Skip to main content
Menu
Our Best Tool For Predicting Midterm Elections Works In Presidential Years Too

If you followed polls of the generic congressional ballot, you knew as early as summer 2017 that signs pointed to a wave election for Democrats in 2018. Now, for your prognosticating pleasure, the FiveThirtyEight generic ballot tracker is back to help you track the battle for the U.S. House in 2020. The congressional generic ballot question asks voters which party or which party’s candidate they’d support for Congress (usually as opposed to the name of the specific candidate they plan to support).1 And as of midday Wednesday, Democrats held a 6.2-point lead over Republicans — a solid advantage but still smaller than the 8.7-point lead they held in our polling average on Election Day 2018.2 This jibes with other early signs that suggest that while Democratic enthusiasm has ebbed a bit since last year, the political environment still favors Democrats.

A 2017 analysis by erstwhile FiveThirtyEighter Harry Enten found that the generic congressional ballot is one of the most accurate predictors of who will get the most votes for Congress in a midterm election. I wondered whether the same could be said for generic ballot polls in presidential election years. Borrowing from Harry’s methodology, I conducted a similar analysis for recent presidential election years and found that their generic ballot polling is just as predictive. Specifically, for presidential election cycles starting with 1996, I compared either Gallup’s final generic-ballot poll of the cycle3 or RealClearPolitics’s final polling average4 with the House national popular-vote margin and found that the final polling missed by an average of 2 percentage points.5 The same analysis for midterm election cycles starting with 1994 found that the final polling in those cycles missed by an average of 3 percentage points.6

Final generic-ballot polls are very predictive

Final generic-ballot polling margin and national House popular-vote margin, for elections since 1994

Midterm Cycle Final generic-ballot margin House popular-vote margin Error
1994 R+7 R+7 0
1998 D+4 R+1 5
2002 R+2 R+5 3
2006 D+12 D+8 4
2010 R+9 R+7 3
2014 R+2 R+6 3
2018 D+7 D+9 1
Average 3
Presidential Cycle Final generic-ballot margin House popular-vote margin Error
1996 D+3 EVEN 3
2000 D+1 EVEN 1
2004 EVEN R+3 3
2008 D+9 D+11 2
2012 EVEN D+1 2
2016 D+1 R+1 2
Average 2

Final generic-ballot polling margin is based on Gallup’s final generic-ballot poll of the cycle for 1994-2000 and the final RealClearPolitics polling average for 2002-18.

Sources: Gallup, RealClearPolitics, U.S. House of Representatives

Harry also found that even early generic-ballot polls, more than a year before a midterm election, have some predictive power. My analysis found the same for early generic-ballot polls in presidential cycles like 2020. Basically, although polls can certainly evolve over time, generic ballot polls have historically proved pretty stable.

The table below compares the national House popular vote in presidential cycles since 1996 with an average of pollster averages7 that I calculated using all the polls I could find from between January and June of the year before the election. And as you can see, it reveals a decent relationship between a party’s performance in early generic-ballot polls and its ultimate performance at the ballot box. That is, the national House popular vote in presidential cycles has usually wound up within a few points of those early polls. In most presidential cycles, the two parties were neck and neck in early generic-ballot polls, as they were in the eventual congressional elections (as measured by the popular vote). The lone exception is 2008, which was a Democratic wave year — something the early generic-ballot polls correctly predicted.

Early generic-ballot polls are pretty predictive, too

The generic ballot polling margin in the first half of the year before a presidential election vs. the party’s national House popular-vote margin in presidential elections, for elections since 1996

Cycle Early generic-ballot margin House popular-vote margin Error
1996 R+3 EVEN 3
2000 D+5 EVEN 5
2004 EVEN R+3 2
2008 D+11 D+11 1
2012 R+1 D+1 2
2016 D+1 R+1 2
Average 3

The early generic-ballot polling margin is an average calculated from polls found via the Roper Center, RealClearPolitics and the FiveThirtyEight polling database. We calculated an average for each pollster and then averaged those averages.

Sources: Roper Center for public opinion research, RealClearPolitics, U.S. House of Representatives

In summary, generic congressional ballot polls — even early ones — are good measures of the national mood. And because the national mood affects not only congressional elections, but also presidential ones, that means generic ballot polling might provide a back door for approximating presidential election results too. As you can see in the table below, final generic-ballot polling — and to a lesser extent early generic-ballot polling — has come pretty close to the results of the national popular vote for president as well as for House since 2000. That’s pretty good, considering these polls are measuring elections for an entirely different branch of government, although it’s worth noting that even a polling error of 3 or 5 points can change the outcome if the race is close.

Generic ballot polls are good measures of the national mood

The presidential popular-vote margin, early generic-ballot polling margin and final generic-ballot polling margin, for elections since 1996

Cycle Pres. popular-vote margin Early generic-ballot margin Error Final generic-ballot margin Error
1996 D+9 R+3 12 D+3 6
2000 D+1 D+5 4 D+1 0
2004 R+2 EVEN 2 EVEN 2
2008 D+7 D+11 4 D+9 2
2012 D+4 R+1 5 EVEN 4
2016 D+2 D+1 1 D+1 2
Average 5 3

The early generic-ballot polling margin is an average calculated from polls found via the Roper Center, RealClearPolitics and the FiveThirtyEight polling database. We calculated an average for each pollster and then averaged those averages.

Final generic-ballot margins are based on the final RealClearPolitics polling average for 2004-16 and Gallup’s final generic-ballot poll of the cycle for 1996 and 2000.

Sources: Gallup, Roper Center for public opinion research, RealClearPolitics, U.S. House of Representatives

It’s yet another demonstration of how partisanship is the dominant force in today’s politics; these days, people overwhelmingly vote for the same party up and down the ticket. Congressional vote preferences may not drive the results of the presidential race (it’s more likely the other way around), but the former still reflect the latter. And if early generic-ballot polls are, as they seem to be, mildly predictive, then the fact that Democrats currently lead them by 6.2 points is a bad sign for President Trump. It suggests that the national political environment is still blue, and given how closely the House and presidential popular votes have lined up this century, that’s an important thing to keep in mind for the 2020 presidential election.

Check out all the polls we’ve been collecting ahead of the 2020 elections.

Footnotes

  1. The exact wording of the question varies by pollster: For example, ABC News/The Washington Post asks, “If the election for the U.S. House of Representatives were being held today, would you vote for the Democratic candidate or the Republican candidate in your congressional district?” And NBC News/The Wall Street Journal asks, “What is your preference for the outcome of the upcoming congressional elections — a Congress controlled by Republicans or a Congress controlled by Democrats?” Occasionally, a pollster will attempt to figure out what district a voter is in and provide them with the specific names of the candidates running in the district, but this is uncommon.

  2. Democrats’ final margin of victory in the 2018 national House popular vote was 8.6 points.

  3. For the 2000 election and earlier.

  4. For elections after 2000.

  5. We limited our analysis to the 1996-2016 cycles because generic ballot polls conducted in those years surveyed likely voters. In 1992 and previous presidential years, Gallup mostly surveyed all adults — a population that is very different from, and not very representative of, likely voters.

  6. Harry’s analysis went back much further than 1994 — all the way to the 1946 midterm election — and found that the average error was even smaller: 2 percentage points.

  7. Specifically, I took an average of the margins in each pollster’s surveys of the generic congressional ballot and then averaged those averages. That way, one pollster can’t skew the numbers if it released a disproportionate number of polls.

Nathaniel Rakich is FiveThirtyEight’s elections analyst.

Comments