Harm Is Key to Judgments That Stealing Is Immoral

In Study 1, we presented American undergraduates from a range of political orientations with vignettes describing resources that were taken for redistribution by a third party without the owner’s permission. In our effort to test boundary conditions, we accounted for the initial resource inequality by explaining that luck left some individuals with more resources than others. In a fully factorial within-subject design, we varied the harm done to owners, the needs of the recipients, and whether an authority or an individual took the resources.

Consistent with theoretical perspectives highlighting the role of harm in moral transgressions (e.g., K. Gray et al., 2012; Turiel, 2015), we hypothesized that both owner harm and recipient need would affect acceptability judgments. Specifically, we expected that when the recipient was not in need, and the owner experienced harm from losing resources, participants would judge taking resources as unacceptable. In contrast, when the recipient was in need and the owner was not harmed by the loss of resources, we expected that participants would judge taking the resources as more acceptable. Moreover, inasmuch as labeling an act as stealing may partially reflect social-conventional considerations centered on laws, authority, and social consensus, we hypothesized that when an authority took resources without permission, it would be characterized as stealing less often than when an individual did so. Using the criterion judgments of justifications and rule independence, we further expected that when there was minimal harm to the owner, taking property without permission would not be condemned on the basis of moral concerns or judged as wrong when allowed by the law. That is, we anticipated patterns indicating that taking without permission would not be not deemed consistently immoral, but that moral considerations are activated by the harm associated with such acts. In addition, we conducted exploratory analyses that examined these judgments across the political spectrum. We report how we determined our sample size, all data exclusions, all manipulations, and all measures in both studies. Our full hypotheses, data collection and analysis plan are here.

Participants were 201 students (after excluding, as preregistered, 86 who did not complete more than half of the survey) recruited from an undergraduate psychology participant pool at a Catholic University in the Eastern United States (see Table 1 for participant demographics). The study meets relevant ethical guidelines.

To determine our required sample size, we conducted a power simulation using pilot data to examine our expected interactions in Mixed Effects Models. Although all hypotheses were two-way interactions or main effects, we tested our three-way within-subject design with effect sizes based on pilot data. The three-way interaction was evident in 99% of 1000 simulations with 46 participants (see our preregistration for more information). We preregistered a minimum sample size of 200 to address exploratory questions as well as confirmatory hypotheses. Finally, we conducted a sensitivity analysis to examine the power of our hypothesis that required the most power, the harm x need interaction. With simulations based on our final sample size of 201, and using the design of Study 1 and the parameters (including effect size) from our data, we simulated 1000 datasets. We tested the power to detect the interaction over those simulations, which resulted in a sensitivity of 1.

In an online survey, participants read 12 vignettes that described acts of taking without permission to benefit a third party. In the vignettes, people in a town needed to store food resources to make it through the winter, although everyone in the town had worked hard, by luck, their yield varied. All town members had enough food to make it through winter alive, but some people ended up with more surplus barrels of food. Having five extra barrels of healthy food was described as allowing them to make it through the winter comfortable and healthy. Each vignette identified three characters: an owner, whose food is being redistributed, a recipient, who benefitted from the redistribution, and a taker, who redistributed the barrels of food. To vary perceptions of harm, owners started with 100, 10, or 5 barrels of healthy food and had 5 taken against their wishes leaving them with: extra (i.e., 95 barrels, much more than enough); enough (i.e., 5 barrels, just enough); or causing harm (i.e., 0 barrels, not enough). To vary perceptions of need, recipients started with either 0 (need) or 5 (no need) barrels of healthy food. To vary perceptions of authority, vignettes described either the town council or an individual taking the resources, in both conditions, the taker gave the resources to a third-party. Protagonists’ names were gender-matched to participants or assigned randomly for non-binary participants. Participants judged all conditions resulting in 12 trials in which Recipient Need (2: present or absent), Harm to owner (3: extra resources, enough resources, harm), and Authority Status (2: town council or individual) were crossed within participant. The order of these trials was randomized within participant. See the Supplemental Online Materials (SOM) for details.

For each vignette, participants responded to three multiple choice, manipulation checks assessing whether participants’ perceptions of harm, need, and authority reflected the details depicted in the vignette (“How much food was [Owner] left with?”), need (”Before he received the food, did [recipient] have enough to make it through the winter comfortable and healthy?”), and who did the taking (“Who took the healthy food?”). Multiple attempts were allowed, but a correct response was required to move on. Following the manipulation checks, participants judged whether the act was acceptable (”Is it OK or not OK for the town to take 5 barrels from [Owner] in this situation?“) on a 6-point scale ranging from not at all OK (1) to definitely OK (6). Next, the participant judged the extent to which six justifications reflecting moral concerns (”The owner/recipients’ need for food,” “Making things fair to everyone”), conventional concerns (“The legality/illegality,” “Whether the [taker] has the authority to take the food”), or property rights influenced their acceptability judgments (from not at all(1) to a lot (5)). They were then asked if the act should be labeled as “stealing” on a 6-point scale ranging from not at all stealing (1) to definitely stealing (6). Finally, in the individual condition, we asked about Rule Independence (i.e., “If the town passed a law permitting taking without permission, would [taker’s] actions be OK or not OK?”) on a 6-point scale that ranged from not at all OK (1) to definitely OK (6) to assess judgments of taking without permission when the law allowed it. For items assessing acceptability, labeling acts as stealing, and rule independence, although participants responded to a scale labeled with the numbers 1 to 6, their responses were subsequently rescaled to facilitate data interpretation, making zero the midpoint and -2.5 to 2.5 the anchors.

We used the lme4 package (2015) in R (2021) to fit four linear mixed-effects models on judgments of i) acceptability, ii) the role of authority status, iii) importance of various justifications, and iv) rule independence. In each case, a chi-square test indicated that the planned mixed-effects model (based on theory) was an improvement on the intercept-only model, ps < .001. We report hypothesized main effects even when they were qualified by interactions. Follow-up tests were corrected using Holm’s correction. All confidence intervals are 95%. For exploratory research questions, we only report descriptive statistics and 95% confidence intervals.

Our first analysis begins to assess the role of perceived harm in judgments of the acceptability of acts of taking without permission. As hypothesized, there was an interaction between Harm to Owner x Need of the Recipient, F(2, 2199.1) = 5055, p < .001, η_p² = .04. Post hoc tests showed that when there was harm to the owner, differences in benefit to the recipient were not significant (see acceptability judgments in column 1, Table 2). However, when the owner was left with enough (column 2; Table 2), or extra resources (column 3; Table 2), then taking without permission was judged as more acceptable when the recipient was in need, as Figure 1 shows. Thus, harm was a central consideration in participants’ evaluations of taking without permission.

To further explore whether participants judged that taking without permission was acceptable in the No Harm to the Owner (Extra condition)/No Recipient Need condition, we calculated 95% confidence intervals on the mean for that condition compared to the midpoint. If the CI for the one-sample t-test was above the midpoint of 0, this would suggest that participants judged this case of taking without permission as acceptable. Indeed, participants judged that taking without permission was acceptable when the scenario did not depict harm to either party, M = .20, CI [.07, inf.].

Exploratory correlational analyses were conducted to examine whether evaluations of the act of taking without permission were the same as labeling an act “stealing.” Correlations (conducted separately for each cell) indicated moderate associations between labeling an act as stealing and judging it to be okay (rs = -.38 to -.61), indicating that the label of stealing is not the same construct as the judgment of taking without permission.

We hypothesized that labeling an act “stealing” would vary by authority status, such that “stealing” would be used less frequently when towns (i.e., authorities) took resources than when individuals did. Indeed, a main effect for authority status, F(1, 2199.4) = 408.78, p < .001, η_p² = .16, indicated that the “stealing” label was used more for individuals (M = 1.38, SE = .06) than towns, (M = .42, SE = .06). We also explored whether the effect of authority status on labeling an act as stealing would only be evident when the act was not described as causing harm to the owner. Our analyses revealed a significant interaction between Harm to Owner and Authority Status on labeling acts as stealing, F(1, 2198.8) = 11.17, p < .001, η_p² = .01. Partially consistent with our prediction, the magnitude of the difference between ratings for individuals and towns was attenuated in the harm to owner condition, although values still differed (see Table 2).

The influence of authority status on the use of the stealing label and judgments of acceptability suggests that taking without permission may be at least partially conventional. To further test this possibility, we examine whether participants used moral reasoning in their judgments of taking resources without permission. Specifically, we consider participants’ reasoning based on two criterion judgments: justifications and rule independence.

Justifications. We expected that participants would judge moral justifications centered on the harm to the owner or the need of the recipient to be more important in guiding their evaluations when taking without permission entailed harm—either because the scenarios depicted that the recipient had a need or the owner was left without sufficient resources. Consistent with this hypothesis, there were significant interactions between Justifications x Recipient Need, F(5,14194) = 27.09, p < .001, η_p² = .01, and Justifications x Harm to Owner, F(10,14194) = 50.35, p < .001, η_p² = .03. Planned contrasts examined these effects (also see Table 3). Relevant moral justifications (i.e., welfare of the owner or recipient) were judged to be more influential in conditions when taking resources without permission entailed harm. Specifically, the Owner Welfare justification was used more when the owner would be harmed as compared to when the owner was left with enough, p < .001, and when the owner was left with extra, p < .001. On the other hand, the Recipient Welfare justification was used more when the recipient was in need than not, p < .001.

Further, in each of these harm-based conditions, the relevant moral justification was endorsed more frequently than the conventional justification: when there was Recipient Need, the moral justification (i.e., Recipient Welfare) was used more than the most influential conventional justification, (i.e., Authority), p < .001, and when there was harm to the owner, the moral justification of Owner Need was used more than the conventional justification of Authority, p < .001. In sum, when situations entailed harm, relevant moral concerns were deemed more influential than conventional considerations. Overall, moral justifications related to welfare were also judged to be more influential than property rights.

By contrast, when there was No Recipient Need and No Harm to the Owner, we expected that moral justifications would be less uniquely influential. Consistent with this hypothesis, the endorsement of the most used moral and conventional justifications did not differ significantly in this condition.

Rule Independence. Rule independence judgments indicated whether participants judged that an individual taking resources without permission was unacceptable even when it was allowed by the law, an indication of a moral violation. As expected, and consistent with the influence of Authority Status on judgments, when situations involved Harm to the Owner or Recipient Need, judgments were not contingent on the law. An interaction between Harm to the Owner and Recipient Need for judgments of rule independence, F(2, 995.99) = 18.78, p < .001, η_p² = .04, revealed that participants judged that taking resources without permission was unacceptable even when the goods were needed and it was allowed by the law—an indication of moral reasoning—but only when it was depicted as harming the owner (M = -.91, SE = .11), compared to leaving the owner with enough (M = .69, SE = .11) and compared to leaving the owner with extra (M = .94, SE = .11), ps <.001. Thus, when an owner had enough or extra and a recipient was described as in need, participants judged taking without permission as acceptable when permitted by law (see Table 2 and Figure 2).

To further explore whether taking without permission is judged to be moral in the absence of harm, we considered the mean and 95% CI for judgments of rule independence in the condition that depicted no harm to either party (No Harm to Owner, No Recipient Need). If individuals’ acts of taking without permission are deemed consistently immoral independent of harm, they should be judged to be unacceptable regardless of whether allowed by law—even when there is no harm. Comparing mean judgments of taking resources without permission when there was no harm to either party to the midpoint indicated that the difference included the midpoint of the scale (M = -.02, CI [-.21, Inf]), an indication that overall participants did not see taking without permission in the absence of harm as immoral. In addition, mean comparisons suggest that participants saw taking in the absence of harm as more acceptable than taking involving harm (Table 2).

Finally, to explore associations between participants’ political affiliations and their evaluations, we examined correlations between acceptability judgments and political orientation (scored from 1 to 7 where lower scores were more liberal and higher scores were more conservative). In the conditions where the owner was harmed, all correlation CIs contained zero or conservatives tended to judge the act as more acceptable (coefficients ranged from .02 to .16). However, when the owner was left with enough or extra resources, results were more mixed, with conservatives overall trending toward lower acceptability ratings (correlations between -.04 to -.36). See SOM for correlation coefficients within each condition.

We also examined these judgments using binned categories, with liberal (1-3 on the political orientation scale, n = 74), moderate (4, n = 63), and conservative (5-7, n = 63) participants. As above, of particular interest were the conditions where there was Harm to Owners, Recipient Need, or neither.

Means indicate that all three political orientations judged that it was similarly unacceptable to take resources without owner permission when it would harm the owner, with CI’s for the means based on difference from the midpoint, in the Harm/Need condition, liberals: M = -1.99, CI [-2.14, -1.83] ; moderates: M = -1.91, CI [-2.08, -1.74]; conservatives: M = -1.88 CI [-2.05, -1.72]. Conversely, individuals taking resources without permission was seen as acceptable across all three political orientations (albeit to varying degrees) when there was no harm to the owner, with CI’s for the means based on difference from the midpoint in the Extra/Need condition, liberals: M = 1.34, CI [1.13, 1.55]; moderates: M = .97, CI [.67, 1.25]; conservatives: M = .49 CI [.22, .76]. Justifications focused most on relevant moral concerns for the owner or recipient (consistent with the condition) for all political orientations in these conditions (see SOM). For example, participants of all political orientations rated Recipient Need as the most important justification when need was present and the owner was left with more than enough.

When scenarios depicted no harm to the owner (Extra condition) and the recipient as not being in need (the No Harm condition), there were different trends across the political spectrum. That is, when taking did not entail harm to either party, conservatives judged taking without permission to be slightly unacceptable (M = -.17, CI for difference from midpoint [-.45, .12], i.e., the midpoint fell within the CI), whereas moderates (M = .09, CI [-.20, .37] judged taking as slightly acceptable (although the midpoint also fell within the CI), and liberals approved of taking without permission (M = .59, CI [.34, .84]). In addition, under these circumstances, conservatives were also more likely than liberals to judge that Property Rights were important in their judgments of justifications, while liberals more than conservative judged that Fairness was particularly important. Finally, in the absence of harm, Rule Independence judgments showed that only conservatives judged that taking resources without permission was still wrong (under the midpoint) even when the law allowed it. See SOM for acceptability judgments and justifications by political orientation.

The findings from Study 1 suggest that most participants judged property violations in relation to harms to owners and recipients that those violations were depicted as causing or preventing. As expected, participants judged that taking without permission was acceptable regardless of the law when the owner was only minimally harmed and when it benefited someone in need. That is, when owners were left with enough, participants made very different judgments of taking resources, depending on the recipient’s level of need. For many, taking without permission was judged to be acceptable even when the recipient had no need, as long as there was no perceived harm to the owner. Moreover, and consistent with the idea that stealing is not a de facto moral violation, the attribution of the stealing label and judgments of acceptability were only moderately correlated.

Further, criterion judgments, including justifications and judgments of rule independence, also indicated that depictions of harm to recipients and/or owners were key to informing judgments. The justifications that were deemed most important were those related to harm—invoking the needs of either the owner or the recipient. Thus, justifications indicated participants were using moral reasoning to evaluate taking resources, and at times supported taking when it minimized harm. Relatedly, when taking was not depicted as harming the owner, rule independence judgments indicated that it was deemed to be acceptable regardless of the rule. Within SDT, moral issues are characterized based on their rule independence (Smetana, 2013; Turiel, 2017), since their consequences are intrinsic to the act and not contingent on local conventions. Thus, our finding that there are contexts in which taking without permission, particularly when doing so benefits others, were not deemed immoral, contrast with dominant theoretical accounts of stealing as an intrinsically moral violation. Indeed, when neither the recipient nor the owner was depicted as substantially harmed, taking without permission was largely judged to be acceptable whether it was allowed by law or not, and the pattern of justifications was ambiguous, not predominantly emphasizing welfare or property rights. Together, the findings of Study 1 indicate that in the absence of harm to the individuals involved, taking without permission is not consistently or unambiguously seen as a moral issue.

Because some research has found that moral judgments and perspectives on property rights may vary by political orientation (e.g., Graham et al., 2009; Kivikangas et al., 2021), we explored the role of political ideology in judgments about property violations. We found that when there was harm to the owner, liberals, moderates, and conservatives all judged that taking without permission to be similarly unacceptable, even when the law allowed it. Thus, we find that under some circumstances, there may be considerable consistency in judgments across the political spectrum.

Conversely, when a recipient was described as in need and the owner could spare resources, participants across all three political groups judged that taking without permission was acceptable, using moral justifications, and most made this judgment independent of laws. However, the degree to which participants judged taking without permission to be acceptable in these circumstances did vary across the political spectrum, with conservatives indicating less approval than liberals.

Participants’ views particularly diverged for vignettes wherein owners had plentiful resources and recipients were not in serious need. Under these circumstances, the different political groups appeared to emphasize distinct concerns and goals (Turiel, 2017). Liberals were more likely to judge that taking without permission was acceptable on the basis of fairness considerations, whereas conservatives indicated that taking was slightly unacceptable, even when the law allowed it, highlighting property rights as an important consideration in and of themselves. Thus, while people across the political spectrum made judgments based on perceived harm to the property owner or the need of the recipient, in the absence of harm, liberals and conservatives emphasized fairness and property rights differently.

We found the label “stealing” was less frequently applied to town councils than individuals, suggesting that this characterization may at least partially be driven by social-conventional considerations; that is, the government is often seen as having the authority to redistribute wealth (e.g., Norton & Ariely, 2011). Note that participants may have assumed that individuals consent to be represented by government officials, but our vignettes indicated that there was not consent to the specific redistribution. However, it is important to contextualize this finding by noting that, in this study, authorities were described as a “town council” whereas individuals were described as acting alone. Some research indicates that participants may more readily apply blame and responsibility to individuals as compared to groups (Mynatt & Sherman, 1975; Susskind et al., 1999). Thus, in Study 2, we sought to address this confound by describing the actions of a sole authority figure. In addition, justifications were limited as they were assessed with a forced choice question. Study 2 also addressed this issue, by asking for justifications with an open-ended question.

Study 2 sought to expand upon the results of Study 1 with three key modifications to the methodological design that further address our core question: is taking without permission judged to be a moral violation independent of perceived harm to owners? First, while maintaining our focus on boundary conditions, we addressed whether the findings in Study 1 also apply to judgments of cases that include other community contexts, especially in regard to the role of authority and different resources. Next, we investigated whether judgments of taking without permission are aligned with definitions of morality within SDT using different methods, including open-ended justifications, and a new criterion judgment (i.e., rule generalizability, Smetana, 2013; Turiel, 1983, 2015). Finally, we include samples from two countries. These changes are detailed below.

For Study 2, the authority figure was changed to be “the mayor” so that the taker was an individual across all conditions, rather than being an individual in the low-authority condition and a group (town council) in the high-authority condition. Additionally, authority status was manipulated as a between-subjects variable to prevent contrast effects. To increase generalizability, vignettes in Study 2 were based on more than one type of necessary resource that was applicable to a real-world context (drinking water and tractor fuel). Although we acknowledge that some key aspects of our vignettes remained necessarily contrived (e.g., the arbitrary distribution of resources), this was in an effort maintain participants’ focus on our key manipulated variables and reduce extraneous factors. The resources were also described in more abstract and intuitive terms to reduce cognitive load (e.g., rather than “five barrels,” vignettes described protagonists as having “more than enough” or “just enough” of the relevant resource).

The moral status of taking without permission was assessed with two criterion judgments (Turiel, 1983): justifications and generalizability. Unlike Study 1, justifications were assessed with open-ended responses that allowed participants to produce their own justifications without priming them with concerns that they otherwise may not have considered. In turn, Study 2 used a new criterion judgment of generalizability to assess whether a moral violation would generalize to different settings (e.g., asking participants “Would this be okay if it happened in your hometown?”), a hallmark of the moral domain (Yoo & Smetana, 2022).

Finally, sampling for Study 2 was conducted at two universities, one in the mid-Atlantic in the US and one in Eastern Canada, to broaden the diversity of the sample. In addition to the geographic and geo-political differences between the United States and Canada, the Canadian sample added slightly older participants (M_age = 23.39 as opposed to 20.15), a more diverse range of household incomes, and greater representation of non-White participants (36% compared to 22%; see Table 1). The samples also differed politically: the US sample was more politically balanced while the Canadian sample leaned more liberal.

Similar to Study 1, we examined how participants balanced concerns for the harm to the owner and the need of the recipient. We expected that participants’ judgments would be more attuned to the needs of the recipient when vignettes depicted no harm to the owner. However, when there was harm to the owner, we anticipated that recipient need would have a smaller effect on judgments of the acceptability of taking resources. We also expected that taking resources would be deemed more acceptable in the authority present condition. Nevertheless, we predicted that the two judgments (stealing label and acceptability) would be only weakly correlated. We further expected that authority status would have an impact on labeling acts as stealing such that taking resources would be less characterized as stealing in the authority present condition.

While there is evidence that property violations also implicate moral concerns with fairness and equity (Elenbaas et al., 2023), we expected moral justifications to be used primarily when taking resources was depicted as harming owners or addressing the needs of recipients. Therefore, we expected that justifications would include more references to moral (rather than conventional) concerns for vignettes when harm is present (in either the Harm to Owner or Recipient Need conditions). However, when there was no recipient need or harm to the owner, we expected that justifications would be more conventional. Justifications centered on property rights were examined separately on an exploratory basis. A complete list of pre-registered hypotheses is available on OSF, here. Similar to Study 1, we also explored associations between participants’ judgments and political orientations.

Participants were undergraduate students recruited from a private Catholic University in the Eastern United States (n = 188) and a public secular university in Eastern Canada (n = 100), for a total of 288 participants (after excluding 10 participants who completed 50% or less of the survey and one 17-year old). See Table 1 for sample demographics.

We conducted a sensitivity analysis to examine the power to detect the expected harm x need interaction. In 1000 simulated datasets, using our final sample size of 288, and the design and parameters of Study 2, the effect was found in all cases, which resulted in a sensitivity of 1.

We varied the depiction of harm done to the owner and recipient within subjects, and the authority status of the taker between subjects. Study 2 also varied the types of resources that were taken by a third party. In some trials, the necessary resource was tractor fuel; in other trials it was drinking water. In both scenarios, the mayor of the town arbitrarily distributed the fuel or water to the citizens based on postal code without considering need. Because of this uneven distribution, the mayor or another individual chose to take some resources from some citizens and give it to others. We varied whether the initial owner was left with enough (no harm) or not (harm) and whether the recipient had a need (need) for the resources or not (no need). Like Study 1, participants were told that the resources have a shelf-life and that surplus resources could not be saved for the future. The vignettes were designed such that each character’s available resource was unrelated to merit or hard work; that is, distributions were based purely on luck. All characters had unisex names and were referred to with gender neutral pronouns (they/them).

The authority status of the taker was manipulated between subjects by varying whether the resource was redistributed by the mayor (authority present) or a citizen (authority absent). In total, each participant saw four vignettes about tractor fuel and four about drinking water. The resource presented first (water or fuel) was assigned randomly, as was the order of each within-resource trial. Participants completed all trials for a given resource before moving on to the other resource.

Similar to Study 1, participants responded to multiple manipulation checks prior to each vignette to ensure that they correctly understood depictions of harm, need, and authority (multiple attempts were allowed, but a correct response was required to move on). After the manipulation checks, participants were asked for three judgments in the following fixed order: (1) if the action was acceptable (e.g., “Is what the mayor did OK?”), (2) if they would label the action as stealing (e.g., “Is what the mayor did stealing?”), and (3) as a measure of generalizability, if the action would be acceptable in their own hometown (“In your town, would this be considered OK or not OK?”). These initial judgment questions allowed only dichotomous responses (“OK” or “not OK” and “stealing” or “not stealing”). Following each initial judgment, participants gave confidence ratings (e.g., “Is what the mayor did sort of stealing or definitely stealing?”). Responses were then recoded on a four-point scale, centered on neutral ratings (-1.5 = definitely not stealing.-.5 = a little not stealing, .5 = a little stealing, 1.5 = definitely stealing).

Immediately following their acceptability ratings, participants were also asked to justify their acceptability judgments in a free-response format (“why do you think [the third party’s] actions are okay/not okay?). These responses were then coded according to three primary categories: 1) Moral if responses included considerations of harm, well-being, equality, or fairness to either the owner or recipient; 2) Property if responses included considerations of property rights or ownership; and 3) Conventional if responses included considerations of laws, procedural concerns, or authority. There was also a category called Other that was used rarely (less than 10% of justifications). Justifications were coded by two graduate students with extensive coding experience who were familiar with Social Domain Theory. The coding scheme was designed and agreed upon by the authors to establish units of analysis. After a training phase consisting of 60 participants’ interviews, the two coders independently coded 20% of participant responses (an additional 60 participants). No responses from the test phase had been previously discussed and coders were masked to each other’s codes. Interrater reliability (Cohen’s Kappa) for these codes was 0.76. The two coders then independently coded the remaining responses. Final codes from the test and practice phases were based on consensus.

Following all eight vignettes, participants completed a demographics survey about their background and political beliefs. Political beliefs were coded on a 10-point scale with 1= very liberal to 10 = very conservative. Scores were recoded into a categorical variable with three levels (1-4 = liberal, 5-6 = moderate, 7-10 = conservative), as preregistered.

Data analysis was conducted similarly to Study 1. We used the lme4 package (2015) in R (2021) to fit linear mixed-effects models on judgments of i) acceptability of acts, ii) whether acts were labeled as stealing, and iii) acceptability in the participants’ hometown, with participant ID and type of resource as random effects. In each case, a chi-square test indicated that the planned, theory-based mixed-effects model was an improvement on the intercept-only model, ps < .001. Post-hoc tests were corrected using Holm’s correction. Justifications were coded dichotomously, and so we used binomial (logit) generalized linear mixed models fit by maximum likelihood for analysis of justifications. All confidence intervals are 95%. For exploratory research questions, we only report means and 95% confidence intervals.

We expected that evaluations of acceptability would vary depending on depictions of harm to the owner and the level of need of the recipient. Specifically, we expected Recipient Need to be more strongly associated with judgments of acceptability when taking resources did not harm the owner. Indeed, we found an interaction effect for Harm to Owner x Recipient Need, F(1, 1980.18) = 42.55, p < .001, η_p² = .02, which indicated that when there was no harm to the owner, taking resources without permission was seen as somewhat acceptable when the recipient was in need, but unacceptable on average when there was no need, p < .001, as Table 4 shows. In addition, 95% confidence intervals of the difference from neutral were positive (in the No harm/Need condition CI [.39, .56] and contained zero (in the No harm/No need condition CI [-.15, .01].

However, when there was harm to the owner, taking resources was seen as unacceptable in both conditions (including CIs for Harm/Need [-1.24, -1.15 and Harm/No need [-1.41, -1.33], though means still differed, p < .001. Table 4 provides all means and Figure 3 provides a visual display of the data.

Based on exploratory findings in Study 1, we hypothesized in Study 2 that participants’ characterizations of an act as stealing and judgments of its acceptability would be only weakly correlated. Associations conducted separately for each condition between labeling an act as stealing and judging it to be acceptable were in the moderate range (continuous scale, rs = -0.32 to -0.58, dichotomous scale, -.26 to -.47), indicating that these are related but distinct constructs.

As hypothesized, a main effect for authority status, F(1, 270.58) = 106.11, p < .001, η_p² = .28, indicated that when an authority figure took the resources, participant labeled it as stealing less (M= .33, SE = .05) than when a citizen did so (M = 1.09, SE = .05).

Evaluations of the acceptability of acts also varied by authority status. As expected, a main effect for authority status, F(1, 280.93) = 40.33, p < .001, η_p² = .13, indicated that when the authority figure took the resources, participant rated it as less unacceptable (M = -.36, SE= .04) than when a citizen did so (M = -.71, SE = .04)¹. These results indicate that judgments of taking resources without permission are contingent on authority status, similar to conventional rules.

To examine whether participants used moral reasoning in their judgments of taking resources without permission, we next examined participants’ reasoning based on criterion judgments: justifications and generalizability. Based on SDT criteria, judgments of moral issues tend to be justified with moral reasons (i.e., considerations of harm, well-being, equality, or fairness to either the owner or recipient) and generalized to new contexts.

Justifications. We hypothesized that the use of moral and conventional justifications would differ by condition; however conventional justifications were used very rarely (ranging from 5% to 8% in different conditions, see Table 5 for means) and the model had a singular solution. Therefore, we chose to examine only moral justifications compared to all others (conventional, property rights and other justifications).

We conducted a binomial GLMM to test for main effects and the interaction for need and harm on moral justifications. For this analysis only, including the type of resource as a random effect did not improve the model, so it was omitted. The model with main effects and the interaction was an improvement on the model with only random effects, Chi-square p < .001. We expected main effects, but we also included the interaction, given the pattern of results for evaluations above. We did not find a main effect for the Harm to the Owner, nor for the Need of the Recipient. However, the interaction described below was significant, p < .001.

Similar to Study 1, in the No Recipient Need and No Harm to the Owner condition, we expected that moral justifications would be less influential. Probabilities, calculated by taking the inverse logit of coefficients, show that in the Harm to Owner/Recipient Need condition, the probability of a moral justification was .89. Moral justifications were also likely in the No Recipient Need/Harm to Owner condition (.91) and No Harm to Owner/ Recipient Need (.92) conditions. However, a series of Wald tests indicated that moral justifications were less likely to be used in the No Harm to Owner/ No Recipient Need (.84) than all other conditions (ps < .001), although they were still common.

Generalizability. Generalizability judgments indicated whether participants judged that an individual taking resources without permission was unacceptable across contexts, an indication of a moral violation. Overall, judgments of taking resources in the hypothetical town (M = -.54, SD = 1.11) were similar to judgments of taking resources in participants’ hometowns (M = -.56, SD = 1.10). Information across conditions is also presented in Table 4 and Figure 4, also illustrating that participants’ judgments of taking without permission in their hometown were not uniformly negative. Hometown judgments, like judgments in the hypothetical town, varied depending on the depiction of harm to owners and the need of recipients. Similar to judgments for the hypothetical town, judgments of hometowns indicated that when the owner was not described as harmed, taking resources without permission was not seen as strongly unacceptable. Confidence intervals of the difference from neutral for both No harm conditions were close to zero (in the No harm/No need condition CI [-.23, -.04] or were positive (in the No harm/Need condition CI [.18, .36].

When vignettes depicted harm to the owner, taking resources was seen as clearly unacceptable in both conditions (including CIs for Harm/Need [-1.18, -1.06 and Harm/No need [-1.37, -1.29]. Thus when there was harm to owners, taking without permission was seen as wrong across settings; when there was no harm, taking without permission was seen as more neutral across settings. Means are depicted in Table 4 and CI for acceptability are above; more details are available in the SOM.

Finally, to explore associations between participants’ political affiliations and their evaluations, we examined correlations between acceptability judgments and political orientation (scored from 1 to 10 where lower scores were more liberal and higher scores were more conservative). In the conditions where the owner was harmed, all correlation CIs contained zero (coefficients ranged from .11 to -.01. However, in conditions where there was no harm to the owner, conservatives trended toward lower acceptability ratings (correlations between -.21 to -.33) and the CI did not contain zero. See SOM for correlation coefficients within each condition.

Binned groups were also examined with liberal (1-4 on the political orientation scale, n = 140), moderate (5-6, n = 99), and conservative (7-10, n = 45) participants. Means (with CIs compared to the midpoint) indicate that all three groups judged that it was wrong to take resources without owner permission when it was depicted as harming the owner, indicating they saw taking as a moral issue when it involved harm (across Recipient Need conditions, liberals: M = -1.31, CI [-1.34, -1.27]; moderates: M = -1.26, CI [-1.31, -1.21]; conservatives: M = -1.27 CI [-1.36, -1.18]. However, the acceptability of taking resources without permission was generally seen as more ambiguous when there was no harm to the owner, even when there was no need from the recipient (liberals: M = .16, CI [.05, .27]; moderates: M = -.28, CI [-.42, -.14]; conservatives: M = -.36 CI [-.58, -.14], with CIs closer to zero.

Exploratory analyses of justifications by political orientation indicated that overall, mean use of the moral justification was higher for liberals (79% of responses) than conservatives (59%), while conservatives (22%) referenced property rights more than liberals (10%), on average. See SOM for details.

With new scenarios which included different resources and a more diverse sample, Study 2 provided additional evidence in support of a number of patterns observed in Study 1. First, the acceptability of taking resources without permission was judged differently depending on whether the vignettes depicted harm to owners and recipients in need. When there was no harm to either owners or recipients, the acceptability of taking without permission was judged to be ambiguous, i.e. not clearly wrong. Findings from Study 2 bolster the interpretation that perceived harm is a central consideration in determining the acceptability of taking without permission. Acceptability judgments of taking resources without the owner’s permission continued to depend on the degree of harm to the owner and the recipient’s level of need. Like Study 1, participants judged that it is acceptable to take resources to address the recipient’s needs but not at the expense of the owner’s well-being. That is, as expected, participants uniformly judged that harm was unacceptable when the owner was not left with enough; but when the owner had enough resources, they judged taking to be more acceptable when the recipient was in need.

Similar to Study 1, we found that labeling an action as “stealing” did not necessarily mean that the action was evaluated as unacceptable, in that the two judgments were only moderately correlated across conditions. Thus, although they are clearly overlapping constructs, participants sometimes labeled acts as stealing while judging them to be acceptable. As in Study 1, we found that participants were less likely to label acts as stealing when an authority figure took the resources. As expected, we also found that acts were deemed less unacceptable when carried out by an authority figure. These findings suggest that conceptions and judgments of stealing are in part informed by social-conventional considerations centered on authority status.

In Study 2, we collected open-ended justifications which allowed us to identify the types of concerns that were salient to participants themselves as they explained their acceptability judgments. In most cases, participants justified their judgments based on moral concerns, typically drawing on concerns with perceived harm to the owner or the recipient, by referencing their respective needs, as well as fairness considerations. This finding was generally aligned with the patterns evident in Study 1, although participants’ relative emphasis on moral concerns was more striking when they were asked to describe their predominant concern in their own words. Overall, this suggests that taking without permission is indeed often understood in moral terms, regardless of the circumstances. However, we also observed variation across conditions, in line with our hypotheses. That is, moral justifications for acceptability ratings were provided least frequently when there was no harm to the owner and no need from the recipient. Conventional references to authority or norms were rare across conditions. Given the methodology used in Study 2, this does not necessarily imply that conventional considerations were not at play, but rather that they were eclipsed by the salience of moral concerns.

To complement the criterion judgment of rule independence assessed in Study 1, we also asked participants to indicate whether their judgments of actions in the hypothetical town would generalize to another setting– their own hometown. Overall, we found that patterns of judgment were similar across the two contexts. In other words, in conditions where participants judged an action to be unacceptable in the hypothetical town, they also judged the same action to be unacceptable in their own hometown. Similarly, some actions were deemed acceptable in both settings. Thus, overall, these findings imply that without perceived harm to the owner, taking without permission is not consistently viewed as a moral violation in either context. That being said, it is worth noting that exploratory results showed that when someone is depicted as being in need, participants judged it more acceptable to take on their behalf in the hypothetical context than in their own hometown. This may be an indication the participants saw the right to property as depending on specific features of the ownership that bears on whether fairness comes into play. For example, they may consider property rights to be more salient when someone is interpreted to have earned the property fairly (as in some cases in our participants’ cultures), but not when they achieved it by luck (as in our hypothetical vignettes).

Finally, as in Study 1, we explored associations with political orientation. Like Study 1, participants across the political spectrum tended to make similarly condemnatory judgments of taking without permission when taking resources was depicted as causing harm to the owner. On the other hand, when there was no harm to the owner, liberals tended to judge taking resources as more clearly acceptable, regardless of the presence of recipients’ need, while moderates’ and conservatives’ responses were more ambiguous, falling between “a little unacceptable” and the neutral midpoint. In this sense, while Study 1 suggested that participants across the political spectrum were attuned to both harm and needs, Study 2 implied that perceived harm to owners was more consistently influential to participants’ judgments. Exploratory analyses also showed that justifications differed by political orientation, with liberals relying more exclusively on moral justifications related to harm, which included fairness as well as welfare, and conservative referencing property rights relatively more often. Thus, taken together, the findings from Study 2 imply that perceived harm to owners guides judgments of stealing across the political spectrum, but the property rights of owners may be more salient to conservative participants in the absence of explicit harm.

Overall, these two studies provide converging evidence that taking without permission is not consistently judged in moral terms across contexts (See Table 6 for a summary of all results by hypothesis across both studies). When owners were not depicted as being harmed by having their resources taken, participants did not judge taking resources without permission (i.e. “stealing”) to be unacceptable, highlighting the role of perceived harm in judgments of stealing, regardless of its label.

It is important to note that acceptability judgments are not, by themselves, an unambiguous indicator of moral reasoning in Social Domain Theory (Smetana, 2013). Nevertheless, the divergence between acceptability and characterizing an act as stealing emphasizes the multiplicity of concerns and interpretations that bear on each type of judgment. Indeed, our findings regarding associations between the authority status of the taker and labeling an act as stealing suggest that act labels may be particularly informed by social-conventional considerations such as authority and rule dependence (e.g., the same act is deemed “stealing” less when it is authority-mandated). Participants were less inclined to describe taking resources without permission as “stealing” when an authority figure did so than an individual did, and they also judged these acts as more acceptable in the authority condition.

Complementary findings based on acceptability ratings, criterion judgments of rule independence, generalizability, and types of justifications provide evidence that taking without permission is not consistently judged in moral terms when it does not implicate perceived harm to the owner or respond to a need. Taking without permission was generally deemed unacceptable when the owner was depicted as experiencing harm, and as acceptable when the owner was unharmed and the recipient benefited. When both conditions are present (i.e., potential harm to the owner and need of the recipient), taking property potentially pits two moral concerns against each other: harm to the owner can conflict with the recipient’s needs (Turiel, 2017). Past research provides other evidence of boundary conditions on the unacceptability of stealing, in that it is sometimes seen as acceptable when it is pitted against other concerns (e.g., when managers treat subordinates unfairly; Umphress et al., 2009) or when the harm is judged to be minimal (e.g., illegal downloading; Olivero et al., 2019). Across the political spectrum, our results were consistent with this research, showing that participants considered moral concerns related to both owners and recipient beneficiaries of “stealing.”

While our findings clearly demonstrate the relevance of welfare considerations in moral judgments of stealing, the status of property rights as a moral concern is worth considering further. Some forms of rights are judged to be moral (see Helwig & To, 2017). Moral rights are often judged in relation to autonomy such as freedom of speech or religion, where individuals make decisions about their own bodies and lives—rights that may be seen as necessary for human welfare and flourishing. Thus, many moral rights may come from a concern for avoiding harm and promoting human welfare. Property rights may be conceptually distinct in that they imply a right to make decisions about resources—and excessive amounts of resources are not necessarily needed to flourish. Our findings are consistent with the notion that property rights are not consistently seen as moral when taking resources does not imply direct harm to this owner. In this respect, most participants appeared to judge the right to property as being less generalizable and more alterable than the right to bodily autonomy and life, since participants judged it acceptable to remove resources from a resource-rich owner when it helps someone in need of resources. Of course, it is possible that other moral considerations may come into play that involve less direct forms of harm implicated by violations of property rights, such as the deterioration of a sense of community safety, but our coding did not indicate that these were salient in informing participants’ justifications for their acceptability ratings (as assessed by their open-ended justifications in Study 2).

Interestingly, only when taking without permission was depicted as not implicating harm to the owner, exploratory analyses suggested that there was some divergence between liberals and conservatives in their relative emphasis on property rights. In general, correlations indicated that liberals viewed stealing as more acceptable than conservatives when it did not harm the owner. In Study 1 specifically, liberals were more attuned to fairness considerations and conservatives were more attuned to property rights of owners when stealing did not cause harm to the owner or address a pressing need. In Study 2, while all participants tended towards moral justifications (welfare and/or fairness) in explaining their judgments, this was particularly true for liberals, whereas conservatives were more inclined to describe property rights. Thus, our research is consistent with past work suggesting that judgments of wealth redistribution and perspectives on property rights may vary by political orientation (Farrelly, 2004; Graham et al., 2009), but our findings also suggest that these political differences are evident only in some circumstances. It would be useful for future research to further explore the understandings and concerns that are underlying different participants’ perspectives on property rights, and how these rights are understood to be contextually situated and weighed against other considerations.

While our findings are well-aligned with the propositions of Social Domain Theory that we tested in our study, our results are also consistent with other theories in moral psychology. For instance, using the Theory of Dyadic Morality (TDM, mentioned above; K. Gray et al., 2012, 2022) as a lens, our findings would suggest that when stealing was judged as unacceptable, the owner was perceived as a vulnerable patient who experienced harm as a result of an intentional agent (i.e., either the Mayor or citizen) confiscating resources. Conversely, when stealing was judged as acceptable, it was in contexts when the dyadic template was likely to be (mostly) inactivated since the situation did not implicate harm to either party. The differences observed between liberals and conservatives can also be adjudicated with TDM, which advocates that the underlying moral machinery is the same across ideological divides (Schein & Gray, 2015). According to TDM, moral condemnation is a matter of perceived, not necessarily objective, harm. The context of the situation impacts perceptions. Liberals and conservatives reliably differ in their perceptions of when stealing—along with other moral concerns—causes harm (Schein & Gray, 2015). Thus, although we were quite intentional about depicting harms and needs explicitly in the context of the vignettes themselves, liberals may have nevertheless focused more on potential recipients as vulnerable patients who were disadvantaged by unfair and arbitrary distributions, whereas conservatives may have interpreted unauthorized taking as causing harm to the owner regardless of the condition; as noted above, it is also plausible that conservative participants may view unauthorized redistribution as placing the community as a whole under threat, although this was not salient in the open-ended justifications in Study 2. In fact, taking without permission could be judged to implicate harm even when very wealthy people lose very little property since if stealing became common, it could affect people more seriously. Given that research has shown that mind perception can be attributed to groups (H. M. Gray et al., 2007), violating property rights may be viewed as damaging to social order. In this respect, TDM also emphasizes the importance of future research that further disentangles the varied concerns underlying participants’ context-sensitive judgments of stealing.

Foremost, this study is limited by its sampling bias. Although our samples were drawn from two North American countries that were politically diverse, and came from many fields of study, they were nonetheless comprised of mostly White undergraduates exclusively from WEIRD (Western, Educated, Industrial, Rich, Democratic) populations (Henrich et al., 2010). This has implications for the generalizability and external validity of the research, as there is cultural variation in how moral judgments are applied in specific contexts (e.g., Awad et al., 2020; Turiel, 2017). For example, some suggest that WEIRD populations may be more concerned with individual freedoms like property rights (Triandis, 2001), and data from such populations should not be generalized to all contexts. Future research should seek to replicate these findings in other cultural and socioeconomic contexts.

Another limitation relates to the fact that all of the vignettes depicted redistribution of a required resource that was intended to benefit a third party. This is a very specific boundary condition, identified to test the limits of judgment of stealing. However, the patterns observed may not generalize to situations in which protagonists take resources for their own benefit or in the context of taking a less critical resource. Indeed, previous work finds that social closeness affects how individuals evaluate third-party theft (Linke, 2012). Other transgressive actions are also interpreted differently in a third-party context; for example, third-party revenge is understood as serving benevolent ends in a way that revenge on behalf of the self is not (Frey et al., 2021). Taken together, examining how the taker benefits (or does not) from the act of taking may further illuminate the moral basis of evaluations of stealing. Furthermore, despite the use of varied contexts and stimuli across studies, all data are based on vignettes that included character actions and resources that may have been viewed as arbitrary and sometimes lacking in ecological validity. For example, all of the vignettes depicted purely luck-based distribution and subsequent redistribution of resources, such as food (Study 1), water (Study 2), and fuel (Study 2). Future work can expand this research by examining variations across a greater variety of ecologically valid contexts and by drawing on different methodologies. It would also be useful for future studies to take a more qualitative approach and interview participants about their viewpoints on stealing and property rights based on their own lived experiences. Additional research using alternative methods, such as a fully between subject design and studying different populations could either corroborate or conflict with our conclusions, further illuminating our understanding of people’s conceptions about stealing in particular and property rights more broadly.

Stealing, or taking without permission, often implicates harm, but harm is not a necessary component of stealing. The current studies present a novel view of taking without permission, with harm to owners and recipients manipulated. In a series of two studies based on politically balanced samples of North American undergraduate students, we found that participants across the political spectrum considered harm to both recipients and owners when judging acts of taking without permission. Our findings emphasize the value of considering the criteria that people use to judge and reason about moral issues. If property violations were de facto judged as a moral issue, independent of harm, they would have been deemed as unacceptable in ways that are generalized across contexts, for moral reasons, and wrong even in absence of the law. Using these criteria, we show that property violations are not consistently viewed in moral terms. Instead, we found that harm is central to moral judgments of these acts.

A data analytic code review with many helpful suggestions was completed by Radu Urian.

We have complied with ethical standards in accordance with ethical guidelines, including attaining IRB approval and using informed consent for our study.

We have no conflicts of interest to declare.

Both studies were preregistered at OSF, and all materials, code and data are also at OSF, here.