Addressing Gender Bias in Small Business Lending

Last registered on December 20, 2024

Pre-Trial

Trial Information

General Information

Title
Addressing Gender Bias in Small Business Lending
RCT ID
AEARCTR-0015025
Initial registration date
December 13, 2024

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
December 20, 2024, 1:03 PM EST

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

Last updated
December 20, 2024, 4:13 PM EST

Last updated is the most recent time when changes to the trial's registration were published.

Locations

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

Request Information

Primary Investigator

Affiliation
University of Illinois at Urbana Champaign

Other Primary Investigator(s)

Additional Trial Information

Status
On going
Start date
2024-11-17
End date
2025-09-30
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
This project investigates gender bias in small business microfinance lending through a framed experiment conducted in Egypt. Loan officers evaluate previously approved loans with randomized applicant names. By comparing whether the same portfolios are rejected more frequently with names suggesting different genders, the study aims to identify the existence of bias. Additionally, the project explores the origins of this bias, examining whether it is inclusionary or exclusionary errors. It also investigates whether the amount of bias varies based on the incentive structure that loan officers face. The project seeks to determine how such bias might be mitigated. For example, it tests whether sensitivity training, AI-assisted decision-making, or higher incentives and penalties for incorrect decisions have a positive impact or not.

External Link(s)

Registration Citation

Citation
Jung, Youngjoo. 2024. "Addressing Gender Bias in Small Business Lending." AEA RCT Registry. December 20. https://doi.org/10.1257/rct.15025-1.4
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 the first stage, loan officers in the control group make either approval or rejection decisions with a smaller incentive and penalty. Loan officers in Treatment Group 1 receive their Gender-Career Implicit Association Test (IAT) score before making their decisions. Loan officers in Treatment Group 2 receive a higher incentive and penalty for their decisions. Lastly, loan officers in Treatment Group 3 receive both treatments simultaneously: they are notified of their IAT score and receive higher incentives and penalties.

In the second stage, loan officers make approval or rejection decisions with different portfolios. The control group loan officers repeat the same approval/rejection process as before. Loan officers in the non-interactive AI group receive the help of ChatGPT, but without the communication feature in their approval decisions. Loan officers in the interactive AI group receive the help of ChatGPT with the communication feature enabled.
Intervention Start Date
2024-11-17
Intervention End Date
2025-09-30

Primary Outcomes

Primary Outcomes (end points)
Existence of bias, following decisions of AI, and types of questions to AI.
Primary Outcomes (explanation)
With the notification of IAT scores, higher incentives/penalties, and the help of ChatGPT, the bias is expected to be smaller compared to the control group. Then, the existence is the main outcome. Also, whether participants follow AI's decisions or not explains how the AI-assisted decision making can reduce the bias.

Secondary Outcomes

Secondary Outcomes (end points)
Heterogeneous effects by loan type (performing/non-performing loans), gender of loan officers, IAT score of loan officers, risk-attitude of loan officers, and the accuracy of AI's decisions. Additionally, whether they spend more time on loan decisions and how certain they are about their decisions by treatment.
Secondary Outcomes (explanation)
Loan officers may have different bias by whether loan portfolios look good or bad. Also, loan officers may have different gender bias by their gender, IAT score, and risk attitude. Also, with higher incentive and penalty, loan officers may spend more time in their decisions. Their certainty about their decisions can differ by treatment, especially with higher incentives and penalties, and with the help of AI.

Experimental Design

Experimental Design
To investigate gender bias in loan decisions and assess interventions to mitigate it, I am conducting a framed field experiment in Egypt. This study involves approximately 400 experienced loan officers. Loan officers make hypothetical decisions about 10 previously approved small business loan portfolios, with randomized five male and five female names. They decide whether to approve or reject loans. Correct decisions, such as approving performing loans or rejecting non-performing loans, will be incentivized, while incorrect decisions will incur penalties.

The experiment consists of two stages. In the first stage, loan officers are randomly assigned to one of four groups. The control group will make decisions without additional interventions. In Treatment Group 1, loan officers receive feedback on their IAT scores before making decisions. Treatment Group 2 involves imposing higher incentives and penalties for their decisions, while Treatment Group 3 combines IAT feedback with the higher incentive and penalty structure.

In the second stage, loan officers evaluate a new set of 10 loan portfolios with assistance from ChatGPT. Loan officers are further randomized into three subgroups: The first group does not get any help from AI. The second group will use non-interactive AI, where recommendations are viewed without communication, while the last group will use interactive AI, allowing officers to ask follow-up questions. This setup enables an analysis of how AI interaction influences decision-making and whether AI-assisted tools mitigate bias.
Experimental Design Details
Not available
Randomization Method
Randomization done in office by a computer
Randomization Unit
Individual
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
400 loan officers
Sample size: planned number of observations
400 loan officers
Sample size (or number of clusters) by treatment arms
100 for each arm.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
IRB

Institutional Review Boards (IRBs)

IRB Name
UIUC IRB
IRB Approval Date
2024-08-09
IRB Approval Number
24-0796
IRB Name
American University in Cairo (AUC) IRB
IRB Approval Date
2024-07-31
IRB Approval Number
2023-2024-235
Analysis Plan

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

Request Information