Why did Clinton survive the impeachment crisis? A test of three explanations.

Silva, Carol L. ; Jenkins-Smith, Hank C. ; Waterman, Richard 等

In Federalist no. 65, Alexander Hamilton wrote that the impeachment process is a decidedly political one, We use data from a national survey conducted in September and October 1998 to examine explanations for Bill Clinton's survival of the impeachment crisis: the robust economy, his own high popular standing with the public, and the concomitantly lower public evaluations of his principal investigators, Independent Counsel Kenneth Starr and the Republican-controlled U.S. Congress. Our analysis demonstrates that the performance of the U.S. economy did directly influence public support for impeachment, resignation, continued congressional hearings, and censure (four potential punishment options), but not in the manner most often hypothesized. More favorable assessments of the U.S. economy actually were related to higher levels of support for punishing the president. Evaluations of the president proved to be a double-edged sword. Although higher evaluations of Clinton's personal job approval ratings and of his personal ethics ratings are related to lower levels of support for punishment, Clinton benefited more from the former than the latter, as his approval ratings were high while the public held considerable doubts about his ethics. Finally, we show that higher public evaluations of Clinton's primary investigators, Starr and the U.S. Congress, were related to higher levels of support for punishing the president. As both actors had an exceedingly low average approval rating, however, this dynamic worked to Clinton's advantage.

**********

On the eve of the Senate impeachment trial of President William Jefferson Clinton, political analyst for the Chicago Tribune William Neikirk wrote, "At this moment, one overpowering fact has emerged: President Clinton's popularity appears to be speaking louder than the evidence" (Neikirk 1999, A1). Despite the December 1998 House vote to impeach him, and despite polls showing that the public was deeply concerned with his personal ethics, Clinton's job approval rating remained in the 60 percent range. In explaining this phenomenon, many political pundits pointed to the performance of the U.S. economy. For example, Kenneth T. Walsh of U.S. News and World Report (1999) wrote, "Most of the disparity between voters' judgment of his character and his job performance is due, quite simply, to the booming economy." A poll by the same magazine found "73 percent of voters feel that because the economy is in such good shape, people will look the other way [on impeachment] as long as Clinton is reprimanded" (ibid.).

Why did President Clinton survive the impeachment crisis of 1998-19997 Was it because the public approved of the job he was doing as president? Was it because the public gave Clinton credit for the state of the nation's economy or believed they were personally better off financially? Or was there another reason, broad public antipathy toward those who investigated the president--Independent Counsel Kenneth Starr and the Republican-controlled Congress? Using data from a 1998 national survey, we test whether the economy, perceptions of Clinton's job performance (along several dimensions), and perceptions of his principal investigators, Judge Start and the U.S. Congress, are related to public support for four possible methods of punishing the president: impeachment, resignation, whether Congress should conduct hearings into the impeachment matter, and censure of the president. Our findings are consistent with Alexander Hamilton's prescient commentary from Federalist Paper no. 65 which suggested that political factors would play a central role in impeachment cases.

Public Evaluations of the President and the Politics of Impeachment

Article I, Section 4 of the Constitution prescribes, "The President, Vice President, and all civil Officers of the United States shall be removed from Office on Impeachment for and Conviction of Treason, Bribery, or other high Crimes and Misdemeanors." Partly because this provision is so seldom employed, particularly with regard to presidents, the impeachment process is a broadly misunderstood constitutional mechanism. It is not a criminal process, but rather a political one. As Walter Ehrlich writes in the preface to his book Presidential Impeachment: An American Dilemma, impeachment "is a political decision (a vote) made by a political institution (the elected members of Congress) to remove a political official (the President, Vice-President, or other civil officers of the United States)." Ehrlich (1974, 39) continues, "Impeachment as understood by the Framers of the Constitution was not a criminal litigation. It was a political procedure used to remove officials who committed either a criminal or a non-criminal wrong." (1)

Likewise, in Federalist no. 65, Alexander Hamilton notes, "The subjects of its [impeachment's] jurisdiction are those offences which proceed from the misconduct of public men, or, in other words, from the abuse or violation of some public trust. They are of a nature which may with particular propriety be denominated POLITICAL, as they relate chiefly to injuries done immediately to the society itself." Hamilton was remarkably prescient in describing the nature of this political process. He continues, "The prosecution of" impeachment cases "will seldom fail to agitate the passions of the whole community, and to divide it into parties more or less friendly or inimical to the accused. In many cases, it will connect with the preexisting factions, and will enlist all their animosities, partialities, influence, and interest on one side or on the other; and in such cases there will always be the greatest danger that the decision will be regulated more by the comparative strength of parties, than by the real demonstrations of innocence or guilt."

Hamilton deduced that impeachment involves not only the various constitutional actors (the House, the Senate, the chief justice of the Supreme Court, and the president) but also the two great evils the Founders feared most: faction and the unregulated passions of the people. What Hamilton forecast in March 1788 is to a remarkable degree a description of the political dynamics of the impeachment of President William Jefferson Clinton. The two articles of impeachment enacted on December 19, 1998, passed the House of Representatives mainly along partisan lines, with only five Democrats voting with the Republican majority. Likewise, the Senate vote on conviction broke down largely along party lines, with all Democrats voting for acquittal. While the House and the Senate voted, politicians and pundits alike kept a watchful eye on another key manifestation of American politics: public opinion polls.

Performance versus Personal Assessments of President Clinton

Polls conducted as early as January and February 1998 indicated that, while the public was concerned with the president's personal behavior, it also continued to approve of his performance as president. A poll conducted for CBS News and the New York Times registered the president's approval rating at an astonishing 68 percent, even though 59 percent of the respondents from the same poll "were inclined to believe the president had an affair." Still, only 21 percent favored resignation and just 12 percent impeachment (Bennet and Elder 1998). Zaller (1998, 182) notes that Clinton actually received a bounce in the polls "in the initial 10 days of the Lewinsky imbroglio," an approximate 10 percent increase. Although the bounce was temporary, Clinton's approval ratings continued to hold steady at about 60 percent in most national polls throughout 1998, despite continuing evidence that a majority of Americans thought him guilty of both adultery and perjury.

As the debate over impeachment grew more heated, we sought data that would permit us to evaluate a range of arguments about the public's assessments of Clinton, his investigators, and the factors that may have had a bearing on preferences regarding the impeachment process. Because most polls on the issue omitted one or more key measures, we fielded a nationwide survey on the state of the presidency using the University of New Mexico Institute for Public Policy's computer-assisted telephone interviewing laboratory. Overall, 1,053 interviews from a nationwide sample were taken between September 19 and October 20, 1998. The sample list of numbers was drawn from a random-digit-dialing frame, (2) and the resulting response and cooperation rates were 54.3 and 59.4 percent, respectively, based on the American Association for Public Opinion Research standard definitions (American Association for Public Opinion Research 1998). The associated refusal rate was 37.1 percent. Given the number of completed interviews, the estimated sampling error is approximately [+ or -]3.5 percent, though the range is larger for subsamples.

Among the respondents to our survey, 16.6 percent rated President Clinton's job performance as "excellent," 38.6 percent as "good," 22.2 percent as "fair," and 22.6 percent as "poor." Not surprisingly given these evaluations, support for impeachment of the president was limited: 68.6 percent opposed impeachment, while only 31.2 percent supported it. Similarly, support for the president's resignation was only 36.2 percent, while 63.8 percent opposed it. There was somewhat greater backing for the idea that Congress should hold hearings to investigate the allegations into possible wrongdoing: 44.4 percent favored hearings while 55.6 percent were opposed to them. Public support was broadest for censure, with 65 percent of our respondents favoring this option and 35 percent opposing it. While the severity of the four options varies, each has one thing in common: each was a means of punishing the president for his actions. The support for the different actions are shown in Figure 1.

Thus, as various polls (including our own) indicate, the president's job approval ratings were high and majorities opposed both impeachment and resignation. At the same time, some form of punishment (e.g., continuing the impeachment process or censuring the president) was supported by a large percentage of the public. Given this dynamic, one possible explanation for Clinton's survival may be that the public saw the events leading up to the impeachment crisis as a largely personal and not a public matter. If so, then perceptions of Clinton's personal ethics also would be relevant. To tap into this dimension, we asked our respondents, "Thinking specifically about President Clinton, please tell me how well you think he rates on the following four leadership qualities." Respondents were asked to rate four qualities, including perceptions of his personal ethics, on a scale ranging from zero (extremely poor) to ten (extremely good). Asked to rate Clinton's personal ethics, 31 percent of our respondents gave Clinton the lowest possible rating. Slightly more than 70 percent rated his ethics as a five (the scale midpoint) or less. Consequently, while Clinton received broad support for his job performance, the public was far less enamored with his personal ethics. We therefore conjecture that two distinct kinds of evaluations of President Clinton were operating to affect preferences in the impeachment crisis: those who believed he was doing a good job would be more likely to oppose punishment, but those who had reservations about his ethics would be more likely to support punishing the president.

Clinton received higher ratings on the other three leadership dimensions that were included in our poll. While the mean value for "high ethical standards" was only 3.59 on the zero-to-ten scale, the mean for his "ability to work well with Congress," even during the midst of an impeachment crisis, was 6.17. With regard to "experience in foreign affairs" the mean value was 6.59 and for "sound judgment in a crisis" it was 6.70. Because the White House had clear incentives to focus the public's attention on the president's other leadership attributes, rather than perceptions of his personal ethics, we also include these measures in our analysis.

Consequently, while our respondents largely disapproved of Clinton's personal ethics, they may have based their judgment about impeachment on their evaluations of his overall job performance and leadership capability as president. The president and his advisors did much to encourage the public to think in this manner. The president was portrayed as actively working on the business of government. The clear message from the White House was that the president had the nation's best interests in mind while congressional Republicans were playing politics with a personal matter (see Blumenthal 2003). Thus, William Schneider (1998, 3042) was able to write, "Most Americans say they want Clinton to remain in office because they think he's doing a good job. They [therefore] set his personal values aside." But did they? That is an empirical question we will address.

It's the Economy, Stupid!

A second alternative is that, because the nation's economy was robust and individuals were financially better off, the public approved of the president's overall job performance and therefore opposed impeachment and the other punishment alternatives despite their reservations about the president's personal behavior. Writing in early 1998, Zaller (1998, 186) speculated about the president's approval ratings and the impeachment crisis. He wrote, "My personal hunch ... 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." If so, then in the words of Democratic political consultant James Carville, the reason that Clinton survived may be, "It's the economy, stupid."

As Clinton reminded the nation in his 1998 State of the Union address, and would remind them consistently thereafter, the nation's economy was in very good shape. In 1998, the unemployment rate was the lowest (during peacetime) since 1957. The inflation rate was the lowest in thirty years. Added to these results, the stock market had hit record levels and the budget had been balanced for the first time in decades. As a result, most Americans said they were pleased with the performance of the economy and the president's management of it. We asked, "Using a scale of one to seven, where one means very poor and seven means very good, how would you rate the current performance of the US economy?"; 76.3 percent of our respondents ranked the economy above the midpoint on the seven-point scale (that is, toward the favorable side), while another 14.4 percent ranked it at the midpoint. In other words, almost 91 percent of our respondents ranked the economy as a four or higher on the seven-point scale. We also asked, "Would you say that your family are financially better off, worse off, or about the same as you were a year ago?" Of our respondents, 32.2 percent said they were better off, while 57.9 percent said they were about the same, and only 9.9 percent said they were worse off. Put differently, those who had gained financially outweighed those who lost by better than three to one.

With public opinion polls showing high levels of support for Clinton's management of the economy, was it the extraordinary performance of the U.S. economy that encouraged the public to oppose impeachment? There is ample support for this possibility in the scholarly literature. Academics have long argued that the economy is an important determinant of presidential approval ratings and electoral outcomes (see Tufte 1978; Kinder and Kiewiet 1979, 1981; Feldman 1982; MacKuen 1983; Ostrom and Simon 1985; Markus 1988, 1992; Kinder, Adams, and Gronke 1989; Erikson 1989; Brody 1991; Brace and Hinckley 1992). If the economy impacts approval ratings and election results, then it also may have driven perceptions of the impeachment crisis.

The President's Investigators

There is a third possible explanation for Clinton's survival. While the public broadly approved of Bill Clinton's job performance, support for those who sought to remove him from office was lukewarm at best. This Is particularly evident with regard to Clinton's primary investigator, Independent Counsel Kenneth Starr. Starr quickly developed a negative image in the minds of many Americans. On the nightly news, he held impromptu meetings with the press while taking his garbage to the curb outside his house. The image created was not flattering: Starr was only interested in digging up the president's garbage. The White House was delighted with this symbolic turn of events (see Blumenthal 2003). It allowed them to build on their image of Bill Clinton as presidential, while Kenneth Starr was portrayed as vengeful, politically motivated, and interested only in pursuing higher office (a Supreme Court appointment). The Clinton image machine thus portrayed Starr as far from an impartial investigator.

Starr's personal behavior continued to raise questions about his fitness. The media reported that Starr had surrounded himself with partisans in the independent counsel's office and that he often fell to his knees and prayed while jogging. The latter allegation suggested that he might even be part of what Hillary Clinton had famously declared "a vast right-wing conspiracy" (see Tobin 1999).

Given the critical attention that Starr received from the White House and many media outlets, it is not surprising that his job approval rating trailed far behind that of the president. When asked how they evaluated Independent Counsel Starr's job performance, only 35.8 percent rated it as either "excellent" or "good." Negative feelings toward the independent counsel were particularly strong: 42.1 percent rated his performance as "poor," the lowest possible rating on our scale. Clinton's rating, by comparison, was 55.2 percent "excellent" or "good," and only 22.6 percent "poor." While the criteria for evaluating an independent counsel are surely quite different from those for a sitting president, the difference in the overall approval ratings is stark.

The Office of the Independent Counsel was not the only one investigating the president's behavior. As Starr investigated possible criminal charges against the president, the Republican-controlled Congress considered impeachment. As it did with Kenneth Starr, the White House public relations operation also portrayed Republicans in Congress as politically motivated. Like the independent counsel, Congress had lower job approval ratings than did the president. On our four-point scale which ranged from "excellent" to "poor," only 35.3 percent rated Congress' performance as either "excellent" or "good," while 19 percent rated its performance as "poor." While a smaller percentage rated Congress' performance as poor compared to the president's, overall its approval ratings lagged behind the president's.

Hypotheses

In sum, there are three viable explanations for Clinton's survival of the impeachment crisis. First, the White House strategy of portraying Clinton as presidential, combined with his own high approval ratings, may have helped him to survive. On the other hand, his personal ethics may have promoted support for punishment of some sort. Second, the state of the nation's economy and the fact that many Americans considered themselves to be better off financially may have led them to doubt the wisdom of rocking the boat during good economic times. Third, the low public support for the independent counsel and the Republican-controlled Congress may have worked to the president's advantage. Of these three explanations, which best explains why Clinton survived in office?

To answer this question, we first need to create an appropriate dependent variable. Examining preferences for impeachment separately raises a serious methodological concern related to an endogeneity problem. We conjecture that higher approval for the independent counsel would result in greater support for impeachment. At the same time, however, greater support for impeachment might lead to greater support for Starr because Starr was seen as the agent charged with establishing the case for impeachment. Unfortunately, given the cross-sectional nature of our data, there is no straightforward way for us to determine which way this effect occurs. We therefore need a modeling strategy that allows us to address this endogeneity issue.

Our strategy consists of three elements. First, we constructed an aggregated dependent variable so that impeachment is but one part of the overall "punishment" construct. We did this by creating a scale that combines the four possible punishment options. The scale runs from 0 (no punishment at all) to 1 (support for all four punishment options). The scale passes the basic threshold test with a Cronbach's alpha value of .71. (3) As shown in Figure 2, for those who preferred to employ only one punishment option, support was chiefly for censure (88 percent). When two punishment options were chosen, large majorities of respondents picked censure and/or hearings. A simple majority chose the combination of censure and hearings (56 percent picked this pair). For those respondents who chose three of the options, large majorities picked resignation, impeachment, and/or hearings. Fully 51 percent of this group chose the combination of hearings, impeachment, and resignation--omitting only the most lenient option (censure). Thus, the combined punishment variable reflects an increasing severity of punishment, ranging from no action to modest punishment to a maximum "throw the book" level of punishment. This dilutes the (plausible) direct link from opposition to the more specific punishment of impeachment to a low approval for Starr. The second element of our modeling strategy was to include the measure of the respondents' views of Clinton's ethics as an independent variable, as discussed above. We do the same for Clinton's overall job approval and other dimensions of leadership. Therefore, in estimating the link between Starr and punishment, we have stripped the respondents' evaluation of Starr of any covariation with evaluations of Clinton. So if the preference for punishment is indeed driving the measure of the evaluation of Starr, it is doing so in the absence of any effect of the level of approval or opprobrium for Clinton. That seems to us to be implausible. (4) Third, we ran the model with and without the evaluation of Starr and found the estimated coefficients to be remarkably stable. These results are shown in Table 1, column 3. As is evident there, none of the estimated coefficients changed signs, none lost or gained statistical significance, and the proportion of variance explained in the punishment variable dropped only slightly (from a pseudo [R.sup.2] of 0.24 to 0.23). And fourth, we remain modest in what we can conclude from the analysis. Should the relationship between approval for Starr and preference for punishment prove to be substantively and statistically significant, the result would be consistent with the hypothesis that dislike for Clinton's enemies reduces the preference for punishing Clinton--but it cannot be taken as definitive evidence.

[FIGURE 2 OMITTED]

With these considerations in mind, we used the punishment scale as our dependent variable for testing six hypotheses. With regard to the president's leadership, we posit the following:

H1: Among respondents, the higher the president's personal job approval rating, the less likely one will be to support punishing the president, which is represented by lower values on the five-point punishment scale.

H2: Likewise, the more positively one evaluates Clinton's personal ethics, the less likely one should be to support punishment (represented by lower values on the punishment scale).

H3: The higher one rates Clinton's performance on other presidential leadership tasks (foreign policy, the ability to work well with Congress, and his overall sound judgment in crisis situations), the less likely one should be to support punishment.

With regard to the economy, we hypothesize:

H4: The more positive one's assessment of the performance of the economy is, the less likely one will be to support punishment.

H5: Likewise, the better the respondent's own financial circumstances, the less likely one will be to support punishment.

Finally, with regard to the president's political investigators, we posit:

H6: The lower the approval rating of Independent Counsel Kenneth Starr, and the lower the approval rating of the U.S. Congress, the less support there will be for punishing the president.

We also control for an array of demographic and political factors. Regarding demographics, Bill Clinton received strong support in the 1992 and 1996 elections from women, the elderly, African Americans, and Latinos. Hence, for each of these groups, we expect lower levels of support for any form of punishing the president. On the other hand, we conjecture that the higher one's income, the more likely one should be to support punishment. We also control for the respondent's level of education, though we offer no directional hypothesis for it.

With regard to political factors, knowledge of the details of the allegations that would form the basis of the impeachment case should be important. "The Starr Report" was transmitted to the House Judiciary Committee along with much media fanfare. It identified in often salacious detail the independent counsel's charges against the president. Because the report's transmittal occurred prior to our survey, we asked how much respondents had read or heard about Ken Starr's report. Only 3.3 percent said they had read or heard nothing at all about the report, while 78.5 percent said they had read or heard "a great deal" or "a moderate amount" about it. Because the report was highly critical of the president, we hypothesize that the more one heard about it, the more likely one would be to favor punishment.

Other political factors also may have had an impact on the impeachment crisis. The more conservative one is ideologically, the more likely one should be to support punishment, while Democrats should be less likely than Republicans and Independents to do so.

Given the array of independent variables included, it is necessary to consider whether multicolinearity may mask or distort the relationships estimated in the model. To address this potential problem, we calculated the tolerance of each of the independent variables. Tolerance indicates the fraction of the variance in each of the independent variables that is not "explained by" the other independent variables. Tolerance ranges from zero (all of the variance for a given variable is accounted for by the other independent variables--or perfect multicolinearity) to one (none of the variance is accounted for). Low tolerances generally lead to unstable and inefficient estimates. The lowest tolerance was, not surprisingly, for Clinton's approval rating. The other independent variables accounted for 34 percent of the variation in this measure, resulting in a tolerance of 0.66. Tolerances for the presidential performance measures were also moderately affected, with scores ranging from .73 (for "sound judgment in a crisis") to .81 ("ability to work well with Congress"). Tolerances of these magnitudes generally do not pose serious multicolinearity problems for models using large data sets, as is the case here. However, given that the measure of approval for Clinton had the lowest tolerance, we include in the analysis a comparison model that excludes this explanatory variable from the model (see the second column of Table 1).

Our Results

Because our dependent variable is ordinal, we employ an ordered logit model. The results are presented in Table 1. We also estimated the magnitudes of the effects of each of the variables that proved to be statistically significant by calculating the average number of punishment types (0-4) that would be preferred by a respondent at the extreme ends of the scales for that variable. For example, we calculated that respondents who rated Clinton's ethics as "extremely good," but had average scores for all other independent variables, would prefer (on average) to impose 0.73 of the punishment categories. But respondents who rated Clinton's ethics as "extremely poor" would impose (on average) 1.88 punishments--making a difference of 1.15 for this variable. (5) These estimates are shown in Table 2.

One of the primary explanations provided for Clinton's survival of the impeachment crisis relates to the president's personal popularity. We find that the higher one rated Clinton's job approval, the less likely one was to support punishing the president. This is consistent with conventional wisdom on the subject, as well as our own first hypothesis (H1). What is more, this relationship accounted for a change from a preference for an average of 2.41 punishments for those rating Clinton's performance as "poor" to 0.82 for those who gave a rating of "excellent." Consistent with hypothesis H2, we also find that the more positively one rated Clinton's ethics, the less likely they were to favor punishing the president. Those giving Clinton the most negative ethical evaluations wanted, on average, to impose 1.88 of the punishments, compared with only 0.73 for those rating his ethics most highly. As we have noted, unfortunately for the president, while he had high job approval ratings, far fewer respondents to our survey gave him favorable ratings on his personal ethics. Consequently, for many individuals, more negative attitudes about his ethics translated into stronger support for punishing the president. To this extent, then, the two factors worked at cross purposes.

On the other hand, White House attempts to shift the debate to other leadership dimensions seem to have failed. Our model provides little support for the idea that public perceptions of Clinton's performance in foreign affairs, his ability to work well with Congress, or his sound judgment in a crisis are independently related to support for punishment (H3). Considered as a whole, then, evaluations of Clinton's performance proved to be something of a double-edged sword, though his overall job approval had greater weight than did assessments of his ethical behavior.

This assessment is confirmed by inspection of the results when Clinton's overall approval rating is removed from the model. The estimated effect of perceptions of Clinton's ethics remains strong and significant. The magnitudes of the coefficients for the other presidential leadership dimensions remain modest, though two of them increase in magnitude sufficiently to reach levels of statistical significance (judgment in a crisis and working well with Congress). The measure of experience in foreign affairs remains statistically insignificant.

If the president's public standing had both positive and negative consequences, then what effect did the perceptions of the economy have? As noted, most of the speculation was that the economy, which was strong, would work to the president's advantage. Our findings do not support this hypothesis (H4). Instead, as shown in Table 2, we find that the more positively one evaluates the performance of the U.S. economy, the more likely one is to flavor punishment. While the magnitude of the effect is modest (shifting the average preferred number of punishments by only 0.45), it is nevertheless statistically significant and runs counter to the direction we had hypothesized. The reason for this counterintuitive finding is that our estimates control for presidential approval, which is affected by perceptions of the economy, thereby removing the indirect effect of the economy on punishment preferences. (6) Once presidential approval is removed from the model, as shown in the second column of Table 1, neither of the economy variables is significantly related to preferences for punishment. Simply put, once we control for presidential popularity, better perceptions of the economy increased the appetite for punishment. Furthermore, as for one's own finances, this variable is not directly related to the punishment scale at all (H5).

What accounts for our counterintuitive findings regarding the economy? We speculate that, while the public considers economic performance to be an important determinant of how well a president is performing his job (Zaller 1998), the impeachment debate was not about economics. It was about presidential ethics and popularity. As such, economic factors contributed to the president's survival indirectly by promoting higher approval ratings, but not directly by diminishing support for punishing the president. Again, this is not a surprising conclusion if we consider that the impeachment crisis was chiefly about ethics, not economics.

If the president's popular support was a double-edged sword, and the robust economy actually led to higher support for punishment, how did the president survive? The third explanation for Clinton's survival is that negative evaluations of his enemies (Independent Counsel Kenneth Starr and the Republican-controlled U.S. Congress) diminished preferences for punishment. We find that higher ratings for Starr's job performance are related to higher levels of support for punishing the president. As shown in Table 2, those who rated Starr most positively preferred (on average) 1.99 punishments, compared to 0.94 for those who rated him most poorly. As we noted, however, the percentage of our respondents who rated Starr's performance as poor was quite high, with over 40 percent rating it as "poor." Thus, overall, Starr's low approval ratings appear to have served to dampen public support for punishing the president. But these results must be interpreted with caution: despite our use of the aggregated "punishment" variable, and our controls for Clinton's ethics and performance measures, we cannot reject the possibility that the effect is bidirectional. Moreover, there is considerable covariation between the Starr and Clinton ratings. Once the presidential approval measure is removed from the model, as shown in the second column of Table 1, the estimated effect of Starr's approval increases modestly (from 0.29 to 0.38). This is as one would expect, but also illustrates the overall stability of the model results.

Consistent with our sixth hypothesis (H6), as was the case with Starr, the coefficient for evaluations of Congress is significant (though only at the one-tailed level), again suggesting that higher approval ratings for one of Clinton's primary investigators is positively related to support for punishing the president. Note that the effect is more muted than was approval for Starr, resulting in a variation in preferred punishments of only 0.65, on average. Again, because Congress received lower approval evaluations than the president, this dynamic worked to Bill Clinton's advantage.

In addition to these results, only one of the basic demographic variables was related to the punishment scale. African Americans were less likely to support punishment, but again the coefficient only meets a one-tailed test of significance. Finally, and not surprisingly, Democrats were less likely to favor punishment than were Independents and Republicans.

Note that the model proved to be quite stable when the measure of presidential approval was omitted. None of the estimated coefficients changed signs, and the magnitudes of the changes were modest. The only changes of statistical note were the (modestly) increased magnitudes of several of the coefficients for the measures of Clinton's presidential leadership. In general, then, correlations among the independent variables employed do not alter the general conclusions of our analysis.

In summary, evaluations of the president proved to be something of a double-edged sword, with his approval ratings undercutting support for punishment while evaluations of his ethics increased support for it. The more favorably one viewed the state of the economy (controlling for presidential approval), the more likely they were to favor increased punishment. Thus, while one can argue that the economy had indirect effects on Clinton's approval ratings, it is difficult to argue that the economy directly served to help Clinton survive the impeachment crisis. Finally, with regard to the president's principal investigators, higher support for both Starr and Congress was related to increased support for punishment. As both actors had relatively low approval ratings (especially in comparison to the president), this factor worked to the president's advantage.

Conclusions

As Alexander Hamilton noted in Federalist no. 65, the impeachment process is a decidedly political one. We have shown that evaluations of President Clinton represented a double-edged sword. While his (positive) job approval ratings were strongly related to lower levels of support for punishing the president, his (negative) personal ethics ratings promoted increased support for punishment. The economy did have an impact, but not in the expected direction. Indirectly, favorable economic views boosted Clinton's approval ratings, which in turn reduced the desire for punishment. But the direct effect was negative: those with a more favorable view of the economy actually were more likely to favor punishing the president. Put another way, once the effects of a robust economy on approval ratings are controlled for, those who benefited most from that economy were more inclined to want to punish the president.

Clinton may well have been both lucky and smart in his choice of political enemies. Evaluations of the president's main political inquisitors, Independent Counsel Kenneth Starr and the Republican-controlled Congress, were related to the level of support for punishing the president (with higher evaluations related to greater support for punishing the president). Because Starr and Congress had much lower approval ratings than did the president, this factor seems to have worked to Clinton's political advantage. The president and his advisors appear to have understood the importance of this political dynamic. They cultivated it throughout the crisis, continually polling the public to determine what it was thinking and casting the independent counsel and his office in a negative partisan light. Meanwhile, the president's advisors created a public relations strategy that presented Bill Clinton as hard at work on the people's business (including maintaining a healthy economy), even when in private he was deeply concerned about his own personal fate. The president even delivered an upbeat State of the Union address in 1999 in the midst of his own impeachment trial, both to show that he was hard at work and to take advantage of a unique opportunity to influence public opinion (Blumenthal 2003). Clinton was rewarded for his political efforts. The public continued to support him, even though, as our data show, it had serious reservations about his personal ethics.

Our conclusions raise some serious concerns. Because our results highlight the importance of evaluations of the president and his principal investigator, they raise the specter of a future presidential investigation in which prosecutors use polling and public relations techniques to sell their case for (or against) impeachment to the public. This would almost certainly prove to be institutionally and politically corrosive, but given our findings it may be an inevitable result and legacy of the Clinton impeachment case, for as Hamilton reminds us, the impeachment process indeed is fundamentally political. We can therefore hope that impeachments take place only rarely and at utmost need.

Appendix

Measurement of Variables

Clinton Measures. With regard to evaluations of the president, respondents were asked, "Using a scale of excellent, good, fair, or poor how would you rate the job Bill Clinton is doing as President of the United States?" As with the other approval questions (e.g., for Starr and the U.S. Congress), 4 = excellent, 3 = good, 2 = fair, and 1 = poor. Regarding specific leadership qualities, respondents were asked, "Now thinking about President Clinton, please tell me how you think he rates on the following four leadership qualities. Using a scale from zero to ten, where zero means extremely poor and ten means extremely good, how would you rate Clinton on sound judgment in a crisis; experience in foreign affairs; high ethical standards; the ability to work well with Congress?" The order of the four leadership qualities was randomly determined.

Economic Measures. To measure sociotropic attitudes, we asked, "Using a scale from one to seven, where one means very poor and seven means very good, how would you rate the current performance of the US economy?" To measure pocketbook attitudes, we then asked, "Would you say that you and your family are financially better off, worse off, or about the same as you were a year ago?" The scale was arranged from worse off to better off. This variable was recoded with "about the same" as the midpoint.

Start and Congress Measures. With regard to evaluations of the independent counsel, Kenneth Starr, respondents were asked, "Using a scale of excellent, good, fair, or poor how would you rate the job Kenneth Starr is doing as the Independent Counsel investigating allegations about President Clinton?" With regard to evaluations of Congress, respondents were asked, "Using a scale of excellent, good, fair, or poor how would you rate the job Congress is doing?"

Demographic Variables, Gender, African Americans, and Latinos are measured as dummy variables (0 = female, 1 = male; 1 = African Americans, 0 = non-African Americans; 1 = Latino, 0 = non-Latinos). We measured income in increments of $10,000. Age was measured by asking, "How old are you?" Education was measured by asking, "What is the highest level of education you have completed?" This included (1) elementary or some high school; (2) high school graduate/GED and trade or vocational certificate; (3) some college/associates degree; (4) college graduate; or (5) postgraduate degree.

Political Variables, A dummy variable was included, with those identifying themselves as Democrats = 1 and all others = 0. Ideology was measured on a continuum from 1 = strongly liberal to 7 = strongly conservative. Regarding knowledge of the Starr report, respondents were asked, "Please tell me how much have you read or heard about the specific details in Starr's report to Congress: a great deal, a moderate amount, only a little, or nothing at all."

Dependent Variables. The scale for the dependent variables was constructed from the following questions: "Based on what you know about Bill Clinton, do you think that he should or should not be impeached and removed from office?" 1 = he should be impeached and 0 = he should not be impeached. They were next asked, "Do you think that Bill Clinton should or should not resign now and turn the presidency over to AI Gore?" 1 = resign and 0 = not resign. Next they were asked, "What do you think that Congress should do with Ken Starr's Report--hold hearings to investigate the charges contained in the report, or take no action on the report and immediately end the investigation into these matters?" 1 = hold hearings and 0 = no hearings. Respondents also were asked, "Regardless of whether or not you think Clinton should be impeached, do you think Congress should or should not vote to censure Clinton--that is, pass a formal resolution expressing disapproval of his actions?" 1 = should vote to censure and 0 = should not vote to censure.

References

American Association for Public Opinion Research. 1998. Standard definitions: Final dispositions of case codes and outcome rates for RDD telephone surveys and in-person household surveys. Ann Arbor MI: Author.

Bennet, James, and Janet Elder. 1998. Despite intern, president stays in good graces. New York Times, February 24, A1, A14.

Blumenthal, Sidney. 2003. The Clinton years. New York: Farrar, Straus, and Giroux.

Brace, Paul, and Barbara Hinckley. 1992. Follow the leader: Opinion polls and the modern presidents. New York: Basic Books.

Brody, Richard. 1991. Assessing the president: The media, elite opinion, and public support. Stanford, CA: Stanford University Press.

Ehrlich, Waiter. 1974. Presidential impeachment: An American dilemma. Saint Charles, MO: Forum Press.

Erikson, Robert S. 1989. Economic conditions and the presidential vote. American Political Science Review 83: 567-73.

Feldman, Stanley. 1982. Economic self-interest and political behavior. American Journal of Political Science 26: 446-66.

Kinder, Donald R., and D. Roderick Kiewiet. 1979. Economic discontent and political behavior: The role of personal grievances and collective economic judgments in congressional voting. American Journal of Political Science 23: 495-527.

--. 1981. Sociotropic politics: The American case. British Journal of Political Science 11: 126-61.

Kinder, Donald R., Gordon S. Adams, and Paul W. Gronke. 1989. Economics and politics in the 1984 American presidential election. American Journal of Political Science 33: 491-515.

MacKuen, Michael B. 1983. Political drama, economic conditions, and the dynamics of presidential popularity. American Journal of Political Science 27:165-92.

Markus, Gregory B. 1988. The impact of personal and national economic conditions on the presidential vote: A pooled cross-sectional analysis. American Journal of Political Science 32: 137-54.

--. 1992. The impact of personal and national economic conditions on presidential voting, 1956 - 1988. American Journal of Political Science 36: 829 - 34.

Neikirk, William. 1999. Pensive senators want fast trial: President's approval rating makes removal unlikely. Albuquerque Journal, January 3, A1, A9.

Ostrom, Charles W., and Dennis M. Simon. 1985. Promise and performance: A dynamic model of presidential popularity. American Political Science Review 79: 334-58.

Schneider, William. 1998. It's a values thing. Or not. National Journal, December 19, 3042.

Tobin, Jeffrey. 1999. A vast conspiracy: The real story of the sex scandal that nearly brought down a president. New York: Simon and Schuster.

Tufte, Edward R. 1978. Political control of the economy. Princeton, NJ: Princeton University Press.

Walsh, Kenneth T. 1999. The price of victory: Acquittal wasn't cheap. Now Clinton and Congress must move on. But will voters forgive and forget? U.S. News and World Report, February 22, 29. Available from http://usnews.com/usnewslnewslarticles19902221archive_000315_print.htm.

Zaller, John R. 1998. Monica Lewinsky's contribution to political science. PS: Political Science and Politics 31: 182-89.

(1.) Italics are included in the original.

(2.) The national random-digit-dialing telephone list was purchased from Survey Sampling of Fairfield, Connecticut.

(3.) One might expect that censure is an alternative to impeachment or resignation and that those who support censure might oppose the other two options. But this was not the case, as indicated by the positive (but very small) correlations among the measures of these options.

(4.) Indeed, in a simple ordinary least squares model using Starr's approval to predict preferences for punishing Clinton, inclusion of measures of Clinton's approval ratings and the evaluation of Clinton's ethics as controls reduced the magnitude of the estimated coefficient of Starr's approval by nearly two thirds, from -0.71 to -0.26.

(5.) The estimates were derived by first calculating the logit for each of the levels of the punishment scale, setting all independent variables at their means except for the variable that was the focus of interest. We then converted the logits into probabilities for each respondent. Finally, we calculated the sum of the products of the probability for each level of punishment (zero through four) for each respondent. This permitted us to calculate the average number of punishments preferred across each of the independent variables, as shown in Table 2.

(6.) Both the perceived performance of the economy and the pocketbook measure are positively and significantly related to the president's approval ratings, explaining roughly 7 percent of the overall variance in approval.

CAROL L. SILVA

University of Oklahoma

HANK C. JENKINS-SMITH

University of Oklahoma

RICHARD WATERMAN

Martin School of Public Policy and Administration

Carol L. Silva is an associate professor at the University of Oklahoma in the Department of Political Science and associate director of the Center for Applied Social Research. Her work is primarily in public opinion research, science and public policy, and economic valuation.

Hank C. Jenkins-Smith is a professor in the Department of Political Science at the University' of Oklahoma and associate director of the Center for Applied Social Research. He does research on risk, security, and public policy.

Richard Waterman is a professor of political science at the Martin School of Public Policy and Administration. He is the author of such books as The Changing American Presidency, Presidential Leadership: The Vortex of Power, The Image-Is-Everything Presidency, and Bureaucratic Dynamics and has also published widely in leading journals.

AUTHORS' NOTE: We would like to thank Kirk Randazzo, Lonna Atkeson, Anthony Coveny, and Regina Lawrence for their useful comments on this manuscript. We also wish to thank the anonymous reviewers, who provided careful and constructive criticism of the early versions of this manuscript.

TABLE 1
Logit Analysis of Punishment Scale

 Punishment Scale Punishment Scale
 (Clinton Approval (Clinton Approval
 Included) Excluded)

Clinton variables

Clinton's approval rating -.77 *** (omitted)
 (.11)

Clinton's ethics -.30 *** -.35 ***
 (.04) (.04)

Sound judgment in a crisis -.04 -.11 **
 (.04) (.04)

Experience in foreign affairs .05 .03
 (.04) (.04)

Works well with Congress -.05 -.08 *
 (.04) (.04)

Economic variables

Sociotropic attitudes .15 ** .09
 (.07) (.07)

Pocketbook attitudes -.00 -.00
 (.12) (.12)

Impeachment investigators

Starr's approval rating .29 ** .38 ***
 (.09) (.09)

U.S. Congress approval rating -.16 * -.11
 (.09) (.09)

Demographic factors

Education .08 .07
 (.07) (.07)

Age -.001 -.001
 (.004) (.005)

Gender .04 .04
 (.14) (.14)

Income -.02 -.02
 (.03) (.03)

African American -.58 * -.67 *
 (.37) (.35)

Latino -.17 -.16
 (.35) (.35)

Political factors

Read or heard of the .05 .03
Starr report (.09) (.09)

Ideology .07 .07
 (.05) (.05)

Democrats -.33 ** -.38 **
 (.16) (.16)

N 763 764
Pseudo [R.sup.2] .24 .21

 Punishment Scale
 (Starr Approval
 Removed)

Clinton variables

Clinton's approval rating .79 ***
 (.11)

Clinton's ethics -.32 ***
 (.04)

Sound judgment in a crisis -.06
 (.04)

Experience in foreign affairs .07
 (.04)

Works well with Congress -.07
 (.04)

Economic variables

Sociotropic attitudes .12*
 (.07)

Pocketbook attitudes .01
 (.12)

Impeachment investigators

Starr's approval rating (omitted)

U.S. Congress approval rating -.25 **
 (.09)

Demographic factors

Education .07
 (.07)

Age -.001
 (.005)

Gender .01
 (.14)

Income -.01
 (.03)

African American -.58 *
 (.35)

Latino -.13
 (.35)

Political factors

Read or heard of the .06
Starr report (.09)

Ideology .09 *
 (.05)

Democrats -.42 **
 (.16)

N 774
Pseudo [R.sup.2] .23

*** Significant at the .01 level; ** significant at the .05 level;
* significant at the .10 level.

TABLE 2
Estimated Magnitudes of Effects of Independent Variables, Holding
All Else Constant at Mean Values

 Average Number of Average Number of
 Punishments at Punishments at
Explanatory Variable Lowest Scale Value Highest Scale Value

Clinton's approval 2.41 0.82
 (1 = poor, 4 = excellent)
Clinton's ethics 1.88 0.73
 (0 = extremely poor,
 10 = extremely good)
Starr's approval 1.99 0.94
 (1 = excellent, 4 = poor)
Congress's approval 1.68 1.03
 (1 = excellent, 4 = poor)
U.S. economy 0.94 1.39
 (1 = very poor, 7 =
 very good)

 Magnitude
Explanatory Variable of Effect

Clinton's approval 1.59
 (1 = poor, 4 = excellent)
Clinton's ethics 1.15
 (0 = extremely poor,
 10 = extremely good)
Starr's approval 1.05
 (1 = excellent, 4 = poor)
Congress's approval 0.65
 (1 = excellent, 4 = poor)
U.S. economy 0.45
 (1 = very poor, 7 =
 very good)

FIGURE 1. Preferred Punishments by Percent of Sample

Impeach 31%
Resign 37%
Hearings 44%
Censure 64%

Note: Table made from bar graph.