Modeling and forecasting pine sawtimber stumpage prices in the US South by various time series models

Abstract

Among the three timberland return drivers (biological growth, timber price, and land price), timber price remains the most unpredictable. It affects not only periodic dividends from timber sales but also timber production strategies embedded in timberland management. Using various time series techniques, this study aimed to model and forecast real pine sawtimber stumpage prices in 12 southern timber regions in the United States. Under the discrete-time framework, the univariate autoregressive integrated moving average model was established as a benchmark, whereas other multivariate time series methods were applied in comparison. Under the continuous-time framework, both the geometric Brownian motion and the Ornstein–Uhlenbeck process were fitted. The results revealed that (i) the vector autoregressive model forecasted more accurately in the 1-year period by the mean absolute percentage error criterion, (ii) seven out of the 12 southern timber regions played dominant roles in the long-run equilibrium, and (iii) conditional variances and covariances from the bivariate generalized autoregressive conditional heteroscedasticity model well captured market risks.