Midterms are often represented as a referendum on the incumbent president. There was endless news coverage in the spring and summer about Biden’s low approval ratings, an ominous sign for the midterms. But several things have happened to cast this interpretation into doubt. The Dobbs decision has made abortion and reproductive rights a huge issue for the midterms, and one that strongly favors Democrats. Trump has inserted himself ever more visibly into the campaign, endorsing one awful candidate after another while his legal woes worsen steadily, which potentially reshapes the election as Democrats vs. Trump/MAGA rather than a Biden referendum. Finally, Biden’s ratings have improved significantly over the past 6 weeks or so, even as he remains underwater. Here are the current net job approval ratings (a negative number means unfavorable ratings outnumber favorable) from the two sites I follow for Biden approval polling:
FiveThirtyEight: -9.6. Seven weeks ago, Biden was at -16.4.
RealClearPolitics: -10.1. Six weeks ago, Biden was at -15.6.
Predicting things is often difficult, though not always. This is true in politics as elsewhere. It is easy to predict that a Senate election in Oklahoma will be won by the GOP candidate — you need to know almost nothing other the party affiliation. It is difficult to predict who will win in Georgia — this may be the most swingy of swing states in the U.S. at this time. In a series of diaries related to the state of the election, I’ll pull from a wide variety of sources to give as broad a picture as possible. These will include sites that provide polling averages; sites that use current polling and other data to predict vote outcomes; and political pundit sites.
Polls themselves have become more difficult to do well. Response rates are far below what they once were. The switch to exclusive cell phone use by so many people has made sampling harder. Likely voter models become more difficult in a political environment with both polarizing issues (Dobbs and abortion rights) and polarizing candidates (the hordes of GOPers pledging themselves to Trumpism and the Big Lie), both of which can encourage marginal voters to turn out in ways they normally wouldn’t for a midterm.
Punditry, of course, relies on perceived expertise, which is often inflated not just by the pundits themselves but by the media ecosystems in which they operate. Experts have their own biases, just like anyone else, and without conscious recognition of our own biases and some thoughtful reflection about how they impact our judgments, experts can be just as misguided as anyone else. Sometimes even more so, because the more you know about a topic, the easier it becomes to produce rationalizations for your personal preferences.
So take all of this with a grain of salt. My rule of thumb for polls (in part, to avoid feeding my own optimistic and pro-Democrat bias) is that they generally overestimate Democratic support, especially in the era of Trumpism when many pro-Trump conservative voters are inclined to ignore or even lie to pollsters. The serious (by which I mean accuracy-driven) political forecasters, like the Sabato or Cook reports, do have to be as accurate as possible to maintain their credibility and following, unlike newspaper-, TV-, or internet-based pundits. But even they will have a hard time predicting the actual impact of wild card events like the overturning of Roe — which can produce once-in-a-generation shakeup in voting.
That being said, we can expect accuracy to increase as the election approaches. There are numerous reasons for this (more people are paying attention and undecideds decrease; more polling takes place, which may increase the accuracy of polling averages; trends such as swings to one party or the other become more evident; etc.). With just over 6 weeks until the election, this is a pretty good point to start looking at the big picture.
I will include both Senate and gubernatorial races, as well as general predictions about the House makeup in the next Congress. State races are extraordinarily important due to the role of states in such issues as redistricting, voting rights, certifying electors, and more. Not all governors have strong roles in all these issues, but I frankly don’t have the time to drill down to all relevant offices (such as secretary of state races). The focus will be competitive races, but if you feel some are being overlooked, please recommend them for inclusion.
The sites I’m following:
FiveThirtyEight: Senate ratings and House forecast. Provides estimates of election-day vote based on both a polling average and a “deluxe” model that includes historical trends and expert ratings. (538)
Electoral-vote: Provides a simple last-week polling average for Senate races. (E-V)
RealClearPolitics: Another polling average, but broader than Electoral-Vote’s. (RCP)
270towin: Yet another polling average and algorithm. (270)
Cook Political Report: Expert ratings of each race (solid, likely, lean, or tossup). (Cook)
Sabato's Crystal Ball: Also expert ratings (safe, likely, lean, or tossup). (Sabato)
Inside Elections: Yet more expert ratings (solid, likely, lean, tilt, or tossup). (Inside)
“But Psychbob” you say. “Why give us 3 or 4 polling averages, and multiple pundit ratings too? Why not just one of each?”
A good question! And the answer is: not all polling averages are the same — nor all expert ratings. The polling aggregators differ in the time frame they cover (from 1 week to a month or more, also depending on availability of polling for each race) and also differ in the pollsters they include (for example, some include partisan pollsters while others do not). Similarly, every pundit has their own “secret sauce” of factors that go into their forecast and how they weigh them. You can find details at most of the specific sites I include.
Now, with all that background to prep you, let’s look at some data points. Numbers were current as of Saturday 9/24:
Senate Races
|
E-V |
RCP |
270 (polls) |
538 (polls) |
538 (deluxe) |
Sabato |
Cook |
Inside |
Az |
D 47-40 |
D +6.0 |
D 47.7-41.3 |
D 52.7-45.2
|
D 51.8-46.0 |
lean D |
lean D |
tilt D |
FL |
R 47-44 |
R +2.8 |
R 47.7-44.0 |
R 50.9-47.0 |
R 52.6-45.3 |
likely R |
lean R |
likely R |
GA |
D 47-45 |
D +0.3 |
D 48.0-45.8 |
D 50.6-47.9 |
D 49.8-48.6 |
tossup |
tossup |
tossup |
NV |
R 46-43 |
R +1.7 |
R 44.4-43.4 |
D 48.8-48.1 |
D 48.8-48.0 |
tossup |
tossup |
tossup |
NH |
D 50-41 |
D +8.0 |
D 51.0-40.3 |
D 53.8-44.2 |
D 53.0-44.9 |
lean D |
lean D |
tilt D |
NC |
R 47-46 |
R +2.0 |
R 46.2-44.8 |
D 49.4-48.5 |
R 49.9-47.9 |
lean R |
lean R |
lean R |
OH |
R 46-45 |
R +2.2 |
D 45.0-44.8 |
D 50.0-50.0 |
R 51.9-48.1 |
lean R |
lean R |
likely R |
PA |
D 50-43 |
D +4.2 |
D 51.0-42.6 |
D 53.2-44.6 |
D 52.0-45.8 |
lean D |
lean D |
tossup |
WI |
R 48-47 |
R +1.5 |
R 48.4-47.0 |
D 50.2-49.8 |
R 50.8-49.2 |
lean R |
tossup |
tilt R |
Senate overview: I’m sure everyone has heard the doom and gloom “Democrats will be crushed” angle in the mainstream media that was very common over the summer. I believe we hear less of it these days, as the political indicators have been shifting in favor of the Democrats for weeks. In the case of the Senate, the Democrats are a clear favorite to retain control. Seats such as Arizona and New Hampshire that had been described as “tossups” by many sources (and still are by some, such as RCP) seem a lot safer, as polling and pundits converge to agree that Democratic incumbents are favored. Kelly in Arizona leads by 6-7 points in every average and forecast; Hassan in new Hampshire, by even more. The two most at-risk seats are Georgia and Nevada, where the polls and pundits agree in describing the races as toss-ups. However, the Democrats are clearly favored to win the GOP-held seat in Pennsylvania, where Fetterman had masterfully trolled Trump-endorsed MAGA charlatan ‘Dr.’ Oz on his way to 7-8 point leads in the polling averages. Furthermore, GOP-held seats in NC, OH, and WI are all clearly in jeopardy. Add it all up, and Democratic control seems likely, with the punditry predicting anywhere from 50-52 seats. For example, 538 estimates the likelihood of Democratic control at 70%, with 51 seats the most likely outcome.
Next week: We add predictions for the party split in the House and for gubernatorial races!