Robustness analyses of logistic regression results.

The table shows hierarchical four-level logistic regression estimates on whether (1) or not (0) a relation is present after an evaluation moment in the four conditions of our experiment. The first columns depict the main results. Subsequent columns show the results if we employ alternative assumptions in our econometric model. The alternative assumptions concern the length of the evaluation moment (how long do subjects take to evaluate a relationship, original assumption being one second), the way the crucial actor is defined in case two actors decide to keep a link (original definition is the one who initiated the link, alternative definition being the one with the lowest earnings from the link), and the definition of the reference group (original assumption being all others, alternatives are only those to whom one is directly connected (Network) or only the subject with whom one is currently considering a relationship (Link)). Random effects are estimated at the level of the session, the decision maker, and the directed dyad under consideration. To keep the table readable, we only report coefficients relating to the hypotheses, the overall number of decisions, and the log likelihood. Definitions of variables are as in Table 2. P-values are in parentheses.