Testing Margin of Conservatism for Scarce AIRB Data

Abstract

Since the go-live of the Basel II Regime in 2007, banks calculating Risk Weighted Assets under an (Advanced) Internal Rating Based Approach (AIRB) face the challenge to proof data robustness for parameter estimation approval (PD, LGD, CCF) from supervisory authorities. The question of how much margin of conservatism (MoC) is adequate is mostly discussed on a case by case basis and practitioners from banks and supervisors often enter an intense dialogue about how much the final suggested value should be above e.g. the sample mean. The most recent EBA guideline on PD and LGD estimation requires objective thresholds to judge model and parameter stability. This article introduces a testing procedure to structure the data robustness assessment. It employs the idea of checking how additional hypothetical but realistic observations would impact the proposed value based on the available observable sample. It should help practitioners to structure the discussion around MoC with internal validation units and supervisors. The test is applied to LGD data to illustrate the concept.

Keywords: AIRB, Parameter Estimation, Margin of Conservatism, Small Sample Testing

JEL Classification: G28,G32