An Exceedance Probability of Financial Return and Its Application to the Risk Analysis

Abstract

This paper studies a new specification of the autoregressive binary choice model for estimating the exceedance probability of return and its application to the risk management tasks, especially for Value-at-Risk calculation. The author proposed a new parametrization of the volatility equation, which implies the presence of an additional random term. Such a model could not be estimated using the methods of classical statistics; therefore the Bayesian NUTS algorithm was chosen as an appropriate toolkit. Estimated exceedance probabilities were applied in calculating VaR. As a data set, it was taken the daily return of PAO «Sberbank» shares and the one-minute return of the USD-RUB currency pair. The results of VaR estimation were tested for asymptotic convergence to the true value by Engle and Manganelli’s dynamic quantile test.