How to Measure Geopolitical Risk: What's a Number and What's a Guess

TL;DR: A handful of indices count geopolitical risk honestly, by tracking newspaper text volume or filtered company disclosures, but they measure attention, not the probability or size of a loss to a specific company. Most of what a risk register calls "geopolitical risk" is a judgment call dressed as a score, and the fix is keeping the two in separate columns rather than blending them into one false number.

Key takeaways:

• The Geopolitical Risk Index, built by Dario Caldara and Matteo Iacoviello, counts the share of geopolitical-threat articles across ten major newspapers each month and has run since 1985.[^1]
• That index hit its highest 2025 reading on 13 June, the day Israel struck Iran, yet US equities fell only about 1.1 percent that week against a historical average drawdown near 7.7 percent after comparable events.[^2]
• MSCI's Geopolitical Risk Indicator filters company-level news into six categories and pairs each name with a rolling "geopolitical beta" measuring how its stock has historically moved against the indicator.[^3]
• Fewer than one in five companies run a dedicated geopolitics function, and even fewer have built routines that connect a geopolitical reading to a specific business metric.[^4]
• The honest split is attention versus consequence: news-volume indices measure how loudly the world is talking about a threat, not what that threat is worth to one company's revenue.

Israel's strikes on Iran on 13 June 2025 pushed a decades-old academic index of newspaper coverage to its highest point of the year.[^2] The stock market barely moved. That gap between the number and the outcome is the whole problem with measuring geopolitical risk, and it is worth sitting with before building a scorecard around any single figure.

What Actually Produces a Number

Two families of measurement exist, and they count different things.

• News-volume indices count words. The Geopolitical Risk Index tracks the share of articles about war threats, military buildups, terror acts and related categories across the Chicago Tribune, Financial Times, Guardian, New York Times, Wall Street Journal and five other papers, going back to 1985.[^1] It is a measure of press attention to adverse events, not a measure of what any of those events cost a given firm.
• Company-filtered indicators are newer and narrower. MSCI's Geopolitical Risk Indicator runs company disclosures through a language model that sorts them into six categories, from cross-border conflict to trade restrictions to energy access risk, then produces a weekly reading and a rolling "geopolitical beta" for each covered stock.[^3] This gets closer to a firm-specific number, but it is still built from what companies choose to disclose, filtered by a model's classification choices.
• Market-based proxies infer risk from price moves in credit spreads, freight rates or war-risk insurance premiums. These are real prices, which makes them verifiable, but they mix geopolitical risk with every other reason a price moves that week: a rate hike, an earnings surprise, a currency move unrelated to any political event.

Each family answers a different question honestly and a fourth question badly. News-volume indices answer "how much is the world talking about this." Company-filtered indicators answer "which category does this disclosure fall into, and how has this stock historically reacted." Market proxies answer "how did a tradable price move this week." None answer "what does this cost my company," because that needs a model of one firm's contracts and revenue lines, which no general-purpose index can build for someone else's balance sheet.

None of these produce a probability of loss for a specific company on a specific exposure. They produce an attention signal, a classification signal, or a price signal. Confusing any of the three with the fourth thing risk registers need, an estimate of consequence, is where most quantification efforts break down.

Where the Number Runs Out and the Judgment Starts

The 13 June 2025 example is instructive precisely because the index and the market disagreed. The Geopolitical Risk Index recorded its 2025 peak that week.[^2] US equities fell roughly 1.1 percent, well below the average drawdown of about 7.7 percent that comparable historical spikes have produced.[^2] The index was right about attention. It said nothing about whether a specific manufacturer's Gulf supplier contracts, insurance renewals or freight routes were actually exposed, and by how much.

That is the honest boundary. What is countable: how much a topic is being written about, how a model classifies a disclosure, how a price has moved. What is not countable without a company-specific model: whether a named counterparty, contract or route is exposed, and what that exposure is worth in revenue or margin terms. The second question is ordinal at best, a ranking of "more exposed" against "less exposed" built from analyst judgment, not a cardinal number with decimal places.

The failure mode is treating the ordinal judgment as if it carried the same precision as the news-count index, then averaging the two into a single "risk score" that looks quantitative on a dashboard and means nothing underneath. A board that sees "geopolitical risk: 62/100" has been given false confidence, not information. Most vendor dashboards fail here in a specific way: they measure what is easy to display, not what the register needs.

Running Two Columns Instead of One

The practical fix is running two columns instead of one.

• Column one, measured: attention indices, company-filtered indicators and market proxies, cited with their source and update frequency, used to flag when to look harder, not to rank exposures against each other.
• Column two, judged: entity-level and route-level exposure assessments, explicitly labelled as analyst judgment, revisited on a fixed cadence, and never averaged into column one.

In practice this looks like a register with two visible fields per line item, not one blended score. A supplier contract in a contested region carries an attention reading, updated automatically from whichever index the team subscribes to, next to an analyst's exposure rating, updated quarterly or triggered early by a defined threshold. The two fields can disagree. When the attention reading spikes and the exposure rating has not been revisited in the quarter, that mismatch itself is the signal worth escalating, not a blended number that hides the disagreement.

This split matters because of how few companies have built the connective tissue between the two. A January 2026 study of 56 senior executives by BCG, IMD Business School and the World Economic Forum found that fewer than one in five companies have a dedicated geopolitics function at all, and that even where geopolitical monitoring exists, "embedded routines to link geopolitical developments to core business metrics and actions remain rare."¹ Most firms are still relying on manual curation or consultant summaries to bridge that gap, which is itself a judgment process wearing the clothes of a system.

Building a measurement practice that survives scrutiny means naming which number in the register is counted and which is assessed, and refusing to let the second kind borrow the first kind's decimal points. That split also decides what belongs in daily monitoring versus what only needs a quarterly look, a separate design question from measurement itself. Teams building either half of this from scratch can start with the Fortius Intel brief.

Fortius Intel note: A geopolitical risk score that cannot tell a board which half of the number is counted and which half is judgment is not a measurement. It is an opinion with a font that makes it look like one.

Methodology: figures drawn from the Geopolitical Risk Index (Caldara and Iacoviello, Federal Reserve and policyuncertainty.com), MSCI's published Geopolitical Risk Indicator and Geopolitical Beta methodology, AMP Capital's Econosights research note on the June 2025 Israel-Iran escalation, and the BCG/IMD/World Economic Forum January 2026 executive study.

---

Footnotes

1. BCG, IMD Business School and World Economic Forum, "Building Geopolitical Muscle: How Companies Turn Insights into Strategic Advantage," bcg.com, published 12 January 2026, retrieved 22 August 2026, https://www.bcg.com/press/12january2026-companies-dedicated-geopolitics-department. Source for: the finding that fewer than one in five companies have a dedicated geopolitics department, the 56-executive interview sample, and the finding that embedded routines linking geopolitical developments to business metrics remain rare. ↩