By Richard Morin
By Richard Morin
It was the union vote that wasn't. Then again, maybe it was. But however many union voters did or did not turn out on Election Day, news organizations continue to report that union voting surged even as researchers continued to say that it probably didn't.
The story broke on election night, when exit polls conducted by Voter News Service (VNS) revealed that slightly more than one in five voters 22 percent was a member of a union household. Four years earlier, only 14 percent of all voters said they lived in a household that contained someone who belonged to a labor union.
That apparent surge was news: CBS featured the apparent increase in the union vote in its election-night broadcast. The Washington Post, The New York Times, the Associated Press and other news organizations also noted it in subsequent election stories and analyses.
One problem: VNS reported two days after the election that there probably wasn't an increase in the union vote at least not as large as the one suggested by their 1994 and 1998 exit polls. If anything, 1998 exit poll results suggest that fewer members of union households may have voted this year than in the last two national elections.
"Much is being made of the increase in union vote from 1994 to 1998," wrote VNS editorial director Murray Edelman to the VNS survey committee, which is composed of the major television networks and the Associated Press. "This appears to be an artifact of our change in the form of the question."
Edelman more succinctly summarized his view in the title of his memo: "The Big Labor Vote NOT" (Despite his Nov. 5 statement, a search of the Lexis-Nexis media database revealed dozens and dozens of instances since his memo was written and since a Washington Post story first disclosed the problem where newspapers and magazines cited national exit poll data as evidence of a big increase in union voting.)
Why the increase that apparently wasn't? Here's what VNS says happened. In 1994, union voters were determined from what VNS called the "grab-bag" question, which asked people to check all the categories that applied to them. The list included checkoffs for member of a union household, "gay, lesbian or bisexual," and whether the voter owned a gun, as well as other categories.
The grab-bag question appeared at the end of the back side of a two-sided questionnaire. Edelman says that a VNS review found that large numbers of exit poll respondents skipped over or otherwise refused to answer the grab-bag questions. One problem, he says, was that the percentage of gays, gun owners and union-household voters were determined on the basis of the total number of people who took the survey.
Since that number likely included people who skipped or refused to answer the grab-bag question and some of these were union members the exit poll underestimated the percentage of union voters in 1994, VNS researchers suspect. The language and the format of the question also may have been a problem, he says.
For these reasons, the so-called grab-bag format was dropped in 1996 in favor of a straightforward yes/no question that asked "Do you or does someone in your household belong to a labor union?" This is the format of the question that was used this year.
VNS instructions to subscribing news organizations did not note the problem with the 1994 question or caution against making comparisons between 1994 and 1998 results. (VNS did provide comparison data for 1996 for many of the exit poll questions. But many news organizations chose to ignore the 1996 exit poll results. That's because 1996 was a high-turnout presidential year; better, these analysts reasoned and I was one of these analysts to compare this midterm with the last midterm election in 1994.)
It's not the first time that the abandoned grab-bag question has risen from the dead to produce an apparent false-positive reading. Four years ago, 2 percent of voters checked the grab-bag category to indicate they were gay, lesbian or bisexual.
But in the past two national elections, a separate question was used and it estimated the gay/lesbian/bisexual vote at 5 percent in 1996 and 4 percent this year. That led the leaders of some homosexual rights groups to claim falsely, Edelman believes "a big increase in gay voting."
The strongest evidence indicating that the way the question was worded, not changes in voting patterns, is responsible for at least some of the apparent increase in union household voting comes from an experiment that VNS included in the 1994 exit poll.
Four years ago, the grab-bag question included "Religious Right" as a category and 6 percent of voters interviewed checked it off. But in another version of the 1994 survey, they asked a full question to establish whether or not the voters considered themselves members of the Religious Right, and 19 percent said they did strong evidence of a "grab-bag" effect that led to substantial underestimates of voting by some subgroups.
Overall, it appears that the actual number of union household voters, gays and members of the Religious Right was actually three times higher than the grab-bag question suggested. That suggests that a grab-bag union question this year would have indicated that about 7 or 8 percent of all voters were members of union households a decline from the past two national elections.
"If one were to argue from extrapolation, it would be that union voting has declined," Edelman said in his memo. "However, given the imperfection of this measurement, it would be far better to just not make any comparison."
Still, there's some evidence that union turnout may have been up from 1994. The 22 percent union household vote reported this year equals union turnout in the 1996 presidential election. Minorities and low-income voters are more likely to turn out in presidential election years as opposed to midterm elections like 1998. Some suspect union voters do, too. The fact that union voting this year nearly equaled turnout in 1996 would suggest a larger than normal union vote.
"There might be an argument there" that the union vote was up from 1994, Edelman says. "But I'd have to have comparable data from other midterms."
© Copyright 1998 The Washington Post Company