Performance measurement of early warning models
提出一种新的度量指标来评估预测银行严重弱化或破产的模型，解决了传统正确分类百分比指标在低比例弱化银行被正确分类时仍可能偏高的问题，并比较了多个近期预警模型的绩效。
The paper presents a new measure to evaluate models which predict severe bank weakness or failure. The conventional measure has been the ‘percentage classified correctly (CC)’. This measure can be quite high even though a low percentage of weak or failed bank is classified correctly. We resolve this problem by weighting CC by two additional factors: (1) banks that actually weakened or failed as a percentage of those that fail a model's ‘hurdle test’, and (2) the percentage of all weak or failed banks correctly classified. The paper then compares the performance of several recently published early warning models using the new measure.