The Debate Pulse:
Rapid-Response, Interactive Polling


by Mark Thalhimer, with assistance from Josef Federman

During the first and last presidential [1992] debates, CNN added a new dimension to campaign coverage by monitoring the political pulse of 480 registered voters via instant interactive polling.

Using a push-button phone, the 480 randomly selected survey respondents were instructed to express their immediate like or dislike for what each of the presidential candidates said throughout the first debate, held on Oct. 11, and the last debate, held on Oct. 19.

Respondents registered their opinions by calling a toll-free 800 number and punching any number from one through nine on their telephone keypads-creating a scale that ranged from a highly negative to a highly positive reaction. Responses were collected in Omaha, Neb., by Call Interactive and fed to Decision Labs, a company based in Chapel Hill, N.C., that specializes in real-time response polling, who were set up for the debates in CNN's Atlanta newsroom.

The voters were classified as Clinton, Bush or Perot supporters or undecided. Their responses were plotted on a graph similar in appearance to a biofeedback chart on which the needle continuously moved as people registered their opinions to the candidates' words.

"This polling method provides the ability to understand much more clearly the way specific pieces of a speech affect people," said Jack Ludwig, vice president and chief methodologist of the Gallup Organization, which cosponsored the project, along with CNN, through the aid of a Markle Foundation grant.

The rapid-response survey found that voters liked Clinton's middle-class tax cut but were not so enthusiastic about his handling of the Arkansas budget. They supported Bush's proposal for allocating 10 percent of income tax revenues toward reducing the budget deficit, but gave mixed reviews for his attacks on Clinton's character. Nearly all voters reacted positively to Perot's performances (though the number of Perot supporters in the sample was so small that the results were not reported ultimately).

The pollsters instantly converted the polled data into a graph, with each candidate's group of supporters and the group of undecided respondents represented by a different colored line. They then superimposed the graph onto a videotape of the candidates being made during the debate so that, after each debate, viewers could see the instant reactions of respondents that occurred at different points during the program.

During the CNN roundup a few minutes after each debate, public opinion analyst William Schneider interpreted several graphs measured at different points during the program. As Schneider analyzed the survey, the network broadcast a tape of the candidate speaking on one side of the screen while the graph of responses to what was being said played on the other side.

The debate reaction poll was the first time a presidential debate had been tracked using this instant-response, interactive method. While television networks, advertisers and political candidates have used these interactive survey methods in controlled laboratory settings to determine audience response to their programming in the past, those scientific methods have not been applied on a second-by-second basis to viewers across the country before.

In previous real-time television viewer surveys, respondents have participated voluntarily, either by calling a toll-free 800 or a toll 900 number. Such call-in surveys are unscientific, and therefore unreliable, because they do not use randomly selected samples. But pollsters for CNN and Gallup did a good job in selecting a random sample of registered voters for their survey, said Ludwig, because it reflected the demographics of other polls conducted at the time.

However, Ludwig cautioned against "projecting too far" with the polling results. Due to the nature of the poll, the sample was prone to a number of flaws. Its small sample size, for instance, which was reduced even further by dividing respondents into various groups, increased the margin of error well above an acceptable rate for a normal survey. Also a problem was the requirement that respondents have a push-button telephone located near a television, which may have excluded certain demographic groups.

There are also questions about what the poll actually measured. Survey instructions asked participants to react "positively or negatively to what you are hearing and seeing during the debate." But were voters deciding on what they saw or what they heard? Were they reacting to platform positions, rhetorical skill or the candidates' taste in ties? The method, Ludwig concluded, "doesn't make it clear what they are reacting to, and it's impossible to tease that out."

Despite the survey's shortcomings, voter preferences did seem to correspond with those in larger, more conventional surveys, Ludwig said. "The movement of the lines does appear to make sense."

According to Ludwig, the positive feedback to the survey indicates that interactive polling could become a permanent fixture in future political debates. "It's an aid that we don't have otherwise for looking at political messages," he said.