NEW UPDATE: Completed trials may now upload and register supplementary documents (e.g. null results reports, populated pre-analysis plans, or post-trial results reports) in the Post Trial section under Reports, Papers, & Other Materials.
Endogenous gender segregation
Initial registration date
June 20, 2020
June 22, 2020 11:47 AM EDT
This section is unavailable to the public. Use the button below
to request access to this information.
University of Oslo
Other Primary Investigator(s)
Additional Trial Information
If individuals perceive preferences as gender-dependent, they might use gender composition in different occupations as a signal of underlying characteristics. The aim of this project is to explore this potential signalling effect of gender composition. To this end, I perform an online experiment where subjects choose between jobs solely based on summary statistics about the individuals in each career path. In this plan I describe hypotheses to be tested, the coding of variables, and the empirical strategy that will be used.
Intervention 1: Subjects in the experiment choose between two jobs. They are not informed about the characteristics of each job, however, they receive summary statistics of regarding a group of people previously choosing between these two jobs, Group Pre. The characteristic of interest is gender. In one treatment, subjects choose between jobs, A and B, where equally many men (women) chose job A as job B. Hence, the gender distribution of previous choices provides no information regarding the gender dependency of jobs. In the other treatment group, subjects choose from a different pair of jobs, in which the majority of females in Group Pre chose job A and the majority of males chose job B. The intervention is meant to provide a proof-of-concept of the effect of gender composition in different jobs on the subsequent decisions of men and women.
Intervention 2: In the second intervention, one treatment group only gets information about the distribution of their own gender in the Group Pre over two new jobs, A and B. Subjects in the other treatment group receive information about both genders. The treatment tests the informativeness of observing both distributions, relative to only observing one.
Intervention 3: The third intervention measures the causal effect of gender composition on peoples beliefs about how gender-dependent a task is. Subjects in both treatment groups are shown a the same job. In one treatment group, subjects are also shown the fraction of men and women in the Group Pre choosing this job. Both groups are then asked to guess whether men or women did better in the Group Pre.
Intervention Start Date
Intervention End Date
Primary Outcomes (end points)
- domjob: Whether a male (female) subject chose the male (female) dominated job (1) or not (0). - risk: Risk aversion of a subject
- domjob2: Whether a male (female) subject chose the male (female) dominated job (1) or not (0). Intervention 3: - menbetter: Whether a subject believed men did better (1) in the Group Pre or not (0)
Primary Outcomes (explanation)
domjob and domjob 2: Suppose the majority of males in Group Pre chose job A and majority of females chose job B. Then a female subject in intervention 1/intervention 2 will be assigned the value 1 on domjob/domjob2 if she chooses job B and 0 otherwise. A male subject will be assigned value 1 if he chooses A and 0 otherwise.
Secondary Outcomes (end points)
sec_ord_beliefs - A male (female) subjects' belief regarding the share of other males (females) choosing the male (female) dominated job.
ownperform - Whether a subject believes he/she would do better than average on the job in intervention 3 (1) or not (0).
womenbetter- Whether a subject believes women did better on the job in intervention 3 (1) or not (0).
Secondary Outcomes (explanation)
The experiment is conducted online and consists of two parts. Part 1 of the experiment was conducted prior to uploading this pre-analysis plan and is solely used to generate different gender compositions of men and women in multiple pairs of jobs. A job consists of a task, such as a word search task or a set of questions regarding a specific topic. The characteristics of the job --- for instance, topic of questions and payment structure --- are observable in Part 1. Subjects in Part 2 choose between some of the same jobs, but the characteristics of the jobs are unobservable. Instead, they observe summary statistics, such as gender, regarding the individuals choosing the different jobs in Part 1. I implement a set of between subjects treatment in Part 2 to examine how people use gender compositions in jobs from Part 1 as decision variables when the characteristics of jobs are unobservable.
Experimental Design Details
Subjects are allocated randomly to treatments by Nettskjema, the survey software.
400 men and 400 women will be recruited. Within gender randomization is done at the individual level.
Was the treatment clustered?
Sample size: planned number of clusters
Sample size: planned number of observations
800 subjects. 400 men and 400 women.
Sample size (or number of clusters) by treatment arms
0.5 probability of being allocated to one of two treatments. Hence, in expectation, 400 in each treatment.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Given a power of 0.8, significance level of 0.05 and a sample size of 800, the minimum detectable effect size on domjob, domjob 2 and menbetter is a Cohen's d of 0.18. The minimum detectable effect size on risk is 0.29.
INSTITUTIONAL REVIEW BOARDS (IRBs)