Univariate modeling and forecasting of economic time series

Univariate Box-Jenkins time-series (ARIMA models), were used for modeling and forecasting the indices of Per Capita Gross Domestic Product (PCGDP) and Gross Domestic Product (GDP) of PR of China over the past five decades. This article shows the relative forecasting accuracy of fitted ARIMA model in comparison to the more traditional time-series methods: winter's exponential smoothing, simple exponential smoothing, Brown's linear exponential smoothing, and linear trend. The optimum empirical ARIMA models obtained for each economic time-series, have a satisfactory degree of statistical validity (low approximation errors). The root mean square error of the optimum fitted ARIMA models of PCGDP and GDP are 5.67 and 5.70, respectively, which are. lowest as compared to other methods, and the mean percent error are 0.001 and 0.01 %, respectively, which are approximately zero. Indices of PCGDP and GDP are estimated to be 106.99 and 108.54, respectively, in the year 2001 by fitted models.