The Levels-of-Analysis Problem

In brief

In Man, the State, and War (1959), Kenneth Waltz (1924 to 2013) asked a deceptively simple question: where do the causes of war lie? He answered that theorists had located them in three distinct places, which he called images. The first image puts the cause in human beings: in aggression, greed, or folly. The second puts it in the internal character of states: whether they are democratic or despotic, capitalist or socialist. The third puts it in the structure of the international system itself, which has no government above sovereign states and so leaves each to fend for itself. Two years later, J. David Singer (1925 to 2009) turned this observation into a methodological rule. In a 1961 essay in World Politics he argued that the choice of level is not a detail you settle along the way. It is a decision made before the analysis begins, and it governs what you will describe, what you will explain, and what you will be willing to count as a cause.

The full treatment

The problem it answers

International politics is dense with events that seem to demand explanation: a war breaks out, an alliance forms, an arms race escalates. The natural instinct is to ask why. But "why" has no single address. You can explain the outbreak of the First World War by pointing to the personalities of the men who ordered mobilization, or to the militarism and domestic pressures inside Germany, Austria-Hungary, and Russia, or to the structure of a multipolar system in which alliances locked great powers into a spiral no single actor controlled. Each answer is coherent. Each isolates a different kind of cause. The levels-of-analysis problem is the recognition that these are not rival answers to one question but answers to different questions, and that the analyst chooses which question to ask by choosing a level.

Waltz's three images

Waltz drew the three images from the history of political thought rather than inventing them. The first image, human nature, he traced through figures such as Augustine, Baruch Spinoza, and Reinhold Niebuhr: if people are flawed, so are the things they build, and war is the outward form of an inward corruption. The trouble, Waltz noted, is that a constant (human nature) cannot by itself explain a variable (why war sometimes and peace other times).

The second image locates the cause in the internal organization of states. Here Waltz placed liberals who believed that republics would be peaceful, and Marxists who believed that capitalism drove states to imperial war. Change the internal order, the second image promises, and you change the external behavior. But states of every internal type have gone to war, which suggests the internal character is not the whole story.

The third image, the one Waltz would spend his career developing, locates the cause in the system. Because there is no authority above states, each must provide for its own security, and no state can be certain of another's intentions. Even states that want only to survive are pushed toward arming, balancing, and suspicion. On this account war has a permissive cause that sits above any particular ruler or regime: anarchy itself. This third-image argument became the seed of structural realism, which Waltz set out fully in Theory of International Politics (1979).

Singer's formalization

Waltz described three places to look. Singer made the sharper claim about method. In "The Level-of-Analysis Problem in International Relations" (1961) he argued that a scholar must choose between two principal levels: the international system as a whole, or the national state as the unit. Each has costs. Explain everything from the system and you gain generality but flatten the states into interchangeable billiard balls, losing the differences between them. Explain everything from the state and you recover richness and detail but lose the ability to say anything general about the system as a system. Singer's warning was against sliding unconsciously between the two: describing behavior at one level while borrowing explanations from another, so that the argument seems to cover everything while actually resting on a confusion. The level, he insisted, must be chosen deliberately and held consistently.

Why the choice comes first

The deep point is that selecting a level is a prior methodological commitment, not a conclusion drawn from data. Before you inspect a single fact about a war, the level you have adopted has already sorted the world into foreground and background. At the system level, the ambition of a particular chancellor is background noise; the distribution of power is the cause. At the individual level, the distribution of power is the fixed stage; the chancellor's misjudgment is the cause. Neither is wrong. But each makes invisible what the other makes central. This is why the problem is genuinely a problem and not merely a menu: whatever you count as explanation depends on a choice you made before explaining anything, and that choice is rarely forced on you by the evidence.

A worked example

Take the Cuban Missile Crisis of 1962. A first-image account attends to the fears, habits, and reasoning of John F. Kennedy and Nikita Khrushchev, the individuals whose choices under stress steered the world toward or away from catastrophe. A second-image account looks inside the two governments: the pull of the U.S. military establishment, the workings of the Soviet Politburo, the bureaucratic routines that shaped what options even reached the top. A third-image account treats the crisis as a predictable episode in a bipolar contest between two superpowers with nuclear weapons, in which the identities of the men involved matter less than the structure they inhabited. Graham Allison's Essence of Decision (1971) made exactly this move explicit, running the same event through several competing models and showing how each produced a different story and different lessons. The crisis did not choose the level. The analyst did.

Lineage

The three-image framing is Waltz's, but the underlying intuition is old. Thucydides already distinguished the immediate grievances that sparked the Peloponnesian War from its deeper cause, which he located in the rise of Athenian power and the fear it produced in Sparta, an early gesture at a structural, rather than personal, explanation. Waltz's contribution in 1959 was to organize the tradition of thinking about war into three clean categories and to press the point that a constant cannot explain a variable. Singer's 1961 essay then lifted the insight out of the study of war specifically and turned it into a general rule for the discipline. Barry Buzan, in a 1995 reconsideration, distinguished the level at which you locate causes from the unit whose behavior you are explaining, arguing that the two had been run together. The vocabulary of "levels" has since become common ground, invoked across otherwise opposed schools of international relations theory.

The strongest case for it

The framework's power is that it disciplines argument. Once you see that "why did the war happen" fractures into three questions, you can no longer treat a first-image answer as if it refuted a third-image one; they are not competing for the same ground. This blocks a common fallacy, the reductionist slide, in which a rich account of one leader's psychology is offered as if it disposed of the structural pressures every leader in that position would have felt. The scheme also works as a check on one's own reasoning: Singer's demand that you name your level and hold it consistently guards against explanations that quietly help themselves to whichever level is convenient at each step. And it is ecumenical. A realist, a liberal, and a constructivist can disagree fiercely about which level does the real causal work while agreeing on the map of possible levels, which makes genuine debate possible rather than mutual incomprehension.

The strongest case against it

The framework has real critics, and their objections cut in different directions.

The first charge is that the levels are not as separable as the scheme implies. Alexander Wendt argued that agents and structures are mutually constituted: states are shaped by the system, but the system is in turn made and remade by what states do, so treating one level as cause and the other as background misdescribes a relationship that runs both ways. To freeze either level as the fixed stage is to lose the very process by which international politics changes.

A second charge, associated with Robert Jervis and the tradition of foreign-policy analysis, is that the neat separation encourages theorists to ignore how the levels interact. Real outcomes, on this view, come from the meeting of structural pressures with domestic politics and individual perception, and a method that insists on picking one level cleanly discards precisely the interaction effects that explain the interesting cases.

A third line questions whether three (or two) is the right number at all. Once you admit the bureaucracy, the interest group, the individual decision maker, and the transnational network as distinct sites of causation, the tidy triad starts to multiply, and it is not obvious where the list of legitimate levels should stop. Buzan's own reconsideration reflects this worry: the more carefully one specifies levels and units, the more the original elegance dissolves into a finer and less settled taxonomy.

Finally, critics note that the scheme describes where causes might live without telling you which level actually does the causal work in a given case. It is a filing system, not a theory. Waltz's own later structural realism supplies a substantive answer (the system dominates), but that answer is contested, and the levels framework by itself remains silent on the question that matters most.

Where it stands now

The levels-of-analysis distinction is now part of the furniture of the field, taught in the first weeks of nearly every international relations course and used as a shared vocabulary even by scholars who reject Waltz's conclusions. Its descriptive scheme survived better than the strong structural theory Waltz built on top of it: many who use "first, second, and third image" as shorthand would not call themselves structural realists. The main developments since have been refinements rather than replacements, sharpening the difference between a level of explanation and a unit of analysis, and folding in the constructivist point that levels co-constitute one another rather than stacking neatly. The core insight has proved durable precisely because it is modest. It does not tell you what causes what. It tells you that before you answer, you have already chosen where to look, and that the choice deserves to be made in the open.

Test yourself

Pick a recent international event you have an opinion about: a war, a sanction, a summit. Notice which level your explanation defaulted to. Did you reach first for the leaders and their motives, for the kind of regimes involved, or for the structure of power around them? Now try to tell the same story honestly from one of the levels you skipped. If the event looks like a different event, you have felt the problem Singer was pointing at.