United States Public Response to Terrorism: Fault Lines or Bedrock?

Transcription

1 Policy Studies Organization From the SelectedWorks of Hank C Jenkins-Smith 2005 United States Public Response to Terrorism: Fault Lines or Bedrock? Hank C Jenkins-Smith, University of Oklahoma Kerry G Herron Available at:

2 United States Public Response to Terrorism: Fault Lines or Bedrock? Hank C. Jenkins-Smith and Kerry G. Herron Texas A&M University Abstract We test traditional assumptions about the volatility of mass opinion in times of national crises using data about views of terrorism from national surveys of the United States general public in 1995 and 1997, findings from a national survey immediately following the terrorist attacks of September 11, 2001 (9/11), and panel data from a follow-up survey in We compare public assessments of the threat of terrorism, willingness to restrict speech to prevent terrorism, support for employing conventional military force against countries that support terrorism, and levels of certainty about culpability required prior to using military force. Results show stable and measured public views prior to 9/11, immediately following the events of that date, and in the subsequent year. Our findings support democratic and modernist theories of public capacities while challenging long-standing traditional precepts about widespread volatility of mass public opinion. Americans were engaged at a visceral level by the terrorist attacks of September 11, 2001 (9/11), and subsequently by American-led combat operations in Afghanistan and Iraq. Political leaders have committed the American people to a long-term war against terrorism, reorganized the federal government, and engaged in trade-offs between individual prerogatives and measures to enhance domestic security. These circumstances provide a unique opportunity to study the nature of evolving public opinion as Americans adapt to the most deadly attack ever against their homeland and organize for a prolonged struggle against terrorism. Did public reactions in the first year following 9/11 expose deep fault lines along which public opinion was fractured, or did reactions to these unique events illustrate belief structures that provide an underlying bedrock for American opinion? Democratic theorists long have debated public capacities and the role of public opinion in times of national crises. By investigating immediate public reactions to 9/11 and how those reactions evolved in the first 12 months of the war on terrorism, we seek to contribute to the continuing study of public opinion and democratic theory. Modern concepts of public opinion in the United States and its proper role in policy processes evolved to an elitist perspective by the middle of the 20th century characterized by the following propositions. (1) Most citizens lack both the cognitive capacities to understand politically sophisticated and technically complicated issues, and they lack sufficient factual information (knowledge) about most policy issues to develop reasoned preferences. (2) Underlying dispositions among the general public are insufficient to systematically structure and constrain views in complex policy domains, especially foreign and security policies. (3) Public opinion at the individual level is unstable, subject to rapid swings, and susceptible to overeaction that is detrimental to coherent, sustained policy. These concepts (combined with criticisms of the validity of survey research methods) gained wide acceptance among political sophisticates, theorists, and Review of Policy Research, Volume 22, Number 5 (2005) 2005 by The Policy Studies Organization. All rights reserved.

3 600 Hank C. Jenkins-Smith and Kerry G. Herron academics, and, together, helped shape the traditional view of limited public capacities that continues to inform many contemporary assumptions about what elites should expect from the general public. Arguments for the traditionalist perspective have been advanced, among others, by Almond (1950/1960, 1956), Bailey (1948), Converse (1964, 1970), Kennan (1951), Lippmann (1922, 1925), Markel (1949), Morgenthau (1948), Weissberg (2001), and Zaller (1992). By the latter part of the 20th century, traditionalist concepts came under critical scrutiny by a growing body of revisionists. Challenges were published in findings by Achen (1975), Caspary (1970), Chittick, Billingsley, and Travis (1995), Herron and Jenkins-Smith (2002), Holsti (1996), Hurwitz and Peffley (1987, 1990), Oldendick and Bardes (1982), Page and Barabas (2000), Page and Shapiro (1992), Peffley and Hurwitz (1985), Shapiro and Page (1988, 1994), Sniderman, Brody, and Tetlock (1991), Sniderman and Tetlock (1986), Wittkopf (1981, 1986, 1990, 1994), and Wittkopf and Maggiotto (1983). Countervailing propositions suggested by this body of revisionist research include the following. (1) Complete information is not required for reasoned policy choice. Ordinary citizens are cognitive misers who choose to disregard most of the information they could acquire while retaining information considered most important to their individual and collective welfare. Selective information processing is an attribute of rationality and cognitive sophistication, not a result of limited capacities to comprehend. 2 (2) Reasonably well-defined belief systems and complex heuristics support and constrain mass public opinions in both domestic and foreign policy domains in much the same ways in which they provide structure and coherence to views of elite publics. (3) While over time variation in individual opinions and issue salience is evident, such variation should not be equated with broadly unstable and volatile public opinion. Longitudinal analyses of aggregate public views show stability in collective opinions that are related to underlying beliefs or that are events-driven. Our Evidence To test the traditionalist premise of overreaction and volatility in mass opinion, we conducted a national telephone survey of 935 randomly selected members of the United States general public between September 12 and November 4, A follow-up panel survey of 474 of those respondents who agreed to be recalled was conducted one year later between September 12 and November 11, Both surveys included a number of questions about terrorism, some of which previously had been asked of separate national samples in 1995 and By comparing general views long before 9/11 with those immediately following 9/11 and those among the same respondents one year later, we provide benchmarks reflecting how members of the public are adapting to the aftermath of 9/11 and its implications for national and international security. Our data also provide evidence relevant to continuing debate about the stability of United States public opinion under stress, which is our primary focus in this article. The Debate About Public Volatility The Federalist Papers are replete with cautions and structural remedies for the passions of the public (Hamilton, No. 15, 110; Hamilton No. 71, 432; Madison,

4 United States Public Response to Terrorism 601 No. 50, 317 4, ; Madison No. 52, ; Madison No. 63, ). From the Founding Fathers to today, traditional concepts of public capacities have warned against the unreliable and unstable nature of mass opinion, especially in times of national crises. Commenting on the limitations of public opinion in the post-world War II era, Almond (1950/1960, p. 239) stated: Perhaps the gravest general problem confronting policymakers is that of the instability of mass moods, and cyclical fluctuations which stand in the way of policy stability. Walter Lippmann (1955, p. 20), a noted critic of public overreaction, offered the following observation of public opinion and its influence on public policy. The unhappy truth is that the prevailing public opinion has been destructively wrong at the critical junctures. The people have imposed a veto upon the judgments of informed and responsible officials. They have compelled the governments, which usually knew what would have been wiser, or was necessary, or was more expedient, to be too late with too little, or too long with too much, too pacifist in peace and too bellicose in war, too neutralist or appeasing in negotiation or too intransigent.... It [public opinion] has shown itself to be a dangerous master of decisions when the stakes are life and death. Also commenting on the implications for policymakers, Almond (1956, p. 376) made the following criticism of shifting public moods: For persons responsible for the making of security policy these mood impacts of the mass public have a highly irrational effect. Often public opinion is apathetic when it should be concerned, and panicky when it should be calm. 7 Converse (1964, 1975) found public opinion to be so unstructured and unconstrained as to be devoid of consistency over time. One implication that reasonably can be drawn from Converse s research (though not explicitly stated) is that the lack of longitudinal stability of public opinion makes it susceptible to perturbations associated with exogenous and temporary events (Holsti, 1996). Zaller (1992) suggested that members of the general public have so many different temporary constructions of views and preferences that instability of opinion (as indicated by varying responses to opinion survey questions over time) reflects the fact that different considerations are salient at different times, and that only the more well-informed and capable members of the public (elites) consistently constrain policy choices in accordance with organized and integrated sets of beliefs. Implicit in Zaller s construct are the detrimental effects of volatility in the absence of well-defined constraints (coherent belief structures and values) among members of the general public. Mayer (1993) cautioned that the public s immediate reaction to a crisis event may be very different from its considered, long-term judgment, and that changes in public attitudes produced by some crises wear off very quickly when the crises end. The implication is a warning against public overreaction in times of national stress. Predictably, revisionists take exception to traditional models of the volatile public. In 1970, Caspary challenged both Almond s premise (low and unstable public attention to foreign affairs) and Almond s conclusions (unstable support for foreign policy commitments). Analyzing a broader set of foreign policy questions, Caspary (1970, p. 546) concluded that American Public Opinion is characterized by a strong and stable permissive mood toward international involvements. Mueller (1973) found that declining public support for and increasing resistance to United States participation in the conflicts in Korea and Vietnam were systematically related to increasing battle deaths, reflecting a logical connection to a critical measure of associated costs. Achen (1975) concluded that when rigorous

5 602 Hank C. Jenkins-Smith and Kerry G. Herron methodologies are applied, evidence shows that the United States general public s foreign policy views are roughly as stable as those on domestic policy. Using a panel study, Peffley and Hurwitz (1992) found that dramatic changes in foreign policy events (the rapidly evolving United States Soviet relationship after the arms summits of 1987 and 1988) had predictable and reasonable effects on American public opinion, but they also uncovered evidence of impressive stability in respondents images of the Soviet Union and public preferences for specific foreign policies. But perhaps the largest and most comprehensive macro analysis of empirical data on the stability of mass attitudes about foreign policy was conducted by Page and Shapiro (1992) and Shapiro and Page (1994). Using responses to thousands of survey questions posed in dozens of opinion surveys collected between the 1930s and 1991, they found ample evidence of a rational public with highly stable views on domestic and foreign policy issues. Stimson (1991, 2002) and Erikson, Mackuen, and Stimson (2002) applied an economic model and a macro approach involving multiple survey questions over many years and found the public to have relatively stable preferences for more government or less government, and that the preferences moved predictably over time relative to inflation and unemployment. They also found that while change among the better educated may be somewhat more rapid and of greater magnitude, members of the general public with lower socioeconomic attributes showed controlled movement in the same directions under the same stimuli as elites. Parallel to the policy-oriented discussions about the stability or volatility of mass opinions, a related debate is underway about the broader health of democratic processes, the degree to which they are in trouble, and what might be done to strengthen them. In this debate, Reason is commonly portrayed as a fragile force for progress, justice, and greater democracy, which requires protection against the intrusive and destructive impulse of emotion (Marcus, 2002, p. 7). Language and political communication, deliberative processes, rational choice theory, decision making, media influence, and the role of elites are key elements of this discussion. The central theme is that if a more perfect and equitable space for public political discourse can be secured, a sometimes less than rational public can be helped (by elites) to derive rational policies (Habermas, 1979, 1984). Fishkin (1991) suggested citizen juries, citizen conferences, and other specialized public fora to inform citizens of the scope of the issues and help educate them to the rudimentary facts in play. Others (Gutmann, 1987; Nie, Junn, & Stehlik-Barry, 1996) advocated a better citizenry by improving reasoning abilities. Delli Carpini and Keeter (1996) contended that one essential component for improving democratic governance is to increase the political knowledge of the electorate. Still others (Warren, 1996) prefer to emphasize the ways authority is best exercised by informed elites. Much research effort has gone into investigating how the public makes political decisions ( Jackson & Marcus, 1975; Krouse & Marcus, 1984; Lodge, Steenbergen, & Brau, 1995, Marcus & Hanson, 1993; Sniderman et al., 1991; Thompson, 1970). Another central focus has been to investigate what constitutes rational decision making and which institutional arrangements best support it (Aldrich, 1993; Foster, 1984; Rabinowitz & MacDonald, 1989; Riker, 1982). While the debate is too wideranging and multidimensional to survey comprehensively here, one element is most relevant to our line of investigation: the relationship between reason and

6 United States Public Response to Terrorism 603 emotion is key to the debate about the future of democratic discourse and to the issue of volatility in mass opinion. Marcus (2002) held that the traditional construct in which reason and emotion are juxtaposed as polar opposites whose individual existence depends on the absence of the other (zero sum) is wrong conceptually. He argued that publics are able to be rational, in part, because they also have emotions. Rather than preventing rationality, emotions enable rationality, thus the effort to exclude passion from politics will undermine public capabilities to reason. 8 While differentiating between positive emotions (such as compassion, tolerance, and generosity) and negative emotions (such as hate, intolerance, and greed), Marcus also noted that emotion not reason is the well-spring of public enthusiasm and the motivation for political action. In the complete absence of emotion, pure reason would almost always result in the status quo, because change depends not only on rational calculations, but emotional commitment to bring about the changes they imply. According to Marcus (2002), public anxiety, such as that deriving from the traumatic events of 9/11, promotes immediate learning while diminishing reliance on the previously learned. Old habits (and old policies) did not provide the desired security, or else 9/11 could not have happened; new information is needed to reevaluate the threat and understand the changed security environment. Under such conditions, Marcus argued that anxious voters may exhibit less habitual associations and may rely less on partisanship and other familiar heuristics while seeking new and more satisfying explanations of the changed circumstances. The acquisition and weighing of information also can be affected by anxiety. Marcus and MacKuen (1993) noted that anxious voters learn far more about candidate positions than complacent voters. The level of accurate political information has been cited as a shortcoming of mass publics (Delli Carpini & Keeter, 1996), and the effects of anxiety arising from events like 9/11 are likely to stimulate public learning and information gathering. Though anxiety does not, in itself, produce any specific decision, it profoundly affects the way people go about deciding. Schwarz and Bless (1991) and Bless (2000) reported that different affective states require different information processing strategies. They found that individuals in positive affective states rely more on general knowledge structures and familiar heuristics, while persons in negative affective states (anxious, fearful) rely less on general knowledge structures and are more likely to focus on data applicable to the situation. Thus in the absence of anxiety, complacent reliance on habit pertains, but under conditions of anxiety, publics are more willing to consider new alternatives. According to Marcus (2002), rather than fueling instability and volatility, anxiety promotes and stimulates rational processes. At the same time, the range of policy changes available for consideration can vary widely and have large implications for open societies. Anxious publics may be willing to consider new and different policy options that would not have been deliberated under more familiar conditions of status quo and complacency. The terrorist attacks of 9/11 and the ongoing war on terrorism provide the kind of security events, foreign policy challenges, and policy choices that are at the heart of the debate about public capacities. No more singularly distinctive event since the demise of the Soviet Union has received greater media attention, created more

7 604 Hank C. Jenkins-Smith and Kerry G. Herron nd WTC Bombing 9/11/ # st WTC Bombing Olympic Bombing Oklahoma City Bombing Bombing of U.S. Embassies in Kenya and Tanzania 0 1/91 1/92 1/93 1/94 1/95 1/96 1/97 1/98 1/99 1/00 1/01 1/02 9/02 Month/Year Figure 1. Monthly Reporting on Terrorist Incidents: public anxiety, and posed such a wide range of options for policy change. As is evident in Figure 1, the sheer quantitative deluge of news items about terrorism following 9/11 stands in striking contrast to prior levels even following earlier terror events such as the Oklahoma City bombing. The raw materials for public reactions are amply present for evaluating the content, coherence and stability of the public s perceptions and reactions. Contrasting Traditionalist and Revisionist Expectations of Public Reactions Given the nature of the events, media attention, and the ensuing national debate about how to respond to 9/11 and how to prevent future such attacks, traditionalist theories would anticipate widespread fear and volatile public moods, with mass opinion swinging widely among unrealistic assessments of the terrorist threat and its implications. In the traditionalist framework, high levels of public anxiety and a sense of vulnerability will inhibit rational calculation and evaluation of policy alternatives. Publics will be vulnerable to overreactive and demagogic policies that curtail individual freedoms, justified on the altar of enhanced security. Public demand for retribution and the employment of coercive force will pressure government to consider rapid military responses. In contrast, revisionist theories about public opinion under such circumstances expect public assessments of the threat of terrorism to increase more realistically and without widespread fear and panic. Initial concerns and threat characterizations will dampen within the near-term (depending on subsequent events and actions). Citizens will be less likely to rely solely on familiar heuristics such as ideology and partisanship and more likely to rationally assess changed circumstances. Adjustments (if any) to core beliefs about relationships between government and citizens, such as restrictions on individual prerogatives, will be temporary and transitory, leaving basic beliefs unchanged. In the absence of subsequent large-scale terrorist attacks, revisionist theory predicts declining tolerance and decreasing support over time for restrictive or intrusive domestic security measures. Public

8 United States Public Response to Terrorism 605 Table 1. Expectations of Public Opinion Following 9/11 Traditionalist Expectations Widespread fear and panic; threat assessments veer between extremes. Anxiety and vulnerability to group-think prevent objective analysis. Publics vulnerable to reactive policies curtailing rights and freedoms; individual prerogatives surrendered for security. Public demands retribution; widespread support for overwhelming punitive force. Revisionist Expectations Opinion responsive to events, but without panic; controlled and rational threat assessments. Less dependence on traditional heuristics; more willingness to examine new evidence. Initial support for intrusive policies to enhance security quickly dampens. Core beliefs about individual rights do not change. Support for use of military force initially increases, but soon subsides toward pre-9/11 levels. support for using United States military force to respond initially will be substantial, but only under limited circumstances (high degree of certainty of culpability and minimal collateral damages), and initial support will subside in the near term. These arguments are summarized in Table 1. To test these contrasting propositions about the volatility of public opinion under conditions of national stress, we conduct two kinds of analyses. In Data Analysis Section 1, we examine our full sample of 935 members of the public who responded in 2001 to our survey beginning the day after 9/11. By comparing responses to key questions relating to terrorist threat assessments, willingness to consider restrictions on free speech, support for using military force to respond, and certainty of culpability required before responding with force, we can compare public reactions at 10-day intervals across the first 54 days (duration of data collection) following 9/11. 9 This method allows us to examine volatility among our respondents during what arguably was the most fluid and changing period immediately following the attacks. In Data Analysis Section 2, we focus on the 474 panel members who participated both in 2001 and 2002 to examine changes over the intervening year among these same key issues by respondent age, gender, education, income, ethnic minority status, ideology, and political partisanship. Dependent Variables We created the following five metrics for use as dependent variables in data analyses. See the Appendix for lead-ins and complete question wording. Terrorist Threat Index Using a continuous scale where zero meant no threat, and ten meant extreme threat, we created a composite terrorist threat index by averaging equally weighted responses to questions asking participants to rate the current threat posed by all kinds of terrorism (Q71), and the projected threat of all kinds of terrorism over the next ten years (Q72). Freedom of Speech Index On a continuous scale where one meant strongly disagree, and seven meant strongly agree, we asked participants if, in order to help prevent future acts of terrorism,

9 606 Hank C. Jenkins-Smith and Kerry G. Herron they would support banning people from speaking on radio or television if they advocate antigovernment violence (Q80), and if they would support banning information about bomb-making from computer networks (Q81). We averaged equally weighted responses to questions 80 and 81 to form a composite freedom of speech index. Conventional Military Force Index We asked a series of questions about preferences for responding to 9/11 given a high degree of certainty about the following three different levels of state culpability: providing personnel or training for the terrorists (Q95); providing funding for the terrorists (Q96); and providing sanctuary for the terrorists (Q97). For each type of culpability, we asked participants if they would support each of the following five United States response options: (1) applying strong diplomatic and political pressures; (2) applying strong economic and trade sanctions; (3) conducting air strikes using conventionally armed weapons such as bombs and cruise missiles; (4) using United States military forces to invade; and (5) using United States nuclear weapons. By assigning a value of one for each yes response to options three and four, we created a continuous index of willingness to use conventional military force to respond to 9/11 ranging in value from zero (strongly oppose) to six (strongly support). Certainty before Using Military Force Our measure of certainty consists of responses on a scale from zero (not at all certain) to ten (completely certain) to question 98 asking participants how certain the United States government should be about complicity in terrorist attacks against the United States before using military force against a country determined to have supported the terrorists. Data Analysis: Section 1 Our survey in 2001 began on September 12 (the day following the terrorist attacks) and continued for portions of 54 consecutive days and nights, ending on November 4. To measure variation and volatility in responses to each of the four dependent variables described above, we separated our respondents into four sequential increments of 10 days each, followed by the concluding increment of 14 days as shown in Table 2. In that table we also indicate the number and percent of respondents who were surveyed in each of the five time periods, and we compare the mean values for each of our five dependent variables and the overall p-value for analysis of variance (ANOVA) for each. Over the course of the first 54 days following 9/11, combined assessments of the current and future threat of all kinds of terrorism expressed on a scale from zero (no threat) to ten (extreme threat) varied only from a high of 8.57 in the first time period (days 1 10) to a low of 8.20 in the fourth period (days 31 40). Variation across the 54-day collection did not reach statistical significance at the 95% confidence level (p = ). Note also that variation in means across the full 54-day

10 United States Public Response to Terrorism 607 Table 2. Mean Indicators at Post-9/11 Intervals: 2001 T-1 T-2 T-3 T-4 T-5 9/12 9/22 10/02 10/12 10/22 9/21 10/01 10/11 10/21 11/04 ANOVA n % Sample Mean Values p-values Terrorism Index (0 10) Speech Index (1 7) Conventional Force (0 6) Certainty (0 10) period was not statistically significant for our speech index or conventional force index. Both were within a narrow range of variation. Only our measure of the degree of certainty preferred before using United States military force varied significantly from a high mean of 9.04 in the second time period (9/22 10/01) to a low mean of 8.49 in time period four (10/12 10/21) ( p = ). United States and allied bombing against Taliban and Al Qaeda targets in Afghanistan began during our data collection on October 7, We completed 438 interviews prior to the start of United States bombing and other overt military actions in Afghanistan with a mean certainty score of 8.88; the mean for those who responded after United States bombing began in Afghanistan was 8.60 ( p = ). While the magnitude of mean change in absolute terms remains small, the preferred level of certainty before employing United States military force declined significantly after overt military actions began in Afghanistan. The overall picture revealed by these data over the course of the first 54 days following the most deadly attack against the United States in history is not one of public volatility and mood swings that traditionalist theory suggests. Change during this period for four of our composite measures was small and statistically insignificant. Our data suggest that once military actions began in Afghanistan the mean level of certainty required for military operations declined modestly but significantly, which is not surprising, since some respondents may have considered the level of certainty already to have been met. We would note, however, that both prior to and after military operations began in Afghanistan, the expressed level of preferred certainty averaged above 8.5 on a scale from zero to ten. This reflects a more measured and careful view of military retaliation and preemption than traditionalist theory would anticipate. To the contrary, these results support revisionist concepts of public capacities and continuity of opinion in a period of national turmoil and stress. Data Analysis: Section 2 In this section, we continue the analysis of the same dependent variables, but we concentrate on the views of our 474 panel members who were interviewed both in 2001 and We further shift to examining individual-level attributes and their relationships to the issue of opinion volatility. We begin by describing various attributes of our panel, then proceed with a table for each of our four dependent vari-

11 608 Hank C. Jenkins-Smith and Kerry G. Herron ables illustrating unpaired comparisons of means within each measurement period across various demographic and belief measures. Each table also shows paired comparisons of means within the same demographic and belief categories that illustrate change across the two measurement periods (2001 vs. 2002). Demographic Attributes of Our Panel In Table 3, we summarize the distributions and means (where appropriate) of our panel members by age, gender, education, annual household income, ethnic minority/majority status, self-rated political ideology, and political partisanship. As expected, these data show that over the course of the panel waves the survey respondents aged by one year and, on average, became slightly wealthier and attained slightly more education. Shifts in core beliefs, such as political ideology, were statistically unchanged on average. 10 Such core beliefs provide a test of the stability of important underlying beliefs, in that they are assumed to be relatively stable and to provide the hierarchical structure that guides more specific beliefs. In the case of our panel, fully 51.5% of the respondents reported no change in their ideological self-placement between 2001 and 2002, and 88.1% moved one point or less on the seven-point scale (chi-square , p < ). Thus our panel Table 3. Demographic Attributes of Panel Members % 2001 % 2002 Mean 2001 Mean 2002 Age > Gender Women Men Education <College College >Bachelor s Degree Annual Household Income $50 60,000 $50 60,000 $0 50, $51 100, >$100, Ethnic Subgroups Black, Hispanic, Am. Indian All Others Ideology (1 7) Liberal (1 2) Center (3 5) Conservative (6 7) Partisanship Democrat Republican All Others Scale values are as follows: 1 = strongly liberal; 2 = liberal; 3 = slightly liberal; 4 = middle of the road; 5 = slightly conservative; 6 = conservative; 7 = strongly conservative. 2 Those who rated themselves strongly liberal or liberal. 3 Those who rated themselves slightly liberal, middle of the road, or slightly conservative. 4 Those who rated themselves conservative or strongly conservative. 5 Those who identified somewhat or completely with the Democrat party. 6 Those who identified somewhat or completely with the Republican party.

12 United States Public Response to Terrorism 609 showed remarkable stability in this basic belief during the tumultuous year following the September 2001 terrorist strikes. Given that underlying stability in ideological orientation, how did perspectives and preferences concerning terrorism evolve over that period? Comparing Panel Views on the Threat of Terrorism The same two questions whose responses were combined to form our terrorist threat index also were asked in 1997, and they provide a baseline for comparison from well before the events of 9/11. The 1997 survey followed the Oklahoma City bombing incident by two years, and the Olympic Games bombing in Atlanta by one year (see Figure 1). In Table 4, we compare mean values on the terrorist threat index for different demographic and ideological groupings in 1997, 2001, and Comparing the combined overall assessments of the current threat of all kinds of terrorism and the projected threat for the next decade, our panel members registered a predictably steep and significant increase in mean threat assessments from the 6.60 assessed by a previous national sample in 1997 to 8.27 among panel members immediately after 9/11. That increase was followed by a smaller (but statistically significant) decrease to a mean of 8.00 one year later. In terms of subgroup comparisons, note that composite assessments of the current and future threat of terrorism increased substantially across all demographic and ideological categories between our initial measurements in 1997 and our 2001 survey immediately following 9/11. For each subgroup, the increase in mean assessments from 1997 to 2001 had a p-value < In 2002, means generally declined from their highs recorded in 2001, but each measurement for each subgroup in 2002 remained significantly higher than the corresponding values in Looking vertically within each category at the unpaired comparisons of mean values in 1997, notice that variations among some of the subgroups across all categories except ideology were large enough to be statistically significant. Next, note in our 2001 survey immediately following 9/11, that, with one exception, differences in means among panelists in 2001 did not vary significantly within each demographic or ideological category. The single exception was a significantly lower mean threat measurement among liberals compared to ideological centrists (p = ). In all other cases, reduced distinctions in mean threat assessments compared to 1997 may be related to implications of 9/11. Notice also that by the time of our survey in 2002, significant differences between some subgroups again were evident. This finding supports Marcus s (2002) contention, noted above, that publics experiencing high levels of anxiety will rely less on habitual associations such as ideology and other familiar heuristics while rationalizing the changed circumstances. One of the immediate effects of 9/11 may have been a temporary muting of most differences in assessments of the current and future threat of terrorism across all our demographic and ideological categories. Looking horizontally at the paired comparisons within each category of attributes among our panel members in 2001 and 2002, note that (again with one exception) mean assessments of the current and future terrorist threat decreased among each of our subgroups between 2001 and The single exception was

13 610 Hank C. Jenkins-Smith and Kerry G. Herron Table 4. Mean Terrorism Index by Demographic and Ideological Groupings Grouping Means: Means: Means: Change Paired n Variables P-Value 01& 02 All ( 97) Panel ( 01, 02) Age (1) (2) >50 (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p = p = p = p = p = p = Gender Women (0) Men (1) Unpaired Comp 0 1 p < p = p = Education <College (1) College (2) >Bachelor (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p < p = p = p = p = p = Income 0 50 K (1) K (2) >100 K (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p = p = p = p = p = p = Minority Blk/Hisp/AI (1) All Others (0) Unpaired Comp 1 0 p = p = p = Ideology 1 2 Liberal (1) Center (2) Conserv (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p = p = p = p = p = p = Partisanship Democrats (1) < Republicans (2) All Others (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p = p = p = p = p = p = a statistically insignificant increase in threat recorded by respondents in 2002 having annual household incomes greater than $100,000. These data suggest that the events of 9/11 initially increased significantly the awareness of the threat of terrorism across all our demographic and ideological subgroups while also muting distinctions previously and subsequently observable among some of those subgroups. Though assessments of the terrorist threat subsided somewhat among some subgroups during the first year after 9/11, none returned near the levels reported in The rate at which some subgroups

14 United States Public Response to Terrorism 611 reassessed the terrorist threat in the 12 months between the 2001 and 2001 surveys varied significantly, and those changes imply that, to the degree support for policies designed to prevent and counter terrorism are related to public perceptions of the prevailing threat, we also may see initially high levels of policy support become more differentiated by demographic and ideological attributes as time passes after terrorist attacks. It is to those questions we turn next. Comparing Panel Views on the Freedom of Speech Index Both questions comprising our freedom of speech index first were asked in 1995, and those responses provide a pre-9/11 baseline for general comparisons. In Table 5 we compare mean subgroup values for the freedom of speech indices measured in 1995, 2001, and Compared to average combined measurements in 1995 (4.70), mean support for curtailing speech on the radio and the Internet in an effort to prevent terrorism remained statistically unchanged at 4.66 immediately following 9/11. By the time of our survey in 2002, mean support for these kinds of intrusive restrictions declined to 4.44, statistically significantly below both prior measurements in 1995 and Contrary to the cautions of some traditionalists, these data do not show volatility around an exogenous event, and the change is not in the direction traditionalist theory predicts (toward compromising individual prerogatives for the sake of increased security). Several patterns emerge from examining unpaired comparisons across subgroups within each measurement period. In absolute terms, mean support for restrictions to speech for the purposes of preventing terrorism was greatest among those in the highest age group and lowest among those between the ages of 31 and 50 for each of the three surveys. Women were significantly more supportive of restrictions to speech than were men in each survey, and support for restricting speech varied inversely with education in each survey. Those in the lowest income categories expressed the highest support for the proposed policies in each measurement period. On average, those who identified themselves as liberal or strongly liberal were least supportive of the proposed restrictions, while those who categorized themselves as conservative or strongly conservative were most supportive in each of the three time periods. In both post-911 surveys, those who identified somewhat or completely with the Democrat Party were least supportive, on average, while those who identified somewhat or completely with the Republican Party were most supportive of the measures for restricting speech to prevent terrorism. Turning to paired measures of how views among our panel members changed in the first year following 9/11, we note that with the exception of participants with less than college-level educations, respondents in 2002 in all other demographic and ideological categories reported lower mean support for restricting speech to prevent terrorism than was reported one year earlier. Not all changes reached a level of statistical significance, but the trend toward lower acceptance of the suggested measures was clear. Finally, we note the overall stability of views before, immediately following, and a year after 9/11. Unlike the sharply revised assessments of the threat of terrorism previously described, most responses on this index averaged between values of four

15 612 Hank C. Jenkins-Smith and Kerry G. Herron Table 5. Mean Speech Index by Demographic and Ideological Groupings Grouping Change Paired n Variables Means: 1995 Means: 2001 Means: P-Value 01 & 02 All ( 95) Panel ( 01, 02) Age (1) (2) >50 (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p = p = p = p = p = p = Gender Women (0) Men (1) Unpaired Comp 0 1 p < p = p = Education <College (1) College (2) >Bachelor (3) Unpaired 1 2 p < p = p < Comparisons 1 3 p = p < p < p = p = p = Income 0 50 K (1) K (2) >100 K (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p = p = p = p = p = p = Minority Blk/Hisp/AI (1) All Others (0) Unpaired Comp 1 0 p = p = p = Ideology 1 2 Liberal (1) Center (2) Conserv (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p = p = p = p = p = p = Partisanship Dems 2 3 (1) Repubs 2 3 (2) All Others (3) Unpaired 1 2 p = p = p = Comparisons 1 3 p = p = p = p = p = p = and five on a scale from one to seven, suggesting that support for restricting speech to help prevent future terrorism was open to consideration, but was neither heavily favored nor opposed in any of the three periods, especially among those respondents from higher socioeconomic groups. In cases where support for restricting speech initially increased immediately following the attacks, it often subsided to below pre-9/11 levels by the time of our survey in 2002.

16 United States Public Response to Terrorism 613 Comparing Panel Views on the Use of Conventional Military Force Immediately following the attacks of 9/11, support among our panel members for responding with conventional United States military force averaged well above midscale at 4.37 on a scale from zero to six as shown in Table 6. When we conducted our second survey in the fall of 2002, United States military operations in Afghanistan were underway, but intense combat operations had not yet begun in Iraq. By that time, mean support for conventional military operations against suspected terrorist supporters had declined significantly to a value of This response pattern suggests that, as some traditionalists might argue, public support for the use of force in response to 9/11 was substantially higher in the immediate aftermath of the attacks when public outrage was highest. A year later, support for the use of force had receded significantly. However, the modal score on the conventional military response index was a value of six (highest scale value) in both surveys, and mean support for the use of military force remained above midscale in both surveys, so the change was not from a position of support to one of opposition, but rather a moderation of the level of support present immediately after 9/ Unpaired comparisons across subgroups within the same survey show, both in 2001 and 2002, that women were significantly less supportive of responding to the 9/11 attacks with United States military force than were men. Support for military retaliation also declined in both surveys with increasing levels of education, with increasingly liberal political ideologies, and among those who most closely identified with the Democrat Party. Conversely, support for using airstrikes and military invasion was significantly higher among men, participants with lower levels of education, those with more conservative ideological orientations, and Republican partisans. Support did not systematically vary by age, income level, or minority and majority ethnicities. As to change between the two measurement periods in willingness to sanction the use of conventional military force, mean support for military options declined significantly between 2001 and 2002 within each of our demographic and ideological subgroups. This illustrates that the overall changes in preferences for military options reported in 2002 were consistent in direction and distributed widely across respondent characteristics. These findings do not support traditionalist expectations of public demands for retribution. They reflect, instead, moderate and declining support for military options over time. Comparing Panel Views on Degree of Certainty Required before Using Military Force Our final dependent variable deals with the preferred level of certainty of a country s culpability in terrorist acts before employing United States military force to retaliate or preempt future acts of terrorism (Q98). As shown in Table 7, panel members preferred high levels of certainty in both measurement periods. On a scale from zero to ten, where zero meant not at all certain, and ten meant completely certain, mean certainty increased from 8.79 in 2001 to 8.93 in 2002, but the change was not statistically significant (p = ). No clear pattern of differentiation is

17 614 Hank C. Jenkins-Smith and Kerry G. Herron Table 6. Mean Support for Using Conventional Military Force by Subgroups Grouping Means: Means: Paired n Variables Change P-Value ( 01 & 02) Full Panel < Age (1) (2) < >50 (3) < Unpaired 1 2 p = p = Comparisons 1 3 p = p = p = p = Gender Women (0) < Men (1) < Unpaired Comp 0 1 p < p < Education <College (1) College (2) < >Bachelor (3) < Unpaired 1 2 p = p = Comparisons 1 3 p = p < p = p = Income 0 50 K (1) < K (2) < >100 K (3) < Unpaired Comparisons 1 2 p = p = p = p = p = p = Minority Blk/Hisp/AI (1) All Others (0) < Unpaired Comp 1 0 p = p = Ideology 1 2 Liberal (1) < Center (2) < Conserv (3) < Unpaired 1 2 p < p < Comparisons 1 3 p < p < p = p = Partisanship Dems 2 3 (1) < Repubs 2 3 (2) < All Others (3) Unpaired 1 2 p < p < Comparisons 1 3 p = p = p = p = apparent from unpaired comparisons of means across categories within each survey, and few changes in paired comparisons of means within each category across the two surveys were statistically significant. We draw two points from these data. First, the high mean values reflect preferences for a cautious and reasoned approach in which culpability is established to a high degree of certainty before employing military force. There is no systematic evidence of overreaction or public pressure for vengeance. Second, as previously noted, United States military action in Afghanistan began on October 7, 2001, after 236 (49.8%) of our 474 panelists had responded to the