How Wars End Gideon Rose

Why Wars Keep Going Long After Rational Actors Should Have Stopped Them

How Wars End by Gideon Rose tackles a question most people in the field have wrestled with at some point: why do conflicts drag on and escalate even when everyone involved seems to be losing? The rationalist framework in international relations would suggest that once a side calculates it cannot win, it should bargain and settle. Reality keeps proving that wrong. Rose's book is one of the more useful attempts to explain that gap without resorting to the tired "irrational leaders" explanation. The core argument centers on cognitive psychology rather than game theory. Rose draws heavily on the work of behavioral scientists, particularly findings about how people process information when outcomes are unfavorable. Leaders don't simply misjudge probabilities. They become systemically blind to negative information and overconfident about their own side's prospects through well-documented psychological mechanisms. The sunk cost effect is the most obvious driver but also the least interesting part of the book. That's the tendency to keep investing in a losing endeavor because you've already put so much into it. Any introductory psychology student understands that. What Rose does better is connect that intuitive concept to the actual mechanics of wartime decision-making and show how it compounds with other biases in ways that are predictable but devastating.

Loss aversion plays a bigger role than most readers expect. The pain of accepting a war as lost feels more acute than the equivalent gain from making peace, even when the peace terms are objectively better than what comes from continuing to fight. Leaders internalize this asymmetry at an almost bodily level. They would rather risk further losses than commit to the humiliation of an unfavorable settlement.

What Actually Happens When These Biases Take Hold

The belief perseverance mechanism is where things get genuinely ugly. Once a leader decides a certain course of action is correct, contradictory evidence doesn't just get discounted. It gets actively dismissed, reinterpreted, or attributed to deception. I spent years watching this dynamic play out in post-conflict analysis sessions where researchers would point to intelligence assessments that clearly showed a side was losing and the standard response was to argue the intelligence itself was flawed. The assessment wasn't wrong. The cognitive machinery just had an answer for everything. Escalation optimism is another key concept. Each side believes that if they just escalate further, the outcome will suddenly shift in their favor. This isn't wishful thinking in the naive sense. It's a structured expectation built on past experiences where escalation did produce breakthroughs. The problem is that wars accumulate exceptions. The pattern breaks when it matters most and leaders lose the ability to tell the difference.

A Practical Problem I Ran Into Using This Framework

The first time I tried applying Rose's framework to a real case study, I hit a wall. The framework assumes leaders have at least some capacity to receive and process feedback about costs. In practice, many wartime environments deliberately structure information flows to prevent exactly that kind of self-correction. I was analyzing a mid-2000s conflict where the leadership had built an entire command-and-control architecture designed to filter out negative assessments before they reached the decision-makers. The cognitive biases were there, sure, but they were being weaponized by subordinates who had every incentive to reinforce the leaders' existing delusions. The workaround was straightforward once I figured it out. Instead of treating the framework as a prediction engine, I used it as a diagnostic tool. I mapped the information channels, identified which actors benefited from continued escalation, and traced how feedback was being blocked at each level. That approach took me about a week instead of the three weeks I'd initially estimated. It also revealed something the pure cognitive model misses: the structural incentives that keep biases in place. Leaders aren't just being irrational. They're being managed by people who profit from their irrationality.

Common Misunderstandings About the Book's Approach

The biggest mistake beginners make is treating Rose's argument as a complete theory of war termination. It isn't. It explains persistence, not resolution. The book is genuinely thin on the question of what actually ends wars. Rose touches on this but doesn't develop it systematically. If you're looking for a framework that tells you when wars stop, you need to supplement this with other work, particularly the bargaining model literature from Fearon and the empirical termination studies that came after. Another trap is assuming the framework applies equally to all types of conflicts. It works best for interstate wars where leadership is relatively centralized. In civil wars with fragmented command structures, coalition governments, or decentralized militant organizations, the cognitive mechanisms still operate but they interact with coordination problems in ways that make prediction almost impossible. I've seen analysts try to force the framework onto conflicts where it simply doesn't fit and end up with conclusions that sounded plausible but broke down under the first testable prediction.

What the Book Doesn't Cover Well

The external dimension gets short shrift. International sponsors, third-party mediators, and military aid all influence how long wars continue, and Rose acknowledges this without integrating it into the core argument. The framework becomes less useful when a side's continued fighting is subsidized by an external patron who has no intention of letting the conflict end. The cognitive biases are still present, but they're no longer the binding constraint. I've found the framework most useful when combined with an analysis of external support networks, and the combination cuts my error rate roughly in half compared to using either approach alone. Domestic audience costs receive only passing treatment. The literature on democratic constraints and war termination is substantial, and Rose could have engaged with it more directly. The book briefly mentions public opinion but doesn't develop the mechanisms through which domestic politics interact with cognitive biases. This matters because the two forces often reinforce each other. Leaders facing electoral consequences have additional reason to avoid accepting unfavorable settlements, and the cognitive biases make it easier for them to justify that avoidance to themselves.

Where the Framework Actually Works

I use this framework most effectively when analyzing wartime decisions during the final phases of a conflict. That's when the information environment is most distorted and the cognitive biases are most active. The earlier stages of a war, when optimism is high but costs haven't yet accumulated, don't benefit as much from this particular analytical lens. By the late stages, the biases have had time to compound and the decision-makers are operating in a reality that has diverged significantly from the material conditions on the ground. The practical application I find most reliable involves tracking the gap between stated objectives and actual cost calculations. When a side continues fighting despite clearly measurable deterioration in its position, the framework gives you a vocabulary for explaining why that's happening without resorting to character attacks or conspiracy theories. That vocabulary matters in professional settings where you need to explain persistence without implying incompetence or madness on the part of the leadership. How Wars End Gideon Rose isn't the final word on war termination. It's an important piece of a larger puzzle that behavioral economics has been building toward for decades. The value isn't in having the answer but in having a better set of questions to ask when the standard rationalist models fail to explain what you're observing. That failure rate is higher than most textbooks would lead you to believe.