These Polls Ain’t Loyal: Why the Generic Congressional Ballot Is the Worst Predictor in American Politics

Nothing amuses me more as a political scientist than people who just don’t trust polls. The skepticism is usually brought on by incredulity about some type of polling analysis—“There’s no way black districts voted more Republican in 2016 than in 2012”—which is usually followed by some nonscientific justification for not believing the results.

Well, no polling company ever called me, so I don’t believe it. Have you ever got a call from a polling company? I thought so! 

To my ears, that’s the equivalent of saying, “I don’t believe in tornadoes because my house never ended up in Oz,” or “I don’t believe in the moon landing because I’ve never been there,” or “I don’t believe in Justin Timberlake because I’ve never seen an extra from Duck Dynasty wearing Jordans before.”

The point is, most polls are legitimate barometers of public sentiment, even if you don’t understand how they’re conducted and the NBC News/Wall Street Journal pollsters have never personally called you to find out your opinion of Sarah Huckabee Sanders’ new haircut. With one caveat: The generic congressional-ballot poll is trash and has always been trash, and as a political scientist, TV watcher and concerned American, I encourage you to ignore it—especially now that the Democrats are freaking out.

Over the last week, several stories have run noting that the Democratic lead in the generic ballot has dropped, from 10 to 12 points last November and December to somewhere between 5 and 7 points in early February. The change has been noted by Republicans as a sign that Trump is finally growing on the country again; the GOP-pushed tax plan is becoming more popular; and maybe that Democratic midterm tsunami that was just off the coast late last year may turn out to be just a trickle by the time it hits the polling booths this fall.

Democrats, of course, have been freaking out like they always do, concerned that maybe they took a big hit from the shutdown (not really), or maybe they don’t have a message (they don’t need one), or maybe they aren’t spending enough time wooing “working class” white voters (just STAHP!). The problem with all of this concern is that the generic congressional ballot is just what its name states: It’s the most basic measurement out there, and its predictive powers are highly overestimated.

While several polling agencies—Rasmussen, Gallup and the Washington Post—all have slight differences and their own unique formulas, the basics of a generic-ballot poll are the same. It’s a general poll of all voters about which party they are more likely to vote for in a midterm year, Republican or Democratic. Mix those together and you have a general sentiment about which party is doing well heading into the midterms.

Of course, the generic ballot usually has no differentiation for region or position being voted for because it’s meant to be a general measure. So while Democrats may hold a 5-point lead overall, that doesn’t mean they’re faring well in individual states or individual races; it just means that the majority of the people being polled at that moment say they’ll vote for a Democrat this fall.

You can see how that is inherently flawed, as the nation saw with Hillary Clinton and Donald Trump in 2016. She won the popular vote by over 3 million, but the distribution of votes across the nation wasn’t enough for her to win the Electoral College. In fact, in several electoral cycles, Democrats may actually earn more votes across the country but still not take the House of Representatives or the Senate, or the majority of gubernatorial races, because of where those votes are located.

Consequently, when the Dems’ generic-ballot lead drops from a temporary 10 points in November 2017 to only 7 points in February 2018, that could mean Republicans are gaining ground in red states and Democrats are losing ground in blue states, or it could just be a reversion to the mean. The point is, the generic ballot is only one of several factors that matter in prediction elections, and it may be one of the most flawed, given the generalized nature of the poll.

Focusing too much on the generic ballot is akin to Match.com making love connections primarily based on your zodiac sign and downplaying race, religion and whether your match agrees that Chance the Rapper is better than Kendrick Lamar. There are just other factors to consider.

Predicting a party’s midterm success is a mixture of how many vulnerable seats are in play, the economy, the president’s popularity, retirements and special elections heading into the midterms. Most of those factors favor Democrats. As a colleague told me last week, the only “poll” that matters is Election Day, and since Donald Trump was inaugurated in 2017, Democrats have flipped 35 Republican seats, from governor’s mansions to state legislatures and even one big, red Senate seat in Alabama.

In the first six weeks of 2018, Democrats flipped two more state legislative seats in Wisconsin and Missouri districts where Trump won by double digits 18 months ago. Wisconsin Gov. Scott Walker is literally refusing to allow special elections in the state. The Republican-led Pennsylvania Legislature might impeach a judge over losing a gerrymandering case, and over 30 Republican members of Congress have retired in the last year rather than face the voters this fall.

None of these are the actions of a team that feels a victory coming—quite the opposite, in fact. None of this guarantees a Democratic wave this fall, but it won’t be the generic ballot that tips us off to the results. America should have learned by now that you can’t trust these polls.