Research Note:

Democratic Leaders, Crises and War

August 17, 2019

Jonathan Renshon1, Keren Yarhi-Milo2 and Joshua D. Kertzer3

Word count: 6373 words in text + 1582 in bibliography = 7955 words

Abstract : Many of our theories of IR argue that leaders and publics alike use regime typeto draw inferences about actors’ behavior in crises and war, with important implicationsnot just for how other actors treat democracies, but also how democracies themselvesbehave. Yet although these beliefs about regime type are important, they are also di�cultto study directly. In this research note, we put democratic leaders’ beliefs directly underthe microscope, fielding a survey experiment on a unique elite sample of members of theIsraeli Knesset. We find that Israeli leaders perceive democracies as no more likely tostand firm in a crisis but more likely to emerge victorious in wars. We replicate ourfindings in six national samples in four democracies (Israel, South Korea, the UnitedKingdom, and the United States), suggesting that similar beliefs about democracies incrises and war hold across a range of democracies.

1Associate Professor & Glenn B. and Cleone Orr Hawkins Chair, Department of Political Science, University ofWisconsin-Madison. Email: renshon@wisc.edu. Web: http://jonathanrenshon.com.

2Professor of Politics and International A↵airs, Columbia University. Email: ky218@columbia.edu. Web:https://polisci.columbia.edu/content/keren-yarhi-milo

3Paul Sack Associate Professor of Political Economy, Department of Government, Harvard University. Email:jkertzer@gov.harvard.edu. Web: http:/people.fas.harvard.edu/˜jkertzer/.

Political leaders and mass publics alike often look at world politics through the lens of regime

type, frequently using states’ domestic political institutions as indicative of their foreign policy

behavior, underlying preferences, or levels of resolve. In discussing his country’s options in dealing

with the rising German threat in the 1930s, the French defense minister Edouard Daladier told

then British Foreign A↵airs minister Anthony Eden that “no democratic country could indulge in”

preventive war (Eden, 1962, 44). Adolf Hitler made similar arguments, famously noting in Mein

Kampf as well as during private deliberations with his advisors that because democracies were

corrupt and weak, they would not take a stand against Germany, writing that they “will be unable

to muster the courage for any determined act” (quoted in Press, 2005, 76). More recently, the

Clinton administration’s democracy promotion agenda was directly tied to the President’s belief —

itself influenced by academic research on the topic — that “democracies rarely wage war on one

another,”1 a sentiment later echoed by President George W. Bush (Gartzke, 2007, 167).

IR scholars have made similar arguments, o↵ering a variety of theoretical models in which actors

use regime type as a heuristic to draw inferences about others’ intentions, capabilities, or resolve.

Both rationalist (Bueno De Mesquita and Lalman, 1992, 156-7) and constructivist (Risse-Kappen,

1995) variants of democratic peace theory argue that the reason why democracies are less likely to

fight each other is because states use regime type as a heuristic for hawkish or belligerent preferences.

Other versions of this same theory pinpoint the public’s beliefs about regime type as critical in

constraining the use of force (Mintz and Geva, 1993a), or argue that decision-makers believe that

democratic states are likely to try harder and expend more resources than comparable autocratic

states (Bueno de Mesquita et al., 1999, 794). And a related literature on democratic credibility

posits that the targets of threats will believe those issued by democracies to be more credible than

those from other regimes (Fearon, 1994; Schultz, 1999). Even work on far-afield subjects such as

terrorism emphasize the importance of beliefs about regime type: Pape (2003), for example, argues

that terrorists target democracies because they believe them to be especially sensitive to casualties.

These beliefs about whether democracy is a blessing or a curse in foreign policy are both theoreti-

cally and politically consequential, with implications not just for how other actors treat democracies,

but also how democracies themselves behave. For example, if democratic leaders and publics believe

democracies are soft targets, democracies will be likely to work harder to signal their resolve in

crises, perhaps taking stronger actions than they would otherwise. Conversely, if democracies see

1https://2009-2017.state.gov/p/io/potusunga/207375.htm

1

themselves as at an advantage in war, they may be more cavalier, confident, or risk acceptant about

entering into war with autocracies and less inclined to seek conflict with other democracies.

In this research note, we study democracy in international politics from a new angle, focusing not

on the question of whether democracies behave distinctively in foreign a↵airs (e.g. Lake, 1992; Reiter

and Stam, 2002), but instead by putting democratic leaders and publics’ beliefs about democracy

under the microscope. Studying whether actors use regime type as a heuristic when assessing resolve

or predicting military outcomes requires disentangling democracy from its correlates — and although

democracy is many things, randomly assigned is not one of them. Democratic states are wealthier,

tend to be found in democratic “neighborhoods,” were united by common interests throughout the

Cold War, and happen to include the current global hegemon. Thus, like Tomz and Weeks (2013),

Mintz and Geva (1993b), Rousseau (2005) and others, we study beliefs about democracy using

experimental methods.

Unlike much experimental work in IR, however, we include evidence from a sample uniquely

positioned to give us direct insight into the beliefs of leaders: 89 current and former members of the

Israeli Knesset. Our participants are not only elite in every sense of the term (ranking all the way

up to that of Prime Minister), but also have a history of making decisions about war and peace,

with over two-thirds of the sample having served on the Foreign A↵airs and Defense Committee.

We take advantage of our participants’ experience and domain-specific knowledge by employing

an experiment directly related to international conflict, letting us test what elite decision-makers

believe about the role of regime type in crises and war — and in a country outside the United States,

particularly important given concerns about the American-centric nature of many of our conclusions

about democracy in IR (Colgan, 2019; Levin and Trager, 2019). We also replicate our study on six

di↵erent national samples across four democracies facing very di↵erent security environments: two

representative samples of the Israeli public, two in the United Kingdom, one in South Korea, and a

nationally diverse sample in the United States. These replications give us insight into whether and

how the beliefs of our sample of leaders di↵er from those of “average citizens” in how they think

about issues relating to war and peace, speaking to an important debate about the utility of mass

public samples for testing theories of elite decision-making.

Our results add to the growing literature on the link between regime type, crisis outcomes,

and war, as well as addressing methodological questions related to elite experiments in IR. We find

that the inferences drawn by world leaders are subtle and contextual; democratic leaders do not,

2

for example, view democracies as all-powerful, but nor are they entirely pessimistic. We find the

e↵ect of democracy is seen as contingent: both our foreign elite decision-makers and the publics

they govern view democracies as no more likely than their autocratic counterparts to stand firm in

disputes, but more likely to prevail in conflicts that escalate to war. Democracies’ reputations, then,

are mixed.

We investigated this pattern further in one of our follow-up studies, exploring potential mech-

anisms drawn from the IR literature on regime type. Our results suggest that our respondents

believe democracies are no more resolved in crises partially because they see democratic threats as

less credible, contrary to the predictions of the democratic credibility thesis. At the same time,

when evaluating war outcomes, the democratic advantage our respondents reported appear to stem

predominantly from beliefs that democracies have higher quality militaries, and to a lesser extent,

a belief that democratic leaders select into winnable wars.

1 Democracies and Beliefs

The beliefs of leaders and the public about regime type are critical to the study and practice of

international politics. George Kennan’s view on the importance of covert operations was directly

tied to his pessimistic view of democracy’s disadvantages in conflict (Gaddis, 2011). Similarly,

neoconservatives’ arguments about democracy promotion are inextricably linked to their concerns

about democratic weakness in war (Caverley, 2010) while the liberal variant of the same belief is

based on the logic— voiced by Clinton’s National Security Advisor Anthony Lake — that democracies

“tend not to abuse their citizens’ rights or wage war on one another” (quoted in Mansfield and

Snyder, 1995, 5).

In other words, democracies in international politics have reputations.2 And just as reputational

considerations more generally shape both how actors are treated by others, and how strategic actors

themselves behave (Brutger and Kertzer, 2018), these beliefs about regime type not only a↵ect how

other states treat democracies, but also how democracies themselves conduct foreign policy. For

example, if democratic leaders buy into the Almond-Lippman consensus’ concerns about democratic

weakness in foreign policy crises, they should strive to insulate the public from foreign policy, and

2We follow Dafoe, Renshon and Huth (2014, 374) in defining reputations as “beliefs about persistent characteristicsor behavioral tendencies of an actor, based on the past actions of the actor” — in this case, beliefs about the behavioraltendencies of a type of system of government.

3

engage wherever possible in private rather than public diplomacy.

In addition to being critical for how actors act and respond to each other in world politics, these

beliefs are at the core of both theoretical and practical debates about democracies in conflict. In

some cases, the critical role of beliefs is readily apparent. For example, coercive threats are only

made credible via “beliefs about the genuineness” of the coercer’s communications (Schultz, 1999,

257). And in the canonical work on audience costs, the claim that leaders of democracies will find

it easier to signal resolve than their autocratic counterparts implies something very specific about

how leaders should calculate the credibility of threats made from di↵erent regimes. In fact, Fearon’s

(1994, 577) primary interest is in the question, “how do leaders come to revise their beliefs about

an opponent?”3

Although the beliefs of leaders play fundamental roles in signaling-based arguments for demo-

cratic distinctiveness, they loom similarly large in other accounts. One of the central insights of

Weeks’s 2014 disaggregation of authoritarian regimes is that “the impact of institutions depends on

the preferences and beliefs of the individuals who populate and lead” them. In fact, even theories

that on first glance seem to be about outcomes carry important implications for what beliefs we

should observe, or make assumptions about how leaders and publics perceive democratic distinc-

tiveness. For example, consider the body of literature focused on the question, “do democracies win

wars at a higher rate?” This is, at first glance, a simple matter of observing and coding outcomes at

the international level. However, nearly all of our theories about why democracies might have these

advantages hinge on beliefs. For example, one set of mechanisms that lie at the heart of the demo-

cratic triumphalist/defeatist debate involves dovish preferences. Democracies might be perceived as

being less likely to stand firm in crises out of the belief that democratic publics or decision-makers

subscribe to norms of liberal pacifism that proscribe the use of force except as a last resort (Schum-

peter, 1955; Kertzer and Brutger, 2016). Similarly, if democratic publics are believed to be especially

sensitive to either the human or financial costs of war (Mueller, 1971; Valentino, Huth and Croco,

2010), it is possible that democratic leaders would be more reluctant to risk escalation.

The democratic peace literature has similar implications for individuals’ beliefs. In some cases,

the beliefs that matter theoretically are those of the public: e.g., whether and how they use regime

type as a heuristic to understand the moral issues involved in attacking other regimes (Tomz and

3See also Prins (2003, 70): “domestic political opposition can only validate leader preferences or intentions if theforeign adversary believes such opposition can in fact impose costs on a leader.” Emphases added.

4

Weeks, 2013) or whether it a↵ects their expectations regarding the scale of casualties (Gartzke,

2001). Institutional variants of democratic peace theory posit a mechanism that relies on leaders’

perceptions that democratic states are likely to try harder and expend more resources than a com-

parable autocratic state (Bueno de Mesquita et al., 1999, 794). Similarly, Downes (2009, 11) notes

that the selection e↵ects argument used to support democratic triumphalism implies that “because

democratic leaders are cautious in selecting only those wars they are highly likely to win, democratic

elites who choose to initiate or enter wars should be confident of victory.” Thus, in their exploration

of democratic behavior in international crises, Gelpi and Griesdorf (2001) routinely make reference

to the beliefs that leaders in any regime type have about democracies: for example, that foreign

leaders are more likely to target democracies “because of their belief that democracies prefer not to

fight” (p. 642).

Theories of democratic di↵usion similarly emphasize the important role of leaders’ and publics’

beliefs about democratic distinctiveness. Gunitsky (2014, 576) argues that democratization occurs in

waves after hegemonic shocks partially because victories on the battlefield change observers’ beliefs

about “relative regime e↵ectiveness”: the wave of democratization that occurred after both the First

and Second World War, he argued, was partially a function of other states seeking to emulate the

regime types that had come out on top. In a similar vein, Elkink (2011) agues that di↵usion of

attitudes among the public can help explain the spatial clustering of democracy and its di↵usion.

Beliefs about democratic distinctiveness similarly play a large role on theories of the democratic

peace that emphasize learning, as in Starr (1992, 210).

More generally, the new wave of literature on the role of leaders in IR consistently points to

beliefs as the mechanism through which individual leaders matter: Yarhi-Milo (2014) shows how

leaders’ beliefs systematically a↵ect what type of costly signals they see as informative indicators of

their adversary’s intentions; for Saunders (2011), it is because leaders vary in their causal beliefs that

they adopt di↵erent military intervention strategies; for Horowitz and Stam (2014), it is because

prior military experience changes leaders’ beliefs about the utility of force that leaders with combat

experience conduct foreign policy systematically di↵erently than leaders without; for Fuhrmann and

Horowitz (2015) it is because of di↵erent beliefs about the importance of maintaining independence

that former rebels are more likely to pursue nuclear weapons.

However, with few exceptions that we can find — mostly about the mass public rather than

political elites — scholars working on democracies in IR have not directly studied beliefs. For

5

example, as Hermann and Kegley (1995, 512) note, despite the mountains of work on democratic

distinctiveness, “this wave of research neglects sources of explanation that lie within the people

involved in making policy decisions.” This gap is understandable given the inherent challenges of

observational data, as well as the genuine di�culty of studying leaders’ assessments and private

beliefs. Still, we argue that there are potentially important advantages that might derive from

testing whether decision-makers believes the things that IR scholars expect them to. The majority

of our theories concerning democracies in crises and war are theories of intentional action; even if

we only care about outcomes, the mechanisms we use to explain them typically implicate the beliefs

of decision-makers (Elster, 2007); if these theories have any Verstehen to them, we should expect a

broad congruence between these theoretical predictions and the judgments of the decision-makers

about whom we are theorizing. While testing these predictions on observational data is beyond

the scope of this paper, our findings provide crucial foundations for moving forward in testing and

refining our theories of conflict behavior and regime type.

If previous work has focused less than perhaps it should have on beliefs, a related question

regards whose beliefs matter. In some cases, we care about the beliefs of leaders—whether their

assessments of credibility are shaped by the regime type of their adversaries—while in others, we

care about the beliefs of the general public. The beliefs of the average citizen matter for several

reasons. In some cases, the theories we are testing directly implicate mass beliefs: for example,

whether or not democratic publics are particularly cost-sensitive during conflicts or see regime types

as di↵erentially threatening. In others, understanding publics’ beliefs about democracy is useful

because of the importance of public opinion for theories of democratic constraint: if publics and

leaders share the same beliefs about democracies in crises and war, it suggests public opinion is

unlikely to act as a constraint on leaders’ abilities to act on those beliefs. It also valuable to use

mass public data to learn how widespread the reputations of democracies are (do they extend across

countries and types of democracy?).

Taken together, this discussion of research on beliefs about regime type and its implications

for democratic behavior and theories of IR suggest obvious ways forward. They call for a unique

research design, one based on a series of experiments, focused on questions related to democratic

distinctiveness and fielded on a combination of elite leader and mass public samples. We describe

exactly that approach in the following section.

6

2 Research Design

We study beliefs about democracies in crises and war by fielding an original experiment in seven

samples, from four di↵erent democracies, across a three year period. The first of our samples

consists of elite decision-makers: 89 current and former members of the Knesset in Israel. As we

show below, this is an unusual sample even by the standards of elite experiments in IR: two-thirds

of our respondents had experience serving as members of the Knesset’s Foreign A↵airs and Defense

committee; our least “elite” participant is a member of parliament; our most, a Prime Minister.

The other six surveys were fielded on mass public samples — two nationally representative samples

in Israel, two in the United Kingdom, one in South Korea, and one nationally diverse (though not

representative) sample in the United States. As we discuss below, by fielding the same experiment in

a range of democracies in very di↵erent foreign policy contexts, we not only probe the generalizability

of our results, but also draw inferences about beliefs about democracies in crises and war outside

the confines of the United States (Narang and Staniland, 2018).

We begin by discussing our elite sample of Knesset members — and why Israel is a particularly

interesting case to study leaders’ beliefs about democratic distinctiveness — before reviewing our

six mass public samples. We then present our experimental design.

2.1 Elite sample: the Israeli Knesset

For our elite sample, we field a survey experiment on current and former members of the Israeli Knes-

set. There are several important substantive advantages to studying the beliefs of Israeli decision-

makers. First, the most direct way to achieve our goal of examining decision-makers’ beliefs about

democratic reputations in war and peace is to sample from a population that has wrestled with

those issues outside the lab. This is, after all, what is unique about studying leaders rather than

the public. “Use of force” decisions are ubiquitous in Israel, and highly salient for Israeli decision-

makers: during our leaders’ time in o�ce (from 1996 onwards), Israel was involved in 16 Militarized

Interstate Disputes (MIDs). Our participants were involved in many of these cases. More broadly,

since earlier experiences of Israeli conflict might have shaped their beliefs, we note that Israel has

been involved in 128 documented MIDs since the state’s inception, and seven militarized compellent

threat episodes (Sechser, 2011).

Second, because of the structure of Israel’s parliamentary system, the vast majority of the

7

executive branch — e.g., the Prime Minister and ministers in the security cabinet — are also elected

members of the Knesset (the legislative branch of government). Thus, the Israeli Knesset — unlike,

e.g., the U.S. Congress — is comprised of policy makers who are directly and intimately involved in

use of force decisions. Because of political norms and relatively short election cycles, it is common for

former members of the executive branch to later become members of the opposition in the Knesset;

conversely, nearly all current members of the executive branch were at some point in their career

members of the opposition in the Knesset. Thus, even the Knesset members in our sample who are

currently part of the opposition have either been members of the executive branch in the past, or

are likely candidates to become members in the future. Put di↵erently, by sampling current and

former members of the Knesset, we are also e↵ectively sampling current, former, and potentially

future members of the executive.

Our elite survey was fielded July-October 2015. Of 288 potential subjects, 89 participated,

leaving us with a 31% response rate, relatively high for surveys of this type. The Knesset sample is

described in Table 1 and our recruitment procedures are described in Appendix §A.2; Appendix §A.3

describes our protocol to increase our confidence that the Knesset members themselves participated

in the study, rather than their sta↵. Of our Knesset participants, 25% were current members; the rest

(75%) were former Knesset members. A great many of our participants had experience in IR-relevant

contexts: 64% had active combat experience, and 67% had experience serving as members of the

Knesset’s Foreign A↵airs and Defense committee. They also had considerable political experience:

on average, our participants had served 3 terms in Parliament, and some had served as many as 9

terms. While 58% of the Knesset subjects had never served as a Minister, 29% had been at least

a Deputy Minister, and fully 12% of our sample was in our highest category of elite experience,

such that our participants include individuals who had served as Cabinet Members, and even Prime

Minister.

In Appendix §C.1, we show that our leader sample is fairly representative of the universe of Israeli

political leaders from the time frame we examined, although not surprisingly, our analysis reveals that

current members of the Knesset were less likely to participate than former members. As is evident,

this is an extremely unusual sample: our least “elite” participant is a member of parliament; our

most, a Prime Minister. In fact, even in many elite studies of decision-making, subjects are often

far removed from the actual decision makers of primary interest to IR theories. Renshon (2015), for

example, uses political and military leaders drawn from a mid-career training program at Harvard

8

Table 1: Knesset Sample (N = 89)

Proportion of respondentsKnesset Member:

Current 25%Former 75%

Exp. on Foreign A↵airs/Defense Committee . . .. . . as backup or full member 67%

. . . as full member 54%Highest level of experience:

. . . not a Minister 58%. . . Deputy Minister 29%

. . . Cabinet Member or higher 12%Male 84%Served in military 95%Active combat experience 64%

Mean SDAge 61.4 10.7Terms in Knesset 3.0 2.1Military Assertiveness 0.61 0.20Political Ideology 0.45 0.24Arab-Israeli Conflict 0.39 0.25International Trust 0.40 0.26

Note: individual di↵erences in bottom four rows scaled from 0-1.

Kennedy School, while Alatas et al. (2009) use Indonesian civil servants, Hafner-Burton et al. (2014)

use “policy elites” (including civil servants, corporate executives, former members of Congress, and

U.S. trade negotiators) and Mintz et al. (1997) use Air Force o�cers. While more elite than college

freshmen, to be sure, the samples used are still somewhat removed from the dictators, presidents,

leaders of the military and foreign ministry, trusted advisors, and generals who are the primary

decision makers in most interstate conflicts.4 This serves as a reminder that the use of quasi-elite

subjects, while interesting and helpful, does not completely obviate the necessity of extrapolating

from one population to another. The research design we employ here is thus perhaps most similar

to Findley et al. (2017), who field experiments about foreign aid on paired samples of Ugandan

parliamentarians and members of the mass public.

9

Table 2: Mass Public Samples (Total N = 9360)

Country Israel I Israel II Korea UK I UK II USAMale 0.53 0.52 0.54 0.44 0.45 0.52Age:<25 0.13 0.13 0.14 0.11 0.14 0.1825-34 0.27 0.26 0.17 0.25 0.25 0.4535-44 0.20 0.20 0.21 0.23 0.23 0.2145-54 0.17 0.17 0.33 0.22 0.20 0.09>55 0.24 0.24 0.15 0.18 0.17 0.07

Education:High School or less 0.35 0.27 0.20 0.54 0.54 0.11Some college 0.22 0.22 0.16 0.11 0.11 0.26College/university 0.27 0.31 0.54 0.25 0.24 0.51Postgraduate 0.15 0.19 0.10 0.10 0.10 0.12

Political Ideology 0.61 0.59 0.52 0.52 0.51 0.39Military Assertiveness 0.58 0.56 0.39International Trust 0.32 0.35Arab-Israeli Conflict 0.63 0.63Pol. Interest 0.45 0.47 0.46 0.56Interest in FP 0.40 0.51 0.51 0.54N 1599 1111 1797 1105 1691 2057

2.2 Mass public samples

While the Knesset sample is valuable both theoretically and methodologically — if we care about

democratic leaders’ beliefs about democracies in crises and wars, we surely especially care about the

beliefs of leaders of a democracy frequently engaged in both — we are also interested in ordinary

citizens’ beliefs about democratic distinctiveness, across a range of countries. We supplement our

elite sample by replicating our experiment in six mass public samples, fielded across four democracies

in very di↵erent foreign policy contexts.

While fielding our experiment on democratic publics allows us to answer questions of theoretical

relevance to debates about democratic distinctiveness and reputations, the pairing of an elite and

mass public samples across multiple countries provides several additional benefits. The first is that

it lets us speak to debates about di↵erences between elites and masses in IR. To that end, two

of the samples are drawn from the Israeli general public. Including mass samples from the same

country as the elite survey provides leverage on the selection of leaders from the general population,

thereby giving us insight into the dimensions on which they di↵er and those which they resemble

4A focus on elites is more common within the literature on foreign policy attitudes, but those works have a similarlyexpansive view of “elites” (e.g., business executives) and were focused on broad foreign policy attitudes, rather thandynamic judgment and decision-making. See, e.g., Holsti and Rosenau (1988) and Wittkopf and Maggiotto (1983).

10

their compatriots. The second is the value inherent in any replication, increasing confidence in

the overall research program and generalizability of our findings. This is particularly germane

for the study of democratic distinctiveness, given Narang and Staniland’s call for IR scholars to

“disaggregate democracies”, and test our theories about democratic foreign policy across a wider

range of democracies, a point echoed by Levin and Trager (2019). We thus replicate our findings in

four other national samples in three other democracies (South Korea, the United States, and United

Kingdom), selected because they vary along multiple dimensions, including region, political system,

relative military capabilities, and the salience of foreign policy. Altogether, our four democracies

vary in a variety of ways: both Israel and South Korea, for example, face regional security threats

and have some form of mandatory military service, while the United States and United Kingdom do

not; both Israel and the United Kingdom are parliamentary democracies, whereas the United States

and South Korea are presidential systems, and so on.

The two Israeli public samples were fielded in September-October 2015, and January 2016, re-

spectively, by iPanel, an Israeli polling firm that has been used e↵ectively by other recent surveys

and experiments (e.g., Ben-Nun Bloom, Arikan and Courtemanche, 2015; Manekin, Grossman and

Mitts, 2015). Both samples are representative of the Israeli Jewish population, and stratified based

upon gender, age, living area and education.5 The third and fourth samples are national samples

in the United Kingdom, and were fielded in May and June 2018 by Survey Sampling International

(SSI). The fifth sample is a national sample in South Korea fielded by the Korean polling firm

Embrain in May 2018. The British and Korean samples were both stratified based on gender, age,

and location. The final sample was fielded on a nationally diverse, although not nationally repre-

sentative, sample in the United States, through Amazon Mechanical Turk in June 2015. In addition

to standard demographic data, participants in each survey completed questionnaires capturing a

variety of political orientations; the list of orientations varied based on the country, but in the Israeli

case included military assertiveness, political ideology, stance on to the Arab-Israeli conflict, and

international trust.6 Descriptive statistics for these six samples, totaling over 9000 respondents, are

presented in Table 2. Our combination of samples, while obviously not comprehensive of all democ-

racies, captures a wide range of democratic security environments, across a wide range of regions

5Our focus on the Israeli Jewish population is due entirely to logistical constraints, specifically the inability ofonline polling companies in Israel to provide anything close to a representative sample of the minority Israeli Arabpopulation.

6See Appendix §B.2 for the complete instrumentation.

11

(North America, Western Europe, the Middle East, and East Asia) and in a manner not usually

addressed with original data in a single study.

2.3 The Experiment

All respondents across all seven samples were presented with the same experiment, albeit in di↵erent

languages (Hebrew in Israel, Korean in South Korea, and English in the UK and the United States).7

Here is the situation:

• Two countries are currently involved in a public dispute over a contested terri-tory. The dispute has received considerable attention in both countries, becauseof the risk that disputes like these can escalate to the use of force.

• Country A is a [democracy/dictatorship]. Country B is a dictatorship.

• Both countries have moderately powerful militaries, with large armies, moder-ate sized-navies, and well-trained air forces.

• Neither country is a close ally of the United States.

• Country A is slightly larger than Country B, though their economies are ap-proximately the same size.

• Country A has moderate levels of trade with the international community.Country B has high levels of trade with the international community.

• The last time the two countries were involved in an international dispute, dif-ferent leaders were in power.

1. Given the information available, what is your best estimate about whetherCountry A will stand firm in this dispute, ranging from 0% to 100%?

2. If the dispute were to escalate and war were to break out, what is your bestestimate about whether Country A will win, ranging from 0% to 100%?

Figure 1: Experimental Vignette

All subjects read a vignette (reproduced in Figure 1), in which the regime type of Country A

was experimentally manipulated. In addition to our regime type manipulation, we described a

number of other characteristics of the individual states, as well as their relationship, both to avoid

“information leakage”8 and to avoid “putting our thumb on the scale”: if respondents are only

7Our experiments were fielded using the Qualtrics platform (aside from the few Knesset members who chose to fillout paper copies). In each study except for the Israel II sample, respondents took part in other, unrelated experimentsas part of their participation. The module orders were fixed because of the multi-mode nature of the Knesset study,which means that we cannot examine potential order e↵ects.

8See, e.g., Tomz and Weeks 2013; Dafoe, Zhang and Caughey 2018, as well as the discussion in Kertzer and Brutger2016.

12

provided information on one dimension, it is far more likely for them to weight that dimension

heavily. We thus described each country’s military, economy, geographic size, and so on, seeking to

avoid demand e↵ects for our treatment by ensuring the two countries slightly di↵ered on multiple

dimensions. Participants then indicated how likely they thought it was that Country A would stand

firm in the crisis, and, if the dispute escalated into war, how likely they thought that Country A

would win.9

3 Results

Table 3: Summary of results across samples

In crises In war[estimates of A standing firm] [estimates of A winning war]

Dictatorship Democracy �democracy Dictatorship Democracy �democracy

Knesset 59.14 53.25 -5.89 44.30 58.58 14.28Israel I 57.86 53.27 -4.59 48.52 53.49 4.96Israel II 59.91 58.24 -1.67 50.06 58.29 8.23Korea 61.15 61.17 0.02 47.43 53.75 6.32UK I 60.83 61.15 0.31 51.90 55.48 3.58UK II 60.91 61.36 0.45 52.70 56.29 3.59USA 66.57 65.42 -1.14 49.81 53.87 4.06

Di↵erences significant at p < 0.05 in bold, calculated via rank-sum tests for the Knesset sample, and t-tests for thepublic samples. Country B is fixed at “dictatorship” in all samples.

We present our results two di↵erent ways: numerically in Table 3, and graphically in Figure 2,

which displays bootstrapped density distributions of the average treatment e↵ects. In both Table

3 and Figure 2(a), four patterns are evident. First, across all seven samples, respondents never

perceived democracies as possessing an advantage in crises. In the Israel I sample, democracies

are actually viewed as 4.6 percentage points less likely to stand firm than dictatorships are, a

statistically significant e↵ect; the Knesset sample features a negative treatment e↵ect for democracy

of similar magnitude (�5.9 percentage points), although because of its small sample size, the e↵ect

is not statistically significant at the 95% level.10 On the whole, then, democracies do not have a

reputation for being particularly distinctive in their crisis behavior, and where they are, democracy is

9In the two Israeli public samples, we included a manipulation check asking participants to recall which regimetype condition they were in. In the Israel I sample, 85% correctly recalled their treatment condition; in IsraelII, the passage rate was 90%. In the analysis below, we include all respondents regardless of whether they passedthe manipulation check or not, although the results do not substantively change if those participants who failed themanipulation check are excluded.

10In Appendix §C.7, we show that these e↵ect sizes are comparable to those from large-N studies of democracyand victory in crisis bargaining drawn from the MID data. Full results for Knesset sample are depicted in Table 5 inAppendix §C.2. Results for the Israeli public are reproduced in Appendix §C.4.

13

Figure

2:Average

treatm

ente↵

ects

anddi↵erences-in-di↵erences

(a)Averagetrea

tmen

te↵

ects

-20

-1001020

...in

cris

es...

in w

ars

Effect of democracy(b

)Di↵eren

ce-in-di↵eren

ce

010

2030

Diff

eren

ce-in

-diff

eren

ce: e

ffect

of d

emoc

racy

in w

ar -

cris

is

Sample

Isra

el I

Isra

el II

Isra

el: K

ness

et

Rep

ublic

of K

orea

Uni

ted

Kin

gdom

I

Uni

ted

Kin

gdom

II

USA

Thedistributionsofav

eragetrea

tmen

te↵

ects

(calculatedusingB

=1500bootstraps)

inpanel

(a)sh

owdem

ocracies

are

nev

erperceived

asbeingatanadva

ntagein

crises

(andin

theKnessetandfirstIsraelisample,are

believed

tobeatasignifica

ntdisadva

ntage),butare

alw

aysbelieved

tobeatanadva

ntagein

wars.The

bootstrapped

di↵eren

ce-in-di↵eren

ceresu

ltsin

panel

(b)sh

owth

atth

ee↵

ectofdem

ocracy

isseen

asbeingsignifica

ntlymore

positivein

warth

anin

crises

across

all

seven

samples.

Note

thatth

egreater

variance

inth

eKnessetsample

isdueto

itssm

aller

sample

size;seeAppen

dix

§C.6

forresu

ltswithdow

nsampledpublic

distributionsth

atfeatu

reasimilarsp

read.

14

believed to be a liability rather than an asset. Second, across all seven samples, respondents believe

democracies possess a significant advantage in wars; the e↵ect was particularly pronounced among

our Knesset respondents (14.3 percentage points), but are positive and statistically significant in

all of the other samples as well. Third, as Figure 2(b) shows, the di↵erence-in-di↵erence between

democracies and dictatorships in crises versus wars is statistically significant across each of our seven

samples: in each country, democracies are perceived as at a significantly greater advantage in wars

than crises.

Fourth, focusing specifically on the Israeli samples, our results reveal a general similarity in the

judgments displayed by our elite decision-makers in the Knesset and the mass public they represent.

As is starkly illustrated in Figure 2, both our elite and our public samples espoused democratic

pessimism in crises (most starkly in the Knesset sample and Israel I), and democratic triumphalism

in wars: democracies were believed to be perhaps less likely to stand firm in disputes, and more

likely to win conflicts that escalated to the use of force.

This congruence is notable given the intensity of debates about the extent to which elite decision-

makers di↵er from members of the mass public (e.g., Mintz, Redd and Vedlitz 2006; Druckman and

Kam 2011; Hafner-Burton, Hughes and Victor 2013; Kertzer 2016, Linde and Vis 2017). In both

cases, any di↵erence between the three samples was in magnitude, not direction: leaders were both

slightly more pessimistic about the odds for democracies in crises, and also a bit more optimistic

about their chances in war. While the congruence between our samples should be interpreted with

caution, our results remind us that di↵erences between elites and ordinary citizens should not be

overstated without a theory as to why we might expect characteristics of elites to moderate the

impact of treatment e↵ects (Renshon, 2015). Indeed, supplementary analyses find relatively little

sign of heterogeneous treatment e↵ects within our Israeli samples: current MKs behave similarly to

former MKs, more hawkish MKs behave similarly to more dovish MKs, and so on.11

11More formally, we estimate a series of simple regression models in which we regress each dependent variable (eitherresolve, or military e↵ectiveness), on the democracy treatment, the respondent-level characteristic under investigation,and the interaction between the two. The interactions lack statistical significance (on resolve: p < 0.13 for currentMK status, p < 0.40 for military assertiveness; on military e↵ectiveness: p < 0.71 for current MK status, p < 0.41 formilitary assertiveness). We present additional tests for heterogeneous treatment e↵ects in Appendix §C.5.

15

Testing The Mechanisms

Thus far, we have shown that, in one elite and six mass public samples, democratic reputations are

mixed; democracy is believed to have a contingent e↵ect, displaying a significantly more positive

e↵ect on winning in war than in standing firm in crises. These results are broadly consistent with

a variety of predictions made in the IR literature, but given the multiplicity of causal mechanisms

discussed in this scholarship, it is unclear which ones are responsible for the e↵ects we see here.

To provide some evidence on the mechanisms that underlie judgments about the resolve and

e↵ectiveness of democratic states, we added a series of questions to the survey instrument fielded on

our Israel II sample. Following the completion of the scenario, respondents were presented with

a series of questions tapping into a series of potential mechanisms, depicted in Table 4.12 For the

purpose of the mechanism questions, we divided our subjects in half, with half being reminded of their

answer to the “resolve in a crisis” answer and being asked questions about the crisis scenario, and

half being reminded of their answer to the “e↵ectiveness in war” outcome and being asked questions

about the war scenario. In addition to this between-subject manipulation, we also included an order

manipulation to avoid potential downstream e↵ects, and estimate within-subject e↵ects for the two

of the domestic constraint mechanisms, described in greater detail in Appendix §B.3.13

In Figures 10-14 in Appendix §C.8, we present the first part of any mechanism analysis: probing

the e↵ects of our treatment on our mediators. To that end, we present density distributions for our

mechanism questions by treatment group, where Country A is either a Democracy or Dictatorship.

This allows us to address the first set of implications drawn from extant theories: do respondents

evince di↵erential beliefs about the characteristics of democratic and autocratic states?

In fact, they do. Our respondents in the Democracy treatment saw Country A’s citizens as

more likely (by 17%) to believe that force should be use as a “last resort,” but also perceived the

country’s leaders as more likely (by 31%) to evince the same pacifist beliefs. They believed that

country was being more casualty-sensitive (by 33%) and more sensitive to the financial costs of war

(by 21%) compared to subjects for whom Country A was described as a dictatorship. Similarly,

subjects believed democracies’ armies had better trained (by 7%) troops with higher morale (11%)

12Our full list of questions and closed-response options is contained in Appendix §B.3.13Despite the broad congruence between the results we find in all seven samples, it is important to note that the

mechanisms we find support for below need not necessarily be the same as those operating amongst elites, or inpublics in other countries — it is possible, for example, for elites and masses, or Koreans and Americans, to reachthe same conclusions about democracies for di↵erent reasons. Nonetheless, these mechanism tests o↵er us a chance togain further insight into the underpinnings of the public’s judgments in one of these cases, letting us test the extentto which their rationales correspond with our theoretical frameworks.

16

Table 4: Mechanism Questions

Crisis Mediators:

Dovish Preferences:

1. What proportion of the citizens do you think be-lieves force should only be used as a last resort?

2. What proportion of the national decision-makersdo you think believes force should only be usedas a last resort?

3. How sensitive do you think its citizens would beto casualties?

4. How sensitive do you think its citizens would beto the financial costs (e.g., increased taxes) offighting?

Credibility:

5. If the leader of Country A makes a public threat,how likely do you think it is that they’ll followthrough?

War Effectiveness Mediators:

Selection e↵ects:

6. If Country A initiated the dispute, how likely doyou think it is that it will prevail in the dispute?

Military Conduct

7. How likely do you think it is that other countrieswould come to Country A’s defense?

8. How well-trained do you think the soldiers ofCountry A are?

9. How strong do you think the morale of the sol-diers is in Country A?

How strong do you think the morale of the sol-diers is in Country A?

Mediators for both Crises and War:

Domestic Contraints:

[Thinking back to the original crisis, if Country A were to back down. . . ][If Country A were to lose thewar. . . ] � [Thinking back to the original crisis, if Country A were to stand firm. . . ][If Country A wereto win the war. . . ]

10. What e↵ect do you think it would have on public support for the government?*

11. What e↵ect do you think it would have on the likelihood of the government remaining in o�ce?*

12. What do you think is likely to happen to the leader?

*Within-subject e↵ects

17

than their autocratic counterparts. And not only did democratic armies have reputations for being

of higher quality, but they also had reputations for being more likely (by 24%) to be able to count

on assistance from allies.

To test whether these mechanisms mediate the relationship between beliefs about democratic

regimes and inferences about their resolve and military e↵ectiveness, we turn to nonparametric causal

mediation analysis (Imai et al., 2011). This allows for estimation of an average causal mediation e↵ect

(ACME), which can be interpreted as the expected di↵erence in the outcome when the mediator

took the value it would realize under the treatment condition as opposed to the control condition,

while the treatment status itself is held constant.14

Figure 3 presents the results of our causal mediation analysis. The top panel presents the

potential mediators for the e↵ect of democracy on beliefs about standing firm in a crisis. As is

evident, the “credibility” ACME is negative and statistically significant, implying that one reason

why our subjects believed democracies were slightly less likely to stand firm was that they believed

democracies were less likely to follow through on their threats compared to autocracies, counter

to the predictions of the democratic credibility literature. In contrast, we do not find evidence

that respondents’ beliefs about standing firm in a crisis were mediated by any of the “domestic

constraints” suggested by the literature on leadership turnover or public support. On the whole, the

varying signs of the di↵erent ACMEs in Figure 3(a) suggest that one explanation for the weak or null

total e↵ect of democracy in crises in some samples may be due to the presence of an “inconsistent

mediation e↵ect,” in which di↵erent causal mechanisms push in di↵erent directions.

The bottom panel of Figure 3 presents our analysis for our war outcome mediators. Here, we

find strong evidence consistent with explanations that focus on democracies’ reputation for military

e↵ectiveness: mediators relating to the morale and training of democracies as well as their likelihood

of being aided by allies all presented significant and positive ACMEs. We also found evidence that

subjects’ inferences were in part mediated by beliefs about the selectivity of democracies, consistent

with a number of extant works (e.g., Reiter and Stam, 2002). However, as before, we found little

evidence that subjects evinced beliefs consistent with the “domestic constraints” school of democratic

e↵ectiveness.

In sum, our respondents in the Israel II sample view democracies as more cautious and peace-

ful, and (perhaps as a result) less credible in their use of coercive diplomacy during crises; however,

14Causal mediation analysis was implemented via the mediation package in R (Imai et al., 2010).

18

ACME

-4 -2 0 2 4

Dovish public

Dovish leaders

Casualty sensitivity

Financial sensitivity

Leader turnover

Public support

Leader's fate

Follow through

Dovishpreferences

Domesticconstraints

Credibility

(a) Crisis mediators

ACME

-4 -2 0 2 4

Pr(Victory | Initiate)

Leader turnover

Public support

Leader's fate

Allies

Training

Morale

Selectivity

Domesticconstraints

Militaryeffectiveness

(b) War Mediators

Figure 3: Causal Mediators for E↵ect of Democracy in Crisis and War (with 95% quasi-BayesianCIs)

19

this perceived weakness appears to subside when the crisis escalates into a full blown war. At that

stage, di↵erent calculations come into play, and the military e↵ectiveness of democracies is seen as

a major asset.

4 Conclusion

If we take leaders at their word, actors in international politics frequently use regime type to draw

inferences about states’ future behavior. Democracies have reputations, which a↵ect both how other

actors treat them, and how they themselves behave. Yet, although there is large and robust liter-

ature exploring the ways that democracy matters in international politics, it largely focuses on the

direct e↵ects of domestic institutional configurations on state behavior, rather than examining this

perceptual pathway, in which institutions a↵ect behavior through actors’ beliefs. In this research

note, we therefore sought to place beliefs about democracies in crises and war under the microscope,

fielding a survey experiment on an unusually elite sample of past and present Israeli Knesset mem-

bers, thereby providing direct evidence on the beliefs of political leaders. We also replicated our

findings on six national samples in four democracies to thoroughly assess the generalizability of our

findings as well as provide evidence on causal mechanisms. Across all seven samples, we find that

democracies have mixed reputations: they are perceived as no more likely to stand firm in crises

than their autocratic adversaries, but are believed to be significantly more likely to prevail on the

battlefield. On a broader level, our results therefore o↵er one more data point demonstrating that

we should not uncritically assume that leaders and the public will di↵er dramatically.

Our experimental design also allowed us to evaluate the mechanisms leading to those assess-

ments, intervening in some prominent debates in the literature. We find support for the notion that

democracies are believed to select into winnable wars, as our participants viewed conflicts that were

initiated by democracies as more likely to succeed compared to ones initiated by non-democracies.

Respondents also evince beliefs consistent with some variants of democratic triumphalism that focus

on the role of higher morale, more e↵ective armies, and greater support from alliance partners.

In contrast, we find mixed or weak support for other mechanisms. In line with norms of liberal

pacifism, participants saw democracies as less likely to use force unless as a last resort, but they

also saw them as no more sensitive to the costs of war, and beliefs about dovish preferences did

not mediate the relationship between regime type and resolve in crises. Our results also cast doubt

20

on theories that rely on leaders’ beliefs about the advantage democracies have in sending credible

public threats in crises. In fact, the opposite appears to be the case: our participants viewed public

threats that were issued by democracies as less credible than ones issued by non-democracies. Taken

together, our results suggest that while some claims about the distinctiveness of democracies accord

with our participants’ beliefs, others did not. While our results by no means constitute the final

say on the empirical validity of these claims, at the very least they point to a potentially large

discrepancy between what our theories assume leaders’ beliefs to be, and what we find when we

examine them directly.

Our results suggest a number of implications. For instance, our finding that democratic leaders

and publics perceive themselves as having no advantage over autocracies in signaling resolve in crises

suggests that perhaps democratic leaders might attempt costlier signals than they would otherwise

to compensate for their perception of weakness in “contests of will.” In addition, our finding that

democratic leaders and their publics see themselves as more likely than other regimes to win wars

suggests that we might see observe overconfidence and increased risk-acceptance among those groups

in the lead-up to war. Finally, if reputations for resolve and war-winning adhere to regime type—

as our findings indicate—we should expect democratic leaders to show systematic di↵erences in

their willingness to engage in conflict with democracies versus autocracies. Indeed, in this sense,

our findings provide microfoundational support for the democratic peace’s primary finding that

democracies are significantly less likely to fight each other.

Our method—paired experimental designs that focus on di↵erent stages of the conflict process—

also allows us to connect our results to contemporary bargaining theories of war, which tend to views

crises and war as one part of a continuous process. Because we ask all of our respondents about

both crises and war, we are able to provide insight on how beliefs about democracies in each phase

of conflict interact. In Appendix §C.8 we show that for five of our seven samples, our respondents’

beliefs about democratic advantages in war swamp concerns about democratic disadvantages in

crises.

The nature of the samples also suggests fruitful avenues for replications and extensions. While

our subjects were drawn from a variety of di↵erent democratic societies, future research should

examine similar questions in non-democratic societies. A more constructivist reading of our results

might also lead scholars to seek to replicate our results among decision-makers in countries where

alternative understandings of “democracies” might be present. Lastly, future research on the role of

21

regime type in international security could formulate and test domain-specific claims about the e↵ects

of democracy in contexts such as counterterrorism, counterinsurgency, and low-intensity conflicts.

22

References

Alatas, V., L. Cameron, A. Chaudhuri, N. Erkal and L. Gangadharan. 2009. “Subject pool e↵ectsin a corruption experiment.” Experimental Economics 12(1):113–132.

Ben-Nun Bloom, Pazit, Gizem Arikan and Marie Courtemanche. 2015. “Religious Social Identity, Re-ligious Belief, and Anti-Immigration Sentiment.” American Political Science Review 109(2):203–221.

Brutger, Ryan and Joshua D. Kertzer. 2018. “A Dispositional Theory of Reputation Costs.” Inter-national Organization 72(3):693–724.

Bueno De Mesquita, Bruce and David Lalman. 1992. War and reason. New York, NY: CambridgeUniversity Press.

Bueno de Mesquita, Bruce, James D. Morrow, Randolph M. Siverson and Alastair Smith. 1999. “AnInstitutional Explanation of the Democratic Peace.” American Political Science Review 93(4):791–808.

Caverley, Jonathan D. 2010. “Power and Democratic Weakness: Neoconservatism and NeoclassicalRealism.” Millennium 38(3):593–614.

Colgan, Je↵ D. 2019. “American Bias in Global Security Studies Data.” Journal of Global SecurityStudies 4(3):358–371.

Dafoe, Allan, Baobao Zhang and Devin Caughey. 2018. “Information Equivalence in Survey Exper-iments.” Political Analysis 26(4):399–416.

Dafoe, Allan, Jonathan Renshon and Paul Huth. 2014. “Reputation and Status as Motives for War.”Annual Review of Political Science 17:371–393.

Downes, Alexander B. 2009. “How smart and tough are democracies? Reassessing theories ofdemocratic victory in war.” International Security 33(4):9–51.

Druckman, James N. and Cindy D. Kam. 2011. Students as Experimental Participants: A Defense ofthe “Narrow Data Base”. In Cambridge Handbook of Experimental Political Science, ed. James N.Druckman, Donald P. Green, James H. Kuklinski and Arthur Lupia. New York, NY: CambridgeUniversity Press pp. 41–57.

Eden, Anthony. 1962. The Eden Memoirs: Facing the Dictators. London: Cassell.

Elkink, Johan A. 2011. “The international di↵usion of democracy.” Comparative Political Studies44(12):1651–1674.

Elster, Jon. 2007. Explaining Social Behavior: More Nuts and Bolts for the Social Sciences. NewYork, NY: Cambridge University Press.

Fearon, James D. 1994. “Domestic Political Audiences and the Escalation of International Disputes.”American Political Science Review 88(3):577–592.

Findley, Michael G., Adam S. Harris, Helen V. Milner and Daniel L. Nielson. 2017. “Who ControlsForeign Aid? Elite versus Public Perceptions of Donor Influence in Aid-Dependent Uganda.”International Organization 71(4):633–663.

Fuhrmann, Matthew and Michael C. Horowitz. 2015. “When Leaders Matter: Rebel Experienceand Nuclear Proliferation.” Journal of Politics 77(1):72–87.

23

Gaddis, John Lewis. 2011. George F. Kennan: An American Life. New York, NY: W.W. Norton.

Gartzke, Erik. 2001. “Democracy and the Preparation for War: Does Regime Type A↵ect States’Anticipation of Casualties?” International Studies Quarterly 45:467–484.

Gartzke, Erik. 2007. “The capitalist peace.” American journal of political science 51(1):166–191.

Gelpi, Christopher F. and Michael Griesdorf. 2001. “Winners or losers? Democracies in internationalcrisis, 1918–94.” American Political Science Review 95(3):633–647.

Gunitsky, Seva. 2014. “From Shocks to Waves: Hegemonic Transitions and Democratization in theTwentieth Century.” International Organization 68(3):561–597.

Hafner-Burton, Emilie M., Alex Hughes and David G. Victor. 2013. “The Cognitive Revolution andthe Political Psychology of Elite Decision Making.” Perspectives on Politics 11(2):368–386.

Hafner-Burton, Emilie M., Brad L. LeVeck, David G. Victory and James H. Howler. 2014. “DecisionMaker Preferences for International Legal Cooperation.” International Organization 68(4):845–876.

Hermann, Margaret G. and Charles W. Jr. Kegley. 1995. “Rethinking Democracy and InternationalPeace.” International Studies Quarterly 39(4):511–533.

Holsti, Ole R. and James N. Rosenau. 1988. “The domestic and foreign policy beliefs of Americanleaders.” Journal of Conflict Resolution 32(2):248–294.

Horowitz, Michael C. and Allan C Stam. 2014. “How Prior Military Experience Influences TheFuture Militarized Behavior of Leaders.” International Organization 68(3):527–559.

Imai, K., L. Keele, D. Tingley and T. Yamamoto. 2010. “Causal mediation analysis using R.”Advances in social science research using R pp. 129–154.

Imai, K., L. Keele, D. Tingley and T. Yamamoto. 2011. “Unpacking the black box of causality.”American Political Science Review 105(4):765–789.

Kertzer, Joshua D. 2016. Resolve in International Politics. Princeton, NJ: Princeton UniversityPress.

Kertzer, Joshua D. and Ryan Brutger. 2016. “Decomposing Audience Costs: Bringing the AudienceBack into Audience Cost Theory.” American Journal of Political Science 60(1):234–249.

Lake, David A. 1992. “Powerful Pacifists: Democratic States and War.” American Political ScienceReview 86(1):24–37.

Levin, Dov H and Robert F Trager. 2019. “Things You Can See From There You Can’t See FromHere: Blind Spots in the American Perspective in IR and Their E↵ects.” Journal of Global SecurityStudies 4(3):345–357.

Linde, Jona and Barbara Vis. 2017. “Do Politicians Take Risks Like the Rest of Us? An ExperimentalTest of Prospect Theory Under MPs.” Political Psychology 38(1):101–117.

Manekin, Devorah, Guy Grossman and Tamar Mitts. 2015. “The Domestic Sources of TerritorialConflict: The Case of Israel.” Working paper.

Mansfield, Edward D and Jack Snyder. 1995. “Democratization and the Danger of War.” Interna-tional security 20(1):5–38.

24

Mintz, Alex and Nehemia Geva. 1993a. “Why don’t democracies fight each other? An experimentalstudy.” Journal of Conflict Resolution 37(3):484–503.

Mintz, Alex and Nehemia Geva. 1993b. “Why Don’t Democracies Fight Each Other? An Experi-mental Study.” Journal of Conflict Resolution 37(3):484–503.

Mintz, Alex, Nehemia Geva, Steven B. Redd and Amy Carnes. 1997. “The E↵ect of Dynamic andStatic Choice Sets on Political Decision Making.” American Political Science Review 91(3):553–566.

Mintz, Alex, Steven B Redd and Arnold Vedlitz. 2006. “Can we generalize from student experimentsto the real world in political science, military a↵airs, and international relations?” Journal ofConflict Resolution 50(5):757–776.

Mueller, John E. 1971. “Trends in Popular Support for the Wars in Korea and Vietnam.” AmericanPolitical Science Review 65(2):358–375.

Narang, Vipin and Paul Staniland. 2018. “Democratic Accountability and Foreign Security Policy:Theory and Evidence from India.” Security Studies Forthcoming:1–38.

Pape, Robert A. 2003. “The Strategic Logic of Suicide Terrorism.” American Political ScienceReview 97(3):343–361.

Press, Daryl Grayson. 2005. Calculating credibility: How leaders assess military threats. Ithaca, NY:Cornell University Press.

Prins, Brandon C. 2003. “Institutional instability and the credibility of audience costs: Politicalparticipation and interstate crisis bargaining, 1816-1992.” Journal of Peace Research 40(1):67–84.

Reiter, Dan and Allan C Stam. 2002. Democracies at War. Princeton, NJ: Princeton UniversityPress.

Renshon, Jonathan. 2015. “Losing Face and Sinking Costs: Experimental Evidence on the Judgmentof Political and Military Leaders.” International Organization 69(3):659–695.

Risse-Kappen, Thomas. 1995. “Democratic Peace - Warlike Democracies? A Social ConstructivistInterpretation of the Liberal Argument.” European Journal of International Relations 1(4):491–517.

Rousseau, David L. 2005. Democracy and war: institutions, norms, and the evolution of interna-tional conflict. Stanford: Stanford University Press.

Saunders, Elizabeth N. 2011. Leaders at War: How Presidents Shape Military Interventions. Ithaca,NY: Cornell University Press.

Schultz, Kenneth A. 1999. “Do Democratic Institutions Constrain or Inform? Contrasting TwoInstitutional Perspectives on Democracy and War.” International Organization 53(2):233–266.

Schumpeter, Joseph. 1955. The Sociology of Imperialism. In Imperialism and Social Classes. Cleve-land: World Publishing Company.

Sechser, Todd S. 2011. “Militarized compellent threats, 1918–2001.” Conflict Management and PeaceScience 28(4):377–401.

Starr, Harvey. 1992. “Democracy and War: Choice, Learning and Security Communities.” Journalof Peace Research 29(2):207–213.

25

Tomz, Michael R. and Jessica L. P. Weeks. 2013. “Public Opinion and the Democratic Peace.”American Political Science Review 107(4):849–865.

Valentino, Benjamin A., Paul K. Huth and Sarah E. Croco. 2010. “Bear Any Burden? HowDemocracies Minimize the Costs of War.” Journal of Politics 72(2):528–544.

Weeks, Jessica L.P. 2014. Dictators at War and Peace. Ithaca, NY: Cornell University Press.

Wittkopf, Eugene R and Michael A Maggiotto. 1983. “Elites and masses: A comparative analysisof attitudes toward America’s world role.” The Journal of Politics 45(2):303–334.

Yarhi-Milo, Keren. 2014. Knowing the Adversary: Leaders, Intelligence, and Assessment of Inten-tions in International Relations. Princeton, NJ: Princeton University Press.

26

Democratic Leaders, Crises, and War

Supplementary Appendix

Contents

A Recruitment protocol 2A.1 Recruitment letter to Knesset Members . . . . . . . . . . . . . . . . . . . . . . . . . 2

Figure 1: Recruitment Letter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2A.2 Recruitment procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2A.3 Participant verification protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

B Study instrumentation 6B.1 Experimental protocol (translated to English) . . . . . . . . . . . . . . . . . . . . . . 6B.2 Individual di↵erence measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

B.2.1 Military assertiveness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8B.2.2 International trust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8B.2.3 Right-wing ideology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9B.2.4 Arab-Israeli conflict . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

B.3 Mechanism questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

C Supplementary analysis 13C.1 Representativeness and survey non-response . . . . . . . . . . . . . . . . . . . . . . . 13

Table 1: Gauging the representativeness of the sample . . . . . . . . . . . . . . . . . 14C.2 Knesset results table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16

Table 2: Regression model: Knesset sample . . . . . . . . . . . . . . . . . . . . . . . 16C.3 Israeli public samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

Table 3: Israeli Public Samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18Table 4: Elite-Mass Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19

C.4 Israeli public results table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20Table 5: Regression models: Israeli public samples . . . . . . . . . . . . . . . . . . . 21

C.5 Heterogeneous e↵ects in the Israeli sample . . . . . . . . . . . . . . . . . . . . . . . . 22C.6 Downsampling Israeli public results . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

Figure 2: Downsampled crisis results . . . . . . . . . . . . . . . . . . . . . . . . . . . 25C.7 Comparison of e↵ect size magnitude with large-N work . . . . . . . . . . . . . . . . . 26

Table 6: The probability of crisis reciprocation: Bilateral MIDs . . . . . . . . . . . . 27Table 7: Predicted probabilities of crisis reciprocation . . . . . . . . . . . . . . . . . 28

C.8 Exploiting the two-stage nature of the experimental design . . . . . . . . . . . . . . 29Figure 3: Exploiting the two-stage structure of the experiment . . . . . . . . . . . . 31

C.9 E↵ect of regime type treatment on mechanism questions . . . . . . . . . . . . . . . . 32Figure 4: Dovish preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Figure 5: Selection e↵ects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33Figure 6: Domestic constraints: leader costs of war . . . . . . . . . . . . . . . . . . . 33Figure 7: Credibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34Figure 8: Military conduct . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

1

A Recruitment protocol

A.1 Recruitment letter to Knesset Members

. , ,

..

:! !" . .

.

: , . .

, :

! . , , . : XXXXX

, ). (! ,

. , . .

' : .

,

, , , !

Figure 1: Recruitment Letter

A.2 Recruitment procedures

We began the recruitment process for the elite sample by compiling a dataset of all 408 individuals

who had served as members of the Parliament of Israel (i.e., the Knesset) from the beginning of

the 14th Knesset in June 1996 through the 20th Knesset (the current Knesset) that was sworn in in

2

March 2015. We compiled a data set that included the following information about our population:

1. full name

2. party a�liation while in Knesset

3. names of all Knesset committees on which (s)he served

4. number of terms served

5. whether (s)he served as a minister in the government, and if so, what portfolios (s)he held

6. whether (s)he was a member of the Cabinet

Contact information for our participants was obtained through a variety of channels, including

the Secretary of the Knesset, the Knesset Channel, the di↵erent parties’ leadership o�ces in the

Knesset and other government o�ces where former Knesset members are currently employed. Email

addresses for all current members of the Knesset were obtained through the Secretary of the Knesset.

To verify whether the contact information we obtained was correct, we either called or emailed all

the former Knesset members from the last twenty years and asked them if they would be interested

in taking a “10 minute electronic survey by a team of professors from leading American Universities.”

29.4% of the initial population was removed from the sampling frame at this stage, either because

the members were deceased, were too sick to participate, or because their contact information was

out of date and newer contact information could not be found. This process left us with a sample

of 288 potential candidates to take our survey. This pool included all 120 current members of the

Knesset along with 168 former members whose contact information was available.

On July 10, 2015, we executed a soft launch of our on-line survey. The survey included a

recruitment email, written in Hebrew (reproduced in Appendix §A.1), a link to our on-line survey,

and an individual six-digit password that was pre-assigned to each member. In the following days,

we emailed the invitation to all current and former members in our dataset. A few weeks later,

we sent a reminder email to those who had not responded to the survey. We sent a third round

of reminders a few weeks later. In between these rounds, we phoned former and current Knesset

members or their assistants to remind them to take the survey. In early August, the Director

of Academic A↵airs at the Knesset, together with the Secretary of the Knesset, sent an email to

all current Knesset members encouraging them to take the survey, repeating essentially the same

information we provided in the introductory email.

3

In addition to the on-line survey, we created identical hard-copy versions of our survey. In

mid-August we sent those who had not responded to our survey a reminder email and attached an

electronic copy of our survey that could be opened in Microsoft Word. Respondents were given the

option of either faxing or emailing the completed survey back to us. That same six-digit code was

the only identifying information on the paper copies of the survey, allowing us to track completion

among our sample population. Members of our research team also traveled to the Knesset on four

separate occasions to invite current members to participate.

The entire recruitment process was done in Hebrew. Two Hebrew-speaking research assistants

and one member of the research team who is a native Hebrew speaker corresponded with the members

of the Knesset or their assistants. Participants were informed that there would receive no financial

reward for taking the survey, but that we would be happy to share with them the results of the

survey. Moreover, participants were promised full anonymity: with the exception of the research

team, participants were assured that identifiable information would not be released or reported.

A.3 Participant verification protocol

We took several steps to increase our confidence that the current and former decision-makers partic-

ipated in the study rather than members of their sta↵. First, in the introductory email we explicitly

indicated that the questionnaire should be fielded by the decision-maker himself, and not by mem-

bers of his or her sta↵. We explained that the code we provided to access the on-line survey was

personal, and should not be shared with others. Importantly, we did not o↵er any material incentives

for filling out the survey, to dissuade decision-makers and assistants for taking the survey for those

material reasons.

Second, in the survey itself we asked the participants to enter their complete date of birth.

This allowed us to compare this information with the date of the decision-makers in o�cial Knesset

records. Third, for the 75% of our sample consisting of former Knesset members, a Hebrew-speaking

research assistant and one of the authors were both in touch with the decision-maker directly via

phone or email, and confirmed with him/her that they were the ones taking the survey. Anecdotally,

our research team found that many of our participants were quite eager for the opportunity to opine

on issues of foreign policy to an outside audience.

In the case of some current Knesset members, after receiving approval from their parliamentary

assistant, a Hebrew-speaking research assistant from our team or one of the authors gave the Knesset

4

members the survey directly and picked it up from them within a two-hour window. However, some

Knesset members wished to maintain their anonymity and thus were not in direct contacted with

the research team.

Finally, although we follow best practices, as is always the case with elite experiments, we should

note that decision-makers who wished to “cheat” and delegate their participation to others could

have probably found ways to do so. However, the combination of the types of questions asked in

the survey, the absence of material compensation for survey completion, our explicit request the

survey not be filled out by others, and the enthusiastic response to our survey from most of the

decision-makers who took the survey leave us confident that the vast majority of them participated

directly.

5

B Study instrumentation

B.1 Experimental protocol (translated to English)

There is much concern these days about the spread of conflict. We are going to describe a situation

the international community could face in the future. For scientific validity, the situation is general,

and is not about any specific countries in the news today. Some parts of the description may strike

you as important; other parts may seem unimportant. After describing the situation, we will ask

you to make predictions about what you think will happen.

Here is the situation:

• Two countries are currently involved in a public dispute over a contested territory. The dispute

has received considerable attention in both countries, because of the risk that disputes like these

can escalate to the use of force.

• Country A [is a democracy/is a dictatorship]. Country B [is a democracy/is a dictatorship].

• Both countries have moderately powerful militaries, with large armies, moderate sized-navies,

and well-trained air forces.

• Neither country is a close ally of the United States.

• Country A is slightly larger than Country B, though their economies are approximately the

same size.

• Country A has moderate levels of trade with the international community. Country B has

high levels of trade with the international community.

• The last time the two countries were involved in an international dispute, di↵erent leaders

were in power.

Given the information available, what is your best estimate about whether Country A will stand

firm in this dispute, ranging from 0% to 100%? [scaled from 0 (A will definitely not stand firm) to

100 (A will definitely stand firm)]

Now, we’d like to ask you a di↵erent question about the same situation. As a reminder, here is

the situation:

6

• Two countries are currently involved in a public dispute over a contested territory. The dispute

has received considerable attention in both countries, because of the risk that disputes like these

can escalate to the use of force.

• Country A [is a democracy/is a dictatorship/recently transition to democracy]. Country B [is

a democracy/is a dictatorship/recently transition to democracy

• Both countries have moderately powerful militaries, with large armies, moderate sized-navies,

and well-trained air forces.

• Neither country is a close ally of the United States.

• Country A is slightly larger than Country B, though their economies are approximately the

same size.

• Country A has moderate levels of trade with the international community. Country B has

high levels of trade with the international community.

• The last time the two countries were involved in an international dispute, di↵erent leaders

were in power.

If the dispute were to escalate and war were to break out, what is your best estimate about whether

Country A will win, ranging from 0% to 100%? [scaled from 0 (A will definitely not win) to 100 (A

will definitely win)]

7

B.2 Individual di↵erence measures

As discussed in the main text, both our Knesset and Israeli public participants completed question-

naires measuring military assertiveness, political ideology, stance on to the Arab-Israeli conflict, and

international trust.1 Military assertiveness (Herrmann, Tetlock and Visser, 1999) classifies individ-

uals along a single “hawk-dove” dimension and has long been considered a key dimension of foreign

policy beliefs. Recent work has also established its continued relevance, as military assertiveness

remains as related to the decision to use force, even among a sample of U.S. leaders, as it was during

the Cold War (Herrmann and Keller, 2004; Herrmann, Tetlock and Diascro, 2001; Kertzer and Brut-

ger, 2016). This construct is equally predictive of foreign policy preferences in non-U.S. samples (see,

e.g., Hurwitz, Pe✏ey and Seligson 1993, Bjereld and Ekengren 1999 and Reifler, Scotto and Clarke

2011). Our political ideology item asks subjects to classify themselves along a single dimensions

from “left” to “right” in politics, while a separate measure asks subjects to do the same with respect

to the Israeli-Arab conflict. Our international trust measure is adapted from work by Brewer et al.

(2004), which finds that generalized trust of other countries in the international system helps to

structure beliefs about the foreign policy arena. The US sample includes these same items, apart

from the Arab-Israeli conflict measure; the Korean and UK samples include the political ideology

measure, but not the military assertiveness, international trust, or Arab-Israeli conflict measures.

B.2.1 Military assertiveness

1. The best way to ensure peace is through military strength

2. The use of force generally makes problems worse

3. Rather than simply reacting to our enemies, it’s better for us to strike first.

All scaled from 1 (strongly agree) to 5 (strongly disagree). Items 1 and 3 reverse-coded such that

higher scores indicated higher level of military assertiveness.

B.2.2 International trust

1. Some people say that Israel can trust other nations, while others think that Israel can’t be

too careful in dealing with other nations. Where would you place yourself on this scale from

1 (Israel can count on other countries) to 7 (Israel cannot count on other countries)?

1Beliefs related to international trust were assessed only in Public I and the Knesset sample.

8

Reverse-scored such that higher scores indicate more higher levels of trust.

B.2.3 Right-wing ideology

1. There is much talk of “left” and “right” in politics. How would you rate yourself on a left-right

scale, from 1 (right) to 7 (left)?

Reverse-scored such that higher scores indicate more right-wing ideology.

B.2.4 Arab-Israeli conflict

1. How would you rate yourself on a scale from “left” to “right” with respect to the Israeli-Arab

conflict where 1 one is on the far right and 7 is on the far left?

Reverse-scored such that higher scores indicate more right-wing ideology.

9

B.3 Mechanism questions

The following mechanism questions were presented following the administering of the resolve and

military e↵ectiveness outcome questions. The questions were presented in a mixed factorial design,

which included both between- and within-subject manipulations. First, the mechanism questions

were presented in three blocks: one consisting of general beliefs about Country A, one soliciting

beliefs about the consequences if Country A were to win the crisis/war, and one soliciting beliefs

about the consequences if Country A were to lose the crisis/war. The order of the win/lose blocks

was randomly assigned, to avoid potential downstream e↵ects. Moreover, participants were either

given the win/lose blocks in reference to the crisis scenario, or the war scenario. Asking questions

about both types of scenarios as a between-subject factor allows us to reduce the length of the

mechanism instrumentation, while including mechanism questions about both positive and negative

outcomes allows us to construct within-subject measures of the e↵ect of losing on leader-level costs

of war, for each participant.

1. What proportion of the citizens do you think believes force should only be used as a last resort?

(1) 0-20% (2) 20-40% (3) 40-60% (4) 60-80% (5) 80-100%

2. What proportion of the national decision-makers do you think believes force should only be

used as a last resort?

(1) 0-20% (2) 20-40% (3) 40-60% (4) 60-80% (5) 80-100%

3. How sensitive do you think its citizens would be to casualties?

(1) Very sensitive (2) sensitive (3) somewhat sensitive (4) not very sensitive

4. How sensitive do you think its citizens would be to the financial costs (e.g., increased taxes)

of fighting?

(1) Very sensitive (2) sensitive (3) somewhat sensitive (4) not very sensitive

5. If Country A initiated the dispute, how likely do you think it is that it will prevail in the

dispute?

(1) Very unlikely (2) unlikely (3) neither unlikely nor likely (4) likely (5) very likely

10

6. If the leader of Country A makes a public threat, how likely do you think it is that they’Aoll

follow through?

(1) Very unlikely (2) unlikely (3) neither unlikely nor likely (4) likely (5) very likely

7. How likely do you think it is that other countries would come to Country A’s defense?

(1) Very unlikely (2) unlikely (3) neither unlikely nor likely (4) likely (5) very likely

8. How well-trained do you think the soldiers of Country A are?

(1) Very well trained (2) somewhat well trained (3)neither well trained nor poorly trained

(4) somewhat poorly trained (5) very poorly trained

9. How strong do you think the morale of the soldiers is in Country A?

(1) Very strong (2) somewhat strong (3) neither strong nor weak (4) somewhat weak (5)

very weak

[Thinking back to the original crisis, if Country A were to back down. . . ][If Country A were

to lose the war. . . ]

10. What e↵ect do you think it would have on public support for the government?

(1) Significantly increase (2) somewhat increase (3) not a↵ect (4) somewhat decrease (5)

significantly decrease

11. What e↵ect do you think it would have on the likelihood of the government remaining in o�ce?

(1) Significantly increase (2) somewhat increase (3) not a↵ect (4) somewhat decrease (5)

significantly decrease

12. What do you think is likely to happen to the leader?

(1) Remain in power (2) voted out of power (3) removed by a coup (4) exiled (5) death

[Thinking back to the original crisis, if Country A were to win. . . ][If Country A were to win

the war. . . ]

13. What e↵ect do you think it would have on public support for the government?

11

(1) Significantly increase (2) somewhat increase (3) not a↵ect (4) somewhat decrease (5)

significantly decrease

14. What e↵ect do you think it would have on the likelihood of the government remaining in o�ce?

(1) Significantly increase (2) somewhat increase (3) not a↵ect (4) somewhat decrease (5)

significantly decrease

12

C Supplementary analysis

C.1 Representativeness and survey non-response

There are two ways of thinking about the representativeness of our elite sample. The first asks how

our participants compare to the complete population of individuals who served in the Knesset from

1996 to the present. The second asks how our participants compare to our sampling frame, a di↵erent

group than the complete population because it does not include members who had passed away,

were too sick to participate, or for whom we were unable to acquire up to date contact information.

Thus, whereas the first quantity explores whether our participants look like the universe of Knesset

members in this time period, the second explores survey non-response. We explore both questions

in Table 1 below, which presents a set of linear probability models comparing our participants to the

universe of Knesset members from 1996-2015 (models 1-2) and to only those Knesset members who

had been sent the survey (models 3-4). The results show that unsurprisingly, current members of

the Knesset were less likely to participate in the survey than former members, but that interestingly,

our participants are not significantly less “elite”, as measured by the proportion of respondents with

experience as deputy ministers, or as cabinet members or higher. If anything, our sample is slightly

more experienced than the universe of decision-makers, though the number of terms in o�ce did not

significantly predict survey response.

Israeli legislative politics features a number of characteristics that makes calculating an sum-

mary partisan representativeness score somewhat complex, including a high degree of fragmentation,

parties frequently splintering and forming, frequent party switching, and the presence of Arab and

religious parties that cannot be cleanly positioned on unidimensional partisan space. As a result,

after coding all parties in the Knesset from 1996-2015 as being either left, center, right, Arab, re-

ligious, or none of the above, we look at partisan representativeness in two di↵erent ways. First,

we focus on the representation of MKs from Arab parties (e.g. Hadash, Ra’am, Ta’al, Balad, the

Joint Arab list, etc.), and religious parties (e.g. Shas, United Torah, Tkuma, Jewish Home, etc.),

which are orthogonal to the left-right spectrum and thus are separated from the main analyses. We

find that although both Arab and religious MKs are included in the sample, they are both slightly

underrepresented: MKs from Arab parties make up approximately 7% of the population of MKs

in this time range, but only 3% of the sample; MKs from religious parties make up approximately

13

Table 1: Gauging the representativeness of the sample

Compared to...All Knesset members Sampling frame

(1) (2) (3) (4)

Current member �0.043 �0.049 �0.210⇤⇤⇤ �0.184⇤⇤⇤

(0.045) (0.057) (0.054) (0.065)Highest level of experience:. . . Deputy minister 0.017 0.044 0.035 0.079

(0.054) (0.071) (0.072) (0.088). . . Cabinet member or higher �0.044 �0.098 �0.075 �0.096

(0.076) (0.098) (0.093) (0.114)Male 0.025 0.081 0.072 0.097

(0.053) (0.063) (0.067) (0.076)Terms 0.011 0.021 0.008 0.013

(0.012) (0.016) (0.015) (0.018)Left-right party membership �0.070⇤⇤ �0.063

(0.030) (0.038)Constant 0.177⇤⇤⇤ 0.312⇤⇤⇤ 0.320⇤⇤⇤ 0.436⇤⇤⇤

(0.054) (0.087) (0.070) (0.108)N 415 295 288 225R2 0.007 0.043 0.063 0.080⇤⇤p < .05; ⇤⇤⇤p < .01

20% of the population of MKs in this time frame, but only 7% of the sample.2 Second, we take the

members who were exclusively associated with parties on the main ideological spectrum (left, center,

and right), and calculate a left-right ideology score averaging across their terms in o�ce within the

sampling period, such that the lowest score (1) is assigned to an MK who consistently represented

a left-wing party, and the highest score (3) is assigned to an MK who consistently represented a

right-wing one. So, for example, an MK that served two terms in a left-wing party and a third

term in a centrist party would be coded as being slightly more conservative (1.33) than an MK that

served all of their terms in a left-wing party (1), and much less conservative than MK who served

their terms in right-wing parties (3). Models 2 and 4 in Table 1 show that our sample is slightly

more left-wing than the population of MKs in this time period as a whole, but interestingly, that

this skew is not due to non-response bias, in that left-leaning MKs were not significantly more likely

2Given the presence of party switching, we code a�liation for Arab and religious parties here using a simple binarydecision rule, in which an MK is classified as having a religious or Arab a�liation if they represented a religious orArab party at any point in time during the time period under investigation. Thus, an MK who served a term asa member of a religious party, but then switched to a non-religious party in a subsequent term would retain theirreligious classification for the purpose of this analysis, and would be excluded from the left-right ideology analysisdiscussed below.

14

to participate in the survey than right-leaning ones. The partisan di↵erence thus appears to stem

from the probability of entering the sampling frame rather than the probability of response.3

3Supplementary analyses suggest that the average left-right ideology amongst deceased MKs is more conservative(2.4) than living MKs (2.0), potentially explaining some of this e↵ect.

15

C.2 Knesset results table

Table 2: Regression model: Knesset sample

Resolve Military E↵ectiveness

(1) (2) (3) (4)

Democracy �5.886 �6.971 14.279⇤⇤⇤ 14.128⇤⇤⇤

(4.324) (4.378) (3.145) (2.969)Male 2.228 6.960

(6.565) (4.404)Military assertiveness �0.107 25.794

(19.128) (12.990)International trust �14.226 8.190

(9.559) (6.423)Combat experience �7.430 �0.590

(4.720) (3.181)L-R ideology 0.178 �31.778⇤⇤

(19.735) (13.274)Hawk (Arab/Israeli conflict) �4.950 18.234

(21.214) (14.435)Constant 59.136⇤⇤⇤ 70.048⇤⇤⇤ 44.302⇤⇤⇤ 27.360⇤⇤⇤

(3.057) (10.820) (2.224) (7.310)N 88 82 86 80R2 0.021 0.096 0.197 0.312⇤⇤p < .05; ⇤⇤⇤p < .01

16

C.3 Israeli public samples

The first study of the Israeli public (Israel I ) was piloted on September 30, 2015 and fielded from

October 6-9, 2015. The second study of the Israeli public (Israel II ) was piloted on January 17,

2016 and fielded from January 18-25, 2016. Descriptive statistics for each of these samples can be

found in Table 3. Descriptive statistics for the Korean, and UK, and US samples are presented in

the text.

Table 4 compares our three Israeli samples along several dimensions, all scaled from 0 to 1 for

ease of interpretability. First, we note that current and former Knesset members do not di↵er very

much on observed covariates; the only statistically significant di↵erence between the two subsamples

is that current leaders are on average 12 years younger than former leaders. The di↵erences along

the ideational dimensions are all small, and none of them are statistically significant. Unsurpris-

ingly, our leader sample averages about 20 years older than our respondents from the Israeli public,

and is generally less conservative, less hardline on the Arab-Israeli conflict, and more trusting of

international institutions than the public at large.

Along the ideational dimensions that we measured, Israeli leaders are notably less conservative,

less hawkish with respect to the Arab-Israeli conflict and more trusting of international institutions

than the general public. For example, with respect to general ideology, Knesset members averaged a

score of 0.45, a score that places them nearly an entire standard deviation less conservative than the

mean score in the Israeli public samples. With respect to the Arab-Israeli conflict, our leaders were

an entire standard deviation below the mean for the public samples. They also scored higher on

international trust, and lower on military assertiveness, though those di↵erences were not significant.

17

Table 3: Israeli Public Samples

Israel I Israel II

Male 53% 52%Education:

No High School degree 2% 2%High School degree 33% 26%

Some college 23% 22%College degree 27% 31%Masters degree 14% 17%Doctoral degree 2% 2%

Military Experience:Did not serve 22% 20%

Served, no active combat 50% 49%Combat experience 28% 31%

Location:Jerusalem 10% 10%Tel Aviv 12% 15%

Central Zone 18% 18%Haifa 15% 14%

Northern Region 11% 13%Southern Region 12% 9%

Lowland 10% 11%Sharon Area 8% 9%

Yehuda and Shomron 3% 2%Religiosity:

Secular 60% 54%Traditional 19% 30%Religious 15% 14%Orthodox 5% 3%

Birth Country:Israel 81% 81%

Former USSR 10% 11%Other 9% 9%

Mean (SD) Mean (SD)Age 41.8 (14.6) 41.8 (14.8)Military Assertiveness 0.58 (0.19) 0.56 (0.20)Right Wing Ideology 0.61 (0.25) 0.59 (0.24)Hawkishness (Arab-Israeli conflict) 0.63 (0.25 ) 0.62 (0.25 )International Trust 0.32 (0.27) —

18

Elite–PublicKnesset Israeli Public Gap

Current Former Overall I II I IIAge 52.2 64.4 61.4 41.8 41.8 19.6 19.5Military Assertiveness 0.61 0.61 0.61 0.58 0.56 0.03 0.05Right Wing Ideology 0.47 0.44 0.45 0.61 0.59 –0.16 –0.14Hawkishness (Arab/Israeli) 0.44 0.37 0.39 0.63 0.62 –0.24 –0.23International Trust 0.37 0.41 0.40 0.32 — 0.08 —

Table 4: Elite-Mass Comparison: Statistically significant di↵erences in means between public sampleand leader (overall) sample depicted in bold; p-values calculated via Wilcoxon rank-sum tests.

19

C.4 Israeli public results table

20

Tab

le5:

Regressionmod

els:

Israelipublicsamples

Resolve

Militarye↵

ectiveness

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Dem

ocracy

�4.591⇤

⇤⇤�4.752⇤

⇤⇤�1.675

�1.857

4.964⇤

⇤⇤4.984⇤

⇤⇤8.227⇤

⇤⇤7.948⇤

⇤⇤

(1.232)

(1.237)

(1.273)

(1.281)

(0.985)

(0.986)

(1.114)

(1.118)

Male

�2.174

�3.176⇤

⇤�1.366

�1.947

(1.405)

(1.449)

(1.119)

(1.265)

Age

0.03

50.028

0.071

0.086⇤

⇤

(0.047)

(0.047)

(0.037)

(0.041)

Education

�3.825

0.258

�2.128

�6.630⇤

⇤⇤

(2.760)

(2.877)

(2.198)

(2.512)

Political

know

ledge

5.028

6.370

�2.491

1.040

(3.635)

(3.540)

(2.896)

(3.088)

Com

bat

0.717

�1.339

�0.980

�0.679

(1.509)

(1.530)

(1.202)

(1.335)

Militaryassertiveness

3.803

0.781

9.819⇤

⇤⇤7.002⇤

⇤

(4.051)

(4.048)

(3.227)

(3.532)

International

trust

�0.780

0.034

(2.474)

(1.970)

Ideology

5.323

2.523

1.694

6.681

(4.452)

(4.781)

(3.546)

(4.171)

Haw

k(A

rab-Israeli)

�1.474

2.239

�1.630

�3.282

(4.334)

(4.552)

(3.452)

(3.977)

Bornin

Israel

0.118

3.355⇤

⇤�0.239

�0.029

(1.631)

(1.682)

(1.299)

(1.467)

Con

stan

t57.864

⇤⇤⇤

50.825

⇤⇤⇤

59.913

⇤⇤⇤

49.681

⇤⇤⇤

48.523

⇤⇤⇤

43.914

⇤⇤⇤

50.062

⇤⇤⇤

44.474

⇤⇤⇤

(0.876)

(4.701)

(0.897)

(4.379)

(0.701)

(3.745)

(0.784)

(3.825)

N1,100

1,089

1,111

1,102

1,100

1,089

1,110

1,101

R2

0.012

0.023

0.002

0.016

0.023

0.042

0.047

0.069

Sam

ple

#I

III

III

III

II⇤⇤p<

.05;

⇤⇤⇤ p

<.01

21

C.5 Heterogeneous e↵ects in the Israeli sample

Our main approach in the paper to test whether the results generalize outside of the Israeli context

is to field the study in a range of other democratic countries (the United States, South Korea, and

the United Kingdom), where we find broadly similar results. Nonetheless, we can also invesitgate

the question of generalizability from a di↵erent angle, following best practices suggested by many

comparativists and qualitative IR scholars (e.g. George and Bennett, 2005), by exploring within-

case variation: though Israel may di↵er from other countries in important ways, we can exploit

within-country variation along those dimensions to provide further insight into our findings’ scope

conditions.

In the case of Israel, three features in particular are worth noting. First, it may the be case

that Israeli citizens’ unique familiarity with the military (nearly all non-Orthodox Jewish Israelis

serve at some point in their lives) a↵ects their attitudes towards the use of force. Consequently,

our sample of public citizens may be more knowledgeable and more experienced with “use of force”

decisions than the typical democratic citizen in another country. Despite compulsory service in the

IDF, however, there is variation in who actually experiences combat. We exploit this variation to

estimate whether there are heterogenous treatment e↵ects with respect to combat experience.Yet

participants’ combat experience does not appear to a↵ect their beliefs about the e↵ects of regime

type on resolve or military e↵ectiveness.4 This suggests that our results are not particular to Israel

by virtue of that country’s unique and intimate familiarity with conflict.

Two additional potentially unique features of Israel present themselves as dimensions on which

our subjects might di↵er systematically from citizens in other democracies: knowledge of interna-

tional politics and the high proportion of citizens who immigrated from former communist countries.

As with military experience, however, the dimensions are ones in which we can use the variation

within our samples to address potential concerns about generalizability. As in our analysis of com-

bat experience above, we find no significant interaction e↵ect between political knowledge and our

treatment (for resolve: p < 0.884; for military e↵ectiveness: p < 0.901). A related concern is that

the waves of new immigrants from the former Soviet Union — nearly a million Jews immigrated to

Israel from the former Soviet states between 1989-2006 — may a↵ect our results if citizens socialized

4More formally, we pool the two Israeli public surveys and estimate a series of regression models, regressing eachdependent variable (either resolve or military e↵ectiveness) on the democracy treatment, combat experience, theinteraction between the two, and a set of demographic controls. The interactions lack statistical significance (onresolve: p < 0.348; on military e↵ectiveness: p < 0.361).

22

in non-democratic states display a distinctive set of political attitudes (Pop-Eleches and Tucker,

Forthcoming). Here again we fail to find heterogenous treatment e↵ects (p < 0.399 for resolve;

p < 0.435 for military e↵ectiveness), suggesting that this aspect of the composition of Israeli society

is not a strict boundary condition for our findings.

Related to issue of generalizability are the potential concerns about confounding. For example,

the repeated conflicts in which Israel has been involved might cause Israelis to see their adversaries

as more resolved than participants in other countries, a particular danger if those subjects also

interpreted the scenario as being about Israel. To preempt this, we designed our studies to invoke

two hypothetical countries with features that are inconsistent with those of Israel: only half the

participants see a country described as a democracy, and that democracy is described as not being

a close ally of the United States. Yet even if participants did have Israel in mind, it is unlikely

to change the interpretation of our findings. For example, if Israelis simply see every adversary as

more resolved (than a Canadian respondent would, for example), this might explain why we see a

negative e↵ect of democracy on beliefs about resolve in crises, but would not explain the positive

e↵ect of democracy on beliefs about resolve in military battles.

Finally, Israeli participants might be more likely to interpret the scenario in the context of the

Middle East, while participants in other regions might be more likely to rely on other analogies or

mental models instead (Khong, 1992). To test whether our findings are contingent on the mental

models participants employ, we presented participants with a list of potential regions (Central Asia,

North America, Western Europe, Sub-Saharan Africa, Middle East/North Africa), and asked them to

rank the regions in terms of their likelihood of experiencing the scenario described in the experiment.

In this manner, we can assess whether participants who imagine the scenario taking place in di↵erent

parts of the world understand the relationship between democracies, resolve, and war di↵erently. We

employ Yilmaz, Aslam and Robertson’s (2008) ⌧ap statistic, a weighted rank similarity measure,5 to

test whether participants who imagine the scenario taking place in di↵erent regions make di↵erent

predictions about the relationship between democracy, crises, and war. Reassuringly, we find no

evidence of heterogeneous treatment e↵ects across rank similarity scores, indicating that the regional

context participants have in mind does not a↵ect the conclusions we reach.6

5We assigned each participant a rank similarity score in reference to a ranking consisting of the average positionof each region. Unlike unweighted rank similarity statistics like Kendall’s ⌧b, weighted rank similarity measures,commonly used in information science, assign di↵erential penalties based on the position in the ranking at which theyoccur. We also replicate our analyses using Kendall’s ⌧b, and find the results hold.

6As before, we estimate a set of regression models, regressing each dependent variable on the democracy treatment,

23

the regional similarity measure, the interaction between the two, and a variety of demographic controls. The interactionlacks statistical significance across all four models (on resolve: p < 0.972 for ⌧ap, p < 0.580 for ⌧b; on militarye↵ectiveness: p < 0.793 for ⌧ap, p < 0.450 for ⌧b).

24

C.6 Downsampling Israeli public results

Figure 2: Downsampled crisis results

In Crises...

-20 -10 0 10Effect of democracy

Density

SampleIsr Public I (Downsampled)

Isr Public II (Downsampled)

Knesset

The figure shows that the heavier tails in the distribution of treatment e↵ects in the Knesset results in the mainpaper is due to the relatively small sample size rather than peculiarities about the sample; when we downsample theresults from the two public samples, sampling N = 89 observations with replacement rather than N = 1000, we find

similar distributions as in the Knesset survey.

25

C.7 Comparison of e↵ect size magnitude with large-N work

The experimental results found fairly consistent patterns across all seven samples. One question

that remains concerns the substantive size of the e↵ects: how important is democracy in crises if

participants only see it as having at most a 6% e↵ect? How does this compare to the observed e↵ect

of democracy in crises as measured in large-N work?

To address this question, we use replication data that Downes and Sechser (2012) provide for

Schultz’s (2001) pathbreaking exploration of democratic performance in disputes. Table 6 presents

four sets of models taken from models 3 and 4 from table 5.5 in Schultz (2001, 146-147). The

data consist of bilateral militarized interstate disputes (MIDs) using version 2.1 of the MID data;

because of missingness, the analysis focuses on the years 1816-1984.7 The dependent variable is a

dichotomous variable indicating whether the MID was reciprocated. The first two columns of Table

6 include data from the two world wars, while the second two do not; each model is estimated using

either a logistic regression, or a linear probability model.

The centerpiece of the models is the first three rows, which indicate the regime type of the

initiator and target in each dispute. Schultz is interested in particular in the coe�cient estimate for

the democratic initiator variable, which is negative and significant across all four models, suggesting

that democracies fare better in crises because their targets “back down without a fight.” (p. 9). The

coe�cients in columns 2 and 4 are already on a probability scale, but to interpret the substantive

e↵ects for the logit model, we calculate the predicted probabilities, presented in Table 7, which

depicts predicted probabilities using models 1 and 3, for contiguous dyads, a non-democratic target

state, and a policy demand, for with all other covariates set to their mean values for each of the four

power configurations. The results suggest e↵ect sizes ranging from -6.3% to -8.5%. In this sense,

they are similar to the results from the linear probability models in columns 2 and 4, which depict

e↵ect sizes for democratic initiators facing non-democratic targets as ranging from -6.6% to -7.9%.

A direct comparison between Schultz (2001) and our experimental results is di�cult, for any

number of reasons: questions about the limitations of the MID data for testing these types of

questions (Downes and Sechser, 2012), questions about the limitations of observational data more

generally in causally identifying the e↵ects of democracy (Tomz and Weeks, 2013), di↵erences be-

tween the setup of our experimental scenario and the covariate profile in the MID data, and so on.

Despite these caveats, and the di↵erent conclusions we draw from our data, it is notable that the

7For more details about the sources of the variables in the model, see Schultz (2001); Downes and Sechser (2012).

26

Table 6: The probability of crisis reciprocation: Bilateral MIDs

(1) (2) (3) (4)World wars World wars World wars World warsincluded included excluded excluded

Democratic Initiator �0.314⇤ �0.066⇤ �0.374⇤ �0.079⇤

(0.181) (0.038) (0.191) (0.041)Democratic Target �0.003 �0.002 �0.051 �0.012

(0.159) (0.034) (0.169) (0.036)Both Democratic 0.364 0.077 0.489 0.102

(0.343) (0.073) (0.361) (0.077)Major Power Initiator-Major Power Target �0.171 �0.036 �0.314 �0.067

(0.285) (0.061) (0.306) (0.065)Major Power Initiator-Minor Power Target �0.254 �0.054 �0.242 �0.054

(0.201) (0.043) (0.210) (0.045)Minor Power Initiator-Major Power Target 0.218 0.045 0.232 0.051

(0.252) (0.054) (0.263) (0.057)Initiator’s share of capabilities 0.070 0.017 0.060 0.012

(0.240) (0.052) (0.247) (0.053)Contiguous 0.501⇤⇤⇤ 0.110⇤⇤⇤ 0.547⇤⇤⇤ 0.120⇤⇤⇤

(0.145) (0.031) (0.155) (0.033)Alliance portfolio similarity 0.143 0.033 0.167 0.039

(0.223) (0.048) (0.234) (0.050)Status quo evaluation of initiator �0.134 �0.026 �0.083 �0.016

(0.192) (0.041) (0.205) (0.043)Status quo evaluation of target �0.219 �0.048 �0.261 �0.057

(0.215) (0.046) (0.229) (0.049)Territory 0.288⇤ 0.069⇤ 0.273 0.063⇤

(0.165) (0.036) (0.174) (0.038)Government 0.327 0.074 0.236 0.053

(0.375) (0.080) (0.383) (0.082)Policy �1.142⇤⇤⇤ �0.259⇤⇤⇤ �1.215⇤⇤⇤ �0.279⇤⇤⇤

(0.149) (0.032) (0.158) (0.034)Other �0.649 �0.151 �0.608 �0.141

(0.569) (0.127) (0.588) (0.132)Constant 0.051 0.513⇤⇤⇤ 0.075 0.521⇤⇤⇤

(0.269) (0.058) (0.281) (0.061)N 1,305 1,305 1,153 1,153AIC 1,646.909 1,447.088Model type Logit OLS Logit OLSWorld war years dummies Yes Yes No No⇤p < .1; ⇤⇤p < .05; ⇤⇤⇤p < .01

27

Table 7: Predicted probabilities of crisis reciprocation

Initiator-Target Nondemocratic initiator Democratic initiator E↵ectWorld wars included

Major Power - Major Power 30.8% 24.6% -6.3%Major Power - Minor Power 31.7% 25.4% -6.4%Minor Power - Major Power 39.7% 32.5% -7.2%Minor Power - Major Power 39.1% 31.9% -7.2%World wars excluded

Major Power - Major Power 27.8% 20.9% -6.9%Major Power - Minor Power 32.2% 24.7% -7.6%Minor Power - Major Power 39.9% 31.4% -8.6%Minor Power - Major Power 39.3% 30.8% -8.5%

Predicted probabilities calculated using the coe�cient estimates from Table 6, columns 1 and 3. The predictionsshown are for a contiguous dyad, a nondemocratic target state, and a policy demand, with all other variables set to

their mean values within each power configuration.

e↵ect sizes we find for democracies in crises in our experiment are similar to those found in one of

the canonical large-N studies of this question. In this sense, our maximum e↵ect size of democracy

in crises of 6% is substantively meaningful in the context of other work that has also explored this

question.

28

C.8 Exploiting the two-stage nature of the experimental design

As noted in the main text, one of the innovations of our experimental design is that we study

respondents’ beliefs about democracies in both crises and war, which not only broadens the scope

of the investigation to democracies’ reputations, but also more closely connects our research design

to contemporary bargaining theories of war, which view crises and war as part of one continuous

process (Smith and Stam, 2004; Powell, 2004). This also enables us to explore questions in the

like the di↵erence-in-di↵erence between democracies vs dictatorships in crises vs wars: one of the

striking patterns we show in Figure 2 in the main text is that across all seven samples, respondents

believe democracies are at a significantly larger advantage in wars than in crises.

But to what extent do democratic advantages in war outweigh potential disadvantages in crises?

We can exploit the two-dimensional structure of our experiment to get at this question more directly.

The top panel in Figure 3 presents a series of mosaic plots. Here, for each of our seven samples, we

dichotomize each dependent variable at the 50% mark (dropping those respondents in each sample

who provided answers right at 50%, such that crisis observations are coded either as standing firm

or backing down, and war observations as either winning or losing). The x axis therefore presents

the proportion of respondents who thought the state in question was more likely to stand firm

versus back down in the crisis, while the y axis depicts the proportion of respondents who thought

the state in question was more likely to win versus lose in the war; in each sample, responses

in the democracy condition are presented in red, and responses in the dictatorship condition in

aquamarine. The mosaic plots show considerable variation across our samples in terms of the

proportion of respondents who believe democracies are at an advantage in both crises and wars (with

most respondents perceiving a democratic advantage in war across most samples, but more between-

sample variation in beliefs about democratic performance in crises). To more clearly illustrate

respondents’ beliefs, then, we conduct a simple bootstrapping exercise, in which, for each sample,

we:

• Create a simulated dataset by sampling from the actual data with replacement

• Using the simulated dataset, calculate the product of both dependent variables, expressed as

a percentage (e.g. if a respondent saw the country as having a 50% chance of standing firm

and a 60% chance of winning, the score would be 0.5 x 0.6=30%)

• Calculate the mean score for respondents in the democracy condition, and the mean score for

29

respondents in the dictatorship condition

• Subtract the mean score for respondents in the dictatorship condition from the mean score for

respondents in the democracy condition, to calculate the overall e↵ect of democracy

• Repeat B = 1500 times to derive the bootstrapped distribution of scores

We plot these scores in the density plots in the bottom panel of Figure 3. They show that for one

sample (USA), democracies’ perceived disadvantage in crises outweighs their perceived advantage

in war, another sample (Israel I), the two phases cancel each other out, and in the remaining five

samples, democracies’ advantage in war outweighs their perceived disadvantage in crises.

30

Figure

3:Exp

loitingthetw

o-stagestructure

oftheexperim

ent

Isra

el I

Isra

el II

Isra

el: K

ness

etR

epub

lic o

f Kor

eaU

nite

d K

ingd

om I

Uni

ted

Kin

gdom

IIUSA

Democracy Dictatorship

Back

Down

Stand

Firm

Back

Down

Stand

Firm

Back

Down

Stand

Firm

Back

Down

Stand

Firm

Back

Down

Stand

Firm

Back

Down

Stand

Firm

Back

Down

Stand

Firm

Lose

Win

Lose

Win

Crisis

War

Reg

ime

Type

Democracy

Dictatorship

-50

510

15O

vera

ll ef

fect

of d

emoc

racy

(Pr(

Sta

nd F

irm) *

Pr(

Win

))

Sample

Isra

el I

Isra

el II

Isra

el: K

ness

et

Rep

ublic

of K

orea

Uni

ted

Kin

gdom

I

Uni

ted

Kin

gdom

II

USA

Thetop

panel

presents

aseries

ofmosaic

plots,in

which

resp

onsesto

each

dep

enden

tva

riable

within

each

sample

hav

ebeen

dichotomized

at50%,to

dep

ictth

eproportionofea

chsample

whoperceives

thestate

inquestionasbeingmore

like

lyto

standfirm

vsback

dow

nin

acrisis

(onth

exaxis),

andwin

vslose

inwar(on

theyaxis).

Thebottom

pan

elpresents

thebootstrapped

distributionofth

eov

eralle↵

ectof

dem

ocracy,ca

lculatedasth

eproduct

ofth

elikelihoodth

atth

estate

inquestion

willstan

dfirm

inacrisis,an

dwin

inawar.It

show

sth

atin

general,dem

ocrac

ies’

adva

ntage

sin

war

areperceived

asou

tweigh

ingan

ydisad

vantage

sin

crises,althoughth

isva

ries

across

samples.

31

C.9 E↵ect of regime type treatment on mechanism questions

Figure 4: Dovish preferences

p < 0 p < 0 p < 0 p < 0

Dovish public Dovish leaders Casualty sensitivity Financial sensitivity

1

2

3

4

5

Density

Regime TypeDemocracy

Dictatorship

32

Figure 5: Selection e↵ects

p < 0

Pr(Victory | Initiate)

1

2

3

4

5

Density

Regime TypeDemocracy

Dictatorship

Figure 6: Domestic constraints: leader costs of war

Pr(Govt stays in office) Public support

p < 0.272

p < 0.068

p < 0.093

p < 0.202

-3

-2

-1

0

1

-3

-2

-1

0

1

Effect of losing crisis

Effect of losing w

ar

Density

Leader's fate

p < 0.021

p < 0

1

2

3

4

5

1

2

3

4

5

Effect of losing crisis

Effect of losing w

ar

Density

Regime TypeDemocracy

Dictatorship

33

Figure 7: Credibility

p < 0.001

Follow through on threat?

1

2

3

4

5

Density

Regime Type

Democracy

Dictatorship

Figure 8: Military conduct

p < 0 p < 0 p < 0

Allies to the defense Soldiers' training Soldiers' morale

1

2

3

4

5

Density

Regime TypeDemocracy

Dictatorship

34

References

Bjereld, Ulf and Ann-Marie Ekengren. 1999. “Foreign policy dimensions: A comparison between the

United States and Sweden.” International Studies Quarterly 43(3):503–518.

Brewer, Paul R, Kimberly Gross, Sean Aday and Lars Willnat. 2004. “International trust and public

opinion about world a↵airs.” American Journal of Political Science 48(1):93–109.

Downes, Alexander and Todd Sechser. 2012. “The Illusion of Democratic Credibility.” International

Organization 66(3):457–489.

George, Alexander L. and Andrew Bennett. 2005. Case Studies and Theory Development in the

Social Sciences. Cambridge, MA: MIT Press.

Herrmann, Richard K and Jonathan W Keller. 2004. “Beliefs, values, and strategic choice: US

leaders’ decisions to engage, contain, and use force in an era of globalization.” Journal of Politics

66(2):557–580.

Herrmann, Richard K, Philip E Tetlock and Matthew N Diascro. 2001. “How Americans think

about trade: Reconciling conflicts among money, power, and principles.” International Studies

Quarterly 45(2):191–218.

Herrmann, Richard K., Philip E. Tetlock and Penny S. Visser. 1999. “Mass Public Decisions to Go to

War: A Cognitive-Interactionist Framework.” American Political Science Review 93(3):553–573.

Hurwitz, Jon, Mark Pe✏ey and Mitchell A Seligson. 1993. “Foreign Policy Belief Systems in Com-

parative Perspective.” International Studies Quarterly 37(3):245–270.

Kertzer, Joshua D. and Ryan Brutger. 2016. “Decomposing Audience Costs: Bringing the Audience

Back into Audience Cost Theory.” American Journal of Political Science 60(1):234–249.

Khong, Yuen Foong. 1992. Analogies at War. Princeton, NJ: Princeton University Press.

Pop-Eleches, Grigore and Joshua Tucker. Forthcoming. Communism’s Shadow: The E↵ect of Com-

munist Legacies on Post-Communist Political Attitudes. Princeton, NJ: Princeton University

Press.

Powell, Robert. 2004. “Bargaining and Learning While Fighting.” American Journal of Political

Science 48(2):344–361.

35

Reifler, Jason, Thomas J Scotto and Harold D Clarke. 2011. “Foreign Policy Beliefs in Contemporary

Britain: Structure and Relevance.” International Studies Quarterly 55(1):245–266.

Schultz, Kenneth A. 2001. Democracy and coercive diplomacy. New York, NY: Cambridge University

Press.

Smith, Alastair and Allan C Stam. 2004. “Bargaining and the Nature of War.” Journal of Conflict

Resolution 48(6):783–813.

Tomz, Michael R. and Jessica L. P. Weeks. 2013. “Public Opinion and the Democratic Peace.”

American Political Science Review 107(4):849–865.

Yilmaz, Emine, Javed Aslam and Stephen Robertson. 2008. A New Rank Correlation Coe�cient for

Information Retrieval. In Proceedings of the 29th Annual International ACM SIGIR Conference

on Research and Development in Information Retrieval. pp. 587–594.

36