Social science research suggests that capital punishment is administered in an arbitrary manner. The term arbitrary has two meanings: death sentences are imposed randomly, or death sentences are influenced by legally irrelevant factors (Reference Baldus, Woodworth and PulaskiBaldus, Woodworth, & Pulaski 1990:14–15). The latter interpretation can be illustrated through an anecdote that Reference BrightStephen Bright, the president of the Southern Center for Human Rights, recounted: “A member of the Georgia Board of Pardons and Paroles has said that if the files of 100 cases punished by death and 100 punished by life were shuffled, it would be impossible to sort them out by sentence based upon information in the files about the crime and the offender” (1994:1840). Put differently, knowing the legal facts of a capital case is not sufficient to predict whether a defendant lives or dies—social facts are also pivotal. Existing research demonstrates that the death penalty is more likely to be imposed on behalf of white victims and female victims (see, e.g., United States General Accounting Office 1990; Reference Baldus, Woodworth and PulaskiBaldus, Woodworth, & Pulaski 1990; Reference Baldus, Woodworth and AckerBaldus & Woodworth 2003; Reference PaternosterPaternoster et al. 2008; Reference Paternoster and BramePaternoster & Brame 2008; Reference HolcombHolcombe et al. 2004).
But scholars have largely ignored the other member of the sociological trinity: social status. Radelet and Baldus appear to be the only death penalty researchers who have investigated the role of victim social status. Reference RadeletRadelet (1989) examined the small number of cases in American history in which a white defendant was executed for killing a black victim. Close examination of the cases revealed that the seemingly anomalous executions could be explained by the presence of a low-status defendant and/or a high-status victim. Reference Baldus, Woodworth and PulaskiBaldus and colleagues (1990, Reference Baldus, Woodworth and Zuckerman1998, Reference Baldus, Woodworth and Grosso2002) have used quantitative methods to examine victim social status. Developing a measure based on occupational categories, Reference Baldus, Woodworth and PulaskiBaldus, Woodworth, and Pulaski (1990) report that death was more likely to be imposed on behalf of high-status victims in Georgia during the 1970s. More recently, Reference Baldus, Woodworth and ZuckermanBaldus, Woodworth, Zuckerman, et al. (1998) found that victim social status influenced both prosecutors and jurors in Philadelphia from 1983 to 1993, as death was less likely to be imposed on behalf of low-status victims, defined by low-skill occupation, unemployment, criminal behavior, gang membership, public housing, dilapidated nonpublic housing, dropping out of high school, and mental retardation. Finally, Reference Baldus, Woodworth and GrossoBaldus, Woodworth, Grosso, et al. (2002) report that the victim's occupation had a substantial influence on death sentencing in Nebraska from 1973 to 1999: Each unit increase in the victim's occupational prestige—from low to medium to high—elevated the chance of a death sentence as much as each additional statutory aggravating circumstance.
The current research contributes to the emerging field by incorporating a theoretical concept of victim social status and examining whether the influence of victim social status holds across different studies with different measurement strategies. In his 1976 classic The Behavior of Law, Black advanced a multidimensional vision of social status—one that includes, but transcends, wealth. Specifically, Reference BlackBlack (1976) argues that social status comprises five elements: vertical status (wealth), radial status (integration in the social life of a community), cultural status (sophistication and conventionality), normative status (respectability), and organizational status (the capacity for collective action). Black's approach harnesses the explanatory power of major sociological traditions: Vertical status draws on the ideas of Marx, radial status on Durkheim, cultural status on Parsons, normative status on Goffman, and organizational status on Weber (see Reference BlackBlack 1976, Reference Black1995). Developed in a book designed to explain legal variation, Black's comprehensive model of social status requires no translation to be applied to capital punishment.
Using multiple indicators to reflect Black's concept, I examine the influence of victim social status on the district attorney's (DA's) decision to seek the death penalty and the jury's decision to impose a death sentence. The data include the population of adult defendants indicted for capital murder in Harris County, Texas, from 1992 to 1999 (n=504). Harris County—home to Houston and surrounding areas—is arguably the capital of capital punishment. With 108 executions in the modern era, defined as the Supreme Court's reinstatement of capital punishment in 1976 to the present, Harris County has often captured the national and international spotlight in the death penalty debate (see, e.g., Amnesty International 2007). To put the number of executions in perspective, if Harris County were a state it would rank second in executions after Texas.Footnote 1 Indeed, Harris County has executed about the same number of offenders as all the other major urban counties in Texas, combined.Footnote 2 The period from 1992 to 1999 is also critical because the number of death sentences in Harris County climbed to historic highs.Footnote 3
The findings indicate that victim social status has a robust influence on capital punishment. The predicted probability of a death sentence is about six times greater if a victim is integrated, sophisticated, conventional, and respectable (as compared to a victim who is marginal, unsophisticated, unconventional, and disrespectable). Such patterns demonstrate that death penalty scholars have given insufficient attention to a central feature of social life: social status.
To be clear, the current research is not a test of Reference BlackBlack's (1976) theory of law. Such a test would require parallel data on the social status of both the defendant and victim. Indeed, the article illustrates how researchers can use Reference BlackBlack's concept of social status to understand sentencing—both capital and noncapital—without necessarily embracing his theory of law (1976) or Reference Blackparadigm of pure sociology (1995, Reference Black2000).
Black's Concept of Social Status
Social status refers to a person's position in society relative to others (Reference BlackBlack 1993:161). According to Reference BlackBlack (1976), social status comprises five elements: vertical status, radial status, cultural status, normative status, and organizational status. Figure 1 illustrates Black's concept of social status, including the elements of social status, the definition of each element, and the empirical indicators and data used to measure each element (empirical indicators and data are discussed in the methods section).
Figure 1. Black's (1976) Concept of Social Status
Vertical status refers to wealth, or the “… material conditions of existence” (Reference BlackBlack 1976:11). In premodern societies, wealth might be gauged according to livestock, food, or perhaps slaves. In modern societies, wealth is a question of finances, such as personal income, inheritances, stock portfolios, home values, and land holdings. Economic prosperity is an integral component of Black's concept of social status, but the concept also includes noneconomic attributes.
Radial status refers to integration in social life (Reference BlackBlack 1976:48–54). The term radial can be seen as a series of concentric circles: Some people are central to the social life of a community, while others remain on the periphery. The most integrated members of a community get married, have children, participate in local politics, volunteer for local charities, and attend religious services. The least integrated members of a community are isolates—uninvolved, unknown, and socially invisible.
Cultural status includes related but distinct attributes: whether a person is sophisticated and conventional (Reference BlackBlack 1976:61–73). Being sophisticated is a question of refinement. Those who have a college degree are more sophisticated than those who have a high school degree. Those who are versed in literature, politics, and the arts are more sophisticated than those who are not. Those who are aficionados of style—from clothing to architecture to food—are more sophisticated than those who are not. Conventionality is defined by the frequency of different forms of culture. Being white in the United States is more conventional than being black because whites are more numerous. Similarly, being Catholic is more conventional than being Baptist because Catholics are more numerous. But being Baptist is more conventional than being Wiccan. Conventionality is not a value judgment but rather a matter of counting; frequency determines conventionality.
Normative status refers to respectability (Reference BlackBlack 1976:105, 111–17). Respectability depends on the extent to which a person has engaged in deviant behavior that evoked social control. The forms of social control that can damage a person's reputation are endless, ranging from minor to severe, such as being admonished for breaking a personal confidence; being suspended from school for cheating; being fired from the workplace due to inappropriate behavior; being abandoned in a relationship due to sexual infidelities; being accused, convicted, and incarcerated for a crime. Some people have an unassailable reputation, some a questionable reputation, and some a spoiled reputation.
Organizational status refers to “… the capacity for collective action” (Reference BlackBlack 1976:85). Some organizations allow members to marshal collective resources in the hour of need, such as religious groups, fraternal orders, families, nations, and international alliances. Those who are members of such groups have a ready network of supporters and allies, but those who are not must fend for themselves.
The five elements define a person's social status (Reference BlackBlack 1993:161).Footnote 4 The highest status is conferred upon those who are wealthy, integrated, sophisticated, conventional, respectable, and organized. The lowest status is conferred upon those who are poor, marginal, unsophisticated, unconventional, disrespectable, and unorganized. Most people fall between such extremes due to a combination of low, moderate, and high status on different elements. Regardless, all people can be placed on the continuum of social status—a person's social position relative to others.
The current research examines the impact of victim social status on two pivotal moments in a capital case: the DA's decision to seek death, and the jury's decision to impose death. The DA's decision is fateful because the die is essentially cast: During the time period under consideration, John Holmes, the long-time Houston DA, pursued death in just a fraction of the eligible cases, but secured death in most attempts. The jury's decision is fateful because it represents the final disposition, and the disposition of death is carried out considerably more often in Texas than other states (Reference BlumeBlume et al. 2004).
The predicted impact of victim social status on the DA's decision to seek death and the jury's decision to impose death can be summarized as follows:
Hypothesis 1: Death is more likely to be sought and imposed on behalf of high-status victims.
Hypothesis 2a: Death is more likely to be sought and imposed on behalf of wealthy victims.
Hypothesis 2b: Death is more likely to be sought and imposed on behalf of integrated victims.
pothesis 2c: Death is more likely to be sought and imposed on behalf of sophisticated victims.
Hypothesis 2d: Death is more likely to be sought and imposed on behalf of conventional victims.
Hypothesis 2e: Death is more likely to be sought and imposed on behalf of respectable victims.
Hypothesis 2f: Death is more likely to be sought and imposed on behalf of organized victims.
Research Methods
Cases and Outcomes
The data included the population of 504 adult defendants indicted for capital murder in Harris County from 1992 to 1999 (the defendants murdered a total of 614 victims).Footnote 5 The Harris County district clerk (HCDC) used the Harris County Justice Information Management System (JIMS) to identify the defendants. The HCDC also provided a JIMS file that contained public information about each case, including whether the case resulted in a plea bargain or trial and the disposition. The Harris County DA's office provided archival documents that were used to verify the list of defendants and determine if the DA sought death.
Figure 2 tracks the cases. The figure reveals that the DA sought death against 129 of the 504 eligible defendants. The 129 defendants in question were adjudicated as follows: 98 were sentenced to death, 29 were sentenced to life imprisonment, one was sentenced to the Texas Department of Corrections (TDC) for some period of time less than life, and one was acquitted.Footnote 6 Of the 98 condemned defendants, 37 have been executed to date and 47 are awaiting execution on death row (as of July 2, 2009). The other 14 condemned defendants will not be executed: 10 were commuted to life imprisonment due to the Supreme Court's 2005 decision in Roper v. Simmons barring the execution of juveniles, and four died of natural causes on death row. The figure also reveals that the DA sought a life sentence against 218 defendants and reached a plea bargain with 157 defendants.
Figure 2. Adult Defendants Indicted for Capital Murder in Harris County, Texas from 1992 to 1999 (status current as of July 2, 2009)
Victim Social Status
Vertical status, or wealth, was measured according to median household income (MHI) in the victim's residential neighborhood. Because the Texas Department of Health's Vital Statistics Mortality File (VSMF) contained census block group numbers for both 1990 and 2000, linear interpolation was used to estimate MHI. If, for example, MHI in the victim's block group increased from $30,000 in 1990 to $40,000 in 2000, then MHI for a 1995 case was estimated to be $35,000. MHI was divided by 10,000 to ensure a meaningful unit of measurement (the transformation means that a one-unit change in MHI reflected a $10,000 change rather than a miniscule $1 change).
Radial status, or integration in social life, was measured according to whether the victim was married or widowed based on data from the VSMF (1=married, widowed; 0=separated, divorced, single). To avoid a cumbersome term, victims who were married or widowed are referred to as married throughout the remainder of the article.
Cultural status was subdivided into sophistication and conventionality.
• Sophistication was measured according to whether the victim had a college degree based on data from the VSMF (1=college degree; 0=not).
• Conventionality was measured according to whether the victim was white or Hispanic, as compared to black or Asian, based on data from the VSMF (1=white/Hispanic; 0=black/Asian). The U.S. Census Bureau's 2006 profile of Harris County reveals the following distribution: 38% Hispanic, 37% white, 19% black, and 5% Asian (http://www.census.gov/). In Houston, whites and Hispanics are more conventional, meaning numerous. The measure of conventionality raises an important question: Is it legitimate to combine whites/Hispanics and blacks/Asians? Prior research using the same data, but with separate indicators for each group, reveals that death is more likely to be sought and imposed on behalf of white and Hispanic victims than black and Asian victims—a pattern that justifies the unusual race/ethnic pairings (Reference PhillipsPhillips 2008). The pairings provided a sound indicator of conventionality that did not obscure intrapair differences.
Normative status, or respectability, was measured according to whether the victim had a clean criminal record (1=clean criminal record; 0=prior conviction). Data were drawn from the Web site http://www.publicdata.com. The site charges users a fee to access public criminal record data compiled from 43 states, including Texas. Searches were conducted on all victims (disaggregating misdemeanors and felonies was not possible because the relevant data fields were often missing on the public data Web site; disaggregating violent and nonviolent offenses was not possible because the number of violent offenses was too small for a separate indicator).
Organizational status could not be measured due to the absence of relevant data. The potential impact of the victim's organizational status is considered further in the conclusion.
A note on race and gender: Reference BlackBlack argues that race, gender, and social status are distinct concepts. Nonetheless, he also argues that race and gender often correspond to social status and, if so, can be used as blunt proxies of social status (1989:108, Reference Blacknote 48; 1993:162). In the current article, victim race was used to measure conventionality, and victim gender was treated as a control (Black has used race to illustrate the relationship between conventionality and law in his own work; see 1976:69–70).
Controls
To control for potential confounders, the multivariate models also examined other social characteristics of the defendant/victim and the legal dimensions of the case. Data for the controls were collected and merged from the following archival sources: JIMS, VSMF, grand jury indictments (GJI), the Harris County medical examiner (HCME), The Houston Chronicle newspaper (HCN), and the Web site http://www.publicdata.com. Table 1 reports measurement strategies, means, and data sources for the controls.
Table 1. Measurement Strategies, Data Sources, and Means for Victim Social Status and Controls
Notes:
1. Abbreviations: GJI=Grand Jury Indictment; HC=The Houston Chronicle; HCME=Harris County Medical Examiner; JIMS=Justice Information Management System; MHI=Median Household Income; PD=http://www.publicdata.com; VSMF=Texas Vital Statistics Mortality File.
2. Means are calculated based on the 446 cases included in the multivariate models.
3. In the rare cases in which the age, race, or sex of the victim was missing in the VSMF, the data were drawn from the HCME.
In terms of the defendant's characteristics, the models controlled for race (dichotomous indicators for white, black, Hispanic, Asian),Footnote 7,Footnote 8 age (dichotomous indicators for teens ages 17 to 19, young adults ages 20 to 29, and adults ages 30 or more), sex (1=male), form of legal counsel (1=appointed counsel for the entire case, 0=hired counsel for some or all of the case), prior violent conviction (1=yes), and prior nonviolent conviction (1=yes).
Controlling for the defendant's prior criminal record was particularly important because Texas juries must conclude that the defendant is a future danger to impose a death sentence. To ensure that the measure was sound, JIMS criminal record data from Harris County were supplemented with information from http://www.publicdata.com. Searches were conducted on all defendants. The inquiries revealed that among defendants who had a clean record in JIMS, 13 had a prior violent conviction and 32 had a prior nonviolent conviction on the public data Web site. Data from JIMS and the public data Web site were merged to measure prior violent and prior nonviolent conviction.
Most of the victim's characteristics were included in the measure of social status, but the models also controlled for sex (1=female) and whether the victim was physically vulnerable due to age (1=6 to 16 or over 60, children less than 6 considered separately as a statutory form of capital murder).
In terms of the legal dimensions of the case, the models controlled for whether multiple defendants were indicted (1=yes), the method of murder (dichotomous indicators for shot, stabbed, beaten, and asphyxiated), the heinousness of the murder (discussed below), and the statutory type of capital murder (discussed below). Controls for heinousness and the statutory type of capital murder require elaboration.
To measure the heinousness of the crime, newspaper articles about each case were collected from The Houston Chronicle's online archive (an average of 6.75 articles per case, for a total of more than 3,400 articles). A research assistant coded the aggravating and mitigating circumstances in each case based on a list drawn from Baldus and colleagues' landmark research on race and capital punishment (Reference Baldus, Woodworth and PulaskiBaldus, Woodworth, & Pulaski 1990:526–35). Table 2 lists the aggravating and mitigating circumstances in question.
Table 2. Aggravating and Mitigating Circumstances Used to Construct Measure of Heinousness
Note: The list of aggravating and mitigating circumstances was derived from Baldus, Woodworth, and Pulaski (1990:526–35).
The following formula was used to construct a scale of heinousness: number of aggravating circumstances minus number of mitigating circumstances (the scale ranged from −3 to +7). The original scale was transformed into three dichotomous indicators: Level 1 heinousness (bottom quartile of scores ranging from −3 to 0), Level 2 heinousness (middle 50 percent of scores ranging from 1 to 2), and Level 3 heinousness (top quartile of scores ranging from 3 to 7).
The heinousness measure included missing data because The Houston Chronicle did not report on 28 cases. To address the problem, missing cases were assumed to be Level 1. This assumption is based on compelling patterns. To begin, the cliché“if it bleeds it leads” encapsulates the media's obsession with sensational crimes. Considering the fact that The Houston Chronicle reported on 476 of the 504 cases, the 28 capital murders that did not attract media attention were almost sure to be the least heinous of all. The DA did not seek death against any of the 28 defendants in question, bolstering the assumption of minimal heinousness. Because the substantive results were the same regardless of whether the missing cases were excluded or coded as Level 1, the models presented in the results section used the revised indicator of heinousness to ensure complete data. Thus, the original scale was transformed into three dichotomous indicators to facilitate a solution to the missing data problem (and because some values on the original scale had no cases or just one case).
Using newspaper articles to code heinousness is not ideal. But all other avenues were closed (the DA denied access to capital murder memorandums as a confidential work product, police reports contained limited information or had substantial amounts of information redacted, and case files/transcripts were not an option because of the number of cases disposed through plea bargains). Nonetheless, focusing on newspaper articles is a reasonable approach. To begin, the measure was reliable. In a reliability test conducted by the author, the aggravating and mitigating circumstances were coded the same in 25 of 30 randomly selected cases, and all 30 cases were categorized as the same level of heinousness. The measure also had face validity: 12 percent of cases coded as Level 1 heinousness received a death sentence, compared to 17 percent of cases coded as Level 2 heinousness and 35 percent of cases coded as Level 3 heinousness. The aggravating and mitigating circumstances listed in Table 2 are also arguably the sort of facts a newspaper would tend to report. Most important, the current research improves upon the vast majority of capital punishment studies, which rely on supplemental homicide report data and therefore do not include any measure of heinousness.
The research assistant also coded heinousness based on a visceral reaction to the facts of the crime. Each case was assigned to Level 1 (relative minimal), Level 2 (intermediate), or Level 3 (extreme). The aggravating/mitigating and visceral measures of heinousness produced the same substantive results. The aggravating/mitigating measure was used in the multivariate models because it was more reliable (a reliability test of the visceral measure conducted by the author revealed that only 33 of 50 randomly selected cases were coded as the same level of heinousness) and produced slightly more conservative estimates of the effect of social status on capital punishment.
Grand jury indictments were used to code the statutory types of capital murder. Of the types of capital murder delineated in the Texas statute, the following appeared in the data: robbery, burglary, kidnapping, rape, remuneration, multiple victims, child 0 to 5 years old, police officer, arson, and obstruction/retaliation. The type of capital murder was measured through dichotomous indicators coded 1=yes and 0=no (other included police officer, arson, and obstruction/retaliation). Because a case can be a capital murder for multiple reasons, the indicators are not mutually exclusive.
Coding Cases With Multiple Victims
The coding procedures accommodated cases with multiple victims. If the data included information for all the victims, then the case could be coded. Specifically, cases were coded according to the victim with the highest value on the variable in question (if, for example, a case included two victims with MHIs of $40,000 and $50,000, then MHI for the case was coded as $50,000; or if a case included a male victim and a female victim, then the case was coded as female victim). However, if the data did not include information for all the victims, then the entire case was considered missing.
Modeling
Logistic regression was used to estimate the impact of victim social status on the odds of the DA seeking death (1=seek death; 0=else) and the jury imposing death (1=death sentence; 0=else).Footnote 9
Recall that the data included a population of cases, not a random sample. The question of whether statistical significance should be applied to population data remains contested (see, e.g., Reference CowgerCowger 1984, Reference Cowger1985; Reference BerkBerk et al. 1995a, Reference Berk1995b; Reference BollenBollen 1995; Reference FirebaughFirebaugh 1995; Reference RubinRubin 1995). The answer depends, in part, on the definition of a population. One interpretation is that an “apparent population” includes the census of relevant events. Therefore, an apparent population should be treated as a true population, meaning that tests of statistical significance are irrelevant. But another interpretation is that an apparent population is just one “realization” of all the populations that could have occurred if historical events were replicated numerous times. Therefore, an apparent population should be treated as a random sample, meaning that tests of statistical significance are relevant (Reference BerkBerk et al. 1995a).
The current research follows Reference BollenBollen's (1995) interpretation that: (1) ignoring statistical significance in population data is legitimate and appropriate if a researcher is attempting to describe the population rather than draw inferences; and (2) researchers should focus more on substantive significance and less on statistical significance. Thus, the central purpose of the research is to describe the magnitude of the relationship between victim social status and capital punishment for the population of cases.
Nonetheless, the use of statistical significance in arguably inappropriate contexts—including convenience samples and population data—is so common that ignoring significance has become unacceptable to most readers. Capitulating to convention, I report significance levels. But I maintain that all the population parameters are meaningful, regardless of significance.
Strengths and Limitations
The research has important strengths, including focusing on the capital of capital punishment, incorporating a theoretical concept of victim social status, including multiple indicators of victim social status, controlling for the heinousness of the crime, and examining multiple stages of the capital punishment process (the DA's decision to seek death and the jury's decision to impose death).
But the findings must also be interpreted through the lens of important limitations, including (1) the inability to examine the charging and indictment decisions, (2) the reliance on reasonable but imperfect measures of victim social status, and (3) the inability to control for strength of evidence. Each limitation is considered in turn.
The path from the commission of a murder to the pronouncement of a death sentence in Harris County includes four decisions: the intake prosecutor's decision to charge a defendant with capital murder, the grand jury's decision to indict a defendant for capital murder, the DA's decision to seek the death penalty, and the jury's decision to impose a death sentence. Despite repeated attempts, collecting the data needed to examine the charging and indictment decisions proved impossible. But the charging decision does not appear to exhibit much variation. To begin, the Texas capital murder statute delineates narrow categories of murder that are death-eligible. The precision of the statute simplifies the charging decision, as opposed to states that define heinous murders as death-eligible (see Texas Penal Code, title 5, chapter 19, section 19.03). Moreover, The Houston Chronicle reports in a February 2001 special series that intake prosecutors have “standing orders” to file capital murder charges in all possible cases (Reference Tolson and BrewerTolson & Brewer 2001:A1). The indictment decision exhibits almost no variation, as data from the HCDC indicate that grand juries returned a “no bill” in just seven capital cases from 1992 to 1999. Because the charging decision does not appear to exhibit much variation, and the indictment decision is a formality, this particular limitation was real but not fatal.
In addition, the indicators of victim social status are imperfect. Using median household income in the victim's neighborhood to gauge the victim's personal wealth was imprecise because the victim could be a neighborhood outlier (nonetheless, median household income is an improvement over past studies that use gender to gauge wealth; for a discussion, see Reference CooneyCooney 2002). The research also relied on proxies: Whites and Hispanics are more numerous in Houston and therefore are assumed to be more representative of mainstream culture, married people are assumed to be more integrated in the social life of the community, and college graduates are assumed to be more culturally sophisticated. To test such assumptions, I examined the Houston Area Survey (HAS), a probability sample telephone survey of Harris County residents conducted annually from 1982 to 2008; http://houstonareasurvey.org/. The HAS findings suggest that the assumptions are sound: Whites and Hispanics are more apt to be Protestant or Catholic and moderate or conservative, the dominant religions and political ideologies in Houston; married people are more likely to have children, register to vote, volunteer, and attend religious services; and college graduates are more likely to follow current events, read the newspaper, visit museums and attend live theater, have access to the Internet, use a computer at home or work, and speak multiple languages (the full results from the HAS analysis are available upon request). Though blunt, data on race, marital status, and education appear to tap the theoretical concepts of conventionality, integration, and sophistication. Using the victim's prior criminal record to gauge respectability was the strongest and most direct measure in the study. In short, the indicators were reasonable and represented an improvement over prior attempts to measure Reference BlackBlack's (1976) concept of social status, but they remained imperfect.
Finally, the research was limited by the inability to control for the strength of evidence in each case, a pivotal legal consideration. But this was not a fatal flaw. Strength of evidence could be related to social status if high/low-status victims tend to be killed in murders that naturally produce more evidence. If, for example, beating a rape victim to death produces more evidence than shooting a robbery victim, and if high-status victims are more likely to be killed in the former scenario than the latter, then apparent status disparities might be a legitimate response to differences in the strength of evidence. But the models controlled for the type and method of murder, so the data included proxies for strength of evidence. Strength of evidence could also be related to social status if the police conduct more thorough investigations on behalf of high-status victims. If so, then controlling for strength of evidence might locate the source of status disparities in the police department's investigation rather than the DA's decision to seek death or the jury's decision to impose death, but it would not eliminate the existence of such disparities. Thus, controlling for strength of evidence would not change the ultimate conclusion—victim social status matters—but might identify a different causal mechanism. It is also worth noting that 496 of the 504 defendants were convicted, suggesting that insufficient evidence was rarely an insurmountable problem for the prosecution.
Results
Black suggests that the elements of social status can be examined separately, or combined to measure a person's overall social status (1993:161). I consider both strategies in turn.
Examining Each Element of Victim Social Status
Is capital punishment influenced by the victim's wealth, integration, sophistication, conventionality, and respectability? To answer the question, Table 3 reports odds ratios from the bivariate and multivariate logistic regression of seek death and death sentence on each element of social status (the number of cases dropped from 504 to 446 due to missing data on victim social status).
Table 3. Odds Ratios From the Bivariate and Multivariate Logistic Regression of Seek Death and Death Sentence on Each Element of Victim Social Status (n=446)
Notes: 1.
* =P<0.1; **=p<0.05; ***=p<0.01
2. Reference categories: defendant race=white; defendant age=teen; method of murder=shot; heinousness=Level 1; type of capital murder=robbery.
3. Other types of capital murder included arson, obstruction/retaliation, and killing a police officer.
4. The odds ratios for Asian defendants in the seek death and death sentence models were 2.63 and 2.16, respectively.
The findings suggest that vertical status does not influence capital punishment. The small bivariate relationship attenuated in the multivariate models because median household income and female victim were partially redundant: MHI was higher among female victims, and death was more apt to be sought and imposed on behalf of female victims. Because vertical status could have a nonlinear effect, the continuous measure was recoded as follows: MHI above the mean, MHI divided into quartiles, and MHI divided into quintiles. MHI was also logged to correct for positive skew. The substantive outcome remained the same regardless of measurement specification.
The pattern for MHI is common in research: a bivariate relationship attenuates in a multivariate model. But a bivariate relationship can also strengthen in a multivariate model. In fact, Table 3 reveals that all the bivariate relationships for the remaining elements of social status were amplified in the multivariate models. This less common but equally valid pattern is called suppression—the true relationship is suppressed in the absence of controls. Suppression occurs if the correlation of each X with Y has a different sign than the product of the remaining correlations. Figure 3, Panel A, illustrates three possible forms of suppression (see examples A, B, and C). Figure 3, Panel B, illustrates the actual forms of suppression for the remaining elements of social status, all of which correspond to example A (for an excellent discussion of suppression see Reference McClendonMcClendon 1994).
Figure 3. The Suppression of Victim Social Status
Beginning with radial status, the odds of seeking death climbed from 1.4 to 1.7 times higher, and the odds of a death sentence climbed from 1.3 to 1.5 times higher, if the victim was married. Suppression occurred because the death penalty was more likely to be sought and imposed on behalf of married victims despite the fact that such victims were less likely to be raped: 3 percent of married victims were raped, compared to 9 percent of nonmarried victims. Thus, the preferential treatment of married victims became more pronounced after accounting for the fact that married victims were less apt to suffer a macabre form of capital murder that tends to attract the ultimate state sanction.
Cultural status—defined as sophistication—was also suppressed in the absence of controls. The odds of seeking death climbed from 1.0 to 1.5 times higher, and the odds of a death sentence climbed from 1.1 to 2.2 times higher, if the victim had a college degree. How is a college degree transformed from meaningless to consequential? The death penalty was sought and imposed on behalf of graduates and nongraduates at the same rate, but graduates were less apt to be killed in the most gruesome murders: just 10 percent of graduates were killed in murders coded as Level 3 heinousness, compared to 22 percent of nongraduates. Controlling for the heinousness of the crime demonstrates how the initial appearance of even-handedness can be misleading. Equal treatment for unequal crimes is unequal treatment.
Cultural status—defined as conventionality—also had a strong influence on capital punishment that became even stronger once controls were introduced. The odds of seeking death climbed from 1.6 to 1.8 times higher, and the odds of a death sentence climbed from 1.6 to 2.0 times higher, if the victim was white/Hispanic. Here, suppression was a product of the relationship between victim race and defendant race: Just 32 percent of white/Hispanic victims were killed by black defendants, compared to 81 percent of black/Asian victims. Thus, the death penalty was pursued more vigorously on behalf of white/Hispanic victims despite the fact that such victims were rarely killed by black defendants, the race group most at risk of the ultimate state sanction.
Finally, the true impact of normative status on capital punishment was suppressed until controls were introduced. The odds of seeking death climbed from 1.6 to 1.8 times higher, and the odds of a death sentence climbed from 1.4 to 1.7 times higher, if the victim had a clean criminal record. The suppression of normative status was a product of several factors. Specifically, victims with a clean criminal record were less likely to be killed by a defendant with a prior violent conviction, less likely to be killed in a murder coded as Level 3 heinousness, and less likely to be killed in a murder with multiple victims. The preferential treatment of respectable victims was even stronger than it initially appeared because such victims tended to be killed in murders that were less likely to include the pivotal legal considerations that are often fateful in a capital case.
Identifying the Stages of the Process That Produced Disparities
Having established the relationship between each element of social status and capital punishment, an important question remains: Is the relationship a product of the DA's decision to seek death, the jury's decision to impose death, or some combination? Table 4 presents predicted probabilities to answer the question (confounders held constant at the mean). The predicted probabilities suggested that status disparities stem from two processes: abatement and accretion.
Table 4. Using Predicted Probabilities to Identify the Stage of the Process That Produced Disparities
Note: Abbreviations: SD=seek death; DS=death sentence; PP=predicted probability.
The process of abatement means that status disparities originate in the decision to seek death but are partially corrected in the decision to impose death. Radial status and normative status are both subject to abatement. Consider the pattern for radial status: If the victim was married the predicted probability of seeking death was 0.24 and the predicted probability of a death sentence was 0.16, so the conditional probability of the jury rendering a death sentence at trial was 0.67 (0.24x=0.16; x=0.16/0.24; x=0.67); but if the victim was unmarried the predicted probability of seeking death was 0.16 and the predicted probability of a death sentence was 0.12, so the conditional probability of the jury rendering a death sentence at trial was 0.75. Consider, too, the pattern for normative status: If the victim had a clean criminal record the predicted probability of seeking death was 0.20 and the predicted probability of a death sentence was 0.14, so the conditional probability of the jury rendering a death sentence at trial was 0.70; but if the victim had a prior criminal record the predicted probability of seeking death was 0.12 and the predicted probability of a death sentence was 0.09, so the conditional probability of the jury rendering a death sentence at trial was 0.75. Under abatement, the DA is considerably more likely to seek death on behalf of married and respectable victims, but jurors are slightly less likely to impose death on behalf of married and respectable victims—meaning that juries reduce but do not eliminate the impact of radial status and normative status on capital punishment. The jurors' partial reversal is presumably a response to the DA occasionally overreaching on behalf of married and respectable victims.
The process of accretion means that status disparities originate in the decision to seek death and are exacerbated in the decision to impose death. Both forms of cultural status—sophistication and conventionality—conform to the snowball pattern. The predicted probability of seeking death was 1.3 times higher on behalf of a victim with a college degree (0.24/0.18), and the conditional probability of the jury rendering a death sentence at trial was 1.4 times higher (0.92/0.67), so the predicted probability of a death sentence was 1.8 times higher (1.3*1.4 or 0.22/0.12). Similarly, the predicted probability of seeking death was 1.6 times higher on behalf of a white/Hispanic victim, and the conditional probability of the jury rendering a death sentence at trial was 1.2 times higher, so the predicted probability of a death sentence was 1.9 times higher. Under accretion, the DA is considerably more likely to seek death on behalf of sophisticated and conventional victims, and jurors are slightly more likely to impose death on behalf of sophisticated and conventional victims—meaning that jurors amplify the impact of cultural status on capital punishment.
Examining Overall Victim Social Status
Is capital punishment influenced by the victim's overall social status? To create a composite measure, the four dichotomous indicators—married, college degree, white/Hispanic, and clean criminal record—were summed (median household income was not included in the composite measure because of the null effect described above; the substantive results were the same if the four indicators were combined based on factor scores, but the summative scale was more interpretable and facilitated the calculation of predicted probabilities).
The composite measure ranged from 0 to 4, with a mean of 2.11 and a standard deviation of 0.89. Table 5 reports odds ratios from the bivariate and multivariate logistic regression of seek death and death sentence on composite victim social status. The impact of composite victim social status in the bivariate models was moderate, producing odds ratios of about 1.3 for both case outcomes. Not surprisingly, the strength of the relationship burgeoned in the multivariate models because of the suppressors described above. Specifically, the multivariate models suggested that each unit increase in the victim's composite social status almost doubled the odds of the DA seeking death and the jury imposing death, with odds ratios of about 1.8 and 1.7, respectively (the impact of composite victim social status remained robust and significant if victim race was removed from the summative measure).Footnote 10
Table 5. Odds Ratios From the Logistic Regression of Seek Death and Death Sentence on Composite Victim Social Status (n=446)
Notes: 1.
* =p<0.1; **=p<0.05; ***=p<0.01
2. Reference categories: defendant race=white; defendant age=teen; method of murder=shot; heinousness=Level 1; type of capital murder=robbery.
3. Other types of capital murder included arson, obstruction/retaliation, and killing a police officer.
4. The odds ratios for Asian defendants in the seek death and death sentence models were 2.54 and 2.19, respectively.
The findings from the multivariate models can also be used to estimate predicted probabilities for levels of victim social status, as presented in Table 6 (confounders held constant at the mean). Considering the extremes demonstrated the magnitude of the relationship: If a defendant killed the highest status victim possible, whose composite score summed to 4 (meaning a white/Hispanic victim who was married with a clean criminal record and a college degree), then the predicted probability of a death sentence was 0.30; but if a defendant killed the lowest status victim possible, whose composite score summed to 0 (meaning a black/Asian victim who was single with a prior criminal record and no college degree), then the predicted probability of a death sentence dropped to 0.05. The predicted probability of a death sentence was about six times greater on behalf of a high-status victim, compared to a low-status victim.Footnote 11
Table 6. Using Predicted Probabilities to Estimate the Magnitude of the Relationship Between Composite Victim Social Status and Capital Punishment
Conclusion
The current article is part of a broader project that examines whether the death penalty is administered in an arbitrary manner in the capital of capital punishment. In a series of studies, I have examined the impact of defendant race (Reference PhillipsPhillips 2008), defendant legal counsel (Reference PhillipsPhillips 2009), and victim social status on the decision to seek and impose death (current work). Considering all the social attributes simultaneously produced a remarkable pattern. If an indigent black defendant killed a high-status victim—one who was integrated, sophisticated, conventional, and respectable—the predicted probability of a death sentence was 0.42. But if a white defendant who could hire legal counsel for some or all of the case killed a low-status victim—one who was marginal, unsophisticated, unconventional, and disrespectable—the predicted probability of a death sentence plummeted to 0.02 (calculated based on the final model in Table 5). To be clear, predicted probabilities regarding combinations of attributes were speculative. The probabilities were not based on cases in the data that had the combination of attributes in question. Instead, the probabilities estimated the impact of a combination of attributes based on the known impact of each attribute in the combination. Nonetheless, the probabilities are instructive. The predicted probability of a death sentence was 20 times higher in cases marked by the most aggravating social facts, providing strong evidence that the death penalty is administered in an arbitrary manner in the most active death jurisdiction in the nation. Presumably, the disparities would be even greater if perfect measures of vertical status, radial status, cultural status, and normative status were available.
The meaning of such disparities depends on whether justice is defined according to the characteristics of a particular capital murder case or the characteristics of the broader institution of capital punishment. Reference Van Den Haag and AckerVan Den Haag (2003) argues that the arbitrary administration of the death penalty does not undermine justice. He begins with the premise that justice is the punishment a person deserves. So if Offender A is executed for murdering a high-status victim, and Offender B is spared despite the identical murder of a low-status victim, the execution remains just—Offender A deserved to die. Indeed, the only injustice is the mere incarceration of Offender B, who also deserved to die. Executing both would be ideal, but only executing Offender A is better than executing neither: Some justice is better than no justice. But Reference McDermottMcDermott (2001) argues that justice cannot be reduced to the characteristics of a particular case. McDermott notes: “Implicit in the desert claim ‘X deserves punishment P’ is the claim that X deserves this punishment from a legitimate authority” (2001:322). McDermott then argues that the government's legitimate authority to execute is undercut by the arbitrary administration of the death penalty. The current research cannot settle the debate: Science cannot answer moral questions such as how justice should be defined. Nonetheless, the debate is intriguing because it raises the question of whether the retribution argument, often seen as unassailable, is actually subject to empirical critique.
Although science cannot answer questions about what should be, it can answer questions about what is. Given the small but growing body of research suggesting that victim social status influences capital punishment (Reference RadeletRadelet 1989; Reference Baldus, Woodworth and PulaskiBaldus, Woodworth, & Pulaski 1990; Reference Baldus, Woodworth and ZuckermanBaldus, Woodworth, Zuckerman, et al. 1998; Reference Baldus, Woodworth and GrossoBaldus, Woodworth, Grosso, et al. 2002), how should scholars proceed? Among the many questions that must be answered to develop a more complete understanding of the relationship, two seem particularly important:
1. What are the causal mechanisms that produce the relationship between victim social status and capital punishment? From a Blackian perspective, partisanship might provide the answer (for more on Black's theory of partisanship, see Reference Black, Baumgartner, Boyum and MatherBlack & Baumgartner 1983; Reference BlackBlack 1993; Reference CooneyCooney 1998; Reference Phillips and CooneyPhillips & Cooney 2005). Partisanship, defined as providing support to one side of a conflict, can be fateful in a capital case. The victim's supporters can fundamentally change the legal landscape by hounding the police to build the strongest case possible, demanding that the DA go for the death penalty, putting pressure on the media to cover the case, and building and sustaining a public outcry—just to name a few forms of partisanship. But what determines partisanship? Black argues that high-status victims attract more partisans, and “partisanship begets partisanship” (1993:127). The victim's organizational status—or capacity for collective action—might also be crucial. Consider the potential role of families, the most important group to which most people belong. Some victims are members of cohesive families that are accustomed to acting as a unit. Other victims are members of fractured families that never exhibit collective action. Still other victims' families fall between such extremes. Where fractured families might become even more splintered and impotent in the wake of a tragic loss, unified families can engage in partisanship that alters the dynamics of a case. Beyond families, friendship networks or work colleagues might serve as a source of posthumous partisanship. So the intervening link might be partisanship—the substantial support that high-status victims receive from family members, friends, work colleagues, and others. Other mechanisms might also be at work. It is possible that the DA and the jurors are more sympathetic to a victim whose social status is similar to their own. Or perhaps the murder of a high-status victim is simply seen as more tragic. Because the relationship between victim social status and capital punishment has been established in multiple studies using different methodologies, one of the next logical steps is to focus on the origin of such disparities. Nonetheless, researchers should be prepared for the possibility that the effect of social status is universal, but the causal mechanisms that produce the relationship might be numerous and might even vary across time and place.
2. What sort of data can be used to advance our understanding of the relationship between victim social status and capital punishment? Unfortunately, standard quantitative datasets on capital punishment contain almost no information regarding victim social status. Even original data collection efforts that draw on multiple archival sources produce imperfect measures, as demonstrated here. One promising approach would be to comb through trial transcripts. Guilt and punishment phase transcripts are filled with clues about the victim's social status. Coding transcripts would allow a researcher to develop a nuanced measure of victim social status and examine whether victim social status predicts case outcomes. The transcript data could also be supplemented with interviews of the victims' family members and friends to investigate the forms of partisanship described above. Because transcripts tend to be quite long, and in-depth interviews are expensive and time-consuming, the project would require significant funding to hire a team of research assistants. But if the research team could develop reliable coding procedures for the transcripts, and gain access to victims' family members, the project would surely produce valuable insights into modern capital punishment.Footnote 12
The concept of arbitrariness suggests that the legal facts of a capital case cannot fully explain the outcome: Social facts also shape the ultimate state sanction. In the capital of capital punishment, death is more apt to be sought and imposed on behalf of high-status victims who are integrated, sophisticated, conventional, and respectable. The notion that high-status victims matter more than low-status victims might be true of all forms of sentencing. But such an argument cannot be made with confidence because scholars have largely ignored the question. To demonstrate the gap in the broader sentencing literature, I used Sociological Abstracts to conduct a content analysis of social science journals. The review revealed a stark pattern: reams of research regarding the impact of defendant social status on sentencing, but almost no research regarding the impact of victim social status on sentencing.Footnote 13 Such an imbalance is troubling—surely punishment depends on who is swindled, violated, or attacked. In this sense, the current research has important implications beyond capital punishment: Reference BlackBlack's (1976) rich multi-dimensional concept of social status not only facilitates insights about the role of victim social status in capital punishment, but also provides a theoretical tool that could be used to address such an important omission in the broader sentencing literature.
