Forecasting Wood Products Consumption

Forecasts of wood consumption have long been used as an aid in production and marketing planning when there is a significant time lapse between the planning period and the anticipated event, or when the production-distribution process creates a time lag. Forecasting is distinguished from estimating on the basis of whether the estimated future values of the explanatory variables are, or are not, within the range of these values as used in developing the projection model. If the values in the target period lie outside the range, the projection is a forecast, if they lie within, the projection is an estimate. The complexity of the model used to make estimates or forecasts of wood consumption depends on the objective of the estimator, the extent of his knowledge of the relationships he is trying to account for, and the availability of the necessary data. If the objective is only to make an estimate of the value of one variable from change in another, and there is no need to understand underlying structure, a satisfactory projection can be made using free hand curves. Or, the relationship between the variable to be projected and those variables which are believed to influence it can be described mathematically in equation form. However, if it is desired to learn more about the true relationship believed to underlie an observed phenomenon, as well as to forecast or estimate, then it is necessary to specify the relevant variables to be included, the expected signs and perhaps the general magnitudes of the expected coefficients in advance of the computations. Forecasting models containing interdependent variables usually require more than one equation to be solved simultaneously or recursively as the case may be. Testing the reliability of regression and correlation statistics, and hence the forecasts themselves is theoretically less meaningful where the values of the independent or dependent variables are selected in other than a random manner from an unspecified population. Regression and correlation results are also affected if the data are subject to error, or if some variables which should have been included are left out of the calculations.

You must be logged in to download any documents. Please login (login accounts are free) or learn how to Become a Member