“Same crime, same sentence?” Disparities in laypersons’ sanctioning preferences for male and female offenders, and the link to respondent gender bias

Abstract

This study examined whether offender gender was associated with disparities in sanctioning preferences, and if these disparities were linked to implicit or explicit gender-bias attitudes. Participants (N = 316, n = 126 male, n = 190 female) completed an anonymous survey, the Implicit Association Test (IAT), the Ambivalent Sexism Inventory (ASI), and were randomly assigned to vignettes followed by sanction options on four crimes; solicitation, theft, child sexual abuse (CSA) and homicide. Half received the vignettes featuring female offenders and the other half with male offenders. Overall, participants selected significantly harsher sanctions for male offenders for three of four crimes: solicitation (d = .45), theft (d = .25), and homicide (d = .61), with a ns difference for child sexual abuse (d = .03). There was no participant gender effect. There was a significant leniency effect towards women, except for CSA. Results indicate a small effect for explicit gender stereotype for only two of the four crimes, solicitation and CSA, and no effect of implicit gender stereotype. This study offers support for the “leniency effect” in relation to women who offend, although these biases may not greatly affect sentencing preferences.

Keywords:

• crime
• gender
• bias
• Implicit Association Test
• Ambivalent Sexism Inventory

1. “Same crime, same sentence?” Disparities in laypersons’ sanctioning preferences for male and female offenders, and the link to respondent gender bias

As a construct, bias suggests that individuals may have underlying preferences against or in favour of different people depending on their group membership (Tajfel & Turner, 1986). Gender bias influences many areas in life including education (Levinson & Young, 2010), employment (Rudman & Kilianski, 2000), family roles (Eagly & Mladinic, 1994; Fine-Davis, 2016; Glick & Fiske, 1996, 1997) and criminal justice (Deering & Mellor, 2009; Mellor & Deering, 2010; Pina-Sanchez & Harris, 2020; Rachlinski et al., 2009; Rodriguez et al., 2006). Studies examining gender bias, whether they use explicit and or implicit measures, have found it to be pervasive (Herzog & Oreg, 2008; Levinson et al., 2010; Rudman et al., 2001).

Criminological research has consistently identified biases influencing arrest rates (Bailey et al., 2020; Cassidy & Rydberg, 2020), stop and search patterns (Peirson et al., 2020), and sentence severity (Brandon & O’Connell, 2018; Doerner & Demuth, 2014; Forney & Lacy, 2022; Nowacki, 2020) depending on people’s social class background, gender, or ethnicity (Levinson & Young, 2010; Rachlinski et al., 2009; Sampson & Raudenbush, 2004). Studies on crimes by women (Bates et al., 2019; Bourke et al., 2014; Denov, 2001; Maher, 2021) highlight that even serious transgressions may be minimised because of a perpetrator’s gender. In some instances, it has been argued that suspect gender may lead to non-prosecution of a case (Leahy, 2020; Leone et al., 2019). In relation to the most serious offences such as homicide (Fridel, 2019; Maher, 2021), sexual violence (Anderson et al., 2020; Bates et al., 2019; Beeby et al., 2020; Higgins & Ireland, 2009), and drug dealing (Fleetwood, 2014), evidence points toward a leniency bias in relation to women. Underpinning this leniency may be the perception that women who offend have less capacity for agency than men. It is argued that women tend to be viewed as chronic victims (Mellor & Deering, 2010; Rooney, 2020; Saradjian, 1999) rather than possible perpetrators of serious crime. In contrast to the leniency hypothesis, it has been argued that women who violate gender expectations by committing male-typical offences, experience a backlash effect (Glick & Fiske, 1997; Rudman et al., 2001) and are perceived more negatively, and treated more harshly by the public (Forney & Lacy, 2022; Murphy & Brown, 2000), and the judiciary (Beeby et al., 2020; Fridel, 2019; Nowacki & Windsong, 2019; Patterson et al., 2019).

Healy (2013, p. 558) defines agency as “the capacity to exercise will during interactions with the social world”. Agency has been implicitly understood as a masculine attribute while communality is associated with feminine traits (Eagly et al., 2020; Rudman et al., 2001; Rudman & Kilianski, 2000). Studies highlight that the gender of an individual actor influences perception of agency, risk, and aggression (Healy, 2020; Mellor & Deering, 2010; Rooney, 2020; Rudman et al., 2001). Research has indicated that women who commit serious sexual offences, and murder, are more often diagnosed with a severe mental disorder than are men (Beeby et al., 2020; Carabellese, et al., 2020; Mellor & Deering, 2010; Deering & Mellor, 2009; Putkonen et al., 2001), and have been labelled as “delicate”, and “sensitive” (Prior, 2005 p .24), more deserving of medical help than legal punishment. Women who stayed congruent to gender stereotypes and had murdered their children have been portrayed as having no agency, being mentally unwell, and elicit descriptions such as “desperately sad case” and “tragedy of enormous proportion” (O’Riordan, 2021) in media headlines. In comparison, men who have murdered their children are described as “deranged” (McMahon, 2014) and, “depraved” (Breen, 2020).

The research literature appears to demonstrate that when partialling out other variables, women are sentenced significantly more leniently than men. This study seeks to investigate whether well validated explicit and implicit gender bias measures predict attitudes for preferred sanctioning severity towards men and women who commit identical crimes. Specifically, the following three questions will be examined:

1. Is greater leniency shown towards women who offend compared to men?

2. Is greater leniency towards women shown by both male and female respondents?

3. Which form of bias (implicit versus explicit) is more predictive of leniency towards women?

2. Method

2.1. Participants

Participants comprised of a convenience sample of adults (18 years and over). Recruitment for this study was through email, social networking platforms, survey exchange sites (surveyswap.com and surveycircle.com), and Prolific (compensated, funded by the authors). A criterion for inclusion in this study was that participants lived in Ireland. A total of 452 participants responded with 316 (126 males, 190 females) fully completing the survey. Only fully completed survey responses were included in the analysis; 136 participants were excluded due to partial completion of the questionnaire. The age of participants ranged from 18 to 68 years of age (M = 35, SD = 11.57).

3. Materials and procedure

3.1. Implicit gender bias

The most well-known implicit measure is the IAT (Greenwald et al., 1998). Its claim is that it measures thoughts and feelings without the respondent required to explicitly state their views. Participants completed the five-step gender stereotype IAT (Greenwald et al., 1998), as used in previous research (Rudman et al., 2001; Rudman & Glick, 2001; Rudman & Kilianski, 2000). Participants distinguished between fifteen male names and female names by pressing right and left keys on the computer keyboard. In addition, they distinguished seven agentic words (e.g., independent, individualistic) and seven communal words (e.g., communal, connected). Participants responded to female names and agentic words (female + agency) with the right key and male names with the communal words (male + communal) with the left key; they then repeated the steps in reverse. The creation, cleaning, and scoring of the IAT was conducted with “Shinyapp” via “Iatgen” (Carpenter et al., 2019) following D-score algorithms (Greenwald & Farnham, 2000). IAT data were cleaned whereby individual trials over 10,000 milliseconds were deleted, as were any trials where more than 10% of the responses were under 300 milliseconds. A negative score indicated stereotypical associations (female and communal, male and agentic), whereas a positive difference score indicated non stereotypical associations (male and communal, female and agentic). The analyses revealed an acceptable internal consistency using a split-half procedure (Spearman-Brown correlation = .74).

3.2. Explicit gender bias

The Ambivalent Sexism Inventory (ASI; Glick & Fiske, 1996) is a well-validated instrument measuring ambivalent sexist attitudes towards women. The ASI (Glick & Fiske, 1996) is a 22-item (α = .90) scale that consists of two subscales: Hostile Sexism (HS) consisting of 11 items (α = .91) (e.g., “Most women interpret innocent remarks or acts as being sexist”) and Benevolent Sexism (BS) consisting of 11 items (α = .81), (e.g., “Many women have a quality of purity that few men possess”). Respondents indicated their agreement with a statement ranging from 0 (strongly disagree) to 5 (strongly agree). Items 3, 6, 7, 13, 18 and 21 were reverse scored. Higher scores represent more hostile and benevolent sexism.

3.3. Punitiveness towards male and female offenders

This study employed two groups of four hypothetical vignettes (Figure 1). Participants’ were randomly assigned to either group A (n = 158) or group B (n = 158). The crime was described identically for both groups, except for the name and gender of the offender. Four offences with different levels of seriousness were used in the vignettes: solicitation, theft, child sexual abuse (CSA), and homicide. Participants selected their preferred sentencing option from 1 to 6 (“In your opinion, if found guilty what sentence would be most appropriate?”) in response to the vignette from the column on the right.

Figure 1. Description of group A and group B vignettes that participants were assigned and sentencing length for the four crimes.

Note: Participants selected their preferred sentencing from 1 to 6 (“In your opinion, if found guilty what sentence would be most appropriate?”).

This study was approved by the Research Ethics Committee of the researchers’ university. Data was gathered online from February 2021 until June 2021. An anonymised computer survey link directed participants to Qualtrics, a secure survey platform.

4. Results

It was examined whether the dependent measure—length of preferred sentence—would vary with: (1) Gender of the offender, (2) Respondent gender and, (3) Implicit and explicit gender bias. To test each of these three questions a series of analyses was conducted: descriptive statistics; t-test to compare respondent mean scores; a mixed model ANOVA, to determine if leniency was shown to women and if this was by both male and female respondents; correlations to examine the relationships between variables; and a series of regression analysis to examine associations between the implicit and explicit measures on the outcomes.

4.1. Leniency effect

An independent t-test indicated that greater leniency was shown to women (n = 158) who offend compared to men (n = 158) for solicitation (t (314) = 3.99, p = .001), theft (t (314) = 2.23, p = .03) and homicide (t (314) = 4.82, p = .001). There was no significant difference for the sentencing of child sexual abuse (t (314) = .350, p = ns). Using a mixed model ANOVA, results of the sentencing for each crime were considered separately (Table 2). Three reached statistical significance (using a Bonferroni adjustment alpha level of .012; i.e., dividing a 5% level of significance by the four crime types), solicitation; (F _{1, 312} = 14.59, p = .001; d = .45), theft; (F _{1, 312} = 6.54, p = .011; d = .25), and homicide; (F _{1, 312} = 26.07, p = .001; d = .61). Child sexual abuse had no significant difference in sentencing between the groups. Results indicate significant leniency effect towards women, except for CSA.

4.2. Respondent gender effect

Respondent gender differences were examined on the IAT, ASI, and its two sub-factors: BS and HS. The effect size ranged from small to medium (Table 1). IAT scores for female respondents (M = −.174, SD = .32) showed a greater difference in gender bias than for male respondents (M = −.107, SD = .30). Male respondents scored higher on gender bias on explicit measures than did females. Table 2 displays respondent gender differences for the four crimes, with men showing moderate to high leniency towards female offenders for theft and homicide, respectively.

Table 1. Frequency, mean, standard deviation, and the intercorrelation for the implicit and explicit measures and the sentencing for the four offences, with respondent gender differences

				Male			Female
	N	M	SD	n	M	SD	n	M	SD	d	1	2	3	4	5	6	7	8
1 Solicitation	316	1.97	1.57	126	1.82	1.40	190	2.08	1.67	0.16	-	0.101	.207**	0.133	0.068	.311**	.339**	.214**
2 Theft	316	2.64	1.42	126	2.71	1.39	190	2.59	1.44	0.07	−0.067	-	.260**	.254**	0.028	0.047	−0.012	0.101
3 CSA	316	5.13	1.12	126	5.14	1.13	190	5.13	1.12	0.008	0.026	.268**	-	.292**	0.096	.230**	0.128	.292**
4 Homicide	316	4.09	1.54	126	4.21	1.58	190	4.01	1.52	0.13	0.035	.434**	.454**	-	0.026	0.092	0.081	0.084
5 IAT	316	−0.146	0.319	124	−0.107	0.308	189	−0.1740	0.3251	0.21	0.009	−0.12	0.067	0.040	-	0.051	0.086	0.001
6 ASI	317	30.43	17.75	126	37.58	17.88	190	25.77	16.08	0.69	0.141	0.050	0.012	−0.030	0.147	-	.907**	.888**
7 HS	317	13.86	11.08	126	17.90	12.29	190	11.24	9.34	0.61	0.122	0.080	0.033	−0.041	0.119	.883**	-	.612**
8 BS	317	16.56	9.11	126	19.67	9.10	190	14.54	8.56	0.58	0.113	−0.009	−0.020	−0.003	0.126	.773**	.384**	-

Note: *p < 0.05; **p < .01 (2-tailed). Correlations displayed for women (n = 190) above the diagonal; correlations for men (n = 126) displayed below the diagonal. Abbreviations: IAT, Implicit Association Test d-score; ASI, Ambivalent Sexism Inventory Total score; BS, Benevolent Sexism; HS, Hostile Sexism; CSA, child sexual abuse.

Table 2. Respondent gender difference and total mean, standard deviation, and effect size on crime sentencing for male and female offenders

	Male offenders		Female offenders
	(n = 158)		(n = 158)
	M (n)	SD	M (n)	SD	d	p	F	ηp2
Solicitation	2.32 (158)	1.75	1.63(158)	1.27	0.45	.001**	14.59	0.05
Male Respondents	2.2 (54)	1.69	1.51 (72)	1.79	0.4	-	-	-
Female Respondents	2.36 (104)	1.79	1.73 (86)	1.44	0.39	-	-	-
Theft	2.81 (158)	1.48	2.46 (158)	1.33	0.25	0.03*	6.54	0.02
Male Respondents	3.07 (54)	1.52	2.43 (72)	1.23	0.46	-	-	-
Female Respondents	2.68 (104)	1.79	2.48 (86)	1.42	0.12	-	-	-
CSA	5.15 (158)	1.12	5.11 (158)	1.33	0.03	0.72	0.123	0.001
Male Respondents	5.27 (54)	0.96	5.04 (72)	1.24	0.21	-	-	-
Female Respondents	5.09 (104)	1.2	5.17 (86)	1.01	0.1	-	-	-
Homicide	4.5 (158)	1.12	3.69 (158)	1.49	0.61	.001**	26.07	0.07
Male Respondents	4.81 (54)	1.31	3.76 (72)	1.62	0.71	-	-	-
Female Respondents	4.33 (104)	1.55	3.62 (86)	1.38	0.48	-	-	-

Note: *p < 0.05; **p < .001. Dependent variable: “In your opinion, if found guilty what sentence would be most appropriate?” (Scores 1 to 6).

Table 1 highlights the correlation matrix for respondents by gender and the results highlighted female respondents had a strong positive correlation between BS with sentencing for CSA (.292) and solicitation (.214). These results indicate that women who scored higher on BS gave longer sentences to male offenders for CSA and solicitation. Men who gave longer sentences for CSA also gave longer sentences for theft and homicide.

4.3. Gender bias effects

Results of the correlation (Table A1) showed that the IAT had a slightly positive relationship between the total ASI score (r = .11, n = 314, p = <.05) and HS (r = .12, n = 314, p = .03). There was no significant relationship between the implicit measure and BS (r = .07, n = 317, p = .19) with sentencing for any of the four offences. The ASI score showed a strong positive relationship with solicitation (r = .20, n = 316, p = .001) and child sexual abuse (r = .13, n = 316, p = .02). BS had a strong positive correlation with the sentencing of child sexual abuse (r = .16, n = 316, p = .005) and solicitation (r = .15, n = 316, p = .009). There was also a strong positive relationship between HS and sentencing for solicitation (r = .21, n = 316, p = .001). These results indicate that higher levels of explicit gender bias were associated with higher sentencing outcomes.

Four different hierarchical multiple regression analysis were conducted on each of the dependent variables: solicitation, theft, CSA and homicide (Table 3). Offender gender and respondent gender was entered into Step 1, explaining 5% of the variance for solicitation, 1% of the variance for theft, none for CSA and 7% of variance for homicide. After entering the ASI and IAT at Step 2, the total variance was 11%, F (4, 308) = 9.74, p = .001 for solicitation; 1%, F (4, 308) = 1.72, p = .14 for theft; 2%, F (4, 308) = 1.96, p = .10 for CSA; and 8%, F (4, 308) = 6.75, p = .001 for homicide. In the final analysis, only the ASI was statistically significant with sentencing on solicitation (sr = .24, p = .001) and child sexual abuse (sr = .13, p = .02). There was no effect for the implicit gender stereotype.

Table 3. Hierarchical Multiple Regression Analysis Summary Predicting Sentencing outcomes with respondent gender, offender gender, ASI and IAT scores

	Solicitation						Theft						CSA					Homicide
	B	SE B	Beta	sr	Change in R^2	R^2	B	SE	Beta	sr	Change in R^2	B	SE B	Beta	sr	Change in R^2	R^2	B	SE B	Beta	sr	Change in R^2	R^2
Step 1
Constant	2.70	.41					3.45	.39			.02	5.23	.31			.00	.00	5.84	.41			.08***	.08
Respondents gender	.17	.18	.05	.05	.05***	.05	−.16	.17	−.05	−.05		−.02	.13	−.01	−.01			−.30	.17	−.10	−.10
Offender gender	−.70	.17	−.21	−.21			−.37	.16	−.13	−.13		−.05	.13	−.02	−.02			−.84	.17	−.30	−.30
Step 2
Constant	1.60	.48			.06***	.11	3.22	.45			.00	4.82	.36			.02**	.03	5.70	.50			.00	.08
Respondent gender	.45	.18	.14	.13			−.11	.18	−.04	−.04		.10	.14	.05	.04			−.25	.20	−.08	−.08
Offender gender	−.68	.17	−.22***	−.22			−.37	.16	−.13*	−.13		−.05	.13	−.02	−.02			−.85	.17	−.30***	−.30
ASI	.02	.01	.26***	.24			.004	.00	.06	.05		.01	.00	.14**	.13			.00	.01	.04	.04
IAT	.14	.27	.03	.03			−.14	.25	−.03	−.03		.27	.20	.08	.08			.153	.27	.03	.03

Note: ASI = Ambivalent Sexism Inventory; IAT = Implicit Association Test; CSA = Child Sexual Abuse; *** p < .001, ** p < .05

5. Discussion

The aim of this study was to explore whether greater leniency was shown towards women who offend compared to men, and if that was shown by both men and women respondents. The study also looked to examine if traditional implicit and explicit gender stereotype measures were predictive of leniency towards women and if so, which measures? Based on the analysis of this study the results demonstrate that there was a leniency effect towards women, except in relation to child sexual abuse. Homicide had the strongest relationship with gender bias, with men receiving significantly longer sentences. This is in line with previous research (Bailey et al., 2020; Beeby et al., 2020; Cassidy & Rydberg, 2020; Deering & Mellor, 2009; Fridel, 2019; Maher, 2021; Mellor & Deering, 2010; Pina-Sanchez & Harris, 2020; Rachlinski et al., 2009) which found that gender bias influences decision making and perception of risk. Comparable with Higgins and Ireland (2009) and Mellor and Deering (2010), this study found that there were no differences between the participants, as both men and women respondents demonstrated the same general biases in sentencing male and female offenders. Casey and O ‘Connell (1999) and Forney and Lacy (2022) both argued that the type of crime is a far greater determinant of sentencing severity than is bias.

The implicit and explicit measures both indicated the presence of gender-stereotypical attitudes, with male participants demonstrating higher gender bias on the explicit measures. These concur with previous research findings: women’s explicit gender attitudes are more egalitarian than men’s (Glick & Fiske, 1996; Rudman et al., 2001). However, female participants showed slightly higher implicit gender bias in relation to female agency than male respondents; this did not relate to sentencing severity. The findings of this study indicate that higher scores on the ASI scale were associated with higher sentencing for men who commit homicide, and higher sentencing for women who commit solicitation. The results showed that there was no effect of implicit gender stereotype on any of the crimes. There was a small positive effect for the explicit gender stereotype measure on sentencing preferences for solicitation and child sexual abuse.

Nosek and Smyth (2007) argue that the IAT measures a related but distinct construct from the multidimensional model of the ASI (Glick & Fiske, 1996). A critique of the IAT has been its strength in predicting behaviour and whether the measure can adequately predict constructs that have real world influence. Evidence shows that the IAT measures fluctuate between test and re-test and it has been suggested it measures “states rather than traits” (Brownstein et al., 2019). Some researchers posit that measures of explicit bias such as the ASI (Glick & Fiske, 1996) have more validity than measures of implicit bias.

Potentially, the benevolent sexism perspective implies a chronic female victim eliciting chivalry from men. In such a framing, women are perceived as weak, sensitive, and needing to be protected from themselves. Alternatively, hostile sexism offers the perspective that women who violate gender expectations are deficient and are in need of medicalisation. Therefore, the perception of agency for women tended to be diminished, regardless of whether respondents prescribed to a benevolent or hostile perspective of sexism.

This study has several limitations that should be considered when interpreting the results. Firstly, this study has a modest sample size that uses lay persons and not criminal justice officials who oversee sentencing. However, this study looked to explore and examine if gender bias explained the disparities that exist in relation to our perception of offending behaviour and its harm. Secondly, the results are from hypothetical rather than real world examples. A future study to explore sentencing disparities would benefit from recruiting judges and real world cases to explore the factors that determine and influence judges’ decision making in relation to the sentencing of men and women who offend. Thirdly, there is an absence of socio-demographic information about the sample, and therefore it is difficult to establish that it is representative of the general population.

6. Conclusion

The current study challenges the assumption that the public are free from gender bias in relation to the criminal justice system. Results revealed that there are substantial variations of perception between men (perceived as agentic) and women (perceived as communal) who commit the same crime. These variations can be explained by both implicit and explicit bias, indicating that gender stereotype attitudes are pervasive. However, in relation to offending behaviour and sentencing the implicit gender stereotype measure had no effect and the explicit gender stereotype offered a small effect on two of the four crimes. These results provide an interesting addition to the study of leniency bias for sentencing preferences. Overall, this study has established a strong leniency bias, however, this bias is not explained by individual differences in measures of implicit and explicit bias. Further research to investigate the causal affect for such variations in sentencing would greatly benefit how and why there is a “leniency effect” towards women who offend.

Open Scholarship

This article has earned the Center for Open Science badge for Open Data. The data are openly accessible at https://osf.io/dfcem/, https://doi.org/10.17605/osf.io/dfcem and https://osf.io/9pvn6.

Acknowledgements

We would like to acknowledge a university seedfunding grant.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that supports the findings of this study are openly available in OSF at https://osf.io/dfcem/, https://doi.org/10.17605/osf.io/dfcem and https://osf.io/9pvn6.

Additional information

Funding

The authors received no direct funding for this research.

Appendix

Table A1. Correlation of the Implicit Association Test, Ambivalent Sexism Inventory with its two sub-factors Benevolent Sexism and Hostile Sexism and the four offences

	1	2	3	4	5	6	7	8
1. IAT	-
2. ASI	.113*	-
3. BS	0.073	.851**	-
4. HS	.120*	.902**	.541**	-
5. Solicitation	0.038	.204**	.146**	.207**	-
6. Theft	0.024	0.058	0.065	0.040	0.039	-
7. CSA	0.086	.131*	.157**	0.081	.140*	.263**	-
8. Homicide	0.038	0.057	0.063	0.040	0.091	.327**	.359**	-

Note: *p < 0.05; **p < .01 (2-tailed). Abbreviations: IAT, Implicit Association Test d-score; ASI, Ambivalent Sexism Inventory Total score; BS, Benevolent Sexism; HS, Hostile Sexism; CSA, child sexual abuse.