相関係数の変動性を考慮した、信用リスク管理方法について —相関係数をランダム行列化したガウシアンコピュラモデルの提案(1)(2)—相関係数の変動性を考慮した、信用リスク管理方法について —相関係数をランダム行列化したガウシアンコピュラモデルの提案(1)(2)— A New Framework for Credit Risk Management with the Randomness of Correlation Matrix
In this paper, the author proposes a new approach for credit risk managementby applying standard method. As a matter of fact, almost all financial institutionsin Japan are using standard method which is called as Gaussian copula for theircredit risk management. They estimate losses from their credit portfolio under thestressed economic scenario and confirm that the losses are able to be covered bytheir own capital margin. Especially after Lehman crisis in September 2008,many researchers have been pointing out insufficiency of the standard methodand started to propose new approaches because losses based on standard were notenough conservative. Common points in these approaches are their utilizing fattaildistributions so as to let the correlations between assets work effectively forhaving satisfactory number of simultaneous defaults. They are t-copula, Gumbelcopula,Clayton-copula, and so on. The new approach in this paper is completelydifferent from these in terms of following points. Firstly, it considers randomnessof correlations in view of actual market fluctuations. Secondly, it can be regardedas an improved version of standard method to preserve its tractability, transparency,and simpleness, which are strongly recommended in the current Basel committeereports. Thirdly, the new approach can output reasonable losses with concisesystem. In this paper, the author explains the theoretical aspects of this approachand the results from numerical experiments with actual financial market data.