State Pollsters Earn Good Grades | Eastern North Carolina Now

    Publisher's note: This article appeared on John Hood's daily column in the Carolina Journal, which, because of Author / Publisher Hood, is linked to the John Locke Foundation.

    RALEIGH     North Carolina has one of the most-polled electorates in the United States. As we head into a general election full of highly competitive races for governor, senator, president, and other offices, those of us who closely follow politics in the Tar Heel State will again appreciate the high frequency of polling available.

    It's not just the frequency that comes in handy, however. It turns out that the quality of polling in North Carolina is quite good, relatively speaking. According to the latest nationwide study by the website FiveThirtyEight.com, all of the pollsters that are most active in our state receive high marks for accuracy - either an A or a B+.

    Two organizations poll North Carolina roughly every month. The Civitas Institute, a conservative think tank based in Raleigh, commissions live-operator polls from a company called National Research (which gets a B+ for accuracy) as well as occasional "flash" polls by the automated-dialing company Survey USA (which gets an A). From the other side of the spectrum, the Democratic firm Public Policy Polling (B+) conducts regular automated surveys of North Carolina voters.

    Survey USA also does polling for state media outlets such as Time Warner Cable News and WRAL-TV. In addition, two universities conduct regular surveys in North Carolina. High Point University (B+) and Elon University (B+) mix electoral questions in with in-depth questions about issues and controversies.

    Other firms do polling in North Carolina on an occasional basis. But these five are the main providers of survey research about politics and public policy in our state. Their work makes up the vast majority of surveys I use to track trends in approval ratings and electoral strength. Of the 18 polls publicly released in North Carolina since the beginning of February, for example, 14 were produced by Civitas/National Research, PPP, Survey USA, High Point University, or Elon University.

    (By the way, those polling averages show Republican Pat McCrory and Democrat Roy Cooper essentially tied for governor, Donald Trump and Hillary Clinton essentially tied for president, and Republican Richard Burr leading Deborah Ross for U.S. senate by five points.)

    It's worth considering how these pollsters fared during past election cycles. Just before the voting ended in the 2014 senate race between Kay Hagan and Thom Tillis, less than a percentage point separated the two candidates on average. Of North Carolina's frequent pollsters, Civitas and Survey USA came closest to the eventual outcome with their final polls, each of which had Hagan and Tillis tied (he ended up winning by a hair). PPP had Hagan up by two points.

    In 2012, Pat McCrory defeated Democratic nominee Walter Dalton in the gubernatorial race by just over 11 points. The final pre-election polling average was actually more generous to the Republican, at a spread of 14 points. Once again, Civitas/National Research came closest to pegging the outcome. Its last released poll of the cycle put McCrory over Dalton by precisely 11 points, although the survey came out several weeks before Election Day. PPP's final survey (McCrory +7) proved to be too generous to Dalton, while Survey USA (McCrory +17) and Elon (McCrory +14) put the spread too far in the other direction.

    Polling is hardly an exact science. Most opinion surveys intended to predict outcomes are designed at a 95 percent confidence level. That is, 19 times out of 20, the poll is expected to generate predictions that are within a margin of sampling error (a few points, give or take) of the actual outcome. When such a poll shows Smith leading Brown by 1 point, and it turns out that Brown beats Smith by 1 point, that's not some outrageous mistake. It's par for the course. Furthermore, roughly once every 20 times, even the best pollster will produce a wildly inaccurate sample.

    To understand the dynamics of a campaign and eventually to form my best guess about the likely victor, I read all the survey research I can find. Fortunately, I can find a lot of it in North Carolina - and it's pretty darn good.