Here is your quant homework for the weekend...
Damiano Brigo and IMI colleagues Andrea Pallavicini and Roberto Torresetti discuss the issue of modelling loss distributions in pools of credit names in this month's Risk. They produce a fairly straightforward GPL model, but the paper leaves some questions open: leave your thoughts in the comments.
Consistency with single names remains an issue whenever one models losses directly, especially to check sensitivities with respect to single-name CDS and related issues. What are the viable paths one can take for this? Is the GPCL extension of the GPL model presented here, based on the common Poisson shock framework, a viable method? What about the alternative and controversial random thinning technique? (Thanks to Dr Brigo for suggesting this question).
The correct pricing and calibration of bespoke CDO tranches or CDOs with optional features (e.g. forward starting CDOs) is still hindered by the lack of correlation data and thin trading. When will the market develop a reliable term structure of correlation to allow us to analytically price these or more exotic CDO products?
Many of the current sophisticated CDO pricing models suffer from the lack of liquidity in forward starting and bespoke tranches. The investment banks may fear to create new products that they cannot price correctly given this lack of data. This could in theory slow down even more the generation of those data that would make their models work: is this a chicken and egg situation?
Meanwhile, for non-quants, file this under "unintended consequences" - insider trading is way up this year, according to the FT. Understandable - cheap capital and the growth of private equity means more scope for buyouts and takeovers, which in turn means more opportunities for criminals to make a quick profit on advance notice of a move. Expect more emphasis, and possibly more intrusive oversight, from the SEC and its fellow regulators later this year.
Subscribe to this blog's feed
