Choice Modeling Using Micro Data: Applications

Course overview

The growing availability of survey data and large datasets has transformed the way economists and social scientists conduct research, both in academia and in professional practice. These datasets often include qualitative or categorical information, making traditional regression models unsuitable for many applications. For example, standard linear models cannot adequately handle outcomes based on discrete choices (e.g., whether an individual has private health insurance), binary statuses (e.g., employment vs. unemployment), or non-negative variables with many zero observations (e.g., household expenditure on specific goods like tobacco).

This course introduces participants to a range of discrete choice and limited-dependent variable models, providing the tools needed to analyze and interpret such data effectively. Special attention will be given to the use of choice models for prediction—including variable selection techniques—and to the integration of machine learning methods for working with large datasets.

The program is organized into two main blocks:

• Block 1 (three sessions): Discrete choice models, including count data models.

• Block 2 (two sessions): Limited-dependent variable models, including sample selection issues commonly encountered in cross-sectional data.

Hands-on sessions will guide participants through estimating models in STATA and interpreting results using real-world data drawn from fields such as demand analysis, health economics, and labor economics. Selected empirical papers will also be discussed to connect theoretical models with applied research practice.