Primary Outcomes (end points)
The key outcome variable is the average probability that the manager will hire the older team member, p_old.
We know from the auxiliary experiment ("The Age Factor I" AEARCTR-0014972) that older workers perform worse on average than younger ones (however, the distribution of performance overlaps). It therefore maximizes expected profits to discriminate statistically against older workers.
From Experiment 1 ("The Age Factor I" AEARCTR-0014972), in which managers' beliefs were elicited, we know that managers in T3 JointInfo and T4 SepInfo statistically discriminate against older employees. However, managers in T3 still overestimate the average performance of older employees, while in T4, average performance is estimated almost exactly correctly.
In this experiment, we now examine how the beliefs in the different information treatments translate into actual behavior. Because T1 NoInfo offers no gain compared to T2 AgeInfo with respect to the hiring decision we are interested in, we do not conduct a hiring analogue of T1. Instead, we introduce a treatment T4b SepInfo, in which the manager's hiring decision influences not only his own payout but also the worker's payout, in order to control for social preferences.
In T2 AgeInfo, the managers have information that one of the two workers is older, but they do not know which. Therefore, on average, p_old should be 50% (H2).
In T3 JointInfo, the managers have information that one of the two workers is older, but they only know the joint distribution of the performance. So here, beliefs about F(X)_old and F(X)_young play a role and due to biased beliefs, we hypothesize that on average p_old=0% (belief based statistical discrimination). (H3).
Finally, in T4a and T4b SepInfo, false beliefs about age-related performance can be corrected. However, since the average performance of older worker is worse in the Word Encryption Task, this leads to perfect stochastic discrimination, p_old=0%. (H4a) In treatment 4b, social preferences could play a role, as workers' payouts are also affected by the hiring decision (H4b).
Thus, our working hypothesis in comparing the treatments is that as soon as the age of the applicants is identifiable, their chances of being hired drops to zero (or close to zero). Furthermore, in this case (comparison of JointInfo vs. SepInfo), it does not matter whether the managers know that older workers perform actually worse on average or only believe this to be the case.
At the within-subjects level, the observation over 10 rounds also tests whether the strength of hiring discrimination correlates with the level of overall performance, i.e., whether the older worker of stronger teams is systematically attributed an even smaller share of success and therefore has a lower hiring probability.
As explained in more detail below in the experimental design, after completing the 10 rounds, managers are presented with two of the hiring decisions again. The actual performance of both workers in a team is revealed, with the older worker performing better in one team and the younger worker performing better in the other. Managers can now revise their original hiring decision or stick with it. Four possible scenarios are conceivable.
1) The manager always hires only the better candidate. => No taste based discrimination (profit-maximization and meritocratic motive go hand in hand).
2) The manager always hires only the younger candidate. => Taste based discrimination against older workers (costly for the manager in the situation where younger worker performed worse).
3) The manager always hires only the older candidate. => Taste based discrimination against younger workers (costly for the manager in the situation where olderr worker performed worse).
4) The manager always hires only the worker who performed worse. => May indicate social preferences, anti-meritocratic attitudes, etc.
Our exploratory hypothesis is that all four outcomes occur, but that overall there is little taste-based discrimination. Comparing T4a and T4b, we hypothesize p_old|T4a < p_old |T4b.
Correlational Hypothesis:
We also hypothesize that both stochstical and taste based discrimination against older workers is more pronounced the younger the manager is.