Bridging Partisan Divisions over Anti-Terrorism

Download data and study materials from OSF

Principal investigators:

Neil Malhotra

Stanford University



Elizabeth Popp

University of Illinois Urbana Champaign


Sample size: 1339

Field period: 3/21/2008-3/31/2008


We examine how changes in perceptions of threat affect individuals’ policy views, as well as the political implications of this relationship. We administered a survey experiment to a representative sample of the U.S. population in which we exogenously manipulated individuals’ perceived likelihood of a future terrorist attack on American soil and assessed subsequent changes in support for terrorism-related public policies. We find that perceived threat substantially increases support for policies intended to reduce terrorism and that this effect is concentrated among Democrats who believe another terrorist attack is likely to occur. These results suggest that increased levels of threat following September 11th may have assisted the Bush Administration in building a bipartisan coalition for its anti-terrorism policies by attracting individuals whose predispositions may have otherwise precluded their support. More broadly, our findings demonstrate how political elites can leverage and manipulate threat to bridge partisan divisions.


Hypothesis 1: Information indicating a higher perceived threat from a terrorist attack will make individuals more supportive of public policies designed to combat terrorism.

Hypothesis 2: The effect of threat information on support for public policies designed to combat terrorism will be stronger among Democrats who believe an attack is likely.

Experimental Manipulations

As the main treatment, we presented respondents with a brief statement about findings from an actual survey of arms control experts conducted by the Senate Foreign Relations Committee.[1] Respondents were asked to read the following passage, in which we manipulated the information respondents received about the percentage chance (X) of a terrorist attack: Assisting the government in preparing for a potential terrorist attack, the Senate Foreign Relations Committee surveyed a group of arms control experts about the likelihood of a terrorist attack on American soil. Based on an analysis of the responses, the experts concluded that there is a [X]% chance that there will be a damaging terrorist attack in the United States in the next five years. X was randomly drawn from multiples of five between 5 and 95, inclusive.


The main dependent variable is an index of attitudes towards anti-terrorism policies. We considered four policies that have been proposed to combat terrorism: (1) “Do you support or oppose the U.S. government using wiretaps to listen in on citizens’ phone conversations in terrorism investigations?”; (2) “Do you support or oppose a law requiring libraries to turn over to terrorism investigators records of what books people have checked out?”; (3) “Do you support or oppose limits on airline passengers carrying liquids or gels (e.g., beverages, toothpaste, shampoo) onto airplanes?”; and (4) “Do you support or oppose the United States launching a military attack against Iran, which American government officials have accused of supporting terrorist organizations such as al Qaeda?”[2] All of these policies impose some sort of costs on the public. To reduce measurement error, we averaged these four items together to construct an index (Cronbach’s a of .76). The scale was coded to lie between 0 (lowest level of support) and 1 (highest level of support).

Summary of Results
Consistent with Hypothesis 1, the threat of terrorism significantly boosts support for anti-terrorism policies. The coefficient associated with “Threat Treatment (X)” (b1 =.08, p=.04, two-tailed) is positive and statistically significant, meaning that increasing the threat level from 5 to 95 percent results in an eight percentage point increase in the range of the dependent variable. As predicted by Hypothesis 2, the treatment effects were concentrated among threatened Democrats. Per theoretical expectations, the coefficient associated with the “Threat Treatment (X) ´ Threatened Democrats” interaction is positive and statistically significant (b3 = .23, p=.02), meaning that the impact of Xi on policy attitudes was stronger among threatened Democrats as compared to the members of the omitted category, which is comprised of Republicans and unthreatened Democrats.[3] Decomposing this coefficient estimate into its constituent elements, we find that among threatened Democrats, increasing X from 5 to 95 percent boosted policy support by .27 units, or over one-quarter of the range of the dependent variable (b1 + b3 = .27, p=.002). Among this subgroup, the effect of threat perceptions is stronger than any of the demographic predictors, including education. Conversely, among the omitted subgroup, X does not significantly affect anti-terrorism policy attitudes (b1 = .04, p=.40).


Malhotra, Neil, and Elizabeth Popp. 2012. "Bridging Partisan Divisions over Anti-Terrorism Policies: The Role of Threat Perceptions." Political Research Quarterly. 65: 34-47.