doi 10.4067/S0718-83582012000100002


Using Social Disorganization Theory to Understand the Spatial Distribution of Homicides in Bogota, Colombia1


Gipsy Escobar2

2 United States. B.A. in Political Science from Universidad de Los Andes (Bogota, Colombia), M.A. in Criminal Justice from the John Jay College of Criminal Justice (City University of New York), candidate for a doctoral degree in Criminal Justice from John Jay College of Criminal Justice and the Graduate Center of the City University of New York. Assistant Professor in the Department of Criminal Justice and faculty member of the Graduate School at Loyola University of Chicago.


The social disorganization tradition has found community disadvantage to be one of the strongest and most consistent macro-level predictors of homicides in urban areas in the United States. However, research conducted in urban areas of Latin America suggests that the effects of disadvantage and disorganization on homicides may be different in that region. This study assesses the spatial effects of community disadvantage and the ability of communities to secure external resources (i.e. public control) on neighborhood homicide rates in Bogota, Colombia using spatial data analysis. Data from several sources including official homicide figures, 2005 census, and interviews were used. Results provide partial support for social disorganization theory: concentrated disadvantage/social isolation, and social disorder seem to predict higher homicide rates, while the availability of basic public services (a proxy for public control) has a negative effect on the dependent variable. Unexpected results were observed: the presence of agencies of social control (i.e. police and conflict resolution) predict higher, while the proportion of young males and population density predict lower homicide rates. The presence of other criminal structures (e.g., gangs, militias, drugs and arms trafficking) and people displaced by the conflict were not significant. Implications for theory and policy are discussed.




Research in the ecological tradition has established that differences in the socio-structural characteristics of communities produce variation in crime and delinquency rates. Shaw and McKay observed that Chicago neighborhoods with higher concentrations of poverty, residential mobility, and heterogeneity of values were more likely to have higher delinquency rates in the early 20th century. They characterized these communities as socially disorganized, a condition that caused neighborhoods to be less efficient in exercising social control and, consequently, more criminogenic than more affluent, stable, and homogeneous communities3.

Later research criticized this approach for focusing exclusively on the internal dynamics of communities, ignoring the external political and economic processes and decisions that shape them4. Bursik and Grasmick proposed a systemic approach to social disorganization aimed at explaining the interactions that occur between a community’s internal networks and the external world in the process of attempting to regulate behavior. The Systemic Model of Crime Control puts forward the idea that social networks exercise social control at three separate, but interconnected levels. The first and most basic level of control takes place within private networks (i.e. families, friends, neighbors) where the expectations for acceptable behavior are transmitted and through which the behavior of children and adolescents is supervised. The next echelon of control, the parochial level, represents a community’s ability to oversee the actions of residents and visitors, and it is exercised by broader interpersonal networks (i.e. neighborhood associations, tenant groups, parent-teacher associations, neighborhood watch groups) and through the participation in local institutions (i.e. churches, schools, voluntary organizations). Finally, the public level of control connects private and parochial ties to a larger system of networks embedded within the ecological structure of a city. Public control thus represents a community’s ability to secure needed services and resources that are managed and distributed by external agencies. In general, these resources are limited and local communities must compete with other neighborhoods for their acquisition5.

Research on the ecology of crime has been largely conducted in the United States and the Anglophone world (Canada, England, and Australia). In recent years, however, there has been an increase in the interest to test ecological theories of crime in other latitudes such as China and Latin America. Indeed, an international conference on violence in Latin American neighborhoods convened by this journal in Santiago, Chile on October, 2011, showcased over forty studies analyzing various indicators of neighborhood violence in the region, about a third of which used an ecological approach. Despite this heightened interest in the ecology of crime in Latin America, there has been very little research using this conceptual framework to study violence in Colombia. Scholars in that country have mainly focused their attention on disentangling political violence from other types of violence at the national, regional and local levels, and on evaluating the effects of local policies on violent crime rates. Moreover, with the exception of Cerdá, Morenoff, Duque and Buka6, studies that look at the potential effect of socio-structural variables on urban violence in that country have not relied on an ecological theoretical framework. In addition, research on the ecology of crime in the United States and elsewhere has mainly focused on examining the effects of social cohesion and collective efficacy—embodied in private and parochial networks—, with very little emphasis on exploring the effects of the public level of control on violent crime.

This study attempts to contribute to this literature by empirically testing the generalizability to the Latin American context of ecological theories of crime developed in the United States. The research employs a variety of data sources to explore the effect of social disorganization indicators such as concentrated disadvantage, social disorder, and residential mobility on the spatial distribution of homicide rates in Bogota, Colombia. In addition, it proposes alternative measures of social disorganization constructs that, though being analogous to those used in the American literature, are more reflective of both social realities and data availability in Colombia. Furthermore, the paper investigates the potential effect of the public level of control on homicide victimization, and it also controls for the potentially confounding role of the presence of illegal groups that may compete with the legitimate authorities in the exercise of local social control, such as gangs and urban militias of irregular groups. Finally, a large number of social disorganization studies have only placed attention to geography through the use of multilevel models controlling for the community context of individual behavior. However, these techniques do not account for the effect that neighboring communities might have on one another. This study uses spatial data analysis to focus explicitly on the location and geographic arrangement of neighborhoods, allowing for the estimation of local autocorrelations among them, and controlling for the effect of neighboring units on one another.


Literature Review

This section summarizes the research on socio-structural predictors of crime focusing on those that have received the most support in the literature and that will be central to this study’s exploration of the relationship between social disorganization and homicide in a Latin American context.


Concentrated Disadvantage

Concentrated disadvantage has generally been defined as the spatial concentration of poverty and other disadvantages, such as illiteracy, unemployment, and family disruption, in a confined number of neighborhoods within a city7. This concept has been hypothesized to foster social disorganization and the consequent likelihood of crime because the high concentration of unemployment rates in low-income communities reduces the amount of positive adult models, and diminishes the availability of potential links to employment for youth. Thus, the intergenerational flow of mainstream values is obstructed in these neighborhoods, whose residents end up putting “a premium on male aggressiveness as a means of dealing with limited opportunity and of providing a social identity”8.

In addition, the concentration of female-headed households with children and of high divorce rates reduces a community’s ability to supervise its young, primarily because the ratio of children in need of supervision to supervising adults is much higher than in more privileged communities9. Moreover, the compounded effect of disadvantage and crime rates further promotes instability and decay in these neighborhoods10. This process ultimately undermines the strength of the neighborhood’s social ties and institutions, harming their political base and their ability to converse with local bureaucracies to guarantee a fair distribution of resources and public services11.

The literature shows that the concentrated disadvantage index is one of the most consistent predictors of violence and homicide victimization in the United States. However, recent studies suggest that the effects of disadvantage may be different in Latin America’s urban centers because of the way low-income settlements developed in those cities. Indeed, since the 1930s a succession of economic crises affecting the agricultural sector, and the advent in some Latin American countries, such as Colombia, of conflicts between state and irregular forces produced several migration waves from rural areas into the major cities of those countries. Unlike in the United States where immigrants settled in the industrial heart of large cities, in Latin America, rural migrants tended to settle in the outskirts of urban areas. Villareal and Silva argue that the residents of these improvised settlements in Belo Horizonte, Brazil—many of which still do not receive basic public services today—were highly dependent on each other to survive, generating dense social networks in these disadvantaged neighborhoods12. In a similar vein, Cerdá et al. contend that these processes “combined with the subsequent need to fight for possession of illegally occupied lands and to get access to water, meant that many poor neighborhoods in Medellin became highly socially organized”13. Nonetheless, these neighborhoods still tend to have relatively high rates of violent crime. In fact, these two studies found social cohesion and collective efficacy respectively to be positively associated with disadvantage, disorder, and violence.

In sum, communities in differing social and cultural contexts may resort to different strategies to cope with conditions of concentrated disadvantage. However, neighborhood disadvantage does seem to increase the chances that a community will experience higher levels of crime and disorder than its wealthier counterparts, regardless of the cultural context.


Residential Mobility

The concept of residential mobility is perhaps one of the oldest in the ecological literature. The early studies by Burgess14 and Shaw and McKay15 argued that residential mobility is an indicator of social disorganization because the generation of strong social ties is a slow process, thus, when people are constantly moving in and out of a neighborhood, residents do not have the time to build meaningful and trustful relationships. The inverse of residential mobility, residential stability, facilitates the intergenerational transmission of mainstream values and the consequent creation of social networks that provide the new generations with opportunities to maintain or improve their social status16. In addition, residential stability increases residents’ stake in the community, thus promoting their participation in the establishment of common goals and the solution of collective problems17.

Residential mobility, generally measured as the percent of residents five years old and older who have changed residence in the past five years, has been found to be negatively associated to social cohesion and community satisfaction18, and positively related to disadvantage19, delinquency rates20, and violent crime21.


Social Disorder

Disorder can broadly be defined as a violation of tacitly agreed upon norms of public behavior. Indeed, social disorder is indicated by the presence of behaviors such as bands of teenagers congregating on street corners, prostitutes and panhandlers, public drinking, verbal harassment of women on the street, and open gambling and drug use, among other things. In this way, visible social disorder is an indicator of community organization because it shows that the residents lack the commitment to work together on the solution of common problems22. Bursik and Grasmick add that, insofar as social disorder affects collective action, it reduces a community’s regulatory capacities and it also decreases its ability to bring in external resources to prevent further deterioration.23

Social disorder has been shown to (1) mediate the effect of socio-structural factors on crime24; and (2) indirectly affect crime rates by decreasing the levels of neighborhood interaction and mutual trust25.


Socio-Structural Conditions and Homicide Studies in Colombia and Bogota

Figure 1 shows a very steep increase in Colombia’s homicide rate per 100.000 starting in 1985 (42), reaching a pinnacle in 1991 (81) and then beginning an uneven descent until finally arriving at pre-1985 rates in 2007 (39). Although historically having a homicide rate lower than the country’s, Bogota experienced a similar pattern, reaching the highest point in 1993 (81) and presenting a sustained decrease until attaining a rate of only 17 homicides per 100.000 in 2009. For many years Colombian scholars attributed these levels of violence to the internal armed conflict in rural areas and to interpersonal violence in urban areas. However, a World Bank report suggested that during the 1990s only about 20 percent of homicides in Colombia could be attributed to the conflict, and that the combination of economic and social harms (i.e. poverty, inequality, rapid urban growth, lack of educational and employment opportunities, family disruption, and situational precipitators such as easy access to alcohol, drugs, and firearms) were responsible for the other 80 percent26.


Figure 1. Homicide Rate Trends in Colombia and Bogota (1980-2009)


On the other hand, a panel study evaluating the predictors of violent crime in Colombia found that homicides (1980-1998) in 711 municipalities are determined by the existence of and interactions between armed groups, illegal activities, and ineffectiveness of the justice system, and that socio-structural conditions (i.e. poverty, inequality, and political exclusion) actually produce in Colombia a similar kind of violence to that generated by equivalent circumstances in other countries27.

In the case of Bogota, Llorente et al.28 found that 50 percent of homicides concentrate in only 14 percent of census tracts where only 25 percent of the population resides. Additionally, they found population composition (proportion of males), illiteracy, the school drop-out rate, the presence of criminal structures and illegal markets (e.g., stolen goods, drugs, and arms trafficking), and the presence of alcohol outlets to be positively related to homicide rates. On the other hand, they found a negative relationship between poverty (Unsatisfied Basic Needs Index) and homicide in Bogota. The study also found that localities with higher public spending (health, roads, security, education, and recreation) per capita also had higher homicide rates29. Nonetheless, another study yielded the opposite relationship whereby public expenditure in the social sector (health, education, and social promotion) actually had a negative, but weak, effect on the homicide rate at the locality level in Bogota30. Finally, using spatial data analysis, Formisano concluded that socio-economic variables did not explain the geographic concentration of high homicide rates in Bogota, but the presence of drug distribution and violent criminal groups did31.

In summary, research on the socio-structural predictors of homicide in Bogota is inconclusive and, in some instances, even contradictory. Clearly, the presence of irregular groups, organized crime and other criminal structures complicates the understanding of violent crime in that city, and the use of a theoretical approach that has shown to be sound elsewhere could shed some light on this debate.


The Systemic Model of Crime Control and the Public Level of Control

There is a plethora of research within the ecological tradition that focuses on the effects of informal social control—embedded in private and parochial social networks—on crime, delinquency, victimization, fear of crime, and social disorder. There is in fact widespread agreement that dense private and parochial ties produce higher levels of social cohesion32 and that higher levels of social cohesion, in general, predict higher levels of community involvement and lower levels of deviant behaviors33. Nevertheless, current social disorganization researchers emphasize that the existence of dense social networks is a necessary, but not a sufficient condition for the exercise of effective systemic social control34.

Even though it has been recognized that linking private and parochial networks to the public level is essential to achieving effective social control, there has been only a handful of studies in the United States, and practically none in the international context, that examine the effects of the public level of control on crime rates. The limited evidence from the United States, however, suggests that those communities that are not effective in reaching out to external agencies and that fail to acquire needed services and resources are more likely to have higher rates of crime, delinquency, and victimization than those that are successful in doing so3. In addition, the available research suggests that the effect of public control36 and access to social institutions37 on crime varies across ecological units and might mediate the effect of disadvantage and social isolation on crime.

Thus, the scant research shows support for a negative association between the public level of control and violent victimization. Additionally, the literature also suggests that residents of urban areas are in fact more likely to call the police to solve community problems than to directly intervene themselves. Alternatively stated, there is some evidence that citizens prefer to activate the public level of control when the public peace is disturbed than to directly exercise informal social control to restore it38.

Bursik and Grasmick contend that the likelihood of crime is higher in those areas where networks of public control fail to effectively provide services to the neighborhood. Two types of public services are related to crime control. The first type involves a relationship between the neighborhood and the police department, which may have a more direct effect on the ability of a community to control crime. Even when negative attitudes toward the police are prevalent in disadvantaged neighborhoods, research shows that residents still consider police intervention as the best way of controlling crime39. The second type, distribution of basic services, may have an indirect effect on the regulatory capacity of neighborhoods as it measures the strength of the ties a community has to the local government.40


Public Control and Illegal Sources of Social Control

Criminal groups may hinder social control and escalate violence in a neighborhood through at least two interconnected processes. The first process implies the cooptation of local social networks to ensure control of the local illegal market. The second process involves criminal groups usurping some state functions in order to (1) gain the support and tolerance of local residents; and (2) facilitate dealings with and reduce attacks from the state. In extreme cases, particularly when state presence is extremely weak, the second process may lead to a total impersonation of the state in these areas.

According to Browning, Feinberg and Dietz, dense ties and frequent contact among neighbors result “in more extensive integration of residents who participate in crime into existing community-based social networks”41. Patillo found that deviant and non-deviant residents are bound to each other in a “system of interlocking networks ... that sometimes paradoxically, and always precariously, keeps the peace”42.

Kubrin and Weitzer suggest that when neighborhoods experience vacuums in formal control (perceived or real) local offenders will take advantage of these voids and impose their own forms of control43. Patillo found that organized gangs helped maintain order in the neighborhood44, and Arias found that organized crime in a Rio de Janeiro favela “provides services to residents to maintain their support in the face of the violence provoked by drug trafficking.”45.

In sum, these findings suggest that when criminal structures co-opt local social networks, neighborhood residents may enter in tacit or even explicit agreements with criminals in order to keep a modicum of peace and order within the community. These agreements undermine the neighborhood’s regulatory capacities and its ability to prevent crime and violence because (1) law-abiding resident networks become embedded with criminal networks, fostering tolerance of illegal behavior; (2) criminal structures routinize the use of violence as a legitimate way to regulate behavior; and (3) law-abiding residents will be persuaded against contacting outsiders, including city authorities, to solve communal problems because such behavior may be considered as a breach of contract and may lead to violent retaliation.


Study Site and Unit of Analysis

Bogota, the capital of Colombia, is the most populous city in that country with a population of approximately seven million people. According to the 2005 census, 69 percent of its residents were under the age of 40, and 13 percent of the population was composed of males between the ages of 15 and 29. The ethnic distribution of the city was rather homogeneous with only 1,7 percent residents self-identifying as belonging to a minority group (Amerindian, Romani, Afro-Colombian). In terms of economic deprivation, 4,56 percent of the population reported that, during the week prior to the census, they had spent one or more days without consuming any food due to lack of money; and 9,4 percent reported having looked for work during the same time period46. Regarding family disruption, the census reports that 15 percent of households with children are headed by a single, separated, or divorced female. There is also a high rate of residential mobility with 32 percent of Bogota residents reporting they changed residences in the five years prior to the census. In fact, 37 percent of Bogota residents were not born in that city, 13 percent of which moved there in the five years before the census. Furthermore, of those who recently migrated to Bogota, 28 percent claimed having difficulties finding a job and six percent having a threat against their lives as their main reason to move.

The city is subdivided into 20 localities47 (see Figure 2) each of which is administered by a democratically elected Local Administrative Board and a local mayor appointed by the city Mayor. This study uses the official neighborhood or census urban sector as the unit of analysis. The 2005 census identifies 664 urban sectors. Fifty-eight sectors that were identified as being rural areas located at the fringes of the city, and two neighborhoods that were islands sharing no boundaries with any other neighborhood were deleted from the analyses. In addition, 35 units with population sizes smaller than 1.000 were merged to a neighboring sector belonging to the same administrative unit in order to avoid extremely inflated rates that could bias the analysis48. In this way, the analyses include 569 neighborhoods.


Figure 2. Bogota Political-Administrative Division


The main limitation of employing official neighborhoods is that the definition of their boundaries might not reflect the cognitive maps of the residents. In fact, official neighborhoods might be clusters of multiple areas informally identified as neighborhoods by residents and visitors alike. However, Haining proposes that administrative regions are ideal for spatial analysis because “[t]hey provide a framework for collecting data, delivering services, [and] distributing government funds”49. The use of official neighborhoods as the unit of analysis in this study is thus appropriate due to the focus on the effect of the availability of public services on homicide rates.

There is variability across neighborhoods for most of the socio-structural characteristics summarized here, particularly those related to issues of disadvantage. By and large, neighborhoods in the northeast area of the city are much more affluent than those located in the south of the city, though there is some internal variation such that spatial patterns can be identified. In general terms, though, the most affluent neighborhoods are located in the Usaquen and Chapinero localities and the least affluent in the San Cristobal, Rafael Uribe, Tunjuelito, Ciudad Bolivar, Bosa, and Kennedy localities. The remainder areas have a mix of middle- and lower-class residents, with some very affluent neighborhoods located in some areas of Suba, Barrios Unidos, and Teusaquillo. Sample characteristics can be observed in Table 1.


Data and Methods

Dependent Variable

The outcome variable of the study is the cumulative homicide rate per 10.00050 residents for the years 2003 to 2005. To create it, the homicide counts for the years 2003, 2004 and 2005 were summed up, divided by the average population size across the three years51 and then multiplied by 10.000. Because homicides are rather rare events, a cumulative approach is preferred to an averaged approach because it allows the researcher “to reduce measurement error and the problem of volatility in homicide counts from one year to the next”52.

One problem with using rates based on population size is the assumption that the underlying risk is located within the population residing in a given neighborhood. However, neither victims nor offenders need dwell in the area were a homicide event takes place. Some researchers propose using densities as a better way to standardize crime data to remove this bias53. Nonetheless, the use of population rates in this study is justified because of its exploratory nature and the fact that most of the homicide literature relies on rates. Thus it is important to use a standardization process that allows for comparison with research conducted in other latitudes. In addition, homicide rates have been found to present spatial patterns of concentration and diffusion that seem to make them amenable to the ecological approach54.

The homicide data are part of a larger dataset collected by the Centro de Estudios sobre Desarrollo Ecómico (CEDE) at Universidad de Los Andes in Bogota. Using the death protocols kept by the Instituto Nacional de Medicina Legal y Ciencias Forenses (INMLCF) in hard (1977-1995) and electronic (1996-2005) form, CEDE created a panel dataset for all of the homicide events that took place in Bogota during the period 1977-2005. The data contain information on the characteristics of the victim and on the circumstances of the event. In addition, based on the address where the murder occurred (or where the body was found by the authorities), the data were geocoded to the X- and Y-coordinate level. To create the neighborhood homicide counts, the data was projected on a map of Bogota using the GIS software ArcMap©, and then the points were joined to the neighborhood polygons to get the aggregated counts. A tabular join procedure was used to merge map, homicide counts, and census attributes.

The deaths caused by two car bombs placed by the FARC in 2003 (Club El Nogal on February 27, 2003, 35 deaths, and San Andresito commercial area on October 8, 2003, 8 deaths) were excluded from the analysis to avoid biasing the spatial patterning of homicide rates. In particular, the terrorist attack against Club El Nogal would have artificially inflated the homicide rate of the Rosales neighborhood, which, otherwise, has no homicides for the study period.


The average cumulative neighborhood rate is 10,82 homicides per 10.000 residents55. The variable was normalized using a natural log transformation to reduce the large positive skewness observed in the raw data.

Table 1. Sample Summary Statistics*

Mean or %


Std. Dev.



Dependent Variable

Homicide Rate per 10.000






Independent Variables

% Population Experienced Hunger






% Population 15+ Illiterate






% Female-Headed Households w/Children






% Homes w/ Phone Service






% Homes w/ Sewerage Service






% Homes w/ Electricity Service






% Population Moved






Presence of Police Stations or CAIs (Yes)


Presence of Conflict Resolution Agencies (Yes)


Rate of Alcohol Outlets per 10.000






Rate of Videogame, Gambling & Lotto Outlets per 10.000






Control Variables

Population Density per Kmr2






% Population Young Males (15-29 age)






% Population Moved due to Threat to Life






Presence of Gangs (Yes)


Presence of Social Cleansing (Yes)


Presence of Hitman "Offices" (Yes)


Presence of FARC Militias (Yes)


Presence of Paramilitary Militias (Yes)


Presence of Drug Markets (Yes)


Presence of Arms Markets (Yes)


* Summary statistics are presented for raw variables prior to transformations and dimension reduction.


Independent Variables

Concentrated Disadvantage and Social Isolation: Concentrated disadvantage has been measured by combining indicators of economic deprivation, family disruption, and racial heterogeneity56. This study creates an index analogous to that commonly used in ecological research in the United States, combining a number of variables from the 2005 census. A Principal Factor Analysis procedure was executed (78,38 percent of variance explained after Varimax rotation), which yielded two factors57. The concentrated disadvantage and social isolation factor has strong positive loadings from percent of population who experienced hunger for more than one day due to lack of money, percent population 15 years old and older who is illiterate, and percent of households with children headed by a single woman, and a strong negative loading from percent of homes with phone service (representing a social isolation element).


Residential Mobility: The association between residential mobility and crime rates is so common place in the literature that Bursik considers that ecological models that do not test for the residential mobility assumption “will not only have limited degree of theoretical power, but they will have an interpretive framework that is generally unrelated to the dynamics of modern urban areas”58. This study uses the common measure of residential mobility indicating the percent of people five years old or older who have changed residence in the past five years using information from the 2005 census (mean=31,89).


Social Disorder: Skogan argues that social disorder is signaled by the presence of disruptive social elements such as rowdy teenagers, prostitutes, public drinking, open gambling and drug use59. Field observations are recommended as the most reliable way of collecting data about social disorder60. In the absence of direct observation, this study utilizes two proxy indicators obtained from the 2005 census: the rate of alcohol outlets per 10.000 residents (mean=13,75) and the rate of video game, lotto and gambling outlets per 10.000 residents (mean=8,18). Both indices were normalized using a natural log transformation to fix significant positive skewness.


Public Control: Residents’ perceptions and satisfaction with the local government’s involvement with the neighborhood61, residents’ willingness to cooperate with the police and the existence of active police-community partnerships62; overall city expenditure per capita63; residents’ connections to city bureaucracies64; and the availability of general services such as parks and playgrounds, neighborhood watch programs, health services, block group and tenant associations, mental health centers, after-school programs, etc65 have all been proposed as measurements of the public level of control.

This study proposes to measure public control by using indicators of the two types of public services suggested by Bursik and Grasmick. First, as mentioned in the section describing the concentrated disadvantage index, the PFA procedure yielded a second factor with large positive loadings from percent of homes with electricity service and percent of homes with sewage service. Although a two variable factor is problematic and should be interpreted with caution, the basic public services index is still useful as a proxy of public control.

The second type of public control involves a dichotomous indicator of the presence of police precincts and Comandos de Atención Inmediata (CAI)66 in the neighborhood (20,9 percent have either a police precinct or a CAI in 2005) and a dichotomous indicator of the presence of conflict resolution agencies (Inspección de Policía, Personería, Comisaría de Familia, and Unidad de Mediación y Conciliación) whose function is to help civilians solve private conflicts outside of the criminal courts (7,4 percent have a conflict resolution agency in 2005). The data on police presence was obtained from the website of the Metropolitan Police Department of Bogota, and on the presence of conflict resolution agencies from locality profiles produced by the Mayor’s Office and available online.


Control Variables

Population Structure and Composition: The literature generally finds that crime rates tend to be positively associated to population density and the proportion of young males in the population. This study follows suit by using census information to control for the population density per Km2 (mean=30.181), and the percent of males aged 15 to 29 (mean=13,37). Population density was normalized using a squared root transformation to improve positive skewness.


Forced Displacement: It is calculated that since 1985 between 3,3 and 4,9 million people have been forcefully displaced by the conflict67. Most displaced people move from rural areas or smaller townships to the largest cities, including Bogota. Following the arguments regarding the effect of heterogeneity on social organization and social control, their arrival to a complex urban ecosystem could potentially destabilize the receiving communities. Having been exposed to violence they may (1) alienate themselves for fear of experiencing violence again, or (2) be more tolerant of the use of violence to solve interpersonal conflicts. It is also possible that their stigma as displaced people makes them easy targets of violence.

A forced displacement index was created using the 2005 census to calculate the percentage of individuals who moved in the past five years due to threats to their lives. The index is not just limited to persons coming from outside Bogota, but it also includes people who moved within the city (mean=0.64). This variable was also normalized using a squared root transformation to solve problems of skewness.


Criminal Structures and Organized Crime: Between 2003 and 2004, the CEDE research group led by Llorente and Escobedo conducted interviews with the commanders of the 19 police precincts of Bogota (or their delegates) to assess the presence of criminal structures and organized crime at the neighborhood level. During the interviews, the researchers showed a list of neighborhoods to the police commanders and asked them to identify those neighborhoods under their jurisdiction where the following criminal structures, organized crime groups or criminal activities were present: gangs (37,2 percent), FARC militias (8,9 percent), paramilitary militias (16,7 percent), hitman offices (oficinas de sicarios—25,1 percent), social cleansing (5,6 percent), arms trafficking (19,3), and drug distribution (72,8 percent). The data is coded as 1 to indicate the presence or 0 to indicate the absence of these groups or acts in each neighborhood. Hitman offices and social cleansing were combined in an indicator of selective murder groups (26,1 percent). The rest of the variables were included in the analysis without combining them in the hopes of evaluating the effects of different types of criminal structures on the neighborhood homicide rates.


Data Analysis

The literature shows that homicide rates in the United States generally present patterns of spatial dependence or autocorrelation68. According to Messner and colleagues, these patterns are generated by so-called “contagious transmission”, a process that uses social networks and communication flows to spread information about the occurrence of violent events in one neighborhood to its surrounding areas, thus influencing violence in those nearby communities in a nonrandom geographic way69.

This study utilizes Exploratory Spatial Data Analysis (ESDA) techniques to examine the spatial distribution of homicide rates in Bogota and determine whether patterns of spatial dependence or spatial heterogeneity exist; and Spatial Regression Analysis (SRA) to explore the effects of social disorganization and public control on homicide rates, controlling for potential spatial dependence.

Exploratory Spatial Data Analysis (ESDA) is a combination of graphical and statistical techniques that allow the researcher to visualize and describe spatial distributions, and detect spatial patterns, spatial clusters, and spatial outliers70. ESDA techniques include the estimation of global and local statistics of spatial autocorrelation such as Moran’s I and Local Indicators of Spatial Association (LISA), and visualization methods such as Moran scatterplots and LISA maps. These methods explicitly take into account the spatial arrangement of the units of analysis through the inclusion of a spatial weights matrix. A row-standardized first-order queen contiguity matrix (neighbors are determined by shared borders and vertices) is thought to be appropriate for the spatial distribution of Bogota neighborhoods, because the wide variation in the units’ areas precludes the use of a distance matrix and because the phenomena under study transcend geographic boundaries.

In addition, it is today widely recognized that classic Ordinary Least Squares (OLS) regression is inappropriate when analyzing area or lattice data, because spatial data usually violate the assumptions of homogeneity of variance, and independence of residuals. The violation of these laws of OLS yields biased, inconsistent and inefficient regression coefficients because the standard errors are overestimated for positive values and underestimated for negative values, biasing the tests of significance toward rejecting the null hypothesis. In addition, R-square estimates are exaggerated and, therefore, inferences are incorrect71.

Spatial Regression Analysis pays explicit attention to the location and arrangement of geographic units by including a spatial weights matrix that reflects the expected geographic processes into the model. A spatial lag model is recommended when the analyst has evidence that a pattern of spatial dependence exists (the values of y in location i are suspected to be influenced by the values of y in i’s neighbors), and the effect is above and beyond other predictors specific to i72. In a spatial lag model the spatial dependence is entered into the model as an additional covariate, “a so-called spatial lag, or weighted average of values for the dependent variable in “neighboring” locations””73. This research employed the open source software GeoDaTM, a trademark of Luc Anselin, to conduct all analyses.



Exploratory Spatial Data Analysis

The first step in the ESDA process is to test the null hypothesis of spatial randomness by estimating whether spatial autocorrelation is present in the data. The global Moran’s I is perhaps the most frequently used method to test for spatial dependence. Estimation and inference in spatial data analysis must rely on random permutations that reshuffle the observed values over space to estimate how likely the actual spatial distribution would be. The randomization exercise produces a reference distribution of possible combinations of values over space. The observed coefficient, in this case Moran’s I, is then compared to the reference distribution to determine the likelihood that it could stem from a random distribution74. This process produces pseudo-significance levels based on the number of permutations performed.

Table 2 below presents evidence of spatial dependence in both the dependent variable and the main explanatory predictors, with concentrated disadvantage/social isolation and homicide rates presenting respectively the strongest spatial autocorrelations.


Table 2. Global Moran's I Statistics*



Homicide Rates (NL)


Concentrated Disadvantage


Residential Mobility


Alcohol Outlets Rate (NL)


Gambling Outlets Rate (NL)


Basic Public Services Public Control


* Empirical pseudo-significance based on 999 random permutations.
** Pseudo-p<0,001.


Although the global Moran’s I helps determining whether there is a general pattern of spatial dependence in the whole study region, it is not very useful in identifying the specific source of the autocorrelation when there is heterogeneity in spatial dependencies. Local Indicators of Spatial Association (LISA) or local Moran statistics disaggregate the global indicator of correlation by calculating indicators of spatial association between each unit and the average of its neighbors75. Local autocorrelations can be visually inspected using LISA cluster maps, which color-code those areas with significant local autocorrelations according to the type of association they exhibit76. Indeed, clusters of high-high correlations (locations with above average values surrounded by neighbors with above average values) are coded red, low-low correlations (areas with below average values surrounded by neighbors with below average values) are coded blue, low-high correlations (units with below average values surrounded with above average neighbors) are depicted in light blue, and high-low associations (areas with above average values located in the midst of below average neighbors) are displayed in pink77.

Figure 3 presents LISA maps for the outcome variable and the main predictors. These maps show that there are spatial regimes within the study region for each of the variables. For instance, map (A) shows that there are four large significant clusters where units with high homicide rates are contiguous to neighborhoods with high rates as well. These clusters are located in the downtown area of Bogota (Martires, Santa Fe, and La Candelaria), the south (Rafael Uribe, and Ciudad Bolívar), and the southwest (Kennedy). In addition, the LISA map shows a few pockets of neighborhoods with low homicide rates surrounded by communities with similar characteristics in the north (Usaquén, Suba. Chapinero, and Teusaquillo), one in the northwest (Engativa), and one more in the south (between Bosa and Tunjuelito).


Figure 3. LISA Maps


The downtown spatial regime has the highest LISA values and it includes infamous neighborhoods such as San Bernardo, San Victorino, Veracruz, La Capuchina, and Santa Ines, the latter one known for the so-called Calle del Cartucho which had been completely taken over by drug dealers, street beggars, prostitutes and arms traffickers by the 1990s. This was known as the most dangerous zone of the city and not even the police dared enter it. In an effort to reduce violence in the downtown area of Bogota, where many of the governmental and cultural institutions of the city concentrate, the administration of Mayor Peñalosa razed these blocks to the ground and built the Tercer Milenio park, inaugurated in 2000, as a way of recovering public space from the hands of criminals. Interestingly, violence still concentrates in these areas by the 2003-2005 period. The main criticism of Peñalosa’s urban renewal project is that it did not include a relocation strategy for the thousands of homeless people who had taken over this neighborhood. Thus, instead of eliminating it, the problem was merely displaced to adjacent areas such as El Bronx, which today is said to supply most of the drugs consumed in the city.

Furthermore, the spatial distribution of concentrated disadvantage and social isolation in Map (B), Figure 3, shows a clear division of the city by social class with most of the neighborhoods in the northern localities of Usaquen, Chapinero and Teusaquillo having no or very low levels of disadvantage, and the southern localities of Santa Fe, San Cristobal, Usme, Ciudad Bolivar and Bosa having the worst levels of disadvantage. The remaining localities have a more mixed distribution. There seems to be an overlap between homicide rates and disadvantage and isolation, particularly in the spatial regimes located in the downtown and south areas of the city. In fact, an inspection of the bivariate correlation between the spatial lag of homicide rates and concentrated disadvantage/social isolation shows a positive significant association (see Table 3).

The spatial distribution of residential mobility in Bogota (see Map (C), Figure 3) is somewhat contrary to social disorganization expectations. It seems that those areas of the city that have higher levels of disadvantage also have the lowest levels of residential mobility. It is possible that in Bogota disadvantage actually traps people. It is not uncommon for middle- and upper-class residents to move often, usually within the same or to better neighborhoods. Poor people, on the other hand, do not have the means to improve their situation and end up effectively being trapped in the same residence potentially for life. In fact, the bivariate spatial correlation between the spatial lag of disadvantage/isolation and residential mobility is significant and negative, suggesting that at higher levels of disadvantage/isolation there are lower levels of mobility (I=-0,22, pseudo-p<.001). Nonetheless, the bivariate spatial association between homicide rates and residential mobility is not significant (see Table 3).

Pockets of both alcohol and gambling outlets rates appear in the downtown localities of Martires, Santa Fe, and La Candelaria, and the locality of Chapinero in the north (see Maps (D) and (E), Figure 3). It is possible that the presence of several universities in Chapinero and Santa Fe is driving the large rates here, but it is also clear that there seems to be an overlap with the distribution of disadvantage (bivariate associations of the spatial lag of disadvantage/isolation with alcohol and gambling outlets were equivalent in value: I=0.16, pseudo-p<0,001) and homicide rates (see Table 3).

Table 1 shows that the average percentage of homes with electricity and sewage services is actually quite high. Most neighborhoods seem to have between an average and a high coverage of basic public services; however, the spatial distribution shows a few pockets of neighborhoods with below average coverage in the downtown area (Martires, Candelaria, Santa Fe), Ciudad Bolivar, San Cristobal, Fontibon and Kennedy (see Map (F), Figure 3). The bivariate association between the spatial lag of homicide rates and the basic public services public control index is weak, but significant and in the expected direction: the higher the coverage of basic public services the lower the homicide rates (see Table 3).


Table 3. Bivariate Moran's I between Spatial Lag of Homicide Rates and Predictors


/ Statistic

Concentrated Disadvantage


Residential Mobility


Alcohol Outlets Rate


Gambling Outlets Rate


Basic Public Services Public Control


* Empirical pseudo-significance based on 999 random permutations.
** Pseudo-p<0,001.


Spatial Regression Analysis

These statistics and visual tools confirm that there is spatial dependence in the data. Therefore, a spatially lagged model is used here to estimate the effect of social disorganization on homicide rates.

Table 4 presents the results of OLS and spatial lag models predicting neighborhood homicide rates in Bogota using social disorganization and public control indicators and controlling for population structure and composition and the presence of variables somewhat idiosyncratic to the Colombian context of violence. Model fit statistics in Table 4 and residual diagnostics in Figure 4 confirm that the spatially lagged model is more appropriate. When controlling for the spatial autocorrelation in the model: (1) explained variance increases; (2) model fit improves; (3) coefficients are more conservative; and (4) the significance level of some variables change. In particular, the percent of young males is not significant in the OLS model, but it becomes significant at the .05 level in the spatial lag model.


Table 4. OLS and Spatial Lag Models for Homicide Rates (NL) (2003-2005)

OLS Model

Spatially Lagged


Coefficient (SE)


Coefficient (SE)



1.91*** (.22)




Spatial Lag (Rho)





Concentrated Disadvantage

.31*** (.06)


.24*** (.05)


Residential Mobility

.01 (.01)


.01† (.00)


Alcohol Outlets Rate (NL)

.12** (.04)


.08* (.04)


Gambling Outlets Rate (NL)

.10* (.05)


.10* (.04)


Basic Public Services

-.10* (.04)


-.08* (.04)



.22* (.09)


.19* (.08)


Conflict Resolution Agencies

.40** (.14)


.28* (.13)


Percent of Young Males

-01 (.01)


-.02* (.01)


Population Density KM2 (SQRT)

-.01*** (.00)


-.01*** (.00)


Forced Displacement (SQRT)

.30 (.18)


.15 (.18)



-.076 (.09)


-.02 (.08)



-.29† (.17)


-.23 (.15)



.07 (.13)


.02 (.12)


Drug Distribution



-.01 (.08)


Arms Trafficking

.05 (.12)




Selective Murder Groups

.34** (.11)


.23* (.10)


Adjusted R2






Log likelihood















Robust LM (lag)



Robust LM (error)



Likelihood Ratio



***p<0,001; **p<0,01; *p<0,05; †p<0,1


Furthermore, Figure 4(A) shows that the residuals are correlated and heteroscedastic in the OLS model, and that the predictors do not do a good job at explaining the spatial regimes in the localities Santa Fe, Martires, Rafael Uribe, Ciudad Bolivar, Kennedy and Barrios Unidos. Residuals in the spatial lag model are still heteroscedstic (Figure 4(B)), though the problem is somewhat smoothed, but they are not correlated and the introduction of the spatial lag accounts for most of the spatial clusters.


Figure 4. Model Diagnostics


The spatial lag model uses a first-order queen contiguity weights matrix, and it explains about 46 percent of the variance in the homicide rate. The coefficient for the spatial lag associated with homicide rates (Rho) is a significant contributor to the model, highlighting the importance of spatial proximity in understanding the relationship between homicide rates and social disorganization. All social disorganization variables, with the exception of residential mobility, significantly and positively predict homicide rates. Moreover, all three measures of the public level of control are significant. The basic public services indicator of public control performs in the expected direction predicting lower homicide rates as the coverage of public services increases. Police and conflict resolution agencies, on the other hand, behave in the opposite direction to what was hypothesized and positively predict homicide rates. A possible explanation of this relationship may be that the allocation of police and conflict resolution resources tends to be more reactive than proactive. In other words, both city bureaucracies and community residents may push for the investment of this kind of resources in the areas with the highest crime rates. This might be particularly the case with the placement of CAIs, since residents can request them from the police, and conflict resolution agencies, since they might be considered to be more needed in high crime areas by city authorities.

Percent of young males and population density are significantly but, surprisingly, negatively associated to homicide rates. It is possible that in areas with low population density there are less capable guardians with the ability to exercise informal social control and, thus, the opportunity for violence and for criminals to dispose of victims murdered elsewhere is greater in these areas. The finding regarding the percent of young males, though having a very small effect size, is more puzzling. Some researchers argue that the inclusion of community population characteristics as covariates in models predicting violent victimization might be misleading because neither the offenders nor the victims need be residents of the area where the incident took place78. Thus, it is possible that the percent of young males in a community does not necessarily account for the levels of violence that neighborhood experiences. In fact, when population density is removed from the model, the percent of young males loses its significant effect.

The presence of selective murder groups has a positive effect on homicide rates. These groups often work at the behest of drug cartels, paramilitaries, and even disgruntled citizens who want to exert retaliation against those who do not act within expectations. These groups also serve as agents of illegal social control when they dispose of “undesirables” (e.g., homeless people, prostitutes, drug addicts, homosexuals, pedophiles, street criminals, social activists, etc.) in the community by exerting the ultimate punishment.

Finally, although the other variables were not significant, the direction of their coefficients is worth noting, particularly in the case of the presence of gangs, FARC and drug distribution. These three variables were expected to produce positive coefficients, yet they yielded negative values, perhaps suggesting that these groups could also be considered as agents of social control, only in their case they are trying to keep the peace in the communities that tolerate their presence instead of eliminating adversaries.


Discussion and Policy Implications

The current study makes important contributions to the ecological understanding of homicide rates in an urban context outside of the United States. First, the study provides some evidence in favor of the generalizability of ecological theories of crime to the Latin American context. In particular, the fact that most social disorganization measures have an effect above and beyond that of criminal structures, organized crime and conflict should bring attention to the fact that disadvantage and isolation are more socially detrimental than the existence of criminal organizations alone. As noted earlier, concentrated disadvantage and social isolation minimize social advancement opportunities, cutting the links to mainstream society, and hindering the generational transmission of mainstream values. Families and other social institutions see their ability to regulate the behavior of children reduced by the constant demand to provide for their well-being with very scarce social and economic resources. Under these conditions, residents of disadvantaged neighborhoods resort to alternative solutions to the social advancement problem, some of which involve engaging in illegal activities. Furthermore, the illegal nature of these alternatives implies that those who engage in them must compete among themselves to gain the control of markets and places. This competition tends to involve the threat and use of violence, routinizing it to secure a more or less stable position within these systems.

Kubrin and Weitzer’s results show “that neighborhoods with higher levels of concentrated disadvantage are especially likely to experience greater numbers of retaliatory than non-retaliatory killings”79 (emphasis in original). The detrimental effects of concentrated disadvantage may also spread to neighboring areas by increasing their homicide rates, independent of their own socio-structural conditions80. In sum, perhaps the most deleterious by-product of concentrated levels of economic, social, and cultural disadvantage in urban areas is the attenuation of mainstream cultural values81 that protect a community from the spread of deviance and violence.

Bogota’s administration and the national government should make a greater effort in reducing inequality by: (1) providing free public education for all children and offering literacy programs for illiterate adults; (2) expanding the eligibility criteria for social programs such as SISBEN by increasing the lower bound of the minimum wage required to qualify as “below the poverty line” to be able to receive public assistance; (3) enhancing the programs offered by the Secretaría de Integración Social (Social Integration Secretariat), in particular those that deal with at-risk children and youth, food policies, and family stability, to make sure that they are reaching the intended population; and (4) increasing the coverage of basic public services to 100 percent of the population. Policies that attempt to deal with urban decay by focusing only in the physical recovery of neighborhoods, but that do not improve the conditions of the residents of these communities in the aforementioned policy areas are bound to simply displace the violence to other communities, as it was the case with Calle del Cartucho and the Tercer Milenio Park.

The results of the present study contradict some of the findings of prior research. A few studies of violence in Colombia have concluded that the presence of illegal armed groups and markets accounts for much of the variance in homicide rates, while socio-structural factors such as poverty and inequality explain only a small amount82. One of these studies (Formisano’s) even used a spatial lag model, thus accounting for the spatial dependence effect. Perhaps, the shortcoming of those studies is the way in which they measured poverty by using indices such as GINI and the Unsatisfied Needs Index. The measure of concentrated disadvantage and social isolation proposed here might be a more realistic reflection of felt poverty, particularly because it includes experienced hunger and lack of phone service. These, in combination with family disruption and illiteracy, might provide a truer depiction of real disadvantage. In this way, the second contribution relates to the construct validity of measures of social disorganization, particularly as they relate to violence in developing countries.

The third contribution is connected to the empirical test of the public level of control, a largely under-tested construct. This study proposes that the availability of basic public services and the presence of agencies of social control might provide a somewhat adequate measure of public control. Although the basic public services index should be interpreted with caution (only two variables loaded in the factor), it produced similar results to more sophisticated measures used by Sanchez and colleagues at the locality level83. In 2004 a city ruling created a new administrative unit known as Unidades de Planeación Zonal (UPZ – Zonal Planning Units). These UPZs are smaller clusters of neighborhoods that share similar demographic and socio-structural characteristics and through which the locality and the city channel resources to the neighborhoods. Information collected at this level might provide better insight into the public level of control and the city administration could require that detailed information about the allocation of public resources be collected by all UPZs in order to evaluate the effectiveness of social programs.

Finally, the positive association between the presence of agencies of social control and homicide rates might actually be interpreted as an exercise of public control. Research in the United States has shown that even when the relationship between the police and the community in disadvantaged neighborhoods is adversarial, residents still consider police intervention as the best way of controlling crime. It is possible that the presence of these agencies in neighborhoods with higher homicide rates is partially explained by an actual exercise in public control whereby community and authorities agree to their placement. In 2007, the Metropolitan Police of Bogota started implementing a mobile CAI program to improve police response to street crimes. The units are regularly relocated based on the prior spatial distribution of crimes. These units might prove effective not just in terms of preventing and reacting to crime, but also in improving police-citizens relationships as the units are also equipped to respond to less serious complaints. The performance of these mobile CAIs needs to be evaluated in a rigorous way to determine if these resources are being appropriately apportioned.


Limitations and Future Research

There are several limitations to this study. First, the ecological model implemented carries a certain amount of aggregation, also known as the ecological fallacy, because the effects of individual residents’ attributes are not being controlled for in the model. Future analyses should consider a spatially lagged multilevel approach. A second shortcoming is the cross-sectional nature of the study. Although it has been argued here that the behavior of homicides as well as of the predictors tends to be stable from one year to the next, changes do take place in the long term. For instance, in the case of Bogota, there have been dramatic changes in the homicide rates in both upward and downward directions since the 1980s (see Figure 1). Therefore, a longitudinal study could arrive to more accurate causal conclusions about the nature of homicides in Bogota. Third, as mentioned above, the construct validity of the measures of public control is suspect in this study and future studies should attempt to collect more sophisticated data relating the allocation of resources and services at the neighborhood level. Fourth, this study did not include any direct measures of social organization such as participation in local organizations and associations, which are theorized to mediate the effects of disadvantage on crime. Future studies should control for this important variable. Fifth, the multivariate analysis did not account for the possible effect of interactions between disadvantage/isolation, public control, and social disorder. The literature argues that social disorder and public control may mediate the effect of socio-structural disadvantages on crime, therefore a better specification of the model should control for these interactions. Finally, there is some debate in the literature as to what is the best way to standardize outcomes and predictors (e.g., rates, density, cumulative counts) to conduct ecological studies. Most studies utilize population standardized rates to create their measures. However, some studies have argued that it is misleading to assume that offenders and victims reside in the area where the criminal event happened. For instance, offender search theory argues that offenders seek and stumble upon crime opportunities as they travel between nodes or areas where they conduct most of their routine activities. These nodes include, but are not limited to their place of residence84. Consequently, some researchers favor the use of cumulative counts85 and others the use of densities86 to study the ecology of crime. Future research should compare the three approaches and assess whether the same kind of predictors are associated to the ecology of homicides.

In sum, this study provides some evidence in favor of the usefulness of social disorganization theories to understand violent crime in Latin American cities. Similar models should be replicated across the region to confirm whether the evidence from Bogota is generalizable to other urban areas in the subcontinent. The results suggest policy implications to reduce disadvantage and increase the public level of control as potentially effective strategies in fighting lethal violence in the region.



1 This study was possible thanks to María Victoria Llorente (Fundación Ideas para la Paz), Rodolfo Escobedo, and the Centro de Estudios sobre Desarrollo Económico at Universidad de Los Andes in Bogota who made the homicide and police interview data available for analysis by this author.

3 Shaw & McKay, 2011 (1942).
4 Heitgerd & Bursik, 1987.
5 Bursik & Grasmick,1993a.
6 Cerdá, Morenoff, Duque and Buka, 2008.
7 Krivo, Peterson, Rizzo and Reynolds, 1998.
8 McGahey, 1986, p. 252.
9 Krivo et al., 1998.
10 Reiss, 1986.
11 Tripplet, Gainey and Sun, 2003.
12 Villareal & Silva, 2006.
13 Cerdá et al., 2008, p. 8.
14 Burgess, 1984 [1925].
15 Shaw and McKay, 2006 [1942].
16 Skogan, 1986.
17 Kasarda and Janowitz, 1974.
18 Sampson, 1991.
19 Taylor, 2001a.
20 See Bursik, 1986; Bursik and Webb, 1982.
21 Kane, 2005.
22 Skogan 1990, 1999.
23 Bursik and Grasmick, 1995.
24 Taylor, 2001a.
25 Snell, 2001.
26 Banco Mundial, 1999.
27 Sánchez and Núñez, 2007.
28 Llorente, Escobedo, Echandia & Rubio, 2001.
29 Localities are the city’s administrative units. There are currently 20 localities in Bogota.

30 Sánchez, Espinosa, and Rivas, 2007.
31 Formisano, 2002.
32 See Gibson, Zhao, Lovrich and Gaffney, 2002; Sampson, 1991; Sampson, Morenoff and Earls, 1999.
33 See Bellair, 1997; Carr, 2003; Lee and Bartkowski, 2004; Pattavina, Byrne and Garcia, 2006; Rosenfeld, Messner and Baumer, 2001; Rosenfeld, Baumer and Messner, 2007; Saegert, Winkel and Swartz, 2002; Saegert and Winkel, 2004; Sampson and Groves, 1989; Sampson, Raudenbush and Earls, 1997; Sampson and Raudenbush, 1999; Simons, Simons, Burt, Brody and Cutrona, 2005; Skogan, 1986; Snell, 2001; Taylor, 2001b; Velez, 2001.
34 See Bursik, 1999; Kubrin and Weitzer, 2003a; Lee and Ousey, 2005; Sampson et al., 1997; Sampson, 2002; Stucky, 2003.
35 See Belnar, Cerda, Roberts and Buska, 2008; Carr, 2003; Pattavina et al., 2006; Stucky, 2003; Taylor 2001a; Velez, 2001.
36 See Velez, 2001
37 See Lee and Ousey, 2005.
38 See Carr, 2003; Pattavina, 2006; Warner, 2007.
39 Bursik & Grasmick, 1995; Carr, Napolitano and Keating, 2007; MacDonald and Stokes, 2006.
40 Bursik & Grasmick, 1995.
41 Browning, Feinberg and Dietz, 2004, p. 510.
42 Patillo, 1998, p. 748.
43 Kubrin and Weitzer, 2003b.
44 Patillo, 1998. See also Taylor 2001a.
45 Arias, 2006, p. 303.
46 Population 15 years old and older. It excludes full-time students, housewives, retirees, disabled people who cannot work, and people in other situations.

47 One locality (Sumapaz) was excluded from the analysis because it is a rural area that was just recently incorporated into the political-administrative structure of the city.

48 See Haining, 2003, and Kubrin & Weitzer, 2003b for a similar approach.

49 Haining, 2003, p. 184.
50 The average neighborhood population size is about 12.000, thus, 10.000 seems like a reasonable base to standardize the data.

51 Population size for 2003 and 2004 was estimated by creating a shrinkage factor of the 2005 census population based on the locality’s reported population growth for those years.

52 Mears and Bhati, 2006:251. See Baller et al., 2001; Blau and Blau, 1982; Messner et al., 2002; Rosenfeld et al., 2001.
53 Pridemore & Grubesic, 2011.
54 See Cohen and Tita, 1999; Fagan and Davis, 2004; Fagan, Wilkinson and Davies, 2007; Kubrin and Weitzer, 2003b; Mears and Bhati, 2006; Sampson, 2003.
55 Note that the rate is inflated due to the summing of three years.

56 See Browning et al., 2004; Kubrin & Weitzer, 2003b; Sampson & Raudenbush, 1999.
57 Output available upon request.

58 Bursik, 1986, p. 59.
59 Skogan, 1999.
60 See Taylor, 2001a; Sampson and Raudensbush, 1999, 2001.
61 Velez, 2001.
62 Taylor, 2001a.
63 Stucky, 2003.
64 Carr, 2003.
65 Belnar et al., 2008.
66 The CAI are police units with jurisdiction over smaller areas than the precincts, and are aimed at providing a faster response to citizen complaints.

67 Serralvo, 2011.
68 See Baller et al. 2001; Cohen & Tita 1999; Kubrin & Weitzer 2003; Mears & Bhati 2006; Messner, Anselin, Baller, Hawkins, Deane & Tolnay, 1999.
69 Messner et al., 1999.
70 Anselin, 1996.
71 Loftin & Ward, 1983.
72 Ward & Gleditsch, 2008.
73 Baller et al., 2001, p. 566.
74 Anselin & Bera, 1998.
75 Baller et al., 2001.
76 Anselin, Syabri & Kho, 2006; Baller et al., 2001.
77 Messner et al., 1999.
78 See Pridemore, 2011; Rosenfeld, Bray, and Egley, 1999.
79 Kubrin and Weitzer, 2003b, p. 169.
80 Mears and Bhati, 2006.
81 Warner, 2003.
82 See Formisano, 2002; Llorente et al. 2001a, 2001b; Sánchez and Núñez, 2007.
83 Sánchez, Espinosa, and Rivas, 2007.
84 Brantingham and Brantingham, 1993.
85 Rosenfeld, Bray, and Egley, 1999.
86 Pridemore, 2011.


ANSELIN, Luc & BERA, Anil K. Spatial dependence in linear regression models with an introduction to spatial econometrics. En ULLAH, Aman & GILES, David E.A. (Eds.) Handbook of applied economic statistics. New York, NY, Marcel Dekker. 1998. p. 237-289. ISBN 978-0585157665.

ANSELIN, Luc. The Moran scatterplot as an ESDA tool to assess local instability in spatial association. En FISCHER, Manfred M.; SCHOLTEN, Henk J. & UNWIN, David J. (Eds.) Spatial analytical perspectives on GIS: GISDATA 4. London, Taylor & Francis.1996. p 111-125. ISBN 978-0748403394.

ARIAS, Enrique Desmond. The dynamics of criminal governance: networks and social order in Rio de Janeiro. Journal of Latin American Studies. 38(2): 293-325, 2006. ISSN 1469-767X. DOI: 10.1017/S0022216X06000721.

BALLER, Robert D.; ANSELIN, Luc; MESSNER, Steven F.; DEANE, Glenn; & HAWKINS, Darnell F. Structural covariates of U.S. county homicide rates: incorporating spatial effects. Criminology. 39(3): 561-590, 2001. ISSN 0011-1384. DOI: 10.1111/j.1745-9125.2001.tb00933.x

BELLAIR, Paul E. Social interaction and community crime: examining the importance of neighbor networks. Criminology. 35(4): 677-703, 1997. ISSN 0011-1384. DOI: 10.1111/j.1745-9125.1997.tb01235.x

BELNAR, Beth. E.; CERDA, Magdalena; ROBERTS, Andrea L. & BUKA, Stephen L. Effects of neighborhood resources on aggressive and delinquent behaviors among urban youths. American Journal of Public Health. 98(6): 1086-1093, 2008. ISSN 0090-0036. DOI: 10.2105/AJPH.2006.098913

BLAU, Judith R. & BLAU, Peter M. The cost of inequality: metropolitan structure and violent crime. American Sociological Review. 47(1): 114-129, 1982. ISSN 00031224. DOI: 10.2307/2095046

BRANTINGHAM, Patricia L. & BRANTINGHAM, Paul J. Nodes, paths, and edges: considerations on the complexity of crime and the physical environment. Environmental Psychology. 13: 3-28, 1993. ISSN 02724944. DOI: 10.1016/S0272-4944(05)80212-9.

BROWNING, Christopher R.; FEINBERG, Seth L. & DIETZ, Robert D. The paradox of social organization: networks, collective efficacy, and violent crime in urban neighborhoods. Social Forces. 83(2): 503-534, 2004. ISSN 0037-7732. DOI: 10.1353/sof.2005.0006.

BURGESS, Ernest W. The growth of the city: an introduction to a research project. In: PARK, Robert & Burgess, ERNEST W. (Eds.). The City. Chicago, IL, University of Chicago Press. 1984[1925]. p. 47-62. ISBN 0-226-64611-4.

BURSIK Jr., Robert J. The informal control of crime through neighborhood networks. Sociological Focus. 32(1): 85-97, 1999. ISSN 0038-0237.

BURSIK Jr., Robert J. Neighborhood-based networks and the control of crime and delinquency. En: BARLOW, Hugh D. (Ed.) Crime and Public Policy: Putting Theory to Work. Boulder, CO, Westview Press. 1995. p 107-130, ISBN 978-0813326788.

BURSIK Jr., Robert J. & GRASMICK, Harold G. Neighborhoods and crime: the dimensions of effective community control. New York, Lexington. 1993. 226 p. ISBN 978-0669246322.

BURSIK Jr., Robert J. Ecological stability and the dynamics of delinquency. En: REISS Jr., Albert J. & TONRY, Michael (eds.). Communities and Crime. Chicago, IL, The University of Chicago Press. 1986. p. 35-66. ISBN 9780226807980.

BURSIK Jr., Robert J. & WEBB, Jim. Community change and patterns of delinquency. The American Journal of Sociology. 88(1): 24-42, 1982. ISSN 0002-9602. DOI: 10.1086/227632.

CARR, Patrick J.; NAPOLITANO, Laura & KEATING, Jessica. We never call the cops and here is why: a qualitative examination of legal cynicism in three Philadelphia neighborhoods. Criminology. 45(2): 445-480, 2007. ISSN 0011-1384. DOI: 10.1111/j.1745-9125.2007.00084.x.

CARR, Patrick J. The New Parochialism: The implications of the beltway case for arguments concerning informal social control. American Journal of Sociology. 108(6): 1249-1291, 2003. ISSN 0002-9602. DOI: 10.1086/377517.

CASAS, Pablo & GONZÁLEZ, Paola. Políticas de seguridad y reducción del homicidio en Bogotá: mito y realidad. [En línea]. En: Casas, Pablo (Ed.) Seguridad Urbana y Policía en Colombia: Ensayos de Seguridad y Democracia. Fundación Seguridad y Democracia. 2005. Disponible en:

CERDÁ, Magdalena; MORENOFF, Jeffrey; DUQUE, Luis Fernando & BUKA, Stephen L. The limits of collective efficacy: investigating the influence of neighborhood context on levels of violence in Medellin and Chicago. [En línea]. Paper presented at the annual meeting of the American Sociological Association, Sheraton Boston and the Boston Marriott Copley Place, Boston, MA. 2008. Disponible en:

COHEN, Jacqueline & TITA, George. Diffusion in homicide: exploring a general method for detecting spatial diffusion processes. Journal of Quantitative Criminology. 15(4): 451-493, 1999. ISSN 07484518. DOI: 10.1023/A:1007596225550.

FAGAN, Jeffrey; WILKINSON, Deanna L. & DAVIES, Garth. Social contagion of violence. En: FLANNERY, Daniel J.; VAZSONYI, Alexander T. & WALDMAN, Irwin D. (Eds.) The Cambridge Handbook of Violent Behavior and Aggression. Cambridge University Press. 2007. p. 688-724. ISBN 978-0-521-60785-8.

FAGAN, Jeffrey & DAVIES, Garth. The natural history of neighborhood violence. Journal of Contemporary Criminal Justice. 20(2): 127-147, 2004. ISSN 1043-9862. DOI: 10.1177/1043986204263769.

FORMISANO, Michel. Econometría espacial: características de la violencia homicida en Bogotá. [En línea]. Documentos CEDE 2002-10. ISSN 1657-7191. Disponible en:

GIBSON, Chris L.; ZHAO, Jihong; LOVRICH, Nicholas P. & GAFFNEY, Michael (2002). Social integration, individual perceptions of collective efficacy, and fear of crime in three cities. Justice Quarterly. 19(3): 537-564, 2002. ISSN 0741-8825. DOI: 10.1080/07418820200095341.

HAINING, Robert. Spatial Data Analysis: Theory and Practice. New York, NY, Cambridge University Press. 2003. 432 p. ISBN 978-0521774376.

HEITGERD, Janet L. & BURSIK Jr., Robert J. Extracommunity dynamics and the ecology of delinquency. [En línea]. The American Journal of Sociology. 92(4): 775-787, 1987. ISSN 0002-9602. Disponible en:

KANE, Robert J. Compromised police legitimacy as a predictor of violent crime in structurally disadvantaged communities. Criminology. 43(2): 469-498, 2005. ISSN 0011-1384. DOI: 10.1111/j.0011-1348.2005.00014.x.

KASARDA, John D. & JANOWITZ, Morris. Community attachment in mass society. American Sociological Review. 39(3): 328-339, 1974. ISSN 0003-1224. DOI: 10.2307/2094293.

KRIVO, Lauren J.; PETERSON, Ruth D.; RIZZO, Helen & REYNOLDS, John R. Race, segregation, and the concentration of disadvantage: 1980-1990. Social Problems. 45(1): 61-80, 1998. ISSN 0037-7791. DOI: 10.1525/sp.1998.45.1.03x0157a.

KUBRIN, Charis E. & WEITZER, Ronald. New directions in social disorganization theory. Journal of Research in Crime and Delinquency. 40(4): 374-402, 2003a. ISSN 0022-4278. DOI: 10.1177/0022427803256238.

KUBRIN, Charis E. & WEITZER, Ronald. Retaliatory homicide: concentrated disadvantage and neighborhood culture. Social Problems. 50(2): 157-180, 2003b. ISSN 0037-7791. DOI: 10.1525/sp.2003.50.2.157.

LAND, Keneth C.; MCCALL, Patricia L. & COHEN, Lawrence E. Structural covariates of homicide rates: are there any invariances across time and space? American Journal of Sociology. 95: 922-963, 1990. ISSN 0002-9602. DOI: 10.1086/229381.

LEE, Mathew R. & OUSEY, Graham C. Institutional access, residential segregation, and urban black homicide. Sociological Inquiry. 75(1): 31-54, 2005. ISSN 0038-0245. DOI: 10.1111/j.1475-682X.2005.00111.x.

LEE, Mathew R. & BARTKOWSKI, John. Civic participation, regional subcultures, and violence: the differential effects of secular and religious participation on adult and juvenile homicide. Homicide Studies. 8(1): 5-39, 2004. ISSN 0031-3599. DOI: 10.1177/1088767903256528.

LLORENTE, María Victoria; ESCOBEDO, Rodolfo; ECHANDÍA, Camilo & RUBIO, Mauricio. Violencia homicida y estructuras criminales en Bogotá. [En línea]. Bogotá, Centro de Estudios sobre Desarrollo Económico, Universidad de Los Andes. 2001. Paz Pública – Documento de Trabajo No. 20. Disponible en:

LOFTIN, Colin & WARD, Sally K. A spatial autocorrelation model of the effects of population density on fertility. American Sociological Review. 48(1): 121-128, 1983. ISSN 0003-1224. URL:

MACDONALD, John & STOKES, Robert. Race, social capital, and trust in the police. Urban Affairs Review. 41(3): 358-375, 2006. ISSN 1078-0874. DOI: 10.1177/107808740528170.

MCGAHEY, Richard M. Economic conditions: neighborhood organization, and urban crime. En REISS Jr., Albert J. & TONRY, Michael (eds.). Communities and Crime. Chicago, IL, The University of Chicago Press. 1986. p 231-270. ISBN 978-0226807980.

MEARS, Daniel P. & BHATI, Avinash S. No community is an island: the effects of resource deprivation on urban violence in spatially and socially proximate communities. Criminology. 44(3): 509-548, 2006. ISSN 0011-1383. DOI: 10.1111/j.1745-9125.2006.00056.x.

MESSNER, Steven F.; RAFFALOVICH, Lawrence E. & SHROCK Peter. Reassessing the cross-national relationship between income inequality and homicide rates: implications of data quality control in the measurement of income distribution. Journal of Quantitative Criminology. 18(4): 377-395, 2002. ISSN 0748-4518. DOI: 10.1023/A:1021169610837.

MESSNER, Steven F.; ANSELIN, Luc; BALLER, Robert D.; HAWKINS, Darnell F.; DEANE, Glenn & TOLNAY, Stewart E. The spatial patterning of county homicide rates: an application of exploratory spatial data analysis. Journal of Quantitative Criminology. 15(4): 423-450, 1999. ISSN 0748-4518. DOI: 10.1023/A:1007544208712.

PATILLO, Mary E. Sweet mothers and gangbangers: managing crime in a black middle-class neighborhood. Social Forces. 76(3): 747-774, 1998. ISSN 0037-7732. DOI: 10.1093/sf/76.3.747.

PATTAVINA, April; BYRNE, James & GARCIA, Luis. An examination of citizen involvement in crime prevention in high-risk versus low- to moderate-risk neighborhoods. Crime & Delinquency. 52(2): 203-231, 2006. ISSN 0011-1287. DOI: 10.1177/001112870528415.

PRIDEMORE, William & GRUBESIC, T.H. Alcohol outlets and community levels of interpersonal violence: spatial density, outlet type, and seriousness of assault. Journal of Research in Crime and Delinquency. May 17, 2011. ISSN: 1552-731X DOI: 10.1177/0022427810397952.

REISS Jr., Albert J. & TONRY, Michael (eds.). Communities and Crime. Chicago, IL, The University of Chicago Press. 1986. 430 p. ISBN 978-0226807980.

ROSENFELD, Richard; BAUMER, Eric & MESSNER, Steven F. Social trust, firearm prevalence, and homicide. Annals of Epidemiology. 17(2): 119-125, 2007. ISSN 1047-2797. DOI: 10.1016/j.annepidem.2006.07.016.

ROSENFELD, Richard; MESSNER, Steven F. & BAUMER, Eric. Social capital and homicide. Social Forces. 80(1): 283-309, 2001. ISSN 0037-7732. DOI: 10.1353/sof.2001.0086.

ROSENFELD, Richard; BRAY, Timothy M. & EGLEY, Arlen. Facilitating violence: a comparison of gang-motivated, gang-affiliated, and nongang youth homicides. Journal of Quantitative Criminology. 15(4): 495-516, 1999. ISSN 0748-4518. DOI: 10.1023/A:1007548309620.

SAEGERT, Susan & WINKEL, Gary. Crime, social capital, and community participation. American Journal of Community Psychology. 34(3/4): 219-233, 2004. ISSN 0091-0562. DOI: 10.1007/s10464-004-7416-2.

SAEGERT, Susan; WINKEL, Gary & SWARTZ, Charles. Social capital and crime in New York City's low-income housing. Housing Policy Debate. 13(1): 189-226, 2002. ISSN 1051-1482. DOI: 10.1080/10511482.2002.9521439.

SAMPSON, Robert J. The neighborhood context of well-being. Perspectives in Biology and Medicine. 46(3): S53-S64, 2003. ISSN 0031-5982. DOI: 10.1353/pbm.2003.0073

SAMPSON, Robert J. Organized for what? recasting theories of social (dis)organization. En WARING, Elin. & WEISBURD, David (Eds.) Crime & Social Organization: Advances in Criminological Theory Volume 10. New Brunswick, NJ, Transaction Publishers. 2002. p 95-110, ISBN 9780765800640.

SAMPSON, Robert J.& RAUNDEBUSH, Stephen W. Disorder in Urban Neighborhoods—Does It Lead to Crime? Research in Brief. [En línea]. Washington, D.C., National Institute of Justice: 6 p. Disponible en:

SAMPSON, Robert J.; MORENOFF, Jeffrey D. & EARLS, Felton. Beyond social capital: spatial dynamics of collective efficacy for children. [En línea]. American Sociological Review. 64(5): 633-660, 1999. ISSN 0003-1224. Disponible en:

SAMPSON, Robert J. & RAUNDEBUSH, Stephen W. Systematic social observation of public spaces: a new look at disorder in urban neighborhoods. American Journal of Sociology. 105(3): 603-651, 1999. ISSN 0002-9602. DOI: 10.1086/210356.

SAMPSON, Robert J.; RAUDENBUSH, Stephe W. & EARLS, Felton. Neighborhoods and violent crime: a multilevel study of collective efficacy. Science. 277(5328): 918-924, 1997. ISSN 0036-8075. DOI: 10.1126/science.277.5328.918.

SAMPSON, Robert J. & WILSON, William Julius. Toward a theory of race, crime, and urban inequality. En: HAGAN, John & PETERSON, Ruth (Eds.). Crime and Inequality. Stanford, CA, Stanford University Press. 1993. p. 37-54. ISBN 978-0804724777.

SAMPSON, Robert J. Linking the micro- and macrolevel dimensions of community social organization. Social Forces. 70(1): 43-64, 1991. ISSN 0037-7732. DOI: 10.2307/2580061.

SAMPSON, Robert J. & GROVES, W. Byron. Community structure and crime: testing social-disorganization theory. The American Journal of Sociology. 94(4): 774-802, 1989. ISSN 0002- 9601. DOI: 10.1086/229068.

SAMPSON, Robert J. Crime in cities: the effects of formal and informal social control. En REISS Jr., Albert J. & TONRY, Michael (eds.). Communities and Crime. Chicago, IL., The University of Chicago Press. 1986. p. 271-311. ISBN 978-0226807980.

SÁNCHEZ, Fabio; ESPINOSA, Silva & RIVAS, Ángela. ¿Garrote o zanahoria? factores asociados a la disminución de la violencia homicida y el crimen en Bogotá, 1993-2002. En: SÁNCHEZ, Fabio (ed.). Las cuentas de la violencia: ensayos económicos sobre el crimen y el conflicto. Colombia, Grupo Editorial Norma. 2007. p 301 – 355, ISBN 978-958-45-0324-4.

SÁNCHEZ, Fabio & NÚÑEZ, Jairo. Determinantes del crimen violento en un país altamente violento: el caso de Colombia. En SÁNCHEZ, Fabio (ed.). Las cuentas de la violencia: ensayos económicos sobre el crimen y el conflicto. Colombia, Grupo Editorial Norma. 2007. p. 25-61. ISBN 978-958-45-0324-4.

SERRALVO, José. Internal displacement, land restoration, and the ongoing conflict in Colombia. [En línea]. Journal of Humanitarian Assistance. Feinstein International Center, Tufts University, 2011. Disponible en:

SHAW, Clifford R. & MCKAY, Henry D. Juvenile delinquency and urban areas. En: CULLEN, Francis T. & AGNEW, Robert. Criminological Theory: Past to Present. 3rd ed. Los Angeles, CA, Roxbury. 2006. p 95-108. ISBN 1-931719-63-2.

SIMONS, Ronald L.; SIMONS, Leslie G.; BURT, Callie H; BRODY, Gene H. & CUTRONA, Carolyn. Collective efficacy, authoritative parenting and delinquency: a longitudinal test of a model integrating community – and family – level processes. Criminology. 43(4): 989-1029, 2005. ISSN 0011-1384. DOI: 10.1111/j.1745-9125.2005.00031.x.

SKOGAN, Wesley. Measuring what matters: crime, disorder and fear. [En línea]. En: LANGWORTHY, Robert H. (Ed.) Measuring What Matters: Proceedings From the Police Research Institute Meetings. Washington DC, U.S. Department of Justice, National Institute of Justice and Office of Community Oriented Policing Services, NCJ 170610 37-53, 1999. Disponible en:

SKOGAN, Wesley. Disorder and decline: crime and the spiral of decay in American neighborhoods. Berkeley, CA, University of California Press. 1990. 219 p. ISBN 0-520-07693-1.

SKOGAN, Wesley. Fear of crime and neighborhood change. In REISS Jr., Albert J. & TONRY, Michael (eds.). Communities and Crime. Chicago, IL, The University of Chicago Press. 1986. p. 203-229. ISBN 978-0226807980.

SMITH, Douglas A. Does gentrification affect crime rates? En REISS Jr., Albert J. & TONRY, Michael (eds.). Communities and Crime. Chicago, IL, The University of Chicago Press.1986. p. 163-202. ISBN 978-0226807980.

SNELL, Clete. Neighborhood structure, crime, and fear of crime: testing Bursik and Grasmick's neighborhood control theory. New York, NY, LFB Scholarly Publishing LLC. 2001. 159 p. ISBN 9781931202077.

STRETESKY, Paul B.; SCHUCK, Amie M., & HOGAN, Michael J. Space matters: an analysis of poverty, poverty clustering, and violent crime. Justice Quarterly. 21(4): 817-841, 2004. ISSN 0741-8825. DOI: 10.1080/07418820400096001.

STUCKY, Thomas D. Local politics and violent crime in U.S. cities. Criminology. 41(4): 1101-1135, 2003. ISSN 0011-1384. DOI: 10.1111/j.1745-9125.2003.tb01015.x.

TAYLOR, Ralph B. Breaking Away from Broken Windows: Baltimore Neighborhoods and the Nationwide Fight Against Crime, Grime, Fear, and Decline. Boulder, Colorado, Westview Press. 2001a. 330 p. ISBN 978-0813397580.

TAYLOR, Ralph B. The ecology of crime, fear, and delinquency. In PATERNOSTER, Raymond & BACHMAN, Ronet (eds.), Explaining Criminals and Crime: Essays in Contemporary Criminological Theory. New York, NY, Roxbury Publishing Company. 2001b. p. 124-139, ISBN 978-1891487323.

TAYLOR, Ralph B. Social order and disorder of street blocks and neighborhoods: ecology, microecology and the systemic model of social organization. Journal of Research in Crime and Delinquency. 34(1): 113-155, 1997. ISSN 0022-4278. DOI: 10.1177/0022427897034001006.

TRIPLETT, Ruth A.; GAINEY, Randy R. & SUN, Ivan Y. Institutional strength, social control and neighborhood crime rates. Theoretical Criminology. 7(4): 439-467, 2003. ISSN 1362-4806. DOI: 10.1177/13624806030074003.

VELEZ, Maria B. The role of public social control in urban neighborhoods: a multi-level analysis of victimization risk. Criminology. 39(4): 837-864, 2001. ISSN 0011-1384. DOI: 10.1111/j.1745-9125.2001.tb00942.x.

VILLARREAL, Andrés & SILVA, Braulio F.A. Social cohesion, criminal victimization and perceived risk of crime in Brazilian neighborhoods. Social Forces. 84(3): 1725-1753, 2006. ISSN 0037-7732. DOI: 10.1353/sof.2006.0073.

WARD, Michael D. & GLEDITSCH, Kristian S. Spatial Regression Models. Thousand Oaks, CA, Sage Publications. 2008. p. 99. ISBN 978-1412954150. Series: Quantitative Applications in Social Sciences

WARNER, Barbara. Directly intervene or call the authorities? A study of forms of neighborhood social control within a social disorganization framework. Criminology. 45(1): 99-129, 2007. ISSN 0011-1384. DOI: 10.1111/j.1745-9125.2007.00073.x.

WORLD Bank. Violence in Colombia: Building Sustainable Peace and Social Capital. Washington, DC, World Bank. 1999. 56 p. ISBN 978-0821343838.


Received: 26.10.11
Accepted: 23-03-12