Professional Documents
Culture Documents
Disputes
Author(s): Paul K. Huth and Todd L. Allee
Source: The Journal of Conflict Resolution , Dec., 2002, Vol. 46, No. 6 (Dec., 2002), pp.
754-790
Published by: Sage Publications, Inc.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to
The Journal of Conflict Resolution
TODD L. ALLEE
Department of Political Science
University of Michigan
AUTHORS' NOTE: We thank the National Science Foundation and the U.S. Institute of Peace for finan-
cial support of our research. The comments and suggestions of Curt Signorino are especially appreciated.
The data sets used in this article can be found at the JCR data site: http://www.yale.edu/unsy/jcr/jcrdata.htm.
754
stages through which international disputes progress and the diplomatic and military
policies adopted by state leaders at each of these different stages. The results of our sta-
tistical analyses provide strong evidence in support of the theoretical and empirical
approach we adopt for studying the democratic peace. In particular, we find that (a)
opposing patterns of accommodative as well as conflictual behavior by democratic
states can be explained by differences in the political accountability of democratic
leaders, and (b) the explanatory power of domestic- and international-level variables
varies in consistent ways across the different stages of a territorial dispute.
We divide this study into the following sections. First, we present a brief review of
the democratic peace literature and highlight the limitations that provide the rationale
for our theoretical and empirical work in this study. We then outline the different stages
of territorial disputes and develop a series of testable hypotheses that drawn on broader
theoretical models. In the third section, we describe a series of statistical tests and pres-
ent the results of such tests. Finally, we conclude by discussing the implications of our
findings for the literature on the democratic peace.
Our theoretical and empirical analyses of the democratic peace are built around the
behavior of states involved in territorial disputes. We assemble a data set consisting of
348 territorial disputes during the period from 1919 to 1995 and examine the diplo-
matic and military behavior of states seeking to change the territorial status quo and
those preferring to preserve it.2
We believe there are several advantages to analyzing a data set of territorial dis-
putes. First, by requiring that a territorial disagreement exists to begin with, we mini-
mize problems of irrelevant "no-conflict" observations. Our data set consists of cases
in which the use of diplomatic or military activity to advance one's claim is always a
possibility. A related benefit is that focusing on state behavior in territorial disputes
allows us to develop and test more specific hypotheses about state behavior. In particu-
lar, we are able to examine the conditions under which states are likely to resolve dis-
putes through negotiations, reach a deadlock in talks, or see their dispute escalate into a
2. See Huth and Allee (2002) for a more detailed description of the coding rules used for identifying
cases of territorial disputes and summary descriptions of territorial disputes.
war. A third advantage of this alternative research design is that by looking at the dif-
ferent stages of a territorial dispute, we can capture the impact of particular variables a
each stage. This helps us to address potential problems of selection bias (e.g., Fearon
1994a; Huth 1996; Reed 2000; Schultz 2001b; Smith 1995). A fourth and final advan-
tage of studying territorial disputes is that they are a central issue over which milita-
rized disputes and wars have erupted. As a result, the study of territorial dispute
should provide a demanding test for the impact of democratic institutions on the con-
flict behavior of foreign policy leaders. If democratic political institutions have the
capacity to discourage military conflict and promote peaceful resolution of disputes
then territorial disputes will push that capacity to its limits.
THEORETICAL FRAMEWORK
If the challenger state initiated a military threat in an attempt to alter the status quo,
then a military escalation stage ensues. The challenger and target are now in a military
confrontation, and each state must now decide whether to escalate a military confron-
tation and risk war. The challenger and target choices can be portrayed as falling into
one of two policy categories: (1) low escalation or (2) high escalation. A policy of low
escalation indicates that a state engages in at most limited military preparations and the
buildup of forces beyond initial military actions, even if the other party escalates to
higher levels of military preparedness. Escalation to high levels occurs when the chal-
lenger or target refuses to offer any territorial concessions and mobilizes for a large-
scale attack or resorts to the large-scale use of force when the other party st
and refuses concessions.
Over the duration of a dispute, decision makers make numerous choices regarding
the maintenance of the status quo, strategies for negotiations and dispute settlement,
and the use of coercive military pressure. These choices regarding diplomacy and mili-
tary force-and the resulting pathways to stalemate, dispute settlement, and war-are
the focus of our analysis in the remainder of this article. We present our hypotheses by
discussing the expected impact of each independent variable on state behavior in each
of these three stages.
THEORETICAL FOUNDATIONS
attempts at military coercion will also mobilize domestic opposition. Opposition elites
and mass publics therefore do not generally hold more dovish diplomatic and military
policy preferences than incumbent leaders. Instead, they simply seek to punish leaders
who adopt controversial or failed foreign policies.3
We expand on this third premise with two specific claims. First, we believe that
democratic leaders are typically more accountable for their foreign policy choices.
This is because political opposition in democratic regimes is generally more capable
of derailing policy programs and removing leaders from power (Bienen and van de
Walle 1992; Gelpi and Grieco 2001; Bueno de Mesquita and Siverson 1995) due to
institutions such as well-organized and independent political parties, regular competi-
tive elections, and independent legislatures. Second, the degree of political account-
ability among democratic leaders can vary depending on political conditions. In par-
ticular, we believe two factors-election cycles and differences in the strength of
opposition parties in legislatures-affect the degree to which democratic leaders are
accountable at any point in time (e.g., Alt and King 1994; Lohmann and O'Halloran
1994; Martin 2000; Milner 1997; Powell 2000).
Premise 4: The impact of political vulnerability on foreign policy. The greater the political
vulnerability of leaders, the higher the political costs to leaders for pursuing controversial
or unsuccessful foreign policy actions.
We believe that this fourth premise provides insights into bluffing and deception in
international disputes. Because bluffing is always a possibility in international dis-
putes, state leaders try to judge whether their adversary is in fact bluffing (e.g., Fearon
1994b). Leaders who are less vulnerable to political opposition should be more willing
to bluff because the domestic political risks of retreating from a clear and firm diplo-
matic or military policy are less extensive.
Hypothesis la: The leaders of democratic states should be less willing to rely
force to achieve territorial goals. As a result, democratic leaders should b
3. We are not convinced that mass opinions are consistently anchored at dovish polic
do we accept the claim that the foreign policy preferences in the mass public are consist
than the preferences of incumbent political elites (e.g., Gaubatz 1995; Holsti 1996; Ninc
Page and Shapiro 1992).
(a) less likely to initiate military threats as opposed to accepting the status qu
likely to challenge the territorial status quo with a call for negotiations rat
threat of military force,
(b) more likely to make concessions in negotiations,
(c) less likely to resort to higher levels of escalation in military confrontat
Hypothesis lb: The greater domestic audience costs of retreating in the face
and military pressure can provide a bargaining advantage to democratic
result, we expect leaders to be
(a) less likely to challenge the status quo with military threats against demo
states,
(b) more likely to offer concessions in negotiations when democratic adversari
signal their resolve to stand firm,
(c) less likely to escalate to higher levels in military confrontations when
adversaries strongly signal their resolve to use military force.
The next two hypotheses build on the general logic supporting hypothesis la, but
now we focus on the unwillingness of democratic leaders to offer concessions when
domestic political costs are expected to be high.
Hypothesis 2b: When ethnic conationals populate disputed territory, democratic leaders will
face stronger domestic pressures to adopt hard-line policies than their nondemocratic
counterparts. As a result, democratic leaders should be
(a) more likely to challenge the status quo with calls for talks and military threats and
especially likely to favor military threats over a reliance on negotiations,
(b) less likely to make concessions in negotiations,
(c) more likely to escalate military confrontations to high levels.
On the basis of earlier findings by Huth (1996) and Saideman (2001), we generally
expect incumbent leaders in all political systems to secure domestic support for back-
ing demands of greater political self-determination when ethnic conationals are
located across international borders in disputed territory. However, we argue further
that democratic leaders will face particularly strong pressures from domestic opposi-
tion groups and public opinion to take forceful initiatives to challenge the status quo
and support their ethnic conationals. We expect the defense of principles of political
In contrast, when the executive's party commands a majority position in the legisla-
ture, then opposition parties should be in a much weaker position to (a) veto the terms
of international agreements they dislike, (b) ensure criticism through legislative debate
and hearings, and (c) threaten removal in the event of a diplomatic or military retreat.
As a result, democratic leaders with greater political security should be more willing to
pursue controversial policies such as making concessions in negotiations or initiating
and escalating military confrontations.
In hypothesis 3b, we argue that secure democratic governments are more likely to
be the targets of calls for talks based on the expectation that they can offer concession
and still secure domestic ratification. It follows, then, that adversaries will view secure
democratic governments as more politically capable partners for trying to achieve a
negotiated settlement. The important task for a country's leaders is to calculate the
most opportune time to put offers of concessions on the negotiating table that will be
reciprocated by their negotiating partner. Political leaders do not want to incur the
political heat at home for offering concessions unless they believe their negotiating
partner can withstand the same type of domestic political pressure in their battle fo
ratification and approval of any agreement. If leaders expect a negotiated agreement to
unravel due to domestic opposition in the other country, then they have few incentives
to expose themselves politically to charges of selling out from their own domesti
opposition.
Because of the greater political flexibility of secure democratic leaders, one might
argue that such leaders are more likely to be targets of threats and higher levels of esca-
lation. The argument would be that if opponents know that secure democratic leaders
are better able to fight off domestic opposition following a military retreat, then these
opponents should be more willing to probe, bluff, and even escalate. Although this
logic is plausible, we emphasize the military risks associated with gambling that a
secure leadership in another country will back down under military pressure or accept
a limited military defeat. We would only expect risk-acceptant states to gamble by
challenging secure adversaries, yet we do not think that most states are consistently
risk acceptant in this way.
The next set of hypotheses considers the effects of elections on foreign policy
choices.
Hypothesis 4a: The more recently national elections have been held, the more likely are
incumbent leaders to
(a) challenge the territorial status quo, especially through a threat of military action;
(b) make concessions in negotiations;
(c) resort to higher levels of escalation in military confrontations.
Hypothesis 4b: The more recently national elections have been held in a democratic state, the
more likely are adversaries of that democratic state to
(a) challenge the territorial status quo with calls for talks instead of threats of military
action,
(b) make concessions in negotiations,
(c) avoid higher levels of escalation in military confrontations.
with such policies, we should expect a pattern to emerge between the timing of elec-
tions and the diplomatic and military policies pursued by leaders in territorial disputes.
In hypothesis 4a, the logic is that the accountability induced by elections should be
greater when democratic leaders expect to face elections relatively soon.4 In contrast,
when elections are not expected for some time, then the threat of electoral defeat
should have weaker political effects (e.g., Gaubatz 1999; Milner and Rosendorf
1997). Therefore, the more recently national elections have been held, the more will
ing incumbents should be to adopt an active foreign policy in which they seek negotia-
tions or threaten the use of force in an attempt to change the status quo. Furthermore,
democratic leaders should be more willing to make concessions in negotiations and
resort to higher levels of escalation in military confrontations in the periods shortly
after national elections.
Precisely because democratic leaders are less constrained in the period shortly after
elections and more capable of withstanding the political fallout of more accommoda-
tive policies, we argue in hypothesis 4b that other states will recognize this and judge
this to be a favorable time to seek a negotiated agreement based on the logic we devel-
oped in support of hypothesis 3b. As a result, it makes political sense for state leaders
to offer concessions to democratic negotiators when they believe that democratic lead-
ers are less constrained by the political risks of concession making (i.e., elections are
not expected to be held soon). Conversely, when democratic leaders face upcoming
elections, they should shy away from territorial concessions in negotiations. Knowing
this, adversaries should be less willing to offer concessions themselves.
When we consider the effects of recent elections in a democratic adversary on the
choices of state leaders to initiate or escalate the use of military force, we encounter the
same opposing arguments about what to expect as discussed above for hypothesis 3b.
Again, although it is possible that other states might be tempted to initiate and escalate
military challenges against democratic governments that have recently held elections
in an attempt to pressure secure leaders into making territorial concessions by military
bluffs or the use of force, we believe this is a quite risky strategy. Instead, we argue that
adversaries would be less likely to initiate and escalate military confrontations
because they are worried that politically secure democratic leaders are more willing to
risk a military conflict to defend their territorial claims.
4. Of course, in presidential systems, the time period between elections is fixed and thus known in
advance. In parliamentary systems, elections must be held within a certain period of time, but exactly when
they will be held is likely to be a function of a range of political conditions. Despite the greater uncertainty
facing leaders in such a situation, the general logic of the argument still holds. Once elections have occurred,
they are unlikely to be held again within a short period of time.
Hypothesis 5: Lower levels of diplomatic and military conflict are expected in terr
putes between two democratic states. As a result, state leaders in democr
should be
(a) less likely to initiate military challenges and more likely to seek negotiations,
(b) more likely to make concessions in negotiations over disputed territory,
(c) less likely to resort to higher levels of escalation in military conflicts over disputed
territory.
Hypothesis 6a: In mixed dyads, military confrontations generally result from nondemocratic
challengers threatening democratic targets.
Hypothesis 6b: In mixed dyads, the resort to high levels of escalation in a military confronta-
tion is most likely by nondemocratic leaders against democratic adversaries.
5. See Huth and Allee (2002, chap. 3) for a more detailed description of these internati
and military variables.
6. In our data set, there are 211 instances in which multiple foreign policy initiatives co
puted territory are initiated within the same year. For example, two states might hold talks twic
a pair of states might engage in negotiations over disputed territory in March, only to see
mobilize troops against the target in October.
month-specific nature of our data also ensures that the sequencing of important events
is captured accurately.
Although we employ a different statistical model for the status quo stage as
opposed to the negotiation and military escalation stages, several model estimation
features are common across the three stages. First, for each stage, we estimate three
separate statistical models to test the various hypotheses from the political account-
ability model. These three models mirror the distinctions between accountability-
based arguments we made in the earlier theoretical section. We estimate an "across-
regimes" model, a "within-regimes" model, and then a "dyadic" model. Dividing the
estimation in this manner makes the most sense conceptually and is the most effective
way to test such a wide range of arguments. When estimating each model, we also
include the set of international political-military variables described at the beginning
of this section. Furthermore, due to the nonlinear nature of all of our models, we also
present a series of predicted probability results to provide a more substantive interpre-
tation of variable effects (see Tables 4, 8, and 12). We estimate the impact of discrete
changes in particular variables on the predicted probability of certain outcomes by
holding all other variables constant. Although the coefficient results for each econo-
metric model generally provide a basic sense of the estimated direction and signifi-
cance of hypothesized relationships, these predicted probability results are often more
substantively meaningful. Finally, we estimate all models using Huber or "robust"
standard errors due to concerns with possible contemporaneous correlation and
nonconstant variances across the units of observation.7
Our first question concerns the decision of challenger states to initiate diplomatic or
military actions in pursuit of their country's claims to disputed territory. There are
three distinct options available to the leaders of challenger states: (1) refrain from any
initiatives and maintain the status quo, (2) propose talks and rely on negotiations in an
attempt the alter the status quo, and (3) resort to threats of military force in support of
territorial claims. Because these three choices are not clearly ordered, we require a
model that treats outcomes as nominal instead of ordinal (Long 1997, chap. 6). As a
result, we estimate a series of multinomial logit models to explain the decisions of
leaders in the challenge the status quo stage. The primary drawback of multinomial
logit is the fairly restrictive independence of irrelevant alternatives (IIA) assumption.
However, we employ the two best-known tests for the IIA assumption and find no evi-
dence to reject IIA in any of our models.8
7. Because the observations in our data set span different numbers of months, we are especially sensi-
tive to concerns with heteroscedasticity.
8. The independence of irrelevant alternatives (IIA) assumption is met when an individual's prefer-
ences among alternatives remain consistent regardless of which choices are or are not available (see
McFadden 1981). We employ both the Hausman and the Small-Hsiao tests to examine the IIA assumption
(Hausman and McFadden 1984; Small and Hsiao 1985). In all cases, we perform the tests after first eliminat-
ing the option of negotiations and then the option of military force. At times, we obtain a negative chi-square
statistic, which Hausman and McFadden (1984, 1226) claim should be interpreted as evidence that IIA has
not been violated.
We end up with 6,542 observations for the challenge the status quo stage. T
1,782 cases in which a challenger initiates a round of talks and 390 cases i
challenger initiates the use of force. For operational purposes, informatio
pendent variables for the challenge the status quo stage is drawn from the
which the challenger initiates a militarized confrontation or a round of ta
Cases in which the challenger does not challenge the status quo are more d
conceptualize and operationalize. We use a "12-month rule," which holds
months have elapsed since a challenger state has threatened force or called
then it is considered to have "done nothing" during that period of time. That
then included in the data set as a case in which no challenge was made.9 E
sive 12 months of activity is treated in the same way until the state once again
new round of talks or threatens force.1' Finally, the temporal structure of t
raises some concerns about serial correlation. In the spirit of Beck, Katz,
(1998), we include a variable to count the number of months since the chal
undertook some activity-whether it was a call for talks or threat of forc
Beck 1998). This variable serves primarily as a control for the impact of tim
history on decisions to challenge the status quo.
The multinomial logit results are presented in Tables 1 through 3. In each of these
models, we set the choice to accept the status quo as the baseline category, which gen-
erates results that compare the options of doing nothing versus both pursuing negotia-
tions and initiating force. However, because we also care about the choice between
challenging through talks versus force, we also present results for this direct compari-
son." Tables 1 through 3 contain three columns of coefficient results that compare each
choice to the remaining two options. The impact of discrete changes in hypothesized
variables on the probability of calls for negotiations and threats of force is summarized
in Table 4.
We first examine the results of the political accountability model hypotheses (see
Tables 1-3). In sum, the very clear overall conclusion we draw is that democratic lead-
ers are much less likely to turn to threats of force in attempt to change the territorial sta-
tus quo and are considerably more likely to challenge the territorial status quo by call-
ing for negotiations. These results hold in both the monadic and dyadic contexts and
9. In this case, we randomly sample one of the months during this period of no challenge and include
it as an observation in the operational challenge the status quo stage data set.
10. We also consider and employ a number of additional specifications for the decision "not to chal-
lenge." One option is to use a modified version of the 12-month rule, in which case the 12-month rule is used
for up to 5 years after some action and 3 or fewer years before some observable action. The logic is that states
are only likely to actively consider contesting territory when the territorial issue is salient as opposed to dor-
mant. Yet another strategy is to randomly sample periods of inactivity from the months in which no action
was taken. We include varying numbers of random "no-challenge" cases and find only negligible differ-
ences based on the number of random observations included. As a result, we find the results for the challenge
the status quo stage to be remarkably stable regardless of the coding rule we employ for including observa-
tions in which no challenge was made.
11. The comparison of talks versus force is a straightforward transformation of the same model (see
Liao 1994).
TABLE 1
sug
atio
neg
dem
dom
Fir
rac
less
com
TABLE 2
dem
(see
cra
ver
dya
ter
12.
catio
TABLE 3
pat
mix
in
How
atte
acr
bot
and
are
and
neg
On
cies
late
TABLE 4
Recent election in target (no -> yes) 28.7 33.3 +4.6 +16.0
Strategic value of territory (no -> yes) 26.9 30.4 +3.5 +13.0
Strategic value of territory (no -> yes) 3.8 5.1 +1.3 +34.2
with ethnic conationals, democratic challengers now become more than twice as likely
to threaten force compared with nondemocratic challengers (see Table 4). According
to our data, democratic leaders in such scenarios stand a 14.4% probability of threaten-
ing force, whereas nondemocratic leaders are predicted to threaten force only 6.5% of
the time.'1
13. These probabilities represent the impact of democracy only in the presence of ethnic ties, remov-
ing the impact of democracy in any other way. All other democracy-based interaction term probabilities are
examined in a similar manner.
Among democratic states, electoral cycles (hypotheses 4a and 4b) also help explain
when challengers are likely to challenge the territorial status quo, especially by turning
to negotiations (see Table 2). Many of the hypothesized relationships about the impact
of election timing are significant and in the predicted direction. As challenger states
begin to approach elections, they become more conservative and less likely to issue
any form of territorial challenge. Democratic leaders in challenger states are nearly
20% less likely to call for negotiations near the end of an electoral cycle and 25% less
likely to threaten force to pursue territorial claims when they face upcoming elections.
Similarly, the electoral cycle in target states also affects the challenger's calculations.
Challengers are 16% more likely to pursue talks right after the target has recently held
an election as opposed to when the target is nearing a general election.
We now turn to the results for the control variables relating to international
political-military conditions. The impact of these variables is robust across the three
different specifications of the political accountability model (see Tables 1-3).'4 On the
whole, these variables provide a strong explanation for decisions by challengers to
employ threats of force to pursue their territorial claims, yet they say very little about
challenger decisions to use negotiations to address territorial claims. Four of the five
variables regarding threats of military force receive strong support. In general, military
threats to the territorial status quo are more likely when the territory has strategic
value, the military balance favors the challenger, and the target is currently involved in
another militarized dispute. For example, challengers are 34% more likely to turn to
military force in at attempt to acquire strategically valuable territory. In addition, chal-
lengers are more than three times as likely to threaten force to acquire disputed terri-
tory when they have a sizable military advantage compared with situations in which
they are at a significant military disadvantage (see Table 4). On the other hand, chal-
lengers are less likely to issues coercive challenges to the territorial status quo when
they share common security ties with the target. In fact, the likelihood of a military
challenge drops by one-third when the two states are in other territorial disputes with a
common opponent (see Table 4).
If the challenger decides to challenge the status quo either through a call
ations or a threat of force, our analysis shifts to examining the outcom
rounds of talks and military confrontations, respectively. Unlike the challen
tus quo stage, here we consider decisions made by both the challenger an
However, we do not analyze rounds of talks and militarized disputes dy
attempt to code or explain a "joint" outcome for a pair of states. Instead
separate outcome (either degree of concessions or level of escalation) fo
challenger and target. This allows us to understand how particular variabl
14. As a general rule, for purposes of consistency, we use the estimated results from
regimes version of the political accountability model to interpret the significance of the
political-military variables and calculate the impact of changes in these variables on the pred
ity of the various outcomes of interest.
15. The use of "directed dyads" is one way to analyze the behavior of two disputants
directional observations are not independent, and the disturbances across the two observat
be correlated.
16. See Greene (1997,906-11) for a general discussion of bivariate probit and Smith (1999) for a spe-
cific discussion of bivariate models and the interrelatedness of state decisions.
17. Bivariate probit is sometimes used as a method for estimating two potentially interrelated deci-
sions of the same actor (see Reed 2000). Yet bivariate probit is also an appropriate method for modeling the
related actions of two separate actors (see Zorn 2002).
18. Recall that the data used to explain challenger decisions to challenge the status quo are taken from
the beginning month of the round of talks or militarized dispute, when the actual "challenge" was first made.
We present the challenger results from the three bivariate probit models in Tables 5
through 7. Table 8 contains predicted probability estimates of the impact of discrete
changes in hypothesized independent variables.
The hypothesis that democracies are typically more likely to make concessions
(hypothesis la) receives some support. The coefficient on democracy level (in the tar-
get concession equation) is positive and statistically significant (see Table 5). In fact,
strongly democratic targets are more than 67% more likely to make concessions in
talks over disputed territory than are strongly nondemocratic targets (see Table 8).
There is some suggestive evidence that democratic challengers also are more likely to
make concessions, although the coefficient on level of democracy in the challenger
equation is not statistically significant. Nevertheless, the impact of challenger democ-
racy does appear positive and statistically significant when we estimate a Heckman
probit model for the challenger to check for the existence of any selection effects.2
19. Formally, the large and statistically significant rho indicates that variables omitted from the equa-
tion to predict challenger concessions are highly correlated with those variables omitted from the equation to
predict target concessions.
20. The estimated rho for the across, within, and dyadic models ranges from .16 to .21. None of the
three Wald tests for dependence between the status quo stage disturbances and the negotiation stage distur-
bances are statistically significant.
21. In fact, in the Heckman model, the coefficient for the impact of challenger democracy on chal-
lenger concessions is positive, and the p value is .04. We are hesitant to place too much emphasis on this
result because it does not account for challenger and target interdependence in any way, yet the finding is
suggestive when placed in a broader context.
TABLE 5
TABLE 6
TABLE 7
TABLE 8
Common security ties (no alliance -> alliance) 40.2 47.3 +7.1 +17.7
there is some support for the idea that concessions are more likely between pairs of
democracies (see Table 7). Target states in democratic dyads are 21% more likely to
offer concessions in talks over disputed territory than are targets in nondemocratic
dyads (see Table 8). The result for challengers in democratic dyads is similarly positive
but is not quite statistically significant, with a p value of .15 for a one-tailed test.
Only one of the control variables, common security ties between challenger and tar-
get, receives strong support. Challengers and targets are more likely to make conces-
sions to one another when they share common security ties (see Table 5 and also
Tables 6 and 7). In particular, the presence of an alliance between the two st
challengers and targets about 18% and 22% more likely to make concession
tively (see Table 8). The only other significant finding is that targets with a
itary advantage are 47% less likely to make concessions than targets with a si
itary disadvantage.
23. The estimated rho for the Heckman probit versions of the three accountability models
-.53 to -.57. Wald tests for the independence of challenger status quo and escalation decisi
standard levels of statistical significance, as p values range from .10 to .12.
24. This finding also holds when we run a model that includes solely the international
ables and the single variable for challenger (target) democracy, without any of the democracy-
tion terms. Furthermore, according to the Heckman probit model, democratic challenger
somewhat more likely to escalate. In this model, the coefficient for the challenger level of
able is positive, and the p value is .18.
result suggests that monadic findings of a democratic peace are attributable largely
the fact that democracies tend to avoid the initiation of military conflict with all states
a finding we discussed earlier in the discussion of status quo stage results.
There is some modest evidence that particular democratic regimes are associated
with decisions to escalate to higher levels of force (see Table 10). Challenger leaders
who are in a position of strength domestically (hypothesis 3a) are more than twice
likely to escalate force than are leaders who are in a weak domestic position (see Tab
12). We also find that challengers are more than twice as likely to escalate when target
face upcoming elections (hypothesis 4b) as opposed to when targets have recently he
elections (see Table 12). Similarly, challengers are less likely to escalate force agains
leaders of democratic target states who are in a strong position domestically (hypothe-
sis 3b). The predicted probability of challenger escalation drops from nearly 19% t
about 12% when target leaders command a clear majority of parliamentary or legisl
tive seats as opposed to when the target leader's party holds only a 30% minority o
seats.
Our final result for the escalation stage is that democratic dyads are very unlikely
see their military confrontations escalate to high levels of force. In fact, we find
cases in which challengers in democratic dyads escalate to higher levels of force
only one case in which a democratic target escalates against a democratic challeng
Recall from our earlier findings that territorial dispute challengers in democratic dy
are less likely to choose the route of military coercion in the first place, as only 16 of
374 military confrontations are between pairs of democratic states. But here we f
that democratic dyads are also less likely to escalate in the rare event of a democr
dyad military confrontation. The Heckman probit model for the challenger also c
firms this result.25 Finally, there is some support for the idea (hypothesis 6b) that n
democratic states in mixed dyads are more likely to escalate than are democratic stat
in mixed dyads (see Table 11). Nondemocratic targets in mixed dyads are more th
83% more likely to escalate force than are democratic targets in mixed dyads (
Table 12). However, we find no discernible difference between the behavior of de
cratic and nondemocratic challengers in mixed dyads (see Table 11).
In general, international political and military variables provide quite strong resul
for escalation decisions, especially compared to their relatively weak performanc
the negotiation stage. Five of the six control variables receive strong support, and fo
of these five supported variables push challengers toward higher levels of escalat
(see Table 9 and also Tables 10 and 11). For example, when the territory is of strat
value, the predicted probability of challenger and target escalation increases by m
than 150% and 70%, respectively (see Table 12).26
Similarly, the presence of a military advantage makes the challenger more likely
escalate military confrontations to higher levels of force.27 Also, both the short-ter
25. According to the results of the Heckman model, challengers in democratic dyads are less likel
challenge the status quo through threats of military coercion (p < .01) and are then less likely to esc
force in military confrontations (p < .01).
26. This substantively large effect remains when we estimate a Heckman probit model.
27. In the Heckman model, the relationship between challenger local balance of forces advantage
escalation remains very strong (p < .001), whereas the results for the short-term balance are only slig
weaker (p value of .12).
TABLE 9
TABLE 10
CONCLUSION
TABLE 11
TABLE 12
Strategic value of territory (no -> yes) 6.7 16.9 +10.2 +152.2
Strategic value of territory (no -> yes) 11.8 20.1 +8.3 +70.3
NOTE: The predicted probabilities represent the marginal probability of challenger (target) escalation, re-
gardless of whether the target (challenger) also escalates. Unless specified otherwise, the territory is consid-
ered to be of strategic value to both states, the military balance is 3:2 in favor of the target, and democratic
governments are assumed to hold a 55% legislative majority and to have won an election 2 years ago. All
other variables are set to median values. TD = territorial dispute.
very different choices are made by state leaders to pursue negotiations and refrain
from using force as bargaining leverage in possible future negotiations over disputed
territory.
In the second part of our analysis, we tested how well each of the models accounted
for decisions by leaders to offer concessions in negotiations. Although realist variables
produced weak results, the findings for the political accountability model were quite
supportive. For example, democratic leaders in challenger states are quite sensitive to
electoral cycles and prefer to offer concessions in periods shortly after national elec-
tions have been held in their own country or in democratic targets. Another finding was
that democratic leaders in challenger states are keenly aware of the domestic politics of
treaty ratification and therefore are unlikely to bring back home territorial agreements
APPENDIX
Local balance of military forces. This measures the forces that each side has mobilized dur-
ing a militarized dispute and therefore can be committed at the outset of an armed conflict. The
local balance is coded as a dummy variable, and a value of 1 is recorded if a state enjoys a mili-
tary advantage of roughly greater than two to one.
Other military dispute involvement of challenger or target. A dummy variable is coded with
a value of 1 when a state is currently involved in another military conflict. The primary data
sources are the COW data set on militarized interstate disputes from 1816 to 1992, as well as ou
data on military confrontations over disputed territory.
Common security ties. Two measures are constructed. The first measure is an alliance
dummy variable for which a value of 1 is recorded if there is a defense pact or entente military
alliance between the challenger and target. The second measure is a dummy variable that indi-
cates whether the challenger and target currently share a common territorial dispute adversary.
Strategic value of disputed territory. A dummy variable is coded as equal to 1 if the territory
is strategically located or if it contained (or was believed to contain) natural resources that wer
used by the state in the production of military weapon systems.
Dummy variablefor democratic status of challenger and target. We also code and use a con-
trol variable that indicates whether a country is democratic based on the democratic accountabil-
ity variable (see above). We consider a country to be democratic if its POLITY net democracy
score is +6 or higher. We assign a value of 1 to these cases and assign a 0 to all other cases.
Democratic leaders and signals of resolve. We generate this interaction term by multiplying
the target democratic accountability variable (see above) and a dummy variable that indicates
whether a target state signals resolve in its initial response to a challenger's threat of force or ca
for negotiations. A signal of resolve in negotiations is coded as the refusal of the target negotia-
tors to offer any concessions at the outset of talks. For military confrontations, a signal of resolv
occurs when the target's initial response to the challenger's threat was to respond with an equa
or greater level of force and to refuse any concessions.
Democratic response to stalemate. This interaction term is created by multiplying the demo
cratic accountability variable (see above) and a second dummy variable that is coded with
value of 1 if a recent round of talks had ended in a stalemate. A recent stalemate has occurred
when a round of talks within the past 2 years resulted in one side offering only limited or no con-
cessions, whereas its opponent offered no concessions at all.
Democracy and ethnic conationals. This is another interaction term created by multiplying
the democratic accountability variable (see above) and a dummy variable to indicate whether
ethnic conationals are located in disputed territory.
Strength of ruling coalition in legislature. In presidential or mixed systems where the presi-
dent is responsible for foreign policy decisions, we collect data on the percentage of seats held
by the president's party (and its coalition allies) in the lower house of the legislature, as well as
the corresponding percentage in the upper house if the upper chamber has effective treaty ratifi-
cation powers (for use in the negotiation stage analysis). In parliamentary or mixed systems in
Timing of elections for challenger and target. This variable measures the numbe
since the previous election for each state based on the dates of all national elect
presidents or for national legislatures (depending on the state's type of politica
operational purposes, we also create and sometimes substitute a dummy variable
an election has been held within the past year.
Mixed dyads and nondemocratic states in mixed dyads. A dyad is considered mixe
democracy score for one of the two states is +6 or higher, and the net democrac
other state is +5 or lower. We also create a dummy variable to indicate that a state
ocratic state in a mixed dyad.
REFERENCES
Alt, James, and Gary King. 1994. Transfers of governmental power: The meaning of ti
Comparative Political Studies 27 (2): 190-210.
Auerswald, David. 2000. Disarmed democracies: Domestic institutions and the use off
University of Michigan Press.
Beck, Nathaniel. 1998. Modeling space and time: The event history approach. In Research
social sciences, edited by Elinor Scarbrough and Eric Ranenbaum, 192-212. Oxford, U
versity Press.
Beck, Nathaniel, Jonathan Katz, and Richard Tucker. 1998. Taking time seriously: T
section analysis with a binary dependent variable. American Journal of Political Scie
88.
Bennett, D. Scott, and Allan Stam. 2000. Research design and estimator choices in the analysis of interstate
dyads. Journal of Conflict Resolution 44 (5): 653-85.
Bienen, Henry, and Nicolas van de Walle. 1992. Of time and power. Stanford, CA: Stanford University
Press.
Bueno de Mesquita, Bruce, James Morrow, Randolph Siverson, and Alastair Smith. 1999. An institutional
explanation of the democratic peace. American Political Science Review 93 (4): 791-808.
Bueno de Mesquita, Bruce, and Randolph Siverson. 1995. War and the survival of political leaders. Ameri-
can Political Science Review 89 (4): 841-55.
Cederman, Lars-Erik. 2001. Back to Kant: Reinterpreting the democratic peace as a macrohistorical learn-
ing process. American Political Science Review 95 (1): 15-32.
Elman, Miriam Fendius, ed. 1997. Paths to peace. Cambridge, MA: MIT Press.
.2000. Unpacking democracy: Presidentialism, parliamentarism, and theories of democratic peace.
Security Studies 9 (4): 91-126.
Fearon, James. 1994a. Signaling versus the balance of power and interests. Journal of Conflict Resolution
38 (2): 236-69.
. 1994b. Domestic political audiences and the escalation of international disputes. American Politi-
cal Science Review 88 (3): 577-92.
Gartzke, Erik. 1998. Kant we all just get along? American Journal of Political Science 42 (1): 1-27.
.2000. Preferences and the democratic peace. International Studies Quarterly 44 (2): 191-212.
Gaubatz, Kurt. 1995. Intervention and intransitivity: Public opinion, social choice, and the use of military
force abroad. World Politics 47 (4): 534-54.
. 1999. Elections and war: The electoral incentive in the democratic politics of war andpeace. Stan-
ford, CA: Stanford University Press.
Gelpi, Christopher, and Joseph Grieco. 2001. Democracy, leadership tenure, and the targeting of militarized
challenges. Journal of Conflict Resolution 45 (6): 794-817.
Goemans, Hein. 2000. War and punishment: The causes of war termination and the First World War. Prince-
ton, NJ: Princeton University Press.
Gowa, Joanne. 1999. Ballots and bullets. Princeton, NJ: Princeton University Press.
Greene, William. 1997. Econometric analysis. 3d ed. Upper Saddle River, NJ: Prentice Hall.
Gurr, Ted Robert, and Keith Jaggers. 1999. Polity 98: Regime Characteristics 1800-1998. College Park,
MD: Center for International Development and Conflict.
Hausman, Jerry, and Donald McFadden. 1984. Specification tests for the multinomial logit model.
Econometrica 52:1219-40.
Heckman, James. 1979. Sample selection bias as a specification error. Econometrica 47 (1): 153-61.
Holsti, Kalevi. 1996. The state, war, and the state of war. Cambridge, UK: Cambridge University Press.
Huth, Paul. 1996. Standing your ground. Ann Arbor: University of Michigan Press.
Huth, Paul, and Todd Allee. 2002. The democratic peace and territorial conflict in the twentieth century
New York: Cambridge University Press.
Jaggers, Keith, and Ted Robert Gurr. 2000. POLITYIII: Regime change and political authority, 1800-1994.
ICPSR XML version. Ann Arbor, MI: Inter-University Consortium for Political and Social Research.
Jones, Daniel, Stuart Bremer, and J. David Singer. 1996. Militarized interstate disputes, 1816-1992: Ratio
nale, coding rules, and empirical patterns. Conflict Management and Peace Science 15 (2): 163-213.
Layne, Christopher. 1995. On the democratic peace. International Security 19 (4): 175-77.
Liao, Tim Futing. 1994. Interpreting probability models: Logit, probit, and other generalized linear models
Sage University Paper Series on Quantitative Applications in the Social Sciences, 07-101. Beverly Hills
CA: Sage.
Lohmann, Susanne, and Sharyn O'Halloran. 1994. Divided government and U.S. trade policy. International
Organization 48 (4): 595-632.
Long, J. Scott. 1997. Regression models for categorical and limited dependent variables. Thousand Oaks,
CA: Sage.
Maoz, Zeev. 1997. The controversy over the democratic peace. International Security 22 (1): 162-98.
.1998. Realist and cultural critiques of the democratic peace. International Interactions 24 (1): 3-89.
Martin, Lisa. 2000. Democratic commitments. Princeton, NJ: Princeton University Press.
McFadden, Daniel. 1981. Econometric models of probabilistic choice. In Structural analysis of discrete
data, edited by C. F. Manski and Daniel McFadden, 198-272. Cambridge, MA: MIT Press.
Milner, Helen. 1997. Interests, institutions, and information. Princeton, NJ: Princeton University Press.
Milner, Helen, and Peter Rosendorff. 1997. Democratic politics and international trade negotiations: Elec-
tions and divided government as constraints on trade liberalization. Journal of Conflict Resolution 41
(1): 117-46.
Nincic, Miroslav. 1992. Democracy andforeign policy: The fallacy ofpolitical realism. New York: Colom-
bia University Press.
Page, Benjamin, and Robert Shapiro. 1992. The rational public: Fifty years of trends in Americans'policy
preferences. Chicago: University of Chicago Press.
Powell, Bingham. 2000. Elections as instruments of democracy. New Haven, CT: Yale University Press.
Reed, William. 2000. A unified statistical model of conflict onset and escalation. American Journal of Polit-
ical Science 44 (1): 84-93.
Reiter, Dan, and Allan Stam. 2002. Democracies at war. Princeton, NJ: Princeton University Press.
Russett, Bruce, and John Oneal. 2001. Triangulating peace: Democracy, interdependence, and interna-
tional organizations. New York: Norton.
Saideman, Stephen. 2001. The ties that divide. New York: Columbia University Press.
Schultz, Kenneth. 2001a. Democracy and coercive diplomacy. Cambridge, UK: Cambridge University
Press.
2001b. Looking for audience costs. Journal of Conflict Resolution 45 (1): 32-60.
Signorino, Curtis. 1999. Strategic interaction and the statistical analysis of international conflic
Political Science Review 93 (2): 279-98.
Small, Kenneth, and Cheng Hsiao. 1985. Multinomial logit specification tests. Internationa
Review 26 (3): 619-27.
Smith, Alastair. 1995. Alliance formation and war. International Studies Quarterly 39 (4): 405-
.1999. Testing theories of strategic choice: The example of crisis escalation. American J
Political Science 43 (4): 1254-83.
Snyder, Glenn. 1997. Alliance politics. Ithaca, NY: Cornell University Press.
Snyder, Jack. 2000. From voting to violence: Democratization and nationalist conflict. New Yor
Zorn, Christopher. 2002. U.S. government litigation strategies in the federal appellate cour
Research Quarterly 55:145-66.