Spectral backtests of forecast distributions with application to risk management
Chief： Prof. Alexander J. McNeil
Time： 2019-11-21 14:00-15:00
摘要： We study a class of backtests for forecast distributions in which the test statistic depends on a spectral transformation that weights exceedance events by a function of the modeled probability level. The weighting scheme is specified by a kernel measure which makes explicit the user’s priorities for model performance. The class of spectral backtests includes tests of unconditional coverage and tests of conditional coverage. We show how the class embeds a wide variety of backtests in the existing literature, and further propose novel variants which are easily implemented, well-sized and have good power. In an empirical application, we backtest forecast distributions for the overnight P&L of ten bank trading portfolios. For some portfolios, test results depend materially on the choice of kernel.
Alexander McNeil has been Professor of Actuarial Science at the University of York since September 2016. Educated at Imperial College London and Cambridge University, he was formerly Assistant Professor at ETH Zurich and Maxwell Professor of Mathematics at Heriot-Watt University. His research interests lie in the development of quantitative methodology for financial risk management and include models for market, credit and insurance risks, financial time series analysis, models for extreme risks and correlated risks and enterprise-wide models for solvency and capital adequacy. He is joint author, together with Rüdiger Frey and Paul Embrechts, of the book "Quantitative Risk Management: Concepts, Techniques and Tools", published by Princeton University Press (2005/2015).