The Polls Are Swinging to the Democrats

Just when you think you have a fix on what polls are telling us will happen in November, everything changes.

In recent weeks, it has looked like the forecast for Democrats’ prospects for the fall had diminished from a wave to a swell. Just today, Vox published a quite informative piece asking experts what to make of the steady decline in the once-formidable Democratic advantage in the generic congressional ballot (a polling question that simply asks which party a respondent supports in the upcoming U.S. House elections). And indeed, on June 2 that advantage had declined to just over 3 percent in the Real Clear Politics averages, not a number consistent with a big November sweep.

But the RCP average lead for Democrats has quickly sprang back up to 7.4 percent, the highest level in two months. At FiveThirtyEight, which uses a different mix of polls and also adjusts for partisan bias, Democrats have a more modest 6.4 percent lead. But neither is consistent with the idea that the Democrats have lost their midterm mojo and might not ever get it back what with the economy booming and Trump’s approval ratings gradually rising.

This very latest trend could be illusory, too, but the lesson is not to get too comfortable with any set of assumptions. One of the polls driving the Democratic statistical recovery, from NBC/Wall Street Journal, suggests that while voters (or at least non-Democrats) are beginning to give Donald Trump credit for the economy, there is also an undertone of sentiment for a Democratic Congress to make sure he doesn’t do anything crazy:

By a whopping 25-point margin, voters say they’re more likely to back a congressional candidate who promises to serve as a check on President Donald Trump, according to a new national poll from NBC News and the Wall Street Journal.


At the same time, six-in-10 are satisfied with the U.S. economy, and a plurality of voters give Trump credit for the economic improvement.


Despite that economic optimism, however, the poll shows that Democrats enjoy a 10-point advantage on congressional preference, with 50 percent of registered voters wanting a Democratic-controlled Congress, versus 40 percent who want a GOP-controlled one.

A big weekly online poll from Reuters/Ipsos also gives Democrats an 11-point (43/32) advantage. Last week, Reuters/Ipsos had the Democratic lead down to two points (39/37).

Now, you can drive yourself crazy by choosing to believe that every twist and turn in individual polls (some, depending on timing, exerting a big effect on averages at any given moment) reflects the Big Trend that will remain in place until Election Day, or you can ignore short-term trends (as Republicans are likely going to do with this latest lurch in the generic balloting) and point to monthly or multi-monthly trends. A lot of people (including myself) did that for a while until it became clear the regular double-digit Democratic general ballot leads of last 2017 were the outlier rather than the more modest leads that began to emerge earlier this year.

For now, it would probably be smart to avoid too much hyperventilating over either the vanishing Democratic lead of the last couple of weeks or its apparent revival at present. But there are some things to keep in mind as the polls roll in:

1) Not all polls are equal in quality, and some pollsters release numbers so often that they are overrepresented in averages. FiveThirtyEight tries to adjust for this problem by “weighting” polls according to its own assessment of pollster reliability, while also correcting for partisan “house effects.” RCP uses the equally valid approach of leaving the numbers alone and letting biases and quality issues cancel themselves out over time — which does, however, mean short-term trends there can be misleading.

2) Most pollsters are now testing samples of registered, not likely, voters, and things could change down the road when they “switch over.” There are two cross-cutting issues on the subject of likelihood to vote. First of all, the “enthusiasm gap” between Democrats and Republicans that everyone has observed since the moment Donald Trump was elected should give the former a boost at the expense of the latter when pollsters start to test “likely voters.” But on the other hand, the kinds of voters who historically tend to participate in non-presidential elections are also disproportionately likely to vote Republican. Even that new NBC/WSJ poll giving Democrats a ten-point lead in the generic congressional ballot shows that the group least interested in the midterms is voters under 35 years of age, who strongly lean Democratic. LV polls should begin to sort out these conflicting factors, but keep in mind different pollsters use different ways to measure likelihood to vote. Those that assume voters who participated in the last two, heavily pro-GOP midterms are the ones that will show up this November may miss some Democratic strength.

3) At this point, late trends are the only trends that matter. One of the hoariest of political clichés is the one that goes: “The only poll that matters is on election day.” Actually, preelection polling, even months and years in advance of an election, can affect fundraising, candidate recruitment, media coverage, activist engagement, and party/candidate strategy quite a bit. But after a certain point, yes, earlier generic congressional polls stop mattering, and late-cycle changes in the environment are entirely possible. In 2010, the last generally accepted “wave” election, Democrats led RCP’s generic ballot average as late as July 3. They ultimately lost the national House popular vote (which is what the generic ballot estimates) by nearly 7 percent.

No poll should be dismissed as entirely meaningless, but no one in either party should get too carried away in the spin wars over individual polls or short-term trends. Things can change with great speed, even if in hindsight the big picture doesn’t change at all.

The Polls Are Swinging to the Democrats Again