Experimental Design
In the baseline three-games design of Cox (2004), Treatment A plays the classical trust game where Players X and Y start with the same endowment. Player X sends an amount of money to Player Y from her endowment, the amount gets tripled, and Player Y decides how much to send back to Player X from the tripled amount. In Treatment B, only Player X has a choice: she sends an amount to Player Y from her endowment, which gets tripled, and Player Y keeps this amount. In Treatment C, Player X has no choice: an amount is taken from their endowment and sent to Player Y. The amount then gets tripled, and Player Y decides how much to send back to Player X from the tripled amount.
We extend this design first to upstream reciprocity. In baseline Treatment A, three players start with the same endowment. Player X decides how much to send from her endowment to Player Y, where the sent amount gets tripled. Player Y then decides how much to send to Player Z from her total endowment, where the sent amount gets tripled again. In Treatment B, only Player A has a choice: she sends an amount to Player Y from her endowment, which gets tripled, and Player Y keeps this amount. Player Z also keeps her initial endowment. In Treatment C, only Player Y has a choice. First, an amount of money is taken from Player X's endowment and sent to Player Y. The amount then gets tripled, and Player Y decides how much to send to Player Z, where the sent amount gets tripled again.
The comparison of Treatments A and B shows whether, in the baseline treatment, the choices of Players X have a strategic component, i.e., whether they expect Players Y to send some of their endowment to Players Z. The comparison of Treatments A and C shows whether Players Y behave differently when their total endowment was a result of a kind action from Players X from when Players X did not have a choice.