Data was analyzed with the following script: Full interaction: UNIANOVA avgunfair BY change blueend /METHOD=SSTYPE(3) /INTERCEPT=INCLUDE /EMMEANS=TABLES(OVERALL) /EMMEANS=TABLES(change) /EMMEANS=TABLES(blueend) /EMMEANS=TABLES(change*blueend) /PRINT=DESCRIPTIVE /CRITERIA=ALPHA(.05) /DESIGN=change blueend change*blueend. Collapsed across endgroup UNIANOVA avgunfair BY change /METHOD=SSTYPE(3) /INTERCEPT=INCLUDE /EMMEANS=TABLES(OVERALL) /EMMEANS=TABLES(change) /PRINT=DESCRIPTIVE /CRITERIA=ALPHA(.05) /DESIGN=change. We collected all 1500 participants, randomly assigned to be invited to participate as part of the first 750 or second 750. In the experiment, all participants were randomly allocated to either the green or blue group, then re-randomized to either the same group or to the other group. Participants then rated how acceptable [1-7-point unnumbered Likert scale, Strongly agree, Agree, Slightly agree, Neither agree nor disagree, Slightly disagree, Disagree, Strongly Disagree] seven different allocations of bonus points were to the groups: one fair [170/170 to blue/green groups] and six unfair [210/110, 235/150, 220/180, 180/220, 150/235, 110/210]. The original authors showed that participants showed greater in-group favoritism if they had stayed a part of the same group opposed to switching groups. Consistent with the replication model of the original investigation, we created an unweighted average of the ‘acceptability’ of the six unfair allocations of points as the measure of in-group favoritism. All analyses were carried out using a 2 [group change] x 2 [ending group] ANOVA. The pre-registered prediction was that there would be no interaction but a main effect of group change. We predicted greater in-group favoritism for those who stayed in their group than those who changed groups. We analyzed the 2nd 750 participants first, followed by the first 750, finally combining them into a larger analysis of the full 1500. Second 750 We found no evidence for any of the hypothesized effects. The overall interaction (F (1, 746) < 1) and both the residual main effects of ending group (F (1, 746) < 1) and whether group was switched (F (1, 746) < 1) were all non-significant (all ps > .38). Collapsing across the ending group did not change this result, people showed the same amount of in-group favoritism if they remained in their group (M = 4.835, SD = 1.005) than if they had switched groups (M = 4.757, SD = 1.016, (F (1, 748) = .843, p > .35, d = .067, 95%CI = .21 to -.076). Thus, we were unable to replicate the original results. First 750 In the 1st 750 participants collected, we again found no evidence for any of the hypothesized effects. The overall interaction was not significant (F (1, 746) = 2.144, p = .144), the residual main effects of ending group (F (1, 746) < 1) and whether group was switched (F (1, 746) = 3.103, p = .079) were all non-significant. Collapsing across the ending group did not change this result, people showed the no different amount of in-group favoritism if they remained in their group (M = 4.868, SD = 1.027) than if they had switched groups (M = 4.739, SD = .958, (F (1, 748) = 3.193, p > .07, d = .131, 95%CI = .274 to -.013). Thus, we were unable to replicate the original results again. 1500 combined data Finally, we combined the two datasets into the full 1500-person study. Once again, we found no evidence for any of the hypothesized effects. The overall interaction was not significant (F (1, 1496) < 1), the residual main effects of ending group (F (1, 1496) = 1.485, p = .223) and whether group was switched (F (1, 1496) = 3.601, p = .058) were all non-significant. Collapsing across the ending group did not change this result, people showed no different amount of in-group favoritism if they remained in their group (M = 4.847, SD = 1.016) than if they had switched groups (M = 4.748, SD = .987, (F (1, 1498) = 3.643, p > .05, d = .099, 95%CI = .2 to -.003). This effect size was not moderated by whether the data came from the 2nd or 1st 750 participants (F (1, 1496) < 1).