Filtered extreme-value theory for value-at-risk estimation: evidence from Turkey

Purpose - The purpose of this paper is to use filtered extreme-value theory (EVT) model to forecast one of the main emerging market stock returns and compare the predictive performance of this model with other conditional volatility models. Design/methodology/approach - This paper employs eight filtered EVT models created with conditional quantile to estimate value-at-risk (VaR) for the Istanbul Stock Exchange. The performances of the filtered EVT models are compared to those of generalized autoregressive conditional heteroskedasticity (GARCH), GARCH with student-t distribution, GARCH with skewed student-t distribution, and FIGARCH by using alternative back-testing algorithms, namely, Kupiec test, Christoffersen test, Lopez test, Diebold and Mariano test, root mean squared error (RMSE), and h-stepahead forecasting RMSE. Findings - The results indicate that filtered EVT performs better in terms of capturing fat-tails in stock returns than parametric VaR models. An increase in the conditional quantile decreases h-step ahead number of exceptions and this shows that filtered EVT with higher conditional quantile such as 40 days should be used for forward looking forecasting. Originality/value - The research results show that emerging market stock return should be forecasted with filtered EVT and conditional quantile days lag length should also be estimated based on forecasting performance.