The experiment involves three phases: a production phase, an allocation phase and a distribution phase. During the production phase, participants perform an effort task, involving the copying of shapes. They are given blocks of squares containing both white and black squares. Next to each block, there is a copy containing white squares and grey squares in the same positions as the ‘original’ block. The goal is to put crosses in the grey squares, in order to copy the shape of the original block. Participants are given a few minutes to go through as many blocks of squares as possible.
In the allocation phase, we group participants into pairs – Player A and Player B – who are playing in different rooms. Player A is endowed with three units (low endowment) and Player B is endowed with seven units (high endowment). We have two treatments. In the luck treatment (our control group), the initial endowments of players is determined randomly based on their participant ID number (which is randomly allocated to participants in the beginning of the session). In the merit treatment, the initial endowments of players are based on their performance in the effort task. Together, each pair has a total production of ten units.
The distribution phase consists of two similar stages, in each of which participants make distributive choices about these ten units between themselves and their pair. In the first stage, both participants make their choices independently and these are not communicated to the other player until the end of the experiment, when the final amounts are paid to all participants.
In the second stage, we maintain the same initial allocation, but induce a change in power between the participants. Faced with an unequal initial distribution of incomes, the player with a low endowment has grabbing power, but the player with the high endowment may veto the final transfers. Thus, the decisions are sequential and Player B has the power over the final distribution. While Player A is the first to propose their distribution decision, this choice is then shown to Player B, who has the highest endowment, and the power to decide whether they agree or disagree with the proposal. The final distribution corresponds to the decision of Player B and we match the payment of both participants according to this decision. In addition to collecting their choices, we ask each participant to anticipate the decision of the other player: Player A will guess the decision of Player B after proposing the distribution, and Player B will guess the proposal of Player A before seeing their decision.
The design of the study is based on between-subject comparison. In each country, we initially run a total of 40 sessions, 20 in rural settings and 20 in urban settings (see more details on the sampling in section 3) and we will add an additional round of 40 sessions, 20 in rural settings and 20 in urban settings, in the South of Vietnam. Within each location (which can be rural or urban), we run an even number of sessions and each treatment is randomly allocated to half of these sessions. The assignment of participants between sessions in each location is random.