Though the election was over a month ago, several close races have been forced into overtime. Additional ballots were counted, recounts were triggered in Florida, and Mississippi held the final round of its special election for the Senate at the end of November (not to mention tight House races across the country that weren’t settled until well after election).1And even still we may not be completely out of the woods… the North Carolina 9th District election may not be settled until well into the new Congress, possibly require an entirely new election in the district, or the House of Representatives to decide for itself. But with all of the dust finally settled, let’s unpack what we got wrong, what we can be proud of, and what we learned for the next go around.

For reference, you can see our ratings and predictions here.

What We Got Right

Our Ratings In General and our Projected Final Count: By and large, we actually did a very good job of analyzing the lean of each state going into the election and the likelihood of each party winning a seat. Every race we rated as “solid” or “likely” going to one party went to that party. We probably should have been a bit more generous with the likelihood of incumbents getting unseated (more on this later), but we only had one race (out of 35) as a “lean” that went to the opposite party, and that was Florida, where Rick Scott unseated Bill Nelson by 0.2%. Adjusted by our ratings, we would project 24.7 seats won by Democrats – in reality it was 24, so we actually did really well here (and probably should have leaned on the math more than our assumptions about incumbency to make the final predictions).                         

Calling our Races: Aside from our ratings, we also went out of our way to “predict” a winner, mostly just for fun and to see how we could do2Ratings. Are. Not. Predictions. Once again, adjusted by our ratings, we would project 24.7 seats won by Democrats – the actual result was quite close at 24. When we said there were three “likely” Democratic races and two “lean” Democratic races, that means we’re taking 3 * 80% + 2 * 70%, for a total of 3.8, which does round to the 4 seats that Democrats actually did win out of these five races. Of the 35 races, we miscalled 3 – Florida (for Nelson), Indiana (for Donnelly), and yeah… Tennessee, where our predicted winner, Phil Bredesen, was handily defeated by Marsha Blackburn, was a misstep. That said, we successfully called the Democratic gains in Nevada and Arizona, the Republican gains in North Dakota and Missouri, as well as incumbents holding on in Montana, West Virginia, and Texas (but just barely – like I said, more on this later). Once more though: predictions are just for fun! They are essentially a guess and all things considered next time I would probably prefer to do a model like this again where we had our ratings, and write a separate article with my predictions (because one is objective, the other is subjective). We guessed well, but for the sake of our own credibility and to encourage both objectivity in ratings and thoughtful review in predictions, the two are probably best done separately from one another.

Candidates Matter: Joe Donnelly in Indiana and Claire McCaskill in Missouri both faced incredibly flawed opponents last time they ran for their seats, making their victories akin to Doug Jones’s 2017 win against Roy Moore in Alabama. Without candidates as problematic opposing them on the ballot in 2018, both McCaskill and Donnelly faced tough odds and couldn’t pull it off. Meanwhile, strong candidates like Beto O’Rourke, Kyrsten Sinema, and Kevin Cramer were able to – if not all win – at least pose quite a credible threat to their incumbent opponents. Jon Tester in Montana stuck to his guns and still came out on top, while more placid candidates like Bill Nelson in Florida weren’t able to hold on. New Jersey’s Bob Menendez won despite his history of corruption charges and baggage therein, but by a much weaker margin than his last statewide run, when he won by almost 20 points. To be sure, we hedged our bets here, overestimating the chances of candidates like Donnelly, Bredesen, and both Joe Manchin in West Virginia and Tester (both of whom won, but by smaller than typical margins) because of their ability to politic at a local level and avoid nationalizing their races. But almost all of the Democrats in these red states underperformed expectations, sometimes significantly. That leads us to conclude that partisanship is an increasingly dominant factor in Senate races;3Though also in House races and the presidency. Governors races have been able to shirk the partisanship more than federal races have. with candidate strength generally unable to overpower it to the degree it used to except in particularly egregious examples (like Alabama in 2017 or Missouri in 2012). This should scare Democrats who barely hung on this time; the ability to truly make their races about their state as opposed to the national sentiment is weakening by the year.

Our Arizona and Nevada Assessment: This is the one I’m proudest of, because we spent an outsized amount of time relative to other states analyzing the races in the Southwest and the Arizona write-up was extensive. We called Nevada making a decisive (by 5%) break against the Republicans and embracing a more metropolitan and modern representation. The call still was more conservative than it should have been; we rated it a toss-up in line with our overstated incumbency advantage, but in general our assessment here was spot on. With Arizona, on the other hand, we hit out of the park – thoroughly capturing the movement in the state and calling a lean-Democratic headwind that resulted in a Sinema victory. Sinema capitalized on a more socially-liberal, pro-business, pro-tech, and moderate Southwestern Democratic environment and seized the legacy of both Trump-critical Arizona Senators before her, John McCain and Jeff Flake, to present herself as the best representation for a transitioning Arizona. These races are perfectly reflective of what we’re seeing throughout the formerly-conservative American Southwest.

Being bullish within reason on Beto: We managed to thread the needle in the high-profile Texas race, and were correct in calling it to be as close as it was (Cruz ended up winning by less than 3%). The polling leading up to the election had been giving Cruz a high single digits lead, and as we predicted did not pan out, which is why we gave the state only a “lean Republican” rating. Indeed, Texas is looking good to fall into our general “Southwestern movement” as noted in the Arizona and Nevada point above. Special points to us because we even predicted and mentioned “incumbency potentially being a disadvantage” in this race. If only we’d have applied that logic nationally…

What We Missed

We overstated the incumbency advantage: This is the big thing that caused all of the misses we did have, and we really should have seen it coming considering how close many races with incumbents were. With the trouble that Dean Heller in Nevada, Claire McCaskill in Missouri, and Heidi Heitkamp in North Dakota all appeared to be in (all of whom we correctly predicted would lose), I should have broadened the application of this pattern nationally… it would have raised more eyebrows in Indiana and Florida, which were our two closest miscalled races. Most of the states actually… just kind of went the way they would be expected considering the partisanship of each respective state and the sizable Democratic headwinds. If we’d have just stuck with that and not rated incumbents so favorably, it would have made for better predictions. FiveThirtyEight had a good post-election piece about the value of incumbency in 2018 and how it has diminished, which captures this well – but the national pattern was there and we should have seen it, especially considering we outright mentioned it in our Texas race analysis.

Our Predicted Final Count: Probably our biggest “overall” miss was the predicted final seat count – we had predicted a 50-50 split, and the reality was Democrats lost two seats and fell to a 47-53 disadvantage that saw Republicans  increase their majority (we’re counting independents Bernie Sanders and Angus King among the Democrats, whom they caucus with). The truth is, our final count meant that Democrats would have to win at least every “lean Democrat” as well as every “toss-up”race for that to be possible, and there were a lot more of those than seemed feasible, considering how many more seats Democrats had up for election this round in general. Republicans didn’t need to play much defense anywhere other than Texas and Nevada, while vulnerable Democratic incumbents were scattered. Capturing a 50-50 split assumes Democrats have almost a “perfect” night if you will, and left no room for error. I made this assumption on the belief that Republicans would lose at least one of their “lean” or “likely” states, without really thinking about Democrats facing the same threat (because most of the Democrats were incumbents); and that was wrong… and even if I’d have not made this assumption, I’d still be off by like one or two races. But that’s what predictions are for.

Florida: I want to make special note of Florida just because it’s only race I had “leaning” towards a party (which is still just a smidge above a toss-up) that did not go for that party. Though our analysis of Florida being increasingly Trump-like and bogged down by too many conflating variables to really size up was actually pretty astute, the premise that Rick Scott had won his last two statewide elections in Florida in years where Republicans did very well nationally and that this would not pan out for him in the Democratic-leaning 2018 was inaccurate. The truth is, Florida has changed over the last eight years, and it’s increasingly moving towards embracing a Trump-style Republican Party, as opposed to the Bush-style Republican Party characterized by its former governor, Jeb Bush.

Closing Thoughts

This was an amusing preamble to what I hope will be more and more columns like this – and more exciting election coverage in 2020. We’ve learned some lessons about incumbency, how important good (and bad) candidates are, and the trends some regions are undergoing that will make for very interesting Senate and presidential maps next cycle. Republicans seem like favorites to keep the Senate for a while now – the most competitive races in 2020 are few and far between, with a few potential Democratic pickups possible in Arizona, Maine, and Colorado… and a major Democratic vulnerability in Alabama. But, for 2018, all things considered let’s be clear, Democrats had a very good and impressive night, and overestimating incumbents in an increasingly partisan environment is the major takeaway to watch out for next time.