Insights from Policy Lab in AMIGDALA

Interview with Dr. Marek Tiits, Policy Lab, on AMIGDALA’s trade modelling role. Dr. Tiits explains how Policy Lab builds machine-learning predictive models of bilateral trade to quantify long-run patterns, green-goods flows, and policy impacts that feed into AMIGDALA’s integrated economic and energy analyses.

Policy Lab contributes to AMIGDALA by developing predictive models of international trade based on advanced machine learning techniques. Our work extends the project’s framework through the integration of detailed global trade data, drawing on historical series that reach back to the 1990s. This enables us to analyse long-term patterns and assess how trade flows may evolve under different policy and market conditions. 

A particular emphasis is placed on trade in green goods that are central to Europe’s green transition. We combine standard determinants of trade, such as economic size and geographical distance, with additional variables reflecting tariffs, environmental policies, and technological change. This data-driven approach supports the estimation of multi-product gravity models using machine learning methods, including random forests and graph neural networks. Such models cover bilateral trade across nearly 200 countries and more than 5,000 product groups.

On this basis, we explore opportunities to contribute to the integrated modelling effort in AMIGDALA. The predictive trade models complement macroeconomic and energy system models, and their insights can be incorporated into the broader framework. This provides policymakers and industry stakeholders with a clearer understanding of the interactions between trade, industrial production, and energy use, which is essential for designing Europe’s pathway towards climate neutrality and resilience.