David M Rothschild on Posted on

Together with Sam Corbett-Davies and Tobias Konitzer, I ran regular polling on mobile-based Pollfish. We used the data to discuss support of public policy and quick reactions to unfolding events. The idea was that in 2016 we would study the data collection and analytics to nurture new processes, answering pressing questions now, and then use them for election forecasts rather than tradition polling when it was tested. While we did not hide it, I made a mistake by not pushing the Pollfish voter intention polling (46 of 51) harder. It persistently pointed to Trump in Pennsylvania and tightness in the rest of the rust belt. While the binary accuracy was similar to top public polling, this experimental poll consistently pointed to Clinton’s trouble in the rust belt that ultimately cost her the election to Trump, with: a more Trump leaning voter population and more support for Trump from key demographics.

The public opinion polling we did throughout the year had an impact on the narrative of what people where thinking about public policy and how they were absorbing the campaign. At the beginning of the cycle we showed in the Washington Post how well Trump’s policies aligned with the GOP voters, but not the GOP elites. As the campaign showed the uniquely targeted impact of free trade and the incredible wedge that immigration and guns drew in the American voters. We dissected the conflicting impact of the election on teenage girls for the New York Times. Showcasing how Clinton was inspiring, but also the backlash was scary. Here is one reason Trump won:

PredictWise20161110a

But, while we disseminated articles about public opinion regularly, we only talked sparingly about the voter intention polling. With so much publicly available polling, we figured that this was a great year to build out the technically, but could we really add to the discourse? Well, it turned out that the amount of binary misses in the public polling is not extraordinary, but the public polling missed three critical states: Michigan, Pennsylvania, and Wisconsin. Overall, the national polling average may be off 2 pp when the votes are all counted; not bad. But, public polling missed something about either about white turnout or support in one region and that cost them the only thing that mattered, the winner of the Electoral College. And the Pollfish polling was able to target detailed geographical and demographic sub-groups. Just look at the September map we had up on Pollfish. We had a 0.5 pp lead for Clinton in the popular vote, but had her losing: FL, NC, PA, OH, WI, and tight in MI. While she ultimately carried NH and ME this map is quite impressive:

PredictWise20161110b

 

Polling does two things: estimate the voter population and the support for each candidate from the voter population. 

The voter population that my colleagues and I created for the Pollfish polling more closely resembled the true voting population than the estimates from the public polling. This is best demonstrated by Nate Cohn’s interesting experiment of giving four different pollsters (and himself) the exact same polling data and see what topline numbers they generate. We certainty absorbed some of Latino population into our estimate of White votes (i.e., we had too few Latino and too much white, compared with the Florida exit polls, for whatever exit polls are worth). But, we ultimately had a more Republican/Trump make-up state-by-state: older, whiter, and less educated, than the what the polls estimated. That is why for the same sentiment data, our projections were constantly more Trump.

PredictWise20161109c

Further, we had support levels for Trump and Clinton that closely resembled the exit polling. We consistently had Republicans supporting Trump at a rate of 2 pp more than Democrats supported Clinton. To be honest, I was concerned; what had we done wrong? But, the Pollfish data was solid there. We had surprisingly strong support from college educated whites, already controlling for age and party identification. White males supported Trump at 64 percent and white females at 52 percent. We had Hispanic support at 75 percent, a few points higher than the exit polls, but, more crucially for the rust belt, we estimated 85 percent support for Clinton from African-Americans. In-line with the exit polls and below the 2012 numbers for Obama. All of this is backed up by the polling data and why our estimates from Pollfish were throwing rust-belt states towards Trump.

Now, more than ever, we need a population that understands each other, how we feel about different topics. And, if you think of this in-terms of market intelligence in general, we need to develop strategies that focus on capturing not just people on average (like the national polling), but really can understand detailed demographics (like White/non-college/Wisconsin). That is not just important for elections, but in an increasingly personalized world, for all marketing. I believe we do can this with mobile-first polling and the right analytics.