Lab 4: Serial Correlation
The file timber.wf1 has annual data on the determinants of harvest of timber in
Oregon during the years 1959-1989.
The variables (compiled by Anthony Hazen) are defined as follows:
HARVEST = Total Softwood Timber harvested from Oregon in a given year.
Data are in billion board feet. (Range 5.1212 - 8.743)
EXPORTS = Volume of timber exports to foreign destinations measured
in 100 million board feet. (1.469 - 13.874)
HOUSTART = Total housing starts in the U.S, in millions.
(Range 1.072 - 2.379)
INDPROD = Index of Industrial Production for paper and wood products.
(Range 4.75 - 15.57)
TIMBPRIC = Stumpage prices for the Pacific Northwest. Prices are measured in
dollars per 1000 board feet. (Range 2.48 - 43.22)
PRODPRIC = Producer Price Index for all commodities. (Range 3.16 - 10.87)
1. Estimate a log-log model of the supply of Oregon timber.
2. Generate a model using the Hendry method
3. When you have your final model, test whether each of the elasticities is
equal to 1 (ignore the sign of the coefficient).
4. Test your model for 1st order Serial Correlation using the LM test.
5. Try to eliminate the serial correlation by using:
a. The Newey-West correction
b. A lag of the dependent variable
c. An AR model