The Economics of Prevention in an Age of Polycrisis

Executive Summary

Prevention is almost always appraised against doing nothing. Its actual alternative is emergency response — which is already being paid for, repeatedly, at the worst available price.

The economics of prevention have been settled for years. The Global Commission on Adaptation put US$1.8 trillion of investment against US$7.1 trillion in net benefits; the World Bank's Lifelines found that specifying resilience at the design stage adds 3 to 5 per cent to an asset's upfront cost while avoiding losses worth 50 to 100 per cent of its value. An earlier report in this series assembled that benefit-cost evidence and asked why the money still does not move.

This report asks a narrower and, we think, more tractable question. Not what is prevention worth? but why does the demonstrably more expensive option keep winning? The answer is not that decision-makers have weighed the returns and declined. It is that prevention and response are bought through different budget instruments, tested against different hurdles, accounted for in different ways, and measured on different outputs — and on every one of those four dimensions, response has the easier path.

Those first two figures are the whole argument in miniature, and they are a cost ratio rather than a benefit-cost ratio, which is what makes them unusually hard to argue with. Resilience is not expensive. It is cheap at the moment of design and ruinous at the moment of repair, and almost all of the world's disaster spending happens at the second moment.

1. The Comparison That Is Never Made

Every appraisal of prevention asks whether it beats the alternative uses of scarce capital. Almost none asks whether it beats the response spending that is going to happen anyway.

A finance ministry considering a flood-protection programme weighs it against a school, a road, a clinic. That is the correct discipline for capital rationing, and it is also the wrong frame, because it treats the money as the only scarce thing and treats inaction as costless. Inaction is not costless. The floods will arrive; the relief will be paid; the reconstruction will be financed. Those payments are as certain as anything in public finance, and they appear in nobody's appraisal of the flood-protection programme.

Set the two side by side properly and the comparison changes character. Prevention is a capital expenditure: it is paid once, it produces a durable asset, that asset sits on the balance sheet, and it lowers a liability for the whole of its life. Response is an operating expense: it is paid every time, it buys consumption — food, shelter, medicine, temporary works — and when it is finished there is no asset and the underlying exposure is exactly what it was before.

That last clause is the important one. Reconstruction, by definition, restores the status quo ante. The status quo ante is the condition that produced the loss. A response programme that succeeds perfectly returns the system to the state in which the next event will do the same damage again. This is why response spending recurs and prevention spending does not.

2. The Response Premium

Buying protection after the fact costs more than buying it before, for five specific and separable reasons. Together they are the response premium.

It is worth being precise about why the same protection is more expensive when bought late, because each component implies a different remedy.

Table 1 — The five components of the response premium

The five are cumulative rather than alternative: a single reconstruction programme typically pays all of them at once.

The crisis-financing component deserves particular emphasis, because it compounds. A country that borrows to reconstruct does so on terms set by its post-disaster condition, and the resulting debt service constrains its capacity to prevent the next event. Response spending therefore does not merely fail to reduce future exposure; it can actively reduce the fiscal room in which prevention would have to be bought.

3. Why Polycrisis Breaks the Response Model

Responding to shocks as they arrive is a coherent strategy under one set of assumptions about how shocks behave. Those assumptions no longer hold.

The ex-post model is not stupid. It is the rational answer to a particular risk profile: shocks that are rare, independent and bounded. Under those conditions a state can hold a contingency reserve sized to roughly one event, draw it down when the event comes, and rebuild it in the quiet years that follow. Prevention against every possible hazard would be wasteful; insurance-like self-financing is efficient.

Each of those three assumptions fails under polycrisis. Shocks are not rare: the frequency of weather, climate and water-related events has risen sharply over the past half-century. They are not independent: a drought becomes a harvest failure, a price spike, a fiscal crisis and a health emergency, and the World Economic Forum's Global Risks Report has for several years described a landscape defined less by any single peril than by the connections between them. And they are not bounded, because a correlated cluster can exceed any reserve a developing economy could plausibly hold.

The UNDRR's Global Assessment Report on Disaster Risk Reduction 2022 makes the institutional version of this point: risk-management systems built for discrete, sectoral hazards are structurally unsuited to systemic risk. The financial version is simpler. If the reserve is never rebuilt between events, it is not a reserve. It is a permanent line item that grows.

3.1 The crossover

The arithmetic is worth making explicit, because it is the entire case and it is not complicated.

Two features of that picture matter more than the exact numbers. The first is that the prevention line is almost flat: having paid once, the marginal cost of surviving each additional event is small. The second is that the response line has no such property — it is linear in the number of events, forever, with no terminal value.

This is what polycrisis does to the calculation. It does not change the cost of prevention or the cost of response. It increases the number of events, and therefore moves the crossover point nearer in time. A government that would have been rational to self-insure against one flood a decade is not making the same decision when it faces three a decade alongside a drought, a price shock and an epidemic.

4. Pakistan 2022: The Model in One Country

A single event, showing every component of the response premium operating at once.

In the summer of 2022 Pakistan flooded. The World Bank's assessment put damages and economic losses at over US$30 billion, with reconstruction needs estimated at a further US$16 billion. Some 33 million people were affected.

Every element of Section 2 is visible in that sequence. The spending was made at the moment of maximum need and minimum bargaining power. It was financed into a balance sheet already under strain — the floods arrived while the country was negotiating with the IMF, and the reconstruction bill deepened a debt problem that has required repeated Fund programmes since. It bought relief and repair rather than protective capital. And it restored, at enormous cost, a landscape whose exposure to the next monsoon was substantially what it had been before.

The fiscal tail is the part that is easiest to miss and hardest to escape. A shock absorbed through borrowing does not end when the water recedes; it persists as debt service for years, and where restructuring becomes necessary it can take years more to complete. Zambia defaulted in November 2020 and did not conclude its external-debt restructuring until 2024, with Ghana and Sri Lanka reaching comparable points only over 2024 to 2026 (IMF Global Sovereign Debt Roundtable). Through all of those years, the fiscal space in which prevention would have to be financed is smaller than it was before the event.

5. What Prevention Actually Buys

The return evidence is settled and is summarised briefly here; the figure that matters most for this argument is a cost ratio rather than a benefit-cost ratio.

The aggregate case is well established. The Global Commission on Adaptation estimated that US$1.8 trillion invested across early warning, resilient infrastructure, dryland agriculture, mangroves and water management could generate US$7.1 trillion in net benefits; the World Resources Institute's 2025 study of 320 real investments found benefits above ten dollars per dollar and more than half of them accruing whether or not a disaster occurred. That literature is assembled in more detail in an earlier report in this series, and is not re-argued here.

The number this report leans on is different in kind, because it compares two costs and therefore does not depend on how one values avoided harm.

A resilient asset costs a few per cent more than a fragile one and avoids losses worth a large fraction of the whole asset. This is not a close call, and it does not require agreement about discount rates or the value of a statistical life. It is a procurement decision, taken at the drawing board, that is cheap when taken then and unavailable at that price ever again.

Which makes the timing decisive. Most of the infrastructure that will carry the emerging world through the 2040s and 2050s is being specified now. Resilience designed in at the outset is bought at 3 to 5 per cent; the same protection retrofitted later, or replaced after a loss, is bought at the prices of Section 2.

5.1 The cheapest prevention is still substantially unbuilt

Early warning is the least capital-intensive prevention there is and among the highest-returning interventions in development. That roughly two countries in five still lack an adequate system is difficult to explain in economic terms at all — which is the subject of the next section.

6. Why the Worse Option Keeps Winning

Six structural asymmetries, none of them about the economics, each of which gives response an easier path through a government than prevention.

If prevention dominates response on cost, the persistence of the response model requires an explanation that is not economic. Six asymmetries do most of the work, and they are worth separating because they imply different fixes.

Table 2 — Why response wins the institutional contest

The asymmetries are analytical categories drawn from public-financial-management practice rather than measured effects; the point is that they are separable and separately fixable.

The second and third rows are the ones most amenable to direct repair, and they are the least discussed. A government that funds relief from a contingency reserve outside its expenditure ceiling, while requiring flood defences to compete inside it against health and education, has built a machine that will systematically choose the more expensive option. No official in that system is behaving irrationally. The rule set is producing the outcome.

The fourth row needs care, because it is easily misread as blaming vulnerable states for their own exposure. It is not that. The expectation of post-disaster assistance is a rational reading of how the system has actually behaved, and the states in question are, almost without exception, fiscally unable to self-fund the prevention that would be optimal. The asymmetry is a property of the architecture. It is fixed by changing when international money arrives — before the event rather than after — not by withholding it.

7. The Instruments Already Exist

The gap is not a design gap. Pre-arranged, anticipatory and contingent instruments have been built; they are radically under-capitalised relative to the liability they address.

An architecture for paying before the event now exists in outline. The IMF's Resilience and Sustainability Trust, capitalised through rechannelled Special Drawing Rights, supplies long-tenor financing for resilience-building — tenor being exactly what a fifty-year protective asset requires. Climate-resilient debt clauses suspend repayments when a qualifying disaster strikes, so that a shock does not compound into a debt crisis. Contingent credit lines and sovereign risk transfer pre-position money against events that have not yet happened. The Fund for responding to Loss and Damage addresses harm that prevention failed to avert.

What is missing is scale, and the scale problem is visible in the simplest possible comparison.

Developing countries paid US$921 billion in net interest in 2024 — a figure that includes, among much else, the compounding cost of shocks already absorbed through borrowing. International public adaptation finance in the same period ran at US$26 billion against an assessed need of US$310 to 365 billion a year by 2035. The system is not short of money moving in response to disaster risk. It is short of money moving before it.

8. Making Prevention Buyable

An agenda aimed at the institutional asymmetries rather than at the economics, because the economics have not been the binding constraint for some time.

8.1 Fix the budget asymmetry

• Put prevention in the same appropriation class as response. Where relief is funded from a contingency reserve outside the expenditure ceiling, prevention should have access to the same facility, or response should be brought inside the ceiling so both face one test.
• Legislate multi-year prevention envelopes that survive the annual cycle, so returns landing beyond an administration's term are not automatically deprioritised.
• Require every appraisal of a protective investment to state the expected response spending it displaces, and to count that displacement as a benefit. At present the counterfactual is silently set to zero.

8.2 Make the recurring liability visible

• Publish disaster response and reconstruction spending as a multi-year series rather than as a sequence of one-off emergencies. A charge that recurs every three years is a liability, and should appear as one.
• Maintain an exposure register — assets and people under protection, and the standard they are protected to — so that prevention has a countable output and can be managed against a target.

8.3 Move international money earlier

• Shift the balance of international support from post-disaster appeals toward pre-arranged and anticipatory finance, so that the assistance a state can reasonably expect arrives before the event and can be spent on protection.
• Make climate-resilient debt clauses standard rather than exceptional, and capitalise the Resilience and Sustainability Trust and the Loss and Damage Fund at a scale proportionate to the liability they address.

8.4 Buy resilience while it is still cheap

• Mandate design-stage resilience standards for publicly financed infrastructure, capturing protection at 3 to 5 per cent of capital cost while the emerging world's build-out is still ahead of it.
• Finish Early Warnings for All. It is the cheapest prevention available and roughly two countries in five still lack an adequate system.
• Require build-back standards that raise the protected baseline rather than restoring the one that failed — the single change that would stop reconstruction spending from guaranteeing its own repetition.

8.5 Conclusion

The case for prevention is usually made as an appeal to prudence, and prudence has not been persuasive. This report has tried to make it as an accounting observation instead. Every government exposed to recurrent hazard is already running a prevention budget; it is simply booked as relief and reconstruction, paid late, at surge prices, on crisis financing, with no asset at the end and no reduction in the exposure that produced the bill. That is not a decision to spend less. It is a decision to spend more, later, for less.

What has changed is the regime. Responding event by event was a defensible strategy when events were rare, independent and bounded, and most of the world's disaster-financing architecture was designed on that assumption. Under polycrisis the assumption fails in all three respects at once, and the crossover point — beyond which responding is simply the costlier path — arrives inside the planning horizon of any government now writing a budget.

The obstacles are institutional and therefore, in principle, cheap to remove. They are a contingency reserve that funds one kind of protection and not the other; an appraisal rule that sets the counterfactual to zero; an accounting convention that expenses one and capitalises the other; and a measurement system with no line for a disaster that did not occur. None of that requires new money or new technology to change. It requires deciding that a flood defence and a flood response are two prices for the same thing — and then declining to keep paying the higher one.

References

Every quantitative claim above is attributed inline. The principal sources are collected here.

• World BankPakistan: Flood Damages and Economic Losses Over US$30 Billion and Reconstruction Needs Estimated at US$16 Billion (October 2022).
• World Bank / GFDRRLifelines: The Resilient Infrastructure Opportunity — a 3–5 per cent design-stage premium against avoided losses of 50–100 per cent of asset value. See also the GFDRR summary.
• UNDRRGlobal Assessment Report on Disaster Risk Reduction 2022: Our World at Risk — risk-management systems built for discrete hazards against systemic risk.
• UNDRREarly Warning Systems Reach New Heights, but Critical Gaps Jeopardize Global Progress — 119 countries reporting a system, about 60 per cent. See Early Warnings for All.
• Global Commission on AdaptationAdapt Now: A Global Call for Leadership on Climate Resilience (2019) — US$1.8tn for US$7.1tn in net benefits.
• World Resources InstituteClimate Adaptation Investments Yield Massive Returns (2025) — 320 investments, benefits above $10 per $1, more than half non-contingent.
• UNCTADA World of Debt — US$921bn of net interest payments by developing countries in 2024.
• UNEPAdaptation Gap Report 2025 — US$26bn of flows against US$310–365bn of assessed need by 2035.
• IMFResilience and Sustainability Trust; and the Boston University GDP Center's assessment of its role in creating fiscal space.
• IMFGlobal Sovereign Debt Roundtable — 6th Co-Chairs Progress Report (April 2026) — restructuring timelines for Zambia, Ghana and Sri Lanka.
• UNFCCCFund for responding to Loss and Damage; first-year pledges as reported by Devex.
• World Economic ForumGlobal Risks Report 2026 — the interconnection of the risk landscape.
• H HeuristicsScaling Climate Adaptation Through Development Finance and Technology Diffusion — the benefit-cost literature and the finance and delivery gaps, treated at length.