You are on page 1of 20

Global Crime

ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/fglc20

The spatial diffusion of homicide in Mexico City: a


test of theories in context

Carlos Vilalta, Pablo Lopez-Ramirez & Gustavo Fondevila

To cite this article: Carlos Vilalta, Pablo Lopez-Ramirez & Gustavo Fondevila (2021): The
spatial diffusion of homicide in Mexico City: a test of theories in context, Global Crime, DOI:
10.1080/17440572.2021.1909480

To link to this article: https://doi.org/10.1080/17440572.2021.1909480

Published online: 08 Apr 2021.

Submit your article to this journal

Article views: 58

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


https://www.tandfonline.com/action/journalInformation?journalCode=fglc20
GLOBAL CRIME
https://doi.org/10.1080/17440572.2021.1909480

ARTICLE

The spatial diffusion of homicide in Mexico City: a test of


theories in context
Carlos Vilalta , Pablo Lopez-Ramirez and Gustavo Fondevila
Centre for Research in Geospatial Information Sciences (Centrogeo), Mexico City, Mexico; Center of Research
and Economic Teaching, Mexico, Mexico

ABSTRACT ARTICLE HISTORY


Homicidal violence has increased substantially in Mexico City in Received 26 July 2020
recent years. In this regard, we ask three questions: First, is there Accepted 24 March 2021
a contagious spread of this violence across neighbourhoods? KEYWORDS
Second, does it spread in association with drug market activity Homicide; drug crimes;
among local criminal organisations? Third, does it spread to neigh­ organised crime; spatial
bourhoods characterised by concentrated disadvantage, disorder, diffusion; mexico
and crime opportunity? Using homicide data aggregated at the
neighbourhood level, we found the contagious spread of homicidal
violence in neighbourhoods already troubled with drug dealing
crimes and concentrated disadvantage. Based on our findings, we
propose that while some theories are able to explain the spatial
clustering of homicide, only social disorganisation theory is capable
of predicting its spatial diffusion. Furthermore, we argue that
advances in criminological theory require the testing of ad-hoc
correlates when studying the Latin American context.

Introduction
In this study we tested three complementary hypotheses with regards to the spatial
diffusion of homicidal violence in Mexico City. First, we tested whether there was
a contagious spread of homicide across neighborhoods. The simplest of the spatial
diffusion hypotheses predicts that fluctuations in homicidal violence in one neighbor­
hood will affect homicide fluctuations in adjacent neighborhoods. Second, we tested for
the contagious spread of homicide to neighborhoods more susceptible to drug market
activity between organized crime groups. Although this hypothesis has been partially
confirmed for the case of aggravated assault crimes in the U.S. (Contreras & Hipp, 2020),
this is an ad-hoc hypothesis for the case of Mexico City which is explained in the next
section. Third, we tested whether homicide spread to more structurally criminogenic
neighborhoods characterized by high levels of social disorganization, broken windows,
signs of disorder, and crime pattern covariates.
We tested these hypotheses using homicide data for the 2016–2019 period in Mexico City
neighbourhoods. We present evidence that homicidal violence has spread to neighbour­
hoods troubled with drug dealing crimes and concentrated disadvantage, yet

CONTACT Carlos Vilalta cvilalta@centrogeo.edu.mx Centre for Research in Geospatial Information Sciences
(Centrogeo), Contoy 137, Lomas De Padierna, Tlalpan, Mexico City, 14240 Mexico
© 2021 Informa UK Limited, trading as Taylor & Francis Group
2 C. VILALTA ET AL.

independently of signs of disorder and crime opportunities. In other words, the spatial
diffusion of homicide is likely to happen in concurrence with drug market competition
among criminal organisations operating in disadvantage neighbourhoods.
These findings are important for two main reasons. First, because they show that social
disorganisation, broken windows and crime pattern theories are able to explain the
spatial clustering of homicidal violence, but only social disorganisation, and more speci­
fically, concentrated disadvantage, can explain the spatial diffusion process behind this
crime. This finding alone offers a shift from the traditional focus of studying homicide
change independently of its spatial spillover effects. Second, the experience of Latin
American cities is largely absent from criminological theory and research. However,
much of the homicidal violence problem is, and is likely to continue to be, a Latin
American problem (Vilalta & Muggah, 2020). Criminological theory thus clearly needs to
be empirically tested and expanded based on the Latin American case. In this sense, while
we do not update any theory, we present one novelty regarding the Mexican context,
namely, that homicidal violence cannot be studied independently of drug dealing crimes.

Homicidal violence in Mexico City


A recent intelligence report by the Mexican Ministry of Citizen Security (SSC) explains that
homicide in Mexico City is driven by three main factors: organised crime conflicts, fights
and alcohol misuse, and armed robbery (El Universal, 2020)1. The first factor is not
a structural factor but a contingency that creates a context for extreme violence which
some neighbourhoods have more difficulty avoiding than others.
The aforementioned analysis estimates that 65% of homicides recorded between
December 2018 and December 2019 in Mexico City were executions (El Universal,
2020). Over the years, organised crime in Mexico City has become more violent and
more public. Current criminal organisations in Mexico City either have ties with or are
implants of older and more powerful organisations outside the city. Among the leading
criminal organisations in Mexico City are ‘La Union de Tepito’, initially formed by members
of the ‘Familia Michoacana’ and ‘Beltran Leyva’ organisations that joined to counter the
presence of other minor organisations linked to the ‘Zetas’, the ‘Sinaloa Cartel’, and the
‘Fuerza Anti-Union’, with links to the ‘Jalisco Nueva Generacion’ cartel. Rising competition
between criminal organisations has been accompanied by armed clashes, leading to
a geography of homicide associated with drug market operations and retaliatory violence.
The same analysis reports that fights, often in connection with alcohol misuse, are
the second driver of homicide in the city. About 14% of homicides can be linked to fights
caused by arguments between neighbours, traffic arguments, family violence, or personal
revenge. A quick look at the 2019 Mexican national victimisation and public security percep­
tion (ENVIPE) indicates that 21% of victims of crime in Mexico City reported that their attackers
were under the influence of alcohol, illegal drugs or both at the time of the crime.
About 10% of homicides were related to armed robbery. Indeed, the same ENVIPE
survey indicates that 78% of all victims of robbery in Mexico City reported that their
attacker was armed with a gun. Also, 20% of all victims of robbery reported that their
attacker was both intoxicated (either by alcohol or drugs or both) and carrying a firearm.
The remaining 11% of homicides in the city were related to killings by stray bullets (1%)
and other unspecified causes.
GLOBAL CRIME 3

The spatial diffusion of homicidal violence


There is wide cross-national evidence that homicidal violence spatially clusters in
a reduced number of neighbourhoods (Dugato, Calderoni, Berlusconi, 2020; Fagan,
Wilkinson, Davies, 2007; Giménez-Santana, Caplan, Drawve, 2018; Lauritsen & Lentz,
2019; Papachristos & Bastomski, 2018; Pereira, Mota, Andresen, 2017; Smith & Sandoval,
2019; Valencia & Sanz, 2017; Valente, 2019; Vilalta, Lopez, Fondevila, Siordia, 2020; Vilalta,
Muggah, Fondevila, 2020; Vilalta & Muggah, 2014). There is also consensus that homicide
dynamics are subject to local structural forces (Culyba et al., 2016; Fagan, Wilkinson,
Davies, 2007; Ferrandino, 2018; Ingram & Marchesini da Costa, 2017; Light & Harris,
2012; Messner & Anselin, 2004; Vilalta & Muggah, 2014) as well as to contingencies,
such as a new drug epidemic, disputes between drug trafficking organisations, changes
in police strategy, or the arrival of paramilitary groups in the city (Vânia Ceccato, Haining,
Kahn, 2007; Fagan, Wilkinson, Davies, 2007; Flores & Rodriguez-Oreggia, 2014; Flores &
Villarreal, 2015; Muggah & Tobón, 2018; Vilalta & Muggah, 2014).
Furthermore, evidence shows that homicide spreads spatially with diffusion properties
(Blumstein, Rivara, Rosenfeld, 2000; Fagan, Wilkinson, Davies, 2007; Loeffler & Flaxman,
2018; Loftin, 1986). Case studies evidence the contagious spread of homicidal violence
within a city (Loeffler & Flaxman, 2018) and across counties (Messner et al., 1999), census
tracts (J. Cohen & Tita, 1999), and even municipalities in the specific case of Mexico (Flores
& Villarreal, 2015). Importantly, Contreras and Hipp (2020) found that in Miami-Dade, drug
arrests on a census block in one month increased aggravated assault incidents in sur­
rounding blocks the next month. Others find that homicidal violence spreads to more
affluent counties, while more rural counties serve as barriers against the diffusion of
homicides (Messner et al., 1999). In Mexico, Flores and Villarreal (2015) speculated that
homicide spread across municipalities may be due to clashes between drug cartels or
groups within the cartels, particularly after the deployment of federal armed forces to
combat these organizations. These studies find both the contagious and hierarchical
spread of homicide, that is, as a function of distance and through shared influences.
One such shared influence is gang violence, which induces the spread of homicidal
violence as a function of drug dealing crimes (J. Cohen & Tita, 1999; Papachristos, 2009;
Papachristos, Hureau, Braga, 2013; G. E. Tita & Radil, 2011).
In this line of research, it is argued that the spatial diffusion of homicidal violence occurs
through social contact (Fagan, Wilkinson, Davies, 2007; Loftin, 1986). Zeoli et al. (2014)
found that gang homicide clusters evolved in one area of the city and spread to other
areas over time as a function of the contagious diffusion of behaviours rather than the
movement of offenders. Another case study in Chicago found that murders spread
through an epidemic-like process of social diffusion, as gangs evaluated the highly visible
actions of others in their local networks (Papachristos, 2009). The same is true for adult
and adolescent gun violence (Bond & Bushman, 2017; Tracy, Braga, Papachristos, 2016) –
adolescents learn aggressive and violent behaviours either by direct experience and/or by
observing others and imitating their behaviour.
Previous studies also report that a high proportion of offenders commit homicides
outside their neighbourhood (G. Tita & Griffiths, 2005) while others show an even
distribution between own neighbourhood homicides and outside neighbourhood homi­
cides (Groff & McEwen, 2006). In this regard, some studies suggest distinguishing between
4 C. VILALTA ET AL.

types of homicides in terms of victims’ and suspects’ journeys – the victims and suspects
of gang homicides travel the longest distances, whereas suspects in domestic homicides
travel the shortest distances (Pizarro, Corsaro, Yu, 2007). In the same vein, previous studies
have found that offenders in general prefer to travel to areas of low social cohesion
(Johnson & Summers, 2015).
To sum up, it seems that the spatial diffusion of homicide may be due to (1) crime
opportunity differentials in connection with gang violence (e.g. new versus saturated
drug markets), (2) emulation and copycat crime (i.e. social diffusion processes), and/or (3)
mobility and networking of offenders (i.e. journey to crime).

Theoretical correlates of homicide


In addition to our ad hoc hypothesis of the contagious spread of homicidal violence in
neighbourhoods more susceptible to drug market activity, this study includes a test of three
criminological theories, namely, Social disorganisation, Broken windows, and Crime pattern
theory. We include them as alternative explanations as they are able to explain social processes
resulting in spatial variations of homicidal violence among other crimes. As such, we view the
impact of drug market activity on homicidal violence as a process that is not independent of
the social structural characteristics of neighbourhoods. Our purpose in including these the­
ories, is to integrate concepts and insights from different theories. The review that follows
focuses on the insights provided by these theories to the study of homicide.
Social disorganisation theory has always been difficult to conceptualise and operationalise
(Kubrin, 2009). Two obvious difficulties are the absence of data at the micro-place level as well
as measurement equivalencies across countries. However, previous studies have shown that
social disorganisation concepts are associated with concentrated disadvantage measures
affecting crime in similar ways (Gerstner, Wickes, Oberwittler, 2019; Graif & Sampson, 2009;
Sampson, Raudenbush, Earls, 1997; Sampson & Morenoff, 2004; Stein, Conley, Davis, 2016;
Ward, Nobles, Youstin, Cook, 2014). The empirical argument is that concentrated disadvantage
is a tangible environmental indicator of the likelihood of residents to take an active role in
preventing crime (Sampson, Raudenbush, Earls, 1997).
Measures of neighbourhood concentrated disadvantage (e.g. poverty, unemployment,
family income, literacy, welfare, marginality) have been associated to spatial variations of
homicide in Mexican cities (Cesar Fuentes & Sanchez, 2015; Merino & Fierro, 2016; Vilalta,
Lopez, Fondevila, Siordia, 2020; Vilalta, Muggah, Fondevila, 2020; Vilalta & Muggah, 2014).
Interestingly, some studies have also shown Broken windows theory measures spatially
associated with concentrated disadvantage (Gau & Pratt, 2010; Sampson & Raudenbush,
2004; Wilcox, Quisenberry, Cabrera, Jones, 2004). Broken windows theory states that run-
down public spaces and visible signs of disorder or incivilities left unattended, will lead to
more serious crime (Lanfear, Matsueda, Beach, 2020; O’Brien, Farrell, Welsh, 2019a; Wilson
& Kelling, 1982). Two central tenets of Broken windows theory are neighbourhood social
disorder and physical disorder. Previous tests of the theory have used measures of social
and physical disorders, such as drinking in the street, selling drugs, prostitution, shoot­
ings, fights, domestic violence, street lightning, unpainted crosswalks, lack of stop signs,
unmaintained vacant lots, empty beer bottles visible in the street, graffiti, litter, aban­
doned cars, drug paraphernalia, condoms on sidewalk and needles/syringes on sidewalk
(Culyba et al., 2016; Gau & Pratt, 2008; O’Brien, Farrell, Welsh, 2019b; O’Brien & Sampson,
GLOBAL CRIME 5

2015; Sampson & Raudenbush, 2004; Taylor, 2018; Vilalta, Lopez, Fondevila, Siordia, 2020;
Williams, Burnap, Sloan, 2017; Yang, 2010). Some studies have used principal component
analyses (PCA) and indexes to create measures of social disorder and physical disorder
(Gau & Pratt, 2008; 2010; O’Brien & Sampson, 2015; Sampson & Raudenbush, 1999; 2004;
Vilalta et al., 2019; Wilcox, Quisenberry, Cabrera, Jones, 2004; Yang, 2010).
With regard to empirical evidence, while some studies have found social disorder to
increase homicide (Alves, Lenzi, Mendes, Ribeiro, 2015; Culyba et al., 2016; César Fuentes,
2015; Patino, Duque, Pardo-Pascual, Ruiz, 2014; Rosenfeld, Fornango, Rengifo, 2007;
Vilalta et al., 2019), others have not found such a relationship (Cerdá et al., 2009;
Sampson & Raudenbush, 1999; 2001).
Crime pattern theory combines rational choice and routine activity theory to help
explain spatial variations of crime activity (Eck & Weisburd, 1995; Wilcox & Cullen, 2018).
Crime pattern theory predicts that land use characteristics create opportunities for crime
(Vânia Ceccato, Haining, Kahn, 2007; L. E. Cohen & Felson, 1979; McCord, Ratcliffe, Garcia,
Taylor, 2007; Vilalta, Castillo, Torres, 2016; Vilalta, Lopez, Fondevila, Siordia, 2020; Vilalta,
Muggah, Fondevila, 2020). Evidence shows that entertainment venues, bars and restau­
rants correlate with more homicide (Vânia Ceccato, Haining, Kahn, 2007; Henderson,
Morgan, Patel, Tiplady, 2005; Hohl et al., 2017; Swart, Seedat, Nel, 2018; Vilalta, Castillo,
Torres, 2016; 2019; 2020). Liquor outlets and convenience stores selling alcohol also seem
to impact violent crime (Norza, Vargas, Avendaño, Rincón, Ospino, 2018; Snowden,
Stucky, Pridemore, 2017; Valasik, Brault, Martinez, 2018) including homicide in Mexico
City (Vilalta, Lopez, Fondevila, Siordia, 2020; Vilalta, Muggah, Fondevila, 2020). Non-
residential land uses also correlates positively with homicide rates (Cesar Fuentes &
Sanchez, 2015). Likewise, mass transportation facilities, such as metro stations or bus
stops, with a concentration of high numbers of motivated offenders with suitable targets,
have been found to be statistically associated with homicidal violence at the census block
level (Vilalta, Muggah, Fondevila, 2020), although not at the neighbourhood level (Vilalta,
Lopez, Fondevila, Siordia, 2020).
Clearly, these theories have been able to explain the spatial clustering of homicidal
violence in cities across the world by providing different views of criminogenic conditions
of places. One need not subscribe to one particular theory to hypothesise some type of
spatial effect associated with the social processes explained by these theories. What
remains to be seen is whether any of these theories can help explain the spatial diffusion
of homicide.

Hypotheses
In this study, we propose the following hypotheses:

● The spatial diffusion by ‘proximity’ Hypothesis (H1): The geographically closer neigh­
bourhoods are, the more positively correlated time fluctuations in homicide will be.
Our expectation is that the spatial lag parameter, Rho, will be positive, indicating that
the closer the neighbourhoods, the more spatially dependent the year-to-year
fluctuations in homicide.
● The spatial diffusion by ‘drug market activity’ Hypothesis (H2): Homicide violence will
not spread to neighbourhoods randomly, but as a consequence of organised crime
6 C. VILALTA ET AL.

groups fighting for drug market territories. Our expectation is that the spatial lag
parameter, Theta, will be positive, indicating that the closer the neighbourhoods
with more drug dealing crimes, the more spatially dependent the year-to-year
fluctuations in homicide.
● The spatial diffusion by ‘criminogenic characteristics’ Hypothesis (H3): Spatial diffu­
sion effects will be strongest in neighbourhoods with similar criminogenic condi­
tions, measured by their levels of social disorganisation, signs of disorder, and crime
pattern opportunities. Our expectation is that the spatial lag parameters, Theta, will
be positive, indicating that the closer and more criminogenic the neighbourhoods,
the more spatially dependent the year-to-year change in the homicide rate.

Data and methods


Case study and units of analysis
The case study was in Mexico City. Years (N = 4) were the temporal units of analysis and
neighbourhoods (i.e. colonias) were the spatial unit of analysis (N = 1,824). The neighbour­
hood polygon spatial data was requested from the National Electoral Institute (INE)2. In
Mexico, a neighbourhood is a set of dwellings that constitute a settlement – it may be
constituted by one or more census blocks and does not have jurisdictional autonomy. As
such, they are often homogeneous in their socioecological structure and easily quantifi­
able. The average Mexico City neighbourhood is 0.47 square kilometres (0.18 square
miles) with a mean of 4,890 residents, according to the 2010 census. The benefit of
using neighbourhoods in crime analysis is that they provide a sense of place and
contextualise a broad range of social issues, such as collective efficacy and social disorder
processes (Vilalta, 2013).

Dependent variable
Criminal investigation data are managed by the local Office of the Attorney General of
Mexico City (Procuraduria General de Justicia de la Ciudad de Mexico). These data can be
downloaded directly from the Portal de Datos de la Ciudad de Mexico of Mexico City3. This
study used 4,510 criminal investigations for homicide (i.e. homicidio doloso) recorded
between January 2016 and December 2019, using geographic coordinates (point data) of
the incident. Homicide data were aggregated to neighbourhood level (polygon data) and
homicide rates were calculated per a population of 1,000 residents. Homicide rates (plus
a value of 1 to adjust zeros) were transformed to their natural logs to reduce skewness4.
Using a first-order queen-type neighbour matrix (i.e. strict contiguity matrix), the global
Moran’s I coefficient indicated significant positive spatial autocorrelation in the natural
logs of the homicide rates (I = 0.170, p = 0.001).

Independent variables
Hypothesis two (H2) states that homicide violence will spread across neighbourhoods as
a consequence of organised crime groups disputing drug markets. To represent drug
market activity, we used the number of criminal investigations for drug dealing crimes (i.e.
GLOBAL CRIME 7

narcomenudeo) also recorded between January 2016 and December 2019 and by geo­
graphic coordinates (point data) of the incident. Previous studies have used similar
measures (e.g. drug arrests) as indicators of local drug activity in city case studies
(Contreras & Hipp, 2020; Warner & Coomer, 2003) including Mexico City (Vilalta, 2009).
After georeferencing, drug dealing crime rates (plus a value of 1 to adjust zeros) were
also transformed to their natural logs to reduce skewness. The source of data is the same
as with the dependent variable. We followed the same georeferencing procedures as
above.
A number of census-based variables were used to test Hypothesis 3 (H3). H3 is the
spatial diffusion by ‘criminogenic potential’ hypothesis. Following previous studies, we
used a variety of measures representing the criminogenic potential of neighbourhoods.
First, a concentrated disadvantage single component was constructed using principal
component analysis (PCA) that included the 15–29 year old (%) population of
a neighbourhood, Native indigenous speaking population (%), population without health
insurance (%), population without elementary schooling (%), and the average number of
persons per bedroom as proxy for crowding conditions5. These 2010 census-based
variables have already been used to measure concentrated disadvantage as a proxy for
social disorganisation in previous studies based in Mexico City (Vilalta, Lopez, Fondevila,
Siordia, 2020; Vilalta, Muggah, Fondevila, 2020). The PCA statistics are given in the
Appendix (Table A1).
Second, neighbourhood signs of disorder were operationalised through two variables:
social disorder and physical disorder (Sampson and Raudenbush, 2004; Yang, 2010). PCA
were also used to simplify the complexity of the social disorder and physical disorder
concepts. The data source for these components was the Mexico City 911 emergency call
system. This approach has already been used to create component measures of disorder
and informal social control (O’Brien, Sampson, Winship, 2015; O’Brien, Farrell, Welsh,
2019a; O’Brien & Sampson, 2015). We identified four types of 911 incidents for the
construction of the social disorder component: public intoxication with drugs, public
intoxication with alcohol, street scandals, and loud partying; and five types of 911
incidents for the construction of the physical disorder component: graffiti signs, drinking
water leaks, sewage spills, gas leaks, and downed power lines. Similar components have
also been used to predict homicide crimes in previous Mexico City studies (Vilalta, Lopez,
Fondevila, Siordia, 2020; Vilalta, Muggah, Fondevila, 2020).
Third, we used PCA to create a crime opportunity index at the neighbourhood level.
This index was calculated based on data from the 2017 National Statistical Directory of
Economic Units (DENUE), the Mexico City Open Data (CDMX) source, and the
OpenStreetMaps data source. The DENUE provided the geographical location of bars,
restaurants, and convenience stores with alcohol sales permits while the CDMX provided
the location of the metro and metrobus stations. These data were aggregated at the
neighbourhood level. The OpenStreetMaps data source was used to create a street
permeability index which was also included in the crime opportunity index. The street
permeability index was obtained by the weighted sum of the streets entering each
neighbourhood. Different weights were used to represent the different street capacity,
where a greater weight (i.e. 2) represents a main avenue while a lesser weight (i.e. 1)
represents a residential street. The higher the value of the neighbourhood index, the
higher its street capacity. The permeability index and metro and metrobus transportation
8 C. VILALTA ET AL.

variables were included in the crime opportunity index for two reasons: (1) to proxy
daytime population concentration, and (2) because mass transit tends to attract and
generate crime as it gathers large crowds of people (Gallison, 2016). The alcohol avail­
ability variables (i.e. bars, restaurants, and convenience stores with alcohol sales permits)
were included in the crime opportunity index as previous studies have found a positive
spatial correlation with homicide in Mexico City (Vilalta, Lopez, Fondevila, Siordia, 2020;
Vilalta, Muggah, Fondevila, 2020).

Statistical analysis
One non-spatial and three spatial panel regression models were used to test our three
hypotheses. The random effects non-spatial panel regression was used as a baseline model
to compare the effectiveness of spatial models given that the residuals of this model were
spatially autocorrelated (I = 0.164, p = 0.001). One spatial panel model (SAR) was employed to
test for the spatial autocorrelation of the DV. Following this, one spatial Durbin (SDM) panel
and one spatial Durbin Error model (SDEM) were used to test for spillover effects of the IVs.
The former includes a spatial lag of the DV (i.e. Rho coefficient) as well as spatial lags of the IVs
(i.e. Theta coefficient), whereas the latter nests the spatial interaction of the IVs with the error
term (i.e. Lambda coefficient) to indicate whether there is spatial autocorrelation in errors
which may be due to the omitted bias. Among the advantages of SDM and SDEM is that they
include both endogenous (direct or local) and exogenous effects (indirect or spillovers),
produce unbiased coefficient estimates even in the case of misspecification, and do not
impose prior restrictions on the magnitude of spatial spillover effects (Elhorst, 2010;
Rüttenauer, 2019). Another advantage of the SDEM is that the spatial dependence of the
error terms is modelled separately from spillover effects.
Random effects were used since our DENUE and census-based independent variables (i.e.,
concentrated disadvantage and crime opportunity components) contain measures that are
not available on a yearly basis (e.g. population between 15 and 29 years old or bars and
restaurants). As such, these measures are time-invariant. One advantage of using random
effects is that we have hypothesised that differences across (rather than within) neighbour­
hoods influence homicide rates6. For instance, H2 tests whether yearly fluctuations in homi­
cide rates across neighbourhoods are due to differences between these neighbourhoods in
terms of previous levels of drug dealing activity (Contreras & Hipp, 2020).
Spatial panel models were tested using the queen first-order neighbour contiguity
matrix. One neighbourhood in the southernmost part of the city is not contiguous to any
other neighbourhood; however, it was forced into the regression analyses. The spatial
weight matrix was row normalised given that the number of neighbours for each spatial
unit was not the same. Row normalisation brought two benefits: first, the total effect of
neighbours remained unchanged across all spatial units no matter how many neighbours
a spatial unit has, and second, the spatial lags have the same metric as their original
variables, allowing for comparable interpretation.
SDM and SDEM regression coefficients are not directly interpretable as marginal
changes because of the infinite feedback loop created by the use of spatial lags (Pace &
Lesage, 2010). Thus, we calculated the direct (own-neighbourhood), indirect (spatial
spillover), and total impacts (i.e. partial derivatives) of each variable on the dependent
GLOBAL CRIME 9

variable. The software utilised were STATA v.15.1 and Geoda v.1.14. Results with alpha less
than 0.05 were considered statistically significant.

Results
Table 1 shows the descriptive statistics. The mean neighbourhood homicide rate
increased by 34% from 2016 to 2019. The coefficient of variation (CV) shows
a decreasing trend in variation in neighbourhood homicide rates over time, suggesting
a convergence trend in homicide risk across neighbourhoods.
Panel regression results of the SAR, SDM and SDEM are shown in Table 2. Estimated
direct, indirect, and total impacts of the SDM are presented in Table 3. With regard to H1,
we see that the spatial lag coefficient for homicide (Rho) is positive and highly significant
in the SAR and SDM, indicating that homicide spread over time to adjacent neighbour­
hoods. This result confirms H1, that is, homicidal violence spread by proximity indepen­
dently of other factors. However, some variables are missing in the model as indicated by
the statistical significance of the Lambda coefficient in the SDEM. The Lambda coefficient
indicates that homicide dynamics during the period of study can be characterised by both
spatial dependence and error (omitted variable) processes.
The Theta coefficient of drug dealing crimes is positive and statistically significant in
the SAR, SDM and SDEM, indicating that homicide spread over time to adjacent and
neighbourhoods more troubled by drug dealing crimes. This result confirms H2, namely,
that homicide spread to neighbourhoods suffering from drug dealing crimes, likely as
a consequence of organised crime groups fighting for drug markets. While the direct
effect of drug dealing is larger than the spillover effect, neither effect is small (see Table 3).
SDM and SDEM panels show positive own-neighbourhood and spillover effects of
concentrated disadvantage on homicide in adjacent neighbourhoods. However,
neither social and physical signs of disorder nor crime opportunity components
had significant spillover effects on homicide. According to previous studies in Latin
American cities, a positive and significant direct effect of concentrated disadvantage
on homicidal violence is expected (César Fuentes, 2015; Cesar Fuentes & Sanchez,
2015; Pereira, Mota, Andresen, 2017; Santos, Barcellos, Sacarvalho, 2006; Vilalta,
Castillo, Torres, 2016; Vilalta, Lopez, Fondevila, Siordia, 2020; Vilalta, Muggah,
Fondevila, 2020; Vilalta & Muggah, 2014).
The standard deviation of the time-invariant neighbourhood-specific effects
(Sigma_u) is about one-third the standard deviation of the residuals (Sigma_e) in
all regression models, suggesting that a small fraction of the variance of the model is
due to unobserved heterogeneity across neighbourhoods. In other words, even with
an omitted variable, the selection of variables provided a good fit to the data. SAR,

Table 1. Descriptive statistics of homicide rates (per 1,000).


Mean Std. Dev. CV Min Max
2016 0.106 0.483 4.552 0 16.949
2017 0.108 0.279 2.579 0 3.727
2018 0.140 0.342 2.438 0 5.582
2019 0.142 0.302 2.135 0 4.717
Total 0.587 0.940 1.601 0 13.019
N = 1,824
10 C. VILALTA ET AL.

Table 2. Panel data regressions.


Model (1) Model (2) Model (3) Model (4)
Non-spatial Panel SAR panel SDM panel SDEM panel
Drug dealing 0.351*** 0.344*** 0.335*** 0.344***
(0.068) (0.068) (0.068) (0.069)
Conc. disadvantage 0.072*** 0.067*** 0.052*** 0.052***
(0.005) (0.005) (0.008) (0.007)
Social disorder 0.031*** 0.030*** 0.030*** 0.030***
(0.009) (0.009) (0.009) (0.009)
Physical disorder 0.017 0.019 0.023** 0.023**
(0.010) (0.010) (0.010) (0.010)
Crime opportunities 0.044*** 0.041*** 0.049*** 0.048***
(0.007) (0.007) (0.008) (0.008)
2017 0.027 0.026 0.016 0.017
(0.023) (0.023) (0.023) (0.026)
2018 0.102*** 0.084*** 0.054** 0.057**
(0.024) (0.024) (0.026) (0.028)
2019 0.121*** 0.106*** 0.078*** 0.085***
(0.024) (0.024) (0.025) (0.028)
Drug dealing*2017 −0.132 −0.140 −0.147 −0.155
(0.081) (0.080) (0.080) (0.082)
Drug dealing*2018 −0.151** −0.151** −0.166** −0.167**
(0.073) (0.073) (0.073) (0.074)
Drug dealing*2019 −0.181** −0.183** −0.194*** −0.201***
(0.074) (0.074) (0.074) (0.075)
Homicide (Rho) 0.122*** 0.120***
(0.017) (0.018)
Homicide (Lambda) 0.113***
(0.020)
Drug dealing (Theta) 0.129*** 0.166***
(0.038) (0.039)
Conc. disadvantage (Theta) 0.024** 0.034***
(0.010) (0.010)
Social disorder (Theta) 0.004 0.008
(0.014) (0.015)
Physical disorder (Theta) −0.021 −0.020
(0.014) (0.015)
Crime opportunities (Theta) −0.021 −0.016
(0.011) (0.012)
sigma_u:_cons 0.243 0.242*** 0.237*** 0.234***
(0.011) (0.011) (0.012)
sigma_e:_cons 0.612 0.615*** 0.615*** 0.617***
(0.006) (0.006) (0.006)
Intercept 0.329*** 0.278*** 0.279*** 0.320***
(0.017) (0.018) (0.019) (0.019)
2
Pseudo R 0.109 0.113 0.113
AIC 14,540.5 14,521.0 14,535.1
Wald´s test of spatial terms 51.8*** 82.3*** 67.4***
Obs. 7,296 7,296 7,296 7,296
Standard errors are in parenthesis. Year of reference is 2016.
*** p < 0.01, ** p < 0.05

SDM and SDEM panel regressions offered similar AIC values, although the value for
the SDM was comparatively lower, suggesting a better fit.

Discussion and conclusion


Loftin (1986) was the first, to our knowledge, to empirically argue for the contagious and
explosive spread of homicidal violence in vulnerable populations. Accordingly, we have
GLOBAL CRIME 11

Table 3. Direct, indirect and total impact estimates and marginal changes in homicide rates (LN).
Model (1)
SDM
Direct Indirect Total
Drug dealing 0.211*** 0.171*** 0.383***
(0.027) (0.042) (0.048)
Conc. disadvantage 0.052*** 0.034*** 0.086***
(0.007) (0.011) (0.008)
Social disorder 0.031*** 0.009 0.039**
(0.009) (0.016) (0.016)
Physical disorder 0.023*** −0.020 0.003
(0.010) (0.016) (0.018)
Crime opport. 0.048*** −0.017 0.032**
(0.008) (0.012) (0.012)
Standard errors are in parenthesis
*** p < 0.01, ** p < 0.05

presented evidence that homicidal violence spread by proximity across Mexico City
neighbourhoods, contingent on neighbourhood drug crime incidence and concentrated
disadvantage – but independent of disorder and crime opportunity. Likewise, we have
confirmed Contreras and Hipp (2020) argument that illegal drug markets can induce
a spatial spillover effect on homicidal violence.
While our selection of variables and PCA components depicting social disorganisation,
broken windows, and crime pattern theories were able to explain the spatial clustering of
homicide, only social disorganisation by means of concentrated disadvantage was able to
predict its spatial diffusion. Thus, only social disorganisation theory explains the spatial
process of homicidal violence – the other theories appear to only explain spatial patterns,
at least in the case of Mexico City.
The increasingly negative interaction between drug-dealing crimes and year in all
models is substantively important as it indicates that the neighbourhood-level relation­
ship between drug activity and criminal homicide weakened over time7. This finding may
suggest that homicides either caused by or associated with organised crime drug dealing
operations declined in number during the period of study when analysed at the neigh­
bourhood level. This remains speculative yet represents an interesting hypothesis for
future study.
In this respect, if we are to test and advance criminological theory based on the Latin
American case, some modifications need to be made that respond to context. For the
Mexican case, we know homicidal violence is strongly associated with organised crime
(Flores & Rodriguez-Oreggia, 2014; Flores & Villarreal, 2015), and organised crime is
strongly associated with drug dealing (Shirk & Wallman, 2015; Vilalta, 2014). In this
study, we confirmed the contagious spread of homicidal violence in neighbourhoods
with higher rates of drug dealing crimes. While this finding is not entirely new, given the
Contreras and Hipp (2020) study, it is nevertheless one of the least discussed processes of
spatial elements of crime. Our study compensates for this lack of research.
One limitation of this study is that while we have estimated the spatial effect of drug
dealing crimes on the spatial diffusion of homicide, we do not know why some neigh­
bourhoods have high rates of drug dealing crimes in the first place. Another limitation is
that our study focuses on a single city. There might be something specific about Mexico
12 C. VILALTA ET AL.

City that explains the results of this study. In addition, we did not include the neighbour­
hoods in the State of Mexico surrounding Mexico City as neighbourhood crime data for
the surrounding area were unavailable. Future research may also provide direct evidence
of the causes of the spread of violence, perhaps through interviews with drug dealers and
drug users.
In conclusion, criminological theories on the geography of homicide have provided
clear explanations for the spatial clustering of homicide, but spatial diffusion processes of
homicidal violence remain extremely understudied. In addition, most empirical evidence
to date is based on the developed world, yet, most of the homicidal violence problem is
essentially a Latin American one. Clearly, criminological theory cannot disregard the
examination of the spatial patterns and processes of crime and violence in the developing
world, given the gravity of the problem and its capacity to provide theoretical insights.

Notes
1. This report was not made publicly and we were not able to access it. However, some
journalists were able to access and report on it.
2. This cartography is a heavy data file, and thus can only be obtained by CD via a specific
request to the INE authority. See: https://portalanterior.ine.mx/archivos3/portal/historico/
contenido/interiores/Detalle_geografia_electoral_y_cartografia_transparencia-id
-02d4aee2ccdf0210VgnVCM1000000c68000aRCRD/
3. See: https://datos.cdmx.gob.mx/pages/home/
4. As suggested by one of the reviewers.
5. We are aware that previous studies include income (e.g. median family income) measures in
Concentrated Disadvantage indexes. As no such data exists in the Mexican Census, we were
unable to do this.
6. In a fixed-effects model, time-invariant variables are taken up by the intercept.
7. As correctly pointed out by one of the reviewers.

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

Notes on contributors
Carlos Vilalta is Professor at the Center for Research in Geospatial Information Sciences (CentroGeo)
in Mexico City. He studies the geography of crime and fear of crime, prevention policies, criminal
statistics, and prison populations. He has been visiting researcher in UC San Diego, U. of Florida,
Cambridge, McGill, U of Missouri in St. Louis, Washington University in St. Louis, U. of Houston and
UNC-Chapel Hill. He holds a PhD in Urban Studies from Portland State University.
Pablo Lopez-Ramirez is coordinator of the master's degree at the Center for Research in Geospatial
Information Sciences (CentroGeo) and responsible for the DataLab project of the National
Geointelligence Laboratory. He has taught the subjects of Spatial Analysis, Geographic
Information Systems and Geoinformatics. His research focuses on developing links between tradi­
tional research in Human Geography, especially from the perspective of Spatial Analysis, and Data
Science.
Gustavo Fondevila is a professor and researcher at the Centro de Investigación y Docencia
Económicas (CIDE) in Mexico CIty, where he concentrates on empirical and comparative quantita­
tive criminology. Specifically, he focuses on criminal justice and prisons in Latin America by use of
GLOBAL CRIME 13

surveys of prisons, court records and more. His most recent work examines the relationship
between prison violence and criminal government within prisons of the region. He also studies
criminal justice institutions from a quantitative perspective, such as prosecution, defense and
courts.

ORCID
Carlos Vilalta http://orcid.org/0000-0002-6030-7018
Gustavo Fondevila http://orcid.org/0000-0003-4818-0584

Bibliography
Alves, L. G. A., E. K. Lenzi, R. S. Mendes, and H. V. Ribeiro. “Spatial Correlations, Clustering and
Percolation-like Transitions in Homicide Crimes.” EPL (Europhysics Letters) 111, no. 1 (2015): 18002.
doi:10.1209/0295-5075/111/18002.
Blumstein, A., F. P. Rivara, and R. Rosenfeld. “The Rise and Decline of Homicide—And Why.” Annual
Review of Public Health 21, no. 1 (2000): 505–541. doi:10.1146/annurev.publhealth.21.1.505.
Bond, R., and B. Bushman. “The Contagious Spread of Violence among US Adolescents through
Social Networks.” American Journal of Public Health 107, no. 2 (2017): 288–294. doi:10.2105/
AJPH.2016.303550.
Cardona, M., H. I. García, C. A. Giraldo, M. V. López, C. M. Suárez, D. C. Corcho, C. H. Posada, and
M. N. Flórez. “Homicidios en Medellín, Colombia, entre1990 y 2002: Actores, motivos
y circunstancias.” Cadernos De Saúde Pública 21, no. 3 (2005): 840–851. doi:10.1590/S0102-
311X2005000300018.
Ceccato, V., A. Uittenbogaard, and R. Bamzar. “Security in Stockholm’s Underground Stations: The
Importance of Environmental Attributes and Context.” Security Journal 26, no. 1 (2013): 33–59.
doi:10.1057/sj.2011.32.
Ceccato, V., R. Haining, and T. Kahn. “The Geography of Homicide in São Paulo, Brazil.” Environment
and Planning A: Economy and Space 39, no. 7 (2007): 1632–1653. doi:10.1068/a38283.
Cerdá, M., M. Tracy, S. F. Messner, D. Vlahov, K. Tardiff, and S. Galea. “Misdemeanor Policing, Physical
Disorder, and Gun-related Homicide: A Spatial Analytic Test of “Broken-windows” Theory.”
Epidemiology 20, no. 4 (2009): 533–541. doi:10.1097/EDE.0b013e3181a48a99.
Cohen, J., and G. Tita. “Diffusion in Homicide: Exploring a General Method for Detecting Spatial
Diffusion Processes.” Journal of Quantitative Criminology 15, no. 4 (1999): 451–493. doi:10.1023/
1007596225550.
Cohen, Lawrence E., and Marcus Felson. „Social change and crime rate trends: A routine activity
approach.„ American sociological review (1979): 588–608
Contreras, C., and J. R. Hipp. “Drugs, Crime, Space, and Time: A Spatiotemporal Examination of Drug
Activity and Crime Rates.” Justice Quarterly 37, no. 2 (2020): 187–209. doi:10.1080/
07418825.2018.1515318.
Culyba, A. J., S. F. Jacoby, T. S. Richmond, J. A. Fein, B. C. Hohl, and C. C. Branas. “Modifiable
Neighborhood Features Associated With Adolescent Homicide.” JAMA Pediatrics 170, no. 5
(2016): 473–480. doi:10.1001/jamapediatrics.2015.4697.
Dugato, M., F. Calderoni, and G. Berlusconi. “Forecasting Organized Crime Homicides: Risk Terrain
Modeling of Camorra Violence in Naples, Italy.” Journal of Interpersonal Violence 35, no. 19–20
(2020): 4013–4039. doi:10.1177/0886260517712275.
Eck, J. E., and D. Weisburd. “Crime and Place, Crime Prevention Studies.” In Monsey, NY: Criminal
Justice Press. EckCrime and Place, Crime Prevention Studies4Monsey1995. 1995.
Elhorst, J. P. “Spatial Panel Data Models.” In Handbook of Applied Spatial Analysis, edited by
M. M. Fischer and A. Getis, 377–407. Heidelberg, Germany: Springer Berlin Heidelberg, 2010.
doi:10.1007/978-3-642-03647-7_19.
14 C. VILALTA ET AL.

Fagan, J., D. L. Wilkinson, and G. Davies. “Social Contagion of Violence.” In The Cambridge Handbook
of Violent Behavior and Aggression, edited by D. J. Flannery, A. T. Vazsonyi, and I. D. Waldman,
688–724. Cambridge, USA: Cambridge University Press, 2007. doi:10.1017/
CBO9780511816840.037.
Ferrandino, J. “Endemic, Outbreak or Epidemic? Geographies of Affliction, Exposure and Homicide
Immunity in Chicago.” Journal of Crime and Justice 41, no. 4 (2018): 347–363. doi:10.1080/
0735648X.2017.1398676.
Flores, M., and A. Villarreal. “Exploring the Spatial Diffusion of Homicides in Mexican Municipalities
through Exploratory Spatial Data Analysis.” Cityscape 17, no. 1 (2015): 35–50.
Flores, M., and E. Rodriguez-Oreggia. Spillover Effects of Homicides across Mexican Municipalities:
A Spatial Regime Model Approach. 2014.
Fuentes, David, and Kevin Ruiz. “Un 65% de Asesinatos En La CDMX Son ‘Ejecuciones’: SSC.” El
Universal, 2020. https://www.eluniversal.com.mx/metropoli/un-65-de-asesinatos-en-la-cdmx-
son-ejecuciones-ssc
Fuentes, C. “El impacto de las viviendas deshabitadas en el incremento de delitos (robo a casa
habitación y homicidios) en Ciudad Juárez, Chihuahua, 2010.” Frontera Norte 27, no. 54 (2015):
171–196.
Fuentes, C., and O. Sanchez (2015). Contexto sociodemográfico de los homicidios en México D.F.: Un
análisis espacial. Revista Panamericana de Salud Pública, 38, 450–456. https://www.scielosp.org/
article/rpsp/2015.v38n6/450-456/es/
Gallison, J. K. “The Line of Crime: Dismantling Fears and Concerns of Crime along Vancouver
SkyTrain’s Canada Line.” Security Journal 29, no. 3 (2016): 485–500. doi:10.1057/sj.2013.42.
Gau, J. M., and T. C. Pratt. “Broken Windows or Window Dressing Citizens’(in) Ability to Tell the
Difference between Disorder and Crime.” Criminology & Public Policy 7, no. 2 (2008): 163–194.
doi:10.1111/j.1745-9133.2008.00500.x.
Gau, J. M., and T. C. Pratt. “Revisiting Broken Windows Theory: Examining the Sources of the
Discriminant Validity of Perceived Disorder and Crime.” Journal of Criminal Justice 38, no. 4
(2010): 758–766. doi:10.1016/j.jcrimjus.2010.05.002.
Gerstner, D., R. Wickes, and D. Oberwittler. “Collective Efficacy in Australian and German
Neighborhoods: Testing Cross-Cultural Measurement Equivalence and Structural Correlates in a
Multi-level SEM Framework.” Social Indicators Research (2019): 1–27. doi:10.1007/s11205-019-
02081-4.
Giacopassi, D., and D. R. Forde. “Broken Windows, Crumpled Fenders, and Crime.” Journal of Criminal
Justice 28, no. 5 (2000): 397–405. doi:10.1016/S0047-2352(00)00054-4.
Giménez-Santana, A., J. M. Caplan, and G. Drawve. “Risk Terrain Modeling and Socio-Economic
Stratification: Identifying Risky Places for Violent Crime Victimization in Bogotá, Colombia.”
European Journal on Criminal Policy and Research 24, no. 4 (2018): 417–431. doi:10.1007/
s10610-018-9374-5.
Gorman, D. M., P. W. Speer, P. J. Gruenewald, and E. W. Labouvie. “Spatial Dynamics of Alcohol
Availability, Neighborhood Structure and Violent Crime.” Journal of Studies on Alcohol 62, no. 5
(2001): 628–636. doi:10.15288/jsa.2001.62.628.
Graif, C., and R. J. Sampson. “Spatial Heterogeneity in the Effects of Immigration and Diversity on
Neighborhood Homicide Rates.” Homicide Studies 13, no. 3 (2009): 242–260. doi:10.1177/
1088767909336728.
Groff, E., and T. McEwen. “Exploring the Spatial Configuration of Places Related to Homicide Events.”
Institute for Law and Justice (2006): 137.
Haney, T. J. ““Broken Windows” and Self-Esteem: Subjective Understandings of Neighborhood
Poverty and Disorder.” Social Science Research 36, no. 3 (2007): 968–994. doi:10.1016/j.
ssresearch.2006.07.003.
Henderson, J. P., S. E. Morgan, F. Patel, and M. E. Tiplady. “Patterns of Non-firearm Homicide.” Journal
of Clinical Forensic Medicine 12, no. 3 (2005): 128–132. doi:10.1016/j.jcfm.2004.10.011.
Hohl, B. C., S. Wiley, D. J. Wiebe, A. J. Culyba, R. Drake, and C. C. Branas. “Association of Drug and
Alcohol Use with Adolescent Firearm Homicide at Individual, Family, and Neighborhood Levels.”
JAMA Internal Medicine 177, no. 3 (2017): 317–324. doi:10.1001/jamainternmed.2016.8180.
GLOBAL CRIME 15

Ingram, M. C., and D. C. M. Marchesini. “A Spatial Analysis of Homicide across Brazil’s Municipalities.”
Homicide Studies 21, no. 2 (2017): 87–110. doi:10.1177/1088767916666603.
Irvin-Erickson, Y., and N. La Vigne. “A Spatio-temporal Analysis of Crime at Washington, DC Metro
Rail: Stations’ Crime-generating and Crime-attracting Characteristics as Transportation Nodes and
Places.” Crime Science 4, no. 1 (2015): 14. doi:10.1186/s40163-015-0026-5.
Johnson, S. D., and L. Summers. “Testing Ecological Theories of Offender Spatial Decision Making
Using a Discrete Choice Model.” Crime and Delinquency 61, no. 3 (2015): 454–480. doi:10.1177/
0011128714540276.
Kubrin, C. E. “Social Disorganization Theory: Then, Now, and in the Future.” In Handbook on Crime
and Deviance, 225–236. New York, USA: Springer, 2009.
Lanfear, C. C., R. L. Matsueda, and L. R. Beach. “Broken Windows, Informal Social Control, and Crime:
Assessing Causality in Empirical Studies.” Annual Review of Criminology 3, no. 1 (2020): 97–120.
doi:10.1146/annurev-criminol-011419-041541.
Lauritsen, J. L., and T. S. Lentz. “National and Local Trends in Serious Violence, Firearm Victimization,
and Homicide.” Homicide Studies 23, no. 3 (2019): 243–261. doi:10.1177/1088767919848665.
Light, M. T., and C. T. Harris. “Race, Space, and Violence: Exploring Spatial Dependence in Structural
Covariates of White and Black Violent Crime in US Counties.” Journal of Quantitative Criminology
28, no. 4 (2012): 559–586. doi:10.1007/s10940-011-9162-6.
Loeffler, C., and S. Flaxman. “Is Gun Violence Contagious? A Spatiotemporal Test.” Journal of
Quantitative Criminology 34, no. 4 (2018): 999–1017. doi:10.1007/s10940-017-9363-8.
Loftin, C. “Assaultive Violence as a Contagious Social Process.” Bulletin of the New York Academy of
Medicine 62, no. 5 (1986): 550.
McCord, E. S., J. H. Ratcliffe, R. M. Garcia, and R. B. Taylor. “Nonresidential Crime Attractors and
Generators Elevate Perceived Neighborhood Crime and Incivilities.” Journal of Research in Crime
and Delinquency 44, no. 3 (2007): 295–320. doi:10.1177/0022427807301676.
Mennis, J., P. W. Harris, Z. Obradovic, A. J. Izenman, H. E. Grunwald, and B. Lockwood. “The Effect of
Neighborhood Characteristics and Spatial Spillover on Urban Juvenile Delinquency and
Recidivism.” The Professional Geographer 63, no. 2 (2011): 174–192. doi:10.1080/
00330124.2010.547149.
Merino, J., and E. Fierro. Violencia en las ciudades de México: Un análisis de tres periodos. CEPAL, 2016.
Messner, S. F., and L. Anselin. “Spatial Analyses of Homicide with Areal Data.” Spatially Integrated
Social Science 12 (2004): 127–144.
Messner, S. F., L. Anselin, R. D. Baller, D. F. Hawkins, G. Deane, and S. E. Tolnay. “The Spatial Patterning
of County Homicide Rates: An Application of Exploratory Spatial Data Analysis.” Journal of
Quantitative Criminology 15, no. 4 (1999): 423–450. doi:10.1023/A:1007544208712.
Muggah, R., and K. A. Tobón. “Reducing Latin America’s Violent Hot Spots.” Aggression and Violent
Behavior 40 (2018): 83–90. doi:10.1016/j.avb.2018.09.003.
Norza, E., N. M. Vargas, B. L. Avendaño, H. Rincón, and M. A. Ospino. “Criminología ambiental
y homicidio en la ciudad de Bogotá (Colombia).” Revista de Estudios Sociales No.35, no. 63
(2018): 55–71. doi:10.7440/res63.2018.05.
O’Brien, D. T., C. Farrell, and B. C. Welsh. “Looking through Broken Windows: The Impact of
Neighborhood Disorder on Aggression and Fear of Crime Is an Artifact of Research Design.”
Annual Review of Criminology 2, no. 1 (2019a): 53–71. doi:10.1146/annurev-criminol-011518-
024638.
O’Brien, D. T., C. Farrell, and B. C. Welsh. “Broken (Windows) Theory: A Meta-analysis of the Evidence
for the Pathways from Neighborhood Disorder to Resident Health Outcomes and Behaviors.”
Social Science & Medicine 228 (2019b): 272–292. doi:10.1016/j.socscimed.2018.11.015.
O’Brien, D. T., and R. J. Sampson. “Public and Private Spheres of Neighborhood Disorder: Assessing
Pathways to Violence Using Large-scale Digital Records.” Journal of Research in Crime and
Delinquency 52, no. 4 (2015): 486–510. doi:10.1177/0022427815577835.
O’Brien, D. T., R. J. Sampson, and C. Winship. “Ecometrics in the Age of Big Data: Measuring and
Assessing “Broken Windows” Using Large-scale Administrative Records.” Sociological
Methodology 45, no. 1 (2015): 101–147. doi:10.1177/0081175015576601.
16 C. VILALTA ET AL.

Pace, R. K., and J. Lesage. “Spatial Econometrics.” Handbook of Spatial Statistics 1, no. 1 (2010):
245–260.
Papachristos, A. V. “Murder by Structure: Dominance Relations and the Social Structure of Gang
Homicide.” American Journal of Sociology 115, no. 1 (2009): 74–128. doi:10.1086/597791.
Papachristos, A. V., D. M. Hureau, and A. A. Braga. “The Corner and the Crew: The Influence of
Geography and Social Networks on Gang Violence.” American Sociological Review 78, no. 3 (2013):
417–447. doi:10.1177/0003122413486800.
Papachristos, A. V., and S. Bastomski. “Connected in Crime: The Enduring Effect of Neighborhood
Networks on the Spatial Patterning of Violence.” American Journal of Sociology 124, no. 2 (2018):
517–568. doi:10.1086/699217.
Patino, J. E., J. C. Duque, J. E. Pardo-Pascual, and L. A. Ruiz. “Using Remote Sensing to Assess the
Relationship between Crime and the Urban Layout.” Applied Geography 55 (2014): 48–60.
doi:10.1016/j.apgeog.2014.08.016.
Pereira, D., C. Mota, and M. Andresen. “Social Disorganization and Homicide in Recife, Brazil.”
International Journal of Offender Therapy and Comparative Criminology 61, no. 14 (2017):
1570–1592. doi:10.1177/0306624X15623282.
Pizarro, J., N. Corsaro, and S. Yu. “Journey to Crime and Victimization: An Application of Routine
Activities Theory and Environmental Criminology to Homicide: Victims & Offenders: Vol 2, No 4.”
Victims & Offenders 2, no. 4 (2007): 375–394. doi:10.1080/15564880701568520.
Pridemore, W. A., and T. H. Grubesic. “A Spatial Analysis of the Moderating Effects of Land Use on the
Association between Alcohol Outlet Density and Violence in Urban Areas.” Drug and Alcohol
Review 31, no. 4 (2012): 385–393. doi:10.1111/j.1465-3362.2011.00329.x.
Rosenfeld, R., R. Fornango, and A. F. Rengifo. “The Impact of Order-maintenance Policing on
New York City Homicide and Robbery Rates: 1988-2001.” Criminology 45, no. 2 (2007): 355–384.
doi:10.1111/j.1745-9125.2007.00081.x.
Rüttenauer, T. “Spatial Regression Models: A Systematic Comparison of Different Model
Specifications Using Monte Carlo Experiments.” Sociological Methods & Research (2019):
0049124119882467.
Sampson, R. J., and J. D. Morenoff. “Spatial (Dis) Advantage and Homicide in Chicago
Neighborhoods.” Spatially Integrated Social Science (2004): 145–170.
Sampson, R. J., and S. W. Raudenbush. “Systematic Social Observation of Public Spaces: A New Look
at Disorder in Urban Neighborhoods.” American Journal of Sociology 105, no. 3 (1999): 603–651.
doi:10.1086/210356.
Sampson, R. J., and S. W. Raudenbush. “Disorder in Urban Neighborhoods—Does It Lead to Crime?:
(512722006-001) [Data Set].” American Psychological Association (2001). doi:10.1037/e512722006-
001.
Sampson, R. J., and S. W. Raudenbush. “Seeing Disorder: Neighborhood Stigma and the Social
Construction of “Broken Windows.”.” Social Psychology Quarterly 67, no. 4 (2004): 319–342.
doi:10.1177/019027250406700401.
Sampson, R. J., S. W. Raudenbush, and F. Earls. “Neighborhoods and Violent Crime: A Multilevel
Study of Collective Efficacy.” Science 277, no. 5328 (1997): 918–924. doi:10.1126/
science.277.5328.918.
Santos, S., C. Barcellos, and M. Sacarvalho. “Ecological Analysis of the Distribution and Socio-spatial
Context of Homicides in Porto Alegre, Brazil.” Health & Place 12, no. 1 (2006): 38–47. doi:10.1016/j.
healthplace.2004.08.009.
Shirk, D., and J. Wallman. “Understanding Mexico’s Drug Violence.” Journal of Conflict Resolution 59,
no. 8 (2015): 8. doi:10.1177/0022002715587049.
Smith, T. A., and J. S. O. Sandoval. “A Spatial Analysis of Homicides in Saint Louis: The Importance of
Scale.” Spatial Demography 7, no. 1 (2019): 57–82. doi:10.1007/s40980-018-00046-8.
Snowden, A. J., T. D. Stucky, and W. A. Pridemore. “Alcohol Outlets, Social Disorganization, and
Non-violent Crimes in Urban Neighborhoods.” Journal of Crime and Justice 40, no. 4 (2017):
430–445. doi:10.1080/0735648X.2016.1176949.
GLOBAL CRIME 17

Sparks, C. “Commercial Density, Residential Concentration, and Crime: Land Use Patterns and
Violence in Neighborhood Context—Buscar Con Google.” Spatial and Spatio-Temporal
Epidemiology 2, no. 4 (2011): 301–309. doi:10.1016/j.sste.2011.10.001.
Stein, R. E., J. F. Conley, and C. Davis. “The Differential Impact of Physical Disorder and Collective
Efficacy: A Geographically Weighted Regression on Violent Crime.” GeoJournal 81, no. 3 (2016):
351–365. doi:10.1007/s10708-015-9626-6.
Swart, L.-A., M. Seedat, and J. Nel. “The Situational Context of Adolescent Homicide Victimization in
Johannesburg, South Africa.” Journal of Interpersonal Violence 33, no. 4 (2018): 637–661.
doi:10.1177/0886260515613342.
Taylor, R. Breaking Away from Broken Windows: Baltimore Neighborhoods and the Nationwide Fight
against Crime, Grime, Fear, and Decline. New York, USA: Routledge, 2018.
Tita, G., and E. Griffiths. “Traveling to Violence: The Case for a Mobility-Based Spatial Typology of
Homicide.” Journal of Research in Crime and Delinquency 42, no. 3 (2005): 275–308. doi:10.1177/
0022427804270051.
Tita, G. E., and S. M. Radil. “Spatializing the Social Networks of Gangs to Explore Patterns of Violence.”
Journal of Quantitative Criminology 27, no. 4 (2011): 521–545. doi:10.1007/s10940-011-9136-8.
Tomsen, S., and J. Payne (2016). Homicide and the Night-time Economy: Report to the Criminology
Research Council. Criminology Research Grants. https://researchdirect.westernsydney.edu.au/
islandora/object/uws:38447/datastream/PDF/download/citation.pdf
Tracy, M., A. A. Braga, and A. V. Papachristos. “The Transmission of Gun and Other Weapon-Involved
Violence within Social Networks.” Epidemiology Review 38, no. 1 (2016): 70–86. https://www.ncbi.
nlm.nih.gov/pmc/articles/PMC4762247/ .
Valasik, M., E. E. Brault, and S. M. Martinez. “Forecasting Homicide in the Red Stick: Risk Terrain
Modeling and the Spatial Influence of Urban Blight on Lethal Violence in Baton Rouge, Louisiana.”
Social Science Research 80 (2018): 186–201. doi:10.1016/j.ssresearch.2018.12.023.
Valencia, E. J. B., and M. I. Z. Sanz. “Homicide and Land Prices: A Spatial Analysis in Santiago De Cali.”
Cuadernos de Economía 40, no. 113 (2017): 147–159. doi:10.1016/j.cesjef.2016.06.001.
Valente, R. “Spatial and Temporal Patterns of Violent Crime in A Brazilian State Capital:
A Quantitative Analysis Focusing on Micro Places and Small Units of Time.” Applied Geography
103 (2019): 90–97. doi:10.1016/j.apgeog.2019.01.006.
Vilalta, C. “La Geografía Local Del Narcomenudeo: Patrones, Procesos Y Recomendaciones De
Política Urbana.” Estudios demográficos y urbanos 24, no. 1 (2009): 49–77. doi:10.24201/edu.
v24i1.1344.
Vilalta, C. “How Exactly Does Place Matter in Crime Analysis? Place, Space, and Spatial
Heterogeneity.” Journal of Criminal Justice Education 24, no. 3 (2013): 290–315. doi:10.1080/
10511253.2012.715659.
Vilalta, C. “How Did Things Get so Bad so Quickly? An Assessment of the Initial Conditions of the War
against Organized Crime in Mexico.” European Journal on Criminal Policy and Research 20, no. 1
(2014): 137–161. doi:10.1007/s10610-013-9218-2.
Vilalta, C., and G. Fondevila (2019). Modeling Crime in an Uptown Neighborhood: The Case of Santa
Fe in Mexico City. Papers in Applied Geography, Published Online. 10.1080/23754931.2018.1554502
Vilalta, C., J. Castillo, and J. Torres (2016). Violent Crime in Latin American Cities (No. 474; Discussion
Paper, P. 126). Inter-American Development Bank.
Vilalta, C., P. Lopez, G. Fondevila, and O. Siordia. “Testing Broken Windows Theory in Mexico City.”
Social Science Quarterly 101, no. 2 (2020): 558–572. doi:10.1111/ssqu.12760.
Vilalta, C., and R. Muggah. “Mugga.” Trends in Organized Crime 17, no. 3 (2014): 161–180.
doi:10.1007/s12117-014-9213-0.
Vilalta, C., and R. Muggah (2020). Future Trends in Homicide: Extrapolations from 2019 to 2030
(SDG16.1; Pathfinders for Peaceful, Just and Inclusive Societies, P. 28). Center on International
Cooperation. https://530cfd94-d934-468b-a1c7-c67a84734064.filesusr.com/ugd/6c192f_
354a5b3c97d44fb4a9dbc8de7f19d0ce.pdf
Vilalta, C., R. Muggah, and G. Fondevila. “Homicide as a Function of City Block Layout: Mexico City as
Case Study.” Global Crime 1, no. 1 (2020): 1–19. doi:10.1080/17440572.2020.1715219.
18 C. VILALTA ET AL.

Ward, J. T., M. R. Nobles, T. J. Youstin, and C. L. Cook. “Placing the Neighborhood Accessibility–
Burglary Link in Social-Structural Context.” Crime and Delinquency 60, no. 5 (2014): 739–763.
doi:10.1177/0011128710364804.
Warner, B. D., and B. W. Coomer. “Neighborhood Drug Arrest Rates: Are They A Meaningful Indicator
Of Drug Activity? A Research Note.” Journal of Research in Crime and Delinquency 40, no. 2 (2003):
123–138. doi:10.1177/0022427802251018.
Wilcox, P., and F. T. Cullen. “Situational Opportunity Theories of Crime.” Annual Review of
Criminology 1, no. 1 (2018): 123–148. doi:10.1146/annurev-criminol-032317-092421.
Wilcox, P., N. Quisenberry, D. T. Cabrera, and S. Jones. “Busy Places and Broken Windows? Toward
Defining the Role of Physical Structure and Process in Community Crime Models.” The Sociological
Quarterly 45, no. 2 (2004): 185–207. doi:10.1111/j.1533-8525.2004.tb00009.x.
Williams, M. L., P. Burnap, and L. Sloan. “Crime Sensing with Big Data: The Affordances and
Limitations of Using Open-source Communications to Estimate Crime Patterns.” The British
Journal of Criminology 57, no. 2 (2017): 320–340.
Wilson, J. Q., and G. Kelling (1982). Broken Windows. Atlantic Monthly, 249(3), 29–38.
Yang, S.-M. “Assessing the Spatial–Temporal Relationship between Disorder and Violence.” Journal
of Quantitative Criminology 26, no. 1 (2010): 139–163. doi:10.1007/s10940-009-9085-7.
Zeoli, April M., Jesenia M. Pizarro, Sue C. Grady, and Christopher Melde. „Homicide as infectious
disease: Using public health methods to investigate the diffusion of homicide.„ Justice quarterly
31, no. 3 (2014): 609–632.

TABLE A1. Independent variables: principal component analysis statistics.


Concentrated Disadvantage
(Explained variance: 62.2%) Component loadings Kaiser-Meyer-Olkin test
Pop. 15–29 years old (%) 0.775 0.877
Native Indian speaking pop. (%) 0.563 0.805
Pop. w/o health insurance (%) 0.806 0.885
Pop. w/o elementary schooling (%) 0.878 0.719
Persons per bedroom (Mean) 0.879 0.730
Broken windows: Social disorder
(Explained variance: 77.5%)
Public intoxication with drugs (Count) 0.869 0.761
Public intoxication with alcohol (Count) 0.918 0.753
Street scandal (Count) 0.892 0.820
Loud partying (Count) 0.839 0.813
Broken windows: Physical disorder
(Explained variance: 56.4%)
Graffiti signs (Count) 0.711 0.859
Drinking water leaks (Count) 0.760 0.834
Sewage spills (Count) 0.667 0.860
Gas leaks (Count) 0.810 0.778
Downed power lines (Count) 0.795 0.782
Crime Opportunities
(Explained variance: 54.1%)
Street permeability (Index) 0.664 0.806
Metro stations (Count) 0.658 0.876
Metrobus stations (Count) 0.702 0.829
Bars (Count) 0.794 0.705
Restaurants (Count) 0.912 0.687
Conv. stores w/alcohol sale permit (Count) 0.643 0.712

Appendix
GLOBAL CRIME 19

TABLE A2. Estimation of sample descriptive statistics.


Mean Std. Dev. Min Max
Homicide rates (LN) 0.443 0.707 0.000 5.139
Drug dealing crime rates (LN) 0.242 0.400 0.000 6.217
Conc. disadvantage 0.000 1.752 −6.609 7.839
Social disorder 0.000 1.760 −1.653 14.763
Physical disorder 0.000 1.679 −1.457 12.230
Crime opportunity 0.000 1.801 −1.368 29.117
N = 7,296

You might also like