Message-ID: <24785341.1075863404905.JavaMail.evans@thyme> Date: Thu, 27 Sep 2001 08:54:00 -0700 (PDT) From: nelson.neale@enron.com To: j.kaminski@enron.com, vasant.shanbhogue@enron.com Subject: Forest Products Consultancy Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit X-From: Neale, Nelson X-To: Kaminski, Vince J , Shanbhogue, Vasant X-cc: X-bcc: X-Folder: \VKAMINS (Non-Privileged)\Kaminski, Vince J\Deleted Items X-Origin: Kaminski-V X-FileName: VKAMINS (Non-Privileged).pst I took a look at the forest products consulting proposal by the group from Texas A&M. It appears that they are proposing to develop a vector autoregression (VAR) model among the various Enron forest products to better understand pricing relationships. I am assuming that they would set up a VAR model that would include pricing for commodities (i.e., pulp (multiple), liner board, lumber, strand board, newsprint) as well as key supply and demand variables. They do put a cap on the number of time series to be investigated at 25. Given the number of Enron P&P commodities and recognizing that each may have a unique set of supply and demand variables, I can easily see the evaluation exceeding the cap. However, the supply and demand terms of a key underlying commodity (say pulp or lumber) may serve as a proxy for supply and demand in the other markets (although I doubt it). The proposal does emphasize the importance of stationary time series and the cointegration approach using an error correction model (ECM) or vector error correction model (see Bessler's CV publications). These remarks offer reassurance that we have probably been taking the right approach in our other modeling work (e.g., diesel fuel-heating oil cointegration, JCC-Brent crude cointegration, etc.). We haven't yet implemented VAR in our price modeling work, but will likely use this approach for the suite of plastics (e.g., polypropylene, propylene, ethylene, HDPE, etc.) as well as for price indexes within the tech sector (e.g., semiconductor equity index, pc equity index, etc.) For the record, we have been working with EIM fundamentals on the development of a newsprint price curve. Cointegration was considered for newsprint price, lumber prompt month futures, and pulp prompt month futures; however, the correlation analyses suggested that the difference relationships were very weak. We instead focused on developing price as a function of newsprint supply and demand terms (e.g., North American and Canadian production, producer inventory, US newsprint demand, total newspaper dailies inventory, consumer days of supply, etc.). In short, I don't have a problem with the TAMU approach or the scope of work. If EIM fundamentals does carry through with this relationship, it might be beneficial to offer some research input along the way, especially since we are somewhat familiar with the technical approach and the needs of the Enron business units. Nelson