A news event-driven approach for the historical value at risk method

Value at Risk (VaR) is a tool widely used in financial applications to assess portfolio risk. The historical stock return data used in calculating VaR may be sensitive to rare news events that occur during the sampled period and cause trend disruptions. Therefore, in this paper, we measure the effects of various news events on stock prices. Subsequently, we identify irregular events using a Poisson distribution, and we examine whether VaR accuracy can be improved by considering news events as an additional input in the calculation. Our experiments demonstrate that VaR predictions for rare event occurrences can be improved by removing the event-generated disturbance from the stock prices for a small, optimized time window.