Improvements in the Measurement of Systemic Risk for Credit Portfolio Management

This paper argues that the correlation calculations embedded in the Basel rulebook are unable to provide useful estimates of the systemic risk factor and therefore estimates of unexpected losses. This is because the nature of systemic risk is dynamic and cannot be estimated from a static model based on historical data. The implication is that systemic risk is not universal but is dependent on the development of credit conditions within each economy. Data collected from a neo-Wicksellian analysis of the US economy suggests that there is a strong relationship between the credit cycle, asset correlation and loss rates. An excess growth in credit will eventually reverse leading to rising correlation and loss rates.Our various attempts at model calibration to estimate the extent of asset correlation based on the growth in credit at a point in time have to date not yielded particular promising results. This is partly due to a paucity of data and partly due to the challenge of modelling linear systems that periodically break down displaying nonlinear characteristics. Approaches using nonlinear dynamics may well be able to provide more insight into such a model calibration