(531b) Oil and Gas Cost Basis Model for Purchased Raw Materials

Kalos, A. - Presenter, The Dow Chemical Company
Rey, T. - Presenter, The Dow Chemical Company

Historically, about two dozen raw materials that The Dow Chemical Company purchases account for about a third of the overall raw material costs. Since these raw materials are derived primarily from oil and/or gas (at 85%+), their costs are highly correlated to oil and gas prices. As a result, it is desired to understand the sensitivity of the raw materials costs to fluctuations of the prices of oil and gas. The importance of this varies seasonally and it becomes paramount when prices begin to exceed $60 pbbl.

In this paper, we describe a methodology and the development of mathematical models for assessing the impact of changes in the price of oil and gas on the costs of raw materials that are primarily derived from oil and/or gas. This methodology relies on the use of published stochiometric unit ratios from the chemical synthesis routes of commercial processes that are actually used to produce these raw materials. The stochiometric network for each raw material takes into account alternate routes (e.g., oil vs. gas), which the manufacturers may choose depending on availability, the price of oil and gas and other considerations at particular points in time. In addition, the costs of the remaining raw materials are estimated on the basis of empirical correlation models, with respect to the oil/gas-based raw materials. In this way, a comprehensive estimate of the total purchased raw materials costs can be made.

Although the focus of this work was not on developing forecasting models (rather, we focused on developing descriptive models), we demonstrated that it is possible to forecast raw material costs, as long as good forecasts are provided for the demand levels of the raw materials themselves, as well as forecasts of the prices of oil and gas. For this, we developed both linear (Box-Jenkins type) as well as non-linear (neural network-based) multivariate time series models. These were then used to generate forecasts for the demand levels of the oil/gas-based raw materials, the rest of the raw materials, and oil & gas prices.

After developing the models, we forecasted the total purchased raw material costs 34 days in advance of the close of two consecutive quarters; the forecasts were within 6.4 % of the actual costs at the end of the quarter. Given the uncertainties in the input data, these results are adequate for planning raw material purchases. All data used in the paper (both amounts and costs) are indexed, in order to protect proprietary information.