Whenever I happen to watch some talking heads on a cable news channel – usually while stuck in an airport – I’m always impressed by how mistaken the basic premise of the conversation is. The pundits will waste lots of words on how Obama’s pivot on FISA might turn off his liberal base, or how McCain’s tax cuts will appeal to working women, or they’ll consider the political implications of whatever issue happens to be in the news that day. The underlying assumption, of course, is that issues matter, that voters are fundamentally rational agents who vote for candidates based on a coherent set of principles. In other words, they assume that my political preferences reflect some mixture of ideology and selfish calculation. I’ll vote for the guy who best matches my geopolitics and tax bracket.
The problem, as political scientist Larry Bartels notes, is that people aren’t rational: we’re rationalizers. Our brain prefers a certain candidate or party for a really complicated set of subterranean reasons and then, after the preference has been unconsciously established, we invent rational sounding reasons to justify our preferences. (Some voters, of course, probably do chose their candidate for “rational” reasons, but I have yet to meet very many of them.) This is why the average voter is such a partisan hack and rarely bothers to revise their political preferences. For instance, an analysis of five hundred voters with “strong party allegiances” during the 1976 campaign found that, during the heated last two months of the contest, only sixteen people were persuaded to vote for the other party. Another study tracked voters from 1965 to 1982, tracing the flux of party affiliation over time. Although it was an extremely tumultuous era in American politics – there was the Vietnam War, stagflation, the fall of Richard Nixon, oil shortages, and Jimmy Carter – nearly 90 percent of people who identified themselves as Republicans in 1965 ended up voting for Ronald Reagan in 1980.
That said, when our preferences in the voting booth can be influenced they are often influenced by completely arbitrary factors. (In a 2004 paper, Bartels argued that “2.8 million people voted against Al Gore in 2000 because their states were too dry or too wet” as a consequence of that year’s weather patterns. In other words, these climatic acts of god cost Gore the election.) Or consider this new PNAS study summarized in the Globe Ideas section:
Several researchers examined voting across different polling locations from the 2000 election in Arizona, where a ballot initiative proposed to raise the sales tax to fund education. The initiative received a couple extra percentage points of support from people whose polling location was a school – a boost not seen with other initiatives unrelated to education. This effect did not go away even when adjusting for the possibility that different types of people might be voting in different types of locations. (Support for the initiative did not depend on how new the school was, suggesting that voters weren’t consciously assessing their surroundings.) The researchers also ran an experiment that randomly assigned people to view images of schools or office buildings and then vote on the same Arizona initiative. Not only did school images increase support for the initiative, but they significantly mitigated the relevance of one’s fiscal philosophy or parental status. None of the participants perceived the images’ influence, even after being told. Of course, schools aren’t the only kind of polling location: 40 percent of the votes in the 2000 Arizona election were cast in churches.
And then there’s the Peter Jennings effect. During the 1984 presidential campaign, Brian Mullen, a psychologist at Syracuse University, excerpted thirty-seven segments of campaign coverage from the nightly news programs of ABC, NBC and CBS. He then showed these short snippets to dozens of people and asked them to rate the facial expressions of each news anchor on a 21-point scale. Dan Rather of CBS was the most neutral anchor. When he talked about Ronald Reagan he scored 10.37, compared to 10.47 when he talked about Walter Mondale, the Democratic candidate. (Higher numbers signify more positive facial expressions, such as a smile.) Tom Brokaw, the anchor at NBC, was similarly unbiased, scoring 11.50 when covering Reagan and 11.21 when covering Mondale. However, the expressions of Peter Jennings, the ABC newscaster, seemed to exhibit a consistent partisan slant. When covering Mondale, his facial expressions got an average rating of 13.38. But when Reagan was the subject, Jennings’ face lit up, so that he scored 17.44.
Mullen then conducted a phone survey of two hundred people across the country that regularly watched the network news. He found that, in every single television market, people who regularly watched Peter Jennings were more likely to vote for Reagan. According to Mullen, this statistically significant discrepancy was caused by the subtle bias of Peter Jennings: viewers of ABC were persuaded to vote for Reagan due to the “operation of his emotional cues”. In other words, the smile of news anchor swayed their vote.
Of course, most voters would deny being influenced by Peter Jennings’ facial expressions, just as they would deny being influenced by the location of their polling place. (For the record, Peter Jennings vigorously denied having a pro-Reagan bias.) We are utterly ignorant of the argument taking place inside our head, clinging instead to the illusion of our rationality.