The geography of femicide in Sergipe, Brazil: matriarchy, human development, and income distribution

REV BRAS EPIDEMIOL 2021; 24: E210016.SUPL.1 ABSTRACT: Aim: This research analyzed a joint spatial distribution and explored the possible associations between epidemiological aspects and feminicide rates, in towns of the Sergipe State in Northeastern Brazil. Methodology: An exploratory ecological study investigated the global spatial autocorrelation of epidemiological aspects with femicide rates from towns in Sergipe State, Brazil, (n = 75), in the 2013–2017 period, using the “global” and “local” Moran statistic method and a multiple spatial regression. The exposure variables included socioeconomic and demographic conditions, services and health condition, and femicide rates. We used the software Stata 11.0, SPSS 18.0 and GeoDa 0.95-i. Results: The spatial distribution of femicide rates was not random and showed high spatial autocorrelation and predominance of significant spatial groupings of towns with the highest mortality rates due to femicide in the central region of Sergipe State. In the multiple regression analysis, the percentage of women in charge of families and the Municipal Human Development Index were positively associated with the femicide rates in towns in Sergipe’s municipalities in the studied period (p < 0.05). The opposite situation occurred between the Income Concentration Index (GINI) and the femicide rates. Conclusion: This is the first study that has analyzed the factors associated with the spatial clusters of femicide rates in a geographical space where there is a predominance of patriarchal culture. There was a femicide increase in locations with the lowest social inequality, the highest human development and authority exercised by women in the family environment.


INTRODUCTION
Femicide is rooted in gender inequality and contributes to perpetuate the subordinate position of women in society at the legal, social, and economic levels. It is considered a serious public health issue that was described by the World Health Organization over 30 years ago 1,2 .
Despite the promulgation of what is known as the "Maria da Penha Law", which aimed to create mechanisms to curb domestic and family violence against women, female mortality from aggression in Brazil has increased more than twice in the last 30 years (Maria da Penha Law, Law n. 11.340, August 7 th , 2006) 6 . It is not known for sure whether the increase in police records of femicide reflects an increase in the number of cases or a decrease in underreporting, considering that the "Maria da Penha Law" is relatively new, so that there may be an ongoing learning process by the judicial authorities 5,7 . Nevertheless, this can have happened because this device allows retaliations that often culminated in femicide 8,9 . In order to change this sad Brazilian reality, another important law called the "Femicide Law" was implemented with the purpose of being even more rigorous in coercing aggressors and making this phenomenon still visible. With the amendment of article 121 of the Brazilian Penal Code, femicide becomes qualified in the list of heinous crimes (Femicide Law, Law n. 13.104, March 9 th , 2015) 10,11 .
Studies have shown an increase in femicides in capitals and in large Brazilian cities, in the three-year periods of 2007-2009 and 2011-2013, mainly among young, poor, black, single and low-educated women 3 . Paradoxically, some studies on femicide have pointed out that poor women are more affected, because they observed a positive association between poverty and female deaths 12,13 . The contexts of women's homicides are diverse and obey distinct social dynamics and peculiarities to the geographic context of origin, in which the gender marker is always present 3 .
In addition, the Northeastern society, especially that of the State of Sergipe, culturally preserves provincial habits and traditions in a peculiar way, whose valorization of families with surnames, titles and economic possessions is an indelible mark in the history of the patriarchy present in the Brazilian Northeast 14,15 . Hence, there is evidence that points to an increase in this type of violence in places where patriarchal culture is still perpetuated 16 , because in more conservative communities, gender norms are more rigid and women remain submissive, strictly fulfilling and supporting for long periods the roles assigned to them by culture 3 .
Thus, the present study aimed to analyze the joint spatial distribution and explore possible associations between epidemiological aspects and femicide rates, in the 2013-2017 period, in towns of the State of Sergipe in Northeastern Brazil.

METHODS
This exploratory ecological study was approved by the Ethics Committee of the Universidade Federal de Sergipe (CAEE 17504319.0.0000.5546), including secondary data from the Sergipe population, which is freely available in Brazilian databases.
Sergipe is the smallest Brazilian state and sixth least populous in the country, with a geographical area of 21,926.908 km 2 and a population of 2,068,017 people. It is divided into eight territories (Agreste Central, Alto Sertão, Baixo São Francisco, Centro Sul, Grande Aracaju, Leste, Médio Sertão, Sul) and 75 municipalities, whose greatest distance from the capital of the state is around 200 km 17 . Sergipe is 20 th in the national ranking of the Human Development Index and fourth in the ranking in this category among the Northeastern states. Its capital, Aracaju, is a true social and economic oasis of opportunities for other municipalities in the state, with the best development rates and indicators in Sergipe. Aracaju, thus, offers the best work and service opportunities. Such aspects that have called the attention of the population residing in the State's countryside to the capital 17 , since more than half of the towns in Sergipe have a low human development index and a high social vulnerability index 18 .
The geographic and spatial data of each municipality in Sergipe, considering its latitude, longitude, perimeter, area and location, were obtained from the website of the Brazilian Institute of Geography and Statistics (IBGE) 17 .
The dependent variable of this study was determined from the female mortality rates due to aggression, calculated according to each municipality in Sergipe, in the period from 2013 to 2017. The coefficients were grouped into five years to decrease the temporal and geographic oscillations common in small magnitude events. Gross mortality data due to women aggression were retrieved from the Brazilian Ministry of Health's Mortality Information System and death causes were considered according to the categories of the 10 th Edition of the International Classification of Diseases (ICD-10) in the range from X85 to Y09 19 . The total resident population data in each municipality in the State of Sergipe in the period corresponding to the study were based on the 2010 population census and estimates for the inter-census years, which were obtained on the website of the IBGE 17 . Femicide rates were calculated by dividing the mortality data related to women's aggression in each of the Sergipe's towns by the data on the total female population residing in the same place.
The independent variables selected for this study were collected in several national databases in the studied period (2013-2017), in which the averages of the data available for each of these variables were estimated. These variables are described in detail as follows: • Univariate exploratory analysis of spatial data was performed to investigate the global spatial autocorrelation of femicide rates in Sergipe's municipalities using the Moran I index, under the assumptions of normality and randomization 21 . The distribution of Moran index values varies between -1.0 and +1.0 and tests whether connected areas present greater similarity in relation to the studied indicator, than would be expected in a random pattern. Thus, the spatial autocorrelation measures the number of nearby objects compared to others using the Moran I index, which can be classified as positive, negative, and without spatial autocorrelation.
The variables in this study were grouped into three blocks to jointly assess the indicators associated with femicide by Pearson's correlation coefficient (r), in which the direction and magnitude of the associations between the independent variables were assessed using a correlation matrix: socioeconomic status; demographic status; and services and health status.
For this analysis, all variables were standardized with a mean of zero (0.0) and standard deviation equal to one (1.0), due to their different dimensions, which could impair their inclusion and interpretation in the model.
Multiple spatial regression was performed in the last analysis phase. The quality of the spatial regression model fit is similar to the traditional linear regression model, which was verified by the residual analysis and also based on the Moran I index 22 . The following criteria were used to include or remove variables from the model: • selection of the variable with the highest statistical correlation; • inclusion of variables that, when analyzed together, obtained a higher F in the simple regression analysis. This inclusion does not prevent variables in the same block from being included as "adjustment variables", regardless of their association; • inclusion of variables that, once in partial correlation, controlled by modeled variables, showed significant correlation with the dependent variable.
The final model exclusion criteria for the variables were p ≥ 0.05. The software used was Stata 11.0, SPSS 18.0, and GeoDa 0.95-i.
There was not a random distribution of this rate with positive spatial autocorrelation (I = 0.0993881; p = 0.001 for 999 permutations). It is possible to observe the occurrence of spatial autocorrelations of the "high-high" type, indicating a grouping of municipalities in Sergipe with the highest mortality rates due to femicide located in the central region of Sergipe (Areia Branca, Malhador, Santa Rosa de Lima, Ribeirópolis) and the "low-low" type, which was considered the group with the lowest rate, represented by the municipalities of Nossa Senhora de Lourdes, Propriá, São Francisco and Japaratuba (Figure 2). Spatial autocorrelations of the "low-high" type were seen in the municipality of Maruim. Thus, the municipality or the spatial group presents low values regarding the rates of femicide, but high rates in the surrounding municipalities ( Figure 2). An inverse condition was observed in the municipalities of Lagarto, Aquidabã and Carmópolis, where spatial autocorrelations of the "high-low" type occurred, a scenario in which these municipalities present high rates of femicide, but their surrounding municipalities have low rates (Figure 2).
Moran Local's Statistics I (Figure 3) is also of great importance for the analysis, since it shows the degree of significance of certain groups. According to the Local Space Association Indicator (LISA), the municipalities with the highest rates of femicide are Santa Rosa de Lima and Aquidabã. The analysis identified significant LISA sector clusters. The significance of this indicator in the period implies the existence of positive multidirectional externalities of the femicide rates in the municipalities of the State of Sergipe.
In the bivariate analysis (Table 1), the percentage of unemployed individuals and women in charge of families (matriarchy) had a statistically significant positive correlation with the rates of femicide (p < 0.01 and p < 0.05, respectively). There were also negative significant correlations between the proportion of the population with less than ½ of minimum wage and femicide rates (p < 0.05). Table 2 illustrates the results of the final spatial regression analysis model. The percentage of women in charge of families (matriarchy) and the Municipal Human Development Index (MHDI) were positively associated with the femicide rates in Sergipe's municipalities in the period between 2013 and 2017 (p < 0.05). The opposite situation occurred between the Income Concentration Index (GINI Index) and the rates of femicide in these municipalities in the studied period.

DISCUSSION
This study has some limitations inherent to the used methodology. The possibility of ecological fallacy cannot be excluded, since an association observed between aggregates does not necessarily mean that the same association occurs in an individual level 23 . The low construct validity is another possibility that cannot be ruled out, because not all explanatory outcomes of the variables may have been included in the methodology used 24 . In order to reduce some of these limitations, this study worked with variables available in several national databases and may present differences in quality inherent to the use of indirect estimates 25 .
The results of the present study show a very uneven distribution of the femicide rates per 100 thousand women in Sergipe's municipalities in the period studied, whose rates standardized ranged between 11.74 deaths/100,000 women and 87.80 deaths/100,000 women ( Figure 1). In order to adjust differences in the population's distribution by sex and, Table 2. Spatial multiple regression model of femicide rates standardized by 100 thousand women in Sergipe's municipalities and associated factors, between 2013-2017.

Variables
Coefficients Standard error t p-value consequently, a possible confounding effect of the latter on the femicide rates in the municipalities of Sergipe, the direct method of standardization of these rates was employed 26 , considering there is a possibility of increase or reduction of the female population to the detriment of the male population in some municipalities. In addition, the selection of women's deaths from aggression with a femicide proxy variable could overestimate this phenomenon. However, a 24% of underreporting deaths in the Northeast region of Brazil should be expected 2,17,27 , which could compensate for underreporting or even the ill-defined diagnosis of the death cause due to female aggression.
There was a predominance of significant spatial groupings of femicide rates in the central Sergipe region (Figure 2), which was confirmed by LISA ( Figure 3). Studies also show the possibility of some peculiar contexts in the central region of Sergipe, despite no spatial dependence on the femicide rates in the Sergipe's municipalities in 2017 28 . However, Grana 29 and Meneghel and Portella 30 suggest that the incidence of femicide is higher in urban agglomerations and cities with a higher population density. It is known that the scenarios of femicides are diverse and present peculiar geographical contexts that are associated with the female gender 30 . There is an increase in femicide in places where there is a predominance of patriarchal culture, an indelible mark of Sergipe society 14,31,32 . This fact could be observed in the groupings of municipalities in Sergipe that presented spatial autocorrelations of the "high-high" type. In these agglomerations there is intense livestock activity, with appreciation of male gender roles, in which courage, strength and virility of men are considered essential to work with cattle, usually extended to gender relations and daily life. Similar conditions were found in other regions of Brazil that also have this economic matrix and still maintain strong traditional patriarchal values and submission of women to men 33 .
The MHDI was directly associated with the femicide rates in the multiple spatial regression. This fact makes evident the contradiction between the increase of femicide in places with higher human development rates 30,34 . These results are contrary compared to other studies, which describe an inverse relationship between the occurrence of femicide with economic development and the quality of life of the population, respectively 35 .
The income concentration, estimated from the GINI index, showed an inverse association with the femicide rates in the final spatial model. Therefore, places with the lowest social inequality present the highest femicide rates. This result is contradicting the findings of Leite et al. 35 , which reported the presence of a direct relationship between social inequality and femicide rates. However, in the same study, it is reported, in a peculiar way, that there are some regional exceptions to the trends of social inequality, such as in the Espírito Santo State, which presented better indicators of inequality and high indicators of female mortality rates due to aggression compared to the other Brazilian States analyzed 35 . It is noteworthy that although socioeconomic issues are linked to femicide, the factors that permeate violence in general go beyond poverty and social inequality. This is reinforcing the multifactorial nature of violence against women, resulting from the interaction of individual, relational, social, cultural, and environmental factors 35,36 .
The matriarchy was positively associated with femicide rates in Sergipe's municipalities in the studied period. Hence, the change in traditional gender roles is a potential generator of conflicts for challenging the patriarchal thinking of society 2,3 . Thus, the achievement of women's socioeconomic autonomy in patriarchal societies is able to tension traditional gender patterns, increasing the risk of femicide 37,38 .
In this context, it is clear that despite the socioeconomic condition being closely related to femicide, this phenomenon has multifactorial characteristics that result from the interaction between individual, relational, social, cultural and environmental conditions 1 . Femicide has its roots in gender inequality, presenting diverse contexts that obey distinct social dynamics and are peculiar to the geographic context of origin. This is the first study that analyzed the factors associated with spatial clusters of femicide in a geographic space where there is a predominance of patriarchal culture. Although public policies to combat femicide have already completed more than a decade in Brazil, this study presents a very uneven distribution of femicide rates that showed a direct association with human development and matriarchy, and an inverse association with concentration of income. These findings reinforce the fact that violence against women is a multifactorial phenomenon, resulting from the interaction of individual, relational, social, cultural, and environmental factors. Thus, it is imperative to carry out further studies in order to unravel the possible factors that predispose high rates of femicides in in the municipalities of the State of Sergipe. Local redirection of Brazilian public policies may be needed to combat femicide, since it has its roots in gender inequality, with different contexts that obey social dynamics and peculiarities to the geographical context of its origin.