NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Anticarrots and Athletic Gym Membership
Last registered on May 28, 2015


Trial Information
General Information
Anticarrots and Athletic Gym Membership
Initial registration date
May 28, 2015
Last updated
May 28, 2015 2:22 PM EDT
Primary Investigator
Yale Law School
Other Primary Investigator(s)
PI Affiliation
University of Amsterdam
Additional Trial Information
In development
Start date
End date
Secondary IDs
This experiment will test whether the experience of resisting temptation to quit an activity or task boosts subjects’ determination to complete that task.

We pursue this question through a collaboration with a gym to randomly offer some new members the chance to give up their membership and have their membership fee refunded. Our hypothesis is that those who are presented with and resist the temptation will come to the gym more often and renew their membership at higher rates. In particular, we want to test whether the group of people who are offered a temptation to quit the gym ends up exercising more – even after including (in our intent to treat analysis) those students who accept the temptation and quit their gym membership.

By rejecting a temptation a person is able to learn something about herself, which in turn will increase her commitment to a certain activity or task. Bénabou and Tirole (2004) map a theoretical framework for “willpower activities” like resisting temptation. They hypothesize that resisting temptation is “hard” information that even people with imperfect recall about their state of mind—for example, being committed to going to the gym—can use to build an enduring image of themselves. Resisting the temptation provides the actor information about his ability to commit to a course of action and hence about the likelihood of success. This new information will, in turn, make it easier and more attractive to stay on the task rather than quitting. Our experiment will test this theory.
External Link(s)
Registration Citation
Ayres, Ian and Giuseppe Mattiacci. 2015. "Anticarrots and Athletic Gym Membership." AEA RCT Registry. May 28. https://doi.org/10.1257/rct.650-2.0.
Former Citation
Ayres, Ian, Ian Ayres and Giuseppe Mattiacci. 2015. "Anticarrots and Athletic Gym Membership." AEA RCT Registry. May 28. http://www.socialscienceregistry.org/trials/650/history/4318.
Experimental Details
On their second visit to the gym within a week of registering, one-third of these new members will be randomly presented at a computer terminal with one of three treatments. Members assigned to each treatment group will see a different initial screen; all members will then be shown a screen asking the member to rate his or her satisfaction with them gym.
Intervention Start Date
Intervention End Date
Primary Outcomes
Primary Outcomes (end points)
Primary: Decision to quit, satisfaction with gym, decision to extend gym subscription after initial gym contract ends, frequency of attending the gym (at week, month, semester, year frequencies), and various measures of subsequent academic achievement and satisfaction. Secondary outcomes will include tests of heterogeneous treatment effects with regard to several pre-treatment variables, including: language, gender, marital status, postcode of residence, work-study status, academic degree, academic major, academic year, academic grades, number of credits per semester, level of difficulty of academic classes, number of classes.
Primary Outcomes (explanation)
Secondary Outcomes
Secondary Outcomes (end points)
Secondary Outcomes (explanation)
Experimental Design
Experimental Design
The experiment focuses on new members at the gym, who have already demonstrated additional interest in exercising by going to the gym twice in the first week of their memberships. We limit our subjects to this population because these new members have a commitment level that is high enough that they are likely to resist our temptations but still malleable and subject to updating based on new information about themselves.
Experimental Design Details
The experiment flow is as follows: 1. A new member signs up at gym and signs a contract that includes a section or clause about academic experiments. This is their consent to participate in the study. 2. On their second visit to the gym within a week of registering, one-third of these new members will be randomly presented at a computer terminal with one of three treatments. Members assigned to each treatment group will see a different initial screen; all members will then be shown a screen asking the member to rate his or her satisfaction with them gym. The treatments are shown in either English or Dutch depending on the member’s initial language selection. The treatments have been described above in the "intervention" section. 3. The experiment design limits the number of “quits” by asking the members at the beginning of their membership while they are still excited about their new commitment. It also does so by asking them as they enter the gym prepared to work out, which may dissuade them from quitting, workout bag in hand. In the data, those who quit will be marked as having 0 visits and as not renewing their contract. 4. The system should block new subscriptions within three months from those who have accepted the offer. There should be a procedure to check that other gyms do not try to bypass this ban. 5. The system will allow the researcher to stop the experiment in three ways: (i) if we decide to do so; (ii) automatically if the number of cancellations for a specific contract type exceeds a certain threshold (the system will allow us to set different thresholds for different contract types); (iii) automatically if the total refunds for cancelled subscriptions exceed a certain threshold (the system will alert the gym and the researchers if total refunds have reached 80% of the limit). We will match gym outcome data with academic survey outcomes from the following national survey in Holland: http://www.studiekeuzeinformatie.nl/wp-content/uploads/2014/11/Vastgestelde-vragenlijst-NSE-2015-Eng.pdf The survey includes sections asking students to rate their satisfaction levels (on a scale of 1 to 5, 1 = very dissatisfied, 5 = very satisfied, and 6 = Not applicable) with academic programs, for example: 1. “The academic skills acquired in your study program/The applied research skills acquired 2. Testing and assessment (e.g. assessment criteria and forms of testing) 3. Program schedules 4. Study load 5. Other facilities and study environment 6. The general atmosphere in your study program 7. The degree to which you are involved in improving your study program” Students are asked a wide range of other satisfaction-related questions; they include, using the same scale as above: 1. The level of your study program 2. The degree to which the content of your study program meets your expectations 3. The feasibility of deadlines 4. Student disabilities, illnesses or conditions The gym data and the university data can be matched using the students' names. Data will be made anonymous after matching is done.
Randomization Method
Randomization is performed by the gym's computer system.
Randomization Unit
The assignment is random and immutable, stratified by contract type with treatment assignments for each contract type made in random permuted blocks of size X (where X is the number of active treatment groups). In practice, here is an example with 3 groups: for each contract type, the system randomly assigns every new incoming customer to group A, B or C. The assignment is made in random permuted blocks of 3: this means that when the first new customer for a particular contract type comes in, the system generates a random permutation of the letters {A, B, C}. For instance, if the generated permutation is {B, A, C}, then the customer is assigned to group B. The next customer is assigned to group A and the next one to group C. If a new customer arrives (the fourth customer), the system generates a new permutation. Say, this time it is {B, C, A}, and the system proceeds as before. The system generates a new permutation for every block of three new customers. This procedure is followed for each contract type independently. We should be able to add / delete contract types easily before the experiment starts and to selectively stop the experiment for some contract types and not for others. The assignment is used only for this experiment and is not communicated / visible to the personnel or the customer.
Was the treatment clustered?
Experiment Characteristics
Sample size: planned number of clusters
8 contract types
Sample size: planned number of observations
1500 new gym members
Sample size (or number of clusters) by treatment arms
We expect to have 1500 participants in the experiment over a twelve month period, and for each treatment group to have approximately 500 participants.

Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB Name
Yale University Human Subjects Committee
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, Papers & Other Materials
Relevant Paper(s)