Approximating Unilateral CVA for Interest Rate Swap Portfolios

Sammanfattning

This thesis investigates whether the unilateral credit valuation adjustment (CVA) of plain-vanilla interest rate swap portfolios can be approximated accurately using a pre- computed library of basis-swap CVAs, avoiding the cost of full Monte Carlo revaluation. The approximation is constructed to require only a yield curve, a swaption volatility surface, and counterparty CDS spreads as inputs. The method is evaluated against a Monte Carlo benchmark for counterparties from three sectors and across four interest-rate models: the Hull–White one-factor, Hull–White two-factor, shifted Cox–Ingersoll–Ross, and a stochastic-volatility Hull–White model. The estimation achieves a mean absolute percentage error of 20–24% across models. Introducing a scalar adjustment to the netting ratio reduces the MAPE to 16.1%. The results indicate that the estimation could have practical use for fast intraday calculation of CVA but as a complement rather than a substitute for full Monte Carlo CVA.