Experimental Design Details
#### GENERAL INTRODUCTION TO THE EXPERIMENTAL DESIGN ####
I use the paradigmatic spectator-worker design to (i) create an experimentally controlled situation of inequality between two workers and (ii) observe how spectators redistribute money between them. The focus of this study is on the redistribution decisions of the spectators in different experimentally created inequality situations between workers.
[WORKERS| I hire 100 US workers on Amazon's online labor market Mechanical Turk for an email collection task. Each worker k earns a piece-rate pi_k and can freely choose how many tasks e_k to complete. Afterward, workers are assigned to pairs. I frequently refer to the two workers in a pair as worker A and worker B.
[SPECTATORS] I invite participants from the general US population, also referred to as spectators, to an online experiment. Spectators can redistribute the worker's earnings.
The central feature of the design is a between-subject comparison of redistributive behavior in two types of inequality situations:
Situation type (a): The circumstances to which worker A and worker B react are identical. That is, they have the same piece-rate expectations. Ultimately, worker A receives piece-rate pi_A, and worker B receives pi_B.
Situation type (b): Worker A reacts to different circumstances than worker B. That is, they have different piece-rate expectations. However, eventually, workers receive the same piece-rates as in situation type (a).
Thus, the design systematically varies the expected circumstances to which workers react but keeps constant which piece-rate they ultimately earn. If workers react to the same circumstances, their effort choices are directly comparable. If workers, however, react to different circumstances, circumstances exert a differential impact on their choices. Contrasting redistributive behavior across these two situation types illustrate whether or not spectators take this into account.
Differences in redistributive behavior in these treatment comparisons jointly derive from two different mechanisms. First, they depend on the beliefs of spectators about the effect of circumstances on effort choices. If, for instance, spectators do not understand that expected circumstances affect choice, treatments 1 and 3 (or 2 and 4) appear identical to them, and no change in redistributive behavior is to be expected. If this is not the case, the second factor, fairness preferences, becomes critical: Is inequality due to choices that derive from randomly assigned circumstances considered fair?
To distinguish between these two mechanisms, two additional treatments exogenously manipulate the beliefs of participants. To do so, I include a new page to the instructions on which I inform spectators that effort choices in the task are strongly context-dependent; i.e., workers typically react to the piece-rate they expect to receive.
See the analysis plan for further details.