Date of Award

Spring 1998

Project Type


Program or Major

Natural Resources

Degree Name

Doctor of Philosophy

First Advisor

Theodore E Howard


Timber price forecasts are important components of timberland investment analysis. Econometric models used in forecasting timber prices can be complex because demand for timber is derived through demand for other products such as paper and housing. In contrast to econometric methods, time series analysis or autoregressive techniques allow price forecasts to be made from the timber price series themselves.

A necessary condition for using time series techniques is that the timber price series be stationary or mean-reverting. The primary hypothesis in this study was that timber prices would not be stationary and time series techniques could not be applied to their analysis. A secondary hypothesis was that shocks to timber prices occur so frequently that prices would not have a chance to revert to a mean and so statistical tests would not show that timber prices are stationary.

Four commonly used tests for stationarity were applied to eleven timber price series to test the primary hypothesis and a list of timber shocks was developed to test the secondary hypothesis. The stationarity tests indicated that all of the price series were either first or second difference stationary. However, the stationarity tests tested only for a constant mean, and not for a constant variance. Charts of all the price series indicate that the variability of each series has changed over time. Since a constant variance is a required condition of stationarity, this result suggests the primary hypothesis should not be rejected.

Operations control chart techniques were used to analyze the changing variances and to determine if recent subsets with fixed mean and variance existed. All eleven price series were first difference stationary over some recent subset of years. The primary hypothesis can be rejected for the most recent subsets of all the price series tested.

The subsets of the price series suggest the existence of breakpoints in the series. It was then hypothesized that breakpoints common to several price series might indicate timber price shocks. Breakpoints were selected on an a priori basis by studying the behavior of the level price series. Sharp changes in direction or volatility were chosen for testing with Chow's breakpoint test.

Common breakpoints were compared to the list of possible shocks developed to test the original secondary hypothesis. The breakpoints did not correspond well to any shocks. This points to a limitation in using time series analysis: changes in the underlying process producing the price series can be identified as having occurred, but it is not possible to determine the cause of that change. Econometric techniques might be useful in identifying the causes.