In Study 2, we aim to validate the findings in Study 1 in other situations. Specifically, we focus on TPP with prisoner’s dilemma game. Previous studies have shown a tendency for punishers to receive positive evaluations[30] . However, it is noteworthy that such literature usually employed a TPP with dictator game, where non-cooperative behavior is assumed to be solely generated by greed. Thus, the effect of the motive for non-cooperation on how the punisher is evaluated awaits further exploration. In Study 2, we tested this by comparing the effects of non-cooperation motives in the PG-P and TPP. Because it is not possible to manipulate fear in a dictator game, the TPP included a two-person prisoner’s dilemma. Based on our own results from Study 1 and previous findings[30], we developed the following two hypotheses:
- Hypothesis 2-1 (replication of Study 1): In the PG-P, the impression evaluations and intention to cooperate scores will be higher for non-punishers (vs punishers), regardless of the order condition.
- Hypothesis 2-2: In the first and last conditions of the PG-P, the impression evaluations and intention to cooperate scores will be higher for non-punishers (vs punishers). In the first condition of the TPP, the non-punisher (vs punisher) will receive higher scores, while in the last condition of the TPP, the punisher (vs non-punisher) will receive higher scores.
Study 2 Methods.
Study 2 was reviewed and approved by the ethical committee at Kochi University of Technology. In accordance with the Declaration of Helsinki, all participants provided written informed consent.
We preregistered Study 2 in the open science framework (for peer review, https://osf.io/2jdmq/?view_only=6b6a391c9e7b4171aa7a7ad02fe3b4ba). Note that the first version of preregistration involved some errors in the SAS code for the analysis, which could not be analyzed properly. The following sections describe the results of our analysis with the corrected code.
Design
In Study 2, we implemented a 2 (Game: PG-P vs. TPP) x 3 (Order of non-cooperative decision: SIM vs. FIRST vs. LAST) x 2 (Target: punisher vs. non-punisher) design. Target was set as a within-factor with counter-balancing, while game and order of non-cooperative decision were set as between-factors.
Participants
Since there was no significant interaction effect in Study 1, we predicted that the effect size of the interaction in Study 2 (Hypothesis 2-2) may have also been small. Specifically, Cohen’s d was set at 0.2, with statistical power set at 0.95 or higher. According to PANGEA, the total required sample size was 600. As such, we recruited a total of 602 participants from the Japanese crowdsourcing service “Lancers” (https://www.lancers.jp). However, we did not obtain data on gender or age.
Procedure
As with Study 1, Study2 was conducted online using Qualtrics. The participants were presented with a hypothetical experimental setting, either involving a PG-P with punishment (PG-P condition) or TPP with prisoner’s dilemma and punishment (TPP condition). To assess the punishers and non-punishers, they subsequently responded with their respective impression evaluations and intentions to cooperate.
In the PG-P condition, a game situation with five players was described. Specifically, four players (A, B, C, and D) made decisions in the first stage (PG stage), while the remaining player (E) engaged in the second stage (punishment stage). The public goods stage was identical to that from Study 1, as was the manipulation of the decision-making order. The punishment stage was also similar to Study 1, except the punisher/non-punisher was an observer in the first stage (i.e., not a PG player). Games in which the observer (E) punished and the observer (F) did not punish were presented in random order. Participants then responded with their impression evaluations and intentions to cooperate with the punisher or non-punisher, respectively.
In the TPP condition, a prisoner’s dilemma with observer was described as the first stage. Players A and B received a fixed show-up fee of 500 yen, while an additional 500 yen was given as an endowment in the first stage. Any portion of the 500 yen endowment that they gave to the counterpart was doubled by the experimenter, then transferred to the counterpart. This exchange occurred only once. Similar to the PG-P condition, the order of decision-making in the first stage of the TPP condition was manipulated. In the SIM condition, the two players made decisions simultaneously, wherein one player offered the full amount, while the other kept the full amount. In the FIRST condition, the first player kept their entire sum of money, while the second player offered their entire sum of money to the first player. In the LAST condition, the first player offered their entire sum of money to the second player, who knew that their counterpart had cooperated, but kept the entire sum of money.
In the TPP punishment stage, the observer in the first stage (i.e., player C) made the decision; this player was also given a fixed show-up fee of 500 yen, and an additional 500 yen endowment. They then decided whether to use the endowment to withdraw money from the other players (A or B). For any amount of money that was spent for withdrawal, three times the amount would be withdrawn from the target person (note that such a reduction was collected by the experimenter, and not given to player C). Any endowment that was not used for punishment would be added to player C’s experimental reward. Each participant was presented with a scenario in which player C decided to punish (i.e., use all 500 yen to deduct money from the non-cooperator) in the first stage, and another scenario in which the observer (this time, player D) decided not to punish. The order of these scenarios was randomized. Finally, the participants responded with their impression evaluations and intentions to cooperate with the punisher or non-punisher, respectively.
Measurements
In Study 2, we used the same measurement items from Study 1 for the impression evaluations, intention to cooperate, and manipulation check for non-cooperation motivations. We also included the following items for exploratory purposes: (a) evaluations of the non-cooperator in the first stage (using the same items for the punisher and non-punisher), (b) intentions to punish the non-cooperator if participating in the hypothetical experiment, and (c) extent to which participants cared about equality among players in the first stage, equality among first-stage and second-stage players, and whether the non-cooperator was punished.
Study 2 Results
Manipulation checks for non-cooperation motivations
Figure 3 shows the mean estimated motivations for non-cooperation in each condition. Our 2 (Game) x 3 (Order) x 2 (Type of motive) ANOVA with estimate of the non-cooperation motive set as the dependent variables showed significant main effects for both order (F(2, 596)=23.76, p < .001, partial η2 = .074) and game (F(1, 596)=4.36, p = .037, partial η2 = .007). There were also significant interaction effects for both type and game (F(1, 596)=16.35, p < .001, partial η2 = .027) and type and order (F(2, 596)=23.98, p < .001, partial η2 = .074). However, there was no significant main effect for type (F(1, 596)=1.86, p = .173, partial η2 = . 003), no significant interaction effect for game and order (F(2, 596)=0.58, p = .559, partial η2 = . 002), and no significant interaction effect for game, order, and type (F(2, 596)=0.66, p = .517, partial η2 = . 002). The simple main effects of type revealed no difference for the type of motivation in the SIM condition (p = .834, d = .021), that fear was higher than greed in the FIRST condition (p < .001, d = .59), and that greed was higher than fear in the LAST condition (p < .001, d = .376), regardless of game type. Based on these results, the manipulation of the non-cooperation motivation was successful.
Hypothesis testing
Figure 4 shows the mean values for impression evaluation and the intention to cooperate in each condition. To test Hypothesis 2-1, we conducted a mixed-factor MANOVA with both the values for impression evaluation and intention to cooperate set as dependent variables; we employed a 3 (Order) x 2 (Target) design with independent variables using the PG-P data only. The results showed a significant main effect only for target (Wilks’s Lambda = .9702, p = .003, partial η2 = .03), meaning that the non-punisher (vs punisher) was positively evaluated. There was no significant main effect for order (Wilks’s Lambda = .9555, p = .099, partial η2 = .045), and no significant interaction effect for order and target (Wilks’s Lambda = .9813, p = .006, partial η2 = .019). These results were essentially the same as those from Study 1, and thus supported Hypothesis 2-1.
To test Hypothesis 2-2, we conducted a mixed-factor MANOVA with both the values for impression evaluation and cooperative intention set as dependent variables; we employed a 2 (Game) x 3 (Order) x 2 (Target) design with independent variables. The results showed a significant main effect for game (Wilks’s Lambda = .9771, p = .008, partial η2 = .023) and significant interaction effect for game and target (Wilks’s Lambda = .9789, p < .001, partial η2 = .021). However, the predicted interaction effect for game, order, and target was not significant (Wilks’s Lambda = .9921, p = .095, partial η2 = .008). Further, there were no significant main effects for order (Wilks’s Lambda = .9901, p = .656, partial η2 = .01) or target (Wilks’s Lambda = .9993, p = .534, partial η2 = .001), nor were there significant interaction effects for game and order (Wilks’s Lambda = .9789, p = .122, partial η2 = .002) or order and target (Wilks’s Lambda = .9971, p = .418, partial η2 = .003). These results did not support Hypothesis 2-2.
To further interpret the significant interaction effect for game and target, we exploratorily conducted a MANOVA with target set as the independent variable, both for the PG-P and TPP. The results showed that punishers were less positively evaluated than non-punishers in the PG-P (Wilks’s Lambda = .9708, p = .003, partial η2 = .029), but more positively evaluated than non-punishers in the TPP (Wilks’s Lambda = .9858, p = .038, partial η2 = .014).