{ insiders’ ADVICE Polls Post Mortem steve vancore explains how pollsters got the presidential election wrong, but not that wrong WINTER 2016 INFLUENCE | 77 number of these polls was down significantly (by one account, over 40 percent since 2012) and the aggregators noted above were simply relying on scant data. Second, in key swing states — those states where the campaigns were engaged — the underlying predictive models were wrong. Without going into too much detail about rural/urban, income and education demographic changes, the simplest explanation is that Republicans ultimately voted in larger numbers and/or Democrats voted in lower numbers than most pollsters thought they would. In Wisconsin, Michigan, and Pennsylvania most notably, the turnout differentials were enough to change the outcome of the election in those states. The polls — what few there were — were based on turnout predictions that simply turned out to be wrong. In Pennsylvania, for example, Democrats turned out about 5 points lower than they did in 2012 and Republicans turned out nearly 9 points higher. Most pollsters who worked in that state AND published their responses, simply did not see that coming. (We simply do not know what the private pollsters were doing because they did not publish their polls.) Does this mean polling is dead? Should you have less trust in your pollster? No and no. Remember, the vast majority of polls you saw or read about leading up to the election were created for public consumption. They were most often produced by media outlets or universities who, due to budget reductions, were more likely to use cheaper methodologies to create their “made-for-prime-time” polls. That alone can, and seemingly did, compromise the final products. But, you should be more careful. Ask more questions. Be an educated consumer of the polls that you commission. Make sure your pollster uses good modeling techniques, invests time and energy understanding your audience and — now more than ever — doesn’t cut corners. Save the killing for another day. Steven Vancore is president of ClearView Research, a political polling and research firm in Tallahassee. Steve has been conducting polls, focus groups and related research projects in Florida for nearly three decades. He can be reached at svancore@vancorejones. com. Why were national polls so wrong? One supposes if Shakespeare were writing “Henry VI” following the 2016 presidential election, he might have penned, “The first thing we do, let’s kill all the pollsters.” Not since “Dewey Defeats Truman” have our nation’s political prophet-eers been so flat out wrong. Or were they? The rolling average, as posted on websites such as RealClear Politics, FiveThirtyEight. com, Huffington Post, Associated Press and similar such sites, showed Clinton with a tightening +3.2-point lead heading into Election Day. She ended up with a final +2-point lead in the popular vote. So, combined, the published polls most of the polling aggregators from the above list were relying on were about a point off. That’s not that bad. But they were consistently off on the Electoral College counts (based on state-level polls) and that directly led to the tsunami of wrong predictions going into the election. They were consistently wrong, it turns out, for two very specific reasons. First, believe it or not, the actual PHOTO: Via NBC.com
19520EE
To see the actual publication please follow the link above