Larry Jacobs at Polls Conference

Do polls accurately measure public opinion? Who drives policymakers' decisions - the public, policy elites, or special interest groups? How do policy elites use poll data on public opinion?

Eighty academics, pollsters, policymakers, journalists, and students gathered at Northwestern on May 13 to address these and other questions at an IPR conference on "Polls, Policy and the Future of American Democracy." Three sets of panelists offered original research and commentary on the importance of public opinion in shaping both domestic and foreign policy and whether politicians "pander" to public views. A fourth panel proposed new methodology to more accurately measure the public pulse.

Impact on Policy. Much of the conference centered around the extent to which public opinion affects policy. The "centerpiece" of this debate, according to Northwestern political scientist and conference rapporteur Benjamin Page, was a paper by Robert Erikson, Michael MacKuen, and James Stimson that proposed a "Model of American Politics."

Using an elaborate time series analysis with 17 equations and 48 variables to measure change in how liberal or conservative the public was feeling in any given year, their model offered empirical evidence that public opinion ultimately influences policy. "Behind all this complexity" said Page, "it all comes down to the simple proposition that the public runs the show. This model is rather breathtaking in its simplicity."

William Domhoff of the University of California at Santa Cruz disagreed. He argued that corporate elites, through policy planning and discussion networks, shape public opinion by funding think tanks, sponsoring studies, speaking on television, etc. They also influence policy activity (i.e., voting on bills) directly and indirectly, through public opinion and election outcomes.

Larry Jacobs of the University of Minnesota and Robert Shapiro of Columbia University think politicians often manipulate public opinion for their own purposes. In a paper based on their book, Politicians Don't Pander, they maintain that policy decisions of presidents and members of Congress have become less responsive to the substantive policy preferences of the average American since the 1970s.

They contend that politicians use public opinion research to find the "language, arguments, and symbols that would win public support and unify Washington elites² in support of their policy goals. This means they use polls not to figure out what the public wants and how to respond, but to figure out what they can sell to the public and how to package the policies they want to sell.

Domestic Policy. Politicians are paying a lot of attention to polls, reported IPR director Fay Lomax Cook, though they may not be interpreting them accurately. In a study with Page and IPR graduate fellow Jason Barabas, she analyzed how policy elites have invoked public opinion about Social Security over the past decade and the extent to which their claims are accurately based on evidence from the polls.

They found the preponderance of these claims had to do with a lack of confidence about the future of Social Security. Though evidence confirms that confidence is low, this is not a new development as the claims would imply‹it has been low since the late '70s. It is unclear whether the cause is lack of knowledge about the workings of Social Security or about what will happen when the program can no longer pay full benefits, Cook said. Also differing from the claims, the data suggest that when respondents have a chance to consider the risk of privatizing Social Security, they become less certain they prefer this direction.

Paul Burstein of the University of Washington reported that public opinion strongly influenced congressional action on equal employment opportunity until 1972, but that its effects since then, while still strong, have been complicated by other influences. These include feedback from other branches of government, Supreme Court decisions, and partisan politics.

Foreign Policy. Shapiro and Jacobs maintain that U.S. foreign policy has become less responsive to public opinion since the end of the Cold War. Not only is there less media coverage‹a reality confirmed by Chicago Tribune commentator Dick Longworth‹but sophisticated White House polling techniques continue to be used to manipulate and change public opinion. "Elites are driving public attitudes" Jacobs said. "They want to shrink critical debate among citizens, to discursively vaporize them."

A comparative study by Steven Kull and Clay Ramsay of the University of Maryland concluded that policymakers incorrectly perceived the public as supporting isolationist policies. To explain this, they speculate that policy practitioners may be failing to seek information about public attitudes, responding to the vocal public as if it were the majority, assuming that Congress and the media reflect the public will, or else just underestimating the public.

Measuring Public Opinion. Martin Gilens of Yale University described a range of innovative survey experiments, suggesting that question wording and context and interviewer relationships all can be experimentally manipulated so different kinds of analyses may be used to extract meaning from randomized survey experiments. He also showed how different modes of data collection (by mail, in person, by telephone) reflect significant differences in response to questions on sensitive issues.

Northwestern economist Charles Manski proposed probabilistic polling methods as a more accurate way of determining the outcomes of elections. In contrast to outcome polls typically used by Gallup and Pew, which use a subjective question design (respondents are asked to gauge their answers based on the likelihood of an outcome), probabilistic polling asks respondents to give the percent chance they will act in a certain way (i.e., choose to vote; vote for candidate A or candidate B). This method offers a greater range of possible responses.

"The probability scale enables respondents to fully express uncertainty about their future voting behavior and enables analysts to aggregate voting expectations into predictions of election outcomes" Manski said.

In a concluding address, Humphrey Taylor, chairman of the Harris Poll, returned to a major theme of the conference by citing the critical role of polls in democratic elections. "Probably the most important contribution of pre-election and exit polls, particularly in new democracies, is that it makes it harder for governments to steal elections," he said.

The conference was sponsored by IPR with support from Northwestern's School of Speech, American Studies program, and departments of sociology, political science, and communication studies. The agenda and papers may be found online at: www. northwestern.edu/IPR/events.