What Polls Can and Can't Tell Us in Presidential Politics

The polling season has started early. Nearly a year and a half before the general election, we are already in the midst of the horserace. On the right side of the aisle, straw polls just in from Iowa have already shown what anyone who can read the nonverbal writing on the wall would have already known from watching the most recent debates (see the Huffington Post blog of the debate): that Mitt Romney and Mike Huckabee can connect with voters.

Polls, even those made of straw, can provide useful data, as can focus groups. Armed with hand-dials and focus groups, Frank Luntz guided Republicans for years toward "words that work," testing alternative wordings of the same idea. His most celebrated, of course, was the "death tax," a brilliant renaming of the boring "estate tax," which brings to mind an immediate sense of unfairness and outrage: "Now they're taxing you for dying?" Of course, it didn't help that Democrats had no comparably compelling language to remind people that the two percent of the population whose estates were large enough to cross the threshold for paying the "death tax" usually hadn't worked for that money themselves and could afford to pay it even after choking to death on their caviar.

But polls and focus groups can mislead as often as they inform. They can misinform the public if voters are unaware of the extent to which what you get out of them depends on what you put into them. If you ask people if they are "pro-life" or "pro-choice," you miss all the nuances that lead two-thirds of voters to believe that we should find some "middle ground" on abortion -- if you happen to ask that question. They also mislead voters in an election when the media repeatedly report national numbers, because we don't elect our presidents in direct elections. If they can't afford to sample enough voters in a state-by-state or region-by-region basis that can approximate likely Electoral College results, the media shouldn't report anything, even on a slow news day, because doing so creates false impressions of how candidates would fare in the election (not in a fictitious national referendum) that create bandwagon effects and bias voters' judgments about electability in early primary states.

Polls and focus groups can also mislead -- and cost elections -- when campaigns don't understand their limits. They led Al Gore's campaign in 2000 to avoid talking about the earth we leave our children (notice that I didn't say "the environment"), even though that was his most enduring passion, because his consultants couldn't find their way from "the environment" (a term that is, in fact, emotionally and electorally deadening) to the voter. They used the polls, like Democratic pollster-strategists have used them in so many elections, to tell the candidate what issues to talk about, instead of using them the way Luntz used voters' responses, namely to help candidates refine the words and imagery to talk about what really matters to them. Gore showed how easily he could have turned his passion about the earth into similar feelings among the electorate in "An Inconvenient Truth," with images of glaciers falling and emotionally powerful words that conveyed -- and activated in the rest of us -- his passion, as he movingly told his listeners, with an intonation in his voice that transmitted just how important the issue really is, "This is our only home."

Polls and focus groups similarly led John Kerry's team to hold back on talking about Abu Ghraib and to put a gag rule on orators at the 2004 Democratic Convention, telling them not to speak ill of an administration that makes most of us long for the days when we had a more honest man in the White House, like, say, Richard Nixon. While Bush was spending 75% of his advertising budget on negative campaign ads -- because his team knew to ignore voters when they said they don't like "negativity" -- Kerry's team believed people's internal spin.

You can't ask people conscious questions about unconscious processes, and most of the gut-level emotional responses that lead people to vote for one candidate or another -- or to respond or not respond to a negative ad -- are generated outside of our awareness. Asking people if they are influenced by negative appeals is like asking them how their liver is responding to the candy bar they just ate. They have no clue, and when they answer, they are as often wrong as not. Why do you find Obama inspiring when you hear him on the stump? And why does he often seem distant, flat, and uninspiring in debates and presidential forums? You can try to dissect these things and take your best guess, and you can pull them apart with greater precision if you have the expertise. But most people who watch Obama on the stump just feel inspired, and most who watch him in the debates just don't. They can't tell you why, and if you ask them, they'll generate rationalizations for their gut-level feelings as often as they'll generate valid reasons.

In the current presidential year, polls have an additional limitation: People have different conscious and unconscious attitudes toward women in positions of power and toward black people, and you never know which will be dominant in the polling booth. So asking voters to describe their preferences in a race that includes one female and one black presidential candidate is risky business.

To get at unconscious reactions to candidates, you have to use different technologies than the ones that can tap into conscious attitudes. Try showing a picture of the candidate on the Internet and then asking potential voters to read a series of words presented in, say, blue, green, red, or yellow. Their task is to ignore the words and simply report the color in which they are printed. This isn't easy, because if the candidate has activated a network of associations -- a set of interconnected thoughts, feelings, images, and emotions -- that includes one of those words, the word will be harder to ignore. The reason is that it's at a heightened state of unconscious activation, so it draws attention more easily. So if you find Hillary Clinton competent, "competent" will slow you down when you're trying to ignore it and simply click on the color in which it's printed. If you find her "shrill," shrill will slow you down. Those responses are measurable -- in milliseconds -- and they can tell you something very important that poll results can't.

Or if you want to know the gut-level feelings and associations Hillary Clinton, Barack Obama, and John Edwards evoke outside of awareness, another way to find out is to use subliminal technologies that experimental psychologists discovered in the 1940s and neuroscientists have rediscovered in the last decade. Tell subjects (potential voters) that they are about to read the words of an anonymous candidate for president, and present them with a series of sentences that reflect standard Democratic positions. Prior to each sentence, present the face of one of the three candidates subliminally -- fast enough for the brain to register, but too quickly to register consciously -- so that one-third are exposed subliminally to Clinton, one-third to Obama, and one-third to Edwards. Then have them rate the anonymous candidate (whose words they believe they have read) on a series of dimensions related to leadership, strength, character, and competence. People will think they have responded to the generic candidate, but any differences will reflect their different attitudes toward the "subliminal candidate."

This kind of "unconscious polling" could have won Al Gore the election in 2000. The Gore team had a vigorous (more like vituperous) debate about whether Bill Clinton would be a help or a hindrance on the campaign trail, with some fearing his associations to impeachment and others pointing to his high approval ratings, particularly for his job performance as President. Depending on what they already believed and how they worded the questions, pollsters could read the tea leaves however they liked, and they did. But when my colleague Joel Weinberger and I presented Clinton's face subliminally on the Internet immediately before the face of the unpopular California governor Gray Davis during his recall election around the time Davis was trying to figure out whether it was a good or a bad idea to have Clinton stump for him, people had no idea they'd seen Clinton, but it affected their feelings toward Davis. Republicans and Democrats were largely unmoved, because their attitudes toward Davis were relatively fixed. But the association with Clinton strongly decreased Davis's numbers among independents swing voters, who votes decide most elections. That's why Hillary Clinton has been wise to ignore conventional punditry about having her husband too publicly involved in her campaign. Most people would give their left arm to have Bill Clinton back in the White House after years of George W. Bush, and he is associated in swing voters' minds with eight years of peace, prosperity, common sense government, and fiscal responsibility.

Are there dangers in using technologies like these? Sure, any technology can be used irresponsibly and unethically -- including the constant polling we have all grown to accept and assume. But as we have seen from the disasters of the Bush years, there are bigger dangers in not using technologies that might inform us scientifically and flying blind or relying on intuition to make decisions that can make or break elections, as when Al Gore kept his best pitcher in the bullpen. There's a big difference between using technologies to measure people's attitudes in more sophisticated ways, and inserting subliminal or otherwise "under the radar" messages into ads to manipulate voters' attitudes or emotions -- as in the subliminal RATS ad run by the Bush campaign against Al Gore in 2000, and the "Harold, Call Me" ad designed to activate negative stereotypes about black men wanting white women that the Republican National Committee ran against Harold Ford in Tennessee in 2006. Subliminal manipulation of people's associations (e.g., by unconsciously pairing Gore and RATS) and stealth racist appeals are unethical, and candidates who use them should be subjected to well-deserved character attacks. And anyone using methods to assess unconscious associations in campaigns or polling for the media should institute the same rigorous safeguards used in all research in psychology and neuroscience using these methods: get informed consent from respondents before beginning the tasks by telling them that they might be exposed to subliminal stimuli so they can decide whether or not they want to participate (most people actually find it fascinating), and "debrief" them at the end of the study about what they just saw. This would avoid any possibility of using these tests like "push-polls" that look like polls but are designed to make voters associate the opposition with things that are unsavory or untrue under the guise of asking, "What would you think if you knew that Harold Ford..."

But what recent discoveries in psychology and neuroscience have made clear is that the limits of polls and focus groups are more substantial than many voters, pundits, and political strategists have realized, and that a range of technologies can lift the curtain on some of the sentiments people can't express in words but often express in votes. And in an election year in which two of the major Democratic candidates are likely to evoke different conscious and unconscious reactions, it's time the media caught up with the science and reported both.

Drew Westen, Ph.D., is professor of psychology and psychiatry at Emory University. He is the author of The Political Brain: The Role of Emotion in Deciding the Fate of the Nation.