Poll Plot

Isn’t it great to see Democrats coming in off that window ledge? Until the polls shifted, congressional Democrats were acting like stockbrokers in 1929. The latest Washington Post poll, for example, shows that likely voters now prefer the Democratic candidate in House races over the Republican by a 51-42 percent margin. Little more than two weeks ago, the Post poll had both parties deadlocked at 47 percent among likely voters.

Before Democrats get the vapors and start fantasizing about “Speaker Gephardt,” Chatterbox feels compelled to cock a skeptical eyebrow. Consider the most important question that the Post survey uses, which is standard among published polls. Voters are asked, regarding the House race in their district, “If the election were being held today, for whom would you vote: the Democratic candidate or the Republican candidate?”

This generic horse-race query seems pretty straight forward. (Press bias monitors should note that pollsters routinely ask half the sample the question with the phrase “Republican candidate” coming first). But the question, which was created by the Gallup organization nearly 50 years ago, has one large flaw. With more than 400 House incumbents on the ballot this year, many voters will be going to the polls to vote for or against Congressman Sleaze, rather than choosing a party as in a parliamentary system.

Most pollsters cavalierly dismiss this objection by citing empirical evidence. The answers to this generic question have reliably forecast outcomes in non-presidential election years. Back in 1994, Gallup called the GOP takeover of the House by predicting accurately that Republican candidates would receive 54 percent of the national vote. (The final 1994 Post pre-election survey, though, badly missed the mark by giving the Democrats a 47-42 percent edge). Gallup also called the 1990 House races with stunning exactitude.

But barometers don’t always work in politics the way they do in weather forecasting. Already, there are cautionary warnings about the reliability of the horse-race question in an odd-duck election like 1998. Andy Kohut, who directs the Pew Research Center polls, says that low voter interest in politics explains why “there’s been so much bouncing around in these generic polls.” He adds, “This is a less certain election than most because there are so many new cross-currents out there: The issue void caused by the Lewinsky matter; uncertainty about the economy; and the scandal itself.”

Pollster G. Evans Witt in an article in the new issue of American Demographics goes even further by suggesting that may be there hidden problems with politicians’ favorite crystal ball. As Witt, the president of Princeton Survey Research Associates, writes, “It is useful (and humbling) to remember that a marvelous technique for predicting House control developed in the early 1970s by Edward Tufte of Yale (based on presidential job ratings and a single economic indicator!) lost its predicative power as the nation, the electorate and politics changed in the 1980s.”

In a follow-up interview, Witt suggested that there was also “a real possibility” that this choose-your-party question was now a flawed indicator. “What’s true about this election,” he said, “which has only been true twice in my lifetime, is that the Republicans are defending a majority in the House. And, remember, the first time was in 1996.” In other words, the political landscape was totally reshaped by the Gingrich takeover in 1994. And since 1996 was a presidential year, when the formula routinely misfires, the generic horse-race question has never been tested in an election like 1998.

The Washington political and pundit establishment craves pseudo-certainty, even in situations in which the only rational response is “damned if I know.” That’s why everyone clings to these national surveys, as if they were a life raft. For without them, the TV talking heads would have no way of handicapping 435 separate House elections. Even in the mostly hotly contested House races around the country, there are few reliable published surveys. And conveniently leaked polls conducted for individual candidates are notoriously self-serving.

That’s why, nearly three weeks before the election, Chatterbox recommends a furrowed brow, a quizzical look and a shrugged-shoulder admission of humility. The battle for control of the House is too volatile, too ill-focused and too confusing for anyone to glibly predict the outcome. If you have to bet, get good odds and gamble that this time the generic horse-race polls will be wrong.

Walter Shapiro