The pollsters were wrong. Again. They are rather like economists, in that they continue to try to portray what they are doing as some kind of science, when it fact it’s more like art. Or at least biology rather than physics.
Just as quantitative traders these days find that using historic data to try to back-test future investment strategies isn’t working, political pollsters are finding that looking at 2016 turnout numbers or voting cohorts as any indicator of what might happen this year was simply a wrong-headed strategy. As any FT reader knows, past results are no indicator of future returns, particularly when the world is changing so fast that both historic economic and political data are more likely to be misleading.
But there’s something else going on with these pollsters, most of whom are well-educated, coastal types who spend much more time calculating data sets and crafting questions than they do tuning into voter psychology. I called up the Republican party strategist and pollster Frank Luntz, the man who helped conservatives change words such as estate tax into “death tax” and global warming into “climate change”, to get his view on the issue.
Luntz started saying a couple of weeks ago that he thought many polls might be over-optimistic about a Democratic sweep, in part because of his own mishap in 2016, when he relied on Fox exit polls rather than his own focus groups. “Back then, I called Dick Cheney to say Trump would lose and Democrats would give up Senate seats. Later, when that turned out to be wrong, I called to apologise.” When Cheney accepted the apology and asked if Luntz wanted to come hunting, he politely declined the offer. Wise man.
The core problem with polling, according to Luntz, is a general lack of empathy on the part of pollsters towards the subjects. “A lot of Trump voters simply don’t do polls, because they feel that pollsters look down on people like them or don’t respect their point of view.” Subjects that pick up on this will shut down and simply refuse to participate, give short and unhelpful feedback or even provide misleading answers.
To combat this, Luntz tries to loop them in, he says, with an opening script that stresses “your chance to be heard” and sprinkles in empathetic language such as “knowing that each community is different”, etc. He also tries to keep his language simple. “I took a journalist on a polling trip once, and he asked a question about social security that took 15 minutes. The subject just stared at him. I made it a 15-second question and he gave a perfectly clear answer.”
The question itself is telling. “Would you be prepared to sacrifice now to know that the social security system would be safer in the future?” Oh, interesting, I said to Luntz — you are using the word “sacrifice” to connotate duty, honour, perhaps patriotism? “No!” he said. “You are making the same mistake some pollsters would make. You are a rich person — to you, the notion of sacrifice connotes something good.” By contrast, an unemployed Trump voter thinks, “I’ve already sacrificed. Let the rich sacrifice.” Excellent point — and mea culpa.
Words matter, but so does body language. “I really much prefer to take 15 minutes with someone in a focus group rather than 15 seconds with a question or two,” he says. “I want to talk to a dozen people, not a thousand. I need to see how they look at me, how they move. What are their motivations?” Luntz says if he had to go back to school, he’d study behavioural economics. I bet a lot of politicos — not to mention economists and traders — would agree.