Summary

This course introduces participants to discrete choice models. These econometric models are used to explain how people choose between discrete outcomes, such as mode of travel to work or type of treatment for pain. The course will cover the subset of discrete choice models known as random utility models, namely the multinomial logit and nested logit. These models are often used in disciplines such as economics, transportation, and public health. No prior knowledge of discrete choice modeling is expected. Hands-on exercises will be conducted in Python.

Why Take This Course?

Random utility models are used across many disciplines. They allow one to use regression techniques to model choices between multiple outcomes, something not possible with many other models. Unlike many other models of discrete outcomes, random utility models are interpretable—it is easy to see which predictor variables are associated with which choices. Random utility models are also consistent with rational economic theory, meaning that properly specified estimates can be interpreted as willingness-to-pay and transformed into dollar amounts to understand the welfare impacts of policy. This course will prepare participants both to estimate these models and to interpret and evaluate them when encountered in practice.

What Will Participants Learn?

This course will combine lecture and hands-on coding experiences. Participants can expect to learn:
The theory underlying random utility modeling
How to interpret estimates from random utility models
Common pitfalls in random utility modeling
How to structure and estimate random utility models
Prerequisites and Requirements

Participants should be familiar with linear regression. Some understanding of binary logistic regression, as well as experience using Python, will be helpful but is not required.