Experimental Design Details
This experiment will have three main stages which we describe below.
1. The questionnaire stage
One week before the start of the course, a survey is conducted among all teaching assistants. The survey consists of questions whose purpose is to record the preferences of the subjects. Specifically, self-reported levels of direct and indirect reciprocity, conditional altruism and reputation concerns/self-worth are recorded. The teaching assistants will be told that the survey is part of a research project on-going at the university, with the purpose of studying how people make decisions.
2. The field experiment stage
In the end of the second week of tutorials the students will be asked to evaluate and rate all the teaching assistants based on their teaching skills.
In the third week of the course, a representative of the employer will make a phone call to both treated and control. During the phone call she will ask the respondents about their experience with some of the services offered by the university to its teaching assistants. At the end of the phone call, those in the treatment group will receive an unexpected message of recognition for their performance.
After the final exam, students will be asked again to fill in the evaluation forms for the teaching assistants that were in charge of their tutorials. The grades on the exams for the course will be recorded as well for all other courses the students attend that block. Differences between intermediate and final evaluation forms and changes in the presence in the tutorial will be used as outcome measures in evaluating the average treatment effect of the intervention. Heterogeneous treatment effects will be explored based on the different preferences stated in the surveys by teaching assistants.
3. The lab experiment stage
After the treatment is administered for the teachers of interest, during a festive gathering organized by the university for all the teaching assistants, all those who taught a course in the past academic year will be invited to also play a number of incentivised games with the purpose of eliciting the same preferences that were self-reported in the survey. We expect that only a sub-sample of those invited will actually show up. For ease of future replications, the games will be based on standard measures in the literature designed to elicit the preferences of interest: direct reciprocity, indirect reciprocity, conditional altruism, reputation concerns.
The experiment will take roughly 30 minutes and the participants will undertake it using their mobile phones, following a link that will be sent to their e-mail addresses by the experimenters. The design of the experiment is closely related to the methodology implemented by Stanca (2009) in his paper "Measuring indirect reciprocity: Whose back do we scratch". The paper defines 3 types of reciprocity: direct (DR), indirect generalized reciprocity (IGR) and indirect social reciprocity (IRS). IGR is is defined as adopting a helpful action towards someone else, at one's own material cost, because some other person's intentional behavior was perceived to be helpful to oneself (if A helps B, then B helps C) which corresponds to our definition of indirect reciprocity (henceforth IR). IRS is defined as adopting a helpful action towards someone else, at one's own material cost, because that person's intentional behavior was perceived to be helpful to some other person (if A helps B, then C helps A) which corresponds to our definition of conditional altruism (henceforth CA).
In the beginning of the experiment, besides a show-up fee the participants will be given an endowment of n tokens. The experiment will take place in three rounds, each design to test one of the mechanisms of interest. In each round participants will be given either the part "A" or the part "B", each with a probability of a half. Participants will be instructed that in the beginning of each round they will randomly be matched in pairs (A,B). The matching will only last for one round (a one-shot interaction) and their identity will remain anonymous in order to prevent strategic choices.
All subjects A_i will be told that they can choose an amount a_i, an integer between 0 and n, that they can send to the player B_i with whom they are matched, in the beginning of each round. The amount will be subtracted from A_i's payoff, multiplied by 3 and added to the payoff of B_i.
Round 1 (DR): Before being informed about the amount a_i, player B_i will be asked to provide the amount b_i they want to send back to player A_i for every possible decision in the decision set of A_i, namely alpha=(0, 1, 2,.....n). In the end, B_i is informed about the amount a_i and the corresponding amount b_i is multiplied by 3 and sent back to player A_i.
Round 2 (IR): Before being informed about the amount a_i sent by player A_i, player B_i will be asked to provide the amount b_i they want to send to another player A_j for every possible decision in the decision set of A_i. B_i is informed about the amount a_i and the corresponding amount b_i is multiplied by 3 and added to the payoff of player A_j.
Round 3 (CA): Player B_i will be asked to provide the amount b_i they want send to another player A_j for every possible decision in the decision set of A_j regarding the number of tokens they chose to send to another player B_j. B_i is informed about the amounts a_i and a_j and the corresponding amount b_i is multiplied by 3 and added to the payoff of player A_j.
To isolate the "reputation concerns" mechanism using monetary incentives we decided to offer all respondents the opportunity of spending some of their show-up fee in order to buy a certificate stating that they were teaching assistants during the academic year. The respondents will have three options: (i) not spending anything, (ii) spending a very small amount of money to receive a print-out of the certificate or (iii) spending some extra money in order to receive a framed copy of the certificate. We believe that for labor market purposes, neither (ii) nor (iii) are really needed since the teachers can just add this information to their CVs without paying for it. As such, choosing (ii) or (iii) as an option captures some of the self-worth and/or reputation concerns that respondents might have (since the -framed- certificate also offers the possibility of displaying one's achievement), with (iii) signaling a higher reputation concern than (ii).
At the end of the academic year, respondents will be asked to answer a couple of questions designed to understand the relationship between teaching assistants and the frequency with which they communicate and share information with each other. Furthermore, an open question will be asked to see if teaching assistants were "unhappy" about any other issues regarding the course. The results to this question will be used as a robustness check to ensure that spillovers between treatment and control do not occur, such that individuals in the control group found out that they would qualify for recognition, but they nonetheless did not receive it.
Also, two questions regarding how they perceive the importance of their job will be asked. This can help us study yet an additional theoretical channel: if some teaching assistants are not aware about the fact that their work is very important to students, they might put little effort in. However, since this is revealed through the treatment, effort can simply increase through an update in the beliefs about the importance of their teaching to students. As such, teaching assistants will be asked "how important they think their job is", and whether "the Tutor Academy has contributed during the course to the perceived importance of their job".