DATES: MAY 30 - JUNE 3, 2006
LOCATION: AMERICAN UNIVERSITY
Objectives and Scope
Discrete choice models can now be found in practice in all the social sciences, in medical research, transport research, the physical sciences, and throughout the research environment where the behavior of individual entities and decision makers is analyzed empirically. The purpose of this class is to provide the background for understanding both the underlying theory and the empirical tools for practical application of discrete choice modeling in a variety of economic/econometric estimation problems. We will survey the broad scope of discrete choice models.
The model building aspect of the course will begin with the rudiments of multiple regression, then quickly proceed to the fundamental building block, the canonical model of binary choice. From there we will proceed to many extensions including the multinomial logit model (MNL) for choice among multiple alternatives. Variants on the MNL will include recent advances in mixed logit (random parameters), nested logit, and so on, for models that are designed to accommodate a wide variety of behavior patterns. Applications will be drawn from economics, marketing and transport research.
We will also examine basic and advanced models for binary choice, including models for sample selection, censoring, and so on. Applications here are drawn from economics and health economics. A third set of results is built around observations on count data, such as frequencies of bankruptcies, accidents, health care utilizations, recreation site usage, etc. Here, we will visit a large variety of specifications, particularly those designed to handle stratification, censoring, truncation and selection.
Recent work in many fields has focused on panel data models. We will examine many different extensions of the discrete choice models to panel data. The course will examine familiar, basic methods and frontier developments in the field.
In addition to the theoretical, model developments, students will spend a large amount of the course time and effort on received studies from the literature and, most importantly, on hands on, empirical applications using ‘live’ data sets. Students will be introduced to NLOGIT, one of the most widely used packages for discrete choice models. Part of the course work will consist of practical applications and development of a model to be discussed on the last day of class.
Preliminary time schedule
Each day will consist of four sessions, two morning and two afternoon. Within each interval, we will spend one of the two sessions on discussing the models and empirical applications. A second session, one in the morning and one in the afternoon, will be spent in the computer lab, where students will learn how to use NLOGIT and will estimate models and do empirical analysis of ‘real’ data sets.
Content and Topics
Below is a tentative topical outline. The order of topics may change. Depending on availability of time, and interest, other topics may be included and some topics may not be covered. In each of these sections, we will consider underlying theory, specific models, estimation, and received applications.
Theory and Model Development
1. Basic Building Blocks
Interpretation: Marginal effects
The analysis of binary choices
Extensions: Panel data and heterogeneity
2. Models for Discrete Choice Among Multiple Alternatives
Underlying theory: The random utility model
Multinomial logit models for multinomial choice
Analysis: Marginal effects
Fit and prediction
Extensions of the multinomial logit model
Nested logit models
Heteroscedasticity and heterogeneity
Mixed logit models and random parameters models
Kernel logit models
The multinomial probit model
Specification issues in discrete choice models
Stated and revealed preference data
Choice sets and attribute sets
3. Model extensions
Multinomial and multivariate probit models
Panel data models
Fixed and random effects
4. Models for counts
Dispersion and heterogeneity: Negative binomial model
Models for panel data
Specification issues in count data models
5. Special topics
Sample selection models
Censoring and truncation
Simulation based estimation
Models for panel data: Random parameter models
Fixed and random effects
Estimation: Bayesian and Maximum likelihood estimation
Computer Lab Sessions and “Hands on Data”
Practical examples and open discussions will take place in the second session in the morning and afternoon meetings. In these sessions we will:
1. Learn how to fit and analyze discrete choice models
2. Discuss philosophical, practical and technical issues.
3. Discuss applications of the techniques that have appeared in the literature
4. Estimate models using real data. Carry out exercises using NLOGIT.
5. Develop applications to be discussed on the last day of class.
The software package used in the course will be NLOGIT, written by the instructor. Various related materials will be distributed.
Target Group and Requirements
The course may be of interest to
1. PhD students interested in new methods of estimation. Students from American University and other universities are welcome.
2. Faculty, professional economists, researchers and econometricians who work in support of decision making in government agencies as well as the private market.
NOTE: Participants should have prior knowledge at the level of an introductory course econometrics or a course in statistical analysis (estimation techniques) at the PhD level.
The course can be taken for three credits or for no credit. To receive the full three credits, the participant needs to complete a research paper. Credits can only be obtained by writing the applied paper. Students can start working on the paper at the end of the course. The paper is due a number of weeks after the end of the sessions.
• The main texts are Hensher, D., Rose, J. and Greene, W., Applied Choice Analysis, Cambridge University Press, 2005 and Cameron, A.C. and Trivedi, P., Microeconometrics, Cambridge University Press, 2005.
• A Reader composed of the main readings for the class will be provided to each participant. This will include selections from Greene, W., econometric Analysis and some articles.
• A detailed reading list will be provided with the Final Syllabus.
• Handouts for NLOGIT will be provided as well.
Three Credit costs for students.
Fixed fee (zero credits) for Researchers.
William Greene is currently Professor of Economics and Entertainment and Media Faculty Fellow, Department of Economics, Stern School of Business, New York University. His primary field of interest is applied econometrics in the areas of discrete choice modeling, production economics and efficiency estimation and panel data. He is also a student of the economics of the entertainment industry. Teaching at New York University since 1982 includes Econometrics, Microeconomics, Macroeconomics and Economics of the Entertainment and Media Markets. Previously Professor of Economics, Cornell University, 1976-1982. Other employment includes visiting lectureships at Oxford, Indiana, University of York, the Chicago Federal Reserve Bank, University of Lund, and consulting to The World Health Organization, Resource Consultants, Inc., Ortho Biotech, National Economic Research Associates, The U.S. Navy and others. He is the president of Econometric Software, Inc.since 1985. He is the author of the textbooks Econometric Analysis and Applied Choice Analysis, software LIMDEP and NLOGIT, and approximately 100 articles in applied econometrics, econometric methods, transportation, health economics, etc. appearing in Econometrica, Journal of Econometrics, Review of Economics and Statistics, Transport Research, Health Economics, The Journal of Productivity Analysis, Econometric Reviews, Empirical Economics, Journal of Political Economy, Journal of Economic Education, Economic Perspectives and the American Economic Review and others. He is currently the editor in chief of Foundations and Trends in Econometrics and is an associate editor for The Journal of Productivity Analysis and the Journal of Economic Education.