Beyond Ability to Pay: Procedural Justice and Offender Compliance With Restitution Orders

Threats to the Fair Treatment–Payment Relationship

Offenders generally perceive economic sanctions to be a severe punishment, primarily because they have few legitimate sources of income. For example, McClelland and Alpert (1985) and Apospori and Alpert (1993) found that recent arrestees believed a US$5,000 fine was about as severe as spending a year in jail. In one survey of offenders, two thirds of the respondents said it was difficult for them to pay their sanctions due to income-related reasons (Ruback, Hoskins, Cares, & Feldmeyer, 2006). In addition, sanctions tend to disproportionately affect poor and minority offenders by further decreasing their limited financial power (Harris, Evans, & Beckett, 2010).

For these reasons, several scholars argue that the current system of economic sanctions—particularly fees, fines, and other court-related costs—represents impractical public policy (Beckett & Harris, 2011; Harris et al., 2010; Harris, Evans, & Beckett, 2011). Unlike compliance with police officer requests, which typically does not involve monetary loss by the participant, restitution compliance involves the payment of economic sanctions when the participant may not have the ability to do so. To address nonpayment due to inability to pay, several European countries use a sliding scale system known as day fines, whereby economic sanctions are progressively based on the ability of offenders to pay the sanctions, similar to an income tax (Hillsman, 1989).

However, restitution, which is intended to recompense victims for the harm suffered rather than recoup administrative costs for the criminal justice system, operates differently. For example, in some U.S. jurisdictions, an offender’s ability to pay may be considered in restitution imposition decisions, but in Pennsylvania, restitution must be ordered in full “regardless of the current financial resources of the defendant, so as to provide the victim with the fullest compensation for the loss” (18 Pa. C.S.A. § 1106). Therefore, the biggest threat to the link between procedural fairness and payment, particularly in places such as Pennsylvania, may be the inability of offenders to comply with court-mandated restitution orders.

As noted, paying restitution can increase the likelihood that an offender will successfully desist from crime. Inability to pay is perhaps the single most important factor that could block payment and, by extension, desistance. In addition to an inability to pay, there are differences among offenders in terms of how open they are to treatment, that is, how “‘ready’ they are to engage in a change process” (Casey, Day, Howells, & Ward, 2007, p. 1428). One aspect of treatment readiness is self-blame or acceptance of responsibility for the harm caused by a crime (Day et al., 2009). Individuals who express more self-blame and remorse for their actions are more likely to engage in treatment programs and should similarly be more likely to pay a greater amount of restitution.

Alternatively, there is the possibility, derived from Martin Seligman’s work on self-blame and attributional styles (Buchanan & Seligman, 2013; Peterson, Schwartz, & Seligman, 1981), that self-blame may be counterproductive. Characterological self-blame is related to self-esteem, unmodifiable personal traits (i.e., personality and character), and a perceived deservingness for past negative events. Individuals who blame themselves based on characterological flaws are more likely to exhibit symptoms of depression (Peterson et al., 1981). Thus, individuals who blame themselves for their offending because they believe they are “bad people” may be less likely to pay, because payment serves as a reminder that their offending is the result of immutable personal characteristics.

Restitution in Pennsylvania

Because policies and practices related to restitution imposition and payment are so varied across jurisdictions (Haynes, Cares, & Ruback, 2015), it is important to provide a brief contextual discussion of restitution practices in Pennsylvania. The defendant’s ability to pay is not a consideration when deciding whether and how much restitution to impose (Ruback, Ruth, & Shaffer, 2005). Restitution is mandated by law in all eligible cases, which are cases where a quantifiable loss or harm was suffered by an identifiable party, including a person, business, or insurance company. However, restitution is ordered only in about two thirds of eligible cases despite the legal mandate. In terms of payment, about half of cases receive no restitution and about a quarter receive payment in full.

Any and all restitution paid goes to the specific victim(s) listed in the case (Ruback et al., 2005). Offenders are also required to pay US$60 into a general victim fund, but this amount is listed as a “fee” that is separate from the ordered restitution amount. That is, 100% of any restitution paid goes to the victim of the crime. It follows that offenders’ attitudes about the payment of restitution reflect payments to the victim(s) in their cases rather than payment to some amorphous class of victims more generally.

Prior Experiment

Ruback et al. (2014) tested the effect of sending letters to probationers on restitution payment. The population of current and former probationers who still owed restitution and who were delinquent on payments were randomly assigned to one of four conditions. The first condition received letters that provided information about the amount of restitution ordered by the court, the amount that had been paid up to the point that the letter was mailed, and the amount that still had to be paid. Probationers in the second condition received letters that provided a justification or rationale for payment (e.g., it helps victims). The third condition combined both information and rationale for payment into a single letter. The control condition received no letters. Letters were mailed monthly for 6 months, and any payments were recorded both between letters and after the mailings ceased. Results indicated that probationers who received informational letters, compared with the other three conditions, paid both more restitution and made more payments during the observational follow-up period.

Sample and Procedure

All participants who were enrolled in the prior experiment were solicited for participation in a follow-up survey (N = 775). To maximize the number of respondents, we used the Dillman (2007) survey method, which involved up to four mailings: (a) an initial contact letter briefly describing the study; (b) a survey with a one-dollar incentive attached; and, if necessary, (c) two follow-up surveys.

Of the 775 experimental participants, 530 (68%) were successfully contacted during the follow-up period with the remainder returned undeliverable, usually because the participant changed addresses. A total of 149 respondents (19% of the total; 28% of those located) completed the survey: 101 returned the first survey, 29 returned the second survey, and 19 returned the final survey. Given the transient nature of the population, the response rate is consistent with other surveys of offenders (Ruback et al., 2006; Wetzel & Lowenkamp, 2011). The survey, which took about 30 min to complete, consisted of a battery of questions spanning several topics. Respondents were mailed a US$20 Walmart gift card upon completion.

Five months after the survey, participants received a debriefing letter about the combined experiment and survey research and were informed that they could withdraw their data if they wished. Four respondents chose to withdraw their data and were eliminated from the data files.

Multiple Imputation, Survey Weighting, Data Quality, and Consent

Because of missing data, data presented in this article were subjected to multiple imputation. ¹ Consistent with recent guidelines, we followed a series of design decisions for the imputation process: (a) all available measures were used as predictors in the imputation models, including the dependent variables, and a large number of datasets (M = 40) were imputed (Graham, 2009; White, Royston, & Wood, 2011); (b) values for categorical and ordinal variables were not rounded post-imputation, as doing so would bias the results (Wu, Jia, & Enders, 2015); and (c) no skewed variables were transformed prior to imputation (Graham, 2009). In addition, we constructed scales after imputing scores on the component items rather than the reverse.

Furthermore, data were adjusted for survey nonresponse (Grau, Potter, Williams, & Diaz-Tena, 2006). First, we ran a logistic model predicting the odds that a respondent could be located (i.e., a letter was not returned as undeliverable) based on available information (age, gender, race, and case characteristics). Next, given that a respondent could be located, we predicted the odds that a respondent would return a completed survey. Finally, we multiplied the two inverse probability weights together to create an overall weight. The weights used in all analyses represent the combined location and completion weights.

In addition, four respondents chose to remove themselves from the study, and we withdrew another case that was determined to be an extremely influential outlier (amount of restitution paid in excess of US$30,000). The final sample is 144 of the 149 participants who completed the survey.

Measures

In the following section, we describe only the measures directly examined in the present study. For composite measures, all indices represent the mean of component items unless specified otherwise, with items reverse-scored when appropriate. To our knowledge, Stata currently lacks the facilities to conduct principal components analysis on multiply-imputed data. Therefore, we were constrained to additive scales and relied on reliability coefficients for their construction.

Dependent Variables

We examined two dependent variables drawn from each respondent’s court case files: the amount of restitution paid and the number of months for which restitution payments were made. In Pennsylvania, all case information is public record and freely accessible on the Internet (via https://ujsportal.pacourts.us/).

Independent Variables

Our primary independent variable reflects procedural justice, or fair treatment by probation officers. We adapted five items from Jackson et al.’s (2012) survey. The original items measured beliefs about police, which we modified to reflect beliefs about the extent to which probation officers (a) “treat me with respect,” (b) “are helpful,” (c) “make fair decisions when handling problems,” (d) “usually act in ways that are consistent with my own ideas about what is right and wrong,” and (e) “can be trusted to make decisions that are right for probationers.” Responses ranged from 1 = not at all to 7 = very much. The five items were combined into a single scale.

Threats to the Treatment–Payment Relationship

Control Variables

Analytic Plan

First, we present the descriptive statistics for the analytic sample, including means and reliability statistics for all the variables used in the regression models. Second, we present the results of a linear regression model predicting restitution payments during the payment tracking period as a function of procedural fairness and the other predictors. Because there is reason to suspect that ability to pay may condition the effects of some other variables (e.g., intent to pay means little if one has no financial resources), we also test for interactions between income and procedural fairness and the potential threats to the fairness–payment relationship. Third, because payments were subject to overdispersion (i.e., 30% of offenders made no payments), we present a negative binomial (count) model that predicts the number of months in which payments were made. Finally, given that the average offender paid very little restitution, we present results from the survey regarding potential mechanisms that offenders said would or would not be likely to increase their payment.

Limitations and Implications for Future Research

Limitations of this study relate to the size and representativeness of the sample. Because offenders are a highly mobile population, it was difficult for us (and for probation staff) to locate the probationers. Nearly a third of the total sample was definitively lost because we received an undeliverable letter from the post office, and an unknown number of additional surveys were delivered but may have been discarded by the tenants or other members of the household. Thus, the responses we received may not be representative of the entire sample, although we used statistical methods, by weighting for nonresponse, to control for both those individuals who were contacted and those individuals who responded.

The modal reason for failure to deliver was that the respondent had moved and left no forwarding address. In addition, four participants contacted us and explicitly requested that we avoid all subsequent communication (e.g., “the crime is in my past and I am over it”). It is imperative that the field of criminology attempt to identify the best channels through which to maintain contact with a highly mobile population that may wish to sever ties with agencies after being released from supervision. Despite multiple follow-ups, we obtained just 144 usable responses, which are sufficient to identify only large effects (Cohen, 1992). Nevertheless, we identified several robust and significant findings.

An ongoing challenge is to achieve balance between research expenses and an adequate sample size for analysis. At around 12 pages of text with a completion time of roughly 30 min, it is possible that the survey design contributed to the low response rate among respondents. However, we attempted to offset the potential burden on respondents’ time by including a US$20 incentive for completing the survey. At a cost of roughly US$2,700, the use of repeated follow-up mailings for non-responders produced an additional 48 respondents, an increase of nearly 50% more than the first mailing alone.

In addition, we measured ability to pay as household income, but income is clearly only one piece of an individual’s ability to pay economic sanctions. On one hand, household income does not capture outstanding debts, necessary monthly expenses, and nature of employment that may decrease an individual’s capability of paying. On the other hand, external sources of support—such as family and friends who provide housing or other financial aid—which may inflate an individual’s ability to pay, are not captured by household income. It is possible, for example, that third parties provided money to probationers to make restitution payments. Future research should consider the possibility that ability to pay takes numerous forms beyond income.

Temporal ordering is a concern due to the cross-sectional nature of the study and timing of the survey administration. To preserve experimental integrity by not alerting participants to the nature of the study, we withheld the survey until after the experiment had ended and recorded payments prior to the collection of attitudinal data. Typically, however, attitudes are recorded prior to behavioral measures. Supplemental analyses revealed no significant differences in attitudes between experimental conditions (including the control group), suggesting that it is unlikely that the experiment changed attitudes (available online).

It is also possible that the causal order is actually reversed, that is, the payment behaviors increased perceived fair treatment; our cross-sectional design cannot disentangle the temporal order of effects. Based on a careful reading of prior literature, we suspect that this possibility is unlikely. According to the theoretical model of procedural justice (Tyler, 2006), perceived fairness of treatment increases the perceived legitimacy of the criminal justice system, which in turn increases behavioral compliance with the law. Longitudinal research has repeatedly confirmed that perceived fair treatment predicts subsequent compliance behavior (Murphy & Tyler, 2008; Tyler, 2006; Wenzel, 2002), though to our knowledge, there are no studies supporting the opposite conclusion. However, in the absence of a longitudinal test, we cannot definitively rule out the possibility for reverse causation in this study.

Finally, the sample in this study may not generalize to other populations. Reflecting the demographic characteristics of offenders in the county, participants in this study were disproportionately white. Therefore, future research should attempt to replicate the findings in more diverse populations.