| |
Abstract:
In the wake of recent failures of risk management, there has been a widespread call for improved quantification of the financial risks facing firms. At the forefront of this clamor has been Value at Risk. Previous research has identified differences in models, or Model Risk, as an important impediment to developing a Value at Risk standard. By contrast, this paper considers the divergence in a model's implementation in software and how it too, affects the establishment of a risk measurement standard. Different leading risk management systems' vendors were given identical portfolios of instruments of varying complexity, and were asked to assess the value at risk according to one common model, J.P. Morgan's RiskMetrics. We analyzed the VaR results on a case by case basis, and in terms of prior expectations from the structure of financial instruments in the portfolio, as well as prior vendor expectations about the relative complexity of different asset classes. It follows that this research indicates the extent to which one particular model of risk can be effectively specified in advance, independent of the model's detailed implementation and use in practice.
|
