If civil war broke out in the US, how bad could it get? Data reveals the unpredictability of war

Photo not related to article. [Getty Images Bank]

Predicting how far a war will escalate is nearly impossible, according to new research.

Interstate wars since 1946 have killed an average of 93,000 people. World War II alone claimed about 16.63 million lives. Wars share a name, but their scale can differ by hundreds of times.

Some analysts say that for every additional year a war continues, there is roughly a 10 percent chance that annual deaths will double and about a 1 percent chance they will increase tenfold. The finding suggests that wars can grind on with little change before rare, sudden events trigger mass casualties.

A research team led by Aaron Clauset, a professor at the University of Colorado Boulder, and Barbara Walter, a professor at the University of California San Diego, published these findings in volume 12, issue 32 of the international academic journal Science Advances.

The team said interstate wars, which had quieted after the end of the Cold War, are on the rise again — reaching their highest level since the late 1980s as of 2026.

How do wars escalate?

The research team drew on battle-death data compiled by the Peace Research Institute Oslo. The dataset covers annual conflict fatalities worldwide from 1946 to 2008 and encompassed 299 civil wars and 22 interstate wars. For interstate wars, the team also used a separate dataset of 95 cases spanning 1823 to 2003.

The team defined an "escalation coefficient" as the ratio of deaths in a given year to deaths in the previous year — a value above 1 indicating escalation and below 1 indicating de-escalation. Because the measure tracks the rate of change rather than absolute numbers, it allows wars of vastly different scales to be compared on the same terms.

The results showed that escalation and de-escalation were nearly equally likely. In 44 percent of all cases, death tolls changed little from the previous year; among the remaining cases, the chances of a tenfold increase and a tenfold decrease were roughly the same.

Notably, the escalation patterns of civil wars and interstate wars were statistically indistinguishable. Since 1946, interstate wars have killed an average of 93,000 people over 2.1 years, while civil wars have killed an average of 21,000 over 3.5 years. Despite differences in scale and duration, the way escalation unfolds was the same across both types.

Real-world cases confirm that duration and scale do not necessarily go hand in hand.

The conflict between the Ethiopian government and the Oromo Liberation Front lasted 25 years as of 2008, yet recorded only 650 deaths. By contrast, the Chinese Civil War produced more than 1 million combat deaths in just four years, from 1946 to 1949.

The team built two war models — one without escalation and one with it — and tested each against actual data. A model based solely on initial scale and duration, with no escalation, failed to reproduce the observed distribution of deaths. Once escalation was incorporated, the model matched real-world data without any additional adjustment.

However, civil wars showed a tendency for deaths to decline the following year once they surpassed a certain threshold. When annual deaths exceeded 500, civil wars tended to shrink; below that level, they tended to grow.

The team interpreted this as a result of limits on rebel recruitment and domestic and international political pressure. In interstate wars, by contrast, death tolls did not predict whether the conflict would grow or shrink.

What if civil war broke out in the US?

Using their model, the research team attempted to project casualties for hypothetical civil wars breaking out in the United States, China, India and Nigeria in 2025 and 2060.

Initial death tolls were drawn from historical cases. The team compiled first-year fatalities from past civil wars, divided by each country's population, and randomly sampled from the resulting distribution. In the Kurdistan Workers' Party conflict that began in Turkey in 1984, for example, the country had a population of about 47.57 million and recorded 442 deaths in the first year.

The projections suggested that China and India, with their larger populations, could see around 50,000 deaths, while the United States and Nigeria could see around 10,000. However, repeated simulations produced a projected range spanning from 1,000 to 1 million deaths, yielding no statistically meaningful result.

The team said the findings underscore the impossibility of predicting how large a war might grow, noting that smaller-population countries can sometimes produce deadlier civil wars than larger ones.

Meanwhile, the team acknowledged several limitations: the annual data cannot capture escalation that unfolds over just a few months; the interstate war sample of only 22 cases made independent analysis difficult; and the model cannot explain what determines the initial scale at which a war begins.

Reference

DOI: 10.1126/sciadv.aeb1705

Aaron Clauset et al., "Escalation dynamics and the severity of wars." Sci. Adv. 12, eaeb1705 (2026).

dbsdn1110@heraldcorp.com