5. DATA SOURCES, METHODOLOGY AND CONSTRUCTION OF VARIABLES5.1. Calculating Return There are two ways to calculate stock returns5.1.1. Rolling Return This is the percentage return that an investor would get if he bought the shares at the end of a particular day/month t-1 and sold them at the end of the next day/month. For day t and stock A, the daily return R At is defined as R At = { In (P At/PA, t-1)}*100 The stock paid a dividend on day t, the total return would be R At = {In ( PA +Divt /PA, t-1)}*1005.1.2. Discrete Return An alternative method for calculating stock returns is defined as R At = {(P At/PA, t-1)-1}*1005.1.3. Continuous Compound Returns and Discrete Returns Using the continuous compound rate of return, it is assumed that Pt = Pt-1 ert where rt is the rate of return during period (t-1,t) and where Pt is the price at time t. If r1, r2,….,r12 are the returns of 12 months, then the price of the security at the end of the 12 months will be P12 = P0 and r1 +r2 +….+r12This representation of prices and returns allows us to assume the i average daily or monthly returns are r = (r1, r2,….,r12)/ 12. Since we can assume that the return data for the 12 months represents the distribution of returns for the following month, it follows that the continuously compounded returns return is the appropriate return measure and not discretely compounded return. (Benninga, 2008)5.2. RETURN PREDICTABILITY TEST In this research study, the methodology consists of four sections based on a set of performance predictability information. The information set can be defined as the past history of stock prices, time patterns, market and business characteristics. The first section consists of the predictability of short-term returns based on the past hi...... middle of paper..... .r ARIMA model, the next step is to check whether the selected model is appropriate. A test of the chosen model is to see if the residuals estimated by the model are white noise, so the chosen model fits the data reasonably well. The Box and Jenkins(BJ) methodology suggests some diagnostic checks to determine whether an estimated model is statistically appropriate or not; if "Yes", go to the last stage, which is prediction; if “No”, go back to the first step and repeat the procedure from parameter identification and estimation to diagnostic check until a good final model is obtained.5.2.2.2.1. D) Phase 4: ForecastingOne reason for the widely accepted ARIMA models is their success in forecasting. Forecasts made with this method are more authentic than those made with other econometric models, particularly for short-term forecasts (Gujarati, 2003).