Aizenman, Desbordes and Saadaoui: Measuring Commercial Destruction from Bombs and Bullets vs. Taxes and Penalties

That’s my definition. From their recently released NBER Working Paper, The Risk of International Conflict and Trade: Military Wars, Trade Wars, and the Noise of Nationalism, by Joshua Aizenman, Rodolphe Desbordes and Jamel Saadaoui:

We construct a monthly conflict index from GDELT v2 event data, averaged
against the human-selected ICEWS dataset using Ridge regression. Although the index measures the hostile part of bilateral relations, not its volume, the measurement remains important: GDELT wrongly categorizes normal relations as hostile, increasing the share of hostility among allies, elite traders who dominate media coverage.

Without correction, these measurement errors may significantly distort the measurements
trade conflict relations. Each observation decomposes into four additional parts:
kinetic warfare, military posture, trade content hostilities, and baseline diplomacy. I
index is directed, allowing for separate measurement of aggressive and retaliatory actions
the answers.

Three sets of findings emerge. First, the decomposed panel tracks the country’s key developments over the past decade. Russia-Ukraine relations are strained long before 2022. The US-China conflict is largely economic. These patterns are not visible in the composite display (Section 3).

Second, in the gravity equation, the aggregate index is negative, large, and statistically
Importantly, however, the decomposition reveals that only two of the four layers drive the effect. Kinetic conflicts (“military war”) and trade context hostilities (“trade war”) are both economically large and precisely measured. Standard communication, despite the dominance of limited hostility, has a commercial effect that cannot be separated from zero. The target structure shows a revenge path that includes losing trades over several months. Against pre-increase trade, the country’s recession since 2015 has put at risk about $334 billion, with the US-China pair accounting for half (Section 4).

Third, in the horse race against IntenSE (Chevalier et al., 2026), the two countries are very close
index, our decay is still statistically and economically significant.
Because the two indicators draw on the same basic GDELT data, the comparison
distinguishes the contribution of our four design options: Goldstein weighting, supervised comparison against a human-selected ground truth, bilateral structure directed (asymmetric),
and sanctions-on re-content of sections (Section 5).

The construction of the indices reveals some interesting patterns in the way conflicts are conducted, including among the “Great Powers” of the last decade.

Their results also show interesting patterns in the use of weapons of economic policy through international coercion:

To me, the interesting result is that democracies are confined to using coercion through military action. At least this is true until 2023. Whether this will be true in 2026 is, to me, an interesting question (closely related to whether the US will still be considered a democracy in 2025).

Using their estimates, the authors conclude that the increase since 2015 has reduced trade by $344 bn by 2023.

This is an equilibrium equation that ignores trade diversion. It overestimates the bilateral welfare losses but gives an indication of the trade that could be destroyed by the increase in conflicts over the past decade: about 2.4% of global bilateral trade in 2015.