Gini coefficient (credit scoring)
In credit scoring, the Gini coefficient measures how well a score ranks goods and bads. It is calculated from the ROC curve as twice the area under the curve (AUC) minus one. A value of 0 means ranking no better than chance, and 1 means every bad scores below every good.
Example
For illustration, suppose a score has an AUC of 0.75. Its Gini is 2 × 0.75 − 1 = 0.50.
Calculating Gini from score bands
Put the score bands in score order and tally the goods and bads in each one. The band counts are used to draw the ROC curve.
The curve captures how the two outcome groups are ranked across the bands. Its area is the AUC input to the Gini calculation.
Reading Gini across products and samples
A Gini depends on how the score ranks goods and bads in the sample being measured. It also depends on who is in that sample: the same score measured only on approved applicants usually has a lower Gini than when measured on all applicants.
Compare values on the same population. A difference between products or samples can reflect who was measured, so it does not by itself show that one score ranks outcomes better.
Questions we get asked
Is the credit scoring Gini the same as the economics Gini?
No. In economics, the Gini coefficient measures income inequality. In credit scoring, it measures how well a score ranks goods and bads.
Who decides which outcomes count as goods and bads?
Goods and bads are the outcome groups in the sample. Gini uses those labels, but does not prescribe a particular definition for them.
Does a Gini score decide who should get credit?
No. It measures how the score ranks goods and bads in a sample. The lender writes the policy that determines how scores are used for credit decisions.
Decisimo’s free Gini coefficient and AUC calculator calculates both measures from goods and bads in score bands and draws the ROC curve.
