Does Moral Conviction Really Override Concerns About Procedural Justice? A Reexamination of the Value Protection Model

We were not able to generate a dataset or analyze a correlation matrix that included all of the variables listed in Skitka’s original correlation matrix (2002, p. 593) because the matrix was not positive definite. We therefore broke up the data by policy domain for both the data generation and the structural equation models. Additional analyses revealed that this is most likely due to multivariate linear dependencies, where several variables together perfectly predict another variable. Specifically, a non-positive definite matrix occurs when trying to predict outcome fairness in the civil rights domain with pre-threat procedural fairness and the three moral mandates. By entering moral mandates two at a time, we found that moral mandates in abortion and homosexual civil rights accounted for 70% of the variance in outcome fairness ratings of a court decision granting rights to homosexuals; moral mandates in immigration and homosexual civil rights accounted for 51% of the variance in outcome fairness ratings; and strikingly, moral mandates in abortion and immigration accounted for 30% of the variance in outcome fairness ratings of a homosexual rights policy decision. Such results suggest that the operationalizations of the outcome fairness and moral mandate variables in the study may have conceptually overlapped.

Another way to investigate whether this measure was indeed tapping outcome fairness is to see how judgments of the procedures at Time 1 relate to judgments of outcome fairness at Time 2 when the perceiver does not have a moral mandate. In this case, we would expect to find the generally robust fair-process effect (Lind & Tyler, 1988; Thibaut & Walker, 1975). We selected participants who had a “moral mandate” score between −1 and 1 of the original mandate measure for each domain. The correlations between pre-threat procedural justice ratings and post-threat fairness ratings showed that evaluations of outcome fairness at Time 2 are almost completely unrelated to evaluations of the procedures at Time 1. This offers further support for the speculation that the operationalization of this measure was not tapping perceptions of fairness.

Our analysis thus only uses one of the three dependent variables evaluated in Skitka (2002). While she uses post-threat outcome fairness, procedural fairness, and moral outrage, we only examine post-threat procedural fairness. As reported in Skitka (2002), we did not find effects of pre-threat ratings of the procedures on post-threat outcome fairness. However, we were hesitant to conclude that pre-threat judgments did not affect post-threat outcome fairness because (1) the operationalization of outcome fairness was empirically indistinguishable from that of moral outrage, and (2) post-threat outcome fairness was not associated with pre-threat fairness ratings even for those without a moral mandate, as reported in Note 2. Using only post-threat procedural fairness ratings is a clear limitation of this study, but we address this issue in Study 2. In that study, we find that the evaluations of the procedures are related to final ratings of both outcome and procedural fairness.

According to Skitka and Mullen (2002a), 17.3% of the participants were classified as having a moral mandate that Elian should stay in the United States (“anti-decision mandate”), 37.5% of participants had a mandate that Elian should return to Cuba (“pro-decision mandate”), and 45.3% of the participants could be classified as having “weaker or no moral mandate (p. 1423).” Thus, we re-ran our analysis with moral mandate dummy-codes based on these exact guidelines. For the anti-decision mandate dummy-code, the bottom 17.3% of the data were assigned “1” and the middle 45.3% were assigned a “0”. Likewise, the top 37.5% of the data were assigned a “1” for a pro-decision mandate dummy-code and the middle 45.3% were assigned a “0”. In two regression models, we examined the effects of procedural justice, moral mandates (either pro- or anti-decision), and the interaction term on post-resolution measures. The results from these analyses were completely consistent with the results reported in Table 4, such that pre-raid procedural justice influences all of the post-resolution measures and these effects were not qualified by moral mandates.