Social scientists and legal scholars have devoted renewed theoretical attention in recent years to the symbolic and social contexts of punitive social control (Reference Blomberg and CohenBlomberg & Cohen 1995; Reference GarlandGarland 1990, Reference Garland2001). An important theme of the new sociology of punishment emphasizes how victims' advocates and elected officials have mobilized support for tougher punishment policies by symbolically reconstructing the punishment of criminals as a victim's right (Reference Beckett and SassonBeckett & Sasson 2000:156–64). In his recent book The Contradictions of Capital Punishment, Reference ZimringZimring (2003) applies a provocative variant of this “social constructionist” theme to the toughest punishment of all: the death penalty.
Reference ZimringZimring (2003) asks why the death penalty was resurrected in the United States, when most of the rest of the world was abandoning it, and why its application is so heavily concentrated in Southern states. His answer is that capital punishment was given new life through its symbolic reconstruction as an acknowledgment of the victim's rights and as a form of compensation for the loss suffered by the victim's family. The symbolic transformation of capital punishment was essential in overcoming a potent obstacle to popular support for the death penalty, Americans' distrust of government, which otherwise would result in opposition to such a harsh and irreversible imposition of state power over citizens. According to Zimring, “vigilante” cultural traditions sustain the idea of harsh punishment as a communal ritual on the victim's behalf and counteract the inhibiting effect of distrust of government on support for the death penalty. Such vigilante traditions are most likely to have taken root in areas with a legacy of lynching, which are heavily concentrated in the South.
Reference ZimringZimring (2003:118) cautions that the data at his disposal are limited, and he calls for further research to substantiate his claims. The purpose of the present article is to probe the interconnections among distrust of government, the historical context, and public support for the death penalty in the United States using a multilevel statistical framework. We translate Zimring's discursive arguments into a formal causal model and derive explicit hypotheses from this model. We then test these hypotheses with data on death penalty attitudes, government distrust, and other attributes for 5,140 white and 1,192 black respondents from the General Social Survey (GSS; Reference Davis and SmithDavis & Smith 1998). The GSS has been used widely to study correlates of public opinion in the United States since the early 1970s, including attitudes toward the death penalty. It is the only nationally representative survey in which data on both governmental distrust and death penalty attitudes are available for relatively large samples of whites and blacks. The GSS survey data have been linked to aggregate-level measures of social context, including data on lynching, which serves as an indicator of a vigilante tradition. Our analyses represent an effort to respond to Zimring's call for further tests of his theory that can “draw us closer to understanding the link between one of the most troubling chapters of the American past and the controversial and distinctive circumstances of execution in the American present” (2003:118).
The Contradictions of Capital Punishment
Capital punishment is becoming less and less common throughout the world. According to Amnesty International (2005), 61 nations were abolitionist in law or practice in 1981. This figure increased to 122 by 2005. A small number of those nations retaining the death penalty account for the vast majority of executions. Amnesty International estimates that 81% of all known executions in 2002 were carried out in just three nations: China, Iran, and the United States. The United States is thus clearly an outlier in its reliance on execution as a means of criminal sanctioning, especially when compared to other advanced Western nations.
Over the years, various explanations for this aspect of American exceptionalism have been offered, including the distinctive features of American federalism and the populist nature of American politics (Reference Hood and TonryHood 1998; Reference Radelet and BorgRadelet & Borg 2000; Reference Zimring and HawkinsZimring & Hawkins 1986). While plausible in many respects, these explanations of why the death penalty survives in the United States fail to account fully for the striking geographic variation in its application. Executions are heavily concentrated in Southern and border states. In 2002, 71 persons were executed in 13 of the 38 states with a death penalty: 33 in Texas; 7 in Oklahoma; 6 in Missouri; 4 each in Georgia and Virginia; 3 each in Florida, South Carolina, and Ohio; 2 each in Alabama, Mississippi, and North Carolina; and 1 each in Louisiana and California (Reference Bonczar and SnellBonczar & Snell 2003). Reference ZimringZimring (2003) advances a highly creative interpretation of this geographic variation in support for and use of the death penalty in the United States.
Reference ZimringZimring (2003) begins his analysis of capital punishment by documenting divergent trends in the United States and Europe over the latter decades of the twentieth century. Whereas European nations have gradually eliminated the death penalty from their legal codes and have reframed capital punishment as a violation of human rights (p. 40), executions returned to high levels in selected states in the United States after the nationwide moratorium following Furman v. Georgia (1972). The resurgence of the death penalty in the United States is quite paradoxical, Zimring observes, because capital punishment obviously entails maximum governmental power over citizens, and the tradition of distrust of an excessively powerful government is as strong in the United States as anywhere else in the world (p. x). How is it possible, then, for large numbers of U.S. citizens to endorse the supreme use of governmental power while at the same time to be characteristically distrustful of government?
Zimring's answer is that support for the death penalty has been rendered compatible with a fundamental distrust of government through a “symbolic transformation” of capital punishment from an exercise of the power of a distant government to a service for crime victims. As a result of legal innovations such as victim impact statements and a new psychological language that depicts capital punishment as providing “closure” for the victim's family, capital trials and executions have been transformed into social practices for assisting victims. Even though executions are ultimately sanctioned and carried out by the state, they are understood as a response to the demands of individuals for punishment for private purposes (p. 64). In essence, capital punishment is rendered palatable, despite a deep suspicion of government, because it has been symbolically “degovernmentalized” and converted into a form of victim compensation (p. 63).
Zimring proposes further that receptivity to this degovernmentalized, “private service imagery” (p. 64) of execution varies greatly across states and regions, reflecting different historical circumstances. Citizens are most receptive to the imagery of capital punishment as an exercise of private justice in those areas with a “vigilante tradition.” Within the vigilante tradition, punishment is viewed as a community responsibility rather than an exercise of governmental power. In Zimring's words, “[t]he vigilante is by definition suspicious of his government, and that is one reason why vigilantes are willing to arrogate the power to punish crime to community groups” (p. 111). The vigilante tradition does not overtly condone or endorse killing, according to Zimring. Rather, by fostering the view that punishment is an expression of “the will of the community rather than the power of a distant and alien government” (p. 89), the vigilante tradition effectively “neutralizes” distrust of government as a reason to oppose capital punishment (p. 99).
Zimring invokes psychoanalytic nomenclature to explain the psychological mechanism underlying this symbolic reconstruction of punishment power. The citizen who embraces vigilante values identifies legal punishments as a communal rather than a governmental activity through “a form of transference, where the affective bond from communal social control in earlier times is transferred to state authority for executions and other serious punishments” (p. 99). In essence, transference is the psychodynamic process through which those socialized into vigilante values are able to sustain the delusion that executions are not what they really are: extraordinary exercises of state power.
In support of these arguments, Zimring points to the association between the historical experience of lynching and the contemporary use of executions.Footnote 1 He concludes: “The lynch mob and the lethal injection are found in the same American neighborhoods. Where a lynching history is absent, there is a lower-than-average chance that executions take place” (p. 118). Zimring is careful to acknowledge the difficulties in making firm causal inferences given limitations of the data, but he suggests that “the circumstantial case” for a connection between a history of lynching and executions is strong (p. 117).
Zimring thus advances a highly complex and subtle interpretation of the relationship between distrust of government and support for capital punishment. On its own, distrust of government should lessen support for the death penalty because distrustful citizens should be predisposed toward limiting governmental power.Footnote 2 However, in the presence of a vigilante tradition, this effect of distrust of government is effectively neutralized because executions have been transposed from an activity of big or distant government into an expression of community will: “… citizen identification with a vigilante tradition produces tolerance for execution behavior by inhibiting distrust of government as a motive for being afraid of or ambivalent about executions” (p. 118). In the language of causal modeling, Reference ZimringZimring's (2003) arguments imply that the strength of the vigilante tradition should moderate the effect of distrust of government on death penalty support, with significantly weaker effects of distrust observed among those who reside in areas where the vigilante tradition is more pervasive. We depict the implied model schematically in Figure 1.
Figure 1. A Model of Moderating Effects of the Vigilante Tradition on Government Distrust and Death Penalty Support
The Significance of Race
Zimring observes that the “victims of lynching were overwhelmingly African American” (2003:90), but he does not address the issue of racial differences in the contradictions of capital punishment. This is not altogether surprising, given the hypothesized psychoanalytic mechanism linking the vigilante tradition with death penalty support. As explained above, the vigilante tradition allegedly entails an affective bond with communal social control that is transferred to government, enabling those who embrace these values to dissociate executions from governmental power. This interpretation is plausible when applied to whites, but since blacks were the primary victims of vigilante acts, there is no reason to expect the same kind of positive affective bond with communal social control that would lead to such transference. Indeed, for blacks, a vigilante tradition might well foster a highly negative view of communal social control, and thus the kind of transference that occurs for whites in such an environment would be particularly unlikely to occur for blacks. Rather than neutralizing the effect of distrust of government, the strength of the vigilante tradition might enhance the inhibiting effect of distrust of government on death penalty support for blacks. This implies that for blacks the vigilante tradition may moderate the effect of government distrust on death penalty support in the opposite manner as predicted for whites.
In the present investigation, we assess the effects of government distrust and exposure to a vigilante tradition on death penalty attitudes in separate analyses of black and white respondents using data from the GSS that have been linked with data on various features of the social context in which respondents reside (Reference Davis and SmithDavis & Smith 1998). We do so by estimating multilevel regression models that evaluate whether the magnitude of the effect of government distrust on support for capital punishment among individuals varies across U.S. states, and if so, whether that variation is a function of the strength of the vigilante tradition. If the causal model derived from Zimring is correct, the expected negative effect of distrust should be significantly weaker for whites in states with a stronger vigilante tradition. The theoretical basis for any moderating role of a vigilante tradition for blacks is uncertain. However, if there is such an interaction effect, the strength of a vigilante tradition should enhance (make more negative), rather than neutralize, the inhibiting effect of government distrust on death penalty support. This would be represented in a significantly larger negative effect of distrust among blacks who reside in areas with a stronger vigilante tradition.
Finally, we propose two additional hypotheses that pertain to the main effects of the vigilante tradition on death penalty support. Zimring is primarily concerned with explaining the paradox of widespread support for state-sanctioned killing among a population inclined to distrust government. He accordingly focuses on the subtle interconnections between government distrust and the vigilante tradition, and he expresses skepticism about the possibility that “… the major influence of the vigilante precedents on contemporary capital punishment comes from any inherited enthusiasm for killing as a form of social control” (2003:98). However, it is also plausible that, for whites, a vigilante tradition will increase the likelihood that violence, including lethal violence, is viewed as a legitimate means of punishment. As Tolnay et alia note, many Southern whites viewed lynchings “as an extreme, but necessary, form of popular justice that guaranteed the swift and severe punishment of black criminals” (1996:789). Thus in some ways capital punishment can be thought of as a continuation of practices that were previously implemented by the cherished local community, which is reflected in Zimring's observation that lynchings and executions occur “in the same neighborhoods” (2003:118; see also Reference SteelwaterSteelwater 2003).
For blacks, in contrast, the lessons to be drawn from a past vigilante tradition are likely to be quite different. Although they were not the exclusive victims of lynching in the United States, blacks were disproportionately terrorized by lynching at the hands of white mobs who claimed to be administering justice (e.g., Reference Brundage and FitzhughBrundage 1997; Reference CutlerCutler 1905; Reference RaperRaper 1933; Reference Tolnay and BeckTolnay & Beck 1992; Reference TolnayTolnay et al. 1996). Given this historical experience, coupled with the “popular justice” rhetoric often invoked to justify lynching, it seems likely that the potential for abuse and unfairness of lethal sanctioning should be especially salient to blacks who reside in communities with a history of lynching. Accordingly, we hypothesize that any observed main effect of the vigilante tradition on death penalty support will be positive for whites and negative for blacks, net of other predictors.Footnote 3
Literature Review
A substantial body of research has examined correlates of death penalty attitudes. Most of this research has relied on pooled samples of whites and blacks and has focused on identifying individual attributes that are associated with attitudes about capital punishment. Although far from uniform, this research generally reveals higher levels of support among whites, older persons, men, wealthier individuals, conservatives, religious fundamentalists, married persons, and those who reside in less-populated areas.Footnote 4 The effects of other individual attributes are less certain, but there is some evidence that fear of crime increases support for the death penalty (Reference RankinRankin 1979; Reference Tyler and WeberTyler & Weber 1982; Reference Seltzer and McCormickSeltzer & McCormick 1987; Reference Keil and VitoKeil & Vito 1991; Reference Longmire, Flanagan and LongmireLongmire 1996), church attendance reduces support (Reference HarveyHarvey 1986), and the effect of educational attainment on support for the death penalty is nonlinear. Low levels of support have been observed among persons who did not finish high school and those who graduated from college, and higher levels of support have been found for those whose educational attainment falls between these points (Reference FoxFox et al. 1991). In addition, some research suggests that persons who reside in the South exhibit slightly higher levels of support than persons from other regions of the United States (Reference Barkan and CohnBarkan & Cohn 1994; Reference BohmBohm 1991; Reference FoxFox et al. 1991; but see Reference BorgBorg 1997).
Race is one of the strongest correlates of attitudes toward the death penalty and other punitive forms of formal social control in the United States: blacks are generally less punitive than whites (Reference Bobo and JohnsonBobo & Johnson 2004; Reference Longmire, Flanagan and LongmireLongmire 1996). Nonetheless, relatively few studies have compared results for blacks and whites. Some research has examined predictors of death penalty support among whites exclusively (e.g., Reference Barkan and CohnBarkan & Cohn 1994; Reference BorgBorg 1997; Reference SossSoss et al. 2003). The results of these studies are generally similar to those based on pooled race samples, but they also reveal the importance of racial prejudice in shaping death penalty attitudes among whites. The few studies that examine models of death penalty support separately for whites and blacks suggest that some factors (e.g., perceptions of police power, religious punitiveness) exhibit contradictory effects on death penalty support across the two groups (e.g., Reference YoungYoung 1991; Reference BorgBorg 1998). But the main conclusion that emerges from these studies is that most of the individual attributes consistently shown to be significant predictors of death penalty support in pooled samples shape death penalty attitudes among whites but not blacks. A similar pattern arises from analyses of other forms of punitive social control (e.g., Reference CohnCohn et al. 1991; Reference Taylor and GarthD. G. Taylor et al. 1979), although this research also suggests that fear of crime may be especially important in explaining punitive attitudes among blacks (see also Reference ArthurArthur 1998).
Despite the substantial attention devoted to individual-level predictors of death penalty support, to our knowledge only one study has examined the role of government distrust in shaping attitudes about the death penalty. Reference SossSoss et alia (2003) include an indicator of government distrust in their study of the effects of racial prejudice on death penalty support among whites. Drawing on literature that highlights problems in the administration of the death penalty, they hypothesize that government distrust should reduce public support for capital punishment. Using data from the National Election Study (NES) for 1992, they find a negative effect of government distrust on death penalty support (i.e., higher levels of distrust are associated with lower levels of death penalty support), net of several other individual and county attributes. However, this effect is not statistically significant using a two-tailed test. Therefore, strong conclusions about the effect of government distrust on death penalty support among whites cannot be drawn from their (2003) study, and they do not report parallel findings for blacks.
A similar gap characterizes the literature with respect to the role of social context in shaping attitudes about the death penalty. Although there is much speculation that various aspects of the social environment influence punitive attitudes in general and support for capital punishment more specifically,Footnote 5 only two studies have considered such effects explicitly. Reference BaumerBaumer et alia (2003) draw on instrumental and socialization theories of punitiveness as well as insights from social constructionist and conflict perspectives in their examination of the effects on death penalty attitudes of several features of the social context. Using data from the GSS and other sources, Reference BaumerBaumer et alia (2003) find that, controlling for several individual-level correlates of death penalty attitudes, persons who reside in areas with higher homicide rates, a larger proportion of blacks, and a more conservative political climate are significantly more likely to support the death penalty. These results are consistent with an integrated objectivist-constructionist theoretical perspective, wherein punishment attitudes are responsive to both objective levels of violence and claims-making activities directed at exploiting concerns about race and crime to promote support for punitive policies. Baumer et alia suggest that “objective levels of crime and the rhetoric and imagery used by elites and the media to frame the crime issue coalesce in shaping public opinion about capital punishment” (2003:867).
Although not the focus of their study, Reference SossSoss et alia (2003) also examine the effects of social context on death penalty attitudes with data from the NES. Consistent with Reference BaumerBaumer et alia (2003), they find that whites who reside in counties with higher murder rates are more likely to support the death penalty. They also find that county racial composition moderates the effect of racial prejudice on attitudes toward capital punishment: antiblack prejudice increases support for the death penalty among whites much more in counties in which blacks compose a larger fraction of the population. Taken together, these studies underscore the importance of examining contextual sources of public support for the death penalty. However, neither study considers the possible role of a vigilante tradition in shaping death penalty attitudes directly or moderating the effect on those attitudes of government distrust.
In summary, a good deal of research has documented individual-level correlates of death penalty support, but relatively few studies make explicit comparisons across race or assess the impact of the social context. Moreover, with the exception of the study by Reference SossSoss et alia (2003), the potential importance of government distrust as a factor shaping attitudes toward capital punishment has been neglected, and no prior studies have considered the possibility of theoretically informed cross-level interactions. The present study addresses these gaps in the literature by estimating the effects of government distrust for whites and blacks separately, by evaluating whether the effect of distrust varies across geographic areas, and by examining whether an indicator of a vigilante tradition moderates the magnitude of the distrust effect or exhibits a main effect on death penalty support.
Data and Methods
We examine the hypotheses outlined above with individual-level data from the GSS that have been linked with information about the social context in which respondents reside. The GSS data used for our research contain geographic codes that identify the states and local areas (metropolitan areas and nonmetropolitan counties) in which respondents reside. Roughly two-thirds of the GSS primary sampling units (PSUs) are single- or multicounty metropolitan areas, and the remaining third are nonmetropolitan counties (Reference Davis and SmithDavis & Smith 1998; see also Reference BaumerBaumer et al. 2003; Reference RosenfeldRosenfeld et al. 2001; Reference TaylorM. Taylor 1998).
Samples
The GSS has been conducted annually or semi-annually since its inception in 1972, and more than 40,000 households have now participated in the survey. The questions asked in the GSS have changed considerably over time, however, and many are available for only a portion of the full period covered by the survey. This is the case for several of the variables relevant to our analysis. Specifically, indicators of a key explanatory variable in our study, government distrust, and some important control variables (e.g., interpersonal trust, egalitarianism, racial prejudice) are available only for a subset of the years in which the GSS has been fielded, and for some of these indicators the years for which data are available do not overlap precisely. We include in the analysis all white and black respondents for whom complete data are available on the measures employed in the study. This yields a sample of 5,140 white respondents from 263 sampling units and 43 states interviewed between 1980 and 1996, and 1,192 black respondents from 170 sampling units and 36 states interviewed between 1980 and 2002.
Measures
The dependent variable for our analysis is derived from a GSS item that asks respondents if they “favor or oppose the death penalty for persons convicted of murder” (Favor the death penalty). We use this item to construct a binary measure of respondents' attitudes toward the death penalty, coded 1 for those who favor the death penalty for persons convicted of murder and coded 0 for those who oppose the death penalty.Footnote 6
The key explanatory variables in our research are the degree of distrust of government expressed by respondents and the strength of a vigilante tradition in the state in which respondents reside. Government distrust combines responses to three questions that tap whether GSS respondents have “a great deal of confidence,”“only some confidence,” or “hardly any confidence” in the people running the federal government, Congress, and the Supreme Court, respectively.Footnote 7 We code the responses from 0 (“great deal of confidence” to 2 (“hardly any confidence”) and sum across the items. The resulting index ranges from 0 to 6, with higher values indicating greater distrust (Alpha=0.66 for both the black and white samples).
The GSS items on government distrust refer explicitly to the branches of the federal government. From a theoretical standpoint, trust in the government of the respondent's state arguably is more directly relevant to death penalty support than trust in the federal government because the death penalty is imposed primarily at the state level. Unfortunately, the GSS does not ask about trust or confidence in state government. However, items in the 1991 survey asked respondents about their confidence in “government departments” and “the courts and legal system,” without specifying their jurisdiction (local, state, federal). According to the survey, moderately strong correlations exist between the expressed level of confidence in these generic governmental entities and confidence in Congress, the executive branch, and the Supreme Court: respondents who distrust the federal government are much more likely than others to lack confidence in government generally. Government distrust appears to generalize across all branches and levels of government.Footnote 8 To operationalize the vigilante tradition, we closely follow Zimring's lead. Zimring defines a vigilante tradition as one in which the punishment of criminals is regarded as a “local concern,” a “community rather than state response” (2003:98). After reviewing several possible indicators of a vigilante tradition, he concludes that historical rates of lynching serve as a useful and valid indicator of the type of vigilante values relevant to understanding the contemporary paradox of public support for capital punishment in the United States.Footnote 9
Following Zimring, we use state-level lynching data to measure the strength of the vigilante tradition in different areas.Footnote 10 We obtained state-level lynching counts for the period 1882–1968 from the Tuskegee Library Archives and append these counts to the GSS data using codes that specify the state in which respondents resided at the time of the interview. We then compute the number of lynchings per 100,000 residents using state population counts (averaged over the historical period) from the U.S. Bureau of the Census. Not surprisingly, the distribution of lynching rates across states is highly skewed. To reduce skewness and minimize the influence of extreme values, we apply a square root transformation to the lynching rate and use the transformed values in the analyses. Higher values on this variable (Vigilante historical context) indicate a greater number of recorded lynchings per capita between 1882 and 1968 in the state where respondents resided.Footnote 11
Our statistical models also include controls for additional variables that might be associated with death penalty attitudes and the key explanatory variables. These control variables encompass individual characteristics of respondents and features of the social context. With respect to social contextual variables, we include several state-level indicators of the nature of the political climate that have been highlighted in recent discussions of the death penalty (e.g., Reference Jacobs and CarmichaelJacobs & Carmichael 2002), as well as measures of various dimensions of the more local environment (conceptualized here as the GSS sampling units, which reflect counties and county groups) that have been emphasized in previous multilevel investigations of the factors that shape attitudes about capital punishment (Reference BaumerBaumer et al. 2003; Reference SossSoss et al. 2003).
Our analysis includes most of the individual-level attributes shown in previous research to be significant predictors of death penalty support. The following control variables are included for both whites and blacks: sex, age, educational attainment, income, marital status, political orientation, religion, church attendance, place size, region, fear of crime, interpersonal trust, egalitarianism, and year of interview.Footnote 12 In addition, we include an indicator of racial prejudice in our regression models for whites. We restrict the racial prejudice variable to the models for whites both because prior research has shown it to be a particularly strong predictor of white attitudes toward the death penalty (e.g., Reference CohnCohn et al. 1991; Reference YoungYoung 1991; Reference Barkan and CohnBarkan & Cohn 1994) and because the prejudice items available in the GSS for the years included in our study refer only to prejudice against blacks.Footnote 13 Our measurement of the individual-level control variables, described in Appendix A, conforms closely to strategies used in prior research.
To account for effects on public attitudes toward the death penalty of state political ideology and partisanship (see Reference Jacobs and CarmichaelJacobs & Carmichael 2002), we include three indicators of the political landscape of the state in which respondents reside: the political party composition of state legislative chambers, the political party affiliation of state governors, and state government ideology. Each of the indicators considered refers to conditions of the respondent's state in the year preceding the interview. We measure the composition of state legislative bodies with a continuous variable (State Republican legislature) that represents the percentage of legislative seats occupied by Republicans. We measure the governor's party affiliation with a dummy variable (State Republican governorship) coded 1 for respondents who reside in states with a Republican governor and coded 0 otherwise. The indicator of state government ideology used is based on combined interest group ideology ratings for each of the major actors in state government, weighted by their share of power (see Reference BerryBerry et al. 1998, for a detailed description of this measure). The variable is a continuous measure, with higher values reflecting more politically conservative state governments (State government conservatism). Consistent with claims that these dimensions of state political climate are conceptually distinct (e.g., Reference BerryBerry et al. 1998), they exhibit weak-to-moderate intercorrelations (see Appendix B for a matrix of correlations for these and other selected variables considered in the analysis).
Finally, following Reference BaumerBaumer et al. (2003), we include as control variables the homicide rate, racial composition, level of income inequality, and degree of citizen conservatism in the PSUs from which the GSS respondents represented in our study were drawn. The homicide rate (PSU homicide rate) represents the number of homicides per 100,000 residents in the year preceding the interview; the measure was constructed with data on homicides from the National Center for Health Statistics (NCHS) and population estimates from the U.S. Bureau of the Census. We measure the degree of conservatism in the local environment by aggregating responses to a GSS question that asks respondents to report where they would place their political views on a scale from 1 (extremely liberal) to 7 (extremely conservative). The contextual variable reflects the mean level of conservatism in respondents' PSUs during the period when the GSS interview was conducted (PSU conservative climate). We use data from the U.S. Bureau of the Census to construct measures of relative minority population size and economic inequality. PSU-level indicators of these socioeconomic conditions in the year prior to the respondents' interview were derived using decennial county-level census data to estimate values for the intercensal years. The relative size of the minority population in the respondents' PSUs is measured with census data on the percentage of residents who identify themselves as black (PSU percent black). To reduce skewness, we use the natural log of this variable in the regression models. The level of economic inequality for the PSUs in our analysis is measured with the Gini index for the relative distribution of family incomes within these areas (PSU income inequality).
Analytic Strategy
Our primary objectives are to estimate the effects of individual government distrust and state-level lynching prevalence on support for the death penalty, controlling for other individual attributes and several features of the social context, and to evaluate whether any observed effect of distrust is moderated by a vigilante historical context. We employ a multilevel modeling strategy to achieve these objectives. Multilevel regression models have become the standard method for analyses of data on individuals nested within communities, and they provide an efficient means to examine the type of cross-level interaction hypothesized here. Due to the binary coding of our dependent variable and the hierarchical nature of our data, we estimate a series of three-level logistic regression models that evaluate support for capital punishment among individuals nested within GSS PSUs, which serve as our level 2 units, and states, which serve as our level 3 units (see, e.g., Reference Raudenbush and BrykRaudenbush & Bryk 2002; Reference Guo and ZhaoGuo & Zhao 2000; Reference Patterson and ProsserPatterson 1991; Reference Wong and MasonWong & Mason 1985).Footnote 14
Given the divergent theoretical predictions for whites and blacks discussed above, we report results separately for the two groups. After presenting descriptive statistics for the two samples, we present multilevel regression models that situate our analyses in the prior literature and evaluate our main hypotheses. We begin with a fully unconditional model, followed by a model that adds in the individual-, PSU-, and state-level control variables. These models describe the basic patterns observed in our data and permit an assessment of their comparability to those from prior research. We then assess our main hypotheses by estimating multilevel models that include all of the control variables and the two key explanatory variables: government distrust and the prevalence of a vigilante tradition. In these analyses, we assess the main effects of the explanatory variables, as well as the hypothesized cross-level interaction between them.
Results
Table 1 presents descriptive statistics for all variables included in the analysis. Overall, 79% of white respondents interviewed during the period report that they favor the death penalty for persons convicted of murder. Support for the death penalty is much lower among blacks, with just 48% indicating that they favor the death penalty. The means for our key explanatory variables, the index of government distrust and vigilante historical context, are not amenable to straightforward interpretation. However, examination of the components of the government distrust index (not shown) reveals that just over one-quarter of whites express distrust (i.e., have hardly any trust) in the executive and legislative branches of government, and about 15% distrust the Supreme Court. Comparatively, blacks express nearly identical levels of distrust in the Supreme Court and Congress, but higher levels of distrust in the executive (33% of blacks have hardly any confidence in the federal government). In addition, black GSS respondents live in states where lynching was more prevalent between 1882 and 1968. On average, black respondents reside in states where there were, on average, seven lynchings per 100,000 residents, whereas whites reside in states in which there were, on average, six lynchings per 100,000 residents.
Table 1. Descriptive Statistics for Variables Included in Analysis of Death Penalty Supporta
a For whites, the descriptive statistics for all individual-level measures are based on 5,140 persons; the statistics shown for PSUs are based 891 cases, and the statistics for states are based on 369 cases. For blacks, descriptives for the individual-level measures are derived from a sample of 1,192 persons; the corresponding statistics for PSUs are based on 447 cases, and the statistics for states are based on 283 cases.
Note: Italics indicates dichotomous variables for which the mean shown equals the proportion coded 1.
Table 2 displays results for two multilevel regression models for white and black GSS respondents. The first model is a fully unconditional model that shows the estimated mean levels of support for capital punishment for each group and the observed variability in support across PSUs and states. The second model for each group adds all the control variables considered in the analysis, including dichotomous variables that identify the interview year (to conserve space, the parameters for the year variables are not shown in tabular form).
Table 2. Hierarchical Logistic Regressions of Death Penalty Support on Control Variables for GSS Respondents.
Note: Standard errors in parentheses. Models also include dummy variables identifying the year of the interview.
* p≤.05, two-tailed test
a Parameter not estimated.
For whites, Model 1 shows that the estimated mean level of support for capital punishment (π0) is 0.792 (exp(1.34)/1+ exp(1.34)). But the random effects variance components at the bottom of the table also reveal that the level of support expressed by whites varies significantly across PSUs (r0jk=0.164, p≤0.05) and states (u00k=0.076, p≤0.05). Model 2 adds the individual-, PSU-, and state-level control variables considered in the study. Comparing the variance components from this model with those obtained in Model 1 indicates that more than 40% (0.42=(0.164−0.095)/0.164) of the observed PSU-level variation and more than 50% (0.53=(0.076−0.036)/0.076) of the state-level variation in death penalty support among whites is accounted for by the control variables.
In subsequent models, we consider the extent to which state differences in the presence of a vigilante historical context also contribute to the overall state-level variation in support for capital punishment among whites.Footnote 15 Before doing so, it is instructive to note that the effects observed in Model 2 for the individual-level controls are largely consistent with previous research on death penalty attitudes, which has focused largely on whites (see Reference BohmBohm 1991; Reference Longmire, Flanagan and LongmireLongmire 1996).Footnote 16 Also consistent with expectations, we find that whites who reside in states with more-conservative governments are more likely to support the death penalty. The results for the other contextual indicators, however, depart from theoretical expectations. Specifically, none of the other indicators of state political party representation and none of the PSU-level measures are significantly associated with whether whites support the death penalty.
Models 3 and 4 present a parallel set of analyses for blacks. Model 3 reveals a much lower estimated mean level of support for capital punishment among blacks (0.484=(exp(−0.061)/1+ exp(−0.061)) than was observed for whites. The model also shows that levels of support for the death penalty among blacks exhibits less variability across PSUs (r0jk=0.010, p≤0.05) than was observed for whites in the unconditional model, and black support for the death penalty does not vary significantly across states (u00k=0.106, p≥0.05). Model 4 shows further that although the effects of most of the individual-level control variables are in the expected direction, only three achieve statistical significance at conventional levels using a two-tailed test. Similar to the patterns observed for whites, the likelihood of death penalty support is lower among blacks who did not finish high school and who attend church more frequently, and it is higher among those with more conservative political orientations. None of the indicators of social context is significantly related to death penalty attitudes among blacks. Interestingly, the results indicate that conservative state governments are important for shaping attitudes about capital punishment among whites, but not blacks.
Are attitudes about the death penalty, as hypothesized, affected by government distrust and the presence of a stronger vigilante tradition? The results shown in Table 3 address these questions. We again present two multilevel regression models for each group. The first model displays estimates of the main effects for the indicators of government distrust and state vigilante historical context, and the second model considers whether they exert a multiplicative effect. The equations that generate the results shown for these models also include the individual-, PSU-, and state-level control variables, but we omit them from Table 3 because their effects are virtually identical to those shown in the previous table.Footnote 17
Table 3. Hierarchical Logistic Regressions of Death Penalty Support on Government Distrust and Vigilante Historical Context
Note: Standard errors in parentheses. Models also include dummy variables identifying the year of the interview, and individual-, PSU-, and state-level control variables.
* p≤.05, two-tailed test
a Parameter not estimated.
Model 1 shows that for whites, the coefficient for the government distrust index is positive and statistically significant (π17=0.054, p≤0.05). Whites who distrust the government are significantly more likely to support capital punishment. In addition, the main effect of the vigilante tradition is significant. Net of a wide array of individual attributes and features of the social context, whites who reside in states where lynching was more prevalent during the latter part of the nineteenth and early part of the twentieth centuries are significantly more likely than others to support the death penalty (γ004=0.115, p≤0.05).
In Model 2, we examine whether the effect of government distrust on white death penalty attitudes varies across states, and if so, whether the effect is moderated by the presence of a vigilante tradition. The estimated variance components for this model reveal that the government distrust slope does not vary significantly across states (u17k=0.010, p>0.05). Given this finding, it is not surprising that the indicator of vigilante historical context does not significantly moderate the effect of government distrust on death penalty support (γ1501=−0.014, p>0.05).
For blacks, the results displayed in Table 3 tell a very different story. Contrary to the pattern observed for whites, Model 3 shows that the estimated coefficient for the government distrust index is significantly negative for blacks (π15=−0.134, p≤0.05), indicating that blacks who distrust the government are significantly less likely to support capital punishment. Also, Model 3 reveals that the effect of the historical prevalence of lynching on blacks' attitudes about the death penalty is not statistically significant (γ004=0.101, p>0.05).Footnote 18 Finally, as shown in Model 4, the effect of government distrust for blacks exhibits little variation across states after controlling for other factors (u14k=0.019, p>0.05), and the state-level variability that does exist is not moderated by the historical prevalence of lynching.
In summary, the findings reveal a significant positive main effect of lynching prevalence only for whites. Further, we observe significant effects of government distrust on death penalty support among whites and blacks that are opposite in sign.Footnote 19 These effects are relatively stable across states and are not moderated by the prevalence of lynching.Footnote 20
The effects reported in Table 3 are robust to a variety of alternative empirical specifications, including multiplicative models that test for the possibility that the patterns shown are variable across time and region. To ensure that the results obtained are not sensitive to pooling several years of GSS data, we estimated a series of models that assessed whether the magnitude of the effects of our key explanatory variables and other variables varied over the period covered in the study. We did so by adding interaction terms for each variable, with year of interview specified as the moderator variable. These supplementary analyses revealed that the effects of government distrust and state vigilante historical context do not vary significantly over the time frame examined in our research, and the parameter estimates for these variables shown in Table 3 are unchanged when interactions between year of interview and the control variables are introduced.Footnote 21
In addition to being robust, it is noteworthy that the magnitude of the significant effects observed for government distrust and lynching is nontrivial. To illustrate, we present in Table 4 predicted probabilities of death penalty support for white and black respondents who exhibit widely different levels of government distrust. These predicted probabilities are computed using the coefficients from Models 1 (for whites) and 3 (for blacks) and assuming mean values for all other variables (see, e.g., Reference Hosmer and LemeshowHosmer & Lemeshow 2000).
Table 4. Predicted Probabilities of Death Penalty Support Among Persons who Express Different Levels of Distrust and who Reside in States with a Stronger Lynching Tradition.
a Estimates not computed because coefficient was statistically non-significant.
As Table 4 shows, the predicted probabilities associated with government distrust imply that levels of support for capital punishment range from about 79% among those who exhibit very low levels of distrust (the fifth percentile on the government distrust scale) to just over 84% among those who exhibit very high levels of distrust. The contrast is much more striking for blacks, among whom levels of support for the death penalty range from about 59% at very low levels of distrust to about 39% at very high levels of distrust. Table 4 also reveals that, net of other factors, levels of death penalty support increase from about 78% among whites who reside in states with no history of lynching (equivalent to the fifth percentile) to about 86% among whites who reside in states with a very prominent history of lynching (i.e., the 95th percentile). Although modest in absolute terms, the magnitude of the lynching effect observed for whites is comparable to or larger than the net effects of many of the most robust predictors of death penalty support, including gender, education level, income, frequency of church attendance, and racial prejudice.
Discussion
Our evaluation of Zimring's account of the contradictions of capital punishment in the United States has yielded mixed results. The analyses offer no support for the argument that a vigilante tradition, as reflected in a legacy of lynching, promotes support for the death penalty among whites by neutralizing the effect of government distrust. We have interpreted this argument as implying a statistical interaction, whereby a negative “inhibiting” effect of distrust on death penalty support is moderated (weakened or made more positive) by residence in states with a history of frequent lynching. Contrary to this hypothesis, the evidence from the multilevel analysis indicates no significant variation across states in the effect of government distrust on support for capital punishment for either whites or blacks that might be statistically explained by a measure of lynching.
Of course, measurement error is always a potential explanation for “null effects.” The dichotomous measure of death penalty used in our research tends to inflate levels of support and fails to capture variability in degrees of support for the death penalty (e.g., Reference McGarrell and SandysMcGarrell & Sandys 1996). It is conceivable that the model of interaction effects would fare better with a more-refined measure of the dependent variable. For instance, perhaps a vigilante tradition would significantly moderate the influence of government distrust on the degree of support for capital punishment, a possibility that cannot be evaluated with the dichotomous measure available in our data. We note, however, that our findings for control variables are generally consistent with previous research that has relied on broader measures that distinguish between levels of intensity in support for capital punishment.
The use of state-level counts of lynching in the past to measure the strength of support for vigilante values is also open to criticism. If vigilante traditions are localized phenomena, states may not serve as the appropriate geographic scale for measuring the frequency of lynching. A more fundamental criticism is that historical information on lynching is at best an indirect proxy for contemporary vigilante sentiments. We accordingly join Reference ZimringZimring (2003:103) in calling for a “survey agenda” to generate the kind of data that would permit more rigorous assessments of the influence of vigilante values on attitudes toward capital punishment.
Despite any limitations of measurement, our analyses do reveal a substantively important main effect among whites for the indicator of a vigilante tradition. White respondents residing in states with a history of frequent lynching are significantly more likely to express support for the death penalty. This effect emerges despite an extensive array of controls for individual attributes and other contextual factors. We caution that the precise nature of the processes linking a history of lynching with contemporary death penalty support among whites cannot be determined with our data. Zimring suggests that a vigilante tradition facilitates the embrace of harsh forms of punishment as a communal ritual in the service of victims. Although our findings are generally supportive of these arguments, further research is needed to substantiate this particular interpretation or to advance alternatives. Nevertheless, our analyses confirm Zimring's general insight that “the troubling chapters of the American past” (2003:118) are highly relevant to an understanding of geographic variation in contemporary support for state-sanctioned executions.
We also uncover intriguing main effects of distrust of government on death penalty support that differ markedly for whites and blacks. For black respondents, government distrust is related to support for capital punishment in the direction consistent with intuition: blacks who are distrustful of government are less likely to confer upon government the supreme power to take a human life. The effect of government distrust is the reverse for whites. Rather than exerting any inhibiting influence on death penalty support, government distrust is associated with an increased likelihood that white respondents will support the most extreme exercise of governmental control, holding constant other potentially confounding factors, such as political ideology and partisanship.
These results for whites thus resurrect the paradox posed at the beginning of the paper: Why is it that the most distrustful whites are actually the strongest proponents of capital punishment? A possible answer is implied in another study coauthored by Zimring dealing with the relationship between government distrust and support for California's highly publicized “three strikes” legislation for mandatory prison sentencing (Reference ZimringZimring et al. 2001). The researchers found that those citizens who were most distrustful of government were most supportive of “broad and extreme penal measures” (Reference ZimringZimring et al. 2001:232). To interpret this finding, Zimring et alia speculate that distrustful citizens typically are suspicious that governmental officials will be sympathetic to offenders rather than victims (“people like us”) and will treat offenders “with inappropriate leniency” (2001:231). As a consequence, they favor policies that sharply limit the discretion of officials in the application of criminal sanctions.
We suggest that Zimring et alia's arguments about mandatory prison sentencing can be extended to capital punishment. Advocates of the death penalty commonly express skepticism about alternative punishments such as life sentences without parole on the grounds that murderers often “get off easy” in practice. Sentences that are ostensibly for “life” do not in fact turn out to be so, or offenders serving life sentences enjoy amenities such as television, weight lifting, and free health care in prison. Citizens who do not trust government officials to carry out harsh, permanent punishment may be, for this reason, more likely than others to support the death penalty, which minimizes the discretionary influence of officials, certainly after it has been imposed. White citizens, that is. The experiences of blacks have evidently shaped their attitudes toward serious criminal sanctioning in ways that differ profoundly from those of whites. An important challenge for future research is to explain more fully the stark racial differences in attitudes toward punitive control.
Our results also have broader implications for knowledge about the features of the social context that shape death penalty attitudes. Although not the focus of the present analysis, the results obtained for three features of the social context—homicide rates, racial composition, and political conservatism—diverge from those reported in prior research. As noted above, Reference BaumerBaumer et alia's (2003) multilevel analyses of GSS data for a pooled sample of whites and blacks reveal significant positive effects of each of these features of the social context on death penalty support, and Reference SossSoss et alia (2003) find a significant positive effect of county-level homicide rates on death penalty support for whites using data from the 1992 NES. Our analyses indicate that none of these contextual conditions affects death penalty attitudes significantly for whites or blacks.
Our supplementary analyses suggest that this divergence may be a function of differences in sample composition. Specifically, we found evidence that death penalty attitudes became significantly more sensitive to the degree of political conservatism in the local environment over the period covered in our study, with whites who reside in such areas becoming increasingly likely to support the death penalty. We found a similar pattern for PSU homicide rates and percent black, although the annual change in the effects of these variables was not statistically significant. The cumulative impact of these patterns is that the influence on death penalty attitudes of levels of political conservatism, homicide, and percent black was significantly stronger in the 1990s than in the 1980s, as reported in footnote 21. Our findings for these contextual indicators diverge from those reported by Reference BaumerBaumer et alia (2003) and Reference SossSoss et alia (2003) because a much larger proportion of our samples were drawn from the 1980s. Future research should explore in greater detail the historically contingent effects on death penalty attitudes of these variables, as well as other factors that may yield differing effects over time.
We close with a final comment on the general implications of our research for theories of punishment. Our analyses call attention to the importance of both historical continuity and contingency in support for punitive criminal justice policies. On the one hand, the findings about lynching suggest that horrific events from the past evidently continue to shape attitudes about the exercise of governmental power for criminal sanctioning almost one hundred years later, at least among whites. On the other hand, the finding that homicide rates and a conservative political climate became salient factors for death penalty support only in the 1990s reaffirms the basic social constructionist insight that public consciousness about crime and criminals is not a simple or automatic reflection of current “objective” conditions, such as chronically high rates of crime. For crime to influence punishment attitudes, it must be “problematized,” and the ways in which this occurs will reflect the sociopolitical environment of the time. The climate of our time is highly receptive to inflexible, irreversible, victim-oriented punishment policies (Reference GarlandGarland 2001), and especially so among whites in conservative locales where homicide rates are high, government distrust is rampant, and the legacy of lynching is most pronounced.
Appendix A. Definitions and Metrics for Individual-Level Control Variables Included in Analysis of Support for the Death Penalty
Appendix B. Correlations for Selected Variables in Analysis of Support for the Death Penalty
Note: Correlations for white GSS respondents are shown in the upper right-hand portion of the matrix; correlations for blacks are presented in the lower left-hand portion of the matrix.
* p≤.05.
