Abstract
Cooperation in the Public Goods Game (PGG) is determined by a mixture of individual differences (e.g. personality, social preferences) and group dynamics (e.g. reciprocation, social norms). However, to our knowledge, no thorough attempt has been made to separate individual and group levels of cooperation and to quantify the variance in cooperation that can be attributed to the group level. In an analysis of 10 open datasets (total N = 4,556, 1,003 groups, 7-50 rounds), we chart the trajectory of individual and group-level variance across rounds of repeated PGGs. We find that the portion of group-level variance increases initially and plateaus around the fifth round, typically at a level between 20 and 50%. In addition, we identify four factors that increase the portion of group-level variance: (i) punishment opportunities; (ii) detailed feedback including all group members' decisions; (iii) small groups (≤4 players); and (iv) groups with homogenous social preferences.