Social Justice Research

, Volume 22, Issue 2, pp 231–240

Conservative Shift among Liberals and Conservatives Following 9/11/01


  • Paul R. Nail
    • University of Central Arkansas
    • York University

DOI: 10.1007/s11211-009-0098-z

Cite this article as:
Nail, P.R. & McGregor, I. Soc Just Res (2009) 22: 231. doi:10.1007/s11211-009-0098-z


Political orientation and political attitudes were measured in two independent adult samples. One sample was taken several months before the terrorist attacks on 9/11/01; the other, shortly after. Liberal and conservative participants alike reported more conservative attitudes following 9/11/01 than before. This conservative shift was strongest on two items with the greatest relevance to 9/11/01: George W. Bush and Increasing Military Spending. Marginally significant conservative shifts were observed on two other items (Conservatives, Socialized Medicine), and the direction of change on eight of eight items was in a conservative direction. These results provide support for the motivated social cognition model of conservatism (Jost et al., 2003) over predictions derived from terror management theory (e.g., Greenberg et al., 1992).


Motivated social cognitionTerror management theoryWorldview defensePolitical orientationLiberalsConservativesPsychological defensesTerrorist threats9/11 Terrorism


Jost, Glaser, Kruglanski, and Sulloway (2003) conducted a landmark meta-analytic review of the literature concerning the relationships among political orientation, personality traits, situational factors, and social attitudes. Following their review, they concluded that a large portion of this literature could be integrated if it was assumed that uncertainties, fears, and threats associated with life events predispose individuals to develop a politically conservative ideology. “People embrace conservatism (at least in part) because it serves to reduce fear, anxiety, and uncertainty; to avoid change, disruption, and ambiguity; and to explain, order, and justify inequality between groups and individuals” (p. 340). Jost et al. and others have referred to their general approach as the motivated social cognition model of conservatism.

One hypothesis that follows directly from the Jost et al., (2003) analysis is that “situations of crisis or instability in society will, generally speaking, precipitate conservative, system-justifying shifts to the political right” (p. 351). This hypothesis has received support from a number of studies that have examined conservative shifts following real-world crises (e.g., Bonanno & Jost, 2006; Echebarria-Echabe & Fernández-Guede, 2006; Ullrich & Cohrs, 2007; see also, Sales, 1972), as well as manipulated threats in the laboratory (e.g., Jost, Fitzsimons, & Kay, 2004; Lambert et al., 2009; Nail, McGregor, Drinkwater, Steele, & Thompson, in press; Thorisdottir & Jost, 2009). After the terrorist attacks on American soil on 9/11/01, for example, Bonanno and Jost (2006) found that “high-exposure survivors” of the attacks reported that they had become more conservative in their thinking and attitudes in the 18 months following the attacks (p. 311). Echebarria-Echabe and Fernández-Guede (2006) observed a similar conservative shift following the terrorist attacks against railways in Madrid, Spain (3/11/04). Support for the motivated social cognition model is even more remarkable in the Echebarria-Echabe and Fernández-Guede study because their participants were from the Basque region of Spain and, accordingly, were far removed geographically, and not affected directly by, the Madrid attacks. Thus, conservative shift as a defense against threat is apparently not restricted only to those who are directly impacted by a threat but can extend also to those who are merely reminded that the world can be a threatening place.

The present study further probes the phenomenon of motivated conservatism by assessing various political attitudes among American participants before and after the terrorist attacks of 09/11/01. We add to the literature by addressing an issue that has not received sufficient attention in previous studies demonstrating conservative shifts following threat—the extent to which self-identified liberals and conservatives alike will adopt motivated conservatism in the face of threatening world events. The Jost et al. model would be most powerfully supported if liberals and conservatives were similarly attracted to heightened conservatism following 09/11/01.

Motivated Conservatism Versus Worldview Defense?

Whether conservative shift following threat occurs among liberals as well as conservatives is an important issue because it highlights contrasting predictions of two contemporary theoretical orientations. The motivated social cognition model of Jost et al. (2003) predicts a conservative shift for liberals and conservatives alike, as both would find their threat-related fears and anxieties mollified by the stable, clear-cut view of the world that conservatism offers (see also Thorisdottir & Jost, 2009). In contrast, terror management theory (e.g., Greenberg, Simon, Pyszczynski, Solomon, & Chatel, 1992) predicts that liberals and conservatives should each exaggerate their own political leanings as a means of coping with threat. Here, solace is hypothesized to be found in more strongly embracing one’s pre-threat worldview (viz., worldview defense). Thus, by this view, conservatives should become more conservative following threat, and liberals more liberal.

The extant literature provides some evidence for both theoretical perspectives. Bonanno and Jost (2006) found comparable conservative shifts among 9/11/01 survivors for self-reported Democrats, Independents, and Republicans alike. Similarly, following a reminder of 9/11/01 terrorism, Landau et al. (2004, Study 3) found that liberals and conservatives both increased their support for Republican President George W. Bush relative to control participants who were not reminded of 9/11/01. Some researchers, however, report data ostensibly more in line with the worldview defense position (e.g., Gailliot, Stillman, Schmeichel, Maner, & Plant, 2008; Greenberg et al., 1992; Jonas et al. 2008). In the Greenberg et al. study, for example, following a mortality salience (MS) threat, conservative participants rated an out-group member (a liberal) more negatively relative to an in-group member (a fellow conservative) as compared to conservative participants not exposed to the MS threat. Thus, MS enhanced the conservative tendency to see out-group members more negatively than in-group members. The nonsignificant trend among liberals, however, was in the opposite direction, with liberals after a MS threat rating a conservative slightly less negatively relative to a fellow liberal. Because tolerance for dissimilar others is a stereotypical liberal value, Greenberg et al. interpreted this trend among liberals as indicating that they had become more liberal in response to threat. It should be noted, however, that this finding could be interpreted alternatively as supporting a type of conservative shift in that liberals in this study responded to a conservative more favorably following threat. What is needed is research directly comparing post-threat changes among liberals and conservatives on political issues that are not as open to alternative interpretation as the measures used by Greenberg et al. (1992).



Two independent samples of adult, non-college students were recruited and tested. The first sample was taken in October and November of 2000, the second during the same months in 2001, one to two months following 9/11/01. All respondents were drawn from the same population and were acquaintances, friends, or relatives of students enrolled in undergraduate psychology courses at a southwestern U.S. state university. We gave the students bonus credit toward their grade for recruiting research volunteers. The 2000 sample included 70 participants (34 males, 34 females, with two not reporting gender or age; M = 39.38 years of age, SD = 10.15); the 2001 sample included 69 participants (25 males, 44 females, M = 42.28 years of age, SD = 13.25). Because of these differences between the 2000 and 2001 samples, all analyses herein include gender and age as covariates.

Procedure and Materials

The data collection sessions took place in a large classroom. We first assessed gender and age, followed by self-reported political orientation, which served as one of our two independent variables (the other being Year, pre- vs. post-9/11/01). Political orientation relied on a single item, “Please circle the label below that best describes your political point of view” (very liberal, liberal, moderate, conservative, very conservative). We later coded the responses from 1 = very liberal to 5 = very conservative. Similar single-item measures have been shown in past research to validly assess political orientation (e.g., Jost, 2006; Landau et al., 2004; Nail, Harton, & Decker, 2003). The eight items that assessed Political Attitudes, which served as our dependent variables, started at the top of the second page of materials. The items included four typically conservative items/targets (Republicans, George W. Bush, Conservatives, and Increasing Military Spending) and four typically liberal items/targets (the American Civil Liberties Union, Liberals, Socialized Medicine, and Feminists), each evaluated on a −5 to +5 scale. The eight items were presented in the following order: (1) The American Civil Liberties Union, (2) Liberals, (3) Republicans, (4) George W. Bush, (5) Socialized Medicine, (6) Conservatives, (7) Feminists, and (8) Increasing Military Spending. Liberal items were reverse scored, and then all items were averaged to create the Political Attitudes scores (α = .85). Thus, negative scores indicated predominantly liberal attitudes, positive scores indicated conservative attitudes, and scores close to zero reflected centrist attitudes. Participants’ self-reported political orientation and Political Attitudes were highly correlated (for the 2000 sample r [70] = .69, p < .0001; for the 2001 sample, r [69] = .77, p < .0001), thus further supporting the validity of single-item measures of political orientation (cf. Nail et al., 2003).

It is important to note that we had no particular scientific or scholarly interest in, or hypotheses regarding the effects of threat on political attitudes when these data were collected. Rather, we were measuring political orientation and attitudes in the context of testing contemporary theories of racism, which made conflicting predictions for politically liberal versus conservative respondents (see Nail et al., 2003; Nail, Harton, & Barnes, 2008). Accordingly, before assessing political orientation and attitudes, we did not overtly prime thoughts of terrorists, terrorism, or political values in any of the sessions either before or after 9/11/01. This is important because several studies that seem to support the worldview defense position have obtained effects only with the priming of liberal or conservative values before measuring response to threat (Gailliot et al., 2008; Jonas et al., 2008). Further, because we had no a priori predictions regarding the effects of threat, the significant differences reported below could not have been due to any type of experimenter demand.

Results and Discussion

To assess the motivated conservatism and worldview defense hypotheses—especially with respect to potential differences between liberals and conservatives—we regressed mean Political Attitudes onto standardized political orientation scores, effect-coded Year (2000 vs. 2001), and the political orientation X Year interaction (see Aiken & West, 1991). The first order effect for political orientation was significant, β = .72, t(130) = 12.85, p < .0001, η2 = .86. As would be expected, conservatives reported more conservative Political Attitudes than liberals (see Fig. 1). More importantly, and in support of the motivated social cognition model, there was also a significant first order effect for Year, β = .24, t(130) = 4.28, p < .0001, η2 = .41. Participants’ Political Attitudes overall were more conservative after 9/11/01 (M = 1.75, SD = 1.91), than before (M = .82, SD = 1.82). Of further importance, the observed conservative shift was not significantly different for liberals versus conservatives, as the political orientation X Year interaction did not approach significance, β = .03, \( \left| t \right| \) < 1.
Fig. 1

Mean political attitudes as a function of year (pre- vs. post-9/11/01 terrorism threat) and political orientation (extreme Liberal through extreme Conservative). Higher political attitudes equal greater conservatism

The standard deviations of Political Attitudes reported above also provide evidence relevant to the motivated conservatism versus worldview defense hypotheses. If threat causes liberals to become more liberal, and conservatives more conservative, in line with the Greenberg et al. (1992) notion of worldview defense, we would have expected greater variability in Political Attitudes after 9/11/01 than before, as left-leaning participants shifted their political attitudes to the left, and right-leaner’s to the right. The fact that the standard deviations of Political Attitudes were almost identical before and after 9/11/01 (Fmax = 1.09, ns) is inconsistent with the worldview defense position.

To further examine the motivated conservatism and worldview defense hypotheses, we conducted a set of planned contrasts as a function of Year at different levels of political orientation. For these contrasts, we grouped extreme liberals and liberals together, and extreme conservatives and conservatives together, to yield reasonable sample sizes (see Fig. 2). The contrast for liberals in 2000 vs. 2001 (M = −2.02, SD = 1.57; M = −1.21, SD = 1.57, respectively) yielded a significant conservative shift, t(128) = −1.76, p < .05, η2 = .02. Similarly, moderates showed a conservative shift (2000: M = −.25, SD = .87; 2001 M = .65, SD = 1.04), t(128) = −2.73, p < .005, η2 = .06; as did conservatives (2000: M = 1.45, SD = 1.39; 2001 M = 2.69, SD = 1.42), t(128) = −3.34, p < .005, η2 = .08. All three of these contrasts provide support for the motivated social cognition model over terror management theory.
Fig. 2

Mean political attitudes as a function of year (pre- vs. post-9/11/01 terrorism threat) and political orientation (Liberals, Moderates, and Conservatives). Higher political attitudes equal greater conservatism

To examine whether conservative shift might be stronger for some of the individual items than others, we analyzed the eight Political Attitude items as a function of Year in a MANCOVA (again, with gender and age as covariates). Consistent with the regression analysis above, the multivariate main effect for Year was significant, F(8, 123) = 16.37, p < .0001, η2 = .52. As Table 1 reveals, however, contributions to this significant multivariate effect were not equivalent for all of the items. Very strong conservative shifts between 2000 and 2001 were observed for two of the items, George W. Bush, F(1, 129) = 79.95, p < .0001, η2 = .38; and Increasing Military Spending, F(1, 129) = 32.51, p < .0001, η2 = .20. Marginally significant shifts were found with respect to two other items, Conservatives, F(1, 129) = 3.63, p < .06, η2 = .03; and Socialized Medicine F(1, 129) = 3.13, p < .08, η2 = .02. The remaining four items all showed shifts in a conservative direction, but none approached statistical significance (F’s ranging from .20 to 1.93). Nevertheless, eight out of the eight items showed a shift in a conservative direction, χ2(1) = 8.0, p < .005.
Table 1

Adjusted means, standard errors, and F ratios for the eight political attitude items separately as a function of year





Pro-conservative items





George W. Bush








Increasing military spending




Mean conservative




Pro-liberal items









Socialized medicine








Mean liberal




* p < .08, ** p < .06, *** p < .0001

The means have been adjusted for gender and age. Negative means indicate liberal attitudes, positive means conservative attitudes

aThe American Civil Liberties Union

General Discussion

The present findings provide a fairly consistent pattern of support for the Jost et al. (2003) motivated conservatism hypothesis over the Greenberg et al. (1992) worldview defense hypothesis. Consistent with previous field work (Bonanno & Jost, 2006; Echebarria-Echabe & Fernández-Guede, 2006), Political Attitudes overall were significantly more conservative following 9/11/01. Further, this conservative shift in Political Attitudes was comparable for self-identified liberals, moderates, and conservatives alike (Fig. 1). Regarding the variability of participants’ self-reported Political Attitudes, there was no difference as a function of year (before and after 9/11/01). This finding contradicts the worldview defense hypothesis that in response to threat liberals and conservatives should each become more extreme in opposite directions.

Considering the individual Political Attitude items, it makes a good deal of sense in retrospect that the items on which participants showed the most pronounced conservative shifts were those most closely linked with the threat posed by 9/11/01 (viz., George W. Bush and Increasing Military Spending). It seems likely that the events of 9/11/01 made people feel vulnerable to attack and that they responded by rallying around the Commander in Chief and the military as a means of feeling less vulnerable to such attacks in the future. This explanation is consistent with the Jost et al. (2003) literature review and model, as well as with research by Lambert et al. (2009) showing that Americans’ changes toward conservative attitudes following 9/11/01 were mediated by anger.

From the perspective of the Jost et al. (2003) model, it is somewhat surprising that large conservative shifts occurred only on the Bush and military spending items. Nevertheless, employing two-tailed tests, we obtained conservative shifts on two additional items that fell just short of the conventional .05 alpha level (viz., Conservatives and Socialized Medicine). These changes are noteworthy because they are consistent with the Jost et al. contention that conservatism serves a defensive function in terms of one’s ego, in-group, and/or established social systems. Had we obtained conservative shifts on only the Bush and military spending items, it might have been suggested that these two shifts were purely strategic in nature rather than defensive as well; that is, they might have been interpreted as logical, rational support for the Commander in Chief and the military in the face of a very real threat from Al Qaeda. It is difficult to see, however, how shifts to the political right regarding increased support for Conservatives in general or (especially) decreased support for Socialized Medicine could represent strategic defenses vis-à-vis Al Qaeda. Thus, the overall pattern of conservative shifts herein appears to support the Jost et al. motivated social cognitive interpretation of such shifts. The reason we obtained significant or marginally significant shifts on only four of the eight Political Attitudes items, however, is an interesting and open question for future research.

Supporting the notion that conservative shifts in response to threat reflect a defensive function is research by Nail et al. (in press). In Study 1, for example, we exposed some participants to threatening information about a case of unfairness in the U.S. criminal justice system. Later, we measured conservative shift in terms of the tendency to judge out-group members more harshly than in-group members, a tendency that has been linked to political conservatism in past research (e.g., Nail et al., 2003). Liberal participants showed no bias at all for in-group versus out-group members in a no-threat control condition. Under high system threat, however, liberals were just as biased as conservatives. Conservatives, in contrast, displayed the same relatively high level of bias regardless of threat level. The important point for the present analysis is that the observed conservative shift toward bias among liberals cannot be reasonably interpreted as a case supporting purely strategic conservatism, because the type of threat, system injustice, was far removed conceptually from the measure of defensive conservatism (i.e., relative bias against out-group members). Rather, these findings support a defensive interpretation of conservative shifts in line with the Jost et al. (2003) motivated social cognition model.

A clear limitation of the present research is that because we obviously could not go back and randomly assign participants to the pre- vs. post-9/11/01 groups, we cannot definitively conclude with the present data set that the terrorist attacks on 9/11/01 caused the observed increase in conservative attitudes following 9/11/01. Yet, numerous laboratory experiments have observed conservative shifts that can be attributed to manipulated threats (e.g., Jost et al., 2004; Lambert et al., 2009; Landau et al., 2004; McGregor, Nail, Marigold, & Kang, 2005; Nail et al., in press; Thorisdottir & Jost, 2009; Rosenblatt, Greenberg, Solomon, Pyszczynski, & Lyon, 1989; Sales & Friend, 1973). Thus, considering the weight of the evidence from studies of both (a) real-world, field studies and (b) laboratory experiments employing threat manipulations, it appears that threats do indeed generally cause shifts in a conservative or right-wing direction.

How might the present findings be reconciled with other findings that seem more consistent with a worldview defense reaction? The most frequently cited study in support of the worldview defense position is Greenberg et al. (1992), summarized in the “Introduction” herein. To reiterate briefly, MS threat caused conservatives to become significantly more biased against a liberal (vs. a conservative) target. Liberals, however, showed a nonsignificant trend in the opposite direction, expressing less bias under threat against a conservative (vs. a liberal). Greenberg et al. interpreted this latter finding as liberals under threat clinging more strongly to the liberal value of tolerance, thus more strongly supporting their liberal worldview. As noted in the “Introduction”, however, it is possible that this change among liberals itself represents a type of conservative shift—increased appreciation under threat for conservatives. Furthermore, liberal participants in this study had declared that they were liberal—which may have primed the value of tolerance—just before they evaluated the liberal and conservative targets. Thus, hypocrisy constraints may also have prevented liberal participants from indulging in derogation of conservatives relative to liberals.

Other studies that would seem to support the worldview defense position are likewise dependent upon the immediate priming of liberal or tolerant values (Gailliot et al., 2008; Greenberg et al. 1992, Study 2; Jonas et al., 2008). Gailliot et al. (2008, Study 1), for example, examined an issue among white college students in America that is quite important if not at the core of liberal political attitudes—acceptance of African Americans. Gailliot et al. found that non-Black participants’ attitudes toward African Americans became significantly more positive following MS threat, but only if the liberal ideal of egalitarianism had been primed first. Without this prime, there was no difference in attitudes between MS participants and dental pain/control participants.

The issue of priming is critical, we believe, because all studies to our knowledge that seem to support the worldview defense position are dependent upon the priming of liberal values (Gailliot et al., 2008; Greenberg et al., 1992; Jonas et al., 2008). In contrast, while one study has found that priming of conservative values enhanced a conservative shift to MS threat (Jonas et al., 2008, Study 3), numerous studies have found conservative shifts in the absence of any overt priming—that is, priming apart from the inherent nature of the threat itself (Bonanno & Jost, 2006; Echebarria-Echabe & Fernández-Guede, 2006; Jost et al., 2004; Lambert et al., 2009; Landau et al., 2004; McGregor et al., 2005; Nail et al., in press; Thorisdottir & Jost, 2009; Rosenblatt et al., 1989; Sales, 1972; Sales & Friend, 1973 Ullrich & Cohrs, 2007). In this context, the reader will recall from our “Method” section that the present data were initially collected for the purposes of testing various theories of racial attitudes. We had no interest at the time in examining responses to 9/11/01 and did nothing to remind participants of either 9/11/01 or broader political issues when the data were collected.

We conclude that a fair test between the motivated conservatism and worldview defense hypotheses occurs under circumstances in which participants respond with no overt politically relevant primes (one way or the other). Under these circumstances, the weight of the evidence considering both the present data and published reports supports the Jost et al. (2003) motivated social cognition model. Further research will be necessary to determine why we obtained larger conservative shifts on certain political issues (viz., Conservatism and Socialized Medicine) as compared to other fairly similar issues (viz., Republicans, Feminists, Liberals, and the ACLU).

Copyright information

© Springer Science+Business Media, LLC 2009