Shifting police strategies: US aid and repression by public safety institutions in Latin America

Abstract

This article studies the international determinants of extrajudicial police killings in Latin America. Human rights violations by security forces are a problem that has plagued Latin American countries, despite broad democratization efforts in the region. Addressing the problem requires a better understanding of the processes that underlie it. Although existing literature has explored the determinants of police repression in Latin America through country-specific studies, there has been a lack of cross-national analyses. In this article we use original data on human rights violations by security forces in Latin America between 1998 and 2018 to argue that, beyond domestic factors, foreign aid provided by the United States in the War on Drugs era has influenced police repression both by encouraging heavy-handed policies and by driving a substitution dynamic between violations by different security forces.

Acknowledgments

We thank two anonymous reviewers, the JHR editorial team, Sam Bell, Chelsea Estancona, Patrick Homan, Wonyong Jung, and the participants at the 2023 LAPSS Annual Meeting at Universidad EAFIT (Medellín, Colombia) and at the Conference on Judicial Politics and Human Rights: Taiwan in Comparative Perspective (Columbia, SC) for feedback and research assistance. All remaining errors are our own.

Notes

1 Although, as noted by an anonymous reviewer, the Carter presidency (1977–1981) was an exception in developing an emphasis on human rights.

2 For example, in 2017 the United States withdrew its military aid and assistance from Myanmar in response to the 2016 violent killings of more than 10,000 Rohingya Muslims by security forces (The Guardian, 2017). In 1997, the United States cut aid to Colombian military units with human rights violations (CRS, 1998, p. 9) and, in 2017, the US Congress added specific conditions to release aid to Colombia calling for the dismissal of military officers involved in the falsos positivos extrajudicial killings scandal (CRS, 2021, p. 22).

3 Although much of the repression was simply shifted to paramilitary groups and other units that did not receive US military training (Kaplan, 2017; Tate, 2011).

4 The Ill-Treatment and Torture (ITT) Specific Allegation Data dataset is an exception to this, but as the name denotes, it is limited to only one specific physical integrity rights violation—torture (Conrad et al., 2014).

5 In this measure, discussed in more detail in Supplemental Online Appendix B, higher values represent more physical integrity rights violations, which is the opposite of the CIRI dataset (Cingranelli, Richards, & Clay, 2014). Alhough both datasets follow similar coding rules, they are not the same. The correlation coefficient between the police repression measure and the CIRI physical integrity rights index is .59 (inverting our measure to match the direction of the CIRI index).

6 All coding was done by a mix of graduate and undergraduate student coders. After a training phase, each country-year was coded by two independent coders, with any disparities in coding resolved by a third coder. See Supplemental Online Appendix B for the dataset’s codebook. We also aggregate the violation scores so that we have an index of annual police physical integrity rights violations that can range from 0 (no physical integrity rights violations by the police) to 8 (the police frequently violated all four physical integrity rights). Although this broader measure is not used in our main analysis, we use it in some robustness checks included in the Supplemental Online Appendix. Note that while the police and paramilitary measures are new to this piece, the military measure has previously been used by one of the authors (Martínez Machain, 2023).

7 Figure A1 in the Supplemental Online Appendix repeats the analysis using the index of police physical integrity rights abuses instead of killings as the dependent variable. It shows similar results but less support for the substitution dynamic, showing that, consistent with Stavro and Welch (2023), it is killings that appear to be driving the increased police repression.

8 We note that our analysis includes both economic and military aid. When disaggregating the forms of aid, our results appear to be driven by economic aid. As this is counterintuitive, we probed these findings further. Given that existing work focuses on US military training as the form of aid that is used to influence human rights practices in recipient countries and that military training is a key form of military aid provided by the US to Latin America (Martinez Machain, 2023), we replicate our analysis using military training in Figure A2 in the Supplemental Online Appendix, and find results consistent with our main findings. As an additional robustness check we also estimated our models distinguishing between DoD aid and aid from other agencies. Results remain similar.

9 Note that the difference in sample size between military and police killings is due to there being a missing value in military killings for the Dominican Republic in 1998.

10 As an additional robustness check, we controlled for levels of civilian control of the military using the measure from Kenwick (2020). Findings remain the same. Given that including this variable drops the N significantly drops (because the data is available until 2010), we do not include this variable in main analysis.

11 As a robustness check, we also we also ran a model (OLS) that treats the dependent variable as continuous. Results remained similar.

12 For details, see Table A1 in the Supplemental Online Appendix.

13 As a robustness check, we repeated the analysis with time fixed effects as well as splitting the sample before and after 9/11. Results remained similar.

14 We present marginal effects from a logit (instead of ologit) model for ease of interpretation and visualization, as they directly represent changes in the probability of our outcome. The logit and ologit models yield similar results, supporting the robustness of this choice.

15 As a robustness check, we repeated the analysis using paramilitary killings instead of police killings as the DV, to see if there was a similar dynamic occurring (Tate, 2011). We found no such effect. Given that in the past countries such as Colombia have faced scandals for using paramilitaries to engage in extrajudicial killings instead of the military, it makes sense that this would no longer be a viable strategy in Latin America.

16 As an additional robustness, we test Hypothesis 1b, and the substitution dynamic, using an alternate modeling strategy. Following DeMeritt and Conrad (2019), we create a variable that is coded 1 if military killings decrease and police killings increase, –1 if military killings increase and police killings decrease, and 0 otherwise. The results, presented in Table A3 of the Supplemental Online Appendix, show support for a substitution dynamic in which military killings are substituted for police killings when US aid increases.

Additional information

Notes on contributors

Paola Fajardo-Heyward

Paola Fajardo-Heyward, PhD, Binghamton University, is an associate professor of Political Science at Canisius University. Her research focuses on human rights, the political economy of development, the politics of the developing world, and Latin America with a special emphasis on Colombia.

Ghashia Kiyani

Ghashia Kiyani, PhD, Kansas State University, is a visiting Assistant Professor of Political Science at Western Illinois University. Her research focuses on human rights and civil-military relations. Her work has appeared in Journal of Conflict Resolution, Conflict Management and Peace Science, Journal of Human Rights, Human Rights Review, and Armed Forces and Society.

Carla Martínez Machain

Carla Martínez Machain, PhD, Rice University, is a professor of Political Science at the University at Buffalo, SUNY. Her research focuses on foreign policy analysis, with a focus on military policy and international conflict. She has a book titled Beyond the Wire: US Military Deployments and Host Country Public Opinion with Oxford University Press and her work has appeared in Journal of Politics, American Political Science Review, and International Studies Quarterly.