Abstract:
Autoregressive Moving Averages {ARIMA (p,d,q)} model consist of autoregressive terms (AR), integrated terms (I) and the Moving averages terms (MA). According to Box and Jenkins, time series can be used to identify previous random errors in a time series which interact to create a pattern hence may be used for future forecasting. This can be identified using autocorrelation and partial autocorrelation graphs and spikes of time series. The number of lags required for data to be stationary I (d) can be identified by differencing of time series and if the data becomes stationary after differencing, then the I(d) terms exist. Similarly, when the autocorrelation and partial autocorrelation graphs are none zero and decays gradually, the AR terms and the MA terms exist hence the model can be used for forecasting the future. ARIMA (p, d, q) guided this study by identifying the model which is suitable for forecasting dividend per share. Furthermore, the model helped in data transformation of dividend per share into stationary. The current study therefore sought to model dividend per share using Autoregressive Integrated Moving Average Model. Reviewed studies focused on modeling variables using individual firm data while the present study modeled dividend per share using industry or macro- level quarterly data. Therefore, the nature of forecasting models when data at macro-level is used was not known; the modeling is therefore imperative. The study revealed that ARIMA (P,d, q) model can be used to predict dividend per share; the identified ARIMA (p, d, q) model is ARIMA (1,1,1). The study concludes that investors, financial analysts and managers should use ARIMA (1, 1,1) to forecast dividend per share. The study is important to policy makers, regulators, investors and scholars.
|