THE EFFECTS OF ESTIMATION ERROR ON MEASURES OFPORTFOLIO CREDIT RISK

This paper uses Monte Carlo simulations to assess the impact of noisy inputparameters on the accuracy of estimated portfolio credit risk. Assumptionsabout input quality are derived from the distribution of historical samplestatistics commonly used in default risk modelling. The resulting estimationerror in the distribution of portfolio losses is considerable. Losses that arejudged to occur with a probability of 0.3% may actually occur with aprobability of 1%. The paper also shows that estimation error leads to biasesin VaR estimates and significance levels of backtests. The biases can becorrected by analysing predictive distributions which average over theunknown parameter values.