Fig 1.
Schematic representation of macro concepts of fairness (shaded in gray) based on the relation between individual endowments and outcomes.
Fig 2.
The relation between individual endowments and final scores in the experiment suggests that outcomes are less dispersed when they are visible (O and EO) and that the distinctions between the least endowed and the rest are most palpable when no information about others is available (NI).
The figure depicts mean final score with 95% confidence intervals for each endowment level per group (different marker symbols correspond to different groups). Scores are centered on the mean score of the players with Endowment = 4.
Fig 3.
The statistical dispersion of outcomes is lower when outcomes are visible (O and EO).
The figure shows means and 95% confidence intervals for the Gini coefficients of players’ scores in the treatments with no information (NI), visible endowments (E), visible outcomes (O), and visible endowments and outcomes (EO). The dashed line represents the Gini coefficient expected from the distribution of endowments.
Fig 4.
Differences between the outcomes of the least endowed (Endowment = 2) and the average endowed (Endowment = 4) are lowest when endowments are visible (E and EO).
The figure shows means and 95% confidence intervals for the Cliff’s delta statistic for scores between players with endowment two and four (left) and four and six (right) in the treatments with no information (NI), visible endowments (E), visible outcomes (O), and visible endowments and outcomes (EO). The maximum value of 1 occurs when all the players with the higher endowment have higher final scores than all the players with the lower endowment; 0 occurs when the players with the higher endowment are equally likely to have higher as lower final scores than the players with the lower endowment.
Fig 5.
We find statistical evidence for higher reciprocity in the absence of information about others (NI), charity towards the poor when outcomes are visible (O and EO), and charity towards the least endowed when endowments are visible (E and OE).
The results also show inequity aversion towards the poorer when outcomes are visible (O and EO) but higher reciprocity towards the better endowed when endowments are visible (E and EO). The figure shows effect estimates and 95% confidence intervals from meta-analyses with fixed-effect regression models of log-odds coefficients and standard errors estimated in temporal exponential random graph models. The vertical bars show group-specific estimates.
Fig 6.
Participants report using simple reciprocity most often when no information is available (NI).
When endowments are visible (E and EO), reciprocity specifically with the advantaged becomes a prominent strategy too. Helping the disadvantaged and giving in order to equalize outcomes but also to prevent oneself from falling behind are most prominent when both endowments and outcomes are visible (EO). The figure shows the proportion of participants in each treatment who mention one of the listed strategies in their free-text-entry response to the question “What was your strategy in the game?” The error bars show 95% confidence intervals.