Unintended discrimination in financial advice

Last registered on April 22, 2025
View Trial History

Pre-Trial

Trial Information

General Information

Title
Unintended discrimination in financial advice
RCT ID
AEARCTR-0014980
Initial registration date
April 17, 2025

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
April 22, 2025, 12:12 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

Country
Germany
Region

Primary Investigator

Name
Markus Eyting
Affiliation
JGU Mainz
Contact Primary Investigator

Other Primary Investigator(s)

PI Name
Florian Hett
PI Affiliation
JGU Mainz
Contact Investigator
PI Name
Christine Laudenbach
PI Affiliation
SAFE Leibniz Institute for Financial Research
Contact Investigator

Additional Trial Information

Status
In development
Start date
2025-04-17
End date
2025-05-31
Keywords
Behavior, Finance & Microfinance, Gender
Additional Keywords
discrimination, risk preferences, financial advice
JEL code(s)
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Wealth inequality across social groups is a pressing concern in modern society (Chancel et al. 2022). Factors like labor market participation or systematic wage gaps are typically considered as key explanatory factors. However, even with initially equalized income and wealth distributions, systematic socioeconomic disparities may emerge due to variations in investment returns and even accumulate over time.
As investment decisions are complex, as they should ideally fit individual circumstances and preferences, many households require expert guidance and advice, which mostly takes place via actual social interaction – human advisors assess the situation of the individual advisee and make their investment recommendations accordingly. This interpersonal nature introduces scope for biases in judgment and decision making, that might negatively affect the quality of particular financial advice. Depending on the nature of these biases, they might differentially affect the advice to specific social groups, hence giving rise to differences in expected returns and ultimately amplifying systematic differences in financial wellness.

This research project investigates the influence of (mis)perceptions and stereotypes in the process of investment advice to better understand the underlying mechanisms driving differences in wealth accumulation. Existing work has looked at discrimination in financial advice with respect to general quality, i.e., whether specific groups have the same access to financial products, receive the same product at higher prices or inferior products at similar prices (e.g., Alesina et al. 2013, Brock and De Haas 2020, Bucher-Koenen et al. 2023, Bhattacharya et al., 2023). In this project, we take another perspective and consider differences in the match quality as the relevant outcome: If financial advice is to a substantial part about finding the right products given the specific preferences of an individual, professionals’ advice may effectively discriminate against specific groups, if the advice is based on inaccurate perceptions of these preferences. If this inaccuracy differs systematically between specific social groups, this mechanism can induce unintended discrimination: Even though professionals give advice “to the best of their knowledge”, differences in this knowledge lead to differential treatment of different groups.

We analyze the existence and accuracy of stereotypes regarding investor preferences of men and women. We ask whether these stereotypes systematically vary in their accuracy, i.e., whether men and women are equally understood in terms of their investment preferences. We also investigate whether professional advisers are particularly prone or resistant to (wrong) stereotyping.
External Link(s)

Registration Citation

Citation
Eyting, Markus, Florian Hett and Christine Laudenbach. 2025. "Unintended discrimination in financial advice." AEA RCT Registry. April 22. https://doi.org/10.1257/rct.14980-1.0
Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
Experimental Details

Interventions

Intervention(s)
In stage 1, we recruit a representative sample of the German population who represent clients seeking financial advice. We elicit two different kinds of information on their risk preferences, as well as various demographic information, such as gender, age, or the clients’ race, as well as open text field information about the clients' looks and hobbies.

In stage 2, we recruit (i) a second representative sample of the German population acting as financial advisors and (ii) a German sample of actual financial advisors. All participants are presented with a sequence of client profiles from stage 1 and asked for their portfolio recommendation for the particular client and how confident they feel about this advice.

In stage 2 we vary the incentive and information structures of the advisors.
Intervention (Hidden)
In stage 1, we recruit a representative sample of the German population who represent clients seeking financial advice. We elicit three kinds of information from these clients: (1) using a standardized questionnaire from financial advisors, we obtain clients’ stated financial risk preferences as it is actually elicited in typical advise processes, (2) using a validated interactive risk simulation method (Kaufmann et al. 2013, Bradbury et al. 2015), we elicit the clients’ actual financial risk preferences in the form of a preferred risky share in a portfolio, (3) we ask for various demographic information, such as gender, age, or the clients’ race, as well as open text field information about the clients' looks and hobbies. We vary the order between (1) and (2).

In stage 2, we recruit a German sample of financial advisors. Hereby, part of our sample consists of practicing financial advisors, whereas the other part consists of laymen who take the role of a financial advisor in our experiment. All participants are presented with a sequence of client profiles, including AI-generated images (based on (3) above), as well as information from (1), and (3). They are asked for (i) their portfolio recommendation for the particular client and (ii) how confident they feel about this advice.

In stage 2 we vary whether the financial advisors are paid out (i) according to how well their advice matches the own portfolio choice of the clients as elicited with the interactive risk simulation tool, (ii) by a fixed fee, irrespective of their advice, as a randomly drawn advice will automatically be implemented for the respective client, or (iii) by a fixed fee, conditional on whether a randomly drawn advice was accepted by the respective client. For each of these payment schemes, advisors are first asked to select 5 of the 10 clients they gave recommendations to, of which one is then randomly drawn. This allows to test for differences between perceived and actual accuracy of advisors' recommendations.

In a cross-intervention to (ii) and (iii), we also vary whether (a) the financial advisors see the outcome of the interactive risk simulation tool (2), or (b) not.
Intervention Start Date
2025-04-17
Intervention End Date
2025-05-31

Primary Outcomes

Primary Outcomes (end points)
- Portfolio recommendation of advisors
- Client risk preferences as measured by the interactive risk simulation tool
Primary Outcomes (explanation)
- Portfolio recommendation of advisors
- Client risk preferences as measured by the interactive risk simulation tool

Do these match differently for men and women?
How does this vary between treatments?

Secondary Outcomes

Secondary Outcomes (end points)
- Risk elicitation using standard survey
- Groupiness measure
- Age, Gender, nationality, education, profession, political party, religion
- Decision times
- Attention Checks
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
We run a two staged online experiment.

In stage 1, we recruit a representative sample of the German population who represent clients seeking financial advice. We elicit two different kinds of information on their risk preferences, as well as various demographic information, such as gender, age, or the clients’ race, as well as open text field information about the clients' looks and hobbies.

In stage 2, we recruit a German sample of financial advisors. All participants are presented with a sequence of client profiles from stage 1 and asked for (i) their portfolio recommendation for the particular client and (ii) how confident they feel about this advice.

In stage 2 we vary the incentive and information structures of the advisors.
Experimental Design Details
In stage 1, we recruit a representative sample of the German population who represent clients seeking financial advice. We elicit three kinds of information from these clients: (1) using a standardized questionnaire from financial advisors, we obtain clients’ stated financial risk preferences as it is actually elicited in typical advise processes, (2) using a validated interactive risk simulation method (Kaufmann et al. 2013, Bradbury et al. 2015), we elicit the clients’ actual financial risk preferences in the form of a preferred risky share in a portfolio, (3) we ask for various demographic information, such as gender, age, or the clients’ race, as well as open text field information about the clients' looks and hobbies. We vary the order between (1) and (2).

In stage 2, we recruit another German sample. All participants are presented with a sequence of ten client profiles, including AI-generated images (based on (3) above), as well as information from (1), and (3). They are asked for (i) their portfolio recommendation for the particular client and (ii) how confident they feel about this advice.

In stage 2 we vary whether the financial advisors are paid out (i) according to how well their advice matches the own portfolio choice of the clients as elicited with the interactive risk simulation tool, (ii) by a fixed fee, irrespective of their advice, as a randomly drawn advice will automatically be implemented for the respective client, or (iii) by a fixed fee, conditional on whether a randomly drawn advice was accepted by the respective client.

In a cross-intervention to (ii) and (iii), we also vary whether (a) the financial advisors see the outcome of the interactive risk simulation tool (2) or (b) not.

We plan to run (i), (iia/iib), and (iiia/iiib) among laymen who take the role of a financial advisor in our experiment (laymen sample).
We plan to independently run (i), (iia), and (iiia) among practicing financial advisors (expert sample).
Randomization Method
randomization by a computer
Randomization Unit
individual
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
clients: 200
laymen sample of advisors: 2000
expert sample of advisors: 400
Sample size: planned number of observations
Given that each advisor gives ten portfolio recommendations, we plan with 2000*10 = 20.000 portfolio recommendations of laymen advisors. Given that each advisor gives ten portfolio recommendations, we plan with 400*10 = 4000 portfolio recommendations of expert advisors. We plan to elicit 500 risk preferences from clients (one each).
Sample size (or number of clusters) by treatment arms
laymen sample of advisors: 400 per treatment arm
expert sample of advisors: 130 per treatment arm
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Supporting Documents and Materials
IRB

Institutional Review Boards (IRBs)

IRB Name
Gemeinsame Ethikkommission Wirtschaftswissenschaften der Goethe-Universität Frankfurt und der Johannes Gutenberg-Universität Mainz
IRB Approval Date
2024-10-30
IRB Approval Number
N/A
Analysis Plan

Post-Trial

Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Intervention

Is the intervention completed?
No
Data Collection Complete
Data Publication

Data Publication

Is public data available?
No

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials