Thursday, December 31, 2015

The More We Poll, the Less We Know

ELIZABETH WILLIAMSON, NYT, DEC. 30, 2015

Polling of Republicans in the early-voting states of Iowa, New Hampshire and South Carolina has reached unprecedented levels, fueled by the number of candidates in the hunt and an obsession with the horse race rather than a meaningful debate over policy, a new Boston Globe study says.

Since the race began, more than 200 surveys have been conducted on the preferences of the thin slice of electorate who, this early, a) bother to respond; and b) can remember who’s running.

So what can they tell us? Not much.

Pollsters agree that at this point in the race, no poll nor group of polls can predict who will be the next president, or even who will be the party’s nominee. “Early surveys, even a month out, are not going to give people the prediction they’re looking for — particularly in primaries,” said Michael Dimock, president of the Pew Research Center and a political scientist who has conducted many polls in his time.

Who gets the most excited by polls? The candidates who lead them. It’s not a Donald Trump rally without the candidate trundling to the podium, crowing about the numbers showing him winning. But those polls do not predict a Trump victory. Here’s why:

The level of support Mr. Trump boasts about (currently, between 30 and 40 percent) is drawn from within the 25 percent of voters who identify themselves as Republicans, according to Nate Silver of FiveThirtyEight, meaning that to secure the nomination he would have to broaden his appeal among Republicans beyond the slice he has — not easy, given the extremism he represents. When Mr. Trump was polling at about 30 percent, Mr. Silver estimated he was pulling between 6 and 8 percent of the American electorate, which he said was about “the same share of people who think the Apollo moon landings were faked.” Also, only a small fraction of those self-identified Republicans will vote.

Primary polls are especially flawed, because most voters don’t make up their minds until a few weeks or even days before their state’s contest. And each primary’s result influences how people will vote in subsequent ones.

As polls have multiplied in this cycle, so too have the chances that they are unreliable: some because sample sizes are too small or unrepresentative; some because pollsters often don’t call cellphones, excluding (mostly younger) people without landlines. Some aggregated polls include both historically accurate and methodologically flawed polls.

(More here.)

0 Comments:

Post a Comment

Links to this post:

Create a Link

<< Home