Statistical Forecasting of Producer Price Index for Foods, Beverages and Tobaccos

Анотація

Food and beverage industry is the most significant industry for the Ukrainian economy: it accounts for nearly 1/4 in the total industrial sales, and nearly 1/5 in the total exports. This raises the importance of forecasting key indicators of the industry performance, with special emphasis on index of producer price for foods, beverages and tobaccos. The objective of the paper is to show methodological tools for statistical forecasting for stationary and non-stationary dynamic series with monthly or quarterly periodicity by use of linear regression methods, chain substitutions and seasonal coefficients, which can be realized in MS Excel. The best method is chosen by indexes of producer price for foods, beverages and tobaccos, computed for January 2007 - September 2016. Dynamic series are divided into stationary and non-stationary ones. A non-stationary dynamic series can consist of three main components: trend, seasonal component for monthly or quarterly data (cyclic component for yearly data), and irregular (random) component. Stationarity of a series is checked by Dickey - Fuller test. Each type of dynamic series uses its own methods of statistical forecasting: for a stationary series the method of chain substitution should be chosen, for a non-stationary series with a trend the linear regression will be the best, for a series with seasonal component the method of seasonal coefficients should be chosen, for a series with trend and seasonal component the method of seasonal coefficients should be used. It is found by Dickey - Fuller test that the series of index of producer price for foods, beverages and tobaccos is a stationary one, i. e. the one requiring the method of chain substitutions for forecasting. Yet, all the three methods were approbated for illustration purposes. Index of producer price for foods, beverages and tobaccos is estimated by the three methods of statistical forecasting for the period of September - December 2016. The computations confirm practicability of the chain substitution method for forecasting stationary dynamic series. It is found that the most significant measure for assessing quality of the forecasts is correlation coefficient.

Класифікація

Схожі роботи