Portfolio Risk Management
Case study
Portfolio Risk Management
#Finance #Risk
🡒 Portfolio management requires risk assessment, and this can be done in a range of ways from very qualitative, to in-depth modelling and scenario simulation.
Managing liquidity is a particular challenge post-financial crisis. Good practice for fund managers dictates that stress testing be conducted regularly for "extreme but plausible scenarios" – this is simple to state, but difficult or unclear how it is implemented. One particular issue is that assets will not respond independently in light of many extreme and plausible scenarios, but there are limited means to estimate this directly. We provide statistical simulation-based stress testing which naturally incorporates relationships between assets. Combined with historic data, expert-elicitation, and scenario planning, probabilistic liquidity requirements can be addressed with associated uncertainties.
Ultimately the problem distils to a probability distribution of loss-return, over a window of time. Data was a combination of historically observed distributions of returns, expert-elicitation for levels of inter-asset correlations and their uncertainties, and additional expert-derived parameters. The loss-return PDF is generated from Monte-Carlo simulation, with copula methods underpinning the simulations. This permits the specification of arbitrary shapes for the marginal asset return distributions, but also an arbitrary inter-asset correlation matrix.
#Finance #Risk
🡒 Portfolio management requires risk assessment, and this can be done in a range of ways from very qualitative, to in-depth modelling and scenario simulation.
Managing liquidity is a particular challenge post-financial crisis. Good practice for fund managers dictates that stress testing be conducted regularly for "extreme but plausible scenarios" – this is simple to state, but difficult or unclear how it is implemented. One particular issue is that assets will not respond independently in light of many extreme and plausible scenarios, but there are limited means to estimate this directly. We provide statistical simulation-based stress testing which naturally incorporates relationships between assets. Combined with historic data, expert-elicitation, and scenario planning, probabilistic liquidity requirements can be addressed with associated uncertainties.
Ultimately the problem distils to a probability distribution of loss-return, over a window of time. Data was a combination of historically observed distributions of returns, expert-elicitation for levels of inter-asset correlations and their uncertainties, and additional expert-derived parameters. The loss-return PDF is generated from Monte-Carlo simulation, with copula methods underpinning the simulations. This permits the specification of arbitrary shapes for the marginal asset return distributions, but also an arbitrary inter-asset correlation matrix.