Much of the press coverage of the 2018 House elections has focused on identifying the most competitive House races and attempting to predict their results. Some academic models have also adopted this tactic, basing their conclusions on an analysis of highly contested districts. Still others run regressions using a combination of presidential approval ratings, generic congressional ballot surveys, and other miscellaneous factors such as disposable income.
We offer a different approach, one based on the three most recent elections that created new House majorities (1994, 2006, and 2010). Boiled down to their essentials, these elections were defined by three variables: (1) How many Democrats, Republicans, and independents were there in the voting-eligible population? (2) What percentage of each group voted Democrat/Republican? (3) Who turned out to vote? Given reasonable estimates of those values, one could have made fairly accurate predictions about the likely outcome in each of those years. We believe that examining the same three variables can provide similar insights in 2018.
The first of these variables is fairly easy to estimate. Every week, pollsters ask Americans about their partisan identity, providing a constant stream of data about the distribution of partisans in the population. Likewise, predicting how partisans will vote post-1990 is reasonably straightforward — about 90 percent will vote their party ID without deviation. However, the final variable, turnout, is hard to pin down. This difficulty partly stems from the high variability of voting rates in off-year elections: Republicans typically vote in larger numbers than Democrats. But not always. Whatever the challenges to measuring turnout, it remains a decisive electoral factor — recent history suggests that Democrats need to at least match their 2006 turnout numbers in order to take the House this year. Thus, turnout is the crucial element of any predictive formula.
Historically, the Democrats’ advantage in party ID has given them a buffer against weaker off-year turnout. From the New Deal through the mid-to-late-1980s, Democrats were the dominant party in terms of identifiers, posting annual advantages of nearly 20 percentage points. Post-1990, though, that margin has been substantially reduced.
Table 1 illustrates this shift. It breaks down party identification in two ways: First, it shows the distribution of responses when people are asked: Do you consider yourself a Democrat, Republican, or independent? Second, it shows the distribution of responses including independent-leaners as partisans (that is, independents who indicate that they typically lean more toward one party than the other). Between 1950 and 1990, Democrats overwhelmed Republicans by either measure, leading by an average of 18 points. Post-1990, however, that has shrunk to an annual advantage of about six points. As a result, Republicans have become more competitive in congressional elections — they have won House majorities in every post-1990 midterm election save 2006. With the Democratic ID edge just a third of its pre-1990 size, Republicans have been able to leverage superior turnout into consistent control of Congress.
Weekly YouGov polls suggest that Democratic ID advantage in 2018 is slightly above its normal post-1990 levels. In September, Democrats held an eight-point advantage without leaners (32.3 percent to 24.2 percent) and 6.7-point advantage with leaners included. In October, Republicans trimmed cut the Democratic lead in the first category to 7.8 points, but slipped to 7.3 points behind in the alternative measure. Accordingly, the October results suggest that Democrats are in a strong position in terms of party ID relative to recent midterm elections.
Contrary to their dwindling lead in party ID, Democrats have gained recently with regard to the loyalty of their partisans. Prior to 1990, the Democrats voted for their own party’s House candidate less than 90 percent of the time. Between 1994-2002, that number dropped further, but Democrats now favor their own party’s candidate at rates of about 90 percent. Meanwhile, Republicans’ loyalty has steadily increased. Prior to 1990, slightly over 80 percent of Republicans would vote for their own party’s House candidate; between 1990 and 2002 Republican loyalty was about 85 percent, and has remained at about 90 percent since then.
Table 3 contains survey estimates of partisan loyalty (and independent vote intentions) post-Labor Day 2018. In mid-September, about 89 percent of Democrats (with or without leaners) said they would vote Democratic, while about 4 to 5 percent indicated that they would vote Republican. Republicans were slightly less loyal, favoring their own party’s candidate at rates of 86 to 87.4 percent. By October, however, partisan sympathies had begun to exert greater influence: Over 91 percent of Democrats and Republican indicated that they would vote their party ID.
What about independents? In the wave elections of 1994, 2006, and 2010, the winning party captured independents, and their votes could once again make the difference this year. In September, that group split almost equally, with Democrats at 38.1 percent and Republicans at 38.2 percent. However, October polls show Democrats leading among independents by about 6.4 percentage points. That shift portends poorly for Republicans. Assessing turnout is thus critical to determining whether the GOP can limit its losses in 2018.
Since the dominant question of this election cycle is whether the Democrats will retake the House, it makes sense to reflect on turnout patterns in past wave elections. Table 4 provides those figures, displaying turnout for strong partisans, weak partisans, and leaners in the last three wave elections. In the Republican Revolution of 1994, GOP voters of every strength voted at higher rates than Democrats, which was enough to overcome the Democrats’ numerical advantage. Likewise, Republicans crushed Democrats turnout-wise in the 2010 wave, running up impressive margins of 15 percent among strong partisans, 16 percent among leaners, and 24 percent among weak partisans.
The sole blue wave in recent years was 2006. In that contest, Democratic turnout approximately equaled that of Republicans — enough to give Democrats a win in the House, given their numerical advantage.
Taken together, the figures suggest that Democrats can narrowly win back the House in 2018 if turnout trends resemble 2006. Our model of turnout, which regresses to historical turnout rates on predictive variables such as presidential approval and economic indicators, suggests that this is a substantial possibility. Right now, we predict turnout among Democrats at about 70 percent among strong identifiers, 58 percent among weak identifiers, and 61 percent among independent-leaners. Among Republicans, those numbers are 68 percent, 58 percent, and 74 percent, respectively.
Combining the turnout estimates with the generic congressional vote and party ID information in Tables 1-3 indicates that Democrats will win approximately 52 percent of the two-party House vote in 2018. Such a number would ordinarily be enough to provide Democrats with 226 seats and the House majority. But Republicans have generally received a “seat bonus” in that chamber as a result of gerrymandering and Democratic saturation in urban areas. As such, a four-percentage-point advantage in the two-party vote might not to push Democrats over the line in the House.
We believe, however, that Democrats will probably outperform this initial prediction. The 52 percent statistic is predicated on so-called “pure” independents remaining at about 15 percent of the population, historically a dubious proposition. As Table 1 indicates, more independents report leaning toward one party or the other as an election draws near, and by Election Day, only about 12 percent of individuals identify as “pure independents.” Taking this trend into account, our prediction for the Democratic share of the two-party vote increases to 52.3 percent.
Moreover, there are reasons to believe that GOP turnout will ultimately fall below the rates we currently predict. For example, of the 77 percent of weak Republicans who approve of Donald Trump’s performance, only 30.3 percent “strongly” approve; the rest only “somewhat approve.” Similarly, among “highly enthusiastic” voters, 30.1 percent of Democrats indicated that a candidate’s partisan affiliation was the most important factor in determining their vote (as compared to only 22.7 percent of highly enthusiastic Republicans). Among highly enthusiastic Democratic women, that number jumps even higher, to 32 percent. Finally, when it comes to fundraising, Democrats have significantly out-raised the GOP ($850 million to $577 million). Democrats have a 3-to-2 cash advantage in competitive seats, which further signifies the hidden weakness of the Republican position.
Collectively, these figures suggest that Republicans will probably not exceed their 2006 turnout rate. If we assume that GOP voters instead vote at the same rate they did 12 years ago – and that Democrats will turn out at their 2006 rate we mentioned above –– our model predicts that the Democrats will win about 54 percent of the two-party vote. A proportional share of seats would give Democrats a 235-200 advantage in the House (a gain of 42 seats). However, given the Republican seat bonus, Democrats will likely only pick up between 25 and 30 seats, giving them a slim majority.