A Forbes article by Jonathan Vanian said it best, “Prior to Trump’s upset win, virtually all national polls showed the businessman and reality television star trailing Democratic nominee Hillary Clinton. Her win was considered inevitable, with prominent pollsters and pundits merely arguing about how big her guaranteed victory would be. And then on Tuesday, voters proved the experts wrong.”
What did Trump do right? To say he ran an unconventional campaign would be a gross understatement. He methodically eliminated 16 primary contenders, countering their talking points with his bombastic personality. He captured an inordinate amount of free TV time by being outlandish. Trump had a sense of what people wanted to hear and recognized that anger among working class white voters ran deep. He played to emotion, not data points.
Steven Bertoni’s article headline in the December 20, 2016 in Forbes magazine read, ”The secret weapon of the Trump campaign: his son-in-law, Jared Kushner, who created a stealth data machine that leveraged social media and ran like a Silicon Valley startup.” Kushner, who had no political experience, committed to Trumps’ campaign in November 2015, after seeing a raucous Trump rally in Springfield, Illinois. On that return trip, Trump and Kushner talked about how the campaign might better use social media.
”At first Kushner dabbled, engaging in what amounted to a beta test using Trump merchandise. ’I called somebody that works for one of the technology companies that I work with, and I had them give me a tutorial on how to use Facebook micro-targeting,’ Kushner says. The Trump campaign went from selling $8,000 a day worth of hats and other items to $80,000, generating revenue, expanding the number of human billboards—and proving a concept. By June the GOP nomination secured, Kushner took over all data-driven efforts. Within three weeks, in a nondescript building outside San Antonio, he had built what would become a 100-person data hub designed to unify fundraising, messaging, and targeting.”
In this case, a lack of awareness of traditional campaigning was an advantage. Kushner was able to look at the business of politics without the constraint of precedent.
Eric Schmidt, the former CEO of Google and one of the designers of the Clinton campaign’s technology system, agrees with Vanian. “Jared Kushner is the biggest surprise of the 2016 election. Best I can tell, he actually ran the campaign and did it with essentially no resources.”
What did Clinton do wrong? According to The Washington Post, Clinton’s campaign used a custom Algorithm called Ada; a complex computer algorithm that staff fed “a raft of polling numbers, public and private” to play a role in most strategic decisions Clinton aides made. According to aides, Ada ran 400,000 simulations a day and a report was generated that gave Robby Mook, the campaign manager, a detailed roadmap of which background states were most likely to tip the race in one direction or the other, allowing them to decide where and when to send the candidate and her surrogates and where to air television ads.
Like much of the political establishment, however, Ada did not accurately predict the turnout of rural voters in Rust Belt states. Pennsylvania was correctly identified as a critical state early on, which explains why Clinton visited it often and closed her campaign in Philadelphia. Other states that Clinton would lose, like Michigan and Wisconsin, either were not identified as at-risk or were deemed so too late.
A number of election post mortems indicate that Bill Clinton, a politician with proven fluency in reading and responding to voter emotion, advocated that his wife’s campaign pay more attention to white working class voters. Perhaps he reasoned that while that group was not within Clinton’s reach, she might draw enough votes to win Michigan, Pennsylvania, and Wisconsin—states that Trump narrowly won instead.
Let’s look at data analytics as it pertains to a presidential election: An article in Wired by Cade Metz stated, “The lesson of Trump’s victory is not that data is dead. The lesson is that data is flawed. It has always been flawed—and always will be…. But this wasn’t so much a failure of the data as it was a failure of the people using the data. It’s a failure of the willingness to believe too blindly in data, not to see it for how flawed it really is.”
Summary: The use of data analytics by presidential campaigns did not begin in the 21st century. Clinton aides believed their work with data was the most sophisticated to date, and while this may be true, it did not translate to a strategic advantage over Trump when all other factors were accounted for. If Barack Obama’s 2012 presidential victory proved big data’s triumph for accurately predicting elections, Donald Trump’s 2016 presidential win could demonstrate the opposite.
According to Nik Rouda, senior analyst at the Enterprise Strategy Group, “Polls aren’t really big data. The sample sizes were certainly good enough for a poll, but maybe didn’t meet the definitions around volumes of data, variety of data, and historical depth contrasted against real-time immediacy, machine learning, and other advanced analytics. If anything, I’d argue that more application of big data techniques would have given a better forecast.”
Professor Samuel Wang, manager of the Princeton Election Consortium, which gave Clinton a 99 percent chance of winning as of the morning of Election Day, stated ”The incorrect forecasts don’t appear to be a problem with the margin of error, the polling resulted in a systematic error. The entire group of polls was off, as a group. This was a really large error, around 4 points at presidential and Senate levels, up and down the ticket.”
Wang went on to say he is still evaluating the results. Late candidate selection by undecided voters may have impacted predictions. Whether predication models can better account for such last-minute decisions or changed minds remains unknown for now.
As the world makes the Internet its primary means of communication, we will be confronted with even more data—so-called “Big Data.” On the Internet, fact, opinion, idle chatter, and humor run together in a common sea where intent is difficult to ascertain and the bots are becoming more indistinguishable from the humans. The old database maxim of “garbage in, garbage out” should guide all efforts to incorporate this data in predication models.
And in the meantime, the even bigger promise is that artificial intelligence will produce more reliable predictions. But even the most sophisticated artificial decision engine remains dependent on imperfect data inputs. Neural networks can’t forecast an election without data—data that is selected and labeled by humans. While such AI systems have become adept at object recognition because people have uploaded millions of photos to places like Google and Facebook already, we lack the same kind of clean, organized data on presidential elections to train neural nets.
Conclusion: Are your data analytics predictions models suffering from the same problems as the models that predicted Hillary Clinton would easily win the U.S. presidential election? Next month we will explore this further.