Femicides: female homicide in Brazil


Femicidios: homicidios femeninos en Brasil



Stela Nazareth MeneghelI; Vania Naomi HirakataII

ICurso de Análise de Políticas e Sistemas de Saúde. Programa de Pós-Graduação em Enfermagem. Universidade Federal do Rio Grande do Sul. Porto Alegre, RS, Brasil
IIGrupo de Pesquisa e Pós-Graduação. Hospital de Clínicas de Porto Alegre. Porto Alegre, RS, Brasil





OBJECTIVE: To assess female homicide rates due to aggression according to sociodemographic and health indicators.
METHODS: Ecological study on female homicides due to aggression in Brazil between 2003 and 2007. Information on 19,459 deaths were obtained form the Brazilian Mortality Database. Standardized female homicide rates due to aggression were correlated with 28 socioeconomic, demographic and health indicators, using Pearson's correlation test. Multiple linear regression was performed including variables with p < 0.20 and excluding those with multicollinearity.
RESULTS: The standardized female homicide rate due to aggression was 4.1/100,000 during the study period. After adjustment, three variables remained significantly associated with female homicides: birth rate (p = 0.072), percentage of Evangelicals (p = 0.019) and male homicides due to aggression (p < 0.001). The model had a predictive power of 69% (r2 = 0.699). The Brazilian states of Espírito Santo, Pernambuco, Mato Grosso, Rio de Janeiro, Rondônia, Alagoas, Mato Grosso do Sul, Roraima e Amapá showed the highest rates during the study period.
CONCLUSIONS: The female homicide rates due to aggression were high in Brazil and varied by region. The association with male homicides due to aggression stresses the importance of reducing structural violence to protect women against violence.

Descriptors: Women. Homicide. Cause of Death. Violence Against Women. Socioeconomic Factors. Ecological Studies.


OBJETIVO: Analizar la mortalidad femenina por agresión según indicadores sociodemográficos y de salud.
MÉTODOS: Estudio ecológico sobre la mortalidad femenina por agresión ocurrida en Brasil de 2003 a 2007. Los datos de 19.459 óbitos fueron obtenidos del Sistema de Informaciones sobre Mortalidad. Los coeficientes estandarizados de mortalidad femenina por agresión fueron relacionados (test de correlación de Pearson) con 28 indicadores socioeconómicos, demográficos y de salud. Se realizó regresión lineal múltiple con variables que presentaron p< 0,20 y se excluyeron las variables que presentaron multicolinealidad.
RESULTADOS: El coeficiente estandarizado de mortalidad fue de 4,1 óbitos/100.000 mujeres en el período. Posterior al ajuste, tres variables permanecieron significativas y asociadas a la mortalidad femenina por agresión: tasa de natalidad (p=0,072), porcentaje de evangélicos (p=0,019) y coeficiente de mortalidad por agresión en el sexo masculino (p<0,001). El modelo posee una capacidad de predicción del resultado de 69% (r2 = 0,699). Espírito Santo, Pernambuco, Mato Grosso, Rio de Janeiro, Rondonia, Alagoas, Mato Grosso do Sul, Roraima y Amapá presentaron los mayores coeficientes en el período.
CONCLUSIONES: La mortalidad femenina por agresión en Brasil fue elevada y no homogénea entre las regiones. Entre las variables asociadas al evento, se destaca la mortalidad masculina por agresión, indicando la importancia de la reducción de la violencia estructural como protección a la violencia perpetrada contra las mujeres.

Descriptores: Mujeres. Homicidio. Causas de Muerte. Violencia contra la Mujer. Factores Socioeconómicos. Estudios Ecológicos.




Violence against women known as gender-based violence is by the World Health Organization (WHO) a public health problem since 1990. Most violent acts occur at home and the victim usually knows the perpetrator. Gender-based violence includes property, sexual, psychological, and physical violence and may result in the woman's death by suicide or homicide.13

Homicides occurring due to gender-based violence are known as femicides, a political and legal term to refer to this type of death. The term femicide was first used by Russell21 in a court of human rightsª and refers to any manifestation of the unequal power relations between men and women that culminated with the death of one or more women because of their gender. This type of crime can occur in several situations including deaths perpetrated by an intimate partner with or without sexual violence, serial murders, sexual violence followed by death, and femicides associated with the death of another person.7

More than half of all female homicides are femicides caused by gender inequalities6 and this phenomenon is present on several continents. It is estimated that between 60% and 70% of femicides in the United States and Canada are committed by current or former partners.7 In Europe, in countries such as Russia, homicides and femicides increased with the collapse of the Soviet Union because of social disorganization, which led to the proliferation of gangs and violent crime.11 Data from human rights organizations indicate that Guatemala, a country that has experienced violent internal conflicts, has one of the highest rates of femicides in the Americas. Between 2003 and 2005, 1,398 women were murdered in Guatemala, 1,320 in El Salvador, 613 in Honduras, more than 400 in Mexico and 269 in Nicaragua.

Male homicides due to violence have increased in several regions of Central America and South America and unlike female homicides both the victim and the perpetrator are men. Male murders are not caused by gender inequalities but rather by street violence: fighting, territory control, gangs, drug trafficking, death squads19 while the majority of female homicides are caused by intimate violence that permeates the interpersonal relations between men and women.

Sociodemographic factors associated with female homicides by partners include poverty, age disparity between partners, and informal marital status. In many countries, one third of the women murdered were trying to separate from their partners, especially in the three months preceding the crime, and had a history of repeatedly experiencing violence and assaults.9 Studies conducted in the United States found a relationship between femicide rates and areas with higher rates of poverty, instability, black population, unemployment and violent crimes.6 Regarding the relationship between religion and violence, it is believed that the religious discourse reinforces misogyny, hegemonic masculinity and tolerance of violence against women as it preaches submission of wives to their husbands.25 Femicides have been strongly associated with gender inequality and discrimination, economic deprivation, and aggressive masculinity and machismo including use of firearms, involvement with organized crime, drug and human trafficking, armed conflicts and high rates of male homicides due to physical violence.12,19

The socioeconomic conditions of the victims vary by country and circumstance. The pattern of femicides that is repeated in most countries suggest that women are at much higher risk than men of being killed by an intimate partner and this risk increases when there are disagreements between the couple.6 In Latin America many murdered women are socially deprived and they are often depicted by the media as prostitutes, factory and transnational automaker workers known as maquilas in Spanish-speaking countries, and gang or drug trafficking members. In short, they are poor young migrants who live in slum areas or irregular housing projects, work in jobs with poor working conditions and are highly vulnerable.19

In Brazil, studies on femicides have showed higher rates among young white women with low schooling who are unskilled workers while the perpetrators are young married men, usually less educated than their victims, with a criminal record, repeated involvement in fighting, and engaging in threatening behaviors and violence against women.3,4

The objective of the present study was to assess female homicide rates according to sociodemographic and health indicators.



Ecological study that correlated female homicides by geographical regions (states) with sociodemographic and health variables. The explanatory variables were obtained from secondary data from the Pesquisa Nacional por Amostra de Domicílios (Brazilian National Household Survey, PNAD) and the Instituto Brasileiro de Geografia e Estatística (Brazilian Institute of Geography and Statistics, IBGE) population census. Data were also obtained from the Brazilian Ministry of Health Database (DATASUS), the National Cancer Institute and the National Secretariat of Public Safety of the Ministry of Justice.

Our initial intention was to assess femicides but this information was not available from either death certificates or the Brazilian Mortality Database (SIM). Thus, we chose to use total female homicides as a "rough indicator" of femicides since 60% to 70% of these deaths are caused by gender inequalities.6,7

Similar studies were not found in the literature. Carcedo & Sagot7 estimated the fraction of femicides among female homicides in Costa Rica and found 70% of femicides during 1990-1999. Another ecological study carried out in the United States12 used as outcome information on femicides from related organizations and 36% of the federal units did not report the percentage of female homicides due to gender-based violence making it difficult to estimate femicide rates. The use of female homicides as a proxy of femicides in the Brazilian population may overestimate the actual rates but it may compensate for underreporting or ill-defined diagnosis of physical violence against women seen in some Brazilian macroregions since it is estimated 23% underreporting of deaths in the North region and 24% in the Northeast in 2009.b

The dependent variable was female homicide rate estimated by state from 2003 to 2007. The rates were grouped into five years to reduce time and geographical fluctuations usually seen in events of small magnitude. Data from SIM were used and female homicides were categorized as X85 to Y09 codes according to the International Classification of Diseases - 10th Revision (ICD-10).

Population data for the denominators were obtained from the 2000 Population Census, the 2006 Population Count and intercensal projections by age group and gender provided by IBGE and available from the Brazilian Ministry of Health DATASUS website. Homicides rates were standardized using the WHO standard world population for 2000 to 2025. The standardization allows comparisons between areas with different demographic and age patterns as is the case of Brazilian macroregions.

The explanatory variables included 28 indicators, most of them disaggregated by sex. Those with no information by sex referred to household or total population and were analyzed globally. All indicators were disaggregated by geographical areas (states). The study variables were grouped into four major thematic blocks:

  • Economic: Gini index (males and females); Human Development Index (HDI); percentage of economically active male and female population employed; average monthly income (males and females); percentage of poor in total population (less than 0.5 monthly minimum wage); percentage of white and black poor population; percentage of unemployment; percentage of female heads of household.
  • Demographic: female marital status (married, separated, and single); percentage of illiteracy (males and females); percentage of people with 15 or more years of schooling (males and females); percentage of people not originally from the city (males and females); urbanization rate; percentage of black population (black and mixed), fertility rate; birth rate; and religion (percentage of Catholics, Evangelicals and no religion).
  • Unionization, communication and public safety: percentage of unionized people (males and females); percentage of households with internet and cell phones; ratio of police to population.
  • Health: female and male life expectancy; mortality rate from AIDS (males and females); standardized mortality rate from cervical and breast cancer; maternal mortality rate and proportional mortality from ill-defined causes (males and females) - an indicator of information quality; number of physicians per 1,000; and male homicide rates.

The data were gathered into a database using Microsoft Excel®. The dependent variable was plotted on the map of Brazil using Epimap.

Bivariate analysis was performed through Pearson correlation test using SPSS v. 10.0. Nine out of the 17 explanatory variables with p<0.20 were included in the multiple linear regression model. The best correlation values for indicators stratified by sex or other variable (poverty, religion, marital status) were included in the regression model. The variables with multicollinearity were excluded from the multivariate analysis. It was used a multiple linear regression model with backward selection method, in which, starting from a model including all variables, those with p>0.10 were progressively removed.

Although the study was based on secondary data available for public consultation, it was approved by the Research Ethics Committee at the Escola de Saúde Pública and State Health Department of Rio Grande do Sul (protocol CEPS-ESP 473/09).



Approximately 20,000 women were murdered in Brazil between 2003 and 2007, a standardized average homicide rate of 4.1 deaths per 100,000. Espírito Santo, Pernambuco, Mato Grosso, Rio de Janeiro, Rondônia, Alagoas, Mato Grosso do Sul, Roraima and Amapá were the Brazilian states with the highest female homicide rates during the study period (Figure 1).

Of 20,000 females killed during the five years studied, most were single young women with low schooling and 50.7% had black and mixed skin color. About 20% of these deaths occurred among children and adolescents under 20. Approximately one third of the deaths occurred in the victim's home (Table 1).



Table 2 shows the variables obtained from secondary sources grouped into four major blocks: economic; demographic; unionization, communication, public safety; and health.

Table 3 shows (Pearson's) correlations between standardized female homicide rates and the variables related to economic condition, demographics, unionization, communication, public safety and health. Significant correlations (p<0.20) were found between the percentage of poor population and HDI for the economic variables; marital status, religion, illiteracy and birth rate for the demographic variables; and households with cell phones and ratio of police to population; proportional mortality from ill-defined causes, female mortality rate from AIDS and male homicide rate for the health variables. The percentage of poor population (total, white and black) and HDI showed multicollinearity and thus were excluded from the multivariate analysis.

Table 4 shows the results of the multivariate analysis according to the nine variables included in the model. After adjustment, three variables remained significant: birth rate (p = 0.072); percentage of Evangelicals (p = 0.019); and male homicide rate (p<0.000). The model had a predictive power of 69% (r2 = 0.699).




Homicides rates have been increasing in Brazil in recent decades, and it is one of the highest in the Americas.23 Females homicides account for around 10% of all homicides, which may be suggestive of secondary relevance in view of few studies on this issue. However, despite its relatively lower rates, female homicides constitute a social problem and its magnitude is not comparable to that in the male population7 because most female homicides are gender-based.19

The fact that one third of all deaths occurred in the victim's home supports the assumption notion they were femicides, i.e., deaths caused by an intimate partner, family member or acquaintance, contrasting with male homicides that largely occur in public spaces.

The distribution of the independent variables and the outcome was highly heterogeneous in Brazil allowing an ecological analysis. The highest female homicide rates were seen in the states of Espírito Santo and Rio de Janeiro in southeast Brazil, where it is also reported high male homicide rates, indicating high levels of urban violence. Pernambuco and Alagoas in the northeast region and some northern and central-west states, areas of migration and land conflicts, also showed high rates of female homicides. Rio de Janeiro, Pernambuco and Espírito Santo have reported higher rates of homicides nationwide and the capital city of Vitória, in the southeastern state of Espírito Santo, has the highest risk of death from homicide, followed by the cities of Recife, São Paulo and Rio de Janeiro.14

Economic indicators and indexes were included in the study to assess the association between female homicides and poor living conditions. A paradoxical inverse relationship was found contrasting with studies that found an association between poverty and femicides9,12 or poverty and homicides.2,15,22 Since the economic variables were strongly interrelated showing high colinearity and potential confounding effects they were not included in the multivariate analysis model.

It is noteworthy the change in traditional gender roles, and the massive entry of women into the formal work force allows many of them to achieve economic independence. This situation potentially creates conflicts as some men no longer have the role of breadwinners and heads of household and may have a strong reaction to this new status generating gender-based violence and even increasing the number of fatal cases.8,20 This change in gender roles is reflected by the growing number of families where the woman is in the labor market and her partner is unemployed. Unemployed men may use violence against women to assert their authority,19 and male unemployment has become a major risk factor for femicide.6 The fact that this change has been seen mostly in large urban centers may explain the lower female homicide rates found in most northeastern states that have more conservative societies characterized by traditional gender roles and lower rates of female entry into the labor market than in other Brazilian macroregions.c However, it does not mean these areas are less violent but rather that violent deaths are either not reported or are included in other categories such as suicides or injuries. Pernambuco is the exception with high prevalence of urban gender-based violence as evidenced in Schraiber et al study.24 These authors reported high prevalence of all types of violence among women in Pernambuco. Although it seems contradictory, in many situations and in more rigid traditional hierarchical societies, women may protect their lives by hiding the fact they are abused and not directly confronting domestic violence.21

It was initially hypothesized in this study that disadvantaged populations with lower education and higher birth and fertility rates would be more exposed to violence. However, in the final model, we found an inverse relationship between femicides and birth rates, i.e., the areas with lower birth rates had higher female homicide rates. This finding is consistent with the hypothesis that spousal violence that may eventually culminate in death is likely to be higher in societies where the traditional gender roles have changed. In other words, more educated women have more opportunities to enter the labor market and can exercise greater control over their reproductive life and consequently have less children.10

Homicide rates were expected to be higher in areas with greater migrant populations. Female migrants are more vulnerable and migrants in general life in poorer living.6,17 In the present study, migration -measured by the percentage of males and females not originally from the city- was not associated with violence.

There was found a positive association between the outcome and the percentage of married women, and an inverse relationship with single women, which was not lost in the final model. In the United States married women were less likely to be murdered than single ones6 and in Canada women living with a partner were eight times more likely to be murdered than married women. Knowing the marital status of a woman can be a protective factor as women trying to separate from their husbands, partners or boyfriends are at higher risk of femicide.9

Religion in general supports the ideal model of a patriarchal family where women are under the authority of their husbands, which may encourage violent male behaviors. Pentecostals view women in a servant, subordinate position with little access to decision making, which makes violence seems natural and makes it difficult to offer resistance and report it.25

The Brazilian religion map was included in the study and the percentage of Evangelicals remained in the final model, suggesting an association between high frequency of Evangelicals and femicides. In Brazil, the largest contingent of evangelical believers are in the northern and central-western states while the Northeast has the lowest number of Pentecostal churches with a high rate of Catholics.1 The growing number of Evangelicals in Brazil can be explained by economic inequality with a strong presence of Pentecostal churches in socially deprived areas. Pentecostals have members among lower income populations and reached areas inaccessible to other religions and characterized by precarious conditions with a complete lack of public power.5 The overlap between the map of Evangelicals and femicides questions not only the association with poverty but also shows that femicides as well as Pentecostals are more prevailing among "the poorest of the poor" or areas of extreme social deprivation and lack of public power. It is thus verified that femicide is the last violation and act of power against women or the most extreme form of sexual terrorism.12

The variable that showed the strongest association and the highest level of significance in the final model was male homicide rate. While male homicide is an indicator of urban violence, female homicide is an indicator of interpersonal violence. Contrasting with other studies,7,12 these two forms of violence were strongly associated in the present study. Homicide rate is an indicator of urban and social violence associated with social and economic inequalities, weak role of the State regarding public policies and ineffective public safety actions - all factors contributing to impunity, organization of death squads and drug trafficking.18

The association between indicators of urban violence and gender-based violence suggests that high-stress societies play an important role in violence against women. High-stress societies are those that go through a transformation process that may be called modernization, with civil unrest, armed conflicts, war or terrorism causing harmful effects on the weakest.16 This explains the strong association between male and female homicides in Brazil predominantly in more violent areas. These areas include urban spaces disputed by drug traffickers and border regions where there are land disputes, land grabbing, intense migration, prostitution and sexual exploitation. Studies with border populations and those living in areas of armed conflicts in Mexico and Central America have showed that these areas have the highest rates of femicides. The border is a place of crossing and migrant inflow where family relationships are strained, structures break down and relations marked by inequalities become more unequal and all of these are directed against women without a face bringing terror17 within an environment of impunity, disintegration of the state and failure of law enforcement.

A study limitation is the use of secondary data. There may have been information of varying quality, death underreporting, inadequate reporting of diagnoses and inadequate completion of death certificates with homicides reported as suicides or injuries. This practice is not widespread but it is mainly seen in northern and northeastern Brazil.c Another limitation of the study refers to its design that may result in ecological fallacy such as the association between lower birth rates in better-off population and femicide, which mostly occurs among poor women. However, despite these limitations, the study results evidenced significant associations especially between structural violence and gender-based violence showing the importance of fighting violence with global interventions.

The present study shows the often underestimated magnitude of female homicides in Brazil. Further studies on femicides are needed, especially in the Brazilian states with growing homicide trends. We stress the importance of using the gender factor while analyzing female homicides. The use of the gender category in the analysis of vital events calls for a new political stand of viewing violent deaths as an unnatural act, not only attributing them to factors of a personal nature and understanding gender hierarchy in the Brazilian society as a factor that expose women to all sorts of vulnerabilities with resultant femicide.



1. Almeida R, Monteiro P. Trânsito religioso no Brasil. Sao Paulo Perspec. 2001;15(3):92-100. DOI:10.1590/S0102-88392001000300012        

2. Barata RB, Ribeiro MCSA, Moraes JC. Desigualdades sociais e homicídios em adolescentes e adultos jovens na cidade de São Paulo em 1995. Rev Bras Epidemiol. 1999;2(1/2):50-9. DOI:10.1590/S1415-790X1999000100005        

3. Blay EA. Assassinato de mulheres e direitos humanos. São Paulo: Editora 34; 2008.         

4. Blay EA. Violência contra a mulher e políticas públicas. Estud Avanç. 2003;17(49):87-98.         

5. Bohn S. Evangélicos no Brasil: perfil socioeconômico, afinidades ideológicas e determinantes do comportamento eleitoral. Opin Publica. 2004;10(2):288-338.         

6. Campbell JC, Glass N, Sharps PW, Laughon K, Bloom T. Intimate partner homicide: review and implications of research and policy. Trauma Violence Abuse. 2007;8(3):246-69. DOI:10.1177/1524838007303505        

7. Carcedo A, Sagot M. Femicidio en Costa Rica 1990-1999. Washington, DC: Organización Panamericana de la Salud; 2000. (Colección Teórica, 1).         

8. Dantas-Berger SM, Giffin K. A violência nas relações de conjugalidade: invisibilidade e banalização da violência sexual? Cad Saude Publica. 2005;21(2):417-25. DOI:10.1590/S0102-311X2005000200008        

9. Dobasch RM, Dobasch RP, Cavanagh K, Lewis R. Not an ordinary killer - just an ordinary guy: when men murder an intimate woman partner. Violence Against Women. 2004;10(6):577-605. DOI:10.1177/1077801204265015        

10. Giffin K. Pobreza, desigualdade e equidade em saúde: considerações a partir de uma perspectiva de gênero. Cad Saude Publica. 2001;18 Supl:S103-12. DOI:10.1590/S0102-311X2002000700011        

11. Gondolf EW, Shestakov D. Spousal homicide in Russia versus the United States: preliminary findings and implications. J Fam Violence. 1997;12(1):63-74. DOI:10.1023/A:1021945917011        

12. Grana SJ. Sociostrutural considerations of domestic femicide. J Fam Violence. 2001;16(4):421-35. DOI:10.1023/A:1012229011161        

13. Krug E, Dahlberg L, Mercy J, Zwi AB, Lozano JA. Informe mundial sobre la violencia y la salud. Washington, DC: Organización Panamericana de la Salud; 2003.          )

14. Lima MLC, Souza ER, Ximenes R, Albuquerque MFPM, Bitoun J et al. Evolução de homicídios por área geográfica em Pernambuco entre 1980 e 1998. Rev Saude Publica. 2002;36(4):462-9. DOI:10.1590/S0034-89102002000400012        

15. Macedo AC, Paim JS, Silva LMV, Costa MCN. Violência e desigualdade social: mortalidade por homicídios e condições de vida em Salvador, Brasil. Rev Saude Publica. 2001;35(6):515-22. DOI:10.1590/S0034-89102001000600004        

16. McWilliams M. Violence against women in societies under stress. In: Dobasch RE, Dobasch RP, editors. Rethinking violence against women. Thousand Oaks: Sage Publications; 1998. p.111-40.         

17. Monarrez Fragoso J. Feminicidio sexual serial en Ciudad Juárez: 1993-2001. Debate Fem. 2002;13(25):279-308.         

18. Minayo MCS. Inequality, violence, and ecology in Brazil. Cad Saude Publica. 1994;10(2):241-50. DOI:10.1590/S0102-311X1994000200011        

19. Prieto-Carrón M, Thomson M, Macdonald M. No more killings! Women respond to femicides in Central America. Gender Dev. 2007;15(1):25-40.         

20. Saffioti HIB. Já se mete a colher em briga de marido e mulher. Sao Paulo Perspec. 1999;13(4):82-91.         

21. Sagot M, Carcedo A. Ruta crítica de las mujeres afectadas por la violencia intrafamiliar en América Latina: estudios de caso de diez países. Washington, DC: Organización Panamericana de la Salud; 2000.         

22. Santos SM, Barcellos C, Carvalho MS, Flores R. Detecção de aglomerados espaciais de óbitos por causas violentas em Porto Alegre, Rio Grande do Sul, Brasil, 1996. Cad Saude Publica. 2001;17(5):1141-51. DOI:10.1590/S0102-311X2001000500015        

23. Schraiber LB, D'Oliveira AFPL, Couto MT. Violência e saúde: estudos científicos recentes. Rev Saude Publica. 2006;40:112-20. Nº espec. DOI:10.1590/S0034-89102006000400016        

24. Schraiber LB, D'Oliveira AFPL, França Junior I, Diniz S, Portella AP, Ludemir AB et al. Prevalência da violência contra a mulher por parceiro íntimo em regiões do Brasil. Rev Saude Publica. 2007;41(5):797-807. DOI:10.1590/S0034-89102007000500014        

25. Souza SD. Revista Mandrágora: gênero e religião nos estudos feministas. Rev. Estud Fem. 2004;12:122-30. Nº espec. DOI:10.1590/S0104-026X2004000300014        



Stela Nazareth Meneghel
Escola de Enfermagem
Universidade Federal do Rio Grande do Sul
R. São Manoel, 963 - Rio Branco
90620-110 Porto Alegre, RS, Brasil
E-mail: stelameneghel@gmail.com

Received: 1/7/2010
Approved: 11/14/2010
Research funded by Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq - Process nº: 472238/2007-8).



Study presented at the meeting "Fazendo Gênero 9", held in Florianópolis, 2010.
The authors declare no conflicts of interests.
a Inter-American Institute of Human Rights. I Regional Report Situation and Analysis of Femicide in the Central American Region; 2006 Aug. San José, CR: IIHR; 2007.
b Brazilian Institute of Geography and Statistics. Estatísticas do Registro Civil 1999/2004/2009. [cited 2011 Mar 18] Available from: http://www.ibge.gov.br/home/presidencia/noticias/imprensa/ppts/0000000205.pdf [sent to the library for harmonization]
c IBGE. Mapa do mercado de trabalho no Brasil: 1992-1997. Rio de Janeiro :IBGE, 2001) [sent to the library for harmonization]

Faculdade de Saúde Pública da Universidade de São Paulo São Paulo - SP - Brazil
E-mail: revsp@org.usp.br