Moral intuitions, punishment ideology, and judicial sentencing

ABSTRACT

Considerable research examines discretion in judicial sentencing. However, little is known about the role of moral values or ideological beliefs in judicial sentencing decisions. The current study draws on insights from moral psychology to propose a model of judicial decision-making in which moral intuitions may inform sentencing both directly and indirectly via ideological beliefs about punishment (including general punitiveness and concern for offenders). We test this model using a statewide survey of Pennsylvania Common Pleas Criminal Court judges (N = 132), which included hypothetical sentencing vignettes. Results indicate that although moral intuitions were related to punishment ideology, moral intuitions were largely unrelated to judicial sentencing decisions, with a few exceptions. We interpret the results as suggesting that while moral and ideological preferences may be relevant under some circumstances, the role of morality in judicial decision-making may be constrained by legal or organizational factors.

KEYWORDS:

• Judicial sentencing
• moral intuitions
• punishment ideology

Disclosure statement

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

Supplemental data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/0735648X.2023.2248085

Notes

1. Although different theories of moral cognition exist in the moral psychological literature, key insights from MFT – that moral cognition is intuition-first and pluralistic, and that moral intuitions can be organized into individual- and group-centered domains – tend to be congruent across different perspectives (see, e.g., Janoff-Bulman and Carnes 2013; Schein and Gray 2018). We focus on MFT in the current study given that its utility (and that of its measurement instruments) has been demonstrated in work on lay sentencing decisions (e.g., Silver 2017).

2. It is important to note that in this context, we are referring to individuals’ personal beliefs about moral wrongness, rather than judgments of illegality (a different form of ‘wrongness’). Whereas judges may vary in the extent to which they personally view different types of actions as morally wrong, we would expect judges to largely agree on the extent to which different actions are illegal.

3. This categorization is also consistent with other ways of typifying offenses, such as mala en se vs. mala prohibita offenses; offenses that have victims vs. victimless offenses; and moral vs. conventional offenses.

4. As noted previously, individualizing moral foundations may also promote retributive punishment toward offenses against individuals, consistent with the observation above that perceived violations of specific moral values may trigger outrage and preferences for harsher punishments (Giacomantonio et al. 2017; Silver 2017).

5. Although power may be reduced for ordinal logistic regression (used to predict sentencing) relative to OLS regression (Taylor, West, and Aiken 2006), sensitivity analyses using different modeling strategies, discussed in the following section, returned similar results.

6. Small numbers of missing values were imputed for the variables measuring sentencing; thus, the analytic samples used in the sentencing models ranges from 128 to 131.

7. Because the vignettes were designed to ‘trigger’ specific moral foundations, the fact patterns are somewhat atypical (e.g., most cases of resisting arrest do not involve employer-employee disputes). However, our pretesting suggested the vignettes provided adequate information with which to make criminal sentencing decisions.

8. Specifically, for the Care vignette 1 = incarceration <60 months, 2 = incarceration 61–72 months, 3 = incarceration 73+ months; for the Fairness vignette 1 = no sentence, restorative sanctions, fines, or probation, 2 = incarceration <6 months, 3 = incarceration 7+ months; and for the Loyalty, Authority, and Sanctity vignettes 1 = no sentence, restorative sanctions, or fines, 2 = probation <18 months, 3 = probation 18+ months, IP/RIP, or incarceration.

9. Departures for the individualizing vignettes were overwhelmingly downward, while departures for the binding vignettes were overwhelmingly upward (as could be expected given that the sentencing ranges for the binding vignettes included no incarceration while the ranges for the individualizing vignettes did). We therefore predicted downward departures for the former and upward departures for the latter.

10. These items were adapted from research assessing punitive attitudes among the public. A key difference in evaluating general punitiveness among judges is that judges work within the criminal justice system and are directly affected by such policies. For example, mandatory minimum sentences constrain judicial discretion in addition to increasing sentence severity across the board, suggesting that opposition to this policy among judges could reflect preferences for leniency or preferences for discretion. As such, we ran supplemental analyses using death penalty support only as a measure of general punitiveness. All results were similar in direction and magnitude to the main results. The significance of two effects changed: death penalty support alone was not significantly predicted by binding moral foundations (b = .245, p = .089); and death penalty support was significantly associated with harsher sentencing in the Loyalty/betrayal vignette (b = .516, p = .014).

11. The full MFQ includes two parts asking respondents to rate the relevance of various concerns to their moral decision-making (Part 1) and to rate their agreement or disagreement with various statements corresponding to each of the moral foundations (Part 2). Given space restraints and the potentially sensitive nature of some questions for judges (e.g., regarding loyalty to the United States), we followed prior research indicating that each part of the scale can be used separately (Graham, Haidt, and Nosek 2009) and included only Part 1 items in the survey.

12. We also measured race/ethnicity. Given that only 8 respondents identified as racial/ethnic minorities, however, we excluded this variable from the models. Supplemental analyses including the measure indicated that race/ethnicity was not significantly associated with any outcomes and its inclusion did not substantively change the results.

13. Philadelphia has approximately 1.6 million residents. The next most populous cities in Pennsylvania are Pittsburgh, with approximately 300,000 residents, and Allentown, with approximately 125,000 residents.

14. Although OLS regression is typically considered appropriate for dependent variables with five or more categories (e.g., Snijders and Bosker 2012, 310), there is some debate over this point. Therefore, we also recoded each punishment ideology variable to have three categories and estimated the outcomes with ordinal logistic regression analyses. Results from these sensitivity analyses were substantively the same as those in the main analysis.