When we created our first model of the presidential electorate before the 2012 election, we had a simple hypothesis: voters were becoming increasingly entrenched in their partisan preferences and that demographics would play an important role in the political success of parties in the short and medium-term. Our model was not intended to predict the future, but instead set a benchmark of what level of support a candidate should receive in order to measure a campaign’s effectiveness. However, nearly four years later, it has proven to be highly adept at projecting expected outcomes and revealing successful campaigns at both the Presidential and Senate levels.
As we have discussed, there is a clear advantage for the Democratic presidential candidate in 2016, while control of the Senate is a toss-up, though there is a slightly higher chance that Republicans will maintain control. As we have pointed out in the past, our PAAR model (Percent Above Anticipated Result) does not factor in specific candidates and other outside influences, such as who the candidates at the top of the ticket are; instead, it sets a benchmark for what a candidate should receive. This then allows us to determine whether a campaign was successful outside of the typical metric of a win or a loss. Using the same principles as our other successful models, we have created a model for the House of Representatives for the 2016 election. Not surprisingly, it confirms what most prognosticators have already been saying: Democrats might gain a handful of seats, but Republicans are in no danger of losing control of the House. What will be interesting to see is which campaigns over perform expectations, and which fall short.
Building the House Model
Despite the success we have had with our other PAAR models, the House has always seemed like the toughest to project for three major reasons: redistricting, volatility in extremely partisan districts, and demographics.
Firstly, all of our models use the performance of the parties in the previous two elections as key factors in the model, and redistricting makes it much more difficult to compare elections in an apples-to-apples manner. The new redistricting maps went into effect in 2012, which means 2016 is the first election when we have two elections to go back to and compare.
From a cosmetic standpoint, projecting results in every race would be unnecessary and very unlikely to be accurate on a percentage point basis at the extreme ends of the spectrum. There are a large number of House seats where a candidate may run unopposed for one election and get 80% in the next election. If our model projected the candidate would receive 90% of the vote, the candidate would over perform one year by 10 points and underperform by 10 points the next, even though they probably barely ran a campaign either year. Clearly this swing does not matter in the context of the election because the seat was never remotely in doubt, but more importantly the results are not an accurate depiction of how a candidate performed. As a result, we will only be comparing the 31 closest races based on the ratings in the Cook Political Report and Larry Sabato’s Crystal Ball.
The third issue is that we rely on demographic changes in our Senate and Presidential model, and it is much harder to get demographic changes down to the congressional district level. We are able to indirectly capture the changing demographics of a district by utilizing a comparison between the current districts and data that is available to us, such as census data as well as voter turnout patterns. Then, the House PAAR score can be calculated using the same ratio as used in our PAAR projections for the state.
2016 House Paar
The House PAAR model shows only a handful of surprises, as most of the expectations are closely in line with other projections. Out of the 31 races selected, only six qualify as true toss-ups with the expected results within 3 percentage points of each other. Republicans are favored in 16 races and Democrats in eight.
The first thing to note is that both Charlie Cook and Larry Sabato list far more races as “competitive” than our model predicts there should be. It doesn’t make a good story to only show a handful of close races, and it certainly doesn’t look great for their projections to show a race as a lock only to have it go the other way, so we understand why the lists tend to be a little larger.
The five races that Cook and Sabato both have listed as at least “Lean Republican” are the five most heavily favored Republican seats according to the PAAR model. In each of the races, the projected PAAR difference between the two candidates is over 14 percentage points. It certainly isn’t impossible for Republicans to lose one of these races, but they should be considered as strongly favored to win rather than the district simply “leaning” in their direction. While we expect the House PAAR to have more volatility, it is worth noting that in 2014 no campaign over or under performed by more than 3 points.
DEMOCRATIC ADVANTAGE SEATS
Although Republican John Katko won New York’s 24th Congressional district by almost 19 points, PAAR views this race as still favoring the Democrats due to Democrat Dan Maffei’s 5-point win in 2012 and President Obama’s strong victories in the district. Clearly the wheels fell off the wagon two years ago, but a Democrat who doesn’t make the same mistakes is still seen as the favorite. The only other “Lean Democratic” race, Rick Nolan in Minnesota’s 8th District, is expected to be just outside a toss-up according to PAAR as well after Nolan’s consecutive narrow victories. However, Nolan will need to at least match the caliber of his past campaigns to hold his seat.
Of all 31 races that Cook and Sabato have as competitive, the two prognosticators and PAAR are only able to agree on three races as pure toss-ups. Leading off with the largest margin among the tossups is Arizona’s 1st congressional district. While Ann Kirkpatrick’s victories the past two cycles have increased Democrats’ odds of success, Obama’s defeats in the district in both 2008 and 2012 keep the race as a tossup.
While Democrats have had more success in Florida’s 26th congressional district during the last presidential election, with Obama winning 53% of the district and Joe Garcia winning with 54%, Republican Carlos Curbelo defeated Joe Garcia in 2014. However, being in a presidential year is not a guaranteed win for Democrats as McCain squeezed out a victory in this district in 2008. Clearly this race is expected to come down to the wire with PAAR giving Democrats a 0.4-point advantage. The last of the three agreed upon toss-ups, Iowa’s first congressional district, has been favorable to Democrats in presidential years; Obama won with significant margins both times and Democrat Bruce Braley had a decisive victory in 2012. However, Iowa demographics are working against Democrats and Rod Blum’s win in 2014 establishes this race as a toss-up.
Interestingly, while you will see that all three models agree on the Republican side, one of the three races that both prognosticators have as “Lean Democratic” and one labeled “Likely Democratic” are expected to be toss-ups in PAAR. Both of the districts are in California: the 7th and 52nd.
Democrat Scott Peters is running for reelection in California’s 52nd after narrowly winning in both 2012 and 2014. PAAR has the Democrat favored by just over 1 percentage point. California’s 7th District is actually projected to be the second closest race in 2016 with Democrats favored by just .35 percentage points. While incumbent Democrat Ami Bera has won the last two elections, he saw his margin in 2014 shrink to less than 1 percentage point.
REPUBLICAN ADVANTAGE SEATS
The majority of the seats that Cook and Sabato have ranked as competitive favor the Republicans according to the PAAR model. The differences between PAAR’s rankings and Cook and/or Sabato’s are mostly due to the fact that President Obama has won in the district, but a Republican representative has also won. For example, both Cook and Sabato rate New York’s 19th district as a toss-up likely because Obama won the district twice and Republican Representative Chris Gibson has also won the district twice (most recently by a 28-point margin). On the congressional level, the Republicans have a clear advantage for this seat. This is also true in Nevada’s 3rd (except for Sabato’s rating), where Obama won in both 2012 and 2008 as has Republican Joe Heck, who greatly increased his margin of victory from 2012 to 2014.
Republicans also have the advantage in districts such as Michigan’s 1st where Republican Dan Benishek narrowly defeated Gary McDowell in 2012. However, given his nearly 7-point victory in 2014 and Romney’s 8-point win over Obama, Republicans are expected to keep this district if they match the caliber of Benishek’s previous campaigns.
Some districts should be easy pickups for Republicans, such as Minnesota’s 2nd congressional district; however, it requires the candidate to match the campaign caliber and favorability of retiring Representative John Kline, which is not necessarily an easy task. While Republicans have some wiggle room, Obama has also narrowly won the district, so it is up to the candidate to live up to the high campaign standard that Representative Kline has set in this district. This is also the case in Pennsylvania’s 8th congressional district where Republicans have nearly a 20-point advantage due to Representative Fitzpatrick’s strong victories in the past. However, with his self-imposed term limit coming up, it is doubtful that another Republican will be able to match the high watermark he has set in the district, particularly since Obama won the district in 2008 and came extremely close in 2012. These two districts are unique cases where a popular incumbent is not seeking reelection and is leaving their hopeful successor with a high standard to match.
While there is a great deal of agreement between the PAAR scores and other projections, there are a few races that PAAR sees quite differently than the conventional wisdom.
In Maine’s 2nd District, both Cook and Sabato rate the race as a toss-up, but PAAR thinks Democrats will have an easy win. The incumbent, Republican Bruce Poliquin, won in 2014 with just 45% of the vote despite Democrats dominating the seat in all of the prior elections. However, even with Democrats’ previous success, it is unlikely his opponent Emily Cain will match the current PAAR score.
The common theme among the outliers is that the state is not very competitive at the statewide level. As a result, when a race that was projected to go 60-40 ends up being 70-30, it can really look like the House candidate was not pulling their weight. As an example, given New York Republicans’ terrible efforts in recent statewide elections, PAAR points to Democrats cruising to victory in New York’s 1st and 24th districts whereas Cook and Sabato say they both are toss-up races. For other outliers, it can be a factor based on an individual, such as John Barrow in Georgia who was probably one of the only Democrats who could win the seat in Georgia. Since we are factoring in one year where he won and another where it was relatively close, the model shows Democrats with a rosier chance than what the reality is. However, given the overall accuracy of the model, we’re okay with a couple of districts where the model is not applicable as they seem to be few and far between.
When the Senate PAAR model was released, we expected a bit more volatility in the results compared to the Presidential level. After all, voters are much less hardened into partisan preferences for state elections than presidential elections. Similarly, the House PAAR should show even more variability. There is much more room for an effective campaign to produce a surprise in a House race with its greatly reduced saliency and news coverage.
Even if Democrats dramatically outperformed their projected PAARs, they would still be well short of winning back the House. Of the above races, PAAR projects only five seats to flip from Democrat to Republican (ME-02, NY-01, IL-10, TX-23, NY-24) which is not impressive considering Democrats have their lowest share of House seats since 1947. What will be interesting to see will be the effect Donald Trump has on House races. If his presence at the top of the ticket moves a point or two away from Republican candidates, it won’t make much difference in the overall outcomes. If the effect is greater, we could see many more Democratic candidates improving over PAAR. In less than 100 days, we will know the answer.
For a PDF version of this memo, please click here.