Do human rights treaty obligations matter for ratification?

Abstract

International relations scholarship assumes that states weigh the costs and benefits of treaty ratification. In human rights, the worse a particular state’s record, the higher the presumptive costs of ratification and the lower the likelihood of ratification. But prior work neglects variation in the extent of obligation that different treaties create. In this article, we argue and demonstrate that (1) human rights treaties differ substantially in the scope and scale of the obligations they contain, (2) this variation can be measured, and (3) it matters for ratification. Treaties that create a larger number of demanding obligations imply greater potential costs of compliance for states. The larger the number of demanding obligations, the more grounds various actors will have to challenge a state’s practices. We analyze innovative data on treaty obligations and commitments for the 10 core global human rights treaties to test our propositions, and we find strong support.

Acknowledgments

For helpful comments on previous drafts, we thank the editors and reviewers at The Journal of Human Rights. Previous versions of this article were presented at the University of Southern California, the 2017 Southern California International Law Scholars Workshop, and the 2018 International Studies Association Convention. The data underlying this article are available at the JHR Harvard Dataverse site.

Disclosure statement

The authors report that there are no competing interests to declare.

Notes

1 Comstock (2021) complicates this picture, arguing that different commitment paths (e.g., ratification with or without prior signature, accession, and succession) also influence the expectation and likelihood of state compliance.

2 One partial exception is Dancy and Sikkink (2012), who classify human rights treaties into three broad categories (physical integrity rights treaties with individual criminal accountability, physical integrity rights treaties without individual criminal accountability, and all others). Our approach differs in that it recognizes that human rights treaties vary in terms of how demanding the obligations they contain are and takes into account the level of “demandingness” of each treaty. Even among Dancy & Sikkink’s three categories of treaties, there is variation in the degree of obligation.

3 It is worth noting that all of the examples Gilligan (2004) uses to illustrate his argument are in the economic or environmental realms—not human rights.

4 We acknowledge that unlike the other treaties in our sample, the Genocide Convention is both a human rights and criminal law treaty and is more limited in scope. In addition, while the other treaties have committees that monitor state compliance, the main compliance mechanism for the GENO is the International Court of Justice (ICJ).

5 Note that our analysis excludes the European Union (recorded by the United Nations [UN] as a party to the CRPD), Palestine (recorded as a party to all 10 human rights treaties, save for the CED and CRMW), Hong Kong and Macau (at one time recorded as parties to the CAT, CERD, CRC, and CRPD), and the Holy See (recorded as a party to the CAT, CERD, and CRC). These actors’ ability to enter treaties is contested and variable across our period of analysis, making them too different to compare to the broader population. In any case, listwise deletions due to missing data for state-level covariates means their exclusion does not affect the overall results.

6 If it is possible to measure the level of demandingness and demonstrate its effect on commitment to human rights treaties, it should be possible to do so in other domains (e.g., security, economics, environment) where the costliness of treaty obligations should be easier to observe and quantify. For security, economic, and environmental treaties, the costs of specific obligations should be more readily measurable, for example, in terms of particular weapons systems, military bases, trade gains and losses in specific industries or even products, and reductions in particular pollutants.

7 Convention on the Rights of the Child (1989), Art. 6(2).

8 This figure replicates Figure 2 in Zvobgo et al. (2020, p. 790).

9 Our definition of precision aligns with Koremenos’s: “an agreement’s degree of precision or ambiguity refers to the exactness or vagueness of its prescribed, proscribed, and authorized behaviors” (2016, p. 160).

10 Genocide Convention (1948), Art. 4.

11 Convention Against Torture (1984), Art. 2(1).

12 Convention on the Elimination of All Forms of Discrimination against Women (1979), Art. 9(1), emphasis added.

13 Convention on the Rights of the Child (1989), Art. 8(1), emphasis added.

14 Convention on the Rights of Migrant Workers (1990), Art. 18(1), emphasis added.

15 Convention on the Elimination of All Forms of Discrimination against Women, Art. 18(1), emphasis added.

16 A partial exception is Boyes et al. (2023) who investigate states’ withdrawal of human rights treaty reservations in response to international social pressure, operationalized as peer state objections and treaty body periodic reviews calling for reservation withdrawals and affirming that a given state is legally bound by the provision to which it had attached a reservation.

17 Of course, states can prosecute specific human rights violations under domestic law in domestic courts. The CAT is unique in requiring states to criminalize torture and prosecute or extradite persons who commit acts of torture.

18 Given that obligations requiring domestic action account for 91% of obligations—and the exclusion of such a requirement would not alter the measure—the question about combining dimensions into one measure centers on whether precision and strength should be jointly included. See the Supplementary Appendix for further discussion.

19 Hill and Watson’s (2019) work challenges this idea. The authors find that regime type does not always condition treaties’ effect on rights respect, at least in the case of CEDAW. To be sure, CEDAW is in many ways unique among human rights treaties and the lack of a conditional effect of regime type on compliance may be due to autocracies’ substantial participation in the treaty’s negotiations (Comstock, 2022) and autocracies’ engagement with and socialization in women’s rights, both during and after CEDAW’s adoption (Comstock & Vilán, 2023).

20 Vienna Convention on the Law of Treaties (1969), Art. 19.

21 Hill (2016) argues that states are more likely to enter reservations when human rights treaties contain standards that are more rigorous than those in their domestic laws, and he tests the argument with respect to the ICCPR. Comparable data on domestic laws relevant to all 10 of our human rights treaties are unfortunately not available. Still, our argument is slightly different from Hill’s. We test the relationship between how demanding a treaty is and the likelihood that states ratify it with reservations as compared to without reservations. We find that states are less likely to ratify more demanding treaties, even when taking reservations into account. The higher the number of demanding obligations, the less likely states are to commit and the less likely they are to commit fully, i.e., without reservations.

22 Version 2 of the IHROC—Treaty Reservations dataset is available via the JHR Harvard Dataverse site. This version increases the sample size by roughly 6%, to 77,821 observations, and the number of reservations to 1,013. We replicate Zvobgo et al.’s (2020) main findings in the Supplementary Appendix.

23 Note, the reservations dataset captures additional observations for amendments to prior reservations and declarations. The dataset also includes observations for “edge cases,” for example Hong Kong and Macau, for which there are recorded reservations and declarations. Because both reservations and declarations are rare, we erred on the side of inclusion for the reservations data and analysis, while we err on the side of exclusion for the ratification data and analysis. Due to missing values on a number of the covariates, observations for Hong Kong and Macau are ultimately dropped in statistical analyses of reservations accounting for state-level factors. The reservations dataset does not include observations for Palestine or the Holy See, which had no registered reservations or declarations through the end of 2014. Due to missing values on a number of the covariates, they too would be subject to listwise deletions. For its part, the European Union, as an international organization, is not germane to an analysis of state ratification or reservation behavior.

24 We are interested in reservations that apply to specific treaty provisions. States sometimes enter reservations regarding a treaty as a whole. Such reservations are not directly relevant to our analysis because they concern a state’s broader political values or goals, for example, its constitutional or religious law, or its relationship with the state of Israel. Whole-treaty reservations do not express a state’s position regarding specific obligations.

25 The average reservation rate across treaties is 1.3%.

26 We note that the GENO first raised the question of the possibility and permissibility of treaty reservations, a question that was answered in a 1951 ICJ advisory opinion. The ICJ’s “object and purpose” criteria laid the ground for Article 19 of the Vienna Convention on the Law of Treaties (VCLT), which further defines the possibility and permissibility of reservations. As a technical point, the VCLT covers the human rights treaties in our sample from CEDAW onward.

27 Version 2 of the IHROC—Treaty Obligations dataset is available at the JHR Harvard Dataverse site. Details on the data are available in the Supplementary Appendix.

28 Each treaty was coded independently by two coders following detailed instructions from the principal investigator. Any differences in coding were resolved by the principal investigator. See the Supplementary Appendix for additional discussion.

29 To identify treaty obligations, the relevant unit of the treaty text is sometimes the article (for example, CAT Art. 11 is a single-paragraph article). More often, an article contains two or more paragraphs, some of which also contain sub-paragraphs. We coded the lowest-level unit available in each instance.

30 We measure the skewness of Demanding obligations at 0.85, which means that the distribution leans moderately to the right.

31 In these models, we exclude observations in which ratification of the ICESCR or the ICCPR is the outcome variable.

Additional information

Funding

This material is based upon work supported by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-1418060. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.

Notes on contributors

Suzie Mulesky

Suzie Mulesky is a senior behavioral scientist at Vanguard.

Wayne Sandholtz

Wayne Sandholtz is John A. McCone Chair in International Relations and Professor of International Relations and Law at the University of Southern California.

Kelebogile Zvobgo

Kelebogile Zvobgo is an Assistant Professor of Government at William & Mary, a faculty affiliate at the Global Research Institute, and the founder and director of the International Justice Lab. Previously, she was Provost’s Fellow in the Social Sciences at the University of Southern California, where she earned her PhD Her work engages questions in human rights, transitional justice, and international law and courts.