Experimental Design
The study intends to follow a 2 (Empathy treatment) by 2 (observability) randomized experimental design within a donation game. The design follows the procedure below:
i. Empathy-altruism hypothesis [validation] • Randomized treatment to empathy.
– Study experimentally induces empathy using the commonly used perspective- taking task (the act of perceiving a situation or understanding a concept from an alternative point of view such as that of another individual.) The idea is to have player A in the donation game think about the feeling/emotion and perspective of player B ex-ante then respond to the donation game questions. The treatment group will be shown a case of a suffering Kenyan listed on the Mchanga platform and who is in need of assistance. A full profile of the individual shall also be provided in text 1.
– The control group observes a collage of pictures showing weather conditions. Then they are asked to choose one that best suits the general weather condition the previous day [yesterday].
ii. Success of empathy treatment [Validation]
• Subjects are presented with three lexicons (words) describing empathetic feelings. To
ensure continuous elicitation of feelings, we assume a slider version of the scale running from 0 to 100. Participants slide the cursor to a point in which they feel empathetic, concerned, and sad. There shall be an example (with unrelated questions) on how to use the slider before the actual survey questions2.
iii. Reputation [Full cross randomization to observability treatment.]
• Before the donation game subjects are randomized in a sequential dictator game round.
Half are randomized to start their round of play. However, before playing, they are made
aware that their decisions will be made public to the other half of the subjects prior to their round of play. The randomization on the round of play each subject is assigned is communicated accordingly on the computer screen. The second-round players are not observed.
iv. Reciprocity
• The context is that within minimal groups, individuals give to their group members and
expect reciprocity not just from the person they have donated to but from the community members who have observed their donating behavior. For this, a typical trust game may be inefficient. We implore proxying reciprocity with the expectations donors have when giving. We, therefore, use an expected reciprocity measure in a hybrid dictator game. Before subjects receive a donation from a dictator, they are asked what they expect to be the donation. This is the measure we use to estimate/proxy reciprocity.
Treatment arms, randomization and Statistical specification i. Stage 1: Set up and empathy inducement
• The sample is randomized to empathy treatment using the perspective-taking task. The treatment group is also primed by a vignette about a member of a community experiencing a medical condition that requires a substantial amount of money in hospital bills. Further, a brief profile of their local dependency ratio (children, spouse and general household needs) is provided. After reading through the textual vignette, the treatment group writes three words about the member’s feeling and how their decision to share some resources/money would affect her. The control group chose between a collage of pictures depicting different weather conditions and then write three statements about yesterday's weather in the morning, afternoon and in the evening.
ii. Stage 2: Success of the empathy manipulation.
• Subjects are presented with three lexicons (words) describing empathetic feelings. To ensure continuous elicitation of feelings, we assume a slider version of the scale running from 0 to 100. Participants slide the cursor to a point in which they feel empathetic, concerned and sad. There shall be an example (with unrelated question) on how to use the slider before the actual survey questions:
EmpAvscore = α1 + β1Treatmentemp + ε
iii. Observability
• A full cross randomization is implemented. Half of the subjects are treated to full observability. That is, they are told that their donation behavior will be made public to their paired subject in the room before their round of play. The second round players’ behavior is not observed.
iv. Effects of empathy
• The donation game proceeds as usual. All subjects are told that they will participate in two games. In the first game (dictator game), they will receive 50 shillings from the experimenter. They will be asked to donate some, all or none of the resources to Player B. They are also made aware that there shall be another round of play for which they will be receivers. The study estimates;
Donationsavg =α2+β1Treatemp+β2Treatemp∗Treatinf +ε
v. Stage 4: Reputation
• Following the information randomization, half of the players are concerned about observability of their behavior. Hence to capture this hypothesis, the study shall seek to estimate:
Donationsavg =α3+β1Treatinf +β2Treatemp∗Treatinf +ε
vi. Stage 4: Exptected Reciprocity.
To estimate, even momentarily, whether individuals could be seen to care about reciprocity in charitable giving, we assume a novel approach to the experimental variant of a trust game.
Formally, reciprocity as measured within a trust game is such that player A interacts with player B in gift exchange. Having been given equal endowments, Player A allocates positive resources to player B for which the experimenter multiplies the donation by a factor (usually 3) before player B receives it. Player B then makes a decision about how much he wants to send back to Player A. The amount sent back is potentially the measure of reciprocity.
However, while giving within welfare groups, individuals who are concerned about reciprocity know that reciprocity is not tied to the exact individual they are now donating to. That is, donating to player B is not limited to his position to reciprocate but extends to other group members. It is also often not tied to the cause3 of the “now” donation but has effects on other forms of occurrences that may require donation.
This confound potentially renders a typical trust game weak in identifying the preferences. We then propose to use a retrospective self-reported measure. That is we inform the subjects that whatever they donated to player B got tripled by the experimenter. Also, they are told that Player C could reciprocate their donations of upto three times what they donated. They are then asked to indicate how much they would expect player C to send back.