No Exit From Exit Polls

Chatterbox is a turnout freak. He believes that the best barometer about the health of American democracy is the percentage of eligible voters who actually go to the polls. And if you were following the turnout numbers on election night and immediately thereafter, you came away with the reassuring news that 1998 was a pretty typical mid-term election.

On ABC’s prime-time election-night special, Peter Jennings intoned, “I was just looking at turnout figures. Our estimate is now that about 38 percent of eligible voters turned out. And as I look back at my graph here, that’s about the same as 1994, higher than 1990, higher than 1930.” That same 38-percent figure cropped up in Wednesday morning’s Washington Post. Dan Balz, one of the Chatterbox’s favorite political reporters, wrote, “Preliminary results by the Voter News Service indicated a 38 percent voter turnout this year, falling just slightly below the 38.8 percent turnout of the last midterm election in 1994. Since 1970, the voter turnout in midterm elections has fluctuated between about 37 percent and 40 percent.” And in Thursday’s New York Times, columnist William Safire waxed Panglossian over this turnout figure: “Despite hand wringing predictions of rampant apathy and turnoff, midterm turnout was 38 percent of the voting-age public. That is as high as the controversial Agnew-driven campaign of 1970.”

There was just one problem with this soothing authoritative analysis–it was completely wrong. The deceptive 38-percent figure was derived, as Balz reported, from the Voter News Service (VNS), the fancy name for the network-and-wire-service consortium that provides all the exit-poll results.

It wasn’t until Thursday that Curtis Gans, who heads the Committee for the Study of the American Electorate, released his own calculations based on actual vote totals rather than exit-poll data. And guess what? Turnout was actually 36.1 percent, the lowest figure for a mid-term election in 56 years. That’s a dramatic drop from 1970–the fire-and-brimstone year that Safire cited–when 46.8 percent of eligible voters went to the polls.

Why was everyone so off base? “The VNS doesn’t know how to analyze turnout,” says Gans. “The VNS has gotten it wrong almost every time.” Turnout was not the only slip-up by the exit-poll people. As Richard Morin reports in Friday’s Washington Post, the exit-poll numbers pointed to a dramatic upsurge in voting by union families, a statistic that had the AFL-CIO breaking out American-made champagne. But it turns out that the surge in union voters may have been largely due to a change in the wording of an exit-poll question. According to a memo by VNS editorial director Murray Edelman, “Much is being made of the increase in union votes from 1994 to 1998. This appears to be an artifact of our change in the form of the question.”

The root of the problem is that journalists and political professionals have become so addicted to exit polls that they have lost other traditional methods of post-election analysis. In the old days, demographers like Richard Scammon could quickly glean trends from raw precinct-by-precinct vote totals. But that useful skill is now about as common as monks who can produce illuminated manuscripts. The sad result is that the political community is reduced to bowing before the exit-poll numbers as if they were the holy writ. But exit polls were designed to do just one thing–to quickly predict who will win individual elections. On that score, the record of VNS is exemplary. But the insatiable demands for instant information have forced the exit pollers to provide answers that go beyond the built-in limitations of the surveys. When the election-day weather is rainy or cold, for example, many voters, particularly the elderly, don’t stick around long enough to finish the entire exit-poll questionnaire. Such partial results can skew the results, especially about secondary matters like the demographic makeup of the electorate.

All this should serve as reminder that the only thing worse than no data is wrong data. And despite this stirring sermon, Chatterbox knows that justified journalistic skepticism about exit polls will again be completely forgotten on election night 2000.

Walter Shapiro