Counterparty Credit Risk

Download the full article

Credit Valuation Adjustement (CVA) measures the counterparty risk embedded in any financial derivative. The main difficulties linked to CVA are of computational order, since its computation relies on heavy Monte-Carlo simulation schemes. The specific issue I worked on was the optimization of the computation of CVA sensitivities. In this field, the standard method is finite difference, where first-order derivatives are estimated neglecting higher order effects. This method provides a biased but consistent estimate. However it is numerically intensive since the function of interest has to be evaluated in two separate points. I implemented several alternative methods (finite difference, likelihood ratio, Malliavin weights, pathwise estimator, Vibrato Monte-Carlo) and compared their precision and performance numerically using Python. My results have been presented to analysts and quants located in London, New-York, Paris, Hong-Kong and Tokyo.