International Influences on the Survival of Territorial Non-state Actors

Abstract

Territorial contenders are political entities that control populated territories but lack recognition as sovereigns. They pose existential threats to their host states by reshaping recognized borders and generating zones of contested authority. States have strong incentives to eliminate them, and yet they persist—developing countries host an average of three territorial contenders within their borders. Understanding why territorial contenders survive and how they die is a critical puzzle in the study of state making. International forces offer important, if overlooked, explanations for these seemingly domestic processes. First, we argue that international rivals perpetuate the existence of territorial contenders by undermining a state's ability to reintegrate them through peaceful negotiations or by force. Secondly, the international human rights treaty regime provides a mechanism by which territorial contenders can galvanize support from potential allies, increasing a state's willingness and ability to resolve these disputes through peaceful reintegration processes.

Governing territory has long been a central requirement to be considered a sovereign state. It remains so today, with states increasingly investing in the structures and symbols of territorial authority (Carter and Poast 2015; Simmons and Kenwick forthcoming). Yet, the inhabitable surface of the earth is fixed, and there has always been competition for territory among political actors. The locus of this competition has shifted. In previous centuries, interstate competition for territory was a primary component of both territorial competition and state-making processes. More recently, international borders have become increasingly fixed and interstate conflict less frequent (Atzili 2007; Fazal 2007; Pettersson, Högbladh, and Öberg 2019; Zacher 2001). Now, the most common challenges states face emanate from within their borders, most dramatically, from rebel groups that control territory and govern it in competition with the sovereign state (Arjona, Kasfir, and Mampilly 2015; Huang 2016a; Mampilly 2011; Stewart 2018). However, rebel groups are only one type of non-sovereign actor threatening sovereign states' control of territory. Consequently, we focus on the broader category of territorial contenders (TCs)—the set of all political entities that wrest control of populated territory within the boundaries of recognized members of the international system, regardless of whether this involves civil war.

TCs challenge the sovereignty of their host states, generating zones of contested authority and often fostering calls for independent statehood or even attempting to usurp their hosts entirely. Given these threats, host states have extraordinary incentives to eliminate TCs quickly. There is no stable equilibrium so long as they persist. Despite strong incentives to eliminate them, TCs are plentiful. There are three TCs for every state in the developing world (Lemke and Crabtree 2020, 523), and many survive for decades before reintegration or promotion to sovereign statehood. Which factors determine whether TCs survive and how they die?

We argue that international factors both enable TCs and constrain the strategies host states use to eliminate them. We focus on two such factors. The first draws upon theories of subversion in linking the presence of international rivals with weakened territorial authority (see, most prominently, Lee 2018; Lee 2020). Our point of departure is twofold. First, rather than empirically examining the absence or weakening of state authority, we focus on the presence and persistence of alternate forms of authority in the form of TCs. This makes explicit that subversion operates not only by creating “ungoverned spaces” or zones of anarchy, but also by cultivating zones of competing state and non-state authority. Secondly, we explore two specific channels through which subversion operates. The first centers upon a military logic: the presence of an external benefactor allows TCs to stave off forceful reintegration by the state. The second mechanism is more subtle, if not less pernicious: foreign rivals reduce the likelihood that TCs will be peacefully reintegrated by their hosts and, in so doing, ensure the persistence of contentious interactions between the two.

Next, membership in the international human rights treaty regime constrains the strategies states use to eliminate TCs. International treaties serve as hand-tying mechanisms with potential to meaningfully alter their signatories' actions (Morrow 2007; Simmons 2000). Within the human rights domain, they attempt to sanction the use of repressive strategies states could otherwise use to forcibly eliminate TCs and repress the populations they contain. The result is that host states must choose between abrogation of their treaty obligations or more peaceful strategies of eliminating TCs, on average, pursuing the latter.

We test these expectations using novel data on TCs in the developing world, uncovering strong empirical support. We show that the presence of rivals abroad reduces the risk TCs will be reintegrated by their hosts, especially through peaceful means. Human rights treaties, by contrast, increase the likelihood TCs will be quickly and peacefully reintegrated. These findings contribute to the literature on state making by accounting for the survival and death of the world's preeminent, if often unrecognized, non-sovereign state-making entities, with their presence accounting for a significant number of the world's civil wars and failed states. That our findings are strongest in predicting the outcomes of peaceful reintegration processes underscores the need for greater attention to non-violent conflict-resolution processes in the study of non-state actors. We also contribute to the human rights literature by identifying a hitherto-unrealized channel through which international treaties foster peace. Existing studies have focused largely on their impact for aggregate levels of repression, or on repression within already-violent conflicts (for example, Conrad and Ritter 2013; Fariss 2018; Hafner-Burton and Tsutsui 2005; Hathaway 2002; Morrow 2007; Simmons 2009). By focusing on TCs, we home in on the constraining effect treaties have in the particularly crucial context where the incentive to repress exists but only sometimes combusts into active violence.

Theoretical Arguments

Interactions between states and TCs are consequential. Tilly (1990, 96) went so far as to refer to “attacking and checking competitors and challengers within the territory claimed by the state” as state making itself. Perhaps less clear is that these interactions are costly for both states and TCs. For host states, TCs present security risks,¹ damage international prestige, and cost vast sums of lost revenue in the form of uncollected rents and lost access to natural resources. TCs exist in an even more precarious position—nearly half are forcibly reintegrated, that is, conquered, by their hosts, and only about 10 per cent ever achieve widespread international recognition (Lemke and Crabtree 2020). Leading a TC is also a risky proposition, over 13 per cent of such leaders are killed or captured by state forces.

Given these costs, the surplus of TCs is puzzling. Analogizing these interactions to interstate war, canonical bargaining models suggest that mutually incurred costs would motivate would-be TCs and hosts to come to peaceful arrangements, preventing the emergence of TCs to begin with. In such bargains, the dissident group that would have otherwise seized territory and become a TC remains non-territorial and the state retains control of all its territory. Yet, war is rare and TCs are not. That so many emerge and persist means something must prevent states and TCs from easily resolving their disagreements. This puzzle is resolved at least partly by recognizing what distinguishes interactions between hosts and TCs from those between sovereign states. Relative to the interstate scenario, we expect host states to place a greater value on resolving disagreements with TCs by force, rather than through peace.

States can seek three outcomes as a means of resolving their TC problem: forceful reintegration, peaceful reintegration, and mutual cohabitation. We assume that states generally prefer forceful reintegration. If they forcefully reintegrate the TC's territory, they end the threat posed to them. Although they pay costs in conquering the TC, they also send a message to other would-be TCs: territorial challenges will result in annihilation.² We further assume that after forceful reintegration, peaceful reintegration is the second preference. Peaceful reintegration involves a negotiated settlement whereby the TC surrenders its control of territory and returns to normal political relations within the state. Surely, TCs only make such agreements if they receive (likely substantial) concessions. Peaceful reintegration might seem attractive because it avoids the costs of conquering the TC. However, it invokes a different cost by also sending a message to would-be TCs: territorial challenges will be met with acquiescence. Thus, peaceful reintegration might involve great downstream costs. Finally, we assume that merely tolerating the existence of a TC is the state's lowest preference. The TC's tolerated persistence denies the state resources and incentivizes the emergence of more TCs because that toleration is easily interpreted as weakness.³

This preference ordering allows us analytic traction in understanding why TCs die more often by forceful than through peaceful reintegration,⁴ but it begs the question of why states fail in their bellicose pursuits and why they might nevertheless sue for peace. As we elaborate in the following, international actors can affect conflict processes by buttressing a TC's ability to combat the state militarily, and their involvement can either increase or reduce the problems of uncertainty that undermine diplomatic negotiations.

Rivalry and Subversion

For many, international security threats are beneficial for state centralization and development. Real or perceived threats from abroad allow states to increase their extractive capacity (Thies 2004). When rivals challenge the territorial integrity of the state, opposition forces may be less willing or able to challenge incumbents, allowing leaders to centralize the state apparatus (Gibler 2012). States with interstate rivals are more likely to commit mass killing against subsets of their civilians (Uzonyi 2018). Even potentially disloyal militaries are more likely to subjugate themselves to civilian rule when their attention is directed toward external, rather than internal, threats (Desch 2001). According to this logic, states are built from the outside-in, with rivals being instrumental to the process.

Yet, rivals also have more deleterious effects on state making—as sources of subversion attempting to directly undermine their adversaries from within. Rivals can, and often do, directly support rebel and terrorist organizations fighting their hosts, particularly when the latter reside in peripheral or trans-border regions (Findley, Piazza, and Young 2012; Salehyan 2007). In the extreme, they may directly intervene in civil wars and prolong their duration, exerting maximal costs on their hosts (Cunningham 2010).

These approaches presuppose the existence of armed hostilities in the subverted host state, which is only sometimes the case in the study of TCs. Lee (2018; Lee 2020) recognizes this problem and generates a broader theoretical framework that directly links the presence of interstate rivals with incomplete spatial governance, regardless of whether violence has occurred. In addition to direct military, diplomatic, and economic pressures, rivals seek to undermine a state's authority within its own territory. They do so by exploiting zones of weak or contested sovereignty. When non-state actors reside in this territory, it may mean providing these groups with diplomatic, material, or other forms of support. This concerted attempt at subversion is intended to “hollow out” host states by denying them access to critical resources necessary for state building and development (Lee 2018, 290).

However, it is not enough for rivals to foster instantaneous challenges to state authority; they also have a direct interest in perpetuating this situation indefinitely. How do they do this, and how, specifically, does this condition the relationship between TCs and their hosts? We argue that rivals perpetuate the existence of TCs through two channels, each of which undermines the strategies states typically use to eliminate them. The first relates to forceful reintegration, or the conquest of TCs. Doing so, of course, requires hosts to overpower TCs militarily, but this process becomes costlier when they are supported from abroad. For example, Chad faced a series of TCs operating out of the country's north and east. These TCs were supported financially, militarily, and sometimes diplomatically by Libya. Libya and Chad were rivals, owing to the former's ambitions to control northern Chad. Libya's support of Chadian TCs was so important that Chad was unable to defeat the former government when it reconstituted itself as a TC across three northern provinces in the 1980s because of Libyan support to that TC (Buijtenhuijs 1998).⁵

An alternative strategy for eliminating TCs is bringing them to the negotiating table and agreeing to peaceful reintegration. In so doing, TCs typically relinquish their territorial authority in exchange for political or economic concessions. Rivals have incentives to undermine these negotiations through whatever means available. This may include threats to withdraw future support if TCs come to the negotiating table at all or promises of future resources to offset the concessions offered by a host.

Consider the TC in the Gaza Strip, governed by Hamas since 2006. The lack of successful diplomatic negotiations between Israel and Hamas is overdetermined, but one serious impediment to such negotiations is Iranian support for Hamas and independent Gaza. Iran (among others) supplies Hamas with financial, military, and diplomatic support (Hroub 2006). Without that support, Gaza would not be able to persist as a TC. Were Hamas to enter negotiations with Israel for a reintegration of Gaza into Israel, Iranian support would likely cease. Without that support, the TC could not persist and then Israel may not need to negotiate; they could more easily invade. The rivalry dynamic influencing this state–TC dyad effectively eliminates the possibility of peaceful reintegration.⁶

Beyond direct attempts to sabotage negotiations, the presence of third parties exacerbates uncertainty, as neither the state nor the TC can anticipate ex ante how much foreign opponents will intercede in relations between the state and the TC once peaceful reintegration is under way. Host states may be reticent to engage in negotiations if they cannot trust a TC to sever ties to rivals aboard, which might allow them access to secret weapons caches. Conversely, for TCs, the typical commitment problems associated with disarmament are compounded by concerns that foreign support might (or might not) evaporate when they sue for peace. Stated generally, our expectations of how rivalry influences TC survival motivate the following hypotheses:⁷

The Constraining Effect of International Treaties

Treaties are a second source of international pressure affecting TC life and death. Early work on the efficacy of treaties oscillated between optimism and pessimism about their impact on state behavior (Downs, Rocke, and Barsoom 1996). While a consensus has remained elusive, theoretical and empirical research provides grounds for expecting that international treaties are more than mere “scraps of paper” (Leeds, Long, and Mitchell 2000; Morrow 2007; Simmons 2000).

Like the broader literature on treaty compliance, there is skepticism about the efficacy of human rights treaties specifically. Several factors underscore this skepticism, chiefly, that membership in many treaties is so ubiquitous as to undercut any semblance of effectiveness—many of the world's most repressive regimes are signatories to treaties whose terms they openly flout without apparent repercussions (Hathaway 2002). Hafner-Burton and Tsutsui (2005, 1378) push these arguments further, arguing human rights treaties will be inefficacious at best and actively counterproductive at worst by providing governments “a shield for increasingly repressive behaviors after ratification, as treaty ratification confers on them human rights legitimacy and makes it difficult for others to pressure them for further action.”

Nevertheless, while critics of human rights law often point to a lack of enforcement, the mechanisms driving compliance often take root domestically and as the result of pressures brought by a variety of actors. Signing human rights treaties creates focal points around which political actors and social movements mobilize, orient expectations, and hold governments accountable. Whether through altering the political agenda, sharpening judicial challenges, or mobilizing dissent, a wide body of empirical research shows that human rights treaties reduce repressive behavior by states (Conrad and Ritter 2013; Fariss 2018; Simmons 2009).

At first glance, TCs might seem unlikely beneficiaries of the international human rights regime, if only because eliminating these actors surely stands at the forefront of any host state's national security agenda. Simmons (2009, 16), otherwise an optimist on treaty efficacy, nevertheless argues: “the more a treaty addresses issues clearly related to the ability of the government to achieve its central political goals, the weaker we should expect the treaty's effect to be”—even more so if hostilities with TCs have erupted into civil war, the single best predictor of where repression will occur (Hill and Jones 2014). In these cases, it would seem the constraining impact of treaties would be swept away by larger incentives to repress.

Nevertheless, the inherent features of TCs actually make them particularly well suited to benefit from human rights treaties. TCs are cohesive, organized, and control territory. Cohesion and organization allow them to more effectively transmit information about abuse and repression to external actors and prospective allies, who may themselves pressure host states. Territorial control improves TCs' ability to grant access to the areas where abuse is likely to occur and the individuals who could attest to having experienced or witnessed abuse in the past. This direct access allays concerns about credibility and objectivity by granting access to private actors who can more credibly bring evidence to international allies and advocates, including organizations like Amnesty International and Human Rights Watch, who can then levy normative pressures against repressive host states (Hafner-Burton and Tsutsui 2005, 1385; Keck and Sikkink 1998).⁸

These features make TCs broadly effective in transmitting information regardless of whether their hosts are signatories to human rights treaties, but the latter would improve the likelihood this information is put to good use. While activist groups and organizations are broadly interested in uncovering evidence of repression, they have a particular interest in finding and broadcasting evidence of treaty violations, which are often central in reports they release. Evidence of treaty violation expands the number of prospective international allies willing to mobilize on behalf of repressed or abused groups. In addition to a normative aversion to repression, however, other states may be equally or perhaps more motivated by maintaining good relations with states that value treaty compliance. Repression in states that are deeply embedded in the international treaty system—that is, states that have signed and ratified many human rights treaties—represents not only violations of human rights, but also a direct threat to the maintenance of international law and order more broadly. Domestically, while nationalists may support repression and activists may oppose it, treaties carry weight with moderates who recognize TCs as a security threat but may be concerned with the deleterious ramifications of treaty abrogation.

In addition, after being armed with information on treaty violation, allies can launch more effective investigations and challenges against host states that attempt to eliminate TCs through violent repression. When governments get away with repression, they often do so by concealing their actions and destroying evidence of abuse (Davenport and Ball 2002; Reuning, Kenwick, and Fariss 2020), and human rights treaties can lack bite when combatting forms of repression that are not easily observed (Lupu 2013a). In the case of TCs in signatory countries, the combination of territorial access and information searches by international organizations can make these attempts at concealment more difficult and raises the risk that treaty violation will be observed and subsequently punished.

In short, TCs are uniquely well positioned to extract the greatest possible impact from human rights treaties. Their ability to provide access and information on human rights abuses allows them to form or solidify alliances with domestic and international actors, and gives greater weight to challenges to host states.⁹ The resulting challenges are variegated, with international law inspiring enforcement through decentralized and creative mobilization strategies (Dancy and Fariss 2017, 15). Drivers of compliance may take the form of formal or informal sanctions from international actors, pressures from social movements, or challenges brought within domestic legal systems.

How does this impact the life and death of TCs? We anticipate that human rights treaties will not increase the lifespan of TCs per se, but instead condition whether they are terminated via forceful or peaceful reintegration into the host state. Forceful reintegration seldom occurs through conventional warfare alone and typically entails some degree of armed repression of civilians, be it through political imprisonment, torture, one-sided killing, or other forms of abuse. Applying these strategies is not only costlier for signatory host states, but may also backfire if they draw international attention to the plight of TCs. In contrast, strategies of peaceful reintegration not only reduce the risk of sanctions, but may also attract positive reinforcement from other members of the treaty regime eager to support the peaceful resolution of potential powder kegs of violence and repression. Cast in bargaining terms, human rights treaties also function as a costly signal by which states make clear(er) their commitment to human rights, reducing uncertainty and (at least partly) assuaging concerns the TC's peaceful disbandment would be taken advantage of by their host. Thus, TCs are more likely to experience a peaceful end when their hosts are committed to human rights treaties:¹⁰

Testing our hypothesis opens an important and unexplored dimension of the impact of the international human rights regime. A significant portion of the world's population exists within TCs and therefore is a core community at risk of experiencing state violence. This is also a particularly difficult testing ground for optimistic theories of human rights compliance because treaties have to alter state behavior in a highly salient issue arena.

Research Design

In our analyses, we study how long TCs persist. TCs are political entities controlling populated territories but lacking sovereign recognition. Their governance of territory and population is sometimes rudimentary, but the same is true for sovereign states in much of the developing world. What qualifies a political entity as a TC is that it can plausibly be argued to be the non-sovereign political authority within a populated territory; that is what is meant by “control” of territory. Such control is evident when a TC removes the sovereign state's representatives from the area, or when it taxes the population and when it provides services in replacement of the sovereign state. Similarly, if a sovereign state undertakes an invasion to reclaim control of territory, that is evidence a TC has been in control. We are able to study a broad sample of TCs thanks to a recently released dataset (Lemke and Crabtree 2020). The dataset was collected by treating sovereign states as sampling units and then pouring over the histories (primarily books and journal articles) of each state in the sample in search of instances where someone other than the sovereign state controlled territory within the sovereign state's borders. Our Online Appendix features a full list of the TCs, the host states in which they resided, the years of their existence, and the manner of their deaths.

Often, the TCs identified are rebels, but about one-fifth of TCs are never in conflict with any other political entities, whether that be the sovereign state or another TC. Of the TCs that do resort to military conflict, they are involved in such conflict only about half of the time they exist. Thus, rather than focusing merely on rebels, we study all instances where a non-sovereign entity controls some part of a state's territory because that existence is a clear threat to the sovereign state, even absent fighting. For example, the persistence of Somaliland is a profound threat to the restoration of a Republic of Somalia, even though there has never been fighting between Somalia and Somaliland. Similarly, the Mon people of Burma have been a TC since 1958 but are only identified by the Armed Conflict Data as a rebel group from 1959 to 1963 and again in 1990. Thus, even when a TC is sometimes a rebel group, a great deal of years in which the sovereign state is challenged are added by considering all TC years rather than just years of active hostilities. Focusing only on territorial rebel groups as “the” threat sovereign states most dread denies us the opportunity to consider all groups that take territory from a sovereign state and all the periods after civil war hostilities end but in which a non-sovereign entity continues to deny the sovereign state control of all of its territory. Focusing on TCs therefore allows us to avoid selecting non-state actors based on whether they are actively engaged in violence. It also moves us further from the bias toward state-centric analyses of territorial governance (Scott 2009). We analyze patterns of TC survival between 1948 and 2010.

Dependent Variables

Our dependent variables record how long TCs survive and the manner in which they die. The two types of death we are concerned with are: (1) peaceful reintegration, which occurs when some political bargain is reached between the TC and host whereby the former relinquishes territorial control; and (2) forceful reintegration, occurring through the direct conquest of the TC by the host state. Of the 145 post-Second World War TCs in the dataset, 43 (29.45 per cent) are peacefully reintegrated, 50 (34.24 per cent) die through forceful reintegration, and 27 (18.49 per cent) still existed in 2010, the final year of observation. The remaining TCs died either through absorption into another TC or by promotion to sovereign statehood (typically after defeating their hosts militarily), and we treat these cases as censored observations in our empirical analysis (for a descriptive analysis, see Online Appendix B).

Explanatory Variables

Our two key explanatory variables are interstate rivalry and embeddedness in the international human rights regime. Rivalry is recorded using Thompson and Dreyer's (2011) dataset, which identifies international dyads where mutual threat perceptions are high.¹¹ From these data, we construct a count of how many rivalries a host state is involved in within any country-year. We rely on a count indicator under the presumption that most rivals have an incentive to support TCs in one another, but the more rivals that are present, the more opportunities exist for subversion. We do not limit ourselves to contiguous rivals because subversion can be, and often is, pursued from a distance.

Next, we identify which states are signatories to human rights treaties whose terms would constrain repressive behaviors that might otherwise be deployed to violently eliminate TCs. No human rights treaty that we are aware of specifically targets the treatment of TCs by their host state, nor do we intend to identify only the subset of treaties whose terms are most specifically related to the treatment of TCs. Instead, we adopt a more expansive approach. States that have only made limited commitment to respecting human rights are also likely to selectively draw exception when repressing TCs or dissident groups. Thus, we identify states that have made the broadest commitment to respecting human rights—these are the cases where treaty commitments hold significant weight in raising the costs of violent elimination of TCs. We use Fariss' (2018) latent variable measure of embeddedness in the international human rights treaty regime based on 23 major treaties (see Online Appendix D). The resulting measure is roughly normally distributed, with a mean in our sample of 0.31.

Control Variables

We control for a variety of factors.¹² The first set of controls capture essential traits of the state and the TC. These include: Sovereign State Capabilities, based on the Correlates of War Composite Indicator of National Capabilities (CINC); the Number of Other TCs within the sovereign host state; and an indicator of whether the TC Resides in Mountainous Terrain.

The second set of controls encompass a broader set of traits that might plausibly covary with the substantive covariates of interest and TC death. Using the Polity data, we also record Host State Failure (see Iqbal and Starr 2016). Host State Development indicates the level of development of the sovereign state and is computed as the ratio of each sovereign state's CINC economic components divided by its demographic components (see Bremer 1992). TC Population is a logged estimate of the number of people residing in a TC's domain. Sovereign State Allies counts the number of allies the sovereign state has for that annual observation (Gibler and Sarkees 2004). TC Loot is a time-varying dummy equal to 1 if the TC funded itself by engaging in trade in contraband or by ransoming hostages. Finally, Sovereign State Neighbors is a count of the number of direct land neighbors of sovereign states playing host to TCs.

Finally, we control for features highlighted in Florea's (2017) study of de facto state survival. Foreign Support indicates whether the TC received foreign military support during that year. Recognition ¹³ counts the number of sovereign states extending diplomatic recognition to the TC. Fragmentation is an indicator of whether a TC is a fragment or splinter from another TC, or whether the TC is the rump from which other splinters fragmented. Host State Executive Constraint is Polity IV's executive constraints measure.

Empirical Identification Strategy

TCs are not randomly assigned to hosts, meaning that potential unobserved confounders will complicate causal interpretations of empirical tests, even after controlling for the variables listed earlier. We therefore pursue multiple, complementary model specification strategies to minimize the risk of unobserved confounding.

First, we adopt a semi-parametric duration modeling framework, testing our expectations using a Cox proportional hazards model (Box-Steffensmeier and Jones 2004), which is built around the concept of a hazard rate h(t,X). This pertains to the rate of TC death at time t given covariates X. The hazard rate is estimated both with respect to a baseline hazard h ₀(t), which reflects the rate of failure (in this case, TC death) among units when all the aforementioned covariates are equal to 0, and with respect to a systematic portion of the model containing all covariates X.

To disaggregate survival by death type, we use an independent competing risks specification (Zorn and Van Winkle 2000). This involves estimating separate Cox models for each of the two types of TC death discussed earlier. In each of these models, TC duration is measured with regard to one specific type of TC death—observations dying in other ways are treated as censored, as are TCs that continued to persist in 2010, the final year of observation.

We use a frailty model specification as a means of addressing unobserved confounders potentially threatening the validity of our inferences. Some hosts may be inherently more “frail”—that is, prone to forceful or peaceful reintegration—than others for reasons not fully accounted for with our control variables. Frailty models incorporate an additional parameter that allows for random shifts in the baseline hazard for a given event occurring. In this case, the parameter v _i is grouped at the country level and shifts the baseline hazard accordingly. The model takes the form:

$$h_i( t ) {\rm \;}\vert {\rm \;}( {\beta {\rm^{\prime}}X_{it}, \;v_i} ) {\rm \;} = h_0( t ) v_iexp( {\beta {\rm^{\prime}}X_{it}} ) \eqno , \;$$

where i indexes TCs and t indexes years of TC survival, X _it is a matrix of the covariates described earlier, and β is a vector of coefficients associated with each variable. The baseline hazard shifts depending on the values of the covariates captured in the systematic portion of the model, X _itβ, and the frailty parameter, v _i.

We test for violations of the proportional hazards assumption (PHA), which underpins all duration models. Violations of the PHA occur when the effect of a covariate changes significantly across the lifespan of the observations. We follow the best practice recommended by Box Steffensmeier and Jones (2004, 133–137) for identifying and correcting violations of the PHA. The test is data-driven and involves generating regression estimates for a standard model specification and then examining the Schoenfeld residuals from each model across the logged values of time (Box Steffensmeier and Jones 2004, 133–7). This test produces p-values associated with each covariate and its relationship with time,¹⁴ and violations are addressed by subsequently including interaction terms between each offending covariate and the natural log of time in the final model specification.

The principal advantage of this approach is that it requires weaker parametric assumptions than most conventional models, while also accounting for unobserved confounders in the data. Nevertheless, a variety of country- and time-specific factors remain a threat to identification. Unobserved features, such as a host country's international neighborhood, might explain both their abilities to eliminate TCs and their proclivity to engage in rivalries. The evolution of the human rights treaty regime throughout the post-Second World War era also means that the treaty embeddedness indicator is necessarily collinear with time.

Our second specification strategy therefore leverages a fixed-effects modeling design to better account for these specific threats to causal inference. Our most comprehensive models take the following form:

$$Pr( {Y_{it} = 1} ) = {\rm \Lambda }( {\alpha + \beta X_{it} + \gamma_j + \delta_t + \varepsilon_{it}} ) \;\eqno , \;$$

where Y _it is an indicator of whether a TC was forcefully or peacefully eliminated in a particular year, and Λ is the logit link function. As before, X _it is a matrix of covariates associated with each TC. We report the results of models that include γ _j, which is a vector of host-country fixed effects, and δ _t, which records calendar-year fixed effects. Thus, this approach accounts for fixed, unobserved confounders related to country units and calendar years.

Finally, we also conduct a series of qualitative investigations of our data to probe the validity of our results. First, we present an in-depth investigation of the Naxalites in India to assess whether our causal mechanisms operate as we would expect. The case was selected as an instance where our causal mechanisms might plausibly be expected to operate. This was not known ex ante, but deeper investigation indicated that this was indeed the case.¹⁵ Secondly, a reasonable concern of reverse causality might exist for our rivalry hypothesis, namely, that the presence of TCs supported by external states leads to a rivalry. We address this by reporting the results of an investigation of all 77 rivalries in our data, finding little evidence that our results could plausibly be explained by reverse causality (see Online Appendix E).

Results

Tables 1 and 2 report the results of our competing risks analysis of TC death via peaceful and forceful reintegration, respectively. Each table contains three specifications that incrementally increase the number of control variables. Positive coefficients correspond to an increase in the baseline hazard of a death type, while negative values indicate the opposite. Effects are reported in terms of the percentage change in the baseline hazard of a given type of death occurring. Like the percentage change in odds, these values have a lower bound at −100 and no upper bound.

Table 1. Competing risks analysis of TC death, peaceful reintegration (1948–2010)

Note: Coefficients reported with standard errors in parentheses. * p < 0.1; ** p < 0.05.

Table 2. Competing risks analysis of TC death, forceful reintegration (1948–2010)

Note: Coefficients reported with standard errors in parentheses. * p < 0.1; ** p < 0.05.

Consistent with our expectations, rivalry significantly reduces the likelihood that a TC will be reintegrated into its host state, either by peace or by force. For peaceful reintegration, the fully specified model identifies a negative and statistically significant relationship between rivalry and the hazard of peaceful reintegration. Each additional rival is associated with a 37.9 per cent decrease in the baseline hazard of peaceful reintegration (95 per cent confidence interval: 3.09 per cent to 60.22 per cent reductions). The reduced model specifications paint a picture with more nuance, at least insofar as they yield evidence of a violation of the PHA.¹⁶ Rivalry is therefore interacted with time, and the resulting effects are reported in Figure 1. In each case, the effect of rivalry is insignificant until a TC has survived past its early years (approximately fifteen years in Model 1 and seven years in Model 2), after which point, the association with peaceful settlement becomes negative, non-trivial, and statistically significant. Why might the pernicious effects of external influence on peace be localized to more mature TCs? As Lee (2020) notes, subversion does not create anarchy from order, but instead relies upon capable agents in the host state. Within the present context, it appears that TCs must survive past their most vulnerable years before external influence undermines the peace process. These temporal dynamics should be interpreted with caution, as they are not persistent across model specifications, though the thrust of results strongly supports the notion that rivals undermine the prospects for peace through subversion by means of TCs.

Figure 1. The effect of interstate rivalry on the hazard of peaceful reintegration, Models 1 (left) and 2 (right).

Notes: Plot reports the effect of a one-unit increase on the international rival variable. 95% confidence intervals are displayed with a dashed line. The effect varies across TC age, reported in logscale. The bottom panels display histograms of observations with instances of peaceful reintegration marked in red.

Results are parallel for forceful reintegration (see Table 2), with rivals reducing the likelihood states will solve their TC problem through force alone. Although insignificant in Model 4, the effect of rivalry becomes marginally significant (p = 0.056) after controlling for essential confounders in Model 5 and more strongly so in Model 6 (p = 0.036). Substantively, each additional rival is associated with a 30.79 per cent reduction in the baseline hazard of forceful reintegration (95 per cent confidence interval: −50.91 per cent to −2.42 per cent). That rivalry operates through both predicted channels in the survival analysis highlights the multiple dimensions through which strategies of subversion are implemented.

The duration analysis also yields clear evidence that embeddedness in the international human rights treaty regime significantly alters state behavior vis-à-vis TCs. The impact, however, is channeled through incentivizing states to peacefully reintegrate, rather than forestalling more coercive strategies. In all three models reported in Table 1, there is a positive, statistically significant relationship between human rights treaty embeddedness and peaceful reintegration. In Models 1 and 2, the effect is time invariant, with a one-standard-deviation increase on the embeddedness score increasing the hazard of peaceful reintegration by 169.40 per cent and 178.16 per cent, respectively. Due to a violation of the PHA, the indicator is interacted with TC age in logged years, with Figure 2 displaying substantive effects. Early on, the impact of signing up to human rights treaties remains substantial. When a TC first appears, increasing its host state's embeddedness in the human rights treaty regime increases the baseline hazard of peaceful reintegration by 239.90 per cent (95 per cent confidence interval: 5.85 per cent to 905.85 per cent). Although the confidence intervals are large, these are some of the strongest positive associations in our data. The effect weakens with time, becoming insignificant at p < 0.05 after about 3.5 years and at p < 0.10 after about six years. Although this timespan might at first appear short, it is critical for TCs, who only live for an average of about ten years.

Figure 2. The effect of human rights treaty embeddedness on the hazard of peaceful reintegration.

Notes: The plot reports the effect of a one-standard-deviation increase in the degree of a host state's embeddedness in the international human rights treaty regime on the hazard of peaceful reintegration. 95% confidence intervals are displayed with a dashed line. The effect varies across TC age, reported in logscale. The bottom panel displays a histogram of observations with instances of peaceful reintegration marked in red.

There is not strong evidence that human rights treaties exert an impact on whether a TC will be forcefully reintegrated. The embeddedness coefficient is negative, as expected, but its standard error is especially wide. That treaty embeddedness does not forestall forceful reintegration is consistent with previous studies finding treaties struggle particularly in combating violent forms of repression (Hathaway 2002; Hill 2010; Lupu 2013b). Unlike many of these studies, however, we do not uncover any evidence suggestive of a counterproductive effect—signing these treaties does not increase a state's propensity to reintegrate coercively. When viewed in conjunction with the positive impact treaties have on the prospects of peaceful reintegration, it appears as though treaties exert their impact more directly in fostering initial strides toward a positive peace, rather than through the direct prevention of coercive reintegration.

Turning to our fixed-effects specification strategy, Table 3 reports the results of logit regression models where the dependent variable is a dichotomous indicator of peaceful reintegration (Models 1–3) or forceful reintegration (Models 4–6). These results are consistent with those reported in the duration analysis, though the relationships pertaining to peaceful reintegration are now stronger for both treaty embeddedness and international rivalry. The presence of international rivals significantly reduces the probability of peaceful reintegration, while human rights treaty embeddedness significantly increases it. Results for both variables are weaker in the models of forceful reintegration. As before, treaty embeddedness has no discernible effect, and the direction of the coefficient is not consistent across specifications. Contrary to the duration analyses, rivalry no longer significantly reduces the probability of forceful reintegration, though the association is always negative. As an additional measure, we also inspect models where the dependent variable is reintegration by peace or by force (Models 7–9). Since rivalry is expected to operate similarly across either death type, it would be worrying if we did not observe significant results in this case. Nevertheless, the rivalry coefficient is negative and significant across all three models. The same is true of the treaty embeddedness indicator, though our theoretical expectations and the aforementioned results suggest this is driven by peaceful reintegration processes.

Table 3. Regression analysis of TC death, peaceful and forceful reintegration (1948–2010)

Note: Logit coefficients reported with standard errors in parentheses. * p < 0.1; ** p < 0.05. Additional covariates include host state capabilities, development, allies, and failure status, the number of other TCs in the host, TC population (logged), loot, foreign support, recognition, and fragmentation.

In sum, both specification strategies provide clear evidence that international influences critically determine the life and death of TCs. These factors have at least as much empirical purchase as variables related to state capacity, which are typically given more emphasis in the literature but are only sometimes significant in our models. For example, host state capability is positively associated with forceful reintegration in the survival analysis but not in the fixed-effects models (see Table C1 in Online Appendix C), and it is never related to peaceful reintegration. The results for host state development are similarly inconsistent across model specifications. One possibility is that these variables are not particularly good proxies for host state capacity. We therefore analyze the relationship using Lee and Zhang's (2017) measure of state capacity (see Table C2 in Online Appendix C). However, again, the core explanatory variables remain significant and in the expected direction, while the capacity indicator is insignificant. These results do not mean that state capacity is unimportant. Instead, we make the lesser claim that capacity alone provides an incomplete explanation for TC survival, which must be contextualized in terms of broader international constraints.

Illustrative Case

Before concluding, we offer a case that nicely elucidates our findings about international treaty embeddedness and forceful reintegration of TCs. The Naxalites (or the Communist Party of India [Maoist]) is a TC that initially seized territory in 1980 and came to control a large swath of east/central India known as “the Red Corridor.” So successful are the Naxalites that in 2006, Prime Minister Singh described them as “the single biggest internal challenge ever faced by our country” (Human Rights Watch 2006). The Indian central government and affected subnational governments responded with aggressive incursions, inflicting considerable harm on civilians within the Red Corridor, but were unable to eliminate the Naxalites.

If our argument is correct, the Naxalites should have worked hard to spread the word about Indian human rights violations against them and against people in their territory. The Naxalites should have gained support from domestic and/or international actors committed to protecting human rights, and there should have been legal repercussions for the Indian government(s), which remained heavily embedded in the international human rights treaty regime.¹⁷ In a democracy like India, this ratification gives the treaty legal status in domestic courts. Human rights supporters can then bring legal cases against a government that violates such treaties. This is exactly what happened in the Naxalite case. The Naxalites have toiled to spread word of Indian human rights violations, they have gained support from domestic and international allies committed to protecting human rights, and there have been legal repercussions for the Indian government(s), all diminishing India's ability to forcefully reintegrate the Naxalites.

To elaborate, the Naxalites devote considerable effort to disseminating information about India's human rights violations, frequently issuing press releases criticizing India's brutality in their efforts to defeat the Naxalites.¹⁸ This information is also reported by human rights allies. In 2009, Human Rights Watch (2009) “documented widespread abuses by Indian government forces, including arbitrary arrests, torture, and unlawful killings, all of them unpunished during previous operations against the Maoists.” Domestic allies are active as well: “Various civil society organizations and individuals have come out strongly against the Government of India for waging a war on its own people” (Srivastava 2011, 96–7).

Domestic human rights activists have brought cases before Indian courts questioning laws promoting anti-Naxalite activities and challenging the Indian government's legal right to raise and employ pro-government militias against people residing in the Red Corridor:

The respected Public Union for Civil Liberties in India has filed suit, alleging that [the Special Public Protection Act, which authorizes military action against the Naxalites] is amenable to gross abuse and misuse, arbitrariness and partiality and can result in harsh and drastic punishment to innocent persons without hearing or remedy and can be abused for the suppression of the fundamental rights of the citizens. (Human Rights Watch 2006)

Human Rights Watch (2006, emphasis added) editorializes that “the government should repeal the new Special Public Protection Act, or amend it to conform to international human rights law.”

A domestic legal case against the use of pro-government militias, like Salwa Judum and the Koya Commandos, was heard in India's Supreme Court, resulting in:

a blow to both the Chhatisgarh government and the Centre, the Supreme Court has declared as illegal and unconstitutional the deployment of tribal youths as Special Police Officers—either as Koya Commandos, Salwa Judum or any other force—in the fight against the Maoist insurgency, and ordered their immediate disarming. (Venkatesan 2011)

These militias had been an important part of the government's strategy against the Naxalites; the Supreme Court's decision diminished the sovereign state's capacity to forcefully reintegrate Naxalite territory.

Finally, while there is no proof the Naxalites communicate directly with domestic and international human rights allies, the Indian government is convinced such connections exist: “Human rights activists have repeatedly come under attack or been arbitrarily arrested on unsubstantiated accusation of Naxalite links” (Human Rights Watch 2009). Further, “The government continues to view such civil rights groups as front organizations of the Maoists and have curtailed the freedom of many such activists” (Srivastava 2011, 97).

In sum, TCs publicize violations of human rights law by states trying to suppress them. This information assists domestic and international actors mobilized to bring pressure on sovereign states to obey international human rights law, even during campaigns to suppress TCs. These efforts result in legal cases, going as high as national supreme courts, being decided against the government, which then inhibits the sovereign state's military options for forcefully reintegrating TCs.

Critically, however, while human rights treaties have increased costs for the Indian government and reduced its ability to forcefully reintegrate the Naxalites, it has never fully abandoned its forceful reintegration strategy, nor has it ever fully embraced a peaceful reintegration strategy, as doing so too willingly would create dangerous precedents for other would-be TCs. The modest effects we observe in our empirical analysis of forceful reintegration thus appear to be reflective of a broader trend: human rights treaties increase the costs of using repressive force against TCs, but these costs are often insufficient to deter use of these strategies altogether.

Conclusion

Our results help explain why some TCs persist for puzzlingly long periods of time, destabilizing sovereign states and threatening peace and stability in the process. More capable states can and do eliminate TCs more quickly, but this quintessentially domestic competition between sovereign states and TCs is heavily influenced by international factors as well.

Our findings highlight the specific contours through which rivalry reshapes the politics of spatial authority. Aiding and abetting TCs constitutes a significant and damaging strategy of subversion that interstate rivals pursue through multiple channels. They undercut what might be the most direct security strategy: to dispatch TCs through forced reintegration. Furthermore, while a substantial proportion of TCs never engage in violent conflict with their hosts, the presence of rivals weighs no less heavily by forestalling or halting peaceful reintegration processes. Rivalries have long been examined as a predominant explanation for interstate war and peace, but our findings yield another clear implication: a world with fewer rivalries would make for more harmonious domestic governance as well.¹⁹

There are also new grounds for optimism about the efficacy of the international human rights treaty regime. Much of the debate surrounding treaty compliance has rightly centered on the prevention of extreme violations of physical integrity rights—a goal that treaties are sometimes thought to fall short of, or even undermine. We did not uncover substantial evidence that treaties constrained or increased a state's tendency to violently eliminate TCs. More consequentially, however, we found that states embedded in the international human rights treaty regime were more likely to peacefully reintegrate TCs early in their existence. That this relationship obtains signals an important and hitherto unexplored dimension through which human rights treaties confer a positive impact on populations at risk of experiencing repression.

In short, the remarkable persistence of TCs in the international system is explained, at least in part, by the international system itself. It presents not only perils in the form of international rivals seeking to subvert their adversaries from within, but also opportunity in the form of an international human rights regime with the potential to propel TCs and their hosts toward a peaceful resolution. Additional research will be necessary to identify the mechanisms through which the influence of rivals can be counteracted and the international human rights regime preserved and supported.