Within the Margin

It’s been a minute! And despite thinking often about the need to do an accountability check on our 2022 work, I haven’t gotten to it…until now. So let’s take a look at each of the three sets of predictions (House, Senate, Governors) and see what was on target and what missed the mark – and importantly, the thought processes driving those calls at the time. That’ll set us up to do some baseline work on 2024, which is shaping up somewhat differently than other presidential years of recent vintage. We’ll start today with the House of Representatives.

Final topline seat prediction: 224 Republican, 211 Democratic (Republican net gain of 11 seats)
Actual result: 222 Republican, 213 Democratic (Republican net gain of 9 seats)

The leading forecasters spent much of the cycle, including in their final predictions, predicting significantly bigger gains for the Republicans. This was in line with typical recent midterm landslides. They did this despite polling throughout the cycle showing closer races. Most of those forecasters, of course, included two- or three-dozen toss-ups, whereas we make a final call on each because, well, you the reader already generally know the races are close; anyone can give you a list of close seats. I think it’s more interesting to actually come down on one side or the other, even if we’re gonna miss on some of them. That’s because getting to that level of commitment forces us to think critically about each race and gives readers more insight into our thought process. Larry Sabato’s Crystal Ball, to their credit, did so: they ended up with a 237-198 edge for Republicans. FiveThirtyEight (in its final election under Nate Silver’s guidance) used “toss-ups” as terminology for races within a certain margin but their model still provided a vote share forecast for each seat, even those characterized as toss-ups. Under their Classic model, their final prediction had the Republicans with 226 seats; the Deluxe model (which incorporates ratings from non-FiveThirtyEight experts in addition to the features of the Classic model) put them at 225.

Decision Desk HQ had fewer undecideds than most; they gave the GOP a 223-200 lead along with a dozen toss-ups. The others (Cook, Inside Elections, Politico, RealClearPolitics, Fox News, Economist) projected the Republicans from 208 (Economist) to 227 (RCP), with anywhere from 20 to 38 of these elusive toss-ups. For what it’s worth, if you think of these as 50-50 chances and simply distribute half to each party, the Economist comes closest to the actual result (221.5 seats for the GOP) and RealClearPolitics is furthest away (246).

On October 21, our House forecast had the breakdown at the eventually-correct 222-213. The two changes from October 21 to the November 2 forecast were a pair of Nevada seats, NV-1 and NV-3, and on November 8 I caveated that moving NV-1 back into the Dem column warranted serious consideration. My recollection is that collaborating analyst Matt Clausen was always more confident about Nevada than me during the ’22 cycle, and the results validated his instinct. But at 224, we can still feel pretty good about how we did and can make a case – depending on how you want to view those toss-ups – for coming in second-best among this group of race-rating entities.

Let’s go under the hood, though – because it also matters which races we nailed and which ones we missed! I’ll start with the misses:

• There’s 16: nine where we predicted a Rep victory but the Dem won; and seven where the opposite occurred.
• None of the seats we identified as Safe for one party were won by the other. One seat we rated as Likely R went to Dems, though, as Marie Gluesenkamp Perez flipped Washington’s 3rd district (Vancouver and southwest Washington). It was Dems’ first win here since 2008, made possible in part by Trump’s fatwa against those Republicans who voted to impeach him – like WA-3 incumbent Lisa Herrera Beutler. She finished third in Washington’s top-two primary behind Joe Kent, a far-right Trump endorsee. Gluesenkamp Perez, a mechanic and small business owner, ran as a very different sort of Dem candidate than we’ve seen much of lately and defeated Kent by a little under one percentage point.
• Three seats rated as Lean to one party went to the other:
  • OH-13 (Akron, Canton): won by Democrat Emilia Sykes by almost five and a half percentage points.
  • NC-13 (southern edge of the Research Triangle region; Goldsboro): won by Democrat Wiley Nickel by a little over three percentage points. To be honest, I think I intended to shift this to no more than Tilt R and simply forgot in the maelstrom of the campaign’s final days.
  • CA-13 (San Joaquin Valley, inc. Merced, Madera and southern Modesto):won by Republican John Duarte by under 600 votes (less than half a percentage point) in one of the country’s closest Congressional races.
• The rest of our misses were seats we had listed as Tilt (our closest rating).
  • Four of these were in NY.
    • As pessimistic as we were once the Hochul campaign’s weaknesses became apparent in the final stretch, we still didn’t quite realize how bloody it would be for Dems. My final House post did note that I was giving strong consideration to moving NY-3 and NY-4 (both primarily in Nassau County on Long Island) into the Republican column. But – noting the weak Republican candidate in NY-3 and the heavy partisan lean of NY-4 – I stuck with the Tilt Dem rating. That Republican candidate, by the way, was one George Santos. That seat has already flipped back to Dems, in a February 2024 special election won comfortably by Santos’ predecessor, Tom Suozzi, following the scandal-plagued Santos’ resignation.
    • In the Hudson Valley’s NY-17, we knew that DCCC chair Sean Patrick Maloney was distracted by national concerns and had never established as firm a grip on his district as one might have thought. His prior margins just weren’t that impressive. And we knew first-term Assemblymember Mike Lawler was a hard-working candidate with a lot of connections throughout GOP politics. But we felt the final indications were that Maloney had stopped the freefall and might still eke out a win in a district Joe Biden carried by ten points in 2020. We were wrong.
    • Further north in the Hudson Valley and out through the Catskills and Southern Tier, NY-19 was incredibly difficult to call politically – and perhaps, for once, made more difficult by geographic proximity. Though not an inch of my home county of Dutchess lies in this district, Dutchess County executive Marc Molinaro was oddly running for this seat. He had spent his (failed) summer special election campaign in his old district lashing out at people tweeting about him, complaining about news coverage, and dodging any attempts to make him say something coherent about the Dobbs ruling that overturned Roe v. Wade. Seeing him squirm so uncomfortably then and in the months that followed may have left me too optimistic. We had it at Tilt Dem, but Molinaro squeaked by with a narrow win.
  • In the other direction, Dems avoided the worst-case scenario in Nevada – not unlike Harry Reid’s legendary 2010 win in the Silver State. This time not only saw Dems hold the Senate seat (now in Catherine Cortez Masto’s hands) against a strong challenge, it featured wins in NV-1 (southeast Las Vegas metro) and NV-3 (western LV metro) which we had labeled as Tilt Republican.
  • The last few:
    • Forecast for Democrats but won by Republicans:
      • CA-27 (northern LA County and the Antelope Valley)
      • VA-2 (Virginia Beach, Chesapeake, Suffolk, southern Delmarva peninsula)
    • Forecast for Republicans but won by Democrats:
      • NM-2 (southwestern NM, plus a small portion of the ABQ metro)
      • OR-6 (Salem, McMinnville and the southwest Portland suburbs)
      • PA-7 (Lehigh Valley and Jim Thorpe/Carbon County)
      • RI-2 (southern and western Rhode Island, downtown Providence, Federal Hill, Olneyville). Matt was always more skeptical that Allan Fung could actually swing this one, and he was right.

You’d rather win ’em all. But on most of these, the other forecasters had them as too close to call – no one’s posting their perfect bracket when it comes to ’22, after all. What are some of the correct calls we should feel particularly good about?

• We were correct on 17 of our 29 “Tilt” calls – those are the seats frequently being called toss-ups by other forecasters. 59% of the supposed coin-flips isn’t too shabby.
• We were 16 for 19 on “Lean” Calls – so 84% in races where we were expressing more confidence. That seems right. Likewise, 32 for 33 (97%) on Likely ratings, and 100% on Safe ratings. Now, on a number of those the margin ultimately warranted a more-or-less confident call in either direction, and that’s something to hone going forward.
• CO-8 (northern Denver metro inc. Westminster, Thornton, Commerce City and Brighton; Greeley). Cook and Politico had this brand-new district as a toss-up; everyone else had it Lean or Likely Republican. We did not: from the beginning, I thought state representative and pediatrician Yadira Caraveo was a strong fit for this district, which has the largest Hispanic population of any Colorado seat. Historically, candidates who have helped deliver people’s babies or keep them alive during their childhood years fare well with candidates. And then there’s the Colorado GOP, who never miss an opportunity to nominate a fool: conservative senator Barbara Kirkmeyer. This anti-abortion, anti-contraception, secession-curious candidate was a marked contrast to the modern, practical Caraveo, whose relative youth could also appeal better to a district full of transplants and young families. In a district most expected to flip to the GOP, Caraveo won by seven-tenths of a percentage point to narrowly vindicate our Tilt Dem rating.
• CT-5 (northwest Connecticut: Danbury, Torrington, parts of Waterbury, and the Litchfield Hills). Two-term incumbent Jahana Hayes faced a GOP rising star (they hoped) in George Logan, a former state senator who also fronts a Jimi Hendrix tribute band. It was a rare battle of two Black candidates in a mostly-white district, and Republican operatives thought they had the momentum as the campaign entered its final weeks. My own drive-by observations found a better-organized Dem operation on the ground, even in Republican-leaning Torrington. We had the race at Tilt Dem for the final call; Sabato had it flipping Republican and most others had it in their toss-up categories. Hayes indeed won, by eight-tenths of a point. She and Logan will face each other in a rematch this year.
• IL-17 (Quad Cities, Rockford, Galesburg, Peoria, Bloomington). This gerrymandered monstrosity was drawn to keep the district in Dem hands despite Cheri Bustos’ retirement. Joe Biden had carried it fairly comfortably in 2020, but Bustos had only won by four points; this terrain has given Democrats problems in tough years. Three of the raters had it flipping to Republicans; two had Dems keeping it and the rest listed it as a toss-up. We kept it in the Tilt Dem column throughout the year in no small part due to meteorologist Eric Sorensen’s local name rec and relentless campaigning. He ultimately won by four percentage points.
• ME-2 (northern and Downeast Maine, inc. Augusta, Bangor and Lewiston). This is the much more rural of Maine’s two congressional districts and shifted sharply to the right after 2012, voting for Trump twice by solid margins. Despite that, Jared Golden was elected in 2018, re-elected in 2020, and struck us as a solid favorite in ’22 against the man he defeated the first time, Bruce Poliquin. We rated it as Lean Dem; six of the nine national forecasters rated it a toss-up. Golden led by four points after the first round of the ranked-choice election, and won comfortably (by slightly over six points) after reallocation. Golden has serious chops: against a strong Republican nominee we’d have hedged more with a Tilt rating, but Poliquin’s weaknesses seemed clear to us. The other forecasters were leaning too heavily on district fundamentals on this one.
• OH-1 (Cincinnati and most of its metro area, inc. Warren County). Historically, Cincinnati was one of America’s largest Republican-voting cities. That reputation was fading by the Obama Era, but decennial Republican gerrymanders splitting the city – and the strong GOP lean of many Cincinnati suburbs – ensured that the various iterations of this seat stayed in GOP hands from 1994 through to 2022, with one blip in 2008. But two factors converged to make the seat vulnerable to a Dem pickup in ’22. First, the Trump Era rapidly shifted many suburbs blue, and Cincy’s eastern suburbs were not immune. Second, the 2022 elections were fought under a map that kept the city unified, giving it a slight Dem edge (D+2 according to Cook’s Partisan Voting Index). Given that Steve Chabot was never hugely popular and had achieved longevity primarily through favorable mapping, we liked this as a potential Dem flip and rated it Tilt Dem. The other forecasters? Four went Lean R; five went Toss-Up. We were correct, as Democrat Greg Landsman defeated Chabot 52.8%-47.2%.
• PA-8 (Scranton, Wilkes-Barre, Hazleton, the Poconos). This might be a cheap one to claim insofar as we were perhaps too optimistic for Dems with our Lean D rating…but two forecasters saw it as Lean R and seven as toss-up, making them collectively a shade too pessimistic relative to Dem chances here. Ultimately, five-term Democratic incumbent Matt Cartwright won re-election by two and a half percentage points, so the “correct” prediction here was probably more like Tilt Dem. Still, that puts us slightly closer to the mark than the collective wisdom.

Alright…I’m happy to have a little closure on the 2022 House results. The next post will take a look at the Senate forecast and results.