NYT: Polls in GA Underrepresent African-American Vote

Over at the New York Times’s new data journalism hub, The Upshot, Nate Cohn has cast doubt on the accuracy of recent polling in Georgia’s US Senate race.

Specifically, he notes:

Recent polls are most likely underestimating the share of voters who are black, along with Ms. Nunn’s share of the vote.

And he has the numbers to back it up. Even if we assume the electorate will remain unchanged from 2010 (the closest midterm election), the distribution of voters in this election would stand at 66.3 percent white and 28.2 percent black.

That’s at odds with most of the polling in Georgia so far.

Cohn notes that recent polls by Rasmussen and the AJC each pegged the black share at 24%, while SurveyUSA opted for 27% instead.

In fact, only two recent polls—one by Public Policy Polling and the other by NYT/CBS/YouGov—have decided to accept the 2010 numbers and use the 28% statistic in their weighted averages.

But a lot has changed in four years, and the white population as a proportion of voting-age adults is shrinking by half a percentage point each year, causing a gradual erosion of the Republican Party’s base here in Georgia.

Accounting for the changes since 2010, Cohn paints a different picture of the 2014 electorate:

Combining the data on registered voters with census data on the voter-eligible population, I expect the 2014 electorate to be about 64.2 percent white and 28.8 percent black.

If pollsters used those figures, Cohn says the US Senate race would be a “dead heat.”

Of course, our friend at the New York Times makes a lot of assumptions. For example, that Nunn will win 90% of the African-American vote, that minorities will turn out to vote at least as much as they did in 2010, and that underrepresented populations were actually able to register to vote in the first place by the Oct. 6 deadline.

But one thing is clear: Georgia’s population is changing, even if our pollsters are not.

9 comments

    • Baker says:

      I am not African-American and I am not a Democrat so I’m just guessing, but was Thurmond really a draw to get out new(ish) voters or drive turnout in 2010?

      • jh says:

        PPP thinks so. I’m not so sure. I think he happened to be their best candidate available.

        Democrats are deliberately filling the bottom of the ticket with a lot of black females designed to drive up the black vote. I doubt it has any impact. I also don’t think Thurmond drew out any black voters any more than Nunn would.

        So basically, PPP could be right as well as all the other national pollsters that are underweighting black voters.

        But given demographic changes, the intense voter registration efforts, and the Bannock Street turnout effort, I think black turnout at 2010 levels would be a minimum for 2014.

      • jh says:

        I think part of the high black turnout of 2010 was due to the recency of the 2008 election as well. Just an unvalidated hypothesis though.

  1. NorthGAGOP says:

    I believe Roundtree/Landmark have been saying that same since they released their first poll this cycle. In a number of polls women have also been underweighted.

  2. Harry says:

    When Obama isn’t on the ballot the black turnout is an open question, let’s wait and see. My advice is give us old white guys one more shot and then you can have it in 2018.

  3. Tom in Dade says:

    Hey Tyler! Let’s get a little more Dade County up in here…not wading in on the main argument, just thought I’d add this evaluation from fivethirtyeight.com, on the relative accuracy of the pollsters who have been doing the bulk of polling in Georgia:

    Survey USA ‘A’
    Abt SRBI ‘B+’
    PPP ‘B-‘
    Landmark ‘C+’
    InsiderAdvantage ‘D’

    http://fivethirtyeight.com/interactives/pollster-ratings/

Comments are closed.