Semiparametric Score-Driven Exponentially Weighted Moving Average Model

Measuring volatility is used in many important financial and economic models. This paper proposes a new semiparametric score driven exponentially weighted moving average (SP SD-EWMA) model to improve the efficiency of the parametric score driven exponentially weighted moving average (SD-EWMA) model when residuals are fat-tailed and possibly skewed. We experimented the efficiency performance of our model on S&P 500 sample daily data. Implementing in-sample and out-sample analysis it is shown that the proposed model outperforms the other competing models (SD-EWMA and Standard EWMA models) as it renders higher log-likelihood function and smaller Akaike Information Criterion as well as smaller prediction errors. To validate further the outperformance of the proposed model we checked the forecast error of one, two and three days-ahead of value-at-risk (VaR) measure over the out-of-sample period to ensure that proposed model renders the smallest rate of violations and close to the expected level. This paper highly recommends using the proposed model to reduce the statistical efficiency loss of the classical/score-driven parametric volatility models and that could help investors and asset managers better adjusting their trading strategies.