What the React index measures

What the index is for

The React index answers the question: is the world changing faster than usual? ‘Changing’ here means the kind of events that alter plans at short notice: large earthquakes, storms, floods and eruptions, armed conflict, and sudden changes to trade or technology. It is global; local security around the Baltic states is covered by the Be Safe index.

Like the other indexes, it is a 0 to 100 reading where 50 is an ordinary day compared with the recent past. How to read a 0–100 reading explains the scale. It is built from four inputs:

• Strong earthquakes, from the US Geological Survey: 20% of the weight.
• Storms, floods and eruptions, from NASA’s Earth Observatory Natural Event Tracker (EONET): 20%.
• Armed conflict reports, from the GDELT Project: 35%.
• Sanctions, trade and tech shock reports, from the GDELT Project: 25%.

All four are counts of events or articles. For each one the site adds up the last 7 days, ranks that 7-day total against up to 365 earlier days, and turns the rank into a score. The index is the weighted average of the four scores. The details are on How it works.

Earthquakes of magnitude 6 and above

What it measures: the number of earthquakes of magnitude 6 or more anywhere in the world in the last 7 days. The data comes from the USGS Comprehensive Earthquake Catalog (ComCat), which collects earthquake parameters contributed by seismic networks around the world^[2].

How many is normal: from 2000 to 2021, the USGS counted on average about 137 earthquakes a year of magnitude 6 to 6.9, about 14 of magnitude 7 to 7.9 and about one of magnitude 8 or more^[1]. That is roughly 150 a year, or about three a week. Years vary: the count of magnitude 6 to 6.9 earthquakes ranged from 104 in 2017 to 185 in 2011^[1]. A week with five or six is noticeably busier than usual; a week with none is not unusual either.

Why it matters to an ordinary person: most magnitude 6 earthquakes happen far from where readers of this site live, and many are at sea. They still matter when they hit manufacturing regions, ports or tourist areas: they can interrupt supply chains for electronics and car parts, delay shipping and change travel plans.

How the site uses it: it asks the catalogue for every earthquake of magnitude 5.5 or more in the last 400 days, keeps those of 6 or more, and counts them by the UTC day on which they happened, not the day they were reported. This is a sensor measurement, so it counts fully in the reliability figure. It carries 20% of the index.

Known weaknesses: every earthquake counts once, whatever its size, depth or location, so a magnitude 6.1 in the open ocean weighs as much as a magnitude 7.5 under a city. Large earthquakes are followed by aftershocks, so one event can raise the count for days. A magnitude of 5.9 does not count at all. Information about each event in the catalogue is versioned and can be updated after it is first published^[2], so a quake near the threshold can move in or out of the count.

Storms, floods and eruptions tracked by satellite

What it measures: the number of new severe storms, floods and volcanic eruptions logged worldwide by NASA’s Earth Observatory Natural Event Tracker (EONET) in the last 7 days. EONET is a catalogue of natural events, each linked to the sources that first recorded it^[3]. It tracks 13 categories, including drought, dust and haze, wildfires, sea and lake ice, and temperature extremes^[4]. React uses three of them: severe storms, volcanoes and floods. Wildfires are left out of the global count because there are so many of them that they would swamp the other categories.

Why it matters: tropical storms, floods and eruptions close airports and ports, damage crops and energy infrastructure, and change prices for food and insurance far from where they happen.

How the site uses it: it asks EONET for events of those three categories from the last 365 days, open and closed, and counts each event once, on the date of its first recorded position. It carries 20%.

Known weaknesses: NASA itself says EONET is meant for visualisation and general reference and should not be treated as an official source for precise location or timing^[3]. Dates are often set to midnight UTC when the original source gave no time^[3]. A tropical storm counts as one event whatever its strength, and an event can be added to the catalogue some time after it began. How complete EONET is depends on the sources that feed it.

Conflict reporting

What it measures: how many English-language news articles per day mention invasions, airstrikes, coups, declarations of war or troop mobilisation, as counted by the GDELT Project. GDELT monitors print, broadcast and online news from almost every country, in more than 100 languages^[5]. The site counts the articles whose headline matches this search:

• ‘military coup’ OR ‘declared war’ OR ‘troop mobilization’ OR invasion OR airstrikes

Quoted phrases must appear as exact phrases, and OR means any one of the terms is enough. Words match whole words, whatever the case.

Why it matters: new or widening wars move energy and food prices, change travel and shipping routes, and can bring sanctions that reach European businesses and households. Coverage tends to rise quickly when fighting starts or spreads, so the count reacts within a day or two.

How the site uses it: GDELT publishes a list of the English-language articles it has seen every 15 minutes. Every two hours a scheduled job downloads the new lists and counts the headlines that match the search, then stores the counts. Each day’s figure is the number of matching headlines per list read that day, times the 96 lists in a day: an estimate of the whole day, even when some lists could not be read. The site reads the stored counts and never searches GDELT itself. GDELT’s translated articles are not counted. The history builds up from these lists; when the counting began, four lists a day (at 00:00, 06:00, 12:00 and 18:00 UTC) were sampled for the last 30 days, and counts are kept for up to a year. It carries 35%, the largest share, and as a media source it counts at 60% in the reliability figure.

Known weaknesses, stated plainly:

• It counts articles, not events. One airstrike reported by 300 outlets adds 300; a war that newsrooms have stopped covering adds little. The grouping of headlines into single events used on the events page does not apply to these counts.
• The words have more than one meaning. A ‘home invasion’ in a local crime report counts as much as an invasion of a country.
• The counts are not shares of all coverage. A busy news day, or a change in the sources GDELT monitors, moves them.
• English-language outlets, many of them in the US and UK, decide what gets written about.
• The history builds forward from the lists the site has read, up to a year, so for now each day is ranked against a shorter period than a full year.
• The current day is left out until it is complete (UTC), so the latest 7-day total ends with yesterday.

Measured, reported, counted once explains how media reports differ from sensor data on this site.

Sanctions, tariffs and tech shocks

What it measures: the same kind of daily article count, for a different set of words:

• sanctions OR tariffs OR ‘export ban’ OR embargo OR ransomware OR ‘global outage’

Why it matters: new sanctions, tariffs and export bans change what goods cost and whether they can be bought at all. Ransomware attacks and large technology outages can stop hospitals, shops, airlines and payment systems for hours or days. These are structural changes: they rarely have a single moment, but the volume of reporting shows when they are being discussed more than usual.

How the site uses it: as for conflict reporting, a daily count of matching English-language headlines from GDELT’s article lists, summed over 7 days and ranked against the history built up so far. It carries 25% and counts at 60% in the reliability figure.

Known weaknesses: all the limits of the conflict count apply. The words are again broad. In British English a ‘tariff’ is also an electricity or phone price plan, ‘sanctions’ are used in sport and welfare policy, and a ‘global outage’ can be a streaming service down for an hour. Trade policy announced through official documents rather than news stories is only seen once it is reported.

Why ‘the world changing’ is measured this way

There is no single public measure of how fast the world is changing. The index combines two kinds of evidence. Sensors and satellites (earthquakes and natural events, 40% together) record that something physical happened, whether or not anyone wrote about it. News counts (conflict and structural reporting, 60% together) record how much attention human events are getting, which is the only free, daily, worldwide measure available for wars, sanctions and outages.

Because media counts are less certain than measurements, they count at 60% in the reliability figure. With all four inputs live, reliability is 76 out of 100. If GDELT does not answer, the two sensor inputs still carry 40% of the weight, which is the minimum the site needs to report a reading, and reliability falls to 40.

Counting the last 7 days, rather than one day, smooths out quiet weekends and single busy days, so the reading moves steadily rather than jumping. The trade-off is that a large event on a single day is spread across the next week.

Blind spots

• Size. Every count treats one event or article as one, whatever its consequences.
• Disease outbreaks, elections, heatwaves, droughts and wildfires are not part of this index.
• Financial markets are covered by the Invest index, and security around the Baltic states by Be Safe.
• Anything that is not written about in English, for conflict and structural reporting.
• Slow change. A trend that builds over years, such as warming or ageing populations, does not show up in a 7-day count ranked against recent months.
• Direction. The index says that more is happening than usual, not whether it is good or bad.

Each source and its licence is listed on the sources page.