This is our ninth blog post for our Job Market Paper Series blog for 2024-2025.
Sakina Shibuya is an applied economics Ph.D. candidate at University of Wisconsin-Madison. She is an applied microeconomist specializing in gender differences in developing economies. You can find her JMP here.
War and sex are inextricably linked, a connection that extends beyond the use of sexual violence as a weapon against enemy combatants. Sexual violence against civilians committed by soldiers is common worldwide. In fact, government soldiers are among the most frequently reported perpetrators of conflict-related sexual violence (Cohen and Nordås, 2014). While individuals of any sex, sexual orientation, and gender identity can be affected, women and girls are disproportionately fall victim to sexual violence in both conflict and post-conflict settings (Cohen et al., 2013).
How Does Soldier Presence Affect Women in Host Communities?
In my job market paper, co-authored with Felipe Parra, we address the question: What are the consequences of soldier presence for women in host communities? We explore this question in the context of Colombia where the armed forces have faced scrutiny over sexual violence against female civilians during Colombia’s prolonged conflict and subsequent peacetime since 2016 (Oquendo, 2020; Reuters, 2020; Turkewitz, 2020).
Broadly, we tackle three main areas. First, we overcome the difficulty of tracking soldier presence by constructing a new dataset of military base presence in Colombia from 1998 to 2016. Second, we quantify effects of soldier presence on sexual violence to assess whether individual cases of abuse by government soldiers reflect a systemic issue. Third, we provide causal estimates of effects of base presence on these outcomes using a difference-in-differences (DiD) estimator for staggered treatment timing in an event study framework.
Solving the Endogeneity Problem
Not surprisingly, the Colombian government didn’t randomly choose locations for military bases. Factors like conflict intensity, territorial control, economic importance, and conflict timing often influenced decisions on where and when to place bases. To address the endogeneity of base locations and timing, we leverage temporal and geographical variation from Colombia’s substantial military expansion between 2000 and 2016. During this period, the number of municipalities with at least one military base increased by 110%.
As municipalities received military bases at different times, we use the DiD estimator proposed by de Chaisemartin and D’Haultfoeuille (dCdH) (2024). This approach avoids the bad comparison between the switchers (municipalities that received a base) to the non-switchers (those that already had one), which can arise with traditional two-way fixed effects models when estimating effects of a staggered treatment (de Chaisemartin and D’Haultfoeuille, 2022; Goodman-Bacon, 2021). The dCdH estimator fits our context well, allowing for temporary base presence (i.e. the treatment is not all-absorbing), and accommodating non-random timing, unlike other modern DiD estimators. We further refine our comparison by analyzing only municipalities that eventually received military bases, omitting those that never did. This leaves us with 153 unique municipalities out of 1,104 municipalities that exist in our analysis period.
Finally, we aggregate outcomes at the two-year level, as sex crimes are relatively rare events. Aggregating by two-year periods helps detect changes. We consider a municipality treated in a two-year period if it had at least one military base at any point within those two years.
Military Base Presence Has Increased Sexual Violence
Our analysis shows that the presence of military bases increases the rate of sexual violence per 100,000 people, based on cases registered at the Colombian Office of the Attorney General. Specifically, sexual violence rates increase by about seven cases in years 2-3 and 4-5 after a base is introduced, with no statistically significant increase in years 0-1. This initial increase then subsides after years 6 and 7 for the remainder of the period of analysis. This initial rise then diminishes from years 6-7 onward. Overall, military base presence results in a cumulative 72% increase in registered sexual violence cases over 16 years compared to the control mean.
Our estimates are supported as causal, with pre-base introduction estimates statistically equivalent to zero. Our results show balanced trends before base arrival, supporting the satisfaction of the parallel trend assumption. Additionally, the lack of pre-introduction difference in sexual violence suggests there was no anticipation effect, aligning with our understanding of rapid base deployment. Soldiers were often deployed with minimal setup, such as tents, immediately after a base decision, leaving little time for community anticipation.
Increase Driven by Less-educated, Lower-paid Drafted Soldiers
Next, we explore which type of soldiers drive this effect. Two categories of military bases are relevant to our analysis The first type is predominantly staffed by drafted soldiers—Colombian men fulfilling mandatory military service who do not need a high school education. Drafted soldiers receive basic necessities and a small allowance, unlike professional soldiers who receive better pay and benefits. The second type is largely staffed with professional soldiers, who are volunteers with prior military experience and at least a high school education. These soldiers are well-compensated and receive a comprehensive benefits package.
Our analysis of heterogeneous effects by base type indicates that bases with more drafted soldiers, rather than highly trained, well-paid professional soldiers, are primarily responsible for the increase in sexual violence. Bases staffed with more drafted soldiers show an immediate rise in sexual violence in years 0–1 following base introduction, with effects starting to decline from years 4–5 onward.
Is It Because of the Presence of Government Soldiers or Something Else?
The rise in sexual violence may be due to changes in population or security conditions, both of which could shift with the presence of military bases. Additionally, this increase could stem from non-state armed actors rather than solely from army soldiers. To address these concerns, we conduct several mechanism checks. We find no evidence of population or security changes linked to military base presence. However, we lack the data to directly measure the presence of non-state armed actors, so we cannot fully test whether the increase in sexual violence is due to non-state rather than official soldiers.
Some army units have collaborated closely with illegal paramilitary groups to combat left-wing guerrilla forces, meaning that military base presence may coincide with paramilitary presence. Thus, we cannot entirely rule out the possibility that paramilitary soldiers contribute to the increase in sexual violence. However, anecdotal evidence suggests that government soldiers are typically the predominant military presence in treated municipalities, implying that non-state actors likely play a smaller role in the observed rise in sexual violence.
Policy Implications
Our results indicate that the establishment of military bases, a government security program, has had unintended negative consequences on communities that host them. Understanding the relationship between soldier presence and sexual violence is particularly important because recent studies have shown that female victims can face lasting economic consequences (Adams et al., 2024; Adams-Prassl et al., 2024; Sabia et al., 2013).
