North-South trade integration and asymmetric effects: Double Machine Learning evidence from EU-Morocco

This study examines a key paradox in North-South trade integration: while Free Trade Agreements aim to boost export opportunities for developing nations, they often yield asymmetric trade responses that strain external balances. We present one of the first causal evaluations of this issue using Morocco's European Union Association Agreement, which covers over 60% of its trade.Using a Double Machine Learning Difference-in-Differences (DML-DiD) with multiple time periods approach on a balanced panel of 78 countries (1995--2023), we exploit the phased FTA rollout across EU countries. Our findings reveal significant asymmetries: EU imports rose by 120.6%, while exports increased by only 26.1%, yielding a 3.4:1 import-to-export Average Treatment Effect on the Treated (ATT) ratio (0.791 for imports, 0.232 for exports in log-points). This imbalance caused a 51.3 percentage point decline in the trade coverage ratio. Mechanism analysis shows that geographic proximity magnifies FTA effects by 25--31 percentage points; robustness checks, including Rambachan-Roth sensitivity analysis, validate our methodology. These results highlight a fundamental trade-off: while preferential liberalization expands trade volume, maintaining sustainable trade balances requires complementary policies addressing supply constraints and asymmetric market access. JEL Classification: F13 , F14 , F15 , C21 , C23