A few hours ago, the U.S. Presidential election concluded. The result was the election of Donald Trump, an outcome that contradicted the results of almost every poll and expert prediction. Considering the amount of time and money that went into this predictive market research, the inability to come close in predicting the outcome can only qualify as an epic fail.
Ironically, the Daybreak Tracking Poll conducted by the LA Times and the University of Southern California was the only significant poll to show Trump leading Clinton in the days before the election. It used a different methodology, but apparently wasn’t considered trustworthy enough to alter the prediction of a Clinton landslide.
Here are a few thoughts on the research process, on Trump’s surge, and how you can apply what we learned to your business..
Trump Surprise: Four #Neuromarketing Takeaways from the 2016 Election Click To Tweet
Why were conventional polling techniques so wrong?
Advocates of neuromarketing have pointed out for years that conducting market research by asking people questions is seriously flawed. People can’t (or won’t) predict future behavior with accuracy, even with simple binary choices. They have even more difficulty with “why” questions that seek to understand their behavior and choices.
The challenge of predicting the outcome of the U.S. election for president is made even more difficult by several factors. For one, a national survey doesn’t mean much when the candidates are relatively close. At this point, it seems likely that Hillary Clinton will finish with a majority of the popular vote, albeit by a lower margin than most pollsters predicted. One really has to predict outcomes on a state by state basis, greatly increasing complexity and sample sizes.
Political pollsters are faced with many challenges – sampling the right people to represent the entire population is a big one. Not all segments of the population are equally accessible by landline telephones, and polling by other means is also problematic.
Plus, the researchers trying to predict election outcomes are really dealing with predicting not one but two behaviors, showing up to vote and choosing a candidate.
I don’t want to engage in a long defense of traditional survey methods, as I find them suspect for most purposes. But, the final point I’d make is that in this kind of election margins of victory can be very small. A percentage difference that would look like a rounding error in commercial market research can determine the outcome of a state and, indeed, the entire election. Commercial market research has no need for the sort of precision where 49% is an entirely different outcome than 51%.
Business Takeaway: Human behavior is hard to predict, particularly by asking small numbers of people what they will do in the future. Fortunately, consumer research doesn’t demand the precision needed to predict the outcome of an extremely close election.
Why Did Trump’s Message Resonate With Voters?
Books will be written about this topic, but there’s little doubt that Trump effectively tapped into the emotions of many voters. He surfaced fear and anxiety about their economic future. He tapped into human tribalism by focusing on differences between people and an emphasis on immigration as a threat.
Above all, it’s my opinion that Trump’s messaging was simple, emotional, and targeted at voters’ System 1 thinking. As Nobel winner Daniel Kahneman explained in his book Thinking, Fast and Slow, System 1 thinking is fast, intuitive, emotional, and energy efficient. System 2 is slow, logical, rational, and hard work for our brains. Humans will avoid System 2 thinking whenever they can, and Trump’s messaging took advantage of that.
For example, look at the debate about immigration.
Hillary Clinton. Her immigration policy has nine bullet points. Here’s the first one:
Introduce comprehensive immigration reform. Hillary will introduce comprehensive immigration reform with a pathway to full and equal citizenship within her first 100 days in office. It will treat every person with dignity, fix the family visa backlog, uphold the rule of law, protect our borders and national security, and bring millions of hardworking people into the formal economy.
Subsequent bullets include, “End the three- and 10-year bars,” “Defend President Obama’s executive actions—known as DACA and DAPA—against partisan attacks,” and more. If I listed them all, you (and everyone else) would likely tune out and leave. This message is highly complex and has to be decoded by System 2, which requires more focus and effort than most people will give.
Donald Trump. His message isn’t nuanced.
I’ll build a wall.
Clearly, immigration reform is a complex issue that requires deep analysis and nuanced policy. But, voters aren’t going to listen to that. They will tune out and remember nothing. I think a word association test would show this. Ask a thousand voters to think of a word associated with “Trump” and “immigration,” and the vast majority would say “wall.” The same question with Clinton’s name would produce a much more muddled result, with the negatively-charged “amnesty” featuring prominently.
Across just about all topics, Trump kept his messaging simple and emotional. “Make America great again!” is perhaps the best example. He rarely dug into policy specifics that would bore or confuse voters. His talking points were aimed at System 1 thinking, and the decision to vote for Trump was likely a System 1 decision for many voters.Donald Trump's messaging appealed to the voters' System 1 thinking. pic.twitter.com/Ff5Kzt8G0f… Click To Tweet
Business Takeaway: All too often, marketers have a tendency to focus on product features, specifications, prices, and other factual details. To resonate with customers, the appeal should be simple and, if possible, focused on a pain point. Make the boring details available for those who need them, but keep them out of view for most customers.
Would Neuromarketing Have Helped Predict the Outcome?
Neuromarketing techniques do have the ability to get beneath what consumers say and can sometimes predict behavior better than traditional survey methods. In the case of the flawed polls, though, I think the failure may have been more a matter of sampling and statistics. But, it is possible that some non-conscious techniques might have spotted ambivalence among some voters who favored Clinton. The scale needed to add to the accuracy of state-by-state predictions would be daunting, particularly if one wanted to do the testing within a couple of days of the election.
Certainly, some selective neuromarketing studies could have been used to tweak the larger-scale traditional poll numbers, although such tweaking would be viewed as subjective.
By the next election, though, who knows? Millions of voters will be wearing biometric sensors, otherwise known as smart watches and Fitbits. Even more millions will have smartphone and other cameras pointed at their faces, making them potential subjects for facial coding analysis. If researchers can find a way to turn these millions of voters into willing subjects, we might actually get accurate predictions.
Putting Big Data information into the mix would allow highly granular segmentation of data. Slicing results by location, ethnicity, income, and many other factors known about each consumer would yield a far better understanding of what the results mean.
And, flipping it around, this could also lead to highly targeted political marketing. Conventional wisdom is to focus on independents and undecided voters, rather generic catgories. Neuromarketing data could identify voters who are on the fence, show which way they are leaning, and suggest the most effective appeal to convert them.
Business Takeaway: Prepare for a world where neuromarketing studies aren’t a few dozen subjects in a lab but can be scaled to massive levels at reasonable cost. Not far behind will be neuromarketing-based personalization of messages.
One Prediction Tool That Worked
I watched the results of the 2016 election into the early hours of the morning. It wasn’t until after 2AM New York time that major press entities called the race for Trump. A fascinating tool from the NYTimes.com, though, made it clear that Trump was headed for victory many hours earlier.
To be clear, the major news networks won’t declare a winner unless is 99.9+ percent certain they are right. Candidates concede or claim victory based on these declarations, and the consequences of making the wrong call are enormous. So, of necessity, they are conservative.
The NYTimes.com tool, though, offered real-time projections based on the newest vote counts. As early as 7:30 PM, their chart showed a shift in momentum. By 8:30 PM, when a small fraction of the total votes were counted, they showed that the probablility of Clinton winning had slipped from a high of 85% just 90 minutes before to a 50-50 dead heat with Trump. Another hour, and Trump’s probability had surged to 90%. By 10PM, the probability of a Trump win was closing in on 95%. While there were a few quick wiggles in the early stages, as more votes came in the predicted percent grew ever more stable.
The clever folks at NYTimes.com also provided statistical error bars for their state projections. When few votes were counted, the bars were wide and, when they overlapped the vertical axis, showed the losing candidate still had a chance. As more votes were counted, these bars grew smaller, eventually turning into just a solid dot.
This was a very nicely presented view of the probable election outcome, and gave viewers useful information hours earlier than the major networks.
Business Takeaway: Use the information tool that combines speed with accuracy adequate for your purposes. Stock, currency, and commodity traders used early but reliable information like this to make “Trump wins” trades hours before the networks were close to calling the race.Don't wait for more precise data when what you have is good enough. pic.twitter.com/5BPsa9p3v8… Click To Tweet
What’s your take on the U.S. election, the failure of the pollsters, why Trump won, etc.? Leave a comment with your thoughts!