Social Sharing
Extended Viewer
On the White Board - August 2008
Aug 15, 2008
Asset Correlations in Finance
Asset correlations play a crucial role in finance and risk management. They are key inputs for economic capital calculations and implicitly specify concentration risks at the name, industry and geographical levels. A reasonable estimate of asset correlations is required not only to capture credit risk at the portfolio level, but also to enable the selection of proper diversification strategies once concentration risks are identified. However, the estimation of asset correlations is difficult since they are not directly observable in the market. Typically, they are estimated either from equity correlations or a combination of both equity and balance sheet information.
An alternative approach is to seek other sources of market data to estimate asset correlations. One such approach is to combine equity information with name specific CDS data. We can then run a structural model (such as CreditGrades) backwards to estimate an intrinsic asset value. In addition, we could also back out an implied leverage estimate from CDS data.
Consider Figure 1, where different plots of correlations for Ford and General Motors are provided: (i) equity correlation, (ii) intrinsic asset correlation, and (iii) implied leverage correlation. The latter two are derived from both equity and CDS data using the CreditGrades model. Correlations are rolling and are computed using a years worth of data with equal weights. In all cases we observe a marked increase in correlations from July 2007 to August 2008. Note that asset correlations are lower than equity correlations and that the trend can move in the opposite direction. In this particular case, the credit market reacted more to Ford's delay of North American schedule for production in March 2003 than the equity market. Another example is provided in Figure 2, where the average asset and leverage correlation for a sample of US banks and brokers [1] is compared to the average equity correlation. As with our previous example, we observe that asset correlations derived from equity and CDS data is smaller than equity correlations. Given the credit turmoil since the summer of 2007, the credit market has experienced a systemic downturn and is reflected in Figure 2, whereby, in all cases we observe an increase in correlations after this period (with the greatest increase arising from leverage). Finally, as the market bottom outs in the next few quarters, it would be interesting to see if asset and leverage correlations (derived from both the equity and credit markets) remains high or reverts to pre-crisis levels.
----[1] The sample consisted of CitiGroup, JP Morgan and Chase, Wachovia, Lehman, Morgan Stanley, Merrill Lynch, and Goldman Sachs.
Figure 1: Ford and GM asset correlations