Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY

Group Polarisation in Reactions to Dishonesty

Steffen Keck INSEAD, Boulevard de Constance, 77305 Fontainebleau Cedex, France [email protected] Tel: +33160729117 Fax: +3316079117

11/07/11

Acknowledgments I thank Neil Bearden, Linda Babcock, David de Cremer, Enrico Diecidue, Natalia Karelaia, Wenjie Tang, Alminas Zaldokas and the participants of the 2011 LBS Trans-Atlantic Doctoral Conference for their insightful comments.

Funding This research was funded by the INSEAD R&D committee and the INSEAD alumni fund.

1

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Abstract This paper compares groups and individuals in terms of their willingness to incur financial costs in order to punish dishonest behavior. Study 1 shows that the use of dishonest bargaining tactics is punished more often by groups than by individuals. At the same time groups are also more willing than individuals to use dishonest tactics themselves, when it serves their financial self-interest. The results of Study 2 demonstrate that groups’ increased willingness to punish dishonesty is mediated by their stronger negative affect directed toward dishonest parties. Overall, these results suggest that being part of a group increases group members’ self-serving tendency to evaluate dishonesty differently depending on whether its use is beneficial or harmful to them.

Keywords: deception, ethics, groups, punishment

2

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY GROUP REACTIONS TO DISHONESTY Lies and other forms of deception are a common feature of daily life (DePaulo, Kashy, Kirkendol, Wyer, & Epstein, 1996) and are frequently employed in negotiations, bargaining, and other situations involving mixed motives (Tenbrunsel, 1998; Thompson, 2004). The use of deception in such cases is not only considered to be unethical (Thompson, 2004) but when discovered, might also lead to a mutually harmful conflict—for example, a breakdown in negotiations. Previous research has repeatedly shown that individuals are, in general, willing to punish dishonest others even when punishment is financially costly and leaves both parties worse-off (Boles, Croson, & Murnighan, 2000; Brandts & Charness, 2003; Wang, Galinsky, & Murnighan, 2009; Wang & Leung, 2010). Yet in organizations, many critical tasks (such as negotiations) are frequently carried out by groups rather than by single individuals. Thus it is important to take group-specific factors into account when seeking to understand how the use of dishonest tactics affects organizational outcomes. The likelihood that deception will be used has been shown to depend on various factors, including expected financial gains (Gneezy, 2005), relative power of the deceiver (Koning, Steinel, van Beest, & van Dijk, 2011), and the interaction’s competiveness (Schweitzer, DeChurch, & Gibson, 2005). Most closely related to this paper are the studies of Sutter (2009) and Cohen, Gunia, Kim-Jun, and Murnighan (2009), who compare groups and individuals in terms of their willingness to send a deceptive message to other participants when this serves their financial interest. Both studies find that groups are more likely to do so than individuals. Cohen et al. (2009) also show that increased deception is fully mediated by groups’ stronger focus on maximizing their financial gains. Consistent with these results, Stawiski, Tindale, and

3

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Dykema-Engblade (2009) report that—in a simulated computer-mediated negotiation—groups were less likely than individuals to reveal information that would decrease their bargaining power, even when specifically prompted to do so by the other party. In contrast to these studies, which focused on the decision to use deception, I explore how groups react when deception is used against them. When individuals discover that they have been the target of deceptive tactics, they are often willing to incur substantial financial costs in order to punish the deceiving party (Boles et al., 2000; Brandts & Charness, 2003; Wang et al., 2009; Wang & Leung, 2010). As previous research has demonstrated, the willingness of individuals to inflict punishment on others is strongly correlated with feelings of negative affect, such as anger and spite (Pillutla & Murnighan, 1996; Wang et al., 2009). Whereas the causes and moderating factors of individual reactions to dishonesty have been widely studied, the role of groups in this context is still unclear. Prior findings in the literature give rise to two contradictory hypotheses. Previous results consistently show that groups behave more competitively in social dilemmas and are less interested than individuals in maximizing joint gains for all parties (Morgan & Tindale, 2002; Wildschut, Pinter, Vevea, Insko, & Schopler, 2003). Moreover, groups react more aggressively when provoked (Meier, Hinsz, & Heimerdinger, 2007). For example, groups are more likely than individuals to deliver electrical shocks to a target person after being insulted (Jaffe, Shapir, & Yinon, 1981; Jaffe & Yinon, 1979), retaliate more when provoked by being allocated a portion of hot sauce to consume (Meier & Hinsz, 2004) and are more likely to engage in verbal aggression and threats when persistently annoyed by another individual during a work task (Pruitt, Parker, & Mikolic, 1997).

4

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY An important reason why groups can be expected to behave more aggressively when provoked by the dishonest behavior of another group is that each group member embedded in intergroup relations might feel more negative affect than a lone individual interacting with another individual. Results of Hoyle, Pinkley, and Insko (1989) suggest that people in general associate more negative traits (e.g., “abrasiveness”) and fewer positive traits (e.g., “agreeableness”) with groups than with individuals. As a consequence, group members might interpret dishonest behavior to be more hostile and provocative when its source is another group and therefore feel more anger (Pruitt et al., 1997). Second, when discussing which decision to make, group members’ attitudes often polarize in the direction of the dominant attitude within the group (Lamm & Myers, 1978). Thus, if all group members feel at least mild anger at the beginning of the discussion, then negative affect toward the other group might become stronger during the discussion. Third, being part of a group— even one whose membership is determined randomly—can result in increased favoritism toward in-group members and increased hostility toward out-group members (Tajfel, Billig, Bundy, & Flament, 1971). In particular, out-groups that are perceived as violating social norms or posing a threat to the in-group are likely to elicit strong negative affect, such as anger and contempt (see Hewstone, Rubin, & Willis, 2002). Wang et al. (2009) demonstrate that, for individual decision makers, negative affect has a strong influence on the decision to punish dishonesty; it is therefore reasonable to assume that negative affect has a similar influence on group decisions. Yet another reason for groups to react more aggressively against dishonest tactics is that group members are less identifiable than individual decision makers and so cannot be assigned individual responsibility for their decisions. Lessened

5

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY identifiability causes group members to be less inhibited in their behavior after a provocation and to react more strongly (Meier et al., 2007). In line with this argument, Schopler, Insko, Drigotas, Wieselquist, Pemberton, and Cox (1995) provide evidence that nonidentifiability is positively correlated with competitive behavior in social dilemmas. In contrast to the reasoning and results described so far, prior research has also found that, compared with individuals, groups tend to act more in line with economic rationality (Bornstein, Kugler, & Ziegelmeyer, 2004; Bornstein & Yaniv, 1998; Robert & Carnevale, 1997). If punitive actions are costly and if the possibility of future interactions is ruled out, then higher levels of rationality should make groups less likely to punish dishonest behavior. In other words, their higher focus on financial self-interest—which induces groups to behave more competitively in social dilemmas (Wildschut et al., 2003) and makes them more likely to engage in deception (Cohen et al., 2009)—should make them less likely to take mutually harmful measures when they’re the target of dishonest tactics. In line with this hypothesis of economically rational self-interest, groups have been found to show lower levels of positive reciprocity when trusted by others (Cox, 2002). We might therefore expect a similar result for negatively reciprocal behavior, such as punishment for dishonesty. Research on group decisions in ultimatum bargaining games points in this direction. Bornstein and Yaniv (1998) find no difference between intergroup and interpersonal bargaining with respect to the mutually disadvantageous rejection rates of unfair ultimatum offers. Yet because the groups on average received less generous offers, these results suggest that groups are actually less likely to punish others for unfair divisions.

6

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY I shall advance two competing hypotheses based on the foregoing discussion. The first hypothesis (henceforth the escalation hypothesis) predicts that groups have a greater desire than individuals to punish dishonest behavior. Thus, groups will be more willing to incur financial costs in order to decrease the payoff of a dishonest target. This effect may be mediated by either negative affect or perceived responsibility. In contrast, the second hypothesis (henceforth the moderation hypothesis) predicts that groups are less willing than individuals to incur financial costs in order to decrease the payoffs of a dishonest target. This effect may be mediated by groups’ higher focus on their financial self-interest. It test these opposing hypotheses in two experiments. In Study 1, participants take part in a two-stage bargaining game. I find that when they are the target of a deceptive tactic, groups are more likely than individuals to spend parts of their payoff on punishing the dishonest party. In line with previous research (Sutter, 2009; Cohen et al., 2009) the results also show that groups—more often than individuals—use deception themselves in order to obtain a higher share of the bargaining pie. In Study 2, I replicate the first result and demonstrate that the greater willingness to punish dishonesty is mediated by the stronger negative affect experienced by members of a group than by individual decision makers.

Study 1 Method Research assistants recruited 250 French-speaking participants (133 men and 117 women, aged 18 to 31 years; mean age, 22 years) at a major university in Paris. The vast majority (94%) of the participants were undergraduate students. Participants received a fixed fee of €4 for their participation and had the opportunity to gain as

7

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY much as €10 extra during the study. The average total compensation per participant was €9. In total, the experiment consisted of 27 sessions with 6–12 participants taking part in each. Upon their entry into the laboratory, a research assistant assigned the participants (either individually or as groups of two) to isolated cubicles equipped with a single computer. Groups were formed by randomly matching two individual participants. Participants in each cubicle were then matched with participants in another cubicle to form a pair of players. Special care was taken to ensure that players in each pair remained anonymous to each other. Participants took part in modified, one-shot version of an ultimatum bargaining game adapted from Boles et al. (2000). Participants (individuals or groups of two) were randomly assigned to the role of either allocator or recipient. Each allocator was matched with a recipient and endowed with €10. Participants were always matched with decision makers of the same type, which means that individuals bargained with other individuals (n = 41) and groups with other groups (n = 42). The instructions informed participants about this procedure, so all participants knew about the type of their opponent in the game. Before the game started participants were required to answer 8 quiz questions which tested their knowledge of the rules of the game. If necessary participants then received further explanations from the experimenter. At the first stage of the game allocators made an offer to recipients about how to split the endowment. Recipients independently decided about their rejection thresholds—in other words, the lowest offer that they were willing accept. If the rejection threshold was lower than the offer made by allocators, then the offer was accepted and both players received their share of the pie as stipulated by allocators. Otherwise the offer was rejected, in which case allocators received nothing and

8

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY recipients received an amount equal to an ostensibly randomly determined “outside option”. At the beginning of the game, recipients were informed about the precise value of their outside option but allocators were informed only that the outside option was uniformly distributed between €0 and €8.1 Unbeknownst to participants, the value of this outside option was not determined randomly but instead fixed at €2. Before allocators made their offers, recipients had the option of sending a computer-mediated message to allocators in which they could suggest a particular division of the €10 and also make claims about the ostensible value of their outside option. All recipients were required to send a particular suggestion concerning how to split the €10, but they had the option to refrain from sending any information to allocators about the value of their outside option. Whether recipients exaggerated the value of their outside option serves as a measure of participants’ willingness to engage in deception. The general information about the rules of the game and procedure up to this point was common knowledge to all participants. After players had made their decisions, the bargaining stage of the game ended and allocators were informed about outcomes. Without the prior knowledge of allocators or recipients, the game continued with a subsequent stage. At this stage, allocators were informed about the true value (€2) of the recipient’s outside option. They then had the opportunity to “punish” recipients by choosing a fixed outcome of €2 for recipients (the value of their outside option) at a cost of €1 for themselves. Whether allocators chose to decrease the payoff of recipients serves as a measure of participants’ willingness to punish dishonest tactics. Figure 1 gives an overview of the structure of the game: [[ INSERT Figure 1 about here]] 1

Under the assumption that all players are solely interested in their own financial well-being, the optimal strategy for allocators in this case is to offer half of the pie (i.e., €5). See Boles et al. (2000) for a more detailed discussion and the proof.

9

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Finally, after allocators and recipients had made their decisions, participants individually completed a paper-based questionnaire.2 Participants rated the ethicality of sending a deceptive message (“How ethically appropriate do you think is the decision of the sender to you a message with an incorrect number?”) and their satisfaction with the bargaining procedure (“How satisfied are you with the way the bargaining with the other player was conducted?”). After completing the questionnaire participants were paid and fully debriefed. 3

Results Recipient decisions In the messages they sent to allocators, groups suggested paying significantly more money to themselves than did individuals (M = 4.46, SD = 1.26 for individuals vs. M = 5.31, SD = 1.52. for groups), t(81) = −2.77, p = .01, d = −0.62. There is no difference in the actual rejection thresholds (M = 3.35, SD = 1.16 for individuals vs. M = 3.47, SD = 1.19 for groups), t(81) = −0.45, p = .66, d = −0.10. In all cases the offers made by allocators were accepted. The results also show that deceptive messages about the value of the outside option were sent significantly more often by groups (25 of 42, 60%) than by individuals (11 of 41, 27%), p < .01 using Fisher’s exact test.4 Only 2% of groups (1 of 42) informed allocators about the true value of their outside option, whereas 32% of individuals (13 of 41) did so, p < .001. There was no significant difference between individuals and groups with respect to remaining silent about the outside option: 16 of 2

For groups, the questionnaire responses from each group member were averaged. Before the game started, participants were not aware about the punishment stage of the experiment, which could only diminish their payoffs. Therefore, I did not actually implement allocators’ punishment decisions and instead paid both allocators and recipients according to the original outcome of the bargaining game. 4 In the following analysis, all p-values for comparisons between proportions are computed using Fisher’s two-sided exact test. 3

10

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY 42 groups (38%) versus 17 of 41 individuals (42%), p = .82. The amount of money that recipients suggest being paid to them is highly correlated with the information they send about their outside option. Recipients who exaggerate the value of their outside option ask for significantly more money (M = 5.82, SD = 1.06) than those who tell the truth (M = 3.57, SD = 1.45), t(48) = 6.07, p < .001, d = 1.26, or those who refrain from sending any information (M = 4.45, SD = 1.17), t(67) = 5.15, p < .001, d = 1.95. These results strongly suggest that recipients used deception strategically to increase their share of the pie.

Allocator decisions On average, allocators offered approximately half of the pie to recipients (M = 4.93, SD = 0.98 for individuals; M = 4.94, SD = 1.19 for groups). A t-test shows that there is no significant difference between individual and group offers: t(81) = −0.06, p = .96. Of those allocators who received information about the recipient’s ostensible outside option (n = 50), only 5 (3 groups and 2 individuals) offer less than the reported value of the recipient’s outside option. However, more than a quarter (21 of 82) of the allocators offer less than what is suggested by recipients, and this behavior is significantly more likely from groups (17 of 42, 41%) than from individuals (4 of 41, 10%), p < .01. None of the individuals or groups decided to punish recipients who were truthful or refrained from sending any information. Figure 2 shows the proportion of groups and individuals who received exaggerated information about the outside option and decided to punish the dishonest recipient. [[ INSERT Figure 2 about here]]

11

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY The results show that, conditional on being lied at by recipients, groups (15 of 25, 60%) chose to reduce the payoff of recipients significantly more often than individuals (2 of 11, 18%), p = .03. The mean amount that allocators subtracted from recipients’ payoffs was €3.60 for groups and €4.00 for individuals. It is important to note that there was no significant difference between individuals and groups with respect to the average financial payoff of dishonest recipients before allocators decided whether to punish those recipients (M = 5.59, SD = 0.58 for individuals vs. M = 5.37, SD = 0.97 for groups), t(34) = −0.69, p = .49, d = −0.24).

Perceived ethicality and process satisfaction All participants rated the extent to which they considered sending a deceptive message to be ethically acceptable and also rated their personal satisfaction with the bargaining procedure. All ratings were made on a Likert 7-point scale. The data reveal that groups in the role of recipients (M = 4.32, SD = 1.58) judged sending deceptive information about their outside options to be more ethically acceptable than did individuals (M = 2.99, SD =1.58), t(81) = −3.81, p < .01, d = −0.85. In addition, more satisfaction with the bargaining procedure was reported by groups (M = 4.73, SD = 1.38) than by individuals (M = 4.04, SD = 1.45), t(81) = −2.22, p = .03, d = −0.49). These results stand in direct contrast to the ratings made by allocators. Groups who received a deceptive message (M = 2.25, SD = 0.95) judged sending false information about the outside option to be less acceptable than did individuals who received a deceptive message (M = 3.01, SD = 0.91), t(34) = 2.34, p = .02, d = 0.54. Similarly, conditional on having been the target of deception, groups (M = 2.46, SD = 1.00) also reported being less satisfied with the bargaining procedure than did individuals (M = 3.14, SD = 0.84), t(34) = 1.94, p = .06, d = 0.76.

12

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Discussion In Study 1, participants took part in a bargaining game that incorporated several important features of real-world negotiations. Participants had the opportunity to suggest a fair allocation of the pie yet also to gain an advantage by lying about their outside option, which can be viewed as a party’s best alternative to a negotiated agreement (BATNA). As argued by Thompson (2004), lies about one’s BATNA are among the most frequent examples of unethical tactics in bargaining. In addition to being unethical, lies about outside options might constitute fraud from a legal standpoint (Shell 1991). This is consistent with the experimental setting in Study 1, in which victims of such dishonest tactics are given the means to inflict some form of punishment on the dishonest party. The results of this study show that, compared with individuals, groups are significantly more likely to engage in deception in order to increase their financial gains. Moreover, as suggested by the escalation hypothesis, groups are also more willing than individuals to punish such dishonest behavior—even if doing so entails financial costs (on average, 24% of their total earnings from the game). Overall, the results demonstrate an interesting asymmetry in the effects of group decision making in the context of ethically questionable behavior. When they can benefit from dishonesty, groups consider dishonest tactics as less ethically inappropriate and are more willing than individuals to use them. However, when they are the victims of such tactics, groups consider them to be more inappropriate and are more willing to incur financial costs in order to punish the perpetrators. A limitation of Study 1 is that the results do not provide insights into the mechanism behind groups’ stronger reactions to dishonesty. Another shortcoming is that groups were always paired with other groups and individuals with individuals.

13

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Hence it is not possible to disentangle the extent to which the increased willingness to punish in the intergroup condition can be attributed to the fact that decisions were made by a group (a source effect) or that the target of the punishment was a group (a target effect). Meier et al. (2007) argue that both types of effects could be important and might operate in different ways. Supportive evidence for this hypothesis is provided by Winquist and Larson (2004), who show that both being in a group and playing against a group have a significant influence on players’ decisions in a Prisoner’s Dilemma game. A final limitation is that each group in Study 1 consisted of only two persons. Whether groups of two (dyads) are in fact comparable to groups of three or more members has been a subject of controversy in the literature (see, e.g., Kipling, 2010; Moreland, 2010). These issues are addressed in Study 2.

Study 2 Method Study 2 follows a 2 (decision maker type: groups vs. individuals) × 2 (target of punishment: groups vs. individuals), between-subject design. Research assistants recruited 118 French-speaking participants (55 men and 63 women aged 18–32 years, Mage = 22 years) at a major university in Paris. The vast majority (92%) of the participants were undergraduate students. Participants ostensibly had the opportunity to earn between €3.50 and €7.50 during the study; in fact, their real payoff was fixed at €7.50. In total, the experiment consisted of 14 sessions with 6–12 participants taking part in each. Upon their entry into the laboratory, a research assistant assigned the participants—either individually (n= 28 individuals ) or as groups of three (n = 30 groups)—to isolated cubicles equipped with a single computer. Groups were formed by randomly matching three individual participants. Participants took part in a two-

14

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY player game adapted from Gneezy (2005). In this game, one player (the “sender”) sends a message, which may be truthful or deceptive, to a second player (the “receiver”). Prior studies have already established that in this game, groups are more likely than individuals to send deceptive messages (Cohen et al., 2009; Sutter, 2009). Hence, in this study I focus solely on the reactions of receivers. Before the game started participants were required to answer 6 quiz questions which tested their knowledge of the rules of the game. If necessary participants then received further explanations from the experimenter. Unbeknownst to them, all participants were assigned to the role of receivers and the role of the sender was played by the computer. In the modified version of the game used in this study, receivers tried to guess a randomly determined number between 0 and 1000. Players who guessed the number correctly ostensibly received €7.50, and others €3.50. Receivers were informed that they were paired with another player (“the sender”) who knew the correct value of the number and would send them a message before they made their guess. Half of the individual players were informed that the sender was another individual, and half were told that the sender consisted of a group of three individuals. Likewise, groups were told that the sender was either another group of three or a single individual. Senders were described to receivers as having a financial incentive to send them the wrong number, since a wrong guess by receivers results in a higher payoff for senders. However, given that the sender’s message was the only information receivers had about the correct number, they still had a strong incentive to implement the number contained in the message.5 In the experiment, all receivers received a deceptive message with a wrong number. After making their guess, participants were 5

Only one group and two individual participants decided to pick a number other than the one contained in the message.

15

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY informed about the true value of the correct number and thus realized that the senders had given them incorrect information. Participants were then informed that the study continued with a second stage and that, as a consequence, the outcome of the game in the first stage did not have any influence on final payoffs. Instead, the computer determined the payoffs for both senders and receivers to be €7.50. Participants were made aware about this possible second stage in the initial instructions. The decision whether the experiment continued with the second stage or ended after the first stage was described as being made randomly. In fact for all participants the game continued with the second stage. In the second stage receivers were given the option of spending some of their payoffs on reducing the payoff of senders. Receivers could spend between €0.10 and €1.00 (in 10-cent intervals) to lower senders’ payoff by 4 times the amount spent. This procedure had the advantage that, prior to the decision about punishments, the payoffs of senders and receivers were equal. As a consequence, there was no possibility of the receivers’ decisions being influenced by disadvantageous inequality in the payoffs. After making their decisions, participants individually completed a paperbased questionnaire.6 In the group condition, each group member started the questionnaire by indicating what decision they would have preferred had they decided alone. Next, all participants rated the extent of their negative affect toward the dishonest senders with respect to six different items (anger, annoyance, irritation, feeling offended, feeling resentment, and happiness [inversely coded]; e.g. “How angry do you feel about the behavior of the other group [individual]?” ); alpha=0.85). Participants then responded to two questions that assessed the extent to which they focused on maximizing their financial gains (“We [I] focused on getting a high

6

For groups, the questionnaire responses from each group member were averaged.

16

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY financial outcome for us [me]” and ”We [I] focused on doing what is financially best for us [me]”; alpha = .89). Next, the participants indicated the degree to which they felt responsible for their decisions ( “To what extent do you feel responsible for the decision you made in the second game?” and “To what extent do you feel like the decision in the second game was up to you?”; alpha = .79). Finally, participants rated the level of trust they placed in the senders’ initial message (“To what extent did you trust the message to be accurate?”) and the ethicality of sending a deceptive message (“How ethically appropriate do you think is the decision of the sender to send you a message with an incorrect number?”). All ratings were made on Likert 7-point scale. After completing the questionnaire, participants were paid and then fully debriefed.7

Results Punishment decisions Figure 3 shows the mean amount which groups and individuals decided to spent on punishing senders. [[ INSERT Figure 3 about here]] A 2 (decision maker type: groups vs. individuals) × 2 (punishment target: groups vs. individuals) analysis of variance (ANOVA) reveals that groups spent significantly more on punishments than individuals, F(3, 54) = 6.30, p = .02, η2 = .10. Whether the targets of punishment were groups or individuals does not significantly affect the average amount spent, F(3, 54) = 0.01, p = .92, η2 < .01. Moreover, there is no significant interaction between the decision-maker type and the punishment target, F(3, 54) = 0.03, p = .86, η2 < .01. Consistent with this result, the data show that groups (18 of 30, 60%) punish dishonest behavior more often than individuals (8 of 7

Because participants had been deceived about the existence of human players in the role of senders, all participants were paid the maximum of €7.50.

17

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY 28, 29%), p = .02, but that groups (14 of 29, 48%) were punished just as often as individuals (12 of 29, 41%), p = .79.

Mediation As predicted by the escalation hypothesis, a 2 (decision maker type: groups vs. individuals) × 2 (punishment target: groups vs. individuals) ANOVA shows that group members (M = 3.98, SD = 1.81) reported significantly higher negative affect than individuals (M = 2.67, SD = 1.56), F(3, 54) = 8 .60, p < .01, η2 = .12. The results do not show a significant effect of the punishment target (M = 3.45, SD = 2.00 for group targets vs. M = 3.21, SD = 1.64 for individual targets), F(3, 54) = 0.32, p = .58, η2 < .01. Nor do the results show a significant interaction effect: F(3, 54) = 1.54, p = .22, η2 = .01. [[ INSERT Figure 4 about here]] Figure 4 presents the results of a mediation analysis (Baron and Kenny, 1986) with negative affect as the suggested mediator between group decision making and higher spending on decreasing the payoffs of senders. As figure 4 shows, negative affect is significantly higher for group members (path a in the figure) than for individuals, and it also significantly influences spending on punishment (path b). Consistently with the suggested mediation via negative affect, the positive effect of group decision making on the level of punishment is reduced—and actually becomes insignificant when I control for negative affect in the regression. A Sobel test confirms that this reduction (path c/c’) is significant, z = 2.79, p < .01. An alternative mediator suggested by the escalation hypothesis is the perceived responsibility for punishment decisions. Consistently with this notion of mediation, perceived responsibility was lower for groups (M = 4.40, SD = 0.90) than

18

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY for individuals (M = 5.66, SD = 0.98), t(56) = 5.01, p < .001, d = 1.36. However, a mediation analysis reveals no evidence that perceived responsibility mediates the effect of group decisions on the level of punishment (z = 1.32, p = .19; Sobel test). Contrary to what is predicted by the moderation hypothesis, there was also no difference between groups (M = 3.27, SD = 0.94) and individuals (M = 3.16, SD = 1.00) with respect to their focus on financial self-concern, t(56) = −0.43, p = .66, d = −0.11.

Individual preferences after group decision After each group made a joint decision, the group members individually indicated what decision they would have preferred had they decided alone. There is no significant difference between the level of spending on punishment preferred by group members (M = 0.41, SD = 0.36) and the actual group decisions (M = 0.40, SD = 0.42), t(29) = 0.22, p = .83, d = 0.02. Group members preferred significantly higher spending on punishment than individuals who actually decided alone (M = 0.15, SD = 0.30), t = −2.89, p < .01, d = 0.79. This indicates that the shift toward stronger punishments is not due solely to the aggregation of individual preferences into a group decision but, in addition, constitutes an actual polarization in the willingness of group members to punish dishonesty. There was no significant effect as a function of whether group members were deciding to punish individuals (M = 0.40, SD = 0.37) or groups (M = 0.42, SD = 0.37), t(28) = −0.16, p = .87, d = −0.06.

Ethicality ratings and trust in message sent by senders A 2 (message recipient: groups vs. individuals) × 2 (source of the message: groups vs. individuals) ANOVA does not indicate a significant difference between individuals

19

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY and groups with respect to trust in the message initially sent by senders (M = 5.34, SD = 0.98 for groups vs. M = 5.04, SD = 1.26 for individuals), F(3, 54) = 1.06, p = .31, η2 = .02. There is also no significant difference in trust regarding messages sent by individuals (M = 5.31, SD = 1.26) compared with messages sent by groups (M = 5.08, SD = 0.98), F(3, 54) = 0.59, p = .45, η2 = .01. Neither is there a significant interaction effect: F(3, 54) < 0.01, p = .96, η2 < .01. A 2 (message recipient: groups vs. individuals) × 2 (source of the message: groups vs. individuals) ANOVA reveals that groups (M = 2.16, SD = 0.76) perceive the behavior of senders to be (marginally significantly) less ethical than do individuals (M = 2.50, SD = 0.75), F(3, 54) = 2.96, p = .09, η2 = .03. Whether the message was sent by a group (M = 2.37, SD = 0.80) or an individual (M = 2.28, SD = 0.74) does not have a significant effect on ethicality ratings, F(3, 54) = 0.02, p = .66, η2 < .01. There is no significant interaction between the type of the recipients and the message source: F(3, 54) = 0.02, p = .66, η2 < .01.

Discussion Consistent with the findings of Study 1 and the escalation hypothesis, the results of Study 2 demonstrate that there are higher levels of mutually harmful spending on punishment when punishment decisions are made by groups. This effect is mediated by the stronger degree of negative affect which group members feel when interacting with a dishonest party. The findings also allow one to rule out such alternative explanations as diffusion of responsibility. Contrary to what is suggested by the moderation hypothesis, there is no evidence that groups focus more than individuals on their financial self-interest when deciding whether to punish dishonest behavior. An important result is that the willingness to punish dishonesty does not depend on

20

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY the target of the punishment; that is, groups are punished just as much as individuals. This indicates that group members’ greater desire to punish dishonesty can be attributed to factors specific to being part of a group rather than factors related to the target of punishment. Confirming the results of Study 1, the findings in Study 2 suggest that there is an asymmetry in the effects of group decision making on the use of and reactions to dishonesty. As shown by prior research, groups are more willing than individuals to send deceptive messages to receivers (Cohen et al., 2009; Sutter, 2009). Yet Study 2 has shown that groups are also more likely to punish the very same behavior and to perceive it as being more unethical.

General Discussion The results of two experiments demonstrated that being part of a group significantly affects group members’ reactions to dishonest behavior. As predicted by the escalation hypothesis, group members directed more negative affect toward a dishonest party than did individual decision makers, and they were more willing to incur financial costs in order to punish dishonesty. This finding stands in contrast to previous research, which has shown that groups behave more in line with economic rationality and focus to a greater extent on their financial self-interest (Bornstein & Yaniv, 1998; Robert & Carnevale, 1997; Wildschut et al., 2003). The stronger willingness to punish dishonesty also persisted in subsequent hypothetical individual decisions, and it did not depend on whether the target of punishment was an individual or a group. This suggests that taking part in the group decision polarized group members’ individual attitudes in a lasting way toward stronger feelings of negative affect and greater willingness to punish dishonest behavior.

21

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY A second important finding is that group membership increases moral hypocrisy. Valdesolo and DeSteno (2007, 2008) demonstrate that individuals tend to engage in moral hypocrisy by judging a transgression committed by themselves to be less unethical than the same transgression committed by another. The findings in this paper indicate that making decisions within a group accentuates this self-serving tendency. As demonstrated in Study 1 although groups were more likely than individuals to punish dishonest bargaining tactics, the former were also more likely to use the very same tactics themselves. Group members also perceived the same dishonest behavior as being more unethical (than did individuals) when the group was its victim but as being less unethical when they could employ dishonesty to their own advantage. The same pattern holds for the results of study 2. Previous research (Sutter, 2009; Cohen et al. 2009) has shown groups are more likely to send deceptive messages in the sender-receiver game employed in study 2. My results show that groups are also more willing to punish the use of this very same dishonest tactic, when it is employed against them. The results of this paper call for caution when delegating negotiations to groups of negotiators. Not only are unethical tactics likely to be employed more often in inter-group negotiations but when their use is discovered groups also react more aggressively. Therefore, in situations characterized by strong informational asymmetries (so that deception and lies provide substantial advantages), negotiations between groups might break down more frequently than interpersonal negotiations, leaving both parties worse-off. In such cases, the negotiation parties might consider an interpersonal negotiation format that involves a single decision maker on each side. One limitation of this research is that, in both studies, the interests of the two parties were directly opposed; however, negotiations in the real world often allow for

22

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY mutually advantageous gains and the creation of synergies (Bazerman, Curhan, Moore, & Valley, 2000). Hence one possible extension of the research reported in this paper is to explore group reactions to the use of unethical tactics in a richer setting that allows for such mutual gains. It would be interesting, in particular, to investigate the extent to which prior results in the literature—which have shown that groups outperform individuals in generating mutually beneficial deals (Polzer, 1996; Thompson, Peterson, & Brodt, 1996)—continue to hold in settings where deceptive tactics are used by negotiators. Another limitation of this paper is that it focuses only on situations in which individuals or groups are the direct target of dishonest behavior. Future research should explore to what extent group membership influences attitudes toward dishonesty (and other unethical behavior) when the victims are third parties. Kim, Cooper, Dirks, and Ferrin (2011) show that groups of judges are generally harsher than individual judges in their assessment of alleged transgressors. This increased harshness was mediated by the extent to which arguments in favor or against the transgressor were put forward during the group discussion. Future studies should, in addition, focus on evaluating hypothetical transgressions committed by the group itself and further explore the extent of moral hypocrisy and the conditions under which it is likely to occur in groups. Another interesting result of this research that is worthy of additional exploration is that group members experience more negative affect (such as anger) than do individual decision makers. Future research might attempt to explore the extent to which group membership polarizes not only anger but other emotions, such as fear or joy, as well. It would also be interesting to explore the effects of such polarization effects on group-specific variables, such as cohesiveness and the occurrence of intergroup conflict. For example, a common experience of

23

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY strong feelings of anger toward an outside target, as evidenced in Study 2, might cause groups to become more cohesive and thereby reduce the likelihood of conflicts within the group.

24

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY References Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. Bazerman, M. H., Curhan, J. R., Moore, D. A., & Valley, K. L. (2000). Negotiation. Annual Review of Psychology, 51, 279–314. Boles, T. L., Croson, R T. A. & Murnighan, J. K. (2000). Deception and retribution in repeated ultimatum bargaining. Organizational Behavior and Human Decision Processes, 83, 235-59. Bornstein, G., Kugler, T., & Ziegelmeyer, A. (2004). Individual and group decisions in the centipede game: Are groups more ‘‘rational” players? Journal of Experimental Social Psychology, 40, 599–605. Bornstein, G., & Yaniv, I. (1998). Individual and group behavior in the ultimatum game: Are groups more ‘‘rational” players? Experimental Economics, 1, 101– 108. Brandts, J., & Charness, G. (2003). Truth or consequences: An experiment. Management Science, 49, 116-130. Cox, J. C. (2002). Trust, reciprocity, and other-regarding preferences: Groups vs. individuals and males vs. females. In R. Zwick & A. Rapoport (Eds.), Advances in experimental business research. Dordrecht: Kluwer Academic Publishers. Cohen, T. R., Gunia, B., Kim-Jun, S. Y., & Murnighan, J. K. (2009). Do groups lie more than individuals? Honesty and deception as a function of strategic selfinterest. Journal of Experimental Social Psychology, 45, 1321-1324. DePaulo, B., Kashy, D. A., Kirkendol, S. A., Wyer, M. M., & Epstein, J. A. (1996). Lying in everyday life. Journal of Personality and Social Psychology, 70, 979995. Gneezy, U. (2005). Deception: The role of consequences. The American Economic Review, 95, 384–394. Hewstone, M., Rubin, M., & Willis, H. (2002). Intergroup bias. Annual Review of Psychology, 53, 575–604.

25

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Hoyle, R.H., Pinkley, R.L., & Insko, C.A. (1989). Perceptions of behavior: Evidence of differing expectations for interpersonal and intergroup interactions. Personality and Social Psychology Bulletin, 15, 365-376. Jaffe, Y, Shapir, N., & Yinon, Y. (1981). Aggression and its escalation in individuals and groups. Journal of Cross-Cultural Psychology, 12, 21-36. Jaffe, Y, & Yinon, Y. (1979). Retaliatory aggression in individuals and groups. European Journal of Social Psychology, 9. 177-186. Kim, P. H., Cooper, C. D., Dirks, K. T, & Ferrin, D. L. (2011). The manifestation of mob mentalities: How individuals and groups differ in their responses to alleged transgressions, Unpublished Manuscript. Kipling D. W. (2010). Dyads can be groups (and often are). Small Group Research, 41, 268-274. Koning , L., Steinel, W., van Beest, I., & van Dijk, E. (2011). Power and deception in ultimatum bargaining, Organizational Behavior and Human Decision Processes, forthcoming Lamm, H., & Myers, D. G. (1978). Group-induced polarization of attitudes and behavior. In L. Berkowitz (Ed.), Advances in experimental social psychology (pp. 145–195). New York: Academic Press. Meier, B. P., & Hinsz, V. B. (2004). A comparison of human aggression committed by groups and individuals: An interindividual-intergroup discontinuity. Journal of Experimental Social Psychology, 40, 551-559. Meier, B. P., Hinsz, V. B., & Heimerdinger, S. R. (2007). A framework for explaining aggression involving groups. Social and Personality Psychology Compass, 1, 298-312. Moreland, R. L. (2010). Are dyads really groups? Small Group Research, 41, 251267. Morgan, P. M., &Tindale, R. S., (2002). Group vs. individual performance in mixedmotive situations: Exploring an inconsistency. Organizational Behavior and Human Decision Processes , 87, 44–65 Pillutla M. M., & Murnighan J. K. (1996). Unfairness, anger, and spite: emotional rejections of ultimatum offers. Organizational Behavior and Human Decision Processes 68, 208–224. Polzer, J. T. (1996). Intergroup negotiations - The effects of negotiating teams. Journal of Conflict Resolution, 40, 678-698.

26

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Pruitt, D. G., Parker, J. C. & Mikolic, J. M. (1997). Escalation as a reaction to persistent annoyance. International Journal of Conflict Management, 8, 252-270

Robert, C., & Carnevale, P. J. (1997). Group choice in ultimatum bargaining. Organizational Behavior and Human Decision Processes. 72, 256–279. Schopler, J., Insko, C. A., Drigotas, S. M., Wieselquist, J., Pemberton, M., & Cox, C. (1995). The role of identifiability in the reduction of interindividual-intergroup discontinuity. Journal of Experimental Social Psychology, 31, 553 – 574. Schweitzer, M. E., DeChurch, L., & Gibson, D. E. (2005). Conflict frames and the use of deception: Are competitive negotiators less ethical? Journal of Applied Social Psychology, 35, 2123–2149. Shell, G. R. (1991). When is it legal to lie in negotiations? Sloan Management Review, 32, 93–101. Stawiski, S., Tindale, R. S., & Dykema-Engblade, A. (2009). The effects of ethical climate on group and individual level deception in negotiation. International Journal of Conflict Management, 20, 287-308. Sutter, M. (2009). Deception through telling the truth? Experimental evidence from individuals and teams. The Economic Journal, 119, 47–60. Tajfel, H., Billig, M. G., Bundy, R. P., & Flament, C. (1971). Social categorization and intergroup behavior. European Journal of Social Psychology, 1, 149–178. Tenbrunsel, A. E. (1998). Misrepresentation and expectations of misrepresentation in an ethical dilemma: The role of incentives and temptation. Academy of Management Journal, 41, 330-39. Thompson L. (2004). The mind and heart of the negotiator. Upper Saddle River, NJ: Prentice Hall Thompson, L., Peterson, E., & Brodt, S. (1996). Team negotiation: An examination of integrative and distributive bargaining. Journal of Personality and Social Psychology, 70, 66–78. Valdesolo, P., & DeSteno, D.A. (2007). Moral hypocrisy: Social groups and the flexibility of virtue. Psychological Science, 18, 689-690. Valdesolo, P., & DeSteno, D. (2008). The duality of virtue: Deconstructing the moral hypocrite. Journal of Experimental Social Psychology, 44, 1334-1338. Wang, C. S., Galinsky, A. D., & Murnighan, J. K. (2009). Bad drives psychological reactions but good propels behavior: Reactions to honesty and deception. Psychological Science, 20, 634-644.

27

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Wang, C. S.,& Leung, A. K. -y. (2009) The cultural dynamics of rewarding honesty and punishing deception. Personality and Social Psychology Bulletin, 36 1529–1542 Wildschut, T., Pinter, B., Vevea, J. L., Insko, C. A., & Schopler, J. (2003). Beyond the group mind: A quantitative review of the interindividual–intergroup discontinuity effect. Psychological Bulletin, 129, 698–722 Winquist, J. R., & Larson, J. R., Jr. (2004). Sources of the discontinuity effect: Playing against a group versus being in a group. Journal of Experimental Social Psychology, 5, 675–682.

28

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Figure 1: Structure of the bargaining game in Study 1

Recipients send a computer-mediated message to allocators containing: a) A suggestion about how much of the pie should be paid to them (0-10€). b) Information about their outside option. Afterwards recipients decide about their rejection thresholds.

Bargaining Stage

Allocators receive the message and decide about how much of the pie to offer to recipients (0-10€).

Offers and rejection thresholds are compared and allocators and recipients are informed about the outcome of the bargaining stage.

Punishment Stage

Allocators are informed about the true value of recipients’ outside option and can choose to punish recipients.

29

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY

Proportion choosing to punish .2 .3 .4 .5

.6

Figure 2: Proportion of allocators choosing to punish dishonesty in Study 1

Group decisions

Individual decisions

Notes: Includes only allocators who were the target of deception.

30

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY

.1

Amount spent on punishment ($) .2 .3 .4

.5

Figure 3: Mean amount spent on punishment in Study 2

Group target Group decisions

Individual target Individual decisions

Notes: Error bars indicate standard errors of the means.

31

Running head: GROUP POLARISATION IN REACTIONS TO DISHONESTY Figure 4: Results of mediation analysis (Baron and Kenny, 1986) Negative affect 1.31** a Decision maker: Groups vs. individuals

0.17** b c / c’ 0.25* / 0.03

Spending on punishment

Notes: N = 58. Groups are coded as 1 and individuals as 0. Numbers are unstandardized coefficients from ordinary least-squares regressions. Letters denote paths in the model. p < .05; **p < .01

32