There is a claim that circulates in strategy discussions and gets repeated because it sounds right: it costs more to attack than to defend, so a country that spends on defence and reinforcement is richer than one that spends on offence.
It sounds right. It is also, as usually stated, unfalsifiable — no number, no unit, no boundary. This post puts a number on it, states where the number comes from, and says which part of the number is the weakest.
First, a correction
The working figure was 500 billion dollars a year. In Spanish, “500 billones” — which is 500 trillion in English notation, and 500 trillion dollars is roughly sixteen times the GDP of the United States. Not a large number: an impossible one.
So: US$500 billion per year, about 1.7% of US GDP. That is large and plausible. It is roughly what the US spends on defence, which is the scale on which a strategic reallocation could actually matter.
The five components
| Component | US$ bn/yr | Nature |
|---|---|---|
| Multiplier differential — defence procurement circulates domestically, offence projects leak abroad | 200 | Arithmetic, one assumption |
| Offensive operations — forward basing, expeditionary logistics, force protection | 180 | Estimate |
| Debt service on the borrowing that funded prior interventions | 60 | Estimate |
| Trade disruption — the cost of the instability that offence generates | 70 | Estimate |
| Reconstruction — what has to be rebuilt afterwards | 40 | Estimate |
| Total | 550 | rounded down |
Only the first line is arithmetic. The multiplier differential is 400 billion of reallocated spending times a 0.5 domestic-multiplier gap. Everything below it is the author’s estimate, and it is labelled as such rather than dressed up as measurement.
That honesty has a cost, and it should be stated: because four of five components are estimates, the result is not a measurement. It is a structured argument with a magnitude. Whether that is worth publishing depends on whether the structure survives being attacked — which is the next section.
The number depends on one parameter
The total is not evenly supported. It is supported mostly by the multiplier differential, and that term is a single coefficient:
| Domestic multiplier gap | Total (US$ bn) | % of US GDP |
|---|---|---|
| 0.25 | 450 | 1.50% |
| 0.375 | 500 | 1.67% |
| 0.50 | 550 | 1.83% |
| 0.625 | 600 | 2.00% |
The whole result moves between 1.5% and 2.0% of GDP on that one number. So the defensible statement is not “500 billion” — it is “of the order of half a trillion dollars a year, and here is the parameter it hinges on.”
And a warning that matters more than the number: the multiplier argument is circular if used alone. If defence spending is assumed to be more productive, it must be cheaper. The argument only has force when the multiplier differential is measured rather than assumed, or when supported by the other four components. Drop the differential entirely and the floor is 350 billion — still large, and built on components that do not depend on the contested coefficient.
What this does and does not show
It shows: the arithmetic of a shift from offence to defence is large and positive under assumptions that are stated, and the conclusion is robust across the plausible range of the one parameter it depends on.
It does not show that defence is always better. The paper concedes the point explicitly: the claim that productivity beats position holds where production is what is at stake. It fails for positional goods — territory, chokepoints, prestige — where the gain to one party is by construction a loss to another, and where a security dilemma can make offensive postures individually rational while collectively ruinous.
Confusing those two cases is how the argument becomes a slogan. A party that values position over production will not be persuaded by a cost model, because a cost model is a statement about production.
Where NEPTUNO fits — and where it does not
The framework’s claim is narrower than “defence is cheaper”. It is that projection requires immunisation, and immunisation requires governance — you cannot operate outward at a scale larger than the resilience of what you operate from.
That is a claim about ordering, not about arithmetic. This cost model is offered as consistent with it, not as evidence for it. An economic estimate does not validate an operating framework, and the paper says so in as many words. What would validate it is a case where projection without immunisation failed in a way the ordering predicted — and that is a different paper.
The point of writing it down
A figure that cannot be checked is a figure that cannot be argued with, and a claim that cannot be argued with is not a contribution — it is a position. Half a trillion dollars a year, one parameter, four estimates and a stated boundary can be attacked, and being attackable is what makes it worth reading.
That is the standard the question deserves. Applied to our own claims first.
Formal version: WP 17 — Defence Is Cheaper Than Attack: A Cost Accounting of Offensive Versus Defensive Posture
This is the readable version. The paper writes out all five components, runs the sensitivity analysis, and has a section listing the conditions under which the result would be wrong.