Military leaders often enter conflicts believing they can control their size, duration and intensity. A new statistical analysis led by researchers at the University of Colorado Boulder suggests that such confidence is frequently misplaced. Wars can expand through a chain of unpredictable decisions, alliances and battlefield reactions, turning limited confrontations into prolonged and far more destructive conflicts. The study finds that escalation is not an unusual outcome of war but a persistent risk embedded in the process itself—especially when fighting occurs between nations.
History offers repeated examples of leaders underestimating how far a conflict might spread. General Douglas MacArthur expected American troops to return from the Korean War by Christmas, only months after the fighting began. Instead, Chinese forces entered the war that winter, transforming the conflict and extending it for years. More recently, President Donald Trump described the war with Iran in March as a “short-term excursion,” yet the conflict has continued for more than five months, with periods of renewed escalation. Such episodes illustrate the danger of treating war as a finite operation with a predictable endpoint.
The research, published in Science Advances, examines how the human cost of wars changes from year to year. Led by computer scientist Aaron Clauset, a professor at CU Boulder, the study analyzes conflicts that took place primarily between 1946 and 2008. Rather than measuring war only by its final death toll, the researchers tracked the number of soldiers killed in battle during each year of a conflict. This year-by-year approach allowed them to study escalation and de-escalation as dynamic processes rather than as simple characteristics of wars that had already ended.
The results challenge the assumption that wars naturally become less intense with time. In large civil wars, violence often tends to decline as combatants lose resources, territory or the capacity to sustain major offensives. Wars fought between countries, however, show a persistent risk of continuing escalation. Across the conflicts examined, the researchers found no universal tendency for wars to become steadily larger or steadily smaller. Instead, violence fluctuated, sometimes falling for a period before rising again with surprising force.
The statistical pattern was striking. Each additional year of fighting carried roughly a 10 percent chance that the severity of the conflict would double, according to the analysis. There was also approximately a 1 percent chance that the level of violence would increase tenfold. These probabilities do not mean that every war follows a predictable mathematical trajectory, nor do they imply that escalation is inevitable. They describe a recurring risk across a broad historical record. Even when the chance of a dramatic increase is relatively small in any single year, the danger accumulates as a conflict continues.
The findings also suggest that a war’s early appearance offers little reassurance about its eventual scale. Conflicts that begin with relatively low losses are not necessarily destined to remain limited, while those that begin with severe casualties do not automatically grow into the largest wars. Duration alone is not sufficient to explain why some wars become catastrophic. Instead, the researchers identify escalation dynamics—the interactions and feedback loops that develop during fighting—as the primary mechanism that can drive an armed conflict toward enormous levels of destruction.
Those dynamics can emerge from events that are difficult to anticipate. A new ally may enter the war, opening another front or supplying weapons and troops that alter the balance between combatants. Fighting may spread into a neighboring region, while a military setback may lead one side to launch a larger offensive. Political leaders can also abandon earlier limits when they believe commitments are no longer credible. A ceasefire may collapse, peace negotiations may fail, or an attack intended to pressure an opponent may provoke retaliation. Each decision changes the strategic environment in which the next decision is made.
This process creates what Clauset describes as a ratcheting effect. Once violence increases, previous agreements and expectations may become harder to maintain. Rising casualties can make leaders less willing to compromise, while battlefield losses can produce demands for retaliation. The resulting actions may intensify the conflict further, destabilizing the commitments that had previously kept violence contained. In this sense, war behaves like a volatile system in which violence can build and feed back on itself. A conflict with only two initial participants and modest losses can therefore acquire new actors, new objectives and a much larger human cost.
The study does not provide a forecasting tool capable of predicting exactly when a war will escalate or how many people will ultimately die. Its importance lies instead in clarifying the limits of prediction. Policymakers who describe a conflict as small, controlled or short-term may be ignoring the statistical risks revealed by historical data. Clauset argues that the findings should encourage leaders to treat every armed conflict as carrying the possibility of substantial escalation, regardless of its initial intensity. More detailed information from individual battles, military orders and political negotiations could eventually help researchers identify the mechanisms behind escalation with greater precision. For now, the historical record delivers a stark warning: once war begins, its final scale may be determined less by the intentions that launched it than by the unpredictable chain of reactions that follows.
Subject of Research: People
Article Title: Escalation dynamics and the severity of wars
News Publication Date: 5-Aug-2026
Web References: Science Advances article; Aaron Clauset, University of Colorado Boulder
References: Science Advances, DOI: 10.1126/sciadv.aeb1705
Image Credits: Signal Corps, U.S. Army
Keywords
War escalation, conflict dynamics, military science, political violence, statistical analysis, civil wars, interstate wars, battle deaths, peace negotiations, international security
