An Analysis of Counterparty Credit Risk Estimation Methods Using Monte Carlo Simulation

Abstract

The collapse of the banking sector during the 2008 financial crisis has been largely attributed to the defaults of collateralized debt obligations and subprime mortgages which spread risk contagion as markets recognized widespread improper calculations of counterparty credit risk. Given that most commercial banks today find their great- est source of risk to be their loan portfolios, the banking sector continuously seeks out methodologies of estimating counterparty credit risk to a higher degree of accuracy. Moreover, any firm that participates in the Over-the-Counter derivatives market is exposed to counterparty credit risk through risk transference. This thesis therefore aims to outline a variety of methodologies to estimate the Counterparty Credit Risk of interest rate swap derivatives using Monte Carlo simulation. The procedures im- plemented to estimate counterparty credit risk have changed over the years due to both economic events that have rendered certain models unreliable as well as regula- tory pressures. Therefore, this thesis will begin by discussing the procedure utilized most widely prior to the 2008 financial crisis that employs a Hull-White One Factor interest rate model in its simulation of Credit Value Adjustment as a proxy estimator for counterparty credit risk. Next, a model that employs a 2-Factor Gaussian interest rate process in its simulations of a CVA estimator will be described to illustrate the flaws inherent in the previous model. This thesis will then expand its discussion of counterparty credit risk estimation methods by describing the procedures for imple- menting Structural and Intensity models and a Dynamic Correlated Default Timing model which estimate expected portfolio losses without using CVA as a counter- party credit risk estimator. These advanced modeling techniques incorporate default dependencies and systemic risk factors into the counterparty credit risk estimation framework and therefore are recognized as more realistic/accurate methods relative to CVA estimation methods. This thesis concludes with a discussion of the advantages and disadvantages of all of the aforementioned models.