Fraud Deterrence Institutions Reduce Intrinsic Honesty
Last registered on September 03, 2019


Trial Information
General Information
Fraud Deterrence Institutions Reduce Intrinsic Honesty
Initial registration date
September 02, 2019
Last updated
September 03, 2019 4:54 PM EDT
Primary Investigator
CNRS - Université de Lyon - GATE
Other Primary Investigator(s)
PI Affiliation
Ca` Foscari University of Venice
PI Affiliation
CNRS - Univesité de Lyon - GATE
Additional Trial Information
Start date
End date
Secondary IDs
Deterrence institutions are widely used in modern societies to discourage rule violations but whether they have an impact beyond their immediate scope of application is usually ignored. Using a natural field experiment, we show that they affect intrinsic honesty across contexts. We identified fraudsters and non-fraudsters in public transport who were or not exposed to ticket inspections by the transport company. We then measured the intrinsic honesty of the same persons in a new unrelated context where they could misappropriate money. Instead of having an educative effect, the exposure to deterrence practices increases unethical behavior of fraudsters but also of non-fraudsters.
External Link(s)
Registration Citation
Galeotti, Fabio, Valeria Maggian and Marie Claire Villeval. 2019. "Fraud Deterrence Institutions Reduce Intrinsic Honesty." AEA RCT Registry. September 03.
Sponsors & Partners

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

Request Information
Experimental Details
We study the spillover effects of deterrence institutions on individuals’ intrinsic honesty across contexts. A research assistant, accompanied by a professional actor, observes whether passengers validate or not a ticket or a pass in public transport, a natural environment, identifying who is a fraudster and who is not. Then, after a targeted passenger exits the bus or tram, the professional actor plays the following scene. The actor seemingly picks a 5-euro banknote on the ground and asks the targeted passenger whether he or she lost the banknote. The rate of acceptance of the banknote, which gives a measure of intrinsic honesty, is correlated with the status of the passenger in public transport (fraudster vs. non-fraudster) and with the occurrence of a ticket inspection in public transport (inspected vs. non-inspected).
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
The key variables are whether: 1) the passenger validated or not a ticket or a pass; 2) a ticket inspection occurred during the journey; 3) the targeted passenger takes or not the banknote proposed by the professional actor on the street.
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
The field experiment is conducted in public transport in Lyon (France) by teams involving a research assistant and a professional actor. The experiment consists of two stages. The first stage takes place on board buses and trams and produces two main natural conditions. In the Inspection condition, a targeted passenger has been controlled by ticket inspectors from the transport company during his or her journey, whereas in the No-Inspection condition no ticket inspection occurred. The occurrence of a ticket inspection is only determined by the public transport company.

The second stage takes place when the targeted passenger gets off the vehicle, on the street. A professional actor, who is part of the experimental team, walks behind the targeted passenger and suddenly bends down to seemingly pick up a 5 euro banknote on the ground. The actor then calls the attention of the targeted passenger by asking whether he or she had lost a 5-euro banknote. We measure intrinsic honesty by recording whether the passenger takes or not the banknote. The actor has to use his or her mobile phone as an audio recording device when playing the scene on the street. We use these recordings to verify that the actors followed the protocol that we gave them to play the scene and as a robustness check to ensure that any minimal deviation from this protocol did not affect the internal validity of our results.

In our experiment there are three treatment conditions (Inspection, No Inspection, No Inspection-Audience) and two categories of passengers (fraudsters vs. non-fraudsters). The No Inspection-Audience condition is similar to the No Inspection condition, except that the research assistant walks by the actor when the latter plays the scene.

The research assistant has to identify the target passenger in the bus/tram (among the first few passengers entering the vehicle, the target passenger is the first of these passengers getting off the bus/tram). The research assistant has also to collect information for each observation: the name of the actor playing the scene, the time of the day, the weather, the treatment condition (Inspection, No Inspection, No Inspection-Audience), the name of the bus/tram line where the subject travelled, whether the subject validated a ticket, a pass or nothing, the approximate number of people on board the bus/tram, whether someone could notice the scene played in the street, whether the subject took or not the 5-euro banknote, the gender, estimated age, estimated economic status based on appearance, and ethnicity of the subject, and whether he or she wore religious symbols. In the Inspection condition, additional information is collected: the number of controllers, whether the inspection was conducted at the tram/bus stop or on board, whether the controllers wore uniform or civil clothes, the gender of the controller who inspected the subject, and, if the subject was a fraudster, whether he or she paid the fine immediately, and whether he or she had an emotional or aggressive reaction.
Experimental Design Details
Randomization Method
The research assistant has to identify the target passenger in the bus/tram, following this procedure. He observes the first few passengers entering the vehicle. He notes whether they validate or not a ticket. Then, the target passenger is the first of these passengers getting off the bus/tram. Thus, randomization is made in the field, depending on the passenger behavior and on the transport company inspections.
Randomization Unit
Randomization is at the individual level.
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
There is no clustering.
Sample size: planned number of observations
Based on an a priori power analysis, we decided to collect 100 observations per group (fraudsters and non-fraudsters) in the Inspection condition. Collecting data in the Inspection condition was much more complicated than in other conditions since we were dependent on the natural occurrence of ticket inspections. Therefore, while we were trying to reach the target of 100 observations for the Inspection condition, we continued to collect data in the No Inspection one, even if we had already collected 100 observations per group in this condition (in order not to waste the actors’ time, since they were paid per hour). This is why we collected, overall, more data in the No Inspection condition.
Sample size (or number of clusters) by treatment arms
In total, the field experiment involved 708 passengers: 358 non-fraudsters (104 in the Inspection condition, 140 in No Inspection condition and 114 in No Inspection-Audience condition) and 350 fraudsters (100 in the Inspection condition, 140 in No Inspection, and 110 in No Inspection-Audience condition).
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
To determine the sample size for both the No Inspection and Inspection conditions, we conducted an a priori power analysis. To form reasonable predictions about the behavior of fraudsters and non-fraudsters in the control group (i.e., No Inspection condition), we built on the results of Dai et al. (2018) who ran an artefactual field experiment in public transport in Lyon using a similar subject pool as ours. Dai et al. (2018) estimated the proportion of dishonest individuals among fare-dodgers and non-fare-dodgers in a die-under-the-cup task. The estimated proportion of fully (partially) dishonest subjects was between 0% and 19% (41% and 60%) for non-fraudsters, and between 9% and 46% (37% and 74%) for fraudsters. Assuming similar proportions of full and partial liars in our field experiment, and assuming that full (partial) liars accept the banknote all (half of) the times, we predicted between 46% and 64.5% (30% and 39.5%) of fare-dodgers (non-fare-dodgers) taking the banknote. Using the midpoints of these intervals and assuming a type-I error rate of α = 0.05 and a power level of 0.8, we computed a sample size of 92 subjects per group (fraudsters and non-fraudsters), which we rounded to 100 to be more conservative. Sample sizes are computed for two-sample proportions tests. With a sample size of 100 observations (for fraudsters and non-fraudsters, respectively), a significance level of 0.05 and a power of 0.8, the minimum detectable effect size of comparing the I and NI conditions is 0.19 for fraudsters and 0.20 for non-fraudsters. This corresponds to a Cohen’s h of approximately 0.4. Hence, a sample size of 100 was large enough to detect a small-medium treatment effect.
IRB Name
CEI - INSERM (IRB00003888)
IRB Approval Date
IRB Approval Number
Post Trial Information
Study Withdrawal
Is the intervention completed?
Is data collection complete?
Data Publication
Data Publication
Is public data available?
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers