THE INTERNATIONAL CONFERENCE ON RECENT TRENDS IN STATISTICS DATA ANALYTICS (ICRTSDA)

Abstract

The key objective of this study is to forecast inflation in Pakistan using an autoregressive integrated moving average with explanatory variables (ARIMAX) model. The studied data is obtained from World Development Indicator (WDI) during 1970 to 2020. This study used Ordinary Least Squares (OLS) method. Various diagnostic and selection criteria for the optimal model are preferred for forecasting of inflation rate in Pakistan. ARIMA (2, 0, 1) model is found suitable because its AIC, SC and HQC are least. Then including explanatory variables in ARIMA (2, 0, 1) model and make it as ARIMAX (2, 0, 1) model. Now multicollinearity test applied which outcomes reveals that there is no problem in it. Further, the residuals of ARIMAX (2, 0, 1) model is normally distributed, serially uncorrelated and free from hetroskedastic problem. The results of model showed that there is a smaller difference between forecasted and original values. Depending upon the lowest values of root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and Theil inequality coefficient ARIMAX(2, 0, 1) model is unique and sufficient for forecasting inflation rate in Pakistan depending upon the Theil’s inequality forecast accuracy measures.
Key Words: Pakistan, WDI, ADF, ARIMA, AIC, SC, HQC, ARIMAX, Forecasting.