Global Journal of Human Social Science, E: Economics, Volume 23 Issue 2

Volume XXIII Issue II Version I 57 Global Journal of Human Social Science - Year 2023 ( )E © 2023 Global Journals Impact of COVID-19 on Stock Market Volatility and Forecast using ARIMA and EGARCH tentative ARIMA models (1,1,1), ARIMA (1,1,4), ARIMA (4,1,1) and ARIMA (4,1,4). It is advised to choose a model that is parsimonious as it gives a better forecast than an overidentified model. Models with the smallest number of parameters to be estimated are usuallyparsimonious. From Table 2, between the contest ARIMA (1,1,1) and ARIMA (4,1,1) which all have 2 significant coefficients, ARIMA (4,1,1) is ideal for the study since it has the lowest volatility, highest adjusted R-square, and lowest AIC and SBIC. Table 2: Determination of Appropriate ARIMA Model Differenced GSECI ARIMA (1,1,1) ARIMA (4,1,1) ARIMA (1,1,4) ARIMA (4,1,4) Significant coefficient 2 3 3 3 Sigma 2 (volatility) 619.0792 599.9418 599.9776 651.5751 Adj R 2 0.080375 0.108803 0.108750 0.032103 AIC 9.275762 9.244577 9.244561 9.327075 SBIC 9.298344 9.267160 9.267144 9.349658 c) ARIMA Model Estimate The final model has been determined as ARIMA (4,1,1) and it is presented in Table 3. Table 3: ARIMA (4,1,1) estimates of the Ghana Stock Exchange Composite Index. Variable Coefficient Std. Error t-Statistic Constant -0.161561 0.976326 -0.165478 AR (4) 0.212911*** 0.014475 14.70849 MA (1) -0.287736*** 0.007387 -38.94937 SIGMASQ 599.9418*** 5.705647 105.1488 Note: *, ** and *** represent 10%, 5% and 1% level of significance respectively d) Forecast analysis of GSE-CI The ARIMA (4,1,1) model is used to forecast the closing price of the Ghana stock exchange composite index from 1 st March, 2021 to 26 th February, 2022. The forecast of the outer year fromMarch 2021 to February 2022 shows a upwards trend of the GSECI. 1,600 2,000 2,400 2,800 3,200 3,600 17 18 19 20 21 22 GSECI Forecast GSECI Figure 3: Forecast of GSECI showing Actual and Forecast The correlogram after estimation of the ARIMA (4,1,1) model indicates there is no information uncaptured since all the residuals are barely flat and do not lie above the standard error bound. Thus, all the lag structures should lie within the 95% confidence intervals or the standard error bounds. Since all the residuals lie