Sampling for Others in a Principal-Agent Setting

Last registered on August 06, 2024

View Trial History

Pre-Trial

Trial Information

General Information

Title

Sampling for Others in a Principal-Agent Setting

RCT ID

AEARCTR-0011849

Initial registration date

July 27, 2023

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published

July 28, 2023, 2:06 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Last updated

August 06, 2024, 9:43 AM EDT

Last updated is the most recent time when changes to the trial's registration were published.

Locations

Country

Germany

Region

Primary Investigator

Name

Christoph Becker

Affiliation

Heidelberg University

Contact Primary Investigator

Other Primary Investigator(s)

PI Name

Stefan Trautmann

PI Affiliation

Heidelberg University

Contact Investigator

Additional Trial Information

Status

In development

Start date

2023-08-08

End date

2024-12-30

Keywords

Behavior, Lab

Additional Keywords

Sampling, principal, information, learning

JEL code(s)

D83, C91

Secondary IDs

Prior work

This trial does not extend or rely on any prior RCTs.

Abstract

When making a choice in real-world situations such as investing or making larger purchases, decision makers often trust others’ experience. Commonly, such as in financial advice, there is a principal-agent relationship between decision maker and advisor, with the advisor collecting information by experience for one or multiple clients. Literature on decisions from experience so far has focused mostly on individual decision making, i.e. situations where decision makers gather information for themselves and then make a choice. We aim to provide insights into the linkage between information gathering and actual decision making, when both processes are handled by different persons. We test the fundamentals of such processes in a controlled laboratory setting. We vary the degree of responsibility agents have to collect information by having them either i) just gather information, ii) also provide a recommendation, or iii) also make a choice for the client. We contrast these results to decision makers collecting information and making decisions for themselves.

External Link(s)

Registration Citation

Citation

Becker, Christoph and Stefan Trautmann. 2024. "Sampling for Others in a Principal-Agent Setting." AEA RCT Registry. August 06. https://doi.org/10.1257/rct.11849-3.0

Sponsors & Partners

Experimental Details

Interventions

Intervention(s)

Different modes of collecting information for self or for others.

Intervention (Hidden)

Participants will sample information from a previously unknown distribution. The collect this information either for themselves to inform their own decision or for other participants who in turn use it to inform their decision.

Intervention Start Date

2023-08-08

Intervention End Date

2024-12-30

Primary Outcomes

Primary Outcomes (end points)

(i) Number of samples drawn in each task.
(ii) Which lottery (out of two possibilities) was chosen in ach task.

Primary Outcomes (explanation)

(i) Participants draw samples from two lotteries. The samples either inform themselves or another participant (depending on the treatment) on which lottery to choose.
(ii) We then record the final choice participants either make for themselves or others.

Secondary Outcomes

Secondary Outcomes (end points)

(i) Demographics (age, gender, field of study, risk attitude)

Secondary Outcomes (explanation)

(i) Answers will be used as control variables.

Experimental Design

Experimental Design Details

Participants in the experiment are faced with five tasks that each consist of a choice between two risky lotteries. Participants initially have no information bout the outcomes and probabilities of the lotteries, but can acquire additional information, depending on their specific treatment (see below). Collection information is done by clicking one two buttons, each representing on of the lotteries. Each time a button is clicked, the software randomly draws an outcome from the respective lottery (with replacement) and shows it on the screen to the participant. Only the last sampled outcome is shown, hence participants have to memorize outcomes they sampled in the past.

The experiment is divided into two broad conditions: The Agent condition and the Client condition. In the Agent condition, participants are randomly assigned to one of four treatments: Self, Sample, Recommendation and Decision. Participants in the Self treatment sample information in the above described manner and then make a final and potentially payoff-relevant decision for themselves. At the end of the experiment, one of the five tasks is randomly selected. An outcome from the lottery chosen by the participant is then realized. The participant receives the realized outcome plus a participation fee.

In the Sample, Recommendation, and Decision treatments, the participants sample information for the participants in the client condition. In the Sample treatment, the sampled information is saved to be transferred to one of the participants in the client treatment later. The Recommendation treatment works similarly, however, in addition to the sample, the participant also is asked to recommend on of the two lotteries to the client. In the Decision treatment, finally, the participant makes a final and potentially payoff-relevant decision between the two lotteries for the client.

Participants in the Client condition receive the sample, recommendation and decision of the participants in the agent condition. Clients play three tasks: In the first one they receive the sample from one of the agents, in the second one they receive the sample plus the recommendation, and in the third, the decision of one of the agents is implemented for them. This design ensures a degree of agency also for the clients, rather having some of them only receive decisions of one of the agents. We match one agent each from the Sample, Recommendation, and Decision treatment to each client. This means that one random task of every agent is for sure used for a client.
Participants take part in the experiment individually and online. Participants in the Agent condition transmit their samples, recommendations, and decisions that are saved and then provided to clients in a later session. Participants are recruited using the subject pool of the Alfred-Weber-Institute for Economics of Heidelberg University.

Randomization Method

Randomization of treatments done by the experimental software.

Randomization Unit

Individual

Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters

500 Laboratory participants

Sample size: planned number of observations

500 Laboratory participants

Sample size (or number of clusters) by treatment arms

100 participants per treatment: 100 participants in Self, 100 participants in Sample, 100 participants in Recommendation and 100 in the Client condition. Participants in the Client condition see data from 3 participants of the Sample, Recommendation, and Decision treatment respectively (hence 3:1 agents to clients).

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)

There is no clear theoretical prediction how many samples should be drawn. In order to identify differences of 5 Samples between the different conditions (assuming a standard deviation of 12.5 based on previous research for participants sampling for themselves), we would need around 100 participants to achieve 80% Power at an alpha=0.05.

Supporting Documents and Materials