Applying Moral Foundations Theory to the Explanation of Capital Jurors’ Sentencing Decisions

Abstract

This study applies moral foundations theory to capital juror decision making. We hypothesized that binding moral foundations would predict death qualification and punitive sentencing decisions, whereas individualizing moral foundations would be associated with juror disqualification and a leniency effect. Additionally, we considered whether moral foundations can explain differences in death penalty application between conservatives and liberals. Respondents from two independent samples participated in a mock-juror task in which the circumstances of a hypothetical defendant’s case varied. Results revealed moral foundations were strong predictors of death qualification. The binding and individualizing foundations were related to sentencing decisions in the expected ways. Supporting our contention that moral foundations operate differently across different types of cases, heterogeneity in the effects of moral foundations was observed. Finally, we found support for the hypothesis that the relationship between sentencing decisions and conservatism would be attenuated by moral foundations.

KEYWORDS:

• Moral foundations
• juror decision making
• death penalty
• capital jurors

Acknowledgments

This research was supported by internal funding from the School of Criminal Justice at Texas State University. The authors are grateful for the support of Dr. Christine Sellers, Aaron Hernandez, and Allison Fernandez in conducting this research.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes on contributors

Tyler J. Vaughan, PhD Department of Sociology and Corrections, Minnesota State University – Mankato.

Lisa Bell Holleran, PhD School of Behavioral and Social Sciences, St. Edwards University

Jason R. Silver, PhD School of Criminal Justice, Rutgers University – Newark.

Cases Cited

California v. Brown, 479 U.S. 538 (1987)

Penry v. Lynaugh, 492 U.S. 302 (1989)

Wainwright v. Witt, 469 U.S. 412 (1985)

Notes

1 Haidt (2012) proposes that Fairness be divided into two foundations: Fairness/cheating, which focuses on proportionality, and Liberty/oppression, which emphasizes freedom and equality. Because “liberty/oppression” is still considered a “candidate” for a moral foundation (Moralfoundations.org, 2018), and because measures of the split foundations are not yet available, we refer to a single-foundation conception of Fairness.

2 An additional orthogonal experimental manipulation was part of the research design in the student sample (the order of questions in the questionnaire differed across questionnaires) and this condition is controlled in all multivariate student analyses.

3 Many authors have studied juror decision making in the capital context using college students (Barnett, Brodsky, & Davis, 2004; Barnett, Brodsky, & Price, 2007; Najdowski, Bottoms, & Vargas, 2009) as well as using community members (Platania, & Konstantopoulou, 2014). Rarely have authors examined two independent samples, although Bell Holleran, et al. (2017) found small mitigating effects of child abuse and neglect in a sample of Texas college students which grew in magnitude when examining the relationship among community members reporting for jury duty.

4 Texas carried out 106 executions between 2010 and the writing of this manuscript, the most in the United States. The remaining 19 states that conducted executions during this time collectively accounted for 187 executions (DPIC, 2018).

5 Examining the CJP respondents closely, 8.5% were black females, 6% black males, 41% white females, and 39% were white males. In the online sample, these percentages were 6%, 2.4%, 50%, and 26.6%.

6 We assessed the extent to which the randomization of the experimental conditions was efficacious by testing for association between experimental condition and the demographic variables. Testing the null hypothesis that the experimental conditions were unrelated to the demographic variables individually, we failed to reject the null at the 0.05 level of statistical significance in each sample for each variable using chi-square tests of association where the demographic variables were categorical (gender, race/ethnicity, political affiliation, frequency of worship, education, and income) and analysis of variance (ANOVA) for age. We concluded the experimental manipulations were independent of the respondents’ characteristics and the randomization was effectively implemented.

7 The binding/individualizing classification, rather than the five foundation classification was used for two reasons. Theoretically, broader moral worldviews are most relevant to understanding death penalty attitudes. Second, similar to other studies (Koleva et al., 2012), there was a high degree of intercorrelation between the Care and Fairness foundations, as well as between the authority, loyalty, and purity foundations. These intercorrelations prevented the disentangling the factors from one another, leading us to collapse the care and fairness foundations, and the authority, loyalty, and purity foundations into two scales.

8 The full results of factor analytic models are available on request.

9 Although the qualification variable could be conceived as ordinal rather than nominal, the assumption of proportionality of odds or parallel regression was tested in ordered logistic models using the Brant test in which we rejected the null hypothesis in both samples and concluded the parallel regression assumption was violated. We employed the alternate, albeit more complex, multinomial logistic model as a result. A standard deviation increase in endorsement of the individualizing foundation increased the relative risk of life-only group membership over qualification by about 40% in both samples (Student z = 3.17, p < 0.001; Online z = 3.73, p < 0.001), while a standard deviation increase in the binding foundations decreased the relative risk of life-only group membership over qualification by 29% in the student sample (z = −2.49, p < 0.05) and 45% in the online sample (z = −5.68, p < 0.001). Moral foundations did not predict death-only group membership in the student sample (perhaps due to the limited number of such respondents, n = 15); however, in the adult sample, substantial effects were detected (binding RRR =2.22, z = 2.83, p < 0.01; individualizing RRR =0.566, z = −2.84, p < 0.01).

10 In sensitivity analyses grouping the death-only group with the qualified group, the results comparing life-only and qualified jurors were substantively identical.

11 Of the 1759 students that completed the survey, 1609 were considered death qualified. Of these 1609 students, 1372 provided complete data. Most of the missing observations were attributed to two variables – 143 failed to render a sentencing decision, and 69 refused to answer the political ideology question. In multinomial logistic regression models retaining respondents who failed to render a sentence as a third category (0 = life, 1 = death, 2 = no decision) the effects of the variables on the sentencing decision were substantively unchanged. However, black students and students in the other racial/ethnic category were more likely than white students not to render a sentencing decision. No other significant relationships emerged explaining why the sentencing decision was missing. In models where a missing category was added to the political affiliation variable (1 = strong liberal… 5 = missing political ideology), we observed no substantively significant changes to the conclusions of the models.

12 Initially, the effect of political affiliation on sentencing decisions was modeled treating the four point political affiliation variable as categorical. However, the differences in the effects between groups (strong liberal, liberal, conservative, and strong conservative) were constant and the effect was therefore modeled while treating the variable as continuous, although strictly speaking it is ordinal.

13 It is possible that Koleva et al. (2012) were unable to detect the effects of the Loyalty, Authority, and Purity dimensions due to multicollinearity (correlations among these foundations in their study were greater than 0.5), whereas these three distinct dimensions were collapsed into a single dimension in this study.