75 Relative Forecasting of Aggregate Inventory to Sales Ratios

This study is the furtherance of forecasting economic and financial variables with the aid of nonlinear neural network models. This work investigates the relative predictability of changes in real aggregate manufacturing and trade inventory to sales ratio using nominal 3-month Treasury bills interest rate. The forecasting of changes in real inventory to sales ratio was done using 1-step and 4-step ahead prediction regimes with nonlinear neural network and linear regression models. The efficacy of the relative forecasts was determined with performance statistics correlation, root mean square error, and Theil inequality coefficient. As expected, the neural network models performed better than the regression models over the tested period of inventory to sales ratio data series. The manufacturing and trade inventory to sales ratio is a lagging economic index of business cycle indicators. It provides information on the state of goods in the economy relative to their current rate of sales.