Author(s): Stanley Onyedikachi Cynthia, Etuk Ette Harrison and Peters Nwagor

Abstract: This study looked into the supremacy of GARCH modelling over ARIMA modelling of Nigerian export commodity price index series. Data was collected from the Central Bank database from January 2000 to December 2020 and was analyzed using EVIEWS statistical software. The time plot of Nigerian export commodity price index (figure 1) shows that the series changes over time indicating that there is clearly a secular trend in it. By the Augmented Dickey-Fuller Test in Table 1, the series is non-stationary with p = 0.8989 > 0.05. Therefore there was need for differencing (see figure 2). After the series was differenced, a stationarity test was ran for the differenced series. The Augmented Dickey Fuller test of Table 2 shows that the differenced series is stationary with p = 0.0000 < 0.05, indicating that the trend has been removed through differencing. The EVIEWS statistical software was used in estimating the model. The correlogram of Figure 3 has spikes at lag 1 for the Autocorrelation function as well as for the partial autocorrelation function suggesting an ARMA (1,1). Then ARMA (1,1) was estimated for the differenced series (table 3) which showed that the series does not follow an ARMA(1,1) as suspected, the coefficients of the parameters are both non-significant statistically. After the failure of the ARIMA(1,1,1) we fitted the ARIMA(1,1,0) and ARIMA(0,1,1) to the original data. ARIMA(1, 1, 0), ARIMA (0,1, 1) and GARCH (1,1) having met the required assumptions as the appropriate model was used to estimate its parameters. From the result obtained, it was reveal that by the AIC, SC and HQ terms, the GARCH (1,1) model when compared to other models has proven to be adequate model for the modelling of differenced Nigerian Export Commodity Price Index series.

Pages: 11-21 | Views: 292 | Downloads: 20

Download Full Article: Click Here