Skip to main content
Dryad

Did tariff exposure bend national emission trajectories? A pre-trends cautionary tale from the US-China trade war

Diwani, Mazen1; Hamid, Mohamed2; Santangelo, Giacomo3

Published Jul 28, 2026 on Dryad. https://doi.org/10.5061/dryad.qrfj6q5zf

Data files

Jul 28, 2026 version files 231.79 KB

Click names to download individual files

Abstract

The 2018 US–China trade war was the sharpest escalation of trade barriers among major economies in decades, yet its consequences for carbon emissions remain little studied. We ask whether tariff exposure altered the national CO2 emission trajectories of affected countries, using propensity score matching across 171 countries over 2015–2019 (162 on common support), with treatment defined on trade-weighted tariff quartiles. A conventional matched comparison appears to show high-exposure countries — predominantly rapidly industrializing economies — growing emissions 4.5 percentage points faster than controls (p = 0.007), a gap stable across alternative algorithm labels and sample exclusions. This apparent robustness is illusory. The properly weighted matched ATT — the intended estimand — is never statistically distinguishable from zero (0.28 pp before a pre-trend control, 0.32 pp after). The significant 4.5 pp figure is instead an unweighted two-sample comparison, the wrong estimator for a matched design, and does not correct itself: it stays 4.4 pp (p < 0.01) even with the pre-trend covariate in the match. A placebo over 2015–2016, before any tariff, finds them already growing 2.4 pp faster than controls (p = 0.009). We find no detectable effect on national trajectories beyond pre-existing dynamics, but the design’s limits are real: the confidence interval is wide (we can exclude effects above roughly 6 pp, not moderate ones), the window is short, and national aggregates miss the cross-border reallocation channel. The transferable lesson is methodological: baseline covariate balance offers no protection against selection on growth dynamics, and placebo and pre-trend diagnostics are part of identification, not optional refinements.