The deepest divide in war is not between democracies and dictatorships, but between states free to leave and those fighting for survival.

In August 2021, the United States withdrew from Afghanistan after twenty years of war. Domestic exhaustion helped drive the decision. So did a mission whose objectives had grown murkier with each passing year, Washington’s strategic pivot toward China, and the collapse of the Afghan government itself. Six months later, Russia invaded Ukraine, and Kyiv faced no comparable menu of options. Negotiated settlement, territorial concessions, and a frozen conflict have all been discussed. None offers an easy escape if Ukraine intends to remain a sovereign, self-governing state.

The difference has less to do with who chose to fight than with what each country remained free to do once the fighting began—particularly whether it could stop. Great powers can often decide that a war has become too costly, too unpopular, or simply no longer worth the trouble. Weaker states confronting conquest rarely enjoy that luxury.

That is a more precise version of a familiar, if overstated, claim: Materially dominant states treat war as an instrument of policy, while weaker states experience it as a condition of survival. The distinction is broadly correct, but it needs two qualifications before it can carry much analytical weight.

The first concerns entry. Weak states choose to start wars all the time. Argentina invaded the Falkland Islands in 1982. Iraq invaded Iran in 1980. Serbia entered the wars of the 1990s by choice. Great powers, meanwhile, sometimes have war forced upon them, as the United States did after Pearl Harbor. What varies systematically with material power is not the decision to enter a war. It is the range of options a state retains once inside one—and, above all, its capacity to leave.

The second qualification concerns what strategic freedom actually buys a strong state. It does not guarantee freedom throughout an engagement. The United States learned that in Vietnam, Iraq, and Afghanistan. France learned it in Algeria. Britain learned it in Northern Ireland. Once a great power commits ground forces to a protracted conflict, occupation dynamics, alliance obligations, sunk costs, and the political price of admitting failure begin to narrow its choices. Strong states generally retain more options before committing to war and more viable exits after doing so. They do not retain unlimited freedom in between.

Those qualifications do not erase the underlying asymmetry. They merely locate it more accurately. James Fearon’s work on “audience costs” helps explain why democratic leaders, in particular, can pay a domestic political price for making public threats they fail to carry out. That price tends to compound as a war drags on without a recognizable victory. Elections approach. News coverage sours. Legislative coalitions fray. A war may remain militarily sustainable yet become politically impossible. A government can therefore withdraw while its forces still hold the field, preferring the humiliation of retreat to the mounting cost of staying.

Weaker states facing existential threats behave differently, although domestic politics never disappears for them. Ivan Arreguín-Toft’s research on asymmetric conflict suggests that when a country’s survival is genuinely at stake, its tolerance for backing down falls sharply because there is no safe version of defeat. Iran’s war with Iraq from 1980 to 1988 illustrates the pattern. Tehran absorbed losses that might have shattered a Western democratic coalition within months. Yet even that endurance had limits. Iran eventually accepted UN Security Council Resolution 598, the framework that ended the war, after Ayatollah Ruhollah Khomeini likened the decision to drinking from a poisoned chalice. Existential stakes can suppress political disagreement and enlarge a state’s tolerance for pain. They cannot abolish either politics or arithmetic. Eventually, even a government fighting for survival must weigh its aims against the cost of pursuing them.

The related assumption that only democracies face domestic constraints on warfare is also only half right. Authoritarian leaders answer to elite coalitions, party institutions, security services, and patronage networks capable of removing them by means other than an election. Research by China scholar Sheena Chestnut Greitens on domestic security, along with Jessica Weeks’s work on accountability in authoritarian regimes, shows that many autocrats face genuine domestic pressure—just not through the ballot box. Caitlin Talmadge’s comparative study of authoritarian militaries, drawing in part on the Iran-Iraq War, traces how fear of a coup can shape officer promotion, training, and command structures in ways that protect a regime at home while degrading its performance on the battlefield. The constraint is real. It is simply easy to miss if one looks only for elections, opinion polls, and hostile headlines.

This is also why treating authoritarian governments as interchangeable produces bad analysis. China’s campaign of repression in Xinjiang emerged from a highly institutionalized party-state with enormous administrative capacity. Russia’s wars in Chechnya reflected a more personalist system in which decisions were concentrated in fewer hands. Iran occupies a hybrid position: Elected bodies exercise limited authority alongside far more powerful unelected institutions, including the Islamic Revolutionary Guard Corps and the supreme leader. All three systems are authoritarian, but the structures that constrain their leaders—and the mistakes those structures encourage—are markedly different.

One variable remains missing, and it may be the most consequential: the stakes. Capability, political will, and institutional constraints describe an actor. Stakes describe what that actor believes is actually on the table, and they vary independently of regime type. The United States fighting a war of choice in Vietnam and the United States responding to Pearl Harbor were the same country, operating under the same Constitution. Yet the wars followed wholly different political logics because Americans did not believe the stakes were comparable. Israel regarded the 1973 war as a struggle for national survival. Saudi Arabia’s intervention in Yemen was a contest over security and regional influence, not the kingdom’s continued existence. Any useful model therefore needs four variables rather than three: capability, constraint, political will, and stakes.

Two further objections deserve direct answers because both are valid and neither overturns the argument. Democracies are not inherently restrained in how they wage war. The firebombing of Tokyo, Britain’s counterinsurgency in Kenya, and France’s conduct in Algeria all came from democratic governments willing to inflict extraordinary violence. Jason Lyall’s research on counterinsurgency outcomes suggests that such brutality is hardly an aberration. Democratic restraint, when it operates at all, works primarily through procedure and time rather than conscience. Elections and legislative oversight may determine how long a democracy can sustain a war and what its public will tolerate. They offer no assurance that the warfare itself will remain humane.

Authoritarian regimes, for their part, remain capable of catastrophic strategic errors. The Soviet invasion of Afghanistan and Russia’s invasion of Ukraine are obvious evidence. Yet their internal constraints remain real. What differs is how those constraints align with sound decision-making. A system without a free press, organized opposition, or independent institutions capable of challenging a leader’s assumptions before troops cross a border generates its own distinctive failures. Those failures exist alongside the errors democracies make; they do not replace them.

Taken together, capability, constraint, political will, and stakes explain far more about how countries fight—and how they eventually stop—than regime type alone. Regime type determines the form a country’s constraints take. It does not determine whether constraints exist. Nor does raw power reveal how much sacrifice a government will accept before changing course.

The most useful question, then, comes before any tally of tanks, troops, or treasuries: Whose survival is actually at risk when a war begins—the state’s, the leader’s, or that of a governing coalition answerable to no electorate at all? The answer does more than explain why wars start. It helps determine how savagely they are fought, how long they endure, and who retains the privilege of walking away.