Table of Contents
Is a Lending Model Fair If Approval Rates Match but Qualified Applicants Are Treated Differently?
Learn what it means when a loan approval model passes demographic parity but fails equal opportunity, and why equal treatment of qualified applicants matters in fairness analysis.
Question
Your loan approval model passes demographic parity (similar approval rates across groups) but fails equal opportunity (different true positive rates for qualified applicants across groups). What does this tell you about the model’s fairness?
A. The model approves similar percentages of applicants from each group, but qualified applicants from some groups are less likely to be approved than equally qualified applicants from other groups.
B. The model is completely fair because demographic parity is the only fairness metric that matters in lending.
C. The model should be immediately deployed because it meets at least one fairness criterion.
D. The model has too many false positives for one group and should be retrained.
Answer
A. The model approves similar percentages of applicants from each group, but qualified applicants from some groups are less likely to be approved than equally qualified applicants from other groups.
Explanation
The model may look balanced at a high level because approval rates are similar across groups, but that does not mean it is fair to qualified applicants. If equal opportunity fails, some qualified people are being approved less often than equally qualified people in another group.
In lending, that matters because fairness is not only about matching approval percentages. It is also about giving applicants with similar creditworthiness a similar chance of approval, regardless of group membership.
Why the others are wrong
B is incorrect because demographic parity is only one fairness metric and does not capture whether qualified applicants are treated equally. C is incorrect because meeting one fairness criterion does not automatically make a lending model safe or fair enough to deploy.
D is too specific because failing equal opportunity points to different true positive rates, not necessarily to false positives. A false-positive issue would relate more directly to equalized odds or false positive rate analysis.