I blew it.
Three months ago in this column, I asked why, in a time of peace and prosperity, voters weren't rallying behind the incumbent presidential party. Would the good times carry Al Gore to victory over George W. Bush? "I don't know, and neither do you," I concluded. "There are no absolute laws in politics or any other social science. Laws operate only within certain parameters and under certain conditions. Those parameters and conditions, in turn, are subject to human intervention."
A month later, with Gore flying high in the polls, I forgot that lesson. In a column titled "Why Bush Is Toast," I wrote, "The numbers are moving toward Gore because fundamental dynamics tilt the election in his favor. The only question has been how far those dynamics would carry him. Now that he has passed Bush, the race is over. … Stick a fork in [Bush]. He's done."
Today, as election officials recount the ballots in Florida, Bush's fate remains in doubt. But one thing is clear: My record as an election prognosticator is cooked. It's just as well. Since its inception, this column has tried to do two things. One is to get outside the box of conventional punditry and explain how broad, hidden biases—not just obvious ideological biases—skew the way we think about current events. The other is to participate in that same punditry. The first enterprise has always been more distinctive and valuable to readers than the second. Most of us pundits suspect we're no better at predicting election results than the average intelligent person is. In my case, there's now proof. But the lesson goes both ways: I ought to spend less time playing the pundit game and more time deconstructing it. With that in mind, let's look at how I screwed up.
1. Overlooked factors. Any theory that tries to predict human behavior on the basis of a few variables is incomplete. When other factors remain dormant or cancel out, the theory appears to work, as the simplistic models developed by academic forecasters have done in some recent elections. But as soon as the excluded factors come back into play, the theory fails. The academic models err because they rely almost exclusively on two variables: economic data and poll ratings. My theory, which incorporated six factors, was more complex, but not complex enough.
The factors I included were 1) peace and prosperity; 2) an "issue terrain" favorable to Gore; 3) the imbalance of negative information about the candidates (voters knew more about Gore's record and flaws than about Bush's); 4) the "moral expectations game" (voters had already factored Gore's ethical liabilities into their evaluations); 5) collateral damage (Bush couldn't drive up Gore's negative ratings without driving up his own); and 6) the Bush campaign's inferior political judgment.
What's notable, in retrospect, is what this theory didn't include. Some missing factors that turned out to be very important—principally, doubts about Bush's intelligence and experience—played to Gore's advantage. But others decisively helped Bush. My theory took no account of ideology, which Bush exploited to reframe the election in late September. His emphasis on trusting people rather than trusting government drove a key wedge between Gore and many swing voters. Other factors I omitted—Republican and Democratic turnout efforts, for example—canceled each other out. I neglected to factor in Gore's superiority as a debater, but I also overlooked his annoying debating style. The mutual cancellation of these factors led to the same result as if my theory were correct. But that was sheer coincidence.
2. Human error. Objective factors such as economic growth don't magically translate into votes. They have to be communicated. That was the point of my Aug. 15 column:
The law's basic formulation—"In a time of peace and prosperity, voters will re-elect the party in power"—assumes at least three conditions. First, voters must remember a time of war or economic difficulty. Otherwise, they won't be able to identify, in relative terms, a time of peace and prosperity. Second, they must believe that politicians can significantly affect the likelihood of peace or prosperity. Otherwise, they won't believe that either party is truly in power. Third, they must think in terms of party. Otherwise, when the incumbent party substitutes one nominee for another—even when the new candidate has been vice president under the old one—voters will see nothing to re-elect. … If I fail to see prosperity in its historical context, or to associate it with elections, or to think of Bush and Gore in terms of their parties, I won't vote the way the theory predicts. Somebody has to step in and make me see things that way.
It's doubly ironic that I forgot that lesson four weeks later. The premise of this column—and the reason it's called Frame Game—is that our perceptions, reasoning, and decisions are shaped by the efforts of political players to make us look at things from one perspective rather than another. Absent subjective framing, objective conditions guarantee nothing. The academic forecasters who predicted a Gore victory on the basis of economic growth are now blaming Gore for failing to frame the election as a referendum on the economy. They think this failure on his part is outside the calculus of predicting elections. But it isn't. By leaving it out of their models, they—and I—predicted the wrong result. Instead of dismissing this inconvenient factor as an anomaly, they ought to incorporate it in their theories.
Will Saletan covers science, technology, and politics for Slate and says a lot of things that get him in trouble.