* This report has been supported by the Pew Charitable Trusts, Russell Sage Foundation and an "Investigator Award" from the Robert Wood Johnson Foundation. This continuing research is being conducted at the University of Minnesota, and at the Barnard-Columbia Center for Urban Policy and Paul F. Lazarsfeld Center for the Social Sciences at Columbia University. We greatfully acknowledge the research assistance of John Bies, Eric Lawrence, Michael Zis, and Eric Ostermeier. The analyses are solely the authors'.

High on the list of problems often touted as afflicting contemporary American politics is the charge that policy makers "pander" to public opinion by formulating significant policies to mirror polls. The presumption is that politicians slavishly follow public opinion; the cost has been an abdication of leadership and of the will to use personal judgement to promote the national interest. President Clinton's first term is often cited as prime evidence of pandering -- and Dick Morris' recent book on his stint as the President's pollster has only inflamed the charges. As Maureen Dowd recently wrote in the New York Times, Clinton's mind meld with Morris "demonstrates that polling has turned leaders into followers."

Despite the nearly universal acceptance of the notion that today's politicians pander, there is no systematic evidence to support it. We set out to investigate the notion of pandering by interviewing over 100 members of the executive and legislative branches during Bill Clinton's first term in office, and by examining evidence from public opinion polls and government decisions on significant policy issues since the mid-1930s. In the interest of full disclosure, we should emphasize that our past research supported the pandering charge by revealing that strong and stable public preferences exerted a strong influence on government policy.

Our research team interviewed 114 people who advised or worked in the executive and legislative branches from 1993 to 1996. We interviewed 23 individuals associated with the Clinton administration and 91 senior legislative staff to members of Congress. The interviews were based on a preset questionnaire and were conducted on a confidential basis to encourage candor. (Many of the interviews were characterized by frank discussion of personal and professional matters that were far more politically damaging than the subject of public opinion.)

Interviews with individuals tied to the Clinton administration suggested that public opinion did not generally drive the formulation of specific social policies. We focussed particular attention on the President's health reform effort and were consistently told that public opinion was not a decisive influence on the formulation of policy. For instance, both interviews and examination of confidential White House records indicate that Stanley Greenberg, the President's primary pollster during 1993 and 1994, was locked out of the process of formulating health care policy.

When we pressed for explanations why the administration adopted a modified version of managed competition or made other important health policy decisions, we were repeatedly pointed to the President's personal convictions and those of his senior advisers. As Stan Greenberg told to us, "our problem is not being political enough. The general impression that the President is too political in fact, is just dead wrong. The opposite's the case. His tendency is to develop policy removed from politics."

We also interviewed a random sample of senior congressional staff that reflected the partisan balance in Congress and the larger size of the House of Representatives. (The number of interviews reported below is at times less than 91 because time constraints prevented some staff from finishing the interview.)

Members of Congress are clearly not addicted to polls, based on our interviews. It is quite the opposite. A consistent theme in our interviews was that members conducted very little polling, and the surveys they conducted were largely limited to district or state election campaigns. Instead, individual members relied on traditional contacts with their constituents -- face to face meetings, mail, and phone calls. When given the opportunity to identify all the ways they monitor public opinion, 21 of 91 staffers (23%) reported conducting private polls; 79 of 91 (87%) mentioned that their offices monitored public opinion by relying on letters or phone calls; 51 (56%) pointed to personal encounters. Fifty-six percent (44 of 79) reported that their offices commissioned no surveys and, not surprisingly, nearly three-quarters (59 of 80) indicated that no one in their office had special responsibility for analyzing them. One reason that members do not rely on opinion surveys is that they do not trust polls. A Republican staffer expressed a pervasive judgment: "we discount any polling -- it all depends on how questions are asked, who is asked. Who designs [a poll] can make it say what they want. We trust direct contact with constituents instead."

Not only do members not rely on polls but they apparently consider public opinion an inappropriate consideration in formulating policy. When we pushed the staff we interviewed on the usefulness of public opinion information in policy making, a common response was that public opinion was not directly considered. For instance, when staffers were asked whether existing opinion today or anticipated future opinion influenced decisions, three quarters (18 of 24) spontaneously challenged the premise of the question and volunteered that public opinion was of no use. In a separate question, a quarter (12 of 44) responded to a question about how public opinion weighed in the member's decisions about specific policies like health care or welfare reform by again challenging the question's implicit assumption and insisting that their member's decision were not guided by public opinion.

The clear message was that the policy positions of members were guided by their personal beliefs, ideology, and judgments. Staffers regularly emphasized their members' determination to "do what's best" according their personal values. As one Republican explained: "On policy, beliefs are more important than public opinion." Another explained that the member "just does what he feels he needs to do. Public opinion is not at all useful in day to day policy making." Nine out of 10 (74 out of 82) acknowledged that public opinion information was used to lobby their offices but argued that it had no influence on the member; their member "stuck" to his or her beliefs and distrusted the results because the "numbers are so easy to manipulate" to serve the interests of the lobbyist -- whether it was an interest group, another member or the White House.

There is a danger with relying on interviews: respondents may simply offer the expected answers to politically loaded questions on whether members "pandered." Several bodies of research, however, corroborate the central tendency in the interviews to discount the impact of public opinion on policy decisions.

Our interviews with individual congressional offices are consistent with a large body of quantitative research on the relationship between individual members of Congress and constituency preferences in the members' districts. The findings of these studies indicate that public opinion has a measurable but modest and highly contingent impact on members' voting decisions and electoral prospects. Part of the explanation for the muted impact of constituents is that their policy preferences have but a small influence on congressional elections; voters give substantial weight to candidates' personal characteristics and party in their voting decisions, and know little about (and often do not act on) members' policy positions or records. Evidence, then, that constituency opinion is not a decisive influence on individual members substantiates the downplaying of public opinion by congressional staff.

Another body of research compares overall government policy with aggregate public opinion over time. Alan Monroe of Illinois State University has studied the consistency of government policies with prior majority support by Americans for policy changes between 1960 and 1991. He asks: Does the government adopt policies supported by majorities in polls? These data indicate that opinion-policy agreement declined overall from 63% consistency in the 1960-80 period to 55% in the 1981-91 period; consistency slipped in the area of social welfare policy from 63% to 51%. While this is a crude test, it is consistent with a decline in responsiveness to public opinion (not an increase as the pandering myth supposes).

Monroe's findings are confirmed by an alternative research method that compared cases of changes in opinion with changes in policy that occurred, or failed to occur, after opinion changed. This approach asks: Were changes in public opinion (as recorded in identically worded polling items) followed by changes in policy? This research found that opinion-policy congruence declined from 67 percent during the 1935-45 period (heavily dominated by wartime issues) to 54 percent during the 1960s, and then increased to 75 percent during the 1970s.

Using this approach, we conducted preliminary research that focussed on four social policies: welfare, crime, social security, and health care. We compared a substantial amount of public opinion data from 1984 to 1994 with available indicators of government policies from 1985 to 1995, dividing our analysis into three time periods that corresponded with the three most recent presidential administrations.

Our preliminary results, which we will be further examining in future research, reveal that policy became increasingly incongruent with previous changes in public opinion. For the 18 cases of opinion change in the 1984-1987 period, 12 cases were congruent (67 percent). In particular, government spending increased in policy areas of growing public support. During the period of the Bush administration, congruence dropped to 40 percent (14 of 35). The drop in responsiveness reflects the lack of policy change in health care and crime, as well as the lack of clear policy measures for some issues (excluding the latter cases produces 43 percent congruence).

Congruence remained low during the Clinton 1992-94 period at 36 percent (14 of 38 cases). What is particularly striking is that the Clinton period's responsiveness could be interpreted as considerably lower: six of our 14 cases of congruence involved health care reform where declining public support for the Clinton reform plan coincided with Congress' subsequent defeat of the plan. In other words, the public's prior expectations and support for acceptable health reform was thwarted. Oppositional elites may have simply responded to public opinion that they helped to create; the health reform case may strain the notion of democratic responsiveness to autonomous public preferences. If we put aside the Clinton health reform plan and its components, opinion-policy congruence drops to 26 percent (8 of 31 cases). This very low frequency of responsiveness also reflects 7 "mixed" cases in which increases in welfare case loads occurred with declines in benefit levels, making opinion-policy congruence judgments ambiguous.

The evidence overall points to a persistent pattern since 1980: a generally low and at times declining level of responsiveness especially during the first two years of the Clinton presidency.

Although policy responsiveness has declined over time, electoral competition continues to create selective periods when officeholders become particularly attuned to public opinion. Just weeks before the 1996 elections a flurry of legislation was signed into law that was clearly supported by the public: an increase in the minimum wage, environmental protection, the provisions of Kennedy-Kassebaum health care reform, and, most important, welfare reform.

In short, systematic comparisons of government policy with polls provides no evidence that policy makers during the first term of the Clinton presidency were persistently preoccupied with pandering. If anything, the evidence suggests that government responsiveness to public opinion has been on the downswing.

We face a puzzle: if public opinion does not normally appear to affect policy decisions, why do government leaders stock a veritable warehouse of information on public opinion? According to interviews and other evidence, the explanation for the apparent disregarding of public opinion amidst studious monitoring is that the Clinton administration and congressional leaders used information about public opinion to craft their political rhetoric. Polls and focus groups are used to identify the language and arguments that presented already decided policies in the most appealing manner.

Administration officials repeatedly told us that the primary purpose of tracking public opinion was not to "pander" but to educate, lead or otherwise influence public attitudes toward the President and his policies. It is telling that Greenberg's services on behalf of the administration's health reform initiative were most heavily used after the policy was largely set and it was time to sell the plan. Leon Panetta, Clinton's former Clinton Chief of Staff, boldly stated in a recent New York Times, "the lesson of the last two years is that the best path to the center is through the President's bully pulpit.... [When they are motivated,] the American people can drive Congress to the center."

Our congressional interviews produced a persistent theme: polls were "useful in terms of framing what you were going to do anyway." Six out of ten congressional staff (53 of 85) indicated that public opinion information had been used to frame already decided policy decisions. Twice as many of those we interviewed indicated that public opinion information was useful for leading or educating Americans (47%, 41 of 88) than for responding to them (25%, 22 of 88). (Of 88 staffers, 15 argued that both were important and 10 others pointed to other uses of public opinion.) Contrary to the notion that pandering politicians obsequiously follow public opinion, congressional officials repeatedly emphasized that polls and focus groups were a "tool to shape our message," "educate the citizens" and to "learn what messages we need to get across." As one Republican staffer commented, opinion information is "not the source of decisions on policy but rather is helpful in showing the kinds of things [a member] is not communicating... and therefore the need to try to educate further."

The notion that today's politician panders to polls when making significant policy is a myth built on scattered anecdotes and ungrounded presumptions. This myth neither enjoys empirical support nor conforms with the current incentive structure in American politics. The comparisons of polls with government policy over time indicate that policy makers' responsiveness to public opinion has not increased over time; if anything, it has declined since 1980.

The dropoff in responsiveness since 1980 is consistent with a trend toward political polarization. For much of the past decade and a half, neither political party has controlled both the White House and Congress. The divisions created by split party control have been intensified by increased ideological divergence especially in Congress; the Democratic Party's loss of its southern conservative base to Republicans has made each political party more homogenous. The result is that political leaders are encouraged to rely on directing public opinion to counteract the political leverage of their opponents and to build support for their own ideological or policy goals.

The charge of pandering politicians is dangerous: it leads to the erroneous conclusion that the public is to blame for the ills that afflict Americas politics. More leeway, it may be implied, ought to be given to the few who have superior training and knowledge. The evidence suggests, however, that the current state of American politics cannot be blamed on excessive responsiveness to public opinion and insufficient leeway for elites. Political leaders already exercise significant discretion. If anything, they should pay more attention to public concerns and preferences.

1. This "change-oriented" or "covariation" method avoids sensitivity to question wording because identically-worded survey items are compared in determining changes in opinion. The analysis focussed on cases in which there were opinion changes of at least 6 percentage points (well beyond the point of random sampling error). Policy is measured by the best available and acceptable indicator for the substantive issue raised by the polling question.
   Comparisons of the measures of opinion and policy change are generally classified according to whether they are "congruent" (i.e. occur in the same direction), "noncongruent," or "uncertain" when an acceptable policy measure was unavailable. When two equally good policy measures were considered, the results were classified as "mixed" if the policy measures moved in opposite directions -- one congruent and the other noncongruent with public opinion. For instance, measures of policy spending per beneficiary were sometimes used along with the number of beneficiaries.
2. Our public opinion data were collected from the Roper Center POLL database as well as on recent "Poll Trends" articles in Public Opinion Quarterly on the issues of welfare, crime, social security, and health care. In addition, most of the measures of government policy involved constant dollar changes in spending (deflated by the CPI or, in the case of health care, the CPI for medical care); spending per benefit recipient; changes in numbers of beneficiaries; or the passage of a law or bill.
3. We investigate the connection between policy making and public opinion in "The Politicization of Public Opinion: The Fight for the Pulpit," in The Politics of Social Policy, ed. by Margaret Weir (Washington, D.C.: The Brookings Institution, forthcoming); we are continuing our investigation in our project, "Leadership and Citizenship," which is supported by the Pew Charitable Trusts.