Who Really Pays for Resilience?

Financing Incidence: Who Pays for Preparedness?

The first question is not simply how much resilience costs. It is how the cost of preparing for disruption enters the financial structure of a system.

Businesses and governments face different versions of this problem.

For a company, maintaining additional inventory can provide a buffer against interruptions in supply. Yet inventory is not costless to maintain. It occupies working capital and storage capacity, and its value depends partly on whether the goods can still be sold when they are needed.

The difficulty is identifying what happens to these costs after they enter a company's accounts.

Do they reduce profit margins? Are they reflected in wholesale prices? Do retailers absorb them, negotiate them with suppliers, or incorporate them into final prices? The answers may differ across industries, products and market structures.

Research on retail inventories offers an important distinction.

In its 2025 working paper, Retail Inventories and Inflation Dynamics: The Price Margin Channel, the Federal Reserve's Mehrotra, Oh and Ortiz examine the relationship between retail inventories and price margins using industry-level data and supply-side variation. Their findings support a relationship between inventory scarcity and retail pricing margins.

A related study by Baslandze and Fuchs, published in the Journal of International Economics in 2026, examines supply disruptions, product availability and pricing dynamics during the pandemic. It finds that delivery shortfalls can affect retail prices, including through shortages experienced by competing suppliers.

These studies help explain why insufficient availability can influence pricing.

They do not, however, establish how much of the cost of maintaining additional precautionary inventory is passed on to consumers.

That distinction is essential. A price increase during a shortage is not evidence that a company has transferred the cost of holding extra stock to its customers. Scarcity, delivery delays and the cost of maintaining a buffer are related questions, but they are not interchangeable explanations.

The same distinction becomes important when governments and public utilities finance infrastructure.

Consider California's electricity system. The California Public Utilities Commission reviews utility spending and determines which costs may be recovered through regulated electricity rates. Its regulatory process provides a real example of how infrastructure-related expenditure can become part of the financial obligations borne by utility customers.

The Public Advocates Office has documented changes in approved wildfire-related revenue requirements for PG&E, Southern California Edison and San Diego Gas & Electric during 2023 and 2024.

But these costs need to be described accurately. They include a mixture of wildfire mitigation, rebuilding and response, insurance, and liability-related expenditure. They cannot all be treated as the capital cost of newly constructed resilience infrastructure.

The records establish a financing channel: approved costs can be recovered through utility rates.

They do not, by themselves, establish the final net burden on households at different income levels.

A study of electricity-grid undergrounding in California illustrates why the choice of financing arrangement matters.

In a 2023 paper published in Nature Energy, Wang and colleagues examine how the costs of undergrounding power lines could be distributed across communities with different socioeconomic characteristics.

The study compares alternative allocation approaches, including arrangements in which costs are borne locally and alternatives involving broader cost sharing.

Its modelling shows that a locally allocated cost can represent a substantially larger burden relative to income in poorer communities. A broader allocation arrangement can change that distribution.

This is an important result, but its status must be understood correctly: the study evaluates proposed or simulated allocation arrangements. It does not document an implemented tariff reform or measure the actual bills subsequently paid by households under those alternatives.

Nevertheless, it exposes a fundamental design question.

Two infrastructure programmes can pursue the same technical objective while distributing their financial responsibilities differently. The engineering outcome may be similar, but the allocation of its cost need not be.

That does not make one financing arrangement automatically fairer in every circumstance. Broader cost sharing may distribute expenses across a larger population, while local financing may more closely connect payment to the communities receiving the investment. The appropriate assessment depends on who benefits, who is exposed, the available alternatives and the wider financing system.

The important point is that the method of financing is not a secondary administrative detail. It is part of the structure through which resilience is delivered.

But identifying who finances protection still does not tell us who actually receives it.

A community may contribute to an infrastructure programme without every household experiencing the same reduction in risk. Conversely, a household may benefit from a public protection measure without directly financing the entire investment.

To understand this second dimension, we need to move from the financial arrangements surrounding resilience to the distribution of its actual or estimated benefits.

Sources: Retail Inventories and Inflation Dynamics: The Price Margin Channel · The Price of Delay: Supply Chain Disruptions and Pricing Dynamics · Understanding Electric Rates · 2023-2024 Wildfire-Related Cost Increases of California’s Three Major Investor-Owned Electric Utilities · Local and Utility-Wide Cost Allocations for a More Equitable Wildfire-Resilient Distribution Grid

Protection Incidence: Who Actually Benefits?

Protection is often discussed in terms of the assets or population a project is designed to protect.

Yet the intended coverage of a programme and the benefits that can be measured afterwards are not necessarily identical.

Flood protection offers a useful example because the available evidence allows us to examine different outcomes: insurance claims, property values and the socioeconomic characteristics of exposed communities.

These measures can tell us something about protection, but they answer different questions.

An insurance claim is not the same as total physical damage. A change in property value is not the same as an avoided loss. And the presence of a protective structure does not establish that every household in the surrounding area receives an equal benefit.

The distinction becomes particularly important when a study reports an estimated reduction in losses.

One study of flood adaptation in the United States provides a particularly useful example.

In a 2024 paper published in Nature Communications, Cano Pecharroman and Hahn examine the effects of community participation in the National Flood Insurance Program's Community Rating System (CRS), a programme that encourages local flood-risk reduction measures.

The researchers use approximately 2.5 million flood insurance claims and report 14,729 community units in their analysis. They apply a model-based approach to estimate how the effects of CRS participation vary across communities with different characteristics.

Their results suggest that estimated reductions in insurance claims payments are not uniform. The authors report differences across community groups, with estimated savings in the approximate range of $5,000 to $15,000 per policy.

The unit of measurement matters.

These figures should be understood as model-estimated differences in insurance claims payments per policy, not as cash received by every household. Although the paper's abstract sometimes uses the expression "per household", its methods and results specify the per-policy measurement.

There are further limitations.

CRS participation is not randomly assigned. Communities that participate may differ from those that do not in ways that also affect flood losses. The researchers use community characteristics and a model-based treatment-effect approach to address these differences, but such adjustments cannot automatically eliminate every source of selection bias.

The analysis also relies on insurance claims. Households without flood insurance are not represented in the claims outcome, even though they may face substantial physical and financial exposure.

The authors additionally account for the possibility that outreach and improved awareness affect the reporting of claims. That adjustment is itself model-dependent.

What, then, does the research establish?

It provides evidence that the estimated benefits of a particular community flood-adaptation programme vary across the insured populations and communities examined.

It does not establish the total welfare benefit received by all residents, the actual cash savings of every household, or the distribution of protection among people who are not insured.

This distinction is more than a technical qualification. It changes the question we should ask.

A programme may reduce insured losses while leaving important parts of the population outside the measurement. Its measured success can be real without representing the complete distribution of protection.

Insurance claims are not the only way to examine the value of protection.

A different study, published in Risk Analysis in 2018 by Beltrán, Maddison and Elliott, examines the relationship between flood defences and residential property values in England.

The researchers use a repeat-sales approach, drawing on more than 12 million property transactions and 1,666 flood-defence projects constructed between 1995 and 2014.

Their findings reveal an important geographical distinction.

The estimated property-price effects were positive in urban areas, with reported increases of approximately 12.6% to 16.7% in the relevant specifications. In rural areas, the estimated effects were smaller and could be negative, ranging from approximately 0.8% to 5.0% below the comparison level.

These results should not be interpreted as a direct measurement of the physical protection provided by each defence.

Property prices reflect more than avoided flood damage. They can incorporate expectations, local characteristics and the way buyers value the surrounding area. A rise in property value may benefit an owner, but it does not necessarily translate into the same financial gain for a tenant.

Equally, a negative property-price effect does not establish that a flood defence made a rural community less safe.

The study is useful precisely because it demonstrates why the economic value associated with protection cannot be reduced to a single, uniform outcome.

Different locations can experience different market responses even when the policy objective is broadly similar.

Taken together, the insurance and property-market studies reveal two different ways of measuring the benefits of flood protection.

One examines estimated changes in insured claims payments. The other examines how protection is reflected in property transactions.

Neither provides a complete account of social welfare.

And neither can, on its own, tell us who remains exposed after protection has been introduced.

That leads to the third dimension.

Sources: Exposing Disparities in Flood Adaptation for Equitable Future Interventions in the USA · Assessing the Economic Benefits of Flood Defenses: A Repeat-Sales Approach

Residual Risk Incidence: Who Remains Exposed?

Resilience investments are generally designed to reduce risk, not to eliminate it.

A reinforced electricity network can still experience outages. A flood-defence system can still be exceeded by an extreme event. A business with alternative suppliers can still face shortages if disruptions affect several sources simultaneously.

The relevant question is therefore not only how much risk has been reduced, but also how much remains---and where it remains.

This is a different question from asking who paid for a project or who received its measurable benefits.

To understand residual risk, we must first distinguish exposure from protection.

Exposure describes who or what is situated in harm's way. Protection describes the measures that reduce the likelihood or consequences of harm. The two may overlap, but they are not interchangeable.

A community can receive public investment and remain highly exposed. Another may have lower baseline exposure without receiving any new protective investment.

Without separating these conditions, it is easy to mistake the allocation of resources for the achievement of protection.

A study of the 2021 floods in Germany by Odersky and Löffler provides one illustration of why exposure deserves separate attention.

Published in the Journal of Economic Inequality, the research links geocoded information on flood-affected buildings with socioeconomic characteristics of surrounding neighbourhoods.

The authors find that flood exposure in the studied event was higher in relatively lower-income neighbourhoods.

This is an observation about the distribution of exposure. It is not evidence that a particular resilience programme caused the difference, nor does it measure how much protection different households received from a specific investment.

The distinction matters.

If a neighbourhood experiences greater flood exposure, several explanations may be relevant: its geographical position, the characteristics of the built environment, historical development patterns and the distribution of protective infrastructure.

A study of exposure alone cannot determine which explanation is responsible.

It does, however, remind us that the people who remain exposed after a disruption cannot be identified simply by looking at the total amount of infrastructure a region has built.

There is also an important counterpoint to the assumption that public resources necessarily bypass vulnerable communities.

Karim and Noy's 2020 study in World Development examines disaster-adaptation and social-protection allocations across 483 subdistricts, or upazilas, in Bangladesh.

Their analysis finds a positive relationship between disaster risk, socioeconomic vulnerability and the allocation of relevant public resources.

This suggests that public allocation can respond to vulnerability rather than simply follow existing economic or political advantages.

But allocation is not the same as protection.

A budget directed towards a high-risk community does not, by itself, demonstrate that the investment was completed, that the intended population could access it, or that household losses subsequently declined.

The distinction between money allocated and risk actually reduced must remain visible.

These findings point towards a more careful way of evaluating resilience.

We should not assume that vulnerable communities always receive less protection. Nor should we assume that the presence of public investment means their remaining risks have been adequately addressed.

The distribution of residual risk is an empirical question, not a conclusion that can be inferred from spending totals or infrastructure coverage alone.

Sources: Differential Exposure to Climate Change? Evidence from the 2021 Floods in Germany · Risk, Poverty or Politics? The Determinants of Subnational Public Spending Allocation for Adaptive Disaster Risk Reduction in Bangladesh

A Diagnostic for Understanding Resilience

The three questions examined above can be brought together through a simple analytical framework: the Buffer Incidence Map (BIM).

The purpose of this framework is to examine how the financial responsibilities, protective benefits and remaining risks associated with resilience are distributed.

It is important to be precise about what this framework represents.

The Buffer Incidence Map is an original editorial diagnostic tool. It is not a peer-reviewed economic theory, a statistically validated index or an empirically established causal model. Its purpose is to organize questions that are often considered separately, not to manufacture answers where evidence is missing.

It has three dimensions.

Financing Incidence: Who Pays?

The first dimension examines the financial responsibility for preparing for disruption.

In a business, this might involve the working capital committed to inventory and alternative supply arrangements.

In public infrastructure, it might involve taxes, public borrowing, regulated electricity rates or other cost-recovery mechanisms.

At the household level, preparation might involve insurance, home improvements or personal financial reserves.

The relevant questions are not limited to the total amount spent.

Who provides the money? Through which mechanism? Are costs shared, subsidized or recovered from a specific group? And does the financing arrangement change the relative burden across participants?

Protection Incidence: Who Benefits?

The second dimension examines who actually receives the benefits of reduced risk.

A flood-defence project may protect a geographical area, but the resulting benefits can take different forms.

Some may appear in insurance claims. Others may be reflected in property values, business continuity or reduced disruption to essential services.

These outcomes must not be treated as interchangeable.

A reduction in insured claims does not automatically represent the welfare gain of every resident. An increase in property values does not necessarily benefit tenants. And a project designed to protect a community does not establish that all households have equal access to its protection.

The measurement must match the claim being made.

Residual Risk Incidence: Who Remains Exposed?

The third dimension considers what happens after protection has been introduced.

Which communities remain outside the protected area? Which households lack relevant insurance coverage? Which businesses still depend on vulnerable suppliers? Which risks are reduced but not eliminated?

This dimension is particularly important because a successful project can still leave significant exposure elsewhere.

The absence of a recorded loss among insured properties, for example, does not establish that uninsured residents experienced no harm.

The three dimensions therefore ask different questions:

Financing Incidence examines the distribution of responsibility. • Protection Incidence examines the distribution of benefits. • Residual Risk Incidence examines the distribution of what remains.

Together, they provide a more complete way of investigating resilience without assuming that the three distributions must either coincide or conflict.

A useful application of the framework would examine one named project, identify the relevant groups and map the available evidence for each dimension.

Where a relationship is only theoretically possible, it should be labelled as a possible channel rather than an established causal connection.

The framework is useful precisely because it makes those unanswered questions visible.

The Counterargument: Resilience Can Benefit Everyone

There is an important objection to this way of thinking.

If stronger infrastructure reduces the probability of disruption, if better preparation prevents shortages, and if public investment protects communities against natural hazards, then resilience can create benefits that extend well beyond the individuals or organizations directly paying for it.

A more reliable electricity network, for example, can support households, businesses, hospitals and public services simultaneously.

The benefits may be widely shared even when the costs are collected through a particular financing mechanism.

The strongest version of this objection is about the counterfactual. The relevant comparison is not between investment and an ideal of evenly distributed costs and benefits; it is between investment and no investment. Even when costs and gains are unevenly distributed, a resilience investment may still leave every group better off than it would have been without that protection. Avoided losses can be real even when they are not shared equally.

The question in this article is how the gains from protection are distributed, not whether gains exist at all. Showing that some groups pay more, receive different forms of protection, or remain more exposed does not establish that they would have been better off without the investment. But the possibility of an improvement over no investment does not settle who benefits, by how much, or who continues to carry residual risk.

There are also reasons not to assume that public investment systematically overlooks vulnerable communities.

Karim and Noy's analysis of disaster-related public spending in Bangladesh found a positive relationship between socioeconomic vulnerability, disaster risk and relevant public allocations.

This suggests that public financing can be directed towards communities facing greater exposure.

Similarly, the California undergrounding study by Wang and colleagues illustrates how alternative cost-sharing arrangements can change the distribution of infrastructure costs. Financing can be designed differently; a locally concentrated burden is not the only possible arrangement.

These examples matter because they challenge an overly pessimistic interpretation.

Resilience is not inherently regressive. Its benefits are not necessarily captured by wealthy property owners or large corporations. Nor is every public infrastructure investment an example of costs being transferred to people with fewer resources.

Indeed, a carefully designed programme could reduce existing inequalities by directing protection towards communities that would otherwise face greater losses.

But this possibility does not make distributional analysis unnecessary.

It makes it more important.

If resilience investments can produce different outcomes depending on how they are financed, targeted and implemented, then the question is not whether resilience is inherently fair or unfair.

The question is what determines the outcome in a particular case.

Sources: Local and Utility-Wide Cost Allocations for a More Equitable Wildfire-Resilient Distribution Grid · Risk, Poverty or Politics? The Determinants of Subnational Public Spending Allocation for Adaptive Disaster Risk Reduction in Bangladesh

What the Framework Cannot Tell Us

Separating financing, protection and residual risk is useful, but it does not automatically reveal an unfair distribution.

A framework can organize evidence. It cannot replace it.

Several questions remain unresolved.

First, the direct relationship between precautionary inventory costs and consumer prices has not been established by the studies examined here. Research on shortages and retail pricing cannot substitute for evidence about the incremental cost of maintaining additional safety stock.

Second, although regulated infrastructure costs can be recovered through electricity rates, this does not establish the net financial burden experienced by households at different income levels. That requires project-specific payment data, subsidies, actual protection outcomes and a credible comparison.

Third, insurance claims, property values, public spending allocations and neighbourhood exposure measure different things. They cannot simply be combined into a single indicator of who benefited from resilience.

These limitations also identify what could challenge the framework.

If high-quality project-level research repeatedly showed that financing responsibilities, protective benefits and residual risks were distributed in essentially the same proportions, the practical value of examining them separately would be reduced.

If evidence across different countries and sectors consistently demonstrated that resilience investments produced progressive net welfare gains, that would challenge any suggestion that their distribution was generally unequal.

If project-level net welfare data showed that additional costs were offset by subsidies, transfers and avoided losses, with no meaningful differences in net outcomes across income, insurance or housing-status groups, the claim of substantive distributional inequality in those projects would be weakened.

And if a reliable system-level measure could accurately predict the financial burden, protective benefits and remaining exposure of different groups, the need for a separate distributional diagnostic would become less compelling.

The framework should remain open to these possibilities.

Its value lies not in predicting that a particular group will lose, but in preventing us from assuming that everyone benefits simply because the system performs better.

A More Complete Definition of Resilience

The pursuit of resilience is fundamentally about preparing for uncertainty.

Businesses seek to maintain operations when supply chains are disrupted. Governments invest in infrastructure that can withstand extreme conditions. Communities develop protection against environmental hazards. Households try to preserve their security when ordinary systems fail.

These efforts can produce substantial benefits.

But the success of a resilience investment cannot be fully understood through engineering performance, aggregate expenditure or the amount of disruption avoided alone.

A stronger network tells us something about the network.

A larger inventory tells us something about a company's preparedness.

A completed flood-defence project tells us something about the infrastructure that has been built.

None of these observations, on its own, tells us everything about the people and organizations connected to that system.

To understand the broader result, three questions must remain visible.

Who finances protection?

Who actually benefits from it?

And who remains exposed when the protection is no longer sufficient?

The answers may reveal that financial responsibility and protection are closely aligned. They may show that public investment distributes benefits widely, or that targeted programmes improve the position of communities facing the greatest risks.

They may also reveal gaps between the people financing preparedness and those receiving its benefits.

None of these outcomes should be assumed in advance.

Perhaps the most important change in how we think about resilience is therefore not to abandon the pursuit of stronger systems, but to expand the definition of what success means.

A system should not be judged only by how well it survives a disruption.

It should also be understood through the distribution of the protection it provides, the resources required to sustain it, and the risks that remain beyond its boundaries.

Because ultimately, a system can become more resilient without making everyone equally secure.

And understanding that difference is part of understanding resilience itself.