Hawks,Doves,and International Cooperation

Past Research on Ideology and Foreign Policy

A growing body of literature has shown that party ideology has a strong influence on the hawkish or dovish nature of a government’s foreign policy. ² One common theme in all of these studies is that parties in office are constrained by the preferences of their electoral base (Palmer, London, and Regan 2004; Koch 2009). Supporters of left-wing parties are more dovish and willing to punish leaders that take a belligerent stance, whereas supporters of right-wing parties will reward more conflictual policies and punish conciliatory behavior. ³ Since deviations from these policies can be politically costly, it is expected that a government’s foreign policy will be consistent with these preferences. That is, governments on the left will be more conciliatory and those on the right will be more conflictual. ⁴

Palmer, London, and Regan (2004) and Arena and Palmer (2009) show that governments on the right are more likely than those on the left to become involved in international conflicts. Koch and Fulton’s (2011) results are also consistent with this expectation, while Koch (2009) and Koch and Sullivan (2010) confirm it for the duration of disputes as well. The hawkish base for right-wing governments insulates them from pressure to withdraw from international disputes. Conversely, the dovish base of left-leaning governments increases pressure to end conflicts, leading them to seek a quick exit from a dispute even if it means accepting a less favorable settlement (Koch and Sullivan 2010). Clare (2010) showed that even minor coalition parties that are ideologically distant from the rest of the government can move the government toward more hawkish or dovish policies, depending on whether they are positioned to the right or left of the government, respectively.

This general argument can be contrasted to some formal theoretic research that expects hawkish leaders to be best situated to negotiate peace. Cowen and Sutter (1998) and Cukierman and Tommasi (1998) both argue that hawkish leaders have greater credibility to sell peace to the domestic public. Since leaders with hawkish reputations, by definition, prefer a hard-line stance against opponents, they can credibly convey the wisdom of accommodation as the best policy. Doves, on the other hand, have a harder time making the case for accommodation because this is seen as their default position anyway.

Schultz (2005) also concludes that hawks are more likely to deliver peace. He argues that left- and right-wing parties contain both extreme and moderate elements. Since the public, in his model, wants to retain moderate leaders, a cooperative action can have different electoral effects, depending on its source. If a leader from a right-wing party makes a cooperative move, it is a sign of her or his moderation. However, the same action by a left-wing leader indicates that she or he is at the extreme dovish end of the party. A cooperative move by the dovish leader is therefore likely to lead to an electoral loss, regardless of whether or not it is reciprocated, whereas the hawkish leader can survive even in the event of unreciprocated accommodation. Consequently, as Schultz argues, a hawk is more likely to deliver peace because she or he is more likely to clear the electoral hurdle and remain in office when his or her cooperative actions take effect. ⁵

Schultz’s (2005) model addresses a puzzle from the previous research; it can explain why hawks would cooperate in the first place by showing that they may gain electorally by taking moderate steps. Yet it is possible to question the electoral motivations of the public in this model. In Schultz’s formal stylization, the public votes after an initial round of cooperation and they will replace a dovish leader, even if a cooperative move is reciprocated, out of mistrust that she or he will be taken advantage of in future negotiations. It is not clear, however, why the public would not retain a dovish leader that extracts concessions from an opponent, but at the same time reelect a hawkish leader who does not elicit reciprocal cooperation.

My argument is that the public will electorally reward or punish an incumbent depending on how successfully she or he can extract concessions from a foreign opponent. This holds regardless of the public’s initial expectations, shaped by the government’s ideological profile, on how it is likely to behave. In this case, the research focus shifts toward an international opponent and whether it will behave differently at the bargaining table depending on who it anticipates to face in future negotiations. Yet the cited studies leave unexplored an international opponent’s preferences for one leader over another and how these preferences influence its willingness to reciprocate. Schultz does show that a hard-line opponent will cooperate with a dovish actor (though not a hawkish one) in the current period in order to take advantage of its accommodation at later times. In this study, I argue that the motivation behind concessions can be quite different: these are made to avoid a tougher opponent in the future rather than to outmaneuver the current one. Smith (2009) is a rare study that takes an international opponent’s preferences into account. He shows that leaders will cooperate with a government in an opposing state that has similar preferences to their own but withdraw this cooperation if a more hostile government comes into power. Although Smith does not connect these preferences to the dovish or hawkish nature of the government in the opposing state, as done here, his model is useful for showing that the preferences over a leader in an opposing state influence cooperation.

In the next section, I develop a theoretical argument that departs from the studies mentioned previously in two ways. First, it moves beyond the previous empirical studies to show how the ideological orientation of the government and its domestic opposition is relevant for understanding not only its own behavior but also its adversary’s willingness to negotiate and make concessions as well. Second, contrary to some previous expectations, I show that doves do not negotiate from a position of weakness, but rather, for domestic reasons, they have greater bargaining leverage than do hawks to extract concessions from an adversary.

Theory and Hypotheses

I argue that the previous expectation that hawks can best deliver peace does not necessarily hold once we take into account an international opponent’s preferences to negotiate with one type of leader over another. My argument is based on four assumptions. First, I share the premise, commonly adopted in the domestic “audience cost” literature, that leaders are motivated by their desire to maintain office (Schultz 2005).

Second, a leader’s tenure depends on the support of two critical groups. On one hand, a leader must maintain the sustained support from his or her electoral base, which only supports a party or a leader that pursues its favored policies (Palmer, London, and Regan 2004; Koch 2009; Koch and Sullivan 2010; Nincic and Ramos 2010). Thus, leaders must pursue the policies of their base, even if they are motivated primarily by office retention (Laver and Schofield 1990; Strom 1990). On the other hand, leaders rarely retain office by serving only the narrow interests of an ideological base. Instead, they must also satisfy the more moderate median voter. ⁶ Two characteristics of these moderate voters are essential: (a) their less ideological nature and (b) the ability to switch support from one party or coalition to another (Schultz 2005). Unlike the base, the median voter is not concerned with the policy that is adopted per se, but instead with its success. Failed policies can lead this crucial middle to switch its support from the government to the opposition, costing a leader her or his tenure. ⁷

Third, unreciprocated cooperation is generally seen as a foreign policy failure that could lead moderates to switch their support from one party to another (Colaresi 2004). Unlike previous studies, I do not assume that making a cooperative move toward an opponent is itself costly. Rather, cooperation is only costly if a leader is unable to extract a reciprocal cooperative move from an opponent. For example, Neville Chamberlian’s appeasement policy, which gave Germany control over the Sudetenland in return for Hitler’s assurance that he would not seek any further territorial changes, was initially approved by an “overwhelming majority of the ordinary people” (Taylor 1965, 430). Public opinion turned against Chamberlain only when it became clear that Hitler would not keep up his end of the agreement. As Kissinger (1994, 316) succinctly explained, “The democratic public is unforgiving in the face of debacles, even when these result from carrying out its own immediate wishes. Chamberlain’s reputation collapsed once it became clear that he had not achieved ‘peace for our time.’”

Fourth, conflicts are costly. This does not mean that states would prefer to compromise on any terms rather than continuing hostile relations. In fact, the third assumption above implies quite the opposite—that cooperating with an intransigent opponent is costly—and we should expect that states will continue to seek the best bargain possible. Rather, this assumption simply suggests that there is some type of mutual cooperation that would make both states better off as opposed to continued stalemate or escalation. In the absence of this assumption, there would be no reason for states to make concessions.

All four premises are important for understanding cooperative behavior, and they lay the ground for developing theoretical expectations about a leader’s decision to make a cooperative move and an opponent’s decision to reciprocate in kind.

As for initial cooperative moves, it is straightforward to see that doves should be more likely than hawks to initiate cooperation. The key lies in the policies that are supported by each leader’s electoral base (assumption 2). A dovish leader’s base should support more cooperative policies, while generally refraining from belligerent or intransigent positions. The reverse holds for a hawkish leader’s base (Nincic and Ramos 2010).

Hypothesis 1: Dovish governments are more likely to initiate cooperation, while hawkish leaders are more likely to initiate conflict.

While this Hypothesis is intuitive, it departs from Nincic’s (1988) and Schultz’s (2005) “politics of opposites.” The authors predict that hawks can get a boost from moderate voters by appearing moderate, and doves from appearing more hawkish (Gadarian 2010). While this is plausible, it seems unlikely that hawkish or dovish leaders would be willing to make significant and sustained departures from the preferences of their base. Instead, the Hypothesis is more consistent with the conventional wisdom from the previous research that doves are more likely than hawks to initiate cooperative moves (Palmer, London, and Regan 2004; Koch 2009; Arena and Palmer 2009; Koch and Sullivan 2010).

While the first Hypothesis was premised on a leader’s need to maintain support from her or his electoral base, understanding whether a leader can extract concessions from an opponent requires the consideration of two additional groups: the median voter and domestic opposition parties. Both of these groups are necessary to understand an international opponent’s preferences for cooperating with a hawkish or dovish leader.

To illustrate the argument, Figure 1 shows three ways that the government and opposition can be situated on an ideological space. In the first panel, the two main parties (or coalitions) are situated on the left and right of the political spectrum, respectively, with the median voter in the center. The second and third panels similarly show the relative positions of the government and opposition, though panel 2 shows a dovish government and opposition, while panel 3 illustrates the case of a relatively hawkish government and opposition. In both cases, the median voter once again divides the space between the two parties.

OPEN IN VIEWER

Figure 1.

Illustrative example of the relationship between the government and its opposition.

The first scenario in Figure 1 shows two types of leaders—a dovish leader (if party 1 controls the government) and a hawkish one (if party 2 controls the government). In this case, we should expect state B’s leader to be more likely to reciprocate cooperation when it is offered by a dovish leader in state A who faces a hawkish domestic opposition, and less likely in the reverse—when a hawkish leader in state A faces a dovish opposition. The rationale behind this expectation is twofold. First, state B’s leader should prefer a dovish rather than a hawkish leader in the opposing state. Doves, by definition, are seen as more receptive to cooperation and are willing to make concessions in negotiations. Hawkish leaders, on the other hand, are generally more intransigent (otherwise, they would not be seen as “hawks”), offering little in negotiations while demanding large concessions. Second, it is plausible to expect B’s leader to know that failing to match a dove’s cooperative move would be seen as a policy failure in state A, with the consequence of weakening the dove’s future electoral position against the hawkish opposition. ⁸ In this case, the median voter, whose support is also necessary for a leader to retain office, plays a critical role. The dove’s inability to deliver a successful foreign policy outcome would lead the median voter to switch from supporting a dovish leader to a hawkish one.

Given state B’s preference for dovish over hawkish leaders in state A, and also the domestic consequences for failing to reciprocate a dovish leader’s cooperation, it is possible to identify two responses to an initial cooperative move. If A’s current leader is a dove, B is better off cooperating in the present period since the alternative is to face an intransigent hawkish opponent in the future. This ensures that cooperative gains are made in the current period and also helps preserve the dove in office for future negotiations. Alternatively, if facing a hawkish leader in state A, state B has an incentive to forego cooperation for the sake of anticipated gains from cooperation with a more dovish leader in the future. ⁹

Hypothesis 2a: Dovish leaders are more likely to have a cooperative move reciprocated when they face a more hawkish domestic opposition.

The Israeli–Palestinian negotiations before the Oslo Accords illustrate these Hypotheses. This case, illustrated earlier, shows that the Palestinians were less willing to compromise for two reasons. First, the prospect of facing a dovish Labor government in the future made the current Likud government a less attractive negotiating partner. Second, any concessions given to Likud could be portrayed as a “resounding victory” that could help it in the elections. A similar pattern was evident in the 1977 negotiations over the Panama Canal Treaty. Facing domestic trouble over the treaty, Omar Torrijos, the Panamanian president, tried to extract last-minute concessions from President Carter. Carter’s response suggestively pointed to the hawkish opposition in the Senate and among presidential candidates like Ronald Reagan. He also “warned that a new congressional election process was about to begin, a period when Congress would be even less inclined to accept a new treaty than it already was” (Habeeb 1988, 138). The possibility of a more hawkish Senate in the future prevented Torrijos from pressing the issue. German–Soviet relations before the 1987 elections in West Germany are similarly informative. Members of the ruling Christian Democratic Party were unsettled that Mikhail Gorbachev was receptive to meet with high-profile members of the opposition Social Democratic Party before the elections (Garton Ash 1993). The fear was that this would give the SPD a boost in the elections, which was possibly Gorbachev’s intention all along. As Garton Ash (1993, 105-106) observes, “there are some indications that Moscow did drag its feet in relations with the centre-right government in the hope of having more congenial partners…in Bonn after the elections.”

Note, however, that this does not imply that dovish governments are always more likely to have their cooperation reciprocated. The remaining two scenarios in Figure 1 show why this might not always be the case. In scenario 2, both the government and the opposition are dovish and would thus seek compromise with state B. In this case, B does not have a strong preference for dealing with one government over the other. Failing to reciprocate a dove’s cooperation would only lead the median voter to switch its support to an equally dovish leader that would also take a compromising position. In fact, it can take an intransigent stance and push for more concessions from A’s dovish leader. In the absence of an electorally viable hawkish opposition, a dovish leader loses the bargaining leverage that the “hawks waiting in the wings” provide.

The predictions are less straightforward when the government and the opposition are equally hawkish (scenario 3). On one hand, hawks are less likely to make cooperative moves in the first place (Hypothesis 1). On the other hand, if cooperation is initiated by a hawk, then state B has no incentive to buy time in anticipation of a preferred dovish government. Failing to reciprocate the current leader’s cooperation would only cause the median voter to switch its support to a similarly hawkish leader who may or may not take a compromising approach. Consequently, state B may reciprocate A’s hawkish leader’s cooperation in anticipation that the alternative hawkish leader would take a harder-line once in office. ¹⁰ Compared to the case of a dovish leader and opposition (scenario 2), state B should be more likely to reciprocate a cooperative move when state A’s leader and opposition are on the hawkish end of the political spectrum. The discussion of such a homogenous party system (hawkish or dovish) highlights the importance of divergent policy stands between the government and its opposition if domestic politics are to provide bargaining leverage in international negotiations. ¹¹

Hypothesis 3a: Dovish leaders are less likely to have their cooperative moves reciprocated in the absence of a viable hawkish domestic opposition.

The present study therefore departs from the argument that hawks are best positioned to bargain with international opponents (Schultz 2005) and also from studies suggesting that doves will seek even unfavorable compromise solutions (Koch and Sullivan 2010). Instead, I expect dovish leaders to be more effective than are hawkish ones in extracting reciprocal concessions from an opponent. If an international opponent has a preference to negotiate with a dovish rather than a hawkish leader, then it has a vested interest in reciprocating a dovish leader’s cooperation, since the failure to do so could lead to the ascent of a more hawkish counterpart in the future. This expectation depends on the presence of a hawkish domestic opposition which, despite its critical importance, has not been analyzed in the extant research.

Research Design

I test the Hypotheses on a sample of territorial disputes during the post–World War II period. Territorial disputes have traditionally been considered the most intractable and adversarial disputes (Huth 1996; Huth and Allee 2002; Senese and Vasquez 2008). ¹² Disputed territory is often vitally important for its strategic, ethnic, or economic value. For this reason, territorial disputes are generally seen as an indivisible issue, making them less amenable to compromise (Senese and Vasquez 2008). In addition, the importance of disputed territory often gives rise to hard-line positions in the government and among domestic constituencies (Senese and Vasquez 2008), making it easier to mobilize public support for militarized action (Huth and Allee 2002). Conciliatory moves, on the other hand, can make the government vulnerable to charges of appeasement over issues of vital interests. All of this means that violence, rather than compromise, is common in territorial disputes because the “expected utilities for disputing territorial claims and escalating territorial disputes are higher than the expected utilities for making concessions or accepting the status quo” (Huth and Allee 2002, 31). A number of studies provide evidence for this claim, showing that territorial disputes are more likely than are disputes over policy issues to lead to militarized crises and their subsequent escalation. ¹³

Given the seeming intractability of territorial disputes that makes them more prone to violence than other types of conflicts, it is critically important to understand the conditions under which states are willing to resort to mutual cooperation rather than fighting in such volatile situations. Moreover, because territorial issues are seen as vitally important national interests in which domestic politics is traditionally expected to be bipartisan, they should provide an especially difficult test for a theory, such as the one developed here, that highlights the importance of domestic ideological differences for negotiating peace with an adversary (Huth and Allee 2002, 31-32). All of this suggests that any findings about conciliation or compromise in these most difficult cases may be generalized to other, less conflictual, disputes as well. The cases of territorial rivals are drawn from Huth and Allee’s (2002) data set. Despite variable data availability of the different data sets used for measuring the principle variables, there is still a sizable sample of thirty territorial rivalry dyads for the 1945–1995 period.

Independent Variables

Testing the Hypotheses requires measuring the hawkish or dovish ideological nature of the government and its domestic opposition. For this purpose, I use data from the comparative manifestos project (CMP; Budge et al. 2001). They are based on a content analysis of the election manifestos of political parties in twenty-five democracies over the period 1945 through 2006. To identify the parties inside and outside of government, I use the ParlGov database (Döring and Manow 2010). Based on these data sets, I constructed the following measures:

1. Government ideology. To measure the hawkish or dovish outlook of the government, I used the standard measure created by Laver and Budge (1993). Laver and Budge identified thirteen left-wing and thirteen right-wing issues in the manifesto data that depict a party’s ideological position. Using the percentage of statements in their manifestos devoted to each of these issues, parties can be placed on a left-right ideological continuum as follows:

P a r t y i d e o l o g y = ( R i g h t − L e f t ) / ( R i g h t + L e f t ) ,

where “Right” is the percentage of manifesto statements falling under the right-wing categories and “Left” is the percentage of statements falling under the left-wing categories. The resulting variable ranges from −1 (left-wing party) to +1 (right-wing party). Values near zero indicate a moderate ideological position on these issues. This variable is then used to create a measure of government ideology. Consistent with previous studies (Palmer, London, and Regan 2004), I use a weighted measure of government ideology that accounts for each party’s share of seats in the government. It is operationalized as follows:

G o v e r n m e n t i d e o l o g y = [ ( I D A × S e a t s A ) + . . . ( I D N × S e a t s N ) ] / T o t a l S e a t s ,

where “ID” is each party’s ideological position, “Seats” is the number of governmental seats held by each party, and “Total Seats” is the number of seats held by all parties in government. For the United States, the only presidential system in the sample, government ideology is coded as the ideology of the party holding the presidency. ¹⁴ The resulting variable ranges from −1 (left wing) to +1 (right wing). To make this measure consistent with the ideological space depicted in Figure 1, this variable was converted to range from −100 to +100. ¹⁵

This is consistent with previous studies that use overall ideology scores to measure the government’s ideology (Palmer, London, and Regan 2004; Koch 2009). I also estimate models that measure ideology based only on positive or negative statements about internationalism and the role of the military. The results are reported in Tables 1 and 2.

OPEN IN VIEWER

Table 1.

Government Ideology and the Likelihood of an Initial Cooperative or Conflictual Move by State A, Multinomial Logit Estimates.

	Model 1: overall ideology scores		Model 2: international components only
	Status quo vs. compromise	Conflict vs. compromise	Status quo vs. compromise	Conflict vs. compromise
Government ideology	.005 (.002)***	.019 (.014)*	.002 (.002)	.014 (.009)**
Ideological distance: government outlier parties	.002 (.001)**	.030 (.007)***	−.001 (.002)	.005 (.003)**
Majority government	−.068 (.255)	−2.179 (.971)**	−.115 (.249)	−1.903 (.673)**
Coalition government	−.473 (.298)*	−.355 (1.089)	−.380 (.300)	1.466 (1.935)
Effective number of parties	.298 (.165)**	1.277 (.253)***	.329 (.174)**	1.166 (.351)***
Relative capabilities	.560 (.273)**	3.341 (1.367)***	.570 (.270)**	1.532 (.873)**
Joint democracy	.392 (.349)	.230 (.882)	.171 (.312)	−1.489 (1.075)*
Relative strategic value of territory	.184 (.123)*	−2.412 (1.667)*	.189 (.108)**	−2.398 (1.710)*
Relative economic value of territory	−.056 (.189)	−.993 (1.042)	.179 (.143)	−.300 (.741)
Relative ethnic value of territory	.270 (.210)*	−.004 (.718)	.139 (.179)	−.009 (.563)
Years since last event	−.032 (.019)**	.079 (.027)***	−.030 (.021)*	.081 (.022)***
Years since last event—squared	.001 (.000)**	−.001 (.000)***	.000 (.000)**	−.001 (.000)***
Years since last event—cubed	−.000 (.000)**	.000 (.000)***	−.000 (.000)**	.000 (.000)***
Constant	3.018 (.484)***	−6.613 (1.818)***	3.146 (.491)***	−5.128 (1.571)***
N	4,700		4,345
Wald χ 2	3,532.15***		5.0e+05***

Note: Numbers in parentheses are robust standard errors clustered on the challenger (state A).

*p < .10. **p < .05. ***p < .01 (one-tailed tests).

OPEN IN VIEWER

Table 2.

Dovish or Hawkish Domestic Opposition in State A and the Likelihood of Reciprocation at the Same or Higher Level by State B, Probit Estimates.

	Model 1: government ideology only	Model 2: largest outside party	Model 3: largest outside party (international components only)	Model 4: previous government	Model 5: all outside parties
Government ideology	−.006 (.004)*	.012 (.006)**	.004 (.007)	.001 (.006)	.004 (.004)
Opposition’s ideological distance from government	–	.017 (.004)***	.011 (.005)**	.005 (.003)**	.009 (.002)***
Ideological distance: government outlier parties	−.004 (.003)*	−.006 (.003)**	−.006 (.003)**	−.004 (.003)	−.006 (.003)**
Majority government	.212 (.407)	−.575 (.411)*	−.259 (.355)	−.408 (.409)	−.609 (.398)*
Coalition government	1.244 (.292)***	1.258 (.322)***	1.115 (.340)***	1.434 (.321)***	1.365 (.341)***
Effective number of parties	−.983 (.199)***	−1.046 (.186)***	−1.071 (.257)***	−1.060 (.180)***	−1.112 (.203)***
Relative capabilities	.984 (.482)**	1.079 (.534)**	.897 (.607)*	.936 (.534)**	1.079 (.498)**
Joint democracy	1.138 (.515)**	1.045 (.551)**	1.374 (.611)**	1.467 (.584)***	1.025 (.543)**
Relative strategic value of territory	−1.021 (.413)***	−1.156 (.407)***	−1.026 (.518)**	−1.549 (.558)***	−1.189 (.427)***
Relative economic value of territory	.779 (.324)***	.862 (.257)***	.903 (.280)***	.819 (.276)***	.963 (.245)***
Relative ethnic value of territory	−.584 (.324)**	−.637 (.331)**	−.578 (.373)*	−.556 (.326)**	−.733 (.321)***
Constant	−1.842 (.657)***	−1.492 (.716)**	−1.836 (.732)***	−1.975 (.676)***	−1.480 (.729)***
N	150	150	140	138	150
Wald χ 2	854.17***	2,233.09***	632.51***	3,010.49***	545.59***

Note: Numbers in parentheses are robust standard errors clustered on the challenger (state A).

*p < .10. **p < .05. ***p < .01 (one-tailed tests).

2. Opposition’s ideological distance from the government. Two aspects of this measure are critical for a valid test of the Hypotheses. First, since I argue that international opponents bargain in anticipation of who they may face in the future, it is crucial to identify the party or parties that could plausibly be seen as a potential replacement to the current government. Second, the measure needs to capture not only the ideological position of the potential future government, but also its ideological distance from the current leadership.

To ensure a rigorous test, I use two methods of identifying the domestic opposition. First, I identify the largest opposition party in parliament. Since this party would likely be called upon to form a new government if the other fails a confidence vote or is otherwise voted out of office, it constitutes a valid indicator of the domestic “opposition.” After identifying this party, I created the variable for the Opposition’s Ideological Distance from the Government by subtracting the government’s ideological position on the −100 to +100 ideology scale from the largest outside party’s ideology on this same scale. A value of zero indicates no ideological difference between the government and the opposition; positive values indicate the opposition’s right-wing orientation compared to the government, and negative values indicate a left-wing opposition relative to the current government.

For an alternative, second, measure I look at the party composition of prior governments. Warwick (1996), for example, shows that participation in a previous government is a strong predictor of whether a party will serve in future governments. This measure thus takes into account the ideological position of all parties that could potentially replace the government and not just the currently largest one. It also ensures that the opposition’s ideological placement is not influenced by parties that have never served in the government. ¹⁶ To create the variable, I identified the government within the past fifteen years that was most ideologically distant from the current one. I then subtracted the ideology score of the current government from that of the most ideologically distant prior government. The variable again ranges from −100 to +100, with larger values indicating a more ideologically distant opposition to the left or right. ¹⁷

Control Variables

I also include a set of variables that account for the domestic and international bargaining environment. ¹⁸ In terms of domestic controls, I include dichotomous measures of the government’s majority-minority and single-party versus coalitional status, both of which were found to influence foreign policy. Majority Government is coded 1 if the government controls a majority of seats in parliament and 0 for minority governments. Coalition Government is coded 1 for multiparty governments and 0 for single-party governments.

Research has also found other characteristics of coalitional governments have a strong influence on their foreign policy. For example, governments with a fewer number of veto players were found to be more conflictual (Palmer, London, and Regan 2004). In addition, ideological “outlier” parties in the government can push it toward more conflictual or cooperative actions, depending on their ideological position relative to the government (Clare 2010). I therefore include two additional domestic control variables in the analysis. Effective Number of Parties in the government is a continuous measure of the number of parties in the government, weighted by their share of seats in the parliament. Ideological Distance: Government Outlier Parties is a continuous variable created by subtracting the government’s weighted ideology score from the ideology score of the government party that is farthest from it. It can assume positive or negative values depending on whether the outlier party is farther to the right or left.

I also control for territory- or dyad-specific characteristics that could influence patterns of cooperation among disputants. First, disputes involving strategically or economically important territory (the Golan Heights and Spratley Islands, respectively) should be more intractable than those that do not (Huth 1996; Huth and Allee 2002). The same may be said if one or both states have ethnic ties to the territory (Northern Ireland). I therefore include three variables measuring the Relative Strategic, Economic, and Ethnic Value of the Territory. Each variable can take one of three values, depending on whether the territory is more important to the challenger (−1), has the same value for both states (0), or is more important to the opponent (+1). ¹⁹

Second, to control for relative power, I include a measure of Power Parity as follows (see Danilovic 2002):

P o w e r p a r i t y = 1 − a b s | ( P o w e r T a r g e t − P o w e r I n i t i a t o r ) / ( P o w e r T a r g e t + P o w e r I n i t i a t o r ) | .

The resulting continuous variable ranges from 0 (power disparity) to 1 (power parity). Military capabilities data are from the Correlates of War data set.

Third, to measure joint democracy, I use the variable suggested by Tucker (1998). It is operationalized as follows:

J o i n t d e m o c r a c y = { S Q R T [ ( P o l i t y 2 A + 11 ) × ( P o l i t y 2 B + 11 ) ] } / 21 .

The variable ranges from 0 (joint autocracy) to 1 (joint democracy). The regime type data are from the Polity IV data set (Marshall and Jaggers 2002).

Lastly, to control for temporal dependence, I include a cubic polynomial of the number of months since the last event (conflictual or cooperative) in the dyad (Carter and Signorino 2010). The models are estimated with robust standard errors clustered on the challenger to account for potential problems of spatial dependence (Beck and Tucker 1997). ²⁰

Empirical Results

In Tables 1 and 2, I report the results of a series of probit models estimating the initiation (Table 1) and reciprocation (Table 2) of cooperation. Both tables show the results of models using the overall ideology scores and those derived from party positions on relevant international dimensions only (statements for and against a greater role for the military and those that support or oppose internationalism). In Table 2, I also report the results using alternative measures of the domestic opposition (models 4 and 5).

The Initiation of Cooperation

Table 1 reports the coefficients from a multinomial logit model estimating the onset of cooperation compared to the status quo and conflict. ²¹ As predicted, the positive and statistically significant coefficients for the Government Ideology variable indicates that both the status quo and conflict are more likely than cooperation as the government moves from a dovish to a more hawkish orientation. This pattern holds regardless of whether I use each party’s overall ideology score (model 1) or only the international components of their manifestos (model 2). The results are consistent with previous studies showing right-wing governments to be more conflictual than left-wing governments (Palmer, London, and Regan 2004; Arena and Palmer 2009; Koch 2009; Koch and Sullivan 2010) and also provide support for Hypothesis 1.

To illustrate the results, Figure 2 shows the predicted probability of cooperative (panel A) and a conflictual (panel B) moves as a function of the government’s ideology. The bottom of each graph also shows the observed distribution of cases at each value of government ideology. The distribution of cases spans nearly the entire spectrum of government ideology, ranging from −87 (dovish governments) to +100 (hawkish governments). As predicted in Hypothesis 1, the left-side panel shows that the probability of cooperation is highest for dovish governments (pr ≈ .068) and steadily decreases as the government becomes more hawkish (pr ≈ .025). In addition, the right-side panel of Figure 2 shows the opposite relationship with respect to the probability of a conflictual move. A dovish government is unlikely to take conflictual actions against a rival, with the probability of conflict increasing for hawkish governments.

OPEN IN VIEWER

Figure 2.

Predicated probability of cooperative and conflictual actions as a function of the government’s ideology.

Reciprocation of Cooperation

The results for the models of the opponent’s reciprocal cooperation are shown in Table 2. ²² Once again the results are consistent regardless of whether I use the overall ideology scores (model 2) or only the international statements in the party manifestos (model 3) and despite different methods of identifying the “opposition” (models 2, 4, and 5). The estimates are therefore robust to variable operationalizations of the independent variables.

The results show little support for the argument advanced in previous studies that hawkish governments are better positioned to achieve cooperation. The coefficient for the Government Ideology variable in model 1 is not statistically significant. However, the results for the variable Opposition’s Ideological Distance from the Government are positive and statistically significant as predicted in Hypotheses 2a and 2b. This indicates that international opponents are more likely to reciprocate a leader’s cooperative move when this leader faces a more hawkish domestic opposition, but less likely to do so when her or his opposition is more dovish.

The results also support Hypotheses 3a and 3b about the probability of reciprocation for leaders who do not face an ideologically distant domestic opposition (when a dovish government faces a dovish opposition or a hawkish government faces a similarly hawkish opposition). Note that when the Opposition’s Ideological Distance is included in the models, the coefficient for the Government Ideology variable is interpreted conditionally as the influence of the government’s ideology when the opposition’s ideological distance equals zero. The positive and statistically significant coefficient for Government Opposition in model 2 means that reciprocation is unlikely for a dovish government facing a dovish opposition (Hypothesis 3a) and more likely when a hawkish government faces a hawkish opposition (Hypothesis 3b).

The findings concerning the probability of reciprocation to state A’s cooperative moves are illustrated in Figures 3 and 4. In Figure 3, I graph the predicted probability of reciprocation as a function of (1) the government’s dovish (left-wing) or hawkish (right-wing) ideology (panel A), and (2) the opposition’s dovish (left) or hawkish (right) ideological position relative to the government’s ideology (panel B). ²³ The bottom of each graph shows the observed distribution of cases at each value of the independent variables. In both cases, the distribution of cases spans a wide range of values: observations for the “government ideology” variable range from −87 (dovish governments) to +100 (hawkish governments), while the observations for the “opposition’s ideological distance from the government” range from −100 (an extremely dovish opposition) to +100 (an extremely hawkish opposition).

OPEN IN VIEWER

Figure 3.

Predicted probability of reciprocated cooperation as a function of the government and opposition ideology.

OPEN IN VIEWER

Figure 4.

Predicted probability of reciprocated cooperation as a function of the government and opposition ideology.

Figure 3 also shows only a marginally statistically significant relationship between a government’s hawkish or dovish nature and an international opponent’s reciprocation. At the same time, the domestic opposition’s ideological distance from the government in state A, either to the right or the left, has a strong influence on state B’s patterns of reciprocation. The probability of B’s reciprocation is lowest when the government in state A faces an extremely dovish opposition and increases as the opposition becomes more hawkish. These patterns are consistent with the expectations in Hypotheses 2a and 2b.

The inference follows that, when an international opponent anticipates the possibility of facing a more dovish government in the future, it has fewer incentives to cooperate with the current government. On the other hand, if the opposition is rather hawkish relative to the current government, then the international opponent would expect to face a hawk in the future should it fail to cooperate with the dove. In this situation, the international opponent is more likely to reciprocate the current (dovish) government’s cooperation. The cases discussed earlier also illustrate that considerations consistent with such inferences do indeed take place in diplomatic negotiations. Thus, the results show that the hawkish or dovish nature of the opposition can indeed influence an international opponent’s behavior, whereas the case studies also reveal that governments are in fact also inclined to use it as bargaining leverage.

In Figure 4, I graph the probability of reciprocation for a cohesive government and opposition—that is, when a dovish or a hawkish government faces an equally dovish or hawkish opposition (panel A). For comparison purposes, I include the graph showing the effect of the opposition’s ideological position relative to the government (panel B) from Figure 3. The bottom of each graph shows the observed distribution of cases at each value of the independent variable. Not surprisingly, panel A shows that cases of a cohesive government and opposition are quite rare, especially at the ideological extremes. ²⁴

As predicted in Hypotheses 3a and 3b, panel A shows a clear difference in the probability of reciprocation to governments facing an ideologically similar opposition depending on whether these are hawks or doves. In this situation, hawkish governments that are facing an ideologically similar opposition are more likely than dovish governments, also facing an ideologically similar opposition, to have their cooperative moves reciprocated. However, a comparison to panel B shows that even under conditions that are most beneficial for hawks to achieve reciprocation—that is, when the opposition is also hard-line—they are actually less likely to have their cooperative moves reciprocated compared to governments that are further to the left but that face a hawkish opposition. ²⁵ For example, while the probability of reciprocation for a hawkish government with a similarly hawkish opposition (located 50 points to the right of center) is roughly .70, the corresponding probability for a “centrist” government facing a hawkish opposition (50 points to the right) is higher (pr. ≈ .85).

The figure shows the opposite pattern for dovish governments without any major hawkish opposition. In this case, doves are disadvantaged when they cannot use the threat of a hawkish replacement as bargaining leverage. The probability of reciprocation is lowest when dovish governments face a similarly dovish opposition. Put together, these results show when doves can extract greater concessions from their opponents: doves may indeed have an advantage in negotiating concessions, but only if the international opponent can reasonably expect that they might face a hawkish government in the future.

Conclusion

This study provides a theoretical argument that dovish leaders have an advantage in negotiating with long-standing adversaries. The main premise is that international opponents do not bargain myopically but instead negotiate in anticipation of how their current bargaining behavior might influence who they face in future talks. If the alternative to an opponent’s present government is more hawkish, they are better off making the deal with the dovish government. These preferences over negotiating partners give doves bargaining leverage in international negotiations that can be used to extract concessions from an opponent. In the opposite case, when a hawkish government faces a dovish opposition, a hawkish leader loses bargaining power, since the international opponent has an incentive to stonewall negotiations in order to negotiate with a dovish alternative government in the future.

The empirical findings support the expectations. Rather than being disadvantaged in negotiations, I show the conditions under which dovish leaders are in a better bargaining position than are their hawkish counterparts to extract concessions from an international opponent. The results from the quantitative tests show that a government is more likely to have its cooperative moves reciprocated at the same or a higher level when it faces a hawkish domestic opposition. At the same time, the theory and results also reveal the very restricted circumstances under which hawks might have greater bargaining power than do doves. Only in the absence of a viable dovish opposition are hawks better positioned to make take-it-or-leave it offers to an international opponent. The opponent is, in turn, likely to accept these offers as the prospects of facing an alternative dovish government are virtually nonexistent. In fact, the failure to reciprocate may lead it to face an even more hard-line opponent in the future.

Similar to the previous studies in this area, my argument is perhaps most applicable to negotiations in bilateral disputes. Although multilateral disputes can pose a number of additional problems for negotiations—such as finding agreements that are acceptable to all parties—there is reason to believe that it can also be applied to these disputes as well. Morrow (1999) suggests that this problem can be overcome by reducing the number of actors involved in negotiations. This can be done by separating the issues and negotiating individual bilateral agreements with multiple adversaries (Kahler 1992). For example, though engaged in a multilateral dispute with several of its neighbors, Israel successfully negotiated separate agreements with Egypt in 1978 and Jordan in 1994. West Germany also settled its multilateral dispute arising from post–World War II territorial changes through a series of bilateral treaties with East Germany, Poland, Austria, and Czechoslovakia. China is taking a similar approach in its multilateral dispute over the South China Sea; it has consistently rejected calls for multilateral negotiations and instead insists on bilateral agreements with Philippines, Vietnam, Malaysia, and Taiwan. The possibility of finding bilateral solutions to multilateral disputes suggests that the argument and findings in this study can also be generalized to multilateral disputes as well. ²⁶

The strong empirical support for the argument that doves, rather than hawks, are better positioned to extract international concessions validates future research in this area. For example, although the theory and empirical tests in this study focused on the domestic opposition in democracies, the argument is generalizable to nondemocracies as well. It would therefore be plausible to examine whether U.S. cooperation during the cold war was influenced by the presence of hard-liners in the factional politics of the Soviet leadership. A similar analysis could apply to the context of current U.S.–Chinese relations.

Overall, the ideological stance of the domestic opposition does indeed have an impact on international opponents and the rate of success in negotiating with international rivals. These dynamics explain why doves may have greater bargaining leverage than their hawkish counterparts in negotiations with an international rival. This study offers one way of thinking about the causal mechanisms behind this expectation. The strong empirical support presented here validates a further expansion of this argument in the future.