The scale
| READING | ZONE | WHAT IT MEANS |
|---|---|---|
| 0–24 | Complacency | Unusually calm. In market terms: the moment nobody expects a shock. |
| 25–44 | Easing | Fewer and milder conflicts in the news, more cooperation. |
| 45–55 | Neutral | A typical day by the standard of recent years. |
| 56–75 | Tension | Conflict and coercion are clearly above normal. |
| 76–100 | Crisis | War, attacks or ruptures dominate the world’s news far beyond their usual level. |
The raw material
The GDELT event database. Every day, the GDELT Project reads news articles from around the world, in more than 100 languages, and codes each event it finds: who did what to whom, where, and how hostile or cooperative it was, on the Goldstein scale from −10 to +10. Riftrate reads this database through Google BigQuery.
The Geopolitical Risk Index (GPR) by Dario Caldara and Matteo Iacoviello, which counts the share of newspaper articles about geopolitical threats and acts. It is published every week, with a few days’ delay; its last value is used for up to 14 days.
How each event is weighed
Not all bad news is equal. Each hostile event receives a weight:
weight = severity × coverage × countries
- Severity: how hostile the act is on the Goldstein scale, doubled for threats or use of weapons of mass destruction.
- Coverage: how many articles mention it (on a logarithmic scale), a little more if it leads the article.
- Countries: 1 for a domestic affair, 1.5 for a conflict between two states or inside a nuclear power, 2 when a nuclear power is involved, 2.5 between two nuclear powers. Civil wars count at least as much as a war between two states.
Each family is then measured against the usual daily volume of world news over the past year, so that a flood of unrelated news (a pandemic, an election) does not dilute the conflicts.
The eight signals
| SIGNAL | WEIGHT | WHAT IT READS |
|---|---|---|
| War & military | 20% | Armed attacks, strikes, military threats and blockades. |
| Geopolitical risk (GPR) | 20% | The Caldara–Iacoviello index of geopolitical threats in the press. |
| Terrorism & mass violence | 12% | Bombings, hostage-taking, mass killings. |
| Escalation momentum | 12% | Whether conflict is gaining on cooperation: the last 7 days against the 90-day trend. |
| Economic coercion | 10% | Sanctions, embargoes, economic threats. |
| Media tone | 10% | The average tone of world coverage, from alarmed to calm. |
| Diplomatic rupture | 8% | Expulsions, broken relations, ultimatums. |
| Domestic unrest | 8% | Riots, coups, crackdowns. |
From signals to one number
- Each signal is averaged over the last 7 days, which removes daily noise and the weekend effect.
- It is compared with its own history over the last 3 years and the last 10 years, half each, so that a long war never becomes “the new normal”.
- That comparison becomes a score from 0 (far calmer than usual) to 100 (far more tense than usual).
- The eight scores are averaged with the weights above.
- A final calibration stretches the scale so that the extremes are reachable, as in market fear & greed indices, without changing the order of days. A light 3-day smoothing gives the published reading.
Why it moved, and the hotspots
Why it moved today. Each signal’s change since yesterday is multiplied by its weight; the two largest effects are named under the gauge.
Hotspots and countries this week. Each headline on the News page that names a country counts for it: 3 for war or terrorism, 2 for unrest, 1 for sanctions, diplomacy or de-escalation, a little more if it is less than 12 hours old. The hotspots are the highest scores of the last 48 hours; the weekly change compares with the same time a week earlier. It measures attention in the news, not events on the ground.
Today looks like…
We compare today’s eight signal scores with every day since 2018, except the last 90 days. The closest day is the one with the smallest average gap between scores, in points. We show three days at least 45 days apart, with the nearest major event within 30 days. Similar signals do not mean that the same thing will happen next: it is a comparison, not a forecast.
How often
Every hour. The current day enters the calculation once it holds a quarter of a normal day’s news (usually from about 5:00 UTC); before that, the page shows the previous day. The 7-day averaging means the needle moves by tenths of a point from one hour to the next, and by several points when something big happens.
Testing
Before each version ships, the index is tested against nine major crises since 2018 (from the Western strikes on Syria to the Iran war) and five moments of détente (such as the Singapore summit or the Israel–Iran ceasefire). The History page shows what the index did around each of them.
Limits
- Riftrate measures the world’s news, not events on the ground: a forgotten war weighs less than a covered one.
- Automatic event coding makes mistakes, and English-language media weigh more in the underlying data.
- Pandemics, financial crashes or natural disasters are not measured unless they turn into conflicts.
- It is a thermometer, not a forecast, and not financial advice.
Sources and credits
- The GDELT Project, gdeltproject.org.
- Caldara, D. and Iacoviello, M., “Measuring Geopolitical Risk”, American Economic Review, 2022. Data at matteoiacoviello.com/gpr.htm (CC BY).
- Headlines on the News page: public RSS feeds of the newsrooms named next to each headline. Links only; the articles belong to their publishers.
Versions
- 2.8: why it moved, today looks like, countries this week, coming up, pages for past crises, embeddable gauge.
- 2.7: new design, light and dark modes, live hotspots on the home page.
- 2.6: new name, Riftrate; the scale now rises with tension (0 calm, 100 crisis).
- 2.5: dial gauge, History, News, Methodology and About pages, automatic news wire.
- 2.4: hourly updates, final calibration, GPR repaired.
- 2.2: signals measured against the usual news volume; civil wars and signed agreements weighed properly.
- 2.0: events weighed by severity, coverage and the countries involved.