Experimental Design Details
Two additional treatments are under consideration for potential mechanisms:
1) A ‘free information’ treatment: PB is told PA’s offer before PB chooses to accept/reject. PB cannot choose not to receive this information. If behaviour here is different to the information treatment, then we can conclude that the information treatment’s behaviour is at least in part due to PB’s [check / don’t check] choice rather than their actual state of information.
2) A ‘pivotal information’ treatment: the rejection payoff is changed to (12 for PA, 7 for PB). A PA who checks and finds out they got an offer of 5 will reject with certainty (because they can earn 7). This increases the value of information to PB, and hence increases the signal sent to PCs by PB’s who choose not to read. PCs may choose to treat PB’s who don’t read but accept more harshly than those who choose to accept when no information was available.
________________
*Planned additions: May 2024
We’ve run the baseline and information treatment. In addition, we’ve run the ‘pivotal’ treatment mentioned in the hidden section of ‘experimental design’. Each treatment has approx. 450 observations. These were completed in January 2024.
A summary of our findings is as follows:
• PA’s behaviour did not change significantly across the baseline and information treatment. PA’s are more generous in the pivotal treatment.
• PC’s behaviour does not change significantly between the baseline and information treatment.
• PC’s behaviour does not change significantly between the baseline and pivotal treatment with regards to PB’s who did not check PA’s offer in the latter treatment. However, PC’s are significantly less generous towards PB’s who checked PA’s offer and accepted the ungenerous $5.
Hence, we find evidence that PCs are sensitive to PB’s outside option, and thus to the meaningfulness of the information available to PB, but only towards those who choose to check.
As PB’s who do not check are not held to a higher standard even in the pivotal treatment, we hypothesize that responsibility for obtaining information by PB may be considerably diffused due to the presence of PA, who becomes responsible for being nice towards PB. While PB’s who obtain information dispel any responsibility from PA, PB’s who choose not to obtain information remain under the ‘care’ of PA. Hence, we propose two additional treatments (approx. 450 obs. each) where PA’s role is removed:
‘No PA’ Baseline
• Computer draws a blue ball (67% chance) or red ball (33% chance). PA gets 11 or 15 respectively.
• Player B (PB) can Continue or Stop without knowing which ball was drawn.
Stop payoff for PB: 7
Continue payoff for PB: 9 if blue, 5 if red
• If the red ball is drawn and PB Continues, Player C (PC) is given a chance to decide the final payoffs for PA and PB. PC chooses one of the following three:
(15, 5); (13, 6); (11, 6)
PC is a spectator and their decision does not affect their own payoff.
• After these decisions, each player is given the same set of questions on Beliefs, Normative Expectations and demographic questionnaire. The Beliefs and Normative Expectations questions are incentivised. The latter is incentivised using the Krupka and Weber (2013) method.
‘No PA’ Information treatment
The only difference between the baseline and information treatment is that prior to deciding whether to accept or reject the offer, PB can choose to check or not check the ball that was drawn. To check, PB has to complete a decoding task that takes on average, 2-3 minutes of their time. If the task is completed, PB finds out which ball was drawn prior to deciding to reject or accept the offer. Otherwise, PB chooses as per the baseline treatment.
Our Primary and Secondary Outcomes (End Points) comparisons remain the same as per the original baseline and information treatments.
_____________
**Minor update: June 2024
We previously indicated in “planned additions: May 2024” that in the new treatments, PC continues to choose from the original allocations: (15, 5); (13, 6); (11, 6). Upon further deliberation, we have decided instead to change the alternatives to: (15, 5); (11, 7). This was done for several reasons:
In the new treatments, PC is unlikely to gain utility from punishing/harming PA. Hence (11, 6) is unlikely to be picked. We changed this to (11, 7). The alternative (13, 6) was removed because helping PB by $1 seems trivial now that harming PA is no longer rewarding. We restricted allocations to PB to be no greater than 7 to ensure that PB’s outside option of 7 remains (weakly) dominant (in a payoff-sense) to PC’s ability to redistribute to PB.