State Capacity, Democracy, and the Violation of Personal Integrity Rights

Abstract

While a large literature explores the effect that regime type has on personal integrity rights violations, few studies have explored a state-centric approach to understanding these violations. I develop an argument that focuses on the leaders of the state and the incentives that they have to protect or violate rights. Moving beyond the democracy-autocracy debate, I claim that state leaders who are more secure, face fewer costs in producing their desired policies, and have more bargaining power vis-à-vis their domestic opponents are less likely to violate their citizen's personal integrity rights. Using a series of econometric models, I find support for many of the hypotheses derived from the argument. Based on the results of the models, I offer some potential policy implications.

Joseph K. Young received his PhD in Political Science from Florida State University and is an Assistant Professor at Southern Illinois University with appointments in the Department of Political Science and the Department of Criminology and Criminal Justice. His research interests primarily relate to political violence and how the interaction of states and dissidents generates certain violent outcomes. His work has appeared in the journals Civil Wars, International Studies Perspectives, International Interactions, and Political Research Quarterly.

I would like to thank Will Moore and Chris Reenock for their invaluable suggestions on previous drafts. I also would like to thank Richard Hiskes and the anonymous reviewers for providing detailed suggestions that made this project exponentially better. All errors are, of course, my own.

Notes

∗Significant at 5%

∗∗Significant at 1%.

1. Like Poe and Tate (1994), Poe, Tate, and Keith (1999) and others, I define human rights in terms of rights to “integrity of the person.” While this is a fairly minimalist definition of human rights, excluding social, cultural, or economical rights, there is sill a great deal of variation among states who respect these rights suggesting that respect for personal integrity rights is far from universal. A more restrictive definition would likely further constrain the number of states that actually respect all these other potential components of a more complicated conceptualization of human rights. Throughout the rest of the paper, when I discuss human rights I refer to the more specific subset of personal integrity rights.

2. The polity scale measures democracy from −10 to 10 with high values representing democratic institutions and low values autocratic ones. Intermediate values suggest mixed or inconsistent regimes. Haiti averaged a 7 on the polity scale for much of the 1990s while Pakistan and Colombia averaged an 8. All three had high levels of state repression according to both the Political Terror Scale (PTS) and the Cingranelli and Richard Physical Integrity Rights scale (CIRI).

3. I use the terms repression, violation of human rights, and violation of personal integrity rights interchangeably.

4. See Davenport (2007) for a discussion of the “domestic democratic peace proposition.”

5. I treat the terms leader and state actor as synonymous. I assume that a state actor is part of the leadership and not a bureaucrat.

6. Levi offers a more general approach and looks at all type of state behavior, I focus more on coercive potential.

7. Levi (1988) calls this the discount rate. I refer to it as “job insecurity” to avoid any confusion with the way that game theorists use the term “discount.”

8. See Young (2008) for a discussion of how this relates to the potential onset of civil war.

9. A third reason offered by Davenport (2006) not discussed here is when repression is a policy tool. For example, Stalin's collectivization attempts in Russia or Hitler's final solution are examples of repression of personal integrity rights as a policy tool.

10. This domain slightly changes depending upon which dependent variable is used. The Political Terror Scale is available from 1976, but the CIRI data begin in 1981. The countries used in the estimation sample are listed in Table 6.

11. Briefly, this measure is created using parametric survival models to estimate the hazard rate of losing office for a leader in a given-country year given a set of covariates.

12. A widely dispersed population, such as in Sub-Saharan Africa, makes extracting revenue extremely costly (Herbst 2000). Dense populations reduce transportation and monitoring costs. Unfortunately, finding reliable measures of population density over time is difficult.

13. Population is also used in other studies to indicate the opportunity that a state has to repress.

14. Other studies use a Civil War variable to proxy for the same concept of a violent behavioral challenge to the state. DISSACT is a better indicator as it identifies specific potential violent challenges to state authority. Civil wars, on the other hand, are when those challenges are already manifest. Whether I create a nominal variable or leave it as a count, the DISSACT variable is always strongly negative and highly significant.

15. I log the variable as I assume that changes in dissident activity at lower levels, from the second act to the third, for example, have a greater impact on the dependent variable than changes from say the 51st violent act to the 52nd. I also estimated models with the raw DISSACT variable and get very similar results.

16. I only display the estimates from the logit models in Table 3. The results are extremely similar for the OLS models.

17. I include the results from the OLS models along with all data and do-files necessary to replicate the results in an online appendix available at http://mypage.siu.edu/jkyoung/publications.html.

18. In the OLS specifications, I employ a panel-corrected standard errors (xtpcse command) model using STATA 10 and use a lagged dependent variable both to correct the serial correlation as well as to model past repressive behavior.

19. In the OLS models, INSECURE is positive and significant across the models. Using Goemans et al. (2009) data in creating the measure has similar effects across the logit models, and the INSECURE variable using this specification also fails to reach statistical significance in several models.

20. In the two models where the MOUNTAINS is not significant at the p < .05 level, it is significant at the p < .10 level.

21. I used the PTS A measure for this table. The results for the PTS S are nearly identical. The spirit of the CIRI results are similar but are more complicated as there are more categories to discuss.

22. All replication materials for these robustness checks are available on the author's Web site. For a detailed discussion of Loess and Binary Decomposition see Davenport and Armstrong (2004).

23. See Long and Freese (2006) for a discussion of the use of these information criteria.