The Senate Races Are Breaking Bad Already

E ach election cycle since the 2014 midterms, I’ve used a metric for tracking statewide races for the Senate and the governorships that I’ve come to call the Breakers. This isn’t a predictive model, just some simple algebra applied to the RealClearPolitics poll averages over time to track where these races are headed and what the Republicans need to win each one. One advantage of the RCP averages is that we have them going back two decades, so it is feasible to compare their predictive power over time. My original study traced patterns back to 2002.

What I found then were two things. First, which might suggest some hope but also even more dread for Republicans, is that mid-September polling is not the last word. There are often significant shifts between where the races are postured heading into the fall home stretch and where they end up. Second, which bodes very poorly for Republicans now, is the pronounced tendency of these races to track the president’s approval rating (or the presidential race) in wave-election patterns. At this writing, Trump’s approval rating in the RCP average is 18.5 points underwater, close to the worst it has been in his second term: 39.7 percent approval, 58.2 percent disapproval. The reasons are multiple and cumulative — he burned through his honeymoon by April 2025 and was double digits underwater by November 2025 — but far less relevant for these purposes than their effect on statewide Republican candidates. If Trump is below 40 on Election Day, it’s grisly news for the GOP.

This cycle, if you take the polls at face value, Republicans right now stand to absorb a punishing net loss of five Senate seats — enough to flip the chamber — potentially with losses in heretofore deep-red precincts such as Texas, Alaska, and Ohio, the defeat of long-entrenched incumbent Susan Collins in Maine, blowout Democratic wins in the Sun Belt in North Carolina and Georgia, and — frustratingly — the three closest races in which they’re trailing all being potential pickups in Michigan, Minnesota, and New Hampshire. That’s even with Ashley Hinson in Iowa clinging to barely more than a toss-up lead.

Before we dive in, a few notes on the methodology and format. For the polling averages, I look at the RealClearPolitics average for each race with particular attention to how far each candidate is from the magic 50 percent mark and what share of the remaining undecided vote would need to break in the Republican candidate’s direction to get to 50. (Or to 50 percent of the two-party vote, if there are third-party candidates being polled.) I typically flag how robust the data are for each race — how many polls make up each state’s average, and how many of those are relatively recent. I’m not making mathematical predictions or odds. I’m just collecting the data we have in one place, asking how much room there remains to move based on current polling, and drawing some conclusions about what races are most winnable if one party or the other has the wind in its favor.

“Last Poll” and my tracking of “Post 9/1” polls both use the last day a poll was in the field. Also, I’m using the final-outcome polling in Alaska and Maine, both of which have ranked-choice voting systems and, in Alaska, multiple candidates (including a guy named Dan Sullivan running as a “Republican” in Alaska solely to confuse voters and siphon votes away from incumbent Senator Dan Sullivan). Incumbents are in bold, and states are colored by 2024 results. “Break” shows what percentage of the remaining voters would need to break to the Republican candidate to win. As we can see, in seven of the nine main contested races, the number is in excess of 60 percent, which is a heavy lift:

We face two immediate problems in reading public polls in 2026. The first is that polling is quite sparse. Out of 35 Senate races, nearly half — 17 of them — have not been publicly polled at all. Seven races have either been polled only once and/or haven’t been polled at all since the end of August. That leaves us with minimal guidance on potential surprise races in Kansas and Montana, as well as in the more closely watched Alaska Senate race. Two other races, in Florida and hotly contested Ohio, have been polled only once since September 1. That means we have a set of only nine races with fresh polling from more than one pollster, and in five of those, at least half of the RCP average is still composed of polls completed before September 1.

Why does that matter? Patrick Ruffini has argued that summertime polls tend persistently to underrate Republicans. Nate Silver was not entirely convinced by Ruffini’s thesis but agrees that summer polling should be taken with some cautions. So, the lack of a robust field of early- to mid-September polls is a red flag for reading the current poll averages too strongly.

The second problem is that historical trends have been more uneven in the Trump era, which may or may not be attributable to Trump’s effect on the electorate and the specific difficulty of polling the sorts of marginal voters who are attracted to Trump. As I explained in 2022:

In 2016, every single Senate race ended up going the same way as the presidential race in that state — the first time that had ever happened in the era of popular Senate elections — including races where Republicans trailed in the fall. . . .

Then, there was 2018 — a Democratic midterm wave election by any definition. And yet, Republicans did something unprecedented: They unseated four incumbent Democratic senators into the teeth of the Democratic wave. . . .

Then, we had 2020, which, again, is hard to put precisely in a midterm-wave framework given that Donald Trump’s approval rating was mediocre, and he lost narrowly in a very high-turnout election. The polls, again, were off across the board, mainly by underestimating Republicans in Maine, North Carolina, South Carolina, Kentucky, Kansas, Oklahoma, Arizona, Michigan, Iowa, Montana, Alabama, and Texas.

So, on the one hand, we have a historic trend that favors Republicans; on the other hand, we have a recent trend of polling underestimations that has also favored Republicans.

In 2022, the Senate races broke heavily in a Republican direction between mid-September and the final polls, as the traditional midterm wave election dynamics would predict. But Republicans fell short of the wave they expected. There were eight races in 2022 where neither candidate had hit the magic 50. Republicans lost seven of the eight, and all eight saw the final results be more favorable than the final poll average (which was also true in contested races won by Republicans in Ohio, Wisconsin, and North Carolina):

The poll misses in New Hampshire, Arizona, Pennsylvania, Georgia, and Nevada were significant. Outcomes were more normal in 2024, however: Every Senate race was won by the final poll leader except for Dave McCormick’s victory in Pennsylvania, although Mike Rogers in Michigan, Eric Hovde in Wisconsin, Sam Brown in Nevada, and even Kari Lake in Arizona all came closer than the final polling suggested. Trump, who had run behind the party’s Senate slate in 2016, definitely drove Republican turnout in 2024.

While mid-September is the baseline I use for the fall, I also typically compare those polls to a pre-summer baseline, less to see how individual races are moving than to get the pulse of the overall environment. Here, again, we have even fewer polls to work with, but we can see that there have been multiple races trending blue (most sharply in Ohio, Georgia, and Michigan, in the latter case doubtless with some Democrats coming home to Abdul El-Sayed after a divisive primary) and only one race (Iowa) moving the other way:

The great imponderable this time around is whether there’s still a pro-Trump vote out there that hasn’t been activated much so far but would still vote Republican if they could be persuaded to vote. But how? Nothing like the Kavanaugh hearings of 2018 or (on the Democratic side) the overturning of Roe in 2022 is likely to emerge as a party-polarizing factor independent of the public’s approval of the president. If anything, Trump has made it much harder for anybody to get distance from his gravitational pull. And from where we stand today, it looks as if that pull is straight down. Republicans will need some good news from the polls over the next several weeks to give us any tangible reason to believe otherwise.