A New Approach For Modelling And Estimating Expected Shortfall
Estimation of the coherent measure Expected Shortfall (ES) has been a crucial but challenging task in portfolio risk management, where moment modeling plays a critical role. Existing ES estimations utilize classical central moments and perform poorly for non-normal case and in the presence of outliers. This paper first derives a flexible multivariate variant for Hoskings (1997) L-moments, the Trimmed L-comoments, and then applies it to the estimation of Expected Shortfall via Cornish-Fisher and Edgeworth expansion. Simulations demonstrate that, at various levels of loss probability, Trimmed L-comoments based estimation for Expected Shortfall works superb in the skewed and heavy-tailed cases which most financial data lie in, and is extremely useful when outliers occur. It also shows remarkable robustness, as it works when classical moments and L-moments do not exist. Out-of-sample forecasts and backtests in an empirical study on Chinas stock indices in the coal and electricity industry chain also indicate superior accuracy and reliability for the new risk estimation approach.
expected shortfall Cornish-Fisher expansion Edgeworth expansion trimmed L-comoments heavytail
QIN Xiao
Antai College of Economics and Management,Shanghai Jiao Tong University, Shanghai, P. R. China, 200052
国际会议
2010 International Conference on Management(2010管理国际大会)
上海
英文
195-201
2010-07-24（万方平台首次上网日期，不代表论文的发表时间）