Machine Learning for Applied Economists — Montréal Summer School 2026

This short course introduces machine learning methods through the predictive econometric workflow, with a sustained focus on what works, why it works, and where ML actually generates value versus reinventing the wheel. The course has been taught at central banks and universities internationally, and is being brought to Montréal for the first time in summer 2026.

Who this is for

Applied economists in the private sector, central banks, and government, as well as PhD/Master students and researchers working with economic data. Past cohorts have mixed all of the above, and the discussions are better for it.

What you'll learn

The course covers regularization (Ridge, Lasso, elastic net), tree-based methods (Random Forests, Boosting), neural network architectures tailored for economic data, LLMs and text as data, forecasting, and impulse response functions, all with economic applications. Interpretability sits at the center of it all, so that improved predictive performance doesn't come at a significant cost to transparency. Particular attention is given to architectures and methods designed specifically for the structure of economic and financial data, not just off-the-shelf models adapted from other domains.

The payoff

Participants leave able to design in-house algorithms tailored to their own problems, not just apply off-the-shelf tools. With Claude Code and related tools, implementation is no longer the bottleneck. Craftsmanship is, and that's what this course is built to provide.

Schedule

The course runs Thursday July 2 through Saturday July 4, 9:00 AM to 4:30 PM each day. Each day includes a 60-minute lunch break and two short coffee breaks.

What's included

• Three days of instruction in downtown Montréal
• All course materials, slides, code, and datasets
• Coffee, tea, and light morning refreshments daily

Prerequisites

Basic familiarity with econometrics, statistics, or data science. No specific programming language required. The course is method-focused, with code examples in R and Python you can run alongside.

Instructor

Philippe Goulet Coulombe is Associate Professor of Economics at UQAM. He holds a Ph.D. in Economics from the University of Pennsylvania. His research develops machine learning methods for economic data, with a focus on interpretability and applications in macroeconomic forecasting. He is Associate Editor at the International Journal of Forecasting and consults for central banks, including the ECB and OeNB. He is a CIRANO researcher, a SUERF fellow, and a research affiliate of the Chaire en macroéconomie et prévisions and the Centre de recherche sur l'intelligence en gestion de systèmes complexes (CRI²GS).

Language

The course is taught in English. Questions and informal discussions can take place in French.

Refund policy

Full refund up to 30 days before the course start date. 50% refund for cancellations 14–29 days before. No refund within 14 days, but registrations may be transferred to another person at no charge.

For employer-funded registration requiring a formal invoice, please email [email protected].