Reading I NEED HELP ON MY HOMEWORK

ClarkandSikkingInformationEffectsandHumanRightsData.pdf

Home >Reading homework help >Reading I NEED HELP ON MY HOMEWORK

Information Effects and Human Rights Data: Is the Good News About Increased Human Rights Information Bad News for Human Rights Measures?

Author(s): Ann Marie Clark and Kathryn Sikkink

Source: Human Rights Quarterly , August 2013, Vol. 35, No. 3 (August 2013), pp. 539-568

Published by: The Johns Hopkins University Press

Stable URL: https://www.jstor.org/stable/24518073

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at https://about.jstor.org/terms

The Johns Hopkins University Press is collaborating with JSTOR to digitize, preserve and extend access to Human Rights Quarterly

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

https://www.jstor.org/stable/24518073

HUMAN RIGHTS QUARTERLY

Information Effects and Human

Rights Data: Is the Good News About Increased Human Rights Information Bad News for Human Rights Measures?

Ann Marie Clark* & Kathryn Sikkink*

ABSTRACT

Changes in quality and availability of information related to human rights violations raise questions about how best to use existing data to assess human rights change. Information effects are discernible both in primary sources of information and data coded by two prominent human rights datasets, the Political Terror Scale (PTS) and the Cingranelli-Richards Human Rights Data Set (C1RI). The authors discuss ways that human rights information has changed for the better, evaluate the scales and their primary text sources for countries in Latin America, and compare them with information drawn from regional truth commission data. Extra caution is advised when using summary data to make inferences about human rights change.

Ann Marie Clark is Associate Professor of Political Science at Purdue University in West Lafayette, Indiana.

f Kathryn Sikkink is Regents Professor and McKnight Presidential Chair in Political Science at the University of Minnesota in Minneapolis, Minnesota. For comments we would like to thank David Cingranelli, Rosalee Clawson, Dara Kay Cohen, Yasmine Ergas, Ryan Goodman, Robert Keohane, Hunjoon Kim, Sophie Lelièvre, James A. McCann, Will Moore, David Richards, Kiyoteru Tsutsui, Jana von Stein, Geoff Dancy, and members of the Fall 2010 New York University Seminar on International Relations and International Law and Columbia University's Institute for the Study of Human Rights, where Ann Marie Clark was a visiting scholar in 2010-2011. Brooke Coe, Geoff Dancy, Raul Danyi, Andrew Grover, Julia Kaspar, Darrah McCracken, Claudia Munoz, and Carrie Walling provided research assistance.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

540 HUMAN RIGHTS QUARTERLY Vol. 35

I. INTRODUCTION

Human rights have become a high-profile issue in global politics and inter national relations scholarship. Moreover, documentation of governments' human rights practices in recent decades has increased dramatically in quantity and quality. Recent reports from nongovernmental organizations (NGOs) like Amnesty International (AI) and Human Rights Watch, among others, and governmental sources like the US Department of State (State Department) now typically contain much more and better information than earlier reports. More recent reports also tend to document a wider range of human rights violations. Therefore, social scientists are accruing ever more data on human rights, permitting a broader range of analytic techniques to be employed in causal studies. The increase in the quality and quantity of information about human rights violations around the world is good news for scholars and practitioners interested in the politics of human rights change. However, the varying availability and precision of country-specific information give rise to two major issues explored below: how information availability has changed; and the implications for how researchers should interpret such information, particularly when it is processed for secondary use in summary data sets.

The choice of data, the artifacts of data gathering, and the information environment within which human rights violations take place all potentially affect the ability to understand variation in levels of violations over time. As this area of research continues to grow, researchers should beware of pos sible "information effects": patterns in the data stemming from the process of information collection and interpretation, rather than the process that actually gives rise to human rights violations or their mitigation. This article brings four questions and related hypotheses to bear on the likelihood of information effects in the data. How good is the source information? Do changes in information supply impact data coding? Do the data sets suit ably register changes in type of repression, as reflected in the source reports over time? Do the data sets and their source reports register worsening and improvement adequately?

Much of the quantitative human rights research up to now has relied on two data sets, the Political Terror Scale (PTS)' and the Cingranelli-Richards Human Rights Data Set (CIRI).2 Both gauge countries' human rights perfor mance using the annual reports of AI and the State Department. Researchers using both quantitative and qualitative data should be on the lookout for

1. Political Terror Scale, available at http://www.politicalterrorscale.org. 2. CIRI Human Rights Data Project, available at http://www.humanrightsdata.org.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 541

information effects, but users of these summary databases face particularly acute issues of interpretation. Patterns related to information availability and how the scales deal with variation in the substantive nature of human rights violations can be embedded in the coding and therefore difficult to separate from patterns related to actual changes in violations.

This is not an arcane concern, because quantitative studies of human rights are one of the fastest growing fields in research on world politics. Between 1999 and 2011, ninety articles using either PTS or CIRI were pub lished in the top political science and human rights journals: eight appeared in Human Rights Quarterly.3 Many of these articles make important causal and theoretical claims on issues including globalization's effect on human rights, the impact of regime type on the use of torture, and hypothesized causal relationships between human rights and foreign aid allocation, de mocracy, transitional justice mechanisms, internal armed conflict, and treaty commitment and compliance. The information patterns and measurement issues have important implications for understanding the dynamics of human rights and how findings are evaluated.

In a recent review essay, Emilie M. Hafner-Burton and James Ron char acterized the results of quantitative human rights studies as "pessimistic" about human rights improvements, in comparison with more "optimistic" case study work, suggesting that mixed large-n-based results belie causal accounts of positive change based on case studies.4 While skepticism is healthy and desirable in all empirical research, the current availability of more informa tion complicates judgments for both case-study and large-n researchers as to the nature and direction of human rights change. In particular, greater awareness coupled with more information can make a phenomenon seem more frequent when its empirical frequency has not actually risen.

This question plagues research on crime rates and the prevalence of cer tain diseases and medical conditions. For example, while better availability of mammography screening for breast cancer is beneficially associated with earlier identification of the disease, higher rates of diagnosis and treatment in younger women,5 it is perhaps less beneficially associated with possible over diagnosis.6 After advances in diagnosis there is a period when increased

3. From top thirty political science journals, as listed in Micheal W. Giles & James C. Garand, Ranking Political Science Journals: Reputational and Citational Approaches, 40 PS: Pol. Sci. & Pol. 741, 743-44 (2007) (research assistance provided by Brooke Coe).

4. Emilie M. Hafner-Burton & James Ron, Seeing Double: Human Rights Impact through Qualitative and Quantitative Eyes, 61 World Pol. 360, 363 (2009).

5. Carol DeSantis, Rebecca Siegel, Priti Bandi & Ahmedin Jemal, Breast Cancer Statistics, 2011, 61 Cal. Cancer J. Clinicians 409, 411 (2011); G. Maskarinec, L. Wilkens & L. Meng, Mammography Screening and the Increase in Breast Cancer Incidence in Hawaii, 6 Cancer Epidemiology, Biomarkers & Prevention 201, 201 (1997).

6. M. Kalager, et al., Overdiagnosis of Invasive Breast Cancer Due to Mammography Screening: Results From the Norwegian Screening Program, 156 Annals Internal Med. 491,491(2012).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

542 HUMAN RIGHTS QUARTERLY Vol. 35

diagnoses cannot readily be linked to "true" increases in cancer rates.7 Therefore, with the improvements in technology and the ability to reach previously underserved populations for screening, the increase in diagnoses may or may not be indicative of an actual increase in breast cancer. Further investigation is required to assure valid conclusions. Cancer researchers believe that once monitoring capacity levels off in a particular population, it is easier to estimate and compare cancer rates and to understand the true causes of increases or decreases in frequency.8

Similarly in the human rights field, increased awareness coupled with better information can create a false perception that a phenomenon has increased in prevalence, or increased more than it really has. Keck and Sikkink identified the human-rights-information-related dilemma as the "information paradox."9 They suggested that new reporting on previously overlooked violations such as rape and other forms of violence against women may contribute to a picture of worsening human rights that may not always be related to worsening practices. Few human rights research ers have devoted sustained attention to how information effects could be

relevant to empirical findings. As research moves forward, it is worth tak ing a critical but constructive look at widely used data sources in order to sharpen awareness and move toward reducing the impact of information effects not systematically related to the processes that actually give rise to human rights violations or their mitigation. This would enable researchers to make better informed comparisons and conclusions using the important, but imperfect, data that we do possess.

The ways information changes have impacted reporting and coding can be identified and investigated from several different angles. A focus on Latin America, below, permits a closer evaluation of how reports have changed in length and scope for countries in this region, as well as whether and how the source texts and data series reflect known improvement. Since the inception of the PTS and CIRI datasets, Latin American countries have experienced severe human rights violations, and changes in the types of violations, as well as notable improvements. To get a general picture of how well cod ing corresponds with the worst physical integrity abuses—which they are intended to measure—PTS and CIRI coding is compared with data from five country-level truth commission investigations in Latin America. Two case countries, Guatemala and Brazil, provide further illustration.

7. E-mail from Sophie Lelièvre, Associate Professor of Cancer Pharmacology, Purdue Uni versity, to Ann Marie Clark, Associate Professor of Political Science, Purdue University (23 Apr. 2012) (on file with author).

8. Id.

9. Margaret Keck & Kathryn Sikkink, Activists Beyond Borders 194 (1998).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 543

The comparison yields evidence of mild-to-moderate information effects. It was expected that the two different coding schemes—PTS and CIRI—might be susceptible in different ways to information effects. However, even though they employ different coding strategies, to a large degree they share similar problems. After controlling for variation related to systematic levels of abuse, the PTS and CIRI indices both correlate to some degree with the amount of information included in a report. The numeric scales tend to be "sticky," i.e., slow to change, after past periods of severe abuse. In Guatemala, increased attention to violations after a period of State Department bias seems to in tensify the human rights assessment, and this is reflected in the coding. In Brazil, increased reporting on new human rights claims related to different forms of violation such as political killings cause the data coding to reflect worse human rights on average for more recent periods.

Recognizing how information effects accrue due to variation in the availability and precision of country-specific information should lead re searchers to interpret study results with care. This becomes especially im portant when researchers make generalizations based on longitudinal and cross-sectional comparisons from data sets like PTS and CIRI, which are derived from contemporaneous reporting. If systematic information effects tend to register human rights deterioration more easily than improvement, as argued below, some of the pessimism that has been expressed about the differences between large-n and case study conclusions may be premature. Such effects could be relevant to causal and theoretical claims made in

dozens of human rights studies. This does not mean that researchers should stop using the data, but it does reinforce the message that researchers should use a variety of methods wisely and seek additional tests of hypotheses about cross-national and longitudinal human rights processes. Comparing the findings of both large-n research and case study work in a well-informed manner is essential for advancing and testing knowledge about the dynamics of human rights change.

SOURCES AND CODING: HOW PTS AND CIRI USE HUMAN RIGHTS REPORTS

The two major human rights data sets mentioned above, PTS and CIRI, rely on the same two sources: the Amnesty International Annual Report/0 and the State Department's Country Reports on Human Rights Practices.11

10. Amnesty Int'l, Annual Report (1975), available at http://onlinebooks.library.upenn.edu/ webbin/serial?id=aireport.

11. U.S. Dep't of State, Country Reports on Human Richts Practices (1979), available at http:// onlinebooks.library.upenn.edu/webbin/serial?id=crhrp (hereinafter Country Reports [year]).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

544 HUMAN RIGHTS QUARTERLY Vol. 35

PTS and CIRI assign numeric values to countries based on the AI and State Department annual report texts and now cover most countries of the world, although coverage has varied. CIRI coding begins in 1981.12 Complete PTS data begins in 1981 and now incorporates the coding that Steven C. Poe et al.13 extended from 1976-1980 for some countries.14 The advantage of sum marizing text reports through a coding scheme is that information can readily be compared and used to test causal hypotheses with statistical techniques in cross-national and longitudinal contexts. However, if artifacts of information effects are created by or embedded in the data, non-random sources of error may be introduced that can bias results and affect the descriptive validity of the comparison. The CIRI and PTS coding schemes are outlined only briefly below, since they are well described on the scales' own websites and have been discussed in print by the datasets' curators, David L. Cingranelli and David L. Richards15 and Reed Wood and Mark Gibney,16 respectively.

The Political Terror Scale (PTS). The PTS coding categories address how broadly state-based political violence, defined as threats to physical integrity rights, extended to a country's population in a given year.17 PTS character izes the content of report entries on a scale of 1 to 5. Each year yields two PTS codes: one assigned to the AI report, and one to the State Department report.18 A coding of 1 represents the prevailing rule of law, with no or very rare incidence of political violence, while 5 represents extended political violence in which "terror has expanded to the whole population."19 The PTS does not attempt to count incidents of repression.

Cingranelli-Richards Human Rights Data Set (CIRI). The CIRI dataset creates a separate score for four types of physical integrity abuse: political imprisonment, torture, political killing, and disappearances.20 It is not a strict count of violations against individuals—that would be impossible—but early coder guidelines indicated that, where possible, the coding should cor respond with reported frequency.21 Values are assigned from 0 to 2, based

The CIRI Human Richts Data Project, supra note 2. Steven C. Poe, C. Neal Tate & Linda Camp Keith, Repression of the Human Right to Per sonal Integrity Revisited: A Global Cross-National Study Covering the Years 7 976-1993, 43 Int'l Stud. Q. 291, 299 (1999). Political Terror Scale, FAQ available at http://www.politicalterrorscale.org/faq.php. David L. Cingranelli & David L. Richards, The Cingranelli and Richards (CIRI) Human Rights Data Project, 32 Hum. Rts. Q. 401, 409 (2010); see also David L. Cingranelli & David Richards, Measuring the Level, Pattern, and Sequence of Government Respect or Physical Integrity Rights, 43 Int'l Stud. Q. 407, 409-10 (1999). Reed M. Wood & Mark Cibney, The Political Terror Scale (PTS): A Re-introduction and a Comparison to CIRI, 32 Hum. Rts. Q. 367, 373 (2010). See Political Terror Scale, supra note 1. Political Terror Scale, About, available at http://www.politicalterrorscale.org/about.php Id.

CIRI Human Rights Data Project, FAQ, available at http://humanrightsdata.org/faq.asp. David L. Cingranelli & David L. Richards, The Cingranelli-Richards (CIRI) Human Rights Data Project Coding Manual at 8, 10, 13-14 (Version 7.30.08 2008), available at http://www. humanrightsdata.org/documentation/ciri_coding_guide.pdf.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 545

on whether violations are "practiced frequently" (or roughly 50 or more violations), coded as 0; "practiced occasionally" (or 1-49 violations), coded as 1; or "have not occurred," coded as 2.22 CIRI's Physical Integrity Index (CIRI-physint) adds the four categories, so its range is 0 to 8.23

III. CHARACTERISTICS OF THE REPORTS USED BY PTS AND CIRI

The AI and State Department reports are issued annually within a year of the events they cover, and the two document series are the closest thing available to ongoing human rights reports of record. AI draws on material collected by AI researchers from domestic human rights organizations, other observers, and sometimes on-site missions.24 The State Department draws on in-country embassy staff reports in addition to information generated by NGOs and other observers.25 Further attention to how the reports are produced, before they are coded numerically, helps to contextuaiize the advantages and disadvantages of the associated data indicators.

All social science data sources come with caveats. Several are specific to human rights data. In the midst of major episodes it can be difficult to identify precisely what is occurring and how many people have fallen vic tim to human rights violations. External human rights monitoring may be hindered by repressive governments that try to hide, downplay, or dismiss information. Widespread repression can also shut down domestic NGOs, keep them from organizing, or block their links to outside contacts. Ftowever imperfect, contemporaneous reports of knowledgeable outside monitors are likely to be better than the self-representations of states at revealing the nature of repression, but under repressive circumstances even the best investigators will produce imperfect reports. The imperfections stem from the nature of the enterprise.

The history of how and why the reports have been produced is also relevant for the detection of information effects in the data. The reporting organizations have not necessarily followed the same formula over the years,

Id.

David L. Cingranelu & David L. Richards, Short Variable Descriptions for Indicators in the Cincranelli-Richards (ClRI) Human Rights Dataset, 3 (Version 11.22.2010 2010), available at http://www.humanrightsdata.org/documentation/ciri_variables_short_descriptions.pdf. in Figures 1-6, the CIRI values are reversed for ease of visual comparison, so that lower scores on the graphs represent less severe violation levels for both scales. Amnesty Int'l, Frequently Asked Questions: How does Amnesty International get its information?, available at http://www.amnesty.org/en/who-we-are/faq#how-ai-gets information.

The State Department outlines how each annual report is prepared in the Overview and Acknowledgments section. See, e.g., U.S. Dep't of State, Country Reports on Human Rights Practices for 2012: Overview and Acknowledgments (2012), available at http://www. state.gov/documents/organization/204296.pdf.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

546 HUMAN RIGHTS QUARTERLY Vol. 35

nor has the level of information remained constant.26 Contingencies such as variation in personnel, available sources, and resources devoted to any given country from time to time also give rise to variation in coverage. All of this becomes especially apparent when comparing more recent annual reports with early ones.

The early reports of both AI and the State Department covered fewer countries. ATs annual report began as a summary of its activities for its mem bers, and the earlier reports are more limited in scope when compared with today's reports. For example, earlier reports tended not to include countries with less problematic human rights records.27 The US Congress called for human rights evaluations on countries receiving US aid in the 1970s, and the reports before 1980 only covered countries receiving US aid.28 The number of countries covered did grow, but incrementally.

Later AI and State Department reports also discuss more forms of viola tions, something that is rarely noted. ATs substantive focus has expanded,29 and State Department reports regularly incorporate new mandates originat ing from Congress and the executive branch.30 Although coverage of physi cal integrity violations (the major focus of the PTS and of some of CIRI's measures) has been a relative constant, the definition of what constitutes torture or state-sponsored killing has expanded over the years. For example, previously, "political killings" referred mainly to situations when the govern ment killed its political opponents on a large scale.3' Today, the concept of political killing or "extrajudicial, summary, or arbitrary executions" extends to police use of excess lethal violence and cases when private actors kill their political opponents and the state fails to stop or punish them.32

Perhaps least apparent to potential data set users is that later reports likely contain more and better information. The bureaucratic and human rights research capabilities of both AI and the State Department have

SD procedures are detailed in an appendix to each recent report. See, e.g., U.S. Dep't or State, Country Reports on Human Rights Practices for 2012: Appendix A - Notes on Prepara tion of the Country Reports and Explanatory Notes (2012) available at http://www.state.gov/ documents/organization/204698.pdf. Al's website has a brief discussion of its sources and procedures in Amnesty Int'l, Frequently Asked Questions, supra note 24. Poe, et al., supra note 13, at 298. Id. at 300.

Peter R. Baehr, Amnesty International and Its Self-Imposed Limited Mandate, 12 Neth. Q. Hum. Rts 5 (1994); Ann Marie Clark, Diplomacy of Conscience: Amnesty International and Changing Human Rights Norms (2001 ); Stephen Hopgood, Keepers of the Flame: Understanding Amnesty International (2006).

Interview with State Dep't official (anonymous at interviewee's request), in Wash., DC (11 May 2011). Clark, supra note 29, at 102. Id.; UN Office or the High Commissioner for Human Rights, Fact Sheet No. 11 (Rev. 1), Ex trajudicial, Summary or Arbitrary Executions, 7 (1997), available at http://www.ohchr.org/ Documents/Publications/FactSheetl 1 rev.1en.pdf.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 547

grown markedly since they began their annual reports. In the early 1970s, before reports were issued, only one person in the State Department had responsibility for human rights.33 At the close of the Carter Administration in 1980, there were twenty professional human rights staff positions in the State Department.34 At the end of the Clinton Administration twenty years later, the State Department's Human Rights Bureau had more than 100 staff members.35 Al's staff doubled between 1975 and 1985 to 205 members.36 In

addition, both organizations now use more information from other sources, especially domestic NGOs. A present-day State Department official described regular communication with NGOs before, during, and after each annual report is issued.37

Finally, a question commonly posed by newcomers to the data when they learn that CIRI and PTS are based on State Department and AI reports is whether the organizations' reports are biased toward political or orga nizational points of view. Bias is sometimes evident, but it has not been constant over time. Political bias in favor of US allies was pronounced in some cases covered by early State Department reports, as illustrated by the case of Guatemala in the early 1980s.38 Al's reports do not evidence similar organizational bias, but the organization is committed to a human rights ethos that places high value on the life of every individual whose rights are violated. Therefore, its reports do sometimes discuss improvement or deterioration, but might be less attuned to numeric reductions in political killings, for example, that would be relevant in a scholarly research context.39 With Al's growth in resources and the State Department's reporting reforms discussed below, organizational biases and differences in information access have diminished. However, any early biases that do exist will be reflected in the databases, since the coding schemes address the textual content of the country reports as written.

Interview with State Department official, supra note 30. Jo Marie Griesgraber, Implementation by the Carter Administration of Human Rights Legislation

Affecting Latin America 103, 106 (Ph.D. dissertation, Georgetown University. 1983). Kathryn Sikkink, Mixed Signals: U.S. Human Rights Policy and Latin America 206 (2004). Stephen Hopgood, Amnesty International's Growth and Development since 1961, in Amnesty International 50 years: Reflections and Perspectives 75, 90 (Wilco de Jonge, Brianne McGonigle Leyh, Anja Mihr & Lars van Troost eds., 2011). Interview with State Dep't official, supra note 30. Country Reports (1982), supra note 11, at 516; Country Reports (1984), supra note 11, at 541.

Because early AI reports concentrated on the worst violators, the global cross-national means of the data sets' coding values are likely to be biased toward worse violations in early years, as discussed by Poe, et al. supra note 13, at 300; see also Steven C. Poe, Sabine C. Carey & Tanya C. Vasquez, How Are These Pictures Different? A Quantita tive Comparison of the US State Department and Amnesty International Human Rights Reports, 1976-1995, 23 Hum. Rts. Q. 650, 655-57 (2001).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

548 HUMAN RIGHTS QUARTERLY Vol. 35

The bottom line is that interpretations of the data should be made in line with the limitations of the original sources and, if possible, go some way toward improving upon rather than exacerbating those limitations. Neither AI nor the State Department constructs the reports with the intent of pro viding a basis for quantitative indices. Accordingly, knowledge of the data's cross-national and temporal characteristics is essential for the interpretation of study results based on such comparisons.

QUESTIONS TO BE ADDRESSED

In light of the general reporting changes and coding approaches described above, four general questions related to information effects are outlined below. Hypotheses are summarized in Table 1.

1. How good was information about past violations? The quality of reporting has improved over time so that reports reflect better information on human rights violations. The longitudinal variability is a form of mea surement error with its source in the report texts themselves. If AI and State Department reports now document violations that would previously have been hidden, errors in comparability within single countries over time could be introduced for both scales. (See row 1 of Table 1). Not every country has been equally accessible, so information quality has also varied by country. As noted above, AI and the State Department have devoted differing levels of critical attention to particular countries in their annual reports at differ ent times. Where they exist, these biases are likely to enter the databases, since the databases emphasize that coding should be consistent with the textual content of the country reports and not attempt to filter the informa tion. This kind of information effect could occur for both PTS and CIRI.

However, because PTS codes the AI and State Department reports separately while CIRI does not, one can better understand relative differences in the reporting sources by scrutinizing PTS-Amnesty International (PTS-ai) and PTS-State Department (PTS-scf) scores within countries. To gauge the differ ences, we elaborate on informational access and reporting issues below, as well as the State Department's early political bias as reflected by its reports on Guatemala in the 1980s.

2. Does greater quantity of information affect coding? Researchers have more information and, often, access to more detailed information about human rights violations now than in the past. If more detailed information drives harsher coding, a correlation between harsher coding and longer re ports after controlling for actual change in repression should be observable. Studies of human information processing suggest that coders may be sensi

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 549

tive to longer and more complex arguments, especially when deductively relating information to categories.40

If information was just as accurate in the past about the general nature of repression, but there is more detail now, there may be variation in how the two scales process these differences. (See row 2 of Table 1). The PTS does not count numbers of violations, but instead characterizes their scope. While its general focus on scope may limit the scale's ability to be specific about amount and types of violations, it also may be less affected by more detailed reports. Conversely, because the CIRI index has some (at least im plicit) count criteria, it may be somewhat more susceptible to information effects based on amount of information or level of detail. To gauge these issues, word counts of AI and State Department reports by country entry and year for the Latin American region were gathered for AI from 1976-2006 and for the State Department beginning in 1981, and for all countries glob ally from 1994-2006. While report length is only a rough proxy for more information, no other study has examined this aspect of the original report texts. Changes in report length serve as an entry point for analysis.

3. Do the scales detect changes in type of repression? PTS and CIRI dif fer in how and whether they record variation in type, scope, and intensity of repression. (See row 3 of Table 1). One would expect PTS to be stronger than CIRI in its resistance to information effects related to the source texts'

reporting length and detail, because its coding scheme is not as related to frequency. However, unlike CIRI, the PTS does not differentiate among types of violations that would be coded in the aggregate at similar levels. Without extended in-depth textual analysis, a comprehensive evaluation of whether and how the types of violations reported have varied is not possible, but differences are illustrated by discussing coding in two Latin American cases: Brazil and Guatemala.

4. How well do the scales register worsening or improvement? Is there evidence of a ceiling effect? For both scales, a ceiling effect would be observed if the scaling is not as sensitive to changes at higher levels of repression as it is to low level changes. (See row 4 of Table 1). It is hypothesized that a ceiling effect will be most relevant to the CIRI data, since the "frequent" violations category of the scale covers a very wide range, and the PTS seems to allow for intermediate variations in scope and intensity. To acquire an independent source of human rights data for comparison, information on deaths and disappearances compiled by national truth commissions of five

See Michael D. Cobb & James H. Kuklinski, Changing Minds: Political Arguments and Political Persuasion, 41 Am. J. Pol. Sci. 88 (1997); Daniel N. Osherson, et al., Category Based Induction, 97 Psychol. Rev. 185 (1990); Caren M. Rotello & Evan Heit, Modeling the Effects of Argument Length and Validity on Inductive and Deductive Reasoning, 35 J. Experimental Psychol.: Learning, Memory, & Cognition 1317 (2009).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

550 HUMAN RIGHTS QUARTERLY Vol. 35

Latin American countries—Argentina,41 Chile,42 El Salvador,43 Guatemala, and Peru45—is referred to and organized by year for each country to match the organization of the CIRI and PTS data.46

V. ANALYSIS AND DISCUSSION

A. Information quality

The first question asks whether and how variation in quality of information enters the texts and affects the coding. It is never easy to collect informa tion on human rights violations, but reporting and information quality have improved since the reports began. Other things being equal, more recent violations are less likely to be hidden from outsiders. These issues, which were exacerbated by early State Department political bias, are apparent in early reporting on Guatemala, as discussed below.47

Repression in Guatemala was at times so severe that it eliminated or silenced any human rights movement, and thus eliminated the information that would have been generated by this movement activity. The worst period of killings and disappearances in Guatemala occurred in the early 1980s:

On Argentina, see National Commission on Disappeared Persons (CONADEP), Nunca Mas (1984), available at http://www.desaparecidos.org/nuncamas/web/english/library/nev again/nevagain_001 .htm. On Chile, see Truth and National Reconciliation Commission, Informe Rettig: Informe de la Comisiön de Verdad v Reconciliaciön (1991). On El Salvador, see From Madness to Hope: The 12-Year War in El Salvador: Report of the Commission on the Truth for El Salvador, U.N. SCOR, The Commission on the Truth for El Salvador, U.N. Doc. S/25500 (1993), available at www.usip.org/files/file/ EISalvador-Report.pdf. On Guatemala, see Commission for Historical Clarification (CEH), Guatemala: Memory of Silence (1999), available at http://www.documentcloud.org/documents/357870-guatemala memory-of-silence-the-commission-for.html; Patrick Ball, Paul Kobrak & Herbert Spirer, Am. Ass'n for the Advancement of Science & Ctr. Int'l por Investicaciones en Derechos Humanos

(CIIDH), State Violence in Guatemala, 1960-1996: A Quantitative Reflection (1999), available at http://shr.aaas.org/guatemala/ciidh/qr/english/en_qr.pdf. In the Guatemalan case, the CIIDH (above) breaks down data by year, using the documentation standards that had been established for the truth commission work but with information pooled from the press and from testimony provided to human rights organizations. We use Ball, et al.'s yearly data, which can be compared more directly to the PTS and CIRI scales. The total number of state-sponsored deaths and disappearances documented by CIIDH for the period 1959-1995 is comparable to that registered by CEH, the UN-organized truth commission.

On Peru, see Comisiön de la Verdad y Reconciliaciön, Informe Final (2003). Data adapted and extended from Sikkink, supra note 35. Stephen C. Ropp and Kathryn Sikkink, International Norms and Domestic Politics in Guatemala, in The Power of Human Rights: International Norms and Domestic Change, 188-9 (Thomas Risse, Stephen C. Ropp, and Kathryn Sikkink eds., 1999).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 551

Table 1: Information Effects and Data Set Differences: Hypotheses

Information Effects and Data Set Differences: Hypotheses

Information component PTS CIRI 1. Quality of information, Improvements since early reports (better coverage, Improvements since early reports (better coverage, retrospective: how good was less bias) mean that coding may not be equivalent less bias) mean that coding may not be equivalent information about past violations? over time across countries; State Dep't bias expected over time across countries; expect difficulties with

to be mitigated by PTS's separate coding of AI and early State Dep't bias. State Dep't reports.

2. Quantity of information and Longer and more detailed reports not expected CIRI's count-related coding may render it sensitive detail: does more information to affect coding because of PTS's standards-based to provision of more detail; expected to be more correlate with harsher coding? criteria. vulnerable to false precision. 3. Amount of information, new Measure may not respond well to changes in context Coding subdivided by type of violation; may types of violation: what happens or violations. respond better to changes in form and intensity of when type of violations changes? violation. 4. Degree of violation: how well Fewer problems with ceiling effect expected in upper May not reflect change in high abuse, again does coding respond to extreme levels than with CIRI. because of implicit frequency count (ceiling effect). worsening or improvement?

DC CD ~ •— w 2 m 5 ~ o

ora~Q_ re" g- >io _c g-E g Ecud ^ c *- 00 DC— O 0) F

5 >-£ $ .En« Q- $ S 3 r~i m ~R '£ "5 CD 'u DC O" ° 58c ; °"o p gj<u

o E

CD CD Q_

i i! i v S o ■ -S u ^ "J 8 >.£ 73 j" ^ 2"° o

<= g "O Q. O™ c Sracojaj p.E ^ . c S u n ■- "O re" o £ w 3j— -o o rc E _ o '-o O- c u fc .2

C m 2 C p ID ^ > "! t En u.Um b=J2 o 'o '■= cd •- =5 75 2 -° . a> q» So5J= J» £ • Q_ ^ OD -Q ■*-; C ^ B m> £ o sx 3°b So

C DC

u F — ^ _o £ •& fo •— . QJ 3 :£ c c oc— CD CD .y 1- o p — n3 CU K-QSC -K " 0"° ofiC fl) _ T3"0 •— C B £0"° c J) «ra

l/> *->

3 c -Q 3

Soo "o <-> _ „ | ~ _ Jre E -go u ̂D

c £ u Q 5 ° b _b t i.i« -§.g . ^ 8 > .2 .1 £ 8 I 2 ^2 I " " S-o">. -"2fc .Eos 3

I fill Sfl 211 If I _ .

~a x <d v5 „ ^ B T, B ° U B, C B C "O C 9" c SP-2 n « (DO 3 5 Ere g-_ "OS u c | £|*< .E ^ g u o-re o o-o jg 2 ^ i- DC X OC B •s s_o--~ s £. $ ®"f °-"9 o S -C g a 5S§8 °2 -5 « o ~ <d 2 t: -1" ° ,2

1. Mii f£ 1 I "3 »- DC • - qj ^ > <~ ^ g- c & S(

c 2 ^ ^ Qj

w-fi ^ C « III 11 .li Ml C^'S Org 5"®^ - °T?.C <s E ° .2-t;». *oE

§ !&& ill 111 lil E ||| Igl 111 1|| O .= -O e >e — \Z ■* > 9J **•

c ° > c >-5ri o 05?® ■S ill lis I;I sit ro 5 »5 *- O ?2 r- rt5Q.g W .. -Sf OO*- bjq2?

I all 511 ly ^ • Id c . « o . >"^ . ^ § ■E 1-2: .5 CN -O U rf^S

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

552 HUMAN RIGHTS QUARTERLY Vol. 35

Guatemala logged 25,928 identified and unidentified deaths and disap pearances in the years 1980-1983, or roughly 79 percent of all deaths and disappearances that occurred between 1970-1995 (see Table 3). This figure, large as it is, is likely to be an undercount. In the mid-1980s, human rights organizations did begin functioning again in Guatemala, although they still faced profound repression.48 The process of re-democratization in Guatemala after 1985 contributed to a more information-rich environment.49

Al's reporting during the most repressive period attributed most human rights violations to state officials or government-sanctioned death squads.50 Although details may not have been available contemporaneously, it is clear that AI fully understood the extreme seriousness of the situation in Guatemala at the time. Its reports highlight the movement of repression into the rural provinces where indigenous Mayans lived. In 1980, AI wrote that "hundreds" had been "held, interrogated, tortured and eventually murdered," and "an estimated 1,800 people" had been "'disappeared' and killed."51

AI reported a worsening situation in 1981 and 1982.52 However, in 1982, the year for which truth commission reports retrospectively revealed the highest levels of state-sponsored deaths and disappearances ever reported in any country in the hemisphere, the State Department dramatically under reported the scale of killing and suggested that it was impossible to know who was responsible for the killings—the government of General Efrafn Rfos Montt, right-wing groups, or the guerrillas.53 The State Department reported:

In Guatemala's cities there has been a marked decrease in killings and disap pearances since Rfos Montt came to power, although some abuses continue to be reported. . . . The situation in the countryside, on the other hand, remained unclear for several months. . . . During this period it is believed that members of both the army and the guerrillas were responsible for killings of civilians. . . . Although there continued to be credible reports of human rights abuses by some military units, the overall conduct of the armed forces had improved by late in the year.54

The difference between the State Department's biased reporting and Al's more accurate reports on Guatemala is reflected in the PTS coding. AI reports painted a different picture, and were much closer to the retrospective truth

Americas Watch, Helsinki Watch, & Lawyers Committee for International Human Rights, Critique: Review of the Department of State's Country Reports on Human Rights Practices for 1982 43-^14 (1983). See, e.g., CEH, supra note 44. Amnesty Int'l, Amnesty International Report 1980 139 (1980). Id.

See Ann Marie Clark, "A Calendar of Abuses": Amnesty International's Campaign on Guatemala, in NGOs and Human Rights: Promise and Performance 55, 63 (Claude E. Welch, Jr. ed., 2001). Country Reports (1982), supra note 11 at 516. Id.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 553

commission reports. The PTS coding scheme assigns Al reports (PTS-a/) the worst score, 5, for the deepest period of the genocide (1980-1983), but also for the entire period around that time, 1977-1985, and again in 1990. The PTS-sd value, based on the texts of the contemporaneous State Department reports, remained at 4 (instead of 5, the worst score) throughout the 1980s.

It is now known that embassies were expected to shape their reporting to the overall policy the Reagan Administration was pursuing in Guatema la,55 and the tone of the State Department's 1982 report appears to reflect that policy. Although the State Department human rights reports expressed uncertainty about who was responsible for deaths and disappearances, a confidential CIA cable from the embassy at the same period did not:

Commanding officers . . . have been instructed to destroy all towns and villages which are cooperating with the EGP [Ejército Guerrillero de los Pobres (Guer rilla Army of the Poor), the main guerrilla organization in Guatemala at the time] and eliminate all sources of resistance. . . . When an army patrol meets resistance and takes fire from a town or village, it is assumed that the entire town is hostile and it is subsequently destroyed.56

In the late 1980s, Congress directed the General Accounting Office (GAO, now the Government Accountability Office) to examine the State Depart ment's human rights reporting practices, including policies and procedures for report preparation, whether the reports were accurate and unbiased, and whether State Department staff possessed the resources and training they needed to prepare the reports.57 The GAO concluded that the reports on El Salvador, Guatemala, and the Philippines had "excused these governments from responsibility for abuses based on their promises of corrective action."58 In Guatemala, even US embassy officials acknowledged past bias.59 As a result of such scrutiny, by the mid-1980s reports became more fact-based, as State Department report preparers were "increasingly willing and able to resist pressures to slant the Reports for political purposes."60 According to the GAO, the 1989 State Department report reflected a more "objective and frank" critical approach that evidenced a break from past policies.61

Interview with F. Allen Harris, Diplomat, U.S. Dep't of State {by Kathryn Sikkink), in Washington, D.C. (5 Mar. 2003). U.S. Central Intelligence Agency, Secret Cable: Counterinsurgency Operations in El Quiché (1982) available at http://www.gwu.edu/~nsarchiv/NSAEBB/NSAEBBl 1/docs/doc 11 .pdf. U.S. General Accounting Office (CAO), GAO/NSIAD-90-224, Human Rights: State Depart ment's Commitment to Accurate Reporting Has Increased 2 (1990), available at http://www. gao.gov/assets/150/149785.pdf. Id. at 13. Id. at 3.

Judith Innés de Neufville, Human Rights Reporting as a Policy Tool: An Examination of the State Department Country Reports, 8 Hum. Rts. Q. 630 (1986). See also Poe, et al., supra note 39. GAO, supra note 57, at 15.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

554 HUMAN RIGHTS QUARTERLY Vol. 35

The increased reliability of State Department reports since those early years means that in addition to changes in information access that impacted the quality of human rights reporting in repressive circumstances, the levels of violations reported for US allies may also be biased downward for political reasons at the beginning of the State Department data series.

B. Information supply

Given that the overall quality and amount of human rights reporting has improved, how might coding data be affected? Figure 1 charts average lengths of Latin American AI reports from 1976-2006 and State Department reports from 1981; global average report lengths (including Latin America) are shown separately, from 1994-2006. The average State Department report entry on a given country has grown longer over the years, and average AI entries have also increased slightly in length. The changes in average report length may reflect a true increase in number of human rights violations, but report length could reflect either knowledge about violations that might have remained hidden previously or more detail about violations that would have been known in the past. If known violations (but not the "true number") are growing, one would expect worsening data scores over time apart from whether more or worse violations are occurring. Longitudinal comparisons would be statistically biased due to better information rather than actual increases. In addition, the question must be asked whether coding scales are vulnerable to effects related to length but possibly unrelated to actual severity of violations.

Existing reports can be used to check for correlation between length and how the data sets have interpreted the reports. This section presents a mini-analysis of the effect of report length on coding, focusing on the Latin American country reports from 1981-2006. To control for systematic, violation-related variation in human rights levels, when testing for effects of AI report length, the State Department's PTS score (PTS-sd) serves as a control variable, and vice versa, and both AI and State Department values serve as controls for CIRI.62 Because the scales differ in range and direction, for this analysis the values for CIRI and PTS were transformed to a zero-to one scale, with zero representing the best human rights performance and 1 representing the worst. A lagged dependent variable controls for the previ ous year's score. The values of the word-count variable are logged due to their distribution.

62. The assumption in this design is that factors that would make a State Department report entry longer or shorter due to a measurement artifact are unrelated to those factors that would affect the length of the same year's AI report, and vice versa.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 555

15000

A • Al, latin America mm. mm. — so, Latin America

8 8"

— Ai, Global SD, Global / v* /

* \ x

ooos /*-**

jf -«% ^ X,-"*

o -

^*•»

1970 1980 1990 2000 2010 UQEir ycaf

Figure 1. Mean Word Count, per Country, of AI and State Department Re port Entries

Estimates are derived using the following equations. For the PTS-a/' score:

(1) PTS_aij( = ß, PTS_ai.fMj + ß2 ln(AI_wordcounti() + ß^ PTS_sdi( + constant., 7

+ u. + e,

For the PTS-sd score:

(2) PTS_s<±( = ß( PTS_sd(ï( I) + ß2 ln(SD_wordcountf[)+ ß2 PTS_a/'.(

And, since CIRI coders use both primary texts, for CIRI-p/rys/'nf:

(3) CIRI_physint.( = ß, CIRI_physint.ft_v + ß2 ln(SD_wordcount(() + ß^ PTS_a/;t + constant., + u + e.

(4) CIRI_physint( = ß( CIR1 _physint/ff J; + ß, ln(AI_wordcount([) + ß3 PTS_sdf constant., + u. + e,

For all, u. refers to country-specific fixed effects and e.(represents the error term.

The results are reported in Table 2. For the Latin American countries, there is evidence of a small but statistically significant effect on coding levels relative to report length. On average, as length of a country's report entry grows, the associated PTS-a/, PTS-sd, or CIRI-phys/nf score is slightly worse, after roughly controlling for severity of violation. The estimates suggest that an increase in the length of an AI report entry from 1200 words to 1900 words (approximately one standard deviation) corresponds to a rise in PTS-a;

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

556 HUMAN RIGHTS QUARTERLY Vol. 35

Table 2: Effects of report length on PTS and CIRI coding, Latin America

0) (2) (3) (4) PTS_ai PTS_sd CIRLphysint CIRLphysint (1976-2006) (1981-2006) (1981-2006) (1981-2006)

PTS_ai,, **.3494 (.0249)

PTS_sdM **.3764 (.0504)

CIRI-physint,, **.3437 **.3193 (.0702) (.0556)

Al word count, **.0820 **.0835 logged (.0159) (.0182)

SD word count, **.0437 .0040 logged (.0121) (.0133) PTS_ai **.3499 **.3450

(.0458) (.0607)

PTS_sd **.3636 **.3123 (.0473) (.0408)

Constant **-.4165 **-.2780 **.1070 **-.4071

(.1130) (.1012) (.1214) (.1251)

observations: observations: observations: observations: 641 597 515 545

groups: 30 groups: 30 groups: 25 groups: 26

: = significant at p<.01; *=significant Results based on pooled cross-national time series regression with fixed effects and robust standard errors.

Note: CIRI coding is inverted for comparability with PTS.

of approximately .04 on average on the 0-1 scale; a rise in the State Depart ment report length of one standard deviation, from 6100 to 11,100 words, corresponds to an estimated .03 increase in PTS-sd. Our initial expectations were that CIRI-pbys/nf would not be as vulnerable to length because it is more event-based, but CIRI also requires coder judgment when frequency is not clear-cut.63 CIRI-pbys/nf is as responsive to Al's report length as PTS-a/': a rise in AI report length of 700 words corresponds to a CIR\-physint score also worse by .04 on average. The coefficient for the association between

63. See Cincranelli & Richards, supra note 22, at 18.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 557

C\R\-physint and the widely varying word count of State Department reports is not statistically significant.64

In summary, an effect on coding related to the amount of reporting on each country is apparent. An effect like this would result if inter-subjective coders respond in part to length and level of detail as an indication of in tensity of violations.

C. Changes in Types of Violations, and new Kinds of Violations

The third question is whether the coding approaches adequately deal with on-the-ground changes in types of violations. To explore this question, dif ficulties posed by changes in the nature of repression in Guatemala and Brazil are brought to bear below.

As mentioned above, CIRI-phys/nf adds scores for four different types of violations (political killings, disappearances, political imprisonment, and torture). Torture, or any of the other categories, can only be registered on its own three-point scale, which becomes an additive component of C\R\-physint. CIRI scores for the more lethal violations, disappearance, and extrajudicial execution range from 0-4 when added together. These two components register the worst possible scores from 1981-1986 for Guatemala, but at the same time CIRI-pbys/nf shows a slight improvement in 1982 and 1983, the most acute period of political killing and disappearances as shown by truth commission data. (Table 3 and Figure 5 display scores for Guatemala.)

The lethal nature of the repression (along with the State Department reports' political bias) may help explain why. In 1982, CIRI gives Guate mala an intermediate score for torture, and in 1983, assigns an intermediate score for political imprisonment, before returning to the worst scores for 1984-1986. CIRI-phys/nf may underrepresent the severity of human rights when egregious violations are located in fewer than all four categories. Later in time, information about somewhat less lethal violations produces some surprising equivalencies in coding. By the 1990s, a larger and more diverse set of human rights organizations operated inside Guatemala.65 A milestone was reached in 1990 when the Roman Catholic Church opened the Archbishopric of Guatemala's Office of Human Rights (ODHAG) in 1990, "after years of silence."66 In the intervening years, ODHAG has be come one of Guatemala's most prominent, professional, and internationally well-connected human rights organizations.67

Due to varying Congressional mandates, more thematic sections have been added to the reports over time, but recent SD country entries have shortened on average since 2000, as Figure 1 reveals. Ropp and Sikkink, supra note 47, at 188-9. Id. Id.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

558 HUMAN RIGHTS QUARTERLY Vol. 35

Table 3: Guatemala, PTS and CIRI Values with CIIDH Data, by Year

CIIDH PTS CIRI

CIRI, CIRI, CIIDH, Political Disappear

CIIDH, Identified Integrity ances and Full Victims PTS-AI PTS-SD Index Killing

Year Count* Only* (scale: 1-5) (scale: 1-5) (scale: 0-8) (scale: 0-4)

1970 301 214

1971 410 263 1972 355 266 1973 286 197 1974 139 79

1975 63 49

1976 173 110 4 4

1977 227 155 5 3

1978 203 27 5 3

1979 181 70 5 3 1980 2,349 804 5 4

1981 3,736 866 5 4 8 4

1982 17,953 2,265 5 4 7 4

1983 1,890 734 5 4 7 4

1984 869 420 5 4 8 4

1985 501 224 5 4 8 4

1986 296 176 4 4 8 4 1987 392 224 4 4 4 2 1988 352 247 4 4 7 3 1989 409 290 4 4 6 4

1990 598 345 5 5 7 4

1991 354 248 4 4 6 4

1992 326 179 4 4 6 3 1993 160 122 4 4 4 3 1994 250 135 4 5 6 4

1995 95 45 4 5 6 4

TOTAL 32,868 8,754

* Yearly CIIDH information comparable to truth commission data, from Ball, et. al., supra note 44, at 119. Note: Cl Ri coding is inverted. PTS: 5-worst; 1=best; C\R\-physical integrity. 0=best, 8=worst; CIRI-cfoap (disappearances) added to CIRI-kill (extrajudicial killings), 0=best, 4=worst.

More information was available, and both AI and State Department were reporting on the range of ongoing human rights problems in this period. Guatemala demonstrates that once attention is focused on a human rights situation, the attention is often sustained over time. NGOs may devote more staff time to that country, developing expertise and contacts that increase their knowledge of human rights conditions. The attention is welcome from a human rights standpoint, but this reputational dynamic could lead to measurement bias across countries or years.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 559

A second case, Brazil, illustrates further how changes in type of repres sion interact with changes in the reporting capabilities of NGOs. By way of background, the substantive view of human rights practices under Brazil's authoritarian military regime (1964-1985) is very different from that of post democratic transition Brazil. To simplify a complicated political story, Brazil experienced authoritarian rule in the 1970s, a transition towards democracy in the 1980s, called the abertura, or "opening," and democratic rule in the 1990s and later. Human rights violations were most severe during the pe riod when General Emilio Médici held power (1968-1974) and improved somewhat during the period of abertura.bS Brazil's complete transition to democracy dates from 1989, when the first elections were held to elect a president by direct popular vote since before the military coup.

Researchers generally agree that Brazil's human rights situation has improved in the democratic period, particularly since the mid-1990s.69 The PTS and CIRI scores, surprisingly, indicate that respect for human rights under the military government was better on average than in the democratic period. The full authoritarian period is not covered by the data sets, but Bra zil shows better average PTS and CIRI scores during the authoritarian and transition periods than for the fully democratic period (see Table 4). After the transition to democracy, serious human rights problems remained, but as described below, the report texts reflect violations of a different character than under the military regime.

Because CIRI compiles the physical integrity score by specific type of violation, one can see how different components of the index affect the scores. In the case of post-transition Brazil, the level of political killings and torture drive the poor CIRI-phys/nf scores (see Table 5). The democratic Brazilian governments have rarely practiced disappearances or held political prisoners, a problem during the earlier period, but the later scores on tor ture and extrajudicial killing reflect use of lethal force and torture by police against criminal suspects, detainees, prisoners and others. An authoritative 2007 Brazilian report listing all known cases of proven deaths and disap pearances before, during, and after the military government reveals that the worst period for deaths and disappearances was the period from 1971 to 1974, and that state killing and disappearing of political opponents was rare after 1979, with none after the 1985 transition to democracy.70

Carlos Santiago Nino, Radical Evil on Trial 33 (1996); Joan R. Dassin, The Brazilian Press and the Politics of Abertura, 26 J. Interamerican Stud. & World Aff. 385, 385-88 (1984). For example, the 2013 annual report by Human Rights Watch notes that Brazil is now "among the most influential democracies in regional and world affairs" but still faces "chronic" problems related to police treatment of criminal detainees. See Human Rights Watch, World Report 2013, Brazil, available at http://www.hrw.org/world-report/2013/ country-chapters/brazil. Secretaria Especial dos Direitos Humanos da PresidJncia da RepOblica, Direto A Memoria e A Verdade: Comissäo Especial sobre Mortos e Desaparecidos Pol(ticos (2007).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

560 HUMAN RIGHTS QUARTERLY Vol. 35

Table 4: Brazil, PTS and CIRI-Physint Averages, by Period

Brazil, PTS a CIRI-physfm averages, by period lAuthorttarlan and ltransitlon,197S-89 Democratic, 1990-2006

PTS-ai 3.43 4.12 PTS-sd 2.54 3.82 CIRI-p/ivsfnf 3.56 4.71

Note: CIRI coding is inverted. PTS: 5= worst; 1 =best; C\Ri-physint 0=best, 8=worst

The most plausible explanation for the appearance that respect for human rights was worse in the 1990s and 2000s than during the military regime is that human rights organizations have expanded their focus from a narrow concentration on direct government responsibility for death, disappearance, and imprisonment of political opponents to a wider range of rights. By look ing at the report texts this hypothesis can be further explored.

In its first entry on Brazil in the 1976 Annual Report, AI focused on gross human rights violations, especially political imprisonment, torture, and summary executions committed directly by state officials.71 The State Department's first section on Brazil in 1981 was short and largely positive, commending the government on improvements. Al's report on Brazil that year was less positive overall, but also mentioned "some improvement."72 The improvements noted in basic rights are consistent with the Brazilian report cited above,73 which shows no state-sponsored deaths and disap pearances in 1981.

By 1987, however, in the period of transition under the government of José Sarney, both AI and the State Department took a harsher tone. The AI report on 1986 mentions Al's investigations of rural killings and detainee mistreatment in prisons and police stations.74 This report reflects Al's move ment from a narrow focus on state-sponsored imprisonment, killing, and torture in Brazil to a focus on the state's failure to prevent, investigate, and prosecute rural killings, with added attention to police brutality and exces sive use of lethal force with criminal suspects. Further evidence of NGOs' expanded focus is evident in the 1987 State Department entry on Brazil, which reports that violent police treatment of criminals has "received con

Al reports mentioned in this and the following paragraph refer to the corresponding entries on Brazil in the yearly reports of Amnesty Int'l, supra note 10. Amnesty Int'l, Amnesty International Report 1981, at 118 (1981 ), available at http://www.am nesty.org/en/library/asset/POL10/001/1981/en/5e4bdcf6-0f75-4ec4-be19-ea619d7a229b/ POL100011981eng.pdf. Secretaria Especial dos Direitos Humanos da Presidència da Repüblica, supra note 70. Amnesty Int'l, Amnesty International Report 1986, 129 (1986).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 561

Table 5: Brazil, PTS and CIRI Values, by Year

Brazil, PTS and CIRI Values, by year PTS CIRI

Physical Political

Integrity Disappear Political Imprison Al SD Index ances Killings ment Torture

Year (scale: 1-5) (scale: 1-5) (scale: 0-8) (scale: 0-2) (scale: 0-2) (scale: 0-2) (scale: 0-2) 1976 3 2 1977 4 3

1978 4 (none) 1979 3 2

1980 3 2 1981 2 2 4 0 2 1 1

1982 3 2 4 0 2 0 2 1983 4 2 3 0 1 0 2 1984 4 2 4 0 2 0 2 1985 4 3 3 0 1 0 2 1986 3 3 3 0 1 0 2 1987 4 4 4 0 2 1 1

1988 3 3 3 0 1 0 2 1989 4 3 4 0 2 0 2 1990 5 4 5 1 2 0 2 1991 5 4 4 0 2 0 2 1992 5 4 5 1 2 0 2

1993 4 4 5 1 2 0 2 1994 4 4 4 0 2 0 2 1995 4 4 6 1 2 1 2 1996 4 4 4 0 2 0 2

1997 4 3 4 0 2 0 2 1998 3 3 6 1 2 1 2

1999 4 4 4 0 2 0 2 2000 4 3 6 0 2 2 2 2001 4 4 5 0 2 1 2

2002 4 4 4 0 2 0 2 2003 4 4 4 0 2 0 2 2004 4 4i 4 0 2 0 2 2005 4 4 4 0 2 0 2

2006 4 4 6 1 2 1 2

Note: CIRI coding is inverted. PTS: 5= worst; 1=best; CIRI physical integrity: 0=best, 8=worst; other CIRI scales: 0=best, 2=worst).

siderable public attention . . . due to the increasing assertiveness of local human rights organizations and the press, as well as growing public concern about criminal violence."75

From the point of view of human rights work, this is an encouraging development. For human rights measurement, the impression can be mislead ing if researchers only compare coding over time. A focus on police killings and torture prevails in reports on Brazil from the 1990s onward. This shift is also reflected in the work of Observatôrio das Violências Policiais (OVP,

75. Country Reports (1987), supra note 11 at 398.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

562 HUMAN RIGHTS QUARTERLY Vol. 35

Observatory of Police Violence).76 Based in Säo Paulo, Brazil's largest city, OVP gathers news reports and other comprehensive information on victims of police violence. In the 1970s and 1980s, it is likely that the Brazilian police were also very violent towards criminal suspects, but such violence was not documented or reported.77

To summarize, an increase in domestic human rights NGOs and increased attention to and scrutiny of State Department reporting led to the produc tion of more information about human rights abuses in Guatemala in the wake of the genocidal violence of the early 1980s. In Brazil, attentiveness to previously unreported abuses affected the level of coding after military rule concluded. This suggests that particularly after repressive periods, some systematic measurement error (rooted in both the coding schemes and con temporaneous texts) may be introduced that can impact accurate assessment of human rights processes.

D. Ceiling effects

The fourth question asks whether the PTS and CIRI scales are appropriately responsive when violations are severe. As a result of retrospective truth com mission investigations, more precise information on the nature, intensity, and scope of human rights violations is available today for the periods of intense violations in Argentina, Chile, El Salvador, Guatemala, and Peru. Figures 2-6 chart violations from these countries and compare them with CIRI and PTS data.76 On visual inspection, the truth commission numbers on the graphs are spiked, while the data sets' values are smoother and stay high over more than one time period. Data values cannot go higher in the most intense periods when violations go from bad to worse. Even after the most lethal violence has ended, the AI or State Department annual report entry may still qualify for the data sets' worst scores.

Observatörio das Violências Policiais-SP, available at http://www.ovp-sp.org. Police Brutality in Urban Brazil, Human Rights Watch (1 April 1997), available at http:// www. u nhcr.org/cgi-bi n/texis/vtx/refworld/rwmain ?page=printdoc&docid=3ae6a 7e74.

The truth commissions categorized information on states' lethal repression of individuals in different ways: either as deaths and disappearances combined (Chile, Guatemala, and Peru), just disappearances (Argentina), or just deaths (El Salvador). In Argentina, few people were killed directly before disappearance, so virtually all of the deaths documented during the period covered by the truth commission happened through disappearance. While the important factual distinctions between deaths and disappearances have been contested politically and legally, most people who study disappearances agree in their understanding that virtually all of the disappeared people in each of these countries were killed. For these reasons and for this limited purpose it is appropriate to consider a disappearance as equivalent to a death, in order to use all of the available truth com mission data here. Data in Figures 2-6 adapted and extended from Sikkink, supra note 35.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 563

4000

3500

3000

2500

2000

1500

1000

500

J f

Truth

Commission, deaths and

disappearances

CIRI disap pearance & extrajudicial execution

-w-CIRI Physical Integrity Index

Note: CIRI scaling inverted for comparison Higher values correspond to worse human rights practices Range: PTS-AI and PTS-SD, 1-5; CIRI disappearance & extrajudicial execution, 0-4; CiRI-physint, 0-8.

Figure 2. Chile: Truth Commission, PTS, and CIRI Data

Countries falling in the scales' upper levels of repression may differ significantly in their actual levels of violence.79 In other words, there is a ceiling effect the scales do not go high enough to accommodate variation in the upper levels. CIRI's reaches its worst score at "frequent" violations. PTS offers more depth on the intensity dimension, but also has difficulties when violations are extreme. An artifact of the ceiling effect is "stickiness," especially at the high ends of the scales,80 limiting the ability to register improvement or deterioration when violations are severe.

VI. CONCLUSION: THE IMPACT OF INFORMATION EFFECTS

We hope to stimulate and advance a more systematic and focused discussion of how and why the data may be subject to measurement error, whether and how the data may give rise to biased inferences, and how such problems might be addressed. General findings related to each hypothesis are sum

79. Hafner-Burton & Ron, supra note 4, at 381. 80. Id. at 381-82; Todd Landman & Edzia Carvalho, Measuring Human Rights 90 (2010); Wood

& Cibney, supra note 16, at 379.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

564 HUMAN RIGHTS QUARTERLY Vol. 35

s 5 I

4500

4000

3500

3000

2500

2000

1500

1000

500

0 -iR-fSHHlHH

Year

Truth Commission,

disappearances

CIRI Physical Integrity Index

—*~CIRI disap^ pearance & extrajudicial execution

•Note CIRI scaling inverted for comparison Higher values correspond to worse human rights practices Range PTS-AI and PTS-SD, 1-5; CIRI disappearance & extrajudicial execution, 0-4; CIRI-physint, 0-8.

Figure 3. Argentina: Truth Commission, PTS, and CIRI Data

4500

4000

3500

Jj 3000 § •5 2500 M

| 2000 <5

"Note CIRI scaling inverted for comparison Higher values correspond to worse human rights practices Range PTS-AI and PTS-SO. 1-5; CIRI disappearance & extrajudicial execution. 0-4. CIRI-physInt, 0-8.

Truth

Commission, deaths

~~#"~CiRi disap pearance a extrajudiaal execution

"***"" CIRI Physical Integrity Index

Figure 4. El Salvador: Truth Commission, PTS, and CIRI Data

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data

•"♦identified victims of death and

disappearance

AI-PTS

—#—CIRI disap pearance & extrajudicial execution

-*-CIRi Physical Integrity Index

Year

*Note: CIRI scaling inverted for comparison. Higher values correspond to worse human rights practices. Range. PTS-AI and PTS-SD. 1-5; CIRI disappearance & extrajudicial execution, 0-4; CIRI-physint, 0-8.

** Yearly data collected by CIIDH [see footnote 29. supra}.

Figure 5. Guatemala: Truth Commission, PTS, and CIRI Data

"Note: CIRI scaling inverted for comparison. Higher values correspond to worse human rights practices. Range: PTS-Ai and PTS-SD, 1-5; CIRI disappearance & extrajudiciai execution, 0-4; CIRI-physint. 0-8.

Figure 6. Peru: Truth Commission, PTS, and CIRI Data

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

566 HUMAN RIGHTS QUARTERLY Vol. 35

Table 6: Information Effects and Data Set Differences, hypotheses and findings

£ a.

nj Q. — OJ d aj

i: O

■£ "E DJO O o c

_p "3 Q

E _Q

o t/T <

tj E o

<-2 3 01 to

n3 >

fD QJ

E c 0) o "

o?co«o J? 2 1 s "5-S St.o^-5 3 £ y E 2 X -o ^S>a« .2 ^ " JS " .£ ,P 9 ™ E2 o^-< a op-p - « o S ? -Ex- ° !T. - ~ S a. $ X ai "a ^ "0 ™ -S 8 s G £ 2 9- o > B " a c-asc-s

5 ^ ™ 'o >2 3 u «£'s ® ? » •i f; o 001,1 8 B 1 s! g*2**i£ i-i^ | i! I°l-y ™e§b t= «s.i-s | <u ^ s.2! IB

»1§= §iS-g? 5E" 8.1 | eb| g-l Qc X ZZ U Q_ U CC 'ui .E 3! C1.2 "ui I— U U > .— d c

9 « cu 00 <" >. E ^ -E c Q_~0 r- • c Ti-m^^.2 ^ a; >- c E £ — u

"8 5f|l5>- s|dB^ i^Hl II ■a "S-soos -55 2--i ^ cn-n"^.— a; 4-^ C C ©-.</>> 9 'sSoSt^ X J| E -a -2 2 « Sk is ^ U-z C t <D <" >: d _Q > E SPo § cSo>o 5 ~ c- m E T3 .. « > §: V « " '5 8 .. ..• FIsSSJ -9 F

E -Q - < > . r t py a; n r x ■_ ^ </> t/i ^ 01 O Q- c • 13 O OJ |£

^Bx^sb15 s s >" >-s ooj c 2 £ -a B & 3 £ £ B 2 = s> ~ Bs» c 9- « o ™ 1 8" ^ « af O S oBSB B 2 2 c o O uEEtjEq. Uoi3fB 3f 2

(D V) -pi <yi a 2 2 •- 2 -s £ ai ■■=;£—_ tr O dte > c S o « .£ n » o f O „ ^ oc " E O z o — e I 2 B 8- o ra £ 9 8" 8 s- 9 B -5 IB -2 u B 5 B 8-6s c £ p s I IB "O ^ U O -t— w 1_) —; p PL v/ -• i_ ~ nj

C .^Q.B^< _ T3 ^ £ O CO _C

I *iY*§ ^'fSD5 X< eBo |g^f.Eg fi-i.ll oc 1|! Jijjl Qiii filxlf lf| 'ill llfllM X XX 0> -Q U U LO _E E CO </> h UD > 1- ^ u

8 <*>3 c ^2 g-2 ?2r mir, ^"n|/1BI -% 5K Ei -D _ y > z 5 ~ ^ 22 5™ ,E o"£ .S c u i^-d 2da)"° oc « —

9 g~°| B. 8 £-£ 5 EsS"uJx7 e2»c g-c x £ 8 U tn E ^ 15 -Q^ T3 44 ,E ni c — _Qcu §■ F n n % co-o^ fc <D o —

>^1^11^ s|8gjS jr -| 18. a S- e S-s:1 8. g 8-12; .S | IS«EzDEuS °f«oii 22 iS.Eu

^ "8 % ■o 2 2

5 rz -§ §

i'i (V. </> C

-a M- C o Ib

O c .© c

£ S o 2

S-T ,_ Qj

>..0

S E B i 8.| <si -5

-2 .©

•- .2 1*^ S E

IV.

a o 2 jg H3

©.

a © © t: «+

.E ** ti

£ '£ «+

CN .E -2 P 8

® c 0 o

^ Qj (V. — 2P „ 1= g- 8 JS 2 .5 B -i ^ 91 .2 -eg

° = > g .2 >. c,= j B = c"|"S »■§ a.? S •2 = £. 0 -f g 3 •£ .2 £ Z s ■$ -2 5 g g « s S E _ p •— ^ o p t! O *, "tzj 5C S

| IIS-^-l SES^gF |l-||-2 $>*£>§£ I Co 2 <« u <J>§:-i q^^q. 8 xlaig. <N =■§ 8 8 ^ B 8 £ B Information PTS CIRI component hypothesized found hypothesized found 1. Quality of Improvements since early Higher quality information Improvements since early Higher quality information information, reports (better coverage, likely over time; Guatemala reports (better coverage, likely over time; retrospective: how good less bias) mean that coding example: early political less bias), so coding may Guatemala example: early was information about may not be equivalent over bias (State Dep't) when not be equivalent over time political bias (State Dep't). past violations? time across countries; State compared with AI report or across countries; expect

Dep't bias expected to be coding. difficulties with early State mitigated by PTS's separate Dep't bias, coding of AI and State Dep't reports.

2. Quantity of Longer and more detailed Some artifact of report CIRI's count-related coding Some artifact of length for information and detail: reports not expected to length for AI and State Dep't may render it sensitive CIRI measures, but only does more information affect coding because measures. to provision of more responsive to AI report correlate with harsher of PTS's standards-based detail; expected to be length. coding? criteria. more vulnerable to false

precision.

3. Amount of Measure may not respond Brazil example: scores Coding subdivided by type Brazil example: scores information, new types to change in context. similar although change in of violation; may respond similar although change of violation: what types of violations reported. better to changes in form in types of violations happens when type of and intensity of violation. reported; additive CIRI violations changes? physint less responsive

to extreme repression of single type.

4. Degree of violation: Fewer problems with ceiling Truth Commission com- May not reflect change in Truth Commission how well does in upper levels than with comparison: PTS responds high abuse (ceiling effect). comparison: hard for coding respond to CIRI. better than CIRI to acute CIRI to register extreme extreme worsening or violations. Still hard to violations and later improvement? register extreme violations improvements: ceiling

and later improvements: effect and stickiness, ceiling effect and stickiness.

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

2013 Information Effects and Human Rights Data 567

marized in Table 6. Neither data set responds well to changes in types of violations over time. PTS is somewhat better than CIRI at registering changes in levels of violation in comparison with the independent truth commission data, but both PTS and CIRI are vulnerable to a ceiling effect. CIRI disag gregates measures of kinds of violations, which can be informative, but the structure of its physical integrity scale makes it difficult to register extreme violations of a single type and could be misleading for a researcher relying only on the physical integrity scale.

In conclusion, four additional points are worth noting. First, both indi ces register deterioration quickly when violations increase from previously infrequent levels. The restricted variation in the upper ranges of the indices also creates a ceiling effect, such that less-than-large-scale improvement or worsening will not be reflected when human rights violations are more intense, and they are "sticky," or remain high, when improvement may be occurring. The ceiling effect caused by restricted variation would in principle also prevent the scale from registering any worsening "above" the ceiling; however, human rights improvement is often more incremental than dete rioration. This is supported by the observation that the data set measures are responsive to a rapid deterioration away from a lower level of violations, as seen in the cases covered by truth commissions, but not as sensitive to reductions in severe physical integrity violations as measured by the truth commission counts.

Second, the texts of the reports probably contain more critical attention by observers once violations have been high and once NGO activity has increased in the country, contributing to the ceiling effect problem once the texts are translated to PTS and CIRI codes. On the coding side, addressing any added coder susceptibility to more length and detail would be desir able, but we would note that post-hoc adjudication of the data coding, such as non-independent arbitration procedures among coders, can introduce its own errors in validity.81 Efforts to prevent this kind of information effect prior to and during the coding procedure are more desirable than post-hoc adjustment.

Third, scholars should keep in mind that the data are based on con temporaneous documents and understand the changing characteristics of the source materials over time. Because of increased quality and quantity of information, the data may skew toward worse scores in later years. As Gary King, Robert O. Keohane, and Sidney Verba suggest, if the direction of bias is known, one can be more confident of inferences if they go against

81. Steven B. Rothman, Understanding Data Quality through Reliability: A Comparison of Data Reliability Assessment in Three International Relations Datasets, 9 Int'l Stud. Rev. 437, 443 (2007).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

568 HUMAN RIGHTS QUARTERLY Vol. 35

the grain of probable bias—even if the precise extent of bias is not known —and levels of confidence in inferences that are consistent with bias should

be adjusted.82 Because of the probable bias toward worse scores over time, it will be more difficult in terms of inference to use these indices to test for

the causes of the kind of change that is occurring when human rights are improving than when they are worsening. Some statisticians assisting local NGOs and investigation teams have devised ingenious ways to accurately assess evidence of violations in the field,83 but some information effects rooted in the historical sources we are discussing may not be correctable.

Finally, it appears that the measures are not well suited to the examina tion of extreme episodes of human rights violations like genocide and mass killing or the relative improvements that happen after such extreme episodes. In general, a selection of individual cases should be studied with great care in addition to assessments based on summary data, for analysts interested in the factors contributing to human rights change. The ceiling effect in the data series and the nature of changes in the source texts may obscure a range of kinds of improvement that is important to the human rights researcher. It would be unfortunate if information effects present in summary data unduly contributed to pessimism about human rights change. We believe there will be many other creative ways to test and further evaluate the hypotheses we have proposed here, and we encourage other researchers to do so.

82. Gary King, Robert O. Keohane & Sidney Verba, Designing Social Inquiry: Scientific Inference in Qualitative Research 188 (1994).

83. Making the Case: Investigating Large Scale Human Rights Violations Using Information Manage

ment Systems and Data Analysis (Ratrick Ball, Herbert F. Spirer & Louise Spirer eds., 2000).

This content downloaded from ��24.254.82.142 on Fri, 17 Dec 2021 23:43:36 UTC��

All use subject to https://about.jstor.org/terms

Contents

p. [539]
p. 540
p. 541
p. 542
p. 543
p. 544
p. 545
p. 546
p. 547
p. 548
p. 549
p. 550
p. 551
p. 552
p. 553
p. 554
p. 555
p. 556
p. 557
p. 558
p. 559
p. 560
p. 561
p. 562
p. 563
p. 564
p. 565
p. 566
p. 567
p. 568

Issue Table of Contents

Human Rights Quarterly, Vol. 35, No. 3 (August 2013) pp. i-iv, 539-816

Front Matter
Information Effects and Human Rights Data: Is the Good News About Increased Human Rights Information Bad News for Human Rights Measures? [pp. 539-568]
Measuring Acceptance of International Enforcement of Human Rights: The United States, Asia, and the International Criminal Court [pp. 569-597]
Sikh Cremations and the Re-Imagining of the Clash of Cultures [pp. 598-630]
Public Theologies of Human Rights and Citizenship: The Case of Turkey's Christians [pp. 631-650]
Ritual Slaughter and the Freedom of Religion: Some Reflections on a Stunning Matter [pp. 651-672]
The Influence of the Feminist Anti-Abortion NGOs as Norm Setters at the Level of the UN: Contesting UN Norms on Reproductive Autonomy, 1995–2005 [pp. 673-700]
Getting the Gunpowder Out of Their Heads: The Limits of Rights-Based DDR [pp. 701-719]
Expression of Justice or Political Trial? Discursive Battles in the Karadžić Case [pp. 720-752]
Interview

The Last Witness to the Drafting Process of the Universal Declaration of Human Rights: Interview with Stéphane Frédéric Hessel [pp. 753-768]

BOOK REVIEWS

Review: untitled [pp. 769-778]
Review: untitled [pp. 778-785]
Review: untitled [pp. 785-791]
Review: untitled [pp. 791-795]
Review: untitled [pp. 796-799]
Review: untitled [pp. 799-804]
Review: untitled [pp. 804-806]

Movie Reviews

Review: untitled [pp. 807-811]

Contributors [pp. 812-815]
Back Matter