Why Pollsters Got the 2016 Election Wrong
Link to source:
http://www.uniondemocrat.com/opinion/4816894-151/how-herding-blinded-the-pollsters-to-a-trump
This opinion piece from the Union Democrat explores how herding blinded the pollsters to a Trump win. In the context of political polling, “herding” is the phenomenon where pollsters publish results that fit the narrative at that point in an election, even if the results of their polling indicated a different result. the effect of this phenomenon is so strong that even though it is well-known among pollsters, they are still reluctant to publish outlier results. When one interviewed pollster was asked about whether pollsters are nervous about shifting the data in the wrong direction, the pollster said: “It’s better to be wrong with everyone.” This shows the profound effect that herding has on the poll results that we consume, and base our opinions of the election on.
This piece clearly demonstrates the herding effect that we learned about in Chapter 16 of the textbook. In Chapter 16.2, we examine a simple herding experiment involving students drawing different colored marbles out of an urn. The experiment shows that an information cascade can start at just the third student, provided the first two students’ marbles were the same color and that knowledge was public. This is same result is evident in polling, with the students being the pollsters, the urn being the people polled, and the marbles being the polls. If the first two major polls come out a certain “color” (same sized margin for the same candidate), the third pollster will assume that that result is more accurate than his own, regardless of what his “marble” (poll) tells him. The close parallel between polling and the marble example proves that the information cascade does happen in polling because it is observed in the marbles example.
Even though herding certainly played an role in the pollster’s errors, there are other factors that have to be considered. The pollsters that were interviewed in the Union Democrat piece added that the lack of blue-collar white voters (Trump’s strongest demographic) who answer their phones led to an under-sampling of that demographic, which shifted polls in Clinton’s direction. Pollsters also urge citizens to look at polls with skepticism because of low response rates. Also, national polls should be looked at with even more skepticism because the national popular vote doesn’t determine the winner. Hillary Clinton actually performed almost as well in the popular votes as the polls suggested, even though she lost the Electoral College handily.