![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
ABSTRACT
The empirical assessment of how signatories of human rights conventions comply with their obligations has, so far, yielded conflicting results, especially regarding the compliance mechanisms that are the most promising to ensure improving human rights records. We argue that this is due to the fact that differences in compliance systems have been neglected and that different compliance mechanisms have been assessed in isolation, without considering possible interactions. To analyze this argument, we propose a novel way to assess the effect of these mechanisms by relying on a Markov-transition model. Our results show that human rights violations are time dependent and that the effect of independent variables is conditional on previous human rights violations as well as on the strength of human rights compliance systems.
KEYWORDS:
Acknowledgments
This article draws on Simone Wegmann’s (Citation2011) master’s thesis. Earlier versions were presented at the Annual Congress of the Swiss Political Science Association (Lucerne, February 2–3, 2012), the Annual Meeting of the Midwest Political Science Association (Chicago, April 12–15, 2012), the 2nd Annual General Conference of the European Political Science Association (Berlin, June 21–23, 2012), and the 6th Annual Conference on The Political Economy of International Organizations (Mannheim and Heidelberg, February 7–9, 2013). We wish to thank the participants at these events and especially Jeff Colgan, Marina Kolb, Eric Neumayer, Niklas Potrafke, and Joseph Wright for their helpful comments.
Supplemental Material
Supplemental data for this article can be accessed on the publisher’s website.
Notes
1 Information on treaty ratification is available online: http://treaties.un.org/Pages/Treaties.aspx?id=4&subid=A&lang=en.
2 Landman (Citation2005a) addresses the endogenous nature of treaty ratification by using instrumental variables but offers, however, no information on the quality of these instruments. Neumayer (Citation2005), on the other hand, approaches this problem by employing a Heckman selection model, criticized by Simmons (Citation2010:290): “Neumayer checks for the robustness of these results with a Heckman selection model, with a curious justification for instruments: He holds that ‘newly independent countries receive greater attention with respect to their human rights record as do former colonies’ (Neumayer Citation2005:949), but the likelihood of scrutiny seems to be precisely the mechanism that drives his results concerning the importance of INGOs and democracies.”
3 He considers, however, the CIRI physical integrity scale as continuous variable and assumes a constant effect for treaty ratification independent on the time since ratification.
4 Using a continuous measure for compliance, as the one proposed by Fariss (Citation2014) and Schnakenberg and Fariss (Citation2014), is not affected in the same way by this problem. For this reason, mainly, we refrain from proposing an analysis of how compliance systems affect human rights compliance based on the proposed continuous measure. Doing so would require an alternative empirical model, extending the item-response theory model used by Fariss (Citation2014) and Schnakenberg and Fariss (Citation2014) to not only cover the measurement issues but also the explanatory aspect. While Fariss (Citation2014) offers in his contribution a simple OLS estimation of the effect of treaty ratification, he acknowledges that the proposed models “… are not designed for causal inference …” (Fariss Citation2014:313).
5 While Richards, Webb, and Clay (Citation2015) correctly control in their replication of Poe and Tate’s (1994) study for the lagged ordinal dependent variable, their empirical specification does not consider these floor and ceiling effects.
6 Kim and Sikkink (Citation2010) refer to having explored such a model in their work on how cases brought before human rights courts affect the respect for human rights. They provide, however, no detailed information.
7 For simplicity’s sake we do not address the related issue of time dependence, namely whether having remained in the same category of our dependent variable affects the likelihood of a transition.
8 While we will consider as one of our independent variables the ratification of the Convention against Torture (CAT), we believe that this very visible convention should affect state behavior beyond the narrow area of torture. For this reason we use the full Physical Integrity Rights Index, which comprises a subcomponent dealing with torture (for a similar approach, see Fariss [Forthcoming]). In the online appendix we provide, however, an analysis focusing only on the torture subcomponent and find substantively identical results (see Table A4 in the online appendix).
9 In 2008, the last year covered in our analysis, roughly a fourth of all countries had not ratified the CAT, 42% belonged to the UN compliance system, 10% to the American, respectively the European one, while 14% belonged to the European Union.
13 We also assessed whether an interaction between the political regime and the ratification decision affect the results reported. This is not the case, which is why we omit this additional complication from what follows.
14 Based on logic, we also exclude from the selection/ratification equation the variable on the compliance regimes, as these are linked to having signed a treaty.
15 For countries with border changes in this period, namely Czechoslovakia, Germany, and Yemen, we used as units the currently existing countries. For the beginning of the period, we used the data from predecessor states, namely Czechoslovakia, West Germany, and North Yemen. In the online appendix we report a robustness check based on omitting these observations. We also report in Table A2 of the online appendix descriptive statistics of all the variables employed in our analysis.
16 Table A2 of the online appendix also shows all mean values of the CIRI integrity rights index and its torture component for different compliance systems.
17 As conflicts are extremely rare in countries with good human rights records, we had to regroup the interactions between the previous state of human rights records and conflict for the three highest categories (i.e., the values 6, 7, and 8.). Similarly, among the European Union member states there are no cases with extremely poor human rights records, which is why the interactions of the EU compliance system and the previous state of human rights records is only reported for the four highest categories (i.e., the values 5, 6, 7, and 8).
18 This is also indicated by the insignificant correlation between the error terms of the two equations (i.e., atan(ϱ)).
19 In the online appendix we also report on a very conservative robustness check, namely by providing the estimates of a fixed effects ordinal logit model. As we will discuss, the effects of some of the compliance regimes are considerably reduced due to this conservative estimation.
20 As there are only minor substantive differences, we will focus our interpretation on the simpler ordered probit model.
21 In terms of the Akaike information criterion (AIC), model 2 (either estimated as ordered probit model or as a CMP) is slightly better than model 3. As the latter is substantively more interesting, we focus in what follows on this model.
22 To do so we relied on clarify (Tomz, Wittenberg, and King Citation2003).
23 We keep all remaining variables at their lowest level except GDP per capita, which we kept fixed at US$4,730 (median value), and the political regime, which was fixed at the value of democracies. We chose this latter option, as a nondemocracy ratifying the CAT and joining the EU compliance regime is an impossibility.
24 This is obviously only an imperfect assessment of the long-term effects of a treaty ratification, as the Markov transition model setup implies that the transition probabilities will affect the likelihood of belonging to any response category in a multiplicative fashion.
25 We chose a time span of 10 years, as the maximum number of years since ratification in our sample is 23, but for the effect of higher numbers of years since ratification our estimates are much less precise.
26 These changes in probabilities were calculated based on 1,000 draws from the estimated coefficients and by generating each time for each year 1,000 matrices of transition probabilities, once under the assumption of no ratification and once under the assumption of a ratification. For each scenario and each year, the respective transition probabilities were multiplied to determine the probabilities of attaining a particular category after a certain number of years. Subtracting the two set of matrices from each other results in the estimated changes, which were averaged over the 1,000 draws. Confidence intervals were also determined but are not depicted here because of the information overload they would create in the graphics.