You Cannot Install a Democracy

TL;DR

A statistical examination of seventy foreign-imposed regime changes across the twentieth century found that target states saw little democratic gain compared with similar states that did not experience regime change, and were no more likely to end up as consolidated democracies.¹² What predicts success isn’t the intervener’s intentions or competence — it’s three conditions inside the target: economic development, societal homogeneity, and prior experience of representative government.¹ Which produces the trap: the states weak enough to invade are systematically the ones least likely to democratise afterwards.² Notably, whether the intervener was the United States or another democracy made no difference to the outcome.²

A ballot going into a box. The act is easy to install; the conditions that make it mean something are not. Photo: W.carter, CC BY-SA 4.0.³

The Premise Worth Testing

For most of a century, a particular argument has justified a great deal of force: that the obstacle to freedom in a given country is its regime, and that removing the regime therefore opens a path to democracy. The argument has been made in good faith by many people, and it has an appealing structural simplicity — a bad thing is in the way, so remove the bad thing.

It’s also an empirical claim, and an unusually testable one. There have been many interventions. We can look at what happened.

What Happened

The most systematic answer comes from Alexander Downes and Jonathan Monten, who examined seventy instances of foreign-imposed regime change in the twentieth century. The headline result: states that experienced imposed regime change showed little democratic gain compared with similar states that did not, and were no more likely to become consolidated democracies.¹²

The comparison is what makes this a finding rather than a mood. It isn’t “intervened-in countries often aren’t democracies,” which would prove nothing, since such countries usually weren’t democracies beforehand either. It’s that they didn’t do measurably better than comparable states left alone. The intervention, on average, did not add democracy.

Two further results sharpen it considerably. First, “decapitating” a regime — removing the leader and little else — rarely produces democratic change; interventions that actually build institutions do better.² Second, and more damaging to a common instinct: the intervener’s identity made no difference. Whether the intervening power was the United States or another democracy had no bearing on whether the target democratised.²

That second point deserves emphasis, because it cuts against both the triumphal and the cynical versions of the argument. The problem is not that a particular country is bad at this. It is that the thing itself does not reliably work.

What Actually Decides It

The useful part of the research isn’t the negative result but the specification of when democratisation does follow. Downes and Monten identify three conditions in the target state:¹

• Economic development — wealthier target states are substantially more likely to sustain democratic institutions.
• Societal homogeneity — ethnically and socially divided societies face significantly higher obstacles.
• Prior experience of representative governance — countries with a democratic past are far more likely to sustain a democratic present.

Note what is absent from that list: the intervener’s sincerity, its military competence, the size of its aid package, the elegance of the constitution it drafts. Their conclusion is that implementing pro-democratic reforms such as sponsoring elections is not sufficient to induce democratisation; without favourable local conditions, interveners meet with little success.¹

There’s a related and clarifying distinction: restoration works better than creation. Returning a previously democratic regime to power — Western Europe in 1945 — has a far better record than manufacturing a democracy where none existed.² Which reframes the celebrated cases. Post-war Germany and Japan are routinely cited as proof that imposed democratisation works; the research treats them as exceptional circumstances rather than typical outcomes.¹ They were developed, relatively homogeneous, and had prior constitutional experience. They satisfied all three conditions before anyone landed.

The Trap in the Middle

Put those two findings together and you get something close to a structural impossibility, and it is the most important thing in this literature.

The states that satisfy the three conditions — wealthy, cohesive, with democratic memory — are generally strong states that nobody can easily invade, and which frequently don’t need external help to democratise. The states that are weak, poor and internally divided enough to be plausible intervention targets are precisely the states where the conditions for democratisation are absent.²

So the selection effect runs against the policy at the point of target choice. Interveners are not choosing badly among available options; the option they want — an invadeable country that will then democratise — is close to a null set. Downes and Monten’s own policy conclusion is accordingly deflationary: policymakers should scale back expectations that democracy will flourish after imposed regime change.¹

The Ledger Nobody Balances

If the expected democratic gain is near zero on average, the cost side stops being a footnote. The same analysis notes the Iraq figures explicitly: roughly 4,500 US military deaths, more than 100,000 Iraqi civilian deaths, and about $4.5 trillion.²

The point isn’t that costs are high — everyone concedes that. It’s that the standard justification implicitly treats the democratic benefit as large and probable, when the evidence says it is small and conditional. A cost-benefit argument in which the benefit term is essentially unsupported isn’t a cost-benefit argument. It’s an assertion with arithmetic attached.

Where the Argument Is Genuinely Contested

Two honest qualifications, and the first is substantial.

This finding has been challenged in print. William Nomikos published a critique of Downes and Monten in International Security, with a reply from the authors⁴ — and the core methodological dispute is worth understanding, because it isn’t trivial. It concerns case selection: which of the seventy interventions should count. If you restrict attention to cases where the intervener explicitly intended to install a democracy, rather than all regime changes including those aimed at installing a friendly dictator, the success rate looks materially better. I could not access the correspondence itself to verify the specific figures, so I’ll state only the shape of the disagreement rather than numbers I haven’t read: whether imposed democratisation looks futile or moderately effective depends substantially on whether you judge it by all regime changes or only by the ones that were actually trying.

That’s a real objection, and it should reduce confidence in the strong version of the claim. It’s also worth noting it doesn’t rescue the policy argument entirely — the three-conditions finding survives it, and the selection trap does too.

Second, “no average effect” doesn’t mean “never works.” Averages hide variance, and a policy with a near-zero mean can still contain genuine successes. The research’s own logic supports this: where the conditions hold, outcomes improve. The conclusion isn’t that intervention can never produce a democracy. It’s that whether it does is mostly determined by facts about the target that were true before anyone intervened — which means the causal credit belongs substantially to the country, not the intervener.

Which is the deflationary version of a very old claim about exporting political forms. Not that other peoples are unsuited to democracy — a genuinely ugly argument the evidence here does not support, since the conditions are economic and historical rather than cultural or racial. Only that democracy appears to be something a society arrives at, on the basis of conditions it has accumulated, rather than something that can be delivered to it by force from outside. The intervener’s role, on this evidence, is far smaller than the intervener’s rhetoric.

---

Footnotes

1. https://www.belfercenter.org/publication/forced-be-free-why-foreign-imposed-regime-change-rarely-leads-democratization ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

2. https://www.belfercenter.org/publication/why-foreign-imposed-regime-change-rarely-path-democracy ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷ ↩⁸ ↩⁹

3. https://commons.wikimedia.org/wiki/File:Ballot_dropped_into_ballot_box_2024_Swedish_EU_election_at_St%C3%A5ngen%C3%A4sskolan%2C_Brastad.jpg ↩

4. https://www.belfercenter.org/publication/correspondence-reevaluating-foreign-imposed-regime-change ↩