelection. Sen. Ted Cruz (R-TX) is also just above that 47% figure in RCP, while Sens. Sherrod Brown (D-OH) and Jacky Rosen (D-NV) are just a hair below. Note that we also included results from a different polling average, Logan Phillips’s RacetotheWH, for comparison (Phillips uses a wider assortment of polls than RCP, and he does some weighting to them. FiveThirtyEight, another major polling aggregator, does not yet have Senate race averages). The averages, unsurprisingly, are similar, although both Brown and Cruz are in better position on RacetotheWH.
The polling averages of Sens. Jon Tester (D-MT) and Rick Scott (R-FL), both currently around 45%-46% in the RCP average, are not currently anything to panic about, since there is plenty of time between now and the election for them to improve their standing. However, if either of these incumbents remain stubbornly at or below 45% as we get closer to election day, history suggests that they could be in danger because 12 of the 17 incumbents who were at or below 45% in the final polling average lost reelection. While we have just been looking at incumbent vote shares in this piece, it also has to be mentioned that Scott is leading his likely opponent, former Rep. Debbie Mucarsel-Powell (D), in recent polls, while Tester has trailed his opponent, Navy SEAL veteran Tim Sheehy (R), in a few recent public polls (all of the other Senate incumbents mentioned in Table 8 are leading their opponent in poll averages).
In addition to the polling data, it is important to remember that in recent election cycles, presidential years have generally not seen as much ticket splitting between the presidential election and Senate races. In both 2020 and 2022, although not necessarily in earlier cycles, incumbents generally saw the greatest improvement from the final polling average to Election Day in states where their party won the 2020 presidential election. Due to this decline in ticket splitting, Scott in Florida can probably be considered to have an easier path to reelection than Brown in Ohio despite Brown currently having a slightly larger share of the vote than Scott in the average, because Trump is likely to carry both Florida and Ohio again.
Our big takeaway here, though, is that measuring incumbents versus the 50% mark in competitive Senate races, whether in mid-summer or even right before the election, presents the incumbents with too high a bar—many incumbents who don’t get to 50% in such polling averages still end up winning.