Chapter 9: Forest Insurance and the Risk Architecture of the Forestry Industry

A stand of Chinese fir in Fujian takes twenty-five years to reach harvestable size. For twenty-five years the owner pays annual tending costs, forgoes the land's alternative use, and waits. Then, in a dry spring, a spark from a field-clearing burn crosses a ridge line and the whole investment is gone in four hours. The owner still owes the tending loans. The carbon that the stand had accumulated over two decades re-enters the atmosphere within a week.

Forestry is a business whose balance sheet is a living organism exposed to weather. That single fact explains why the sector's green transition is a risk problem that becomes a financing problem. A bank asked to lend against a twenty-five-year biological asset will price the fire probability, the pest probability, and the policy risk into the loan, and then decline. Chapter 2 described this as credit rationing under information asymmetry. Forestry is the textbook case: long duration, illiquid collateral, and a loss distribution heavily skewed toward rare, total losses. Forest insurance exists to change that loss distribution. This chapter examines how, where it works, and where it does not.

Forest insurance is a layered architecture: a public catastrophe layer for fire, a commercial layer for named perils, an index layer for broad weather risk, and a parametric layer for carbon delivery risk. Each layer addresses a different market failure, and none works alone. The design question for policy is which layer is missing, and what the missing layer costs the credit system.

9.1 The Risk Landscape of Forestry

9.1.1 What Kills a Forest

Fire is the most visible loss driver. China's 1987 Great Khingan fire burned roughly 1.1 million hectares and killed 211 people; that single event dominates the national historical loss record and illustrates the core actuarial problem, because a century of data can be shaped by one observation. In a typical year, forest fires in China affect somewhere between 10,000 and 30,000 hectares, with enormous year-to-year variance. I have not found a single authoritative annual series that covers the whole country at county resolution; the figures most often cited come from the National Forestry and Grassland Administration's annual statistical yearbooks, which report burned area by province but with inconsistent definitions across years. Fire losses are not independent across space. A single dry front can burn a hundred separate holdings in one afternoon, which destroys the diversification an insurer relies on when it writes many small policies.

Pests and disease are slower and larger in aggregate. The pine wood nematode (Bursaphelenchus xylophilus) has spread through pine forests in eastern and southern China since the 1980s, and control requires clear-cutting infected stands plus quarantine buffers. In the western United States, bark beetle outbreaks affected tens of millions of hectares during the 2000s, according to US Forest Service forest health reports; the exact figure depends on the survey year and the species included. Insect damage is a slow catastrophe: it accumulates over two to four years, is hard to attribute to a single event, and produces a claims pattern that looks less like fire and more like a chronic illness. The difficulty for an insurer is moral hazard at the margin. A stand already stressed by drought is more likely to succumb to beetles, and the owner knows the stand's condition better than the underwriter does.

Climate risk operates through the first two. Warmer winters reduce beetle mortality. Longer dry seasons raise fire probability. More frequent extreme wind events, typhoons in the south and derechos in the north, cause blowdown that creates fuel loads and breeding habitat for pests. The 2022 drought across the Yangtze basin cut the water level of Poyang Lake to a record low and stressed large areas of plantation forest. The actuarial consequence is that the historical fire and pest record is no longer a reliable guide to the future. A twenty-five-year policy written in 2024 is priced against a climate that will not exist in 2049. This is the central technical problem of forest insurance and it has no clean solution; the honest response is to shorten the repricing interval and to build parametric triggers that respond to weather rather than to historical loss averages.

Wind is systematically underpriced. Typhoons in Fujian, Guangdong, and Zhejiang cause blowdown that is often more damaging than the fire that follows. A single typhoon can flatten a plantation in a way that makes salvage expensive and reforestation mandatory. Wind risk is highly correlated within a province, which again defeats diversification.

A fifth risk is institutional rather than physical: policy risk. Forest land tenure in China was reformed in the 2000s through the collective forest tenure reform, which assigned long-term use rights to households. Those rights are real but they are administratively mediated. A change in zoning, a new protected area, or a revision of the harvest quota changes the value of the asset without any physical loss. An insurer cannot write a policy against a regulatory taking, and a lender cannot foreclose on a use right that the state may re-designate. The most important risk in forestry is uninsurable, and the credit system knows it.

9.1.2 The Loss Distribution and Why It Breaks Standard Insurance

Standard property insurance is built on the assumption that losses are independent and that the law of large numbers applies. Write ten thousand factory policies and the aggregate loss converges to the expected value. Forestry violates both assumptions.

Losses are spatially correlated. Fire spreads, wind is regional, pests move in fronts. The correlation coefficient between losses on adjacent forest holdings is far higher than between adjacent factories, which means the effective number of independent exposures is much smaller than the policy count suggests. An insurer with 50,000 small forest policies across a province may have the diversification of a few hundred independent bets.

Losses are heavy-tailed. Most years, most stands lose nothing. In a bad year, a province loses a meaningful fraction of its forest cover. The loss distribution is not normal; it is skewed with a long right tail, exactly the shape described in the credit portfolio literature. The expected loss is small and stable; the unexpected loss is large and rare, and it is the unexpected loss that determines solvency capital.

Losses are hard to verify. A fire is observable. A pest infestation is observable only after it has progressed. A slow climate-induced decline in growth is observable only through repeated measurement. The verification cost drives up the expense ratio and creates disputes. For a smallholder with five hectares, the transaction cost of adjusting a claim can exceed the claim itself.

These three features mean that forest insurance cannot be priced on pure actuarial grounds at the smallholder scale. The premium that would cover the tail risk is higher than the smallholder can pay, and the premium the smallholder can pay does not cover the tail. This gap is the market failure that public intervention exists to close, and it is the reason every serious forest insurance system in the world is a public-private hybrid.

9.1.3 The Risk Architecture Concept

Call the set of instruments, public programs, and contractual arrangements that together determine who bears which loss, at what trigger, and with what basis risk the risk architecture of a forest asset.

flowchart TD
    A[Forest Asset Risk] --> B[Frequency Layer]
    A --> C[Severity Layer]
    A --> D[Catastrophe Layer]
    A --> E[Institutional Layer]
    B --> B1[Small fire, pest, wind<br/>Retained by owner]
    C --> C1[Commercial named-peril policy<br/>Insurer balance sheet]
    D --> D1[Public cat fund / reinsurance<br/>Government or global market]
    E --> E1[Policy risk, tenure risk<br/>Uninsurable, borne by owner and lender]
    B1 --> F[Credit System]
    C1 --> F
    D1 --> F
    E1 --> F
    F --> G[Loan pricing and availability]

The architecture matters because the credit system does not care about any single instrument. It cares about the residual risk that the borrower carries after all instruments have done their work. A stand with a commercial fire policy but no catastrophe cover still has a tail that a lender must price. A stand with a public cat fund but no commercial policy has a frequency problem: small losses erode cash flow every few years and the borrower cannot service debt. The rating model in Chapter 5 sees the composite, not the components.

Insurance supports the green transition of forestry by reallocating risk to the layer that can bear it most cheaply. Fire frequency belongs with the owner, who can manage fuel loads and access. Fire severity belongs with the insurer, who can diversify across regions. Catastrophe belongs with the state or the global reinsurance market, which has the capital to absorb a bad year. Institutional risk belongs with no one, which is why it is the binding constraint.

9.2 Forest Insurance Products and Mechanisms

9.2.1 The Product Menu

Forest insurance products fall into four types, distinguished by what triggers a payout.

Indemnity insurance pays the actual loss, assessed after the event. It is the oldest form and the easiest to explain. The insurer sends an adjuster, measures the burned area and the damaged volume, and pays the assessed value up to the sum insured. Indemnity insurance has high verification cost and high dispute risk, and it is vulnerable to moral hazard because the owner controls many of the loss-prevention decisions.

Named-peril insurance restricts cover to a list of causes, typically fire, wind, and sometimes pest. It is the dominant form in China's policy-based forest insurance, which since 2009 has been rolled out as a subsidized program covering fire and selected perils. The restriction to named perils is a response to the verification problem: fire is observable, so the adjuster's job is bounded.

Index insurance pays when an external index crosses a threshold, regardless of the actual loss. The index can be weather-based (rainfall, temperature, wind speed) or area-yield-based (average loss in a defined region). Chapter 8 developed the logic for agriculture. The advantage is that the payout is fast and the verification cost is near zero, because the index is measured by a public authority. The disadvantage is basis risk: the index may trigger when the individual stand is fine, or fail to trigger when the stand is destroyed. Basis risk is the price of low verification cost.

Parametric catastrophe insurance is a special case of index insurance used at the portfolio level. A government or a large forest company buys a contract that pays a fixed amount when a defined event occurs, such as an earthquake above a magnitude or a typhoon above a wind speed in a defined box. The payout is not tied to the buyer's actual loss. It is a liquidity instrument, not an indemnity instrument, and it is used to recapitalize a program after a catastrophe rather than to compensate individual owners.

Table 9-1 summarizes the trade-offs.

Table 9-1: Forest Insurance Product Types and Their Trade-offs

No single product dominates. The choice depends on which cost the buyer is trying to minimize. A smallholder with five hectares cannot afford indemnity verification costs; an index product with moderate basis risk is better than no cover. A large plantation with professional management can absorb the verification cost and prefers indemnity because it eliminates basis risk. A provincial government cannot insure individual stands but can buy a parametric cat layer to protect its fiscal position after a bad fire year.

9.2.2 The Public-Private Model in China

China's policy-based forest insurance is the largest such program in the world by area covered. It is worth describing in institutional detail because it shows how the layers are actually assembled.

The program rests on premium subsidies. The central government subsidizes a share of the premium for public welfare forests and for commercial forests in designated regions, and provincial and county governments add further subsidies. The subsidy structure varies by province but the typical pattern is that the central government covers 30 to 50 percent, the province covers 20 to 30 percent, and the county covers 10 to 20 percent, leaving the owner with 10 to 30 percent. The subsidy is paid to the insurer, not to the owner, which means the owner sees a subsidized premium and the insurer sees a blended premium that covers part of the actuarial cost.

The coverage is named-peril, primarily fire, with wind and some pest cover added in specific provinces. The sum insured is set administratively, often at a level tied to replanting cost rather than to the market value of the standing timber. This is a deliberate choice. Replanting cost is verifiable and stable; market value fluctuates with timber prices and is subject to the harvest quota. Setting the sum insured at replanting cost keeps the program simple and reduces the incentive to over-insure, but it also means the program does not fully protect the owner's investment. A stand worth 200,000 yuan in timber value may be insured for 60,000 yuan of replanting cost. The residual 140,000 yuan is uninsured and sits on the owner's balance sheet, which is exactly the exposure a lender must price.

The delivery model uses the existing rural financial infrastructure. The insurer is typically a subsidiary of one of the large state-owned insurance groups, and the policy is sold through the same channels that deliver agricultural insurance: village-level coordinators, rural credit cooperatives, and in some provinces the supply and marketing cooperatives described in Chapter 10. The loss adjustment is delegated to a joint team of insurer staff and forestry bureau technicians, which uses the forestry bureau's stand maps and harvest records. This is a smart institutional design. The forestry bureau already holds the data needed for loss adjustment, and using it reduces the insurer's information cost. The trade-off is that the insurer depends on a public agency for its core underwriting data, which creates a coordination problem when the bureau's priorities differ from the insurer's.

The program's performance is mixed and the numbers tell the story. Coverage area has grown steadily, and forest insurance now covers a substantial fraction of China's plantation forest. But the loss ratio has been volatile, and in several provinces the program has run at a loss ratio that would be unsustainable without the subsidy. The reason is the one identified in section 9.1: fire losses are spatially correlated, and a bad fire year in one province can produce a loss ratio above 100 percent for the whole provincial book. The subsidy absorbs the average cost but does not solve the tail problem. The tail is transferred upward, to the reinsurance market and ultimately to the state.

9.2.3 Reinsurance and the Catastrophe Layer

The tail problem is where reinsurance enters. A primary insurer writing forest policies in a province retains the frequency layer and cedes the severity layer to a reinsurer. The reinsurer, in turn, cedes part of its exposure to the global catastrophe market through cat bonds or retrocession.

The global reinsurance market has limited appetite for Chinese forest risk. The reasons are data and correlation. The historical loss record is short and dominated by a few events, which makes it hard to price. The correlation with other perils in the same region, particularly typhoon, is high, which means a reinsurer writing both forest fire and typhoon cover faces a compound tail. The result is that the catastrophe layer is thin and expensive, and the state absorbs more of the tail than a pure market solution would.

This is not unique to China. After the 2017–2021 fire seasons in California, several private insurers withdrew from wildland-urban interface markets, and the state-backed California FAIR Plan expanded as a residual market. The pattern is general: when the tail is large and correlated, the private market retreats and the public sector becomes the insurer of last resort. The policy question is how to design that public backstop so that it prices risk correctly and does not crowd out private cover for the frequency layer.

A parametric cat layer is one answer. A provincial government can buy a parametric contract that pays a fixed amount when a fire event exceeds a defined severity threshold, and use the payout to recapitalize the insurance program. The advantage is speed: the payout arrives within weeks, before the loss adjustment is complete. The disadvantage is basis risk at the government level: the contract may pay when the province does not need it, or fail to pay when it does. The design of the trigger is the whole game, and it requires a long and clean historical record of fire events, which is exactly what many provinces lack.

9.2.4 The Smallholder Problem and the Case for Index Products

The policy-based program works best for large holdings and for collective forests managed by village committees. It works least well for the smallholder with a few hectares, who is the majority of forest owners by number.

The smallholder problem has three parts. First, the transaction cost of indemnity insurance is high relative to the sum insured. Sending an adjuster to a two-hectare stand costs the same as sending one to a two-hundred-hectare stand, but the claim is a hundred times smaller. Second, the smallholder's loss is often below the deductible, so the policy pays nothing in most years and the owner rationally concludes that the premium is wasted. Third, the smallholder's dominant risk is not fire but the combination of slow growth, pest, and price, which named-peril fire cover does not address.

Index insurance addresses the first two problems. A rainfall-deficit index or a fire-weather index can be measured at a weather station and paid automatically, with no adjuster visit. The payout is fast and the administrative cost is low, which means the product can be offered at a premium the smallholder can afford. The cost is basis risk: the index may not match the individual loss. Chapter 8 made the same argument for crops, and the empirical literature on index insurance in agriculture is directly relevant. The consistent finding is that demand for index products is lower than expected, and the main reason is basis risk. Farmers who experience a payout that does not match their loss distrust the product and drop it.

For forestry, the basis risk problem is worse than for crops because the loss is slower and the index is coarser. A fire-weather index built on temperature, humidity, and wind speed can predict fire danger, but it cannot predict whether a specific stand burns. A greenness index built on satellite imagery can detect a loss after the fact, but the resolution is coarse and the attribution to a specific cause is uncertain. The design challenge is to find an index that is correlated with the individual loss at a level the owner finds acceptable, and to combine it with a small indemnity component for the residual.

A second design option deserves more attention than it gets: group or area-yield index insurance sold through a cooperative or a village committee. The unit of insurance is the group, not the individual, and the payout is based on the average loss in the group's area. This reduces basis risk at the individual level because the group average is closer to the individual loss than a distant weather station, and it reduces verification cost because the group's loss can be assessed once. The trade-off is that the group must have a governance structure that can distribute the payout fairly, which is a real constraint in villages where the committee is weak. Chapter 10 takes up this question in the context of cooperative finance.

9.2.5 Forest Insurance Compared with Agricultural Insurance

Forest insurance is harder than crop insurance for reasons that show up in the structure of the two programs.

Crop insurance covers an annual asset. The policy is written in the spring, the loss is assessed at harvest, and the cycle resets. Forest insurance covers a perennial asset. The policy must run for years, the loss can occur at any point in a twenty-five-year cycle, and the value of the asset changes as the stand grows. A fire in year three destroys a small investment; a fire in year twenty-three destroys a large one. The sum insured must therefore be adjusted over time, which adds administrative cost and creates disputes about the value at the time of loss.

Crop insurance has a clear yield benchmark. Forest insurance does not, because the harvest is not annual and the volume depends on the thinning schedule. The loss adjustment must estimate the volume that would have been harvested, which requires growth models and stand inventories. The forestry bureau has this data, but it is not always current.

Crop insurance benefits from a long actuarial record in the major producing regions. Forest insurance does not, because the events are rarer and the record is shorter. The result is that forest premiums are more uncertain and the insurer must hold more capital per unit of cover.

These differences do not make forest insurance impossible. They make it more expensive and more dependent on public data and public subsidy. The comparison is useful because it shows that the success of crop insurance in China cannot be replicated in forestry simply by extending the same program design. The program needs a different architecture.

9.3 Integrating Insurance with Forest Carbon Projects

9.3.1 Why Carbon Changes the Insurance Problem

A forest carbon project converts a stand of trees into a stream of credits. The project developer measures the carbon sequestered above and below ground, registers the project under a standard such as the China Certified Emission Reduction (CCER) scheme or an international standard, and sells the credits to a buyer who uses them to offset emissions. The revenue from credit sales is what makes the project financially viable, and it is often the marginal revenue that justifies reforestation on marginal land.

The carbon stream changes the insurance problem in three ways.

First, the insured asset is no longer the timber. It is the carbon stock and the future credit flow. A fire that destroys the timber also destroys the carbon stock and releases the sequestered carbon into the atmosphere, which reverses the credits that have already been sold. The buyer of the credits has a claim, and the project developer must replace the reversed credits or refund the payment. This is the reversal risk, and it is the defining risk of forest carbon projects.

Second, the time horizon is longer than the timber rotation. A carbon project may be registered for thirty or forty years, and the credits are generated over that period. The insurance must cover the whole period, which is longer than any commercial forest policy currently written.

Third, the loss is not just the physical loss. A project that loses its carbon stock also loses its registration, its verification, and its reputation. The transaction cost of restarting a project after a loss is high, and the insurance must cover that cost if the project is to be restored.

The result is that carbon projects need a different insurance product from the one that covers timber. The product must cover the carbon delivery risk, not just the physical loss. This is the frontier of forest insurance and it is where the most interesting design work is happening.

9.3.2 The Reversal Risk and the Buffer Mechanism

The standard mechanism for managing reversal risk is the buffer pool. Under the major carbon standards, a project developer must deposit a share of its credits into a pooled buffer account rather than selling them. If the project later suffers a reversal, the buffer pool is drawn down to compensate the buyers. The buffer share is set by a risk score: projects with higher reversal risk contribute more.

The buffer mechanism is a form of self-insurance at the portfolio level. It works because reversals are relatively rare and the pool can absorb them if the contributors are diversified. It has two weaknesses. The pool is small relative to the tail: a large fire in a concentrated region could exhaust it. And the buffer is not priced to the individual project's risk; it is a flat contribution based on a risk score, which means low-risk projects cross-subsidize high-risk ones.

Insurance can complement the buffer. A project can buy a policy that pays into the buffer pool or directly to the buyer if a reversal occurs. The policy converts the tail risk into a premium, which is more efficient than holding a large buffer if the project's risk is diversifiable. The design question is the trigger. An indemnity trigger based on the measured carbon loss is accurate but slow and expensive to verify. A parametric trigger based on fire occurrence or drought index is fast but carries basis risk. In practice, a combination works: a parametric trigger for the fast payout and an indemnity true-up for the residual.

The buffer mechanism also interacts with the credit rating. A project with a well-funded buffer and an insurance backstop has a lower reversal probability, which lowers the risk premium the credit system charges. This is the integration point with Chapter 5. The rating model should see the buffer and the policy as risk mitigants, and the model must be told they exist.

9.3.3 Insuring the Carbon Stream: Product Design

A carbon delivery insurance product has four components: the insured amount, the trigger, the term, and the exclusions.

The insured amount is the value of the credits that would have been delivered over the remaining project life, discounted to the present. The valuation is complicated by the carbon price, which is volatile and which varies by standard and by vintage. A conservative approach is to insure the volume of credits rather than the value, with the payout set at a reference price. This separates the physical risk from the price risk and makes the product easier to price.

The trigger is the event that causes the reversal. For fire, the trigger can be the burned area exceeding a threshold, verified by satellite imagery and the forestry bureau's records. For drought, the trigger can be a rainfall deficit index. For pest, the trigger is harder because the loss is slow and the attribution is uncertain. In practice, most carbon insurance products cover fire and wind, and treat pest as an exclusion or a separate product.

The term is the project life, which can be thirty years. Very few insurers will write a thirty-year policy because the climate and the carbon market will change. The practical solution is a rolling policy with a repricing interval of three to five years, combined with a contractual commitment to renew. The renewal commitment is the valuable part; the repricing protects the insurer against climate drift.

The exclusions are the standard list plus two carbon-specific ones. Reversals caused by the project developer's failure to follow the management plan are excluded, because they are a moral hazard. Reversals caused by a change in the carbon standard or by a government decision to revoke the registration are excluded if they are treated as a political risk, which is generally uninsurable.

Table 9-2 lays out the components.

Table 9-2: Design Parameters for Carbon Delivery Insurance

9.3.4 The Interaction with Credit and the Rating Model

A forest carbon project is a borrower. It has a loan for establishment costs, and it has a revenue stream from credit sales. The rating model assesses the probability that the borrower defaults. The insurance policy changes that probability in two ways.

It reduces the variance of the cash flow. Without insurance, a fire in year ten wipes out the project and the loan defaults. With insurance, the fire triggers a payout that covers the reversal and the restoration cost, and the project continues. The default probability falls because the left tail of the cash flow distribution is truncated.

It changes the recovery rate. If the project does default, the insurance payout is an asset that the lender can claim. The recovery rate rises, which lowers the loss given default.

Both effects should be reflected in the rating. The model needs three inputs: the existence of the policy, the trigger design, and the basis risk. A policy with a parametric trigger and high basis risk is worth less to the rating than an indemnity policy with low basis risk, because the payout may not arrive when the loss occurs. The rating model must therefore incorporate a basis-risk discount, which is a parameter that the analyst must estimate. This is not a mechanical adjustment; it requires judgment, and the judgment should be documented.

The buffering layer changes the risk profile that the rating measures, and the rating must be told what the buffering layer is. A firm with an index policy has a different default probability than the same firm without one, and the difference depends on the basis risk of the index. The rating model that ignores this will misprice the loan, and the mispricing will be largest exactly where it matters most: in the marginal projects that need the finance to proceed.

9.3.5 A Worked Example

The following is a stylized model, not a real project. The parameters are chosen to be plausible for a subtropical CCER reforestation project; the point is the structure of the calculation, and the reader should substitute project-specific data before drawing conclusions.

Take a 1,000-hectare reforestation project in a subtropical province, registered under CCER, expected to generate 8,000 credits per year for twenty years. At a reference price of 60 yuan per credit, annual credit revenue is 480,000 yuan. Establishment cost is 15 million yuan, financed by a 10-year loan at 5 percent, with annual debt service of about 1.94 million yuan. Credit revenue alone does not cover debt service in the early years; the project also harvests thinning timber from year eight.

Without insurance, the project's default probability is driven by fire risk. Suppose the annual fire probability for an unmanaged stand in the region is 0.5 percent, and a fire destroys the project. The cumulative probability of at least one fire over ten years is about 5 percent. The lender prices this into the loan, and the interest rate rises to 7 percent, which raises annual debt service to about 2.14 million yuan and pushes the project closer to the margin.

With a fire insurance policy that pays the restoration cost and the reversed credits, the default probability falls to the residual risk that the policy does not cover: the basis risk plus the events above the policy limit. Suppose the residual is 1 percent over ten years. The lender can then price the loan at 5.5 percent, and the project becomes viable.

The insurance truncates the tail and makes the residual measurable. The rating model can then price the residual, and the loan can be made. This is the buffering function in its simplest form.

9.3.6 Limits and Failure Modes

Forest insurance has boundaries that no product design will move.

Institutional risk sits outside the boundary. A change in harvest quota, a re-designation of forest land, or a revision of the carbon standard can destroy the project's value without any physical loss. These risks are political and they are generally uninsurable. The lender must bear them, and the rating model must treat them as a separate factor that insurance does not touch.

Basis risk is a permanent feature of every index product, and it is largest exactly when the need is greatest, because a widespread event that triggers the index also stresses the insurer's capacity. The honest response is to measure the basis risk and to disclose it, not to pretend it is zero.

Correlation at the insurer level is the third limit. A provincial book of forest policies is concentrated, and a bad year can produce a loss ratio above 100 percent. The insurer needs reinsurance, and the reinsurance market has limited appetite. The state absorbs the residual, which means the ultimate risk bearer is the taxpayer. This is a legitimate arrangement, but it should be explicit rather than implicit.

Finally, insurance prices risk; it does not remove it. A project on marginal land with a high fire probability will pay a high premium, and the premium may make the project unviable. That is the correct outcome. The insurance is doing its job by revealing the risk, not by hiding it.

These limits are not arguments against forest insurance. They are arguments for designing the architecture with the limits in view. The layers must be assigned to the bearer who can handle them, and the residual must be measured and priced.

9.3.7 The Institutional Complement: Forestry Bureaus, Cooperatives, and Data

The technical design of insurance products is necessary but not sufficient. The products must be delivered through institutions that have the data and the trust to make them work.

The forestry bureau is the key data holder. It maintains stand maps, growth models, harvest records, and fire statistics. It also has the field staff to verify losses. The insurer that partners with the bureau gets access to data that would otherwise be prohibitively expensive to collect. The partnership is not automatic; it requires a data-sharing agreement and a clear division of responsibilities. In provinces where the bureau is well-resourced, the partnership works. In provinces where it is not, the insurer must build its own data capacity, which raises costs.

The cooperative is the key delivery channel for smallholders. A cooperative can aggregate demand, negotiate a group policy, and distribute payouts. It can also monitor members' forest management, which reduces moral hazard. The supply and marketing cooperative system described in Chapter 10 has the rural network to reach smallholders, and in some provinces it has been used as the delivery channel for policy-based forest insurance. The cooperative model works best when the cooperative has real governance and real financial capacity; a shell cooperative adds a layer of cost without adding value.

The data infrastructure is the binding constraint. Forest insurance needs three data streams: a stand inventory, a weather and fire record, and a carbon measurement system. The first two exist in some provinces and not in others. The third is new and expensive. Chapter 13 takes up the digital infrastructure question in detail; the point here is that forest insurance and carbon accounting share the same data foundation, and investment in one supports the other.

9.3.8 The Policy Question

The analysis converges on a set of design choices that are narrower than the usual "expand forest insurance" recommendation.

The frequency layer is best handled by the owner and the commercial insurer, with a subsidy to make the premium affordable for smallholders. The subsidy should be tied to the adoption of risk-reducing management practices, such as fuel-load reduction and thinning, so that it does not simply transfer risk to the insurer without changing behavior.

The severity layer is best handled by the insurer with reinsurance, and the reinsurance market needs better data to price the risk. The public investment with the highest return is the construction of a clean, long, geocoded loss record that a reinsurer can use. This is unglamorous and it is the single most important enabler of the whole architecture.

The catastrophe layer is best handled by the state through a parametric backstop, designed to recapitalize the program after a bad year rather than to compensate individual owners. The backstop should be priced, not free, so that it does not crowd out private cover.

The carbon layer is the newest and the least developed. It needs a standard trigger design, a reference price for valuation, and a clear rule for how the buffer pool and the insurance interact. The CCER scheme's rules on reversal and buffer are the starting point, and the insurance industry should be at the table when they are revised.

The institutional layer is the residual, and it is where the credit system must be most careful. A lender that relies on insurance to cover institutional risk will be disappointed. The rating model should treat institutional risk as a separate factor, and the loan structure should include covenants that give the lender a claim if the forest land's designation changes.

Each of these is a specific design choice that follows from the analysis of the risk landscape. A national forest insurance fund or a mandatory scheme, the single-instrument approach that policy writing tends to favor, will fail because the risk is layered. The layers have to be built separately and priced separately, and the credit system has to be told which layers are in place.

9.4 Conclusion: What the Buffering Layer Buys

Forest insurance is the clearest case of the buffering function in the book's chain because forestry is the asset class where the un-buffered risk is largest relative to the return.

What does the buffering layer buy? It buys a measurable residual. Before insurance, the lender faces a loss distribution with a long tail and no way to price it, and the rational response is to decline the loan. After insurance, the lender faces a truncated distribution with a known basis risk, and the rational response is to price the residual and lend. The transition from decline to lend is the contribution of the buffering layer. In the worked example above, that transition shows up as a 150-basis-point reduction in the loan rate and a project that clears its debt-service coverage ratio instead of failing it.

It also buys time. A forest carbon project needs twenty years to deliver its credits, and the insurance gives the project and its lenders the confidence to commit to the full period. Without the insurance, the project is a series of short-term bets that no one wants to hold. With the insurance, it is a long-term asset that can be financed.

The limits are real. The institutional risk is uninsurable, the basis risk is nonzero, and the catastrophe layer depends on the state. But the limits are known, and a known limit can be priced. An unknown limit cannot. The value of the architecture is that it converts an unknown risk into a known one, and the credit system can then do its arithmetic on the residual.

The next chapter takes up the cooperative layer, where the same logic applies at a smaller scale. A cooperative that shares risk among its members is a form of mutual insurance, and it faces the same correlation problem that a commercial insurer faces. The difference is that the cooperative's members are also its owners, which changes the incentive structure. The question is whether that change is enough to make mutual risk-sharing work where commercial insurance does not.