PDF Ebook Asset Correlations: Shifting Tides
There is ongoing pressure from regulators, investors and rating agencies on financial institutions to build appropriate models that allow measurement of the different risks faced. For the financial institutions the implementation of these models is often a first step towards developing what is now often called an enterprise wide risk framework, which can support and reward management on an enterprise-wide basis by integrating all risk components. As far as credit risk is concerned it follows from Sklar’s theorem that one is able to assess the risk of the entire loan portfolio provided that the dependency structure or copula is known (as well as the marginal distributions of the individual credit losses). However, whilst the assessment of the marginal risks is now relatively well under control, fitting a copula in a credit context is a difficult exercise due to the relative scarcity of observed defaults. Therefore, most institutions limit themselves to using default correlations when assessing the dependencies and use a variance-covariance framework to asses the credit portfolio risk. It must be noted that in the literature on default correlations it is now standard to use the concept of asset correlation to discuss and to compare the different findings. Indeed, an assumption on joint asset movements (typically using a Gaussian copula) allows one to back out the implied asset correlations from the default correlations; we refer to Crouhy et al (2000) for more details.
Asset correlations are also an important component of the Basel II Accord for regulatory capital requirements of credit risk portfolios. In the Basel Committee on Banking Supervision (BCBS) document of January 2001 (BCBS (2001a), asset correlations were assumed to take a value of 20% for all obligors. The modification later that year (BCBS (2001b)) assumed that asset correlation declined with PD: for the lowest PD the asset correlation was 20% and for the highest PD the asset correlation was 10%. Then finally in the Consultative Document of 2006 asset correlations for sovereigns, banks and corporates were principally assumed to be between12% and 24%, once again depending on the probability of default. We note that for small and medium sized corporates an extra downward firm-size adjustment up to 4% is made and this brings the effective range of corporate asset correlations between 8% and 24%.
There was limited research at the time of the first Basel II document and as such, it seems that these estimates were at least partly based on industry perception and practice and this is admitted in the document (BCBS (2001a).
In the subsequent years there have been numerous papers using different approaches and different datasets, the majority of which report dramatically different results. These can be largely separated into two categories. The first uses observed default data to calculate single and pairwise default frequencies from which default correlations are derived. Papers in this category include Gordy (2000), Frey and McNeil (2003), Dietsch and Petey (2004), Jobst and de Servigny (2005) and de Servigny and Renault (2002).
Posted in :
