Volatility Forecasting Performance of Smooth Transition Exponential Smoothing Method: Evidence from Mutual Fund Indices in Malaysia

Abstract

This paper aims to empirically compare the performance of the smooth transition exponential smoothing (STES) method against the well-known generalized autoregressive conditional heteroskedasticity (GARCH) model in one-step-ahead volatility forecasting. While the GARCH model captured most of the stylized facts of the financial time series, threats of outliers in the leptokurtic distributed series remain unresolved. The study compared volatility forecasting performance of a total of 22 models and methods comprising STES, GARCH, and some ad-hoc forecasting. The daily returns of seven mutual fund indices (derived from 57 individual equity mutual funds) under two different economic conditions (sub-periods) were applied across all competing models. Findings revealed that the STES method with error and absolute error as transition variables emerged as the best post-sample volatility forecasting model in both sub-periods with and without financial crisis impact, as verified by model confidence set (MCS) procedure. The implications based on the results are: (1) both the sign and size of yesterday’s news shock have an impact on today’s volatility; (2) the STES method is resilient to outliers, and hence superior to GARCH and other volatility forecasting approaches examined. This study contributes an empirical approach in forecasting the risk of mutual funds investment for investors and fund managers, as well as extending the scope of volatility forecasting literature into the less explored mutual funds.

A Study on Assessing the Forecasting Performance of ARIMA and Regime Switching Model for AUM of Various Types of Mutual Funds in India

“…Hota, Pati, & Satapathy (2021) in their paper tried to forecast the NAV value of two mutual funds 1 day and 5 days ahead by using FLANN-FA model and prediction accuracy has been checked by using RMSE and MAPE evaluation measures and concluded that the proposed model outperforms FLANN model. Kin et al (2021) in their paper tried to compare GARCH and Smooth Transition Exponential Smoothing Method by taking the daily returns of 7 mutual fund indices under two different economic conditions and found that the Smooth Transition Exponential Smoothing Method emerged as the best model.…”

Forecasting the High-Frequency Exchange Rate Volatility with Smooth Transition Exponential Smoothing

“…Liu et al (2020) used STES in forecasting eight main stock indices has also demonstrated that the STES method performed well, regardless of the magnitude of the outliers, compared to other standard methods such as the ES (Exponential Smoothing) and GARCH methods. A recent work by Wan et al (2021) applying STES methods on Malaysia Mutual Fund Indices has showed that the STES method outperformed GARCH model.…”

“…The robustness of the STES method has been tested in stock markets (Taylor, 2004a;2004b;Ung et al, 2014;Liu et al, 2020), Malaysian real estate stocks (Gooi et al, 2018) and Malaysia Mutual Fund Indices (Wan et al, 2021). This present study extends previous research (where Taylor (2004a) used realised weekly volatility as the estimator and proxy while Liu et al (2020), Gooi et al (2018) and Wan et al (2021) focused on daily returns as the estimator) by applying the STES methods on the exchange rate series using daily RV, weekly RV, and monthly RV. The hourly returns are used to construct the RV as other higher frequency data are not available.…”