Skip to content

FRM Part II · FRM Exam Part II · Credit Risk

A bank builds a logistic regression scorecard for consumer loans. During development, the team includes a variable that is only recorded after a customer has been delinquent for 60 days. The model shows an extremely high Gini coefficient in-sample. What is the most likely problem?

The most likely problem is information leakage. The variable is recorded only after delinquency, so it reveals the outcome and inflates in-sample discrimination. It is unavailable when scoring new applicants, so real out-of-sample performance will be much worse than the high Gini suggests.

  1. AInformation leakage, so the model will perform poorly on new applicantsCorrect
  2. BUnderfitting, so the model lacks enough variables
  3. CLow discriminatory power masked by high calibration
  4. DSurvivorship bias that understates the Gini

Explanation

A variable only observed after delinquency is effectively an outcome proxy, not available at application. It inflates in-sample discrimination and fails at origination. This is leakage, not underfitting, and it overstates rather than understates performance.

Did you get it right without looking?

One question tells you little. A timed set on Credit Risk shows your real accuracy, how long you take and where you lose marks.

Practise Credit RiskFree diagnostic mock

More Credit Risk questions