New Des Moines Register poll shows why so many polls during primary season are weak

The Des Moines Register released a poll of the Iowa Democratic and Republican caucuses over the weekend, showing Hillary Clinton building on her lead over Bernie Sanders and Ted Cruz surging into a comfortable 31-21 lead over Donald Trump — by far the worst margin for Trump in any recent GOP poll.

While these topline results are interesting in their own right, they aren’t as interesting as the methodology that produced them. There’s a reason why the Des Moines Register poll is considered to be the only poll that matters in the Iowa caucuses: while they don’t poll as frequently as the major national polling firms, their polls are simply better.

As I’ve written before, polling primary elections is hard. Polling caucuses is even harder. It’s incredibly difficult to predict which subset of the electorate is going to be energized to turn out in a partisan primary — let alone energized to the point of navigating the complex and demanding rules of a caucus. That being the case, traditional polling methodologies are all but guaranteed to make critical errors, producing wild results.

There’s a reason the Register showed Clinton leading Sanders by 9 points and not 41, as a previous Monmouth poll that made some shaky assumptions about the caucuses’ likely electorate had shown.

For starters, the Register’s vision of the size and scope of Iowa’s caucus electorate presents a somewhat plausible picture of reality. Many don’t. For example, CNN published a poll last week showing Donald Trump leading Ted Cruz in Iowa 33-20. The poll contacted 2003 Iowan adults, identifying 552 likely Republican caucus-goers and 442 likely Democratic caucus-goers. In other words, CNN is projecting that turnout for the Iowa caucuses is going to be roughly 50 percent. That’s absurd. Turnout in the caucuses generally hovers around 20 percent — 2008’s Democratic primary saw historically high turnout, and it still didn’t come close to what CNN was projecting for next year. The Des Moines Register, on the other hand, projects turnout close to 30 percent — more reasonable given slightly-but-not-insanely high enthusiasm headed into 2016.

In short, the Des Moines Register’s sample at least stands a chance of being representative of the electorate that will show up on February 1st. CNN’s, along with samples from many other polling firms, simply doesn’t.

Next, the Register takes into account what many public pollsters willfully ignore: non-voters are liars. Most public opinion polls call adults, ask them if they are registered to vote and then (maybe) apply a likely voter screen (Has the self-described registered voter voted in two of the last four elections? On a scale from one to ten, how excited are they to vote in the upcoming election? And so on). Few polls actually go back to check and see if those self-described registered voters are, in fact, registered to vote. Many aren’t. Given that voting is a social norm in democracies, non-voters often exhibit social desirability bias and exaggerate their voting histories. According to a working paper from political scientists Stephen Ansolabehere and Brian Schaffner, 63 percent of non-voters in 2010, when asked, self-reported that they had in fact voted. This being the case, polling firms can’t take their respondents’ self-reports at face value. In an attempt to register a socially-desirable response, citizens who aren’t even registered to vote will self-report that not only do they have an opinion about the upcoming election, but also that they are excited and likely to vote.

These faux-voters are not distributed randomly. As previous polls have shown, there’s a sizable chunk of the American public that a) identifies as Republican; b) is enthusiastic about voting for Donald Trump; and c) would not be allowed to vote were an election held today. This being the case, polls that don’t match their respondents to lists of active registered voters — as the Des Moines Register does — run the risk of drastically over-reporting Donald Trump’s support.

All this is to say that political observers are right to say that when it comes to Iowa, it’s better to wait for the Des Moines Register than it is to draw conclusions based on polls from pretty much anyone else. The corners they don’t cut give them an edge over their competitors.

When they say that Ted Cruz is surging in Iowa, Ted Cruz is surging in Iowa.

UPDATE: Keep all of the above in mind when the punditsphere freaks out over this Monmouth poll, released this morning, showing Donald Trump with a 27-point lead in the national GOP primary.

Jon Green graduated from Kenyon College with a B.A. in Political Science and high honors in Political Cognition. He worked as a field organizer for Congressman Tom Perriello in 2010 and a Regional Field Director for President Obama's re-election campaign in 2012. Jon writes on a number of topics, but pays especially close attention to elections, religion and political cognition. Follow him on Twitter at @_Jon_Green, and on Google+. .

Share This Post

9 Responses to “New Des Moines Register poll shows why so many polls during primary season are weak”

  1. Grant Saw says:

    You do realize statisticians only need a minimum of 32 people to make a statistically relevant point. 1000 people is acceptable and normal. There is greater spreading for standard deviation, but still reliable results.

  2. eenaer6yah says:

    ❝my neighbor’s aunty is making $98 HOURLY on the internet❞….A few days ago new McLaren F1 subsequent after earning 18,512$,,,this was my previous month’s paycheck ,and-a little over, $17k Last month ..3-5 h/r of work a day ..with extra open doors & weekly paychecks.. it’s realy the easiest work I have ever Do.. I Joined This 7 months ago and now making over $87, p/h..Learn More right Here….
    ➤➤➤ http://GlobalSuperEmploymentVacanciesReportJobs/GetPaid/$98hourly… ❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦.❦

  3. Jon Green says:

    They talked to roughly 2000 random Iowans (their initial sample) and determined that roughly 1000 of them were likely to vote. Your likely voter screen is absolutely a statement about how large you think the electorate will be.

  4. Peep says:

    One of my favorite dumb journalist comment’s this primary season was Donald Trump’s people won’t turn out for the caucuses in frigid weather. What? The other candidates’ voters are more likely to dress appropriately in the winter?

  5. Grant Saw says:

    “The poll contacted 2003 Iowan adults,…CNN is projecting…roughly 50 percent. ”

    How many people you poll has nothing to do with expected turnout. You could poll 1% , 10% or 90% of the electorate and it won’t tell you who will show. But regardless, a poll with that many people can give accurate results to who they will vote.

  6. Jon Green says:

    Um, yes. I think we agree?

  7. judybrowni says:

    But does reflect on who might win the nomination, thru the primaries.

  8. Jon Green says:

    In recent history, yes, Iowa Republicans have gone to social conservatives whose appeals were too narrow to win the nomination, let alone the general (Santorum in ’12, Huckabee in ’08…Pat Buchanan only lost to Bob Dole by 3 points in ’96). But that’s a separate issue from whether current polls showing Trump in the lead there, as opposed to Cruz, are accurate.

  9. judybrowni says:

    But doesn’t Iowa generally reward whoever appeals most to the fundies?

    And therefore their winning candidate generally crash and burn in the general?

© 2020 AMERICAblog Media, LLC. All rights reserved. · Entries RSS