Who War! is it Good For? The Relationship between Regime Type, the Fate of Leaders and War
Alexandre Debs and H. E. Goemans September 2, 2008
Abstract We propose and test a formal model of war and domestic politics, consistent with recent evidence on the relationship between regime type, the outcome of war and the probability and consequences of losing oce. Our model builds on two fundamental and hitherto neglected dierences between regime types: the cost of replacing leaders and the post-tenure fate of leaders. We show that war is less likely as 1) the cost of replacement depends less on the outcome of war and 2) the consequences of losing oce are less punitive for leaders. Compared to non-democratic leaders, the cost of replacing a democratic leader depends relatively little on the war outcome and democratic leaders fare relatively well after losing oce. Our model therefore offers an intuitive explanation for the democratic peace. To examine the scope of our theory, we test and nd support for it within the sample of non-democratic regimes. Keywords: war, democratic peace, two-level game
We thank Patrick Kuhn for help with the data of non-democratic regimes (and, by transitivity, Jennifer Gandhi). We also thank Bruce Bueno de Mesquita, Thad Dunning, Barbara Geddes, Gretchen Helmke, Thorsten Janus, Stu Jordan, Shawn Ramirez, Ken Schultz, Curt Signorino, Scott Wolford, participants in the Wallis Political Economy Working Group and Comparative Politics Workshop at the University of Rochester and the 2008 MPSA and APSA meetings and in particular Bob Powell for extremely helpful comments.
1
1
Introduction
The democratic peace, a proposition that democracies do not ght one another in wars, is a well established empirical nding. It is so broadly accepted that Levy (1988) called it the closest thing to an empirical law in international relations. However well-documented it may be, it is fair to say that no explanation for the democratic peace has reached a consensus among scholars.1 We oer a new theory to explain the conict propensity of a very broad array of potential regime types which generates an intuitive explanation for the democratic peace. We focus on two fundamental and hitherto neglected dierences between regime types. First, we can usefully distinguish regime types by the eect of the outcome of war on the costs of replacing the leader. Second, we can usefully distinguish regime types by the post-tenure fate of their leaders; e.g., how leaders fare after they lose oce. These two features systematically dier between democracies and non-democracies. In democracies, leaders are typically replaced by elections or term limits, whereas leaders in non-democratic regimes are replaced through violent means and face additional punishment. (See Table 2 in section 4.1.2 below.) Since the costs of voting are low and do not depend on the outcome of a war, for democratic leaders the probability of losing oce should depend little on the outcome of the war. In contrast, the costs of removing an autocrat can be quite high and depend on the outcome of war. For example, it would be relatively costly to replace an autocrat who wins a war, since he is likely to be surrounded by well-armed cronies, willing to ght to ensure his survival. On the other hand, if the autocrat loses the war, his military resources are likely to be depleted or demoralized, and it should be easier to overthrow him. Consistent with our theory, the survival rate of autocrats is aected signicantly more by the war outcome than is the survival rate of democrats. (See Chiozza and Goemans (2004, 610) and Table 1 in section 4.1.2 below.) Current theories of the democratic peace t poorly with this nding, as they start from the assumption that democrats are more sensitive to the war outcome, or appeal to un-modeled selection eects. We develop a formal model which exploits these fundamental distinctions between regime types. The model predicts that peace should occur when the cost of replacing the leaderand therefore his survival probabilitydepends relatively little on the outcome of the war and when the net gain of staying in oce is relatively small. In those circumstances, rulers are likely to accept an international settlement that avoids the destruction of war, since their survival is not at risk and since the consequences of losing oce are relatively benign. The model yields predictions consistent with the recent literature on the democratic peace (Russett and Starr, 2000; Bueno de Mesquita and Ray, 2004), in that it predicts that war
For a review of the evidence on the democratic peace, see, among others, Ray (1998) and Russett and Oneal (2001). Critics of the nding include Gowa (1998) and Rosato (2003).
1
2
is least likely when both countries are democratic, slightly more likely when only one of the countries is a democracy, and nally most likely when both countries are autocratic. Moreover, since the theory is built on two observable regime characteristics (the eect of the war outcome on the survival rate of leaders and the consequences of losing oce), we can use the model to predict the war-proneness of a broader array of regime types. To that end, we examine the sample of non-democratic countries in the post World-War II period, using the classication introduced in Cheibub and Gandhi (2004) and Gandhi and Przeworski (2006). We show that while dictators do not systematically dier in their sensitivity to the war outcome, civilian dictators fare signicantly better after they leave oce than do military dictators and monarchs. The theory therefore predicts that civilian leaders should be least war-prone, a pattern which is conrmed in the data. We proceed as follows. Section 2 presents the set-up of the model that guides our analysis and section 3 solves the baseline model. Proofs of the formal results are relegated to the appendix. Section 4 situates the model in the relevant literature and highlights where and why our approach diers from previous models that link regime type and war. Section 5 develops and tests a novel prediction about the war proneness of dierent types of non-democracies. Section 6 concludes.
2
Set-Up of the Model
We develop and analyze a one-shot game with two countries and four players, i.e. a leader L(i) and a population P (i) for each country i {A, B}. Leaders divide an international pie of size 1. They can either agree on a division of the pie or decide to launch a war. Let I w {0, 1} designate whether war occurs (where I w = 1 designate that countries go to war, I w = 0 otherwise). We allow for a general bargaining mechanism, where the leader of country i picks an action ai within an action set Ai . We assume that a war can be declared unilaterally by any leader (i.e. there is an element aw in action set Ai that triggers a war). An i international war produces a cost cj for any actor j. It is won by country A with probability p (and by country B with probability 1 p). After a country wins a war, it gets the whole international pie. Let zA characterize the allocation of the international pie after resolution of the international conict (either through peaceful bargaining or war), with zi going to country i (zB = 1 zA ). The utility of any actor j is linear in its countrys share of the international pie, valued at rate vj 0 by j.2 After resolution of the international conict, the population of each country decides whether to replace
Note that we do not impose any restriction on whether the international pie is a public or a private good. It could be a private good, where L(i) and P (i) get a share vL(i) and vP (i) of zi , respectively (where vL(i) + vP (i) = 1). Or it could be a public good, valued by actor j at rate vj .
2
3
its leader. Replacing a leader imposes a (domestic) cost d for the population of country i and brings a benet bi . For reasons which we explain in section 4, we focus our attention on the costs of replacing rulers, abstracting from the eect of war on the benet of keeping the incumbent. The cost d represents any eort expended and any personal risk incurred by the population in removing the ruler, as well as any economic loss produced in the transition from one ruler to the next. We assume that d is a function of the share of the international pie going to country i, its regime type ti {D, N } (where D stands for democracy and N for non-democracy) and the way in which the payo zi was obtained (peacefully or through war). First, we assume that it is more costly to replace a ruler who obtains a better outcome on the international scene (either because he has more resources to buy o supporters, or because he benets from increased legitimacy). Second, we emphasize that the process of replacing a ruler diers signicantly by regime type. In a democratic regime, a ruler is replaced through an election, which is a relatively costless process. Moreover, democracies are generally buttressed by strong bureaucracies, which minimize any loss associated with a transition. In a non-democratic regime, a ruler is replaced through violent means. This process is relatively costly, since insurgents risk losing their life in a revolt against the ruler. Also, transition losses should also be larger in non-democracies, with less developed bureaucracies. Third, given that a dictator relies on his armed forces to remain in power, while a democrat is replaced non-violently, the war outcome should have a greater eect on the cost of replacing non-democratic rulers. (We justify this assumption in greater detail in section 4.) In addition, we assume leaders are easier to replace if they have expended resources in a war. Summarizing the discussion, the function d(zi , ti , I w ) satises the following properties: d(zi , ti , I w ) zi d(zi , N, I w ) d(1, N, 1) d(0, N, 1) d(zi , ti , 1) 0 ti , I w > d(zi , D, I w ) zi , I w > d(1, D, 1) d(0, D, 1) d(zi , ti , 0) zi {0, 1} (1) (2) (3) (4)
The benet bi represents a (net) preference shock for a domestic challenger. It can be a function of the performance of the ruler on dimensions which are outside the current model (for example, the domestic economy). It is unknown at the start of the game and follows a uniform distribution U [l, h], where l 0 and h > 0 is arbitrarily large. Write F as the cdf of bi . The uncertainty about bi is lifted just before the population decides whether to replace the ruler.3
Note, again, that the benets of replacing the leader do not depend on the outcome of the war. We explain this choice in more detail below.
3
4
Rulers strictly prefer to remain in power. Let an incumbent ruler receive a payo Oi if he is ousted and Ii if he is kept in oce. We assume that Ii and Oi depend on the regime type of country i. As mentioned above, the replacement processes dier in democracies and non-democracies. Democrats can expect successful careers in the private sector and give speeches for exorbitant fees. Autocrats, in stark contrast, can be much less sanguine about their prospects after they lose oce as over a third is exiled, jailed or killed in their rst year out of power. Moreover, the amount of rents that leaders can accrue while in oce is generally more limited in a democracy than in a non-democracy. Summing up, we assume that Ii (ti ) and Oi (ti ) are such that Ii (N ) Oi (N ) > Ii (D) Oi (D) > 0 (5)
We can now make the timing of the game and the solution concept more explicit.
2.1
Timing of the Game
1. Leaders (L(A) and L(B)) pick their actions (aA and aB ). 2. Countries get their share of the international pie (zA and zB ) 3. Uncertainty over bi is lifted in each country i 4. Population (P (A) and P (B)) decide whether to replace its leader (they pick rA and rB , respectively).
2.2
Payos and Solution Concept
Players are risk-neutral and we can write the value of the game to the leader and the population in country i, VL(i) and VP (i) , as follows: VL(i) = vL(i) zi I w cL(i) + ri Oi (ti ) + (1 ri )Ii (ti ) VP (i) = vP (i) zi I w cP (i) + ri [bi d(zi , ti , I w )] We solve for a subgame-perfect equilibrium. We assume that a war obtains if and only if there is no division of the international pie (zA , zB ) that both leaders prefer to going to war.
3
Solution of the Model
We can solve the game by backward induction. The solution is fairly simple. The population does not replace its ruler if and only if the cost of replacement exceeds the preference shock for the domestic challenger, i.e. ri = 0 if and only if bi < d(zi , ti , I w ) 5 (6)
Therefore, a ruler stays in power with probability F (d(zi , ti , I w )). Assumption (2) therefore rationalizes the evidence on the average (unconditional) survival rate of leaders. More precisely, it implies that democrats are in general more accountable than autocrats, i.e. they lose power at a faster rate, for example if they commit blunders on the domestic scene which enter into the determination of bi . Moreover, assumption (3) rationalizes the evidence that autocrats are more sensitive to the war outcome, since it implies that F (d(1, N, 1)) F (d(0, N, 1)) > F (d(1, D, 1)) F (d(0, D, 1)) Keeping the crucial distinction between the unconditional rate of survival and the conditional rate of survival (i.e. conditional on the war outcome) in mind, we continue to solve the game by backward induction. Moving up, the leader of any country selects his action ai , anticipating the probability with which he would be replaced. We can then show one of our main results: Proposition 1. I w = 0 if, for i {A, B}, [F (d(1, ti , 1)) F (d(0, ti , 1))] [Ii (ti ) Oi (ti )] < cL(i) Proof. See the appendix This proposition states that if the survival function is not too sensitive to the outcome of the international conict and the gain to remain in oce is not too large, then rulers can always avoid war. In particular, rulers strictly prefer to get their expected share of the pie ((zA , zB ) = (p, 1 p)) and avoid the cost of going to war. As we argue above, democracies are more likely to satisfy condition (7), since the process of replacing leaders depends relatively little on the war outcome. In some sense, such regimes are closer to the ideal case of a unitary state. Indeed, models of war which assume that states are unitary players implicitly impose condition (7), by assuming that there are no private benets of war for the leader. (In the language of our model, this would correspond to the case where the utility of being in oce is equal to the utility of being out of oce, Ii (ti ) = Oi (ti ), or where the outcome of international conict does not aect the leaders survival probability, F (d(zi , ti , I w )) = (ti ) zi , I w ). As we know, we should not expect war to happen in such a model, if there is no commitment problem, asymmetric information or issue indivisibility (Fearon, 1995).4 To illustrate, Figure 1 oers a numerical example, plotting the net benet of peaceful settlement at zi (net of the expected value of going to war) for certain parameter values that satisfy condition (7). We see that in such cases, there is a peaceful settlement that both countries prefer to war.
For a convincing argument that issue indivisibility does not constitute a separate rationalist explanation for war, see Powell (2006).
4
(7)
6
Figure 1 here While proposition 1 is useful, it only oers a sucient condition for peace. We would like to make a stronger statement and nd necessary and sucient conditions for peace. First, we parametrize F (d(zi , ti , I w )) so that we can make meaningful comparative statics. We assume: F (d(zi , ti , I w )) = (ti ) + (ti )G(zi , I w ) (8)
where (ti ) (0, 1), (ti ) > 0, G1 (zi , I w ) > 0 (where G1 (zi , I w ) denotes the derivative with respect to the rst argument of the function G(zi , I w )). In this framework, (ti ) parametrizes the leaders unconditional likelihood of losing ofce (or the level of the survival function). Everything else being equal, a higher (ti ) describes a regime where leaders are in general less accountable, since they lose power for a given vector (zi , I w ) at a slower rate. On the other hand, (ti ) parametrizes the conditional likelihood of losing oce (as a function of the war outcome). Everything else being equal, a greater (ti ) describes a regime where the war outcome has a greater eect on the leaders survival. Second, we want to focus on cases which do not produce a trivial prediction for war and peace. The following conditions suce, while imposing a minimum level of symmetry between the two countries: F (d(p, tA , 0)) < pF (d(1, tA , 1)) + (1 p)F (d(0, tA , 1)) F (d(1 p, tB , 0)) < pF (d(0, tB , 1)) + (1 p)F (d(1, tB , 1)) (9) (10)
To understand these conditions, note that if neither hold, then a peaceful division of the pie where each ruler gets his expected share of the pie ((zA , zB ) = (p, 1 p)) is necessarily preferred to war. In other words, peace would obtain for any value of (ti ), (ti ), Ii (ti ) Oi (ti ) and we could not explain which critical regime characteristics aect the decision to go to war. We want to rule out such an obvious solution, and see whether leaders can arrive at a peaceful settlement. Note also that this set of conditions is relatively mild (and is weaker than imposing that F (d(zi , ti , I w )) is convex in zi ti ).5 Then we can show: Proposition 2. Let F (d(zi , ti , I w )) be as given in (8) and satisfy (9) and (10). Then i {A, B}, and everything else being equal, (a) I w is independent of (ti ).
It is easy to show that if we restrict attention to a general set of functions F (d(zi , ti , I w )), which are either strictly concave, strictly convex or linear in zi , then we obtain non-trivial comparative statics with respect to (ti ), (ti ), Ii (ti ) Oi (ti ) only if F (d(zi , ti , I w )) is strictly convex in zi , and the results of proposition 2 still hold.
5
7
(b) There is a cut-o R+ {}such that I w = 0 (ti )[Ii (ti ) Oi (ti )] Proof. See the appendix. This proposition characterizes necessary and sucient conditions for war under a very general functional form for F and minimum degree of symmetry between countries A and B.6 First, it states that the unconditional likelihood of losing oce does not aect the decision to go to war. This is straightforward. Only the marginal impact of the war outcome and the cost of peaceful concessions should matter. Previous research, which has assumed that the general level of accountability of leaders inuences their decision to go to war, is therefore misguided. Second, it states that war happens if and only if the private stakes for the leader are suciently salient (either because the conditional likelihood of losing oce, (ti ), or the dierence in utility between being in and out of oce, Ii (ti ) Oi (ti ), is suciently large). This result is slightly more subtle. Note that conditions (9) and (10) imply that there is a conict between leaders as concerns their private benet of going to war. More precisely, both leaders want more than their expected share of the pie in war to equate their private benet in peace to the level of their private benet in war. This conict in private benets is more likely to lead to war as private benets become more salient. It may be easier to sharpen our intuition with the help of numerical examples. 1 Figures 2 and 3 generate numerical examples with G(zi ) = zi 2 100 I w , drawing the benet of peaceful settlement at zi (net of the expected value of going to war) for certain parameter values.7 Figure 2 here Figure 3 here In Figure 2, the private benet for war for the leader of country A is small. Therefore, this leader accepts a large range of peaceful settlements to avoid the cost of war (cL(A) ), and peace is possible. In Figure 3, the private benet for war for the leader of country A is large and, by the reverse logic, war is inevitable.
It also provides conditions for which < . A micro-foundation for this functional form would be to assume that the preference shock for the domestic challenger, F , follows a uniform distribution in [0, h], with h 1, and that the cost of replacing the ruler is quadratic in the share of the international pie (d(zi , ti , I w ) = 1 2 a(ti )+b(ti )[zi 100 I w ]; where a(ti ) = h(ti ), b(ti ) = h(ti )). In other words, there are increasing marginal costs in replacing the ruler as he gets more of the international pie.
7 6
8
As we argue above, we consider that democracies are those regimes with smaller private benets for war and greater willingness to make peaceful concessions.8 To further understand this result, it is important to note that victory and defeat in war correspond to two, extreme, peaceful divisions of the pie, minus the resources spent in war. We assume that the leaders sensitivity to war outcomes is indicative of the leaders sensitivity to peaceful divisions of the pie. In other words, the more victory improves the leaders survival relative to defeat, the more a good peaceful settlement should improve the leaders survival relative to a bad peaceful settlement.9 As the dierence between victory and defeat has a greater impact on the leaders survival ((ti ) is larger), concessions in peaceful bargaining become more costly. Therefore, leaders become more dicult partners in international bargaining and they are less likely to nd a peaceful compromise. By the same logic, leaders who care more about remaining in power (Ii (ti ) Oi (ti ) is larger) insist on getting a larger share of the pie in peaceful bargaining and hence are more likely to generate war. Thus, our model diers in signicant ways from the existing rationalist literature on the democratic peace. In the next section, we extensively discuss this literature and, where appropriate, examine the empirical validity of our approach.
4
Our Model Situated in the Literature
The democratic peace proposition goes as far back as Kant (2006). According to Kant, democracies spread a culture of liberal peace and are unlikely to ght, since citizens, who bear the cost of war, will be careful to avoid war if possible. Following his lead, explanations of the democratic peace typically fall into two camps: cultural/normative and institutional/structural explanations (Maoz and Russett, 1993). Cultural/normative explanations stress that leaders should apply the same norm of peaceful conict resolution they employ domestically to other states. Democracies tolerate internal opposition peacefully, and so should attempt to solve conict with other democracies similarly peacefully (Doyle, 1983; Russett, 1993; Dixon, 1994; Weart, 1994). Structural explanations argue that democratic leaders should be constrained in their use of force, since they must obtain the consent of legislators and, eventually, the electorate (Russett, 1993). Put dierently,
In that regard, it is striking to note that the leaders who made truly signicant concessions to avoid war, such as Thorvald Stauning of Denmark, Seyss-Inquart of Austria, and Emil Hacha of Czechoslovakia all led democracies. Stauning (until he died in oce) and Hacha remained the nominal heads of their countries. Seyss-Inquart stayed in oce as governor of the new Austrian provincial administration until April 30, 1939 and later became governor in Poland and eventually Commissioner of the occupied Netherlands. 9 It is empirically impossible to estimate the marginal impact of a better peaceful division of the pie on the leaders survival. Thus, we must empirically anchor our model on our estimates of the eects of the war outcome on the leaders survival.
8
9
in a democracy, the interests of the population (who bears the cost of war) are aligned with those of the decision-maker (Kant, 2006; Babst, 1964). Rational choice theorists have recently begun to formalize some of these institutional/structural explanations of the democratic peace. For example, Jackson and Morelli (2007) modeled the intuition that the bias between leader and population may be lower in democracy which then explains their lower likelihood to wage war. Other rational choice explanations of the democratic peace emphasize how dierent regime types generate and process information (Levy and Razin, 2004; Patty and Weber, 2006). While insightful, these institutional explanations do not fully specify the differences in institutional constraints between regime types, leaving open two very important questions. First, why should rational (forward-looking) voters condition their decision to retain or replace a leader on the outcome of war? Second, why should the same outcome produce dierent decisions of voters in dierent regime types? Answers to these questions are by no means obvious. One set of answers might revolve around the potential benets of replacing the leader. It could be argued that the outcome of war reveals the competence of leaders; victory reveals competence, defeat incompetence. Voters might then decide to replace the defeated leader, in the expectation his replacement will be more competent. Smith (1998)s model of crisis bargaining follows that logic. His model, however, does not address our concerns. Smith (1998) does not answer the question why should voters care about the foreign policy competence of the leader after the (last) conict has ended? Nor does his model address the question why voters in dierent regimes should care dierentially about the competence of their leaders. (To be fair, this question is not really relevant for his purposes in that paper because, by implication, he focuses on audience costs in democracies.) The model proposed by Jackson and Morelli (2007) exhibits this same drawbacks. In an extension to their baseline model, they propose that the domestic audience chooses the bias of its leader, reecting his cost/benet calculus for going to war. But since this decision is taken before leaders negotiate and ght a war, the model does not esh out why rational audiences should carry out these plans nor why their decision systematically diers across regime types.10 In their seminal paper, Bueno de Mesquita et al. (1999) propose an answer to the question of why the benets of replacing a leader dier systematically by regime type. In their set-up, regimes dier by the relative size of their winning coalition (i.e., the set of people who must support a ruler so that he remains in power) as a fraction of the selectorate (i.e., the set of people with political power). Leaders oer a mix of private and public goods to their supporters and must thwart
10 To name another reference, Baliga, Lucca and Sjostrom (2007) present an insightful contribution on the eect of limited democratization on war. In their model, however, dictators always stay in power (by assumption) and citizens in a democracy use retrospective voting rules.
10
the threat of a challenger. Members of the current winning coalition run the risk of losing their access to private goods by supporting the challenger. The bigger is the winning coalition, the more likely they are to be included in the challengers winning coalition. Therefore, rulers with a larger winning coalition enjoy a smaller incumbency advantage in the provision of private goods and they must survive through the provision of public goods. Assuming that the war outcome is a public good, such leaders are reluctant to enter into conict unless they are certain to win. Moreover, they are unlikely to be chosen as targets, since they expend a lot of eort in war. Taken together, Bueno de Mesquita et al. (1999) claim, these factors explain the democratic peace. Let us explore this argument in more detail, to assess how it addresses the questions we raised above. The timing of their game is very similar to ours: 1) leader of nation A chooses between war and peace and, if war is selected, picks a war eort gA ; 2) leader of nation B observes the leader of nation As action and chooses how hard to ght (picks gB ); 3) nature determines the outcome of war; 4) domestic audiences in each country decide whether to retain their leader or defect to a political rival. Then, Bueno de Mesquita et al. (1999) claim that members of the winning coalition support the incumbent if and only if (1 gi ) Ri Ri + i + Vi (z) ci + Wi Si (11)
where gi is the war eort, Ri is the amount of resources in country i, redistributed among the winning coalition (of size Wi ) if the leader stays in power and redistributed randomly among the whole selectorate (of size Si ) if the challenger comes to power, ci is the expected quality of the challenger, i is the performance of the leader on all policy dimensions and Vi (z) is the utility from the war outcome z. We see that if Wi /Si is relatively large, then the incumbency advantage in the provision of private goods is relatively small, and the survival of the leader depends to a greater extent on the war outcome z. This approach raises some diculties. First, it is debatable whether the war outcome is a public good, and it would be much preferable to rely on an explanation that does not assume the war outcome is a public good.11 Second, it does not explain why rational (forward-looking) citizens should condition their replacement decision on the war outcome. By the time members of the selectorate make their decision, the war eort has already been expended and the war outcome has already been realized. Therefore, there is no reason that the war eort and outcome should aect the decision to support the ruler (particularly given that the war outcome is a public good, by denition non-excludable). Hence, their model does not fully address our two central questions: why should rational (forward-looking) voters condition their decision to retain or replace a leader on the outcome of
Note that we provide such an explanation. What we call the public benet could be interpreted either as a public or a private good.
11
11
war? And, why should the same outcome produce dierent decisions of voters in dierent regime types?
4.1
Regime Type and the Costs of Replacement
We present an alternative set of answers to the two questions we raised above. While the literature has focused on the dierent benets, we focus on the dierent costs of replacing rulers in dierent regimes. This approach has both theoretical and empirical advantages. First, a focus on costs allows for a simpler model: a one-shot game of complete information. Focusing on benets would require an innitely-repeated game of incomplete information. Second, the costliness of replacing leaders is a useful, and, indeed, traditional, way to distinguish regime types. Third, we can intuitively explain why the outcome of war fundamentally and dierentially aects the costs of replacing leaders in dierent regimes. Our approach has the distinct advantage that we can anchor it in observable characteristics of regimes. This allows us to explain the war experience of dierent regime types, looking at the sample of non-democratic countries, as we do in section 5. 4.1.1 A Dening Feature of Democracy
The idea that the costliness of replacing rulers is a signicant, if not the most signicant dierence, between dictatorship and democracy is certainly not new. For example, Popper (1963, 124) states: For we may distinguish two main types of government. The rst type consists of governments of which we can get rid without bloodshed for example, by way of general elections; that is to say, the social institutions provide means by which the rulers may be dismissed by the ruled, and the social traditions ensure that these institutions will not easily be destroyed by those who are in power. The second type consists of governments which the ruled cannot get rid of except by way of a successful revolution-that is to say, in most cases, not at all. I suggest the term democracy as a short-hand label for a government of the rst type, and the term tyranny or dictatorship for the second. It is clear that the competitive struggle for the peoples votes, as emphasized in the traditional Schumpeterian denition of democracy, would be meaningless if candidates were not protected against the arbitrary power of the state (Schumpeter, 1950). Riker (1982, 6, 7) argues: Almost everything [. . . ] that we think of as civil liberties (the rights of a speedy trial, habeas corpus, and security against unreasonable search and seizure, for example) originated to protect politicians who feared 12
prosecution if and when they lost oce. Thus the historic purpose of these fundamental democratic liberties has been not to provide freedom as an end in itself, but to render eective both political participation and the process of choice in voting. In the same spirit, in their seminal contribution, Przeworski et al. (2000, 15) adopt a Schumpeterian denition of democracy, and emphasize that regimes with no regularized competition among conicting views enjoy a monopoly of power buttressed by the threat or actual use of force. This dening feature of democracy, central in our model, was also emphasized by the early students of the democratic peace. Weart (1994) states: If required to reduce it all to a single statement, we could isolate anocratic and authoritarian regimes as those where demands for loyalty are so concentrated on a leader (or family or clique) that any citizen who works to have the leader replaced is risking severe punishment. (Weart, 1994, 308). Maoz and Russett (1993) argue that, empirically, the treatment of political opponents (i.e. the annual number of violent political deaths and the number of political executions) is the single most important factor to explain the democratic peace, along with the stability of a regime (i.e. the length of that a political system or regime existed without fundamental change). This leads them to favor normative accounts of the democratic peace.12 No rational choice theory of the democratic peace, however, has taken this distinctive feature seriously. 4.1.2 Sensitivity of Replacement Costs to the War Outcome
We next explain why the outcome of war fundamentally and dierentially aects the costs of replacing leaders in dierent regimes. Specically, we suggest a causal mechanism by which the survival of an autocrat leader is more sensitive to the outcome of a war. We argue that the fundamental processes whereby democratic leaders lose oce are well-institutionalized and do not fundamentally depend on the outcome of conict. Typically, democratic leaders lose oce through elections and term limits. The length of a term of oce and the timing of election are by and large exogenously determined and do not depend on the outcome of war.13 Moreover, a wide range of factors other than the outcome of war typically play an important role in democratic elections. Thus, even after Turkish Prime Minister Ecevit (Cyprus), British Prime Minister Churchill (WW II) and U.S. President George H.W. Bush (Gulf War I) led their countries to victory and signalled their
See also Ray (1995), who takes the historical precedent of a peaceful transfer of power as one of two criteria for a state to be democratic, along with surage for more than half the population. 13 A handful of democracies allows for the endogenous timing of elections. Even among those, only Thatchers calling of elections after Britains victory against Argentina in the Falklands/Malvinas War would suggest that Thatcher called the election early because of her victory. Even in that case, however, Thatcher would have had to call elections soon, no matter the outcome.
12
13
foreign policy competence, both lost the following elections because of the electorates doubts about their economic competence. In sharp contrast, non-democratic leaders often lose oce through violent means (rebellions, civil wars and coups (see author)). In such cases, the outcome of conict can fundamentally aect the process whereby non-democratic leaders are removed from oce by several paths. First, defeat in war aects the ecacy of the leaders repressive apparatus, and thereby diminishes his or her grip on power, making any attempt to overthrow the leader more likely to succeed. Second, defeat in war can aect attempts to coordinate the opposition. In case of (major) defeat, opponents of the current leader may expect that others will support them in an attempt to overthrow the leader, not just because of demonstrated incompetence, but because of a widespread belief that sucient numbers will rise and protest will not be punished. The outcome of conict can thus tip a largely passive populace into a mass uprising or coup attempt (Kuran, 1991). Similarly, victory makes such coordination less likely and thereby diminishes the non-democratic leaders probability of losing oce. Third, autocrats typically rely more on their foreign policyin particular war-ghtingcompetence to shore up their domestic legitimacy. Success in war bolsters their legitimacy, whereas defeat undermines any claims to hold power to protect the nation. In contrast to democratic leaders, thus, autocratic leaders depend on the outcome of war for their legitimacy. In summary, we assume that the costs of replacing and, therefore, the tenure of democrats is less sensitive to the outcome of war. We should note, though, that the conventional wisdom in the literature assumes the very opposite: e.g., that the tenure of democratic leaders is more sensitive to the outcome of war (see, among others, Rousseau et al. (1996); Bueno de Mesquita et al. (1999); Bueno de Mesquita and Ray (2004); Bueno de Mesquita (2006a,b); Filson and Werner (2004, 2007); Peceny and Beer (2003); Ray (2003); Reiter and Stam (2002); Russett and Starr (2000)). Since our assumption is a key component of our model, and fundamentally contradicts what seems to amount to a consensus in the eld, we discuss our disagreement in more detail below.
4.2
The Empirical Record
The fundamental question whether leaders democracies of or non-democracies are more sensitive to the war outcome is, of course, ultimately an empirical question. Previous research has addressed this question and claims to have found that democrats are more sensitive. However, this previous research suers from some fundamental aws. First, the previous literature fails to distinguish between the unconditional likelihood of losing oce and the probability of losing oce conditional on the outcome of a war. While it is plausibleand indeed empirically supported (Chiozza and 14
Goemans, 2004)that democratic rulers are more likely to lose oce in any year, it does not follow they are more vulnerable to the outcome of a war. Second, the current consensus is based on shaky empirical foundations. Scholars who argue that democratic leaders are more sensitive to the outcome of war invariably cite the seminal papers by Bueno de Mesquita, Siverson and Woller (1992) and particularly Bueno de Mesquita and Siverson (1995). The results in those articles, however, do not warrant that inference. For example, the baseline equation estimated by Bueno de Mesquita and Siverson (1995) is Leaders Post-War-Onset Political Survival = a + b1 TenureL b2 TenureL*Demo b3 (Battle Deaths/10K)L +b4 Win i where TenureL is the logarithm of a leaders total time in oce prior to the war (plus 1), Demo is a dummy variable for democracy.14 From this approach Bueno de Mesquita and Siverson (1995) cannot conclude that democrats are more sensitive to the war outcome. Indeed, they impose by assumption that the war outcome has the same impact on the leader of any regime, by including a single variable (win) which does not interact the war outcome with the regime type. Second, they do not show that democrats are more sensitive to war. Indeed, they analyze a sample including only leaders who have gone to war (with no leader who stayed at peace). By interacting regime type and prior tenure in oce they show only that democrats lose power at a faster rate. This is the unconditional likelihood of losing oce, which is independent of the sensitivity to war and, we claim, irrelevant for the decision to go to war. In more recent research on a sample of all leaders 19192003, including both leaders who fought wars and leaders who stayed at peace, Chiozza and Goemans (2004) came to a fundamentally dierent conclusion, contradicting the earlier claims of Bueno de Mesquita and Siverson (1995). Chiozza and Goemans (2004) provide evidence that war aects the tenure of autocratic leadersin particular, defeat in war increases the hazard of losing ocebut war, no matter whether victory, defeat or a draw, does not signicantly aect a democratic leaders hazard of losing oce.15
In a second estimation, they also control for Nonconstitutional Overthrow. The implications of this recent research have not been fully appreciated. For example, Bueno de Mesquita (2006b, 640), re-asserts that defeat in war is often less costly politically for autocrats than it is for democrats (Chiozza and Goemans, 2003, 2004), making them more willing to ght with relatively poorer prospects of victory than are democrats. To be fair to Bueno de Mesquita (2006b) and others, Chiozza and Goemans (2003) are partly responsible for the confusion. They state: We nd that democracies are overall more peaceful than other regimes. Our ndings suggest this is because democratic leaders are more likely to lose oce and therefore less likely to initiate a crisis ((Chiozza and Goemans, 2003, 445). The idea that the unconditional probability of losing oce aects the decision to go to war was no more than a conjecture. In fact, proposition 2 shows that it is wrong.
15 14
15
Rather than rely on competing claims in the earlier literature, we provide our own test of whether and how war aects the hazards of losing oce for leaders of dierent regimes. Table 1 reports the results from tests to empirically examine whether the tenure of leaders of democracies or autocracies is more sensitive to the outcome of war. This table replicates the ndings of Chiozza and Goemans (2004), but adds an additional independent variable (the manner of entryomitted for brevity) and combines the dierent types of democratic and non-democratic regimes into a single category (parliamentary and presidential democracies are combined into a single democratic category; non-democratic and mixed regimes are combined into a single non-democratic category).16 The results show that compared to peace (the excluded category, since we include all conict variables in the model), war signicantly aects the survival probability of non-democratic rulers but not of democratic rulers. Table 1 about here Table 1 reveals that Victory in war decreases the hazard of losing oce while Defeat in war increases the hazard of losing oce for non-democratic leaders. However, neither Victory nor Defeat in war signicantly aects the tenure of democratic leaders. Point estimates corroborate our central point, that autocrats are more sensitive to the outcome of war than democrats. Moreover, Wald tests show that the dierence between victory and defeat is statistically signicant for autocratic leaders (p < 0.013), but not for democratic leaders (p < .58). To counter our conclusion that the tenure of autocrats, rather than democrats, is more sensitive to the war outcome, scholars may appeal to the logic of selection eects. However, as Signorino (2002) has shown, while the presence of strategic interaction introduces selection, it does not necessarily introduce selection bias.17 While there exist many dierent scenarios in which strategic selection could aect our sample (such as diversionary war), strategic selection could strengthen, rather than weaken our conclusion. Assume one set of potential wars, which are attractive for the democratic leaders: these are the wars which democratic leaders are almost certain to win and for which the democratic leader will be rewarded, win or lose. Autocrats will then select themselves out of these wars. Then posit a second set of potential wars, which are unattractive for democratic leaders: these
The data on leaders comes from Archigos version 2.8 (Goemans, Gleditsch and Chiozza, 2005); its temporal domain is January 1, 1919December 31, 2003. Tests revealed that only the Democracy variable violated the proportional hazard assumption of the Cox model. As recommended by Box-Steensmeier, Reiter and Zorn (2003) we therefore include a variable that interacts Democracy with time: Democracy ln(t). Tests revealed that multicollinearity is not a problem. 17 As long as the error term in the war equation is not correlated with the error term in the tenure equation. Note moreover, the conclusion of Signorino (2002, 112) that the typical selection model is most appropriate when the process essentially does not matter.
16
16
are the wars democratic leaders are almost certain to lose and as a result lose oce or democratic leaders will be punished no matter the outcome. Democratic leaders will then select out of these. As a result, the wars we would actually observe should be those that are neutral for democratic leaders, i.e. those in which the outcome is uncertain, and for which the fate of the leader depends on that outcome. If such strategic selection determines the sample we observe, and the fate of the leader depends on the outcome, we should expect to see this dependence in the sample of wars that are actually fought. Since the tenure of democratic leaders does not appear to depend on the outcomes of the wars that are foughtrecall, even Victory does not paythere is reason to think there exists no relationship between the tenure rewards and punishment and the war outcome for democratic leaders.18
4.2.1
Consequences of a Loss of Oce
The second property of our model is generally accepted: the consequences of losing oce are much worse for autocrats than for democrats. While this has been shown to hold for the sample of leaders who selected themselves into war (Goemans, 2000), to date no study has examined the consequences of losing oce for the whole sample of all leaders (both those who stayed at peace and those who engaged in conict) or included this information in their explanation of the democratic peace. Intuitively, democratic leaders, especially in modern times, can look forward to a protable retirement. As President Bush put it when asked about his plans after he leaves the White House, Ill give some speeches, just to replenish the ol coers. In Book, Bush Peeks Ahead To His Legacy, The New York Times, September 2, 2007, p.1).19 In contrast, non-democratic leaders such as Saddam Hussein, can and indeed do anticipate signicant punishment in case they lose oce. Table 2 about here Table 2 reveals the post-tenure fate of democratic and non-democratic leaders from 1919 to 2004. We see that the great majority of democratic leaders (93%) suffered no signicant punishment in the year after they lost oce. In contrast, about 41% of non-democratic leaders faced signicant additional punishment.20 Thus, the net gain of staying in oce surely is lower for democrats than for autocrats.
We thank Kenneth Schultz for his advice and insight on this issue. Bush added:I dont know what my dad gets its more than 50-75 thousand dollars a speech, and Clintons making a lot of money. In their tax lings for the 2008 presidential campaign, the Clintons revealed they had made over $109 million since 2000. The New York Times, April 5, 2008, Clintons Made $109 Million in Last 8 Years. For a broader discussion of the prospect of retiring leaders, see Into the sunset. How ex-leaders adjust to life with less power. Financial Times, Thursday, December 27, 2007, p. 7. 20 Data on the manner of losing oce and the post-exit fate of leaders comes from the Archigos data (Goemans, Gleditsch and Chiozza, 2005).
19 18
17
5
Disaggregating Non-Democracies
One attractive feature of the model is that it explains not just a known nding the democratic peacebut also suggests new, previously unlooked for, patterns of war participation. Proposition 2 states that war is more likely when the leaders survival function is more sensitive to the outcome of international conict and when the net gain of staying in oce is larger. In this section, we expand on the logic of proposition 2 to propose additional, and novel, predictions about the likelihood of war as a function of the type of nondemocratic regime. While the comparative politics literature has long discussed the structural dierences among non-democratic regimes, the consequence of these theories for the war involvement of non-democratic regimes has been relatively neglected. The study of non-democratic regimes in the post-war period has generally followed Huntington (1991)s typology, distinguishing between military dictatorships, one-party states and personal dictatorships (see, for example, Geddes (1999, 2003)). According to this literature, the democratization experience of these regimes diered signicantly. Ruling cliques in military dictatorships and oneparty states were able to control the process of democratization, securing guarantees for themselves after the regime change, while personalist dictatorships crumbled or were overthrown. But the implications for the fate of the leader after any turnover (whether it leads to democratization or not) are unclear. Moreover, the theory is silent as to whether the tenure of certain leaders should be more sensitive to the outcome of the war.21 Peceny, Beer and Sanchez-Terry (2002), Lai and Slater (2006) and Weeks (2007) present a rst step in studying the war experience of non-democratic regimes. Peceny, Beer and Sanchez-Terry (2002) ask whether the empirical ndings of the democratic peace translates to dictatorial sub-types, i.e. whether certain sub-types of dictatorial states are less likely to engage into conict with one another than with other states. They nd weak support for an overall dictatorial peace. They do nd, however, that single-party states are signicantly less likely to engage into conict with one another, although these comprise the only homogeneous dyad to have produced war. They also conjecture that there is no systematic dierence in the institutional constraint of leaders by dictatorial subtype and argue that such a variable is unlikely to account for their war experience. Weeks (2007) contends that certain criteria used by Geddes (1999) give us good proxies about the degree of, again, the unconditional accountability of leaders. She argues that the leaders who are most accountable should be least likely to
For recent advances in the formal study of non-democratic regimes, see among others, Acemoglu and Robinson (2006); Acemoglu, Ticchi and Vindigni (2007); Besley and Kudamatsu (2007); Debs (2007a,b,c); Egorov and Sonin (2006); Gandhi and Przeworski (2006); Myerson (2008); Padro i Miquel (2007); Svolik (2006).
21
18
initiate military conict, and presents evidence in support of her claim. Lai and Slater (2006) build on their own typology of non-democratic regimes and suggest that the leaders who are least secure in power should be most likely to initiate conict. They test this hypothesis and nd empirical support for it. While enlightening, these explanations have some shortcomings and leave room for signicant improvements. First, the logic of Lai and Slater (2006) is contradicted by the democratic peace literature: democrats are least secure in power and they are least likely to initiate conict. Second, as we show above, the unconditional safety of the rulers in power should have no eect on the likelihood of engaging in military conict. Instead, we should examine the sensitivity of the leaders survival to the outcome of the war. Neither of these studies present convincing evidence that the war outcome aects the tenure of leaders as their theory suggests (Peceny, Beer and Sanchez-Terry (2002) and Lai and Slater (2006) do not present this statistics, and although Weeks (2007)s point estimates are consistent with her story, the dierences are not statistically signicant). Third, none of these studies examines the fate of the leaders after they are removed from oce; the consequences of losing oce are assumed to be the same for all regime types. We seek to address these shortcoming by disaggregating autocracies into dierent sub-types, according to the classication suggested by Gandhi and Przeworski (2006). To that end, we merge their data with Archigos, taking care to attribute regime type to the appropriate leader. This produces a data set of non-democracies with monarchs, military, or civilian leaders, from 1946 to 1996. We assess these leaders on the two fundamental factors driving our model: their sensitivity to war outcomes and their post-tenure fate. We rst examine whether the dierent types of autocracies were sensitive to the outcome of war. Table 3 reports the results. For brevity, we omit the coecients of Age at entry (positive and signicant at the 1% level), manner of Entry (negative and not signicant), Civil War (positive and signicant at the 1% level), GDP per capita (negative and not signicant), Growth in GDP per capita (negative and signicant at the 1% level) and Population (negative and signicant at the 10% level).22 Table 3 about here Table 3 shows that among the Autocratic leaders who remained at peace, Civilians as well as Monarchs face signicantly lowers hazards of a removal from oce than did Military Autocrats. The longer they stay in oce, however, the more Monarchs become like Military Autocrats. Turning to the eect of war, we
Note that an interaction between Monarch and time is included since the Monarch variable violated the proportional hazard assumption of the Cox model.
22
19
notice that while the eect of Victory is to increase the tenure of leaders in all three Autocratic regimes, in each case this eect is associated with high degrees of uncertainty, resulting in insignicant overall eects. Defeat, on the other hand, signicantly increases the hazard of losing oce for leaders of all three Autocratic regime types. Comparing the coecients on Victory and Defeat for each of the regime types allows us to examine the sensitivity of the dierent regimes to the outcome of war. Wald-test show that both Military and Civilian autocrats are sensitive to the outcome of war. The dierence between Defeat and Victory is statistically signicant at the 10% level for both Military (p < .083) and Civilian (p < .053) Autocrats. For Monarchs, however, the dierence between Victory and Defeat does not reach statistical signicance (p < .243). Notably, the eects of Victory and Defeat are not signicantly dierent between the regime types. Hence, with the potential exception of Monarchs, the sensitivity to the outcome of war does little to distinguish the incentives of these Autocrats. We therefore turn to the post-tenure fate of these leaders. Table 4 reports a simple cross-tabulation of the three types of Autocratic leaders and their posttenure fate. Table 4 about here Table 4 shows that Civilian Autocratic leaders are by far most likely to enjoy a safe retirement. We also see that only 23 Monarch lost oce.23 Letting a regime type ti {M o, C, M i}, where M o stands for Monarchical dictatorship C for a military dictatorship and M i for military dictatorship, we therefore assume: max{Ii (M o) Oi (M o), Ii (M i) Oi (M i)} > Ii (C) Oi (C) > 0 In line with proposition 2, we expect that the regime types with the biggest contrast in their leaders post-exit fate should face signicantly dierent incentives for war. Conjecture 1. Everything else equal, Military and Monarchical Dictatorships should be signicantly more war-prone than Civilian dictatorships.
To deal with the low number of Monarchs who lost oce, we also estimated models where we combined Monarchs with Civilian leaders, and Monarchs with Military leaders. In both these combinations, Non-Military and Non-Civilian Autocratic leaders faced signicantly dierent hazards of losing oce as a result of the outcome of war. Crucially, and arming our predictions, in the probit model on War reported below, the Non-military and the Civilian leaders were signicantly less likely to become involved in war.
23
20
We test this prediction in Table 5 on the Archigos data merged with the data from Gandhi and Przeworski (2006).24 Table 5 about here
The results from Table 5 show that Civilian Autocrats are indeed signicantly less likely to become involved in war than Military Autocrats. Moreover, while Monarchs are not signicantly dierent from Military Autocrats, Civilian Autocrats are signicantly dierent from Monarchs (Wald test: p < .031). We marshaled our theory to investigate the incentives to go to war of dierent types of non-democratic regimes. As demonstrated by the empirical results, our theory allows us to exploit two dimensions of the calculations of leaders: the probability as well as the consequences of losing oce. When leaders are both more sensitive to the outcome of conict and face worse consequences of losing oce, our theory unambiguously predicts they will be less inclined to accept peaceful settlements short of war. Our empirical dierentiation of non-democratic regimes allows us to conclude that leaders who face a similar probability of losing oce, but a signicantly worse fate after losing oce take their post-tenure fate into account in their decisions whether or not to get involved in war. Thus, while the previous literature focused exclusively on the relationship between regime type, war and the probability of losing oce, our empirical evidence allows us to conclude with condence that the consequences of losing oce also play a signicant a role in structuring the incentives of leaders.
6
Conclusion
Starting from the intuitive assumption that leaders at least partially base their decisions for or against war on how war in turn aects their fate, we introduced a new model that captures the relationship between regime type, the fate of leaders and war. Our model explains several known empirical patterns such as the democratic peace and the tendency of democracies to be generally more peaceful. In addition, our model suggests new avenues of research and new, heretofore unexplored, empirical patterns. In particular, we explored how war aects the fate of three types of non-democratic leaders, and in an extension of our model, used this nding to predict their relative peacefulness. We found that Civilian autocrats are signicantly better o after losing oce than either Monarchs or Military
We adjust the standard errors for within country correlation. We also examined the sensitivity, post-exit fate and war proneness of dierent non-democratic regimes using the typology and data suggested by Geddes (1999, 2003). Mirroring our ndings here, we found that Single Party Autocratic regimes had the least to fear post-exit punishment and were, therefore, signicantly less war-prone than Military and Personalist Autocratic regimes.
24
21
Autocrats. Our model predicts that Civilian autocrats should therefore be more peaceful, and empirical tests corroborated this prediction. We believe that similar extensions of our model have the potential to identify and explain signicant new patterns about war.
22
7
Appendix
Proof. (Proposition 1) The net benet of peace at zA for L(A) is NA (zA ) NA (zA ) = vL(A) (zA p) + cL(A) (12) + [F (d(zA , tA , 0)) {pF (d(1, tA , 1)) + (1 p)F (d(0, tA , 1))}] [IA (tA ) OA (tA )] The net benet of peace at zA for L(B) is NB (zA ) NB (zA ) = vL(B) (zA p) + cL(B) (13) + [F (d(1 zA , tB , 0)) {pF (d(0, tB , 1)) + (1 p)F (d(1, tB , 1))}] [IB (tB ) OB (tB )] By assumption, I w = 1 zA |NA (zA ) 0, NB (zA ) 0 Now note that F (d(p, tA , 0)){pF (d(1, tA , 1))+(1p)F (d(0, tA , 1))} F (d(0, tA , 1))F (d(1, tA , 1)) and F (d(1p, tB )){pF (d(0, tB , 1))+(1p)F (d(1, tB , 1))} F (d(0, tB , 1))F (d(1, tB , 1)) so that NA (p) 0, NB (p) 0 if (7) holds. Proof. (Proposition 2). Use the functional form for F to rewrite (12) and (13) as follows: NA (zA ) = vL(A) (zA p) + cL(A) (15) + (tA )[IA (tA ) OA (tA )] [G(zA , 0) [pG(1, 1) + (1 p)G(0, 1)]] NB (zA ) = vL(B) (zA p) + cL(B) (16) +(tB )[IB (tB ) OB (tB )] [G(1 zA , 0) [pG(0, 1) + (1 p)G(1, 1)]] Recall (14), then (a) clearly follows. Now let us show (b) for i = A (the proof for i = B follows the same steps). AB Dene (zA , zA ) such that
A G(zA , 0) = pG(1, 1) + (1 p)G(0, 1) B G(1 zA , 0) = pG(0, 1) + (1 p)G(1, 1) B A By (9) and (10), zA < p < zA . Consequently, if A (17) cL(B) cL(A) [vL(A) + vL(B) ](zA p) A +(tB )[IB (tB ) OB (tB )] pG(0, 1) + (1 p)G(1, 1) G(1 zA , 0) A A then NB (zA ) NA (zA ) > 0 and I w = 0 (tA )[IA (tA ) OA (tA )].
(14)
23
Now assume that (17) does not hold. If cL(A) cL(B) [vL(A) + vL(B) ]p (18) +(tB )[IB (tB ) OB (tB )] [G(1, 0) [pG(0, 1) + (1 p)G(1, 1)]] then there is a cut-o s.t. NA (0) NB (0) (tA )[IA (tA ) OA (tA )] . Since NB (0) > 0, then we have I w = 0 (tA )[IA (tA ) OA (tA )] . Now assume that NA (0) < NB (0) (either because (i) (18) holds and (tA )[IA (tA ) OA (tA )] > or (ii) (18) does not hold). It is clear from (15) and (16) that NA (zA ) > 0 > NB (zA ). A A A Therefore NA (zA ) > NB (zA ) and NA (0) < NB (0) imply that z A (0, zA ) such that NA (z A ) = NB (z A ). By (14), I w = 1 NA (z A ) = NB (z A ) < 0 Note that dNA (z A ) NA (z A ) z A NA (z A ) = + d(tA )[IA (tA ) OA (tA )] z A (tA )[IA (tA ) OA (tA )] (tA )[IA (tA ) OA (tA )] We calculate
z A (tA )[IA (tA )OA (tA )]
by the implicit function theorem:
(NA (z A ) NB (z A ))/(tA )[IA (tA ) OA (tA )] z A = (tA )[IA (tA ) OA (tA )] (NA (z A ) NB (z A ))/z A G(z A , 0) [pG(1, 1) + (1 p)G(0, 1)] = (vL(A) + vL(B) ) + (tA ) [IA (tA ) OA (tA )] G1 (z A , 0) + (tB ) [IB (tB ) OB (tB )] G1 (1 z A , 0) so that, using G1 (zA , 0) > 0 and NA (zA ) > NB (zA ), dNA (z A ) d(tA )[IA (tA ) OA (tA )] < 0 G(z A , 0) [pG(1, 1) + (1 p)G(0, 1)] < 0
A A A zA > z A NA (zA ) > NB (zA )
which clearly holds, since (17) does not hold.
24
References
Acemoglu, Daron, Davide Ticchi and Andrea Vindigni. 2007. A Theory of Military Dictatorships. M.I.T. mimeo . Acemoglu, Daron and James A. Robinson. 2006. Economic Origins of Dictatorship and Democracy. New York: Cambridge University Press. Babst, Dean V. 1964. Elective Governments: A Force for Peace. The Wisconsin Sociologist 2:914. Baliga, Sandeep, David O. Lucca and Tomas Sjostrom. 2007. Domestic Political Survival and International Conict: Is Democracy Good for Peace? Northwestern University Mimeo . Besley, Timothy and Masayuki Kudamatsu. 2007. Making Autocracy Work. London School of Economics mimeo . Box-Steensmeier, Janet M., Dan Reiter and Christopher J. W. Zorn. 2003. Nonproportional Hazards and Event History Analysis in International Relations. Journal of Conict Resolution 47(1):3353. Bueno de Mesquita, Bruce. 2006a. Central Issues in the Study of International Conict. In Oxford Handbook of Political Economy, ed. Barry R. Weingast and Donald A. Wittman. New York: Oxford University Press pp. 831851. Bueno de Mesquita, Bruce. 2006b. Game Theory, Political Economy, and the Evolving Study of War and Peace. American Political Science Review 100(4):637642. Bueno de Mesquita, Bruce, James D. Morrow, Randolph M. Siverson and Alastair Smith. 1999. An Institutional Explanation of the Democratic Peace. American Political Science Review 93(4):791807. Bueno de Mesquita, Bruce and James Lee Ray. 2004. The National Interest Versus Individual Political Ambition: Democracy, Autocracy, and the Reciprocation of Force and Violence in Militarized Interstate Disputes. In The Scourge of War: New Extensions of an Old Problem, ed. Paul F. Diehl. Ann Arbor: University of Michigan Press pp. 94119. Bueno de Mesquita, Bruce and Randolph M. Siverson. 1995. War and the Survival of Political Leaders: A Comparative Study of Regime Types and Political Accountability. American Political Science Review 89(4):841855. Bueno de Mesquita, Bruce, Randolph M. Siverson and Gary Woller. 1992. War and the fate of regimes: A comparative analysis. American Political Science Review 86(3):638646. 25
Cheibub, Jos Antonio and Jennifer Gandhi. 2004. Classifying Political Regimes: e A Sixfold Classication of Democracies and Dictatorships. Paper presented at the Annual Meeting of the American Political Science Association. Chiozza, Giacomo and H. E. Goemans. 2003. Peace through Insecurity: Tenure and International Conict. Journal of Conict Resolution 47(4):443467. Chiozza, Giacomo and H. E. Goemans. 2004. International Conict and the Tenure of Leaders: Is War Still Ex Post Inecient? American Journal of Political Science 48(3):604619. Debs, Alexandre. 2007a. Media and Divide-and-Rule. University of Rochester mimeo . Debs, Alexandre. 2007b. Political Strength and Economic Eciency in a MultiAgent State. University of Rochester mimeo . Debs, Alexandre. 2007c. The Wheel of Fortune: Agency Problems in Dictatorships. University of Rochester mimeo . Dixon, William J. 1994. Democracy and the Peaceful Settlement of International Conict. American Political Science Review 88(1):1432. Doyle, Michael W. 1983. Kant, Liberal Legacies, and Foreign Aairs. Philosophy and Public Aairs 12(3):205235. Part I. Egorov, Georgy and Konstantin Sonin. 2006. Dictators and their Viziers: Endogenizing the Loyalty-Competence Trade-o. Available at SSRN: http://ssrn.com/abstract=630503 . Fearon, James D. 1995. Rationalist Explanations for War. International Organization 49(3):379414. Filson, Darren and Suzanne Werner. 2004. Bargaining and Fighting: The Impact of Regime Type on War Onset, Duration, and Outcomes. American Journal of Political Science 48(2):296313. Filson, Darren and Suzanne Werner. 2007. Sensitivity to Costs of Fighting versus Sensitivity to Losing the Conict: Implications for War Onset, Duration, and Outcomes. Journal of Conict Resolution 51(5):691714. Gandhi, Jennifer and Adam Przeworski. 2006. Cooperation, Cooptation, and Rebellion under Dictatorships. Economics and Politics 18(1):126. Geddes, Barbara. 1999. What Do We Know About Democratization After Twenty Years? Annual Review of Political Science 2:115144.
26
Geddes, Barbara. 2003. Paradigms and Sand Castles. Theory Building and Research Design in Comparative Politics. Ann Arbor: The University of Michigan Press. Goemans, H. E. 2000. War and Punishment: The causes of war termination and the First World War. Princeton, NJ: Princeton University Press. Goemans, H. E., Kristian Skrede Gleditsch and Giacomo Chiozza. 2005. Archigos, a data set of leaders 18752004. University of Rochester and Essex University, UK. Gowa, Joanne. 1998. Politics at waters edge: Parties, voters, and the use of force abroad. International Organization 52(2):307324. Huntington, Samuel P. 1991. The Third Wave: Democratization in the Late Twentieth Century. London: University of Oklahoma Press. Jackson, Matthew O. and Massimo Morelli. 2007. Political Bias and War. American Economic Review 97(4):13531373. Kant, Immanuel. 2006. Toward Perpetual Peace: A Philosophical Sketch. In Toward Perpetual Peace and Other Writings on Politics, Peace, and History, ed. Pauline Kleingeld. New Have, CT: Yale University Press pp. 67109. Translated by David L. Colclasure, with essays by Jeremy Waldron, Michael W. Doyle and Allen W. Wood. Kuran, Timur. 1991. Now Out of Never: The Element of Surprise in the East European Revolution of 1989. World Politics 44(1):748. Lai, Brian and Dan Slater. 2006. Institutions of the Oensive: Domestic Sources of Dispute Initiation in Authoritarian Regimes. American Journal of Political Science 50(1):113126. Levy, Gilat and Ronny Razin. 2004. It Takes Two: An Explanation for the Democratic Peace. Journal of the European Economic Association 2(1):129. Levy, Jack S. 1988. Domestic Politics and War. Journal of Interdisciplinary History 18(4):653673. Maoz, Zeev and Bruce Russett. 1993. Normative and Structural Causes of Democatic Peace, 1946-1986. American Political Science Review 87(3):624638. Myerson, Roger. 2008. The Autocrats Credibility Problem and Foundations of the Constitutional State. American Political Science Review 102(1):125139. Padro i Miquel, Gerard. 2007. The Control of Politicians in Divided Societies: The Politics of Fear. Review of Economic Studies 74(4):12591274.
27
Patty, John and Roberto A. Weber. 2006. Agreeing to Fight: An Explanation of the Democratic Peace. Politics, Philosophy & Economics 5(3):305320. Peceny, M., Caroline C. Beer and S. Sanchez-Terry. 2002. Dictatorial Peace? American Political Science Review 96(1):1526. Peceny, Mark and Caroline C. Beer. 2003. Peaceful Parties and Puzzling Personalists. American Political Science Review 97(2):339342. Popper, Karl R. 1963. The Open Society and Its Enemies. Vol. I. The Spell of Plato. New York: Harper & Row Publishers. Powell, Robert. 2006. War as a Commitment Problem. International Organization 60(1):169203. Przeworski, Adam, Michael E. Alvarez, Jos Antonio Cheibub and Fernando e Limongi. 2000. Democracy and Development: Political Institutions and WellBeing in the World, 1950-1990. Cambridge, UK: Cambridge University Press. Ray, James Lee. 1995. Democracy and international conict: An evaluation of the democratic peace proposition. Columbia, SC: University of South Carolina Press. Ray, James Lee. 1998. Does Democracy Cause Peace? Annual Review of Political Science 1:2746. Ray, James Lee. 2003. A Lakatosian View of the Democratic Peace Research Program. In Progress in International Relations Theory, ed. Colin Elman and Miriam Fendius Elman. Cambridge, MA: MIT Press pp. 205243. Reiter, Dan and Allan C. Stam. 2002. Democracies at War. Princeton, NJ: Princeton University Press. Riker, William H. 1982. Liberalism Against Populism. Prospect Heights, Ill.: Waveland Press, Inc. Rosato, Sebastian. 2003. The Flawed Logic of Democratic Peace Theory. American Political Science Review 97(4):585602. Rousseau, David L., Christopher Gelpi, Dan Reiter and Paul K. Huth. 1996. Assessing the Dyadic Nature of the Democratic Peace, 1918-88. American Political Science Review 90(3):512533. Russett, Bruce. 1993. Grasping the Democratic Peace: Principles for a Post-Cold War World. Princeton, NJ: Princeton University Press. Russett, Bruce and Harvey Starr. 2000. From Democratic Peace to Kantian Peace. In Handbook of War Studies II, ed. Manus I. Midlarsky. Ann Arbor, MI: University of Michigan Press pp. 93128. 28
Russett, Bruce and John R. Oneal. 2001. Triangulating peace: Democracy, interdependence, and international organizations. New York: Norton. Schumpeter, Joseph A. 1950. Capitalism, Socialism, and Democracy, Third Edition. New York: Harper & Brothers Publishers. Signorino, Curtis S. 2002. Strategy and Selection in International Relations. International Interactions 28(1):93115. Smith, Alastair. 1998. International Crises and Domestic Politics. American Political Science Review 92(3):623638. Svolik, Milan. 2006. A Theory of Government Dyanmics in Authoritarian Regimes. University of Illinois at Urbana-Champaign mimeo . Weart, Spencer. 1994. Peace Among Democratic and Oligarchic Republics. Journal of Peace Research 31(3):299316. Weeks, Jessica. 2007. Accountability, Non-Democracies and International Conict: Four Empirical Tests. Stanford University Mimeo .
29
Figure 1: Net Benet of Settlement as a Function of As Share of the Pie: Peace Between Unitary States
Ii (ti ) Oi (ti ) = 0 1 vl(A) = vl(B) = 1 , p = 1 , cL(A) = cL(B) = 2 . 3 2
30
Figure 2: Net Benet of Settlement as a Function of As Share of the Pie: Peace Between Democracies
1 1 G(zi , I w ) = zi2 100 I w , (ti ) = 10 , [Ii (ti ) Oi (ti )] = 15 1 vL(A) = vL(B) = 1 , p = 1 , cL(A) = cL(B) = 2 3 2
31
Figure 3: Net Benet of Settlement as a Function of As Share of the Pie: War Between a Democracy and an Autocracy
1 1 G(zi , I w ) = zi2 100 I w , (tA ) = 1 , [IA (tA ) OA (tA )] = 50 (tB ) = 10 , 5 1 1 1 [IB (tB ) OB (tB )] = 15 vL(A) = vL(B) = 3 , p = 2 , cL(A) = cL(B) = 2
32
Table 1: Conict Outcomes, Regime Types and the Loss of Oce Tenure b Std.Err 0.248 0.006 . . . 0.755 0.008 0.233 0.216 0.191 0.018 0.893 0.379 0.030 0.626 0.002 0.524 0.244 0.663 2.407 0.839 0.622 .276 .042 . . . .226 .144 .267 .265 .361 .272 .356 .932 .132 2.03 .273 .560 .560 .538 1.15 .361 .634
Variables Democracy Democracy ln(t) . . . Challenger Target Inheritor Democracy Crisis Victory Democracy Crisis Defeat Democracy Crisis Draw Non-Dem. Crisis Victory Non-Dem. Crisis Defeat Non-Dem. C. Defeat ln(t) Democracy Crisis Draw Democracy C. Draw ln(t) Democracy War Victory Democracy War Defeat Democracy War Draw Non-Democracy War Victory Non-Democracy War Defeat Non-Democracy War Draw No. Obs No. Subjects No. Failures Log-likelihood Wald-test
9336 1831 1667 9684.5 497.8 p < 0.001 0.419
p < .01, p < .05, p < .1. The frailty parameter measures the variance of a Gamma distribution with mean equal to 1. Co-variates for Civil war, GDP per capita, GDP growth, Trade Openness, the Change in Trade Openness, Population, Age at entry, number of Times in Oce, and Entry have been omitted for convenience. A full table is available on the authors website.
33
Table 2: Regime Type and The Fate of Leaders OK Non-Democracy Democracy Total 610 (59%) 711 (93%) 1,321 (73%) Pearson Exile 237 (23%) 22 (3%) 259 (14%) Chi2 (3) Jail 118 (11%) 20 (3%) 138 (8%) Killed 74 (7%) 10 (1%) 84 (5%) TOTAL 1,039 (58%) 763 (42%) 1,802 (100%)
= 268.6 p < 0.001
34
Table 3: Conict Outcomes, Autocratic Types and the Loss of Oce Tenure b Std.Err 0.405 1.890 0.147 . . . 1.552 0.030 0.012 0.506 0.824 0.842 8.190 2.059 0.219 0.616 0.826 1.003 2.583 2.463 1.585 2.413 0.991 0.186 1.233 2.871 0.326 4120 637 488 2541.96 147.8 0.247 .132 1.538 .194 . . . .438 .270 .447 .551 .594 .599 5.824 .781 .895 .657 .484 .559 2.252 .748 1.884 2.315 1.445 1.472 3.372 1.038 1.535
Variables Civilian Monarch Monarch ln(t) . . . Initiator Defender Inheritor Military Crisis Victory Military Crisis Defeat Military Crisis Draw Military War Victory Military War Defeat Military War Draw Civilian Crisis Victory Civilian Crisis Defeat Civilian Crisis Draw Civilian War Victory Civilian War Defeat Civilian War Draw Monarch Crisis Victory Monarch Crisis Defeat Monarch Crisis Draw Monarch War Victory Monarch War Defeat Monarch War Draw No. Obs No. Subjects No. res Log-likelihood Wald-test
p < 0.001 .085
p < .01, p < .05, p < .1. The frailty parameter measures the variance of a Gamma distribution with mean equal to 1. Co-variates for Civil war, GDP per capita, GDP growth, Population, Age at entry, and Entry have been omitted for convenience. A full table is available on the authors website.
35
Table 4: The Fate of 3 Types of Non-Democratic Leaders OK Monarch 6 (26%) Military 88 (38%) Civilian 112 (55%) Total 206 (45%) Exile Jail Killed TOTAL 23 (5%) 234 (51%) 205 (44%)
13 (57%) 1 (4%) 3 (13%) 71 (31%) 47 (20%) 27 (12%) 45 (22%) 33 (16%) 15 (7%)
130 (28%) 81 (18%) 45 (10%) 462 (100%)
Pearson Chi2 (6) = 24.57 p < 0.001
36
Table 5: Non-Democratic Regime Types and War War Std.Err .176 .271 .135 .192 .018 1.129 .067 3.083 .132 .038 .187 .400 .227 .291
Variables Civilian Monarch Entry Civil War GDP per capita GDP Growth Population CINC Military Mobilization Number of Borders Time since Previous Onset Time since Previous Onset - squared Time since Previous Onset - cubed Constant No. Obs. Log-likelihood Pseudo R-square Wald-test
b
0.503 0.148 0.021 0.232 0.006 2.415 0.171 5.038 0.572 0.065 0.677 1.261 0.656 2.392
3992 141.34 0.153 104.7 p < 0.001
Coecients of each regime type must be interpreted relative to the excluded category, Military Autocrats. Standard errors are clustered on country-code.
37