Introduction

Imagine a business executive who helped his company double its sales last year but was given very little bonus for his hard work and personal sacrifice. One day, he was sitting alone in his office deciding whether to use some company funds to pay for a European cruise vacation that he had always wanted to take. Surely, he may lose his job and face the legal consequences of embezzlement if he gets caught. However, the extent to which he would worry about potential damage to his social reputation may depend on how easily he can move around in his social environment.

Over the past decade, a great deal of interest has focused on the role of reputational systems as a viable solution to the problem of cooperation. According to Emler [1], reputation refers to “that set of judgments a community makes about the personal qualities of one of its members” (p. 171). Research has shown that humans have a keen sensitivity to reputational information and are more likely to comply with social norms when they feel that they are being observed by others [2–4]. Indeed, a number of studies have consistently found that reputational concerns can facilitate cooperative behavior and deter people from advancing their own interests at the expense of the common good [5–6].

One explanation for why reputational concerns have such powerful effects on cooperation and deviance is that reputation can influence individuals’ payoff in future social interactions. Studies from evolutionary game theory and economics suggest that people with good reputations are more likely to be selected as interaction partners in future social exchanges [5] and to be helped by others [7–8]. People known to be selfish, on the other hand, are at risk of being socially excluded and even punished [9–10].

Although reputational concerns can incentivize prosociality and deter deviance, the strength of these effects can vary according to socio-ecological constraints [11–12]. Residential mobility and relational mobility are two types of socio-ecological constraints that may influence the effects of reputational concern on pro- and anti-social behaviors. Residential mobility refers to the frequency with which people change residence, whereas relational mobility refers to the number of opportunities individuals have in a given context to voluntarily form new relationships and to terminate old ones [13]. These two concepts have been used to explain a variety of important behavioral and psychological tendencies, including identification with reputable groups and avoidance of negative social reputations (see Oishi et al. [13] for a comprehensive review on residential and relational mobility).

Previous research suggests that reputational information may be more useful in residentially and relationally stable settings where the cost of non-cooperation is much higher. In those settings, there are fewer opportunities to form and reform relationships [14]. Once people are socially excluded for misconduct, they may not be able to find alternative groups or relationships to join. Because relationship networks in less mobile contexts tend to be more stable and close-knit, a stronger emphasis is placed on the collective relative to the individual, as shown by greater commitment to group or community welfare [15] and greater willingness to use punishment against wrongdoings [16].

Scholars have posited that systems of mutual monitoring and sanctioning encourage and sustain a high level of group-based cooperation in collectivistic societies [12, 17]. Monitoring and sharing of reputational information are mechanisms within groups that are designed to detect prosociality and deviance, and are therefore considered useful forms of social control [18–19]. This idea is consistent with empirical evidence showing that accountability and public evaluation enhance cooperative behavior among collectivists [20]. Given the resemblance of the socio-ecological conditions between collectivistic societies and residentially/relationally stable environments [21], monitoring of reputational information is expected to sway individuals toward cooperation and away from misbehavior in contexts that are low in residential or relational mobility. Conversely, monitoring of reputational information may be a less effective means of social control when mobility is high, because the consequences of wrongdoings are less severe as individuals can change social circles to avoid social exclusion and conflict. Past research also indicate that individuals from mobile societies tend to feel less obligated to monitor and respond to dishonest behavior than those from stable societies [16].

Three studies examined whether the usefulness of social monitoring in preventing socially deviant behavior differs according to the given socio-ecological context. Specifically, we tested the hypothesis that the presence of social monitoring will be effective in deterring social deviance when residential or relational mobility is low but not when residential or relational mobility is high. The first study examined these factors at the community level, and the latter two studies examined these factors at the individual level. In Study 1, we examined within-nation, regional variation in crime rate as a function of the number of journalists (social monitoring agents) employed and levels of residential mobility. Given that social monitoring matters more in residentially stable communities, the negative relationship between number of journalists and crime rate should be stronger in those communities than in communities that are residentially mobile. In Study 2, we measured subjective perceptions of relational mobility at the individual level and examined their relationship to lying when a social monitoring cue was either present or absent in the laboratory. In Study 3, we manipulated both individuals’ perception of relational mobility and the presence vs. absence of a social monitoring cue in the laboratory and compared their interactive effects on cheating behaviors.

Study 1

To test whether the effects of social monitoring are likely to be stronger in residentially stable communities where damages to one’s reputation matter more, we examined crime rates in U.S. cities that differ in rates of journalist employment (an index of the amount of social monitoring that exists in a given community) and levels of residential mobility. The power of the news media to influence the reputation of individuals as well as organizations is well documented [22–23]. By raising public awareness of social problems and threats, social monitoring by journalists may play a role in keeping social deviance under control in residentially stable communities. Based on previous research on relational and residential mobility, we predicted that having more monitoring agents like journalists will have greater effects on crime rates in cities that are low in residential mobility than those that are high in residential mobility.

Results

We tested our hypothesis using hierarchical linear regression. The dependent variable, violent crime rate, was transformed using natural logs so that the residuals fit the Gaussian distribution. Results from the Kolmogorov–Smirnov test and the Shapiro–Wilk test indicated that the distribution of the residuals were normal after log transformation (ps > .179). To reduce multicollinearity between the predictors, we centered them via z-transformation before creating an interaction term between number of reporters and level of residential mobility. Per capita GDP, unemployment rate, and population size were entered as covariates in Step 1 of the regression. Number of reporters and residential mobility were entered in Step 2. The interaction between number of reporters and level of residential mobility was entered in Step 3.

The regression results suggest that unemployment rate and residential mobility predicted violent crime rate (unemployment rate: B = .13, SE = .04, t = 3.25, p = .002, 95% CI = [.052, .214]; residential mobility: B = .09, SE = .04, t = 2.39, p = .018, 95% CI = [.016, .169]). Consistent with our hypothesis, we found an interaction between number of reporters and level of residential mobility. The overall model explained 14% of the variance in violent crime rate, R² = .14, F(6, 116) = 3.16, p = .007. The interaction explained an additional 3.1% of the variance in violent crime rate beyond the first-order effects (_ΔR² = .03, p = .041) and yielded a moderate effect size (f² = .16). As seen in Fig 1, in cities that are low in residential mobility, number of reporters was associated with lower rates of violent crime. In cities that are high in residential mobility, violent crime rate did not change as a function of number of reporters employed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Residential mobility moderates the link between number of reporters employed and violent crime rate.

https://doi.org/10.1371/journal.pone.0167053.g001

Study 2

The main goal of Study 2 (and Study 3) was to conceptually replicate Study 1 by examining the same factors at the individual level. Although we found the expected moderating effect of residential mobility in the relation between social monitoring and deviant behavior, these findings were at the level of city, not individuals. To avoid the ecological fallacy, therefore, it is important to test whether the observed interaction could be replicated at the level of individuals [24]. In addition, violent crime is a relatively infrequent deviant behavior. Thus, it is important to test our hypothesis with more common behaviors. We conducted Study 2 to address these limitations. Specifically, we manipulated the presence vs. absence of social monitoring cues in the laboratory using a surveillance camera and examined whether it would have differential effects on cheating depending on individual variations in subjective perceptions of relational mobility. The presence of a camera is an established and well-validated cue that has been found to induce public self-consciousness [25, 26]. To measure cheating, we used the die-under-cup task, which is a procedure that has previously been used to detect cheating at the individual level [27–28].

Study 3

Because low and high relational mobility individuals were found to be differentially influenced by social monitoring in Study 2, we reasoned that priming relational mobility should produce a similar pattern of results. Instead of completing the die-under-cup task, participants in Study 3 played an online maze game that provided them an opportunity to win extra cash for the number of mazes they indicated as solved within the time period allotted. As in Study 2, we hypothesized that the presence of social monitoring cues would reduce the likelihood of cheating when relational mobility is low but not when relational mobility is high.

Results

Table 1 reports the number of mazes participants actually solved, the number of mazes they indicated as solved, and the frequency of various cheating behaviors by condition. There was no difference between the low and high mobility conditions or between the monitoring and control conditions in the number of mazes solved (mobility: t[153] = 0.67, p = 0.505, 95% CI [-.934, 1,890]; monitoring: t[153] = 1.12, p = 0.266, 95% CI [-.612, 2.204]) or indicated as solved (mobility: t[153] = 1.25, p = 0.212, 95% CI [-.509, 2.278]; monitoring: t[153] = 1.36, p = 0.176, 95% CI [-.435, 2.349]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Descriptive statistics of mazes indicated, mazes solved, and cheating behavior as a function of relational mobility and social monitoring.

https://doi.org/10.1371/journal.pone.0167053.t001

To test our hypothesis that cheating behavior would decrease significantly from low to high monitoring when mobility is low but not when mobility is high, we created a hypothesis-specific, focused contrast code (+1 = low mobility and low monitoring, -1 = low mobility and high monitoring, 0 = high mobility and low monitoring, 0 = high mobility and high monitoring) and regressed the dependent variables on this contrast code. First, we examined whether participants counted the mazes they solved after the timer went off towards their total score. Even though none of the main effects reached statistical significance (mobility: B = -.69, SE = .47, Wald = 2.14, p = .144, 95% CI [-1.778, .289]; monitoring: B = -.72, SE = .47, Wald = 2.33, p = .127, 95% CI [-1.919, .164]), the contrast analysis was statistically significant, B = .75, SE = .34, Wald = 4.90, p = .027, Exp(B) = 2.113, 95% CI [.109, 1.665], supporting our hypothesis that the presence of a social monitoring cue reduced cheating in the low mobility condition but not in the high mobility condition.

Next, we examined the effects of mobility and monitoring on the use of function keys. As Table 1 indicates, very few participants used the function keys to help simplify or speed up the solution. Hence, we found no main effect of mobility or monitoring on this type of cheating behavior (mobility: B = -.68, SE = .88, Wald = .56, p = .441, 95% CI [-18.677, 1.360]; monitoring: B = .01, SE = .83, Wald = .00, p = .987, 95% CI [-17.997, 17.984]) and the result of the contrast analysis was not statistically significant (B = .01, SE = .58, Wald = .00, p = .982, 95% CI [-1.781, 1.763]). With respect to the difference in number of mazes solved and number of mazes indicated as solved, participants in the low mobility condition over-reported more than those in the high mobility condition on a marginally significant level, t(107) = 1.82, p = 0.072, d = .293, 95% CI [-.037, .850]. Moreover, even though the social monitoring cue had a greater impact on deterring cheating in the low mobility condition compared to the high mobility condition, the difference did not reach statistical significance, B = .15, SE = .16, t = .91, F(1, 153) = 0.83, p = .364, 95% CI [-.190, .631].

Discussion

Does residential and relational stability heighten individuals’ sensitivity to social monitoring? To answer this question, we examined the impact of social monitoring on deviant behavior at both the community level (Study 1) and the individual level (Studies 2 and 3). Across the three studies, we found that social monitoring exerted greater influence on curbing social deviance in more stable environments. Although avoiding being identified as a norm violator is a major concern for humans in general, monitoring of reputational information may be greater when moving around is a more complicated and costly affair.

Previous research has suggested that sensitivity to reputational information plays a key role in human cooperation [5–6], as well as in the deterrence of deviance [34–36]. However, our studies show that the power of social monitoring on social deviance can vary according to social-structural constraints such as residential mobility and relational mobility. In contexts that are similar to our evolutionary ancestors, i.e., low in residential or relational mobility [37–38], social monitoring may be an effective mechanism of social control. People’s willingness to monitor each other’s merits as well as misdeeds may play a crucial role in the functioning of residentially and relationally stable societies. By contrast, the effects of social monitoring and the appeal to reputation might not be as strong in residentially and relationally mobile contexts. Our results are consistent with recent findings on rural-urban differences in interpersonal regret [11]. The thought of rural living evoked greater reputational concern, which in turn increased the intensity of interpersonal regret, relative to the thought of urban living. Together these data suggest that the monitoring of reputation as a mechanism of informal social control may be stronger in smaller, residentially/relationally stable communities than in larger, residentially/relationally mobile communities where there is always an exit option if one’s reputation is damaged.

In the present research, we focused on just one mechanism of social control. Exploration of alternative mechanisms would be worthwhile, particularly those that can shed light on how social order is effectively maintained in residentially or relationally mobile contexts. In societies where individuals change groups frequently, continuously attending and adjusting to new group norms and regulations can be difficult and taxing. Therefore, we propose that members from mobile societies are more apt to follow universal rules of conduct (e.g., law) and to develop an independent, internal code of ethics. This notion is consistent with cultural psychological theories and research on rights-based vs. duty-based moral codes [39–40]. Specifically, the moral judgments and behaviors of individuals living in the US, which is a highly mobile society, are guided by personal preferences [41] and individually-oriented moral codes [40]. Future studies examining whether social control in residentially or relationally mobile settings can be achieved via internalized moral values and standards hold promise in advancing this area of research.

It is worth noting a few limitations of the study. First, convenience samples of undergraduate students comprise a narrow database, creating problems for the generalizability of our findings to larger populations of interest [42–44]. Second, we relied on official crime statistics from a government security agency, which may not accurately reflect crime rates in a given community due to issues such as under-reporting and under-recording [45–48]. Third, even though the power of the news media in shaping a person’s reputation has been well documented [22–23] and that its role in shaming people when they commit minor offences has intensified in recent years [49], the validity of using number journalists employed as a measure of the prevalence of social monitoring needs to be further established. The work of Baum and Potter [50] suggests that the media plays a critical role in controlling the flow of information from whistle blowers to the public. Finally, sample sizes for Studies 2 and 3 were small, making interpretations of the findings more difficult. The reliability of the key finding (i.e., social monitoring deters deviant behavior in low mobility contexts) needs to be examined using larger samples in future research, especially since the pattern was observed with only some of the cheating indices. Aside from the issue of reliability, possible floor effects involving cheating behaviors that were too obvious and difficult to justify could have contributed to the null findings. Previous studies have shown that people are more willing to cheat when they can justify their unethical behavior [28, 51]. For example, participants in Study 3 may have felt that counting mazes solved after the time limit towards their total score was legitimate, given that they actually did solve them on their own (as opposed to reporting ones they never solved). In contrast, using function keys and reporting mazes they never saw are obvious, difficult-to-justify lies, which might explain why very few participants engaged in these forms of cheating and thus creating floor effects that made the effects of social monitoring on cheating hard to detect.

As human societies become more interconnected, issues related to trust and cooperation among individuals and groups deserve greater attention than ever before. Any viable solution to the problems of crime, deception, and mistrust may benefit from considering particular features of the environment in which individuals and groups inhabit. Recent perspectives in cultural psychology—including the dynamic constructivist approach [52], the institutional approach [53], the intersubjective approach [54], and the socio-ecological approach [55]—have highlighted the importance of the social ecological environment in shaping humans’ thoughts, feelings, and behavior. By examining how residential and relational mobility influence individuals’ reaction to cues of social monitoring, we hope to facilitate a potentially fruitful avenue for future research on the psychology of social cooperation and deviance.