Association between racial residential segregation and homicide mortality in municipalities in Minas Gerais, Brazil

Abstract

This article aims to evaluate the association between racial residential segregation and homicide mortality in the state of Minas Gerais (MG), Brazil. We conducted an ecological study in which the units of analysis were municipalities in MG. The outcome was homicide deaths between 2008 and 2012 and the exposure variable was residential segregation measured using the racial interaction index, calculated using data from the 2010 Demographic Census. The covariables were per capita family income and the Gini index. The variables were presented in tables and thematic maps and associations were measured using Bayesian hierarchical models. The results of the model adjusted for per capita family income showed a negative association between the racial interaction index and homicide mortality (coefficient=-1.787; 95%CI=-2.459; -1.119). Homicide mortality was lower in municipalities with higher levels of racial interaction.

Key words:
Social segregation; External causes; Spatial analysis; Health disparities

Resumo

O objetivo desse artigo é avaliar a associação entre segregação residencial racial e mortalidade por homicídios em Minas Gerais (MG). Trata-se de estudo ecológico, com os municípios de MG como unidades de análise. O desfecho foi óbitos por homicídio ocorridos de 2008 a 2012 e a exposição a medida de segregação residencial índice de interação racial, estimada para 2010 a partir dos dados do censo demográfico; outras covariáveis foram renda média per capita e índice de Gini. As variáveis foram apresentadas em mapas temáticos e a associação entre elas foi investigada por modelos hierárquicos bayesianos. Houve associação negativa entre o índice de interação racial e a mortalidade por homicídios (coeficiente=-1,787; IC95%=-2,459; -1,119), em modelo ajustado pela renda per capita. Os municípios de MG com maior interação racial tiveram em média uma menor mortalidade por homicídios.

Palavras-chave:
Segregação social; Causas externas; Análise espacial; Disparidades nos níveis de saúde

Introduction

One of the social determinants of health that has drawn interest from researchers recently is residential segregation, which is the separation of people and exclusion of segments of society for economic, social, cultural, racial, ethnic, or other reasons¹. The perception that individual factors alone do not fully explain the health-disease process generated interest in the study of the context in which people live. The study of neighborhood characteristics allows scholars to study this context because neighborhoods possess both social and physical attributes that are potentially determinants of health. An area of special interest is the study of social and racial inequalities in health, because place of residence is strongly patterned by social status position and ethnicity. To have an impact on living and working conditions, intersectoral health promotion initiatives should focus on the context in which people live². Measures of local residential segregation have high sensitivity for detecting associations with health outcomes³.

Residential segregation is a multidimensional phenomenon. While segregation measures show the extent to which groups are separated throughout a space, dimensions illustrate different aspects of this separation, although there may be a high degree of overlap between measures at an empirical level. The dimension “exposure” refers to the degree of potential contact between social group members within residential areas and one of the most widely used measures of exposure in the interaction index⁴.

In Brazil, external causes are the third leading cause of death, with homicide and transport accidents being the most common external causes of death among the one to 49-year age group. Brazil has the third highest homicide rate in South America, with rates being up to 10 times higher than in middle and high-income countries⁵.

Violence has a substantial impact on health systems, the economy and physical and emotional health. Homicide rates vary greatly across the country, with the highest rates being found in the North and Northeast. Mortality due to external causes disproportionately affects young men and is compounded in contexts of structural inequality and criminality^5.6. There are also substantial spatial variations in homicide and violent crime rates in the state of Minas Gerais⁷.

Despite the importance of this issue, we were unable to find any empirical studies in Brazil investigating the relationship between residential segregation and homicide mortality. Considering the importance of this cause of death and its relationship with social context, the aim of this study was to estimate the association between the racial interaction index and homicide rate in municipalities in Minas Gerais (MG), a large state characterized by substantial regional differences that reflect to some extent national differences.

Methods

We conducted a cross-sectional ecological study in which the units of analysis were municipalities in Minas Gerais (MG). The dependent variable (outcome) was the total number of homicide deaths in the period 2008 to 2012. This data was obtained from the Mortality Information System (SIM) via the website of the country’s national health information system, DATASUS, accessed on 27/11/2019. We used a five-year period for mortality data to reduce small area-related random variation, because many municipalities have small populations and few deaths.

The exposure measure was racial interaction index (RII), which represents the exposure dimension of residential segregation⁴. The indicator, which ranges from a low of 0 to a high of 1, was estimated for each municipality using data from the 2010 Demographic Census. The data were downloaded from the statistics page of the site of the Brazilian Institute of Geography and Statistics (https://www.ibge.gov.br/estatisticas/downloads-estatisticas.html) on 10/11/2019. On this page, we selected the directory Census, followed by Demographic_Census_2010, Universe_Results, and finally Breakdown_by_Census_Tract. From the latter, we downloaded the compacted file “MG_20171016.zip”, using the file “People03_MG.csv” to obtain the total number of residents by race/color (variables v001 to v006) in each census tract. The data were then aggregated into municipalities, because the calculation of the municipal racial interaction index requires data at both census tract and municipal level. Race/color was divided into two categories: “black” (total number of “black” and “brown” individuals); and “white”. RII was calculated using the following equation⁸:

$RII=\frac{{\displaystyle \sum _i^n\left|\frac{b_i}{B}\times \frac{a_i}{t_i}\right|}}{2}$

Where: RII=racial interaction index; a_i=population of group A (white) in sub-area i; b_i=population of group B (black) in sub-area i; B=population of group B in the larger area; t_i=total population in sub-area i; sub-area=census tracts; larger area=the municipality.

The covariables per capita family income and Gini index in 2010 were used in the analyses to control for confounding. This data was also obtained from the 2010 Demographic Census.

The statistical distribution of the variables was presented in tables. We also created thematic maps showing the distribution of indicators across quintiles and estimated global Moran’s I, which is a measure of general spatial correlation⁹.

We estimated standardized mortality ratios (SMR) by dividing the number of observed deaths (O_i) by the number of expected deaths (E_i) in each municipality⁹. To account for the differences in age structure between municipalities, expected deaths were estimated based on the total expected deaths in each age group obtained by multiplying the age-specific rates for the state (r_j ^s) by the population of the specific age group in each municipality (n_j):

$E_i={\displaystyle \sum _{j=1}^m{r}_j^{(s)}}{n}_j$

Where: E_i=expected deaths in each municipality i; r_j=mortality rate for each specific age group j; n_j=population in each specific age group j.

To determine the association between RII and homicide deaths, we ran Bayesian hierarchical models, estimating unstructured random effects (to adjust for overdispersion and capture the effects of unmeasured factors) and structured spatial random effects (to adjust for spatial dependence between observations). The spatially structured component was modelled with the conditional autoregressive (CAR), which smoothes the data according to a certain adjacency structure given by a neighborhood matrix that specifies two areas are neighbors if they have a common boundary¹⁰.

Expected deaths (E_i) were used as an offset variable in the models so that SMR could be estimated for each municipality, calculated by dividing observed deaths (O_i) by expected deaths (E_i) and adjusted by the covariables. Bayesian estimation was performed using the Integrated Nested Laplace Approximation (INLA) approach with the INLA library (http://www.r-inla.org/) of the R program (https://cran.r-project.org/). We stipulated a non-informative prior distribution, which is used when data is expected to exhibit dominance in the adjusted models. We ran models with and without the covariables (per capita income and Gini index). The tables present the means and 2.5 and 97.5 percentiles of the posterior distribution (95% credible interval - CI95%) for each coefficient. The deviance information criterion (DIC) was used to select the final model. The model is expressed as follows:

$y_i|{\phi }_i,{\upsilon }_i~NB({\mu }_i,{\theta }_i)$

$\log \left({\mu }_i\right)=\log \left(e_i\right)+\log \left({\rho }_i\right)=\log \left(e_i\right)+{\beta }_0+{\beta }_1{x}_{i1}+{\phi }_i+v_i$

${\phi }_i~N(\mathrm{0,}{\sigma }_{\phi }{}^2)$

${\upsilon }_i~CAR({\sigma }_{\upsilon }{}^2)$

$SMR={\rho }_i=e{\beta }_0+{\beta }_1{x}_{i1}+{\phi }_i+v_i$

Where: y=observed response variable (deaths); µ=estimated mean of the response variable; θ=estimated model parameters; e=expected value for the response variable; SMR=ρ=estimated standardized mortality ratio; x=explanatory variable; β=coefficients estimated by the model; σ=variance; φ=unstructured random effect; υ=structured spatial effect.

The data were tabulated in TabWin 4.1.52 and all analyses were performed using R 3.5.2.

Results

The SMR for homicide ranged between 0 and 4.67 (median 0.50). Per capita family income ranged between R$ 168 and R$ 1,710 (mean R$ 482), while the mean Gini index was 0.48 (standard deviation 0.05). The mean racial interaction index (RII) was 0.43 (standard deviation 0.05). All the variables had positive spatial autocorrelation, with higher values for RII, per capita family income and SMR (Table 1).

Table 1
Statistical distribution and spatial autocorrelation (Moran’s I) of the variables.

	Minimum	Q1	Q2	Mean	SD	Q3	Maximum	MI
Standardized mortality ratio	0.00	0.20	0.50	0.59	0.55	0.80	4.67	0.44
Racial interaction index	0.08	0.27	0.41	0.43	0.18	0.56	0.90	0.87
Per capita income (R$)	168	351	467	482	170	582	1710	0.55
Gini index	0.33	0.44	0.48	0.48	0.05	0.51	0.78	0.15

• Q1: percentile 25. Q2: median. Q3: percentile 75. SD: standard deviation. MI: Moran’s I; p-value<0.05 (obtained with 1,000 permutations).
• Source: Authors.

All the indicators showed an uneven spatial distribution. The SMR was higher in municipalities in the north and center of the state, which were also areas with a lower racial interaction index and per capita family income (Figure 1).

Figure 1
Spatial distribution maps of the variables in municipalities in Minas Gerais.

The selected regression model (the model with the lowest DIC) was model C2, which included the variables RII and per capita family income and unstructured random and structured spatial effects. There was a negative association between the RII and risk of death, where the greater the racial interaction at municipal level the lower the risk of death from homicide (Table 2). Exposure variable-adjusted SMR had a similar distribution to the non-adjusted SMR, suggesting that the model effectively captured the distribution (Figure 2). Figure 3 presents the posterior distribution of the coefficients from model C2, with most of the distribution lying outside the null value, corroborating the significance of the effects.

Table 2
Results of the regression models: mean and percentiles 2.5 and 97.5 of the posterior distributions of the model coefficients.

Models	IIR	Income	Gini	DIC
A1	-2.229 (-2.336; -2.123)			9543.68
A2	-2.930 (-3.052; -2.809)	0.728 (0.691; 0.765)		8439.37
A3	-3.612 (-3.747; -3.479)	1.370 (1.307; 1.434)	-4.729 (-5.093; -4.366)	7818.15
B1	-1.993 (-2.297; -1.690)			4106.45
B2	-2.576 (-2.936; -2.219)	0.988 (0.666; 1.312)		4089.10
B3	-2.563 (-2.951; -2.177)	0.977 (0.632; 1.323)	0.100 (-0.960; 1.164)	4089.50
C1	-1.427 (-2.080; -0.776)			4019.45
C2	-1.787 (-2.459; -1.119)	0.815 (0.520; 1.112)		4001.63
C3	-1.785 (-2.459; -1.113)	0.806 (0.427; 1.186)	0.043 (-1.065; 1.150)	4002.39

• RII: racial interaction index. DIC: deviance information criterion. Models A: without random effects. Models B: with unstructured random effect. Models C: with unstructured random and structured spatial effects. Best model: C2 (lowest DIC).
• Source: Authors.

Figure 2
Spatial distribution maps of crude (SMR) and adjusted (SMRa) standardized mortality ratios in municipalities in Minas Gerais.

Figure 3
Posterior distributions of the coefficients from the C2 regression model.

Discussion

This was the first study in Brazil to identify an association between greater racial interaction at municipal level and lower homicide mortality. In a literature review of the contextual causes of homicide, the studies analyzed showed that a wide range of social factors, including the Gini index, income and indicators of poverty, were associated with homicide rates, but none investigated indicators of residential segregation¹¹.

The social context in which people live and work is an important determinant of health. In her classic article, Schwartz pointed out how the methodological individualism in epidemiological studies led to a relative ignorance of the importance of this factor, consequently compromising the effectiveness of disease prevention and control measures¹². In another classic work, Diez-Roux called for the development of disease causation models incorporating multiple levels of determination, include the structural factors influencing the health-disease process¹³. More recently, other authors have made considerations along the same lines, highlighting the need for epidemiological research to investigate the structures and networks of causes that underlie disease, rather than the isolated effects of single exposures¹⁴. The authors of a systematic literature review highlighted the recent growth in research on the effects of neighborhood context on health, focusing on areas such as mental health and violence¹⁵. Some studies have shown that socioeconomic and racial residential segregation is an important contextual factor causing health disparities^3,16-18.

The leading external cause of mortality in high-income countries is suicide, followed by transport accidents and falls. In contrast, the leading causes in middle and low-income countries in the Americas are homicide and transport accidents. Evidence suggests that poverty is a risk factor for external cause-mortality, manifested in the form of poor living, work and transport conditions¹⁹. Studies in the United States have observed that risk factors for violent death include belonging to minority groups, low community cohesion, criminality, and use of substances, primarily alcohol²⁰. Another study found an association between residential segregation and increased fatal interactions with the police²¹. In addition, neighborhood factors such as criminality, low income and Gini index, high proportion of Afro-Americans and single parent families, and unemployment rate have a negative influence on criminality. In contrast, number of families living in their own home, social cohesion and informal participation have been shown to reduce violent crime²². In low and middle-income countries, income inequality was associated with homicide, self-reported assault and robbery²³. A study of 10 Latin American countries, including Brazil, Colombia, Ecuador and Mexico, found that if deaths from external causes were prevented, these countries would increase the number of years lived in adulthood in the future²⁴.

Various individual factors are associated with higher frequency of cases of violence and deaths from external causes in Brazil, including being male, young, black and single, having a low income and level of education, and poor living conditions. From a social context perspective, engagement in illegal activities such as drug trafficking, receiving stolen goods, and land invasion, and lack of state presence are some of the main factors responsible for increased rates of death form external causes in Brazil⁶. Residential segregation results in the formation of neighborhoods with high levels of poverty, contributing to an increase in mortality²⁵. Other factors include intrafamily violence, unemployment, lack of public security, police violence, and urbanization. Living in favelas has also been associated with higher rates of homicide mortality, regardless of sex or age⁶.

In addition, austerity policies, especially in a context of economic crisis, can affect outcomes related to external causes of death. In this regard, a study reported an increase in homicide mortality and suicide across Brazil during the financial crisis starting in 2014, with increases being more pronounced in the North and Northeast²⁶.

In Brazil, external causes are the third leading cause of death⁵. The country witnessed a downward trend in homicides among men between 2005 and 2016, despite regional differences: stable rates in the North and Mid-West, increased rates in the Northeast, and reductions in the Southeast and South²⁷. It is possible that public policies such as the Disarmament Statute have contributed to curbing the rampant growth in mortality from firearms in the country²⁸. In contrast, however, the homicide rate among young people and number of female homicides increased between 2007 and 2017²⁹. The country has witnessed a trend towards the “interiorization” of homicide mortality, with large cities in the Southeast like Rio de Janeiro and São Paulo seeing a substantial reduction, while inland states such as Minas Gerais have observed a rise, especially in smaller cities in the mid-north of the state close to the border with Bahia. Smaller municipalities have seen an increase in homicide rates, while those with more than 100,000 inhabitants have witnessed the opposite³⁰.

A study in three state capitals (including Belo Horizonte, the capital of Minas Gerais) investigated population distribution by race/color and level of education using the index of dissimilarity. The findings show that the population of urban peripheries was predominantly black, while the population of areas with better infrastructure in city centers was predominantly white. The study also reported that people who self-declared as black or brown were more likely to live in areas with poorer socioeconomic indicators and higher levels of segregation³¹. Another study comparing residential segregation by race in 40 of the country’s biggest metropolitan areas reported that residential segregation between white and brown people and brown and black people reduced between 1980 and 2010, especially in areas with poorer housing. However, levels of segregation remain high in high-income groups, who have access to a wider variety of housing options³².

Although we did not find publications addressing the association between residential segregation indices and homicide mortality in Brazil, there are studies investigating other health outcomes. One of these studies, which performed an analysis adjusted for other sociodemographic factors, reported an association between residential segregation and higher prevalence of hypertension and diabetes mellitus. The findings show that black and brown people were more likely to live in segregated neighborhoods than white people. According to the authors, residential segregation is a cause of income and racial health inequalities and has a negative impact on health, as segregation is not restricted to people, but also encompasses the segregation of infrastructure and health resources, limiting access to health services among the segregated population¹⁸.

A study that explored the relationship between lethal police violence and socioeconomic development in São Paulo reported the existence of racial and social biases in this association. The findings demonstrate the existence of spatial clusters with high levels of mortality and different characteristics: the victims’ place of residence tended to be located in areas with lower HDI, while the place of occurrence of violent events tended to be in areas with higher HDI³³.

Some studies have assessed the effect of public policies on residential segregation and its social and health consequences. Approaches focusing on early childhood interventions have demonstrated that starting early in local communities can have life-long positive impacts on health and the social determinants of health. The evaluation shows that poor and minority families respond to interventions and take advantage of real opportunities, resulting in an increase in income and improved safety and physical and mental health³⁴. On the other hand, housing policies can drive residential segregation, through the construction of more affordable housing in more isolated areas that tends to be bought by lower income groups. Improvements in infrastructure and transport in these areas commonly fail to be implemented, making the integration of populations living in these areas and those in more central regions even less viable³⁵. Studies suggest that relocating people from their neighborhood is not the ideal policy. Rather emphasis should be placed on improving living conditions in poor areas.

The main limitation of this study is the use of secondary data that may have been subject to shortcomings in the data collection and/or processing stages. It is also possible that deaths were underreported, especially in municipalities located in poorer regions of the state. However, it is important to stress that the sources used were DATASUS and the Demographic Census, both of which are official government databases and widely used in epidemiological studies and subject to constant data quality monitoring. One of the advantages of using these sources is that they encompass the whole population of the state of Minas Gerais, thus reducing the likelihood of selection bias. Another limitation in that cross-sectional studies are limited in their ability to determine the cause-and-effect relationship between variables. In this regard, the associations found by the present study raise the hypothesis that there is a causal relationship. This relationship should be tested by future longitudinal studies.

This study observed the presence of racial residential segregation in municipalities in Minas Gerais and that segregation was associated with homicide mortality, showing that this factor should be taken into account in violence prevention and control polices in the state.

Referências

• ¹
  Smets P, Salman T. The multi-layered-ness of urban segregation. Habitat Int 2016; 54:80-87.
• ²
  Diez-Roux AV, Mair C. Neighborhoods and health. Ann N Y Acad Sci 2010; 1186:125-145.
• ³
  Krieger N, Feldman JM, Waterman PD, Chen JT, Coull BA, Hemenway D. Local Residential Segregation Matters: Stronger Association of Census Tract Compared to Conventional City-Level Measures with Fatal and Non-Fatal Assaults (Total and Firearm Related), Using the Index of Concentration at the Extremes (ICE) for Racial, Economic, and Racialized Economic Segregation, Massachusetts (US), 1995-2010. J Urban Health 2017; 94(2):244-258.
• ⁴
  Massey DS, Denton NA. The Dimensions of Residential Segregation. Soc Forces 1988; 67(2):281-315.
• ⁵
  Malta DC, Minayo MCS, Soares Filho AM, Silva MMA, Montenegro MMS, Ladeira RM, Morais Neto OL, Melo AP, Mooney M, Naghavi M. Mortalidade e anos de vida perdidos por violências interpessoais e autoprovocadas no Brasil e Estados: análise das estimativas do Estudo Carga Global de Doença, 1990 e 2015. Rev Bras Epidemiol 2017; 20:142-156.
• ⁶
  Modesto JG, Alves AYM, Santos LV, Archanjo CCC, Araújo GS. Fatores que influenciam na mortalidade de jovens por causas externas no Brasil: Uma revisão da literatura. Rev Multidebates 2019; 3(2):137-155.
• ⁷
  Gomes CE, Evangelista TF, Lima RL, Parré JL. Determinantes do crime nos municípios de Minas Gerais e seus possíveis spillovers espaciais. Econ Ensaios 2017; 31(2):201-228.
• ⁸
  Sparks CS. Calculating indices of residential segregation [Internet]. [cited 2021 maio 19]. Available from: https://rpubs.com/corey_sparks/473785
  » https://rpubs.com/corey_sparks/473785
• ⁹
  Ord JK. Estimation methods for models of spatial interaction. J Am Stat Assoc 1975; 70:120-126.
• ¹⁰
  Moraga P. Small area disease risk estimation and visualization using R. RJ 2018; 10(1):495-506.
• ¹¹
  Sousa CAM, Silva CMFP, Souza ER. O efeito do contexto sobre a incidência de homicídios: existem evidências suficientes? Interface (Botucatu) 2018; 22(66):915-927.
• ¹²
  Schwartz S. The fallacy of the ecological fallacy: the potential misuse of a concept and the consequences. Am J Public Health 1994; 84(5):819-824.
• ¹³
  Diez-Roux AV. Bringing context back into epidemiology: variables and fallacies in multilevel analysis. Am J Public Health 1998; 88(2):216-222.
• ¹⁴
  Keyes KM, Galea S. Commentary: The Limits of Risk Factors Revisited: Is It Time for a Causal Architecture Approach? Epidemiol Camb Mass 2017; 28(1):1-5.
• ¹⁵
  Arcaya MC, Tucker-Seeley RD, Kim R, Schnake-Mahl A, So M, Subramanian SV. Research on neighborhood effects on health in the United States: A systematic review of study characteristics. Soc Sci Med 2016; 168:16-29.
• ¹⁶
  Kershaw KN, Albrecht SS. Racial/ethnic residential segregation and cardiovascular disease risk. Curr Cardiovasc Risk Rep 2015; 9(3):1-12.
• ¹⁷
  Krieger N, Waterman PD, Spasojevic J, Li W, Maduro G, Van Wye G. Public Health Monitoring of Privilege and Deprivation With the Index of Concentration at the Extremes. Am J Public Health 2016; 106(2):256-263.
• ¹⁸
  Barber S, Diez-Roux AV, Cardoso L, Santos S, Toste V, James S, Barreto S, Schmidt M, Giatti L, Chor D. At the intersection of place, race, and health in Brazil: Residential segregation and cardio-metabolic risk factors in the Brazilian Longitudinal Study of Adult Health (ELSA-Brasil). Soc Sci Med 2018; 199:67-76.
• ¹⁹
  World Health Organization (WHO). Injuries and violence: the facts 2014 [Internet]. Geneva: WHO; 2014 [cited 2021 maio 19]. Available from: https://apps.who.int/iris/handle/10665/149798.
• ²⁰
  Jack SPD. Surveillance for Violent Deaths - National Violent Death Reporting System, 27 States, 2015. MMWR Surveill Summ 2018; 67(11):1-32.
• ²¹
  Johnson O, Vil CS, Gilbert KL, Goodman M, Johnson CA. How neighborhoods matter in fatal interactions between police and men of color. Soc Sci Med 2019; 220:226-235.
• ²²
  Brisson D, Roll S. The effect of neighborhood on crime and safety: a review of the evidence. J Evid-Based Soc Work 2012; 9(4):333-350.
• ²³
  Wolf A, Gray R, Fazel S. Violence as a public health problem: An ecological study of 169 countries. Soc Sci Med 2014; 104:220-227.
• ²⁴
  Calazans JA, Queiroz BL. The adult mortality profile by cause of death in 10 Latin American countries (2000-2016). Rev Panam Salud Publica 2020; 44:1-9.
• ²⁵
  Araújo EM, Costa M CN, Oliveira NF, Santana FS, Barreto ML, Hogan V, Araújo TM. Spatial distribution of mortality by homicide and social inequalities according to race/skin color in an intra-urban Brazilian space. Rev Bras Epidemiol 2010; 13(4):549-560.
• ²⁶
  Machado DB, Pescarini JM, Araújo LFSC, Barreto ML. Austerity policies in Brazil may affect violence related outcomes. Cien Saude Colet 2019; 24(12):4385-4394.
• ²⁷
  Brasil. Ministério da Saúde (MS). Saúde Brasil 2018 uma análise de situação de saúde e das doenças e agravos crônicos: desafios e perspectivas. Brasília: MS; 2019.
• ²⁸
  Waiselfisz JJ. Mapa da Violência 2016: Homicídios por arma de Fogo no Brasil. Rio de Janeiro: Editora Flacso Bra; 2016.
• ²⁹
  Instituto de Pesquisa Econômica Aplicada (Ipea). Atlas da violência 2019 [Internet]. Brasília: IPEA; 2019 [acessado 2021 maio 19]. Disponível em: http://repositorio.ipea.gov.br/handle/11058/9489.
• ³⁰
  Soares Filho AM, Merchan-Hamann E, Vasconcelos CH. Expansão, deslocamento e interiorização do homicídio no Brasil, entre 2000 e 2015: uma análise espacial. Cien Saude Colet 2020; 25(8):3097-3105.
• ³¹
  Amaral EFL. Racial and socioeconomic segregation: an analysis of three Brazilian Metropolitan areas. Redes St Cruz Sul Online 2013; 18(1):248-262.
• ³²
  Valente RR, Berry BJL. Residential Segregation by Skin Color: Brazil Revisited. Lat Am Res Rev 2020; 55(2):207-226.
• ³³
  Ryngelblum M, Peres MF. Social Segregation and Lethal Police Violence in the City of São Paulo (2014-2015). Cien Saude Colet 2020; 26(9):4275-4286.
• ³⁴
  Williams DR, Cooper LA. Reducing Racial Inequities in Health: Using What We Already Know to Take Action. Int J Environ Res Public Health 2019; 16(4):606.
• ³⁵
  Silva JCTP, Cruz MCC, Souza MF. Análise geoestatística da segregação socioespacial em Divinópolis, Minas Gerais: o Programa Minha Casa Minha Vida em perspectiva. Rev Esp 2017; 6(2):27-35.

• Funding

  This study was supported by the Universidade Federal de Juiz de Fora in the form of a scientific initiation grant.

Edited by

Chief editors:

Romeu Gomes, Antônio Augusto Moura da Silva

Publication Dates

• Publication in this collection
  15 Aug 2022
• Date of issue
  Sept 2022

History

• Received
  23 June 2021
• Accepted
  12 Apr 2022
• Published
  14 Apr 2022