Fig. 1: Partner choice under inequality.

a Participants differed on two characteristics: endowment (H = 75 Units, L = 25 Units) and productivity (H = 1.7, L = 1.3), creating four different participant types: high endowment and high productivity (HH), low endowment and high productivity (LH), high endowment and low productivity (HL), and low endowment and low productivity (LL). b In each round, two participants were paired and decided how many units of their endowment to contribute to their shared public good. Each contributed unit was multiplied by the productivity factor of the participant and the resulting investment was divided equally between the pair. c Between treatments, we manipulated how pairs were formed. In the assigned partner condition, participants were pseudo-randomly assigned to different types across rounds, such that everyone interacted with each partner type equally often. In the partner choice condition, participants, on each round, first indicated their partner preference by ranking all types from 1 (most preferred) to 4 (least preferred). Based on their ranking, participants were paired with a partner to play the public goods game with, for that round.