AI Playbook for Credit Card Limit Disparate Impact

A credit card issuer's fair lending team identified that initial credit limits for Black cardholders are on average $1,200 lower than for white cardholders at the same FICO score band. The issuer uses an algorithmic model for initial line assignment. The model was last validated 3 years ago.

When to use this playbook

• Use this playbook when the decision looks like the situation above: A credit card issuer's fair lending team identified that initial credit limits for Black cardholders are on average $1,200 lower than for white cardholders at the same FICO score band.
• It is a fit when you have source files in hand and need a structured, reviewable analysis — not a generic chat answer about "Credit Card Limit Disparate Impact".
• Do not use it as a substitute for licensed, legal, clinical, or authorized official judgment in the domain.

What you'll need

• Credit card origination file (all accounts, past 24 months: initial limit, FICO, income, utilization, geographic data)
• Model documentation for the credit limit assignment algorithm
• Last model validation report (3 years old)
• Bureau attribute data used by the model
• CFPB UDAP and ECOA guidance on credit limit setting

Attachments: Documents (Documents)

You are a fair lending analyst investigating credit card limit disparities at a credit card issuer. I am attaching:

Work only from the attached source files. If a conclusion is not supported, say so.

Produce:
1. Control for all legitimate limit-setting factors (FICO, income, bureau utilization, time at address, employment) and quantify the residual limit disparity by race.
2. Identify which model features are most correlated with the disparity—are there proxy variables (geography, bureau tradeline types) that disproportionately disadvantage Black applicants?
3. Assess the model validation gap: what should have been updated in the 3 years since last validation and what the fair lending testing should have caught.
4. Calculate the harm to affected cardholders: how much less credit access did they receive and what is the potential remediation cost?
5. Tell me the regulatory exposure and whether this requires a voluntary self-disclosure to the CFPB before the next examination.

Call out where independent models are likely to disagree, and list follow-up documents a reviewer should request.

What to expect

• Regression-controlled limit disparity by race
• Proxy variable identification
• Model validation gap analysis
• Harm quantification and remediation cost estimate
• Regulatory exposure and voluntary self-disclosure recommendation

Review before you act

• Validate this output against source files before relying on it: Control for all legitimate limit-setting factors (FICO, income, bureau utilization, time at address, employment) and quantify the residual limit disparity by race.
• Validate this output against source files before relying on it: Identify which model features are most correlated with the disparity—are there proxy variables (geography, bureau tradeline types) that disproportionately disadvantage Black applicants?.
• Validate this output against source files before relying on it: Assess the model validation gap: what should have been updated in the 3 years since last validation and what the fair lending testing should have caught.
• Validate this output against source files before relying on it: Calculate the harm to affected cardholders: how much less credit access did they receive and what is the potential remediation cost?.
• Confirm every cited figure, date, counterparty, or requirement against the attached originals — models compress and can drop a qualifier.
• Treat disagreement between models as a review item, especially on classification, materiality, and recommended next action.
• Do not authorize an operational, clinical, legal, credit, or enforcement action solely because the models agree.