Foundations for integrating the democratic and territorial peace arguments

Democratic and territorial peace: a misguided debate?

Democratic dyads are less likely to go to war than non-democratic dyads—that is, dyads containing at least one non-democratic state (e.g. see Chan, 2012; Russett and Oneal, 2001). Yet the pacific effects of democracy extend beyond war. Democratic dyads are also less likely than non-democratic dyads to threaten, display, or use force against one another (i.e. engage in militarized interstate disputes—or MIDs; see Ghosn et al., 2004) and to experience international crises with one another (Brecher and Wilkenfeld, 2000; Maoz and Russett, 1993). In short, democratic dyads treat one another more peacefully and view one another as less threatening than members of non-democratic dyads, thereby creating the foundation of the democratic peace argument.

A series of related mechanisms may explain democratic dyads’ peaceful behavior. First, democratic institutions can constrain leaders’ behavior—for example, by precluding unilateral action (through checks and balances, including public opinion) or by encouraging negotiation (because democracies take time to mobilize, have incentives to fight harder to win any battles that occur, or face audience costs that will compel them to fight after a certain threshold is crossed; see Bueno de Mesquita et al., 2003; Fearon, 1994; Slantchev, 2006). Second, opposition groups found in democracies may signal to opponents whether resolve for fighting is high (Schultz, 1999). Such signals allow leaders to minimize war via misunderstanding. Finally, democratic norms might encourage democracies to resolve political disputes peacefully. Elections encompass this norm domestically, but the norm may also affect foreign policy decisions (Maoz and Russett, 1993; Mitchell, 2002).

These mechanisms raise two points worth noting for the current study. First, neither mechanism depends upon the issues over which states fight; each encourages democracies to behave more peacefully with each other than with non-democracies, regardless of the issue under dispute. ³ Second, the mechanisms suggest that the democratic peace involves more than war involvement. Norms might persuade democracies to forego any use of the military when dealing with other democracies, even in cases where war is unlikely. Similarly, institutions may alter the ways in which democracies can use the military when confronting one another, even in cases short of war. Both points facilitate the analysis that follows.

Although the democratic peace features prominently in interstate conflict research, it is not the only strong empirical finding within the subfield. Indeed, as scholars have studied the issues over which states fight, findings (re)affirm that territorial disputes are unique and dangerous. Territorial disputes are more likely to be handled militarily than non-territorial disputes (Senese, 2005). Furthermore, once militarized, these disputes are subsequently also more likely than non-territorial disputes to recur and escalate to war (Vasquez, 2009; for an overview, see Tir and Vasquez, 2010).

Vasquez (2009) proposes that these patterns result from territoriality—that is, the tendency of humans to divide the world into distinct territorial units and then to defend these units by force. Moreover, Vasquez proposes that this tendency is a learned behavior; states learn that force is an appropriate allocation mechanism for handling their territorial—but not non-territorial—disputes, leading to more aggressive behavior when territorial issues are at stake (e.g. militarization and war). If true, Vasquez (2009: 154–155) offers two predictions relevant for this study. First, many states will engage in interstate conflict over the delimitation of territorial units (i.e. their borders), thereby transforming borders into a series of conflict fault lines. Second, if territoriality is a learned behavior that sees aggression as an appropriate means of handling territorial (border) disputes alone, then dyadic interstate behavior should become substantially more pacific after a dyad settles its mutual border(s), regardless of the other issues over which disagreements arise. Both predictions garner empirical support (Gibler, 2012; Huth, 1996; Owsiak, 2012).

Gibler (2012), however, extends Vasquez’s proposition further. In addition to affecting conflict behavior, Gibler argues that border settlement may contribute ⁴ to the appearance of joint democracy within a dyad as well. This prediction derives most clearly from the war-making/state-making tradition (e.g. see Tilly, 1992). According to this tradition, a border dispute represents a salient external threat that challenges the identity and existence of a state. To combat this threat, states build their military and centralize power in a (stronger) executive; the former offers better capabilities with which to defend the state, while the latter permits quicker military decisions. Thus, external threats encourage a regime with more non-democratic characteristics. If this is true, however, then the regime holds only so long as the external threat remains in place. Herein lies Gibler’s novel insight: once states settle their borders, the external threat dissipates, and the non-democratic pressures ease. More specifically, when the external threat eases, domestic groups no longer see the need for a strong(er) executive and begin trying both to constrain the executive’s power (i.e. through checks and balances) and to secure greater individual rights for themselves—both of which correspond with more democratic regimes.

Gibler’s (2012) argument constitutes the “territorial peace,” and it comprises two premises: after a contiguous dyad settles its mutual borders, the dyad is both less likely to fight and more likely to contain two democratic states. ⁵ Based on this, Gibler argues that the democratic peace argument is spurious or epiphenomenal; in effect, he claims that the territorial peace subsumes the findings associated with the democratic peace (see also Thompson, 1996). Thus, Gibler essentially defines the democratic and territorial peace arguments as competitors—either one or the other is accurate (Gibler, 2014; Owsiak, 2012; Park and Colaresi, 2014).

Although the democratic and territorial peace arguments may be competitors, I argue that this debate is misplaced and masks avenues for future theoretical advancement. Indeed, as Park and James (2014: 86) note, “One explanation does not have to be at odds with the other. Each could remain meaningful even in a context where the former is conditioned by the latter and vice versa.” To entertain this possibility fully, one must specify “what relationships should hold when,” thereby theorizing about subsets (or groups) of observations (Most and Starr, 1989: 113–114). This is the foundation of the current study, which seeks to uncover both the theoretically unique domains in which the two arguments might operate and the ways in which they might interact.

Such a foundation offers two advantages. First, it may help reconcile seemingly contradictory findings (see above). Differing results may simply derive from the inability of scholars to articulate the unique and integrative effects of each argument, as I show below. Second, it advances preliminary work done by Park and James (2014). These authors uncover a pacific interactive effect between territorial claims and democratic dyads (i.e. democracies with a territorial claim are less likely to experience a MID). Although this finding pushes research in the right direction, it is unhelpful for considering domain-specific theoretical possibilities for two reasons. First, the authors do not separate contiguous and non-contiguous dyads, even though the territorial peace theoretically specifies contiguous dyads as its domain (Gibler, 2012, 2014; Owsiak, 2012). As a result, the study necessarily overlooks the unique domains in which the democratic and territorial peace arguments might operate. Second, the interactive effect uncovered in the study may have an alternative explanation. The territorial peace maintains that contiguous dyads settle their borders before becoming democratic. ⁶ Thus, dyads will have fewer (and less salient) territorial disputes when regime type enters the equation. A more telling integrative test must therefore examine whether democracy exerts a pacific effect on interstate conflict behavior after border settlement occurs, for it is here that the territorial and democratic peace arguments likely reinforce one another. ⁷

Foundations for integration

One can derive theoretical predictions that both isolate the unique domains of the democratic and territorial peace arguments and offer room for their further integration. The foundation for such predictions emerge from Owsiak (2012), who uncovers a strong, bivariate relationship between border settlement and joint democracy within dyads (χ² = 274.256, p < 0.001; γ = 0.645). Additionally, Owsiak finds that the majority of contiguous dyad-years analyzed (70.08%) contain either settled borders or jointly democratic dyads, but not both. This latter finding offers a critical opening. The territorial peace argument proposes that conflict within a dyad should be significantly lower when the dyad’s mutual borders are settled (as opposed to when they are unsettled), regardless of the member states’ regime type. The democratic peace advances a similar argument: conflict within a dyad should be significantly lower when both members are democratic (as opposed to when they are not), regardless of whether the dyad members have settled their mutual border. ⁸ In other words, the arguments predict that:

Two observations about these hypotheses merit attention. First, because the territorial peace concerns border settlement and only land contiguous states have borders to settle, the above hypotheses, as well as many that follow, apply only to land contiguous dyads. I relax this constraint later (see Hypothesis 7). Second, these hypotheses examine situations in which dyads contain either settled borders or two democratic states (but not both). They therefore isolate unique predictions from the individual arguments and allow us to evaluate their merits in distinct domains.

Nonetheless, conflict might still arise within democratic dyads or dyads with settled borders. When it does, each argument offers insight about how the dyad should handle those conflicts. The democratic peace predicts that democratic dyads may have conflicts of shorter duration and less severity than non-democratic dyads. Shared norms, opposition groups, and a free press could facilitate a quicker, more credible exchange of information within democratic dyads, which helps them find a negotiated solution more easily than non-democratic states (Fearon, 1995). At the same time, institutional constraints on democratic leaders might force them to concentrate their efforts on negotiation, provide them with space to conduct such negotiations (i.e. mobilizing takes more time), and motivate them to conclude the conflict before public sentiment locks them into a war (via audience costs; see Fearon, 1994). Thus, democratic dyads should rely less on violence and more on negotiation during their conflicts than non-democratic dyads, regardless of whether these dyads have settled their borders.

The foundations of the territorial peace argument contain very similar predictions, albeit with the proviso that the (un)settled status of borders—rather than regime type—creates a shift in conflict behavior. The territorial peace argument rests upon the notion that states learn that force is an appropriate allocation mechanism for handling (border) territorial, but not non-territorial, disputes via violence. We should therefore see a stark change in behavior when comparing pre- and post-border settlement periods. In the pre-settlement period, norms permit the use of violence as a means of managing conflict. After border settlement, however, there are significantly fewer issues that states can appropriately handle with violence; states should therefore shift from violence to more peaceful allocation mechanisms. This line of reasoning points toward the same general conclusion: dyadic conflict episodes should be shorter and less severe after border settlement occurs.

The above discussion therefore anticipates that:

As before, these hypotheses isolate unique domains of the two arguments. Yet there is no reason why the arguments cannot reinforce one another as well. Territorial disputes, for example, might clear salient, contentious issues from a dyad’s foreign policy agenda before it becomes democratic, thereby yielding greater peace. The (perhaps) subsequent emergence of democratic regimes within both dyad members might bring norms, institutions, or actors that then foster an even greater level of dyadic peace. A similar conclusion could be reached if, alternatively, dyads become democratic before border settlement occurs. In both cases, we might expect that:

A second series of hypotheses derives from the relative timing of democratization and border settlement. The territorial peace proposes that border settlement (i.e. resolving certain territorial claims; see Owsiak, 2012) increases the likelihood of observing both peace and joint democracy in a dyad (Gibler, 2012; Owsiak, 2013). Regardless of whether border settlement is an underlying or proximate cause of peace and joint democracy, the argument’s cause–effect set-up must anticipate that border settlement chronologically precedes the emergence of joint democracy (i.e. there is a clear sequencing argument in the territorial peace). This suggests two predictions. First and foremost:

Three observations are worth noting about this hypothesis. First, we could derive corollaries from it. For example, the hypothesis implies that democratic dyads with unsettled borders should be rare. Such corollaries are consistent with the broader hypothesis. Second, although the territorial peace offers a prediction about the chronological ordering of border settlement and the appearance of joint democracy in a dyad, it does not indicate how close (or far) these events are to (or from) one another. Nevertheless, I will discuss the relative “distance” of these two events to each other in the sections that follow.

Finally, the democratic peace argument may make the opposite prediction. Morrow et al. (2006: 51) review the aims pursued by dyads within MIDs, after which they “speculate that the spread of systems with large winning coalitions [i.e. democracies] was key to the decline of territory as an issue of dispute.” In other words, democratic dyads might have caused territorial claims to subside, which implies that democratic dyads preceded (border) territorial claim settlement. Yet, it is also possible that the authors might instead mean simply that once a democratic dyad exists, it is less likely to revise the territorial status quo militarily (i.e. raise new claims). This latter interpretation could reinforce the above hypothesis (4T), but it need not. For example, a democratic dyad might inherit unsettled borders from past regimes and elect to settle them. If it does this without fighting militarily, then it has not “revised the territorial status quo,” according to Morrow et al. (see also Mitchell and Prins, 1999). It is also clear, however, that, in such cases, border settlement does not precede the emergence of a democratic dyad. Given the ambiguity in this argument, I therefore derive no hypothesis for the democratic peace argument here.

A second prediction generated by the territorial peace’s chronological ordering follows from the first. If dyads settle their borders before both members become democratic and border disagreements constitute a large number of territorial claims (this is true, particularly within the contiguous dyads upon which I am focused here; see Huth and Allee, 2002), then:

The democratic peace argument reaches a similar prediction, albeit via a different route. ¹¹ Bueno de Mesquita et al. (2003) and Morrow et al. (2006) propose that leaders of democratic states are less interested in territorial expansion than non-democratic leaders because they must satisfy a large constituency (i.e. winning coalition) to retain office. The size of this constituency makes it challenging to buy members off with private goods (e.g. money), as each member would receive so little. Consequently, democratic leaders resort instead to a more cost-effective strategy: the provision of public goods—that is, those that benefit both constituency members/non-members alike (e.g. education, civil liberties, or political rights).

Territorial expansion, however, often does not help leaders provide public goods. Gaining territory often yields economic resources (e.g. minerals), but these resources cannot be distributed in a way that makes members of large constituencies sufficiently better off. ¹² Democratic leaders therefore find territorial expansion unattractive (and perhaps irrational). Non-democratic leaders, in contrast, remain accountable to a smaller constituency. Thus, it is more likely that any tangible resources gained by territorial expansion can be distributed to individual constituency members in a way that makes these members marginally better off. Furthermore, non-democratic leaders can retain resources for use in discretionary projects—something democratic leaders do with less ease. Given these institutional incentives, democratic dyads should pursue territorial claims less frequently than non-democratic dyads.

To complicate matters, Bueno de Mesquita et al. (2003: 412–414) introduce further nuance. The above logic focuses on territory with economic resources, such as minerals (e.g. gold, silver, copper, chromium, diamonds) or oil (Huth, 1996: 257). The argument’s logical conclusion is therefore that democratic dyads will experience fewer territorial claims involving economically valuable land. Yet territory can also have strategic value—for example, when it offers a militarily beneficial position (e.g. control of narrow straits, shipping lanes, or principal attack routes, as well as access to the sea; Huth, 1996: 256). This strategic value contributes to national security, a clear public good. Since democratic leaders rely more heavily on public goods provisions, Bueno de Mesquita et al. therefore conclude that:

Finally, the territorial peace argument theoretically applies only to contiguous dyads, as these dyads have borders to settle. ¹⁴ It therefore makes no prediction about the behavior of non-contiguous dyads. ¹⁵ The democratic peace, on the other hand, contains no such spatial restriction. Democratic dyads should treat one another more peacefully regardless of how close or far they are from one another. This suggests a unique domain for the democratic peace, as well as a corresponding prediction:

In order to facilitate the discussion that follows, Table 1 summarizes the hypotheses presented above. I turn now to their empirical evaluation.

OPEN IN VIEWER

Table 1.

Hypothesis summary

Hypothesis	Democratic peace argument (D)	Territorial peace argument (T)
1	Peace without settled borders	Peace without democracy
2	Shorter/less severe conflict without settled borders	Shorter/less severe conflict without democracy
3	Greatest peace with settled borders, joint democracy, and contiguity
4	No expectation	Border settlement before democracy
5	Democracies with fewer territorial claims
6	Democracies with fewer (more) economic (strategic) territorial claims	No expectation
7	Peace in non-contiguous dyads	No peace in non-contiguous dyads

Research design

In order to test the hypotheses outlined above, I conduct a series of logistic regression models. The unit of analysis in these models is the dyad-year during the period 1816–2001. ¹⁶ Using EUGene, I first create one observation for each year that any two given states are both members of the international system (Bennett and Stam, 2000). I then limit the analysis to either contiguous or non-contiguous dyads. For this study, contiguous dyad members share an inland or river boundary (Stinnett et al., 2002). All other dyads are non-contiguous. I split the data in this way for theoretical and empirical reasons. Theoretically, only dyad members that share an inland or river boundary have mutual borders to settle together. ¹⁷ The remaining, non-contiguous dyads are separated by gulfs and seas; for these members, their mutual borders (if we can speak of any) will be determined by the bodies of water that separate them, rather than an international agreement. Thus, non-contiguous dyads do not “settle” their mutual borders—a key component of the territorial peace. Despite this theoretical challenge, I employ a monadic level measure of border settlement in my analysis of non-contiguous dyads. The distinct dyadic measure of border settlement for contiguous dyads and the monadic measure for non-contiguous dyads also empirically preclude me from including all dyads within the same model. I discuss these measures further below.

I employ two dependent variables in my analyses, which measure different types of interstate conflict. First, I measure dispute onset with a dichotomous variable that indicates whether a MID began among a dyad’s members in a given year. A MID occurs when a state threatens, displays, or uses force against another state (Ghosn et al., 2004). To retain consistency with other conflict studies, only dyad members involved in a MID on its first day receive credit for that dispute’s onset. Second, I measure crisis onset with a dichotomous variable that denotes whether an interstate crisis began among a dyad’s members in a given year. A crisis occurs when state actors perceive a threat to basic values, a finite time period in which to make decisions, and an increased likelihood of militarized confrontation (Brecher and Wilkenfeld, 2000; see also Hermann, 1963). Both dependent variables come from EUGene (Bennett and Stam, 2000), and I exclude observations containing an ongoing conflict episode from the analysis.

There are three key independent variables of interest. First, settled borders exist when neighboring states have signed an international agreement that delimits the entirety of their mutual border(s). This dichotomous, dyadic variable comes from Owsiak (2012), and a full list of coding criteria (and examples of each) can be found there. Of note for this study, states that settle their borders cannot unsettle them (e.g. Ecuador and Peru after the 1942 Rio Protocol), even if they leave and return to the international system (e.g. France during the Second World War). Moreover, former colonial states may contest borders delimited by colonial agreements during the first year after their independence, which balances the principles of self-determination and uti possedetis. ¹⁸ The cumulative effect of these coding decisions is to bias the variable against the hypotheses explored in this study (see Owsiak, 2012 for a similar argument). ¹⁹

For non-contiguous dyads, I employ a different measure of border settlement, since these dyad members have no mutual borders to settle. Here, I rely upon a monadic dichotomous measure created by Owsiak (2013): all borders settled. For each dyad member, I use the monadic measure of whether the member settled its borders with all contiguous neighbors prior to the year in question. I then create a dyadic measure of whether both dyad members settled their borders with all contiguous neighbors prior to the year in question. The measure therefore captures whether either dyad member has any external territorial threat in the form of unsettled borders. In so doing, it considers the possibility that the presence of any external, territorial threat causes force to dominate a state’s foreign policy orientation. Although territorial peace proponents would not endorse such a measure (because their theory applies only to contiguous dyads; Gibler, 2012), it is the only option I see for incorporating a similar conceptualization of territorial threat into models assessing conflict in non-contiguous dyads.

The second key independent variable of interest is joint democracy, which indicates whether both members of the dyad in a given year score +6 or higher on the Polity IV autocracy–democracy index (Marshall and Jaggers, 2009). ²⁰ This operationalization conforms to existing conflict research. It also ensures that the variable captures the most stable, strongest democracies—precisely those to which the democratic peace proposition applies best (Russett and Oneal, 2001). Finally, I create a dichotomous interaction term that captures joint democracies with (all) settled borders. This variable denotes dyads within a given year that have settled (all) their borders and whose members each score +6 or higher on the Polity IV index. Including this interaction variable allows me to use the lower-order terms to isolate the individual effects of joint democracy and (all) settled borders (i.e. explore one characteristic in the absence of the other) and the higher-order term to consider integrative effects.

In addition to the variables noted so far, I include a limited number of control variables that can theoretically affect both interstate conflict behavior and border settlement. Dichotomous variables control for whether at least one dyad member is a major state (Correlates of War Project, 2008) or dyad members are allied to one another (Leeds et al., 2002). ²¹ I also include a capability ratio (ln), which measures the ratio of the weaker to the stronger state’s capabilities (Singer, 1987). To account for diminishing effects at extreme values, I use the natural log of this ratio. Finally, I include a variable in each model that captures peace years—or the number of years prior to the current dyad-year during which the dyad experienced no MID (or crisis) onset. This variable accounts for the development of peaceful dyadic relationships. It also provides the foundation for the cubic spline variables that I use to control for temporal interdependence (Beck et al., 1998). ²² To control for the possibility that dyadic observations are not independent of one another, I also cluster the analysis on the dyad. ²³

Finally, certain hypotheses make predictions about territorial claims and conflict episodes. For territorial claims data, I rely on the Provisional Version 1.0 of the Issue Correlates of War (ICOW) territorial claims data (Frederick et al., 2016). For a claim to enter the ICOW dataset, officials of one or more state governments must explicitly claim sovereignty “over a piece of territory that is claimed or administered by another state” (Hensel, 2001: 90). The version of the data I employ includes all territorial claims throughout the world during the period 1816–2001.

I define conflict episodes in two ways: MIDs and international crises. When I examine dyad members’ behavior within these episodes, the unit of analysis shifts to the MID or crisis (respectively). Within each MID (Ghosn et al., 2004), I track three behavioral indicators: (a) MID duration (in days); (b) the severity of MIDs according to a measure created by Diehl and Goertz (2000: Appendix B; range 0–200, where 200 = most severe); and (c) the most severe action taken by either state during the MID (range 0–21, where 0 = no action and 20–21 = war). For crises, I capture a similar set of behavioral indicators collected by the International Crisis Behavior project (Brecher and Wilkenfeld, 2000). These include: (a) crisis duration (in days); (b) the highest level of violence (severity) achieved in the crisis (scale 1–4, where 1 = none and 4 = war); and (c) how the crisis was handled (scale 1–8, where 1 = negotiation and 8 = violence). These variables offer insight into how long dyadic conflict episodes last, how severe they are, and whether states employ tools other than violence when managing them.

Empirical results

I begin the empirical analysis with conflict behavior, since this constitutes the heart of both the democratic and the territorial peace arguments. Table 2 reports a series of multivariate logistic regressions that explore conflict onset—defined here as MID onset—within contiguous dyads. In the table, the primary independent variables convey the information necessary to evaluate Hypotheses 1D/T and 3. For instance, the coefficient on the settled border variable relays the effect of settled borders on the likelihood of a MID, given that the dyad is not democratic (Hypothesis 1T). Similarly, the coefficient on the joint democracy variable reveals the effect of a democratic dyad on the likelihood of MID onset, given that the dyad does not have settled borders (Hypothesis 1D). Finally, the interaction term, joint democracies with settled borders, considers the cumulative effect of joint democracy and settled borders on MID onset (Hypothesis 3). The reference category throughout the table is therefore a non-democratic dyad without settled borders.

OPEN IN VIEWER

Table 2.

Logistic regression of MID onset in contiguous dyads, 1816–2001

	Model 1	Model 2	Model 3	Model 4	Model 5
Time period	1816–2001	1816–2001	1816–2001	1900–2001	1945–2001
Restriction	None	None	No world wars	None	None
Settled borders	−0.4722***	−0.4582***	−0.4640***	−0.6257***	−0.7175***
	(0.1177)	(0.1175)	(0.1211)	(0.1266)	(0.1601)
Joint democracy	0.0405	0.2018	0.2553	−0.0036	0.2167
	(0.3335)	(0.3428)	(0.3287)	(0.3983)	(0.3332)
Joint democracy with settled borders	−0.5745	−0.7459*	−0.7457*	−0.5835	−0.6342
	(0.4242)	(0.4282)	(0.4144)	(0.4705)	(0.4296)
Major state	–	0.0334	−0.0141	0.3242***	0.1226
		(0.1182)	(0.1220)	(0.1177)	(0.1770)
Allied to one another	–	−0.0315	−0.0230	−0.1658*	−0.2244**
		(0.0926)	(0.0938)	(0.0972)	(0.1074)
Capability ratio (ln)	–	0.1317***	0.1398***	0.1475***	0.1598***
		(0.0361)	(0.0359)	(0.0333)	(0.0392)
Peace years	−0.2881***	−0.2855***	−0.2888***	−0.2919***	−0.2981***
	(0.0229)	(0.0226)	(0.0231)	(0.0234)	(0.0264)
Constant	−0.9767***	−0.7828***	−0.7635***	−0.5385***	−0.3105
	(0.1306)	(0.1491)	(0.1511)	(0.1597)	(0.2003)
Observations	16,845	16,845	16,293	14,031	11,053
LR χ2	473.77***	497.58***	491.76***	533.12***	480.19***

Notes: Robust standard errors presented in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01; spline coefficients are not reported in the table.

The models in Table 2 present clear evidence in support of the territorial peace, a conclusion backed by three main findings. First, the settled borders variable remains negative and statistically significant across all model variations, including models that exclude controls (Model 1), contain controls and observations from the entire study period 1816–2001 (Model 2), omit the potentially anomalous World War years (Model 3), and limit the observations studied to two different time periods common to other conflict studies (Models 4 and 5). Much as the territorial peace predicts (Hypothesis 1T), the likelihood of conflict (here, MID) onset is significantly lower in dyads that have settled their mutual borders, even when the dyad is non-democratic. Second, in contrast to the first finding, the joint democracy variable never achieves statistical significance in these models and often contains the wrong sign as well. That is, democratic dyad-years are no more or less likely to experience conflict onset when these dyads’ borders remain unsettled—contrary to the prediction contained in Hypothesis 1D. Finally, the coefficient on the interaction term (joint democracies with settled borders) is also generally statistically insignificant across the models in Table 2.

At first glance, this suggests that joint democracy does not significantly affect MID onset, even during dyad-years in which borders were previously settled (i.e. the interaction term). Such a conclusion, however, would be misguided, as coefficients on interaction terms do not always lend themselves to accurate substantive interpretation (Brambor et al., 2006). I therefore generate a series of predicted probabilities for Models 2–5 of Table 2 to investigate the potential substantive effects of border settlement and regime type (see Online Appendix). When calculating these probabilities, I hold all control variables at their mean values.

Figure 1 displays the predicted probabilities for Model 2 of Table 2. ²⁴ The probability of MID onset in a given dyad-year for a contiguous dyad that is not democratic and has unsettled borders is 0.0563 (point estimate; confidence intervals are depicted as lines in the figure). For contiguous dyads that settle their borders—but still remain not democratic—the probability falls to 0.0363 (a 35.5%, statistically significant decrease). ²⁵ In contrast, if borders remain unsettled, but the contiguous dyad is democratic, the likelihood of MID onset in a given dyad-year increases slightly over the initial probability to 0.0680. This increase, however, is not statistically significant. Finally, within dyad-years in which the contiguous dyad has settled its borders and achieved joint democracy, the likelihood of MID onset falls to 0.0214. This is a statistically significant 62% decrease over the initial probability (of 0.0563) and a statistically significant 41% decrease over the probability for dyad-years that contain only settled borders.

OPEN IN VIEWER

Figure 1.

Probability of MID onset in contiguous dyads, 1816–2001 (Model 2, Table 4).

In the end, the figure suggests three conclusions about contiguous dyads. First, border settlement significantly reduces the likelihood of MID onset, even in the absence of joint democracy. This finding is important; not only does it lend strong support to one of the territorial peace’s predictions (Hypothesis 1T), but also, unlike the finding associated with the interaction term, the democratic peace cannot explain it. Second, joint democracy does not significantly reduce MID onset in the absence of settled borders, contrary to the expectation contained in Hypothesis 1D. Critics may argue, however, that there are few dyads with such characteristics—which creates a wide confidence interval, thereby reducing the likelihood of obtaining a statistical finding. I agree with this criticism, and the critical question becomes: why is this so? I return to this point below. Finally, joint democracy does reduce the likelihood of MID onset, but only in tandem with border settlement (i.e. only after a dyad has settled its borders). This suggests strong support for the interactive effect proposed in Hypothesis 3; contiguous, democratic dyads with settled borders experience greater peace than contiguous dyads that lack one or both characteristic. It also implies that the democratic peace is conditional upon border settlement, at least within contiguous dyads.

I repeat the above analysis with international crises as well. Owing to space constraints, I relegate this analysis to the Online Appendix. Nonetheless, the same statistical and substantive results obtain. Border settlement significantly reduces the likelihood of crisis onset in a given contiguous dyad-year, even in the absence of joint democracy, which supports the territorial peace’s Hypothesis 1T. In contrast, joint democracy does not alter the likelihood of crisis onset in a given contiguous dyad-year when borders remain unsettled (contrary to Hypothesis 1D). Finally, an interactive effect emerges as well; contiguous, democratic dyads with settled borders experience a significantly lower likelihood of crisis onset than contiguous dyads that lack one or both characteristics—in support of Hypothesis 3.

Conflict behavior

Once a conflict occurs, how do contiguous dyads behave? Table 3 considers this question under three sets of conditions: contiguous, non-democratic dyads before and after border settlement (columns 1–3 of Table 3); contiguous dyads with unsettled borders before and after they become democratic (columns 4–6 of Table 3); and contiguous, democratic dyads with settled borders (the final column of Table 3). These conditions both isolate predictions within the unique domains advanced by the individual arguments (Hypotheses 2T/D) and consider potential integrative effects.

OPEN IN VIEWER

Table 3.

Changes in crisis and MID behavior within contiguous dyads

	Non-democratic dyads (territorial peace predictions)			Dyads with unsettled borders (democratic peace predictions)			Democratic dyads with settled borders
	Before settled borders (average)	After settled borders (average)	Difference (t-value)	Before joint democracy (average)	After joint democracy (average)	Difference (t-value)	(Compared with non-democratic dyads after settled borders, column 2)
Militarized interstate disputes
MID duration (days)	202.76	129.29	3.44***	202.76	192.67	0.12	58.00*
Severity of MID (Diehl/Goertz)	68.64	51.49	4.55***	68.64	74.72	−0.42	35.06**
Highest action (either disputant)	14.52	13.83	2.42***	14.52	14.83	−0.31	13.21
International crises
Crisis duration (days)	266.24	170.13	2.98***	266.24	347.60	−0.72	181.30
Severity of violence	2.62	2.07	3.22***	2.62	3.40	−1.38*	1.40**
How crisis is handled	6.23	5.68	1.75**	6.23	7.40	−1.16	3.40**

*

p < 0.10, **p < 0.05, ***p < 0.01.

The first few rows of Table 3 examine behavior within MIDs—particularly MID duration and severity. As the table reveals, MIDs are significantly shorter in contiguous, non-democratic dyads after border settlement. In contiguous, non-democratic dyads, a MID lasts about 203 days when the dyad has unsettled borders. If the borders are settled, however, the MID’s duration is significantly shorter—roughly 129 days (a 36% decrease). I observe, however, no similar, significant decline after joint democracy appears in contiguous dyads with unsettled borders. As above, MIDs last an average of 203 days if the dyad is not democratic and has unsettled borders. If the dyad becomes democratic, MID duration declines 5% to 193 days—a statistically insignificant decrease.

Similar findings emerge with respect to MID severity. Using the Diehl and Goertz (2000) severity measure, the mean level of MID severity falls from 68.64 to 51.49 in contiguous, non-democratic dyads between the pre- and post-border settlement periods. These figures suggest that the average MID in both periods does not involve fatalities; rather, the severity depends entirely on how hostile actors behave in their dispute (Diehl and Goertz, 2000: Appendix B). The remaining statistic on MID behavior confirms this last point. It measures the most hostile action taken by either disputant during the MID. As the table shows, I find that severity significantly decreases in contiguous, non-democratic dyads between the pre- and post-border settlement periods—from a value of 14.52 (corresponding to an action close to occupying territory or seizure) to a value of 13.83 (corresponding to an action closer to a blockade). It seems, therefore, that border settlement changes the behavior of non-democratic states.

In contrast, the appearance of joint democracy does not seem to affect contiguous dyads with unsettled borders significantly. In fact, using two different measures of severity, I find that the emergence of joint democracy may even increase MID severity in contiguous dyads with unsettled borders—although this increase is never statistically significant. Using the Diehl and Goertz (2000) measure, for example, the average severity of a MID for a contiguous dyad with unsettled borders rises from 68.64 to 74.72 when dyad-years are non-jointly and jointly democratic respectively. As before, these MIDs do not seem to involve fatalities, but reflect the hostility with which dyad members interact. This increased hostility appears again when I consider the most hostile action taken during the MID. For dyads with unsettled borders, I observe an increase in the average level of hostility in a MID when dyad-years are non-jointly and jointly democratic—from 14.52 to 14.83 respectively (both corresponding to an action close to occupying territory or seizure).

A similar pattern of behavior occurs within international crises as well, the analysis of which appears within the last rows of Table 3. For contiguous dyads that contain neither joint democracy nor settled borders, the average crisis duration is 266 days. This falls to 170 days (a 36% significant change) if the dyad remains non-jointly democratic and settles its borders, but increases to 348 days (a 31% insignificant change) if the dyad persists with unsettled borders and becomes jointly democratic. Yet border settlement does more than shorten crises; it may also make them less severe. When looking at the “most intense use of violence as a primary management technique” (i.e. severity of violence) and the highest (or most violent) crisis management method used by crisis actors (i.e. how a crisis is handled), similar conclusions emerge (Brecher and Wilkenfeld 2000). Contiguous, non-democratic dyads that settle their borders experience significantly less severe crises than similar dyads with unsettled borders. Contiguous, jointly democratic dyads with unsettled borders, however, experience no significant change in crisis severity over similar dyads that are non-jointly democratic. Thus, Hypothesis 2T receives consistent strong support, while Hypothesis 2D does not—for I repeatedly find that border settlement changes the conflict behavior of non-democratic dyads, while joint democracy does not significantly alter the conflict behavior of dyads that lack clearly delimited borders.

As a final consideration, the last column of Table 3 considers the behavior of contiguous, democratic dyads after border settlement occurs. In this last column, I compare contiguous, democratic dyads with settled borders to undemocratic dyads with settled borders. I do this because border settlement (columns 1–3) consistently alters non-democratic dyads’ behavior, while a change in regime type does not do the same for dyads with unsettled borders (columns 4–6). In effect, this allows me to ask: given that border settlement occurred, does the behavior of contiguous dyads change further after becoming democratic? The answer seems to be yes, in further support of Hypothesis 3. In dyads with settled borders, the emergence of joint democracy corresponds with shorter MIDs (from 129 to 58 days), less hostile MIDs (from 51.49 to 35.06 on the Diehl and Goertz 2000 scale), less frequent use of violence as a primary management technique in international crises (from 2.07 to 1.40; i.e. roughly from minor to no clashes), and less frequent use of violence overall in international crises (from 5.68 to 3.40; i.e. roughly from non-violent military acts to mediation). I therefore once again conclude that joint democracy promotes peace within contiguous dyads, but only after border settlement occurs.

Territorial claims

The above analyses offer insight into conflict behavior, but they omit one important question: might democratic dyads behave more peacefully toward one another because they have cleared the most contentious (i.e. border territorial) issues from their foreign policy agenda before becoming democratic? This is what the territorial peace predicts (Hypothesis 4) and the empirical record supports. Owsiak and Vasquez (2016) analyze contiguous dyads to determine when they first experience joint democracy or settle their borders. Of the 112 contiguous dyads that experience border settlement and at least one year as a democratic dyad, only eight cases exist where joint democracy precedes border settlement. A similar trend appears within non-contiguous dyads as well. In roughly 71% of these dyads, dyad members settle all of their borders (with all contiguous neighbors) before becoming jointly democratic (and this may understate the trend; see Owsiak and Vasquez, 2016). Such findings strongly support Hypothesis 4; in the vast majority of cases, border settlement precedes joint democracy. Indeed, observing joint democracy before border settlement in contiguous dyads is a relatively rare event.

To investigate further the possibility that democratic dyads have cleared contentious issues from their foreign policy agenda before becoming jointly democratic, Table 4 examines the presence of territorial claims in democratic and non-democratic contiguous dyads. Of the contiguous dyad-years in which a given dyad is not jointly democratic, 71.44% contain no territorial claim; the remaining 28.56% contain such a claim. These figures shift when we consider jointly democratic dyad-years. Of the contiguous dyad-years containing two democratic states, 80.31% contain no territorial claim. The remaining 19.69% of jointly democratic, contiguous dyad-years do contain a territorial claim. The difference between these percentages is statistically significant (χ² = 79.598, p < 0.001) and moderately negative (γ = −0.2399; a.s.e. 0.026). These results offer support to both the democratic and territorial peace’s prediction, as embodied by Hypothesis 5. Democratic, contiguous dyads are significantly less likely than their non-democratic counterparts to have a territorial claim.

OPEN IN VIEWER

Table 4.

Bivariate relationship of territorial claims and democracy in contiguous dyads

Territorial claim	Jointly democratic dyad (column percentage)
	No	Yes
No	12,057(71.44%)	1836(80.31%)
Yes	4821(28.56%)	450(19.69%)
Observations		19,164
χ2 (1 d.f.)		79.598 (p < 0.001)
γ (a.s.e.)		−0.2399 (0.026)

The democratic peace, however, makes three additional predictions. First, it predicts that non-contiguous democratic dyads should also experience fewer territorial claims. Bivariate analysis (not shown) does not support this prediction. When looking at either all dyads (both contiguous and non-contiguous) or non-contiguous dyads alone, democratic dyads are somewhat more likely to experience territorial claims (not shown; γ = 0.0632 and 0.1765 respectively). Second, the democratic peace predicts that democratic dyads should have more territorial disputes over strategic territory and fewer disputes over economically valuable territory (Hypothesis 6). This, however, also receives no empirical support. When analyzing all dyads years containing a territorial claim, I find that democratic dyads are more likely to experience territorial disputes over economically valuable territory (χ² = 36.395, p < 0.001; γ = 0.2011; a.s.e. 0.033) and less likely to experience disputes over strategically valuable territory (χ² = 28.597, p < 0.001; γ = −0.1736; a.s.e. 0.032)—the exact opposite of what the democratic peace predicts (see Online Appendix). Such findings fail to support Hypothesis 6. They also suggest a need to reconsider the scant theoretical arguments that link democratic dyads to territorial claims.

Non-contiguous dyads

As a final step, I consider conflict behavior within non-contiguous dyads. The territorial peace argument claims not to apply to these dyads. Nonetheless, I find it important to create a model for non-contiguous dyads that mirrors that for contiguous dyads as closely as possible. I therefore construct a measure of whether both members of the dyad have settled their borders with all of their neighbors. This seems to be a fair way to capture the territorial peace logic, since any external threat (i.e. any unsettled border) should promote an aggressive foreign policy toward the sources of that threat (i.e. neighbors), which could then prioritize force as an allocation mechanism within a state’s foreign policy.

With this in mind, Table 5 presents models of both MID and crisis onset in non-contiguous dyads. Regardless of whether I exclude (Models 1 and 4) or include (Models 2, 3, and 5) the border variables, when borders remain unsettled, joint democracy always significantly reduces the likelihood of conflict onset (see the lower order term). Yet the converse also seems to be true, for the models suggest that non-democratic dyads that settle all their borders also decrease their likelihood of conflict onset (see lower order term). Finally, unlike in contiguous dyads, there appears to be no interactive effect within non-contiguous dyads; that is, non-contiguous, democratic dyads that settle all their borders do not significantly reduce their likelihood of conflict onset further than when they possessed only one of these characteristics (see interaction term).

OPEN IN VIEWER

Table 5.

MID and crisis onset within non-contiguous dyads

	MID onset			Crisis onset
	Model 1	Model 2	Model 3	Model 4	Model 5
Time period	1816–2001	1816–2001	1816–2001	1919–2001	1919–2001
Restriction	None	None	No world war years	None	None
All borders settled	—	−0.1856	−0.3173**	—	−0.9126***
		(0.1167)	(0.1272)		(0.2433)
Joint democracy	−1.0347***	−0.6910***	−0.6107**	−2.0200***	−1.4144***
	(0.1506)	(0.2422)	(0.2463)	(0.4325)	(0.5450)
Joint democracy with	—	−0.4627	−0.3754	—	−0.4914
all borders settled		(0.3956)	(0.4111)		(1.1564)
Major state involved	3.4593***	3.1029***	3.0013***	3.8132***	3.4022***
	(0.1289)	(0.1295)	(0.1420)	(0.2381)	(0.2377)
Allied to one another	0.5475***	0.6531***	0.7434***	0.9435***	1.1659***
	(0.1235)	(0.1400)	(0.1516)	(0.2206)	(0.3001)
Capability ratio (ln)	0.3039***	0.2225***	0.2715***	0.3927***	0.3573***
	(0.0307)	(0.0334)	(0.0369)	(0.0599)	(0.0620)
Peace years	−0.2824***	−0.2587***	−0.2618***	−0.2892***	−0.2462***
	(0.0228)	(0.0244)	(0.0259)	(0.0492)	(0.0580)
Constant	−5.0124***	−5.0023***	−4.8781***	−6.6173***	−6.4135***
	(0.1559)	(0.1589)	(0.1609)	(0.2274)	(0.2573)
Observations	654,124	514,197	503,973	487,976	398,721
LR χ2	1156.37***	1233.26***	1004.44***	511.83***	397.76***

Notes: Robust standard errors presented in parentheses; *p < 0.10, **p < 0.05, ***p < 0.01; spline coefficients are not reported in the table.

As before, I also consider the substantive effects derived from the models that appear in Table 5 (see Online Appendix). These results reveal that joint democracy drives peace in non-contiguous dyads—the exact opposite of the finding for contiguous dyads. Non-contiguous, non-democratic dyads do not substantively reduce their likelihood of conflict onset by settling all outstanding borders. In contrast, dyads with at least one unsettled border do significantly reduce their likelihood of conflict onset by becoming democratic. Finally, non-contiguous, democratic dyads that have settled all borders are no less likely to experience conflict onset than those possessing only one of these characteristics. These findings hold across multiple measures of conflict (MIDs and crises). They also extend to conflict duration and severity as well (see Online Appendix). I therefore conclude that joint democracy brings peace in non-contiguous dyads, including a reduced likelihood of conflict, as well as shorter and less severe conflict episodes when conflict occurs—in strong support of Hypothesis 7. Border settlement does not bring such effects in non-contiguous dyads, either independently or in conjunction with joint democracy. Such results, when aggregated with the other results presented in this section, suggest that both the territorial and democratic peace contribute to peace, but these effects vary by domain (non/contiguous) and can interact.

Conclusion

At the outset of this study, I asked whether the democratic and territorial peace could both be accurate. Answering this question requires that we eschew the tendency to pitch these as competing arguments. If the arguments do not compete, then we must think more carefully about how they can be integrated. Two avenues seem particularly fruitful. First, the two arguments might reinforce one another (within the same domain)—that is, a possible interactive effect. Previous research rarely considers such an effect, but where it does, this work aims to build upon it (e.g. see Park and James, 2014). Second, the democratic and territorial peace arguments might operate in somewhat unique domains (Most and Starr, 1989). This would allow each argument to exert an independent effect for which the other cannot account. In addition, it takes us beyond merely thinking in terms of “interaction effects.”

I pursue both avenues in this work, in an attempt to set the foundation for further theoretical advances. On the one hand, I propose two types of unique domains. First, the territorial peace argument claims not to apply to non-contiguous dyads, while the democratic peace asserts no such claim. The democratic peace therefore clearly operates within a spatial domain where the territorial peace does not. Second, within the same spatial domain (i.e. contiguous dyads), each argument can operate in the absence of the other. Non-democratic, contiguous dyads, for example, can settle their borders, and the territorial peace makes predictions about these dyads’ behavior. Similarly, contiguous dyads that lack settled borders can become democratic, and the democratic peace offers predictions about these dyads’ behavior. Isolating these unique domains therefore sheds light on the relative contributions the individual arguments might make in an integrated theory. On the other hand, I also consider the possibility that the two arguments reinforce one another. Just because, for example, contiguous dyads generally settle their borders before becoming democratic does not mean that joint democracy can exert no effect on dyadic relations. It may instead be the case that both characteristics yield greater peace than either can alone.

After developing and testing predictions that cover a broad range of characteristics and behaviors—including the timing of border settlement relative to democratization; the emergence of territorial claims, militarized disputes, and interstate crises; and dyadic behavior during conflict episodes—I uncover significant support both for the idea that the arguments operate in distinct domains, as well as the claim that the two arguments reinforce one another. Within contiguous dyads, the territorial peace seems to exert a stronger effect than democracy. Border settlement almost always precedes joint democracy in contiguous dyads; this explains why contiguous democracies have fewer territorial claims. Furthermore, even when a dyad remains non-democratic, border settlement reduces the likelihood of conflict onset, and both shortens and decreases the severity of conflicts that occur. The appearance of joint democracy, however, has no similar, significant, pacific effect in the absence of settled borders.

This, however, does not imply that democracy plays no role in promoting peace. In fact, two influences emerge. First, within contiguous dyads with settled borders, joint democracy further decreases the likelihood of conflict onset, further shortens militarized disputes, and further reduces the severity of militarized disputes and crises. This is evidence of a clear interactive effect. Second, within non-contiguous dyads, the democratic peace (rather than the territorial peace) dominates. When a non-contiguous dyad contains at least one unsettled border (with some neighbor), becoming democratic significantly reduces the likelihood of conflict. Border settlement, however, exerts no substantive effect—either in the absence or presence of democracy (i.e. an independent or interactive effect).

Such findings suggest a complementary relationship between the democratic and territorial peace arguments—one not yet explained by existing theory. Future work might therefore build upon the foundation laid here along two fronts. First, we need better theoretical arguments for how the territorial and democratic peace fit together. The work that attempts such arguments makes predictions that are unclear, underdeveloped, or unsupported empirically. For example, although selectorate theory predicts that democracies will experience greater territorial claims involving strategic territory and fewer territorial claims involving economically valuable territory (e.g. see Bueno de Mesquita, 2003:412–414; Morrow et al., 2006), the opposite seems to be empirically true (see Online Appendix). This suggests a need to revisit how issues and regime type interact with one another. Second, democracies may settle land borders only to move to maritime claims (Mitchell and Prins 1999). We could therefore use a better understanding of how non-border territorial claims might be incorporated into these two arguments. Pursuing these avenues of research may not only give us greater insight into the limits and interactions of the two theories, but also help us devise additional empirical tests of them as well.

Nonetheless, although work might continue, the results of this study are clear. The territorial peace plays a strong role in contiguous dyads, while the democratic peace generates an interaction effect in contiguous dyads and dominates non-contiguous dyads. This alone offers an important contribution. Yet the study’s importance extends beyond this in three ways. First, its conclusions help us make sense of empirical puzzles like Colombia and Peru (discussed at the outset of this work). The democratic peace cannot explain the 50 years during which this dyad remained at peace despite the absence of democratic norms and institutions. Incorporating the territorial peace, however, helps us understand not only why this occurred, but also why democratization followed border settlement. Second, it contributes to a number of prominent, ongoing research programs, including the democratic peace (see Chan, 2012 for a review), territorial peace (Gibler, 2012), steps-to-war (Vasquez, 2009), and issue-based approach (e.g. see Hensel, 2001). Finally, the findings carry implications for policy-makers. Many policy-makers operate under the assumption that promoting democratization will help bring peace, but this study sounds a note of caution. The pacific effects of democracy within contiguous states seem conditional upon the status of international borders; if the goal is to build a more peaceful world, then this study suggests that we should think as hard about the settlement of territorial conflicts as we do about the promotion of democracy.

Supplemental Material

Supplemental Material