It’s safe to say that Democrats had high hopes for the 2016 Senate elections. While Republicans had 24 seats to defend this cycle, Democrats only had 10 seats to protect. Many of these seats were in swing states where Democrats figured they could win by utilizing the turnout advantage we witnessed in the past two presidential election years. However, as we saw the night of November 8, this map was not nearly as promising for Democrats as they had previously hoped.
As we have argued in the past, looking only through the prism of wins and losses gives us only a one-dimensional picture of what happened. With this in mind, we developed our PAAR (Percent Above Anticipated Result) model which tries to set the level that a generic Republican and Democratic candidate “should” receive if running against each other. This model, based mainly on past election results, demographic changes, and whether an incumbent is running, has proven to be highly accurate in both Senate and Presidential races. Indeed, the PAAR model typically shows that candidates have a tough time exceeding the PAAR score by more than 3 percentage points, and it also predicts results more accurately one year away from the election than many aggregators and prognosticators on the day before the election.
In September of 2015, we ran the model on the upcoming U.S. Senate races and the end result was a prediction that that the most likely outcome was Democrats ending up with 49 seats this election cycle, falling just short of winning back the majority. While other models (such as FiveThirtyEight, Sam Wang of the Princeton Election Consortium, and the Upshot) that use information up to Election Day, projected Democrats would regain the upper chamber, we all are aware that Democrats ended up falling short of these projections by winning just 48 seats. Regardless, when looking at the PAAR model, Democrats and Republicans ended up about where the model expected them to be a year and a half out.
In this memo, we look back at our Senate PAAR model to see who lived up to our expectations, who exceeded them, and who was left behind.
While PAAR performed better than most in regards to how many seats Democrats would hold beginning in 2017, it is worth looking at the explicit details to find out what went right and wrong. It should be noted before we start that Alaska was excluded from our breakdown because of the unique nature of this year’s election where four independent candidates received a combined total of over 40 percent. In the 33 races we did analyze, Democratic candidates were off from their PAAR score by an absolute average of 3.8 percentage points while Republicans were off by a similar 3.6 point margin. This is slightly better than PAAR’s predictions in 2014, which on average had Democrats off by 4.2 percentage points and Republicans by 3.9 points.
Given Democrats’ underperformance nationwide this cycle, it is unsurprising that Democratic candidates typically had negative PAAR scores. Indeed, 20 of the 33 Democrats received negative PAAR scores, compared to 16 Republicans. What we did find more surprising was that the scores were unevenly distributed; in some states where Democrats were expected to do well (Pennsylvania, Ohio, and Wisconsin), they underperformed, but they also exceeded expectations in solid red states such as Kentucky and Alabama.
By far, the Democrat who exceeded PAAR’s expectations (red state or not) was Ron Crumpton in Alabama – his score was 11.4 percentage points. Despite that stellar performance bump, he still lost to incumbent Senator Richard Shelby by 30 points. However, before we congratulate Mr. Crumpton, we should point out that Shelby spent $12 million compared to Crumpton, who spent just under $30,000. It’s hard to say there are lessons to be learned here other than that Shelby did not spend his money wisely; he wins the award for least successful campaign of 2016…congrats?
Other Democratic campaigns that overperformed included Jim Gray, who received a score of 5.6 points in his race against Senator Rand Paul in Kentucky, despite losing to the former presidential candidate by 13 points on Election Day. In Louisiana, Foster Campbell received a score of 1.5 points, ultimately losing out to John Kennedy (not the Kennedy you’re thinking). And, albeit by a smaller margin, the race in Arkansas also saw a positive Democratic PAAR score (0.6 points).
As far as who performed the “worst” or, said in another way, the Democrat who underachieved by the largest margin was Eliot Glassheim in North Dakota; his PAAR score was -20 points. Many factors contributed to this staggering deficit, including Senator Hoeven’s extremely high favorability ratings, the Trump wave, and Senator Heitkamp raising the bar for Democrats in 2012. The silver lining, if you can call it that, for Democrats is that a win in North Dakota, an extremely red state that carried Trump 63% to 27% against Clinton, wasn’t exactly expected. Additionally, Glassheim only spent $32,000 while Hoeven spent almost $2.8 million. While Hoeven probably did not need to spend nearly that much, he outperformed expectations with a PAAR score of 17.6, a result which gave Hoeven a much better return on investment compared to his Republican counterpart in Alabama.
There were also a handful of Democrats who fell below expectations in states that, while not expected places to win, the party had been eyeing since the beginning of the election cycle. When we take safe states North Dakota and Vermont off the table, the two states in which Democrats fared the worst were Ohio and Iowa. In the Buckeye State, where Senator Rob Portman seemed vulnerable at the beginning of the cycle, Ted Strickland saw a tough loss and ultimately received a score of -10.9 points. And in Iowa, where Senator Grassley faced his toughest reelection bid in recent years, Patty Judge received a PAAR score of -8.1 points.
It is worth pointing out that these are two states that President Obama won in 2008 and 2012, but Trump won in 2016. Clearly this shift did not help the two Democratic candidates, but the underperformance was well above average.
Furthermore, Democrats underperformed and received negative PAAR scores in Oregon, Maryland, and Washington, states considered generally “safe” for Democrats. With that being said, these seats were not in question and the respective Democrats won without breaking a sweat.
Beyond these states, unsurprisingly, Democrats generally did poorly in red states like Oklahoma, South Dakota, Utah, and South Carolina and performed well in blue states such as Connecticut, Hawaii, and New York.
PAAR in Swing States
If you look at the ten states that were estimated to be the closest races heading into Election Day, PAAR did an even better job at forecasting the eventual outcome; it was off by an average of 2.16 percentage points for Democrats and 1.96 points for Republicans. These results are, again, quite similar to some other projections. We had the same call as RealClearPolitics’ final polling average in all ten states and of those ten, there were five states where the PAAR margin was closer to the actual result than FiveThirtyEight’s predictions. That being said, we should remind everyone that our scores were released in 2015 before some candidates were known while the others mentioned here were using polling numbers up to Election Day.
There were also a couple of swing states that were viewed as winnable for Democrats based on the candidates and the perceived weakness of the Republicans having Trump at that top of the ticket. At the end of the day, only one of these Democratic candidates exceeded their expected result.
In Missouri, where many people were continuously impressed by Jason Kander’s campaign, PAAR predicted almost exactly the final spread. PAAR projected the race would end in incumbent Roy Blunt’s favor, 49.3% to 46%; Blunt ended up winning on Election Day 49.4% to 46.2%. Kander exceeded his PAAR score by just 0.2 points and while Blunt exceeded his by only 0.1 points. In the end both campaigns spent $24,536,032 collectively, and that is not including any outside groups or independent expenditures, which is a lot of money for neither side to change the most likely outcome.
In Indiana, Democrats celebrated once they had convinced former Senator Evan Bayh to run for his seat again, but he ended up doing far worse than the polls or pundits forecasted. However, PAAR managed to foresee the realities of the race and predicted that the Republican candidate (in this case, Todd Young) would pull ahead of the Democrat to win 51% to 45.2%. Young ended up winning on November 8th by a slightly larger margin, 52.1% to 42.4%. In the end, Bayh spent $14,061,996 and underperformed by 2.8 points while Young spent $21,326,003 and overperformed by 1.1 points.
Another state that surprised a lot of people on Election Night was Wisconsin. Former Senator Russ Feingold was considered all but a shoo-in to replace current incumbent Ron Johnson but in the end lost 50% to 47%, disappointing Democrats with a -4.5 PAAR score. As we discussed in an earlier piece dissecting state-by-state results, Wisconsin flipped dramatically in 2016 from 2012, with Clinton underperforming Obama by 231,740 votes overall. However, as we will discuss in a future piece, Feingold scored lower than Clinton did in the Badger State.
Beyond these three races, the other swing states were split evenly has far as which Democrats saw positive PAAR scores and which received negative scores. Deborah Ross did particularly well with her score of 1.8 points in North Carolina, which was the highest score among Democrats in all ten swing states. Katie McGinty had the worst night of the ten, receiving the lowest score in the group (-4.6 points).
Campaigns of note
The two highest scoring Republican campaigns were – not all that unexpectedly – in North and South Dakota. Still, Senators John Hoeven (R-ND) and John Thune (R-SD) turned races that were already expected to be blowouts into absolute beatings: Hoeven won by 62 percentage points (79% to 17%) and Thune won by 44 percentage points (72% to 28%). However, the Republican who deserves the most credit for a well-run campaign is, without question, Rob Portman of Ohio. The first-term Senator had originally been on many short-lists of those most likely to lose their seats and as an establishment figure in a decidedly anti-establishment year, he seemed to be swimming against the current of the 2016 cycle. However, he turned what had previously thought to have been a close race (PAAR projected the race as a 50-48 win for Portman) to his advantage and ended up winning by nearly 21 percentage points (58% to 37%).
On the Democratic side, there weren’t very many candidates with overly positive performances. Each of the four Democrats to get a PAAR score greater than 3.0 points – meaning they went significantly above expectation – was also in a race that was decided by 15 percentage points or more. However, five of the ten Democrats in swing races did earn positive scores: Kirkpatrick in Arizona, Murphy in Florida, Kander in Missouri, Hassan in New Hampshire, and Ross in North Carolina. While Kander, Hassan, and Ross were crowd favorites during the election cycle, sadly, only Hassan did enough to win as she flipped the race from a projected Democratic loss to a win.
In November of 2015, we warned that Democrats were far too optimistic about the likelihood of regaining the Senate. Our PAAR model showed them coming up short, and even PAAR ended up overselling Democrats’ likelihood of winning. PAAR was again quite accurate for a tool that doesn’t factor any information about candidates, which further reinforces the idea that partisan preferences are being hardwired into the electorate and candidates have less and less of an opportunity to alter the political gravity in Senate races. It is also showing that most of our Senate campaigns are spending a lot of money and doing very little to change the chances of winning.
Appendix I: States Ranked by DEM PAAR SCORE
Appendix II: States Ranked by GOP PAAR SCORE
For a PDF version of this memo, please click here.