Monday, July 24, 2017, 11:30 AM – 1 PM EDT
Donna Brogan, Professor of Biostatistics Emerita, Rollins School of Public Health
"Why the 2016 U.S. Presidential Polls Were ‘Wrong’: Implications for Future Polling"
After the election’s surprising outcome there has been much discussion in the media about why the polls were “wrong.” In this presentation, Donna Brogan, Professor of Biostatistics Emerita, will focus on post-mortems among professionals who work in sample survey methodology, the area of her own particular expertise. She will share the results of a report the AAPOR (American Association for Public Opinion Research) published last May on their review and assessment of the polls. Other organizations and individuals such as Nate Silver have offered explanations based on data analyses since the election or their own opinion. The polling "errors" seem to fall into two categories denoted by Nate Silver as “technical” and “journalistic.”
Among the many possible “technical” errors, the primary culprits are likely to be those generated by the dual frame RDD (random digit dialing) sampling method used in both public opinion and election polling. The methodological problems of this approach have multiplied in recent decades, and those related to statewide polls as opposed to national polls are especially acute. This makes the data that emerges from such polling somewhat questionable, or at least more uncertain than is typically reported. “Journalistic” errors can and do often compound that problem. Examples of less than adequate reporting among media authors presenting poll results for public consumption are easy to identify, in hindsight, at least, among them the reporting of estimated percentages and estimated Clinton lead over Trump with no indication of their uncertainty.
Perhaps the suggestions of survey methodology professionals for better sampling, analysis and modeling methods, as well as better interpretation of the poll results, will lead to better polling the next time around—and fewer people stunned on election night.