As Time Goes By: Daily Opinion Change During the Persian Gulf Crisis

PoliUcal Communication. Volume 10, pp. 353-367 Printed in ttw UK. All rights reserved 1058-4609/93 110.00 + .00 Copyright ©1993 Taylor & Franca As Time Goes By: Daily Opinion Change During the Persian Gulf Crisis LEE SIGELMAN JAMES LEBOVIC The George Washington University CLYDE WILCOX Georgetown University DEE ALLSOP The Wirthlin Group Abstract The period between the August 1990 Iraqi invasion of Kuwait and the beginning of the U.S.-ted military counteroffensive in January 1991 provides the focus for an analysis of the effect of daily events on U.S. public support for U.S. policy, President Bush's handling of the situation, and optimism that the crisis would be peacefully resolved. Three 'good news and bad news' interpretations of the linkage between events and opinion are tested. Of these, only one, which emphasizes the erosive effect of the accumulation of events in a foreign policy crisis that drags on unresolved, proves consistent with the data. Keywords George Bush, crisis, Persian Gulf crisis, public opinion trends On August 2, 1990, Iraqi forces invaded Kuwait. For the next 51/2 months—that is, until the U.S.-led counteroffensive got underway on January 16, 1991—the American public careened along on a veritable rollercoaster of daily events. This was a period of high drama, fueled by heated rhetoric from the opposing camps. The American president, George Bush, repeatedly vowed that the "intolerable" invasion would not stand and likened the Iraqi leader, Saddam Hussein, to Adolf Hitler. Saddam Hussein, who behaved in ways that sometimes seemed perversely calculated to bring the wrath of the rest of the world down on himself and his nation, grew sterner in his resolve to stay in Kuwait as the pressure to withdraw his troops increased; he vowed a "holy war" against the United States and promised "the mother of all battles" if and when war came. Throughout the period leading up to the air war, the news columns and airwaves were crammed with news from the Gulf; for example, during the first three months following the Iraqi invasion, the Thanks are due to Lance Bennett, David Grier, Fred Joutz, and Jarol Manheim for their valuable suggestions. The opinion data analyzed here were collected by the Wirthlin Group; the authors alone are responsible for the analyses and interpretations. 353 354 Lee Sigelman et a/. New York Times published an average of 25 news stories, editorials, or columns every day about the crisis (Dorman & Livingston, 1992). However, the news often seemed very puzzling, indeed. One day's events might seem to hold out promise of a peaceful resolution; the next day a shooting war might seem imminent; and for a day or two thereafter a luil might set in. The signals from Washington, Baghdad, and other world capitals were numerous, dramatic, and confusing—by turns frightening and hopeful, active and passive, maddening and reassuring, shocking and soporific. The question addressed here is how, on a day-to-day basis, the American public interpreted and responded to these signals. Tracking Opinion Trends From August 22, 1990 (three weeks after the Iraqi invasion of Kuwait), through the end of 1990 (two weeks before the expiration of the UN deadline for Iraqi withdrawal and the onset of the air war), the Wirthlin Croup, a private survey research firm contracted by Citizens for a Free Kuwait, conducted a daily tracking poll of residents of the United States concerning the situation in the Gulf. Interviewees were selected by random-digit dialing, with quotas based on gender and state. Approximately 200 telephone interviews a day were conducted until November 1; thereafter the number was reduced to 100 a day. Telephone calls were made every day except Fridays and holidays (Labor Day, Thanksgiving, Christmas Eve, Christmas Day, and New Year's Eve). Overall, 16,903 people were interviewed on 109 days during the 131-day period between late August and the end of the year. The issue of concern here is why public opinion about the crisis varied from day to day and over the course of the crisis period. That is, why did a certain percentage of those who were interviewed on one day, but a different percentage the next day, express a given view, and why did a certain percentage of those who were interviewed in August, but a different percentage in December, express a given view? A mundane but inescapable part of the answer involves sampling error. Because of sampling error, even if nothing happened to change opinions from one day to the next, an observed jump from 43 percent to 57 percent from one day to the next in the opinions expressed by daily samples of 200 randomly selected respondents might reflect nothing more momentous than chance fluctuations around a constant 50 percent parameter. Naturally, cutting the size of the daily samples to TOO a day widened the 95 percent confidence interval even more, to 20 percentage points in the example just given. Accordingly, we must anticipate a good deal of random error in the daily estimates of public opinion. Does this mean that the tracking poll results are too riddled with random error to warrant serious attention? The most direct way to address this issue is to compare tracking poll estimates to estimates based on larger samples (Allsop & Weisberg, 1988). A question asked on all 109 days of the tracking poll enables just such a comparison; "From what you have heard or read, do you approve or disapprove of the way President Bush is handling the situation in the Persian Gulf?" During the same period, respondents in 20 Gallup polls were asked whether they approved or disapproved *of the way George Bush is handling this current situation in the Middle East involving Iraq and Kuwait."^ The Gallup samples ranged between 750 and 1,250 respondents, several times larger than those used in the tracking poll, but the tracking poll estimates of public approval of Bush's handling of the situation closely paralleled Gallup's: the simple correlation between the two sets of estimates is .89. Daily Opinion Change During the Gulf Crisis 355 We interpret this correlation as an indication that even though the daily tracking poll results contain substantial noJse, the size of their daily samples does not invalidate them as measures of public opinion trends during the period that immediately preceded the air war. We employ three measures of opinion about the Gulf crisis. On all 109 days of the tracking poll, respondents used the four-point scale we just compared to its Gallup poll counterpart to rate President Bush's handling of the situation in the Gulf. Our first measure is the mean daily score on this scale. Each day tracking poll respondents were also asked to assess U.S. policy: 'What do you think about the actions the U.S. has taken so far in the Persian Gulf? Do you think U.S. actions have been too tough, not tough enough, or just about right?" For this question we use the daily percentage of respondents who said U.S. actions had been "just about right."^ Finally, a week after the tracking poll began respondents were first asked the following question, which was repeated every day thereafter: "From what you have heard and read, do you believe the conflict in the Persian Gulf will be resolved through peaceful negotiations or a military conflict?" The percentage who said on a given day that they expected the crisis to be resolved through peaceful negotiations constitutes our third opinion measure. Thus one measure captures trends in public assessments of President Bush's handling of the situation; one focuses somewhat more broadly on assessments of U.S. policy in the Gulf; and one looks toward the anticipated outcome of the crisis. Figure 1 presents daily time lines for these three series.^ According to Figure l(a), approval of President Bush's handling of the Gulf situation waned perceptibly during the crisis period. During the first month of the poll. Bush's daily mean on the 0-3 scale tended to range upward from 2.25, indicating widespread support. However, his ratings then began to fall, dropping into a band .25 either way from 1.75 by mid-October and then leveling off. Thus the decline in Bush's ratings occurred in two phases: first, a relatively rapid drop during the autumn that bottomed out in October and November, and then a sideways movement in December. As a consequence, the time line of these ratings is parabolic—indicating the same trend that Brody (1992) noted in monthly data on support for Bush during the same period. The course followed by "just about right" responses to the question of whether U.S. actions had been too tough, not tough enough, or just about right closely paralleled that of mean daily evaluations of Bush's handling of the situation, as comparison of Figure 1(a-b) suggests and as the correlation between the two series (r - .74) confirms. In late August, three weeks after the Iraqi invasion of Kuwait, two-thirds or more of those polled on a given day considered U.S. actions appropriate. But this proportion declined gradually over the next three months, finally bottoming out around 50 percent in late November and early December and then turning very slightly upward. Again, then, the time line is parabolic. Thus both of the time-series on public evaluations of how well the situation in the Gulf was going followed a parabolic course. That is, during this highly charged period public support for President Bush's handling of the crisis and for U.S. actions themselves did not plunge indefinitely. Rather, support levels dropped, reached a floor, and stayed close to that floor thereafter. However, it is important to note that the support floor was actually rather high; in other words, although support for the president and for administration policy declined perceptibly during the period considered here, "hard-core* support nonetheless remained fairly widespread. c •o c ro xzen 2.75 - (a) 2.5 - 2.25 - m 2 - a 1.75- (D O 1.5 - OQ. 1.25 20 40 60 80 100 120 100 120 Day en BO - (b) IX 70 60 50 - I 40 30 1 20 40 60 80 Day 60 - (c) so 50 - * • • 40 30 a u03 o 0 o 20 - o o 0 * a. X LU 10 \ 1 r 20 40 6b 8b Day 100 120 Figure 1. Tracking data on three key indicators of Gulf conflict opinion, August 22 through December 31, 1990. 356 Daily Opinion Change During the Culf Crisis 357 President Bush's oft-stated goal of expelling Iraqi forces from Kuwait and reinstating the deposed Kuwaiti government immediately raised the question of how the liberation of Kuwait could be achieved. One possibility was a unilateral Iraqi withdrawal, but from the outset it seemed clear that this was unlikely. Another possibility was a negotiated settlement in which the Iraqis bowed to the international forces arrayed against them, but still achieved some of the alms that had propelled their move into Kuwait. Although this was the ostensible goal of much public diplomacy during the autumn, the prospects for a negotiated settlement grew progressively more distant as the situation wore on. Realistically, then, the idea that a restoration of the status quo ante would require an American-ted military offensive had to be taken seriously from the very outset. A substantial portion of the public immediately grasped this possibility. According to Figure 1(c), during the first month of the tracking poll only 40 to 50 percent expressed confidence that the crisis would be settled through peaceful negotiations, leaving 50 to 60 percent who did not share this optimism. Thereafter, public expectations of a negotiated settlement fluctuated widely from day to day, but generally declined with the passage of time. By the end of the year, with the UN resolution's January 15 deadline drawing ever nearer, only about 25 percent still clung to the optimistic expectation that the situation would be peacefully resolved. War, most Americans had come to believe, was close at hand. In sum, by arraying opinions on a time line we have seen that public opinion in the United States not only fluctuated from day to day, but changed over the course of time. However, even though arraying opinions on a time line serves as a logical first step toward understanding opinion change, time per se is not an explanatory factor (Kernell, 1978). That is. Figure 1 leaves no doubt that something happened between August 22 and December 30 that shaped public opinion, but the time lines say nothing about the cause of it. We know that, whatever it was, it was associated with time. The challenge is to identify events and processes that could have shaped the course opinions followed over time. Tracking Crisis Events The obvious place to look for sources of daily opinion change is toward the everchanging situation in the Persian Culf itself, as reported by the news media. From news stories printed in the Nev/ York Times between August 2, 1990 (the day of the Iraqi invasion of Kuwait), and January 16, 1991 (the day the UN deadline for Iraqi withdrawal expired and the U.S.-led military counteroffensive began) we compiled a comprehensive catalog of events related to the situation in the Cuif. We defined an event as a discrete activity directed by one or more actors (nations or groups of nations, such as the United States or the UN) toward one or more targets (nations or groups of nations, such as Iraq or the Persian Gulf nations). In all, we cataloged 801 such events, though here we consider only events that were reported in the Times by the last day of the tracking poll (December 30) and were initiated by either the United States or Iraq. For present purposes, the key questions are whether an event revolved, in part at least, around a diplomatic issue and whether it consisted, in part at least, of military activity. We coded an event as revolving around a diplomatic issue if it centered on intemational deliberations, consultations, or negotiations. No matter how we coded an event on the issue dimension, we coded it as military on the 358 Lee Sigelman et al. activity dimension if weapons or military personnel were used or if their use was considered or threatened/ We coded the United States as the initiator of 62 events that revolved around a diplomatic issue between August 2 and December 30, Iraq as the initiator of 39; for the same period we coded the United States as initiating 82 events that involved military activities and Iraq 24. We summed the number of events of each type that occurred on a given day in order to produce separate daily tallies of U.S.- and Iraqi-initiated events that revolved around diplomatic issues (referred to hereafter as diplomatic events) and of U.S.- and Iraqi-initiated events that involved military activities (referred to hereafter as military events).^ Through December 30, the Times reported as few as zero and as many as three daily events of each type, except for Iraqi-initiated military events, which ranged only from zero to two. The frequencies of U.S.- and Iraqi-initiated diplomatic events on a given day were moderately interrelated (r - .32); none of the other correlations between daily frequencies of U.S.- and Iraqi-initiated diplomatic and military events exceeded .10. In tallying U.S.- and Iraqi-initiated diplomatic and military events, we drew no distinctions based on the scope or magnitude of an event, as we lacked any a priori criteria for making such distinctions. An alternative analytic strategy, which we did not pursue, would have been to specify, from all the events we identified, a relatively small number of events that had potentially "major" consequences, and then to home in on the observed consequences of each. Though attractive in certain respects, such a strategy would have run afoul of some extremely difficult conceptual and measurement issues, so we ultimately opted for the simpler approach we have just outlined. The Link Between Opinion and Crisis Events Why might one expect a connection between the ebb and flow of crisis events on the one hand, and daily opinion change on the other, and what type of connection might one expect? The fundamental premise underlying our search for linkage between events and public responses is that people tend to react positively to good news and negatively to bad news (Brody, 1991). In order to apply this premise, we must be clear about what constitutes good and bad news. One obvious criterion for defining good and bad news is the desire to maintain peace and avoid war. As applied to the Gulf crisis, this idea suggests that the public may have responded positively to daily reports of U.S.- or Iraqi-initiated diplomatic events and negatively to daily reports of U.S.- or Iraqi-initiated military events. Though plausible, this interpretation is by no means certain, for counterinstances spring readily to mind. For example. President Bush's great success in stirring anger and resentment about the Iraqi invasion and in vilifying Saddam Hussein made it likely that a substantial portion of the American public would have greeted as good news a report that the United States had just launched a military strike against the Iraqis. Indeed, President Bush and his closest White House advisors are said to have viewed unilateral Iraqi withdrawal as undesirable, for they believed that Saddam Hussein would emerge from this "nightmare scenario" unpunished, unrepentant, and with his military power Intact—outcomes the Bush administration urgently hoped to avoid (Smith, 1992, pp. 175-176). Another possibility, almost equally simple and perhaps more convincing, is that what people find aversive during a crisis is a feeling that nothing is happening that has the potential to end the crisis. For example, during the latter stages of the Daily Opinion Change During the Culf Crisis 359 Vietnam conflict, both "hawks" and "doves" experienced mounting frustration as the war dragged on with no end in sight. Large segments of the public wanted the war to end and were willing to give the administration a broad mandate to choose among various means, tough or conciliatory, of ending it (Russett, 1990-1991, p. 517; Robinson & Jacobson, 1969). Numerous historical episodes suggest that the can-do American mentality lends itself to an *all out or get out" perspective that is not easily reconciled to limited hostilities and temporizing measures. Applied to the Culf crisis, this idea suggests that, relative to its response to Inaction, the public may have have reacted favorably to reports of any U.S. action, diplomatic or military, and to reports of Iraqi-initiated diplomatic action (but not, of course, to reports of Iraqi-initiated military action, which would have been interpreted as prolonging the crisis). According to either of these two ideas, the public responds on a given day to the events of the preceding day: whenever the public receives a certain amount of good or bad news {however defined), it adjusts its views accordingly. Realistically, though, the events of each new day build on the events of ail the preceding days, and on this longer time horizon the events of any particular day may comprise a relatively minor part of the action. Thus, for example, during the active phase of U.S. military involvement in Vietnam, there was no connection between the number of American casualties in a given month and public support for the war during that month or the month following; that is, public support did not rise and fall as a function of monthly fluctuations in the casualty count. But this does not mean that the public was insensitive to American casualties, for the rising cumulative tide of casualties over time in Vietnam was closely related to declining public support for the war {Mueller, 1973). By the same token, public opinion during the Gulf crisis may have moved less in response to particular events that occurred on a given day than in response to the accumulation of events dating back to the first day of the crisis. As the principals in the crisis undertook more and more actions without resolving the crisis or even seeming to move it perceptibly closer to resolution, public impatience would naturally mount, and the accretion of events that were not resolving the underlying crisis would itself be interpreted as progressively bad news. It follows that approval of President Bush's handling of the situation, endorsement of U.S. actions as 'just about right," and anticipation of a negotiated settlement should have declined as the number of crisis events accumulated from day to day. The daily accumulation of crisis events is measured here simply as the total number of U.S.- or Iraqi-initiated diplomatic and military events recorded prior to the events of a given day, from August 2 through December 30.^ Statistical Results The statistical results are summarized in Table 1. Each set of estimates has certain nuances, but overall these estimates point in the same direction. Mean daily approval of President Bush's handling of the situation can be fairly accurately predicted on the basis of the variables in the model.^ Some 62 percent of the daily variance in these ratings is accounted for by the variables in the model, but none of the daily events variables registers a significant impact. That is, evaluations of Bush's performance did not fluctuate from one day to the next as a function of the occurrence of U.S.- or Iraqi-initiated military or diplomatic actions. Bush's ratings did display some tendency to rise after the U.S. undertook a new action and 360 Lee Sigelman et a/. Table 1 Summary of Statistical Results Approval of Bush's Handling Daily events U.S. military Approval of U.S. Actions Expectation of Peaceful Settlement .013 .70 .462 .68 -.761 -.86 .003 .13 .863 1.09 -.313 -.30 -.043 -1.10 -.325 -.23 -2.338 -1.24 Iraqi diplomatic -.014 -.64 -.024 -.03 .527 .46 Cumulative events -.015 -6.76 -.529 -8.65 -.076 -2.77 U.S. diplomatic Iraqi militaty Cumulative events (squared) Constant R= Number of cases Standard error of the estimate Mean of dependent variable Durbin-Watson statistic probability .0001 5.57 .002 7.15 2.910 22.68 86.860 24.98 43.698 12.63 .619 109 .143 1.93 2.11 >.O5 .572 109 5.10 53.79 2.10 >.O5 .320 101 7.21 34.01 2.08 .63 to fall after Iraq did the same, but these tendencies were not nearly predictable enough to be statistically significant. On the other hand, the cumulative growth of U.S. and Iraqi actions was strongly associated with Bush's ratings, which declined— albeit at a diminishing rate—as the two sides initiated more and more actions without bringing the crisis to an end. From the start of the tracking poll period. Bush's mean daily ratings consistently fell as the cumulative number of events rose, finally bottoming out after the accumulated stock of U.S. and Iraqi events had reached 150. At that point Bush's ratings began to inch gradually upward, but by the end of the crisis period the rebound was not nearly sufficient to have restored the level of public support Bush had enjoyed before the crisis became a stalemate. The second column of Table 1, which gives the results for the daily level of Daily Opinion Change During the Gulf Crisis 361 agreement that U.S. actions during the crisis had been "just about right," tells virtually the same story. The percentage of respondents who described U.S. actions as "just about right" on a given day did not rise or fall significantly as a consequence of newly initiated U.S. or Iraqi military or diplomatic events. What decisively shaped daily affirmations of U.S. policy was the stage the crisis had reached by a certain day. Early in the crisis, the addition of just two events to the cumulative total was associated with a 1-point drop in the percentage of respondents who considered U.S. actions *ju5t about right"—a very substantial effect. By the middle of the crisis period, it took the addition of five new events to the cumulative total to produce the same 1-point drop, and thereafter the pace slowed even further until very late in the crisis period, at which point it stopped altogether and did not move further in either direction. Clearly, then, the performance of the quadratic term does not contradict the idea that prompted us to employ a cumulative term in the first place; it simply helps specify the character of the cumulative effect, which was to produce decline in approval of U.S. actions at an ever-declining rate until approval reached a floor beneath which it moved no further. We saw in Figure 1{c) that optimism that the crisis would end peacefully rather than leading to war fluctuated fairly erratically from day to day. It thus comes as no great surprise to see in the third column of Table 1 that the fit of the model for expectations of a negotiated settlement is hardly precise (R' - .32)." In other respects, though, the estimates in the third column resemble their counterparts in the first and second columns. Again, none of the daily event variables has a significant effect on positive assessments, and all that really matters is the cumulative events variable. The coefficient of -.076 for the cumulative events variable means that the percentage of respondents who expressed optimism about a peaceful resolution of the crisis declined by roughly 1 point for every addition of 13 events to the cumulative total of events. Over the full course of the tracking poll period, a 10 percentage point drop in expectations of a negotiated settlement can be attributed to the progressive deepening of the crisis. Discussion What do these findings suggest about opinion dynamics during an international crisis? Guiding our analyses were two "good news/bad news" interpretations of daily opinion dynamics. The first posited that the public would respond favorably after events that betokened a peaceful resolution of the crisis and unfavorably after escalatory actions, and the second posited that during a crisis situation any U.S. action, diplomatic or military, would be more warmly received than inaction was. The lack of any statistically significant coefficients in Table 1 for the daily events variables means that neither U.S.- nor Iraqi-initiated diplomatic or military events were consistently followed by movements in either direction in support for U.S. actions, endorsement of President Bush's handling of the situation, or optimism about the prospects for peace. These results obviously provide no indication that good or bad news, understood as news of conciliation or conflict, respectively, had any immediate bearing on public assessments of the crisis. The signs of almost all the coefficients for the daily events variables are as predicted by the other "good news/bad news" interpretation, but these coefficients uniformly fall short of statistical significance. Thus it seems fair to characterize an "impatience with stalemate" interpretation of good news and bad news as somewhat more consistent with the 362 Lee Sigelman et al. data than a "conflict avoidance" interpretation, but the main point, again, is simply that the statistical results provide no significant support for either of these perspectives on the effect of daily events.' From this one might draw the substantive lesson that, even during a period of extraordinary tension when the public's attention is drawn to an unusual degree to the international arena, public opinion does not consistently move from day to day in a manner that seems to follow consistently from the character of the events that are transpiring. That is, even though one might concur with those who contend that the American public tends to respond in a reasonable fashion to the broad tenor of national and international affairs (Shapiro & Page, 1988), one might still concede that it is unrealistic to expect the general public to closely monitor and carefully weigh each day's developments. It follows that analysis of the responsiveness of the "attentive public" to daily events might turn up stronger linkages than those isolated here. A related possibility is that changes consistent with the interpretations advanced above may have occurred within certain demographic subgroups, but in a manner that canceled out over the general public. Perhaps, for example, women greeted conciliatory events as good news, while men viewed military actions more favorably; or perhaps members of different social strata reacted differently to different types of events. Such effects, which are consistent with demographic differences isolated in prior studies of U.S. involvement in conflict situations (see, for example, Fite, Genest, & Wilcox, 1990), could have offset one another, reducing aggregate daily trends to fluctuations induced by sampling error. However, because the daily tracking poll samples were so small in the first place, we were reluctant to subdivide them in order to test models of the responsiveness of certain portions of the public to daily events. Analysis of daily changes in subgroup opinions must await the availability of tracking polls with larger daily samples. Alternatively, one might draw the narrower lesson that even if the public does not monitor and weigh all the events undertaken by the principal opponents in a crisis, it could still respond to events that seem especially momentous. We cannot reject this possibility out of hand. Indeed, we consider it quite plausible. However, testing it will necessitate a rather different, more quasi-experimental type of research design than the one employed here. Such a design would immediately confront the difficult problem of how to specify, a priori, which events were momentous and which were not, according to criteria that do not, in circular fashion, take actual or anticipated public opinion into account.'° That is a formidable challenge and one that would carry us far afield from the focus of the present study. This brings us to the cumulative effects of events during the Culf crisis. Here our guiding idea was that the deepening of U.S. involvement in an ongoing international crisis, rather than the occurrence of any specific crisis events, would be the key to understanding the public's response to the crisis. During World War II the American public established its willingness to support a long-term military effort (Smith, 1971), and on numerous occasions it has rallied behind limited military engagements of a rapid, "surgical" character; for example, approval of President Reagan shot up after the successful American foray into Grenada in 1983, and President Bush's popularity soared to historic levels early in 1991 when the U.S.-led counteroffensive encountered unexpectedly mild Iraqi resistance. However, a tense international situation that becomes stalemated can be a very different matter. As stated earlier, public support for American military involvement in Vietnam eroded during the late 1960s and early 1970s. Something similar oc- Daily Opiriion Change During the Culf Crisis 363 curred during the year after the takeover of the American embassy in Tehran in late 1979, when virtually everything the Carter administration tried, including an abortive hostage rescue attempt, failed. At the time of the takeover, Carter's standing in the polls shot up, but as the crisis dragged on the public's patience wore thin and support for Carter plummeted from 69 percent in December to 36 percent in October (Sigelman & Conover, 1981). Brody (1992), analyzing entirely different data than those considered here, has recently concluded that in the early stages of the Gulf crisis, the cumulative balance of positive and negative news had no significant bearing on public support for President Bush, but that as the crisis wore on the cumulative news balance did come to exert a significant effect on evaluations of Bush's handling of the crisis. Our own main findings are not identical to Brody's, but sound the same general theme. That is, as the crisis in the Persian Gulf continued unresolved, our data analyses suggest that the progressive accumulation of U.S. and Iraqi military and diplomatic moves and countermoves itself came to be regarded as bad news. The further the two nations slipped into the crisis, the more negative the American public became about U.S. actions, President Bush's handling of the situation, and the prospects for peace. Dissatisfaction with American policy and with President Bush's handling of the situation did not continue indefinitely on a downward arc. Rather, after the crisis had deepened to a certain point, further U.S.- and Iraqi-initiated diplomatic and military events depressed public support no further; apparently, just as was true of opinion during the Vietnam era, hard-core backers of the president and of American policy in effect erected a floor underneath which support did not fait. Some might interpret the strong effect of the cumulative events variable merely as an indication that nothing but the passage of time really mattered." The basis of this assertion is the observation that, as Figure 2 makes clear, military and diplomatic events accumulated as a linear function of time. That much is undeniable. Diplomatic Events Military Events 300 - 250 - 200 - i 5 300- 50 20 40 60 BO 100 120 Day Figure 2. Running total of events, August 22-December 30, 1990. 364 Lee Sigelman et al. but we see the issue quite differently: the strong effect of the cumulative events variable on public opinion, we believe, provides a plausible—indeed, a compelling—answer to the question of why time mattered. We freely concede that timedependent processes, other than the accumulation of U.S.- and Iraqi-initiated events, might have produced the opinion timelines shown in Figure 1. But, having conceded this as a possibility, we must add that we know of no other time-dependent process that provides an equally compelling account of why American public opinion evolved as it did during the crisis period. During the Culf crisis, we believe, opinions changed over time—criticism grew, support declined, optimism faded— not merely because time went by, but because each passing day was marked by more and more actions that failed to produce the "good news" the public longed to hear, the news that the crisis was over. The data we have examined are highly consistent with this interpretation; although they could be consistent with other interpretations as well, we believe that a cumulative events-based interpretation provides the best account of the opinion trends we observed. It is well understood that the president possesses a unique capacity to shape public opinion, particularly in times of international crisis. The findings reported here suggest that this capacity has a down side that can be no less important, though it is certainly less well understood. The president can label a situation a crisis and expect it to be immediately understood as such; when he assures the public that the matter is being handled in an appropriate manner he can expect that his assurances will be widely accepted; and when the president conveys a sense of urgency by asserting that rapid, decisive action is required immediately (Bostdorff, 1991), he can generally count on the public to "rally around the flag" (Mueller, 1973; Brody, 1991), But the president's ability to convey such urgency is a twoedged sword. Precisely because of the president's ability to demonize the opposing side and to portray a decisive U.S. response as an urgent national priority, a crisis that is not rapidly resolved will engender frustration over the ineffectual attempts that have been made to resolve it, resentment against those responsible for these failures, and pessimism about whether it will be satisfactorily resolved. In his classic treatment of war moods, Richardson (1948a,b) coined the term "warweariness" to refer to the stage of a conflict when, after a lengthy exchange of hostilities, the public's yearning for an end of the war becomes a (or even the) dominant consideration. The growing restiveness of the American public during the autumn of 1990, like the public's reaction a decade earlier to the Iranian hostage crisis, indicates that such weariness is not restricted to shooting wars. The extraordinary attention the public brings to bear on a new international crisis and the unusual tension such a crisis generates will, our findings suggest, eventually give way to "crisis weariness" if a crisis continues unchecked. Indeed, presidents themselves may Inadvertently contribute to such crisis weariness, as an unintended consequence of their success in elevating the nation's collective temperature to a fever pitch. In the weeks following the Iraqi invasion of Kuwait, polling results suggested to President Bush's advisors that the public was not rallying around the themes of democracy and human rights in Kuwait, but that the image of Saddam Hussein as the enemy was evoking a powerful public response. Thus, for the remainder of the pre-war period the White House's public communication strategy centered on reinforcing these sentiments by publicizing Iraqi atrocities and demonizing Saddam Hussein (Manheim, 1992; Schuman & Rieger, 1992; Spellman & Holyoak, 1992). By hammering away on this anti-Saddam theme. Daily Opinion Change During the Gulf Crisis 365 President Bush probably kept public support for himself and for U.S. policy from dropping too far. However, by keeping Saddam Hussein so much in the public eye. Bush also created a gap between his robust rhetoric and the reality of U.S. measures that were not producing the desired result of doing away with Saddam and the threat he posed. Thus a public that had been led into a frame of mind that demanded harsh, punitive action against Saddam Hussein grew restive waiting for American actions to produce tangible results. That public opinion had not been more volatile during the preceding months indicates, we believe, that the public was in a "wait and see" mode, awaiting, with mounting impatience, some action that seemed capable of bringing the crisis to a close. When such an action finally did occur, in the form of the initiation of the air war in mid-January, the public immediately responded in a way that more than restored sagging confidence in President Bush's leadership. In recent American history, some other presidents, having successfully rallied the public behind them in the face of what they declared to be a crisis, have been less fortunate. Notes 1. The Callup data are from Mueller's (1992) catalog of public opinion concerning the Culf War. In five Gallup surveys, a slight variant of this wording was used; "Do you approve or disapprove of the way President Bush is handling the current situation involving Iraq and Kuwait?' The date used here for each Callup survey is the midpoint of the period in which field work was conducted. For the tracking poll, the approval figure is the daily mean on a 0-3 scale, where 0 signifies strong disapproval and 3 means strong approval. For the Callup poll, it is the percentage of respondents who voiced approval of Bush's handling of the situation in a given survey; the Callup question offered only "approve" and 'disapprove* as response options. Neither Gallup nor any other polling organization repeated other questions often enough to permit meaningful comparison with other tracking poll items. 2. This percentage is virtually the complement of the percentage of respondents who considered U.S. policy 'not tough enough'; the correlation between the daily percentage of tracking poll respondents who criticized U.S. actions as 'not tough enough" and the daily percentage who considered these actions 'just about right' is -.90. Only rarely during the tracking poll period did respondents who deemed U.S. policy 'too tough" exceed 10-12 percent of those surveyed. Accordingly, concentrating on the "just about right" response occasions a minimal loss of information. 3. See Mueller (1992) for a comprehensive catalog and description that draws on a wide array of surveys conducted during and after the Culf crisis. 4. On each dimension, an event was actually assigned to as many as four categories; besides diplomatic and military, these were economic and political. 5. The dating of an event proved to be a fairly complex matter. In the first place, the events analyzed here took place in various time zones. For example, the Iraqi invasion began at 2:00 a.m. on August 2 in Kuwait, 7:00 p.m. on August 1 in Washington, DC. To complicate matters further, although most events reported in the Times of a given day occurred the day before, some occurred during the early hours of the day of publication. For present purposes, the crucial consideration is not the date on which an event actually occurred, but the date on which it was reported in the Tf'mes. Thus, in assigning a date to an event, we treated events reported in the Times of a given date as events of that date. 6. It would have been preferable to use a less aggregated measure, or a set of less aggregated measures, such as the cumulative number of Iraqi-initiated military actions, the cumulative number of U.S.-initiated diplomatic actions, and the like. However, the extremely high correlations between the various cumulative time series made this strategy 366 Lee Sigelman et al. impossible; that is, the fact that each separate set of events cumulated over time toward a progressively higher total meant that we could not empirically distinguish which particular cumulative process might be more important and which might be less so. 7. The OLS estimates for this model and for the model of expectations of a peaceful settlement displayed significant firstnarder autocorrelation; the Durbin-Watson statistic for each model fell well outside the zone in which the null hypothesis of no serial autocorrelation can safely be rejected (dw - 1.48 and 1.24, respectively). Accordingly, we estimated the coefficients in these two models via an autocorrelation procedure {AUTO in the SHAZAM package [White et al., 1988]), using the grid-search technique. 8. The number of cases for this analysis is 101 rather than 109, the number for all the other analyses. Seven cases are lost because the expectations question was not asked during the first week of the tracking poll, and one additional case must be dropped because its value on the dependent variable (59 percent late in the crisis period) is such an extreme outlier that it inordinately influences the regression estimates. 9. We experimented with a wide array of alternative models that incorporated various leads and lags into the linkages between daily events and opinion, and we applied various weight and discount functions to the daily events variables to account for potential increases or decreases in the effect of these variables over time. None of these alternative models produced materially different results than those reported in Table 1. 10. Mueller's (1973, pp. 208-208) comments on this problem are highly instructive. 11. This is the way Stimson (1976) stated the issue in his analysis of the dynamics of presidential popularity. References ' Allsop Dee, & Herbert F. Weisberg. {1988). Measuring change in party identification in an election campaign. Americar) lourr^al of Political Science 32, 996-1017. Bostdorff, Denise M. {1991). The presidency and promoted crisis: Reagan, Grenada, and issue management. Presidential Studies Quarterly 21, 737-750. Brody, Richard A. (1991). Assessing the president: The media, elite opinion, and public support. Stanford, CA: Stanford University Press. Brody, Richart A. (1992). Crisis, war, and public opinion: The media and public support for the president in two phases of the confrontation in the Persian Gulf. Paper prepared for the News Media and Foreign Policy Working Group, Social Science Research Council, Washington, DC {December 11-13). Dorman, William A., & Steven Livingston. {1992). Historical content and the news: The case of the 1990-91 Persian Gulf crisis. Paper prepared for the News Media and Foreign Policy Working Croup, Social Science Research Council, Washington, DC (December 11-13). Fite, David, Marc Genest, & Clyde Wilcox. {1990). Gender differences in foreign policy attitudes: A longitudinal analysis. American Politics Quarterly 18, 492-513. Kernell, Samuel. {1978). Explaining presidential popularity. American Political Science Review 72, 506-522. Manheim, Jarol B. (1992). Strategic public diplomacy: Managing Kuwait's image during the Gulf conflict. Paper prepared for the News Media and Foreign Policy Working Group, Social Science Research Council, Washington, DC {December 11-13). Mueller, John E. {1973). War, presidents and public opinion. New York: Wiley. Mueller, John E. (1992). American public opinion and the Gulf war: Trends and historical comparisons. Paper prepared for the annual meeting of the Midwest Political Science Association, Chicago (April 9-11). Richarson, Lewis F. {1948a). War-moods: I. Psychometrika 13, 147-174. Richardson, Lewis F. {1948b). War-moods: II. Psychometrika 13, 197-232. Robinson, John P., & Solomon C. Jacobson. (1969). American public opinion about Viet- Daily Opiriion Change During the CulfCrisis 367 nam. In Walter Isard (Ed.), Vietnam: Some basic issues and alternatives (pp. 63-79). Cambridge: Schenkman. Russett, Bruce. (1990-1991). Doves, hawks, and U.S. public opinion. Political Science Quarterly 705, 515-538. Schuman, Howard, & Cheryl Rieger. (1992). Historical analogies, generational effects, and attitudes toward war. American Sociological Review 57, 315-326. Shapiro, Robert Y., & Benjamin I. Page. (19B8). Foreign policy and the rational public. journal of Conflia Resolutior) 32, 211-247. Sigelman, Lee, & Pamela J. Conover. (1981). The dynamics of presidential support during intemational conflict situations: The Iranian hostage crisis. Political Behavior, 3, 303318. Smith, Jean E. (1992). George Bush's war. New York: Henry Holt. Smith, Robert B. (1971). Disaffection, delegitimatlon, and consequences: Aggregate trends for World War II, Korea and Vietnam. In Charles C. Moskos, Jr. (Ed.), Public opinion and the military establishmerit (pp. 221-251). 8everely Hills, CA: Sage. 5pellman, Barbara A., & Keith J. Holyoak. (1992). If Saddam is Hitler, then who is George Bush? Analogical mapping between systems of social roles, journal of Personality and Social Psychology 62, 913-933. Stimson, James A. (1976). Public support for American presidents: A cyclical model. Public Opinion Quarterly 40, 1-21. White, Kenneth J., Shirley A. Haun, Nancy G. Horsman, & S. Donna Wong. (1988). SHAZAM econometrics computer program user's reference manual version 6. h New York: McGrawHill.