Doctor of Philosophy (PhD)
Renewable Natural Resources
The development of cointegration theories and the presence of nonstationarity in time series raised serious concerns about possible spurious estimations in forest products models. Based on the results of Hsiao (1997a, 1997b), all the virtues of two-stage least square (2SLS) hold if there are sufficient cointegration relations. Stationary null and nonstationary null unit root tests and monthly seasonal unit root tests were applied to the time series used in this dissertation. Cointegration tests with exogenous variables were performed to justify the 2SLS. A regional error correction model (ECM) with four regional lumber supply and demand equations and a U.S.-Canada supply and demand ECM were estimated. CUSUM tests did not find any structural changes. Both estimated models showed that the imported Canadian lumber and the U.S. lumber are substitutes. The estimated long-run and short-run own-price elasticities for demand and supply are inelastic for all the equations but the short-run supply equation for the West Coast. The long-run lumber supply equations have significant trends: annually -3% for the Inland West and 2% for the other regions. The popular maximum likelihood estimation for the restricted ECM cannot pass the test for the restrictions and is, therefore, not used for the regional structural lumber model. A series of univariate and multi-equation models were used as forecasting models. A combination of univariate model were shown to be the best forecasting models for lumber prices, and a combination of univariate and multi-equation models were shown to be the best forecasting models for lumber quantities. The selected combinations of models were shown to be the best with additional observations. It was also shown that lumber quantities could be forecasted better than lumber prices.
Document Availability at the Time of Submission
Release the entire work immediately for access worldwide.
Song, Nianfu, "Structural and forecasting softwood lumber models with a time series approach" (2006). LSU Doctoral Dissertations. 3049.
Sun Joseph Chang