The impact of delegation on the frequency of lying: Theory and experimental evidence

How to cite

Bieberstein, Frauke von; Feess, Eberhard. 2022. "The impact of delegation on the frequency of lying: Theory and experimental evidence." AEA RCT Registry. AEARCTR-0009443. https://doi.org/10.1257/rct.9443-1.1

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

Frauke von Bieberstein* , Eberhard Feess

* Primary investigator

Registry ID
AEARCTR-0009443
DOI
https://doi.org/10.1257/rct.9443-1.1
First registered
Last updated
Status
Completed
Trial dates
2022-05-17 – 2024-05-31
Countries
Switzerland (), New Zealand ()
Registry record
http://www.socialscienceregistry.org/trials/9443

Abstract

Subjects in our online experiment are randomly assigned to groups of three. All group members observe the same outcome of a lottery, which can be HIGH or LOW. The group receives the winning price if and only if the group member in charge reports HIGH. Our experiment has three treatments. (i) In treatment SINGLE, each of the three group members is asked to report the outcome. Each of their reports is then chosen with the same probability of one third to determine the payoff. (ii) In treatment DELEGATION, one subject is randomly chosen to be the first mover (FM). The FM can either report HIGH or LOW, but can also delegate the report to a second mover (SM) who is also randomly chosen. This SM needs to decide between HIGH and LOW.SMs are asked for their report under the assumption that they can actually make a report (i.e. that the FM delegates). (iii) In treatment MULTISTAGE DELEGATION, the SM can delegate as well. If so, the third mover (TM) is asked for the final report. Our main research question is how the frequency of HIGH final reports differs among treatments. We have developed a game theoretical model with moral costs, which occur if and only if the final report is HIGH when the outcome of the lottery is LOW. In our model, moral costs are highest when the own report is HIGH compared to a situation where another group member reports high. This sets incentives to delegate the decision. Our model yields the result that the order of the frequencies among the three treatments depends on the parameters. In particular, it is theoretically possible that the order is non-monotone. Our study is hence not hypothesis-based but explorative.

Intervention

Online experiment. Between-subject design. In treatment SINGLE, all three subjects are in the same role. In the other two treatments, subjects are randomly assigned to the three roles. Each subject makes just one decision, reporting HIGH or LOW, or delegating the report (for FMs in treatment DELEGATION and for FMs and SMs in treatment MULTISTAGE DELEGATION). Subjects also fill out a non-incentivized questionnaire where we ask for gender, age, the belief about how many subjects choose HIGH or LOW when the outcome is LOW (and delegate when applicable), preferences for delegating and being delegated to, and willingness to take risks.

The study will be conducted on Amazon MTurk, a large online platform for surveys and market research. We restrict the sample to participants whose country of residence are the USA, who have at least 500 approved assignments, and whose submissions have been approved in at least 95% of the cases. We impose these restrictions to ensure a high data quality.

Experimental design

Subjects in our online experiment are randomly assigned to groups of three. All group members observe the same outcome of a lottery, which can be HIGH or LOW. The group receives the winning price if and only if the group member in charge reports HIGH. Our experiment has three treatments. (i) In treatment SINGLE, each of the three group members is asked to report the outcome. Each of their reports is then chosen with the same probability of one third to determine the payoff. (ii) In treatment DELEGATION, one subject is randomly chosen to be the first mover (FM). The FM can either report HIGH or LOW, but can also delegate the report to a second mover (SM) who is also randomly chosen. This SM needs to decide between HIGH and LOW.SMs are asked for their report under the assumption that they can actually make a report (i.e. that the FM delegates). (iii) In treatment MULTISTAGE DELEGATION, the SM can delegate as well. If so, the third mover (TM) is asked for the final report.

Design details

Randomization method
Computer (online experiment)
Randomization unit
Experimental session for the treatment and individual for the specific role
Clusters
Same as observations.
Observations
About 4’200 finishing the task We will exclude subjects that: - do not complete the MTurk task within 45 minutes of starting; - exit and then re-enter the task as a new subject (as these individuals might see multiple treatments); - are not approved for any other reason (e.g. not having a valid MTurk ID); - do answer two or more control questions incorrectly: these subjects do not take part in the decision task and are transferred to a dropout page. This procedure is explicitly described at the start of the instructions.
Arms
600 (treatment SINGLE) and 1’800 (both delegation treatments); given that previous literature found gender differences in lying behavior, we aim to have about 50% women and 50% men in each treatment.
Primary outcomes
All our outcomes relate to the case when the outcome of the lottery is LOW (we can observe the outcome of the lottery, which is revealed to the subjects). Our primary outcome is the final frequency of HIGH reports, thereby taking the frequencies of delegation and the behavior of delegates into account.
Secondary outcomes
Our first secondary outcome is the comparison of the behavior of FMs in treatment DELEGATION and SMs in treatment MULTISTAGE DELEGATION (these subjects are, from an incentive perspective, in identical situations). Our second secondary outcome is the comparison of the behavior of SMs in treatment DELEGATION, TMs in treatment MULTISTAGE DELEGATION and the subjects in treatment SINGLE (these subjects are, from an incentive perspective, in identical situations We control for gender, age, the belief about how many subjects choose HIGH or LOW (and delegate when applicable) and for other variables taken from the questionnaire as well as for passing an attention checker in the questionnaire.

Cited by

No citation data available for this registration yet.

Keywords