Header Ads

Breaking News

Primary Season Was Full of Surprises. Here’s Why the Polls Missed Some of Them.

Category: Political News,Politics

Ayanna Pressley’s blowout victory over a 10-term congressman last week was the latest example of a trend that has become abundantly clear: 2018 is the year of the progressive insurgent, from Alexandria Ocasio-Cortez to Andrew Gillum.

But if you relied exclusively on polls, you may not have seen these results coming. A survey conducted in late July, more than a month before the election, showed Ms. Pressley losing by 13 percentage points; she ended up winning by 17. The last poll conducted by Representative Joseph Crowley showed him beating Ms. Ocasio-Cortez by 36 percentage points, but she ended up beating him by 15 in June.

We asked four experts what could have made the polls seem so wrong. Some of it, they said, had to do with capturing the grass-roots energy of 2018. But much of the trouble is more mundane: Primaries are just really hard to predict.

Partisanship doesn’t apply

For all the talk of converting swing voters, a vast majority of Americans who are registered with a political party will vote for that party’s candidate. This provides a sort of backbone for polling in general elections. In primaries, that backbone is missing, making polling much more difficult.

“The things that drive the vote in November are not the same things that drive the vote in the summer, or whenever the primary is being held,” said Charles Franklin, director of the Marquette Law School Poll. “You don’t have the strong structuring of partisanship within a primary that you do in a November election.”

Without the constraints of partisanship, voters’ allegiance is more fluid. They may start out supporting the incumbent but be persuaded to change their minds when they learn more about a challenger. This means that huge shifts in opinion can happen very quickly, in a way you are very unlikely to see in general elections.

Take the poll that showed Ms. Pressley’s opponent, Representative Michael Capuano, leading by 13 points. It was conducted more than a month before the election, and “at the time, that was probably an accurate depiction,” said Patrick Ruffini, a Republican pollster and co-founder of Echelon Insights. “But things change very quickly in primary electorates, where voters are unmoored from party preferences. I’ve seen numbers move 20 points or more in a week.”

[Read more about how The Times is polling the midterms]

The polls were too early

Unlike in presidential races, voters tend not to pay much attention to congressional primaries until the last weeks or even days before they vote, so early polling may be way off base. And if one of the candidates is running a grass-roots campaign, in which personal engagement is decisive, accurate early polling becomes even harder.

“The engagement comes late, the paid communication comes late, so what happens in the last two or three weeks matters a lot,” said Anna Greenberg, a Democratic pollster at Greenberg Quinlan Rosner. “If there’s a heavily grass-roots campaign, the impact of that is going to be felt really close to Election Day and on Election Day.”

In a race like Ms. Pressley’s or Ms. Ocasio-Cortez’s, “the whole point is their candidacies are mobilizing Americans who are not participating in these kinds of elections,” said Mark Blumenthal, head of election polling at SurveyMonkey. “They may not themselves know that they’re likely to vote until a few days, maybe a week before the election.”

The electorate changed

Turnout in nonpresidential primaries is low, often less than 20 percent. So if you want accurate numbers, you can’t just take a random sample of all registered voters — you have to make some educated guesses about who is likely to vote and who isn’t. That creates lots of room for error, and even more so when the district in question has not had a competitive primary in many years, as was the case in Ms. Pressley’s and Ms. Ocasio-Cortez’s districts, and in many others this year.

“We had a lot of primaries this cycle where there’s not been a competitive primary ever in the current district lines,” Dr. Greenberg said. “So knowing who’s going to come out is next to impossible.”

Careful consumers of polls know to look at the margin of error, but that only reflects how much the composition of the sample is likely to differ from the composition of the full population being surveyed. If you don’t even know what the full population looks like, that’s a whole other source of inaccuracy, Mr. Blumenthal said. And this, too, is more of an issue with grass-roots campaigns.

“In general, you would say primaries are made up of older voters who are more intensely partisan, probably have participated in more elections in the past,” Dr. Franklin said. “But the whole point of a successful insurgency is to overturn that by turning out young people or boosting Hispanic or African-American turnout, for example. By definition, you don’t know those things for sure until after the election is over.”

Some voters are hard to reach

Many of the upsets this year have been driven by groups of voters who are very difficult for pollsters to reach, in a primary race or a general election. While the most reliable polls include cellphones in their samples, many people simply don’t answer calls from unknown numbers.

This means young voters and minorities are often underrepresented in polls, even when they are properly represented in the initial samples. For instance, when The New York Times — as part of a live polling project for the general election — polled Kentucky’s Sixth District on Thursday, it had a success rate of 1 in 41 among voters 65 and older. But among voters 18 to 29, the success rate was 1 in 671.

“The progressives are mobilizing voters who in particular may be more difficult for pollsters to get hold of,” Mr. Blumenthal said.

Source link

No comments