Abstract
A population-based case-control study was conducted to assess the relationship between genocide exposure and homicide perpetration in Rwanda. A sample of 150 homicide perpetrators who were charged with and confessed to having committed homicide between 1 May 2011 and 31 May 2013 and 450 controls were enrolled. Cases were matched to controls by neighborhood, age and sex. Socio-demographic, background and genocide-related information was collected from study subjects’ next of kin. Four characteristics of genocide exposure were: genocide survivor, genocide perpetrator, having lost a first-degree relative to genocide and having a first-degree relative convicted of genocide. We assessed the impact of each genocide-exposure variable using conditional logistic regression. Of the 150 cases, 124 (82.7%) were male and 26 (17.3%) were female. The mean age of the alleged homicide perpetrators was 33 years, with a peak in the age group 20-29 years (39.3%). After adjusting for socio-demographic characteristics and past common criminal records, having a first-degree relative who had been convicted of genocide crimes was a significant predictor for homicide perpetration (odds ratio [OR] = 14.4, 95% confidence interval [CI] = 1.6-129.4). Being a genocide perpetrator, a genocide survivor and having lost a first-degree family member to genocide were not identified as risk factors for homicide perpetration. In Rwanda, young people who experienced early exposure to trauma by witnessing their first-degree relatives’ active participation in the genocide, are more likely to commit homicide. Socio-economic and psychotherapeutic programs targeting this population group are needed to rehabilitate these young people for violent behavior change.
Introduction
Homicide is legally defined as the unlawful or deliberate actions of a person resulting in the death of another person (United Nations Office on Drugs and Crimes, International Homicides Statistics, 2004). According to scientific literature and crime reports from Western countries, homicide perpetrators are often young males, with a low level of education and a high prevalence of unemployment, poverty, family instability, and psychological disturbance (Bourget & Gagne, 2012; Brodie, Dady, Crandall, Sklar, & Jost, 2006; U.S. Centers for Disease Control and Prevention [CDC], 2012; Ludwig & Cook, 2000; Marzuk, Tardiff, & Sjirsch, 1992; Putkonen, Komulainen, Virkkunen, Eronen, & Lönnqvist, 2003; Putkonen, Weizmann-Henelius, Lindberg, Eronen, & Häkkänen, 2009; Vaughan, DeLisi, Beaver, & Howard, 2009). Other factors related to homicide perpetration include alcohol abuse, crime recidivism, owning a firearm or other weapon, drug use, and gang affiliation (Weibe, 2003; Federal Bureau of Investigation, 2002; Kellermann & Heron, 1999; Loeber et al., 2005; Roberts, Zgoba, & Shahidullah, 2007). It is well established that homicide perpetration cannot be predicted by a single risk factor but rather is a result of exposure to an accumulation of various negative influences (Weibe, 2003).
The distribution, determinants, and definition of homicide vary widely by region and country (Chon, 2007). The availability of national homicide data differs substantially from country to country based on the capacity to survey for, collect, and analyze homicide data. Difficulty finding reliable homicide data is particularly pronounced in much of the developing world, including Africa. Current estimates of homicide rates in developing countries are provided by the World Health Organization (United Nations Office on Drugs and Crimes, International Homicides Statistics, 2004) and the CDC (2014), among others, but such estimates are not particularly reliable and are based primarily on data gathered from urban areas. For most African countries, including the East African country of Rwanda, data on homicide are particularly scant and unreliable in part because the criminal justice systems that typically generate such data are often disorganized and/or underfunded, understaffed, and lack the knowledge, skills, and financial support to collect, analyze, and disseminate findings (Ziraba, Kyobutung, & Zulu, 2011). Moreover, few studies have explored risk factors for homicide perpetration in the region.
The recent history of Rwanda has been shaped by the 1994 genocide, in which an estimated 1.5 million Rwandans were incited by the Hutu government to murder approximately 1 million Tutsi and moderate Hutu (Human Rights Watch, 2011). In the aftermath of the genocide, a large number of Rwandans were charged with criminal offenses, and judicial overload and prison overcrowding were a pressing national issue. In the spirit of fostering national unity and reconciliation, the post-genocide Rwandan government reintroduced a traditional legal system, referred to as Gacaca, in which alternatives to lengthy incarceration were available for many confessed offenders. By 2005, approximately 12,000 community-based Gacaca juries had been established countrywide, and hundreds of thousands of confessed participants in the genocide were reintegrated into Rwandan society (Hansen, 2005; Kavuro, 2011). As a result, in present-day Rwanda, confessed killers and survivors live side-by-side in towns and villages throughout the country. The situation has raised societal concern that the current rate of homicide is related to lingering effects of the 1994 genocide. Until the present study, however, no data on demographic and other characteristics of homicide offenders in Rwanda have been available, as annual police reports only provide victim counts (Rwanda National Police, 2014). In this article, we investigated whether the type and degree of exposure to genocide is a predictor of homicide perpetration, using a population-based case–control design.
Method
Study Population
The selection of study subjects was conducted nationwide with the cooperation of the Rwanda National Police Department of Operations, which provided names and addresses of charged perpetrators associated with a homicide that occurred between May 2011 and May 2013. One hundred fifty (n = 150) adult homicide perpetrators of Rwandan nationality, aged 18 years and older who were arrested and charged with murder or manslaughter (both defined as homicide under the Rwandan criminal code) were eligible for inclusion in the study as cases. Exclusion criteria included alleged homicide perpetrators who did not confess to having committed the crime and suspected homicide perpetrators whose next of kin could not be traced.
Control subjects (n = 450), aged 18 years and older, were drawn from residents of a neighboring village to minimize information bias that may result from concerns by interviewed subjects about association with an ongoing criminal investigation. Controls were matched to cases by age (±5 years) and sex. The selection of three controls for each case was guided by the input of the village administrator who maintains records of all village dwellers.
Sources of Data
Information on case and control subjects was collected from study subjects’ next of kin in the presence of the village administrator who could corroborate the information, especially on sensitive questions regarding personal life behavior and genocide exposure. Information was collected within at least 1 week of the death of the victim.
Prior to data collection, informed verbal consent was sought and obtained from the surveyed family members of both cases and controls. A detailed validated questionnaire was administered to the next of kin (first-degree relatives, where possible) of both cases and controls, and a verbal interview in a face-to-face setting was then conducted. The questionnaire covered socio-demographic variables, including age, sex, marital status, number of surviving parents, having children, education level, employment status, religion, use of alcohol, use of drugs and homemade illicit brew, past criminal record, whether the subject had been charged with genocide-related crimes, and other information regarding genocide exposure. Several of these factors are established risk factors for homicide perpetration in other countries; however, others were included because we hypothesized that the factor may be associated with both homicide perpetration and genocide exposure and thus needed to be considered as potential confounders in the analysis. Most of these additional factors are stabilizers in the Rwandan culture and traditions, and among these, religion was of particular interest because it had been identified as a factor in neighboring Tanzania (Kibusi et al., 2013).
Genocide exposures were designated by proxy of reliable and easily measurable characteristics resulting from the 1994 genocide. Two primary genocide-exposure categories were a “survivor,” defined as a Rwandan who was living in Rwanda during the genocide period (April 6-July 4, 1994) in government-controlled territories and who was targeted for violence because of his or her ethnicity or political affiliation, or “perpetrator,” defined as having been convicted of genocide acts. Additional categories of genocide exposure included the loss of a first-degree relative during the genocide and having a first-degree relative convicted of genocide.
The information was collected by trained interviewers and recorded on a standard questionnaire form. The questions were tightly structured and closed ended. The principal investigator (W.R.) reviewed all questionnaires for verification, quality control, and data entry. All information was kept in a secure location, and personal or other identifiable information of the study subjects was removed prior to analysis. The study protocol was approved by the National University of Rwanda, School of Public Health’s institutional review board.
Sample Size and Statistical Analysis
The target sample size of the investigation included 150 cases of alleged homicide perpetrators and 450 living controls, the minimum sample to have 95% power to detect an odds ratio (OR) of 2.0, considering a control population genocide-exposure probability of 50% and a 5% alpha-error for a two-sided test. Descriptive statistics were used to characterize the distribution of alleged homicide perpetrator characteristics. The distribution of socio-demographic, criminological, and genocide-exposure characteristics of interest among cases and controls was examined using a cross tabulation. Next, potential socio-demographic and criminological predictors were identified via univariate conditional logistic regression analysis, in which variables associated with homicide perpetration (α = .1 significance) were retained. The identified predictors were further assessed by Mantel–Haenszel test of homogeneity analysis to identify interactions between predictors and hypothesized genocide-exposure characteristics, and no interaction was identified. Once potential predictor variables were identified, a multivariate conditional logistic regression analysis was performed. Separate models were fitted for each hypothesized exposure variable to genocide characteristics, with the exposure variable forced into the model, adjusting for other significant identified predictive factors as possible confounders. Stata v11.2 (StataCorp, College Station, TX, USA) was used for the analysis.
Results
Of the 150 homicide perpetrators investigated, 124 (82.7%) were male and 26 (17.3) were female. The mean age of homicide perpetrators at the time of the crime was 33 years (SD = 10 years), with 39.3% between the ages of 20 and 29 years. The vast majority of homicide perpetrators (n = 126, 84%) were residents of a rural area. Of the five regions of the country, 41 (27.3%) homicide perpetrators were from the Southern Province, 37 (24.7%) from the Northern Province, 34 (22.7%) from the Eastern Province, 21 (14.0%) from the Western Province, and 17 (11.3%) from the Central Province (Kigali capital region). Fourteen (9.3%) homicide perpetrators had been convicted of genocide, and 10 (6.6%) were survivors. The remaining 126 (84%) were neither a genocide survivor nor convicted of genocide acts. Eleven (7.3%) homicide perpetrators had lost a first-degree relative to genocide while 53 (35.5%) homicide perpetrators had at least one first-degree family member who had been convicted for genocide.
The bivariate analysis of socio-demographic and criminological variables among homicide perpetrators and controls shows significant covariates for homicide offending (Table 1). The following characteristics exhibited significant differences (p < .10) between homicide perpetrators and controls: marital status, belonging to a religion, employment status, alcohol consumption habits, use of drugs or homemade illicit beverages, and past criminal records, other than genocide. In the bivariate analysis, there was a significant association between homicide perpetration and having a first-degree relative convicted of genocide (p < .001; Table 2). A third of homicide perpetrators (35.3%) had a first-degree relative convicted of genocide compared with 17.1% of controls. On the contrary, there were no significant differences between proportions of the remaining genocide-exposure variables between homicide perpetrators and controls.
Univariate Conditional Logistic Regression Analysis of Demographic and Criminological Variables Among Homicide Perpetrators (n = 150) and Controls (n = 450).
Note. CI = confidence interval.
Conditional Logistic Regression Analysis of Hypothesized Genocide Characteristics for Homicide Perpetrators.
Note. Adjusted for number of marital status, belonging to a religion, employment status, alcohol-drinking pattern, drug user and/or banned homemade brew drinker, and past criminal record, other than genocide variables. CI = confidence interval.
The multivariable conditional logistic analysis demonstrated that the strongest genocide characteristic for homicide perpetrators was having a first-degree relative convicted of genocide in criminal or traditional Gacaca courts (adjusted OR [aOR] = 14.4, 95% confidence interval [CI] = [1.6, 129.4]; Table 2). None of the remaining genocide-exposure variables were significant. For being a genocide survivor, there was a non-significant increased odds in homicide perpetration for survivors after controlling for possible confounders (OR = 2.5, 95% CI = [0.5, 13.2]). Similarly, being a genocide perpetrator and having lost a first-degree family member to genocide were not risk factors for committing homicide (aOR = 1.1, 95% CI = [0.1, 14.7] and aOR = 1.1, 95% CI = [0.2, 6.6], respectively).
Discussion
To our knowledge, this is the first epidemiological investigation evaluating the effect of genocide exposure on homicide perpetration. In this study on the lingering effects of the 1994 genocide in Rwanda, we found that people who had a first-degree family relative convicted of genocide were more likely to be homicide offenders than other Rwandans with no such record. The exact effect and mechanism of this exposure should be studied more in the future, and readers should be cautious when interpreting the point estimate for this result due to the wide CIs. In contrast, being convicted of genocide was not associated with committing homicide. The vast majority of the 150 homicide offenders in this study were unlikely to have actively participated in the killings during genocide because they would have been minors in 1994. However, it is plausible that these individuals were exposed to different violence at a young age, particularly if a first-degree relative was convicted of genocide. Furthermore, these individuals could have had family instability or disruption as a result of incarceration of the first-degree relative following genocide conviction in Gacaca or criminal courts. These possible reasons for an increased association of homicide perpetration in individuals with a first-degree relative convicted of genocide are consistent with studies in youth that demonstrated increased violent behavior among children who witnessed family instability and violence, such as domestic violence, parental suicide, and imprisonment of a family member (Drury Satcy et al., 2014). Furthermore, the finding is consistent with the theory of intergenerational transmission of violence, which suggests that experiencing or witnessing an awful traumatic event is transmitted from adult to children in families and communities (Atkinson, Nelson, & Atkinson, 2010). Therefore, one could assume that in the case of Rwanda, violent behavior resulting from posttraumatic stress disorders could have been transmitted from parents to children following the 1994 genocide. This situation should be of great concern, and a recent study revealed that 26.1% of Rwandan people were presented with posttraumatic stress disorder symptoms (Munyandamutsa, Nkubamugisha, Gex-Fabry, & Eytan, 2012).
Neither being a genocide survivor nor losing a first-degree relative in the genocide was associated with increased risk of homicide perpetration. Although this has not been studied well within the context of genocides, numerous studies show that people exposed to mass trauma, such as wars, subsequently developed posttraumatic stress disorders that lead to aggressiveness and violent behavior in their families or neighborhoods (Elbogen et al., 2013; Winkle & Safer, 2011; Wolf et al., 2013). One reason that we may not have observed a similar effect in Rwanda is that in the aftermath of the 1994 genocide, the Government of Rwanda adopted numerous policies that were intended to mitigate the effects of post-genocide trauma among survivors and improve their socio-economic welfare. In that framework, various programs designed to alleviate the suffering of genocide survivors were implemented. These programs include establishment of the genocide survivors’ assistance funds and the creation of associations to support different survivor groups, such as widows and students. These support groups could help in fostering forgiveness and national reconciliation, and therefore decrease the risk of revenge crimes.
Other risk factors for homicide perpetration identified in this study were generally similar to other studies (Bijleveld & Smith, 2006; Delisi & Scherer, 2006; Dobash, Dobash, Cavanagh, & Lewis, 2004; Dobash, Dobash, Cavanagh, Smith, & Medina, 2007; Farrington, Loeber, & Berg, 2012). Dealing in illegal activities was identified as risk factors for homicide perpetration (DeLisi & Walters, 2011; Dolan & Smith, 2001; Heide, 2003) as was alcohol abuse, drug use, and/or drinking illicit homemade intoxicating beverages (Ziraba et al., 2011). In addition, in this study, the characteristic of having been previously involved in any other crime was strongly associated with a high risk of committing a homicide (Delisi & Scherer, 2006; Dolan & Smith, 2001; Farrington et al., 2012). Furthermore, we found that living in a polygamous relationship was associated with an increased risk of being a homicide perpetrator; however, this trait has not been identified in other African studies (Ziraba et al., 2011). Our investigation showed that belonging to a religion was protective against homicide perpetration, as has been demonstrated in a study on socio-cultural factors on homicide, conducted in Dar es Salaam (Kibusi et al., 2013). Because our cases and controls were matched on age and gender, we could not look at the effect of these factors on the risk of homicide perpetration. However, the majority of homicide perpetrators were young men, consistent with international scientific literature (Loeber et al., 2005; Roberts et al., 2007; Vaughan, DeLisi, Beaver, & Howard, 2009).
The strengths of this study were threefold. First, it was a 2-year population-based national-matched case–control study, where data were prospectively collected on incident cases of homicide. The study was mainly based on the close collaboration between police investigators, trained medical interviewers, family members, and local administration in accessing sensitive information on a range of characteristics that were considered to be potential risk factors for homicide perpetration. Finally, the sample size was large enough to detect differences, and study subjects were selected from all regions of the country, therefore making the sample representative of the study population.
However, while interpreting our results, some limitations should be considered. First, selection bias should be considered, as the selection of cases of homicide perpetrators was not based on criminal court conviction, but rather we relied on police investigations and subsequent charges of murder or manslaughter. To minimize the bias, we only included in our study cases of homicide perpetrators where police investigators had sound evidence to such an extent that they had confessed to the crime. Another limitation is that a 5-year interval used in matching controls to cases may be large enough that there may still be age effects within the matched pairs. However, in this study, the age distributions of cases and their corresponding controls were very close (32.6 years, SD = 10.1 years and 32.5 years, SD = 9.9 years, for cases and controls, respectively), and therefore we think this effect is minimal.
Second, information bias could have occurred because the information was obtained from next of kin and village leaders who may have limited information on some sensitive characteristics, such as alcohol consumption, drug use, and even genocide-related information. We could not directly interview the perpetrators immediately after the homicide event because it might have influenced their criminal case, and we did not believe they would give unbiased information. In piloting our study, we found that next of kin knew many of these details on the person they were responding about, and that each village administrator had the responsibility to know the socio-economic status and behavior of his constituents for security purposes. Therefore, we believe that the information bias was limited.
In conclusion, in addition to common risk factors for homicide perpetration that have been documented in other studies, we discovered that in post-genocide Rwanda, individuals whose first-degree relatives were convicted of participating in the 1994 genocide had a significantly higher risk of being a homicide perpetrator. Further studies are recommended to confirm this unique finding and to better understand the mechanism that drives this result. However, these individuals should be given special attention and support to address lingering effects of early exposure to violence as a critical part of strategies aimed at fighting homicide perpetration in Rwanda, 20 years after the genocide.
Footnotes
Acknowledgements
The authors extend their sincere thanks to the Departments of Operations, Criminal Investigations, and the Directorate of Medical Services of the Rwanda National Police for their valuable participation in, and contribution to, this study.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was sponsored by the World Health Organization. W. R. completed this article as part of his PhD, which was funded by the Centers for Disease Control and Prevention through financial support to the capacity building of the School of Public Health of the University of Rwanda. B. H. G. was supported by the Harvard Medical School Department of Global Health and Social Medicine Research Core.