Re-evaluating 2016 with our Senate WAR model

Among Democrats, arguably no cycle was greeted with as much hope for the Senate map as the one during the 2016 cycle. At the beginning of it all, strategists across the nation thought the majority was theirs for the taking, and Democrats were salivating at the prospects of unseating incumbent freshmen like Pat Toomey in Pennsylvania and Ron Johnson in Wisconsin, among others. And yet, arguably no cycle ended with as much grief and disappointment as this one for them, with the party picking up only two seats and falling well short of the 5 needed to secure the majority.

In the aftermath, a lot of blame was pointed in many different directions. Some chose to blame the candidates themselves and felt that the they didn’t run the right races with the proper message or strategy. Evan Bayh and Russ Feingold, among others, came in for some of the harshest criticism from burned Democratic leaders and voters. And yet, when examining the data with the benefit of hindsight, there’s an argument to be made that perhaps this failure had less to do with candidates and more to do with the forces of polarization and partisanship than anything, a take that is also borne out by quantitative analysis.

About three weeks ago, we published our 2020 Senate Wins-Above-Replacement (WAR) model, in which we took a look at the Senate map nationwide and tried to figure out how races deviated from a generic D-vs-R matchup, given the national environment and the results observed across the nation. Today, we’ll use this to take a look at the 2016 elections. We’ll control for the same factors as we did in the 2020 model — fundraising, the state presidential partisan lean, demographics, and incumbency. On top of this, we’ll also add in one more feature we’ve developed — a state’s racial polarization, quantified by taking the ratio of the Black percentage to the Democratic vote share (this helps us control for states like Alabama, where the racial dynamics of the state mean Democrats have a very high floor and a very low ceiling, because they continuously get 90% of the Black vote and only 20% of the white vote).

With that done, what are the results we get?

The 2016 Senate Overperformance Map. Louisiana and California are omitted due to their primary systems complicating the general election, while Alaska is omitted due to it mainly being a Republican-vs-Libertarian scenario.

The results can also be found in tabular form below.

The 2016 Senate rankings, in tabular form

There are some standouts in this piece worth a special mention. On the Republican side, North Dakota’s John Hoeven put up possibly the strongest performance of the cycle; part of this undoubtedly has to do with the minimal effort put up by Democrats, but it is certainly not enough to explain a 26 point divergence between the presidential and Senate results! Similarly, Rob Portman and Jonny Isakson blew past their state partisan leans in defeating their challengers. In Isakson’s case, his victory is particularly impressive given Georgia’s higher levels of racial polarization; outrunning Trump by 13% in majority-Black DeKalb county is, suffice to say, nothing short of exceptional. In Portman’s case, he successfully cast Strickland as “weak” and “low energy” and capitalized on a collapsing candidacy to cruise to re-election, forcing Democrats to dump millions into the state before cutting ties and ceding the race. This is the reverse of Rand Paul, who was incredibly unpopular in Kentucky and underperformed his baseline by nearly 10 points, though he still won handily in the end.

On the Democratic side, Chuck Schumer put up an abnormally strong performance for an incumbent Senator, beating the predicted margins by around 9 points, with his brand of retail politics and long-term incumbency playing exceptionally well in New York. War veteran Tammy Duckworth, meanwhile, was the strongest successful Democratic challenger, overperforming the expected baseline by approximately 6 points in dumping Mark Kirk out of the Senate. Popular New Hampshire governor Maggie Hassan, meanwhile, defeated an incumbent senator in Kelly Ayotte by handily outperforming the expectations, which is no mean feat in a closely decided state, let alone one like New Hampshire, which has a famous affinity for ticket splitting. And Jason Kander’s numbers speak for themselves, with the then-Secretary of State for Missouri putting up the strongest performance by any Democratic challenger in any of the cycles or races Split Ticket has examined, coming within 3 points of stunning Roy Blunt in a Trump +19 state.

In terms of the more macro-level takeaways, a few things stand out. Firstly, given the results of the cycle, no Democratic candidate was actually expected to win in a state that Hillary Clinton lost. Whether with Katie McGinty in Pennsylvania, Russ Feingold in Wisconsin, or Evan Bayh in Indiana, none of the Democratic candidates were expected to pull off the win in a state that was carried by the opposition party’s presidential candidate — in fact, all 3 of the aforementioned ones actually overperformed the baseline that a generic Democrat-vs-Republican race would have been expected to produce! The lack of ticket-splitting is a pattern that extended to 2020 as well — the only state in which the presidential and Senate winners diverged in partisanship was Maine, where incumbent Republican Susan Collins put up a performance for the ages in defeating the Democratic Speaker of the House, Sara Gideon.

In fact, looked at another way, the Democrats actually ended up winning a race that they probably should have lost, as a generic Democrat should probably have been defeated handily by a generic Republican candidate in New Hampshire. That incumbent governor Maggie Hassan won against Ayotte was, in large part, due to her own exceptional strength as a candidate and due to Ayotte’s decision to split from Donald Trump and disavow his candidacy, which likely drew enough voters away from her to condemn her to a ~1,000 vote loss.

Between Tammy Duckworth, Jason Kander, and Maggie Hassan, Democrats put up three of the strongest challengers they could find in three critical states. The biggest disappointment for them was actually probably North Carolina, where a Hassan-level candidate could almost certainly have defeated Richard Burr, who was expected to win by only 0.4 points but actually won by 5.8. But while one may argue that candidates of Kander or Duckworth’s quality would have won Wisconsin and Pennsylvania, this is not the type of situation a party should be expected to win a seat under, lest of all defeat an incumbent in, and losing the Senate election should come as little surprise if the presidential candidate loses the state as well. In the modern age of polarization, it is exceptionally difficult to achieve this split result, and in 69 Senate races across 2016 and 2020, it has happened in exactly one.

That is probably not an accident. And that may be a very concerning thing for Democrats in 2024, where they’re defending Senate seats in the states of Wisconsin, Michigan, Pennsylvania, Nevada, Montana, West Virginia, and Ohio — all of which were more Republican than the nation in 2020.

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.

%d bloggers like this: