Concentration of Urban Violence in Fortaleza and Strategies for Crime Prevention

Abstract

Despite the continued prevalence of violence in Latin America, there is a relative dearth of research investigating both spatial patterns of violent crimes and the effectiveness of evidence-based crime prevention policies in Brazil. This study aims to address this gap in extant knowledge by creating a Spatial Violence Index and a Restrictive Ambient Index to investigate the spatial dynamics of violent crimes and urban vulnerabilities in Fortaleza. Both exploratory spatial data analysis and spatial regression models were employed to visualize the associative patterns and measure the correlation between the two indexes. The results demonstrate how locations characterized by high levels of violence are spatially correlated with more vulnerable locations in terms of both socio-economic-demographics and urban disorder. Overall, the study identified 124 vulnerable micro-territories that would benefit from the allocation of resources in an effort to reduce violence in the city by enhancing the efficiency of policing and prevention strategies.

Keywords

urban violence spatial analysis local vulnerabilities concentration of violence Latin America Fortaleza

Introduction

In 2018, Brazil registered the highest number of homicides in the world, along with ranking among the seven countries with the highest homicide rate at the global level (UNODC, 2021). According to prior studies and a recent systematic review, both an evidence-based approach to crime prevention and targeted strategies of predictive policing at micro-places can help to reduce crime as well as allocate resources more effectively and efficiently (Braga et al., 2019; Braga et al., 1999; Meijer & Wessels, 2019; Ratcliffe et al., 2011). Research has also demonstrated how this approach could also be expedient for tackling violent crimes in Latin America (de Oliveira et al., 2019; Muggah & Tobón, 2019). Nevertheless, despite the widespread violence in Latin America’s urban contexts, there is a scarcity of studies that have analyzed spatial patterns of violent crime in this region and/or delineated potentially effective and efficient crime prevention strategies.

This study aims to address this gap in extant knowledge and literature by investigating the micro-dynamics of violent crimes in the city of Fortaleza. In 2018, Fortaleza registered around 70 murders per 100,000 inhabitants, which ranked it among the 10 most violent cities in the world. The city is the state capital of Cearà and is located in the Northeastern region of Brazil. With a population of more than 2.5 million people, it is the fifth largest city in Brazil. According to the most recent crime trends, although the homicide rate there decreased in in 2019, Fortaleza nevertheless remains one of the most violent cities in the country (Seguridad, Justicia y Paz, 2021). Both drug trafficking offenses and lethal and intentional violent crimes appear to be largely concentrated within specific areas of the city where precarious settlements (e.g., tenements, slums, and irregular subdivisions of low-income residents) are located, thus pointing toward a potential spatial association between these crime types and places characterized by social disorganization and urban vulnerability. This study constructs two indexes using Principal Component Analysis (PCA) to measure the level of violence at micro-places (Spatial Violence Index [SVI]) as well as the level of vulnerability of those micro-places in terms of socio-economic-demographics and disorder point of view (Restrictive Ambience Index [RAI]). The spatial and statistical relationship between these indexes are then measured through an exploratory spatial data analysis (ESDA) and a set of statistical models (i.e., ordinary least square [OLS], spatial auto regressive model, and spatial error model [SEM]). The concentration of violence largely occurs in vulnerable areas, and there is a spatial association between the concentration of violent crimes and highly vulnerable places in Fortaleza. The study identified 124 vulnerable micro-territories which would benefit from the allocation of resources in order to reduce violence in the city. Given that those areas are characterized by high levels of socio-economic-demographic vulnerability and disorder that, in turn, make them fertile ground for violence, as part of the process of carrying out a cost-benefit analysis balance, crime prevention policies should seek to prioritize those areas to maximize the results in terms of violence reduction.

In the first section, this article provides an overview of previous relevant analyses of crime concentration at the microlevel within different urban settings. The second section delineates both the empirical strategy and methodologies that were adopted to investigate the relationship between violence and place-based vulnerabilities. The third section presents the results of the spatial and statistical analysis that was performed to test this relationship. The fourth section puts forward suggestions for crime prevention and the allocation of resources, while the fifth section concludes the article.

Literature Review

From the early 1980s up until the present historical juncture, there has been an emergent theoretical interest in the micro-dynamics of crime, with a substantial number of studies delineating the ways in which crime tends to cluster within specific places. In particular, studies on the spatial concentration of criminal events have proliferated over the last two decades. Indeed, understanding the micro-dynamics of crime has become a predominant area of scientific inquiry within the field of criminology in light of the emergence of strong evidence supporting the fact that crime concentrates within micro-spaces of urban settings. The findings of these studies have substantial implications for crime forecasting and police resource allocation models, which ultimately aim to develop successful crime prevention programs within specific places (Braga & Weisburd, 2010; Braga et al., 1999; Johnson, 2010; Sherman & Weisburd, 1995). Indeed, according to extant literature, the places in which crime concentrates are invariably characterized by specific features that make them particularly criminogenic. For instance, Caplan and Kennedy (2016) posit that the level of crime risk of each place is determined by a combination of their corresponding level of spatial vulnerability and exposure to crime. “Vulnerability” derives from the combined influence of a set of urban features that either enable or deter crime occurrence, while “exposure” refers to patterns of crime concentration and repeat victimization on a short- or long-term basis. Consequently, by policing and proposing targeted prevention policies within high crime risk places, the overall crime rates of urban settings can be reduced.

The majority of studies on the spatial concentration of crime in micro-places and crime forecasting have been carried out in cities of the United States (see, e.g., Braga & Clarke, 2014; Gill et al., 2016; Haberman et al., 2016; Hipp & Kim, 2016; Kennedy et al., 2011; Levin et al., 2016; Weisburd, 2015; Weisburd et al., 2012) and in few other cities across the world such as, for example, Vancouver, The Hague, Milan, and London (Andresen et al., 2016; Bowers & Johnson, 2005; Ceccato & Oberwittler, 2008; Dugato, 2013; Dugato et al., 2018; Favarin, 2018; Steenbeek & Weisburd, 2016).

More recently, a growing number of scholars have begun to focus their attention on crime concentration and crime forecasting patterns in Latin American cities (Chainey & Monteiro, 2019; Chainey & Muggah, 2020; Chainey et al., 2019; de Melo et al., 2018; Escudero & Ramírez, 2018; Giménez-Santana et al., 2018; Muggah & Tobón, 2019). Chainey et al. (2019) affirmed that the study of crime concentration at micro-places has hitherto had a very western-industrialized focus, despite the fact that crime appears to be concentrated at higher levels in Latin American cities compared to Western-industrialized cities. Homicide and violent assaults are especially concentrated in Brazil (Ceccato, 2005; Ceccato et al., 2007; de Melo et al., 2018; Muggah & Tobón, 2019; Pereira et al., 2017). It follows that data-driven policing and targeted crime prevention strategies can help to reduce violence in Brazil’s urban contexts. However, to be able to do so, further analyses are needed that shed light on both the spatial patterns of crime and factors that cause the concentration of violent events within specific areas.

The present article aims to better understand the spatial dynamics of violent crimes in Brazil, specifically in Fortaleza, before proceeding to propose prevention policies based upon the characteristics of urban spaces and the territorial behavior of the offenders in the city. Pioneering studies in Brazil has highlighted the spatial association between violent crimes and several characteristics of the Brazilian context at micro-places. Poverty, inequality, social disparities, competition for scarce resources, bureaucratic indifference, absence of responsible adults, differences in land use, proximity to drug markets, availability of firearms, proximity to transport nodes, and poor electrification policies in the rural areas emerged as relevant determinants of violence in Brazil (Arvate et al., 2018; Cardia et al., 2003; Ceccato, 2005; Ceccato et al., 2007; Minamisava et al., 2009). Thus, the extant research in Brazil highlights the importance of both social disorganization factors (e.g., poverty, inequality) and situational factors (e.g., proximity to drug markets, availability of firearms) as determinants of violence in several Brazilian urban settings (i.e., São Paulo, Campinas, Goiânia). de Oliveira et al. (2019) started to investigate the spatial patterns of violent crime in Fortaleza and their relationship with local development. Specifically, they highlighted how “Fortaleza responds, on average, for 43.8% of all violent crimes in the State of Ceará, and for 3.0% of all violent crimes in Brazil between 2009 and 2015” (de Oliveira et al., 2019, p. 150). The main findings of their study are that high violent crime rates are related to low-income neighborhoods, high spatial isolation of poor households, and low access to urban infrastructure, along with the high prevalence of illiterate adolescents and young Black males (de Oliveira et al., 2019). The goal of the present article is to expand extant knowledge on the spatial distribution of violence in the city of Fortaleza by both focusing on a smaller unit of analysis and using more recent data (up to 2018) in order to investigate where violence concentrates at the microlevel and whether these crime concentrations are related to vulnerable areas in terms of socio-economic-demographic disorder conditions. In addition, the study pinpoints specific vulnerable areas in which crime prevention policies should seek to focus their attention.

Empirical Strategy of the Present Study

To conduct the analyses, both geo-referenced data (e.g., lethal and intentional violent crime [LIVC], Drug trafficking offenses) and census data (e.g., socio-economic-demographic factors) were collected from multiple sources and aggregated at a larger entity called “aggregate sector.” As a starting point, this research used the map of census sectors that corresponds to the 2010 census survey, which was obtained from the official website of the Brazilian Institute of Geography and Statistics (www.ibge.gov.br). This map divided the city of Fortaleza into 3,043 census sectors. The census sectors are smaller units in comparison to the neighborhoods, and, as such, they can suffer from variability in terms of either the incidence of crimes from 1 year to the next or between one space and another. For this reason, the 3,043 census sectors were aggregated into 493 aggregated sectors. These aggregate sectors were formed out of clusters of contiguous census sectors. As a criterion, favelas, precarious settlements, and clusters of violence were incorporated into the same aggregate space, in an effort to make it more homogeneous in relation to the phenomenon studied. The GeoDa Skater tool was used to aggregate and intersect the 3,043 census sectors and 841 precarious settlements defined by the City Hall of Fortaleza in 2016 (www.observatoriodefortaleza.fortaleza.ce.gov.br).

The empirical strategy adopted in this study to (a) investigate spatial patterns of violence and vulnerable micro-territories and (b) assess the statistical and spatial relationship between violence and vulnerabilities in micro-territories combines multiple data and methods in order to investigate spatial patterns and, in turn, put forward possible solutions.

Investigate Spatial Patterns of Violence and Vulnerable Micro-Territories

To investigate the spatial patterns of violence and vulnerable micro-territories, this study (i) constructs the SVI to measure the concentration of violence at the aggregate sectors level, (ii) constructs the RAI to measure socio-economic-demographic disorder vulnerabilities in the aggregate sectors, and (iii) conducts ESDA of the indexes using a wide range of techniques to highlight potentially similar patterns and correlations.

Construction of the SVI

The SVI measures violence in the city of Fortaleza. The SVI was constructed by running a PCA on the variables included in Table 1. The data were retrieved from the Secretaria de Segurança Pública e Defesa Social do Governo do Estado do Ceará (www.sspds.ce.gov.br) in absolute values. Drug trafficking refers to reports of drug trafficking activities that were confirmed by on-site policing in 2016, 2017, and 2018. The LIVC was retrieved for this 3-year period, with the majority of crimes included in the LIVC being homicides. Indeed, in 2018, homicides accounted for 98.2% of the LIVC in Fortaleza. In addition, we also included in the PCA the home location of 12,565 prisoners who were arrested in Fortaleza in December 2018. It is assumed that criminals operate close to their homes and where they have knowledge of the environment. The positive correlation (ρ = .60) between LIVC2018 and PrisonAddr2018 indicates that the addresses of prisoners correlate with the most violent areas.

Table 1.

Variables Used to Construct the Spatial Violence Index.

Label	Description
Traffic2016	Drug trafficking (2016)
Traffic2017	Drug trafficking (2017)
Traffic2018	Drug trafficking (2018)
LIVC2016	Lethal and intentional violent crime (2016)
LIVC2017	Lethal and intentional violent crime (2017)
LIVC2018	Lethal and intentional violent crime (2018)
PrisonAddr2018	Home address of Fortaleza prisoners (2018)

Source. Authors’ elaboration of data retrieved from the Secretaria de Segurança Pública e Defesa Social do Governo do Estado do Ceará.

To check if the variables were compatible with the common variability needed to perform the PCA, the Kaiser–Meyer–Olkin (KMO) and Bartlett Test were used. The KMO test indicates the degree of explanation of the data based on the factors found in the analysis (Regazzi, 2001).

In order to construct the index with the relevant factors, the respective eigenvalues of the factor loadings were weighted with the weights, creating the weighted average, the Gross Index.

I B = \frac{\sum_{i = 1} (w_{i} F_{i})}{\sum_{i = 1} w_{i}},

where IB = Gross Index, w_i = proportion of variance explained by each factor, and F_i = factor scores.

The IB index is standardized, causing both a zero average and a standard deviation equal to 1. The index will be in the range between 0 and 1; the closer to 1, the stronger the violence. The adequacy of the sample used by the KMO test with a value of 0.825.

The variables used in the construction of the SVI, defined in Table 1, were condensed into two factors that, according to Table 2, explain 66.6% of the total variance of the data. They are: F-1 (LIVC and home address of prisoners) and F-2 (drug trafficking). The rotation method used was Varimax with Kaiser Normalization.

Table 2.

Percentage Explained for each Factor and Accumulated Variance.

Factor	% Variance	% Accumulated
F1	50.19	50.19
F2	16.41	66.60

Construction of the RAI

The RAI measures the degree of restrictive ambience, which is characterized by vulnerable urban spaces and areas that are fertile ground for violence. The index includes the main socio-economic-demographic and disorder variables utilized in extant literature in order to identify potential vulnerable and risky environments. Table 3 presents the main variables that were used to construct the RAI, along with a description of each variable and the corresponding bibliographic references.

Table 3.

Variables Used to Construct the Restrictive Ambient Index (RAI).

Theme	Variable	Description	Bibliographic References
Income	Income <= ½ mw	Persons aged 10 years or over with a nominal monthly income until 1/2 minimum wages	Briceño-León (2002), Cerqueira & Lobão (2003), Batella & Diniz (2010), Cahill & Mulligan (2003), Ribeiro & Cano (2016), and Plassa & Parré (2019)
	IncomePercp	Average per capita nominal monthly income—permanent private households
	House <= 1 mw	Private households and a nominal monthly income per capita—0–1 MW
Income inequality	IncomeIneq	Income Inequality Index	Morenoff et al. (2001), Briceño-León (2002), Cerqueira & Lobão (2003), Ribeiro & Cano (2016), Plassa & Parré (2019), and Chainey & Muggah (2020)
Education (literate)	ResponsNLitP	Responsible not literate people	G. Becker (1968), Batella & Diniz (2010), Ribeiro & Cano (2016), and Chainey & Muggah (2020)
Poverty—structure	Up to 1WC	Permanent private households with up to one bathroom for the exclusive use of residents	G. Becker (1968), Batella & Diniz (2010), and Chainey & Muggah (2020)
Population structure	Youth	People from 15 to 19 years old	Glaeser & Sacerdote (1999), Cahill & Mulligan (2003), Fallahi et al. (2012), Ribeiro & Cano (2016), and K. L. Becker & Kassouf (2017)
	Men	Men residing in private households and collective households
	Density	Density = inhabitants per hectare
Disorder	Drug (16 + 17)	Drug use = sum 2016 + 2017	Kelling & Wilson (1982), Skogan (1990), and Cahill & Mulligan (2003)
	Drunk (16 + 17)	Drunkenness = sum 2016 + 2017
	Disturb (16 + 17)	Disturbance to the quietness of others = sum 2016 + 2017
	Dis (16 + 17)	Disorder = sum 2016 + 2017

Source. Authors’ elaboration of data retrieved from Socioeconomic Data IBGE—2010 census, SSPDS-Ce, and Integrated Security Operations Center (ISOC).

The demographic variables were measured in absolute values with the exception of “Density” (Inhabitants per hectare) and “IncomeIneq” (indicator from −1 to +1). The variables related to income were used to verify the distribution and the level of income. Income values were restricted to a minimum wage for variables related to income per household and per capita in order to capture areas that are more vulnerable. The variable “IncomePercp” can be considered a protection variable, insofar as it is assumed that the higher the per capita monthly income of households, the lower the incidence of violence. The variable “up to 1WC” aims to spatially mirror the most precarious households. The variables “Men” and “Youth” (15–19 years old) attempt to capture the fact that, according to the Secretariat of Public Security and Social Defense of the State, 93% of homicide victims in Ceará in the period under consideration were male, while 64% of them were young people. Previous studies have indicated that education is related to urban violence, which is why the variable “ResponsNLitP” was inserted, as we would expect there to be a positive relationship with violence.

According to Morenoff et al. (2001), the “IncomeIneq” variable is an indicator of income inequality within the census sector and is calculated according to the following formula:

I d_{renda} = \frac{P - R}{R + P},

where P is the total number of people responsible for the household with an income up to one minimum wage, while R is the total number of people responsible for the household with an income higher than fifteen minimum wages. The “IncomeIneq” index ranges from −1 to +1, with −1 representing the extreme of wealth and +1 representing the extreme of poverty. It is expected that an increase in the variable representing the extreme of poverty is positively related to violence. The variables related to urban disorder (i.e., drug use, drunkenness, disturbance to the quietness of others, physical disorder) are measured in absolute values (number of events) and were collected by the Integrated Security Operations Center through complaints and subsequently confirmed by police patrols. In order to both decrease the fluctuation across geographical area and better capture the relationship between disorder and local vulnerabilities, the events that transpired in 2016 and 2017 were added.

The adequacy of the sample used by the KMO test was verified with a value of 0.914. The 13 variables that were used in the construction of the index were condensed into three factors: F.1—social vulnerability, F.2—disorder, and F.3—income inequality. These factors accounted for 89.1% of the total data variance. The rotation method used was Varimax with Kaiser Normalization as shown in Table 4.

Table 4.

Percentage Explained for Each Factor and Accumulated Variance.

Factor	% Variance	% Accumulated
F1	66.5	66.5
F2	14.4	80.9
F3	8.2	89.1

ESDA

ESDA are techniques for describing and visualizing spatial distributions as well as discovering associative patterns (Anselin, 1995). One of the ESDA techniques is the Moran’s index that was used as follows to assess whether the hypothesis that spatial data are randomly distributed was true:

I = \frac{\sum_{i = 1}^{n} . \sum_{j = 1}^{n} w_{i j} (z_{i} - \overset{´}{z}) (z_{j} - \overset{´}{z})}{\sum_{i = 1}^{n} {(z_{i} - \overset{´}{z})}^{2}},

where n is the number of areas, z_i is the attribute value of area i, $\overset{´}{z}$ is the average attribute value in the region, and w_ij are the elements of the normalized spatial proximity matrix. Positive values indicate a direct correlation, while negative values indicate an inverse correlation.

In order to check specific patterns within a large number of areas, the Local Moran’s Index was used as follows:

I_{i} = \frac{z_{i} \sum_{j = 1}^{n} w_{i j} z_{j}}{\sum_{j = 1}^{n} z_{j}^{2}} .

With regard to the identification of local spatial clusters, the method known as Local Indicators of Spatial Association was used. BoxMap is Moran’s Mirroring Diagram. “High-high” correlations show sectors with high proportions of the indicator surrounded by sectors with high proportions of the same indicator; “Low-low” correlations show sectors with low proportions of the indicator surrounded by sectors with low proportion of the same indicator; “high-low” and “low-high” are sectors of transition of the indicator. A bivariate global spatial autocorrelation coefficient was also used to discover if the value of an observed variable in a given area is related to the values of the variable observed in neighboring areas (Almeida, 2012).

Assess the Statistical and Spatial Relationship Between Violence and Vulnerabilities in Micro-Territories

To assess the statistical and spatial relationship between violence and urban vulnerabilities at the aggregate sectors level, we ran spatial autoregressive (SAR) model and SEM. The SAR model verifies whether the dependent variable y is influenced by the values observed in the neighboring regions (Wy; Almeida, 2012):

y = ρ W y + β X + ∊,

where y is the dependent variable, W is the spatial proximity matrix, and Wy is a vector n-by-1 that expresses the spatial lags and is determined by the average of the values of the dependent variable observed in the neighborhood. ρ is the SAR coefficient, X is the set of exogenous explanatory variables, while ∊ is the error term that translates the random influence.

In the SEM, the spatial dependency is residual with a first-order autoregressive structure in the term error:

y = β X + ∊ com ∊ = λ W ∊ + ν,

y = β X + ∊ com ∊ = λ W ∊ + ν,

where y is the dependent variable, W is the spatial proximity matrix, and λW∊ is the autoregressive term that measures the degree of spatial dependence on the estimated model’s error term. λ is the autoregressive coefficient and β is the vector of parameters related to the independent variables. ν is the error term that translates the random influence.

Results

Investigate Spatial Patterns of Violence and Vulnerable Micro-Territories

There is a strong Pearson correlation (ρ = .73) between vulnerable areas (represented by RAI) and violent areas (represented by SVI). Figure 1A and B show the spatial distribution of RAI and SVI via the thematic map that separates the aggregated sectors into four equal deciles. There is a spatial association between 25% of the most violent areas (Figure 1B—red) and 25% of the most vulnerable areas (Figure 1A—red). Similarly, one can discern a correspondence between highly violent aggregate sectors (Figure 1B—red), highly vulnerable aggregate sectors (Figure 1B–red), and precarious settlements (Figure 1C—brown).

In order to assess the existence of patterns of spatial distribution, the Kernel density estimator can be used. Given that there is a strong positive correlation (ρ = .69) between the RAI and the LIVC, the Kernel density was calculated on the LIVC offenses registered in the years 2012, 2014, 2016, and 2017 (Figure 2). This analysis gives a clear impression of the spatial inertia of violence in the city of Fortaleza, insofar as LIVC offenses appear to be concentrated in the same areas despite changes in the intensity of the hot spots (red areas).

Figure 1.

Thematic map of (A) Restrictive Ambient Index, (B) Spatial Violence Index, and (C) precarious settlements.

Figure 2.

Lethal and intentional violent crime Kernel density as a proxy of homicides (Years 2012, 2014, 2016, and 2017).

When comparing Figure 2 and Figure 1A, it becomes evident how violent hot spots tend to concentrate in the most vulnerable areas where the RAI index is higher. This might be indicative of a potential spatial association between violent crimes and highly vulnerable places. However, further statistical tests are required in order to confirm the existence of this spatial relationship.

With a significant Bivariate Moran Global I of 0.29, it was found that the value of the SVI, in a given area, is related to the values of the RAI in neighboring areas. The variables that represent the violence of the SVI (i.e., LIVC, PrisonAddr18, Drug Traffic 17) are positively self-correlated in space with the RAI, achieving a significance of 1% for all. This means that aggregate sectors with a high level of RAI are associated with neighboring areas characterized by high levels of LIVC and drug trafficking.

Variables representing vulnerable areas (i.e., Disturb [16 + 17], Density, ResponsNLitP, IncomePercp, IncomeIneq) are positively correlated in the space with violence (SVI), achieving a significance of 1% for all. This means that aggregate sectors characterized by a high level of violence (SVI) are associated with neighboring vulnerable areas, which is to say that these vulnerability variables thus can potentiate the occurrence of violence.

Map A in Figure 3 shows the bivariate Moran’s I for the SVI-RAI relationship, where the aggregate sectors with the greatest level of violence are those that include a highly vulnerable neighborhood (RAI), type HH (high-high). In addition, these areas are spatially close to areas of transition from low violence to areas of high violence according to quadrant LH (low-high). Map B in Figure 3 shows the BoxMap for the SVI index. There is a spatial pattern in the clusters for the index plotted on the respective map. Relationships of the type high-high in red depicts aggregate sectors with high levels of violence surrounded by aggregate sectors that are also highly violent.

Figure 3.

BoxMap: (A) Bivariate Moran’s I between Spatial Violence Index (SVI)–Restrictive Ambient Index (RAI) and (B) univariate Moran’s I for SVI.

Assess the Statistical and Spatial Relationship Between Violence and Vulnerabilities in Micro-Territories

The RAI is composed of three vulnerability factors: F.1—social vulnerability, F.2—disorder, and F.3—income inequality. These factors represent the explanatory variables, while SVI was used as the dependent variable. OLS, SEM, and SAR models were estimated. As part of the model selection process, the existence of spatial autocorrelation was verified using the Lagrange multiplier and the robust Lagrange multiplier to ensure that the most appropriate spatial model was adopted. The Moran Index of 0.251 was significant for the residues of the OLS model, thus confirming spatial autocorrelation. According to Table 5, the diagnosis of the likelihood ratio was significant for the models, which demonstrates that both the spatial error and spatial lag models can be used to replace the classic model. According to the Akaike info criterion (AIC), the SEM has the best fit and explanatory power as a result of having the lowest AIC and highest value in the likelihood function.

Table 5.

Results of the Econometric Analyses.

	OLS		SEM		SAR
R2	.58		.64		.62
Variables	Coefficients	t Value	Coefficients	z Value	Coefficients	z Value
W_IVE ( $ρ$ )					0.2830	6.1829
Constant	−0.0659	−2.7056	−0.0673	−2.1502	−0.0790	−3.4304
F.1—Social vulnerability	0.4399	23.2072	0.4363	22.8200	0.3970	19.9780
F.2—Disorder	0.1669	8.1613	0.1777	7.7068	0.1350	6.5891
F.3—Income inequality	0.2094	8.2616	0.2075	6.3801	0.1520	5.8676
Lambda ( $λ$ )			0.4526	8.0082
Tests		Prob.		Prob.		Prob.
Log likelihood	632.52		663.35		652.32
Likelihood ratio			61.65	0.000	39.58	0.000
Akaike info criterion	−1.257.05		−1,318.71		−1,294.60
Lagrange multiplier		0.000	85.26	0.000	49.68	0.000
Robust LM		0.000	35.97	0.000	0.39	0.532

By observing the results of the SEM, the F.1—social vulnerability shows, due to the marginal effect, a 44% impact on violence for each percentage change in the level of social vulnerability. In other words, the vulnerable socioeconomic structure that predominates in the RAI is a major generator of violence, and variations in this social situation have a strong impact. The factors F.2—disorder and F.3—income inequality are also significant and impacting. The λ coefficient was positive and significant, thus indicating that the factors not observed in a given sector not only affect the rate of violence in that location but also in neighboring sectors.

The SAR had a relevant and significant ρ, thus suggesting that the level of violence within a given aggregate sector positively influences the values of violence in the neighborhood as a whole. There is, then, a correlation between local vulnerabilities to violence.

Suggestions for the Allocation of Resources and Potential Crime Prevention Strategies

According to our results, the concentration of violence is largely located in vulnerable areas in Fortaleza. There is a spatial association between violent crimes and aggregate sectors of strong vulnerability in the city, which was confirmed by the statistical correlation between the SVI and the different components of the RAI. Hence, from a crime prevention perspective, targeting the most vulnerable micro-territories in the city would lead to an overall reduction in violent crime rates due to the fact that violence and urban vulnerabilities are strongly associated in this context. Therefore, according to a cost-benefit analysis and a strategic allocation of resources, the 124 aggregate sectors that belong to the highest RAI decile should constitute the primary target of crime prevention policies in Fortaleza (Figure 4). Given the scarcity of resources and the attendant need to maximize the benefits of crime prevention policies, it may prove beneficial to target only very specific high-risk aggregate sectors were both a high level of violence and high urban vulnerabilities are concentrated. These places represent 21% of the aggregate sectors in Fortaleza and account for 50% and 45% of LIVC and drug trafficking offenses, respectively (Figure 4). For these reasons, they should be considered as the primary areas of interest for policy makers and police authorities seeking to engender an overall decrease in violent crime trends in the city.

Figure 4.

Sociodemographic and violence composition of high-risk aggregate sectors.

Fortaleza suffers from the spatial inertia of violence, insofar as LIVC offenses appear to remain concentrated in the same areas, despite changes in the intensity of violent hot spots in certain areas (Figure 2). In such situations of spatial inertia, violence tends to remain restricted and constant across space and time, until some force either increases or decreases its intensity. This force can either be positive—such as, for example, through an efficient public policy that aims to reorganize the urban space, the inhibition of local disorders, or the restoration of collective efficacy—or it can be a negative, such as the rise to dominance of a criminal faction or an escalation of violence due to a drug trafficking war. The key to both decreasing the intensity of these violent hot spots and developing a solution to the crime problem in the city cannot be unanimous, but rather should seek to employ a multifaceted strategy that, in this particular instance.

That is to say, public policies must focus on the 124 identified micro-territories where violent crimes and local vulnerabilities are concentrated. Different types of policies should orbit around these 124 micro-territories to both enhance the power of crime prevention and reduce crime opportunities. Given that vulnerable environments generate manifold unfavorable conditions, a targeted strategy for crime prevention should combine (i) ostensible policing in order to increase offenders’ perceptions of the costs of crime; (ii) community policing in order to establish a police presence and facilitate the development of a trusting relationship within the community; (iii) public policies, such as, among other things, investing in improving urban infrastructure and developing more dignified social conditions; (iv) fomentation of the social capital of the territory in which the active participation of local organizations and the wider community can play a vital role in terms of social control, facilitating communication between people and generating commitment toward the social good; and (v) public policies designed to strengthen the welfare state and educational programs.

Discussion of the Results

Violent crimes are concentrated in micro-territories in Fortaleza. This finding aligns with the results of previous studies that also observed high level of crime concentration in Latin America and Brazil (Ceccato, 2005; Chainey et al., 2019; de Melo et al., 2018; de Oliveira et al., 2019). The exploratory spatial analysis conducted in this study demonstrated the existence of spatial dependence as well as the spatial association between violence (SVI) and urban vulnerabilities (RAI). From this observed spatial dependence, this study thus verified the correlation between different types of vulnerabilities and violence, by virtue of assessing how local vulnerabilities constitute one of the main determinants of both violence and its concentration within the micro-territories in Fortaleza.

Our analyses corroborate the correlation highlighted by de Oliveira et al. (2019) between social vulnerability, income inequality, and violent crimes in Fortaleza. Indeed, these authors found that high violent crime rates are related to low-income neighborhoods, high spatial isolation of poor households, low access to urban infrastructure, along with the high prevalence of illiterate adolescents and young Black males. In our analyses, the components related to the vulnerable socioeconomic structure that predominates in the RAI are a major generator of violence, and variations in this socioeconomic situation have a strong impact on the SVI. In addition, our findings stress also the importance of situational factors such as disorder as a determinant of violence in the city. Measures of disorder included in the construction of our RAI were reported complaints of drug use, drunkenness, disturbance to the quietness of others, and physical disorders. Thus, both situational and socioeconomic characteristics seem to play a central role in explaining spatial distribution of violence in Fortaleza. These results are in line with the international literature on spatial patterns of violent crime (e.g., Andresen et al., 2016, Escudero & Ramírez, 2018; Giménez-Santana et al., 2018) as well as with prior studies conducted in Brazil (e.g., Ceccato et al., 2007). Ceccato et al. (2007) found the existence of several independent dimensions that explain the spatial variation in homicide rates in São Paulo (i.e., poverty, situational factors such as transport nodes, criminogenic conditions generated by the availability of weapons or drug-related activity, and the spillover effect of homicides). Evidences from our analyses also support the association between concentration of violence and income inequality, situational factors such as disorder, and the spillover effect of violence (i.e., micro-territories with high levels of violence are spatially close to micro-territories with high levels of violence in Fortaleza).

By considering prior studies conducted in the Brazilian context, future analyses focusing on Fortaleza should include measures related to mobility (e.g., proximity to bus stops and railway stations) and availability of weapons that have proven to be important explanatory variables in other Brazilian cities such as São Paulo as well as the low access of the population to urban infrastructure (e.g., sidewalks and street lighting) that has been associated with high violent crime rates in Fortaleza (Ceccato et al., 2007; de Oliveira et al., 2019). Future research might also expand the analysis presented in this article to include different types of crime. Indeed, other patterns of crime concentration could emerge (e.g., concentrations in different areas of the city) or, alternatively, hot spots of property crime could converge in the same places where violent hot spots are located. In this respect, spatial dependence, spatial auto-correlation, and the statistical correlation between concentration of other crimes and urban vulnerabilities should be assessed to better analyze patterns of crime and causal relationships in the city. By identifying spatial patterns of crime and understanding the urban vulnerabilities connected to crime concentrations, it would be easier for policy makers to make choices that are more efficient and enhance the chances to cost-effectively reduce crime. In order to be able to fully do so, future research should also consider the importance of temporal variations of killings in Brazilian cities. Indeed, Ceccato (2005) stressed how homicides in São Paulo take place particularly during vacations, evenings, and weekends. This finding is also confirmed by Pereira et al. (2016) who found statistically significant increases in homicides during the evenings and weekends in Recife.

Conclusions

This work contributes to the scarce literature on spatial concentration of violence in Brazil by analyzing not only the distribution of violence in Fortaleza but also the distribution of different urban vulnerabilities (i.e., social, economic, and situational) and by constructing composite indexes to better measure the complexity of violence and vulnerabilities in the city. In particular, this study creates a SVI and a Restrictive Ambient Index (RAI) to investigate the spatial dynamics of violent crimes and urban vulnerabilities in Fortaleza. Both ESDA and spatial regression models were employed to analyze the associative patterns and measure the statistical correlation between the two indexes. In doing so, this study expands the extant knowledge on spatial patterns of violence in Fortaleza by both focusing on a smaller unit of analysis and using more recent data compared to prior studies.

The results demonstrate how locations characterized by high levels of violence are spatially correlated with more vulnerable locations in terms of both socio-economic-demographics and urban disorder. This evidence raises numerous possibilities in terms of crime prevention strategies and predictive policing in Fortaleza, Brazil, and Latin America. Applications of evidence-based policing, data-driven policies, and community-oriented strategies to reduce crime have hitherto primarily been conducted in the United States and other Western-industrialized cities, but they could also find fertile ground in Latin America. More specifically, our study proposes the use of a targeted strategy that would combine a wide range of public policies in the 124 micro-territories where there is a high-risk of concentration of violent crimes and urban vulnerabilities. These areas should constitute the primary target of public policies designed to reduce vulnerabilities and enhance socioeconomic conditions in conjunction with the implementation of both preventive policing and community-based policing measures. Indeed, a strategic allocation of resources based on a cost-benefit analysis approach would help at first to prioritize specific areas to maximize the results in terms of violence reduction.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Almeida

(2012). Applied spatial econometrics. Alínea Publisher.

Andresen

M. A.

Linning

S. J.

Malleson

(2016). Crime at places and spatial concentrations: Exploring the spatial stability of property crime in Vancouver BC, 2003–2013. Journal of Quantitative Criminology, 33, 1–21.

Anselin

(1995). Local indicators of spatial association—LISA. Geographical Analysis, 27(2), 93–115.

Arvate

Falsete

F. O.

Ribeiro

F. G.

Souza

A. P.

(2018). Lighting and homicides: Evaluating the effect of an electrification policy in rural Brazil on violent crime reduction. Journal of Quantitative Criminology, 34(4), 1047–1078.

Batella

W. B.

Diniz

A. M. A.

(2010). Spatial analysis of violent crime constraints in the State of Minas Gerais. Society & Nature Magazine, 22(1), 151–163.

Becker

(1968). Crime and punishment: An economic approach. Journal of Political Economy, 76, 169–217.

Becker

K. L.

Kassouf

A. L.

(2017). An analysis of the effect of public spending on education on crime in Brazil. Economy and Society, 26(1), 215–242.

Bowers

K. J.

Johnson

S. D.

(2005). Domestic burglary repeats and space-time clusters. European Journal of Criminology, 2, 67–92.

Braga

A. A.

Clarke

R. V.

(2014). Explaining high-risk concentrations of crime in the city: Social disorganization, crime opportunities, and important next steps. Journal of Research in Crime and Delinquency, 51(4), 480–498.

10.

Braga

A. A.

Turchan

Papachristos

A. V.

Hureau

D. M.

(2019). Hot spots policing of small geographic areas effects on crime. Campbell Systematic Reviews, 15(3), e1046.

11.

Braga

A. A.

Weisburd

(2010). Policing problem places: Crime hot spots and effective prevention. Oxford University Press.

12.

Braga

A. A.

Weisburd

Waring

E. J.

Green

Spelman

Gajewski

(1999). Problem-oriented policing in violent crime places: A randomized controlled experiment. Criminology, 37(3), 541–580.

13.

Briceño-León

(2002). The new urban violence in Latin America. Sociologies, 4(8), ano 4. jul./dez, 34–51.

14.

Cahill

M. E.

Mulligan

G. F.

(2003). The determinants of crime in Tucson, Arizona. Urban Geography, 24, 582−610.

15.

Caplan

J. M.

Kennedy

L. W.

(2016). Risk terrain modeling. Crime prediction and risk reduction. University of California Press.

16.

Cardia

Adorno

Poleto

F. Z.

(2003). Homicide rates and human rights violations in São Paulo, Brazil: 1990 to 2002. Health and Human Rights, 6(2), 14–33.

17.

Ceccato

(2005). Homicide in São Paulo, Brazil: Assessing spatial-temporal and weather variations. Journal of Environmental Psychology, 25(3), 307–321.

18.

Ceccato

Haining

Kahn

(2007). The geography of homicide in São Paulo, Brazil. Environment and Planning A: Economy and Space, 39(7), 1632–1653.

19.

Ceccato

Oberwittler

(2008). Comparing spatial patterns of robbery: Evidence from a Western and an Eastern European city. Cities, 25(4), 185–196.

20.

Cerqueira

Lobão

(2003). Determinants of crime: a review of theoretical models and empirical results (p. 956). Text for Discussion-IPEA.

21.

Chainey

S. P.

Muggah

(2020). Homicide concentration and retaliatory homicide near repeats: An examination in a Latin American urban setting. The Police Journal: Theory, Practice and Principles. 1–21. https://doi.org/10.1177/0032258X20980503.

22.

Chainey

S. P.

Monteiro

(2019). The dispersion of crime concentration during a period of crime increase. Security Journal, 32(3), 324–341.

23.

Chainey

S. P.

Pezzuchi

Guerrero Rojas

N. O.

Hernandez Ramirez

J. L.

Monteiro

Rosas Valdez

(2019). Crime concentration at micro-places in Latin America. Crime Science, 8(1), 5.

24.

de Melo

S. N.

Andresen

M. A.

Matias

L. F.

(2018). Repeat and near-repeat victimization in Campinas, Brazil: New explanations from the Global South. Security Journal, 31(1), 364–380.

25.

de Oliveira

V. H.

de Medeiros

C. N.

Carvalho

J. R.

(2019). Violence and local development in Fortaleza, Brazil: A spatial regression analysis. Applied Spatial Analysis and Policy, 12(1), 147–166.

26.

Dugato

(2013). Assessing the validity of risk terrain modeling in a European city: Preventing robberies in Milan. Crime Mapping, 5(1), 63–89.

27.

Dugato

Favarin

Bosisio

(2018). Isolating target and neighbourhood vulnerabilities in crime forecasting. European Journal on Criminal Policy and Research, 24(4), 393–415.

28.

Escudero

J. A.

Ramírez

(2018). Risk terrain modeling for monitoring illicit drugs markets across Bogota, Colombia. Crime Science, 7(1), 3.

29.

Fallahi

Pourtaghi

Rodríguez

(2012). The unemployment rate, unemployment volatility, and crime. International Journal of Social Economics, 39(6), 440–448.

30.

Favarin

(2018). This must be the place (to commit a crime). Testing the law of crime concentration in Milan, Italy. European Journal of Criminology, 15(6), 702–729.

31.

Gill

Wooditch

Weisburd

(2016). Testing the “Law of crime concentration at place” in a suburban setting: Implications for research and practice. Journal of Quantitative Criminology, 33, 1–27.

32.

Giménez-Santana

Caplan

J. M.

Drawve

(2018). 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(4), 417–431.

33.

Glaeser

E. L.

Sacerdote

(1999). Why is there crime in cities? Journal of Political Economy, 107(6), 225–258.

34.

Haberman

C. P.

Sorg

E. T.

Ratcliffe

J. H.

(2016). Assessing the validity of the law of crime concentration across different temporal scales. Journal of Quantitative Criminology, 33, 1–21.

35.

Hipp

J. R.

Kim

Y.-A.

(2016). Measuring crime concentration across cities of varying sizes: Complications based on the spatial and temporal scale employed. Journal of Quantitative Criminology, 33, 1–38.

36.

Johnson

S. D.

(2010). A brief history of the analysis of crime concentration. European Journal of Applied Mathematics, 21(4–5), 349–370.

37.

Kelling

G. L.

Wilson

J. Q.

(1982). Broken windows. The police and neighborhood safety. The Atlantic, 29–38.

38.

Kennedy

L. W.

Caplan

J. M.

Piza

E. L.

(2011). Risk clusters, hotspots, and spatial intelligence: Risk terrain modeling as an algorithm for police resource allocation strategies. Journal of Quantitative Criminology, 27(27), 339–362.

39.

Levin

Rosenfeld

Deckard

(2016). The law of crime concentration: An application and recommendations for future research. Journal of Quantitative Criminology, 33, 1–13.

40.

Meijer

Wessels

(2019). Predictive policing: Review of benefits and drawbacks. International Journal of Public Administration, 42(12), 1031–1039

41.

Minamisava

Nouer

S. S.

de Morais Neto

O. L.

Melo

L. K.

Andrade

A. L. S.

(2009). Spatial clusters of violent deaths in a newly urbanized region of Brazil: Highlighting the social disparities. International Journal of Health Geographics, 8(1), 66.

42.

Morenoff

J. D.

Sampson

R. J.

Raudenbush

S. W.

(2001). Neighborhood inequality, collective efficacy, and the spatial dynamics of urban violence. Research Report, 39(3), 517–558.

43.

Muggah

Tobón

K. A.

(2019). Reducing Latin America’s violent hot spots. Aggression and Violent Behavior, 47, 253–256.

44.

Pereira

D. V. S.

Andresen

M. A.

Mota

C. M. M.

(2016). A temporal and spatial analysis of homicides. Journal of Environmental Psychology, 46, 116–124.

45.

Pereira

D. V. S.

Mota

C. M. M.

Andresen

M. A.

(2017). The homicide drop in Recife, Brazil: A study of crime concentrations and spatial patterns. Homicide Studies, 21(1), 21–38.

46.

Plassa

Parré

J. L.

(2019). Violence in the state of Paraná: a spatial analysis of homicide rates and socioeconomic factors. Economy and Development. 31(1), 01–16.

47.

Ratcliffe

J. H.

Taniguchi

Groff

E. R.

Wood

J. D.

(2011). The Philadelphia foot patrol experiment: A randomized controlled trial of police patrol effectiveness in violent crime hotspots. Criminology, 49(3), 795–831.

48.

Regazzi

A. J.

(2001). Multivariate analysis. Federal University of Viçosa, Exact and Technological Sciences Center. IT department.

49.

Ribeiro

Cano

(2016). Lethal victimization and inequality in Brazil—Evidence at the municipal level. Civitas, 16(2), 285–305.

50.

Seguridad, Justicia y Paz. (2021). 2019 Ranking Bulletin of the 50 most violent cities in the world. Security, Justice and Peace. Citizen Council for Public Safety and Criminal Justice A. C. Retrieved July 15, 2021, from http://www.seguridadjusticiaypaz.org.mx/sala-de-prensa/1590-boletin-ranking-2019-de-las-50-ciudades-mas-violentas-del-mundo

51.

Sherman

L. W.

Weisburd

(1995). General deterrent effects of police patrol in crime “hot spots”: A randomized, controlled trial. Justice Quarterly, 12(4), 625–648.

52.

Skogan

W. G.

(1990). Disorder and decline: Crime and the spiral of decay in American neighborhoods. Free Press.

53.

Steenbeek

Weisburd

(2016). Where the action is in crime? An examination of variability of crime across different spatial units in The Hague, 2001–2009. Journal of Quantitative Criminology, 32(3), 449–469.

54.

UNODC. (2021). Data UNODC. Statistics online. Retrieved July 15, 2021, from https://dataunodc.un.org/

55.

Weisburd

(2015). The law of crime concentration and the criminology of place. Criminology, 53(2), 133–157.

56.

Weisburd

Groff

E. R.

Yang

S.-M.

(2012). The criminology of place. Street segments and our understanding of the crime problem. Oxford University Press.