Modelling and Forecasting of GDP in Bangladesh: An ARIMA Approach

Abstract

This paper aims to model and forecasting on GDP data of Bangladesh for the period of 1960 to 2017. To test the stationarity of the series graphical method, correlogram and unit root test were used. The time series plot of GDP shows a non-stationary pattern and overall this is like exponential curvature shape. Hence the data have been differenced twice to convert the data from non-stationary to stationary. From the autocorrelation function (ACF) and partial autocorrelation function (PACF) we obtain the order of the time series model. The chosen model was autoregressive integrated moving average ARIMA (1, 2, 1). The model has been fitted on data to estimate the parameters of autoregressive and moving average components of ARIMA (1, 2, 1) model. For residual diagnostics, correlogram, Q-statistic, histogram, and normality test were used. Also, Chow test was used for stability testing. Using model selection criterion and checking model adequacy, wesee that the model is suitable in shape. It is found that the forecast values of GDP in Bangladesh are steadily improving over the next thirteen years.