Monica Lewinsky's Contribution to Political Science

by JOHN R. ZALLER

PS: Political Science & Politics, June 1998, volume 31, no. 2, pages 182-189

The bounce in President Clinton's job ratings that occurred in the initial 10 days of the Lewinsky imbroglio may offer as much insight into the dynamics of public opinion as any single event in recent memory. What it shows is not just the power of a booming economy to buttress presidential popularity. It shows, more generally, the importance of political substance, as against media hype, in American politics. Even when, as occurred in this case, public opinion is initially responsive to media reports of scandal, the public's concern with actual political achievement reasserts itself. This lesson, which was not nearly so clear before the Lewinsky matter as it is now, not only deepens our understanding of American politics. It also tends, as I argue in the second half of this article, to undermine the importance of one large branch of public opinion research, buttress the importance of another, and point toward some new research questions.

Whatever else may have transpired by the time this article gets into print, the Lewinsky poll bounce is something worth pondering. In a half-dozen commercial polls taken in the period just before the story broke, Clinton's job approval rating averaged about 60%. Ten days later, following intensive coverage of the story and Clinton's State of the Union address, presidential support was about 10 percentage points higher.(1) The fact that no analyst of public opinion could have credibly predicted this outcome makes the poll bounce especially important to examine. It is, in statistical parlance, a high leverage case.

I begin my analysis with an attempt to establish the parameters of the initial public response to the Lewinsky matter. Toward this end, the results of some three dozen commercial polls, gleaned from published sources, are summarized in Table 1. Although question wordings differ somewhat, all poll results refer to approval of Clinton's job performance as president. Also reported in Table 1 are the results of a content analysis of network TV news coverage during this period.

The content analysis, as shown in the top three rows of the table, gives average minutes of each network news program that were favorable or unfavorable to Clinton. Favorable references include Clinton's denials, attacks on Independent Prosecutor Ken Starr, statements of support for Clinton, and any other information (including non-scandal information) that might tend to enhance public support for the president. Unfavorable references include all statements indicating that the president had an affair with Lewinsky or tried to cover it up, attacks on Clinton or [TABULAR DATA FOR TABLE 1 OMITTED] defense of Starr, and any other information that might tend to undermine public support for Clinton. I emphasize that, although journalists played a major role in creating the Lewinsky imbroglio, other actors, notably politicians, initiated some of the information that was reported.

What the content analysis shows is that the frenzy began with two days of heavily negative coverage, but that coverage was relatively balanced after that (given that the matter continued to attract media attention at all). In fact, if the first two days are removed, the remaining period has about as many positive minutes as negative ones, including two days on which Clinton's coverage was decidedly positive.

I have divided the poll data into four partially overlapping periods. As the table indicates, the first two days of heavily negative scandal reportage had a considerable impact on public opinion. On the basis of a half-dozen polls, Clinton's public support seems to have dropped about six or seven points.(2)

The scandal broke on a Wednesday, with the most heavily negative coverage on that day and Thursday. From Friday on, coverage was more balanced and public support for the president rose. By Monday, Clinton had regained everything lost in the first two days, and in Tuesday's Gallup poll, support for the President rose above pre-Lewinsky levels. There were two notable events in this period, both of which were amply reported on TV news. The first was Clinton's appearance on camera on Monday to make an emotional denial of a sexual relationship with Lewinsky; the other was Hillary Clinton's appearance on NBC's Today Show on Tuesday morning, where she charged the existence of a right-wing conspiracy against her husband.

If there is any particular spike in the data, it is the Tuesday Gallup poll, which was taken between six and nine in the evening and was therefore able to reflect news of Hillary Clinton's appearance on the Today Show that morning. Indeed, the poll was taken just as or just after many Americans were getting news of Mrs. Clinton's appearance; it may therefore, as other polls hint, have overstated its lasting importance on opinion. This poll showed a gain of eight percentage points from the day before, a difference that is statistically significant on a two-tailed test.(3)

Clinton's State of the Union address occurred on Tuesday evening, the end of the seventh day since the Lewinsky story broke. The speech attracted an unusually large audience, presumably because people wanted to see how the crisis-stricken President would perform. According to virtually all the pundits, he performed extremely well. "Good speech, too bad," as one commentator put it.(4)

Two national surveys were taken immediately after the speech. From baselines on the day before the speech - and therefore before Hillary Clinton's charge of right-wing conspiracy - one survey showed no change and the other showed a gain of five points, for an average gain of 2.5%.(5) There was also a CBS poll involving reinterviews with a panel of respondents who had been asked by telephone to watch the speech so that they could be polled afterward. This survey found that Clinton's post-speech job approval rating was 73%. No immediate prespeech baseline for this poll is available, but if we take the best baseline we have - Clinton's 57% job approval in the CBS-New York Times poll from the day before the speech - then the combination of the speech and Mrs. Clinton's defense netted the president some 16 percentage points in support.

A little back-of-the-envelope arithmetic shows that these two sets of post-speech results - an average 2.5% gain in two polls and a 16-point gain among those asked to watch the speech - are not as far apart as they might seem. According to the Nielsen research firm, 53.1 million Americans saw the speech ("TV Ratings for Speech," 1998). This is a lot of people, but only about 25% of the adult population. If 16% of the 25% who watched the speech became more supportive of the president, the overall increase in public support would be only 4.0 percentage points (.16 x .25 = .04). If we assume that viewership of the speech was higher than 25% among those asked to watch it in preparation for a survey but still well under 100%, there is no real disagreement among the three polls on the size of the "speech plus Hillary" effect.

From the bottom panel of Table 1, it appears that public support for Clinton rose another three or so points after the State of the Union, perhaps in response to favorable news coverage of that event. But this gain, if real, is apparently small in relation to gains that had already occurred.

It is tempting to pursue more detailed analyses of particular events, but I have already pressed dangerously close to the limits of the data. Instead, I will step back and offer a somewhat less detailed and, I therefore hope, safer summary: In response to sharply negative media coverage of the Lewinsky matter, public support for the president fell. But support rebounded and then surpassed its initial level as the president, his wife, and their allies fought back.

One point seems especially clear and important: In the period in which Clinton's support fell about 7 percentage points, media coverage was sharply negative, but in the period in which he gained back those 7 points and added an additional 8 to 10 points of support, coverage was essentially balanced. Thus, while media coverage of the Lewinsky matter explains part of the opinion change that occurred, it cannot explain all of it. In particular, the notion that the public responded mechanically to media coverage cannot explain how Clinton ended up with higher job approval ratings than he began with. Additional explanation is needed.

An obvious possibility is to argue that the public makes a distinction between approving the way the president does his job and approving of the president as a person. There is, as it happens, some evidence for this view, but not a great deal. The president's personal favorability ratings fell more sharply than his job approval ratings and also recovered less well. In three NBC News polls, Clinton's favorability ratings were 57% before the Lewinsky matter broke, 40% after three days of scandal coverage, and 50% after the State of the Union. In what is apparently the only other set of surveys that made three such soundings of opinion, Time-CNN found that Clinton's favorability ratings went from 60% to 50% and then back to 60%.(6)

These data on favorability seem to me to do little to alleviate the mystery of Clinton's bounce in job approval ratings, since they show essentially the same trend. Even if we were examining the favorability data alone, we would still be hard-pressed to explain why Clinton, who looked nothing like a teflon president when he was pressing for gays in the military and health reform, stood up to the scandal coverage as well as he did. Nor could we explain why, amidst continued media attention to scandal, he actually recouped most of his initial loss.

Another argument might be that Clinton's specific defense against the allegations of sex and cover-up was simply very persuasive. But I find this hard to swallow - not because I disbelieve Clinton, but because he presented so little evidence to support his side and got so little support from witnesses that were in a position to give it. In particular, Clinton got no help from Lewinsky herself, who was semi-publicly negotiating a plea bargain with the independent prosecutor throughout this period. As I parse Clinton's defense, it has consisted of two flat assertions: "I didn't do it" and "my enemies are out to get me."

If the public believed this defense, it was because it wanted to. I suggest, therefore, that we consider the political context that presumably made the public want to believe Clinton's defense, namely, his record of achievement in office. Clinton made an excellent statement of this record in his State of the Union address. Although the address reached too few people and came too late to explain the bulk of Clinton's recovery in the polls, it is reasonable to suppose that the presidential record that the speech touted was well-known to the majority of the public.

Clinton speech was, first of all, a celebration of a list of "accomplishments" that would be any president's dream: The economy was the strongest in 25 years, the federal budget was on the verge of balance for the first time in 20 years, crime was falling for the first time in living memory, and the country was at peace. In the main section of the speech, the president proposed a series of programs designed to appeal to the ideological center, as exemplified by a plan to use surplus funds to put Social Security on a sound footing, improve public education, and build more highways. Thus, what the president trumpeted in his speech - and what he would presumably continue by remaining in office - was a record of peace, prosperity, and moderation. Or, more succinctly, it was a record of "political substance." This record was so unassailable that, to much of what the president said in the State of the Union, the Republican leadership could only offer polite applause.

Can political substance, thus defined, move public opinion? Certainly it can. Thanks to a distinguished series of studies - including Key (1966), Kramer (1971), Mueller (1971), Fiorina (1981), and Rosenstone (1983) - political scientists have been aware of the importance of "bottom line" politics for some time. Brody's (1991) work on presidential popularity, which stresses the effects of "outcomes" news coverage on approval, points in the same direction. In light of this, it seems entirely plausible to suggest that the poll bounce that Clinton got at the time of the Lewinsky matter was driven by the same thing that drives presidential election outcomes and presidential popularity in general - political substance. It was not admiration for Bill Clinton's character that first buttressed and then boosted his approval ratings. It was the public's reaction to the delivery of outcomes and policies that the public wants.

This argument is much more than a claim that "It's the economy, stupid." In fact, Clinton's economic performance has been only middling through most of his presidency. Taking the average four-year growth in Real Disposable Income (RDI) for every president elected from 1948 on, Clinton's first term economy ranks tenth of 13. If presidential terms are rank-ordered by growth in the 12-month period prior to Election Day, Clinton's first term is still a mediocre tenth of 13 since World War II. Only recently has Clinton's economic performance become as strong as he described it in his State of the Union.

If Clinton's economy cannot by itself explain why he won by nine percentage points over Bob Dole in 1996, neither can it explain trends in his approval ratings. One big but easy-to-overlook factor is peace, which is a virtual prerequisite for popular support. Popular support for Presidents Truman and Johnson was so damaged by bloody wars that, despite reasonably good economies, they chose not to run for reelection.(7) Clinton's administration has not only avoided war, it has enjoyed a very notable success in Bosnia, for which the President was nominated for a Nobel Peace Prize.

The other big and also easy-to-overlook plus for Clinton is his ideological moderation. This is a factor that scholars, with the exceptions of Rosenstone (1983) and Alesina, et al. (1993), have too often ignored. Let me first show anecdotally how moderation affected Clinton's support and then, insofar as possible, make a systematic case.

Since gays in the military and the debacle of health care reform, Clinton has hewed to centrist policies, including ones, like welfare reform (and NAFTA earlier on), that are hard for Democratic presidents to endorse. In his confrontation with the Republican Congress over balancing the budget, it was the president, rather than the Republicans, who held middle ground. And finally, after two decades of massive budget deficits, the president has, by means of an initially unpopular budget package in his first term, helped bring the centrist goal of a balanced budget within apparent grasp.

Consistent with the notion that moderation matters is this fragment of hard evidence: President Clinton's approval ratings were weaker at the midpoint of his first term, when the economy was stronger but he identified himself with noncentrist policies, than at the end of his first term, when the economy was weaker but he had remade himself as a policy moderate. Clinton's average job approval rating in Gallup polls taken in the sixth, seventh, and eighth quarters averaged 44.3% and the average percent change in RDI in these same quarters was 4.7%. In quarters fourteen through sixteen, these figures were 55.5% and 1.5%.

Systematic evidence that policy moderation affects presidential popularity is, as far as I know, non-existent. But as regards presidential elections, the evidence, though limited, is clear. The only published evidence comes from Rosenstone's Forecasting Presidential Elections (1983), which finds centrism to be a major determinant of cross-state and cross-time voting. In another cut at this problem, my research assistant rated each of the candidates in elections from 1948 to 1996 on a seven-point scale, running from liberal (+3) to conservative (-3). The ratings of each pair of candidates were then summed to produce a measure of relative distance from the center - i.e., a measure of relative extremism - such that higher scores indicated greater relative distance from the midpoint by the candidate of the incumbent party. For example, Lyndon Johnson was rated +2 in 1964 and Barry Goldwater as -3, so that Goldwater was one point further from the center than Johnson. Obviously, such ratings are subject to error and bias. But I note that they were developed in connection with another project (press bias in presidential primaries), and that they correlate highly with the ratings of Rosenstone with which they overlap. These ratings also correlate well with a new set of ideological location scores produced by Poole (forthcoming) for presidential candidates who earlier served in Congress.(8)

The results for a standard voting model are shown in Table 2 [omitted]. War is coded as "1" in 1952 and 1968 and "0" otherwise. Economic performance is measured as average percent change in RDI in the four quarters prior to the election; that is, in the 12th through 15th quarters of each term. As examination of the regression coefficients in Table 2 shows, ideological extremism rivals economic performance as a determinant of vote for the incumbent party. Being one point closer to the center on a seven-point ideology scale (as Johnson was in 1964) is worth about 3 percentage points of the vote; by way of comparison, each additional percent of growth in RDI is worth about 2.1 percentage points. Finally, war costs the incumbent party about 4.5 percentage points of the vote.(9)

From all this I conclude that peace, prosperity, and moderation very heavily influence the dynamics of presidential support, probably in matters of presidential popularity and certainly in general elections, for Clinton as well as for other presidents. What the Lewinsky bounce adds to this conclusion is confidence. Although evidence of the importance of political substance has been accumulating for some three decades, no one could have predicted that Clinton would survive the opening round of the Lewinsky affair nearly so well as he did. This is because it has never been quite so starkly clear just how relentlessly the majority of voters can stay focused on the bottom line. Nor, to my knowledge, has it ever been quite so clear that it is possible for public opinion and media opinion to go marching off in opposing directions.

To argue, as I am, that the public stays focused on a bottom line consisting of peace, prosperity, and moderation is not to say that the public is either wise or virtuous. For one thing, its sense of substance seems, in the aggregate, rather amoral - usually more like "what have you done for me lately" than "social justice." Nor is it clear that its decision criteria are very sophisticated. Suppose, for example, that the Watergate investigation of Richard Nixon had taken place in the context of Bill Clinton's booming economy rather than, as was the case, in the context of gasoline shortages and "stagflation" (the combination of high inflation and high unemployment). Would Nixon have been forced from office under these circumstances? Or, if Clinton were saddled with Nixon's economy, would Clinton be, at this point, on the verge of impeachment? These are, I believe, real questions, and the fact that they are does not speak well for the public's wisdom or virtue.

Perhaps future events will shed clearer light on these questions. From the vantage point of early April, when this essay is being finalized, I am keenly aware that issues relating to Lewinsky, Whitewater, and Paula Jones have by no means reached a conclusion. If clear evidence of sexual harassment, perjury, or obstruction of justice emerges, the public might still turn on Clinton. If so, one's judgment of public opinion would need to be more favorable: It waits for clear evidence before reaching a verdict, and it is, after all, concerned with higher values. My personal hunch, however, is that public support for Clinton will be more affected by future performance of the economy than by the clarity of the evidence concerning the charges against him.

I said in opening this article, the Lewinsky affair buttresses some work in political science and undermines the importance of other work. The tradition of studies on economic and retrospective voting, which maintains that the public responds to the substance of party performance, seems strengthened by the Lewinsky matter. On the other hand, the tradition of studies that focuses on the mass media, political psychology, and elite influence, including such diverse studies as Edelman's Symbolic Uses of Politics (1964) and my own Nature and Origins of Mass Opinion (1992), seems somewhat weaker. It is reasonable to contend that the ground has shifted beneath these two traditions in a way that scholars will need to accommodate. However poorly informed, psychologically driven, and "mass mediated" public opinion may be, it is capable of recognizing and focusing on its own conception of what matters. This is not a conclusion that comes naturally to the second tradition.

Let me amplify the nature of the aspersion I have just cast. A major development in American politics in the last 50 to 100 years has been the rise of what has been variously called The Rhetorical Presidency (Tulis 1987), the "political spectacle" (Edelman 1988) and, more simply, Media Politics. This form of politics stands in contrast to an older model of politics, Party Politics. The defining feature of what I prefer to call Media Politics is the attempt to govern on the basis of words and images that diffuse through the mass media. This communication - whether in the form of presidential speeches, press conferences, TV ads, media frenzies, spin, or ordinary news - creates a sort of virtual reality whose effects are arguably quite real and important. Typical of the attitude that prevails in this style of politics is Republican strategist Frank Luntz's assessment of the events I have just analyzed: "The problem with [the Lewinsky matter] is we are not going to learn the real impact for years. . . . It's going to leave an indelible mark on our psyche but I don't know what the mark will be. . ." (quoted in Connolly and Edsall 1998). Freely translated, what Luntz is saying is: "It may take us in the spin business a little time to figure out how to play this, but you can be sure we'll keep it alive until we come up with something that works for our side."

As a Republican strategist, Luntz has an obvious partisan interest in taking this view. But his occupational interest is equally great. He and his colleagues in both parties have an interest in "constructing" a public discourse in which events like the Lewinsky affair are important and in which political substance - in the sense of peace, prosperity, and moderation - is unimportant, except insofar as it is useful to emphasize it.

A sizeable part of political science has organized itself to study this new political style. My analysis of the Lewinsky affair, however, suggests that political science not go too far down this road, since old-fashioned political substance of the kind that party competition brings to the fore is not only thriving in the media age, but quite likely still dominant.

This is not to say that the new style of Media Politics is without importance. If only for the resources it consumes and the public attention it commands, Media Politics matters. More, perhaps, than we would like, Media Politics defines our political culture. But beyond that, the effects of Media Politics on political outcomes must be demonstrated on a case-by-case basis, because sometimes the effects are real and lasting and other times they are not.

One illuminating example of Media Politics that produced lasting effects is Gerald Ford's pardon of Richard Nixon in 1974. Coverage of the event was, of course, overwhelmingly negative. On the basis of the same coding categories as in Table 1, Ford got 11 minutes of negative coverage on the network news on the night following the pardon, as against two minutes of positive coverage. The next night, these figures were 10 and 2 minutes. Reporters were by no means the only source of the bad news. In the first two news days after the pardon, 12 Democratic members of Congress, including the House Speaker and Senate Majority Leader, were quoted on the network news attacking Ford, and within the first week the Democratic Congress passed a resolution condemning the pardon. Three Republican leaders also criticized Ford. In these circumstances, Ford's approval rating fell 17 percentage points in the first two days and about 30 points over the longer run.

The contrast with the Lewinsky case is striking. In the first two days of this case, only three Republican members of Congress, none from the leadership, were willing to be quoted on network TV news attacking Clinton - and not for want of opportunity. Reporters were scouring Capitol Hill for volunteers, but politicians (including Democratic politicians) were playing it safe. Thus, the media were forced to shoulder a much larger part of the Lewinsky story on their own. In these quite different circumstances, Clinton suffered limited short-term damage and made gains over the longer run.

It is a tempting conclusion that when the partisan opposition joins a media frenzy, the two together can move public opinion, but that the media alone cannot do it. But even if systematic research were to establish that this pattern is general, there would still be an obvious concern: Namely, that opposition politicians attack when they see an opportunity to score points and hold fire otherwise. By this account, Democratic politicians attacked Ford because they knew the attacks would play well, but Republican politicians refrained from attacking Clinton because they feared the attacks would backfire. If this argument is considered plausible, as I think it must be, it further underscores the central claim of this essay: That American politics tends to be driven more by political substance - in this case, public disapproval of the pardon of Nixon - than by the antics of Media Politics. It also points to a difficult future research problem: Sorting out whether partisan attacks and other media messages are the causes of public attitudes or their hidden (i.e., endogenous) effects. Surely, the answer is some of both.

Another media frenzy from the Ford administration is worth a brief look. When Ford stated in the second presidential debate that "there is no Soviet domination of Eastern Europe," the mass audience hardly noticed but reporters instantly saw the remark as a gaffe. Polls showed that, citizens polled immediately after the debate judged 44% to 33% that Ford had won. Once the media frenzy of this famous gaffe had run its course, however, the public's judgment was reversed: Several days after the debate, the public thought by a margin of 62-17% that Carter had won. More significantly, Ford also lost ground in straw poll surveys on how people intended to vote. But by about 10 days later, Ford's poll standing had recovered and the gaffe was left for political scientists to ponder (Chaffee and Sears 1979; Sabato 1993, 127-29).

According to a study by Daron Shaw (1995), this pattern is typical. Media frenzies over gaffes and alleged gaffes in presidential campaigns do affect public support for the candidates, but only briefly. The time it takes public opinion to bounce back may, as in the Ford example, disrupt a candidate's momentum and perhaps thereby affect the election, but the lasting direct effect of most media frenzies tends to be nil.

One way to think about this pattern is to assume that there is some "natural" level of support for candidates that is determined by political fundamentals such as the strength of the economy, the candidates' positions on issues, and other such matters. Media frenzies can briefly undermine a candidate's natural level support, but cannot permanently lower it. Thus, what happened to Clinton in the Lewinsky matter is similar to what happens to candidates who misstep in elections; he recovered from the initial attack. The fact that Clinton gained back more support than he lost is harder to explain in these terms, but I offer the following conjecture: In non-electoral periods, the public tunes out from politics, failing, inter alia, to keep its evaluation of presidential performance fully up-to-date. But when, as in the early days of the Lewinsky matter, Clinton's capacity to remain in office came into question, the public took stock and reached a conclusion that led to higher levels of overall support for the threatened leader.

These observations suggest a rough generalization about when media frenzies have lasting effects on opinion and when they don't: The closer media frenzies get to what I am calling political substance, the more likely the effects are to be lasting. The example of Ford's pardon of Nixon would seem to fit this pattern. To take one other example, it seems likely that sympathetic press coverage of attacks by racist southerners on peaceful civil rights protesters in the 1960s had an important effect on northern opinion and thereby congressional action. This was exactly what the Rev. Martin Luther King Jr. expected to happen, and it had lasting importance.

One lesson for political science from the Lewinsky poll bounce, then, is that more attention needs to be given to the general question of when Media Politics (in the sense of trying to mobilize public support through mass communication) matters and when it doesn't, and to do so in a manner that doesn't presuppose the answer. A current research project of Larry Bartels shows how this can be done: With a measure of the "real economy" from the Commerce Department and a measure of the "media economy" from content analysis of media coverage, he hopes to find out which has more influence on presidential approval. Among the auxiliary variables whose impact on presidential approval he will assess is the white-collar unemployment rate in Manhattan. The results will be interesting however they come out.

Another lesson for political science from the Lewinsky poll bounce is that the public is, within broad limits, functionally indifferent to presidential character. "Don't Ask, Don't Tell," as my colleague Art Stein summarizes the mass attitude. Given this, it seems appropriate to consider carefully whether research on the public's assessment of presidential character really helps us to understand the dynamics of American politics.

Contrary to this suggestion, it might be argued that private sexual misbehavior is different from public character, especially in light of changing sexual mores in this area, and that voters' assessments of public character will remain important. Perhaps. But if we view the character issue more broadly, it seems unlikely that voter concern about character has ever been very great. For example, Richard Nixon's peculiar shortcomings were deeply felt by a large number of voters from the moment he stepped onto the national stage in the 1940s. Further, the concerns about Nixon's public character were more serious than any that have been raised about Clinton's. Yet Nixon was elected to the presidency twice, once over Hubert Humphrey, a man whose sterling character has been almost universally acknowledged. Nixon's campaign against Humphrey was, of course, framed by urban riots and a stalemated war in Asia, and in these circumstances, Nixon chose to emphasize substance rather than character. "When you're in trouble," he told voters, "you don't turn to the men who got you in trouble to get you out of it. I say we can't be led in the '70s by the men who stumbled in the 60's."(10) Voters agreed with this emphasis, as they almost always do.

Notes

Thanks to Larry Bartels, Dick Brody, Mo Fiorina, Fred Greenstein, John Petrocik, and Daron Shaw for advice on early drafts of this paper.

1. Documentation of the sources of polls cited in this paper may be found in a PC Excel 5.0 file labeled "Lewpols" on my web page (www.sscnet.ucla.edu/polisci/faculty/zaller). The polls used in determining the overall effect of the Lewinsky matter are: ABC News-Washington Post, January 19 and 31, job approval rates of 59% and 67%; ABC News, January 13 and 30, job approval of 62% and 69%; CBS News, January 18 and February 1, 58% and 72%; Newsweek, January 18 and 30, 61% and 70%; Time-CNN, January 15 and 31, 59% and 72%; U.S. News and World Report, January 11 and February 1, 58% and 66%. In cases in which polling occurred over several days, the date given is for the final day. Although wordings of the questions differ, all refer more or less directly to Clinton's job performance rather than to the Lewinsky matter per se.

2. Gallup conducted a poll on the afternoon of the first day of the episode, prior to the evening news. This poll showed Clinton's support rising to 62% from 60% two days earlier. However, I do not count this poll on the grounds that, although the story had broken at the time of the poll, few Americans could yet have learned about it.

3. The sizes of the two surveys were 864 and 672.

4. Peter Jennings, quoting an anonymous politician.

5. The baseline for the ABC poll was actually January 25-26, with a sample of 1023. The size of the ABC postspeech survey was 528. The NBC pre- and postspeech surveys both have reported sizes of 405.

6. For sources of these and other favorability polls, see the PC Excel file labeled "Lewpols" on my webpage (www.sscnet.ucla.edu/polisci/faculty/zaller).

7. President Bush showed that short, successful wars that cost few American lives do not harm popularity; but neither are they much help over the longer run.

8. Full details of the ideological coding are available upon request.

9. Though going beyond the direct evidence, this analysis suggests that Clinton's confrontation with Congress over the budget in early 1996, in which he reestablished his reputation as a defender of centrist policies, may have been as important to his November win as the economy.

10. Quoted in Newsweek, November 4, 1968, p. 28.

References

Alesina, Alberto, John Londregan, and Howard Rosenthal. 1993. "A Model of the Political Economy of the United States." American Political Science Review 87(1):12-33.

Brody, Richard. 1991. Assessing the President: The Media, Elite Opinion, and Public Support. Stanford: Stanford University Press.

Chaffee, Steven, and David Sears. 1979. "Uses and Effects of the 1976 Presidential Debates: An Overview of Empirical Studies." In The Great Debates: Carter vs. Ford, ed. Sidney Kraus. Bloomington: University of Indiana Press.

Connolly, Cecil, and Thomas B. Edsall. 1998. "Political Pros Looking for Explanations: Public Reaction Seems to Rewrite the Rules." Washington Post, February 9, A6.

Edelman, Murray. 1964. Symbolic Uses of Politics. Urbana: University of Illinois Press.

-----. 1988. Constructing the Political Spectacle. Chicago: University of Chicago Press.

Fiorina, Morris. 1981. Retrospective Voting in American National Elections. New Haven: Yale University Press.

Key, V. O. Jr. 1966. The Responsible Electorate: Rationality in Presidential Voting, 1936-60. New York: Vintage.

Kramer, Gerald H. 1971. "Short-Term Fluctuations in U.S. Voting Behavior, 1896-1964." American Political Science Review 65(1):131-43.

Mueller, John. 1973. War, Presidents, and Public Opinion. New York: Wiley.

Poole, Keith. Forthcoming. "Recovering a Basic Space from a Set of Issue Scales." American Journal of Political Science 42 (August).

Rosenstone, Steven. 1983. Forecasting Presidential Elections. New Haven: Yale University Press.

Sabato, Larry. 1993. Feeding Frenzy: How Attack Journalism Has Transformed American Politics. New York: Free Press.

Shaw, Daron. 1995. "Strong Persuasion? The Effect of Campaigns in U.S. Presidential Elections." Ph.D diss. UCLA.

Tulis, Jeffrey. 1987. The Rhetorical Presidency. Princeton: Princeton University Press.

"TV Ratings from Speech." 1998. Los Angeles Times, January 29, A11.

Zaller, John. 1992. Nature and Origins of Mass Opinion. Cambridge: Cambridge University Press.

About the Author

John R. Zaller is coauthor (with Herb McClosky) of American Ethos (1984) and author of Nature and Origins of Mass Opinion (1992). He is completing a booked called Theory of Media Politics: How the Interests of Politicians, Journalists, and Citizens Shape the News (University of Chicago Press). He received his Ph.D. from the University of California, Berkeley in 1984.