Risk Analysis via Generalized Pareto Distributions

• DOI: 10.1080/07350015.2021.1874390
• Publication Date: 2021-02-17T21:19:58Z
• Authors: Yi He; Liang Peng; Dabao Zhang; Zifeng Zhao
• Keywords: Quantile; Delta method

We compute the value-at-risk of financial losses by fitting a generalized Pareto distribution to exceedances over threshold. Following common practice setting threshold as high sample quantiles, we show that, for both independent observations and time-series data, asymptotic variance maximum likelihood estimation depends on choice threshold, unlike existing study using divergent also propose random weighted bootstrap method interval VaR, with critical values computed empirical absolute differences between bootstrapped estimators estimator. While our results unify inference non-divergent thresholds, finite studies via simulation application real data that derived confidence intervals well cover true VaR in insurance finance.