/ 5-minute read /
(Caveat: I am not involved in political polling. I leave that to the pros like Sarah Dutton and Anthony Salvanto at CBS.)
Blood in the Polling Waters
Amid the handwringing over how the 2016 election polls incorrectly predicted the winner, several findings have emerged:
- Three states were among the most critical to Trump’s success: Michigan, Wisconsin and Pennsylvania;
- Those classified as ‘undecided’ in pre-election surveys tended to favor Trump in the actual election;
- By contrast, Trump voters were either passionately engaged or silently pulled the lever on Election Day without having previously declared themselves to pollsters.
Uncertainty Won the Day
Post-Election Day analysts immediately seized upon the existence of a segment of respondents who answered “uncertain” or “undecided” to pre-election surveys when they really meant Donald J. Trump – The New York Times:
“Perhaps undecided voters broke for Mr. Trump; maybe there really were “silent” voters for him, people who were reluctant to tell pollsters that they backed him.”
This speculation has been quantified by the best and brightest in the market research community. Dr. Aaron Reid of Sentient Decision Science, Inc. gave an insightful talk at a GreenBook / ARF Webinar: Election Results. He presented data that examined the differences between Hillary and Trump in the aforementioned 3 key swing states (MI, WI, PA) with regard to “what they said” pre-election versus “what they did” post-vote. Post results were higher for Trump but unchanged for Hillary.
- Implication: Undecided voters were really unwilling to commit to Mr. Trump in pre-election polling.
Dr. Reid suggests that pollsters can no longer rely exclusively on what people say in order to truly understand what they will do. Rather, attention must be paid to what he labels “won’t say” social behavior and “can’t say” behavior to which consumers don’t have conscious access.
- Not surprisingly, his firm offers Implicit Association Testing to capture non-verbal feedback.
What Would Procter & Gamble Do?
Fortune 500 companies like Procter & Gamble routinely bet the ranch on achieving incremental market share measured in decimal points, not whole numbers. P&G cannot afford to go to market with purchase interest data saying “don’t know.” They’ve learned to marry ad hoc results (such as election polls) with several other streams of known behavioral measures – “Big Data.”
The future of political research will increasingly need to rely on efforts that go on behind-the-scenes prior to the election … like those of Dr. Aaron Reid’s that got very little attention and were, in fact, predictively accurate. For all the brouhaha about undecideds, they did play a pivotal role at the end of the day.