Modelling and Forecasting Long Memory and Volatility in Nigeria’s Consumer Price Index Using an Arfima-Figarch Approach
Auwalu Muhammad Sabo
Tasi’u Musa
Akeem Adepoju
Huzaifa Abdurrahman
Abstract
Commodity prices in Nigeria exhibit long memory characteristics, which lead to high risk of price fluctuations. This study aims to model and forecast the impact of long memory on cereal prices index in Nigeria using a hybrid time series model. The data used for this study are secondary monthly CPI data obtained from the Central Bank of Nigeria (CBN) covering the period 1990-2025. The study employed Kwiat-Kowski Phillip Smidth Shin test to check for stationarity and found that the variables were stationary after taken fractional differencing. The study employed GPH test to check for long memory in the variables and it was found in the variables. The study utilized scatter plot to check for heteroscedasticity and it was found in the residuals of ARFIMA (1, 1.2, 1) models. In addition, ARFIMA (1, 1.2, 1)-FIGARCH (9, 1) was found as the best model with least AIC, MAE, MSE, and RMSE when compared with standalone models. The study employed Ljung-Box and ARCH-LM tests to diagnose the models. Hence, there is no excess correlation in the residuals of best models. The forecast results shows that the forecasted volatility of the CPI variable increases over time, this indicates the rising prices may impact food affordability and accessibility in the nation. Moreover, Government may need to intervene to stabilize prices and ensure food security.
References