Forecasting Volatility in Pakistan Stock Exchange

Abstract

Analysis of time series is used to develop simple models which are able to forecast, interpret and analyze the result concerning its eld of application. The main purpose of this study was to check and model the volatility in Pakistan Stock Exchange (PSX) for the near future. The stock exchange data are highly volatile and there are many factors which a ect the daily market. For example, political stability or instability, dollar prices, Monday charisma, last working day of the week, etc. Thus, it becomes very di cult to forecast these returns due to continuous e ect of above and many other factors. Generally, in the presence of these factors, the naive or simple methods, like, Autoregressive (AR) and Moving Average (MA) and sometime Autoregressive integrated moving average (ARIMA) models are unable to give accurate forecasts as these models are unable to account for the volatility present in the series. In accordance with the main aim of this work, we model and forecast PSX 100 index series using ARIMA and other models, namely, Autoregressive conditional heteroscedastic (ARCH) and Generalized Autoregressive conditional heteroscedastic (GARCH), that are able to account for heteroscedasticity in the time series. The data used in this work were obtained from PSX for the period 1st January 2013 to 30th April 2018. The rst ve years are used for model estimation while the last four months are kept for one- day-ahead out-of-sample forecast using expending window technique. To assess the performance of each model, Mean absolute percentage error (MAPE) and Root Mean square error (RMSE) are used. The results suggested that although the di erences among the results from all models are not very huge, the GARCH model produces superior results in term of MAPE and RMSE.