Commitment discourages contagious dishonesty

Last registered on August 14, 2024

Pre-Trial

Trial Information

General Information

Title
Commitment discourages contagious dishonesty
RCT ID
AEARCTR-0014175
Initial registration date
August 13, 2024

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
August 14, 2024, 2:49 PM EDT

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

Locations

Region

Primary Investigator

Affiliation
Soka University

Other Primary Investigator(s)

PI Affiliation
Hosei University
PI Affiliation
University of Tokyo
PI Affiliation
Osaka University

Additional Trial Information

Status
In development
Start date
2024-08-21
End date
2024-09-09
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
In this study, we investigate the spillover mechanisms of cheating by conducting an incentivized online experiment. In this experiment, the participants worked in pairs on tasks that benefited from cheating. We investigate the extent to which one of these participants becomes more likely to behave honestly as the other participant becomes more likely to behave honestly. We focus on introducing commitment devices that can eliminate the option of cheating in advance. The introduction of commitment devices curbs tempting behavior–in this case, cheating. By randomly dividing participants into a treatment group with a commitment device and a control group without it, we estimate the effect of introducing commitment devices on curbing the chain of cheating. Furthermore, we estimate the extent to which participants’ beliefs about their partner’s dishonesty while performing the cheating task influences the creation of a cheating chain.
External Link(s)

Registration Citation

Citation
Hayashi, Reona et al. 2024. "Commitment discourages contagious dishonesty." AEA RCT Registry. August 14. https://doi.org/10.1257/rct.14175-1.0
Experimental Details

Interventions

Intervention(s)
Intervention Start Date
2024-08-21
Intervention End Date
2024-09-09

Primary Outcomes

Primary Outcomes (end points)
The aim is to determine the extent to which the introduction of commitment devices discourages cheating and the chain of cheating. To do so, (i) we will measure the degree of cheating in the control and treatment groups by the participants’ win probability in the cheating task. (ii) Participants will engage in four types of cheating tasks: individual cheating tasks, cheating tasks with a computer partner, cheating tasks with a human partner, and cheating tasks with a human partner with sequential feedback. To measure the extent to which the introduction of the commitment device discourages cheating behavior, we will use the participants’ win rates in individual cheating tasks and cheating tasks performed with a computer partner. To measure the extent to which the introduction of a commitment device curbs the chain of cheating, we will use the participants’ win rates in cheating tasks performed with a human partner, with and without sequential feedback.
Primary Outcomes (explanation)
Details are provided in the Pre-Analysis Plan.

Secondary Outcomes

Secondary Outcomes (end points)
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
We will conduct an online survey, with the first survey scheduled for August 22–26, 2024 and the follow-up survey scheduled for August 29 to September 2, 2024. We intend to collaborate with a survey company, My Voice Communications, Inc., to recruit participants and collect data. The target participants are men and women from the age group of 20s–70s. To ensure representativeness of the survey, the sample is designed to align with the most recently available data on the age and gender distribution of the Japanese population.
The target sample size for the main experiment is 1,500 participants. There will be approximately 750 participants in each of the two groups (control and treatment groups).

Main survey/experiment

We intend to ask the participants several questions before the cheating task to collect individual background information, including gender, age, education level, spouse, children, employment sector, income class, device used to answer the survey, and experienced well-being.

In the cheating task, we randomly assign participants to the control group, in which they have no commitment device, or to the treatment group, in which they have commitment devices at all times. In the control group, the participants play matching pennies against a computer. The computer shows Heads with a probability of 1/2 and shows Tails with a probability of 1/2. Thus, regardless of the strategy used, the participant wins with a probability of 1/2 and receives a reward and loses with a probability of 1/2 and receives no reward. However, the participants can change the flipped side of their coin after seeing the flipped side of the coin chosen by the computer. Even if a participant loses, the participant can still receive a reward by dishonestly reporting a different side of the coin from the one that the participant had originally chosen. This means that the more dishonest participants are, the more likely they are to win by more than 1/2, and this degree is measured as the degree of dishonest behavior. In the treatment group, participants will play similar matching pennies, differing only in that participants have a commitment device by which they can determine the side of their coin before they see the computer’s side of the coin.

Participants will then work with their partners during the cheating tasks. In the control group, both the participants and partners will play matching pennies with the computer. Both the participants and their partners receive a reward if they win, and will not receive a reward if either or both lose. First, the participants will be assigned a computer partner who always honestly reports the coin side and wins with a probability of 1/2. Thus, if a participant honestly reports on the side of the coin, the win rate is 1/4. If the participant dishonestly reports the side of the coin, the win rate is 1/2. In the treatment group, the participants will play similar matching pennies, differing only in that they have a commitment device for the individual cheating task.

First. all the participants are briefed on the details of the cheating task and practice individual cheating and cheating tasks with a computer partner. No reward is given for this practice. Subsequently, the participants will work on individual cheating and cheating tasks with a partner, each with a reward of 5, or 10 points for winning (1 point is worth 1 yen), four times in total. The order of the four tasks will be changed randomly for each participant.

Further, the participants will work on a cheating task with a human partner who also participated in the survey. The task will be the same as before; if both the participant and partner win, the participant receives a reward. However, owing to the design of the web survey, a partner will be chosen randomly after all participants complete the survey; therefore, their reward will be determined at the end of the survey. Initially, participants will perform two cheating tasks without knowing any information about their partner: one cheating task with a reward of 5-points and the other with a reward of 10-points.

Participants will then perform a cheating task in which they will be provided with information about their partners’ dishonesty. However, because partners will be determined after all participants have responded, it is impossible to inform them about partner dishonesty. Therefore, we avoid this problem by designing the experiment as follows. Depending on the results of the individual cheating task and the cheating task performed with a computer partner, there will be four types of partners. In the control group, the partner types will be divided according to the winners and losers of the cheating tasks, whereas in the treatment group, the partner types will be divided according to whether the commitment device is used. For example,

Control group
The individual task with 5-points rewards—winners
The task with a partner with 5-points rewards—losers

Treatment group
The individual task with 5-points rewards—answers before the coin side appears on the screen
The task with a partner with 5-points rewards—answers after the coin side appears on the screen.

In the control group, there will be four outcomes for individual tasks and tasks with partners: win/win, win/lose, lose/win, and lose/lose. In the treatment groups, there will be four types of cases, with or without the use of commitment devices for individual tasks and tasks with partners: before/before, before/after, after/before, and after/after. Participants will be asked to assume the above four types and work on a cheat task with a reward of 5-points and a cheat task with a reward of 10-points. Thus, they will work on the cheating task eight times. While rewards were generated in all tasks earlier, these eight tasks will generate rewards in only two of them. These two tasks are one of the four tasks with a reward of 5-points and one of the four tasks with a reward of 10-points. Rewarded tasks will be determined by the partner type, which is randomly determined after all participants complete their responses; in the control group, by the results of partners’ cheating tasks winning or losing; and in the treatment group, by whether partners use commitment devices.

After the cheating task, we will ask the participants about their experienced well-being. Further we will conduct a cognitive reflection test (CRT) and ask participants about their cheating tasks: their perceptions of cheating possibilities, anticipation of exposure to cheating behavior, views on the cheating tasks, anticipation of participants’ views on the cheating tasks, and experiences participating in similar experiments.

Follow-up survey/ experiment

We will conduct a follow-up survey approximately one week after the first survey to measure the feedback effect of the outcomes for cheating tasks with partners. We will ask the participants about their well-being at the beginning and at the end. Between these questions, the participants will perform cheating tasks with a human partner. The partners’ responses at this time will be based on the results of the first survey, which makes it possible to send sequential feedback to participants regarding the outcomes of the task. Partners will be randomly selected from among the participants, and the participants will be notified of their partner type and response results based on their partner’s first survey results.

Without knowing the partner type, the participants will perform the cheating task four times: two cheating tasks with a 5-point reward and two cheating tasks with a 10-point reward. Participants will also be notified of their partner’s type and will perform four cheating tasks: two cheating tasks with a 5-point reward and two cheating tasks with a 10-point reward. Thus, the total number of cheating tasks in the follow-up study will be eight. In each task, before the participants respond with their side of the coin, they can anticipate the honesty of their partner.
Experimental Design Details
Randomization Method
Randomization done by a computer.
Randomization Unit
Individual
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
1500
Sample size: planned number of observations
1500
Sample size (or number of clusters) by treatment arms
750
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Given the sample size, we will conduct a power analysis to compute the minimum detectable effect. Specifically, we will apply a two-sample t-test, which assumes equal and unknown standard deviations. At a significance level of α = 0.05 and a power of (1-β) = 0.8, the minimum detectable effect size is 0.1285 standard deviations. Implying that, with 750 participants per group, we will have an 80 percent power to identify an effect size as small as 0.1285 standard deviations between the groups. In the follow-up experiment, we expect the reinterview rate to be 67% and above. With a reduced sample size of 500 individuals per group, the smallest detectable effect size will be 0.1574 standard deviations. Since the minimum detectable effect of less than 0.2 standard deviations has been widely accepted in practice, the power analysis suggests that our sample size is sufficiently large to produce reliable treatment-effect estimates.
IRB

Institutional Review Boards (IRBs)

IRB Name
Soka University Institutional Review Board for Human Research
IRB Approval Date
2024-07-22
IRB Approval Number
2024047
Analysis Plan

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials