Exploring the decay parameter for the exponentially weighted moving average volatility methodology

Abstract

Volatility estimation is a crucial task for financial institutions, as it affects various aspects of their operations, such as risk management, capital allocation, investment strategy and derivative valuation. However, the traditional method of using equally weighted moving averages to estimate volatility can be inaccurate and incorrectly used, especially in volatile market conditions. It yields financial losses for financial institutions in that the volatility estimates do not correctly reflect financial markets in real time. In this dissertation, we implement the exponentially weighted moving average model instead, which assigns more weight to recent data than older data. We explore how the choice of the decay factor λ influences the performance of the exponentially weighted moving average model in different market scenarios. The optimal value of λ varies depending on the market volatility. We therefore demonstrate that the model can provide more reliable and timely volatility estimates than the equally weighted moving average model. These are useful for different applications in financial, such as Value at Risk or Expected Shortfall.