According to Basel II guidelines, all financial institutions should have proper quantitative tools to manage and control the credit risk. CreditRisk+ model developed by Credit Suisse Financial Products (CSFP) is one of the most widely used models for calculating credit value-at-risk. CreditRisk+ is a good choice for many banks for calculating portfolio credit value-at-risk, because this model requires limited amount of inputs and is formulated with limited assumptions. There has been a large amount of research work on CreditRisk+ to enhance its capabilities and features. This paper aims to suggest a parallel computing-based Monte-Carlo simulation, for a very large portfolio using the CreditRisk+ infrastructure, to calculate the portfolio loss distribution.

Supervisory framework changes in the Basel II Capital Accord have tried to bridge the gap between the regulatory capital and the economic capital.1 Banks have developed many sophisticated tools that enable them to manage and collect the data in more efficient ways than earlier. The proposed internal rating system (Basel II) would allow banks to collect data such as exposures and loss given default in an efficient and effective way. These developments in the regulations would allow banks to create new tools to assess their risks and quantify them in a more appropriate fashion. There has been continuous development in the credit risk modeling area. There are three ways to model the credit risk: (1) Ratingsbased such as CreditMetrics2; (2) Structural or firm-value-based such as KMV3; and (3) Reduced form such as CreditRisk+4 (Crouhy et al., 2000).

In this paper, we will focus on CrediRisk+(CR+) model. CR+ is the credit risk model based on actuarial science framework and was developed by Credit Suisse Financial Products (CSFP). This model has become one of the favorite models amongst the supervisory and regulatory community (Austrian Financial Market Authority and Oesterreichische Nationalbank, 2004; and Gundlach and Lehrbass, 2004). Few of the advantages associated with this model are: (1) Requires limited amount of input data; (2) Requires limited assumptions for the formulation of the model; (3) Uses the same data as required by IRB approach in Basel II; and (4) Addresses concentration risk.

Original CR+ model with independent sectors uses Panjer algorithm to compute the analytical solution. Gordy (2002) showed that this recursive algorithm is numerically unstable for large portfolios with many risk factors. There are two alternative methods that can be used to compute the loss distribution analytically.6 The first one is a recursive algorithm suggested by Giese (2003). Haaf et al. (2003) showed that this algorithm is numerically stable; the precision errors are not propagated and amplified by the recursive formulas. The other method suggested by Melchiori (2004) is based on the Fast Fourier Transform (FFT) and it works well in the portfolio with large number of obligors and independent sectors (Avesani et al., 2006).

Financial Risk Management Journal, Credit Risk Model, Credit Suisse Financial Products, Recursive Algorithm, Fast Fourier Transform, Monte-Carlo Simulations, Software Industry, Automobile Industry, Gamma Distributions, Computational Resources, Bernoulli Distribution, Compound Gamma Model.