Experimental Design
We run a RCT with student participants. The experiment investigates whether demand-side gender discrimination exists in the mutual fund investing context, and distinguishes statistical from taste-based discrimination by employing information provision treatment.
In Stage 1, participants make choices between lotteries to elicit individual risk preferences.
In Stage 2, participants receive shared training material with general information on the mutual fund industry. Then, the sample is divided into three groups to receive a gender information treatment:
The positive gender treatment group receives science-based information indicating that male and female fund managers perform similarly.
The control group receives neutral information unrelated to fund manager gender.
The negative gender treatment group receives information suggesting that male fund managers are top performers.
At the end of Stage 2, participants perform a short financial literacy test to measure their financial literacy level.
In Stage 3, all participants will be presented with two comparable index funds from the Morningstar database, both managed by mixed-gender teams. We employ a “gender vignette design” in which, for each fund, half of the participants see a male team member as manager shown in the management section, while the other half see the same fund with a female team member. We anonymize the fund names by referring to them as Fund A and Fund B, use AI-generated photos to indicate the genders of the fund managers (ensuring similarity on age/attractiveness/race), and replace managers’ real names with common male and female names. Additional fund details - such as size, fund strategy, expense ratio, and past performance - will be provided.
In this stage, we also introduce a “fund information treatment” to test whether the provision of detailed fund performance data can reduce potential statistical discrimination against female managers. Participants from the “positive gender treatment” group and the control group are divided into two subgroups based on how much fund performance information is presented. The fund information control subgroup receives fund performance information for 2024 only, whereas the fund information treatment subgroup receives additional historical fund performance information. Participants from the “negative gender treatment” group are assigned the control fund information, as their data are used primarily for robustness tests.
After reviewing the fund information, participants complete the following tasks:
Task 1: Participants are given an endowment and are asked to allocate it between the two funds.
Task 2: Participants are asked to choose one fund to recommend to a hypothetical investor.
Task 3 & 4: At different points during the investment period, participants are shown the actual realized fund return for the respective period, first negative and later positive, and are asked whether they would revise their recommendation to the hypothetical investor.
In Stage 4, participants are presented with two comparable active funds from the Morningstar database and are again subjected to the fund information treatment. They are asked to perform similar investment tasks to Stage 3. The same procedures and presentation standards as in Stage 3 are applied.
In Stage 5, participants are presented with two real charities, one supporting a women-associated cause and the other supporting a men-associated cause. They vote for one of the two charities, and the charity receiving the most votes will receive an actual monetary donation.