Figure 3. The effect of generalized reciprocity on the willingness to donate in the no-observation condition.

To facilitate interpretation of the odds ratios, the figure shows the estimated donation rate and 95% confidence intervals based on a random-intercept linear regression model with robust standard errors corresponding to the random-intercept logistic model in Table 1A. The robust standard errors adjust for possible heteroskedasticity with a binary dependent measure. The dashed line shows the baseline donation rate among seeds in the no-observation condition. The donation rate is significantly higher among invitees than among seeds after receiving one donated invitation but does not continue to increase with receipt of additional invitations.