An analytical framework for causal decision-making in international trade

In international trade, selecting an Incoterm is a critical decision impacting costs and risks. Previous quantitative studies relied on traditional econometric models imposing strict linearity, limiting causal interpretation. This study addresses this gap by applying a causal inference framework to identify Incoterm choice drivers among Colombian exporters. The methodology combines (i) directed acyclic graphs to encode causal assumptions; (ii) nested, cross-validated random forests to capture complex, nonlinear relationships; and (iii) G-computation to estimate unbiased causal effects. Results show that transactional features dominate selection. Transitioning from bulk to containerized cargo increases the probability of selecting an E/F-group term-such as Ex Works (EXW), Free On Board (FOB), or Free Carrier (FCA), where the seller’s obligations end upon delivery to the buyer-by 13.4 percentage points. Additionally, shipment weight and value exhibit nonlinear threshold effects, where E/F terms become significantly more probable only after shipments surpass specific size thresholds-a pattern missed by prior linear models. Relational factors also matter; sharing a common language with the trade partner increases E/F-term probability by 6.5 percentage points. This framework provides managers with a tool to predict Incoterm choices and quantify the causal impact of altering shipment profiles, enabling better-aligned, lower-risk trade contracts. • Apply a causal framework to model trade term selection in real-world customs data. • Capture nonlinear effects of shipment size and value using advanced learning models. • Identify primary trade drivers through empirical analysis of firm-level transactions. • Estimate causal impacts of logistics choices with robust decision tools. • Support international trade planning with data-informed strategic insights.