ABSTRACT

We show how redemption risks of mutual funds can be modeled using the peaks-over-threshold approach from extreme value theory. The resulting risk measure liquidity-at-risk is adapted to cover issues arising when fund redemption data from the real world is used, and we give guidelines for what should be considered in practice. We also provide an automated and easily applicable procedure for determining the threshold parameter of a generalized Pareto distribution by means of a given data set. Moreover, we supplement our findings with a thorough backtesting analysis.