Overview of Vigilantism

Before turning to existing theories that seek to explain the causes of vigilantism, it is important to first examine the phenomenon itself. Scholars have identified a wide range of vigilantism types, which vary based on who is involved (community groups vs criminal organizations), where the act takes place (community based vs online), the punishment of the target (non-violent vs violent), and the underlying motivation (violation of laws vs violation of norms).These differences have led to a diverse set of definitions and conceptualizations, which often depend on the specific context or setting under study (Moncada Reference Moncada2017, Reference Moncada2020).

At one end of the spectrum, some scholars have adopted broad definitions—such as Black’s (Reference Black1998) characterization of vigilantism as “any type of social control that involves handling grievances by unilateral aggression.” At the other end, others offer narrower definitions. For example, Denkers (Reference Denkers1985) proposed that vigilantism must meet several strict criteria: the act has to be spontaneous, not involve the police, target someone suspected of a crime, and the participant must either be the direct victim or a direct witness of the criminal act. More recently, the field has moved towards a middle ground, where the definition of vigilantism revolves around an extralegal act by private citizens in response to a (alleged) crime (Haas et al. Reference Haas, de Keijser and Bruinsma2012; Moncada Reference Moncada2017; Bateson Reference Bateson2021). Just as the definition of vigilantism varies, so too does the conceptualization and associated dimensions that are utilized. This variation can be based on the number of people involved, the degree of violence used, the location of the act, the level of organization involved, and the motivation and justification of the actors (Bateson Reference Bateson2021; Moncada Reference Moncada2017).

This article adopts Bateson’s (Reference Bateson2021) definition of vigilantism as the extralegal prevention, investigation, or punishment of offenses. The conceptualization of vigilantism will focus on collective groups that use the highest form of violence (murder) in a public setting that is motivated by an (alleged) criminal act. This conceptualization aligns with a subtype of vigilantism known as lynchings (Jung and Cohen Reference Jung and Cohen2020). While lynchings are commonly associated with racially motivated violence in the American South, this is not always the case. Lynchings are “lethal, extralegal group violence to punish offense to the community” (Jung and Cohen Reference Jung and Cohen2020). The use of lynchings aligns with the punishment component of Bateson’s definition while including key features that can be leveraged in exploring the role of the state in vigilantism.

To investigate the potential impact of the state’s inadequate provision of security on community engagement in acts of vigilantism, it is essential to begin by examining the current literature that studies the underlying causes of vigilantism and what leads citizens to engage in an illegal violent act. The first theory of interest centers around a path-dependency approach, which suggests that the occurrence and persistence of vigilantism in a society can be traced back to historical events and conditions that have shaped the socio-political landscape (Godoy Reference Godoy2002). According to this argument, certain historical factors, such as a civil war or violent conflicts, create a path or trajectory that influences the emergence and perpetuation of vigilantism (Esparza Reference Esparza2017). The path-dependency argument asserts that the enduring legacy of civil war, marked by pervasive violence, state repression, and the erosion of law and order, has played a significant role in normalizing vigilantism as a response to persistent security challenges (Bateson Reference Bateson2013). Scholarship centered on this approach argues that massive violence causes social trauma to individuals and communities, and the residue of state terror may outlive its perpetrators (Godoy Reference Godoy2002). Moreover, this theory asserts that historical experiences of violence led communities to resort to self-help measures and take the law into their own hands, perpetuating a cycle of vigilantism. These arguments highlight the influence of past events and conditions on the present-day prevalence and acceptance of vigilantism within a society.

An additional theory that has sought to understand vigilantism revolves around cultural norms or customary law, suggesting that vigilantism is deeply rooted in societal customs. Scholars such as Sieder (Reference Sieder2011b), García Fernández (Reference Fernández and Cristina2004), and Acuña (Reference Acuña2009) have explored this perspective. Cultural arguments are often associated with Indigenous communities, positing that their unique cultural or political identities may act as drivers for vigilantism (Morales Reference Morales and Adolfo2012; Handy Reference Handy2004; Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) 2011; McConkie Reference McConkie2023). For example, Handy (Reference Handy2004) argues that there is a demand for a revitalized system of community-controlled justice. Within this scholarly discourse, it is argued that vigilantism enables Indigenous communities to establish their own sense of normalcy in their lives when the government-led justice system has failed them (Girón Reference Girón2007).

Scholars have also analyzed factors such as state capacity, high-crime rates, and economic inequality. For instance, Baker (Reference Baker2004) argues that in Africa, failing state capacity and rising crime foster vigilantism. However, findings on crime rates and state capacity vary. For example, Phillips (Reference Phillips2017) finds that vigilantism doesn’t necessarily exist in municipalities with the weakest governments or the highest crime rates in Mexico. Similarly, in countries with weak state capacity, vigilantism may not emerge at all, even in high-crime areas (Ungar Reference Ungar2007; Godoy Reference Godoy2004). An alternative explanation for vigilantism involves economic inequality, where scholars suggest that in economically unequal communities, relative deprivation leads to vigilantism (Gurr Reference Gurr1970). For example, in urban neighborhoods with stark income disparities, feelings of insecurity or deprivation can motivate marginalized residents to organize vigilante groups (Phillips Reference Phillips2017).

An emerging literature has focused on citizens’ trust and perception of legitimacy or effectiveness of the state and policing (Weitzer and Tch Reference Weitzer and Tuch2005). When citizens perceive the state as incapable of maintaining order, they may begin questioning the legitimacy of its institutions and choose to bypass them altogether (Arias Reference Arias2006; Arias and Barnes Reference Arias and Barnes2016). Many of these studies utilize public opinion data or experiments to identify factors that lead citizens to either support or oppose the use of vigilantism. The findings suggest that when citizens perceive the state or police as ineffective, they are more likely to approve of the use of vigilantism (Haas et al. Reference Haas, de Keijser and Bruinsma2012). This perceived ineffectiveness can then lead citizens to question the legitimacy and the ability of the state to provide essential services such as security (Tanebi et al. Reference Tankebe, Reisig and Wang2016). Similarly, when citizens have lower levels of trust or legitimacy towards the police, they are more likely to support citizens taking matters into their own hands (Asif Reference Asif2023; Dow et al. Reference Dow, Levy, Romero and Fernando Tellez2024). When police are more effective and responsive, they are more likely to be viewed as legitimate and the preferred option for citizens (Haas et al. Reference Haas, de Keijser and Bruinsma2014). This literature has made an important contribution to our understanding of how citizens’ perception of the state can increase support for extralegal action from non-state actors. Yet, citizens approving or supporting vigilantism is different from citizens engaging in vigilantism due to limitations of the state.

Theory of State Absenteeism

In contrast to existing theories of vigilantism, this article introduces the theory of “state absenteeism” as a compelling and comprehensive framework for understanding the phenomenon. The theory of state absenteeism posits that when the state distributes public goods and essential services, such as security institutions and police stations, it can deliberately withhold or underprovide goods to specific areas or population segments. The core of the theory centers on the unequal distribution of state-provided goods, while remaining agnostic about the specific motivations behind why the state is withholding or underproviding goods to a certain area. By de-emphasizing the state’s intent, the theory of state absenteeism can be applied to a wider range of cases. It requires only that the state is withholding or underproviding goods from the areas most in need, while allocating resources to less needy regions. Because states have finite resources, they must select how to distribute these resources; certain areas will be prioritized over others and sometimes at the expense of others with higher need. Even in developing countries with limited resources, the state still allocates some amount of goods such as police stations, even if they are not able to provide equitable levels of goods to all citizens. This uneven allocation can be viewed as a type of selective state-building, in which the state fails to serve all citizens equally—not solely because of resource constraints, but because the resources that are available are directed away from high-need areas.

This lack of adequate state presence in certain regions or communities can create a vacuum in security provision, leading to an environment ripe for vigilantism to emerge. Specifically, the under-provision of goods (police) can decrease citizens’ perception of state or police legitimacy as well as the overall effectiveness of the police in providing security. This can result in higher levels of crime, which in turn further undermine legitimacy. When citizens perceive the police as ineffective and illegitimate, they become less willing to seek help from the state and more likely to mobilize within their communities—who share similar insecurities—to maintain order through acts of vigilantism.

Engaging in vigilantism is not without costs, as it requires citizens to overcome issues of collective action. When the state maintains an adequate police presence, it will be more effective and responsive to criminal activity. This will reduce the number of citizens who feel insecure and increase the perceived risk of punishment for illegal actions like vigilantism, thereby deterring such behavior. In contrast, when the state is absent or ineffective, the collective action problem becomes easier to overcome. The costs of vigilantism is lower due to the decreased likelihood of punishment, and a shared sense of insecurity within the community can facilitate mobilization. This dynamic suggests that even if a state cannot fully resolve all security concerns, it can prioritize areas with the greatest need to enhance police effectiveness and responsiveness. Doing so may reduce the appeal of vigilantism. When the state is absent and fails to address citizens’ security needs, however, people may resort to extrajudicial measures to protect themselves and their communities. In the absence of reliable law enforcement, vigilantism may be seen as a practical response to widespread insecurity and a means of restoring order and justice.

To study the theory of state absenteeism, I examine the level of state presence at the subnational level by analyzing the establishment of police. Because police presence is meant to promote the rule of law and project state power, citizens in areas with more security are expected to feel more secure. However, when the state neglects certain areas —engaging in state absenteeism—acts of vigilantism may emerge in response to the absence of formal security institutions. If the state creates both hot and cold spots of security through the uneven placement of security institutions, this deliberate or systematic neglect could lead neglected communities to take the law into their own hands. While the theory of state absenteeism can be applied to any country, it is particularly relevant for developing countries and post-conflict states, where institutional weakness and limited state capacity can make uneven provision of goods more likely (Bateson Reference Bateson2013). Developing countries and countries impacted by civil conflict face economic and political issues that can foster a cycle of distrust, leaving many communities disillusioned with the government’s ability to provide security and protect their rights. In such a context, state absenteeism exacerbates existing social fractures, making vigilantism an appealing alternative for enforcing norms in the absence of functional state institutions. Importantly, state absenteeism is not limited to regions with a significant Indigenous population as cultural arguments often suggest—but may occur in any country where the state’s presence is sporadic or inadequate. For example, socio-economic inequality and political marginalization can also intensify feelings of abandonment, prompting both Indigenous and non-Indigenous citizens to address security concerns through extrajudicial means.

One essential aspect of this framework is the examination of the presence or absence of police stations at the subnational level. Analyzing contemporary data on the distribution of police stations offers insights into how state resources are allocated and how this allocation correlates with incidents of vigilantism. Areas with limited police presence may experience heightened insecurity, prompting individuals or groups to resort to extrajudicial measures. Conversely, areas with substantial state presence and effective law enforcement may witness fewer incidents of vigilantism, underscoring the state’s critical role in shaping local responses to security threats. This leads to the following hypothesis:

State Absenteeism Hypothesis: The likelihood of vigilantism occurring in a region diminishes as the state’s presence strengthens. Specifically, an increase in the number of police stations per capita correlates with a reduced probability of vigilantism.

However, an argument could be made that the absence of the state and lack of adequate security provision are not the only factors that can influence citizens’ decisions to engage in vigilantism. Corruption within law enforcement could also shape public preferences. When corruption is widespread, and especially when it enables informal protection networks backed by criminal organizations (COs), state presence may no longer serve citizens’ security needs (Asif Reference Asif2023). In such cases, a “gray zone of criminality” may emerge—where police protect a CO by turning a blind eye to its activities or by selectively enforcing the law to harm rivals (Trejo and Ley Reference Trejo and Ley2020). Here, even with a physical state presence, the perceived illegitimacy or partiality of law enforcement may erode public trust.

Still, corruption alone is not sufficient to produce these “gray zones.” Following Snyder and Duran-Martinez (Reference Snyder and Duran-Martinez2009), the police must also possess the capacity to credibly enforce the law. This creates a paradoxical situation: the state must be present and capable enough to pose a credible threat of enforcement to COs, while also being sufficiently corrupt to collaborate with them. In these scenarios, even though the state generates insecurity, citizens may be deterred from engaging in vigilantism due to the heightened risks of arrest or violent retaliation by a protected CO. Thus, even when state presence fosters insecurity, the ability of citizens to overcome collective action barriers and engage in vigilantism may still be lower than in contexts where the state is altogether absent.

A significant strength of the state absenteeism theory is its adaptability to a variety of contexts. Rather than focusing narrowly on a single explanatory factor —such as ethnic composition or civil war legacies—this theory accommodates a broad range of socio-economic conditions, political environments, and levels of institutional capacity. This inclusive approach allows researchers to identify common patterns and unique dynamics that contribute to extrajudicial justice mechanisms in diverse settings. By acknowledging the multifaceted nature of vigilantism and extrajudicial violence, policymakers can design targeted interventions that address the specific root causes of vigilantism in each context, fostering sustainable peace, security, and democratic governance.

The theory of state absenteeism makes a meaningful contribution to the literature on vigilantism by centering the role of modern state-building and police presence. By focusing on the state’s uneven provision of security infrastructure, this framework highlights how institutional neglect—and not just weak capacity—can shape local decisions to resort to extrajudicial justice. This holistic approach complements existing theories and advances our understanding of how citizens respond to insecurity when state protection is unequal or absent.

State Absenteeism in Guatemala

Guatemala presents an ideal case study for evaluating the theory of state absenteeism, as it shares many of the characteristics that existing theories of vigilantism rely on. Prior research has linked vigilantism in Guatemala to its post-conflict legacy, high levels of inequality, and the central role of ethnic identity—particularly in Indigenous communities that mobilize to defend their rights and autonomy in the absence of reliable state protection (Sieder Reference Sieder2011b; Godoy Reference Godoy2002; Englehart Reference Englehart2009). These accounts emphasize both historical trauma and structural marginalization as key drivers of extrajudicial justice. This article builds on these insights by introducing an additional explanatory framework: state absenteeism, in which the state neglects or fails to provide essential public goods—most notably, security. Guatemala offers a valuable context to evaluate this theory alongside existing explanations, allowing for a more comprehensive understanding of the conditions under which communities turn to vigilantism.

Guatemala is uniquely positioned for evaluating multiple theories of vigilantism, including the new framework of state absenteeism. First, Guatemala has had to engage in state-building activities following the aftermath of a violent civil war that lasted from 1960 to 1996. This enables theories stemming from path-dependency/civil war to be tested as data has been readily available to capture both the civil war era as well as the rebuilding of the country following the civil war. Second, the rich ethnic and racial diversity in Guatemala, particularly the experiences of Indigenous communities, introduces a compelling dimension to explore cultural factors that, according to prior theories, could contribute to vigilantism. The historical marginalization of Indigenous populations, coupled with their disproportionate impact by violence and human rights abuses, creates a context where cultural elements could play a pivotal role in the emergence and perpetuation of vigilantism (Pallister Reference Pallister2013).

Furthermore, Guatemala presents a compelling case for examining vigilantism due to the abundant data available from diverse sources. This article leverages established data sources used in prior studies alongside an original dataset meticulously curated through petitions and direct inquiries to the Guatemalan government and the Policía Nacional Civil de Guatemala (PNC). While prior scholarship has leveraged data collected from sources such as the United Nations Verification Mission to Guatemala (MINUGUA) and the Historical Clarification Commission (CEH), focusing on vigilantism in the aftermath of the civil war, this article introduces a distinctive dimension. The original dataset, specifically collected for this study, not only covers historical variables but also provides detailed contemporary information on the exact location and date of establishment of police stations throughout the country, offering a comprehensive view of police station development. This project stands out by bridging the temporal spectrum, utilizing data spanning the civil war, the post-civil war period, and the reconstruction of the state’s security apparatus in the contemporary era. This unique timeline not only facilitates the testing of established theories of vigilantism but also enables an exploration of how modern-day state-building and security provision may influence the prevalence of vigilantism today.

Policing in Guatemala

The Guatemalan police also known as the Policía Nacional Civil (PNC) was established after the end of the Guatemalan Civil War. The PNC is controlled centrally by the state and falls under the orders of the President and Minister of Interior. While the PNC was initially founded in 1997, it took over two years for the PNC to have a presence in all 22 departments of Guatemala. As of 2019 the PNC had 27 main police stations, 125 police posts, 509 substations, seven mobile units, and one municipal police force in Guatemala City. While the number of police stations appears high for a small country, these police stations are spread across 340 municipalities and are responsible for the protection of over 17.2 million people across more than 109,000 km² of geographic area (World Development Indicators). Additionally, Guatemala struggles with providing an adequate number of police personnel. For example, the United Nations argues that an adequate police-citizen ratio is 222 police officers per 100,000 people (Alda Reference Alda2014). However, as of 2009, Guatemala had a ratio of 162 police officers per 100,000 people (Alda Reference Alda2014).

The small number of police officers coupled with the fact that Guatemala has one of the highest homicide rates in the world raises serious concerns for both citizens and policymakers (International Crisis Group 2012). This chronic insecurity is reflected in public opinion perspectives, where nearly half (49%) of Guatemalans report feeling unsafe in their own communities (LAPOP 2016). Moreover, when asked how long it would take for the PNC to respond to a robbery at their home, 46% of respondents said it would take more than an hour, while an additional 5% reported that police either would not come at all or were simply unavailable in their area (LAPOP 2018).

This insecurity coupled with issues of police corruption and discrimination has resulted in most citizens not trusting the PNC (LAPOP 2018). In response, the Guatemalan government has attempted to address these issues by implementing wide-scale reforms designed to improve police training, remove corrupt officers, and recruit more women and Indigenous citizens to the PNC, but the general view of the PNC remains unfavorable. The PNC’s struggle to control crime and violence has resulted in citizens’ criticisms about their ability to provide security throughout the country since security has been argued to be a fundamental “public good” in democratic societies (Carillo-Florez et al. Reference Carrillo-Flórez, Echebarría, Payne, Allamand Zavala, Jarquín Calderon, Freidenberg and Zovatto2007).

However, there are signs of institutional improvement. While Guatemala continues to grapple with one of the highest homicide rates in Latin America, the rate has declined since 2009. Notably, this decline coincides with a steady increase in the number of police stations across the country. As shown in Figure 1, the gradual expansion of police infrastructure over time reflects a growing, albeit uneven, state presence. These developments offer an opportunity to assess the relationship between police presence and the prevalence of vigilantism: as the state incrementally fills security gaps through institutional expansion, it may begin to reduce the appeal or necessity of extralegal justice in some communities.

Figure 1. Homicide Rates and New Police Stations Over Time.

Research Design

To test the state absenteeism hypothesis, I constructed an original dataset and combined it with existing data sources. I collected the original data by submitting multiple petitions and in-person requests to the Guatemalan government and the Policía Nacional Civil de Guatemala (PNC). Once the original data was collected, I was then able to identify each existing and newly established police station in the post-civil war era which allowed for the identification of subnational variation in the number of annual acts of vigilantism.Footnote ² The original data was supplemented with existing data sources, such as data from the Historical Clarification Commission (CEH), which provides information on the number of civil war massacres (CEH 1999). From these data sources, a municipal-level dataset was constructed containing information on whether an act of vigilantism occurred in a municipality each year, as well as the number of acts that occurred each year. The unit of analysis for this study is municipality-year. Guatemala is divided into 340 municipalities, which are grouped into 22 departments.Footnote ³ Data for this project is taken from 325 municipalities as the spatial domain to conduct a cross-sectional analysis. The difference between the total number of municipalities and those included in this analysis is attributed to redistricting over the years. To focus on the effect of the civil war, the analysis was restricted to municipalities that existed during this time period. The temporal domain for this analysis extended from 2004 to 2017 due to limitations in data availability.Footnote ⁴

Dependent Variable

This study conceptualizes vigilantism primarily through the lens of lynchings, which are fatal acts of extrajudicial violence carried out publicly by a group of citizens (Jung and Cohen Reference Jung and Cohen2020). Lynchings represent a particularly extreme and visible form of vigilantism and thus provide a concrete and measurable indicator of state failure to provide security. To test the state absenteeism hypothesis, two dependent variables are employed: Vigilantism and Count Vigilantism. The Vigilantism variable indicates whether a municipality experienced at least one lynching in a particular year, with municipalities where at least one lynching occurred coded as 1 and those where no act occurred coded as 0. Approximately 5% of the municipality-years in the sample had at least one lynching. The Count Vigilantism variable represents the exact number of lynchings that occurred within a municipality per year, ranging from 0 to 6. Utilizing these two variables, factors contributing to municipalities engaging in vigilantism and those increasing the frequency of these acts will be identified. The data for this project are sourced from an original dataset collected during fieldwork efforts from 2017 to 2019, supplemented by data from the Ombudsman (Procuradoria de Derechos Humanos, PDH), the National Civil Police (Policía Nacional Civil, PNC), and specific reports from civil society organizations such as the Centro de Acción Legal en Derechos Humanos, CALDH.

Independent Variables

Three primary independent variables will be used in this analysis: Police Stations, Civil War Massacres, and Indigenous Municipality. These variables will allow the theory of state absenteeism to be compared against the existing path-dependency and cultural arguments which have been heavily focused on in the Guatemalan context. The state absenteeism hypothesis will require the identification of the level of state absence or presence from the central government at the subnational level in the post-civil war period. To operationalize the concept of state absenteeism, I will utilize a measure that identifies the number of police stations per capita. This approach will count the total number of police stations that exist in a municipality as of the prior year and divide by the population which is then multiplied by 10,000. By utilizing this approach, I can examine the depth of the state’s presence by leveraging how many police stations are present in a municipality adjusted for the number of citizens the stations serve. The number of police stations in each area can be viewed as a proxy for the state’s commitment to a given area, as police stations provide security to the population and can ensure compliance with established laws. If the state absenteeism theory holds true, an increase in the number of police stations per capita in a municipality should correspond with a decrease in acts of vigilantism. This would imply that acts of vigilantism are less likely to occur when the state is present and providing security within the municipality.

To apply the path-dependency theory, it is essential to identify government-initiated massacres at the subnational level that occurred during the Guatemalan Civil War. According to the CEH, a massacre is defined as ‘an indiscriminate attack involving the execution of five or more people in the same place…whose victims were in an indefensible state’ (Mezquita Reference Mezquita, Ball, Spirer and Spirer2000, 208). I employ two approaches to operationalize this concept: a count measure and a binary measure. The count measure tallies the total number of government-initiated massacres occurring in each municipality throughout the civil war. Meanwhile, the binary measure serves as a robustness check and can be seen in the appendix. If the path-dependency theory holds true, the occurrence of a massacre in a municipality during the civil war should correlate with increased vigilantism in the modern day. This suggests that areas witnessing extreme violence decades ago are more likely to experience acts of vigilantism today.

To test the cultural argument, it is necessary to determine the proportion of Indigenous people living in a particular municipality. The data from the Guatemalan national census (Instituto Nacional de Estadística 2002) provides the necessary information. By calculating the percentage of Indigenous citizens in a given area, it can be determined whether regions with a majority Indigenous population are more likely to engage in vigilantism, as posited in the cultural argument’s literature. A binary measure is employed to classify municipalities as either majority Indigenous (1) or minority Indigenous (0). As a robustness check an additional measure is included in the appendix that identifies the percentage of Indigenous population in a municipality as well. These variables are then linked to the cultural argument, which suggests that the practice of vigilantism may stem from historical Mayan Indigenous culture. If this argument holds true, it will imply that areas with a high concentration of Indigenous people are more inclined to engage in vigilantism.

Control Variables

The literature on vigilantism has identified several additional theories that may influence the probability that a community will engage in such acts. The control variables that are included in this analysis are meant to identify municipal-level factors that should influence the likelihood that a municipality will experience an act of vigilantism in a given year. Specifically, it is crucial to account for the social environment, as well as the various demographics of the included municipalities in this study. This section will offer concise descriptions of each additional variable slated for inclusion in the statistical analysis and the rationale behind their incorporation. The first variable, homicide, serves as a proxy for crime rate, as heightened crime or homicide levels could directly impact the probability that a municipality will engage in acts of vigilantism. A measure of the homicide rate within the municipality is included and is calculated with the following equation: (Number of homicides/population) × 100,000. Homicide data is sourced from the Policía Nacional Civil (PNC), while population data is drawn from the Guatemalan National Census (Instituto Nacional de Estadística 2002). Municipalities with elevated homicide rates signal a need for increased security measures, suggesting a heightened risk of vigilantism.

Subsequently, a population control measure is introduced, sourced from the Guatemalan National Census Data (Instituto Nacional de Estadística 2002). This control measure is essential because municipalities with larger populations may exhibit heightened probabilities of vigilantism due to the increased potential for conflicts arising from more considerable size differences among inhabitants. As population increases, crime rates and interpersonal interactions also tend to rise. This increased activity can strain governmental resources, making it challenging to provide adequate security measures. Moreover, a control variable accounting for the male population percentage is included. Males are statistically more predisposed to engaging in violent behaviors and criminal activities compared to females. Consequently, communities exhibiting higher male proportions are candidates for crime and violent behaviors. The next control variable for this study will be the number of vigilante acts in a municipality in the previous year. The inclusion of a past act of vigilantism is important, as there may be theoretical reasons to expect that municipalities that engage in acts of vigilantism will continue these practices. Additionally, a measure of municipality size is included. This measure accounts for the geographic land area of each municipality and is measured in square kilometers (KMs). Larger municipalities may have more remote areas that could make vigilantism more attractive due to less proximity to means of resolving conflict via the legal system.

Furthermore, I leverage census data to craft an indicator of economic development within each municipality. This metric is derived from the proportion of households within each locality equipped with electricity. It is plausible that municipalities characterized by lower economic development are more prone to vigilantism, given the heightened likelihood of individuals resorting to criminal activities out of economic necessity. To bolster the robustness of my findings, two supplementary measures of economic development—access to running water and a sewage system—are also included in the article’s appendix. Moreover, a gauge of income inequality is incorporated, drawing from insights in the literature on marginalization and relative deprivation theories, which posits that regions marked by pronounced inequality are susceptible to heightened levels of vigilantism (Phillips Reference Phillips2017).

The final set of control variables is derived from a distance matrix initially compiled by Sullivan (Reference Sullivan2012). The first distance variable gauges the distance from each municipality to the national capital. This measure aids in mitigating potential issues related to state capacity. In countries with weaker states, like Guatemala, the government may struggle to enforce the rule of law in distant communities. Consequently, municipalities farther from the capital might resort to vigilantism due to the state’s perceived weakness. Similarly, rural communities may encounter similar challenges. The second distance variable assesses the distance from each municipality to the department capital, serving to differentiate between rural and urban areas. As one moves away from urban centers, rural communities may become more prone to vigilantism as the state’s capacity diminishes in these remote regions. With increasing distance from the department capital, state capacity declines, potentially prompting communities to fill the void left by the state through vigilantism.

The anticipated direction of the relationship for each of the included control variables is outlined in Table 1 below, while descriptive statistics for these variables can be found in the appendix. As there is potential for multicollinearity, Figure 2 presents a correlation matrix which has been added in order to identify if any of the key variables are highly correlated. None of the primary independent variables shows strong correlations with one another and the majority of the control variables appear in line with expectations. The variables that do exhibit higher correlation suggest that a Variance Inflation test may be warranted during the model estimation process.

Table 1. Variables and Expected Relationships

Figure 2. Correlation Matrix.

Empirical Strategy and Results

To empirically evaluate the proposed state absenteeism hypothesis, a combination of logistic regression models and negative binomial regression models will be employed. The first dependent variable being binary necessitates logistic regression models, enabling the testing of the state absenteeism theory. The second dependent variable, being a count variable, requires a count model such as the negative binomial model. Furthermore, a variety of robustness checks have been incorporated, detailed in the appendix for thoroughness.Footnote ⁵

Table 2 presents the results of the logistic regression models and the negative binomial models. This table first presents the logistic regression model specifications, with and without the state absenteeism variable, aiming to assess the existing theoretical arguments as well as the impact of including the theory of state absenteeism on vigilantism. Additionally, the state absenteeism model (model 2) has department-fixed effects, time-fixed effects, and two-way fixed effects added. Last, a negative binomial model is estimated utilizing two-way fixed effects. Including department-fixed effects controls for unobserved attributes that do not vary over time (i.e., geographic terrain). The inclusion of the time-fixed effects controls for any unobserved attributions that are constant across all the municipalities but vary over time (i.e., a new president). Finally, the inclusion of two-way fixed effects allows for the addition of department and time-fixed effects, which account for both department-specific (time-invariant) and time-specific (unit-invariant) factors. These approaches help ensure that the observed relationships are not driven by endogeneity due to potentially relevant omitted variables.Footnote ⁶ Table 3 presents the substantive effects of the state absenteeism logistic regression by examining the change in predicted probability for each of the key independent variables.

Table 2. Logistic Regression and Negative Binomial Models

Note: *p<.1, **p<.05, ***p<.01

Table 3. Substantive Effects for Logistic Regression Model

The findings regarding the state absenteeism hypothesis, which explores the link between vigilantism and the presence of the state via the number of police stations per capita, are robust. Across all models, the police station variable exhibits a consistently negative and statistically significant association. This suggests that the presence of more police stations per capita in a municipality does indeed correlate with the occurrence of vigilantism and the frequency of vigilante acts. Specifically, as the number of police stations per capita increases, the likelihood of vigilantism occurring in a municipality decreases. This indicates that in areas where the state is absent, communities are more likely to take the law into their own hands due to the lack of security provided by the state.

The substantive effects of the logistic regression model can be interpreted by calculating the change in predicted probability. To calculate this change, the variable of interest is first set at 0 and then moved to a value of 1 to estimate the change in predicted probability, while holding all others at their mean besides the dichotomous variables that are set as 0.Footnote ⁷ The decision to use the value 1 is based on the distribution of the key independent variables and to allow for an easier comparison of the effect of each variable. Moving from the 0 to 1 police station per capita results in a 4.6% decrease in the predicted probability of vigilantism occurring. While this might seem like a small effect, it is important to note that the baseline probability is 5%. This would suggest that establishing additional police stations and overcoming state absenteeism could dramatically reduce the probability that citizens would engage in vigilantism.

Next, I investigate the path-dependency argument using the civil war massacres variable. Although the massacre variable shows a positive association, it fails to achieve statistical significance in any of the logistic regression models or the negative binomial model. This indicates that the path-dependency argument may not be supported, suggesting that communities affected by violence, such as massacres during the civil war, are not necessarily more inclined to engage in vigilantism in the present day.

The positive and statistically significant relationship between the occurrence of vigilantism and whether a municipality is majority Indigenous is consistent across all models. This suggests that the ethnic demographic of a municipality is a key factor that influences the probability that vigilantism will occur. The substantive effect can be discussed by examining the change in predicted probability. Moving from a non-indigenous municipality to a majority indigenous municipality results in a 3.6% increase in the predicted probability of an act of vigilantism occurring. This is a 60% increase compared to the baseline probability. While this finding supports the cultural argument, it is essential to note that this relationship may not solely be attributed to culture and warrants further attention.

I will now briefly discuss the substantive effects of the negative binomial regression model for each of the three independent variables using an incident rate ratio approach. For each additional police station per capita that is added to a municipality, the number of vigilante acts will decrease by 58% on average when holding all other variables constant. For each additional civil war massacre that took place in a municipality during the civil war, the number of vigilante acts will increase by 3%. Similarly, municipalities that have a majority Indigenous population will have on average 133% more cases of vigilantism.

Several of the included control variables reached statistical significance across multiple model specifications. Distance to Capital was positive and statistically significant in each of the models. This suggests that as one moves further away from the capital, the probability of vigilantism occurring increases. This finding suggests that the state may be less able to project its power and maintain order, leaving citizens to take matters of security into their own hands. The population of the municipality is also positive and statistically significant. Larger numbers of people lead to more potential points of conflict and can lead to a higher probability of vigilantism occurring. Last is the level of crime rate in a municipality. Again, this is positive and statistically significant, which suggests that areas where citizens are more threatened with insecurity are more likely to engage in vigilantism. Interestingly, there was no correlation between past and current vigilantism. This may suggest that vigilantism is not an ingrained behavior repeated over time but rather a response to current conditions.

To understand the impact of the police station per capita variable on the overall model, a series of additional tests was conducted. These tests can be found in the appendix. The logistic regression models (model 1 and 2) were compared with and without the police station per capita variable. The limited model was compared to the state absenteeism model to determine whether the inclusion of the state absenteeism variable improved the model’s performance. To quantify the improvement of the model, the Area Under the ROC Curve was compared. The inclusion of the police station per capita variable improved the Area Under the ROC Curve by 2%, going from 0.74 to 0.76. Next, the AIC for the limited and state absenteeism model was compared. The AIC improved from 1291 to 1277 when the police station per capita variable was added. Additionally, a Likelihood-Ratio Test (Chi-Square) was conducted to ensure that the addition of the police station per capita variable resulted in a statistically significant model. The result was indeed statistically significant at the 95% confidence level. Finally, a Variance Inflation test was conducted to ensure that the model does not suffer from multicollinearity. The full results can be seen in the appendix, but no variable had a VIF higher than 4, suggesting that the model is not influenced by multicollinearity.

These tests confirm that incorporating the theory of state absenteeism enhances existing empirical models on vigilantism. This result, coupled with the substantial effects of state absenteeism having a larger magnitude than those suggested by the cultural theory, indicates that future research on vigilantism should consider the theory of state absenteeism as a critical factor.

Robustness Checks

To examine the robustness of these previously discussed results, several additional models with various specifications can be found in the appendix. First, each of the primary independent variables and several key control variables are replaced with different measures to ensure that the results are not reliant on any given indicator. This is accomplished by changing one specific measure at a time to identify if any specific variable alters the findings. For the cultural argument, a measure that captures the percent of a municipality that is Indigenous is used in place of the binary measure that captures if a municipality is majority Indigenous or not. This can account for the potential difference between a municipality that is 51% Indigenous compared to 90% Indigenous Next, a binary measure that identifies if a municipality experienced any civil war massacres is used in place of the count measure used in the path dependency argument.

Turning to the control variables, the economic indicator originally represented by the percentage of the municipality with electricity is replaced with the percentage of the municipality that has running water. This new indicator is used because fewer citizens have access to running water compared to electricity (an average of 72% vs. 78%). Additionally, a new measure is introduced to capture the percentage of the municipality with a sewage system, which averages 25% nationwide. The crime rate control, initially represented by the homicide rate, is replaced with the total number of homicides in a municipality each year. This change is important because the homicide rate is influenced by the population size. For instance, a municipality with ten murders and 10,000 people have the same murder rate as one with one murder and 1,000 people. Using the total number of homicides ensures that the measure of crime is not skewed by population size. Lastly, the Urban-Rural measure is replaced with a measure of population density. The rationale is like the Urban-Rural measure: higher population density indicates a more urbanized area. This change addresses potential limitations with the distance to the department capital measure, where a municipality could be far from its department capital but still contain a large city.

Next, an additional measure is added that captures the number of languages spoken in the municipality. The inclusion of this additional control can account for potential differences between cultures that can lead to heightened tension that could cause a rise in vigilantism. However, existing literature also mentions that increasing ethnic diversity in an area may lead to collective action problems, which could potentially reduce vigilantism (Mendoza Reference Mendoza2007). The inclusion of this variable addresses both potential possibilities. Next, Guatemala City, which has 67 police stations, was removed from the sample to ensure the results were not skewed by this potential outlier. Finally, a rare events logistic regression was estimated to ensure the results from the main logistic regression were not a result of small sample bias related to the lower percentage of vigilante acts. The results of this sensitivity analysis remain consistent across all specifications, providing robust evidence for the state absenteeism theory.

Additional Evidence and Potential Mechanisms

While empirical evidence supports the state absenteeism hypothesis in Guatemala, there are several questions that remain due to limitations with observational data. One example of this is the data on police stations and their locations does not directly capture the intention of the state’s decision to withhold or underprovide critical public goods. So, while the empirical findings cannot directly address the state’s intentions, the data can provide insight regarding if the communities that are most at need for security are the ones who are being targeted for additional provision of public goods. As the creation of new police stations increases the government’s ability to maintain order and control, “a first order condition for choosing where to allocate security (police) resources should (arguably) be those areas that violence and disorder is most likely to occur” (Dow Reference Dow2022). Dow calls this the ideal type of police institution, as the priority is the protection of citizens, the reduction of violence, and areas that have the highest security needs will be more prioritized by the government. However, the state can also use the provision of security for other means, such as rewarding political supporters or monitoring political opposition, which would shift some of the distribution of goods from those with the highest need to those with less need, resulting in the deliberate under-provision of security to certain parts of the state. So while the empirical findings for the state absenteeism hypothesis do not directly show the decision-making process or intentions of the state, the original data that has been collected can be leveraged to see if the state is prioritizing areas with the most need or if they are deliberately withholding or underproviding policing to those with the highest need.

To provide support for the assumption of state absenteeism, I explore the variation of the distribution of police stations within Guatemala by identifying where and when a new police station was opened. Currently there are 42 municipalities out of 340 that do not have a police station and between 2004–2017 there were 18 municipalities that had their first police station established. During this same time period the Guatemalan government created a total of 120 new stations across 48 municipalities. This shows that the state has the resources to provide at least some additional security, even if they aren’t able to meet the needs of all of their citizens. One of the key arguments of the theory of state absenteeism is that the state is selectively choosing where to allocate its limited resources and prioritizes certain areas over others. While observational data such as this cannot fully capture the state’s decision-making process, it can show trends that can support the assumption that the state is prioritizing certain areas but not others. Out of the 55 municipalities that received a new police station, 17 municipalities that had below average crime received 20 new stations, 15 that had average levels of crime received 25 stations, and 23 that had high levels of crime received 41 stations.Footnote ⁸ This excludes Guatemala City which had 34 new stations built. So, while half of the municipalities that received a new station were affected by high crime, over a quarter of the stations that were created went to low-crime areas, and the remaining quarter to areas with average levels of crime. It is worth noting that there were 147 municipalities that were considered high crime in at least one of the years in the sample and 83 that were high crime for at least half of the sample. Even focusing only on the persistently high-crime municipalities, the state had the ability and resources to build a new police station in every high-crime municipality, but less than one-third actually received one. While this is an oversimplification of a complex issue, this insight supports the assumption of state absenteeism, that the state can deliberately withhold or underprovide critical public goods to specific areas while prioritizing other areas.

To provide further evidence of the state absenteeism hypothesis, I explore whether municipalities with no police stations experienced vigilantism at a higher rate than municipalities that had state presence. Municipalities where the state was completely absent experienced vigilantism in 6% of sample, compared to municipalities that had at least one police station experiencing 5%. While a 1% difference in experiencing vigilantism does not appear large, it does represent a 17% increase. I then focused specifically on 18 municipalities that had no police stations and then had their first station constructed. While the nature of the data, specifically a short time period and small N, limits the ability to empirically quantify the impact of receiving a first police station there is qualitative evidence supporting the state absenteeism hypothesis. Out of the 18 municipalities that started with no police station, nine did not experience any acts of vigilantism, five experienced vigilantisms prior to the creation of the police station and no vigilante acts after, and four experienced vigilantisms after a police station was constructed. The municipalities that had experienced vigilantism prior to the creation of their first police station but had none after providing useful insight into how state absenteeism can influence vigilantism. For example, the municipality of Sumpango had two acts of vigilantism prior to their first police station being established in 2006, but after its creation did not experience any vigilante acts through 2017 when the data ends even though it has above average levels of crime. On the other end of the spectrum are the municipalities of Nento (one prior occurrence) and San Mateo Ixtatan (one prior occurrence) who typically have relatively low-crime rates but saw spikes in crime in the years that the vigilantism occurred. Without state presence, the citizens of these municipalities responded to increased insecurity by engaging in vigilantism, but once the state was present, they no longer acted outside of the law.

While the empirical analysis presented is only able to identify if state presence in the form of police stations correlates with vigilantism, additional analysis was conducted utilizing the newly released “Lynching in Latin America” (LYLA) dataset to provide preliminary evidence of potential mechanisms that may be at play. LYLA includes both fatal and non-fatal lynchings as well as characteristics for each event that took place. A detailed breakdown of key information can be found in the appendix. For example, in the empirical analysis of state absenteeism the Indigenous variable was statistically significant which suggests Indigenous communities are more likely to engage in vigilantism than non-Indigenous. However, the LYLA data indicates that when Indigenous communities/citizens engage in lynchings they are five times more likely to be fatal lynchings than non-fatal, but non-Indigenous citizens are more likely than Indigenous citizens to engage in non-fatal lynchings. This would suggest that Indigenous citizens are not actually more likely to engage in vigilantism in general but are more likely to escalate to higher levels of violence, which provides additional insight into the previous findings. This would suggest that vigilantism isn’t driven by the Indigenous culture.

Another feature of LYLA is that it captures police involvement including if the police arrested either the target or the vigilantes, if the police prevent the onset/escalation of violence, if the vigilantes cooperate or resist the police, and if the police are in favor of the lynching. This information can provide valuable information on the potential mechanisms that connect police presences to acts of vigilantism by showing if citizens that engage in vigilantism are actually being arrested by the police, if they are cooperating with the police, and if police presences prevent the onset of violence. The data indicates that the citizens engaging in vigilantism are almost never arrested and that the police almost always arrest the target. This would suggest that police presence is not deterring vigilantism with the threat of arrest. The police do, however, largely prevent the onset or escalation of violence, which suggests that police presence might not actually be stopping vigilantism from occurring, but their presences and responsiveness may be able to stop it from becoming fatal. Most interestingly, vigilantes cooperate with the police roughly half the time while resisting the other half. This presents an interesting scenario, where certain communities may be more likely to cooperate under certain situations more than others and is a worthwhile future direction for research.