Fig. 4: Consequences of partner choice on cooperation and resource distribution.

a Cooperation (measured as average contributions to the public good as a percentage of participant’s individual endowment) remained more stable in the partner choice condition (green line, n = 21 groups) compared to the assigned partner condition (black line, n = 21 groups; left panel). The right panel shows that participants differed in their cooperation rate towards partner types. While participants contributed more to the public good when paired with participants who were assigned a high endowment and high productivity factor (HH type), they cooperated less when paired with participants who were assigned a low endowment and low productivity factor (LL type) in the partner choice condition (n = 21 groups). The dotted line shows the average share of units invested in the public good in the partner choice condition (60.9%). b Partner choice increased resource disparity. Upper panel: accumulated resources per participant type (i.e., total number of units earned at the end of the experiment) in the partner choice (green bars, n = 21 groups) and assigned partner condition (grey bars, n = 21 groups). HH types benefitted from partner choice (p < 0.001), whereas LL types earned significantly less (p = 0.005) under partner choice compared to the assigned partner condition. Differences in accumulated resources were tested using a multilevel model (two-sided, Bonferroni corrected for multiple testing). Lower panel: Ex-ante resource distribution per type (left), received share of generated resources per type in the assigned partner condition (middle), and received share of generated resources in the partner choice condition (right) per type (blue = LL, red = HH, dark grey = HL, light grey = LH). Error bars and bands indicate the standard error of the mean. **p < 0.01, ***p < 0.001. Dots show averages (a) and sums (b) per participant.