About the lectures
Mathematical modeling of choice behavior: from theory to practice
Abstract: Choice behavior is relevant in many fields of applications, including marketing, transportation, health and energy. The understanding and predicting the behavior is of primary importance to evaluate the impact of new policies. In this short course, we will show how to derive mathematical models of choice behavior which are operational, and illustrate their use based on some case studies in transportation.
Inaugural lecture: "Choice modeling: an introduction"
The first lecture will be non technical. The first part will consist in an overall introduction to the underlying theory, as well as a description of the data needed to develop operational models. Then, four concrete projects will be presented, in the context of marketing and transportation. The results that will be presented will illustrate the high practical impact of the methodology.
Lecture 1: "Choice theory"
In this lecture, we will elaborate on the theory of choice that will lead us to operational models. Starting from standard concepts in micro-economics, we will show how demand analysis can be performed using choice models. The key idea will be to extend the classical theory of utility derived for continuous goods to the case of discrete goods. This theory was proposed by Daniel McFadden, who received the Nobel prize in Economics in 2000.
Lecture 2: "Binary choice"
The second lecture will pave the way from theory to practice. In the simple context of the choice between two alternatives, an operational model, based on well defined hypothesis, will be specified. This model is called the "binary logit model".
Lecture 3: "Choice among many alternatives"
The third lecture will generalize the results presented in the second one for cases with more than two alternatives. The emphasis will be on the specification of the model, and how concrete behavioral assumptions can be translated into mathematical formulations.
Lecture 4: "Forecasting"
The last lecture will describe how the models are applied to forecast behavior, to forecast market shares, to calculate willingness to pay, as well as some other indicators.
