Editor's Note: This column appears in a point-counterpoint regarding whether polls are useful in predicting elections. To find its counterpoint, click here.
The most cited examples of polling error this election cycle have been the U.K.’s referendum to leave the EU and the failed polling of the Colombia-FARC peace treaty. While these are valid reasons to question the accuracy of modern surveys, there are a host of reasons why the U.S. election will be better predicted.
It is unlikely that the problems with polling in the U.K. and Colombia will appear in U.S. election polls. Unlike “Brexit” polling — in which results frequently bounced between Leave and Remain — pollsters are painting a clear, blue picture of this election that contains almost no volatility. Unlike in Colombia, 2016 primary polls and results don’t suggest that a Trump-styled “spiral of silence” or Bradley/Wilder effect even exists. Our turnout models also have a good track record.
The question then is: Will Brexit or Colombia-style error happen in America come Nov. 8? Can that error be enough for Trump to close that gap?
Probably not. Moving back in time, the popular story of polling inaccuracy starts in 1948 with the famous “Dewey Defeats Truman” misfire. The error in 1948’s pre-election polls arose from the use of quota sampling, a method that has since been proven to have major bias toward Republicans. Pollsters now use probability sampling — a more accurate measuring tool than quota sampling.
This suggests that methodological improvements to polls have given surveyors better voter preference information, a suggestion echoed by Gallup’s archive of trial heat polls. On top of this, there are other pollsters that are even better measurers. Moving even further, predictive power in polling is increased when we pool polls. Indeed, aggregates of polls have been used by election forecasters as the foundation for statistical models that correctly predict election outcomes.
The Upshot election blog all but proved this by showing that variance in polls can occur even with identical data. Although the four polls they reported had a spread of five points, an average of the polls showed a clear 1.6 percentage point lead for Hillary Clinton — on par with other reliable averages at the time.
In sum, we have all the reason to believe that polls this year will be accurate, as they were in 2008 and 2012, especially with the added predictive power of election forecasts. Arguments otherwise may fall victim to wishful thinking or analyses that aren’t holistic, or both. Polls are clearly telling measures of voter preference — much more so than relying on punditry, as epitomized by Trump’s rise in the 2016 GOP Primary.
Morris is a government junior from Port Aransas. Follow him on twitter @gelliottmorris