Forecasting accuracy of stochastic volatility, GARCH and EWMA models under different volatility scenarios

Abstract

The forecasting of the volatility of asset returns is a prerequisite for many risk management tasks in finance. The objective here is to identify the volatility scenarios that favour either Generalized Autoregressive Conditional Heteroscedasticity (GARCH) or Stochastic Volatility (SV) models. Scenarios are defined by the persistence of volatility (its robustness to shocks) and the volatility of volatility. A simulation experiment generates return series using both volatility models for a range of volatility scenarios representative of that observed in real assets. Forecasts are generated from SV, GARCH and Exponentially Weighted Moving Average (EWMA) volatility models. SV model forecasts are only noticeably more accurate than GARCH in scenarios with very high volatility of volatility and a stochastic volatility generating process. For scenarios with medium volatility of volatility, there is little penalty for using EWMA regardless of the volatility generating process. A set of return time series selected from FX rates, equity indices, equities and commodities is used to validate the simulation-based results. Broadly speaking, the real series come from the medium volatility of volatility scenarios where EWMA forecasts are reliably accurate. The robust structure of EWMA appears to contribute to its greater forecasting accuracy than more flexible GARCH model.

Forecasting VaR models under Different Volatility Processes and Distributions of Return Innovations

“…In addition to the above volatility models, we examine the out‐of‐sample performance of the exponentially weighted moving average (EWMA) volatility model often used in practice to calculate VaR by RiskMetrics. This model can be thought of as a special case of the GARCH model (see, for example, Ding and Meade, ).…”

Asset volatility forecasting:The optimal decay parameter in the EWMA model

“…Moreover, the decay parameter should be adjusted in function of the time horizon. Ding and Meade (2010) point that EWMA exhibits a greater forecasting accuracy to real data compared to GARCH and SV models. González-Rivera et al (2004) found EWMA brings good results in the option pricing context.…”

Machine Learning-based Anomaly Detection of Ganglia Monitoring Data in HEP Data Center

“…We take the last k (k = 6, 12, 18, 24, 30, 36) data points of time series x as time series Y (collections of time series) respectively. Fitting features are these values which are the difference between the last element of time series x and the smoothed values of each time series of Y after Moving Average Algorithm [13], Weighted Moving Average Algorithm [14], Exponential Moving Average Algorithm [15] and Double Exponential Moving Average Algorithm [16]. There are over 30 statistical features after time-series features extraction.…”