Bill Clinton may be dominating the headlines with the wild gyrations of the Monica Lewinsky contretemps, but the president is not on the ballot this fall. No doubt Democrats are breathing a sigh of relief at that stroke of luck-more than 400 party stalwarts are on the ballot, seeking governorships and seats in the U.S. House and Senate. But most are still deeply worried that the chaos surrounding Clinton might translate into big losses in November.
With public opinion on Clinton decidedly schizophrenic after he admitted in August that he lied about the Lewinsky affair, predicting the outcome of the fall elections might seem an impossible task, especially for the House of Representatives. Few of us have the resources to follow and tally the polls for each and every one of the 435 House races.
George Gallup, Sr., one of the early giants of polling, had just that problem in mind when he created the generic congressional ballot. Just as he developed the presidential job rating as an alternative measuring stick during the years when there was no presidential election, Gallup devised a question that asked about all the races for the U.S. House in a national poll. Starting in the 1950s, The Gallup Poll began to ask: If the elections for Congress were being held today, which party's candidate would you vote for in your congressional district-the Democratic candidate or the Republican candidate?
It's simple, it's direct and it avoids all kinds of complexity.
And lo and behold, it seems to work-at least for off-year elections like those being held next month. That is, if you add up all the votes cast for Democratic candidates for the House and then add up all the votes cast for the Republican House candidates, the relative percentages look a lot like the generic ballot numbers. (This calculation just ignores the minor party candidates.)
The folks at Gallup say that the average error in their final pre-election poll has been just a shade over 1 percentage point per election since 1950. If you wanted a warning that the Republicans were going to take the U.S. House in 1994, the last Gallup Poll just before Election Day gave about 54 percent of the vote to the GOP-almost exactly what they eventually racked up.
Of course, control of the U.S. House is not about percentages of the overall vote, but about winning at least 218 seats in the lower chamber. Still, Gallup analysts make the case that their numbers were also a remarkably accurate predictor of the GOP seat gain in 1994.
But wait a minute. Aren't there all kinds of reasons that this relationship between the national poll and adding up 435 races should not work?
Absolutely. The calculation doesn't work nearly as well in presidential election years, for example. The Democrats got a sliver more of the actual vote in 1996 than the Republicans, for example, but the GOP kept control of the House. Many other factors can sway the returns. For one thing, turnout drops spectacularly in off-year elections. And percentages can be skewed by wide variations in the number of uncontested races. After declining since the 1960s, there are more than 80 uncontested races this year.
All the criticisms and potential problems are valid, yet the generic ballot has prevailed-so far. 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 predictive power as the nation, the electorate, and politics changed in the 1980s.
So what does the generic ballot say? In a late August ABC News/Washington Post poll, it was Democrats 48 percent, Republicans 45 percent, among registered voters. Not enough of a margin to give the nod to the Democrats. And plenty of time for things to change.