Chivalry and the Moderating Effect of Ambivalent Sexism: Individual Differences in Crime Seriousness Judgments

Abstract

Previous studies have shown that female offenders frequently receive more lenient judgments than equivalent males. Chivalry theories argue that such leniency is the result of paternalistic, benevolent attitudes toward women, in particular toward those who fulfill stereotypical female roles. Yet to date, studies have not examined whether such leniency is indeed associated with paternalistic societal attitudes toward women. The present study goes beyond the investigation of demographics and employs Glick and Fiske's (1996) concepts of hostile and benevolent sexism. We use these concepts to highlight the role of individual differences in attitudes toward women as a key to our understanding of lenient attitudes toward female offenders. Eight hundred forty respondents from a national sample of Israeli residents evaluated the seriousness of hypothetical crime scenarios with (traditional and nontraditional) female and male offenders. As hypothesized, hostile and benevolent sexism moderate the effect of women's “traditionality” on respondents' crime seriousness judgments and on the severity of sentences assigned.

Principles of justice require that arrestees, suspects, defendants, and sentenced offenders (henceforth offenders) be treated equally. Accordingly, offenders' characteristics, such as their demographics, ought to be disregarded. However, a vast amount of theoretical and empirical research suggests that in practice this is not the case (see Daly & Tonry 1997). Most of this literature reports discrimination against disadvantaged social groups, such as African Americans, and individuals who are financially underprivileged.

Even though women are members of a socially weak group, several studies have demonstrated that female offenders tend to receive more lenient treatment than male offenders who have committed the same crimes (see Daly 1989; Daly & Tonry 1997; Spohn 1999). Chivalry theory has arisen as the primary theoretical framework for understanding these findings, suggesting that protective and benevolent societal attitudes toward women lead (predominantly male) decision makers throughout the criminal justice system to take a relatively lenient approach toward female offenders. However, despite the theory's logical appeal, several of its assumptions remain untested. First, most studies of chivalry theory maintain their focus on the characteristics of the offender (e.g., Johnson & Scheuble 1991; Spohn & Beichner 2000; Bickle & Peterson 1991), whereas a test of the theory's assumptions actually requires examination of the characteristics of the evaluators of crime.

In addition, the majority of studies on crime judgments adopt a sociological perspective and take into account demographic characteristics, such as gender (e.g., Allen & Wall 1993; Coontz 2000). The premise of this sociological perspective is that demographics represent the underlying attitudes that ultimately guide crime judgments. However, even if members of different demographic categories (e.g., men and women) tend to form different crime judgments, such a perspective forgoes the possibility of exploring individual differences among members of the same demographic group. When trying to explain differences in crime judgments, there is no reason to presume that all men or all women will hold the same attitudes. Therefore, instead of restricting the investigation to demographics, a direct assessment of attitudes toward women is more likely to provide meaningful insights for our understanding of the chivalry phenomenon.

Finally, because the theory proposes societal norms and attitudes as the basis for the differential treatment of offenders, one would need to study crime judgments among the general public in order to complement extant evidence from concrete decisions taken by law enforcement employees (e.g., police officers, prosecutors, judges). Even though some studies have revealed corresponding judgment patterns for law enforcement personnel and society as a whole (e.g., Corbett & Simon 1991; Levi & Jones 1985; McCleary et al. 1981), chivalry theory has not been tested on a sample of the general public, which would more closely represent overarching societal norms.

The need to examine public opinion in this context is further supported by claims for the direct influence of public opinion on the formation of judicial decisions. Sentencing policies and judicial decisions are strongly influenced by what is considered to be a “community standard” (e.g., Rossi & Berk 1997; Samuel & Moulds 1987; Seron et al. 2006). Judgments by law enforcement and judicial personnel are made in the context of a given society, with its particular set of norms and shared values. These norms and shared values hold a key role in formulating a community standard of justice, which, whether consciously or not, influences the decisions ultimately made by police officers, lawyers, and judges. A study of the general public can empirically inform what the community standard in a given society may be. In the present study, a representative sample of the Israeli public was used to account for individuals' attitudes in explaining differential judgments of male and female offenders.

Chivalry Theory

Although the number of female offenders convicted and incarcerated has increased dramatically in recent years, the biased lenient approach toward them has persisted and has been empirically demonstrated at practically every stage of the judicial process. Studies have shown that wherever discretionary decisions are made, women are less likely than men to be detected, arrested, charged, convicted, and sentenced (e.g., Daly & Tonry 1997; Demuth & Steffensmeier 2004; Spohn 1999; Spohn & Beichner 2000; Stolzenberg & D'Alessio 2004). If sentenced, women are likely to receive milder sentences than men (e.g., O'Neil 1999; Steffensmeier, Kramer, et al. 1993; Steffensmeier, Ulmer, et al. 1998). These results have held, albeit somewhat attenuated, even after controlling for important legal variables, such as the type of offense and the offender's criminal history (e.g., Farnworth & Teske 1995; Spohn 1999; Spohn & Beichner 2000).

To explain this pattern, a number of researchers have developed what is known as chivalry theory. Of several explanations offered within this framework (for a review see Steffensmeier 1980), the most prevalent is termed true chivalry (Edwards 1989:168), according to which protective and benevolent societal attitudes toward women are responsible for the lenient approach toward female offenders. The theory suggests that patriarchal cultures tend to identify women as weak, submissive, childlike, and defenseless, and as not being fully responsible for their actions. In this context, “well”-socialized individuals come to believe that female offenders need to be protected rather than punished (e.g., Kulik et al. 1996; O'Neil 1999; Scheider 2000; Steffensmeier, Kramer, et al. 1993; Stolzenberg & D'Alessio 2004). Contrary to how these individuals would view a male offender—as an independent and mature individual who is responsible for his actions—female offenders are often considered victims of an environment that has failed to provide the necessary guidance and supervision that women generally deserve (Steffensmeier 1980).¹ Thus chivalry theory suggests that society's view of women as weak and defenseless leads to an overall lenient approach toward female offenders.

Nevertheless, a number of studies have found no empirical backing for chivalry arguments, failing to discover gender differences in the treatment of criminal offenders, or even finding women treated more harshly than men (see Daly & Tonry 1997; Spohn & Beichner 2000; Steffensmeier, Kramer, et al. 1993). Some researchers have suggested that the differential treatment of female offenders is not applied to all female offenders, but rather depends on the type of women involved (e.g., Corley et al. 1989; Crew 1991; Farnworth & Teske 1995). We now turn to review these studies.

Selective Chivalry

In light of research inconsistencies regarding the leniency shown toward female offenders, more elaborate formulations of chivalry theory suggest that only women who meet a certain set of social criteria will benefit from preferential treatment (e.g., Crew 1991; Farnworth & Teske 1995; Johnson & Scheuble 1991). The chivalry effect can be thought of as a form of exchange in which society grants female offenders more lenient treatment in return for maintaining “appropriate” and traditional gender-role attributes and behaviors (Steury & Frank 1990). Specifically, in order to enjoy chivalrous treatment, women need to be socially and economically subordinate to their male partners and are expected to fulfill utilitarian familial functions: they should be married, preferably with children (Bickle & Peterson 1991; Daly 1987, 1989); live with their husband; serve as a housewife; and be in paid employment for only a few hours a day at most (Corley et al. 1989; Crew 1991).²

Accordingly, the selective chivalry argument is that only female offenders who conform to traditional gender roles (henceforth “traditional” women) are entitled to the protection granted by society in the form of leniency (see Bickle & Peterson 1991; Spohn & Beichner 2000). However, female offenders who fail to conform to such traditional roles (e.g., by being single, careerist, or feminist; henceforth “nontraditional” women) forfeit the advantages normally granted to traditional women and in some cases draw even harsher treatment than men (e.g., Johnson & Scheuble 1991; O'Neil 1999; Steury & Frank 1990). According to this perspective (also known as the “evil woman” thesis; see Crew 1991; Daly 1989; Spohn 1999), evaluations of crimes take into consideration not only what the woman has done but also who she is with respect to her position in the family and in society (Laster 1994). In this sense, nontraditional women are accused of a double deviance: once for the crime they have committed, and once more for departing from what is considered gender-appropriate behavior (e.g., Bickle & Peterson 1991; Steury & Frank 1990). Thus it appears that societal attitudes toward female offenders are not so much a result of chivalry as they are a result of an attempt to enforce sex-role expectations (e.g., Crew 1991; Edwards 1989).

This portrayal is consistent with postmodern gender theory (e.g., Butler 1990), whereby women's performances are expected to adhere to socially constructed regulative discourses. These regulative discourses, or disciplinary regimes, predetermine the acts that will appear as coherent and “normal” in a given society. Thus, gender roles consist of a rehearsed set of acts that serve as scripts by which women behave. In paternalistic societies, the common scripts by which women are considered “normal,” and whereby their behavior is deemed appropriate, portray a submissive, family-oriented role. From this perspective, crime judgments can be considered one possible means by which women are led to abide by expected scripts.

In support of selective chivalry theory, a number of studies found that nontraditional women were treated more harshly than men by the criminal justice system (e.g., Daly 1987, 1989; Farr 2000; Johnson & Scheuble 1991; Laster 1994; Spohn & Beichner 2000).³ Nevertheless, even after taking women's traditionality into consideration, the expected leniency toward women is still not always found (see Bickle & Peterson 1991). Contrary to this theory's propositions, women's parental status did not predict offenders' sentence outcomes (Steffensmeier, Kramer, et al. 1993), pretrial release decisions (Kruttschnitt & Green 1984), or the likelihood of being incarcerated (Spohn 1999). Thus the picture appears to be even more complex than has been conceptualized in selective chivalry theory. An examination of the proposed sources of chivalry may help explain the inconsistent findings in both chivalry theories, neither of which addresses the presumed sources of this differential approach. Although it is part of the theory's main argument, previous research has not yet tested the assumption that differential treatment stems from protective attitudes toward women in society, and in particular toward traditional women.

While chivalry theories argue that chivalry is contingent on the type of offender (i.e., traditional or nontraditional women versus men), we suggest that any form of chivalry (selective or not) is also contingent on characteristics of the person evaluating the situation (e.g., a sexist or egalitarian individual). When judging the actions of female offenders, some evaluators may exhibit chivalry, while others may not. Beyond demographic influences, chivalrous judgments will depend on the extent to which decision makers hold sexist attitudes toward women. Although it is without explicit reference to chivalry theory, the literature on ambivalent sexism offers a deeper understanding of the chivalry phenomenon and presents a possible explanation of the inconsistencies in current findings.

Ambivalent Sexism

Sexism is commonly defined as the endorsement of discriminatory or prejudicial beliefs based on gender (e.g., Campbell et al. 1997:89). This definition typically involves negative discrimination based on hostility toward, and negative stereotyping of, women, and the endorsement of traditional gender roles (i.e., confining women to roles that are less powerful or esteemed than those of men; Glick & Fiske 1997). Challenging this common view, Glick and Fiske (1996, 1997) argue for a reconceptualization of both the nature and measurement of sexism. In their opinion, sexism and traditional attitudes toward women are not necessarily negative or hostile but may also involve subjectively positive and benevolent feelings, as embodied in traditional beliefs such as protectiveness, paternalism, and chivalry (Glick, Fiske, et al. 2000:765). In other words, sexist attitudes toward women are fundamentally ambivalent, encompassing both hostile and benevolent attitudes.

Glick and Fiske (1996, 1997) note that despite the contradictory emotions involved, the two types of sexism share a paternalistic approach to women. Hostile sexism explicitly regards women as inferior. While less explicit, benevolent sexism also reinforces patriarchy by portraying women as weak and needy. Both hostile and benevolent sexism justify and maintain patriarchal social structures that emphasize men's domination of women (Glick, Fiske, et al. 2000). Nevertheless, the two are still distinct, and several studies have found them to correlate only moderately (Glick & Fiske 1996, 1997; Glick, Fiske, et al. 2000).

Glick and Fiske's framework closely corresponds with tenets of the chivalry arguments. Almost by definition, the benevolent sexist approach toward women implies lenient and forgiving judgments, whereas the denigration involved in hostile sexism suggests harsher judgments. More specifically, Glick and Fiske (1996, 1997) propose that the two types of sexism constitute a form of “carrot and stick” aimed at maintaining women's traditional gender roles (Glick & Fiske 1997:129). Hostile sexism, they argue, “may be directed most strongly at women who challenge men's power (e.g., feminists) and status (e.g., career women)” (Glick, Fiske, et al. 2000:765). Conversely, benevolent sexism is usually associated with affection for women who fulfill traditional female stereotypes, such as stay-at-home mothers. In this case, the benevolent approach toward traditional women excuses the hostile approach to nontraditional women, who allegedly “deserve” it. “Whereas hostile sexism serves to punish women who fail to conform to (male-defined) acceptable roles, benevolent sexism represents the rewards women reap when they do conform” (Glick, Fiske, et al. 2000:765).

Indeed, the two forms of sexism have been found to correlate with men's views of traditional and nontraditional women. Some studies have found hostile sexism to correlate positively with negative attitudes toward women in nontraditional roles: in one study, hostile sexism correlated positively with negative attitudes toward women managers (Sakalli Ugurlu & Beydogan 2002). A second study found hostile sexism to be associated with negative evaluations of female job candidates for a masculine-typed occupational role (Masser & Abrams 2004). In yet another study, benevolent sexism was positively associated with attitudes toward women whose behavior was consistent with a “positive” (i.e., chaste) female subtype, whereas hostile sexism was associated with attitudes toward women who depicted a “negative” (i.e., promiscuous) subtype (Sibley & Wilson 2004).

The Present Study

In light of these formulations, we propose that benevolent and hostile sexism have a central role in explaining chivalry effects. The leniency toward female offenders, and toward traditional female offenders in particular, may not be so much a general phenomenon as specifically characteristic of evaluators who hold benevolent attitudes toward women (i.e., those who espouse benevolent sexism). On the other hand, the reprimanding approach that, according to selective chivalry theory, is displayed toward nontraditional women, may in fact be characteristic only of evaluators who tend to hold negative and hostile views of women (i.e., those who espouse hostile sexism). The present study aims to test the claim posited in chivalry theory that paternalistic attitudes are the source of chivalry effects. At the same time, by incorporating the concepts of benevolent and hostile sexism, we also challenge the underlying assumption that such attitudes are uniform across individuals.

Unlike most empirical research into chivalry theory, which is based on data collected from decisions made by law enforcement officials, in the present study we focus on crime seriousness judgments among the general public. This is important for several reasons: first, research has shown that judicial decisions are influenced by the social, cultural, political, and economic contexts in which criminal justice systems operate (e.g., Dixon 1995). Furthermore, sentencing policies are typically framed with a “community standard” in mind (e.g., Rossi & Berk 1997). It is therefore important to first establish what this community standard is. Finally, because hostile and benevolent sexism are believed to stem from people's socialization within their culture, and are not particular to members of a particular profession (e.g., Glick, Fiske, et al. 2000), an investigation of judgments among the general public can offer a direct assessment of the public's role in the formation of crime seriousness judgments. Accordingly, in the present study respondents from a national sample were asked to evaluate the seriousness of hypothetical crime scenarios and to propose the appropriate punishment for these crimes as committed by (traditional and nontraditional) women and men. Our goal was to assess the relationship between the offender's gender and gender-role “traditionality” and the crime seriousness judgments assigned as a function of respondents' sexist attitudes toward women.

As a first step, to test chivalry theory we wanted to see if positive attitudes toward women—such as those manifested in benevolent sexism—are associated with a lenient approach toward female offenders, and if negative attitudes toward women—such as those manifested in hostile sexism—are associated with a lack of leniency or with an even harsher approach toward female offenders than toward their male counterparts. If paternalistic sexist attitudes are indeed responsible for the differential treatment of men and women, the absence of sexist attitudes would be expected, at the very least, to reduce such differential treatment. We thus posit our first two hypotheses:

Hypothesis 1: Crimes committed by women will be judged as less serious, and will be assigned a lighter sentence, than the same crimes committed by men, for respondents with high, but not low, benevolent sexism scores.

Hypothesis 2: Crimes committed by women will be judged as more serious, and will be assigned a harsher sentence, than the same crimes committed by men, for respondents with high, but not low, hostile sexism scores.

Note that these hypotheses do not take into consideration the extent to which these women conform to stereotypical traditional gender roles, as selective chivalry theory suggests. Even if our first two hypotheses are supported, the question still remains whether all female offenders receive differential treatment, or whether it depends on the “type” of woman the offender is. In line with the extensions proposed by selective chivalry theory, our next two hypotheses refer to the differential reactions that traditional and nontraditional female offenders elicit, compared with those toward male offenders. We expect positive attitudes toward women (i.e., among individuals with high benevolent sexism scores) to be associated with a particularly lenient approach toward traditional female offenders, but not toward nontraditional female offenders. Correspondingly, we expect negative attitudes toward women (i.e., among those with high hostile sexism scores) to be associated with a notably harsher approach to nontraditional female offenders, but not toward traditional female offenders. Hence:

Hypothesis 3a: Crimes committed by traditional women will be judged as less serious, and will be assigned a more lenient sentence, than the same crimes committed by men, for respondents with high, but not low, benevolent sexism scores.

Hypothesis 3b: Differences between the seriousness judgments of nontraditional women's and men's crimes will not be moderated by benevolent sexism.

Hypothesis 4a: Crimes committed by nontraditional women will be judged as more serious, and will be assigned a harsher sentence, than the same crimes committed by men, for respondents with high, but not low, hostile sexism scores.

Hypothesis 4b: Differences between the seriousness judgments of traditional women's and men's crimes will not be moderated by hostile sexism.

Method

Data were collected from a random sample of the adult Israeli population (n=840). Although the majority of studies in the field have been conducted on U.S. samples, several studies of crime seriousness (e.g., Herzog 2003a, 2003b), sexism (e.g., Lieblich & Friedman 1985; Seginer et al. 1990), and judicial treatment of female offenders (e.g., Erez & Hassin 1997) with Israeli samples have produced findings comparable with those of U.S. studies. Furthermore, our guiding theoretical framework suggests that chivalry effects occur because of sexist attitudes toward women. We therefore expect our findings to be relevant for other societies besides the U.S., where sexist attitudes may be prevalent.

The most recent Israeli home telephone directories (2003), covering all geographical regions, provided the study's sampling pool. Official data from the Ministry of Communications indicate that 98 percent of Israeli households are hooked up to the phone system and are listed in the telephone directories. To boost response rates, respondents who could not be initially reached were contacted again. This ultimately led to a high response rate of 62 percent. Overall, the sample presents a close fit to the official data on the Israeli population on a large variety of demographic and social variables (CBS 2005).

Data were collected through personal questionnaires, administered by means of a telephone survey,⁴ between January and March 2004. A content analysis of Israel's major national newspapers was conducted in this period and revealed no outstanding crimes that were liable to affect responses to our questionnaire. Terrorist, military, and criminal acts directly related to the Israeli–Palestinian conflict were excluded from the scenarios used in the study. The study focused solely on criminal acts typically committed in Israel and abroad, thereby enhancing the potential to generalize our findings to other national contexts. Although ideally we would like to have gathered as much information as possible from each respondent, in order to obtain respondents' willingness to participate in the phone surveys and to maintain their focus throughout the interview we limited each questionnaire to five crime scenarios (see more details in the next section), one sexism scale (i.e., benevolent or hostile), and a small number of demographic variables. Each telephone interview lasted approximately five minutes. The wording in the questionnaire was kept as simple as possible, and the students who served as surveyors were carefully trained by the researchers to minimize potential biases. Furthermore, the questionnaire was pretested with a small number of respondents (n=80) to provide an initial test of the measures' reliability and to reveal unexpected response patterns (none were found). The questionnaire was originally written in Hebrew; however, because the Israeli population includes many Arabic- and Russian-speaking individuals, questionnaires were translated to Arabic and Russian through a translation-back-translation process. A comparison of the original questionnaire with the back-translated versions revealed very close similarity between versions.

Because of limitations involving survey length, each respondent answered either benevolent sexism or hostile sexism questions, but not both. The assignment of sexism scale was conducted randomly, and differences between the two groups in respondent demographics were negligible. Those who were assigned the hostile sexism scale comprised a slightly greater number of women (55 percent, compared with 47 percent among those assigned the benevolent sexism scale), and a slightly younger age (mean age of 35.2, compared with 36.6 among those assigned the benevolent sexism scale). No other significant differences were found in respondent demographics across the two groups.

Research Scenarios

Previous studies of crime seriousness judgments regarding a variety of offenses have often used the scenario methodology, in which respondents are presented with an evaluative task that approximates real-life situations and leaves less room for interpretative variation than do standard poll surveys (e.g., Herzog 2003a, 2003b; O'Connell & Whelan 1996; Rossi et al. 1974; Sellin & Wolfgang 1964). However, a main weakness of the simple scenario approach is that it does not allow for the systematic and simultaneous examination of the effects of multiple contextual factors on respondents' judgments (e.g., Cochran et al. 2003; Jacoby & Cullen 1999; Rossi & Berk 1997). We employed factorial design methodology, which overcomes this weakness while retaining the advantages of the simple scenario approach.

The factorial design method uses short multidimensional scenarios presented in a form that combines the benefits of controlled and randomized experimental designs and conventional surveys (e.g., Rossi & Anderson 1982; Rossi & Berk 1997). Accordingly, our crime scenarios were created by the random selection of values from each of several variables (one value per variable per scenario) until each variable was represented and a complete scenario was devised (see sample scenarios at the end of the Appendix). In each scenario, the type of crime and the offender's and victim's characteristics (gender, family status, etc.) were randomly assigned.⁵ As a result, the scenarios created represented a random sample of all possible scenarios, employing all possible values of the selected variables.⁶

Dependent Variables

Respondents were asked to provide two judgments for each scenario. First, in line with the majority of studies on judgments of crime seriousness, respondents were asked to evaluate the perceived seriousness (on a Likert scale from 1=“Not serious at all” to 11=“Very serious”) of each of the crimes described in the scenarios. To supplement the seriousness judgments, respondents were also asked to determine “the most appropriate punishment” for each of the crimes (punishment options were life sentence, any chosen number of years in prison, a monetary fine, community service, or probation: see Appendix).⁷ Research has consistently reported a high correspondence between subjective evaluations of seriousness and judgments of appropriate sentences, such that higher evaluations of seriousness are generally associated with more severe punishment recommendations (see Blumstein & Cohen 1980; Jacoby & Cullen 1999; O'Connell & Whelan 1996). To increase the uniformity of the evaluative task, respondents were instructed to base their responses on their subjective evaluation of the scenarios rather than on any personal legal knowledge they may have had (e.g., Rossi et al. 1974).

Independent Variables

In line with the study's hypotheses, hostile and benevolent sexism were two of the study's independent variables. To keep surveys brief, rather than using Glick and Fiske's (1996) full 22-item ambivalent sexism scales, we used one of their benevolent sexism subscales to measure benevolent sexism: the Complementary Gender Differentiation (CGD) dimension (“Many women have a quality of purity that few men possess”; “Women, compared to men, tend to have a superior moral sensibility”; and “Women, as compared to men, tend to have a more refined sense of culture and good taste”). Because our hypotheses regarding benevolent sexism suggest that those who are high on this scale are likely to hold a more positive view of (traditional) women, we focused on this subscale, which directly addresses women's perceived advantages over men.⁸ Furthermore, the three benevolent sexism subscales have consistently yielded high intercorrelations (e.g., Glick & Fiske 1996, 1997). Indeed, a pretest of the benevolent sexism scale with a sample of 205 undergraduate students yielded correlations of r=0.62 and r=0.55 between the CGD dimension and the two remaining benevolent sexism subscales. The correlation between the CGD subscale and the full benevolent sexism scale was r=0.79, which further suggests that the essence of this construct can be captured by the single dimension.

As for the hostile sexism component, an alternative sexism scale with fewer items—the five-item Old Fashioned Sexism Scale (OFSS; Swim et al. 1995)—was employed. Previous studies have demonstrated high correlations between this and the hostile sexism scale, suggesting that these two scales essentially measure the same construct (e.g., Glick & Fiske 1996, 1997; Glick, Fiske, et al. 2000). Sample items of the OFSS scale include “Women are generally not as smart as men,” and “It would be just as comfortable having a woman versus a man as a boss” (reverse-coded; see Glick, Fiske, et al. 2000). Both the OFSS and the CGD benevolent sexism subscale yielded marginally acceptable reliability coefficient alphas (0.68 for both).⁹

Another set of independent variables involved the gender and traditionality of the offender, which incorporated the offender's family and work statuses.¹⁰ These variables were derived from the randomly assigned factorial dimensions within the crime scenarios (see Appendix). Although judgments regarding women's conformity to stereotypical gender roles rely on a variety of variables, two of the main factors that have been shown to determine whether a woman is perceived as “traditional” are family and occupational status. Conservative gender roles typically emphasize women's role in the family and at home as stay-at-home mothers (e.g., Daly 1987, 1989; Spohn 1999). Thus in the present study, in the various scenarios presented in the questionnaire, traditional women were operationalized as being married with children and as not working in a full-time job. Nontraditional women were operationalized as those who were single and held full-time jobs.

This kind of operationalization of traditionality adheres to arguments and findings in studies of selective chivalry (e.g., Crew 1991; Kruttschnitt & Green 1984; Steffensmeier 1980). Overall, these studies have argued that the sentencing of women is often influenced by economic dependency and motherhood, attributes associated with the traditional female role. Although by definition female offenders do not represent the ideal stereotypical traditional woman, selective chivalry findings nevertheless demonstrate that there could be degrees of traditionality. Despite not being entirely “traditional,” offenders who are married with children have been shown to be judged more leniently than single female offenders without children (e.g., Daly 1987, 1989; Kruttschmitt & Green 1984; O'Neil 1999).

Control Variables

Our study centered on the differential treatment of offenders based on gender and adherence to normative sex roles; however, studies in the United States have found crime seriousness judgments to be influenced by the offender's race as well. The majority of these studies found a main effect, as well as an interaction effect, with race (Chesney-Lind & Shelden 2004; Steffensmeier, Ulmer, et al. 1998; Stolzenberg & D'Alessio 2004). Nevertheless, the effect of gender is generally not conditioned by race, and more-lenient treatment of women is found for both racial minorities and whites (Spohn & Beichner 2000). The effects of race exceed the scope of this article, but we still wished to control for it in our analyses. Because our data were collected in Israel, where race does not constitute a meaningful divider, we used ethnicity (i.e., Jews and Arabs) as the Israeli equivalent to race in the United States. Israeli Arab society has often been described as a nonassimilating, disadvantaged minority (18 percent of the general population), separated from the Jewish majority in almost every aspect of social life. Given that the Jewish–Arab divide constitutes one of the central political, social, and class conflicts in Israel, the social dynamics of ethnicity are often equivalent to the dynamics of race in countries where race constitutes a central defining variable (e.g., Herzog 2003b). We therefore controlled for offenders' ethnicity in our analyses. In addition, we controlled for several other variables that could potentially influence respondents' crime seriousness judgments: the offender's criminal record (yes/not stated), the victim's gender (male/female), and the respondent's gender (male/female), age (interval: in years), income (five categories, ranging from less than 5,000 NIS to more than 9,000 NIS), education level (interval: years of education), family status (eight categories: single, married, divorced, widowed, each of the former four categories with/without children), and employment status (four categories: unemployed, limited number of hours per week, part-time [50 percent] job, and full-time job).

Analyses

Each of our hypotheses involves a two-way interaction effect. Accordingly, we used the OLS regression procedures indicated by Aiken and West (1991). To test each of the hypotheses, we regressed the two independent variables (in line with Aiken and West, continuous variables were centered around their mean before including them in the analysis) and their interaction term on respondents' seriousness judgments. For example, to test Hypothesis 1—that respondents with high, but not low, benevolent sexism scores would judge crime scenarios to be less serious when the perpetrators were women—we regressed offenders' gender, respondents' benevolent sexism scores, and the interaction between gender and benevolent sexism on respondents' seriousness judgments. Similarly, to test Hypothesis 3a—that respondents with high, but not low, benevolent sexism scores would judge crime scenarios when committed by traditional women as less serious than when committed by men—we regressed the traditional (married and not employed) female versus male offender contrast, as well as respondents' benevolent sexism scores, and the interaction between the contrast variable and benevolent sexism, on respondents' crime seriousness judgments.¹¹ Corresponding analyses were applied for each of the hypotheses, with control variables included in all of the analyses.

Results

We started by testing a baseline model in which only the control variables were regressed on crime seriousness judgment scores (Table 1, first column). As expected, offenders with a criminal record received harsher seriousness judgments than offenders with no stated record. Similarly, in line with previous works on offender's race and ethnicity (Daly & Tonry 1997; Sampson & Lauritsen 1997), crimes where the offender was a member of the minority group (i.e., Arab) received harsher seriousness judgments. Also in line with previous works (e.g., Ruback et al. 1999; Seelau et al. 2003), the victim's gender had a significant effect on seriousness judgments, with crimes against female victims receiving harsher judgments. In addition, a number of respondent characteristics had significant effects on crime seriousness judgments. Consistent with previous findings (e.g., O'Connell & Whelan 1996), women tended to report higher seriousness judgments than men. Furthermore, harsher judgments were provided by respondents with a higher employment status (e.g., full-time rather than part-time employees) and by respondents who are married with children. A possible explanation for these findings is that individuals who lead a more conservative form of life, with full-time jobs and families, may also exhibit more conservative views toward crime. A deeper understanding of these findings would require a separate study, with these variables as its focus.

Table 1. OLS Unstandardized Coefficients (with standard errors in parentheses) of Crime Seriousness Scores on Variables of Hypotheses 1, 2, 3a, and 4a

^* p<0.05;

^** p<0.01;

^*** p<0.001.

Sexism and Chivalry

Before testing our hypotheses, we wanted to examine the relationships among benevolent sexism, hostile sexism, and respondents' demographics. In line with previous findings (e.g., Glick, Lameiras, et al. 2002), both benevolent sexism and hostile sexism correlated negatively with level of education (−0.38 and −0.06, respectively). Similarly, both sexism scales correlated negatively with income (−0.21 and −0.06, respectively). Benevolent sexism also correlated significantly and positively with respondents' age (0.19).

Because each subject received only one sexism scale (either benevolent or hostile), this sample cannot provide information about the relationship between benevolent sexism and hostile sexism. However, in an independent representative sample of 3,149 Israeli respondents (these data were collected for a different study), where both benevolent sexism and hostile sexism data were collected from all respondents, the correlation between benevolent sexism and hostile sexism was 0.32 (p<0.01). Furthermore, in line with factor analytic work among U.S. samples, the two scales in this independent sample loaded on two distinct factors.

Analyses of variance (ANOVAs) were used to compare the two sexism scales in this study across values of the respondents' demographic variables. Men proved significantly higher (p<0.01) than women on hostile sexism; in particular, single men were higher on hostile sexism than married women with children (p<0.05). In addition, fully employed women were significantly lower (p<0.01) on hostile sexism than unemployed women and men working part-time. No differences were found in benevolent sexism across the different employment groups.¹²

Next we turned to test our hypotheses. Results of these analyses are presented in columns 2–5 of Table 1. As expected, the interaction effect proposed in Hypothesis 1 was significant (β=−0.262, p<0.001, see second column in Table 1). In addition, offenders' gender (β=−0.105, p<0.01) and benevolent sexism (β=0.447, p<0.001) also yielded significant effects. The gender effect indicates that, on average, when the same crimes were committed by women, they were judged significantly less serious than when they were committed by men. Because hypotheses were not raised regarding possible main effects of benevolent sexism or hostile sexism, and as such effects are not central to this article, we do not elaborate on them.

To interpret the significant interaction effect, we dichotomized the benevolent sexism variable into high- and low-benevolent-sexism groups (one standard deviation above or below the scale's mean) and plotted the interaction effect (see Figure 1a). All the interaction effects in the following analyses were plotted using the same procedure. As can be seen, the relationship outlined supports Hypothesis 1. While among respondents who were low on benevolent sexism, crimes when committed by women were judged as only slightly less serious than when committed by men, high-benevolent-sexism respondents judged crimes committed by women as substantially less serious than the same crimes committed by men.

Figure 1. (a) Interaction Effect of Offender's Gender and Respondent's Benevolent Sexism on Crime Seriousness Judgments. (b) Interaction Effect of Offender's Gender and Respondent's Hostile Sexism on Crime Seriousness Judgments.

In testing Hypothesis 2—that respondents with high, but not low, hostile sexism scores would judge crime scenarios to be more serious when the perpetrators were women than when they were men—beyond the significant interaction term (β=221, p<0.05, see third column in Table 1), both gender and hostile sexism yielded significant effects. As before, the gender effect (β=−0.638, p<0.01) indicated that on average, when crimes were committed by women, they were judged significantly less serious than when they were committed by men. The interaction effect is plotted in Figure 1b. The relationship outlined supports Hypothesis 2. Whereas low-hostile-sexism individuals judged crimes committed by women as less serious than the same crimes committed by men, high-hostile-sexism respondents judged them more serious when committed by women than by men.

Sexism and Selective Chivalry

Our primary hypotheses concerned the interaction between sexist attitudes and selective chivalrous judgments. Results of the analyses testing Hypothesis 3a—that respondents with high, but not low, benevolent sexism scores would judge crime scenarios when committed by traditional women as less serious than when committed by men—are presented in the fourth column of Table 1. In addition to the expected significant interaction effect (β=−0.279, p<0.001), the traditional female versus male contrast also yielded a significant effect (β=−0.320, p<0.001), indicating that crimes committed by traditional women were on average judged to be significantly less serious than when committed by men. The interaction effect is plotted in Figure 2a. As can be seen, the relationship outlined presents strong support for Hypothesis 3a. While among respondents who were low on benevolent sexism, crimes committed by traditional women were judged as seriously as when committed by men, among those who were high on benevolent sexism, crimes committed by traditional women were judged to be substantially less serious than when committed by men. As expected in Hypothesis 3b, the interaction between gender/nontraditionality (nontraditional women versus men) and benevolent sexism did not have a significant effect on seriousness judgments.

Figure 2. (a) Interaction Effect of the Traditional Female versus Male Offender Contrast and Respondents' Benevolent Sexism on Crime Seriousness Judgments. (b) Interaction Effect of the Nontraditional Female Versus Male Offender Contrast and Respondents' Hostile Sexism on Crime Seriousness Judgments.

Results of the analyses testing Hypothesis 4a—that respondents with high, but not low, hostile sexism scores would judge crime scenarios when committed by nontraditional women more harshly than when committed by men—are presented in the fifth column in Table 1. The interaction term (β=0.120, p<0.01), offenders' gender/traditionality (β=0.112, p<0.01), and hostile sexism (β=−0.086, p<0.05), all yielded significant effects. The gender/traditionality effect indicated that, on average, crimes were judged to be more serious when committed by nontraditional women than when carried out by men. As can be seen in Figure 2b, the interaction effect supports Hypothesis 4a. For individuals who were low on hostile sexism, there was no difference between crime seriousness judgments of crimes committed by nontraditional women and those committed by men. On the other hand, those who were high on hostile sexism considered crimes more serious when committed by nontraditional women than when committed by men. Last, as expected in Hypothesis 4b, the interaction between gender/traditionality (traditional women versus men) and hostile sexism did not have a significant effect on seriousness judgments. A summary of the results is presented in Table 2.

Table 2. Summary of Selective Chivalry Findings (Hypotheses 3a, 3b, 4a, 4b)

To test our hypotheses another way, we reran the analyses using the respondents' choice of punishment as the dependent variable instead of seriousness judgments. With the exception of Hypothesis 4a, where the finding using the punishment variable was in the predicted direction but was not significant (p=0.079), all the hypotheses were reconfirmed.

Discussion

We set out to examine the role of individuals' attitudes toward women in explaining chivalry phenomena. Previous studies demonstrated the existence of differential treatment of male and traditional female offenders, and assumed that such differences can be explained in terms of chivalrous attitudes. Our study, by contrast, addresses the process by which such effects are likely to arise. In line with the formulations of chivalry theories, we demonstrated empirically that paternalistic attitudes toward women are one likely source of chivalry effects. But beyond this, our study challenges underlying implications of such theories—namely, that paternalistic attitudes are uniform across individuals, and in particular across men.

In keeping with chivalry theories, we found that female offenders tended to receive more lenient seriousness judgments than men and that judgments were contingent on the extent to which female offenders assumed traditional gender roles. However, the further complexity of the chivalry phenomenon is only revealed when the evaluators' underlying attitudes are considered. Our findings suggest that the extent to which crime evaluators hold benevolent or hostile attitudes toward women is a key factor in forming crime seriousness judgments. Chivalry effects were substantial among respondents who held benevolent attitudes toward women but were negligible among those who did not. Conversely, the disfavor in judgments of women was offered only by respondents who held hostile attitudes toward women.

Our findings were particularly substantial for the benevolent component of sexism. While nonsexist respondents (i.e., those scoring particularly low on the benevolent sexism scale) did not exhibit any difference in the crime seriousness judgments of female versus male offenders, individuals with high benevolent sexism scores judged traditional female offenders qualitatively differently than male offenders; male offenders received extremely high seriousness scores, while traditional female offenders received only moderate seriousness scores. The fact that the effect for hostile sexism was somewhat weaker than the effect for benevolent sexism was likely due to the smaller variance and non-normal distribution in hostile sexism. The lower variance may reflect extant social pressures against the expression of overt hostility toward women. Research on sexism has in fact suggested that paternalistic and sexist attitudes toward women have become less overt (as in the case of hostile sexism) since the early 1970s, and that other constructs, such as modern sexism or benevolent sexism, may better reflect sexist attitudes toward women today (e.g., Glick & Fiske 1996, 1997; Swim et al. 1995).

In line with studies that link socioeconomic variables with liberalism (e.g., Adorno et al. 1950; Rice & Coates 1995), higher levels of education and income were generally negatively associated with sexist attitudes. Sexism also varied across some of the other demographic categories. For example, men tended to exhibit more hostile sexism than women; in particular, single men tended to score higher on the hostile sexism scale than married women with children. Although our findings indicate that some of these socio-demographic variables are related to crime seriousness judgments, the support for our hypotheses was established after controlling for the effects of these variables. Beyond any effect that social characteristics such as age, education, and family status may have on people's crime seriousness judgments, individual differences in benevolent and hostile sexism appear to be quite meaningful for understanding differences in judgments of male versus female offenders.

Such individual differences may also be related to the type of crime prototypes that people embody (Smith 1991, 1993). According to Smith (e.g., 1991, 1993), certain crime categories appear more typical than others. For example, robberies with an armed perpetrator are perceived as more typical than those with an unarmed perpetrator. Similarly, robberies of homes are more typical than robberies of workplaces. In her studies, Smith has shown that individuals' crime judgments are influenced by the extent to which the crime is similar to their crime prototypes. Although not previously tested, it is plausible that the perpetrator's gender, and gender traditionality, may also be part of people's crime prototypes. Furthermore, our findings raise the possibility that crime prototypes may also be a function of individuals' overall attitude toward women. Thus a crime committed by a nontraditional woman may be perceived as more typical, and thus judged more harshly, by individuals who are high, rather than low, on hostile sexism.

Overall, our findings reveal a much more complex picture than has been previously portrayed in theories of differential judicial treatment. Chivalry theses rest on the assumption that chivalrous treatment results from society's paternalistic view of women. Rather than using gender as a proxy for the evaluator's view of women, we focused on what seems to be the actual source of chivalry—the particular benevolent or hostile attitudes toward women. This new focus gains additional support from the fact that respondents' gender was not significant and did not alter the results of our analyses (see Table 1).

Furthermore, in the light of our results, it is less surprising that previous chivalry studies, which have not addressed individuals' attitudes toward women, failed to produce consistent support for the theories. It is possible that in studies where chivalry was revealed, samples comprised a majority of individuals who were high on benevolent sexism. On the other hand, where chivalry was not found, and in particular where the “evil women thesis” (i.e., harsher judgments of nontraditional women) was observed, samples may have had a majority of decision makers who were low on benevolent sexism, or even high on hostile sexism.

Although our findings shed light on the phenomenon of the differential treatment of men and women, because our sample comprised members of the general public rather than judicial practitioners, they cannot be conclusively applied to judicial personnel, whose training and experience on the job distinguishes them from the average respondent. Nevertheless, the aim in this article was to test the central premise on which chivalry theories are based, whereby chivalrous treatment is based on a paternalistic approach in society at large. That said, the next step in this line of research could be to directly examine the relationship between sexist attitudes and chivalry effects among judicial practitioners in order to consolidate our conclusions here.

Future studies should consider additional variables for conceptualizing nontraditional women and additional social divides, such as race or ethnicity, and their interplay with gender and social position. Effects were found in the present study even with nontraditional women being conceptualized simply as being unmarried and fully employed. It may very well be that differential treatment across the various gender/traditionality categories will be even more extreme when race is added to the equation. In line with our current perspective, we suggest that individual differences in racist attitudes may help explain when and why differential treatment is observed across racial groups.

It is somewhat ironic that the literature on chivalry theories, which highlights the role of stereotypical views of women, appears to have adopted a somewhat simplistic and stereotypical view of men. Although one's gender may be somewhat associated with one's attitudes toward women (see, for example, social identity, self-categorization, and defensive attribution theories), the assumption that men in general hold paternalistic and sexist attitudes toward women is just as sexist as the assumption that women are weak and helpless. While such a view may have been better founded years ago, there is sufficient evidence today to suggest that gender-based attitudes have experienced substantial shifts in recent decades, and that men's and women's views of society are now far more heterogeneous, and far less based on the observer's gender, than they were in the past (Glick & Fiske 1996, 1997; Swim et al. 1995).

Although the present study focused on differential judgments of crime seriousness, our findings could explain similar patterns of discrimination in other contexts as well. Our main argument is that instead of using social categories as proxies for people's judgments, whenever possible researchers should focus on the actual attitudes that people hold in the context of the phenomenon studied.