The news tells you where the trouble is. Not who pays for it.
The standard measure of geopolitical risk counts newspaper articles: which countries the news is about. That is a different question from which economies bear the cost, and the two answers barely overlap.
Same world. Two entirely different answers.
On the left, the twelve countries the news reports as most at risk. On the right, where each of those countries sits once you ask instead whose economy is wired to that risk. Every crossing line is a country whose position changes.
Loading…
Where the trouble is
This is the Caldara–Iacoviello Geopolitical Risk index by country. It counts how often ten anglophone newspapers write about geopolitical threats and acts involving each country, rebased so that 100 is that country's own long-run average.
It is a good measure, openly licensed, and versions of this map already exist. It answers one question well: where is the trouble?
Who carries it
Now the same risk routed through dependency: trade, energy, raw materials and shipping chokepoints. This is Intensity: not where trouble happens, but whose economy is wired to it.
Mexico carries heavy exposure with no domestic instability at all, because most of its trade runs to a United States sitting above its own historical average.
Across every month since 2020, the first with properly lagged dependency weights, the rank correlation between a country's own risk and its exposure is 0.17, close to unrelated. That gap is the whole reason this project exists.
Whose newspapers?
Here is the risk index again without rebasing: raw article shares on a common scale. One country dominates, and it is not a country in crisis.
The United States tops this list because American and British newspapers write about the United States. The index measures how visible a risk is to a Western readership, not how dangerous it is. Salience is not threat.
A limitation, not a bug. Weighting by trade does not remove it, and Intensity inherits it in full. It is set out in §5.1 of the methodology, where it belongs.
February 2022
A measure that cannot fail a test is not worth much. So the index is scored against episodes where the answer is known in advance, with the expected countries written down before looking.
For the invasion of Ukraine the prediction was: European importers of Russian energy. The energy channel's largest movers that month were Finland, Poland, Sweden and Denmark, as written.
The tests it failed
The same procedure failed on shipping chokepoints. The prediction was European Red Sea traders; what appeared was a near-uniform rise almost everywhere, the signature of a channel carrying one common factor instead of separating one economy from another.
The diagnosis was concrete. The routing table wrongly sent South-East Asia–China trade through the Taiwan Strait, and Russia–East Asia trade through Suez. It also routed Hong Kong north of the Taiwan Strait, when the Pearl River delta lies to its south. And the channel fed an economy's own risk back through its own strait: Saudi Arabia's Hormuz term was built from Saudi Arabia's GPR alone, Egypt's Red Sea term partly from Egypt's. Each has been corrected — an economy's own GPR is now excluded from the littoral mean, and where nothing remains, that strait contributes nothing to that economy. Even so, that channel is still marked provisional.
The failures ship with the successes.
A measure whose failures are hidden should not be trusted with the ones that pass. All four tests below run from the repository, and their report is committed alongside the data.
It is not a re-plot
Rank correlation between a country's own risk and its exposure averages 0.17 across 80 months. Intensity carries information the source-side index does not.
One channel barely earns its slider
Raw materials correlates 0.92 with trade, and value-added 0.94. The chokepoint channel discriminates; those two mostly restate gross trade.
Routing errors and a leak
The chokepoint event studies missed their pre-registration. Each failure localised a checkable bug rather than a vague miscalibration: two routes sent the wrong way, Hong Kong on the wrong side of the Taiwan Strait, missing Red Sea and Malacca legs, China treated as wholly north of the strait — and an economy's own risk fed back through its own strait. All corrected; the channel stays provisional.
The ranking moves, partly
A two-way fixed-effects fit leaves 0.14 of the variance to economy-specific movement, and the cross-sectional ranking correlates 0.51 with itself a year earlier. The mixed index moves; the chokepoint layer (0.96) is a fixed structure recoloured by world risk.
Now put it to your own question.
The map lets you set the channel mix yourself, switch between a country's own historical baseline and a common scale, and compare gross trade against value-added dependency. There is no single correct weighting, so the weighting is yours to move. Read the result as a sensitivity analysis rather than a ranking.