OUR FUNDAMENTAL PHILOSOPHY

Since its founding in 1984, the philosophy at the Cook Political Report has been that forecasting congressional elections is both an art and a science—that a combination of both qualitative and quantitative analysis works best in ascertaining what is likely to happen nationally and in individual states and districts.

Our approach starts by carefully examining what we can readily quantify: population demographics, historic and recent voting patterns in states and districts, and the partisan performance of states and districts relative to national outcomes in recent presidential elections (a measurement we introduced in 1997 as the Cook Partisan Voter Index or PVI). From these, we make an extremely preliminary judgment of where competitive races are likely to take place.

Then, we weave in the less quantifiable. Looking at a field of candidates, is there an incumbent? If so, how strong has that incumbent run in past years and what polling is available, such as job approval ratings, to suggest the incumbent is unusually strong or weak? Any scandals or recent political problems are then considered. For challengers and other nonincumbent candidates, we look at their backgrounds, both political and professional. In many cases, we conduct hour-long interviews, usually in our offices, with the candidates as they make the rounds in Washington. At one point the New York Times Magazine referred to candidate stops at the Cook Political Report and the Rothenberg Political Report as “one of the stations of the cross” that candidates must pass.

Next, we watch the campaigns as they develop, following press coverage and interviewing campaign consultants both working for the candidates directly, or indirectly through party committees, Super PACs, or independent expenditure efforts. We talk constantly with pollsters for both parties and various groups involved in the campaigns, and often they share their data with us, although we cannot directly use or attribute their intelligence. All of these methods help give us the best possible perspective of who is likely to prevail.

Our seven-point ratings are the primary manner in which we categorize each race, with Solid Democrat and Solid Republican reserved for those races that, at that point, seem virtually certain to go in that particular direction. Likely Democrat and Likely Republican are for races that have a high probability of going in that particular direction, but there are one or more reasons why they are kept on a watch list, to see if those races become competitive. Our competitive race categories are Lean Democrat, Toss Up, and Lean Republican. Toss Ups are races where one candidate may or may not have a small lead, usually in the low or middle single digits, that is not sufficiently large to put into a category that plainly leans one way or the other.

THE TRICKY PART

Even if it was possible to quantify the strengths and weaknesses of incumbents and the quality of challengers as objectively as we can quantify the partisan leanings of districts, we would not claim to possess a surefire “secret formula” for predicting outcomes. The most delicate aspect of our approach may be the subjective task of deciding how to weigh each of these factors against each other. This consideration is carefully and constantly recalibrated during each election cycle.

For example, in past “non-wave” election cycles—that is to say, cycles when parties’ national strength and popularity have been near parity—most competitive races have been in states and districts with PVI scores of between D+5 and R+5, and incumbents’ and challengers’ individual strengths, and sometimes local issues as well, have mattered a great deal in determining outcomes. But in “wave” election cycles when one party has suffered the brunt of voters’ displeasure, the playing field has shifted, and most competitive races have taken place in districts between R+2 and D+8 (if Democrats are suffering) or D+2 and R+8 (if Republicans are suffering). In addition, incumbent and challenger quality has mattered much less in wave situations; even very strong incumbents can lose and very weak challengers can win.

What is most remarkable is that we have not witnessed a truly “localized” or non-wave” election since 2004. Democratic wave elections in 2006 and 2008 were followed by a Republican wave election in 2010. In 2012, voters gave both parties low marks, and amid a fierce, nationally charged presidential race, relatively few states (6 of 33) and districts (26 of 435)^{Footnote 1} split their tickets between the leading presidential candidate and the winning Senate or House candidate. This begs this question: will we return to “localized” elections soon, or will we and others need to recalibrate our approach to adapt to a new “nationalized” normal?

LEARNING TO ADAPT

Countless evolutions in both voting patterns and campaigns, some fleeting and some more durable, have altered the way we evaluate races since the Cook Political Report’s1984 inception. In particular, three trends have either contributed toward or occurred in parallel with the “nationalized” nature of recent elections we have observed: unprecedented partisan geographic sorting of the electorate, a transition away from split-ticket voting to more straight-ticket voting, and a marked rise in the proportion of noncandidate-driven spending in congressional elections. These trends merit more discussion in terms of their implications for both our forecasting methods and those of the larger congressional forecasting community.

GEOGRAPHIC CLUSTERING

First, and somewhat paradoxically, although the proportion of self-identified independents in polls is higher than ever^{Footnote 2}, partisans are increasingly clustering together in like-minded states, districts, counties, and even neighborhoods. Several factors contributed to this sorting pattern beyond simply party-conscious residential settlers: voters who have shifted partisan allegiances (for instance, white southerners switching from Democrat to Republican), parties’ intense focus on registering and mobilizing their “best areas” to get to the polls, and the life cycle (for instance, old New England Republican voters exiting the electorate). There is no denying this sorting: it has been highly visible in presidential elections (see figure 1) and has substantially reduced the number of “swing” states and congressional districts (see figures 2 and 3).

Figure 1

Percentage of USVoters Living in “Landslide Counties”

(Color online.)

Figure 2

Senate Seats by Partisan Voter Index Score, 1994–2014

Credit: Loren Fulton, Cook Political Report.

Figure 3

House Seats by Partisan Voter Index Score, 1994–2014

Credit: Loren Fulton, Cook Political Report.

In 1992, 38.6% of all voters lived in counties that gave either major party presidential nominee at least 60% of the vote. In 2012, 50.1% of all voters did:^{Footnote 3}

Between the 1994 and 2014 cycles, the number of Senate seats in “swing” states—those between D+5 and R+5 on the PVI—has fallen from 58 to 28:

And, between the 1994 and 2014 cycles, the number of “swing” House seats—those between D+5 and R+5 on the PVI—has fallen from 164 to 90:

The evaporation of “swing” seats thanks to this geographic sorting seems to argue for diverting some of our report’s attention from November general elections to primaries, which are tantamount to election in an increasing number of states and districts. However, no discernible pattern exists of competition having “moved” from general elections to primaries between 1992 and 2012 (see figure 4), perhaps a sign that more incumbents are voting strategically in Congress to preempt a serious primary challenge.

Figure 4

House Members Winning 55% or Less in Primaries and Generals, 1992–2012

Credit: Loren Fulton, Cook Political Report.

Instead, this geographic sorting should prompt us and others to pay more attention now to patterns of demographic settlement and their long-term political implications. For example, a California seat with fast-growing Latino and Asian shares currently may be fairly Republican at R+6 on the PVI, but if it had a score of R+8 in 2008 and R+10 in 2004, it may be a competitive seat in 2016 depending on whether the same migration patterns continue. This aspect of forecasting is not new, but it is more relevant than ever.

DECLINE IN SPLIT-TICKET VOTING

Concurrent with a decline in “swing” states and districts, the share of voters willing to cast a ballot for a presidential or congressional candidate opposite their own party has dropped to levels not seen for over a generation. According to data from American National Election Study, in 1992, 68% of partisans voted for their party’s candidate for president, 73% for their party’s candidate for Senate, and 70% for their party’s candidate for House. In 2008, 85% of partisans voted for their party’s presidential nominee, and 81% voted for their party’s Senate and House nominees.^{Footnote 4}

In the Nixon landslide of 1972, 44.1% of all House districts split their ticket between the presidential race and their congressional race. In the competitive 1992 presidential election, the share declined to 23%.^{Footnote 5} In 2012, just 6% (26 of 435) of all House districts were carried by presidential candidates and House candidates of opposite parties, the lowest share since 1920, when just 3.2% of districts featured split outcomes.

Clearly, voters increasingly see the two parties in black and white; fewer voters take the time or have the ability to distinguish between a party’s candidates at the presidential and congressional levels. Why? To be sure, parties have become much more ideologically homogenous since 1992. But one theory worthy of further investigation may be that a decline in voter consumption of local news and a rise in consumption of online, nationalized, and partisan-oriented media has diminished voters’ opportunities to consider local candidates’ backgrounds and has fostered more straight-ticket voting.

Regardless of causes forecasters must pay attention to whether straight-ticket voting continues to increase, remains steady, or subsides during the next few cycles to properly weigh the importance of quantitative factors (partisanship of a state or district) against qualitative ones (such as candidate quality). In addition, particularly in midterm years, handicappers may need to focus just as much on candidates’ infrastructure for motivating their own base to turn out as their ability to persuade voters outside their own party.

RISE IN NONCANDIDATE-DRIVEN SPENDING

In the “old days” of congressional campaigns (really, as recently as 2006), a competitive House general election race might feature (depending on the cost of the advertising market) each candidate spending between $1 and $3 million dollars, with affiliated party committees (the Democratic Congressional Campaign Committee and the National Republican Congressional Committee) spending a comparable or perhaps slightly larger amount to provide backup, mostly in the form of negative ads, so as to allow the intended beneficiary to “stay above the mud.”

Today, thanks in part to party committees’ innovation and backroom maneuvering in the aftermath of the Supreme Court’s 2010 ruling in Citizens United v. FEC, that calculus has changed. Instead of four active advertisers in a given district, often six or more are included: two candidates, their affiliated party campaign committees, and their parties’ affiliated Super PACs. This has contributed to a race-by-race environment in which candidates and their campaigns are no longer the leading men or ladies alone in their own lonely spotlight; they are more like central players with large supporting casts they cannot control.

Between 1992 and 2012, candidate spending in Senate and House elections tripled from $523 million to $1.5 billion. Spending on the part of party campaign apparatuses (such as the DCCC, NRCC, and their Senate counterparts) rose fivefold from $46.1 million to $235 million. But the real force revolutionizing district-by-district combat has been “Non-Party Independent Expenditures,” (including Super PACs), which spent only $6.3 million in 1992 but $457 million in 2012, a 72-fold increase.^{Footnote 6}

The $457 million spent by nonparty groups may not sound like a lot compared to the $1.5 billion spent by candidates, but the bulk of Super PAC dollars flowed into a handful of the most competitive Senate and House races, in some cases dwarfing candidate spending. In one extreme 2012 example, New York City Mayor Michael Bloomberg’s Independence USA PAC spent $3.3 million in California’s 35th District to (successfully) pulverize the incumbent opponent of Democratic candidate Gloria Negrete McLeod, who spent only $281,000.

The dizzying ascent of outside groups in House and Senate races has left forecasters like us no choice but to interact with many more interested parties than simply candidates, their pollsters, and their supporting party committees. Whereas in past cycles, an off-the-record meeting with a candidate and his or her consultant might have offered us a telling glimpse into the kind of themes voters in his or her district would hear about in an ensuing campaign, it is now essential to survey the strategies of other operators who are barred by law from coordinating with candidates but could strongly influence a race’s outcome.

LOOKING AHEAD

It is easy to look at overarching trends in the electorate and congressional campaigns and conclude that candidates control their own destinies less than ever before. In an era when even deeply flawed Republican former Rep. Mark Sanford can win his old House seat back fairly easily, some would argue that qualitative factors—the strengths and weaknesses of candidates, the suitability of their biographies for their states and districts, and the potency of their campaigns—take a backseat to quantitative factors such as state and district partisanship.

Here is another way of looking at the same situation: if indeed it has become more difficult for candidates to overcome obstacles such as self-sorting, straight-ticket voting, and outside spending, which candidates are uniquely strong enough to “defy the odds?” After all, there were still plenty of Senate Democrats who won “Romney states” in 2012 (Senators Jon Tester (MT), Heidi Heitkamp (ND), and Joe Donnelly (IN) to name a few) and plenty of House Republicans who won “Obama districts” (Representatives David Valadao (CA-21), Mike Coffman (CO-06), and Frank LoBiondo (NJ-02) for example), for nuanced reasons that simply cannot be quantified.

These “against the grain” races, although increasingly rare, usually make the marginal difference between one party sitting in the majority or minority of the House or Senate. And they continue to demand analysts who can put a variety of local and personal factors under a microscope. Our challenge will be to continue to “recalibrate” and refine our approach as each campaign cycle adds new electoral and technological twists, while adhering to a basic model that has served our readers well for nearly 30 years. To draw on a baseball analogy from the terrific book and movie “Moneyball,” every Major League Baseball team employs some form of Sabremetrics, but there are good reasons none have fired all of their old-fashioned scouts.