1 Introduction

NGOs’ human rights shaming strategies are linked with their donors’ interests (Cooley and Ron 2002; Oestreich 2007; Tallberg et al. 2015), while the United Nations’ Commission for human rights’ (UNCHR) shaming is primarily norm-driven (Lebovic and Voeten 2006). Which of these applies to the UN’s treaty bodies’ shaming? The UN’s treaty bodies are fundamentally different from both the UNCHR and NGOs. Unlike the UNCHR they are composed of human rights legal experts, not member states’ representatives. Unlike NGOs they depend on the member states, those which they shame, for funding. Furthermore, the treaty bodies depend on the reviewed countries’ cooperation much more heavily than the UNCHR and NGOs. Starting from these insights, this article examines whether linkages with powerful states help explain shaming by the UN’s human rights treaty-bodies. More specifically it aims to determine whether such shaming is associated with voting with the UN’s main donor — the U.S. — at the UNGA and with having a military alliance with the US. To substantiate these questions let us consider a few examples: Despite the high level of state oppression in Ukraine in 2014 the country received little criticism from the CAT committee. By contrast, that same year, Japan was criticized extensively despite the very low level of oppression exercised by the Japanese government. What accounts for this difference? At the time of the review, Japan was twice as likely as Ukraine to vote against the U.S. at the UNGA. Furthermore, Yemen, Syria and Russia were all criticized by the CAT at much higher rates than Ukraine despite being similarly oppressive. These countries voted against the U.S. at the UNGA much more often than Ukraine. This article shows that a significant proportion of the variance in UN CAT shaming is associated with countries’ diplomatic affinity with the United States. The article argues that this finding is linked to the fact that the UN is an IGO and identifies fundamental differences between NGOs’ and IGOs’ human rights shaming which help to make sense of the association between amity with the U.S. and variations in shaming levels.

Research on shaming and monitoring in policy areas as diverse as environmental protection (e.g. Gray and Shimshack 2011), counter-terrorism (e.g. O’Donnell 2006), macroeconomic policy (e.g. Lombardi and Woods 2008) and human rights (e.g. Murdie and Davis, 2012) has been blooming. Much of the research has focused on identifying the conditions under which shaming is effective and induces policy change. At the same time, shaming has itself become a subject of inquiry (Hendrix and Wong 2014; Stroup and Murdie 2012). Shaming is carried out by human rights advocates who are principled strategic actors (Hendrix and Wong 2014; Keck and Sikkink 2014; Murdie and Urpelainen, 2015). Their advocacy is based on a combination of normative and organizational considerations (Carpenter 2007; Cooley and Ron 2002; Hill et al. 2013; Meernik et al. 2012; Murdie and Urpelainen, 2015; Ron et al. 2005; Sell and Prakash 2004). This combination of principled — ethical and pragmatic — utilitarian shaming rationales has been observed both in NGOs and in IGOs including in the UN and its agencies (Oestreich 2007). However, the specific factors which explain IGO shaming are different from those which explain NGO shaming. Like NGOs, IGOs have their organizational interests, mobilization considerations and funding considerations. However, in addition to these considerations, and since the member states are the donors, IGOs must consider states’ interests. In any given shaming context, the IGO must prioritize between those sets of interests and considerations. Understanding how and why IGOs do this, advances research on shaming, on international organizations, and on balancing norms and interests in international politics more generally. This article takes up these issues by looking at the factors which are associated with the UN’s CAT committee’s shaming. The UN CAT treaty body’s concluding observations country reports are analyzed to measure positive and negative points in the CAT committee’s assessment of the human rights situation in a given country (United Nations 2018). The committee’s working methods cannot be described in detail here. Fortunately, O'Flaherty (2006) has written an excellent piece on the subject. In the current context, three technical aspects regarding the committee’s work need to be explained: first, the practice of issuing and publicizing concluding observations reports was initiated in 2001 and the first report was publicized in 2002. Since then the CAT committee issues reports on all the ratifying countries discussed in the committee. The second aspect is the issue of reporting cycles. Countries are obliged to submit a report to the committee within one year of ratification and every four years after that. The state report is the basis for the discussion which ultimately results in the concluding observations report. However, more often than not reports are delayed, mostly as a result of state parties not fulfilling their reporting duties on time. Furthermore, reports are published online as soon as they are written and approved. It is therefore quite difficult to identify structured reporting cycles or waves among the published reports. The third aspect to remember is the relatively fixed structure of the report. The UN’s CAT committee divides its concluding observations reports into three sections: positive points, points of concern, and recommendations. Each section has a heading, and heading wordings are virtually identical across documents. Each issue in each section is numbered or bulleted, making the identification and quantification of positive and negative points quite straightforward.

The next section discusses previous research on donors’ interests in international organizations and explains how donors’ interests are linked with human rights shaming. The third section describes the newly compiled dataset on the UN’s CAT committee’s shaming, and the data used to analyze shaming. The fourth section describes and discusses the results of several analyses, applying various regression and specification approaches. The analyses show that the more a country votes with the U.S. in the UNGA, the less that country is shamed. Therefore, the section argues that in carrying out its human rights reporting duties, the CAT committee considers the UN’s main donor’s — i.e., the United States’ — interests and concerns. The fifth section draws theoretical conclusions and considers some policy implications.

2 Donors’ interests and other factors associated with human rights shaming

Scholars of international financial institutions have long noticed the impact of donors’ concerns on many aspects of the policies and practices of these organizations: The probability of receiving an IMF loan increases if a country votes against its policy preference and along with the U.S. in the UNGA (Andersen et al. 2006b). The sums of the World Bank’s IDA loans are positively associated with voting with the U.S. at the UNGA (Andersen et al. 2006a). Countries which import more from the U.S. receive larger shares of the World Bank’s loans (Fleck and Kilby 2006). The more a country votes with the U.S. in the UNGA the fewer the conditions to the loans it receives from the IMF, and, conditional upon voting with the U.S., countries facing elections receive loans with even fewer conditions (Dreher and Jensen 2007). The conditions attached to World Bank structural adjustment loans are not enforced on countries which align their votes with the United States’ vote in votes that are important to the U.S. (Kilby 2009). The duration of program suspension by the International Monetary Fund as a means of enforcement is significantly shorter for countries that are important to the U.S. (Stone 2004). Among Middle-Eastern countries signing a peace treaty with the United States’ closest ally in the region (Israel) improves a country’s chance of receiving a loan from the World Bank and from the IMF (Harrigan et al. 2006). Similar trends have been identified in regional organizations (Kilby 2011). In sum, donors’ interests are correlated with many aspects of the policy cycle of international financial institutions’ lending. This is indicative of a deep process of politicization of the work of these organizations, wherein finance is used to promote powerful countries’ agendas and to increase their international influence. Since shaming influences aid, investment and trade (Barry et al. 2013; Dietrich and Murdie 2015; Lebovic and Voeten 2009; Murdie and Peksen, 2013; Peterson et al. 2016), the UN’s shaming can be seen as yet another way of inflicting financial punishments upon countries that are not supportive of powerful countries’ international policy agenda. This piece of research begins to examine this question by looking at the association between countries’ diplomatic and military ties with the U.S. and their shaming by the UN treaty body. The next few paragraphs discuss the links between donors’ interests and human rights shaming.

In order to use publicized information to promote human rights, actors must possess moral authority (moral leverage) or be able to harness material authority (material leverage) by mobilizing powerful actors. The objectives of establishing authority and mobilizing powers shape NGOs shaming strategies alongside the general ethical objective of promoting human rights (Keck and Sikkink 2014). However, human rights actors also consider donors’ and members’ interests. In this respect, the exponential increase in the amounts of human rights donations made both by governments and by private actors in recent decades is a double-edged sword. While it enables more elaborate and comprehensive human rights work, it is also a constraint. Empirical analysis has shown that an NGO’s funding structure and considerations of donors’ interests are at least as important as its core mission and considerations of public mobilization in choosing a shaming strategy (Powers 2014). Advocates consider the needs and interests of financiers in carrying out their work. They do this regardless of whether donors require them to do so (Oestreich 2007), with two objectives in mind: to secure future donations and to ensure continued interest in the organization’s activity (Cooley and Ron 2002; Tallberg et al. 2015). Empirical evidences indicate that NGOs take these considerations into account when choosing a shaming approach and that they prioritize human rights issues, targets, or causes which are important to donors and to paying members (Carpenter 2007; Hendrix and Wong 2014; Hill et al. 2013; Murdie and Urpelainen, 2015). Such considerations arguably also explain UN treaty committees’ reporting. While it may seem as though IGOs have policymakers’ attention by definition, policymakers’ interest in the work of the UN CAT committee, their appreciation of its work as significant, and their continued contribution into its budget are by no means guaranteed.

The U.S. provides between 25 and 30% of the UN’s annual budget (known as ‘assessments’). The remaining 70–75% is divided among the other 192 members. The next largest contributor after the U.S. is Japan which contributes around 12% of the UN budget (UN Committee on Contributions 2017). The dominance of the U.S. in UN funding is obvious. As noted above human rights advocates factor donor-interest into their shaming decisions regardless of whether or not donors require them to do so (Oestreich 2007). Moreover, there are several recent manifestations of the United States’ preparedness to withdraw from UN bodies citing unfair treatment of its allies as the reason for the withdrawal. In October 2017 the U.S. withdrew from UNESCO for this reason precisely (Nauert 2017). In 2008 the Bush administration announced its intention to deduce its share of the human rights Council budget from its 2008 contribution to the UN due to what it viewed as an unjust focus on one of its allies. Previous and successive U.S. administrations expressed similar concerns and warnings (Blanchfield 2013) indicating the United States’ willingness to withhold funding due to criticism of its allies’ human rights practices. To get a sense of whether UN human rights officials’ are likely to be concerned about such funding cuts one need not look further than the official website of the office of the high commissioner for human rights where it is stated that the office is underfunded and relies on voluntary governments’ donations for 60% of its mandatory activity budget (OHCHR 2017). In other words, the office is obliged to carry out assignments for which it does not have sufficient funds. Any additional funding is contingent upon the benevolence of the member states. This makes the office an easy and tempting target for member states’ budget cuts and structures the office in a beggar’s position as manifest in its official annual appeal for funds. The 2013 appeal, for example, mentioned the monitoring work of the treaty bodies as an activity which will be hurt by proposed budget cuts (OHCHR 2013). There is concrete evidence that the UN’s human rights bodies are concerned about member states cutting funding and the U.S. has made it quite clear that it associates the funding it provides with criticism of its allies’ human rights practices. It is therefore worth examining whether the main donor’s concerns and the need to keep the main donor engaged and financially committed are helpful in making sense of UN treaty committees’ reports.

The donor-interest hypothesis stipulates that UN treaty committees’ shaming is negatively and significantly associated with the shamed country’s strategic affinity with the U.S.

3 Data

3.1 The dependent variable

The analysis of the UN treaty bodies’ concluding observations reports and the counting of positive and negative points are conducted using the Scrapy web crawling software and a Python script. The web-crawler takes the results of a document type search on the office of the high commissioner for human rights’s online search engine as its input. It parses the HTML code and identifies the country reviewed, the year the report was issued and the link to the full text of the CAT committee’s concluding observations report. The code then follows the link and opens the report in HTML format. The script runs through the HTML code of each report and identifies the positive points section and the points of concern section using the regular expression search function. It then counts the number of points in each section by paragraph tags and numbers. Finally, the script produces a dataset containing the number of positive and negative points on a country-year basis. Where no report is published for a country-year unit, a zero value is given for both positive points and negative points.

The extent to which and the ways in which such automated methods can reliably replace human coding depends on the research question at hand (Fariss et al. 2015; Grimmer and Stewart 2013; Lucas et al. 2015). Automated analysis need not (and cannot) extract all the information contained in the text as humans can. But it can extract enough reliable information to provide an answer to the research question in the current application. Four features of the concluding observations reports make the automation of their analysis relatively straightforward: the first is that the documents are highly structured into separate sections, each section has a heading and heading wordings are virtually identical across documents. The second is that the documents are structured by sentiment, differentiating between ‘positive’ points and points of ‘concern’. Taken together features one and two solve the major problem of sentiment identification and analysis. The sentiment is embedded in the document’s structure. The third feature of the reports which further simplifies their analysis is that the documents are online enabling access to the HTML code which contains the structure and sentiment in machine language. Fourth, the documents are retrievable through a search engine based on the documents’ types (in the current application concluding observations). This solves the major problem of ‘separating the wheat from the chaff’ (D’Orazio et al. 2014). The machine needs to perform three fairly simple tasks: the first is to retrieve metadata from specifically designated fields in the HTML code including the name of the country and the year of the report. The second is to identify the structural elements relying on the titles of chapters and using the regular expression search function. The third is to count the elements in the identified section using format characters (bullets and numbering) which like all of the above are contained in the HTML code. The Online Appendix available at the Review of International Organizations’ webpage gives an example report showing the various sections in the reports and illustrating the way they are counted.

These tasks do not require turning the document into a bag of words, nor formulating categories, nor building dictionaries (Fariss et al. 2015; Grimmer and Stewart 2013; Lucas et al. 2015). The resulting parsimony of the code is perhaps one of the project’s greatest virtues (Quinn et al. 2010). What is more, running bag-of-words sentiment analysis, in this case, would discard crucial information. Namely, the CAT committee’s definition of positive and negative issues and the message the committee sends to the reviewed country through this categorization. Ignoring the committee’s categorization would be counterproductive and would undermine the purpose of the project. Furthermore, a qualitative analysis of the language and terminology used in the reports found incoherence and ambiguity in the strength of words used (O'Flaherty 2006). This ambiguity suggests that the structure of the report is probably a better indicator of the experts’ view of the situation in the reviewed country than the terminology used.

Despite the relative simplicity of the automated analysis used in this project, readers can be forgiven for expecting an indication of the code’s reliability and performance. Considering the simplicity of the functions as described above, using 100 manually coded documents should be enough to validate the results (Hopkins and King 2010). To test the code’s reliability, 114 documents are coded manually according to the method described above. The automatically generated variables are measured against the manually generated variables, using a paired t-test to test the hypothesis that the manual and the automated data are different, to validate the automatically generated data. The results of both the positive and the negative variables do not reject the null hypothesis that the manual and automated data are statistically identical (p = .676; p = .439). Furthermore, as a robustness check, the results are replicated on the manually generated sample.

The dependent variable is the ratio between the number of positive points and the number of negative points in each report. The ratio is preferable to simple counts because it captures the overall evaluation of a country considering both improvements and problematic issues, and because it gives a continuous dependent variable. As a robustness check, model four takes only the count of negative points as the dependent variable (more on this below). Table 1 gives the summary statistics of the dependent variables in both specifications. Figure 1 plots the distribution of the negative points and Fig. 2 plots the positive points’ distribution.Footnote 1

Table 1 Summary statistics of the dependent variables 2002–2017
Full size table
Fig. 1
figure 1

Histogram of negative points

Full size image
Fig. 2
figure 2

Histogram of positive points

Full size image

The main independent variables which test the donor-interest hypothesis measure strategic affinity with the U.S. (Hendrix and Wong 2014). Strategic affinity is operationalized using two separate variables roughly corresponding to diplomatic affinity and military affinity: voting with the U.S. at the UNGA (Voeten et al. 2009) and having a nonaggression pact with the U.S. (Gibler 2009). The latter is lagged by one year to test a delayed association.Footnote 2 Together these variables measure the degree to which a country is likely to be viewed as a friendly country by the UN’s largest donor-the US.

3.2 Control variables

Research shows that shaming increases as media coverage of human rights issues increases (Hill et al. 2013) and that human rights organizations’ (HRO) presence is associated with shaming, although there is debate whether this association is positive (Meernik et al. 2010) or negative (Hill et al. 2013). There is also empirical evidence indicating that the UN treaty bodies rely on information from domestic and transnational NGOs (Egan 2013; Krommendijk 2015). Accordingly, data on HRO shaming and media coverage of human rights abuse is included in the models (Meernik et al. 2012). Both variables are lagged one year allowing for a delayed link between their activity and UN shaming.

Empirical evidences indicate that developing a reputation as sources of reliable information on human rights abuse is a high priority for human rights NGOs (Hendrix and Wong 2014; Hill et al. 2013). Establishing such an authoritative status is also important to the UN treaty committees (Egan 2013; Gaer 2007; O'Flaherty and O'Brien 2007). To hold this effect constant, the model needs to include a measurement of oppression. There are several such measurements available. This article uses the Amnesty International reports based political terror scale (Wood and Gibney 2010) because its coverage runs later than other measurements’ coverages and this is an important advantage due to the relative scarcity of non-zero observations of the dependent variable. The political terror scale measures state oppression on a scale of one to five. Lower scores indicate lower levels of oppression. A score of one is given to countries in which political imprisonment and torture are extremely rare. A score of five is given to countries whose leaders pursue personal or ideological goals through every means including extensive imprisonment and torture. This variable is lagged one year because the treaty body’s reports review past developments.

Research has found that shaming which targets countries with links to powerful states is more likely to mobilize powerful countries to exert pressure on the shamed country than shaming which targets countries without such links (Hendrix and Wong 2014; Hyde 2011; Levitsky and Way 2006). If inflicting reputational costs is one of the committee’s considerations, then the reviewed country’s sensitivity to external pressure should be significant to shaming. To test whether the target countries’ potential sensitivity to pressure from outside is associated with shaming the analysis includes logged aid divided by the logged population size, logged FDI and trade to GDP as right-hand variables. Data is taken from the World Bank databank (World Bank 2018).

UNGA voting patterns should be taken in the context of the general power configuration within the UN. It is possible that UN human rights monitoring is influenced not by one power’s interest but by geopolitical considerations more generally (Cowan and Billaud 2015; Cowell and Milon 2012; McMahon and Ascherio 2012; O’Donnell 2006; Smith 2013). To control for the influence of the global power configuration the estimated models also include voting with China and with Russia (Voeten et al. 2009). To hold constant countries’ substantive positions the analyses also include the ideal point variable which calculates a country’s position towards the global liberal order based on it votes in the UNGA in the observed year (Bailey et al. 2017). China, Russia and the U.S. are all excluded from the estimated models to avoid spurious results (for example Russia has the highest possible score of one in the voting with Russia variable, but this obviously does not reflect actual diplomatic alliance because a country cannot ally with itself).

UN reports are influenced by the availability of information. Conflict reduces the quality of available information but increases the amount of international attention particularly in the media (Hill et al. 2013). Therefore the analysis includes a conflict measurement which equals zero if a country experienced no conflict in the observed year, it equals one if a country experienced either an external conflict or an internal conflict and equals two if a country experienced both. The data for this variable is taken from the correlates of war data project (Palmer et al. 2015). GDP per capita influences shaming because rich countries are shamed more often than poor countries (Ramos et al. 2007) thus the analysis includes GDP per capita data from the world bank databank as well as logged population size (World Bank 2018). The analysis also controls for regime type as measured in the Polity dataset (Marshall and Jaggers 2016). Only the variables that are expected to be linked with shaming with a delay are lagged. Those variables include: the political terror scale, HRO presence and media coverage.

4 Estimation and results

Model one analyzes the manually generated data. This model only includes the main independent variables of interest without the control variables because there are insufficient observations to estimate the full model. It is an OLS model with period dummies to control for all changes that are constant across countries at a given year. The model does not include unit dummies because there is only one observation per country. The dependent variable is the ratio between positive and negative points in the reports which captures the overall judgment of the observed country by the CAT committee as explained above. The results suggest that voting with the U.S. in the UNGA and having a non-aggression treaty with the U.S. are both positively associated with the ratio between positive points and negative points and these results are statistically significant. These findings are indicative, but they require further exploration using the automatically generated data.

Model two aims to address the reciprocal causality between treaty ratification and the evaluation of the observed country’s human rights policy (Hathaway 2007; Von-Stein 2005). A number of methods have been used to mitigate the problem of endogeneity between policy and international agreements. One approach is to exclude groups of countries based on their values on the endogenous variable as a robustness check (Haftel 2010). This approach is inappropriate for the current analysis because the suspected endogenous variable is treaty ratification and only ratifying countries get reports so they cannot be excluded from the analysis. Another approach is to run bivariate reverse regression to test the reverse causality with one to five-years lags (Peksen and Blanton 2017). Running this test with 1,2,3,4 and 5 years-lags between treaty ratification as the dependent variable and the ratio between positive and negative points as the independent variable with period dummies and fixed effects, and also using random effects regression, there is never any statistically significant reverse correlation. But the suspicion that reverse causality is concealed by the failure to include control variables lingers. Differencing can be used to mitigate endogeneity because while the levelled variable may be endogenous changes in it may not be (Dreher et al. 2015). But this approach cannot be used in the current application because changes in ratification of the CAT are few and far between.

Many researchers use instrumental variable regression. This is a promising approach provided that one can find a good instrument which is associated with the endogenous variable but not with the outcome variable. In studies which are not concerned with human rights policies (for example studies of the impact of environmental treaties on environmental policy), ratification of other international agreements and other indicators of progressive universalism are used as instruments for treaty ratification (Aichele and Felbermayr 2012; Slechten and Verardi 2016). But measurements of progressive universalism are inappropriate for the current context because they are likely to be strongly correlated with human rights practices.

Recent research on human rights policies used regional or global treaty ratification rates as instruments for human rights treaty ratification (Cho and Vadlamannati 2012; Cole 2012; Cole and Ramirez 2013; Peksen and Blanton 2017; Simmons 2009). Model two uses the second difference of regional rates of ratification of the CAT as an instrument for the observed country’s ratification of the CAT. The regional categorization is taken from Norris’ (2009) dataset and includes eight different regions.Footnote 3 While changes in regional rates of ratification are likely to influence countries’ ratification of human rights treaties through contagion, their link with other country-level variable should be much looser and they are unlikely to influence the UN treaty body’s review of that country, particularly since the regional categorization is not the UN’s categorization but a more refined geo-political-cultural categorization. The first stage cluster-robust F value for model two is 12.06, well above the threshold of ten enabling rejection of the hypothesis that the instrument is weak. This model assumes that treaty ratification is associated with subsequent human rights policies and therefore the ratification dummy is lagged one year. All of the second stage controls as specified in the previous section, including the period and unit effects, are also included in the first stage as control variables. The second stage includes the controls and independent variables specified in the previous section as well as period and unit dummies. The estimation uses the limited information maximum likelihood estimator with cluster robust standard errors. The dependent variable of the outcome equation is the ratio between positive and negative points in the reports which captures the overall judgment of the observed country by the CAT committee as explained above.

As seen in Table 2 the results indicate that voting with the U.S. is positively associated with the ratio between positive and negative points: the more a country votes with the U.S. at the UNGA, the more positive points it will get in its report in relation to each negative point. The ratio between positive and negative points is not associated with voting with China nor is it associated with voting with Russia thus the possibility that the positivity of reviews is explained by countries’ tendencies to vote with powerful countries is not supported. Model two corroborates the donor-interest hypothesis which stipulates that strategic affinity with the UN’s main donor is associated with shaming.

Table 2 Analyzing the UN’s CAT committee shaming
Full size table

In substantive terms the results of the model mean that all other things being equal, a country voting with the U.S. at the rate of the highest 99th percentile (81.5% of annual UNGA votes) is associated with it getting about 35 fewer negative points per every positive point, than a country which votes with the U.S. at the rate of the lowest 1st percentile (4% of annual UNGA votes). In stark contrast to expectations, the PTS score is positively associated with the positivity of the reports, indicating that the higher the level of oppression the more positive the report. In line with the theoretical expectations the model indicates that there is a negative association between coverage of human rights in the reviewed country by Western media and the positivity of the reports. The more coverage, the more critical the report.

Model three is a Heckman selection model. It addresses the fact that the committee only reviews countries which have ratified the CAT treaty, by first calculating the likelihood of ratifying the CAT treaty and then estimating the model. The selection equation is identified by the inclusion of two variables which should be correlated with the likelihood that a country would ratify the CAT. The first is the regional rates of ratification of the CAT (Cho and Vadlamannati 2012; Cole and Ramirez 2013). These regional ratification rates are unlikely to influence the UN treaty body’s review of that country, particularly since, as explained in the data section, the regional categorization used to calculate rates of ratification is not the UN’s geographical categorization but a more refined geo-political-cultural categorization. The second variable is a measurement of the treaty ratification hurdles in a country-year unit.Footnote 4 The selection equation also includes all of the control variables included in the outcome equationFootnote 5 and period dummies. All of the variables in the selection equation are lagged by two years to reflect assumptions regarding sequencing of the selection process. To account for unit root all of the independent variables in the selection equation are differenced except for dummy variables. The outcome equation includes the independent variables and controls specified in the previous section as well as period dummies. In the outcome equation unit roots are dealt with by differencing all the variables except for dummy variables. The model is estimated with cluster robust standard errors.

As seen in Table 2, the results regarding donor interests are confirmed in this model. The ratio between positive and negative points is positively associated with voting with the U.S. at the UNGA.

Model four uses an alternative specification of the dependent variable, taking the count of negative points in the reports. In this specification, the dependent variable shows signs of zero inflation and over-dispersion. Furthermore, from a theoretical point of view, the mechanisms which determine whether a country is shamed are different from the mechanisms which determine the extent to which a country is shamed. Therefore, zero-inflated negative binomial estimation is used. This method first runs a logit regression to estimate a country’s likelihood of being reviewed and then runs a count model to estimate the number of negative points. The model uses the CAT ratification indicator and the total number of points (also included as a control variable) in the report as the zero inflators. The model is estimated with cluster robust standard errors. The results indicate that an increase in a country’s rate of voting with the U.S. is associated with a decrease of almost four units in the number of negative points that country receives in its annual report. The model also suggests that there is a positive association between trade volumes and the count of negative points which supports the leverage theory.

The Heckman selection model is widely used to analyze non-random data, but it suffers from a number of flaws, most notably the fact that minor specification changes can sometimes alter the results significantly (Harrigan and Wang 2011; Neumayer 2003). Although this is not the case in the current application, and following in the footsteps of previous research (Egger et al. 2006; Hollyer and Rosendorff 2012; Mazumder 2016) model five uses the approach suggested by Ho and his colleagues for handling non-random data (Ho et al. 2006). Accordingly, the data are pre-processed using the R based MatchIt software.Footnote 6 The treaty ratification indicator is balanced on the regional rate of ratification of the CAT, the square root of IGO membership (Cole and Ramirez 2013), a measurement of treaty ratification hurdles,Footnote 7 the logged FDI volumes and GDP per capita (World Bank 2018). The data is processed using the nearest neighbor matching method. The percent improvement of balance is given in Table 3. Model five is estimated with the balanced treaty ratification variable and the variables used to balance the dataFootnote 8 as well as the same control variables used in the previous models. The model is estimated using fixed effects estimation with period dummies and cluster robust standard errors. The results are in line with the findings of the previous models. Voting with the U.S. is associated with a more positive evaluation by the CAT committee. This further supports the donor interest hypothesis. The positive association between oppression and the positivity of the report holds in this model as well. The model also indicates that there is a positive association between a country’s polity score and the positivity of the report indicating that democratic countries are reviewed more positively.

Table 3 Percent improvement of balance following matching
Full size table

To refute the suspicion that these results are driven by the chosen control variables, model six includes additional alternative specifications of the control variables which have been used in previous research on human rights shaming as well as the balanced treaty ratification variable, the variables used to balance the data and the same control variables used in the previous models. Data gathered by Hendrix and Wong (2014) is used to hold constant the trade and investment linkage with the U.S. specifically. The ill-treatment and torture allegations data (ITT) gathered by Conrad et al. (2013) is used to hold constant the observed country’s human rights violation allegations.Footnote 9 All of the variables are differenced to address clustering. The model is estimated using OLS regression with period dummies and cluster robust standard errors. As seen in Table 2 this model also corroborates the donor interest hypothesis. The ratio between positive and negative points in the concluding observations report is positively associated with voting with the U.S. at the UNGA. The association between oppression and the positivity of the report holds in this model. In addition the model indicates that there is a negative association between FDI and the positivity of the reports indicating that countries with high FDI levels receive more negative reports. This further supports the leverage theory.

The issue of reciprocal causality is a thorny one, and therefore, despite having used various methods for dealing with it, the results presented here should be interpreted with a degree of caution. Because of the inherent difficulty of estimating models with endogenous variables, the results are not, and should not be, interpreted causally. The results show a robust correlation between diplomatic ties with the U.S. and shaming by UN treaty bodies. I make sense of this finding theoretically by using the concept of donor interest.

The findings of this article are somewhat different from those found by Lebovic and Voeten (2006), who found that shaming is determined by normative factors rather than by states’ interests. The reason for this difference may be that Lebovic and Voeten studied shaming by the UNCHR which is composed of UN member states’ representatives, while this study focuses on shaming carried out by the treaty body which is composed of UN officials. It may be that the concern of pleasing the main donor is not shared by the states’ representatives at the UNCHR and is specific to UN officials who are part of the UN’s organizational structure and are therefore more likely to take organizational considerations into account.

5 Conclusions

One of the motivations for this piece of research is to determine whether the factors associated with IGOs’ shaming strategies are similar to the factors associated with NGOs’ shaming strategies. Previous research has established that NGO shaming is explained by donors’ concerns and interests, by the need to prioritize resources and focus them on the countries which are most sensitive to external pressure, by actors’ ambitions to establish their reputation as reliable and professional human rights reporters and by the availability of information from HROs present on the ground and from the media. The analysis finds that the CAT committee’s shaming is similar to NGOs’ shaming in that it too is associated with the concerns of the main donor and more specifically with the diplomatic affinity of the reviewed country with the UN’s main donor — the United States. This finding suggests that the prioritization of funding (and assurance of its flow by not upsetting donors) may be common to various kinds of human rights actors including IGOs. This is why some human rights NGOs take care not to receive funding from governments (Wong 2012). However, IGOs cannot do this. The fact that the targets are the donors complicates IGOs’ human rights promotion (Gourevitch and Lake 2012). IGO shaming is also similar to NGO shaming in that there is evidence indicating that it too is associated with information from the ground provided by the media. Some of the analyses presented here suggest that the contents of the UN treaty body’s reports are associated with media coverage of human rights in the observed country. This may be due to the fact that as media coverage and HRO activity increase, so does the amount of information regarding human rights and this information finds its way into the reports. If this is true then this trend is common to all types of organizations promoting human rights.

Previous research on UN treaty bodies reporting has indicated that there is a certain degree of trade-off between providing an accurate account of the human rights situation in a country and taking input from human rights advocates (Merry and Conley 2011). The findings presented in this article suggest that the CAT committee is more strongly linked with the latter. There is no evidence that the CAT committee of experts’ reports are associated with other experts’ evaluations of human rights. Furthermore some of the models suggest that more oppressive countries are reviewed more positively. Thus, if there is a tradeoff between providing an accurate account of the human rights situation in a country and taking input from human rights advocates, the evidence presented here suggest that the latter has the upper hand at the CAT committee.

UN shaming is also similar to NGO shaming in that there is some evidence that the UN treaty body prioritizes its resources in a way which focuses them on the countries where they are most likely to have an impact — i.e., countries which are most sensitive to pressure from outside. Despite the fact that the CAT committee does not choose the targets of its reports but merely produces reports on the basis of a predetermined cycle, there is some indication that the committee is able to adjust the report and make it more critical when reviewing countries that are sensitive to economic pressure from outside.

These conclusions give reason to question whether producing the country reports is the best possible use of UN treaty bodies’ intellectual, financial and political capital. If treaty bodies’ reports are more closely linked with the reviewed country’s diplomatic ties with the U.S., than they are with providing a factually accurate review of the human rights situation and of compliance with the treaty obligations, then maybe committee members are not making full use of their legal and policy expertise. It is perhaps appropriate to devise other, more effective, ways of using the committee’s expertise to push countries towards compliance with their treaty commitments, perhaps through twinning projects, technical assistance, and training of national administrations and judiciaries. A possible solution to the donor-interest problem could be to structure the OHCHR’s funding differently in such a way that no single member state dominates it. A more modest suggestion would be to switch from an approach which lists progress and shortcomings in the implementation of the entire treaty to an approach which provides deep qualitative feedback on the few most pressing issues taking a less judgmental posture (‘positive’ and ‘negative’ points-focused) and a more constructive and solution-focused approach. This might help bypass the problem of pleasing or upsetting donors. Another possible solution is to avoid the complexity of monitoring donors and their allies by delegating this task to third parties with professional credentials. Experienced human rights advocates such as human rights INGOs seem the obvious candidates for the task.

Considerations of donors’ interests emerge from this piece as a significant feature of the policy process not just in international financial institutions but in international organizations more broadly. This suggests that the donor interest phenomenon, which was initially identified by scholars of financial institutions, with regard to development assistance, might actually be a significant variable in other aspects of international public policy. Future research could look into democratization assistance, peace building projects etcetera, in order to assess the relevance of the donors’ interests to other policy areas. Furthermore, in line with previous findings which indicate that the impact of donors’ interests works both through formal and through informal channels, the current analysis indicates that the link between donors’ interests and policy outcomes is significant even when the donor is not involved in the policymaking process, since the U.S. is not involved in writing up the treaty body’s reports, and even when the policy involves mere symbolic rewards and punishments.

But the significance of donors’ interests stretches beyond the symbolic realm. Shaming can have an impact on international aid and trade. Donors’ interests are potentially linked with a country’s ability to get financial assistance both directly and indirectly; Directly through the impact of donors’ interests in international financial institutions (Andersen et al. 2006b; Dreher and Jensen 2007; Harrigan et al. 2006; Kilby 2009; Stone 2004) and indirectly because donors’ interests are linked with shaming and shaming influences aid and trade flows to the shamed country (Barry et al. 2013; Dietrich and Murdie 2015; Lebovic and Voeten 2009; Murdie and Peksen, 2013; Peterson et al. 2016). Future research could look into the relevance of donors’ interests in making sense of other monitoring and shaming practices. In sum this piece contributes to the accumulated evidence which suggest that donors’ interest is an important variable across policy areas, through various policymaking stages, within symbolic and material policy instruments, and among various kinds of international actors — NGOs and IGOs.

Lastly, regarding the ongoing debate on the difference between norm-driven actors and self-interest driven actors, the results presented in this article do not univocally support either. Some important differences between the two types of actors emerge from the analysis. For example the need to provide an accurate account of the human rights situation is an important factor in NGOs’ shaming but is not associated with IGOs’ shaming. The need to please donors is an important factor associated with both types of actors’ shaming, and this emerges as the most robust and widely applicable finding regarding the factors linked with human rights shaming by both types of actors.