Ricardo Daziano (Cornell) - Statistical Inference on Consumers’ Preferences and Willingness-to-Pay
Friday, November 9, 2012 at 3:30pm
Frank H. T. Rhodes Hall, 253
Understanding individual choice behavior is critical for several disciplines that need to account for demand dynamics. Microeconometric discrete choice models represent the cognitive process of economic decisions, are based on the theory of random utility maximization, and allow researchers to better describe consumers’ preferences and willingness-to-pay. Discrete choice models are widely used in applied economics, marketing, urban planning, and in some fields of civil engineering including transportation analysis. Discrete choice analysis is used to forecast demand under differing pricing and marketing strategies and to determine how much consumers are willing to pay for qualitative improvements. In transportation engineering, these models allow researchers, firms, and policy-makers to predict demand for new alternatives and infrastructure (e.g. a light rail or a new highway), to analyze the market impact of firm decisions (e.g. merge of two airline companies), to set pricing strategies (e.g. road pricing, toll definition, revenue management), to prioritize research and development decisions (e.g. ultra low emission vehicles) as well as to perform cost-benefit analyses of transportation projects (e.g. building a new bridge).
To model uncertainty in the determination of demand, discrete choice models are intrinsically probabilistic. Marginal utilities representing consumers’ preferences and tastes are unknown parameters of a statistical model. To estimate these parameters different tools can be applied, including Bayesian statistics, simulation-aided inference, nonlinear optimization, and advanced numerical methods. Using an empirical application of forecasting market shares of different automotive technologies under several competitive scenarios, in this talk we will overview the fundamentals of discrete choice models and the complexities associated with the estimation problem.