One of the oft-repeated narratives coming out of the last election cycle is “The polls were wrong.”
While not entirely accurate, it is true that few analysts predicted a Trump victory.
Suggested reasons for such widespread inaccuracy include unthinkability bias, groupthink, faulty assumptions, dishonest responses, a shockingly unorthodox candidate and, of course, poor sampling.
One excuse is structural: it’s becoming harder and harder to gather interview responses via phone and anticipate who is going to vote. Whatever the cause, it has fanned the flames over whether polling is more art or science. Either way, it’s clear both have to improve.
Luckily, as it relates to the science side things have been getting better. A lot better.
Today, through data science, we can examine individual preferences, beliefs, and choices at a granular level and ultimately use these findings to predict future behavior.
Similar to the way companies use data to determine who is likely to choose Coke or Pepsi, it’s possible to predict with a high level of confidence who will vote, what messages will motivate or persuade, which policies have support, and, yes, which candidate(s) will win or lose.
Predictive models give us insights that sample surveys alone cannot, and merging the two tools allows us to make polling more accurate as well by fully accounting for vote likelihood.
Further, models can perform in ways no poll ever could. Rather than just delivering data to a candidate informing that, say, “roughly 60% of a county supports school choice”, predictive models inform specifically who which voters those supporters are, beyond just their demographic composition.
This kind of information better instructs a campaign on who to turnout, who to persuade, and who to avoid. It means the use of resources become highly efficient, delivering more the right message to the right voting bloc and ultimately eliminating the waste of critical resources.
At WPA, we believe analytics is the future of research. If you’re not including analytic tools in your campaign research, you’re not in the game. You’re using a Walkman in an Smartphone world.
To stay ahead of the shift, we have invested significantly in making data analytics an underpinning staple of our work. The smartest decisions are no longer coming from the graybeards poring over crosstabs (the political equivalent of the baseball scout discounting a prospect over his choice of girlfriend), but PhD’s writing algorithms and modeling a race.
Whether it’s business, politics, marketing, or public affairs, better intelligence means better decision-making and that means a higher level of success for our clients.
As part of our rebranding, WPA Intelligence’s new logo includes a rendition of Leonardo da Vinci’s icosahedron, a 20-sided geometrical figure. Perhaps the world’s most famous Renaissance Man, da Vinci was equal parts artist and scientist.
da Vinci’s illustration of the icosahedron aptly symbolizes the intersection of these disciplines; an effort to envision the rigor of geometry. It is in this space — the border between the creative and scientific — where WPA Intelligence finds its focus. Although we did take liberties with da Vinci’s illustration, adding neural networks.
For more than a decade, WPA — now WPAi — has been developing, honing, testing, innovating, and implementing various strategies using data science and predictive analytics.
Successes reach back to Utah Senator Mike Lee’s 2010 unexpected victory over a long-time incumbent, 2012’s upset by attorney Ted Cruz over an establishment favorite, to our 2016 work on behalf of the Club for Growth to assist in Wisconsin Senator Ron Johnson’s come-from-behind victory.
As Director of Research, Analytics and Digital Strategy for Ted Cruz’s presidential campaign, WPA Intelligence’s Founder Chris Wilson, along with WPAi Chief Research Officer Bryon Allen, helped build more than 1,800 models nationwide, propelling the campaign to a stunning win in Iowa’s primary (where they used 167 different modeled segments to target specific voter groups).
There were few instances as high profile of the failures of traditional polling as a stand alone predictor as that of the public surveys released in advance of the Iowa caucuses. While public data predicted a significantly different outcome, our internal Cruz models provided confidence and clarity in the final decisions made leading up to caucus night.
And just recently, WPA Intelligence’s data models demonstrated the much-maligned House Freedom Caucus in fact curbed a potential fatal error in legislative governance that would have yielded significant ramifications in the midterms by opposing the American Health Care Act.
Like any tool, data science doesn’t guarantee victory or success. But certainly the better intelligence the campaign or the organization has, the more likely it is to make better decisions, use resources wisely, and produce results.
Today, companies wouldn’t be caught dead without an in-house IT department. Likewise, it is time for the data team to become as integral on campaigns as media consultants, fundraisers, and even energetic volunteers.
All the polls might be wrong, but data never lies. We’re betting the future on it.