Experimental Design
Our experiment design is closely aligned with Cappelen, Falch, and Tungodden (2019), henceforth CFT, and builds upon a previous experiment which was preregistered in December 2020 (Haeckl, Möller, and Zednik 2020).
Setup
Our study consists of two experiments: the workers experiment and the spectators experiment. For logistical reasons the spectators experiment will be conducted before the workers experiment.
In the spectators experiment the participant (spectator) is randomly assigned a pair of workers who both completed a knowledge task about sports and games. Of the two workers, the one who answered more questions correctly has earned £6 and the other has earned £0. The spectator learns the gender of the two workers and their earnings from the competition. Then the spectator chooses how much R in [0, 6] of the winner's earnings to redistribute to the loser.
In the workers experiment participants (workers) answer a quiz with 20 multiple choice questions about sports and games. Subsequently each worker makes choices about how his/her payoff is determined and has the opportunity to send one of two messages to the spectator who will decide about the workers' payoffs. The choice set for how the payoff is determined available to the workers varies based on the treatment.
The final payoff of the workers who competed is determined by the decision of a spectator. We therefore first collect the decisions of spectators. Subsequently, we conduct the workers experiment and match each pair of workers with a randomly selected spectator decision from the respective treatment. Spectators are told in advance that a subset of all spectator decisions is randomly selected to be applied to the payoff of participants from the workers experiment. All workers are told that, depending on their decisions, their final income may be determined by the decision of an impartial spectator.
Treatments
Each spectator is randomly assigned a pair of workers who competed against one another. Half of the spectators will be assigned a pair in which the loser is a woman; the other half will be assigned a pair in which the loser is a man. We conduct the following three between-subject treatments:
In the Mandatory Competition (MC) treatment we replicate CFT. Spectators redistribute incomes of workers who do not have a choice but are always matched with another worker for a competition which determines the workers' payoffs.
In the Voluntary Competition (VC) treatment spectators redistribute incomes of workers who have chosen the competition over a piece-wise payment scheme.
In the Selfish Competition (SC) treatment spectators also redistribute incomes of workers who have chosen the competition over a piece-wise payment scheme. In addition, the losing worker has decided to buy a sabotage coin for 10 Pence. This sabotage coin gave the worker a 50% chance to win the competition irrespective of his or her performance. However, the worker has lost nevertheless.
On top of these three between-subject treatments, we also conduct a within-subject treatment. After the initial redistribution decision described above, spectators are randomly sorted into either the High-Dominance Message (HDM) treatment or the Low-Dominance Message (LDM) treatment. In the HDM treatment, spectators receive a message from the loser which signals high dominance (and thereby is agentic). In the LDM treatment, the message sent by the loser signals low dominance. The message is displayed to the spectator after he/she has made an initial redistribution decision. After having received the message, the spectator has the chance to change the redistribution decision previously made. This way, we collect two redistribution decisions from each spectator, one made before and one made after having received a message from the losing worker.
Finally, we implement a veil of risk (Exley 2016; Coffman, Exley, and Niederle 2016) across all treatments. This means that all spectators are informed that, for any amount of money they choose to transfer from the winning worker to the losing worker, there is a 99% chance that the loser receives the money and a 1% chance that the money is lost, i.e. neither the winner nor the loser receives it.