Experimental Design Details
This trial consists of 5 artefactual field experiments run on the online platform Amazon Mechanical Turk (mTurk). The aim of the trial is to manipulate the salience of gender identity cues in subjects so as to test the predictions of identity theory on choice under risk.
The subject pool is restricted to mTurk Workers with a US tax identification number and who have been reviewed positively Requesters in at least 95% of the tasks they completed on mTurk. At sign up, subjects answer a few demographic questions, which include the country of residence to verify the US restriction is works.
Experiment 1 and Experiment 3 have a within-subject design. The experimental procedure consists of three steps. In the first step, subjects' risk tolerance is elicited using lottery choices a la Holt and Laury (2002). In the second step, subjects face the experimental manipulations. The design is a 2 (male, female) X 3 (control state, priming state, threatening state) experimental design. After facing the manipulations, which consist in reading online blog excerpts that include gender identity role cues, subjects recall and describe in detail in a short essay (5-10 sentences) a situation in which they behaved in line with the blog description. The content of the short essays is used to (i) verify subjects are not automated bots, and (ii) provide a manipulation check for the experimental procedure. In the third step, risk tolerance is elicited again at the subject level.
Experiment 2, Experiment 4, and Experiment 5 have a between-subject design. The experimental procedure consists of two steps -- exposure to the experimental manipulations and performance of a set of choices. In Experiment 2, the design is a 2X3 factorial design as discussed above. The choices aim to elicit subjects' own beliefs about winning if they take part in a lottery whose objective probabilities of winning are common knowledge. Beliefs are elicited by asking subjects to report how many times they think they would win if participating 10 times in a lottery that on average wins 50% of the time. Subjects should then report how many times they think their peers would win if participating 10 times in the same lottery. The difference between the beliefs of self-success and peer-success constitute a measure of subjects' "better-than-average" beliefs.
In Experiment 4, the design is a 2X3 factorial design, but the two treatment states are both priming states. Choices here are framed as investment games a la Gneezy and Potters (2009) with varying levels of risk and expected value. In particular, subjects are provided with 100 experimental dollars (whose exchange rate with actual dollars is known by subjects throughout the experiment) in each of three period. Each period, subjects decide if they want to allocate any money to the opportunity, and if yes how much. Subjects can only allocate values between 0 (cannot pay to avoid making a choice) or 100 (cannot invest more than their endowment in each period). Gains and losses across periods cannot be reinvested in other opportunities. The first opportunity has a probability of success of 1/2 and gains 3 times the invested amount in the case of success, zero otherwise. The second opportunity has a probability of success of 1/3 and gains 2.5 times the invested amount in case of success, zero otherwise. The third opportunity has a probability of success of 1/6 and gains 8 times the invested amount in case of success, zero otherwise.
In Experiment 5, the design is a 2 (male, female) X 2 (control, priming) factorial design. Choices are also investment games a la Gneezy and Potters (2009) like the ones described above but framed as delegated investment opportunities. Despite the delegation frame, the compensation of subjects is a pre-specified fraction of the principal's outcome and hence the interests of the agent and the principal are perfectly aligned.
At the end of each experiment, subjects face a debriefing procedure, obtain their payment, which include a fixed and a random component based on luck that guarantees choices throughout the experiments are incentive compatible. No subjects are recruited to take part in more than one experiment.