Gender Discrimination in Financial Advising: Evidence from the Chinese Market

This audit study aims to investigate how financial advisors at securities firms offer gender-differentiated financial advice and commission discounts, and to explore the underlying mechanisms behind these behaviors.

To establish causal relationships, we use a RCT design. Auditors are randomly assigned to distinct “investor roles” based on a combination of individual characteristics such as gender, local status, risk preference, and other observable traits. The roles are assigned randomly to ensure that the only difference between auditors is their gender. This randomization allows us to infer that any differences in financial recommendations or commission discounts received by auditors are due to gender-based discrimination by the financial advisors.

Auditors are then sent to various branches of securities firms, where they are served by financial advisors who are also randomly assigned by the firms. This randomization of both auditors and advisors enables us to isolate the effect of gender on the financial advice and commission offers provided.

The audit study is conducted in two rounds. Round 1 takes place during a typical bearish market, while Round 2 is conducted during a 'bullish' market influenced by an unexpected positive policy shock. This two-round approach allows us to assess how financial advice and commission offers may change in response to shifts in market conditions.

Additionally, we survey financial advisors to better understand the motivations and factors influencing their behavior. These surveys provide valuable insights into the mechanisms driving gender-differentiated financial advice.

This audit study investigates whether financial advisors at securities firms in China offer gender-differentiated financial advice and commission discounts and explores the mechanisms underlying such behavior. To achieve this, we first construct 32 unique investor profiles based on five characteristics: gender (female vs. male), local status (local vs. non-local), risk preference (high vs. low), trading frequency (high vs. low), planned investment amount (high vs. low). We then assign confidence in making financial decisions (high vs. low)} conditionally based on gender: half of the auditors within each gender group are assigned high confidence, and the other half are assigned low confidence. This randomization with conditional assignment approach ensures that the full range of investor profiles is captured without unnecessary redundancy, while also maintaining a balance between randomization and operational feasibility.

This audit study establishes causal effects by using randomization and controlled conditions to isolate the impact of gender on outcomes like financial advice and discounts. Female auditors serve as the treatment group, while male auditors are the control group. The random assignment ensures that any observed differences in financial advice or commission offers are due to gender and not other confounding factors. To further control for external influences, interactions are standardized (e.g., identical scripts, timing, and individual profiles), ensuring that the treatment variable is the only difference between groups. In other words, if male and female auditors with identical profiles receive different commission discounts, the observed difference can be attributed to gender discrimination.

Once assigned, auditors are sent to different branches of securities firms, where they are served by financial advisors who are randomly assigned by the firm. This dual randomization of both auditors and financial advisors ensures that the observed differences can be attributed solely to the treatment variable—gender—allowing us to infer the causal impact of gender on financial advice and commission discounts.

The study is conducted in four rounds. In Rounds 1 and 2, the auditors present themselves as long-term investors, while in Rounds 3 and 4, they present themselves as short-term investors. Rounds 1 and 3 are conducted during a typical bearish market, and Rounds 2 and 4 during a ‘bullish’ market influenced by a surprising positive policy shock. In all rounds, the auditors follow strict protocols for their initial consultations with financial advisors and conduct follow-up meetings within two days to assess any changes in advice or commission offers.

Additionally, a survey of financial advisors is conducted in both rounds to explore the motivations behind their behavior, providing valuable insights into the observed patterns and mechanisms of gender-differentiated financial advice.

2024-04-15

2025-04-23

The primary outcomes examine whether financial advisors provide differentiated investment suggestions based on the gender or other characteristics of potential investors. Since the commission charged varies with the investment bundles recommended, we also investigate whether the discounts offered by financial advisors differ based on the investor's characteristics.

Investment bundles refer to financial recommendations provided by financial advisors, including suggested allocations to investments across stocks, fixed-income products, ETFs, and other assets.
A commission discount refers to the percentage reduction in the commission offered by the financial advisor compared to the default commission rate.

The second outcome examines the mechanism behind the observed discriminatory behavior of financial advisors towards potential investors.

The second outcome examines the mechanism behind the observed gender-based discriminatory behavior of financial advisors towards potential investors. We will explore the second outcome by surveying financial advisors in both the target city and major financial hub (Shanghai, Beijing).

Our study investigates the investment advisory practices of securities firms in a southern city in China. We identified the full universe of securities firms and their branches using publicly available data platforms and confirmed their eligibility for the study through direct contact. Auditors, resembling typical retail investors, visited branches to interact with financial advisors. The study comprised four rounds, with the first round spanning four months, the second conducted over two weeks, and the third and fourth rounds each lasting less than three weeks. Auditors were trained with standardized guidelines and conversation templates to ensure consistency in their interactions. During the visits, auditors collected detailed information about investment recommendations, commission practices, and other related advice. Compensation for auditors ensured their commitment to the study.

This study investigates the investment advisory practices of securities firms in a southern city in China. Using detailed corporate data from a leading business registry platform, we identified 42 branches (45 advisory teams) from 26 securities firms as the full sample universe. In Round 1, auditors visited all branches in the universe, while in Round 2-4, operational changes reduced the sample to 41 branches across 25 securities firms. Before the study, research assistants confirmed the eligibility of branches by contacting them directly to verify that they provided advisory services to retail investors.

Auditors were carefully selected to represent typical retail investors based on specific demographic and financial criteria, such as moderate stock trading experience, moderate level of financial literacy, and local or non-local residency. Each auditor was assigned a randomized “investor role” characterized by varying attributes such as risk preference and planned investment amount. These roles were designed to ensure diversity while maintaining authenticity, minimizing the risk of auditors being perceived as “acting.”

The study was conducted in four rounds. Round 1 lasted four months (from April to August 2024), with auditors visiting 11–12 branches per month, while Round 2 was compressed into two weeks (in October 2024). Rounds 3 and 4 lasted approximately two and three weeks, respectively (from March 17th to April 3rd, 2025, and from April 8th to April 23rd, 2025). Each auditor visited about 20 branches per week.

To avoid raising suspicion, auditors mentioned being referred by a friend during their interactions with financial advisors. They documented details of the advice provided, including investment recommendations, commission charges, and any potential discounts. A distinct set of 32 auditors was deployed to conduct the visits across all rounds.

Auditors underwent a comprehensive training program to ensure standardized interactions. The training included reading detailed manuals, participating in simulated practice sessions with industry professionals, and verifying their understanding through quizzes. During visits, auditors engaged in initial consultations with financial advisors and conducted follow-up meetings to assess whether advice or commission structures had changed. Auditor compensation was structured to ensure their commitment and included payments for training, branch visits, and transportation.

Please refer to Study Design attached for the detail.

Random assignment of auditor characteritics was done by a computer.

Individual (auditor)

No

N/A

In total, there are 10624 visits.

In each round, 32 auditors with unique randomized characteristics visited each branch for two meetings. In Round 1, they visited 42 active branches across 26 securities firms, completing 2,880 total visits. In Round 2, auditors visited 41 branches across 25 firms, completing 2,624 visits. In Rounds 3 and 4, they visited 40 branches across 25 firms, completing 2,560 total visits. Since four distinct sets of auditors were dispatched across the four rounds, a total of 128 unique auditors were hired for the entire study.

Based on our prior investigtion, the standard deviation is 18.98 for males and 21.88 for females, with a mean difference in discount between females and males of 10.01%. This results in an effect size of 0.5. Assuming 80% power, a significance level of 0.05, and an allocation ratio of 1, the minimal required total sample size is 128 for an effect size of 0.5. If the effect size is reduced to 0.2, the required sample size increases to 788. For an effect size of 0.1, the required sample size is 3,142.