Guilt Aversion in Multi-Agent Settings

How to cite

Lenz, Lisa. 2018. "Guilt Aversion in Multi-Agent Settings." AEA RCT Registry. AEARCTR-0002901. https://doi.org/10.1257/rct.2901-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Lisa Lenz*

* Primary investigator

Registry ID
AEARCTR-0002901
DOI
https://doi.org/10.1257/rct.2901-1.0
First registered
Last updated
Status
In development
Trial dates
2018-04-17 – 2018-07-01
Countries
Private
Registry record
http://www.socialscienceregistry.org/trials/2901

Abstract

This research project addresses experimentally as well as theoretically whether decision makers state more selfish decisions if they are protected by the anonymity of the mass and the consequences of their decisions affect a collective of economic agents compared to a situation where they interact with solely one agent?

This presumed increase in selfish behavior is backed up by predictions derived from an adapted versions of the simple guilt and the guilt from blame model by Batigalli and Dufwenberg (2007). In other words, differences in how decision makers experience feelings of guilt and shame for not living up to the expectations of others may differ in multi-agent settings. The concept of guilt offers four different behavioral channels for a potential increase in selfish behavior.

A dictator game experiment is ran on mturk to approach the main research question as well as to disentangle the different potential behavioral channels. In order to address the main hypothesis, the number of dictators & recipient within one group is varied. In order to discriminate between different behavioral channel, the strategy method (dictators state their decisions contingent on different expectations of recipients) is applied, the post-experimental feedback is varied and first order beliefs about the behavior of other dictators are elicited.

Intervention

I randomly assign participants to the different treatment conditions.
The exogenous varied variables are the group size of the dictator setting as well as the post experimental feedback concerning the dictators' decisions. Furthermore, I let dictators state their decisions contingent on different potential beliefs of recipients by applying the strategy method.

Experimental design

Baseline Condition:
--------------------------

Participants will be assigned either type A (dictator) or type B (recipient).
Every type A is randomly matched with a B-participant. Every participant A can decide how much he/she wants to take of the pot that contains $1.00. Participant typ B earns whatever is left in the pot.

Treatment Condition 1 :
--------------------------

8 participants will be assigned either type A or type B. The 4 A-participants and the 4 B-participants interact.
There is a common pot that contains $4.00.Every A-participant can decide how much he/she wants to take out of the common pot up to an amount of $1.00 per participant.The amount of money that remains in the pot after every of the 4 type A participants made the decision will be split equally among the participants of type B.
Participants of type B cannot act.

Treatment Condition 2 :
--------------------------

8 participants will be assigned either type A or type B. The 20 A-participants and the 20 B-participants interact.
There is a common pot that contains $20.00.Every A-participant can decide how much he/she wants to take out of the common pot up to an amount of $1.00 per participant.The amount of money that remains in the pot after every of the 20 type A participants made the decision will be split equally among the participants of type B.
Participants of type B cannot act.

Further Variations:
--------------------------

In Treatment Condition (1) and (2) I further vary the post-experimental feedback. In one condition, B-particpants receive information about how much an average A-participants get or receive information about the distribution of A-players' decisions.

Design details

Randomization method
randomization done in office by a computer
Randomization unit
individual level
Clusters
5 different treatments with 80 dictators and 80 recipients each = 800 participants in total
Observations
400 dictators with 11 different decisions = 4400
Arms
5 different treatments with 80 dictators and 80 recipients each = 800 participants in total
Primary outcomes
dictators' distribution condition, recipients' beliefs about dictator's distribution condition
Secondary outcomes
dictators' beliebst about other dictator's distribution conditions

Cited by

No citation data available for this registration yet.

Keywords