Content uploaded by Rose Mcdermott
Author content
All content in this area was uploaded by Rose Mcdermott
Content may be subject to copyright.
Dead Certain
Confidence and Conservatism Predict Aggression in Simulated
International Crisis Decision-Making
Dominic D. P. Johnson &Rose McDermott &
Jon Cowden &Dustin Tingley
Published online: 24 March 2012
#Springer Science+Business Media, LLC 2012
Abstract Evolutionary psychologists have suggested that confidence and conservatism
promoted aggression in our ancestral past, and that this may have been an adaptive
strategy given the prevailing costs and benefits ofconflict. However, in modern environ-
ments, where the costs and benefits of conflict can be very different owing to the
involvement of mass armies, sophisticated technology, and remote leadership, evolved
tendencies toward high levels of confidence and conservatism may continue to be a
contributory cause of aggression despite leading to greater costs and fewer benefits. The
purpose of this paper is to test whether confidence and conservatism are indeed
associated with greater levels of aggression—in an explicitly political domain. We
present the results of an experiment examining people’s levels of aggression in response
to hypothetical international crises (a hostage crisis, a counter-insurgency campaign, and
a coup). Levels of aggression (which range from concession to negotiation to military
attack) were significantly predicted by subjects’(1) confidence that their chosen policy
would succeed, (2) score on a liberal-conservative scale, (3) political party affiliation,
and (4) preference for the use of military force in real-world U.S. policy toward Iraq and
Iran. We discuss the possible adaptive and maladaptive implications of confidence and
conservatism for the prospects of war and peace in the modern world.
Hum Nat (2012) 23:98–126
DOI 10.1007/s12110-012-9134-z
D. D. P. Johnson (*)
Department of Politics and International Relations, University of Edinburgh, 15a George Square,
Edinburgh EH8 9LD, UK
e-mail: dominic.johnson@ed.ac.uk
R. McDermott
Department of Political Science, Brown University, 36 Prospect St., Providence, RI 02912, USA
e-mail: Rose_McDermott@Brown.edu
J. Cowden
Department of Social Work, SJSU, San José, CA 95192, USA
e-mail: tuberadioguy@gmail.com
D. Tingley
Department of Government, Harvard University, 1737 Cambridge St., Cambridge, MA 02138, USA
e-mail: dtingley@gov.harvard.edu
Keywords Confidence .Overconfidence .Conservatism .Aggression .Evolution .
Politics
You can’t fake it. You have to believe it. And I believe it. I believe we’ll
succeed.
—President George W. Bush, 2006
Today, large-scale violence between societies is widely regarded as costly and
destructive for little gain. Although many people recognize the necessity of conflict in
certain circumstances (e.g., defense against invasion), war is generally seen as the
result of a failure of some better alternative, such as a negotiated settlement (Fearon
1995). When wars occur, therefore, they are often seen to result from poor informa-
tion or miscalculations (Levy and Thompson 2010).
An evolutionary approach suggests a different explanation. If human decision-
making strategies were shaped by the costs and benefits of our ancestral past (as
evolutionary psychologists argue), then evolved proximate mechanisms prompting
aggressive strategies may be triggered today even when cost-benefit ratios are
unfavorable. Although humans are obviously able to evaluate contemporary costs
and benefits somewhat rationally, our judgment and decision-making are not immune
from the influence of evolved, often subconscious, heuristics and biases as well
(Barkow et al. 1992; Fiske and Taylor 2007; Wilson 2004).
In short, we might expect a large degree of “mismatch”between evolved tenden-
cies and the (evolutionarily novel) causes and consequences of those tendencies in
modern contexts. Two particular aspects of mismatch have been suggested by
evolutionary biologists as important to understanding aggression.
First, Robert Trivers (2000,2011) and Richard Wrangham (1999) have suggested
that overconfidence—confidence beyond that warranted by the evidence—may have
been an adaptive strategy in our ancestral past because it improved morale, resolve,
persistence, and/or helped to bluff opponents. A recent evolutionary model showed
that overconfidence could indeed evolve in competition with accurate or under-
confident strategies, under a wide range of conditions (Johnson and Fowler 2011).
Individuals with high levels of confidence may also have been better able to recruit
fence-sitters to side with them, increasing their probability of victory by increasing
coalition size. Today, however, overconfidence is likely to be misplaced because its
original signaling and feedback mechanisms are lost in the modern context of mass
armies, modern weapons, and military leaders far away from the battlefield.
Second, Aaron Sell and colleagues find an association between conservative
political preferences and the endorsement of the utility of force among physically
stronger men with a history of fighting (Sell et al. 2009; see also Sell et al. 2012). This
may have been an adaptive strategy in the past if physically stronger individuals were
more likely to prevail in situations of coercion, competition, and combat, potentiating
a socioecological niche in which aggression brought selective advantages—at least
for some individuals. Today, however, powerful and conservative individuals may
tend to be belligerent even in evolutionarily novel situations in which aggression may
not help them to achieve their objectives and are likely to incur significant costs
instead.
Hum Nat (2012) 23:98–126 99
This study aims to test whether confidence and conservatism are indeed associated
with greater aggression in an explicitly modern political setting. As outlined below,
confidence and conservatism already play a major role in the political science
literature, but there have not been any tests of whether they independently constitute
specific causes of aggression in a controlled experimental setting.
Confidence
One of the things political scientists have learned from the past few decades of
research is that there is no single cause of war. Instead, a multitude of factors have
been shown to contribute to decisions that lead to the use of force, including
individual-level factors such as misperception (Bennett and Stam 2004; Levy and
Thompson 2010; Vasquez 1993,2000). Nevertheless, some particularly pervasive
and influential phenomena are repeatedly identified on the eve of war, even though
they are difficult to define, quantify, or test. One such factor is overconfidence.
Several historians and political scientists have come to the conclusion that, whatever
other causes of war may be present in a given case, overconfidence on one or more
sides appears to be extremely common and represents a causal factor in the decision
to fight (Blainey 1973;Jervis1976; Johnson 2004; Johnson and Tierney 2011;
Lebow 1981; Levy 1983; Stoessinger 1998; Tuchman 1984; Van Evera 1999; White
1968). Indeed, the recent U.S. experience in Iraq and elsewhere has brought the
puzzle of why overconfidence occurs and how it can be averted to the forefront of
both academic and media attention (Draper 2007; Fallows 2004; Jervis 2003; Walt
2011; Woodward 2005).
When one looks beyond political science, however, the association of overconfi-
dence and war should be no surprise. Evidence from several disciplines shows that
overconfidence represents a widespread and powerful bias across a vast range of
human activities, ranging from sports to leadership to economics (Camerer and
Lovallo 1999; Dunning et al. 2003; Ehrlinger et al. 2008; Kanter 2004; Kruger and
Dunning 1999; Salvador et al. 2003; Taylor 1989; Van den Steen 2004). Not least, the
recent banking crisis highlighted an important role of overconfidence among finan-
cial decision-makers (Akerlof & Shiller 2009; Barber and Odean 2001;
Ben-David et al. 2006). Psychologists have long documented overconfidence across
a variety of judgment and decision-making domains (Peterson 2006; Taylor and
Brown 1988), and there is now a considerable literature on its causes, consequences,
and sources of variation (Armor and Taylor 1998; Baumeister 1989; Gollwitzer 2011;
Lim 1997; Taylor and Brown 1994; Taylor et al. 2003). Nobel prize winner Daniel
Kahneman recently concluded that all of the psychological biases he and his col-
leagues have uncovered over the past 40 years of the “cognitive revolution”promote
hawkish decision-making—and many of these biases lead to hawkishness because
they produce overconfidence (Kahneman and Renshon 2006). As Dale Griffin and
Amos Tversky summed it up, “although overconfidence is not universal, it is
prevalent, often massive, and difficult to eliminate”(Griffin and Tversky 2002:248).
However, while it is tempting to extrapolate from studies in psychology and other
disciplines to invoke overconfidence as a cause of war, we have almost no experi-
mental data on whether the overconfident biases reported by psychologists also occur
100 Hum Nat (2012) 23:98–126
in modern politically relevant contexts (for one recent exception see Johnson et al.
2006). Typically, such studies examine everyday personal events, such as health,
driving, or exam performance (for a range of examples see Peterson 2006; Taylor
1989). Several case study analyses suggest that overconfident biases are present in
political contexts (Johnson 2004; Van Evera 1999), but since many other factors are
present as well it is difficult to know if overconfidence is causal, or how important it
is relative to other variables, in promoting conflict. What has been lacking in
addressing this issue is controlled, experimental studies to test such hypotheses while
holding other factors constant. Laboratory experiments offer one such method.
We previously conducted networked wargames in which subjects played political
leaders resolving a dispute. The results clearly showed not only that men (but not
women) were overconfident about the probability of success, but also—and critically
for a causal argument—that greater levels of overconfidence were associated with a
greater likelihood of attacking their opponent (Johnson et al. 2006). This was a
significant step in validating a link between overconfidence and aggression. Never-
theless, this earlier study left open the question of whether overconfidence would be
associated with aggression in real-world political scenarios, rather than in competitive
wargames in which people may have been trying to win, rather than selecting what
they felt was the most appropriate strategy. The present study reports the results of
people’s decision-making behavior in the context of plausible real-world inter-
national crises. Beyond this, the added value of the present study is to explore
the relationship not only between confidence and aggression, but also between
both of these variables and the all-important political variable of conservatism.
Accounting for political preferences is crucial to understanding political behav-
ior, including aggression.
Conservatism
A key factor we expect to affect policy choices in crises is conservatism. Political
ideology and political party identification are associated with large individual differ-
ences in attitudes to domestic and foreign policy, and there is a long history of
research on the causes and consequences of political preferences. For example,
differing political preferences powerfully affect voting behavior, support for war
(Gartner 1997; Mueller 1973), and perceptions of wars and crises (Jervis 1976;
Johnson and Tierney 2006).
Although criticized on methodological grounds, classic older studies directly
linked conservatism with aggression (Adorno et al. 1950; Altemeyer 1988), and
since then a large number of studies related political preferences, and conservatism
in particular, to personality variables. Jost et al. (2003) recently conducted a compre-
hensive review of this literature, including a statistical meta-analysis of factors that
had been linked with conservatism. The authors concluded that conservatism was
consistently and strongly predicted by a range of social-cognitive motives, including
dogmatism, intolerance of ambiguity, lack of openness to experience, mortality
salience, and system instability. Also significant but weaker correlates were uncer-
tainty avoidance, integrative complexity, need for order, structure and closure, and
fear of threat in general. Jost et al. argued that the key factors underlying social-
Hum Nat (2012) 23:98–126 101
cognitive motives and resultant conservative attitudes derived from responses to basic
environmental stimuli, including fear, threat, and uncertainty.
However, none of the studies reviewed in Jost et al. (2003), other than the famous
texts on “right-wing authoritarianism”(Adorno et al. 1950), directly addressed
aggression, especially in the context of political decision-making. The picture is
likely to be complicated because we know that other variables also influence people’s
preferences regarding the use of military force. For example, Feaver and Gelpi found
that military leaders were less tolerant of casualties, while civilian leaders were more
hawkish (Feaver and Gelpi 2004). Preference for the use of force has also been found
to vary with physical and psychological health and illness (McDermott 2007), stress
(Rosen 2004), age (Horowitz et al. 2005), and framing (Johnson and Tierney 2007).
We therefore wanted to test explicitly for a link between conservatism and aggression.
Some early work indicated that such a relationship may exist. Etheredge (1978a,b)
examined a group of senior U.S. State Department officials and found a relationship
between various aspects of their personality (including some related to conservatism)
and their foreign policy choices and attitudes. Specifically, he found that officials who
embraced higher dominance strategies toward their subordinates and family members
proved more likely to endorse the threat of use of military force by the United States in
49 foreign policy crises between 1898 and 1968 (compared with officials who espoused
lower dominance personality characteristics). Indeed, he could predict policy preference
among individuals based on this one trait with more than 75% accuracy.
This study is timely because it follows several new analyses suggesting that there are
fundamental differences in how liberals and conservatives react to threats and make
decisions. For example, physiological reactions to threatening stimuli (as measured by
eye blinks and skin conductance) increase with the degree of subjects’conservatism
(Oxley et al. 2008). Other work has found significant differences in the activation of
key brain regions between liberals and conservatives when performing identical
decision-making tasks. For example, more conservative people showed lower activity
in the anterior cingulate cortex—a brain region involved in cognitive control and self-
regulation—and were consequently less able to alter habitual response patterns to deal
with a novel task (Amodio et al. 2007). Another study found that Republicans showed
significantly greater amygdala activation (a brain region associated with fear) and Dem-
ocrats showed greater insula activity (a brain region associated with conscious attention to
internal physiological and affective states) when performing a risk-taking task (Schreiber
et al. 2009). Finally, as noted in the introduction Sell et al. (2009,2012) found that
physically stronger men with a history of fighting were more likely to adhere to
conservative political preferences and to endorse the utility of force in politics.
Although there is a large literature on the social, psychological, neurological, and
biological correlates of conservatism (see also Alford et al. 2005), there have been few
studies of whether there is a direct association between conservatism and preference
for military aggression. We undertake such an examination in the following study.
Methods
We collected data from 130 male students at the University of California atSanta Barbara
in May 2007. Although we have examined sex differences in aggression in previous
102 Hum Nat (2012) 23:98–126
studies (Johnson et al. 2006; McDermott and Cowden 2001,2008; McDermott et al.
2007), we used only male subjects here to maximize statistical power in revealing the
effects of other (likely weaker) variables of interest. Age range was 19–43 (mean 0
22.0, median020.5, s.d.04.7). Subjects included some university employees as well
as students, the latter of which were studying a wide range of disciplines across
the humanities, social sciences, and life sciences. Ethnicity (N083; not all
subjects gave all data) was 59.3% Caucasian, 16.9% Asian, 11.9% Hispanic, 1.7%
African American, and 10.2% Other. Religious affiliation (N083) was 37.3% Atheist/
Agnostic, 20.3% Protestant, 13.6% Catholic, 8.5% Buddhist, 6.8% Jewish, 1.7%
Hindu, 11.9% Other. Party identification (N0130) was 42.0% Democrat, 15.3%
Republican, 26.0% Independent, 8.4% Other, 8.4% Don’t Know.
Subjects sat alone at desks and completed a range of paper and pencil question-
naires under supervision. No talking or conferring was allowed. The experiment
asked subjects how they would, as U.S. president, respond to three international
crises: (1) a “Hostage Scenario”(in which a commercial airliner was hijacked); (2) a
“Central Asian Scenario”(a resurgence of the Taliban in Afghanistan); and (3) a
“Latin American Scenario”(a possible military coup to oust a socialist leader).
Scenarios were counter-balanced, such that the order in which subjects were pre-
sented with each of the three scenarios was randomized. Building on previous
experience in running experimental wargames (McDermott and Cowden 2001,
2008; McDermott et al. 2007), each crisis was described in around 200–400 words
with realistic detail, and each one was presented on a separate sheet of paper (see
Appendix for full text of scenarios).
In each case, subjects could choose from three options: some form of concession
or appeasement; some form of negotiation; or some form of military attack. In our
design, but not in the order of presentation, we strove to word these options so that
aggressiveness increased incrementally, from appeasement to negotiation to attack.
For example, in resolving the Latin American scenario, subjects could choose to (a)
“let this leader nationalize American interests, but cut diplomatic ties,”(b) “attempt to
negotiate fair restitution for American companies,”or (c) “allow US special forces
and the CIA to assist the local military leader in staging a coup.”The order of the
policy options was also counterbalanced across scenarios.
Each crisis was carefully written in order to resemble a plausible scenario given the
political climate at the time of the study. We also used a number of techniques to
ensure that the whole range of policy options remained reasonable, so that military
aggression was in no way the “best”or “only”realistic option. First, subjects were
explicitly told, in underlined text that, “there is no right answer to any of these crisis
scenarios.”Second, scenarios included a fair amount of complexity, so that many
consequences and actors could be considered in reaching a decision. Third, even the
“appease”option was not entirely unattractive (e.g., in the Central Asian scenario,
withdrawing U.S. troops included the benefit of “concentrating your military effort on
Iraq”). Finally, the aggressive options were not simply a matter of taking a
tough line, they were risky military options that entailed potentially high costs
for one or both sides (e.g., special forces storming a hijacked aircraft, use of
daisy cutter bombs in Afghanistan, and sending U.S. troops to aid a Latin
American coup). With these characteristics, subjects were not expected to
choose the aggressive options lightly.
Hum Nat (2012) 23:98–126 103
Dependent Variable: Aggression
For each crisis, “aggression”was scored as 1 if subjects chose the appease option, 2 if
they chose the negotiate option, and 3 if they chose the attack option. The main
dependent variable in our statistical analyses is the level of aggression of the subjects’
chosen policy options summed over all three crisis scenarios (thus a variable ranging
from 3 to 9; for the purposes of analysis, we subtracted 3 from all aggression data to
scale the variable from 0 to 6).
Independent Variables
Confidence
For each scenario, subjects were also asked about their level of confidence in their
chosen policy as follows: “On a scale of 0 to 10, how likely do you think it is that
your chosen course of action will succeed? Please circle your response, where 0 is
very unlikely and 10 is very likely.”
This is obviously different from other measures of confidence in the social
psychology literature, which tend to focus on psychological, self-esteem-related
measures, or estimates of personal effectiveness at a given task (e.g., Taylor et al.
2003). We used the above measure of confidence because we were interested in
people’s confidence about the efficacy of their specific policy choices, not their
general or personal levels of confidence in everyday life. Our measure was deemed
more likely to be associated with their policy decisions in the crisis scenarios. For
purposes of comparison, we did take measures of a standard personality inventory for
“optimism,”which was (positively but not significantly) correlated with our measure
of confidence (see “Results”).
Finally, note that we are using a measure of confidence, not overconfi-
dence. Overconfidence can be defined as an expectation that is not warranted,
given the true likelihood of some outcome.
1
By this criterion, there is no way in
our hypothetical scenarios to establish whether or not people are overconfident (their
policy may or may not have worked in reality, which was an essential element of the
experimental design). Therefore, our measure only offers a way of distinguishing
individuals who are more or less confident from others in the sample. Note that
this does not detract from the ability to test our hypothesis, however, which is
simply that subjects who choose more aggressive military solutions to hypo-
thetical international crises will demonstrate greater confidence that their chosen
policy will succeed. Overconfidence can be assumed to be more likely at the
higher end of this continuum.
1
Following previous work (Johnson 2004; Johnson and Tierney 2011), we define confidence as the
perceived probability that a specified outcome will occur. For example, low confidence may equate with
a belief that one has a 25% chance of victory, while high confidence may equate with a belief that one has a
75% chance of victory. Overconfidence is defined as a level of confidence that exceeds the true likelihood
of an outcome. For example, if a tennis player expects to win 75% of their matches but loses them all, this
would imply overconfidence.
104 Hum Nat (2012) 23:98–126
Conservatism
We also asked people about their political party identification, and their political
ideology. For party identification, we employed a standard measure, asking subjects:
“Generally speaking, do you think of yourself as a Republican, a Democrat, an
Independent, or what?”(to which they could respond Republican, Democrat, Inde-
pendent, Other, or Don’t Know).
For political ideology, we administered a standard seven-point liberal-conservative
scale (Zaller 1992) which asked: “When it comes to politics, do you usually think of
yourself as very liberal, liberal, slightly liberal, moderate or middle of the road, slightly
conservative, conservative, or very conservative?”This variable also included a “Don’t
Know”category, which was excluded where appropriate in statistical analyses. We
recorded data on both party identification and political ideology because they may
reflect different underlying characteristics, although of course they are, unsurprisingly,
correlated (in our data, Republicans had significantly higher liberal-conservatism scores
than Democrats; Mann-Whitney U-test: Z05.31, N054,19, p<0.0001). Briefly, ideol-
ogy is thought to be a composite measure of people’s political beliefs, whereas party
identification is thought to be in large part due to socialization that occurs before
political beliefs are established (Achen 2002). Recent work by Hatemi et al. (2009,
2011) and Fowler et al. (2009) also suggests that while party identification may
derive from socialization, partisan intensity may result from heritable traits.
Personality Inventories
We also administered a set of nine standard personality inventories from the “Inter-
national Personality Item Pool”(ipip.ori.org). We included them because they were
deemed potentially relevant to how different individuals may respond to a hypothet-
ical crisis (McDermott 2004,2007), and were thus important to include as control
variables. These variables were extroversion, agreeableness, conscientiousness, emo-
tional stability, intellect, optimism, leadership, courage, and locus of control.
2
The
variables are scored as follows. Each inventory lists ten statements (five for locus of
control), such as “Get irritated easily,”and subjects are asked to indicate how
accurately each statement describes them on a five-point scale ranging from “very
accurate”to “very inaccurate.”A single score for each inventory is calculated as the
sum of the responses to all ten statements (some items are reverse-coded as per
standard practice).
Statistical Methods
Levels of aggression summed across all three scenarios followed an approximately
normal distribution as shown in Fig. 1. However, a Kolmogorov-Smirnov test rejected
a strictly normal distribution (Z01.99, N0129, p< 0.001), and transforming the data
2
Locus of control is a measure of how much control people perceive they have over the world. People with
a relatively “internal”locus of control believe that they and their actions can influence events, whereas
people with a relatively “external”locus of control tend to be more fatalist and to believe they cannot
influence events, which are instead seen to be controlled by the environment, other people, or some higher
power.
Hum Nat (2012) 23:98–126 105
to correct for a slight negative skew was not successful (e.g., square root, reciprocal).
Therefore, in much of our subsequent analyses we use nonparametric statistics.
Aggression (as well as the liberal-conservative scale) is an ordinal variable. For all
correlations we therefore use Kendall’sτ
B
test, which is more appropriate than
Pearson’s or Spearman’s Rank tests for correlations between variables with ordered
categories. All tests are two-tailed (which can be considered conservative given our a
priori directional hypothesis: confidence, conservatism, and aggression were all
expected to be positively related).
For multivariate analyses we use ordered probit models. Ordered probit allows the
dependent variable, aggression, to be an ordinal variable. Note that it does not make
the assumption that a difference between, say, 3 and 4 on the dependent variable is the
same as the difference between 5 and 6. In ordered probit, although the dependent
variable is assumed to be normal, slight departures of the dependent variable from
normality are unlikely to bias results.
Results
We found no discernable effects of ethnicity or religion on responses to the crisis
scenarios (whether pooled or examined individually). We also found no effect of age
on confidence, aggression, or conservatism.
Incidence of Aggression
Mean aggression was 3.71 (median04, range01–6, s.d.01.14). Figure 1shows the
distribution of summed levels of aggression, and Table 1shows the breakdown of all
responses in each scenario. Across all scenarios combined, negotiate was the most
common option, followed by attack and then appease. This ranking of responses was
exactly the same within each individual scenario as well—negotiate, then attack, then
appease. The frequency of responses to the Hostage and Central Asian crises were
Fig. 1 Aggression summed
across scenarios appears to be
an approximately normal
distribution, but fails a
Kolmogorov-Smirnov
test (Z01.99, N0129, p<0.001)
106 Hum Nat (2012) 23:98–126
remarkably similar, whereas in the Latin American crisis a much larger majority of
people chose to negotiate. Table 2shows descriptive data for the independent
variables used.
Aggression and Confidence
There was a highly significant relationship between the level of confidence
subjects ascribed to their chosen course of action and how aggressive it was
(Kendall’sτ
B
00.19, N0129, p00.006; see Fig. 2). That is, people who chose more-
aggressive options (e.g., military action) were also more likely to believe that their
chosen policy would be successful. On the contrary, people who chose less-
aggressive options (e.g., appease) were less likely to believe their chosen policy
would be successful.
Within individual scenarios, the correlation was positive in all cases, significant in
the Central Asian crisis (Kendall’sτ
B
00.17, N0130, p00.022) and the Latin Amer-
ican crisis (Kendall’sτ
B
00.18, N0129, p00.018), but not significant in the Hostage
crisis (Kendall’sτ
B
00.07, N0130, p00.39).
Table 1 Frequency of chosen policy options for each crisis scenario and all scenarios combined
Scenario Response
Appease Negotiate Attack
Hostage Crisis 15 (11.5%) 61 (46.9%) 54 (41.5%)
Central Asian Crisis 13 (10.0%) 61 (46.9%) 56 (43.1%)
Latin American Crisis 12 (9.3%) 95 (73.6%) 22 (17.1%)
Total 40 (10.3%) 217 (55.8%) 132 (33.9%)
Table 2 Descriptive data for independent variables used in the analyses
N Mean Median Minimum Maximum SD
Ideology
a
129 3.65 3 1 8 1.86
Confidence 129 19.00 19 9 26 3.86
Optimism 124 38.46 39 16 50 6.71
Locus of Control 129 15.64 15 9 25 3.19
Courage 127 33.73 34 13 49 5.98
Leadership 129 33.88 34 10 50 7.40
Agreeableness 128 38.66 39 20 50 5.90
Conscientiousness 126 34.43 34 17 50 7.67
Emotional Stability 128 35.55 36 16 50 7.54
Extraversion 127 33.08 33 10 49 7.24
Intellect 128 38.70 39 23 50 5.98
a
The 7-point liberal-conservative scale
Hum Nat (2012) 23:98–126 107
One interesting question was whether subjects’“optimism”scores, a measure of
general optimism in everyday life, would reflect their confidencein their chosen solutions
to the crisis scenarios. Optimism was correlated (at a marginal level of significance) with
confidence (Pearson’sr00.16, N0125, p00.075). This suggests that everyday opti-
mism scores that psychologists report may be an important contributing factor for
understanding confidence in political contexts. However, “optimism”by itself was
not related to aggression (Kendall’sτ
B
0−0.06, N0124, p00.42), so the relationship
found between confidence and aggression—as we originally hypothesized in our
research design—appears to be specific to the context (the crisis scenarios and the
chosen policies), rather than a result of individuals’general levels of optimism.
Aggression and Party Identification
Figure 3shows that aggression was higher among self-declared Republicans (N019,
mean04.32) than Democrats (N054, mean03.33) or Independents (N033, mean0
3.94). The difference between Republicans and Democrats was significant (Mann-
Whitney U-test: Z03.18, N=54,19, p< 0.001) but the difference between Republicans
and Independents was not (Mann-Whitney U-test: Z01.42, N= 33,19, p00.16). This
reflects the fact that it is not simply that Republicans are more aggressive than all other
people in the sample. Rather, Republicans are somewhat more aggressive than average,
while Democrats are somewhat less aggressive than average (the difference between
Democrats and Independents was significant in the other direction—Mann-Whitney U-
test: Z02.32, N=54,33, p00.020).
Within individual scenarios, Republicans were more aggressive than Democrats in
all cases (Fig. 4). This was not far from significance in the Hostage crisis (Mann-
Whitney U-test: Z01.67, N=55,19, p00.095), highly significant in the Central Asian
Confidence
25201510
Aggression
7
6
5
4
3
2
1
0
Fit line for Total
0
2
4
6
8
10
Scale
Fig. 2 Aggression increases with confidence that the chosen policy will succeed (Kendall’sτ
B
00.19, N0
129, p00.006). Dot size represents the number of individuals with identical data values
108 Hum Nat (2012) 23:98–126
crisis (Z03.61, N=55,19, p<0.001), and not significant in the Latin American crisis
(Z00.77, N=54,19, p00.44).
Aggression and Political Ideology
We also found a strong correlation between aggression and the seven-point liberal-
conservative scale (Kendall’sτ
B
00.31, N0121, p<0.0001; Fig. 5). Category 8 “Don’t
Know”responses were excluded from these analyses. Apart from the “Moderate/
Middle of the Road”category, there was a perfectly monotonic increase in aggression
along the liberal to conservative spectrum.
Fig. 3 Aggression and party
affiliation. Republicans were
significantly more aggressive
than Democrats (Mann-Whitney
U-test: Z03.18, p<0.001) but not
Independents (Mann-Whitney U-
test: Z01.42, p00.16)
Fig. 4 Aggression and party affiliation by crisis scenario. Republicans were more aggressive than
Democrats in all cases. This was marginally significant in the Hostage crisis (black bars; Mann-Whitney
U-test: Z01.67, p00.095), highly significant in the Central Asian crisis (gray bars; Z03.61, p<0.001), but
not significant in the Latin American crisis (white bars; Z00.77, p< 0.001)
Hum Nat (2012) 23:98–126 109
Figure 6shows the pattern within each scenario. Aggression was significantly
correlated with ideology in the Hostage crisis (Kendall’sτ
B
00.18, N0122, p00.022)
and the Central Asian crisis (Kendall’sτ
B
00.35, N0122, p<0.0001), but not in the
Latin America Crisis (Kendall’sτ
B
00.05, N0121, p00.54).
Fig. 5 Aggression increases significantly with score on the liberal-conservative scale (Kendall’sτ
B
00.31,
N0121, p<0.0001; “Don’t Knows”excluded from statistical test)
Fig. 6 Aggression and political ideology by crisis scenario. Aggression was significantly correlated with
ideology in the Hostage crisis (black bars; Kendall’sτ
B
00.18, N0122, p00.022) and the Central Asian
crisis (grey bars; Kendall’sτ
B
00.35, N0122, p<0.0001), but not in the Latin America Crisis (white bars;
Kendall’sτ
B
00.05, N0121, p00.54). Dotted bars represent the “Don’t Know”category
110 Hum Nat (2012) 23:98–126
To summarize, people who chose more aggressive options were significantly more
likely to have conservative leanings, whether measured by political party affiliation or
a more general liberal-conservative scale.
Aggression and U.S. Foreign Policy Preferences
Do policy choices in these crisis scenarios tell us anything about subjects’real-world
foreign policy preferences? To answer this question, we asked subjects about their
opinions on Iraq (prior to the 2007 “surge”) and Iran. The first question was “The United
States is sending about 20,000 troops to Iraq. Do you think this will improve security and
reduce killings, or that this will not improve security and reduce killings?”Those who
responded that the surge would “improve security and reduce killings”had significantly
higher aggression scores in the crises scenarios than those who thought that it would not
(Mann-Whitney U-test, Z03.24, N031,69, p00.001; Fig. 7). This was also significant
or marginally significant when only using data from the Hostage crisis (Z01.74, N0
31,70, p00.08), Central Asian crisis (Z02.53, N031,70, p00.01), or Latin American
crisis (Z01.72, N031,69, p00.09).
We also asked “If it is proven that Iran is helping the Shiites in Iraq, would you
favor or oppose bombing Iran over this?”Preference for bombing Iran was positively
correlated with aggression scores in the crisis scenarios (Kendall’sτ
B
00.28, N0114,
p<0.001; Fig. 8). This relationship was also evident when only using data from the
Hostage crisis (Kendall’sτ
B
00.25, N0115, p00.004) and the Central Asian crisis
(Kendall’sτ
B
00.29, N0115, p00.001), but not the Latin American crisis (Kendall’s
τ
B
00.002, N0114, p00.98).
Interestingly, confidence expressed in the crisis scenarios was also significantly
higher among those who thought the surge would help in Iraq (Z02.62, N031,70, p0
0.009), and positively and marginally significantly correlated with a preference for
bombing Iran (Kendall’sτ
B
00.15, N0115, p00.07).
Unsurprisingly, liberal-conservative scores predicted both preference for the
surge in Iraq (Mann-Whitney U-test, Z03.33, N070,33, p00.001) and bombing
Iran (Kendall’sτ
B
00.32, N0117, p<0.0001).
3
This pattern was also highly signifi-
cant when comparing Republican and Democrat categories (surge: χ
2
015.8, df01, p
<0.0001; bombing Iran: χ
2
019.5, df03, p<0.001).
4
Relationship Between Confidence and Partisanship
Confidence and the liberal-conservative scale were correlated, though not significant-
ly, with each other (Kendall’sτ
B
00.95, N0130, p00.15), such that more conserva-
tive people tended to be more confident that their chosen policy (whatever it was)
3
Here, comparing the four-level preference for bombing Iran across the seven-level liberal-conservative
scale might be better tested using a chi-square test, but the result is the same: χ
2
051.2, df018, p< 0.0001.
4
Note that since Republicans/conservatives were more aggressive in the crisis scenarios, it may seem
unsurprising that aggression should also correlate with preferences in Iraq and Iran—given that both were
associated with President George W. Bush’s presidency. While certainly a plausible explanation for these
specific results, it further highlights the central finding that conservatives were more aggressive in the crisis
scenarios.
Hum Nat (2012) 23:98–126 111
would succeed. This association was more striking when comparing the two main
party affiliations. Republicans were significantly more confident than Democrats
(Mann-Whitney U-test: Z02.74; N019,55; p00.006).
If we look within party-affiliation categories (note the smaller sample sizes), then
the correlation between confidence and aggression was positive but not significant
among Republicans (Kendall’sτ
B
00.04, N019, p00.82), Democrats (Kendall’sτ
B
0
0.05, N055, p00.66), and “Other”(Kendall’sτ
B
00.13, N011, p00.62); positive and
significant among Independents (Kendall’sτ
B
00.33, N033, p00.02); and negative
but not significant among “Don’t Knows”(Kendall’sτ
B
0−0.06, N011, p00.80).
This suggests that, apart from Independents, the strong relationship between confi-
dence and aggression (Fig. 2) is driven by difference: across political preferences. In
the next section, we conduct multivariate analyses to determine whether partisanship
and confidence can independently explain variance in aggression.
Fig. 8 Aggression in the scenarios was higher among those who favored bombing Iran if it was found to be
assisting Shiites in Iraq (Kendall’sτ
B
00.28, N0114, p<0.001)
Fig. 7 Aggression in the scenar-
ios was significantly higher
among those who thought that
the “surge”would improve se-
curity and reduce killings in Iraq
(Mann-Whitney U-test, Z03.24,
N031,69, p00.001)
112 Hum Nat (2012) 23:98–126
Before moving on to the multivariate analyses, it is worth exploring one alternative
hypothesis. One might expect that people—of any party—should show greater
confidence as policy options become more aggressive, because if more aggressive
actions are potentially more costly, one should perhaps be more confident about the
outcome before engaging in them (we call this the “cost hypothesis”). We see three
reasons to reject this hypothesis.
First, it can be argued that the indirect costs of concession or negotiation may be
just as high as the immediate and more obvious costs of military conflict, and we have
no reason to believe that this did not factor into subjects’decisions.
Second, even if the cost hypothesis is partly explaining the results, it cannot account
for the difference in confidence between people of different political preferences. If
people who choose more aggressive options are more confident because such options
are deemed more costly, then why is this more true for Republicans or conservatives
than Democrats or liberals?
Third, we can test the cost-hypothesis by asking whether more-conservative
individuals are more confident (than less-conservative individuals) when comparing
people who chose the same specific policy (concession, negotiation, or military
attack) in a given scenario. If confidence were determined primarily by how aggres-
sive the chosen policy option is, then we should not expect a difference between, say,
Republicans and Democrats when comparing their confidence in the same policy
choice. We looked at this by comparing Republicans and Democrats split by both
scenario and by each specific policy choice (asking for example, whether Repub-
licans were more confident than Democrats when they both chose to negotiate in the
hostage scenario). Republicans were more confident in every single case but one (the
exception being that Republicans were less confident than Democrats about the attack
option in the Latin American scenario). None of these differences were statistically
significant (Mann-Whitney U-tests: all Z< 1.75, all p>0.08). However, note that the
lack of significance here is likely due to the sample sizes becoming very small once
we split the data both by scenario and by policy choice (range: 8–53 subjects per test),
and because these analyses were limited to Republicans versus Democrats (Indepen-
dent, Other, and Don’t Know subjects were excluded).
Overall, there is no compelling evidence for the cost hypothesis, and Repub-
licans were systematically, though not significantly (in small sample tests),
more confident across the board in all of their policy choices than their
Democratic counterparts.
Multivariate Analysis
We next used an ordered probit model to examine the effect of multiple independent
variables on aggression. Table 3shows the results of ordered probit models for (A)
aggression pooled across all scenarios, (B) the Hostage crisis, (C) the Central Asian
crisis, and (D) the Latin American crisis. Each of these models has three versions:
model 1 includes confidence but not ideology or party identification, model 2
includes confidence and ideology, and model 3 includes confidence and party
identification. All models include all nine personality inventories as controls, but to
ease presentation we omit these results (few of these variables were significant in any
of the models, but wherever they were significant this is stated in the text below).
Hum Nat (2012) 23:98–126 113
Table 3 Ordered probit models for aggression pooled across all scenarios (model A1,2,3), the Hostage crisis (B1,2,3), the Central Asian crisis (C1,2,3), and the Latin American
crisis (D1,2,3). Models only include subjects that answered the ideology question (i.e., they exclude “Don’t Knows”). Political party identification entered as a dummy variable
(10Republican, 0 0Democrat). Personality inventories were included as independent variables but are omitted from the table for clarity of presentation
Model A Model B Model C Model D
Pooled1 Pooled2 Pooled3 Hostage1 Hostage2 Hostage3 Asian1 Asian2 Asian3 Latin1 Latin2 Latin3
PooledConfidence 0.072** 0.065* 0.066+
[0.026] [0.026] [0.039]
HostageConfidence 0.008 −0.041 −0.066
[0.063] [0.070] [0.105]
AsianConfidence 0.109+ 0.096 0.129
[0.066] [0.064] [0.087]
LatinConfidence 0.134* 0.137* 0.079
[0.060] [0.061] [0.114]
Ideology 0.244** 0.222** 0.312** 0.033
[0.066] [0.081] [0.089] [0.090]
Party 0.874* 0.579 1.205** 0.441
[0.340] [0.392] [0.447] [0.439]
cut1
Constant −1.116 −0.541 −1.388 −2.401+ −1.992 −2.812 −0.287 0.481 0.192 0.120 0.243 0.018
[1.306] [1.348] [1.968] [1.361] [1.372] [1.833] [1.319] [1.453] [1.998] [1.044] [1.051] [1.469]
cut2
Constant 0.075 0.657 0.098 −0.887 −0.422 −1.198 1.330 2.233 2.083 2.620* 2.747** 2.447
[1.188] [1.235] [1.843] [1.343] [1.349] [1.812] [1.307] [1.447] [2.003] [1.046] [1.051] [1.491]
cut3
Constant 1.114 1.752 1.146
114 Hum Nat (2012) 23:98–126
Table 3 (continued)
Model A Model B Model C Model D
Pooled1 Pooled2 Pooled3 Hostage1 Hostage2 Hostage3 Asian1 Asian2 Asian3 Latin1 Latin2 Latin3
[1.166] [1.216] [1.842]
cut4
Constant 1.885 2.587* 2.024
[1.162] [1.222] [1.852]
cut5
Constant 3.030* 3.802** 3.354+
[1.209] [1.300] [1.924]
Observations 111 111 67 112 112 68 112 112 68 111 111 67
bic 392.346 385.820 249.737 261.436 259.056 179.433 253.400 244.119 160.264 208.110 212.659 148.732
Standard errors in brackets
+p<0.10, * p< 0.05, ** p<0.01.
Hum Nat (2012) 23:98–126 115
Excluding these personality inventories altogether from our multivariate analyses had
no substantive effect on our results.
Confidence is significant in the pooled data (A1) and Latin American crisis (D1),
marginal in the Central Asian crisis (C1), and not-significant in the hostage crisis
(B1). Locus of control and agreeableness were also of marginal significance in the
pooled model (A1). In the pooled data, confidence remains significant when ideology
is added (which is also significant itself; A2), and marginally significant when party
identification is added (which is again significant itself; A3).
Coefficients in nonlinear models such as ordered probit are not directly interpret-
able. Instead, we calculated the substantive effects generated by changes in “ideolo-
gy”and “confidence”variables for the second pooled model (A2) using the “Clarify”
algorithm (King et al. 2000). Here we ask: what is the change in predicted probability
of being in each category of the dependent variable (aggression) when one variable
(either ideology or confidence) moves from its sample 25th percentile level to the
75th percentile level and all other variables are held at their median? We plot these
predicted changes along with associated confidence intervals in Fig. 9. For example,
for both the ideology and confidence variables, the probability of being in category 4
(that is, scoring 4 on aggression) increases by nearly 10%. Increasing the change in
the independent variable of interest (e.g., from the minimum to maximum) of course
−.2 −.1 0 .1 .2
Probability Change
1 2 3 4 5 6
Aggression Category
Ideology
−.2 −.1 0.1 .2
Probability Change
1 2 3 4 5 6
Aggression Category
Confidence
Fig. 9 Effects of changes in ideology and confidence for the pooled ordered probit model (model A2 in
Table 3), using the “Clarify”algorithm (King et al. 2000). The lines show the change in the predicted
probability of being in each category of the dependent variable (aggression) when the ideology variable
(upper panel) or confidence variable (lower panel) moves from its 25th to 75th percentile level and all other
variables are held at their median. For both variables, the probability of being in category 5, for example,
increases by nearly 10%
116 Hum Nat (2012) 23:98–126
increases these estimates. This gives us further evidence that these variables are not
only statistically significant but also produce reasonable effect sizes.
Robustness Checks
We emphasize the ordered probit results because the dependent variable represents
ordered categories. However, the seven categories of the dependent variable (aggres-
sion) could be treated as a scalar variable. Therefore, here we check results using an
ANOVA model, which is relatively robust to deviations from normality—with
appropriate caution in interpreting the results. All models included all nine person-
ality variables.
To mirror the ordered probit models above, ANOVA Model A included confidence
but neither party nor ideology. The overall ANOVA model was marginally significant
(F02.22; df010,105; p00.061; adjusted R
2
00.07), with a single significant predictor
variable: confidence (F07.68; df01,105; p00.007). Agreeableness was of marginal
significance (F03.63; df01,105; p00.059).
Model B included confidence and ideology (but excluded party identification).
This was the best-fitting ANOVA model, significant overall (F03.36; df011,99; p0
0.002, adjusted R
2
00.17), with two significant predictor variables: the liberal-
conservative scale (F011.36, df01p00.001) and confidence that the policy would
succeed (F05.68, df01p00.019).
Model C included confidence and party identification (but excluded ideology).
Party was a categorical variable constrained to Democrat, Independent, and Repub-
lican. The overall ANOVA model was significant (F02.22; df012,82; p00.018,
adjusted R
2
00.13), with three significant predictor variables: party (F03.91; df 0
2,82; p00.024), confidence (F04.58; df01,82; p00.035), and aggreeableness (F0
4.25; df01,82; p00.043).
To summarize, a variety of methods confirmed the same basic result: people’s level
of aggression in responding to hypothetical crises was significantly predicted by their
confidence that their chosen policy would succeed, by their score on the liberal-
conservative scale, by their political party affiliation, and by their preference for the
use of military force in real-world U.S. policy toward Iraq and Iran. These results
were supported whether looking at individual variables, multivariate ordered probit
models, or alternative ANOVA models.
Discussion
Our study identified two consistently significant factors in explaining aggression:
confidence and conservatism. Subjects who chose more-aggressive options—military
action rather than negotiation or negotiation rather than appeasement—were signif-
icantly more likely to believe that their favored policy would be successful, to have
conservative leanings, and to declare themselves as Republicans. These aggressive
policy preferences were reflected in their views on real-world current events as well.
Subjects who chose more-aggressive options (and who were more confident) in the
hypothetical scenarios were more likely to believe that the surge in Iraq would
improve security, and more likely to favor bombing Iran.
Hum Nat (2012) 23:98–126 117
There are obviously many limitations of this kind of experiment. First, hypothet-
ical scenarios can rarely be completely divorced from real-world political events. In
some ways this is useful because it makes the scenarios more salient and realistic. On
the other hand, subjects may have associated a given scenario with the party then in
power (e.g., Republican George W. Bush’s administration) and thus favored whatever
strategy they perceived their own political party to support at the time of the study.
The experiments, conducted in May 2007, may also have been influenced by the U.S.
election campaign and an increased likelihood of partisan thinking. However, this
may have served to increase salience and political knowledge rather than introducing
a bias in any particular direction.
Second, because our findings are correlational, we do not know whether people
who were more disposed to confidence were, as a consequence, more likely to chose
the most aggressive option or, alternatively, whether people who chose aggressive
options subsequently became more likely to believe it would work. Because we asked
people’s level of confidence after they had made their policy choice, at first glance
this suggests a role for cognitive dissonance in rationalizing decisions that have
already been made (Cooper 2007; Festinger 1957; Jervis 1976). However, that does
not predict greater confidence in decisions that were more aggressive, only greater
confidence in whatever decision was made (aggressive or otherwise). Therefore,
cognitive dissonance cannot explain our results.
Despite some important limitations, the present study suggests that confidence and
conservatism are associated with aggression in crisis decision-making. Confidence
has long been claimed to underlie aggression in historical case studies of war (Blainey
1973; Johnson 2004; Van Evera 1999), but quantitative and experimental evidence
has been lacking. We previously found support for such a link in controlled wargame
experiments (Johnson et al. 2006), but we did not know until the present study
whether decision-makers’levels of confidence would predict aggressive policy
preferences in plausible real-world crisis decisions (rather than in a competitive
game against another player). As we show here, although Republicans and
conservatives were more confident than Democrats and liberals, confidence
explains variance in aggression independently of the variance explained by
partisanship. Confidence, it seems, is a powerful factor that exerts an influence
beyond political ideology. Interestingly, even where political preferences failed
to explain variation in aggression—in the Latin American scenario—confidence
still was a significant predictor of aggression (see bivariate and probit model
results).
The Latin American crisis was the one exception to an otherwise consistent pattern
of results. Absolute levels of aggression in this scenario were lower than in the other
scenarios—across all political parties and across all levels of the liberal-conservative
scale. Military action may thus have been seen by all subjects, across partisan divides,
as potentially expedient and effective in the Hostage and Central Asia cases but not in
Latin America. Subjects may have approached policy options in this scenario more as
a reflection on past U.S. administrations’policies in the region, rather than on the
basis of their own ideology. Of course, it could also be the case that subjects did not
find the region of Latin America to be as important, or to require as active an
intervention, as more strategicially critical regions such as the Middle East and
Central Asia. Alternatively, since Latin America is geographically closer to our
118 Hum Nat (2012) 23:98–126
subjects, participants may be more concerned about potential spillover effects from
military action in this region affecting their own security and quality of life. Given the
number of Hispanics in this sample (11.9%), some individuals may also have had
family members in this region whose safety or interests remained prominent in
thinking through which options they might support.
It is common to associate conservatism with aggression in the present political
climate given the Bush administration’s foreign policy record and the 2002 National
Security Strategy. Conservatism, however, has historically been associated with isola-
tionism, containment, and offshore balancing (Dueck 2008; Mearsheimer 2001), and
many conservative thinkers express the opinion that the United States should not be
the world’s“policeman”. Thus, conservatives are suspicious of the use of force in
certain domains. Nevertheless, recent research suggests that conservatives are psy-
chologically, physiologically, and neurologically more sensitive to uncertainty and
threats in the environment. Our study adds to this literature by demonstrating that
conservatives are also significantly more likely to react to threats aggressively, even
in differing political contexts. Robert Draper summed up the essence of George W.
Bush’s presidency, dominated by the wars in Afghanistan and Iraq, in two words:
“dead certain”(Draper 2007). Bush’s decision-making style was characterized by
both strong conservative ideals and an unwavering belief in ultimate success. Our
study suggests that this phenomenon is not confined to Bush or the neo-cons.
Confidence and conservatism are strong predictors of aggression.
Of course, although confidence and conservatism may lead to more aggressive
behavior, this does not mean that such behavior is necessarily ineffective or undesir-
able. The use of force may sometimes be a better policy than negotiation or appease-
ment (e.g., against Hitler). Mainstream international relations theory stresses the fact
that in international politics there is no world government or Leviathan to control
what states do. In such an “anarchic”world, aggression may be unfortunate but
essential for survival (Mearsheimer 2001; Waltz 1979). For similar reasons, aggres-
sion is likely to have been an adaptive strategy in human evolutionary history to avoid
exploitation and maximize survival and reproductive success (Gat 2006; Keeley
1996; LeBlanc and Register 2003; Wrangham and Peterson 1996). High levels of
confidence and conservatism may have been an advantageous trait in this evolution-
ary history of intra- and intergroup conflict at least for some individuals, increasing
the probability of victory by boosting resolve, morale, perseverence, and the credi-
bility of deterrence (Johnson 2004; Johnson and Fowler 2011; Johnson et al. 2011;
Trivers 2000; Wrangham 1999; Sell et al. 2009).
A problem arisis because human proximate mechanisms underlying aggression
were calibrated for the prevailing distributions of costs, benefits, and risks in the
social and physical environment of the Pleistocene. In today’s very different world,
high levels of confidence or conservatism may be maladaptive when triggered in
modern settings involving mass armies, modern weapons, and decisions made far
from the battlefield—none of which existed in the ancestral environment in which our
brains evolved.
Confidence and conservatism might continue to be adaptive, however, if they are
triggered in settings that are evolutionarily relevant, where evolved signals and
feedback mechanisms work in similar ways to the way they did in the past. For
example, garnering public support to resist invasions or persevere with long wars
Hum Nat (2012) 23:98–126 119
might fall under this category. Evolutionarily relevant signals might, therefore pro-
mote acts of military and political aggression that improve one’s strategic position or
combat effectiveness. As Carl von Clausewitz wrote in the early nineteenth century,
“Boldness in war . . . must be granted a certain power over and above successful
calculations involving space, time, and magnitude of forces, for wherever it is
superior, it will take advantage of its opponents’weakness”(von Clausewitz
1976:190). The relevant question, therefore, is not whether conservatives are “too”
aggressive and “too”confident, but rather: when are conservative hawks or liberal
doves more likely to make the most effective policy choices? And what conditions
are most likely to engender support from both sides of the political divide? These
pose interesting new avenues for future experiments to further our understanding of
human aggression.
Acknowledgments We thank David Carrier and three anonymous referees for their help in improving the
manuscript. DJ thanks the Branco Weiss Society In Science Fellowship for funding. Finally, we especially
thank Elizabeth Cashdan for her encouragement and advice on the manuscript.
Appendix
International Crises Scenarios
For each of the following scenarios, please read the description carefully. Then please
make a decision about what you would do in this situation. As much as possible, try
to really imagine being the leader confronted with this decision. Do not act as if you
were any particular leader, but rather act as yourself in the role of leader. In each
decision, you will be asked to mark your most preferred decision option. There is no
right answer to any of these crisis scenarios.
[New page in original]
Hostage Scenario
A commercial airliner en route from New York to Tel Aviv, Israel, is hijacked and
forced to land at the Riyadh International Airport in Saudi Arabia. The passengers are
predominantly Israeli, but a number of Americans and other nationals are on board as
well.
You know that the hijackers are Palestinians, but their specific affiliation is
not clear at this point. They are demanding the release of a number of
Palestinians held in Israeli jails. They also want safe passage for themselves
to an undisclosed location.
The relevant parties are Israel, the Palestinian hijackers, the Saudis, and the
Americans. Each group has a set of objectives. The Israelis primarily want the safe
return of the passengers. They also want to maintain the credibility of their deterrent
against terrorism and to make sure that their strategic interests remain unchanged. The
Palestinian hijackers primarily want the release of their nationals from Israeli prisons.
They also want to damage Israel’s image internationally and to assure their own safe
passage out of Saudi Arabia. The Saudis primarily want to demonstrate their control
120 Hum Nat (2012) 23:98–126
of the situation and to maintain control among their own people. They would also like
to maintain the status quo in their relations with other Arab nations. The Americans
would like to assure the safety of their passengers, deter similar terrorist actions in the
future, and broker this situation to a successful conclusion to demonstrate
international leadership.
QUESTION A
As the president of the United States, faced with this situation, what is your most
preferred option for action?
__ 1. Encourage a Special Forces assault in an attempt to rescue the hostages.
__ 2. Act as a broker between Israel and the Palestinian hijackers to arrange a
trade, in which passengers are released in small groups, in return for the release of a
few Palestinian prisoners.
__ 3. Put pressure on the Israelis to release the Palestinian prisoners in order to
secure the release of the passengers unharmed.
QUESTION B
On a scale of 0 to 10, how likely do you think it is that your chosen course of
action will succeed? Please circle your response, where 0 is very unlikely and 10 is
very likely.
012345678910
[New page in original]
Central Asian Scenario
The newly established democratic government of Afghanistan has become increas-
ingly unpopular due to the continuing strength of Islamic fundamentalism, economic
decline, internal political dissension and violence, and their crackdown on the highly
profitable heroin trade. An armed rebellion by millennia-old clan-based [groups led
by] warlords has been continuing in the southern and eastern regions. When govern-
ment troops encircle some of these rebel forces, the resurgent Taliban, in concert with
the local warlords, intervene with a sizeable military force. On the third day of the
operation, it seems possible from the size and direction of these attacks, that these
Taliban forces will once again take over the capital city of Kabul. The resurgent
Taliban are calling for an “international brotherhood of Islamic states,”while the
Russians warn that “Islamic fundamentalism will not be allowed to set up positions of
strength along the Russian border.”
U.S. pentagon military strategists are pessimistic about the ability of the United
States to successfully control this military action given the number of American
troops that are currently pinned down in Iraq. They see a slim hope for success only if
the United States seizes a window of opportunity to interdict key mountain passes and
to forestall the resurgent Taliban from setting up air defenses and logistics to support
their further operations. Renewed major U.S. military intervention might require the
insertion of special mountain forces, conventional bombing, and the possible use of
15,000-ton “daisy cutter”bombs. The U.S. dominates in aerial forces, but is in an
unfavorable position in the local military balance on the ground. The latest intelli-
gence data suggests that the mountainous border region between Afghanistan and
Pakistan continues to shelter most of the top leaders of Al Queda, including Osama
Hum Nat (2012) 23:98–126 121
bin Laden. Many Taliban leaders support these terrorists in their campaign against
American forces in the region.
The Russians’main goal is to prevent an unstable situation on its border, but the
Russian government would not mind extending its influence throughout the region as
well. The United States is reluctant to let the Russians do all the work, in case they
take over the territory entirely and move into nearby areas that control important oil
pipelines necessary to both the U.S. and Western Europe.
QUESTION A
As the president of the United States, faced with this situation, what is your most
preferred option for action?
__ 1. Withdraw American forces from the increasing difficulties in Afghanistan and
concede the area to the local Taliban forces, concentrating your military effort on Iraq.
__ 2. Use special mountain forces, conventional bombing, and the possible use of
15,000-ton “daisy cutter”bombs to block the key mountain passes in Afghanistan.
__ 3. Stress diplomatic solutions to the crisis and act to sponsor ceasefire nego-
tiations between the local government, important warlords, and the Russian and
Pakistani governments.
QUESTION B
On a scale of 0 to 10, how likely do you think it is that your chosen course of action
will succeed? Please circle your response, where 0 is very unlikely and 10 is very likely.
012345678910
[New page in original]
Latin American Scenario
A country in Latin America recently elected a socialist leader who has just socialized the
nation’s industries, including a number of important American corporate concerns. This
country is the leading exporter of coffee, cocoa, and bananas to the United States. In
addition, it is an important supplier of certain raw materials, including copper. As a
result, American companies stand to lose billions of dollars, and adequate supplies of
these goods into the United States could be disrupted, likely leading to critical shortages
of coffee, chocolate, fruit, and plumbing material in America. A strong military leader
who is quite sympathetic to American interests is waiting in the wings, eager to take
power from the elected socialist leader, but he needs outside help in order to accomplish
this goal. As president of the United States, you have been under a great deal of pressure
from major corporate businesspeople and important political donors to do something
about this situation, and to roll back the government takeover of these American
companies. These leaders have threatened to back your opponent in the next election
unless you do something to reverse this seizure of American property.
QUESTION A
As the president of the United States, faced with this situation, what is your most
preferred option for action?
__ 1. Attempt to negotiate fair restitution for American companies.
__ 2. Let this leader nationalize American interests, but cut diplomatic ties.
__ 3. Allow US special forces and the CIA to assist the local military leader in
staging a coup.
122 Hum Nat (2012) 23:98–126
QUESTION B
On a scale of 0 to 10, how likely do you think it is that your chosen course of
action will succeed? Please circle your response, where 0 is very unlikely and 10 is
very likely.
012345678910
References
Achen, C. H. (2002). Parental socialization and rational party identification. Political Behavior, 24(2), 151–
170.
Akerlof, G. A. & Shiller R. J. (2009) Animal Spirits: How Human Psychology Drivers the Economy, and
Why it Matters for Global Capitalism (Princeton: Princeton University Press).
Adorno, T. W., Frenkel-Brunswik, E., Levinson, D. J., & Sanford, R. N. (1950). The authoritarian
personality. New York: Harper and Row.
Alford, J. R., Funk, C. L., & Hibbing, J. R. (2005). Are political orientations genetically transmitted?
American Political Science Review, 99(2), 153–167.
Altemeyer, B. (1988). Enemies of freedom: Understanding right-wing authoritarianism. San Francisco:
Jossey-Bass.
Amodio, D. M., Jost, J. T., Master, S. L., & Yee, C. M. (2007). Neurocognitive correlates of liberalism and
conservatism. Nature Neuroscience, 10, 1246–1247.
Armor, D. A., & Taylor, S. E. (1998). Situated optimism: specific outcome expectancies and self-
regulation. Advances in Experimental Social Psychology, 30, 309–379.
Barber, B. M., & Odean, T. (2001). Boys will be boys: gender, overconfidence, and common stock
investment. Quarterly Journal of Economics, 116, 261–292.
Barkow, J. H., Cosmides, L., & Tooby, J. (Eds.). (1992). The adapted mind: Evolutionary psychology and
the generation of culture. Oxford: Oxford University Press.
Baumeister, R. F. (1989). The optimal margin of illusion. Journal of Social and Clinical Psychology, 8,
176–189.
Ben-David, I., Graham, J. R., & Harvey, C. R. (2006). Managerial overconfidence and corporate policies.
Fuqua School of Business, Duke University, Working Paper No. 81.
Bennett, D. S., & Stam, A. C. (2004). The behavioral origins of war. Ann Arbor: University of Michigan
Press.
Blainey, G. A. (1973). The causes of war. New York: Free Press.
Camerer, C., & Lovallo, D. (1999). Overconfidence and excess entry: an experimental approach. The
American Economic Review, 89(1), 306–318.
Cooper, J. (2007). Cognitive dissonance: 50 years of a classic theory. New York: Sage.
Draper, R. (2007). Dead certain: The presidency of George W. Bush. New York: Free Press.
Dueck, C. (2008). Reluctant crusaders: power, culture, and change in American grand strategy.
Dunning, D., Johnson, K., Ehrlinger, J., & Kruger, J. (2003). Why people fail to recognize their own
incompetence. Current Directions in Psychological Science, 12(3), 83–87.
Ehrlinger, J., Johnson, K., Banner, M., Dunning, D., & Kruger, J. (2008). Why the unskillled are unaware.
Organizational Behavior and Human Decision Processes, 105(1), 98–121.
Etheredge, L. (1978a). A world of men: The private sources of American foreign policy. Cambridge: MIT
Press.
Etheredge, L. (1978b). Personality effects on American foreign policy, 1898–1968. American Political
Science Review, 72(2), 434–451.
Fallows, J. (2004). Blind into Baghdad. The Atlantic (January/February), 53–74.
Fearon, J. D. (1995). Rationalist explanations for war. International Organization, 49(3), 379–414.
Feaver,P., & Gelpi, C. (2004). Choosing your battles: American civil-military relations and the use of
force. Princeton: Princeton University Press.
Festinger, L. (1957). A theory of cognitive dissonance. Stanford: Stanford University Press.
Fiske, S. T., & Taylor, S. E. (2007). Social cognition: From brains to culture. New York: McGraw-Hill.
Fowler, J. H., Settle, J. E., & Dawes, C. T. (2009). The heritability of partisan attachment. Political
Research Quarterly, 62(3), 601–613.
Gartner, S. S. (1997). Strategic assessment in war. New Haven: Yale University Press.
Gat, A. (2006). War in human civilization. Oxford: Oxford University Press.
Hum Nat (2012) 23:98–126 123
Gollwitzer, P. M. (2011). Mindset theory of action phases. In P. A. M. Van Lange, A. W. Kruglanksi, & E.
T. Higgins (Eds.), Handbook of theories of social psychology. London: Sage.
Griffin, D. W., & Tversky, A. (2002). The weighing of evidence and the determinants of confidence. In T.
Gilovich, D. Griffin, & D. Kahneman (Eds.), Heuristics and biases: The psychology of intuitive
judgment (pp. 230–249). Cambridge: Cambridge University Press.
Hatemi, P. K., Hibbing, J., & Alford, J. (2009). Is there a "party" in your genes? Political Research
Quarterly, 62(3), 584–600.
Hatemi, P. K., et al. (2011). A genome-wide analysis of political attitudes. Journal of Politics, 73,1–15.
Horowitz, M., McDermott, R., & Stam, A. (2005). Leader age, regime type and violence. Journal of
Conflict Resolution, 49(5), 661–685.
Jervis, R. (1976). Perception and misperception in international politics. Princeton: Princeton University
Press.
Jervis, R. (2003). The confrontation between Iraq and the US: implications for the theory and practice of
deterrence. European Journal of International Relations, 9(2), 315–337.
Johnson, D. D. P. (2004). Overconfidence and war: The havoc and glory of positive illusions. Cambridge:
Harvard University Press.
Johnson, D. D. P., & Fowler, J. (2011). The evolution of overconfidence. Nature, 477,317–320.
doi:10.1038/nature10384.
Johnson, D. D. P., & Tierney, D. R. (2006). Failing to win: Perceptions of victory and defeat in
international politics. Cambridge: Harvard University Press.
Johnson, D. D. P., & Tierney, D. R. (2007). In the eye of the beholder: victory and defeat in U.S. military
operations. In J. Angstrom & I. Duyvesteyn (Eds.), Understanding victory and defeat in contemporary
war (pp. 46–76). London: Routledge.
Johnson, D. D. P., & Tierney, D. R. (2011). The Rubicon theory of war: how the path to conflict reaches the
point of no return. International Security, 36,7–40.
Johnson, D. D. P., McDermott, R., Barrett, E., Cowden, J., Wrangham, R., McIntyre, M., et al. (2006).
Overconfidence in wargames: experimental evidence on expectations, aggression, gender and testos-
terone. Proceedings of the Royal Society of London, Series B, 273(1600), 2513–2520.
Johnson, D. D. P., Weidmann, N. B., & Cederman, L.-E. (2011). Fortune favours the bold: an agent-based
model reveals adaptive advantages of overconfidence in war. PLoS One, 6(6), e20851.
Jost, J. T., Glaser, J., Kruglanski, A. W., & Sulloway, F. (2003). Political conservatism as motivated social
cognition. Psychological Bulletin, 129, 339–375.
Kahneman, D., & Renshon, J. (2006). Why Hawks Win. Foreign Policy (158, January/February), 34–38.
Kanter, R. M. (2004). Confidence: How winning streaks and losing streaks begin and end. New York:
Crown Business.
Keeley, L. H. (1996). War before civilization: The myth of the peaceful savage. Oxford: Oxford University
Press.
King, G., Tomz, M., & Wittenberg, J. (2000). Making the most of statistical analyses: improving
interpretation and presentation. American Journal of Political Science, 44, 341–355.
Kruger,J. M., & Dunning, D. (1999). Unskilled and unaware of it: how difficulties in recognizing one's
own incompetence lead to inflated self-assessments. Journal of Personality and Social Psychology, 77,
1121–1134.
LeBlanc, S., & Register, K. E. (2003). Constant battles: The myth of the peaceful, noble savage. New York:
St. Martin's Press.
Lebow, R. N. (1981). Between peace and war: The nature of international crisis. Baltimore: Johns Hopkins
University Press.
Levy, J. S. (1983). Misperception and the causes of war: theoretical linkages and analytical problems.
World Politics, 36(1), 76–99.
Levy, J. S., & Thompson, W. R. (2010). Causes of war. Oxford: Wiley-Blackwell.
Lim, R. G. (1997). Overconfidence in negotiation revisited. International Journal of Conflict Management,
8,52–79.
McDermott, R. (2004). The feeling of rationality: the meaning of neuroscientific advances for political
science. Perspectives on Politics, 2(4), 691–706.
McDermott, R. (2007). Presidential leadership, illness, and decision making. Cambridge: Cambridge
University Press.
McDermott, R., & Cowden, J. (2001). The effects of uncertainty and sex in a simulated crisis game.
International Interactions, 27, 353–380.
McDermott, R., & Cowden, J. (2008). The role of hostile communications in a crisis simulation game.
Peace and Conflict, 14, 151–167.
124 Hum Nat (2012) 23:98–126
McDermott, R., Johnson, D. D. P., Cowden, J., & Rosen, S. (2007). Testosterone and aggression in a
simulated crisis game. The Annals of the American Academy of Political and Social Sciences, 614(1),
15–33.
Mearsheimer, J. J. (2001). The tragedy of Great Power politics. New York: Norton.
Mueller, J. E. (1973). War, presidents and public opinion. New York: Wiley.
Oxley, D. R., Smith, K. B., Alford, J. R., Hibbing, M. V., Miller, J. L., Scalora, M., et al. (2008). Political
attitudes vary with physiological traits. Science, 321, 1667–1670.
Peterson, C. (2006). A primer in positive psychology. New York: Oxford University Press.
Rosen, S. P. (2004). War and human nature. Princeton: Princeton University Press.
Salvador, A., Suay, F., González-Bono, E., & Serrano, M. A. (2003). Anticipatory cortisol, testosterone and
psychological responses to judo competition in young men. Psychoneuroendocrinology, 28(3), 364–
375.
Schreiber, D., Simmons, A. N., Dawes, C. T., Flagan, T., Fowler, J. H., & Paulus, M. P. (2009). Red Brain,
Blue Brain: Evaluative Processes Differ in Democrats and Republicans. Paper presented at the
American Political Science Association annual meeting, Toronto. Available online at http://dmschrei-
ber.ucsd.edu/Publications/RedBrainBlueBrain.pdf.
Sell, A., Tooby, J., & Cosmides, L. (2009). Formidability and the logic of human anger. Proceedings of the
National Academy of Sciences, 106(35), 15073–15078.
Sell, A., Hone, L., & Pound, N. (2012). The importance of physical strength to human males. Human
Nature, 23(1), doi to be added in proofs. doi: doi:10.1007/s12110-012-9131-2
Stoessinger, J. G. (1998). Why nations go to war. New York: St. Martin's.
Taylor, S. E. (1989). Positive illusions: Creative self-deception and the healthy mind. New York: Basic
Books.
Taylor, S. E., & Brown, J. D. (1988). Illusion and well-being: a social psychological perspective on mental
health. Psychological Bulletin, 103, 193–210.
Taylor, S. E., & Brown, J. D. (1994). Positive illusions and well-being revisited: separating fact from
fiction. Psychological Bulletin, 116(1), 21–27.
Taylor, S. E., Lerner, J. S., Sherman, D. K., Sage, R. M., & McDowell, N. K. (2003). Portrait of the self-
enhancer: well adjusted and well liked or maladjusted and friendless. Journal of Personality and Social
Psychology, 84(1), 165–176.
Trivers, R. L. (2000). The elements of a scientific theory of self-deception. Annals of the New York
Academy of Sciences, 907,114–131.
Trivers, R. L. (2011). Deceit and self-deception: Fooling yourself the better to fool others. London: Allen
Lane.
Tuchman, B. W. (1984). The march of folly: from Troy to Vietnam. New York: Knopf.
Van den Steen, E. (2004). Rational overoptimism (and other biases). American Economic Review, 94,
1141–1151.
Van Evera, S. (1999). Causes of war: Power and the roots of conflict. Ithaca: Cornell University Press.
Vasquez, J. A. (1993). The war puzzle. Cambridge: Cambridge University Press.
Vasquez, J. A. (Ed.). (2000). What do we know about war? Lanham: Rowan & Littlefield.
von Clausewitz, C. (1976). On war (M. Howard & P. Paret, Trans.). Princeton: Princeton University Press.
(Originally published in 1832)
Walt, S. M. (2011). Wishful thinking: Top 10 examples of the most unrealistic expectations in contempo-
rary U.S. foreign policy. Foreign Policy, April 29, http://www.foreignpolicy.com/articles/2011/2004/
2029/wishful_thinking.
Waltz, K. N. (1979). Theory of international politics. New York: McGraw-Hill.
White, R. K. (1968). Nobody wanted war: Misperception in Vietnam and other wars.NewYork:
Doubleday.
Wilson, T. D. (2004). Strangers to ourselves: discovering the adaptive unconscious. Cambridge: Belknap.
Woodward, B. (2005). State of denial. NY: Simon & Schuster.
Wrangham, R. W. (1999). Is military incompetence adaptive? Evolution and Human Behavior, 20,3–17.
Wrangham, R. W., & Peterson, D. (1996). Demonic males: Apes and the origins of human violence.
London: Bloomsbury.
Zaller, J. (1992). The nature and origins of mass opinion. Cambridge: Cambridge University Press.
Dominic Johnson is a professor in politics and international relations at the University of Edinburgh, UK.
He received a DPhil from Oxford University in evolutionary biology and a PhD from Geneva University in
political science. Drawing on both disciplines, he is interested in how new research in biology and
Hum Nat (2012) 23:98–126 125
evolution challenges theories of politics and international relations. His current work focuses on the role of
evolutionary dynamics, evolutionary psychology, and religion in human conflict and cooperation. Addi-
tional information is available at http://dominicdpjohnson.com/.
Rose McDermott is a professor of political science at Brown University. She holds a PhD in political
science and an MA in experimental social psychology from Stanford University. She has held fellowships
at the Stanford Center for Advanced Study in the Behavioral Sciences and the Radcliffe Institute for
Advanced Studies at Harvard University.
Jonathan Cowden holds a PhD in political science from Yale University. He has taught at the University of
Miami, Cornell, and the University of California at Santa Barbara.
Dustin Tingley is an assistant professor of government at Harvard University. He received his PhD in
political science from Princeton University. His current research focuses on experimental tests of bargain-
ing models, the role of emotion in strategic interaction, international political economy, and statistical
methodology. Additional information is available at http://scholar.harvard.edu/dtingley.
126 Hum Nat (2012) 23:98–126
