Can international human rights treaties (IHRTs)

Transcription

1 Treaties, Tenure, an Torture: The Conflicting Domestic Effects of International Law Courtenay R. Conra Emily Hencken Ritter University of California, Merce University of California, Merce International human rights treaties are argue to increase both the likelihoo of omestic mobilize issent an juicial constraint. These pressures pull leaers in conflicting irections: mobilize challenges unermine a leaer s position in power, increasing incentives to repress; courts raise the probability of litigation, ecreasing incentives to repress. We argue authorities balance these pressures base on their job security. Politically insecure leaers, esperate to retain power, repress to control the estabilizing effects of issent. Secure leaers are less likely to fall to citizen pressures, but the probability of facing an effective juiciary weighs heavily in their expecte costs. Consequently, they repress less to avoi litigation. We fin empirical support for the implications of our formal theory using ata on commitment to the UN Convention Against Torture. Treaties have no effect on repression in states with insecure leaers but have a positive effect on rights protection in states heae by secure leaers. Can international human rights treaties (IHRTs) improve state human rights practices? Scholars argue commitment to IHRTs rarely leas to increase protections (Hafner-Burton an Tsutsui 2007; Hathaway 2002), an treaty-boun states have even been known to violate rights more than otherwise expecte (Hill 2010; Vreelan 2008). Although committe states repress less than their uncommitte counterparts (Neumayer 2005), this fining is often attribute to the screening effect of IHRTs: states with goo rights practices select into easy obligations (e.g., Hathaway 2007). Are there conitions uner which IHRTs actually constrain repression? 1 Commitment to international law alters omestic incentives to repress in two ways. Effective omestic courts constrain state repression (e.g., Keith 2002), an courts are enable an embolene to sanction violators in the wake of IHRT accession (Powell an Staton 2009; Sloss 2009). Treaty commitment also creates new opportunities for citizens to mobilize against the state (Simmons 2009; Vreelan 2008), but this shift may incentivize repression: when a population mobilizes to challenge the state, authorities frequently respon with increase rights violations (cf. Davenport 2007a). The omestic effects of IHRT obligations are thus at os: increasingly effective juiciaries constrain obligate states into reucing repression whereas increase mobilize challenges may incentivize increasing repression. When oes IHRT commitment lea to improve rights practices, given the omestic incentives to repress? We argue that state authorities respon to threats to their position of power ifferently epening on their job security. As leaers become increasingly likely to remain in power, they have more to lose an so repress more to control popular threats. The more authorities repress, the more costs they incur, particularly as theyareincreasinglylikelytobebroughttocourt.an international legal obligation to protect rights which augments the probability the state will face costly litigation for repression has the greatest impact when the leaer faces the highest incentives to repress to remain in power an thus is subject to higher costs: commitment to an IHRT will lea to a reuction in repression when leaers are secure in power. By comparison, vulnerable leaers will o what they must to remain in power, repressing without consieration of the court s effectiveness or the impact of the international obligation. We evelop a formal theory of repression an mobilize issent to examine the effects of IHRT obligations on government rights practices. Commitment to 1 A supplementary appenix is available at Data an supporting materials to reprouce the numerical results are available on the authors websites. The authors contribute equally to this research. The Journal of Politics, Vol. 75, No. 2, April 2013, Pp oi: /s Ó Southern Political Science Association, 2013 ISSN

2 398 courtenay r. conra an emily hencken ritter an IHRT yiels some benefit to the state, but it also increases the probability that a repressive leaer will face costly omestic litigation. Given this potential constraint, state authorities anagroupengageinaconflict over popular emans. Our theory yiels novel expectations over the effect of IHRTs on state repression: commitment to an IHRT has no effect on state repression when leaers are vulnerable to turnover, but it has a negative effect on repression when leaers are secure in power. Using an empirical strategy aligne closely with the assumptions of our formal moel an ata on commitment to the Unite Nations Convention Against Torture (CAT) an the state s propensity to torture, we assess how omestic factors affect ecisions to repress in IHRTobligate states as compare to nonobligate states. We fin support for our hypothesis: leaers who fear turnover respon to a mobilizing population regarless of institutional pressures, but secure leaers repress less than they otherwise woul when committe to an IHRT. This stuy is among the few to fin empirical support that IHRTs can lea to improve human rights protections. Surprisingly, treaties have the greatest impact on reucing violations when repression is expecte to be at its highest. Our work is also at the forefront of those investigating how political survival affects state ecision making, suggesting that institutional pressures iffer from societal pressures in their effects. We argue that insecure leaers, fearing turnover, prioritize the salience of behavioral pressures such as a mobilizing population an largely ignore even large costs from constraining institutions like the omestic juiciary. Secure leaers are more likely to feel the impact of institutional pressures an respon accoringly. A wealth of scholarship suggests that emocracies an ictatorships iffer in both state responses to international law an the ecision to repress; our focus on leaers allows us to uncover variance on these outcomes both across an within regime type. In total, this project combines insights from international an omestic stuies of human rights an contributes meaningful knowlege as to the conitions uner which states will repress their citizens. Domestic Effects of International Treaties Scholars have selom foun IHRTs to positively impact rights practices. 2 The ineffectiveness of IHRTs 2 For a sample of stuies of IHRTs an human rights with pessimistic conclusions, see Hafner-Burton an Tsutsui 2007; Hathaway 2002; Hill 2010; Neumayer 2005; an Vreelan results in large part from their lack of enforcement mechanisms (e.g., Hafner-Burton 2005; Ramcharan 1989), an foreign states are rarely willing to take coercive action on behalf of represse citizens (e.g., Lebovic an Voeten 2009; Neumayer 2003). But we shoul not be so quick to eem international law irrelevant: IHRTs may not constrain violations irectly, but they can alter omestic politics in favor of rights protection (e.g., Simmons 2009). More specifically, IHRTs bolster the constraining role of omestic courts an may incentivize social mobilization. Effective omestic courts consistently lea to increase human rights protections. Powell an Staton (2009, 154) efine an effective juiciary as one that constitutes a genuine constraint on state behavior, meaning one that is both willing an able to rule against state actors. 3 The expectation that the court can constrain the state is more important in changing state behavior thanitsactualrulings:victimsaremorelikelytobring allegations before the court when they believe it to be effective (Powell an Staton 2009). Litigation is costly for accuse authorities in resources an opportunity costs, even if they manage to avoi a negative ruling, an the potential costs eter those facing effective juiciaries from repressing (Cross 1999; Keith 2002). State commitment to an IHRT increases the probability that a repressive actor will incur litigation costs beyon those resulting from an effective juiciary alone. Ratifiers must aopt IHRT terms into omestic law if they o not alreay exist, refine extant laws so that they align with the international obligation, an in common-law systems, appropriate international preceents (Hill 2012; Simmons 2009). 4 IHRT ratification also focuses attention on state practices an increases legitimacy for rights-relate cases, both of which make victims increasingly prone to litigate. A state that commits to international law is more likely to face costly litigation than one that oes not commit, regarless of its initial level of omestic juicial effectiveness. Leaers who want to remain in power must also consier threats from societal challenges. For our purposes, mobilization is a coorinate attempt by nonstate actors within the territorial jurisiction of the state to use collectively controlle resources to influence political outcomes of any type (Tilly 1978). 3 Courts are effective when they are free from manipulation (e.g., Cross 1999), an when omestic actors are willing an able to punish noncompliance (e.g., Vanberg 2005). 4 Even in states that o not implement treaty terms into omestic law, treaties can have similar inirect effects on litigation (Hill 2012).

3 treaties, tenure, an torture 399 Citizens can challenge the state through state sanctione (institutional) an unsanctione (mobilize protest) channels; we focus on noninstitutional challenges. 5 The action may be legal or illegal, violent or nonviolent, creating irect costs for the state, amaging infrastructure, or isrupting normal societal activities. Popular challenges also unermine authorities legitimacy to rule, suggesting to observers that leaers cannot control the population. 6 To counter the threat from mobilize challenges, state authorities consistently turn to repression (Davenport 2007a, 7 8). 7 Repression unermines the group s will an capacity to mobilize against the state, whether by restricting their resources an abilities to organize or creating an atmosphere of fear that prevents action (Davenport 2007b). IHRT obligations may influence the likelihoo of social mobilization, an scholars have foun both positive an negative treaty effects. The new, internationally legitimize stanars an the increase focus on rights protections can lea citizens to form new or join existing movements to pressure the state for omestic changes (Hill 2012; Keck an Sikkink 1998; Simmons 2009). Even if the state oes not inten to comply with the letter of an IHRT, groups may believe a ratifying state to be more likely to respon to emans (Vreelan 2008). Scholars have also foun instances of ecrease mobilization after a state commits to a treaty (e.g., Hollyer an Rosenorff 2011). In the next section, we moel the effect of treaty commitment on the interaction between a group an the state, eriving rather than assuming the effect that the treaty has on mobilization, an consequently, on the incentive to repress. 8 To summarize, state authorities consier conflicting costs from omestic institutions an societal pressures when they make ecisions about repression, an international human rights treaties magnify those pressures. The increase probability of litigation in an IHRT-committe state makes repression more costly. But citizens may believe that there is increase probability of receiving their emans uner IHRT obligations, 5 Although mobilization can take institutional forms (e.g., litigation), we capture such costs separately from behavioral threats in our theoretical moel an empirical estimation. 6 Mobilization represents an increasing threat to the leaer s position as it becomes wier in scope, more violent, or more irectly threatening in the group s emans (Davenport 1995). 7 Authorities are better able to control policy an maintain power by using coercive tactics against a mobilizing population (Gurr 1988). 8 In other work, we investigate the effect of IHRT commitment on social mobilization. leaing them to make threatening challenges. How o authorities balance these conflicting incentives? A Theory of Domestic Treaty Effects We specify a formal moel of omestic conflict between state authorities an a group of citizens uner the incentives of an international legal obligation. This theory allows us to erive implications as to how state authorities navigate both omestic an international cross-cutting pressures in the choice of repression. Moel Specification We moel an interaction between a Leaer (L) 9 an a Group of citizens (G). At the outset, the Leaer ecies whether to commit the state to a human rights treaty, with the expectation that oing so will amplify (e) the extant probability he will experience costly litigation (f) for a given level of repression. After committing to the treaty or not, the Group ecies how much to mobilize (m) aroun a eman, an the Leaer simultaneously chooses how much to repress (r), though both of these ecisions entail resource costs (2m an 2r) that make the actors want to minimize their expenitures. Finally, their ecisions conition the probability (u) that the Leaer remains in power at the en of the game. Equations (1) an (2) present the players expecte utility functions, an we iscuss the notation below. The Leaer s payoffs are: U L ¼ f m r f þð1 mþr m Þuþð if uncommitte to IHRT mþr Þu k r ðfþeþþð1 m m mþrþuþðmþr Þu k þ m if committe to IHRT an the Group s payoffs are: U G ¼ mþ 1 m ð1 uþ mþ r m þ 1 u m þ r k ð1þ ð2þ 9 The leaer represents an iniviual or group acting uner authority of the state. Although most physical integrity violations entail a form of principal-agent problem (e.g., Conra an Moore 2010), our moel is appropriate to explain the effects of a treaty via its impact on the probability that the leaer faces costly litigation for repression regarless of how it occurs.

4 400 courtenay r. conra an emily hencken ritter Base on the iscussion in the previous section, we make the following assumptions: Repression (r $ 0) requires resources, represente by 2r in both utilities in equation (1). The more the state represses (or the more severe its action), the more resources it expens. Mobilization (m $ 0) requires resources, represente by 2m in equation (2). The more the Group mobilizes (or the more severe its action), the more resources it expens. The chosen levels of repression an mobilization affect the probability the Group receives its emane policy or goo allocation m mþr. This eman can be anypolicyorallocationanneenotbelimiteto rights-relate concerns. Using this relational specification, we capture the iea that the Group is more likely to receive its emans asmobilizationincreases an less likely as repression increases. 10 If the Group oes not receive its eman, the Leaer remains in power (with benefits equaling 1) with probability 0 # u # 1, which represents the leaer s baseline or a priori job security. If, instea, the Leaer gives in to the Group s emans, the leaer remains in power with a lower probability: u k, such that k. 1. k can represent the scope of the accommoation; the more the Leaer gives in costing him resources or legitimacy the more he risks his position of power. He is k more vulnerable to turnover if he loses the conflict (which he oes m mþr with probability ). An if he loses office, L receives 0. Although the baseline probability of remaining in office is exogenous, both the Group s an the Leaer s conflict ecisions conition the ex post probability L will lose office. The Leaer s costs for repression also inclue the probability (0, f # 1) of incurring litigationrelate costs (value at 1). The probability of being brought to court is a function of juicial effectiveness: if citizens believe the court is a viable place to fin reress from state abuse, they will be more likely to litigate. Committing to an IHRT makes the Leaer a small amount (e, such that 0, e, 1 2 f) more likely to incur court-relate costs. The increase probability of litigation may be ue to new laws, 10 The government coul accommoate a group s emans rather than repressing. Our specification allows the leaer to choose a level of r5 0, which ensures the group receives its emans m mþ0 ¼ 1. That being sai, the theory primarily speaks to situations in which repression is a possible choice for the leaer, which is most often the case. heightene focus on state abuses, increase NGO activity, etc. The Leaer receives a benefit economic or political, omestic or international, physical or reputational forcommittingtoanihrt,representebytheterm m Thismoelallowsustoerivepreictionsastohow state authorities respon to the omestic political effects of IHRT commitment. We assume that the treaty s only irect impact is to increase the probability the leaer coul experience litigation by a small amount, but this effect changes the way citizens expect leaers to act regaring repression. The omestic actors ajust their conflict behavior to account for these expectations. How oes the Leaer balance the increase pressure from the court against incentives to control the Group s emans? Equilibrium Analysis The moel solution is a unique Subgame Perfect Equilibrium, such that there is one optimal choice for any given combination of parameter values. Proposition 1 states the equilibrium solution; proofs can be foun in the Supplementary Appenix available online. Proposition 1. The following strategies constitute the Subgame Perfect Equilibrium: (1) when L oes not commit to an IHRT, G mobilizes at level m U an L represses at level r U, efine as m U [ ðk 1Þuf kð1 þ fþ 2 an r U [ ðk 1Þu kð1 þ fþ 2 ; (2) when L commits to an IHRT, G mobilizes at level m C an L represses at level r C, efine as m C [ ðk 1Þueþ ð fþ kð1 þ e þ fþ 2 an r C [ ðk 1Þu kð1 þ e þ fþ 2 ; an (3) L commits when m > u 2k 2 k 1 1 þ ð Þ ð1 þ fþ 2 þ 2 1 þ f! 2 k 1 ð Þ ð1 þ e þ fþ 2 : Whether or not the state commits to the treaty, the leaer an group engage in conflict simultaneously, conitioning their choices on what each expects 11 For scholarship consiering normative, material, an strategic incentives to commit to IHRTs, see Hathaway 2003; Hollyer an Rosenorff 2011; Moravcsik 2000; an Vreelan 2008.

5 treaties, tenure, an torture 401 the other to o. The more a group mobilizes, the more amage it causes to state authorities, an the more likely it will be to receive its emans. However, higher levels of mobilization require more resources. This means the group tries to mobilize just enough to have a strong chance of receiving its emans while minimizing resource costs. Similarly, the leaer can best control the group an keep it from changing the status quo by repressing severely an wiely, but the costs of oing so limit state authorities to repressing only what they can affor. The actors also consier the probability that authorities will incur juicial costs, regarless of commitment status. Victims are more likely to bring authorities to court for rights violations when they believe that the omestic juiciary is effective. As state in equation (1), repression becomes more costly as both the level of repression (r) an the probability of litigation (f) increase. To counter the increase probability of being brought to court, authorities can repress less an avoi the potential for institutional punishment. The leaer must therefore ecie whether to increase repression to maintain the status quo or ecrease it to avoi litigation. 12 Authorities conition repressive ecisions on their ability to remain in power. The actors have an ex ante or baseline expectation of the probability the leaer will remain in power base on aspects such as the leaer s time in office, institutional means of removal, an policy outcomes. When a leaer submits to a challenge an accommoates the group s emans, he loses the policy itself as well as unermining his legitimacy, both of which subvert his job security. Leaers who want to remain in power thus have to prioritize potential threats to their removal. The more secure a leaer is in power, the more he represses to protect that position from increasing mobilize issent. Recall from equation (1) that the Group wins its emans with probability m mþr, such that the Leaer will remain in power with the ex post probability u k. The greater the leaer s baseline job security (u) is, the 12 In the moel, we assume the probability that repression will involve court-relate costs to be exogenous. By oing so, we assume that the leaer an citizens have a common expectation about the juiciary s effectiveness. However, in many cases, authorities can an o exercise control over the juiciary, such that they can influence this probability of litigation. If we were to allow L to set or influence f, we assume the leaer woul set it as low as possible to minimize his risk. In such a case, we woul be far less likely to see the juiciary constraining state repression. As long as there are empirical instances of the juiciary constraining the state rather than the state constraining the juiciary, however, our theory can be use to explain state behavior. more the leaer stans to lose as mobilization increases. Therefore, as the leaer s baseline job security increases, the leaer represses more in equilibrium. However, this willingness to engage in increasingly severe repression as the leaer becomes more secure also makes the leaer more vulnerable to costly litigation. The more authorities repress, the more likely they are to be subject to legal costs there are more victims to bring suit, there is more evience that the state is in violation that coul result in a trial, an attention is more likely to be rawn to the abuse. If the leaer is likely to be brought to trial for violating rights, increasing the violations will leave the leaer even worse off. Formally, the probability a repressive leaer will be brought to court (f) is multiplie by the negative level of repression; as repression increases, litigation has a greater negative impact on the leaer. Since leaers repress more as their baseline job security increases, they will also respon with a greater reuction in repression uner the expectation of juicial effectiveness when they are more secure in power. What, then, is the effect of IHRT commitment on respect for human rights? In orer to properly erive preictions of state abuse, we enogenize both repression an issent (cf. Pierskalla 2010; Ritter 2013). Whilewecanerivecomparativestaticsovertheeffects of IHRT obligations on both state an group behavior, we present only the implications for repression here an those for mobilization in a istinct paper. 13 In our moel, the only irect effect of commitment on the choice of repression is the increase in the probability a repressive leaer faces litigation, e. The international obligation can lea to new or aapte laws, increases in NGO ai to victims bringing suit, an increase belief inthelegitimacyofacaseoralaw,etc.hence,astate obligate to international human rights law is e more likely to experience negative costs for violating it, which constrains authorities to repress less in expectation of these consequences. Surprisingly, this effect of treaty obligations only constrains authorities from rights violations when they aresecureinpower. While all states can experience the legal effects of commitment to an IHRT, these changes impact vulnerable an secure leaers ifferently. Secure leaers are willing to repress severely to maintain their valuable position of power, an the severe mobilization they are likely to face incentivizes them to o so. This makes them increasingly subject to costly litigation, an all the more so as the citizenry 13 Figures illustrating comparative statics of the equilibrium level of repression across the possible range of political survival are available in our supplementary appenix.

6 402 courtenay r. conra an emily hencken ritter believes the juiciary to be increasingly effective. It is uner these conitions, when the leaer woul otherwise repress so severely as to have very high attenant costs, when treaty commitment can have the greatest impact. Boosting the probability of costly litigation will be most amaging to the leaer when the leaer woul otherwise repress the most an this is when the leaer is most secure in power. 14 Politically insecure leaers, by comparison, o not respon to the effects of the treaty when consiering repression. Such leaers face a higher probability of litigation if they are committe to an IHRT, but this ifference is minor consiering the low amount of repression an insecure leaer is likely to use. In equilibrium, when u (job security) is relatively low, the group will not engage in much mobilize issent, since oing so requires resources that are not necessary when the leaer is alreay likely to lose office. The leaer represses minimally to counter the group s action. At these low levels, the possibility of litigation is not a compelling threat, an neither is the boost from the international legal obligation. Insecure leaers repress as necessary to stay in power, an their ecision epens little on treaty commitment status. From this theory, we erive the following hypothesis: H1: IHRT commitment has no effect on repression when job security is low. As job security increases, the effect of IHRT commitment on repression becomes negative. Empirical Analysis We estimate the state s propensity to repress as a function of commitment to the Unite Nations Convention Against Torture (CAT), using ata on state torture practices in 148 countries from 1984, when the CAT opene for signatories, until Investigating the effect of CAT commitment on state torture is an appropriate manner by which to test our theory for several reasons. First, torture is an action specifically intene to unermine citizens capacity an willingness to issent (Rejali 2007) as we conceptualize repression oes in our theory. Secon, torture is a prevalent repressive tactic, an authorities commonly justify it as a necessary evil for the maintenance of orer (Rejali 2007). It thus constitutes a 14 The constraining effect creates incentives not to commit to the IHRT in the first place, yet the short- an long-term benefits lea even these states to o so theoretically an empirically. ifficult test of the influence of constraints on state respect for human rights. Thir, the CAT focuses on improving human rights for a single type of violation. We prefer a clear measurement link between IHRT manates an abuses, an the tight linkage between the CAT an torture meets this criterion. To estimate the conitions uner which state leaers torture, we nee a measure that represents the state s chosen pattern of abuse. Base on content analysis of Amnesty International (AI) an U.S. State Department torture allegations, Cingranelli an Richars (2010) (CIRI) coe a state s annual environment of torture, accounting for whether the state generally tortures a lot, some, or not at all in a given year. Because the conceptualization of repression in our theory is continuous, we collapse CIRI s trichotomous measure of torture incience tocreateameasure of Systemic Torture, coe 1 if a government is reporte to have engage in a lot of torture an 0 otherwise. 15 Estimating the likelihoo of Systemic Torture thus approximates the continuous concept. CAT Commitment is coe 1 in the year in which a county ratifies (or accees to) the CAT an 1 every year thereafter. A measure of Juicial Effectiveness must account for two concepts. First, it shoul inicate whether juges are free to rule as they see fit an whether their rulings are translate into political outcomes (Ríos-Figueroa an Staton 2013, Staton an Moore 2011). Secon, the measure shoul reflect the extent to which the population believes the court to be effective in its ability to rule against the state; this captures the iea that iniviuals are more likely to bring litigation to an effective court (Powell an Staton 2009). To measure juicial effectiveness, we use a new inicator from Linzer an Staton (2012). 16 Recognizing that extant measures are inicators of an unerlying concept, Linzer an Staton (2012) use a heteroskeastic-grae response item response theory (IRT) moel to combine information from eight existing measures to create a latent measure of Juicial Effectiveness. The final continuous measure inclue in our moels ranges from 0 to 1, where higher values on the scale inicate higher levels of effectiveness Our results are robust to using a binary measure of Any Torture coe 1 if there is even a single violation in a given country-year. 16 For a iscussion of measuring juicial effectiveness, see Ríos-Figueroa an Staton (2013). 17 Our results are robust to several alternative measures of juicial effectiveness an the inclusion of a measure of elections from Cheibub, Ganhi, an Vreelan (2010).

7 treaties, tenure, an torture 403 We require information on the willingness an ability of nonstate actors within the state s territory to engage in mobilize issent against the government. Various measures of internal conflict are available crossnationally for our temporal omain an may seem appropriate as a measure of issent. Unfortunately, these measures typically inclue information on state repression, making them inappropriate for our purposes. Therefore, we turn to the Cross-National Time-Series (CNTS) Data Archive (Banks 2010), which captures opposition acts against the government, but oes not inclue information on state responses. This ata captures low-intensity actions against the state, enabling us to stuy the effects of even minor anti-government mobilization. Mobilization is coe 1 in a given year if a state experiences at least one anti-government emonstration, general strike, riot, revolution, or act of guerilla warfare. 18 We argue that executives make ecisions about torture base on their expectations about remaining in office. Although it is ifficult to measure leaers actual beliefs about job security, we follow Cheibub (1998), who argues that political an economic factors affect the executive s probability of remaining in power. 19 To represent the executive s probability of political survival, Cheibub (1998) uses parametric survival moels to create empirical measures of job insecurity base on the leaer s time in office, previous trens in leaership change, an annual economic growth. Because Cheibub s (1998) measure is limite geographically an temporally, we create our own estimate of the executive s likelihoo of remaining in office using ata from Goemans, Gleitsch, an Chiozza (2009). 20 The resultant measure ranges from 18 Approximately 40% of the state-years in our ata experience mobilization. Our results are robust to a count of these events, changes in this conceptualization (e.g., omitting emonstrations or riots), an an alternative measure of the number of human rights naming an shaming events (i.e., allegations of abuse) in a given country year from Bhasin an Murie (2011). 19 We assume that estimates of Job Security track with leaers beliefs about their tenure. The factors use in the creation of the measure represent latent expectations about job security an are inepenent of the behavioral threats capture in our measures of mobilization. 20 Following Young (2008), we also create two aitional measures of job insecurity. Because leaership change in emocracies is arguably ifferent than leaership change in autocracies, our first alternative measure of job insecurity accounts for previous trens in irregular leaer change, the age of the leaer, an the level of emocracy of the state. Our secon alternative measure of job insecurity accounts for the Cheibub (1998) covariates, as well as previous trens in irregular leaer change, the age of the leaer, an the level of emocracy of the state. 0 to 1 (highest probability of leaership turnover). We reverse the scale to create our final measure of Job Security. 21 Empirical Moel Specification CAT commitment is etermine in part by the covariates preicting torture (Hill 2010; Powell an Staton 2009; Vreelan 2008). If states commit to the CAT only when they face certain combinations of omestic juicial effectiveness, popular mobilization, an executive job security, traitional probit moels make it ifficult to etermine whether CAT-committe states lessen (or heighten) torture as a result of international commitment or as a result of the omestic conitions (cf. Przeworski an Vreelan 2000; von Stein 2005). As such, selection into IHRTs like the CAT is likely to be nonranom; the elements of the torture ecision for which we o not account are likely to be correlate with the errors of the commitment ecision. An estimator must account for the lack of inepenence across the ecisions to commit to the CAT an torture. The most common solution to this problem is to use a selection moel. In a stanar Heckman (1979) moel, the selection stage (here, commitment to the CAT) etermines membership in the outcome stage (systemic torture) because ata for nonselecte units are typically unobserve. This characteristic prevents us from comparing the effects of institutions in committe states to those in uncommitte states. However, we argue that omestic conflict happens in both committe an uncommitte states in our theory, an we observe the outcome of interest (repression) in both types of states in our ata. We therefore face an interesting observability problem: it is impossible to observe (1) the level of repression that woul have occurre in nonsignatory states ha they committe to an IHRT, an (2) the level of repression that woul have occurre in signatory states ha they faile to commit. We use a treatment moel written by von Stein (2005), which allows us to examine the effect of CAT commitment on torture, as well as the effect of omestic institutions on state torture in committe an noncommitte states. The estimator is similar to a traitional selection moel in that it accounts for observe factors that affect the commitment ecision 21 Because state leaers face a low probability of losing office, the ata are right-skewe. Using estimates as an inepenent variable introuces the measure s own error structure into the primary moel. To ajust the variance-covariance matrix, we bootstrap our stanar errors.

8 404 courtenay r. conra an emily hencken ritter (von Stein 2005, 617). Unlike a typical Heckman (1979) moel, which estimates separate selection an outcome equations, the treatment moel allows for the estimation of three equations an equation preicting selection into the CAT, an equation preicting Systemic Torture in signatory states, an an equation preicting Systemic Torture in nonsignatory states. Consequently, we are able to test our hypothesis about the effect of CAT commitment on the likelihoo of repression, accounting for both the observe an unobserve factors that affect state ecisions to ratify. 22 The factors that lea states to commit to IHRTs often lea them to repress, so we inclue measures of Juicial Effectiveness, Mobilization, an Job Security in the selection stage to etermine the effect of CAT Commitment on Systemic Torture. To meet the exclusion restriction of the selection moel (Sartori 2003, 112), we inclue a variable known to affect CAT commitment but not torture. Our instrumental variable is the number of intergovernmental organization (IO) memberships a state maintains uring a given year, incluing NATO, the European Union, etc. (Ulfeler 2011). Commitments in one area of international relations (e.g., trae or conflict) capture a state s affinity for international interepenence; we expect states will be more likely to commit to other areas of international law, incluing human rights law, as this count of memberships increases (Mansfiel an Pevehouse 2006, 149). Importantly, however, we o not expect compliance in one of these areas to be connecte to compliance in another. The incentive structures of alliances or trae agreements, for example, are generally not available to state parties of human rights agreements. Inee, while allies comply with their obligations most ofthetime(lees2003),statesrarelycomplywith IHRT obligations (Hathaway 2002). 23 Finally, our hypothesis about the likelihoo of repression is conitional, requiring the inclusion of interaction an constituent terms in the selection an outcome equations of our moels. We preicte that Job Security will interact with IHRT Membership in 22 Selection moels are sensitive to moel specification; as shown in our Supplemental Appenix, our results are robust to a myria of moel specifications, as well as the inclusion of cubic splines an a counter of prior failures (Beck, Katz, an Tucker 1998) or a thir-orer polynomial time counter (Carter an Signorino 2010) to control for temporal epenence. 23 Because international treaty commitment is argue to be in part riven by regional an global norms (Gooliffe an Hawkins 2006), we check the robustness of our results by also incluing in our selection equation two measures from Powell an Staton (2009) inicating the percentage of states in the region an the worl that have committe to the CAT in a given year. its effects on Systemic Torture, as a function of the levels of Juicial Effectiveness alreay present in the state. 24 The treatment moel accounts for CAT Commitment in the selection stage, leaving us to interact the two concepts that are exogenous in our theory in both the outcome an selection equations: Juicial Effectiveness an Job Security. Empirical Results an Discussion Our hypothesis centers on how authorities respon to the IHRT s effects on omestic politics, particularly as compare to what they woul have one absent commitment. Table 1 presents estimates of the effects of the inepenent variables on Systemic Torture using the von Stein (2005) estimator. The first column of Table 1 lists our results for CAT-signatory states an the secon column for nonsignatory states. The parameter r measures the extent to which unobservable factors not capture in the selection stage affect the likelihoo of Systemic Torture in both signatory an nonsignatory states. 25 We present the results of the selection stage in the bottom half of Table 1, but focus ouriscussionontheresultsintheoutcomeequation estimating the probability of Systemic Torture. 26 The treatment moel allows us to examine the effects of our inepenent variables on IHRT-obligate an nonobligate states. It is impossible to observe in our ata the level of Systemic Torture that woul have occurre in uncommitte states ha they chosen to commit to an IHRT, as well as the level of Systemic Torture that woul have occurre in IHRT-obligate states ha they remaine outsie the treaty. To etermine the effect of CAT commitment, we must imagine two counterfactuals: one in which all countries that have not committe to the CAT are force to ratify, an one in which all countries that ratifie the CAT were 24 Although the irection of our preiction for state repression oes not iffer across juicial effectiveness, the magnitue of the effect oes. Because of this, an to follow as closely as possible to our theoretical moel, we inclue the interaction term in our empirical moels. 25 Likelihoo ratio tests of r for committe compare to uncommitte states lea us to reject the null hypothesis that their joint effect is zero (p ), which inicates selection effects. 26 Our theory suggests that states in which IHRTs are likely to have the largest effects are unlikely to commit. Yet a significant number of states with secure leaers are obligate to IHRTs, either because they electe to o so for short-term benefits or because they came to office with the obligation alreay in place. Thus, we are able to raw inferences about treaty effects having moele the enogeneity of the treatment.

9 treaties, tenure, an torture 405 TABLE 1 Effect of CAT Commitment on Systemic Torture Outcome DV: Systemic Torture Signatories Nonsignatories Juicial Effectiveness t * (1.969) Job Security t (1.556) Juicial Effectiveness t x Job Security t (2.372) Mobilization t 0.606* (0.082) Constant 3.004* (1.321) Selection DV: CAT Commitment Juicial Effectiveness t 4.597* (1.426) Job Security t (1.080) Juicial Effectiveness t x * Job Security t (1.745) Mobilization t (0.052) IO Membership t 0.124* (0.013) Constant * (0.892) r (0.203) Pseuo Log-Likelihoo N * (2.387) * (1.491) (2.827) 0.565* (0.072) 4.089* (1.235) 0.900* (0.269) FIGURE 1 Effect of CAT Commitment on the Probability of Systemic Torture as Job Security Increases Note: * Significant within 95% CI. Sample size: 148 countries from 1984 to r ranges from -1 to 1 an estimates correlation between the error terms of the selection an outcome equations. force not to ratify. We can estimate the ifference in the likelihoo of Systemic Torture across these outcomes to illustrate the effect of commitment to the CAT regarless of selection status (von Stein 2005, 619). Figure 1 shows the effect of CAT Commitment on the probability of Systemic Torture across the range of Executive Job Security for each of these counterfactuals. To create these graphs, we compare the preicte probabilities of Systemic Torture for the observe values of our variables to the preicte probabilities that woul have occurre ha a committe state faile to commit an a noncommitte state been force to commit. For Figure 1(a), we followe the technique escribe in Brambor, Clark, an Goler (2006) an estimate the preicte probability of a state engaging in Systemic Torture using the parameters estimate for uncommitte states, with the values of our inepenent variables set at the means from the uncommitte subsample of states. We then use the estimate parameters from the subsample of states that i committothecattopreictwhatthe uncommitte states woul have one with the same mean values of the inepenent variables but a ifferent treatment. Figure 1(a) plots the ifference in these preicte probabilities between committe an uncommitte states across the observe range of Job Security. This ifference is known as the Average Treatment Effect for the Controls (ATC). By estimating the probability with which a state woul engage in Systemic Torture by varying only one imension its commitment status we can illustrate the preicte effect that committing to the CAT woul have ha on a nonobligate state. Figure 1(b) shows the opposite counterfactual: the preicte probability of Systemic Torture using the estimate parameters for committe states with the

10 406 courtenay r. conra an emily hencken ritter values of the inepenent variables set at the means of that subsample. After estimating the probability of Systemic Torture that woul have occurre ha committe states faile to commit to the CAT, we plotte the ifference in these values across the range of Job Security. In short, Figure 1(b) shows what woul happen if we force all CAT-committe states to renege on their commitment, representing the Average Treatment Effect of the Treate (ATT). Thus, as we o when eriving preictions from our theoretical moel, these figures illustrate how the choice to repress by a given state (represente by its values of the inepenent variables) iffers epening on whether it is committe to the treaty. In both Figures 1(a) an 1(b), the soli lines plot the preicte effect of CAT Commitment on the probability of Systemic Torture for each level of Job Security across its observe range. The estimate effects are statistically significant when the upper an lower bouns of the ninety-five percent confience intervals (shown with ashe lines) o not encompass the horizontal zero lines. OurtheorysuggestsCATcommitmentwillhavea variable effect on repression as a function of job security. When a leaer is politically insecure, that leaer will repress as neee an will be neither eterre nor encourage by any omestic effects of an IHRT. As a leaer s job security increases, however, commitment to a treaty can result in higher prospects for costly litigation that will lea the executive to repress less than woul otherwiseoccur.figure1supportsthesepreictions. Consier the Average Treatment Effect of committing to the CAT for uncommitte states epicte in Figure 1(a). When job security is relatively low, the ashe lines of the estimate change in the probability of systemic torture resulting from commitment surroun the zero line of no effect, suggesting that commitment has no effect on whatever torture ecision authorities make when they are vulnerable to removal. As we move from left to right an executive job security increases, state authorities become significantly less likely to engage in systemic torture as a result of the omestic effects of CAT commitment. Otherwise put, the average uncommitte state in which the leaer has a strong hol on power woul repress less if it were committe to the CAT. More precisely, CAT Commitment ecreases the likelihoo of Systemic Torture by nearly ten percent. This is a very large substantive effect, particularly given the notable absence of positive effects on human rights practices in previous IHRT stuies. The ATT in Figure 1(b) also supports our preictions. This figure represents the change in the probability of Systemic Torture if a committe state were not committe to the CAT. For leaers vulnerable to turnover, such a change woul have no effect, suggesting leaers of this type are not consiering the omestic political effects of the treaty in their torture ecisions. As leaers become more secure, however, not being committe to the treaty leas to a statistically an substantively significant increase in the likelihoo of engaging in Systemic Torture as compare to what the authorities woul o uner the international obligation. A brief illustration highlights the processes we fin in our more generalize cross-national stuy. Contrast the use of torture by Côte Ivoire with that of Bolivia following their respective commitments to the CAT. Côte Ivoire ratifie the Convention in 1995 uner Henri Konan Beie, who was relatively secure in office. 27 High levels of torture in 1995 fell in the two years following CAT commitment an increase only in the year before Beie lost power in a military coup. This suggests authorities willingness to reuce torture uner the obligation while Beie s office was secure, even uner a fairly ineffective juiciary (Côte Ivoire averages a low in the Linzer an Staton (2012) scale of juicial effectiveness). By comparison, Bolivia ratifie the CAT uner Hugo Banzer Suarez, a comparatively insecure leaer. 28 Miling levels of torture in 1999 increase slightly in 2000, returning to pre-2000 levels in the following years. In other wors, commitment to the treaty seeme to o little to alter torture ecisions in Bolivia. Although not a systematic case stuy, this vignette supports our theory: secure leaers torture less uner obligations to the CAT, while insecure leaers fail to respon to the treaty. 29 Conclusion an Implications In this article, we examine how state authorities navigate conflicting omestic an international pressures when eciing how much to repress. States repress to control mobilize challenges, but authorities must consier the constraining potential of the omestic 27 Accoring to our ata, Executive Job Security equalle in Côte Ivoire in Our measure of Executive Job Security is coe for Bolivia in Although Bolivia has more effective omestic courts than the Ivory Coast (Bolivia is coe on the Linzer an Staton (2012) measure of juicial effectiveness, while Côte Ivoire averages a low ), there is little change in juicial effectiveness over time in either state. Thus, the cross-national ifference in job security suggests that the treaty is responsible for these ifferences in human rights practices.

11 treaties, tenure, an torture 407 juiciary on which IHRTs have a small but meaningful effect. We argue that authorities balance these incentives epening on their job security: Leaers vulnerable to turnover will respon to increase mobilization with increase repression, even though it is likely they will be hel accountable for their actions in a omestic court. Conversely, leaers sitting securely in power will repress less in light of their treaty commitments to avoi potential court costs. We fin that IHRT commitment has no effect on the state s ecision to repress when leaers are vulnerable to removal but leas to lower levels of repression as leaers become increasingly secure in power. Following Pierskalla (2010), we present a theory that enogenizes repression an issent for committe an uncommitte states to yiel preictions as to how treaties affect the incentives to repress uner the expectation that changes will also impact the ecisions of groups to mobilize. If the expectation of costs resulting from the international obligation will constrain leaers from violating rights, this alters citizens incentives to mobilize, which in turn effects repression choices. Treaties thus impact not only human rights practices but the conflict of repression an issent as a whole. Although we focus our empirical preictions an tests on the implications of the theory for repression, these implications are rawn from this ynamic theory. Future work will focus on its implications for the effects of treaty commitment on mobilization. Unlike the ominant tren of scholarship on international human rights law, we fin that IHRTs have a positive effect on rights protections when leaers aresecureinoffice.scholarshavefounthatomestic institutions ecrease repression, but there has been very little support that international institutions o so. Many scholars have explicitly foun treaties to have either no effect on rights practices or even to lea to an increase in violations. In contrast, we fin that, even if international law has a small impact on omestic politics, these effects can yiel a substantively meaningful reuction in rights violations when leaers are secure in office. We also contribute to the scholarly unerstaning of the role of political survival on omestic political processes. Tenure consierations affect the state s willingness to repress (Ritter 2013; Young 2009), as well as its tactics. However, to our knowlege, scholars are only beginning to examine how job insecurity meiates the effect of institutions on state action. Our theory implies that job security has a mitigating effect on the ability of omestic institutions to constrain authorities. In particular, we argue an fin that institutional constraints iffer in salience from behavioral pressures. Authorities must prioritize one over the other as a function of the threat each type of influence represents to his position in power. Interestingly, authorities who sit securely in power are more likely to respon to the prospect of institutional constraint than more vulnerable leaers. Insecure leaers will repress to control angerous mobilization in orer to stay in power an international treaty obligations cannot eter vulnerable leaers from this response of self-preservation. Although secure leaers will be comparatively prone to high levels of repression, these are the conitions uner which treaties are most likely to have an effect, such that actors are able to evelop omestic an international constraints to a point of effectiveness in a more stable environment for change. In fact, the effect of the treaty is greater in magnitue in states with weak courts; increasing the probability of litigation via international commitment has greater substantive effect when the juiciary is ineffective than the same international boost when the omestic court is alreay effective. Although we focus on the effect of omestic courts in constraining repression, executives shoul also consier the potential for international ajuication. If effective international courts begin to systematically try violators of international human rights law, executives may be issuae to violate rights even if their omestic courts are relatively ineffective. Uner the CAT s universal jurisiction clause, for example, executives may also nee to consier omestic court costs in states other than their own. In February 2011, former U.S. Presient George W. Bush cancelle a trip to Switzerlan, an action that may have occurre over concerns about being arreste in Geneva for allege torture in Guantanamo Bay. 30 If states anticipate the international costs of violations resulting from their obligations to protect rights some of which may be borne even after their removal from office they may be less likely to engage in repression. Acknowlegments We are grateful to the Department of Political Science an Public Aministration at the University of North Carolina at Charlotte an the Department of Political Science at the University for Alabama for support uring the course of this research. Thanks to Phil 30 Bush Cancels Visit to Switzerlan Due to Threat of Torture Prosecution, Rights Groups Say. Huffington Post 5 February Available from URL: 02/05/bush-switzerlan-torture_n_ html.