Modeling and Analysis of Demand for Commodities and a Case Study of the Petrochemical Market

This thesis aims to establish a demand model for commodities that takes all crucial influencing factors into account. \nTo begin with, we analyze the dependency of the demand on prices, market parameters, and specific characteristics of the customers in order to provide the mathematical framework for a general demand model. In particular, this approach takes account of effects that are caused by price-based substitution of products irrespective of their availability. \nFundamental market models that include supply-demand interactions gain importance in the context of commodity pricing. We explicitly develop demand models for petrochemical products that are applicable within the profit maximization problem of a monopoly in order to determine optimal price and sales decisions. The solvability of the market optimization problem requires additional restrictions on the modeling. \nBasically, the model displays the nonlinear demand-price relationship. Model extensions incorporate the changes of macroeconomic indices, which quantify changes in the economic situation. Moreover, our approach to modeling demand comprises the impacts of varying prices of substitutable and complementary products, and establishes a connection to the characteristics of the consumer's side. \nIntegrating the demand models to real market models necessitates the identification of the demand parameters. We discuss the difficulty to get reliable parameter estimates in the situation of incomplete data and investigate two methods based on additional assumptions in order to estimate the demand parameters. For the demand model that includes the dependency on the price and macroeconomic indices, a heuristic methodology firstly determines parameters based on market simulations. The second approach creates an inequality constrained parameter identification problem, where the constraints reflect additional assumptions on the shape of the demand model function. This problem can be solved using the generalized Gauss-Newton method. \n