The impacts of the age of majority on the exposure to violent crimes

Abstract

This research evaluates the impact of the criminal majority law by analyzing the distribution of homicide victims with ages close to 18 years from roughly 20,000 Police Reports of the Brazilian state of Minas Gerais. Considering young men, we find an increase in the density of murders after the age of criminal responsibility. Supposing that participation in criminal activities increases the probability of being a victim of violent crimes, especially among young men, our results are the opposite of what is expected with the increase in punishment. We outline data suggesting that the changes in lifestyle and the worsening of socioeconomic conditions after the age of 18 may offset the deterrence effect due to increases in risk behavior and relative gains of criminal activities. Finally, we estimate the offenders’ ages to evaluate the effects of lowering the age of criminal responsibility (in 2 years), based on the relationship between the victims and perpetrators and simulating hypothetical age differences among them. At most, 9% of all the murderers and 3% of femicide perpetrators would have more severe sentences. Overall, the results indicate that a reduction in the age of criminal responsibility would not deter criminal activities and would target a small share of criminals.

Appendices

Potential effects of the age of majority on the chance of committing or being a victim of homicide

Our identification strategy relies on the reduced form effect of the age of majority, considering minor variations in the ages of the victims near the cutoffs. We argued that criminal sanctions are an essential modification after the majority, but there are other changes. To better interpret the potential effects related to the age of majority, we define a single structural equation linking the chance of victimization to i) the behavior toward the risk of victimization, ii) the behavior toward participation in crime as the perpetrator, iii) the deterrence effect due to the severity of punishment and the risk of being arrested, and iv) the effects of incapacitation when more criminals are isolated from the rest of society.²⁵

Criminal responsibility affects the incentives to get into the illegal market as a perpetrator. The civil age of majority can increase exposure to homicides as the victim, but can also increase participation in crime as the perpetrator due to deterioration of social conditions. On the other hand, more severe punishment and more incarceration reduce the number of criminals and the potential supply of offenders in society. We consider that reaching the age of criminal responsibility increases the prison sentence length that a potential homicide will face, but does not alter the probability of going on trial (even under the juvenile rules) and being isolated from society.

Let us suppose a simple scenario in which the individual probability of being a victim of homicide (T=1) depends on the behavior toward the risk of the victim (R), which includes lifestyle and habits like the use of drugs and hanging out at nightclubs and bars with friends. Association in criminal activities and propensity to be a criminal (P) increases the chance of being a victim—due to a reaction of the victim, the police or rival criminals, for example. Consider the relationship between the chance of being murdered, the risk behavior and the criminal behavior:²⁶

$$\begin{aligned} T = \beta _0' X + \beta _1 R(\mathrm{Age})+ \beta _2 P(G(\mathrm{age}), L(\mathrm{age}, \lambda )) + \beta _3 (1- \lambda ) \sum P(L, G) + \epsilon \nonumber \\ \end{aligned}$$

(7)

where X is a vector of factors related to violence: demography (proportion of male and young adults), geography/neighborhood, overall and local conditions of the economy, among others. $\epsilon$ is a random idiosyncratic shock in the individual probability of being murdered.

R is a function indicating the risk behavior toward exposure to a violent crime, which may increase after the civil age of majority due to the possibility of attending bars and consuming alcoholic beverages, for example. $\beta _1$ is the effect of risk exposure on the probability of being murdered, expected to be positive.

P is the individual propensity/behavior toward exposure to crime as the perpetrator. It is a function of the costs and revenues of criminal activities, including social norms and cultural conditions (not explored in this paper) and the objective punishment of being in prison and having a criminal record.

Therefore, G is a function indicating the gains from crime, which we consider a positive function of social vulnerability (income, unemployment, and other socioeconomic conditions). Sachsida et al. (2010) show that income inequality, as well as urbanization and unemployment rates, is correlated with homicide rates in Brazil. Gould et al. (2002) show that the job market conditions of young and unskilled adults, as well as unemployment and wages, helped explain the crime trends in the 1980s and 1990s in the USA. Many papers making international comparisons across countries suggest a significant influence of economic growth and income inequality on violent deaths (Soares 2004; Fajnzylber et al. 2002; Lee et al. 2014).

L represents the punishment to the offenders according to criminal law, and $\lambda$ is the proportion of offenders convicted and sentenced to prison, which we consider a proxy of the probability of being arrested for homicide. This variable is related to the judiciary system and the criminal investigation capacity of the police. There is a lack of evidence that this would impact the Brazilian environment. However, international surveys indicate that the risk of being sentenced could be more influential in deterring participation in crime than the severity of punishment (Bun et al. 2020).

We consider that $P=1$ if $G(\mathrm{age}) - \lambda L(\mathrm{age})>0$, and zero otherwise. The functions R(.) and P(.) are censored variables as they regard the behavior and propensity of each individual to be the victim and the perpetrator of crimes and are only observed when these situations happen. We consider the Age $=$ 1 if a person is older than 18 years (reaches the civil age of majority and the age of criminal responsibility), and zero otherwise, which impacts both exposure to victimization and criminal participation. Even regarding crime participation, the net effect on $G(\mathrm{age}) - L(\mathrm{age})$ is not clear, as the gain from crime as well the punishment increase after the age of majority.

$\beta _2$ is the effect of criminal activities on the probability of being a victim. According to the overlap assumption between the offenders and the victims’ characteristics and the correlation between the distribution of victimization and offenders, $\beta _2$ should be positive. The literature also supports the overlap assumption for crimes of homicides (Broidy et al. 2006; Taylor et al. 2008; Katz et al. 2011).

$\sum P(L, G)$ is the total of criminals in the society, and $\beta _3$ reflects the effects of the number of offenders on the probability of being murdered. Gaulez et al. (2018) find no effect of incarceration on the homicide rates in Brazil, while Shikida et al. (2014) find that an increase of 50 prisoners in the judiciary system reduces the homicide rate by 1 per 100,000 inhabitants. $\lambda$ is the proportion of criminals that were incarcerated. According to Nery and Nadanovsky (2020), 25% of homicides in Brazil are solved with the suspects going to trial.

As we use individual data, a person’s age of majority does not lead to changes in the aggregate of criminals ($\sum P(L, G)$) and the incarcerated population ($\lambda \sum P(L, G)$). However, reducing the age of criminal responsibility could raise incarceration and then reduce the number of offenders in society. This scenario allows us to isolate the deterrent effect from the incapacitating effect.

We also consider that $\lambda$ cannot be manipulated by the criminal and its distribution is smooth around the age of criminal responsibility for severe violent crimes. As long as there is some sub-notification of homicides (even less than other crimes), we do not believe that the police and the judicial system would privilege an individual who is 18 years old over someone else who is younger. Moreover, the vast majority of homicide investigations in Brazil are not concluded. There is no evidence that this is correlated with the age of the offender.²⁷ So the only aspect to be increased would be punishment severity (more years in prison if convicted).

Thus, at the individual level, harsher punishments reduce the relative gains from criminal activities and thus reduce the chance of participating in illegal and potentially violent activities, decreasing the chances of being a victim of murder (according to the parameter $\beta _2$). On the other hand, the civil age of majority increases exposure to risk, R, due to social deterioration, drinking, freedom to hang out in bars and nightclubs at night, and moving out of the parents’ house. It can also worsen economic conditions, reducing the opportunity cost of criminal activities and raising the crime rate and, consequently, the chance of victimization.²⁸ Thus, the civil age of majority can increase the probability of being murdered both due to increased risk exposure as a victim and due to increased gains from crime activities compared to the licit job market.²⁹

Typology of homicides and caveats of data

Brazil does not have a national record system for crimes. For crimes of homicides, there are two sources of information: the National System of Mortality (SIM), reported by all the healthcare institutes in Brazil, and the Police Reports.

Brazilian law 6.015/1973 requires that all deaths have a declaration of death to be completed by a physician mentioning the primary cause of death, disclosed by the healthcare institution and recorded in an electronic system. The cause of death is classified according to the International Classification of Diseases (ICD), followed by all healthcare institutions in Brazil and includes unnatural or violent deaths (homicide, suicide, or accident). If there are signals of violence, the police should be called, and the victim’s body should be taken to the judiciary medical center for an autopsy.

In this case, a Police Report is also completed, the information from the healthcare system and the police should be similar. However, de Castro et al. (2003) compared information from these two databases from Minas Gerais and found that the Police Reports have better coverage in small towns, while SIM coverage is better in large cities. Some possibilities for these differences include i) different forms of filling in the information and classifying the deaths, ii) different processes and delay in updating the Police Reports, and iii) errors in filling in the Police Reports or death certificates completed by healthcare professionals.

Police Reports refer to the fact, place, time, and victim of a criminal activity, and crime description does not necessarily match the legal classification, nor does it follow the terminology of the ICD. In the case of homicides, they are divided into attempted and consummated. On the other hand, healthcare institutions receive patients who demand their services, including those victims of violence, who survive and then die after hospitalization. If there is already a police report for attempted murder, it should be updated after death. Moreover, many deaths are classified as “deaths from unknown causes” or other forms of violent death, such as in cases of suicide.³⁰

We decided to collect data from Police Reports only about committed homicides. In this case, there is certainty that the homicide was indeed committed, with little chances for error: for example, reporting suicide as homicide or reporting an attempt of homicide as a serious bodily injury. Each State Secretariat has its own system and standards for collecting and classifying this type of crime. Most states do not provide information about the alleged relationship with the victim, for example. Furthermore, information structure is not standardized, i.e., each Brazilian state can store its information in a customized way. Thus, we chose to check only the state of Minas Gerais.³¹ The Police Reports from Minas Gerais provide a unique piece of information: the probable relationship between the victim and the perpetrator. This information reveals certain identification of the causes and motivation for the homicide.

Costa et al. (2018) analyze the data from arrests in Rio de Janeiro and comment on the possibility of crimes being under reported, especially against property, committed by minors. Moreira et al. (2018) estimated an under-reporting rate of homicides in Minas Gerais of 4.47% for the period of 2005 to 2009, while Cerqueira (2013) finds that the “true” national homicide rate should be 18% higher than the one reported by the official statistics. There can be some under-reporting for homicides, but it should be much smaller than for crimes like robbery. It is unlikely that there will be a depreciation in situations where the victim (and even the aggressor) is a minor. Contrary to Costa et al. (2018), we are looking at data on victims and crimes against life are considered the most serious ones.

Moreover, a Police Report is mandatorily completed if a body is found (Moreira et al. 2018). There is no argument or evidence that the police would under report homicide victims because they are minors, meaning that any error or manipulation of the data must be “almost-random” around the age of 18 years. Although the offenders face different sanctions and justice systems, whether they are older or not, the investigation is the same regardless of the victim’s age.

In general, homicides can be divided according to the underlying situation (Engel 2015; UNODC 2016). The first refers to murders related to “Gangs and drugs,” typically involving young men in vulnerable regions, often associated with many murders in Brazil. Indeed, Beato et al. (2001) shows that homicides in Belo Horizonte, the capital of the state of Minas Gerais, are concentrated on slums, usually dominated by drug dealers in Brazil. In this situation, we conjecture that the perpetrator’s alleged relationship with the victim should be either known (some relationship or even friendship) or unknown. Gang participation is linked to social and cultural norms of youth and male stereotypes of violence. It is also related to the use of chemical substances and a greater chance of victimization (Katz et al. 2011; Taylor et al. 2008).

The second classification of homicides refers to violence in crimes against property. It is generally a robbery followed by a murder (called “latrocínio” in Brazil. It usually happens because the victim or the police reacted). It represents a small number in the total number of homicides in Brazil (less than 3% of the total of homicides in 2020 (IPEA 2017a)). In these situations, we conjecture that the alleged relationship with the victim should be (usually) unknown (strange).

The third cause of death is interpersonal violence, which includes quarrels in bars and parties, where the presumed relationships are of being an acquaintance or even friend, but also can include crimes among neighbors or colleagues from work. Unpremeditated crimes are linked to social vulnerability, violent culture, and alcohol and drug abuse (Murray et al. 2013).

We can find some evidence that interpersonal violence accounts for a significant share of the homicides in Minas Gerais. Using our Police Report database, we show the distributions of time of day and the day of the week when the homicides happen. We see in Fig. 5a that there is a large share of crimes against life occurring after 8:00 PM (after the business hours), and Fig. 5b shows that the days with more homicides are Saturdays and Sundays. Indeed, 10% of the homicides happen during the dawn in the weekend and on Mondays (between midnight and 6 AM). The share is the same when we consider crimes among friends only. The fact that most crimes occur at night during the weekends shows that they must be a consequence of increased risk exposure during leisure activities, presumably associated with alcohol-related brawls.³²

The last cause of homicides is the one that mainly affects women: domestic violence. Generally, the probable offender is the spouse, love partner or ex-partner. Zilberman and Blume (2005) and Murray et al. (2013) point to the association between domestic violence and drug abuse.

Fig. 5

Source: Police Reports obtained from the Public Security Observatory of Minas Gerais state

Distribution of day of the week and time of homicides in Minas Gerais. Note Distribution of murders in Minas Gerais state between 2012 and 2016 according to a time of the day (subdivided by classes of 1 h of duration: 0–1 o’clock, 2–3 h, ...23–24 h), and b the day of the week when the crime happened

Policy recommendation: costs and benefits of reducing the age of criminal responsibility

³³We concluded in the paper that the effects of reducing the age of criminal responsibility from 18 to 16 years on mortality rates would be low since the deterrence effects of harsher penalties were not detectable in our environment. Also, only a tiny portion of potential murderers would be in this age range—which does not imply that there are proportionately few underage criminals, given the population in this age range, but that the change is too small to have any effect on overall rates (the variation of the denominator in the age histogram is narrow, and so is the area).

However, we observed a discontinuous increase in the density of homicides just after the age of 18, which we assume to be related to the effects of civil age on the chances of being the perpetrator or victim of crimes. The histograms in Fig. 1 show that the peak of murders occurs immediately after the age of majority, and the number of victims is negatively correlated with age for minors. Our sample shows 1209 homicides with victims between 16 and 18 years old, 1660 victims aged between 18 and 20, and 1614 victims aged between 20 and 22. The number of victims is reduced to 392 when considering young people between 14 and 16 years old. The populations in these age ranges represent a similar percentage of the whole population (3.5%).³⁴

Thus, just reducing the age of the majority to 16 years (or even further) would not change the current crime scenario in Brazil. In this sense, public policies and debates should focus on other overlooked ways of reducing criminality. Especially, the results suggest that improving social conditions for young adults could help diminish homicides.

Nonetheless, even a slight reduction in a scenario of considerable violence in Brazil can represent thousands of lives being saved. Indeed, Table 6 shows that up to 1433 murderers would be between 16 and 18 years of age (under the usual assumption that criminal behavior is associated with the risk of being murdered). Thus, if the deterrence effect was fully efficient in deterring homicides, up to 1433 lives could be saved in 5 years with a reduction in the age of majority to 16 years (without increasing the number of prisoners). Under any conjecture about the value of human life, such a policy would be advantageous.

The point is that a reduction in the age of criminal responsibility would not induce a reduction in violent behavior (and, therefore, in the number of victims) since the deterrent effect of the majority law is low in our setting. The incarceration effect is also potentially low, as only a tiny proportion of homicides are solved. On the other hand, the costs of expanding the conventional justice system to minors between 16 and 18 years can be high: prosecuting even those who are acquitted, influenced or not by the deterrence effect, for criminals who would or would not commit crimes again.

As we do not find any discernible deterrence effect of harsher penalties,³⁵ the only way that reducing the age of criminal responsibility would impact crime production is through the incarceration effect, under the assumption that convicted murders have more chances of recidivism. However, many studies point out that going to prison increases the chances of recidivism (CNJ 2019; Loeffler and Nagin 2022; Mears et al. 2012). Thus, recidivism can be more strongly associated with the criminal influence in prison than with the idiosyncratic behavior of the criminals. The main obstacle to evaluating the costs and benefits of such a policy is that public and credible data about murderers’ convictions and recidivism are scarce in Brazil.

According to a survey conducted by the “Sou da Paz” Institute with data from 17 states, the national homicide resolution rate was 44% in 2018. This rate changed from 40% to 46% in Minas Gerais between 2015 and 2018 (de Carvalho Silva and Gaudêncio 2021). The state is the third with the highest number of incomplete police investigations (12,032), with only half resulting in prosecution in court. Only 3.24% of all the police inquiries were concluded (ENASP 2012). Nery and Nadanovsky (2020) computed 25,631 homicides in Minas Gerais between 2009 and 2014, but only 9512 individuals were arrested in 2016 for this crime (37%).

Regarding recidivism, a national survey conducted in 2015 by the National Justice Council (CNJ 2019) shows that 42% of the individuals prosecuted in 2014 had reentered the judicial system by 2019. Meanwhile, the rate of reentry of minors into the socioeconomic system was 23.9%. All the states have a higher recidivism rate for adults, except Minas Gerais—9.5% for adults against 13.5% for minors.³⁶

Therefore, considering conservative estimates of the murder resolution rate in Brazil, around 40%, and that 9.5% of murderers would repeat this offense if not arrested, we can use our data to calculate the economic costs and benefits of reducing the age of criminal responsibility. According to Table 6, at most 286 murderers (1433/5 years) would be between 16 and 18 years in a typical year. So then, we have a maximum of 40%*9.5%*286 $=$ 11 lives would be saved annually in Minas Gerais.

However, we would have to prosecute and process all the 286 murderers in the judicial system, independently of whether they would commit a homicide again. Also, 40%*286 = 114 would be convicted and arrested in prison. From an economic perspective, the question is whether the cost that society is willing to pay for reducing 11 deaths per year would surpass the costs of prosecuting and sending 114 individuals to jail. The economic Value of Statistical Life is defined as the total amount that the society is willing to pay to get marginal decreases in the risk of death. Considering that the economic value of living ranges from R$1 to R$4 million in Brazil (Pereira et al. 2020; Lavetti and Schmutte 2016; Ferrari et al. 2019),³⁷ that means the society is willing to pay R$11 million to R$44 million at most for increasing punishment and avoiding 11 homicides.

We now turn to the social costs of reducing the age of criminal responsibility. According to data from the National Justice Council (CNJ 2021), the annual cost of a prisoner in Minas Gerais is R$28,650. Considering the minimum penalty of 6 years in a closed regime for intentional homicide and the maximum of 20 years,³⁸ the total cost per prisoner varies between R$171,720 to R$858,600, disregarding the costs of investigation and prosecution of the person convicted until the final decision of the court.³⁹ Therefore, for the 114 individuals convicted in a year, the total amount varies from around R$19,576 million to R$97,880 million.⁴⁰

There are also various costs related to subjective probabilities in the criminal process: the chance to convict an innocent person or the prisoner being co-opted by gangs. The criminal does not earn money when in jail, and there is evidence that future earnings also drop when a person goes to jail (Selbin et al. 2018). Furthermore, some studies correlate imprisonment to an increased chance of committing crimes (Loeffler and Nagin 2022). Underaged criminals receive educational and psychological support that can reduce the risk of recidivism without the stigma (and economic losses) of being filed in the justice system. Finally, reducing the age of criminal responsibility would increase the demand for investigations, prosecutions, and arrests by the police and the justice department. As a large share of homicides is not solved, an increased number of unsolved homicides can reduce even further the credibility of the judicial system.

An accurate evaluation of the costs and benefits of modifying the age of criminal responsibility would depend on more precise nationwide information about the victims and offenders, the costs of the process, the loss of future earnings when one person is arrested and convicted for murder, and the effects of prison and socio-educational systems on the probability of recidivism. Without this information, reducing the age of criminal responsibility—potentially increasing social vulnerability, crime participation, and risk behavior—would be only a “shot in the dark.”

In our perspective, first, more effective crime-fighting policies should improve information availability about the murderers—identifying and tracking recidivism rates. Secondly, integrate information about the state criminal systems to be possible to have a national picture of the victims and murderers by age. Finally, the clearest strategy to deflect the homicide epidemic in our environment is to increase the velocity and certainty of prosecuting the perpetrators of homicide. Increasing the certainty of punishment can imply more coercive deterrence effects for the entire age distribution with more pronounced effects on the homicide rate.

Future economic research could focus on the impacts of labor market on criminal behavior.⁴¹ Other non-economic issues should naturally be part of the debate when discussing this policy and are beyond the scope of this article. In particular, the legal aspects regarding the constitutionality of reducing the age of criminal responsibility (ANADEP 2015), the psychological aspects regarding the age at which the youngsters have conscience of their acts, the effects of prison on recidivism, and the extent to which rationality can still be present in a scenario of high vulnerability like the one in Brazil (Jacobs and Wright 2010).⁴² Likewise, there are several variants of the proposal, for example, reducing the age of criminal responsibility for heinous crimes only. This article contributes by providing ex-post and ex-ante estimates of the potential impacts of this kind of policy.