Can We Frame the Terrorist Threat? Issue Frames, the Perception of Threat, and Opinions on Counterterrorism Policies

Terrorism and Political Violence, 18:545 559, 2006 Copyright Taylor & Francis Group, LLC ISSN: 0954-6553 print=1556-1836 online DOI: 10.1080/09546550600880625 Can We Frame the Terrorist Threat? Issue Frames, the Perception of Threat, and Opinions on Counterterrorism Policies DONALD P. HAIDER-MARKEL, MARK R. JOSLYN, AND MOHAMMAD TAREK AL-BAGHAL Department of Political Science, University of Kansas, Lawrence, KS, USA In political disputes, issue frames set parameters for debate and shape which view dominates. This study expands issue framing research to examine the influence of frames on the perception of future terrorist threats as well as subsequent support for related counterterrorism policies. We test several hypotheses using data from an experimental field poll. We find that issue frames clearly influence perceptions of threat. However, our frames, which posit specific terrorism threats, only have a limited influence on respondent preferences for counter-terrorism policies. We consider a variety of explanations for these results and provide direction for future research. Keywords counterterrorism, framing, opinion, terrorism, threat Following the September 11, 2001, terrorist attacks on the U.S., the Bush administration appeared to rethink its approach to foreign as well as domestic policy. One resulting change was the beginning of a war on terrorism that included increasing homeland security. Many existing problems, both new and old were reframed by the administration as national security threats that must be corrected in order to win the war against terrorism. Indeed, the 2003 war in Iraq was in large part framed as a necessary evil given the threat posed by Iraqi weapons of mass destruction. And public opinion polls strongly suggest that the administration s efforts to frame the broader war on terrorism, as well as an implied role played by Iraq in terrorism, had a significant influence on the public s knowledge of terrorism-related issues. 1 The administration s framing efforts also appear to have influenced public perception of a connection between Iraq and the attacks of September 11, public support Donald P. Haider-Markel is Associate Professor of political science, University of Kansas (dmarkel@ku.edu). Mark R. Joslyn is Associate Professor of political science, University of Kansas (mjoz@ku.edu). Mohammad Tarek Al-Baghal is a Research Assistant, Insitute of Policy and Social Research, University of Kansas (talbaghal@gmail.com). A previous version of this paper was presented at the annual meeting of the American Political Science Association, August 2003. The Policy Research Institute at the University of Kansas supported this research. Address correspondence to Donald P. Haider-Markel, Department of Political Science, 1541 Lilac Lane, 504 Blake Hall, University of Kansas, Lawrence, KS 66044. E-mail: dhmarkel@ ku.edu 545

546 D. P. Haider-Markel, M. R. Joslyn and M. T. Al-Baghal for war in Iraq, and public knowledge concerning the discovery of weapons of mass destruction in Iraq following the war. 2 Meanwhile, media coverage of terrorism-related issues has been extremely high since September 11, 2001, and most Americans paid considerable attention to media coverage of terrorism in the weeks following the attacks. 3 John Zaller 4 and others argue that although the media frame issues in a manner similar to political elites, such frames may reinforce or conflict with those presented by political elites. 5 The media and political elites may then present competing frames, leaving citizens conflicted or confused. But our knowledge of the impact of issue frames is hardly sparse. A considerable body of research has demonstrated the impact of elite frames on public opinion, 6 the use and influence of issue frames by the mass media, 7 and the influence of issue frames on opinion in experimental studies. 8 However, none of these studies has examined the specific issue of terrorism, either in terms of opinions concerning the problem of terrorism or policy solutions to the problem. Given the policy implications of the September 11 attacks, and the efforts by elites to frame the issue, our research addresses this gap in the literature. We use a series of experimental issue frames concerning terrorism embedded in a random sample survey of the adult population in a Midwestern state following the September 11 attacks. Our study specifically examines whether issue frames influence perceptions of potential terrorist targets and subsequent support for related policy proposals designed to thwart terrorist attacks. Our empirical results do indeed suggest that issue frames influence respondents predictions concerning future terrorist attacks. However, the results concerning the impact of frames on related policy proposals is less conclusive and suggest that the impact of frames may be reduced over time and by respondent predispositions towards government action. Issue Framing and Public Opinion Public opinion researchers have long known that question wording and presentation can influence responses to survey questions. And a growing body of literature has found that frames can influence responses. 9 How conditions in society are defined by public actors and how these definitions are received by citizens is of interest simply because elected officials, the media, and interest groups, among others, consistently try to shape how citizens think about issues in an effort to influence public opinion. 10 Furthermore, individuals have a limited cognitive ability to simultaneously consider all of the possible problems in society or all of the potential dimensions of any given issue. 11 Frames, or specific presentations of issues, have the potential to prime certain ways of thinking about an issue, making these considerations more accessible in memory, and subsequently influencing how the receiver understands the given issue. 12 Given that many citizens are often ambivalent about most issues, the presentation of frames by elites can assist citizens in taking positions on issues, 13 or in assigning blame for specific conditions in society. 14 Thus, if opinion can be manipulated simply by focusing on one element of an issue versus another, then actors in the public arena are more likely and able to shape broad public opinion and subsequently be successful in the policymaking process. 15

Can We Frame the Terrorist Threat? 547 However, one should not assume that all frames are effective or that all persons exposed to a frame are influenced in the same manner. Indeed, individuals are not empty vessels waiting to be filled with incoming messages they are free to incorporate or reject incoming information. And an individual s receptivity to an issue frame is conditioned by the content of the issue frame and its consistency with the individual s established beliefs and values. 16 Put simply, some individuals will be more likely to adjust their opinions based on the information in an issue frame. 17 Existing research on the impact of issue frames has examined a wide variety of issues. Most often individuals are exposed to competing frames that provide alternative ways for thinking about one issue. Although much of this literature has focused on politically volatile issues such as welfare, affirmative action, civil liberties, civil rights, and gun control, among others, 18 no existing studies have attempted to determine whether issue frames can influence respondent predictions about future events and, subsequently, respondent support for policy proposals that might minimize or prevent these future events. Our research addresses this precise issue. We examine whether alternative frames concerning potential future terrorist attacks influence predictions about future terrorist attacks. In addition, we explore whether exposure to issue frames influences respondent support for a variety of policy proposals designed to thwart terrorism, including increased security and travel restrictions. We also explore the possibility that individual predispositions and the content of policies are likely to enhance or inhibit the influence of issue frames on support for anti-terrorism policies. The Study: Frames and Predictions of Terrorist Attacks We begin by explaining our data and methods for examining the influence of frames on predicting terrorist attacks. Following these empirical tests, we review the conditional influence of issue frames and clarify our expectations concerning the influence of terrorism frames on levels of support for anti-terrorism policies. Survey and Design All adults in Kansas represented the population of interest. The population was randomly sampled and potential respondents were contacted for participation in the survey from November 6, 2001 to February 8, 2002. The survey was presented as a general population survey concerning political issues facing the state and the country. A total of 1,641 adults agreed to participate in the survey for a response rate of 62 percent. Respondents were randomly assigned to one of four (treatment) groups, each receiving a slightly modified questionnaire. Individual respondents in each of the treatment groups were exposed to one of four frames, or treatments, each of which was designed to influence perceptions of potential actions by terrorists against the U.S. Additionally, some respondents were randomly assigned to a control (nontreatment) group of 195 respondents; individuals in this group were not exposed to any issue frames (see Appendix for descriptive statistics). The control group was asked the following question: To insure our safety, government officials have been reporting on the various possibilities for future terrorist attacks. What do you think the terrorists are most likely to do next? Respondents were able to provide any answer they wished, and the responses were recorded verbatim. By contrast, respondents exposed to the issue frames were read an

548 D. P. Haider-Markel, M. R. Joslyn and M. T. Al-Baghal additional sentence in the above question. The modified questions were asked as follows: Frame 1 To insure our safety, government officials have been reporting on the various possibilities for future terrorist attacks. Terrorism experts suggest that the terrorists will most likely attempt an attack on America using hijacked airplanes. What do you think the terrorists are most likely to do next? Frame 2 To insure our safety, government officials have been reporting on the various possibilities for future terrorist attacks. Terrorism experts suggest that the terrorists will most likely attempt an attack on America by creating a smallpox epidemic. What do you think the terrorists are most likely to do next? Frame 3 To insure our safety, government officials have been reporting on the various possibilities for future terrorist attacks. Terrorism experts suggest that the terrorists will most likely attempt an attack on America by targeting commercial nuclear reactors. What do you think the terrorists are most likely to do next? Frame 4 To insure our safety, government officials have been reporting on the various possibilities for future terrorist attacks. Terrorism experts suggest that the terrorists will most likely attempt an attack on America using truck bombs at shopping malls. What do you think the terrorists are most likely to do next? Each frame invokes a potential target and=or event being discussed by political leaders and the mass media at the time of the survey. Although perhaps the media focused more attention on threats posed from airlines and smallpox, all four threats were discussed along with many others, including chemical agents and dirty nuclear bombs. 19 Following the completion of the survey fieldwork, the open-ended responses were assigned values so that response comparisons could be made and an evaluation of the impact of the frames conducted. Table 1 displays the distribution of coded responses by respondent exposure to frames. The frequencies reported in Table 1 clearly show that, relative to the control group, respondents exposed to an issue frame were more likely to predict an attack consistent with that frame. Furthermore, simple Chi-square tests demonstrate these differences are statistically significant. However, our results would be more conclusive if they also emerged in a multivariate model predicting respondent answers. Multivariate Analysis: Measurement Dependent Variables: For this portion of our study, responses from each exposure group are compared to the control group. For each dependent variable, we focused only on those responses related to our issue frames, including those related to attacks using airplanes, a smallpox epidemic, attacks on nuclear facilities, and attacks involving truck bombs and shopping malls. For example, our first model has a dichotomous dependent variable simply coded one if the respondent predicted some type of attack using airplanes and zero for any other valid response. 20 Thus, we have four binary dependent variables and each is examined using logistic regression. Independent Variables: Given the lack of previous research in this area, we have no strong expectations as to what the best predictors might be of a respondent s belief of likely future actions by terrorists. Our main focus is whether the exposure

Can We Frame the Terrorist Threat? 549 Table 1. Percentage distribution of respondent prediction for future terrorist acts and targets by full sample and exposure to issue frames Responses Full sample Control group Airplane frame Smallpox frame Nuclear plant frame Truck bomb frame Use 4.4 3.1 7.8 1.9 2.5 4.2 airplane=hijacking Smallpox 5.6 1.0 4.1 3.8 1.2 1.2 Nuclear plant=reactors 10.1 9.2 10.1 18.1 21.6 4.8 Truck bombing=mall 15.0 11.3 14.7 6.2 6.8 25.7 Other transportation 2.4 3.1 1.8 1.4 3.7 2.4 Utilities 2.1 1.0 0.9 4.3 4.3 0.0 Water supply 7.0 8.2 6.0 7.1 6.8 7.2 High-profile events 4.0 6.2 3.2 2.9 1.2 6.6 Economy (general) 1.4 3.1 0.0 0.5 1.9 1.8 High population areas 5.9 9.2 6.0 1.0 8.0 6.0 Natural resources 1.4 0.5 0.9 1.0 3.1 1.8 (gas, oil) General biological 15.1 8.7 23.5 17.6 14.2 9.6 or chemical General infrastructure 2.5 2.6 4.1 1.9 1.2 2.4 Scare tactics 1.1 1.5 0.5 1.0 1.9 0.6 Assassinate high-level 0.9 1.5 0.9 0.5 1.2 0.6 officials Agriculture 6.1 4.6 5.1 4.3 6.8 10.8 Landmarks 0.5 1.5 0.0 0.0 1.2 0.0 Nothing=No attacks 2.8 3.6 0.9 3.3 4.3 2.4 Governmental structure 3.6 5.6 4.1 1.9 1.9 4.2 Small town 1.5 1.5 1.4 1.4 1.2 1.8 Suicide attacks 0.6 1.0 0.9 0.5 0.6 0.0 Anthrax=Mail 0.6 0.5 0.5 1.4 0.0 0.6 Children=Education 0.6 0.5 0.5 0.0 1.9 1.2 Computers=Technology 0.4 0.0 0.0 1.0 0.6 0.0 Olympics 2.3 7.2 0.5 0.5 0.0 3.6 Other 1.9 3.6 1.4 1.9 1.9 0.6 Chi-square 4.83 37.59 11.63 12.78 Number of cases 951 195 217 210 162 167 Notes. Two tailed significance test ¼ Sig. <.05; Sig. <.01; Chi-square tests are conducted between the control group and each frame group separately; Respondents with no opinion were not included; combined with respondents who did not fully complete the survey, 689 respondents had no answer for this question. to an issue frame might influence respondent predictions. As such, our key independent variable for each of the four models is whether or not the respondent was exposed to the relevant frame (airplane attack, smallpox, nuclear facility, or truck bomb). Each frame variable is coded one if the respondent was exposed to the relevant frame and zero if the respondent was part of the control group.

550 D. P. Haider-Markel, M. R. Joslyn and M. T. Al-Baghal In addition, it is possible that certain groups in the population may have been predisposed to predict certain kinds of terrorist attacks over others. Although the literature offered us no guidance in this regard, we included control variables for the following: respondent fear of being a victim of a terrorist attack, feelings towards President George W. Bush, education level, gender, attention to coverage of September 11 attacks and subsequent developments, political ideology, level of patriotism, and level of political knowledge (see Appendix for descriptive statistics). Multivariate Analysis: Results The results of our binary logit analysis of the four dependent variables appear in Table 2. In each case, exposure to the issue frame significantly increased the probability that a respondent would predict an attack consistent with the issue frame, relative to the control group. These results held even after controlling for a variety of demographic characteristics, political preferences, and media exposure to the issue. In only a few cases do any of the control variables have a significant influence on respondent predictions of terrorist actions and the models do not have particularly strong fit statistics. Nevertheless, developing predictors of how respondents might predict future targets was not our goal. We hoped to test whether the impact of our issue frames was not simply an artifact of respondent characteristics that may have varied between the frame exposure groups and the control group. Our findings suggest the impact of the frames is not an artifact and provide fairly convincing evidence that issue frames can be used to manipulate public perceptions of possible future problems. Even though our results are strong and consistent, one might argue that it matters little if issue frames cannot subsequently influence preferences concerning policy solutions that might address those problems. We agree, and the final portion of our manuscript examines precisely this issue. Issue Frames and Policy Preferences Given our results, we are left wondering how the impact of frames might be translated into support or opposition to policy proposals. For example, if a respondent is exposed to a frame that suggests terrorists may use airplanes in future attacks, and the respondent subsequently predicts this type of event, will this respondent be more likely to support policy proposals that may make it difficult for terrorists to use airplanes in future attacks? On its face one might answer yes. However, respondent preferences for specific policy proposals are also likely to be strongly shaped by the nature of the specific proposal, his or her beliefs and predispositions concerning the effectiveness of the proposal, as well as predispositions concerning the role of government in addressing the problem. Our analysis below examines these issues with a variety of dependent variables from our survey. Multivariate Analysis: Measurement Dependent Variables: For this portion of our study, surveyors asked respondents about a series of policy proposals designed to limit or eliminate the threat of future terrorist attacks. Each of these questions was posed to the respondent immediately after the question concerning predictions about future terrorist attacks. As such,

Table 2. Determinants of predictions for type=target of next terrorist attack: Effect of specific issue frames Variables Airplane attack Smallpox attack Nuclear reactor Mall bomb Exposure to target specific frame 1.496 (.552) 3.729 (1.022).916 (.350) 1.166 (.337) Patriotism score.372 (.176).144 (.129).091 (.125).118 (.117) Male.289 (.504).066 (.384).087 (.358).507 (.334) Bush feeling thermometer.017 (.014).007 (.010).011 (.007).002 (.008) Attention to media coverage of 9=11.078 (.288).172 (.218).058 (.192).588 (.190) Political knowledge score.649 (.263).025 (.222).005 (.201).527 (.188) Education level.114 (.150).107 (.114).501 (.115).189 # (.104) Ideology (liberal to conservative).114 (.177).004 (.141).251 # (.130).061 (.124) Age.008 (.015).011 (.013).011 (.013).003 (.010) Fear of terrorist attack.259 (.271).102 (.230).002 (.010).243 (.212) Constant 3.166 (2.327) 4.229 (1.960) 3.369 (1.610).527 (1.450) Percent correctly predicted 93.6 89.4 85.3 83.9 Pseudo R squared.14.24.09.19 Chi squared 19.34 45.596 15.371 40.294 Log likelihood 144.341 196.908 240.501 260.353 Number of cases 345 359 307 322 Notes. Coefficients are Logit coefficients. Standard errors are in parentheses. Two tailed significance test ¼ Sig.<.05; Sig.<.10; Question used for the dependent variable: To ensure our safety, government officials have been reporting on the various possibilities for future terrorist attacks. What do you think the terrorists are MOST likely to do next? (see Appendix for variable coding). 551

552 D. P. Haider-Markel, M. R. Joslyn and M. T. Al-Baghal issue frames may have still been relatively accessible in their minds. All respondents were asked the following: Please tell me if you would favor or oppose each of the following as a means of preventing terrorist attacks in the United States. Please keep in mind that any of these measures may lead to tax increases or increases in the cost of consumer goods or services. 1. Banning airline passengers from carrying on board any luggage including purses, computers, and briefcases. 2. Increasing security at critical infrastructure facilities, such as commercial nuclear power plants. 3. Increasing security in public places, such as shopping malls and government buildings. For each question, respondents were allowed to indicate whether they would strongly favor, favor, oppose, or strongly oppose each proposal. Responses to each question contained a fair amount of variation, with opponents or supporters not lopsided one way or the other (see Appendix). These three policy questions make direct mention of areas of vulnerability in three of the alternative issue frames. We used each of these questions as a dependent variable for our multivariate analysis. Given the limited nature of each dependent variable, analyses were conducted with both OLS regression and ordered logit. The results were similar for both. For ease of comparison, here we present only the OLS results. Independent Variables: We made use of the same set of independent variables as in our analysis above (see Appendix). Because each question asked respondents about policy proposals being discussed in the media, and each proposal would likely require government action and the use of taxpayer money, we expect that respondent characteristics, especially political knowledge and level of media exposure, to play a greater role here than in the models predicting future terrorist events. Most importantly, we include our same dichotomous variable for exposure to an issue frame. Those respondents exposed to the relevant issue frame should be less opposed to the policy proposal. Multivariate Analysis: Results The results of our multivariate regression models are displayed in Table 3. In each model, males and those exposed to greater media coverage of terrorism tended to be less opposed to the security measures proposed. And in the case of increasing security at public places, respondents who had positive feelings toward President Bush and who had a greater personal fear of terrorist attacks tended to be less opposed to the security measures. However, overall exposure to the issue frames appears to have had little influence on respondent preferences. Only in the case of the airplane frame and banning on-board luggage on airplanes did the issue frame approach statistical significance. In other words, in two of the three cases we cannot reject the null hypotheses that the frames have no influence on policy preferences. How is it, then, that frames influence predictions about future terrorist attacks, but have little or no influence on support for policies that might reduce or eliminate specific types of terrorist attacks? Although our analysis is preliminary, below we conduct some alternative tests to explore this question in more detail.

Can We Frame the Terrorist Threat? 553 Table 3. Determinants of opposition to counter-terrorism measures Variables Airlines measures Security at reactors Security at malls Exposure to policy specific.121 # (.071).029 (.054).036 (.049) frame Patriotism score.063 (.025).005 (.019).007 (.018) Male.257 (.076).128 (.056).143 (.053) Bush feeling thermometer.002 (.002).001 (.001).003 (.001) Attention to media.122 (.041).065 (.030).020 (.029) coverage of 9=11 Political knowledge score.077 # (.040).033 (.030).003 (.028) Education level.026 (.024).029 (.019).005 (.016) Ideology (liberal to.015 (.026).019 (.019).041 (.018) conservative) Age.000 (.002).000 (.002).002 (.001) Fear of terrorist attack.067 (.045).011 (.033).053 # (.031) Constant 3.852 (.329) 2.202 (.247) 1.880 (.228) R squared.06.04.05 Standard error.805.583.551 F 3.389 1.980 2.665 Number of cases 516 488 505 Notes. Coefficients are OLS regression coefficients. Standard errors are in parentheses. Dependent variables from survey questions: 1) Banning airline passengers from carrying on-board any luggage including purses, computers, and briefcases. 2) Increasing security at critical infrastructure facilities, such as commercial nuclear power plants, and 3) Increasing security in public places, such as shopping malls and government buildings. Range for each question: 1 ¼ strongly favor, 4 ¼ strongly oppose. The control group is the baseline (see Appendix for variable coding). Additional Analysis and Discussion Most studies of issue frames only examine support for a political position or policy proposal and subsequently make use of frames that are designed to elicit support or opposition for the political position or policy proposal. Our frames were not so directly linked to specific policy proposals and the lack of such a link may have reduced the potential impact of information contained within the frames. However, given that framing theoretically leads the receiver to consider some information element over others, there should still have been a potential impact. We simply suspect that the frames may not have been clearly linked to one type of policy proposal. To investigate this issue, we reestimated our model for the question on increasing security in public places. We had argued that this proposal should be linked to the truck-bombing frame. For our reestimated models, we remove the variable for exposure to the truck-bombing frame and alternatively include a dichotomous variable coded 1 if the respondent was exposed to the smallpox frame and zero if the respondent was in the control group. In addition, we estimated a second equation with a dichotomous variable coded 1 if the respondent was exposed to any of the four frames and zero if the respondent was in the control group. For a third test, we reestimated our model of opposition to increasing security at critical infrastructure

554 D. P. Haider-Markel, M. R. Joslyn and M. T. Al-Baghal facilities with a dichotomous variable coded 1 if the respondent was exposed to any of the four frames and zero if the respondent was in the control group. The results are shown in Table 4 and do in fact suggest that the information in our frames may not have been strongly linked to specific policy proposals in the minds of respondents. In model M1, first column, the results indicate that respondents exposed to the smallpox frame were less opposed to increasing security in public places than respondents in the control group. Given a belief that a biological attack would be launched in a public place, such as a shopping mall, perhaps respondents exposed to the smallpox frame may have become more inclined to support this proposed measure. Furthermore, as models M2 and M3 indicate, respondents who were exposed to any of the four frames were less opposed to increasing security at public places and to increasing security at critical infrastructure facilities than respondents in the control group. These results, when combined with the influence of the airplane frame on the airplane baggage proposal, suggest that information frames that prime potential terrorism threats can influence subsequent policy preferences. However, the results also indicate that primed threats should be more directly linked to policy proposals Table 4. Determinants of opposition to increasing security in public places, such as shopping malls and government buildings, and of increasing security at critical infrastructure facilities Variables M1: Public: smallpox frame model M2: Public: exposure to any frame model M3: Critical: exposure to any frame model Exposure to frame.160 (.050).104 (.039).081 (.038) Patriotism score.0001 (.0180).017 (.011).023 (.011) Male.111 (.053).096 (.034).061 # (.033) Bush feeling thermometer.002 (.001).002 # (.001).0002 (.0010) Attention to media.086 (.029).063 (.018).075 (.018) coverage of 9=11 Political knowledge score.009 (.029).023 (.018).008 (.018) Education level.003 (.016).005 (.010).014 (.010) Ideology (liberal to.026 (.018).026 (.011).009 (.011) conservative) Age.0007 (.0020).0008 (.0010).0005 (.0010) Fear of terrorist attack.008 (.032).027 (.020).001 (.020) Constant 2.208 (.231) 2.158 (.148) 2.128 (.147) R squared.05.04.03 Standard error.592.562.563 F 3.142 5.715 3.747 Number of cases 558 1282 1295 Notes. Coefficients are OLS regression coefficients. Standard errors are in parentheses. M1 and M2 dependent variable from survey question: Increasing security in public places, such as shopping malls and government buildings, M3 dependent variable from survey question: Increasing security at critical infrastructure facilities, such as commercial nuclear power plants. Range 1 ¼ strongly favor, 4 ¼ strongly oppose. The control group is the baseline (see Appendix for variable coding).

Can We Frame the Terrorist Threat? 555 Table 5. Determinants of opposition to banning on-board luggage on airlines; ideology and exposure to airplane frame Ideology Airplane frame coefficients T-score R-square F Liberals.386 (.144) 2.680.25 4.005 Moderates.030 (.158).190.05.639 Conservatives.051 (.096).526.07 2.363 Notes. Coefficients are OLS regression coefficients. Standard errors are in parentheses. Dependent variable from survey question: Banning airline passengers from carrying onboard any luggage including purses, computers, and briefcases. Range 1 ¼ strongly favor, 4 ¼ strongly oppose. The control group is the baseline (see Appendix for variable coding). Only the results for the issue frame variable are shown, but the results are based on the full model estimation. related to that threat. We believe that most of our survey questions on policy proposals failed to make strong links to specific terrorist threats. A second possible explanation for the weak impact of frames concerns respondent political predispositions. 21 Ideology, partisanship, and support for political leaders all might lead a respondent to reject the information contained in a frame. To examine this proposition, we separated respondents into data subsets based on their ideology and reestimated the full models. 22 This method allows us to control for variations in the influence of frames by ideological predisposition. The results are shown in Table 5 for the policy question on airplane luggage. The results do indeed suggest that at least for the question on airplane luggage, ideology conditioned the influence of the airplane frame. In this case, liberals were the only ones influenced by the frame, with the frame lessening opposition to the proposed ban on luggage. Although this finding supports the notion that respondent predispositions are conditioning the influence of the frames, analysis not shown for the other two questions, increasing security at public places and increasing security at critical infrastructure cites, such as nuclear power plants, does not support the conditional hypothesis. In these two cases, no ideological group was significantly more likely to be influenced by the issue frame. Thus, we can only offer the predisposition hypothesis as a partial explanation for the lack of framing effects. Conclusion and Implications Our research question centered on whether issue frames can influence predictions concerning future terrorist attacks and whether these same frames influence preferences toward a variety of counterterrorism policies. Taking advantage of the high political and media salience of terrorism issues in the late fall of 2001, we embedded a series of experimental issue frames concerning terrorism in a random sample survey of the adult population in a Midwestern state. The frames were designed to prime specific ideas in the minds of respondents and to subsequently influence their perceptions concerning potential terrorist threats, as well as their preferences regarding counterterrorism policies. Our study became especially relevant as the Bush administration increasingly couched its justification for anti-terrorism policies and the war in Iraq based on viewing the world through a post-september 11 window. 23 The results of our analysis do indeed suggest that a variety of frames influence respondent predictions concerning future terrorist attacks. However, our results

556 D. P. Haider-Markel, M. R. Joslyn and M. T. Al-Baghal concerning the impact of frames on respondent policy preferences are less than conclusive. We suggest that the impact of frames in this portion of our study may be reduced by a lack of direct connection to specific policy proposals and by respondent predispositions towards government action. Nevertheless, our analysis does make clear that exposure to any frame regarding a terrorist attack does enhance support for increased security measures. Our research furthers the study of issue frames in a number of ways. First, this study provides evidence of the effect alternative frames have when exploiting the stimulus of real-world events. Prior research has often relied on hypothetical issues and frames to test the influence of issue frames. Second, this study examines issue frames in an environment that is not replicable and has not occurred in recent history. The intense media coverage, posturing by elected officials, and citizen emotion related to the September 11 attacks has few corollaries in American history. Yet, even in this hyper-intense environment, our study still found that issue frames could shape opinion. Finally, our results also suggest that frames are most effective when they are strongly linked to specific policy proposals. If the receiver is left to fill in the blank, he or she is less likely to rely on the information provided in the issue frame. These results are compelling because much of the rhetoric used by the Bush administration to justify a broad war on terrorism and for invading Iraq relied precisely on innuendo and rarely provided the direct linkage between threat and policy response. Nevertheless, the Bush administration appears to have successfully sold most of its anti-terrorism policies to a majority of the American people. Therefore, a possible intervening variable here is the credibility or status of the messenger who provides the frame to link threats with specific policies. Further research is clearly needed on the importance of issue frame content, the specifics of policy proposals, and the role of messengers, or those who provide the issue frames. Notes 1. Frank Davies, U.S. Ignorance of Iraq Startles Pollsters: Study Finds Large Numbers With Mistaken Beliefs About 9 11 And Alleged Weapons, Lawrence Journal World, 15 June 2003, 1A. 2. Davies, U.S. Ignorance of Iraq Startles Pollsters, p. 1A; Steven Kull, Clay Ramsey, Evan Lewis, Misperceptions, the Media, and the Iraq War, Political Science Quarterly (2004): 569 598. 3. Jim Rutenberg, Suffering News Burnout? The Rest of America Is, Too. The New York Times, 11 August 2003. 4. John Zaller, The Myth of Massive Media Impact Revisited: New Support for a Discredited Idea, in Diana C. Mutz, Paul M. Sniderman, and Richard Brody, eds., Political Persuasion in American Politics (Ann Arbor, MI The University of Michigan Press, 1996). 5. See also Timothy E. Cook, Governing with the News: The News Media as a Political Institution (Chicago: University of Chicago Press, 1998). 6. See Cook, Governing with the News, 1998; William G. Jacoby, Issue Framing and Public Opinion on Government Spending, American Journal of Political Science (2000): 750 767; William G. Jacoby, Levels of Measurement and Political Research: An Optimistic View, American Journal of Political Science (1999): 271 301; and Thomas E. Nelson and Donald R. Kinder, Issue Frames and Group-Centrism in American Public Opinion, Journal of Politics (1996): 1055 78. 7. See James N. Druckman, On the limits of framing effects: Who can frame? The Journal of Politics (2001): 1041 1066; Marjorie Randon Hershey, The Constructed Explanation: Interpreting Election Results in the 1984 Presidential Race, The Journal of Politics

Can We Frame the Terrorist Threat? 557 (1992): 943 976; Mark R. Joslyn and Donald P. Haider-Markel, Guns in the Ballot Box: Information, Groups, and Opinion in Ballot Initiative Campaigns, American Politics Quarterly (2000): 355 78; Benjamin I. Page, Who Deliberates? Mass Media in Modern Democracy (Chicago: University of Chicago Press, 1996); and Zaller, The Myth of Massive Media Impact Revisited (see note 4 above). 8. See Thomas E. Nelson, Rosalee A. Clawson, and Zoe M. Oxley, Media Framing of a Civil Liberties Conflict and its Effect on Tolerance, American Political Science Review (1997): 567 83; Donald R. Kinder and Lynn M. Sanders, Mimicking Political Debate With Survey Questions: The Case of White Opinion on Affirmative Action for Blacks, Social Cognition (1990): 73 103; and Michael L. Roberts, Peggy A. Hite, and Cassie F. Bradley, Understanding Attitudes Towards Progressive Taxation, Public Opinion Quarterly (1994): 165 190. 9. For examples see Druckman, On the limits of framing effects (see note 7 above); Nelson, Clawson, and Oxley, Media Framing of a Civil Liberties Conflict and its Effect on Tolerance; (see note 8 above) and Kinder and Sanders, Mimicking Political Debate With Survey Questions (see note 8 above). 10. William H. Riker, The Art of Political Manipulation (New Haven, CT: Yale University Press, 1986); and Deborah A. Stone, Policy Paradox and Political Reason. (Glenview, IL: Scott, Foresman, 1988). 11. Susan T. Fiske and Shelley E. Taylor, Social Cognition, 2nd edition (New York: McGraw-Hill, 1991); Bryan D. Jones, Reconceiving Decision-Making in Democratic Politics: Attention, Choice, and Public Policy (Chicago: University of Chicago Press, 1994); and Hebert Simon, Models of Man (New York: Wiley, 1957). 12. Dennis Chong, Creating Common Frames of Reference on Political Issues, in Diana C. Mutz., Paul M. Sniderman and Richard A. Brody, eds., Political Persuasion and Attitude Change, (Ann Arbor, MI: University of Michigan Press, 1996). 13. Chong, Creating Common Frames of Reference on Political Issues (see note 12 above) Stanley Feldman and John R. Zaller, The Political Culture of Ambivalence: Ideological Responses to the Welfare State, American Journal of Political Science (1992): 268 307; and Robert Lane, Political Ideology: Why the American Common Man Believes What He Does (New York: Free Press, 1962). 14. Assignments of blame have been found to influence both policy preferences and political behavior, including voting and collective protest activity. See Debra Javeline, The Role of Blame in Collective Action: Evidence From Russia, American Political Science Review (2003): 107 121. 15. Jones, Reconceiving Decision-Making in Democratic Politics (see note 11 above). 16. Chong, Creating Common Frames of Reference on Political Issues (see note 12 above); Philip E. Converse, The Nature of Beliefs Systems in Mass Publics, in D. E. Apter, ed., Ideology and Discontent (New York: Free Press, 1964); and Donald P. Haider-Markel and Mark R. Joslyn, Gun Policy, Opinion, Tragedy, and Blame Attribution: The Conditional Influence of Issue Frames, The Journal of Politics (2001): 520 543. 17. Chong, Creating Common Frames of Reference on Political Issues (see note 12 above); Shanto Iyengar and Donald R. Kinder, News That Matters: Television and American Opinion (Chicago: The University of Chicago Press, 1987); and John R. Zaller and Stanley Feldman. A Simple Theory of the Survey Response: Answering Questions Versus Revealing Preferences, American Journal of Political Science (1992): 579 616. 18. See Chong, Creating Common Frames of Reference on Political Issues (see note 12 above); Donald P. Haider-Markel and Mark R. Joslyn, Framing Effects on Personal Opinion and Perception of Public Opinion: The Cases of Physician-Assisted Suicide and Social Security, Social Science Quarterly (2003): 690 705; Haider-Markel and Joslyn, Gun Policy, Opinion, Tragedy, and Blame Attribution (see note 16 above); William G. Jacoby, Public Attitudes Toward Government Spending, American Journal of Political Science (1994): 336 361; Jacoby, Levels of Measurement and Political Research, (see note 6 above); Kinder and Sanders, Mimicking Political Debate With Survey Questions (see note 8 above); Nelson and Kinder, Issue Frames and Group-Centrism in American Public Opinion (see note 6 above); Nelson, Clawson, and Oxley, Media Framing of a Civil Liberties Conflict and its Effect on Tolerance (see note 8 above); Roberts, Hite, and Bradley, Understanding Attitudes Towards Progressive Taxation (see note 8 above); and Tom Smith, That Which We

558 D. P. Haider-Markel, M. R. Joslyn and M. T. Al-Baghal Call Welfare Would by Any Other Name Smell Sweeter: An Analysis of the Impact of Question Wording on Response Patterns, Public Opinion Quarterly (1987): 75 83. 19. Darren K. Carlson, What About the War on Terrorism? Gallup News Service, 11 February 2003; and Jeffrey M. Jones, Fear of Terrorism Increases Amidst Latest Warning, Gallup News Service, 12 February 2003. 20. Respondents who refused to answer the question were not included in the analysis. 21. Kimberly Gross, The Limits of Framing: How Framing Effects May be Limited or Enhanced by Individual Level Predispositions. Paper presented at the Annual Meeting of the Midwest Political Science Association, Chicago, 2000; and Haider-Markel and Joslyn, Gun Policy, Opinion, Tragedy, and Blame Attribution (see note 16 above). 22. Liberals were counted as Strong Liberal, Not so strong liberal, and Moderate liberal. Moderates were simply those responding Moderate. And conservatives were those responding Strong conservative, Not so strong conservative, and Moderate conservative. 23. Davies, U.S. Ignorance of Iraq Startles Pollsters (see note 1 above). Appendix. Variables and descriptive statistics Variables Response percentages Mean= Median Total number DV: Mention airplane 5.0 48 DV: Mention smallpox 6.1 58 DV: Mention nuclear 10.3 98 DV: Mention truck bomb 15.0 143 Exposure to airplane frame 22.8 217 Exposure to smallpox frame 22.1 210 Exposure to nuclear frame 17.0 162 Exposure to truck bomb frame 17.6 167 Control group: No exposure to frames 20.5 195 Patriotism score 1 high patriotism to 12 low patriotism 6.186 1519 6.000 Male 1 Female 63.4 1608 2 Male 36.6 Bush feeling 81.28 1585 Thermometer 0 100 90.00 Attention to media Coverage of 9=11 1 none to 5 A lot 4.07 1561 4.00 Political knowledge Score (sum) 1.981 1620 0 correct to 3 correct 2.000 Education level 1 less than 9th grade to 8 graduate degree 4.46 4.00 1562 (Continued)

Can We Frame the Terrorist Threat? 559 Table 1. Continued Variables Response percentages Mean= Median Total number Ideology 1 Strong liberal to 7 strong conservative 4.66 1454 5.00 Age 49.14 1559 Birth year minus 2001 48.00 Fear of terrorist Attack 1 not worried at all to 4 very worried 1.82 1571 2.00 Banning airline passengers from carrying on-board any luggage including purses, computers, and briefcases 1 strongly favor to 4 or strongly oppose Increasing security at critical infrastructure facilities, such as commercial nuclear power plants 1 strongly favor to 4 or strongly oppose Increasing security in public places, such as shopping malls and government buildings 1 strongly favor to 4 or strongly oppose 2.66 3.00 1.66 2.00 1.92 2.00 1512 1538 1525 Notes. indicates not applicable. The patriotism score is the sum of responses to the following four questions, with the third question reversed: 1) A person who doesn t stand when the Star Spangled Banner is being played could still be a good American, 2) It is O.K. to criticize the government, 3) Although at times I may not agree with the government, my commitment to the U.S. always remains strong, and 4) It bothers me to see children made to pledge the allegiance to the flag or sing the national anthem or otherwise induced to adopt such strong patriotic attitudes. For each question, respondents answers were coded as 1-strongly agree, 2-agree, 3-disagree, and 4-strongly disagree. Those with lower scores would be considered more patriotic. The political knowledge score is the sum of the correct answers to following three questions: 1) Do you happen to know what job or political office is now held by Dick Cheney?, 2) Whose responsibility is it to determine if a law is constitutional or not...is it the President, the Congress, or the Supreme Court?, and 3) What majority is required for the U.S. Senate and House to override a presidential veto?