THE AUTHOR 



MICHAEL J. NICCOLUCCI is an economist with the 

 Economics Research Work Unit, Intermountain Research 

 Station, Forestry Sciences Laboratory, Missoula, MT. He 

 received academic training in economics at the University 

 of Montana, where he earned his M.A. His research in- 

 cludes timber sale design and appraisal. 



RESEARCH SUMMARY 



Timber sale planning is a complex, expensive process. 

 Developing a sale to the point where it is ready for auction 

 requires the efforts of many natural resource specialists, 

 many hours, and many dollars. The common expectation 

 is for the sale to sell at initial auction. But many timber 

 sales receive no bids, meaning they do not sell at their 

 initial offering. Given the large investment involved, the 

 occurrence of unsold sales is not desirable. Unsold sales 

 also raise the question of organizational competence. 



Knowing the likely outcome of a timber sale offering, 

 particularly in early design stage, is important to the man- 

 ager. This information can be used to modify the timber 

 sale thereby increasing its likelihood of selling. The 

 research reported here developed and compared two 

 approaches to statistical classification, intended to predict 

 salability at various points in the sale planning process. 

 Classification results were statistically compared based on 

 geographical zone models, models at various points in the 

 timber sale planning process, and the classification meth- 

 ods used. 



Data used in this study came from a sample of 389 sold 

 and unsold timber sales in the Northern Region of the 

 Forest Service, U.S. Department of Agriculture. The region 

 was further divided into two geographical zones — east and 

 west of the Continental Divide. Discriminant analysis and 

 logistic regression were used to develop statistical equa- 

 tions to classify timber sales into groups of sold and unsold. 

 Equations were designed to be used at three points in the 

 "Gates" timber sale planning process, all before the actual 



bidding. The quantity and quality of information increases 

 as the sale proceeds from one gate to another. 



The accuracy of the equations increased as the timber sale 

 progressed through the Gates process. Equations at the first 

 gate, approximately 7 to 10 years before the auction date, 

 correctly classified about 65 percent of the sales, based on 

 only general sale characteristics. The equations for the next 

 gate, 1 to 3 years before the auction, correctly classified 74 

 percent of the sales. Equations at the final gate before the 

 auction correctly classified about 84 percent of the sales. 



Statistical analyses were conducted to test for differences 

 in classification success between geographical zones within 

 the Northern Region, between statistical modeling techniques, 

 and between phases in the timber sale planning process. 

 Statistically significant results were found between geographi- 

 cal zones and sale phases, but no statistical difference in 

 classification success could be found between statistical 

 modeling techniques. 



CONTENTS 



Page 



Introduction 1 



Methods 1 



Timber Sale Preparation — the Gates Process 1 



Classification Methods 2 



Study Design 3 



Analytical Procedures 3 



Results and Discussion 4 



Gate 1 5 



Gates 2 and 3 6 



Gate 4 7 



Statistical Evaluation 8 



Classification Results and Cutoff Points 9 



Management Implications 9 



References 10 



Appendix A: Cutoff Points 1 1 



Appendix B: Discriminant Analysis Classification 

 Equations 13 



The use of trade or firm names in this publication is for reader information 

 and does not imply endorsement by the U.S. Department of Agriculture of 

 any product or service. 



Intermountain Research Station 

 324 25th Street 

 Ogden, UT 84401 



