For the longest time, one of the biggest incongruities about this cycle was the divergence between national polling, which showed a close national environment, and the state polling, which showed a Democratic blowout in the battlegrounds. Our initial temptation was to explain this away through candidate quality, given that in Arizona, Georgia, Ohio, and Pennsylvania, Republicans had nominated fairly unpopular candidates whose favorables were consistently underwater. But even accounting for that, some of the polling we got simply did not reconcile with the trends observed over the last few cycles, which showed candidate quality declining in importance.
Perhaps for this reason, it is no surprise that the Senate polling has finally converged with the indicators given by the national generic ballot polling. Pennsylvania, Nevada, and Arizona now all poll within a couple points of each other, and do so at margins more in line with their presidential partisanships. New Hampshire has now also moved into play, as you would expect it to in a tight national environment. Lastly, Ohio, North Carolina, and Wisconsin have now moved more firmly into the Republican column, to a point where upsets are plausible but unlikely for Democrats, and would likely only happen in a world in which they were already winning the “core four” of Georgia, Nevada, Pennsylvania, and Arizona.
Importantly, however, the battlegrounds all rest on a knife’s edge, with virtually any combination of outcomes in the core battlegrounds being plausible in an environment with this amount of uncertainty. While our site forecasts an R+2 environment, our error bands are wide and our range of plausible outcomes ranges from a 51-49 Democratic majority to a 54-46 Republican majority.
Below is our projected forecast for the Senate; our best current guess is that this will come down to a runoff in Georgia, where we do not think it is wise to favor either party until we get more clarity as to what the national and state-level environment actually is on Tuesday. With that said, if we had to pick which party we think will be in the majority come January, it would probably be the Republicans. There are simply many “easier” flips for them than there are for Democrats; the GOP is on far more solid ground in Wisconsin, Ohio, and North Carolina than the Democrats are in Nevada, Arizona, or Georgia.
That leaves Pennsylvania, and while we think Democrats could plausibly end up with 51 seats on a good night by flipping Pennsylvania and holding Nevada and Georgia, this is not something we would bet on. If we had to guess, we would also say that a GOP win is just a shade more likely in Nevada than a Democratic win is in Pennsylvania; that is, Republicans are more likely to end up at 51 seats by the end of the night than Democrats are to end up at 50, and the best chance that Democrats have to keep their majority would lie in holding ground in Nevada and Arizona while pushing Georgia to a runoff, giving them 49 seats with a chance at 50.
For those of you more tabularly inclined, here is a table summarizing the changes from our previous set of ratings. New Hampshire Florida, and Pennsylvania move in the direction of Republicans, while Nevada moves towards Democrats.
We’ve had the Keystone State at Leans Democratic since August, but at the time of the previous ratings change, the environment was materially better for Democrats in both state and national polling — our aggregator had the national generic ballot at D+3.6 at its peak, while the state polling had Fetterman up by as much as 11 in mid-September. Since then, our tracker has shown the generic ballot receding to a tie, and the state polling has done so as well for the Senate race. Given this, and given Pennsylvania’s status as a state that is slightly more Republican than the nation as a whole (Biden won it by 1 in 2020 despite winning nationally by 4.5), an R+2 year would see Republicans favored to carry this open seat if all else were equal.
There are several factors complicating this, however. Firstly, Mehmet Oz (R) and John Fetterman (D) are not ordinary candidates. Fetterman’s popularity generally hovers around breakeven in most surveys, and there are some real questions that many voters have after an underwhelming performance in a widely-watched debate. For example, in a New York Times poll fielded during the week of the debate, majorities polled before the event said he was healthy enough to serve, but a plurality contacted afterwards said he was not fit enough (though he narrowly led Oz even here). And if there are questions about Fetterman, then Oz himself has a tsunami of unpopularity to overcome; the former teledoctor has some of the highest unfavorables we have ever seen among any candidate, hitting 50% in most surveys.
For Democrats, one heartening thing will be that Josh Shapiro looks set to cruise to the governorship against state senator Doug Mastriano, who appears to be toxic to most of Pennsylvania. Shapiro’s coattails may help Fetterman, especially among independent and suburban voters who associate Oz with Mastriano. Polls also show that enthusiasm to vote in Pennsylvania is extremely high among both parties, and this likely reduces any turnout penalty Democrats may have suffered for it being a midterm year with a Democratic president; in other national polls, Republicans generally enjoy a slight enthusiasm edge, but the data we’ve seen in both the mail return rates and the polls do not indicate this to be the case in Pennsylvania.
Both candidates have matched each other blow for blow on the airwaves as well, with external groups from each party spending roughly $100M each on this race, and the main strategy for everyone appears to be negative advertising. Republicans have hit Fetterman repeatedly on the issues of crime and his fitness for public office in the wake of his stroke, while Democrats have hammered Oz on his New Jersey residency, his wealth, and his animal research scandals.
All of this makes the impacts of candidate quality extremely difficult to gauge. To our eyes, this is the result of two flawed candidates squaring off against each other, and the result is truly anyone’s guess. Polls have narrowed as Oz has coalesced the once-hesitant Republican base, and the averages now suggest a functional tie. But there is a big divergence between the nonpartisan pollsters (e.g. NY Times, Emerson, Marist, Suffolk) and the more Republican ones (co/efficient, Insider Advantage, Trafalgar); the general consensus among the former is that Fetterman leads by an average of 3 points, with only Emerson having Oz up, while the partisan pollsters with more questionable data collection techniques and lower transparency standards have Oz clearly ahead.
Part of the issue, however, is that polling in 2016 and 2020 massively overestimated Democrats in Pennsylvania, as pollsters struggled to reach low-social trust voters that are more inclined to vote Republican than not. Polling now suggests that there are some warning signs of a similar effect. Given this, we are wary of taking polls at face value and put a bit more weight on fundamentals. We do not dismiss the possibility of polls being accurate this time around, as they were in 2018; we simply think that it’s worth being cautious and considering what the overall picture conveyed really is. In our view, polls simply suggest a close race within the margin of error, and so it is worth evaluating which side we expect the race to narrowly end up on. The question we keep coming back to, then, is this: Do we think the Fetterman/Oz candidate quality gap is large enough to overcome what the fundamentals suggest?
Our guess is no. In an R+2 year nationally, Pennsylvania, which was 3 points to the right of the nation in 2020, is likely to be a Republican leaning state. We do not think Oz is a great candidate, and in fact, his unfavorables suggest that he is probably among the weakest ones anywhere this cycle. But Fetterman isn’t an incredible one either, and in the absence of a clear spending advantage for either side, we would not bet against the party out of power in a midterm cycle, especially given the state’s expected partisan lean.
This is not to say Fetterman cannot win — the race was kept at Leans Democratic until now for a reason. Shapiro’s coattails and Oz’s unpopularity could easily see Fetterman buck national trends and eke out a victory over Oz, and we don’t expect either party to win this seat by more than a five-digit vote margin. The easiest thing for us to do would have been to leave it as a tossup and let things fall where they may. But we believe in forecasting an outcome in every race, and in this one, we are inclined to go with the party out of power in a state more Republican than the nation overall, and so we rate this seat as LEANS REPUBLICAN.
There is a massive divergence between partisan and nonpartisan polls emerging in this race. The former shows a clear Herschel Walker (R) advantage with the Republicans in striking distance of winning without a runoff, while the latter shows the reverse for Raphael Warnock (D). For example, Republican firms InsiderAdvantage, Echelon Insights, Trafalgar, co/efficient, and Moore all have Walker up by 3 to 4 points and at 48 or 49 percent of the vote. Unfortunately for Walker, they also happen to have him getting 20 to 25 percent of the Black vote, which we think is far-fetched, to say the least — this simply has not happened for Republicans in Georgia since the turn of the century (with the exception of Jonny Isakson, who was a multi-term incumbent going up against a no-name challenger in 2016).
Meanwhile, most high-quality nonpartisan polls of the race (New York Times/Siena, Emerson, SurveyUSA) show the exact opposite of whatever the above polls suggest, implying that it is incumbent Democratic senator Raphael Warnock who has a healthy lead instead. Warnock’s favorables also appear to be relatively fine in most surveys, and as pastor of Atlanta’s famous Ebenezer Baptist Church, his appeal to the Black community should help him turn out a crucial Democratic demographic on Tuesday.
Georgia is not a state in which polling averages generally tend to overestimate the Republicans, and here, we think Walker’s path to the 50 percent threshold needed to win without a runoff is much harder than the conventional wisdom would suggest, especially given the presence of a libertarian on the ballot. In sharp contrast to Warnock, Herschel Walker’s favorables happen to be downright terrible and are often underwater by double digits; if polling and focus groups are to be believed, the ex-football star’s numerous abortion scandals likely did some very real damage to his image. With that said, Brian Kemp’s strength and presence on the ticket will likely provide some coattail boost to Walker and should help him avoid losing outright. We believe this race’s outcome LEANS RUNOFF; neither candidate is likely to win an outright majority here, and so it may be that control for the US Senate is only decided in December, especially if the rest of our picks turn out correct.
In September, we moved this race to Likely Democratic on the basis of public and private polls showing a massive lead for Maggie Hassan (D). Since then, however, the race has quite clearly tightened to where brigadier general Don Bolduc (R) now has a clear shot at dethroning Hassan, the former governor and current incumbent senator. Most nonpartisan polls have Hassan now leading by low single digits, with a lot of variance — the University of New Hampshire has her up by 2, the University of Massachusetts at Lowell has Hassan +10, Emerson Hassan +5, and the University of St. Anselm Bolduc +1. The average suggests that Hassan has a moderately-sized advantage, but polling errors of this size happen all the time, and if the national environment gets worse, Bolduc could pull off a win in this Biden +7 state.
The problem for Bolduc is that he only began advertising late in the game due to New Hampshire’s late primary, and he has made a number of extremely controversial statements regarding the 2020 election and abortion in a pro-choice, Biden-won state. His unfavorables are quite high in most polls, and while Hassan isn’t exactly popular among the electorate, Bolduc makes her look like Chris Sununu. Bolduc’s high unfavorables may serve to be a wall that is too difficult to scale, as undecideds and independents are more likely to break for Hassan. However, we want to stress that this is not a given, especially if the bottom falls out for the Democrats. While we maintain that Hassan remains the clear favorite, the race has evolved into a real one now and thus is downgraded to LEANS DEMOCRATIC. A Bolduc upset, while more unlikely than not, is clearly now within the realm of possibility, at the very least, whereas no serious person was even entertaining the idea of this a month ago.
There are two problems here that square off against each other. On one hand, the environment could easily get bad enough for Democrats to lose several seats, and Arizona, a Biden +0.3 state, is ground zero for a Democratic loss. On the other hand, a Mark Kelly (D) matchup against Blake Masters (R) is exactly the type of election that could see the incumbent Democratic senator hang on by the barest of margins, due to a combination of candidate quality, incumbency, and fundraising.
To begin with, Kelly was possibly the single strongest victorious Democratic recruit in 2020, outrunning Joe Biden by a full 2% despite ticket-splitting plunging to an abysmally low rate. By our measurements, the candidate quality gap between Mark Kelly and Martha McSally was worth a full 4.4 points above replacement in 2020 — i.e. the race was four points more Democratic than it should have been given the other dynamics at play (spending, presidential topline, the nature of an open seat, etc). Now, Kelly comes into his race as an incumbent, and he has spent more money than Blake Masters could ever hope to have, a discrepancy that Split Ticket partner Clare Considine noted was fueled primarily by small-dollar and campaign fundraising rather than independent expenditures. Given that ad rates are cheaper for campaigns than outside groups, such a discrepancy amplifies Kelly’s spending advantage.
Put in simpler terms, Kelly was able to afford more ads on the airwaves for longer periods of time than Masters. This, combined with incumbency (which we measured as being worth roughly two percentage points in two-way vote share), should give Kelly enough juice to hang on even in an R+2 year, especially if even half of his residual strength from 2020 holds against Masters, and so we rate this race as LEANS DEMOCRATIC.
We are not ruling out a Masters win. If the environment deteriorates further for Democrats, it is quite likely that Masters would come out victorious here. But he is also a candidate with abysmal favorables who has not yet led in a single poll, and so we cannot make an argument that he would be the favorite to win the election. Masters clearly has an excellent shot at winning simply because of the environment, which should be around six points less Democratic than in 2020. But we would not bet on him upsetting the popular incumbent senator who also happens to be awash with money (which Split Ticket has found to still move the needle, even if the impact is less than before).
This may end up being the tightest election in the nation, whether by raw vote margin or by percentage. Polls have told us the same story for months now: this is a functional dead heat, and picking the winning party correctly is truly a 50/50 proposition. Nonpartisan outlets like the New York Times, Suffolk, and OH Predictive Insights have found incumbent Catherine Cortez Masto (D) leading former Attorney General Adam Laxalt (R) narrowly within the margin of error, and Emerson is really the only outlier here, having found a Laxalt lead of 5 points.
Democratic early voting turnout has not been reminiscent of the stellar 2020 levels, but it has not been horrible either. All signs point to an exceptionally close race, but it should be noted that the turnout aspect is very difficult to properly assess ex-ante. This is because voting habits have changed significantly in Nevada with the introduction of universal vote-by-mail and the receding of the pandemic, both of which diminish the value of comparisons to 2018 and 2020 and make apples-to-apples analogies impossible. However, Jon Ralston at the Nevada Independent has assembled a set of excellent models that describe the different possible scenarios well, and after the big recent Clark mail drops, he believes Cortez Masto may have a slight edge due to Laxalt’s relative weakness as a candidate. We are not inclined to disagree with him, given his track record, and so we rate this race as LEANS DEMOCRATIC, though we stress that this is the call of ours that we are the least certain of.
The Democratic lead statewide by about 7,000 votes at the moment, with an urban “firewall” of 38,000 votes, assuming a tie with independents. In 2018, this firewall was just shy of 50,000 votes, and it is entirely possible that the 2022 figures swell to this number by Tuesday, especially if mail-in ballots continue their recent late surge with Democratic-leaning voters. If Democrats hit this number, we think Cortez Masto would be the slightest of favorites, and given that Jon Ralston is suggesting that she is favored, we see no clear reason to bet against someone with his track record. At this rate, if mail continues like this, Republicans will likely enter election day with a 15,000-20,000 vote deficit among registered partisans, and given that they won election day by only 16,000 votes with record turnout in 2020, this would make a Laxalt victory more difficult unless he begin to win independents at a higher rate.
Of course, independents in Nevada are actually likely be voting Republican at a higher rate than before, and this meaningfully complicates any analysis we can do. As Ralston notes, virtually no poll or serious analysis has Democrats winning independents this time around; for example, the New York Times has Cortez Masto losing registered non-partisan voters by 4% and self-identified Independents by 8. Thus, in order to win, she will need to limit her losses here and take advantage of Nevada’s larger Democratic base. In other words, Democrats need a great election day in terms of base turnout in Clark County, but given the number of outstanding votes that they still have, there is reason to believe they can get it, especially given that their turnout operations remain strong in this state thanks to the machine built by former Democratic leader Harry Reid.
For the sake of simplicity, we won’t overanalyze the mix of rain and snow about to hit much of Nevada, but we will point out that inclement weather is proven to have some effect on elections by dampening turnout — the question is just whether Democratic areas get hit harder than Republican ones, and whether Republican voters turn out anyways because they are more enthusiastic. We will avoid taking a side on this, but it is something to keep in mind, as the GOP will hope that the rain and snow forecasts in their base areas don’t hurt their turnout much, especially given that they need good turnout from these places to offset Democratic leads in other areas.
We do wish to stress that this election is as close to a coin flip as one can possibly get. In fact, it may come down to the mail ballots in Clark County that must be “cured” (all ballots being cured would currently result in an 800-vote Democratic boost, but many voters simply do not respond to the outreaches made, meaning their vote will not be counted). On the basis of the evidence available, we think Cortez Masto takes this race by the barest of margins, but if you flip a coin, you’ll probably have as good of a chance at predicting the winner here as we do. All else being equal, we simply don’t think it wise to bet against Jon Ralston, and so we rate this race as LEANS DEMOCRATIC.
Part of the problem with election forecasting is that we often find ourselves in the trap of simply predicting the last election. But there is no true historical analogue to this cycle, and it is extremely unusual for the party in power to poll this strongly heading into an election. Economic and approval-based fundamentals indicate that Democrats should have no chance at keeping their majority, but a combination of candidate quality, polarization, and the Supreme Court ruling in Dobbs v. Jackson appears to have given them a fighting chance. We’re of the opinion that Republicans are slightly more likely than not to regain the majority, but we wouldn’t be surprised to be proven wrong.
I’m a software engineer and a computer scientist (UC Berkeley class of 2019 BA, class of 2020 MS) who has an interest in machine learning, politics, and electoral data. I’m a partner at Split Ticket, handle our Senate races, and make many kinds of electoral models.