QRR CAT: A credit analysis tool
Performs the calculation, at different levels of aggregation, of the portfolio risk measures (economic capital, expected loss, ...).
Allows the optimization of the portfolio using the maximization of economic profit subjet to given restrictions.
- The theoretical background of the analysis is based on interdependant gaussian risk factors, this places the model in a Merton-Vasiseck multifactor type of framework for credit risk.
- Two approaches for the evaluation of the risk metrics are included in this tool: Approximate analytic model and Monte Carlo simulation model.
- The approximate analytic approach is based on the formulation given by Pykhtin for the estimation of the VaR (value at risk) and the CVaR (conditional value at risk or expected shortfall), that includes the multifactor and portfolio granularity effects.
- On the contrary to Monte Carlo based approaches, the approximation has the advantage of using analytical formulas. This implies a reduced amount of time spent for each calculation.
It allows the estimation (using the Euler method) the contribution of each component of the portfolio to the risk measures, which allows a method to assign the capital to each component.
- The Monte Carlo approach is based on the simulation of every risk factor in the model, including the dependencies among them. Risk metrics are obtained using these simulations.
Despite of its computation cost, the Monte Carlo approach allows using richer models, specifically:
- Using a stochastic LGD which depends on the risk factors. It allows to incorporate correlation between the LGD and the default events.
- Incorporating large counterparties with high exposure (granularity effect).
The calculation of the credit risk measures is performed at different aggregation levels (global, by geography, by sector and individually), taking into account both the effects of diversification as the idiosyncratic risks of individual positions.
Maximizes the economic profit of the portfolio subject to given restrictions.
The user can configure what restrictions he wishes to maintain active in each optimization.
The restrictions can be applied at different levels of aggregation (individually or by geography)
The data is uploaded via Excel� files.
Using the interface, the user can modify the portfolio configuration and the optimization parameters (restrictions, ...).
Reports with the analysis and optimization results can be exported to files in the Excel� format.
- Web application developed in Java using a client-server 3-tier arquitecture, consistent with the J2EE (Java Enterprise Edition) standard.
- Use of efficient c/c++ optimization libraries for optimal numerical performance.
- Monte Carlo analysis implemented on CUDA� (GPGPU), a booming technology for massively parallel processing.