James Fearon’s 1995 paper remains one of the most assigned texts in international relations

In his 1995 article, Fearon seeks to answer why wars happen if they are so costly.

He begins by outlining three possible explanations:

1. Irrational - people, including leaders, do not act in their best interest because they misunderstand the consequences of their actions, due to factors like emotions or cognitive biases.
2. Rational - leaders who decide to go to war have personal self-interest (private incentives) that is not aligned with the population and are isolated from negative consequences.
3. Rational - leaders consider the negative consequences of war yet still decide to engage in it.

The article seeks to explain the mechanisms that result in the third scenario. What prevents rational actors from coming to an agreement, which is almost always preferable to war for both sides?

Definitions and assumptions:

• Rational actor = an actor that has stable preferences, seeks to maximize them, and incorporates new information
• Unitary state actor = a state acts as a single decisionmaker (domestic politics and deliberations are ignored or collapsed)
• States are risk averse and prefer non-violence

What doesn’t explain war initiation

Fearon first examines existing explanations in IR literature about outbreak of war, revealing some of the factors to be incomplete, insufficient, or special cases of something else:

1. Anarchy - lack of international enforcement, i.e. no world police
2. Expected benefits of war greater than expected costs
3. Rational preventive war
4. Rational miscalculation due to lack of information
5. Rational miscalculation or disagreement about relative power

He argues that (1) and (2) do not explain war because they do not, by themselves, explain why rational actors don’t come to an agreement that is more preferable to fighting.

(3) is better understood as a special case of a commitment problem.

(4–5) address mechanisms for war, but they do not fully explain why rational actors fail to communicate credibly or engage in diplomacy.

What mechanisms explain why rational actors initiate war?

Fearon proposes three explanations for why rational actors decide to initiate war. These explanations are not necessarily novel, but rather mechanisms that expand on or reclassify existing ones:

1. War due to private information and incentives to misrepresent
2. War as a consequence of commitment problems: at least one side has concerns about another side honoring agreements and future behavior
3. Issue indivisibilities (less common)

(1) must have two components:

• First is uncertainty or misperception about an opponent’s strength or resolve. However, that’s not enough, because a rational actor can engage in different forms of communication to obtain that information.
• The second component is failure to resolve these private information problems. One possibility is private information unavailable to the other side, combined with bluffing or misrepresentation. One incentive to bluff is being weaker than the opponent believes. Another incentive to misrepresent your true strength or resolve is that you think the other side will not believe you.

An example of asymmetric information contributing to war is Russia-Japan in 1904. Russia pursued an aggressive policy in East Asia and believed it wouldn’t face significant resistance from Japan. Japan had stronger military capabilities and resolve than Russia expected, but saw no credible way to convince Russia.

Asymmetrical information (1) can reduce the bargaining range.

(2), commitment problems, have several types:

• Preemptive war or first-striker advantage, with no credible agreement that can be made to avoid it
• Preventive war - one side prefers war now because it fears a less desirable outcome in the future. Another side is not trusted to uphold a bargain in the future and may have an incentive to renege (e.g. Germany and Austria’s concerns about Russia in 1914).
• Appeasement and strategic concessions - reaching a settlement is not preferable to war, because another side may ask for more in the future (either because the other side is already strong, and/or the concessions will give them even more of an upper hand). Example: Winter War between Finland and Soviet Union in 1939.

Commitment problems make agreements less credible. This is especially true when there’s no outside enforcement mechanism. It also contributes to prolonged civil wars.

Conclusion

When a non-violent settlement is preferable to each side, there is a possibility of avoiding war. When a rational actor nevertheless initiates war, it must stem from asymmetrical information or mistrust about the other side’s commitment to honor any potential agreements or future behavior. The mechanisms can also interact and compound.

Note: a formal way to express bargaining is provided at the end.

What is this good for? And some limitations

Prediction

Fearon presents a comprehensive analytical framework that can be used for retrospective analyses of why rational actors engage in wars. However, as Fearon himself said elsewhere, prediction is still a foundational problem. Existing prediction models rely on structural causes proposed by Fearon and others - incredibly useful for knowing where to look for conditions that increase risks of war. But the “why now?” question remains a challenge. For example, see “War is in the Error Term” by Gartzke (1999).

Russia-Ukraine viewed through the rationalist lens

Private (intangible) incentives such as ideology or status of an unchecked leader and failure to resolve information problems have been raised as explanations for Russia’s 2022 invasion of Ukraine. Another contributing factor may be a commitment problem, though not related to security. Instead, the future concern, from Russia’s view, may be that Ukraine becoming increasingly democratic and aligned with the West. The last thing worth mentioning is that autocrats may dislike democracies because they illustrate that a better way is possible. This can affect regime stability, though less likely in this case.

In this scenario, the unitary state actor collapses to a single unchecked leader.

(Lack of) unifying models

Fearon speculates that Hitler is a possible exception to the assumption of rational actors being risk averse and not wanting to fight. In his excellent book Why We Fight, conflict scholar Chris Blattman likewise concedes that there's no unifying framework. It’s striking that even in retrospective analyses, we still don’t have strong unified models.

Given all the information that is already known about conflict, I wonder how much of the gap is a reflection about missed opportunities in theoretical development, synthesis, and multidisciplinary collaboration among scholars who study violence.

Note: an earlier version of this post was published here in July 2025.

---

Formal expression of bargaining

Suppose you have side A and side B, where:V_a is the value or benefit of war for side A, and V_b is the benefit of war for side Bp_a is the probability of side A winning, and p_b = (1-p_a ) is the probability of side B winningc_a is the cost of war for side A, and c_b is the cost of war for side B

The expected net benefit of war for side A is: E_a = V_a * p_a - c_a

The expected net benefit of war for side B is: E_b = V_b * p_b - c_b

Let T = 100 = total value of the issue at dispute (e.g. territory)t_a is the portion of T that goes to side A in a settlementt_b = 100-t_a is the portion that goes to side B in a settlement

Then, for a positive bargaining range to exist (i.e. preference for non-violent resolution):E_a + E_b < 100

Side A prefers a settlement if: E_a < t_a

Side B prefers a settlement if: E_b < t_b

Other ways to express the bargaining range, from the perspective of A:E_a < t_a < 100 - E_b, or the open interval (E_a, 100 - E_b)

Implications/TLDR:When a positive bargaining range exists, there is a possibility of a non-violent settlement.Asymmetrical information narrows the bargaining range.Commitment problems make agreements less credible, especially when there’s no outside enforcement mechanism.Costs of war do NOT necessarily have to exceed benefits, for one side to prefer a settlement. The calculation must also account for probability/optimism of winning.A non-violent settlement will NOT necessarily be fair.