The Belief in State Legitimacy and Homicide: A Cross-National Analysis

ABSTRACT

This study examines whether perceived state legitimacy contributes to the variation in the level of homicide across countries. I perform a statistical test of this relationship by creating a new cross-national time series data set of 86 countries. The analysis suggests a strong, negative, and robust relationship between widespread belief in state legitimacy and the homicide rate, controlling for the level of democracy and other relevant variables. The results provide evidence that subjective measures of state legitimacy are an important determinant of cross-national homicide rates, independent of the characteristics of the state.

KEYWORDS:

• State
• legitimacy
• homicide
• violence

Acknowledgments

I would like to thank Matthew Lange and the anonymous reviewers for their constructive feedback. I would also like to acknowledge the contribution of my research assistants (Megan Butryn, Kirsty Federico, and Alison Pratt) in assisting with the data collection.

Notes

1. This assumes that the nonpayment of “quasi-voluntary” taxes (i.e., taxes on income, profits, capital gains, and property) is not considered a crime, but rather a function of low state capacity.

2. The WHO estimation procedure draws on a number of variables, including the Gini index, the proportion of the male population age 15–30 years old, and lagged gross national income (WHO 2014c). Variants of the latter three variables are included in the present statistical analysis (the Gini coefficient, the proportion of the total male population ages 15 to 24, and gross domestic product per capita). Consequently, the coefficients of these three control variables will likely be biased (i.e., increasing the likelihood of Type 1 error) in the models using the WHO homicide data presented below.

3. Technically, only the 2000 and 2012 homicide data are directly comparable due to measurement (i.e., estimation) differences with the 2004 and 2008 data. However, given the modeling strategy outlined below, there is no reason to believe that any bias as a result of the inclusion of the 2004 and 2008 homicide data would make Type 1 error (i.e., statistical false positives) more likely. This was confirmed by running models limited to the 2000 and 2012 homicide data, the results of which (not shown) are substantively identical to the models using data from all four years (2000, 2004, 2008, and 2012) presented below.

4. Note that I have reversed the order of the original WVS/EVS response categories in order to make the interpretation of the results more intuitive. All other answer categories (Not applicable; No answer; Don’t know; Missing; and Not asked in survey) are excluded from the analysis.

5. Alternatively, one could measure confidence in specific state institutions and combine them to create an index of overall state legitimacy. The WVS and EVS measure confidence in various state institutions such as the police, the courts, and the civil service. Both a composite measure of state legitimacy that combines the country means of confidence in the police, confidence in the courts, and confidence in the civil service and a second composite measure that includes the same components in addition to confidence in government and confidence in parliament result in findings for state legitimacy (results not shown) substantively similar to the models using only the confidence in parliament indicator. These results indicate that the findings presented below are not a statistical artifact of choosing confidence in parliament as the primary measure of state legitimacy.

6. Using the “cluster” option in STATA.

7. Using the “ipolate” command in STATA.

8. The standard method of interpreting the effect of an independent variable in log-linear models is to exponentiate the coefficient of interest, subtract 1 and then multiply by 100 for the percent change in Y for each unit increase in X. Thus, the coefficient of state legitimacy in Model 2 (–0.798) is exponentiated (e^–°^.798 = 0.45), then I subtract 1 from this result and multiply by 100 for the percent change in the homicide rate for each unit increase in state legitimacy [(0.45 – 1) * 100] = –55 percent.

9. The lowest observed state legitimacy score in the dataset is 1.50 (Macedonia, 2001); the highest score is 3.77 (Vietnam, 2006).

10. That is, countries with a standard deviation greater than 0.8 are coded as 1, otherwise they are coded as 0. The results hold when using alternate thresholds and when measuring standard deviation as a continuous variable.

11. The size of the effect, expressed as a percentage difference between the high standard deviation category and the reference group, is calculated as follows: [(e°^.279 – 1) * 100] = 36 percent.

12. This result is achieved despite the potential suppression of the explanatory power of state legitimacy due to the increased likelihood of artificially inflated effect sizes of income inequality and the log of GDP per capita.

13. For every one-unit increase observed in state legitimacy in Model 5, the homicide rate is on average roughly 49 percent lower.

14. In Model 7, for every one-unit increase observed in state legitimacy, the homicide rate is on average roughly 44 percent lower.

15. An alternative model specification to test the causal order would be to use cross-lagged panel models. However, as a result of limited time series data availability for both state legitimacy and the homicide rate, too many cases are lost if the lagged value of the dependent variable is also included in the models.

16. Political culture, broadly defined, is the shared set of attitudes, beliefs, and norms that give order and meaning to politics, affecting orientations toward the polity (Aronoff 2001; Pye 1968).

Additional information

Notes on contributors

Andrew Dawson

Andrew Dawson is Associate Professor of Sociology at York University, Glendon Campus. His research primarily focuses on the causes of the rule of law, with a particular interest in violence. He is currently working on a comparative historical research project examining the relationship between political culture and violence.