Forecasting in a Changing World
New program
Standard forecasting pipelines assume stable relationships in the data. When those relationships shift — as they regularly do in macroeconomics and finance — these pipelines can break down silently, producing overconfident forecasts from outdated models.
We are developing forecasting methods and end-to-end workflows designed for unstable environments, where model performance must be continuously monitored and forecasting strategies must adapt when relationships shift.
What We Are Building
- A practical guide to forecasting under structural change — combining methods with governance principles for when and how to revise forecasting strategies
- A Python forecasting toolkit for evaluation, monitoring, and adaptive procedures
- Live forecasting illustrations and reproducible case studies demonstrating performance and breakdown risk
Building on Our Foundational Research
This program builds on our foundational research on structural change in macroeconomic relationships. Our work on detecting instability, testing for structural breaks, and regime-aware estimation provides the econometric foundations that forecasting under structural change requires.
Our open-source regimes Python package provides the break detection, stability testing, and regime-aware estimation tools that underpin this work.