Region-Specific Structural Covariates of Homicide Rates in Latin America: State Legitimacy and Remittances

Abstract

The goal of this study was to examine region-specific structural covariates of homicide rates in Latin America and the Caribbean (LAC). LAC nations possess 8% of the global population but 33% of homicides, yet the region receives limited attention in studies of social structure and violence. Prior literature suggests two separate social forces particularly relevant to the region, state legitimacy and monetary remittances. Theory from multiple fields provides distinct pathways through which each may influence LAC violence rates, suggesting a negative legitimacy-homicide association but competing hypotheses about the remittances-homicide association. Our unit of analysis was the nation-year, and our sample included 16 LAC nations between 2000 and 2018. We obtained homicide data from the World Health Organization and measures of state legitimacy and remittances from the Latinobarómetro, World Bank, and Organization for Economic Co-operation and Development. Controlling for several traditional structural covariates of cross-national homicide rates, results from panel fixed effects models indicated that state legitimacy and remittances are, respectively, negatively and positively associated with LAC within-nation homicide rates over time.

Keywords:

• State legitimacy
• remittances
• Latin America and the Caribbean
• homicide
• violence

Acknowledgments

These findings were presented at the 2021 annual meeting of the American Society of Criminology, Chicago. We thank David McDowall, Meghan Rogers, Theodore Wilson, and Matthew Ingram for valuable feedback on earlier drafts, and Miguel Rondon for his assistance with data visualization.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 The definition of LAC is contested by area specialists. Some argue that non-Spanish- or non-Portuguese-speaking nations should be excluded due to linguistic, cultural, and political differences. While French-speaking, Haiti is sometimes excluded. English-speaking nations in the Caribbean—Belize, Guyana, Jamaica, and Trinidad and Tobago—are sometimes defined as the West Indies and not part of LAC due to their similarities in language, culture, politics, and overwhelmingly African diaspora (Colburn, 2002).

2 Sovereign nations of the Lesser Antilles are Antigua and Barbuda, Bahamas, Barbados, Dominica, Grenada, Saint Kitts and Nevis, Saint Lucia, and St. Vincent and the Grenadines. Other islands are not sovereign nations or are territories of other nations.

3 An initial remittances-only model contained 19 nations because it was not dependent on Latinobarómetro data. Due to missing data, however, the additional nations added only 24 nation-year observations. Results for that model were substantively the same as with 16 nations, so we present the latter for the remittances-only model to remain consistent with the other two models.

4 Our homicide estimates are the raw data from the WHO Mortality Database, not those from WHO’s Global Health Observatory, for which WHO imputes estimates (Kanis et al., 2017; Rogers & Pridemore, 2023).

5 We believe we are the first to analyze each of these three dimensions separately. We did so because of the immense variation in legitimacy in the LAC region between and within nations over time, and because each dimension could plausibly exhibit independent effects on homicide rates. These results, available from the authors, revealed no significant effects on homicide rates of these individual subtypes.

6 Gilley (2006) produced three state legitimacy scores: unweighted, weighted, and attitudinal. Unweighted scores treat each subtype equally. Gilley (2006, p. 510) argued, however, that views of justification should be overweighted because it “underwrites the laws and rules that govern so much of the rest of social and economic life.” For his legitimacy scale he weighted justification 50% and legality and consent 25% each. We agree with this assessment in principle, but also believe arguments about weighting, especially the exact 50–25–25 weights, require more evidence and thus we weighted each index equally. For our attitudinal scale, α = 0.75, but we did not use it because, as others noted (Gilley, 2006; Nivette & Eisner, 2013), it does not contain an item for views of consent to complete the legitimacy scale.

7 The other responses are “In some circumstances, an authoritarian government may be preferable,” “To people as one, we do not care about a democratic regime or an undemocratic one,” “Does not know,” and “Does not answer” (Latinobarómetro, 2022c).

8 This index does not include homicide and thus does not share information with our outcome variable.

9 During the review process, we also estimated effects using averages for three-year groupings instead of four. This increased thenumber of observations by 30%, and effect sizes and p-values for state legitimacy and remittances were nearly identical in all three models to those shown below in Table 3."

10 We transformed the data here to a 1–10 scale for conceptual ease of understanding. We used the original scale values for our analyses.

11 We conducted a series of post-estimation tests on the normality of residuals. In all three models, residuals were normally distributed. The HDI variance inflation factor was elevated, but results were substantively the same for all variables whether HDI was included or excluded, and thus we retained it in the model.

12 As an aside, our results show that poverty and income equality, structural factors commonly associated with national homicide rates, were non-significant. These results may tap into two themes. Substantively, this may be indicative of the so-called LAC paradox for income inequality and poverty (Bergman, 2018) that we mentioned earlier. Other scholars also found these two variables were not associated with homicide in the region (Chainey et al., 2021; Elias et al., 2022; Navas & Navas, 2022; Ponce et al., 2021; Rivera, 2016). Methodologically, we used panel fixed effects models, and many structural variables–including poverty and income inequality–are often unassociated with homicide rates within nations in longitudinal models.

13 Omitted variable bias could occur in remittances when migrants self-select based on unobservable characteristics like crop failure, sound economic policy, etc. (McKenzie & Sasin, 2007). Selection bias on remittances may result from differences in who is likely to migrate. Healthier, more educated, and wealthier households, for example, may be more likely to migrate (McKenzie & Sasin, 2007). In the LAC region, however, there is large variation between nations and types of migrants in terms of education and socioeconomic background (Fajnzylber & López, 2008).

Additional information

Funding

This research was supported by a Presidential Doctoral Fellowship for Research Training in Health Disparities, awarded to the first author, with funding from the National Institute on Minority Health and Health Disparities (#MD003373) through the Center for the Elimination of Minority Health Disparities at the University at Albany—SUNY.