Reconsidering the costs of commitment: Learning and state acceptance of the UN human rights treaties’ individual complaint procedures

Abstract

How do states react to adverse decisions resulting from human rights treaties’ individual complaint procedures? While recent scholarship has shown particular interest in states’ reactions to international court judgments, research on state behavior vis-à-vis an increasing treaty body output remains scarce. I argue that states generally want to avoid the costs implied by adverse decisions, or ‘views’. Rising numbers of rebukes lead them to update their beliefs about the costliness of complaint procedure acceptance in a Bayesian manner. As a result, states become less inclined to accept further petition mechanisms under different human rights treaties. I test these assumptions on an original dataset containing information on individual complaint procedure acceptance and the distribution of 1320 views for a total number of 169 countries ranging from the year 1965 to 2018. Results from Cox proportional hazards regressions suggest that both the number of views against neighboring states and against the examined state itself decrease the likelihood of acceptance of most of the six individual complaint procedures under observation. I also find evidence that this effect is exacerbated if states are more likely to actually bear the costs of implementation. Findings indicate that the omission of further commitment can be a negative spillover of the treaty bodies’ quasi-judicial output.

1 Introduction

In 2004, the United Kingdom’s government declared that it considered the ratification of the Optional Protocol of the Convention on the Elimination of Discrimination against Women (CEDAW-OP) as an “experiment” with the aim to “enable [the government] to consider on a more empirical basis the merits of the right of individual petition which exists under a number of UN treaties” (JCHR, 2006). Specifically, government officials feared that the “extensive use of individual petition […] could lead to a significant cost to public funds” (JCHR, 2005). Furthermore, the UK government vindicated its inaction towards ratification of other UN human rights treaty individual complaint procedures (ICPs), such as the Optional Protocol of the Convention on the Rights of Persons with Disabilities (CRPD-OP), by putting forward the argument that the “empirical evidence” garnered during the CEDAW “experiment” would have to be taken into account when deciding on the ratification of the ICP at question (JCHR, 2015; Lords Hansard, 2007).

The United Nations treaty bodies’ individual complaint procedures (ICP) serve as a crucial mechanism for individuals to seek justice and hold states accountable for their human rights violations (Fox Principi, 2017; Harrington, 2011; Schoner, 2023). However, we know little about how states react to the mounting number of views arising from these procedures (Çalı & Galand, 2020). After all, they can imply substantial action by state authorities ranging from the payment of financial compensation to legislative changes (Fox Principi, 2017). The above anecdote from the UK suggests that governments can be quite sensitive toward the risk of such costs. What is more, the UK’s practice of using information on the costs and benefits of one ICP before deciding on the acceptance of another ICP under a different human rights treaty suggests that ex-post costs caused by one or more ICPs may have a considerable influence on future ICP acceptance. Does the output of ICPs affect states’ willingness to accept more of those monitoring mechanisms?

Focusing on state commitment to the ICPs under six UN human rights treaties,¹ this article advances the notion that states react to mounting criticisms from the treaty bodies by withholding their acceptance of further ICPs. As a causal mechanism linking ICP output and state commitment, I propose that states undergo a learning process that resembles a Bayesian updating framework. Building on the literature on rational and bounded learning in international investment law (Haftel & Thompson, 2018; Poulsen & Aisbett, 2013), I argue that before committing to ICPs states make cost-benefit assessments that are affected by the information about increasingly active treaty bodies issuing more and more decisions in individual cases. Once they experience the cost implications of an increasing number of adverse decisions, state officials will become wary of their previous cost assessment and adapt it according to the newly found facts. The pivotal implication of this learning process is that new information on costs associated with these decisions, or ‘views’,² can discourage states from further accession to ICPs. The main aim of this article is not to provide a general account of ICP commitment but to better understand the role of the suggested learning mechanism in states’ acceptances of international obligations.

I evaluate the above propositions using survival analysis on data from an original dataset that contains information on state acceptance of all active ICPs, and the distribution and implementation of 1320 treaty body decisions against a total of 169 countries between 1965 and 2018. The results of the empirical analysis suggest that states learn from views that are issued against other states. First-hand experience of views, on the other hand, is only associated with a lower probability of ICP acceptance if the concerned state is also likely to comply with the substance of the views.

The findings of this study have major implications for the literature on commitment to and compliance with international human rights treaties and their monitoring bodies (Comstock, 2021; Hathaway, 2002; Huneeus, 2013; Simmons, 2009). By showing that there is a trade-off between compliant behavior and further commitments, the article highlights possible conflicts between the provision of individual justice and state commitment to international human rights law. In addition, this study speaks to scholarship on nation-states’ contestation of the global liberal order in that it provides novel evidence for the notion that states are highly susceptible to the growing intrusiveness of international organizations into what nation-states regard as their domestic affairs (Börzel & Zürn, 2021; Eilstrup-Sangiovanni & Hofmann, 2020; Tallberg & Zürn, 2019). Specifically, the results indicate that state actors may react to this intrusion not only by employing the usual tactics of backlash (Alter et al., 2016; Madsen et al., 2018) and withdrawal (Pauwelyn & Hamilton, 2018; Soley & Steininger, 2018) but also by refraining from further commitments to the respective legal regime. The suggested learning mechanism sheds light on the limits of individual petition in international human rights law specifically (Steinert, 2024; Montal & Pauselli, 2023; Stiansen & Voeten, 2020), and on a novel aspect of the ongoing process of ‘de-judicialization’ of international politics more generally (Abebe & Ginsburg, 2019).

2 The benefits and costs of individual complaint procedures

As of 2023, eight out of the nine core UN human rights treaties comprise an active petition mechanism that allows individuals to file complaints if they deem their guaranteed rights violated by state entities. If the respective committee has rendered a decision on the merits and established a violation of the underlying treaty in an individual case, it issues a ‘view’ that asks the respondent state to provide redress and reparation to the complainant(s).³ These complaint procedures have become ever more popular among stakeholders with treaty bodies currently processing several hundred communications per year.⁴ Acceptance of ICPs by member states is optional and done by ratifying or acceding to an Optional Protocol or by making a declaration under a specific convention article.⁵ If they are optional, why do states accept such monitoring mechanisms that allow individuals to challenge their authority?

2.1 ICP acceptance

There are two stages at which treaty commitment provides different costs and benefits to the acceding state: (1) when the state enters the treaty and (2) later when it complies with the treaty (van Aaken & Simsek, 2021, p. 206). As this applies, ceteris paribus, also to state commitment to ICPs, this section discusses the costs and benefits involved in the ICP acceptance itself before engaging thoroughly with the costs and benefits of compliance with the ICP output in the next section.⁶

At the initial stage of entering an ICP, the benefits of acceptance may seem to outweigh the costs. Commitment to international human rights norms and their legal manifestations serves as a marker of legitimate statehood and status within the liberal international order (Donnelly, 1998; Geisinger & Stein, 2007; Reus-Smit, 2013). States may be able to gain legitimacy and esteem among domestic or international audiences by committing to a strong mechanism that ensures compliance with the norm(s) enshrined in the treaty (Comstock, 2021; Lohaus & Stapel, 2022) and allows for systematic monitoring and ranking of member states (Kelley & Simmons, 2015). A better reputation among other states may be an end in itself or it may serve as a way to gain more tangible benefits from ICP acceptance, such as development aid or more favorable trade relations. Smith-Cannoy (2012), who was the first to study the puzzle of the acceptance of human rights treaty ICPs as distinct from treaty ratification, argues that states ratify ICPs to gain legitimacy among external and internal pressure groups, especially during times of economic hardship. In a study on the CCPR-OP, Schoner (2023) contends that repressive governments ratify the ICP to increase their economic benefits from aid and trade relations with the European Union. She finds, however, that this positive effect of economic benefits on ICP ratification depends largely on the absence of domestic constraints, that is, domestic institutions that would increase the costs of commitment by pressuring the government to later comply with the committee’s decisions. In other words, positive inducements towards ICP commitment at the entry stage are overshadowed by the costs that emerge at the compliance stage. Thus, reputation benefits gained through ICP acceptance can only be sustained through compliance later on.

The importance of compliance costs for states’ ratification decisions is well understood in the literature on commitment to human rights treaties. In her seminal work Mobilizing for Human Rights, Simmons (2009, p. 64) argues that “governments are more likely to ratify human rights treaties […] with which they can comply at a reasonable cost” and finds that cost avoidance is a fundamental logic behind ratification or non-ratification of several human rights treaties and the CCPR-OP. Hathaway (2003, p. 115) points out that countries considering joining an ICP not only take into account future compliance costs but also “the probability that the costs of compliance will actually be realized.” She finds support for this assertion in ratification patterns of the CCPR-OP and CAT Article 22 (Hathaway, 2007).

These studies show that ICP acceptance implies certain utility in and of itself but that any cost-benefit assessment that does not go beyond the initial commitment stage and also looks at the costs and benefits of compliance later on remains incomplete. Scholarship has thus shifted away from static, macro-level explanations for commitment and compliance and instead acknowledged that “regularly receiving and implementing recommendations from human rights institutions forces states to re-evaluate and revise the extent of their commitment to and compliance with those human rights instruments, institutions, and underlying norms” (Haglund & Hillebrecht, 2020, p. 650).

Despite widespread agreement on the relevance of future compliance costs to treaty ratification, there has hardly been any systematic examination of how states react to the immediate costs of the treaties’ compliance monitoring mechanisms. As a basic principle, I expect the logic of costs to become more important the more effectively a treaty obliges states to align their actions and policies with the treaty provisions. In this regard, ICPs are arguably among the most rigorous monitoring provisions available to UN human rights treaty bodies (Cole, 2012; Dutton, 2012, p. 34). Unlike other monitoring mechanisms, such as the treaties’ reporting procedures, ICPs create very narrow and specific obligations for state parties, if only in individual cases. More precise obligations, then, also create costs that are more targeted and appear more urgent than the diffuse costs stemming from general provisions (Çalı & Koch, 2017, p. 49; Fox Principi, 2020). This should make ICP acceptance particularly prone to cost-benefit arguments. After all, as Dinah Shelton (Shelton, 2015b, p. 28) has noted aptly, “States do not like to be criticized and they do not like to lose cases”. I argue that in order to identify the costs and benefits of ICPs for states one needs to look at their procedural outcomes – the views. They call on the respondent state to take concrete measures to remedy the harm done and rectify the deficiencies in its practices. Such measures can include the payment of financial compensation, medical treatment, reopening of proceedings, or more general recommendations such as the training of judges and law enforcement, and law amendments. Although treaty body views are not legally binding, they are the outcome of a “quasi-judicial” process and are difficult for the involved states to ignore (Ford, 2022; von Staden, 2018). Law-making by the UN treaty bodies can translate into a normative quality that requires states to give up their sovereignty – to some degree – to international non-state actors (Reiners, 2021; Ulfstein, 2014).

I argue that, with the oldest ICP being active for over 60 years and with well over 1400 adverse views issued under the eight active ICPs (Ullmann & von Staden, 2023, p. 5), there is enough information available to states that want to assess the actual costs of commitment.

2.2 ICP compliance

The question of how costs shape state behavior in the area of human rights treaties has been picked up by several studies, albeit with varying levels of theoretical elaboration (Chayes & Chayes, 1998; Cole, 2012; Downs et al., 1996; Gauri, 2011; Goodliffe & Hawkins, 2006; Mulesky et al., 2020; Vreeland, 2008). All these studies share the notion that high compliance costs in the future are likely to constitute an impediment to states’ ratification of human rights treaties in the present. But with what kind of costs—or benefits for that matter—do states face once they have accepted an ICP and become subject to adverse views?

In light of the present study’s objective, it is helpful to differentiate the costs of views along two dimensions: (1) the materiality of the costs and (2) their dependence on compliance. Table 1 classifies four types of costs along these two dimensions. The different types express different levels of intrusiveness into the domestic affairs of the responding state which results in higher costliness. To be clear, all views entail sovereignty costs for states in that they constitute an attempt by an international expert body to alter domestic rules and procedures. However, if a state chooses not to comply with a view this intrusion remains hypothetical, as it did not lead to any change in state behavior. A state that does not comply will have to fear mainly reputation costs – a prospect that should not deter all states equally (Adler-Nissen, 2014; Debre, 2022). If a state chooses to comply, on the other hand, it is directly and concretely confronted with both the tangible fiscal costs that stem from implementation and the possibly more momentous non-tangible political costs that arise with a narrowing of its policy options. Equally, views that aim at more general state practices will have stronger cost implications for the responding state than views that demand, for example, only damages in the individual case. Generally, I expect views to be perceived as more costly by states if they are complied with and if they entail consequences beyond the provision of material remedies. I will describe the four types of costs in more detail below.

Table 1 Types of costs implied by views

Sanctions

A state’s non-compliance with views might result in material costs through shaming by INGOs or other states. There is evidence that naming and shaming of international human rights bodies is associated with a decrease in aid flows (Dietrich & Murdie, 2017; Lebovic & Voeten, 2009; Schoner, 2023), trade (Peterson et al., 2018), and foreign direct investment (Vadlamannati et al., 2018). Since the treaty bodies’ reports and views can be understood as naming and shaming devices (Kahn-Nisser, 2021; Schoner, 2022), it is plausible that they have material consequences for the shamed state. The treaty bodies’ annual and follow-up reports are publicly available and provide first-hand information on a state’s ability and willingness to protect and promote human rights within their own territory. The emerging practice among treaty bodies to assign grades in their follow-up reports on views could further enhance the visibility of non-compliance and mobilize third actors (Kelley & Simmons, 2015). However, it is difficult to trace this mechanism in practice, and to date, there is no empirical study exploring a possible link between treaty body views and sanctions. The relatively low visibility of the follow-up reports and the open-ended nature of the implementation process make the tracing of such a link rather difficult. Ultimately, the idea that the treaty bodies’ views and reports can lead to material sanctions remains hypothetical.

Reputation costs

Views constitute a public attestation that a state has violated the obligations it has imposed on itself through treaty ratification. Violating international commitments signals a willingness to ignore international law and therefore comes with a loss of reputation (Guzman, 2005). Committing to a human rights treaty while later not complying with its provisions or its monitoring body’s decisions could harm a state’s reputation among domestic (Kuzushima et al., 2023) and international audiences and lead to stigmatization and outcasting (Adler-Nissen, 2014; Carraro et al., 2019; Kelley & Simmons, 2015) and, hence, restrict its future scope of action. A loss of reputation can translate into a shrinking of future opportunities for cooperation (Guzman, 2005) and a decrease in soft power to influence and persuade other states (Koh, 2007). Reputation costs can occur (1) when the Committee establishes a violation and, thus, finds that the State party did not comply with the treaty provisions in a specific case, and (2) when a state does not implement the remedies requested in a view, in which case they can be understood as the alternative to fiscal and political costs, which emanate directly from the implementation of remedies.⁷ Reputation costs are akin to audience costs, which arise when “audiences punish policymakers for committing to one policy and then reneging on that promise” (Chaudoin, 2014, p. 235). Such behavior is costly because it gives domestic and international audiences “the opportunity to deplore the international loss of credibility, face, or honor” (Fearon, 1994, p. 581).

Fiscal costs

These are the most evident cost implications that stem from the specific remedies the treaty bodies demand from the violating state. Such costs arise, for example, in the form of direct payment of financial compensation to victims, the performance of concrete remedial actions such as the granting of medical treatment, and the reopening of proceedings in individual cases. The mere organizational costs of setting up a reply to the treaty body’s follow-up mechanism may be regarded as unpleasant by state authorities. However, general remedies such as training courses for judges can also fall under this category. Such costs can be translated into monetary expenses and, thus, are easy to pin down.

As an illustration of how direct costs work in this context, consider the view regarding communication L.N.P. v. Argentina in which the Human Rights Committee (HRCttee) found that Argentine public officials’ treatment of the claimant in the aftermath of a sexual assault amounted to several violations of the Convention on Civil and Political Rights (CCPR/C/102/D/1610/2007, 2011, para. 13.9). As an individual remedy, the HRCttee requested the full implementation of an amicable settlement procedure between the state and the author of the communication (CCPR/C/102/D/1610/2007, 2011, para. 14). About a year later, the State party informed the HRCttee that it had taken steps to implement the required individual remedy (HRC, 2013, pp. 143–144). These included the payment of US$ 53.000 as compensation to the author, a monthly life pension, the construction of a house for the author and her family, the granting of a scholarship, and the sanctioning of the doctor who failed to properly examine the author. These are clear-cut expenses for the State party that arose through the ICP.

Political costs

These costs arise when a view has implications that restrict or modify a government’s policy space and as a consequence “weaken its domestic hold on power” (Sandholtz et al., 2018, p. 161). This category captures most of what others have coined “limited flexibility” (Goodliffe & Hawkins, 2006, p. 22), and “sovereignty costs” (Abbott & Snidal, 2000; Krasner, 1999; Moravcsik, 2000), both of which denote the possibility of interference of other states or international organizations into a state’s ‘sovereign’ freedom to set policies as it sees fit. Yet, political costs refer only to non-pecuniary measures that evince more far-reaching implications in the responding state’s internal affairs.⁹ Remedies formulated in treaty body views may imply different degrees of political costs. Individual measures, such as freeing a publicly known political prisoner, can have political consequences for the government that will exceed the mere financial or organizational costs of the release. General measures, such as legislative change concerning a certain issue, are even more likely to substantially restrict the government’s freedom to choose the policies it regards as optimal for years to come.

An example of political costs stemming from UN treaty body views is the communication Yuri Bandajevsky v. Belarus. In 1999 Bandajevsky, then professor and rector of a public medical institute, was arrested on corruption charges and later sentenced to 8 years of imprisonment. The victim turned to the HRCttee, claiming that his conviction was politically motivated and likely triggered by a critical report of his on the effects of the Chernobyl disaster on Belarus, “which was very different from the official position of the Government“ (CCPR/C/86/D/1100/2002, 2006, para. 3.9). The assertion that Mr. Bandajevsky was a political prisoner due to his open criticism of state authorities was also confirmed by Amnesty International (Amnesty International, 2001). In March 2006, the HRCttee found Belarus to have violated several treaty provisions through the arbitrary arrest of Mr. Bandajevsky (HRC, 2006). Five months after the issuance of the HRCtte’s view, the victim informed the committee that he was released from prison without compensation (Human Rights Committee, 2007, p. 660). The view implied high political costs because, for the Belarusian government, the freeing of a public figure who dares to openly criticize government decisions constitutes a potential threat to its autocratic rule.

By creating a broad spectrum of liabilities, ICPs make it hard for states to escape the costs of views, particularly if they choose to comply with the latter. Granted, the costs implied by a single view should be (mostly) negligible for states, which makes them low-impact events.¹⁰ This means that the experience of a single view may not have a measurable effect on state behavior concerning the acceptance of ICPs but that the accumulation of several views and their implied costs should make an impression on states. Given that compliance with views usually takes several years (Ullmann & von Staden, 2023, p. 11), the costs of multiple views against the same state are likely to build up over the years. Since I expect states to have a vested interest in avoiding these costs, I suggest that views provide information for states on the ‘true’ costliness of ICPs. The next section will set out the mechanism through which state actors process the information that views entail about the costliness of ICPs and what this may mean for states’ willingness to accept them.

3 Treaty body views and state learning

I have shown above that the treaty bodies’ views can imply very specific costs for the respondent state. Another characteristic of the procedure is the high uncertainty as to whether costs will occur and when. Consider that, empirically, the distribution of views among member states casts doubt on the notion that better human rights practices will lead to fewer rebukes. In fact, it is even possible that a country’s overall respect for certain human rights and its likelihood of receiving views are not related at all.¹¹ Over the years, countries such as South Korea, Australia, Denmark, and Uruguay have received a considerable number of views despite their democratic credentials and their reputation as human rights-respecting states. Thus, states receiving high numbers of views do not necessarily have a bad human rights performance in general. Since ICPs concern only singular cases, their acceptance can turn out to involve considerable costs even to states whose laws and practices are generally in line with treaty provisions. Additionally, a significant backlog in the processing of complaints makes it even more difficult for states to anticipate when and how often they are going to receive a view (Callejon et al., 2019). This creates uncertainty which, in turn, has strong implications for supposedly rational state actors trying to calculate the costs and benefits of ICP acceptance. That is because if a state cannot realistically extrapolate the volume of complaints it will receive from the ICP under consideration,¹² that state will find it impossible to make a rational choice on its acceptance. Even for states with democratic credentials and high legal and bureaucratic capacities, making such calculations can be challenging and can result in errors (Putnam, 2020, p. 34). This is why I expect states to be highly susceptible to the information about the costs of ICP commitment implied by views.

I argue that, at an early stage of the ratification process, each state makes its own assessment of the costs of commitment by extrapolating the probability of being targeted by views from a place of uncertainty. At this stage, the benefits involved in commitment mentioned earlier may override fears of compliance costs and lead states to accept one or more ICPs. However, after the state has witnessed a growing number of views against itself or others, it can update its earlier conclusion and expect higher costs as a result of which its ratification of further ICPs becomes less in line with its preference for cost avoidance and, thus, less likely. Certainly, the costs implied by ICP jurisprudence can be understood as costs that are difficult to avoid ex post by the states receiving the views, as the decision to ratify the ICP generating these views is not easy to reverse.¹³ Yet, I argue that once a new ICP under a different human rights treaty is opened for ratification states will incorporate in their decision-making the information on future costs provided by the precedents of earlier ratified ICPs, as exemplified by the UK case above.¹⁴

More precisely, I expect this process to resemble a Bayesian updating mechanism where the probability of a state accepting an ICP depends on the state’s updated belief about the costliness of acceptance. Such updated belief is the result of the state’s prior belief about the costs and benefits times the new information on the likelihood of costs provided by views. In other words, in an uncertain environment, views can be seen as carriers of information about the true costliness of ICP acceptance. Building on its initial cost assessment, the state takes this information into account and makes a new cost-benefit analysis. This updated cost-benefit assessment informs subsequent state action. If, in the new assessment, the costs exceed the benefits for the state, it will not accept a further ICP. The more views a state receives over time, the higher its certainty regarding the probability of costs. The Bayesian updating mechanism thus serves mainly as an ideal type for the basic idea of a rational actor updating its beliefs based on observed events in a situation of uncertainty (Checkel, 2001, pp. 555–557; Meseguer, 2006; Dunlop et al., 2020).

As an ideal type, the updating framework is subject to several qualifications in practice. Most importantly, although states make rational choices according to their preferences, one of which is to avoid costs that exceed related benefits, this “intendedly rational” behavior is bound by a “human cognitive/emotional architecture” (Frieden, 1999, p. 41; Jones, 1999, p. 298).¹⁵ I posit that the imperfect processing of new information does not invalidate the underlying rational updating mechanism but only exposes it to some constraints.

First, bounded rationality means that the avenues through which states receive information on ensuing costs are likely to play a role. Since the treaty bodies regularly publish their views in follow-up reports and also include them in their annual reports to the UN General Assembly, states can gain access quite easily to information on views against other states. States could learn from other states’ cost experiences and in light of new information could reassess their own risk of receiving costly views (Acemoglu et al., 2011; Smith & Sørensen, 2000). Poulsen and Aisbett (2013) have traced such a learning mechanism in the case of bilateral investment law. They have found evidence that states are less likely to sign bilateral investment treaties the more states within their region have been affected by litigation resulting from these treaties. There is no compelling reason why states should not learn from other states also in the area of human rights treaties. For example, there is evidence that European states closely monitor judgments of the European Court of Human Rights against other countries to avoid future litigation against themselves (Küçüksu, 2023). It seems plausible that states would also take into account the treaty bodies’ decisions against other states.

States may monitor treaty body views against any state to receive information about the potential of receiving views against themselves. In this case, the overall activity of treaty bodies under their ICP may already convey important information for states considering ICP acceptance. Yet, availability heuristics could attenuate the impact of all views issued by the treaty bodies on state behavior as state actors are likely to use cognitive shortcuts and limit the information they process using geographical, cultural, and temporal boundaries (Weyland, 2008, p. 292). Boundedly rational states could thus be more strongly affected by the experiences of states within their immediate neighborhood or of like-minded states than by the experiences of all states globally. Again, research on bilateral investment treaties has shown that state behavior is affected by the experiences of other states in relative geographical proximity (Poulsen & Aisbett, 2013) but not by the experiences of all other states (Haftel & Thompson, 2018). Geographical proximity might not be the only catalyst for learning, however. Research has shown that International Organizations (IOs) are “important channels through which states can influence, and be influenced by, the behavior of others” (Greenhill, 2015, p. 3). The experience of states that are members of the same regional IO as the learning state, may thus also have an impact on ICP acceptance. Finally, states may be particularly prone to learn from like-minded states. For example, Pegram (2010) has shown that democracies mainly emulate the design of national human rights institutions from states with a similar political regime. Learning from like-minded states is not restricted to democracies, however, as research on cooperation between autocracies in international fora has shown (von Soest, 2015; Debre, 2022).

Thus, I expect that:

H1a: A state is less likely to accept an ICP, the more views the UN human rights treaty bodies have issued against any other states.

H1b: A state is less likely to accept an ICP, the more views the UN human rights treaty bodies have issued against other geographically or politically proximate states.

Second, bounded rationality also implies that first-hand information, i.e. the experience of receiving views against oneself, will probably have an even greater impact on a state’s cost-benefit analysis than learning about the costs through the experiences of other states. Research on experiential learning has emphasized the essential role of direct experience in learning processes (Levy, 1994, p. 305; Haftel & Thompson, 2018). The expectation that states are most affected by direct experience of views is expressed in the following hypothesis:

H2: A state is less likely to accept additional ICPs, the more views the UN human rights treaty bodies have issued against that state.

These first two hypotheses express the varying degrees of immediacy of the information on costs at the global, regional, and state levels. I expect that as immediacy increases, the new information on costs will make a bigger impression on states and increasingly affect the likelihood that they will withhold future ICP acceptances.

Third, my learning argument is based on a logic of consequences, which implies that different cost/benefit structures have different effects on states. I have argued above that the costliness of views depends on both the materiality of the remedies and the degree of compliance. The determining factor behind the materiality of views could be the nature of the remedies. General remedies that demand actions beyond the single case typically have high political as well as reputation costs that outweigh most individual remedies. Immaterial remedies such as legislative amendments are typically subsumed under the ‘general’ category because they target collective actors rather than individuals.¹⁶ Tackling structural issues that go beyond the respective individual human rights violation should be perceived as particularly costly and thus should have a stronger impact on a state’s decision to ratify another ICP.

H3a: The more views a state has received that demand general remedies, the less likely that state is to accept additional ICPs.

The assertion that the costliness of views depends on the state’s propensity to comply with the remedies that they contain can be ascertained more easily. Both fiscal and political costs will only occur once a state takes concrete measures to implement the remedies. On the other hand, although non-compliers may face audience costs and possibly sanctions, costs of non-compliance are less direct and less likely to occur.¹⁷ I contend that the actual implementation of remedial measures should be more costly than neglecting them and will likely have a more profound learning impact on state actors. Thus, I expect the direct negative effect of views on the responding state’s further ICP ratification behavior to be strongest if that state actually complies with them. The enhancing effect of direct costs on the suggested mechanism is expressed in the following hypothesis:

H3b: The more likely a state is to comply with the remedies contained in the views it has received, the less likely that state is to accept additional ICPs.

4 States’ acceptance of ICPs: empirical analysis

To test my hypotheses, I examine states’ acceptances of six of the eight active ICPs under the UN human rights treaties.¹⁸ With a total number of 168 countries and data from the years 1965 to 2018, my primary data set is composed of 10,476 possible country-years. However, the observation periods and composition of states are different for each ICP, which lets these numbers vary between analyses. The reason for this variance is that I only examine for each ICP those country-years that are ‘at risk’ of accepting the ICP. A state enters the risk set, i.e. is observed, the year it has ratified the human rights treaty underlying the respective ICP and drops out of the risk set the year after it has accepted the ICP.¹⁹ This allows me to take into account the selection effect of treaty ratification and examine only those states that have a certain probability of accepting the ICP at some point in time. For example, France, Mexico and Belize have all ratified the CAT in 1986. However, France made the declaration under CAT Article 22 in 1988, Mexico in 2002 and Belize has to date not accepted the ICP. The years each state is under observation for the analysis of the CAT ICP thus are 2 years (France), 16 (Mexico), and 32 (Belize). Consequently, I create for each ICP a separate data set that is based on the primary data set but includes a unique set of country-years and is analyzed in separate regression models.²⁰

The main challenge with the data described above is that it contains right-censored observations, i.e. all the states that have ratified the treaty but not (yet) the corresponding ICP at the end of the observation period. Common linear regression models would treat these observations as non-acceptances and thus lose important information. To account for this, I employ survival analysis, which considers the time between a fixed starting point (treaty ratification) and a terminating event (ICP acceptance) (Bradburn et al., 2003). By incorporating information from right-censored observations, the chosen method avoids underestimation of the true but unknown time to event. More specifically, I use Cox proportional hazards regression, which allows me to determine how my independent variables influence the occurrence of the event of ICP acceptance at any point in the observation period.²¹

4.1 Dependent variable

I examine a state’s decision to accept an ICP by either ratifying a treaty’s optional protocol or making a declaration of acceptance under the article that establishes the ICP. The binary dependent variable takes a value of 1 for each year a country has accepted an ICP and a value of 0 if otherwise.

4.2 Independent variables

The independent variables of main interest are three measures that capture, first, the cumulative number of views issued by all treaty bodies against all countries other than the observed country (Views Globally). This is the first measure that is intended to capture the effect of new information on views against other states, as expressed in H1a. Second, the effect of cost information flowing from the more immediate neighborhood is measured by the cumulative number of views issued against all other countries within the region²² of the observed country (Views per Region). As alternative operationalizations of this variable, I also test the cumulative number of views against all states within the regional organization with the highest authority the respective state is a member of (Hooghe et al., 2017)²³ and the yearly number of views against like-minded states based on the regime type measure by Maerz et al. (2023). Third, the cumulative number of views issued by all UN human rights treaty bodies against an individual country in each year (Views per Country) is intended to reflect the costs of views concerning each country individually to test the effect suggested in H2. All measures of views are lagged by one year. Data for these variables is taken from the Treaty Body Views (TBVD) Dataset (Ullmann & von Staden, 2023), which contains information on 1320 communications that resulted in views issued by the eight treaty bodies with active ICPs between 1979 and 2018. I use the cumulative number of views as a measure since I expect the costs of one single view to be relatively marginal whereas repeated rebukes should increase the costs and be more strongly felt by states. The fact that the implementation of views can take up to thirty years and more (Ullmann & von Staden, 2023) adds to the notion that the costs of views at the state level can accumulate over the years. To account for the possibility that the impact of views on states’ decision-making decreases after a certain number of views, I additionally run all main models with the natural logarithm of the views variables.²⁴

The effect of the number of views per country could be subject to some qualifications. First, views vary in the degree to which they touch upon issues that are sensitive to certain states and thus may be more or less strongly felt by the responding state. To measure the issue sensitivity of the views, I create a Sensitive Views variable that only counts the views that concern issues that should be least popular for the responding state as deduced from its normative position in the Human Rights Council’s universal periodic review (Terman & Búzás, 2021).²⁵ Second, views that have been rendered against the responding state within the more recent past may weigh more strongly in states’ decisions to ratify an ICP than views that have been issued multiple years ago. To account for the possible importance of temporal immediacy, I test measures that only count the number of views a state has received within the last three and five years preceding the year of observation.

H3a stipulates that views that provide for general remedies that go beyond the reparation and restitution for the individual violation(s) will have a stronger negative effect on ICP acceptance than views that do not require such remedies. To test this claim, I create a Weighted Views per Country variable that reflects the Views per Country variable but multiplies all views that include at least one general remedy by two.

H3b expresses my expectation that receiving views will be more costly if the state complies with them. A straightforward way to approach this question in my model is the addition of an interaction term between the Views per Country variable and a variable that measures implementation. To obtain a measure that reflects the propensity of states to comply with the views they receive, I resort again to the TBVD Dataset (Ullmann & von Staden, 2023), which includes an indicator reflecting the degree to which the state party implemented the remedies of respective views until the respective year. This indicator is based on the information provided in the treaty bodies’ follow-up reports. More specifically, I use the TBVD’s aggregate compliance all indicator which considers all available compliance information and a lack of compliance information as non-compliance. Although this is a suboptimal approximation of the true state of compliance, it better captures compliance with views than proxies such as a state’s general human rights performance. The Average Compliance score is calculated by taking the mean of the TBVD’s aggregate compliance all scores for all views a country has received from the year of the first view to the observed year. It ranges from 0 (no view has been complied with) to 1 (all views have been fully complied with) and is available for each country that has received at least one view from the treaty bodies since 1979. Since the Average Compliance score is only available for states that have received views within the observation period, its interaction with the Views per Country variable weights the number of views for each country according to the degree to which they are complied with.

It should be noted that the large majority (76%) of all views were rendered by the HRCttee, which monitors compliance with the CCPR. What is more, most other ICPs became active only years after the majority of CCPR-OP ratifications had already occurred. This means that there is very low variation in the Views per Country variables in the case of the CCPR-OP – and possibly CERD Article 14 – which can create the problem of monotone likelihood because one or several levels of the Views per Country variables only receive censoring times (Heinze & Schemper, 2001). Consequently, the parameter estimates will converge toward infinity and cannot be interpreted.²⁶ I will not report the results for the Views per Country variable(s) for the CCPR-OP and CERD Article 14 where this is the case.

The first view establishing a violation was issued in 1979 by the HRCttee and the number of views rose constantly over the years to around 80 per year in the 2010s. The cumulative number of views increased most distinctly within the last 30 or so years with over 70% of all violations found only between 1997 (212) and 2018 (1320). It follows that there is a risk of serious correlation between my independent variables and time, which would violate the proportional hazards assumption underlying the Cox regression model and potentially bias model estimates. Thus, I controlled for non-proportional hazards using Schoenfeld residuals and added interactions with time to all variables with a p-value lower than 0.1.

4.3 Control variables

The analysis also includes other variables shown to influence commitment to human rights treaties and their ICPs to eliminate possible confounders. Summary statistics and a correlation matrix on the main variables used can be found in Table 2.A and Fig. 1.A in the Appendix.²⁷

The degree of democracy practiced by a country is a widely accepted determinant of both treaty ratification and ICP acceptance (Cole, 2012; Hathaway, 2007; Sandholtz, 2017; Simmons, 2009). The substantive argument is that democracies are staunch supporters of the normative substance of human rights treaties, which does not oblige them to drastically adapt their laws and practices. I use V-Dem’s electoral democracy index as a measurement of a country’s adherence to democratic principles.²⁸ Following Simmons (2009), I use a squared term of the democracy index in the analysis of the CCPR-OP to better capture her argument about the heightened propensity of more mature democracies to ratify the two ‘bill of rights’ treaties.

ICP acceptance may not only depend on the nature of the political regime overall but also on the occurrence of regime change. Newly democratized states may want to lock-in human rights policy (Moravcsik, 2000; Simmons, 2009), and newly autocratic states could want to signal continuity to domestic and international audiences (Hong, 2016; Smith-Cannoy, 2012) but may also be particularly opposed to ICP acceptance due to their general rejection of the contents of the treaties (Meyerrose & Nooruddin, 2023). Democratic regime changes should thus increase the probability of ICP acceptance, while the effect of autocratic regime changes is ambiguous. To account for the impact of regime changes, I include in the models the regime transformation indicator by Maerz et al. (2023a), which reflects whether an autocratic (-1) or democratic (1) regime change occurred, and lag it by one year.

I also control for the human rights performance of a country, as measured by V-Dem’s ‘Gender equality in respect for civil liberties’ Index for the CEDAW-OP, and Fariss’ physical integrity rights scores (Fariss, 2019) for the remaining ICPs.²⁹

The effects of regional socialization and peer pressure (Hathaway, 2003; Smith-Cannoy, 2012; von Stein, 2016) are accounted for by including the regional ratification rate, i.e. the percentage of states that have accepted the respective ICP among all states within the main regional organization.³⁰ The regional ratification rate constitutes a possible mediating variable since more views may affect the willingness of other states in the regional organization, which decreases the willingness of the observed state to accept. To account for such a possible posttreatment effect, the variable is lagged by two years (Dworschak, 2023).

As more affluent societies are generally more open to ICP acceptance (Simmons, 2009), all models also factor in the log of GDP per capita.

A further contention posits that countries with a common law tradition are less inclined towards ICPs because, in those countries, governments fear the active incorporation of treaty body jurisprudence by independent national courts (Simmons, 2009). I control for this by including an ordinal variable that reflects whether a state follows a common law tradition (2), has a mixed system with common law elements (1), or follows another law tradition (0).

5 Results

Table 2 shows the hazard ratios with the corresponding 95%-confidence intervals for the three independent variables that capture the cumulative number of views at the global, regional, and country levels. The complete results of the Cox regressions are reported in Table 4.1.A to 9.3.A in the Appendix. Note that I report the hazard ratios, which are the exponentiated coefficients. A hazard ratio above 1 indicates that the variable is positively associated with the probability of ICP acceptance, whereas a hazard ratio below 1 indicates that as the value of the variable increases, the probability of ICP acceptance decreases.

Table 2 Results for the main variables of interest

5.1 Learning from others

The results reported in the first row of Table 2 lend support to my expectation that the number of views the treaty bodies are issuing against all other states makes the acceptance of ICPs less likely. More precisely, the Views Globally variable is associated with a decrease in the ‘risk’ of acceptance of the CCPR-OP by 0.3%, of CERD Article 14 by 0.8%, of CAT Article 22 by 0.5%, the CEDAW-OP by 4.5%, the CRPD-OP by 5.7%, and of the CRC-OP3 by 4.2%.³¹ The next row reveals a similar pattern, in which an increase in the number of views against other states within the observed state’s region decreases the state’s risk of acceptance of all observed ICPs. This finding is consistent across the two alternative measures of proximity: views against other states in the main regional organization and views against states with a similar regime type.³² This supports H1b as states seem to be affected by the experiences of neighboring and like-minded states. Note that in three of the six instances, the negative association of the Views per Region variable with ICP acceptance is stronger than in the case of the Views Globally variable. This lends only partial support to the expectation that states are affected more strongly by more immediate cost information. The discrepancy between the estimates for the two variables is particularly wide in the cases of CERD Article 14 and CAT Article 22 on the one side and CRPD-OP and CRC-OP3 on the other. States accepting the first two ICPs are more strongly affected by views against other countries within their environment than by any other states whereas states accepting the latter two ICPs are less strongly affected in their decision by views against neighbors. This pattern also emerges for the other variables measuring views per group: ratification of the CRPD-OP and the CRC-OP3 is even positively associated with the Views per Regional Organization and Views per Regime Type variables. This may be explained by the varying membership compositions of these ICPs. Among the observed ICPs, CERD Article 14 and CAT Article 22 are the ICPs with the highest average democracy score among its member states (V-Dem polyarchy score of 0.74 and 0.72 respectively) whereas states that have ratified the CRPD-OP and CRC-OP3 have the lowest average democracy score (0.60 and 058 respectively). Democracies could be more likely to cooperate with and learn from other states, especially through regional organizations (Linos, 2011; Mansfield & Pevehouse, 2008), than authoritarian regimes, whose cooperation is more strongly driven by ideas of national sovereignty and non-interference (Debre, 2022). To test this claim, I plot the marginal effects of the Views per Regional Organization variable across a binary regime type variable that distinguishes between autocracies (AUT) and democracies (DEM) taken from the ERT dataset (Maerz et al., 2023).³³ Figure 2 supports my assertion of a stronger learning mechanism from fellow regional organization member states among democracies for all ICPs except the CCPR-OP and CRC-OP3.³⁴ Why do democracies ignore information coming from neighboring states in the latter two cases? An ad hoc explanation may be that democracies consider civil and political, and children’s rights as too central to their own identity so that adverse views against other states do not discourage them from ICP ratification.

5.2 Learning from direct experience

While the aforementioned findings underline the role of learning from other states, H2 expresses my expectation that the first-hand experience of states with findings of violations against themselves is an important driver of the posited learning effect. The results in the third row in Table 2 seem to substantiate this assertion only partially. The hazard ratios for the Views per Country variable point in the hypothesized direction in the cases of CERD Art. 14, CAT Article 22, and the CRC-OP, although they only reach levels of statistical significance in the first two instances. Hence, it is only in those two cases that I find evidence that a state’s subjection to views is a strong predictor of its additional ICP acceptances per se. Note, however, that – as conjectured – the lower hazard ratios suggest a stronger negative impact of direct experience of views relative to views against other states. Results for the alternative measures of the Views per Country variable that only count the views within the last three and five years and those concerning sensitive issues are reported in Table 4.3.A to 9.3.A in the Appendix and reveal a very similar pattern.

The results for the CEDAW-OP in Table 2 even run contrary to my expectations and suggest a positive association between the number of views a state has received and that state’s willingness to accept the ICP. Why would states be encouraged to ratify the CEDAW-OP when they have received views from the other treaty bodies? This might be due to the nature of the rights covered by this treaty. Although women’s rights are still largely violated in most states around the world (Richards & Haglund, 2015; WomenStats, 2024), they seem to be a popular human rights issue that many different states can agree on irrespective of their ideological or cultural background (Terman & Búzás, 2021; Comstock & Vilán, 2024). Thus, states may, in general, be more open toward ratifying the CEDAW-OP, than toward other ICPs. States that use ICP ratification as a signaling device to internal and/or external audiences may be particularly likely to ratify the CEDAW-OP when they learn that ICPs ‘do bite’ as this makes ratification more costly and thus more effective (Hafner-Burton et al., 2015; Schoner, 2023). Research has shown that autocracies ratify the CEDAW because the women’s rights treaty allows them to signal adherence to democratic norms while bearing only relatively low political costs (Donno et al., 2022). This would mean that the CEDAW-OP attracts a relatively large number of ‘insincerely committing’ governments that understand views against themselves primarily as a way to make their commitment more convincing, e.g. to shield them against domestic opposition forces. Surely, this logic only makes sense if these states keep actual costs low by not complying with the views they receive. Such insincere commitments would bias any direct effect of views because if states do not experience views as particularly costly, they will probably not make a lasting impression on them.

5.3 Learning conditional on remedial scope and compliance

H3a and b state that the effect of the Views per Country variable may depend on two intervening factors—the amount of costs that the responding state has to bear via implementation and the scope of remedial action. The results of the Cox regressions reported in Fig. 2 provide supportive evidence for one assertion but not the other. The plot reports for each ICP the hazard ratios first of the Views per Country variable, second of the Weighted Views per Country, and third of the interaction term between the Views per Country and the Average Compliance variables. The biased results for the CCPR-OP are excluded intentionally – as explained on page 19. It reveals that the Weighted Views per Country variable yields quite similar hazard ratios as the basic Views per Country variable. Thus, giving more weight to views that contain general remedies has no impact on the effect of the number of views a state receives on its decision to accept either of the five ICPs. In the case of CAT Article 22, general remedies even seem to weaken the negative effect of views. Since the CAT governs human rights norms that are often perceived as a threat to regime survival (Conrad & Ritter, 2013), views that target the laws and practices of the respondent state concerning physical integrity rights may be more readily rejected than views that only aim for restitution in an individual case. Views that are not implemented, in turn, will likely be less effective in changing state behavior towards ICP acceptance.

Figure 2 provides supportive evidence for this assertion. It shows that there is a negative association of the interaction between the Views per Country and the Average Compliance variable with the acceptance of CERD Article 14, CAT Article 22, the CEDAW-OP, and the CRPD-OP. The correlation between the interaction term and ICP acceptance is also considerably stronger now as expressed by lower hazard ratios compared to those obtained for the Views per Country variable alone. However, the interaction effect fails to reach statistical significance in the case of the CRC-OP3.³⁵ An ad hoc explanation for this finding could be that it is the only ICP in our sample that remained virtually inactive during the observation period.³⁶ Based on this information, states could extrapolate a comparatively low risk of being scrutinized under this ICP, despite their first-hand experience of the costliness of views from other ICPs. Either way, the CRC case is also the least informative one, as it offers only very low numbers of events (acceptances) per variable. This means that coefficients are prone to bias, and thus, statistics to test their significance can lose validity (Peduzzi et al., 1995). Interestingly, in the case of the CEDAW-OP, the interaction term indicates a strong negative association with acceptance while the other two Views per Country variables suggest a positive association. In other words, states that receive views and do comply with them have a significantly lower probability of accepting this ICP than states that do not receive views or receive views but do not comply with them. This lends support to my assumption that the positive correlation between the number of views per country and CEDAW-OP ratification that I found earlier is mainly due to a high number of ‘insincere commitments’ that are not followed by compliance.

To make more sense of the interaction effect, I illustrate it in Fig. 3 using marginal effects plots.³⁷ The plots confirm that the higher the Average Compliance value, the stronger the negative effect of the Views per Country variable on the acceptance of the four ICPs. Put differently, the more likely a state is to implement the views it receives, the stronger the inhibiting effect of those views on the acceptance of most of the examined ICPs. Complying states also have a lower hazard of accepting the ICPs of the CAT, and the CRPD the more views have been rendered against countries within the same region or regional organization, as can be seen in Fig. 5.A and 6.A in the Appendix. This underscores the importance of compliance costs (fiscal and political) in triggering the learning mechanism, as expressed in H3b.

6 Discussion

I find evidence for the hypothesized learning effect for some of the examined ICPs but not all. In line with H1, states seem to be deterred from accepting all six examined ICPs by the information provided by the total number of views that are ‘out there’ independent of whether they are subject to any such decision themselves. Equally, states react to the number of views that are issued against like-minded and neighboring states, as conjectured in H1b. For three ICPs, the learning effect is amplified by geographic and political proximity. In sum, the findings provide encouraging support for my claim that states use the information on the prevalence of views against other states to recalibrate their cost assessments concerning additional ICP acceptances.

As to my second hypothesis, I find only partial evidence that states also learn from their own experiences. The results of the survival analysis suggest that the more views states had to respond to directly, the less likely they are to accept the ICPs of the CERD and the CAT. Crucially, I find that this association between views and ICP acceptance depends to a large part on the propensity of the respondent state to comply with the views it has received. Indeed, my findings support H3b, which states that being subject to views only affects states in their choice of ICPs when the views trigger concrete fiscal and political costs for the states through the implementation of remedies. By implication, this means that the indirect costs resulting from non-compliance, such as audience costs and economic sanctions, are relatively toothless in preventing states from ICP acceptance.

For some of the ICPs examined, however, I could find no evidence to support the claim that the learning effect is most effective when states are directly confronted with the views. The non-findings add some important qualifications to the results of this study. For one, they point to the fact that the ICPs in my sample vary in terms of the rights they cover. However, the nature of the rights that are monitored by each ICP are consequential for the learning mechanism. The implications of views on state behavior may vary from little or no impact to very strong reactions depending on which state is addressed and which rights are invoked. For example, the protection of physical integrity as well as civil and political rights is perceived by several states across the globe as sovereignty-threatening but is actively promoted by others – mostly Western, democratic, and liberal states (Meyerrose & Nooruddin, 2023). Additionally, in democracies, a stronger influence of civil society on decision-making processes could be counteractive to the suggested learning effect: domestic stakeholders could be encouraged by views to pressure their government and legislature into more ICP acceptance. By implication, one would expect that states that adhere to principles of liberal democracy are less susceptible to the acceptance-decreasing effects of rising numbers of views. Yet, the interactions depicted in Figs. 1 and 2.A (Appendix) insinuate that the more democratic a state the more likely it is to learn from views against itself and other states when it comes to the acceptance of the ICPs of the CERD, the CEDAW, and the CRPD. The fact that the proposed learning effect is particularly strong among democracies deciding on ICPs that protect vulnerable populations and weakest for ICPs that monitor ‘classic’ civil and political, as well as physical integrity rights may be explained by a higher ‘tolerance of pain’ among (liberal) democracies when it comes to the latter group of rights. Furthermore, relatively potent human rights groups in the area of physical integrity rights could be particularly successful in pressuring democratic decision-makers into the acceptance of CAT Article 22 if the state has already received views under the CCPR-OP.³⁸ This would suggest that democracies are better learners but only in issue areas that are sensitive to them.³⁹

Fig. 1

Interaction between views per regional organization and regime type. Note: Points show simulated relative hazards of views per regional organization on ICP acceptance for different regime types based on 1000 simulations

The results for the interaction terms depicted in Figs. 2 and 3 reveal another interesting pattern: the association between the number of views per (complying) country and ICP acceptance becomes weaker and loses in statistical significance the more recently the observed ICP became operative. This is another indication that states do make a difference between different ICPs. Since the majority of all views issued by the treaty bodies stem from the CCPR-OP, it is striking that the views correlate most strongly with the acceptance of ICPs that overlap with the CCPR-OP either temporally or substantively, or both. CAT Art. 22 is connected to the CCPR-OP in that it also covers physical integrity rights. CERD Art. 14, which became active in 1982, is temporally closest to the CCPR, which became active in 1976. The other three ICPs are younger and cover different sets of rights than the CCPR. Thus, past activities of the HRCttee are not necessarily interpreted by states as an indication of the future activities of the CRPD and the CRC, for example. This would make the acceptance of the latter ICPs less dependent on the views issued under the former ICP. In more general terms this would mean that the proposed learning mechanism depends on a high degree of comparability between the institution issuing the rebukes and the institution under consideration, both in terms of substance and temporal development.

It should be noted that alternative explanations for ICP commitment may have higher explanatory value than the feedback effect of views. The results from all regression models suggest that any effect of the number of views on ICP acceptance will be much smaller than other determining factors such as the degree of democracy or the nature of the law system in the responding state. Also, the number of views could possibly have a countervailing effect on ICP acceptance in that they encourage domestic and international stakeholders to pressure governments into accepting further ICPs, for example through recommendations in the Human Rights Council’s Universal Periodic Review. The fact that, despite these circumstances, the patterns outlined above are traceable across the majority of the six examined ICPs despite existing differences between the treaty bodies regarding their issue area, processes, and legitimacy is noteworthy and speaks in favor of the robustness of the proposed learning mechanism.

Fig. 2

Predicted ICP acceptance - views per country and compliance interaction. Note: Points represent the hazard ratios for the depicted variables from separate regressions. Grey bars show 95% confidence intervals. Results of the full models can be found in Table 4.1.A to 9.1.A in the Appendix

Fig. 3

Marginal effect of views per country on ICP acceptance - dependent on compliance

7 Conclusion

How do states react to interferences by international human rights bodies? Focusing on the individual complaint procedures of the UN human rights treaty bodies, this study has shown that states become less willing to bind their hands through such institutions the more information they have on mounting costs that ensue after commitment. Specifically, it has presented evidence that the outcomes of the ICPs can be understood as carriers of information about the ‘true’ costs involved in ICP acceptance which can lead states to update their cost-benefit assessments and become more reluctant to accept ICPs in the future. The results of the analysis further underline that this learning mechanism is driven by a logic of cost avoidance as the number of views only affects a state’s decision to accept more ICPs if that state also bears the costs of their implementation.

These findings have implications for several research agendas within International Relations. First, this study finds evidence that some costs triggered by treaty body views matter more to states than others. Reputation costs or economic sanctions that result from non-compliance with views are less effective in triggering state learning than the fiscal and political costs stemming from compliance. Thus, in the context of human rights ICPs, the former costs do not seem to have the same impact on states as in other areas (Chaudoin, 2014; Tomz, 2007). The direct costs stemming from the implementation of remedies, on the other hand, can have a substantial impact on states’ willingness to commit to ICPs. This finding is in line with research on bilateral investment treaties that shows that the learning effect of states is stronger the more immediately states experience the costs of their commitments, albeit only if they have to suffer the actual costs (Haftel & Thompson, 2018; Poulsen & Aisbett, 2013; Thompson et al., 2019). More generally, this implies that states are particularly careful about relinquishing sovereignty to IOs if they are under strong pressure to comply.

Second, this study speaks to scholarship on state engagement with international legal institutions. The proposed learning mechanism can be understood as an undertheorized reaction of states to the unexpected consequences of treaty commitment beyond known strategies such as the inclusion of flexibility mechanisms or reservations (Helfer, 2013; Hill Jr, 2016; Koremenos, 2005), and withdrawal (Pauwelyn & Hamilton, 2018; Soley & Steininger, 2018). If states learn about circumstances that alter their previous cost-benefit assessment so that the initial commitment is no longer in their interest but cannot—or do not want to—target the institution that is creating the costs directly, they may turn away from comparable institutions in the future. The result is a quiet disengagement from the monitoring regime overall. This may be read as a novel ‘shadow effect’ of international courts and tribunals that also stems from a politics of resistance to judicial authority but ends not in preemptive reform but in the non-acceptance of adjacent legal instruments (Pavone & Stiansen, 2022; Küçüksu, 2023). Additionally, the study adds to research on international courts’ strategies to react to backlash (Larsson & Naurin, 2016; Stiansen & Voeten, 2020) by suggesting that judicial restraint towards compliant states could be a strategy of international judicial bodies not only to mitigate backlash but to avoid disengagement from individual complaint mechanisms altogether.

Third, the study has implications for research on international human rights law. It sheds light on a potential conflict between the provision of individual justice through compliance with views (Fox Principi, 2017; Murray, 2020; Ullmann & von Staden, 2023) and the goal of a global net of international human rights institutions (Comstock, 2021; Goodman & Jinks, 2013; Simmons, 2009) because it shows that more compliance can lead to less commitment. The results of this study should not serve to play these two objectives off against each other but to recognize that the road to more effective and encompassing human rights protection is not always straightforward. Crucially, this study does not set out to provide a general account of ICP acceptance but rather aims for a better understanding of state learning from the outcome of international monitoring mechanisms. A more positive conclusion from the results may be that states take these mechanisms very seriously and adjust their future decisions accordingly.

Surely the fact that states can accept multiple ICPs under the UN human rights treaty body system is a critical precondition for the proposed form of state reaction to interventions by international monitoring bodies. The UN treaty bodies show a high degree of procedural and organizational similarity that may not be met by comparable monitoring institutions, for example, in other issue areas. Yet, it is conceivable that whenever the possibility to commit to a similar institution exists, states may resort to this form of quiet resistance long before they engage in more visible backlash strategies (Börzel & Zürn, 2021; Walter, 2021). In broader terms, this study suggests that over time states learn about the implications of their international commitments and adjust their decisions to join similar arrangements accordingly. Interference by one institution can decrease the chances of commitment to another institution that governs the same issue area and/or was formed at the same point in time. Should future research establish the proposed state reaction in other issue areas as well, this would have strong implications for an international system that is increasingly defined by overlap and redundancy between institutions (Alter et al., 2016; Haftel & Lenz, 2022; Panke & Stapel, 2023).