Q: What steps are pollsters taking so they don’t repeat the mistakes of 2016?
A: Let’s clarify that “the pollsters” were by-and-large accurate in their national predictions back in 2016. The national pollster the Guru follows is ABC News’ Gary Langer and his final poll had Clinton up by 3% and she won the popular vote by 2%. No problem there. The problem was with the state polls in three fiercely contested upper mid-western battleground states. These are typically conducted by colleges who want to make a name for themselves – think Gonzaga’s basketball program – who use students as interviewers and don’t have big polling budgets. What could go wrong, right?
A 2016 post-mortem was conducted and two points were advanced: conduct more polls in battleground states closer to the election, and include more lower- education voters in the sample. On the first point, where’s the money for that coming from? Pollsters are paid by schools and media outlets and don’t control how many polls are conducted. As for better sampling of less educated voters, the Guru thinks that’s a legitimate area for quality improvement.
But why stop with education? Just because high school-educated whites surged in 2016 doesn’t mean they’ll do so again. Specifically oversampling them would then lead to overcorrection. The better solution is to do everything better and that means a whole host of things from better questionnaire design, to better sampling, to a multi-mode data collection approach, to better training, to better weighting of the sample and, finally, better analysis of the data.
And here’s the punchline: CERC’s 2016 polls on the presidential contest were right on. That’s because we do the things that lead to greater accuracy, even though they cost a bit more. We don’t cut corners and we take our jobs seriously. Ultimately, America needs serious pollsters.