Criminal Justice Outcomes in Intimate and Non-intimate Partner Homicide Cases


3.6 Control variables: Do other factors play a role?

Research has shown that intimate partner homicides are distinct in some ways from non-intimate partner homicides (Silverman & Kennedy, 1993). Therefore, if accused persons are treated differently based on the type of relationship they shared with their victims, it may be that such treatment is warranted. In other words, if lighter sanctions appear to be evident in cases of intimate partner homicide, it may not be the nature of the relationship itself that leads to different outcomes; rather it may be the characteristics of the homicide and/or the characteristics of those involved that justify different court outcomes. To capture this possibility, this section describes a number of legal and extra-legal factors that research has shown are associated with criminal justice outcomes in cases of homicide and other violent crime that, in turn, may also be associated with the type of victim-accused relationship. The variables are grouped into two categories: (1) legal variables; and (2) extra-legal variables, including characteristics of the accused, characteristics of the victim, and characteristics of the homicide incident. These variables are described in more detail below (see Table 3.3 for coding and descriptive information).

Legal variables

A number of legal variables are included in the analyses:[20] Prior criminal record of the accused, the role of the accused in the homicide, the number of accused involved in the homicide, the number of counts of homicide the accused was charged with (i.e. number of victims), the initial prosecution charge, the mode of conviction and, finally, the severity of conviction.[21]

Criminal record

The criminal history of the accused is captured using a three-category variable that distinguishes among those who had no prior record, those who had a non-violent record, and those who had a violent record. Drawing from previous research, it is expected that the existence of a prior criminal record will significantly affect the severity of the sentence (see Blumstein et al., 1983; Hagan & Bumiller, 1983; Klepper et al., 1983; Kruttschnitt, 1982). In addition, it may also have some effect on early decisions such as initial charge and mode of conviction. In the Toronto sample, 43 percent of the accused had a prior record for non-violent offences whereas 13 percent had a prior record for violent offences.

Number and role of those involved

The next set of legal variables pertains to the role of the accused and the number of accused persons and victims involved in the homicide. First, research has shown that the role of the accused may mitigate perceived blameworthiness if it was determined that he/she was a follower in the incident rather than a leader or organizer (Steffensmeier et al., 1998). And, even though an individual who aids and abets a homicide is legally indistinguishable from the primary accused, their relative degree of involvement in the offence may be important in determining the length of sentence (Grant et al., 1998). For example, judges may impose a shorter period of parole ineligibility if the role of the accused was not that of the leader. The role of the accused may also be relevant at earlier decision points in the system because prosecutors, when charging and negotiating pleas, as well as judges and juries at trial, may perceive an accused who played a minimal role, or at least a less crucial role in the killing, to be less culpable for the crime. To capture this, a dichotomous measure distinguishes between those who were the primary and those who were the secondary offender(s). The majority of the accused were primary offenders (85 percent) because, as documented next, the majority of homicides involved single offenders and single victims.

The next two legal variables capture the number of accused persons and the number of victims in each homicide. Research has shown that when more than one accused is involved in a homicide and/or more than one victim is killed, more severe criminal justice responses result (Black, 1976; Huang et al., 1996; Myers, 1980). For example, some research has shown that the seriousness of the prosecution charge increases and the probability of a trial increases when there are multiple victims (Myers, 1980). As such, while a sentence for first-degree murder cannot exceed 25 years, where sentencing discretion is possible (e.g. in second-degree murder cases), judges may increase the parole ineligibility period because the case involved multiple victims. Dichotomous variables indicate whether there were multiple accused persons and/or multiple victims in each case. In this sample, 16 percent of the homicides involved more than one accused and four percent of the cases involved more than one victim.

Outcomes at earlier stages

The remaining three legal variables are dependent variables that become controls at later stages in the court process in response to research that has shown that the effect of early decisions on later outcomes needs to be considered to adequately understand the criminal justice process. For example, while research is contradictory (Brerton & Casper, 1981-82; LaFree, 1985; Nardulli et al., 1988), there is a common belief that guilty pleas or plea bargains often result in charge reductions and/or lighter sentences (Mather, 1979; Nardulli, 1979; Neubauer, 1974; Newman, 1966; Uhlman & Walker, 1979; Vetri, 1964). One reason for this belief is that guilty pleas are often entered for the sole purpose of obtaining a charge reduction. Moreover, it is commonly assumed that offenders who accommodate the system by pleading guilty and saving the expense of a trial are often rewarded with lighter sentences (Dawson, 1969; Rosett & Cressey, 1976; Ruby, 1999). However, it has also been argued that the initial rewards of a guilty plea at one stage of the process, such as a reduced charge, may cancel out any further rewards at later stages, such as sentencing (Eisenstein & Jacob, 1977; Smith, 1986). Similarly, increases in the severity of the charge and the conviction may have an impact on later outcomes. For example, the severity of the initial charge may increase the likelihood that a case will go to trial, that the accused will be found guilty at trial, or that he/she will be convicted of murder. Thus, where applicable, control variables are used to capture decision outcomes at early stages when predicting outcomes at later stages. Mode of conviction is included as a dichotomous measure that distinguishes between cases that went to trial and those that did not while severity of initial prosecution charge and severity of conviction are included as interval-level variables.

Characteristics of the accused

Research on punishment disparity has also documented the effect of a number of extra-legal factors on criminal justice outcomes. In addition to the gender of the accused discussed above, race and age are also important social statuses by which Western society is stratified and differentiated (Steffensmeier et al., 1998). An abundance of studies have examined the independent effects of these two variables on various court outcomes. To date, findings have been inconsistent in research examining the effects of race on sentencing (see reviews, Kleck, 1985; Kramer & Steffensmeier, 1993). Some studies show that blacks receive more severe sanctions than whites (Lizotte, 1978; Petersilia, 1983; Spohn, 1990; Spohn et al., 1981-82); others show that blacks receive more lenient sentences than whites (Bernstein et al., 1977); and still others find few race differences (Klein et al, 1988; Wilbanks, 1987) or mixed results (Dixon, 1995; Kramer & Steffensmeier, 1993). Research has also found that the presence and size of race effects vary across courts with different contextual characteristics (Myers & Talarico, 1987; Tonry, 1995). In this analysis, a dummy variable captures whether the accused is white or non-white. While such a measure is common in criminal justice research on race, using a dichotomy to measure the race/ethnicity of an accused loses much information because a wide variety of race and ethnic groups are included in each of these broad categories. However, due to data limitations, more refined categories were not possible. In the Toronto sample, the majority of accused were white (56 percent).

Findings on the age-sentencing relationship are sparse and recent research reveals that the association is more complex than traditionally believed. On the one hand, most analyses of sentence outcomes control for age as a continuous variable, assuming a linear effect; these analyses typically report a small or negligible age effect (e.g. Klein et al., 1988; Myers & Talarico, 1987; Peterson & Hagan, 1984). On the other hand, several studies find – when the data are partitioned into ‘older’ versus ‘younger’ offenders – that older offenders (e.g. 50 and over) are treated more leniently than younger offenders (e.g. offenders in their 20s; see Champion, 1987; Cutsall & Adams, 1983; Wilbanks & Kim, 1984). Notably, those accused who are in their 60s and 70s appear to benefit the most from the overall greater leniency extended to older persons (see Steffensmeier & Motivans, 2000). Providing some clarification of the role of age in sentencing, Steffensmeier et al. (1995) found a non-linear or inverted U-shaped relationship when the full range of adult ages was included, from late teens to young adulthood through middle and old age. This curvilinear pattern was largely due to the more lenient sentencing of youthful (aged 18-20) compared to young adult offenders (aged 21-29). Younger offenders, aged 18-20, received sentences on par with offenders in their 30s whereas offenders in their 50s and older received the most lenient sentences. The age-sentencing relationship, then, becomes strictly linear from about age 30 into old age. In the present study, the age of the accused is measured in two ways. First, for descriptive and bivariate analyses, a five-category age variable is used that categorizes accused persons into the following age groups: 18-24, 25-34, 35-44, 45-54, and 55 and older. These age groups are consistent with official statistics documenting aggregate patterns in adult criminal court outcomes (CCJS, 2003a). In the multivariate analyses, a continuous variable that captures the reported age of the accused is used. Table 3 shows the distribution of these age groups in the Toronto sample as well as the average age of the accused (31 years).

Employment status and marital status may also be considered social status variables because, similar to the victim-accused relationship, they may be seen as indicators of ‘social morphology’ or the degree to which individuals participate in social life (Black, 1976). For instance, employment status may be relevant to court outcomes because holding a job represents, not only economic power (Turk, 1969), but also social integration (Black, 1976; Landes, 1974). Thus, accused persons who are employed may be treated more leniently at some stages of the criminal process than those who are unemployed (Boris, 1979; Reskin & Visher, 1986). A dummy variable that indicates whether an accused is employed or unemployed shows that, in the Toronto data, 30 percent of the accused were employed.[22] Marital status also represents a form of social integration (Black, 1976; Myers, 1980) and, to capture this, a variable is included that distinguishes accused persons who were married or not married when the homicide occurred.[23] Just over 40 percent of all accused persons in the Toronto sample were married.

Finally, whether or not an accused had a history of psychiatric treatment may be relevant to his or her treatment within the criminal justice process. Thus, a variable captures whether an accused had received inpatient or outpatient treatment for a psychiatric disorder(s). In the Toronto data, where information was available, 10 percent of the accused had received some type of psychiatric treatment.

Characteristics of the victim

Previous research on criminal justice decision-making has often considered only the characteristics of the accused as possible determinants of court outcomes. However, the social structure of a case depends on the identity of both the victim and the accused (Baumgartner, 1999). Thus, considering the characteristics of only one of the parties involved may produce misleading results. Paralleling the variables that capture the characteristics of the accused, measures are included for the victim’s race/ethnicity, age, employment status, marital status as well as prior psychiatric and criminal history (again, see Table 3.3 for coding and descriptive information).

Various social status characteristics of the victim have been shown to be important determinants of criminal justice outcomes in cases of violent crime (Horwitz, 1990; Williams, 1976). First, a dummy variable captures whether the victim was white (coded 0) or non-white (coded 1). In the Toronto sample, 58 percent of the victims were white. The age of the victim has also been found to be relevant at some stages of the criminal justice process, particularly at the initial charging stage (see Williams, 1976). Thus, measures for victim age parallel those included for the accused. The average age of the victim in the Toronto sample was 35 years. The victim’s employment status may also be relevant as research has demonstrated that the criminal justice system’s response to men who kill employed women is different from the response to offenders who kill unemployed women (Crawford and Gartner, 1992). For example, killers of employed women were more likely to be charged with and convicted of first-degree murder and, thus, received longer sentences. Overall, cases that involved employed victims were found to be prosecuted to a significantly greater degree than if victims were unemployed (Boris, 1979). To capture this, an employment status variable similar to that included for the accused is incorporated, showing that 40 percent of the victims in this sample were employed. The marital status of the victim may also be relevant to criminal justice decision-making (Horwitz, 1990). Victim’s marital status is measured the same as the accused person’s marital status, demonstrating that 45 percent of the victims were married. Finally, two variables are included to capture the victim’s psychiatric and criminal history. In this sample, six percent of the victims had received some type of psychiatric history and 30 percent had a prior record.

Characteristics of the incident

Several characteristics of a homicide incident have also been shown to affect criminal justice decision-making. For example, weapon use has often been used as a measure of offence seriousness and has been found to be associated with more severe criminal justice treatment (Hagan et al., 1980; LaFree, 1980; Lizotte, 1978). When that weapon is a gun, some U.S.-based research has demonstrated that more severe criminal sentences result (Cook & Nagin, 1979; Loftin et al., 1983; Wright et al., 1983). In contrast, other research has shown that firearm homicides are treated more leniently, possibly because guns kill more quickly and efficiently than other methods (Givelber, 1994). While there is no comparable research in Canada, anecdotal evidence suggests that judges do not appear to label gun killings as particularly brutal nor is there any evidence that they impose more severe criminal sanctions in these cases (Grant et al., 1998). To control for the potential effects of gun use, however, a variable distinguishes between gun killings and those that involved some other method (e.g. stabbing or beating). Guns were used in close to one-quarter (24 percent) of the Toronto homicides.

With respect to the location of a homicide, research suggests that the public nature of a crime may be seen as a threat to the maintenance of social order whereas crimes that occur in private are less likely to be perceived as such (Lundsgaarde, 1977). In Canada, according to the most recent figures, almost two-thirds (63 percent) of the 544 homicide incidents that occurred in 2002 took place in a private residence (CCJS, 2003a). More specifically, the majority of spousal and other familial homicides (93 percent) occurred in private locations (CCJS, 2003a). Consistent with national figures, about 64 percent of the Toronto homicides took place in private whereas approximately 36 percent took place in a public or semi-public location. A dichotomous measure distinguishes between private and public killings.

Finally, research has demonstrated that alcohol and/or drug use by the victim may affect court outcomes because of the way it may influence perceptions of some victims as blameworthy or potentially responsible for their own victimization (see Williams, 1976). In contrast, alcohol or drug use by the accused may reduce his or her culpability to some degree because intoxication may act to diminish or reduce the perceived intent of the accused (Grant et al. 1998). To capture the potential effects of substance use on court outcomes, separate measures for accused and victim alcohol and/or drug use at the time of the homicide is included, distinguishing between those who had been using alcohol and/or drugs and those who had not. Accused persons had been using substances in 55 percent of the Toronto homicides whereas victims had been drinking or using drugs in 45 percent of the incidents.[24]

Date modified: