Experimental Design
We conduct an experiment to identify individuals’ redistributive preferences in situations where luck completely determines whether a person can work or not. The experiment includes two types of participants: agents and spectators. Our main focus is on the decisions of impartial spectators who choose to redistribute income between two agents receiving unequal initial earnings.
The agents all start by trying out an Encoding Task for one minute in Part 1 of the experiment. After that, in Part 2, the agents are paired and one in each pair is assigned a high earning of $1.20 (agent H) and the other is assigned low earnings $0.20 (agent L). The agents are given one of two assignments at random: work (do the encoding task again for another minute) or not work (wait for one minute). Each spectator is matched with a pair of agents, observes their roles and assignments and decides if, and how much, to redistribute from H to L.
Our main treatment (Treatment 1) will allow us to document redistribution decisions in situations where who can and who cannot work is assigned at random. In Treatment 1, spectators make redistribution decisions between pairs of agents one of who works and the other waits. The ‘worker’ receives the high initial earnings ($H) while the ‘non-worker’ receives the low initial earnings ($L). Agents are told, another participant who has not taken part in the survey will learn of their roles and work status and determine their final payment.
To benchmark how spectators in Treatment 1 tradeoff initial luck with subsequent work status, we compare the results of Treatment 1 with two benchmark treatments - Treatment 2 and Treatment 3. In Treatment 2 both agents wait while in Treatment 3, both agents work. Hence, in the benchmark treatments, initial earnings is still assigned at random (as in Treatment 1), but now both agents have the same work status. If spectators in Treatment 1 condition decisions solely on the fact that work status was exogenously assigned (without regard to the difference in work status between agents), then redistribution decisions across treatments 1, 2 and 3 would look similar. Alternately, if spectators base their decision on different work status (albeit assigned at random), we expect to observe lower redistribution in treatment 1 as compared to treatments 2 and 3.
Participants are recruited in the US and Sweden and the surveys are administered by the market research firm Respondi.
***** MAJOR UPDATE *****
As part of the revision process, two treatment variations (of our original experiment) will be conducted with participants recruited through Prolific. The treatments will vary from the original experiment on two dimensions - a) the stakes offered to agents to complete Part 2 of their survey, and b) the piece-rate payment offered to all participants (agents and spectators) for every letter encoded in Part 1 in the Encoding Task.
The new sessions will also re-run old Treatment 1 and Treatment 3 using Prolific participants as a robustness check and to benchmark the results of the new treatments. Below we list the 3 (2 new and 1 old) versions of the experiment that will be conducted -
1. Treatment 1 and Treatment 3 (Low Wage-Low Stake) : These two treatments are identical to Treatment 1 and Treatment 3 in our original experiment. It offers a piece-rate wage of $0.12 per letter encoded in Part 1. It additionally assigns Participant H with $1.40 and
Participant L with $0.20 as in the original experiment.
2. Treatment 1 and Treatment 3 (Low Wage-High Stake) : These two treatments differ from Treatment 1 and Treatment 3 in our original experiment on one account - it assigns Participant H with $7 and Participant L with $1. Hence, the spectators allocation options are ($7,$1) ; ($6.50, $1.50); ($6,$2) ; ($5.50, $2.50); ($5,$3) ; ($4.50, 3.50); ($4,$4) where the first term inside each bracket denotes the amount allocated to Participant H and the second term denotes the amount allocated to Participant L.
3. Treatment 1 and Treatment 3 (High Wage-High Stake) : These two treatments differ from Treatment 1 and Treatment 3 in our original experiment on two accounts.
First, it assigns Participant H with $7 and Participant L with $1. Hence, the spectators allocation options are ($7,$1) ; ($6.50, $1.50); ($6,$2) ; ($5.50, $2.50); ($5,$3) ; ($4.50, 3.50); ($4,$4) where the first term inside each bracket denotes the amount allocated to Participant H and the second term denotes the amount allocated to Participant L.
Second, the piece-rate wage for encoding in Part 1 is increases from $0.12 per letter to $0.60 per letter.