Outcome Bias and risk taking in a principal agent setting 3

How to cite

Loewenfeld, Moritz. 2025. "Outcome Bias and risk taking in a principal agent setting 3." AEA RCT Registry. AEARCTR-0015445. https://doi.org/10.1257/rct.15445-1.1

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

Moritz Loewenfeld*

* Primary investigator

Registry ID
AEARCTR-0015445
DOI
https://doi.org/10.1257/rct.15445-1.1
First registered
Last updated
Status
Completed
Trial dates
2025-03-05 – 2025-03-31
Countries
Austria (Vienna)
JEL codes
D81
Registry record
http://www.socialscienceregistry.org/trials/15445

Abstract

I consider a setting of delegated risk taking. Agents choose between a first-order stochastically dominant and a dominated lottery. Principals observe choices and outcomes of both lotteries and then decide whether to award a bonus payment to the agent. The goal of this experiment is to study whether outcome bias (OB), that is a tendency to condition bonus payments on outcomes, can shape the incentives faced by agents and thereby their choices. In particular, I seek to address the following research questions. 1) Can outcome bias in bonus decisions create incentives to choose suboptimal actions? 2) Do agents anticipate the OB of principals correctly and do they adjust their choices accordingly, i.e. can outcome bias induce more choices of sub-optimal actions and thus decrease welfare? I further seek to better understand the mechanisms driving outcome bias.

Intervention

I run an online experiment on prolific. For details on the design, see the document "design_pap.pdf".

Experimental design

Agents make a number of decisions between two lotteries on behalf of the principals. Principals decide on bonus payments. Participants are randomly and permanently assigned to the role of either principal or agent. There are two treatments, reward-before and reward-after. In the reward-before treatment, principals make bonus decisions conditional on the agents' choices. In the reward-after treatment, principals can also condition bonus payments on the realized outcomes.
In addition, there are two conditions. In the robustness condition, interactions are one-shot. In the consequences condition, agents are repeatedly matched to the same principal, which implies that principals can influence the agents choices with their bonus decisions.

Design details

Randomization method
Participant are randomized into different treatments and conditions by a computer.
Randomization unit
The individual.
Clusters
The unit of observations is the individual.
Observations
I aim for 880 valid submissions.
Arms
There are two conditions (robustness, consequences), two roles (agent and principal), and two treatments (reward-before and reward-after). The target number of participants is 110 valid submissions per role and treatment in the conditions robustness and consequences.
Primary outcomes
Bonus decisions of participants in the role of principal and choices and beliefs of participants in the role of agent. For details, see the document "design_pap.pdf".

Cited by

No citation data available for this registration yet.

Keywords