Forecasting Monthly Orders for Southern Pine Lumber

Monthly orders for southern pine lumber over the interval from January 1964 to December 1971 were analyzed for the purpose of developing a short-range predictive model. Multiple linear regression was used as the predictive technique and a step down procedure was employed to determine the best regression equation. The independent variables retained in this equation, in the order of their apparent importance, were as follows: the monthly seasonal indices of orders; the price index of structural clay products, lagged one month; the price index of softwood plywood, lagged one month; the price index of Douglas-fir lumber, lagged one month; the price index of softwood plywood, lagged three months; the prime interest rate, lagged one month; the price index of. Douglas-fir lumber, lagged three months; the price index of softwood plywood, lagged two months; and the number of residential housing permits issued in the south, lagged one month. The coefficient of determination (R2) was .8704, indicating that 87 percent of the observed variation in orders was accounted for by the predictive equation. The standard error of the estimate was 30.06 which represented 5.25 percent of the mean response. The Durbin-Watson “d-statistic” was non-significant indicating no problems with serial auto-correlation. When the model was employed to make predictions of monthly orders for the first nine months of 1972, reasonably accurate predictions were obtained. The average forecast error was 5.96 percent.

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