Venus Khim-Sen Liew, Kian-Ping Lim & Chong-Yi Lai. 2004. Predictability of the KLCI Price Movement: Evidence from the Time Series Models. INTI Journal, Vol. 1, No. 4, 239 – 248. Malaysia: INTI College.
Abstract: This study utilizes autoregressive integrated moving average (ARIMA) time series models to predict the price movement of Kuala Lumpur Composite Index (KLCI). ARIMA-GARCH models, which are ARIMA time series models with GARCH errors (relaxing the normality assumption), are also considered. All fitted models excluding those that exhibit nonstationary autoregressive roots are utilized to generate out-of-sample forecast over the forecast horizons of 1 day, 1 week, 1 month, 3 months, 6 months, 9 months and 1 year. Our forecasting evaluation exercise is based on the commonly use mean absolute percentage error (MAPE). Our results show that all the 17 forecasting models produce MAPE of less than 10% for forecast horizons of 3 months or shorter. For forecast horizons of 6 months and above, the forecast errors ranging from at most 7% in the case of autoregressive integrated (ARI) models to 16% in the case of integrated moving average (IMA) models. The best performing models are the ARIMA (11, 1, 0) models, with and without GARCH errors, which dominate over the others for 5 out of 7 forecast horizons. The results of study also show that ARIMA (1, 1, 0) model is the next best alternative model as far as parsimonious is concerned. To sum, this study demonstrates that securities market forecasters may resort to time series approach, which is less resource demanding, for satisfactory short-run forecasts of stock prices.
