OSCILLATIONS AROUND A PLATEAU : DEATHS BY HOMICIDE IN BRAZIL – 1979

This text focuses on the analysis on homicide/assault rates in Brazil over three decades: 1979-2013. It disaggregates the information by area – Great Regions in Brazil and provides separate statistics for young adults and young adult males, the most vulnerable group. Rates are calculated as the ratio of registered deaths by causes from the Ministry of Health and the population estimated by the Central Statistical Office IBGE. To further enhance the comprehension of the process at a lower ecological unit, we focused on the Southeast Region, which presented the sharpest drop. The analysis is replicated for all four states in the Region. Results showed that some short lived downward trends were just part of oscillations around a plateau. This plateau was the consequence of opposing trends among the several regions/states in Brazil. Additionally, a hypothesis raised in sociological literature, that the proportion of young adults in the population is positively correlated with homicide rates is ruled out by our data.


INTRODUCTION
Several statistics referring to the beginning of the first decade of the century indicate a drop in violent deaths in specific areas of Brazil and for the country as a whole: years of life lost (BELTRÃO andDELLASOPPA, 2011a, 2011b), age-specific mortality rates from health registers (BRASIL, 2006), intentional violent crime registers (BRASIL, 2007), young adult violent deaths (FERREIRA, 2005;WAISELFISZ, 2011) among others.Some authors include Brazil as part of an international decline in crime rates (see TSELONI et al, 2010).Nonetheless, the subsequent years showed that this short-lived downward trend was just part of what is now perceived as oscillations around a plateau.The plateau for the combined crude mortality rate by homicides for both sexes in Brazil is around 27.3 deaths per one hundred thousand population (between 1998 and 2013), while for young adults it doubles to 54.5 (between 2000 and 2013) and for young adult males it doubles again to 101.7 (same period).
It is worth noting, though, that these recent figures for Brazil are very high by international standards, well above the 10 per 100 thousand level classified by the World Health Organization as epidemic (apud PNUD, 2013).The Global Study on Homicide (UNODC, 2011c) states that the "… homicide rate in Africa and the Americas (at 17 and 16 per 100,000 population, respectively) is more than double the global average (6.9 per 100,000), whereas in Asia, Europe and Oceania (between 3 and 4 per 100,000) it is roughly half."Considering countries individually, shows the thirty highest homicide rates per 100 thousand using Criminal Justice sources (UNODC, 2011a): countries from the Americas are denoted as red bars (Brazil in black), African countries as blue bars and European countries in green.Among the 30 countries with the highest rates, 23 are from the Americas, 6 from Africa and only 1 from Europe.Asian and Australia/Oceania countries are not included among these thirty.Rankings using different international sources could vary accordingly.Homicide rates from health organizations are usually remarkably different from those originated from Criminal Justice sources (see UNODC 2011b for a comparison of the two sources).More elaborate indicators which include interpersonal violence, such as the Violent Societies Index proposed by Karstedt (2012)  A large group of hypotheses is found in the international literature to explain recent changes in criminality and homicide figures in several developed countries (AEBI and LINDE, 2010;BARBAGLI, 2000;BARKER, 2010;BAUMER, 2008;BLUMSTEIN and WALLMAN, 2006;FOX, 2006;HAGAN, 1994;HAGEDORN and RAUCH, 2004;HUME, 2007;KILLIAS and AEBI, 2000;MORO, 2006;PEARCE, 2006;RENNISON and PLANTY, 2006;RODGERS, 2006;WALLMAN and BLUMSTEIN, 2006;WILKINSON, 2005;ZIMRING and FAGAN, 2000).As is usual in Social Sciences, no consensus can be verified among the authors about the capacity of these hypotheses to explain factors related to the evolution of criminality and homicide/assault, an opinion shared by Tseloni et al (2010).Farrell (2013) singles out and analyzes 15 of those hypotheses related to: "…lead poisoning; abortion legalization; drug markets; demographics; policing numbers and strategies; imprisonment; strong economies; the death penalty; gun control; gun concealment; immigration; consumer confidence; the civilizing process; and improved security."Farrell proposes a battery of five tests as a first filter for those hypotheses.This paper will consider one of those hypotheses: the effect of the proportion of young adults in the population on crime (the demographic hypothesis).
None of these hypotheses considered individually seem to satisfactorily explain homicide rates in Latin America.Analyzing the specific case of Brazil, we can observe two fundamental importance of crime against property in the developed countries; second, the homicide rate in Brazil is not following a downward trend and stabilizing around lower values, but after a long period of increase, it is oscillating around a plateau at a very high level.Brazil also differs from other Latin American countries regarding the interplay of homicide rates and economic development.The stability occurs while other countries present increasing trends in crime and violence, paradoxically concomitant with economic development and reduction in inequality, poverty and unemployment (PNUD, 2013).
In Brazil, crude mortality rates for the country as a whole have remained basically stable since 1995 (external causes around 72.4 per 100 thousand) or 1998 (homicide/assault around 27.3).These plateaus were the result of opposing trends among the several regions/states in Brazil, some showing sharp decreases counterbalanced by sharp increases in other regions/states.As already shown in the literature, the ecological unit considered exerts a strong influence on the level and pattern of the variable analyzed ( VAN WILSEN, 2004).The continental size of the country, more than 3 million square miles, may explain the lack of homogeneous behavior with respect to cultural and sociopolitical issues.Though these figures for Brazil are high by international standards, they are smaller than those of several other countries in Latin America, the most violent region in the World.
All these curtailed lives result in huge human, social and economic costs.To better understand the importance and historical variations of these deaths, this text focuses on the analysis of homicide/assault rates in an extended period, over three decades: 1979/2013 (the latest available information).
At a comparative level, it must be noted that there are profound and nontransferable cultural differences among countries, which are synergic to the evolution of these processes: urban ecologies, lifestyles of young adults at risk, and social, cultural and economic factors (like macrolevel changes in capital investment, distressed communities with residential segregation, racial inequalities and concentration of poverty), and the perception of these inequalities, the access to goods and services (like sneakers, cellphones, internet and social nets) and changes in the economics and trends of the use of illegal drugs.Another influence is the always changing geographical distribution and relations of power among the organizations dealing with production and distribution of illegal drugs (MOELLER and HESSE, 2013).Often, the borderline between deaths caused by wars/guerrillas and homicide is not clear (SOUZA et al., 2012).All these __________________________________________________________________ Revista de Direito da Cidade, vol.07, nº 4. Número Especial.ISSN 2317ISSN -7721 pp.1923ISSN -1943 1928 Department of each Brazilian state.This text uses data from the Ministry of Health with external causes further disaggregated as homicides and other external causes classified according to ICD, while the IBGE uses solely a dichotomous classification of "violent/non-violent." From the inception of the mortality statistics collecting system in 1979 up to 1996, the Ministry of Health system used ICD-9 to classify causes of death and in the remaining years, ICD-10.
The equivalence between classifications was done following the Pan-American Health Organization (PAHO, 1999) recommendations.Disaggregated data by homicide/assault does not exhibit a discontinuity with the changes introduced by the 10 th Revision.The same happens for other groups of causes (see BELTRÃO andDELLASOPPA, 2011a, 2011b).For non-census years, population at risk is estimated by linear interpolation of data by sex and age groups counted in census years (IBGE 1982(IBGE , 1996(IBGE , 2001(IBGE , 2011)).It must be noted that the analysis of the years previous to 1979 is unfeasible since data collection by the Ministry of Health system only started this year.
Ministry of Health data have some quality drawbacks: i) incomplete death coverage; ii) possible misclassification of causes of death as well as the fact that "Deaths due to signs, symptoms and ill-defined conditions are presented in a separate category…" (PAHO, 1999); and iii) the presence of registers with unknown age and/or sex.Due corrections for all these situations as described in Beltrão andDellasoppa (2011a, 2011b) are used in the present text.No correction was performed to adjust for the first shortcoming.Assuming that death coverage has been changing smoothly (no drastic changes occurred in the period, neither on regulation, nor by public campaigns), the impact would be of an increasing trend.It is worth noting that some states showed no undercoverage 5 .For the disaggregation used in this text, an upper limit 6 for the second shortcoming was estimated using information contained in Mello Jorge et al. (2002a, 2002b).Beltrão andDellasoppa (2011a, 2011b), analyzing information up to 2010, verified that completeness of age and sex information on death registers was, with almost no exception among the Great Regions 7 and states, higher than 98%.This figure is 96% for deaths by "External causes" and 95% for deaths by homicide.Following the procedure defined in Beltrão and Dellasoppa 5 Estimates for the state of Sao Paulo indicated full coverage for 1991, 2000 and 2010.For 1980, coverage was 99/98% (respectively for males and females). 6There is some regional discrepancy with a greater proportion of ill-defined causes of death in the North and Northeast, but not enough to jeopardize our analysis. 7Brazil's 27 states are divided into 5 allegedly homogeneous regions denominated Grandes Regiões (Great Regions), namely: North (with the states of Rondônia, Acre, Amazonas, Roraima, Pará, Amapá and Tocantins), Northeast (Maranhão, Piauí, Ceará, Rio Grande do Norte, Paraíba, Pernambuco, Alagoas, Sergipe and Bahia), Southeast (Minas Gerais, Espírito Santo, Rio de Janeiro and Sao Paulo), South (Paraná, Santa Catarina and Rio Grande do Sul) and Midwest (Mato Grosso do Sul, Mato Grosso, Goiás and the Federal District).__________________________________________________________________ Revista de Direito da Cidade, vol.07, nº 4. Número Especial.ISSN 2317ISSN -7721 pp.1923ISSN -1943ISSN 1929ISSN (2011aISSN , 2011b) ) deaths with unknown sex and age were proportionally distributed, in that order, according to the known distribution.Proportional mortality by ill-defined causes, while ICD-9 was under use, fluctuates around 6%.With the implementation of ICD-10 in 1996, there was a slight increase in these proportions, up to around 7%.
Crude mortality rates by homicide/assault for each sex were estimated for the years between 1979 and 2013, both for all age groups and for the young population (between 15 and 24 years of age).Similar statistics were also calculated for both sexes combined.Dealing with each sex separately helps to control possible errors (statistics for women not shown in the text).Besides the estimates for the country as a whole, separate estimates were computed for each Great Region and, separately, for all the states in the Southeast region.

Homicide rates for the Brazilian population
To observe the recent stabilization at such high levels in context, it is adequate to weigh against the evolution in a more extended period for which mortality information by cause is    In Brazil, as in most Latin America countries (PNUD, 2013), young adults are the most affected by homicides, both as perpetrators as well as victims.Young adult deaths represent around a third of all homicides during the period under analysis.Rates are even higher for young males.
__________________________________________________________________ Revista de Direito da Cidade, vol. 07, nº 4. Número Especial. ISSN 2317-7721 pp.1923-19431933 shows the cause specific mortality rate by homicide of young adults for Brazil as a whole and Great Regions from 1979 to 2013.For Brazil and regions, patterns for young adults are quite similar to the ones observed for all ages, but encompassing a larger dynamic range and much higher values for the rates.It is also worth mentioning the absence of the plateau observed for the Southeast between 1999 and 2003.In the last stretch between 2004 and 2013, while rates for Brazil remain almost stable (since 2000), trends for each region follow different patterns, similar to the ones observed for the total population: the Southeast presents a strong declining trend up to the last two years, against a strong increasing trend for the Northeast, North and Midwest regions and a rebound of the South reaching values close to those of the Southeast.This rebound in the South may be interpreted as a 5-year belated decrease (with respect to the Southeast).Regions, 1979-2013 shows cause specific mortality rate by homicide of young adult males (15-24) for Brazil and Great Regions in the 1979-2013 period, and the plateau since 2000 with a level twice the value observed for both sexes combined: 100.1 deaths per 100 thousand.For this group, the plateau is also the result of divergent trends, though similar to the ones observed for the population as a whole: decreasing for the Southeast, increasing for the North, Northeast and Mid-West and a bump for the South.As already mentioned this bump for the South can be interpreted as a belated decrease with respect to the Southeast.Regions, 1979Regions, -2013 Source: Authors' calculation Note: Rates calculated without correction for coverage __________________________________________________________________ Revista de Direito da Cidade, vol. 07, nº 4. Número Especial. ISSN 2317-7721 pp.1923-19431936For young adult males, Brazil's mortality rates by homicide stabilizes slightly higher than that of young adults of both sexes combined.The same happens for the breakdown by region.The ratio between mortality rates for the young adult males in the last and the first year is slightly larger than the one observed for young adults of both sexes: North (4.6), Northeast (11.5),Southeast (2.7),South (4.6) and Midwest (9.0).
shows the same information of the previous Figure but disaggregated by the corresponding states of the Southeast region.Again, one observes a similar pattern at much higher values than those observed for all ages ( ) and for young adults of both sexes (Erro!Fonte de referência não encontrada.).
available: 1979/2013.Crude homicide mortality rates (deaths by 100 thousand population) between 1979 and 2013 are displayed in Erro! Fonte de referência não encontrada.. Erro! Fonte de referência não encontrada.displays the corresponding index rates with 1979 equal to 1.For Brazil as a whole (black line), the rate exhibited an upward trend (with oscillations) from 1979 to 2003, interrupted in 2003 by a small drop up to 2005, followed by fluctuating but stable high homicide rates between 2005 and 2007 and a slight upward trend from then on 8 .Between 1998 and 2013, one could consider that values are oscillating around 27.3 deaths per one hundred thousand.When considering Brazil's Great Regions, the situation is more complex with no homogeneous trends: the homicide rate drop begins in the Southeast Region (violet line) in 2003 after a plateau period, from 1999 to 2002, when homicide rates stabilize around 37 deaths per 100 thousand, falling to a figure 44% smaller, 20.8.All other regions show an upward trend with heterogeneous oscillations from 1979 on, reaching, in three of the Great Regions, values that almost double the __________________________________________________________________ Revista de Direito da Cidade, vol.07, nº 4. Número Especial.ISSN 2317-7721 pp.1923-1943 1930 one for the Southeast: North (35.7),Northeast (39.4) and Midwest (37.1).The exception is the South with a homicide rate of 20.8 deaths by one hundred thousand.Changes are of different magnitude among the regions.The increase in the other regions was proportionally higher than the decrease in the Southeast, but because of the much larger population of the Southeast the final result for Brazil is the relative stability of the national homicide rate observed since 1998.The population in the Southeast encompassed 43.5% of the total population in 1979.This proportion has been rather stable over time with figures of 42.0% in 2013.

Figure 2 -
Figure 2 -Cause specific mortality rate by homicide, Brazil and Great Regions, 1979-2013

Figure 4 -
Figure 4 -Cause specific mortality rate by homicide, Southeast Region and corresponding states, 1979-2013