Chapter 12: Policy Framework for an Integrated Green Finance Ecosystem

A provincial agricultural credit union in eastern China ran a quiet experiment between 2019 and 2021. Its portfolio held tea cooperatives, most of them small, most without audited accounts, most asking for working capital at planting time and again at harvest. Loan officers had historically priced these loans on collateral—a house, sometimes a processing shed—and the collateral requirement kept about two-thirds of applicants out. Then the union began scoring cooperatives on a short list of environmental and operational variables: whether the cooperative held a forest management certificate, whether it maintained a buffer strip along watercourses, whether it had switched to organic fertilizer in the previous two seasons, and how its yields had varied across the last five years. It also bought a weather index product that paid out when rainfall during the spring flush fell below a trigger. Approval rates rose. The share of loans secured only by movable assets rose. The default rate on the scored segment stayed below the union's existing book. None of this required a new bank, a new law, or a new development institution. It required a scoring template, an insurance contract, and a decision to treat environmental performance as information rather than as decoration.

That experiment is the practical referent for this chapter. The preceding chapters established that the green transition cannot be financed through a single channel. Credit alone cannot identify which technology firms are genuinely green and which are merely well-marketed; insurance alone can buffer weather and forest risks but cannot direct capital toward innovation; venture and guiding funds can catalyze green technology but cannot absorb the idiosyncratic risks that make early-stage green projects unbankable. The integrated mechanism—rating-based identification, insurance-based risk buffering, and sci-tech finance catalysis—emerges from the interaction of these three functions, not from the perfection of any one of them.

A policy framework that supports this mechanism has to supply instruments at each layer and coordinate the layers so they reinforce rather than duplicate one another. It also has to govern the resulting system so that public money is spent on public purposes and private capital is not crowded out. This chapter sets out that framework, layer by layer, and closes with the governance question that determines whether the framework survives contact with administrative reality.

12.1 Policy Instruments for Green Finance: The Layered Toolkit

12.1.1 Why a Single Instrument Fails

Start with the arithmetic that motivates the whole exercise. By the end of 2020, China's green loan balance stood at nearly RMB 12 trillion. That is a large number in isolation. Against total loan balances it is roughly 5 percent. The other 95 percent of the credit system—loans to cement plants, logistics firms, chemical producers, and the small manufacturers that make up the bulk of the borrowing universe—was not classified as green and had no mechanism for becoming green. The policy question is not how to expand the 5 percent. It is how to move the 95 percent, or some meaningful fraction of it, onto a transition path.

The instruments available to a government for this purpose fall into four families, and each family addresses a different failure. Table 12-1 lays them out.

Table 12-1: Instrument families and the failures they address

The table is a starting point, not a taxonomy to be defended to the death. The important claim is that these families are complements. A taxonomy without a risk-transfer layer produces classification without lending, because banks will not lend to a green-labeled borrower whose cash flow is destroyed by a drought. A risk-transfer layer without a taxonomy produces insurance sold to anyone, including borrowers whose practices are environmentally harmful, which is how agricultural insurance programs occasionally end up subsidizing land conversion. Catalytic capital without information infrastructure produces venture funds that cannot screen green technology and therefore default to generalist technology bets with a green label attached.

The tea cooperative case illustrates the complementarity in miniature. The scoring template is an information instrument. The weather index contract is a risk-transfer instrument. The union's willingness to lend against the scored profile is the credit outcome that neither instrument could produce alone. Remove the scoring and the insurance pays out to an undifferentiated pool; remove the insurance and the scoring identifies good risks that the union still cannot prudently underwrite because a single bad season wipes out the cooperative's repayment capacity.

12.1.2 Information Instruments: Taxonomy, Disclosure, and Scoring

The first layer of any green finance framework is definitional. A jurisdiction must be able to say what counts as green, and it must be able to say it in a way that a loan officer can apply at a desk without a consultant. Three questions structure the policy design: how do we define green industries and green customers (whether something is green)? How do we rate greenness (how green is it)? And are all green firms equally green over time (how long does the greenness last)?

Most existing taxonomies answer the first question and gesture at the second. The EU Taxonomy Regulation establishes technical screening criteria for economic activities that contribute substantially to climate mitigation or adaptation, and it does so with numerical thresholds—grams of CO₂ per kilowatt-hour, kilograms of CO₂ per tonne of product. That level of specificity is a strength for bond issuance and a burden for bank lending, because a small manufacturer cannot easily demonstrate compliance with a threshold calibrated for large installations. The practical response in most banking systems has been to build simplified scoring templates for small and medium borrowers, derived from the taxonomy but not identical to it.

Disclosure rules operate at a different point in the information chain. They govern what firms report, not how lenders classify. The EU's Corporate Sustainability Reporting Directive and the sustainability disclosure standards issued by the International Sustainability Standards Board have moved the reporting burden from voluntary to mandatory for large firms, and the supply-chain consequence is that smaller suppliers face disclosure demands from their customers whether or not they face them from regulators. For green finance policy, the relevant design question is proportionality: a reporting regime calibrated for listed companies will, if applied without modification to a tea cooperative, produce either non-compliance or fabricated numbers.

The scoring layer is where information instruments become operational for credit. Chapter 3 and Chapter 4 treated the construction of these scores in detail; the policy point here is that scoring standards have public-good characteristics. A scoring methodology developed by one bank and validated on its own portfolio is a private asset. A scoring methodology validated across institutions, with published weights and periodic back-testing, is infrastructure. The transition from the first to the second is a policy choice, and it is the choice that determines whether green credit expands beyond the few institutions with the resources to build their own models.

12.1.3 Risk-Transfer Instruments: Insurance, Guarantees, and First-Loss

The second layer addresses risks that private markets will not absorb at prices borrowers can pay. Two categories matter most for the green transition in agriculture and forestry.

Covariate weather risk is the first. A drought affects every cooperative in a county at once, which is precisely the event that makes conventional indemnity insurance expensive and slow. Weather index insurance solves the moral hazard and verification problems by paying on an objective trigger—rainfall below a threshold, temperature above a threshold—rather than on assessed loss. Chapter 5 set out the design; the policy question is how the product is subsidized and distributed. China's agricultural insurance program has expanded to cover a large share of planted area, and the central and provincial governments together bear a substantial portion of the premium. The subsidy is the policy instrument. Without it, the premium for a weather index contract on a small tea plot exceeds what the cooperative will pay voluntarily, because the cooperative discounts low-probability events and because the basis risk—the gap between the index trigger and actual loss—makes the product feel unreliable.

Forest risk follows a similar path in China, with public premium subsidies and a growing role for commercial insurers in the underwriting. The policy design issue differs from weather index, because forest losses are more readily assessed and because the asset has a long rotation period that makes annual insurance contracts an awkward fit. Multi-year policies and coverage tied to forest management certification are the instruments that align the insurance with the investment horizon.

Guarantees and first-loss tranches belong to the same family but address a different failure. A credit guarantee does not transfer weather risk; it transfers default risk from the lender to the guarantor, which allows the lender to lend to a borrower whose creditworthiness is adequate but whose collateral is insufficient. For green technology firms, the collateral problem is acute: the assets are often intangible, the revenue is often pre-commercial, and the standard lending model has nothing to attach to. A guarantee fund that absorbs the first 20 to 30 percent of losses on a defined green loan portfolio changes the lender's calculus without requiring the lender to build a new underwriting capability.

The design principle that connects these instruments is that risk transfer should be priced and targeted, not universal. A guarantee that covers all losses on all green loans is a subsidy to lending that would have happened anyway. A guarantee that covers the first loss on a defined segment—early-stage green technology, or smallholder forest management—pays for risk that the market genuinely cannot price.

12.1.4 Catalytic Capital: Guiding Funds and Venture Co-Investment

The third layer fills financing gaps that neither credit nor insurance can fill. Early-stage green technology has a characteristic profile: long development timelines, high failure rates, intangible assets, and a public-good component in the eventual diffusion of the technology. Private venture capital underinvests in this profile relative to its social return, and the underinvestment is worse in jurisdictions where exit markets are thin.

Government guiding funds are the instrument most widely used to address this gap. The design questions are well known and frequently answered badly. A guiding fund that invests directly in companies competes with private venture capital and crowds it out. A guiding fund that invests as a limited partner in private venture funds leverages private screening capacity but may simply subsidize funds that would have been raised anyway. A fund-of-funds structure with a requirement that private capital match public capital at a specified ratio, and with a mandate to invest in defined green technology segments, threads between these failures—but only if the matching requirement is enforced and the green mandate is defined precisely enough to be auditable.

Consider how the matching requirement plays out in practice. China's National Integrated Circuit Industry Investment Fund, established in 2014, raised RMB 138.7 billion in its first phase and RMB 204.2 billion in its second, drawing in provincial funds and private co-investors alongside the central contribution. The model is not a green fund, but its capital structure illustrates the leverage arithmetic that green guiding funds should replicate: a central commitment that is large enough to anchor a deal but small enough that private limited partners carry the majority of the risk. When a green fund is capitalized at a ratio of one public yuan to two or three private yuan, the public money is buying screening capacity and a mandate, not substituting for private capital. When the ratio inverts—public money at 70 or 80 percent of the fund—the fund is a budget line with a venture label.

The performance measurement problem is more difficult than the capital structure problem. A guiding fund that cannot report the private capital it mobilized, the technology firms it financed through to commercial revenue, and the environmental outcomes attributable to its portfolio has no way to distinguish itself from a generalist fund. Chapter 8 treated the incentive design and performance evaluation of these funds in detail. For the policy framework, the relevant point is that catalytic capital is the layer with the highest fiscal cost and the longest payback period, and it should therefore be the layer with the strictest performance measurement. The measurement has to be designed at the fund's inception, not retrofitted after the first vintage of investments has matured, because the metrics that matter—graduation to commercial credit, environmental outcome per unit of public capital—require data that the fund must begin collecting on day one.

There is a further design question that receives less attention than it deserves: the relationship between the guiding fund and the credit system. A fund that finances a green technology firm through demonstration and early commercialization creates a firm that, if successful, will eventually need bank credit at a scale the fund cannot provide. If the fund's scoring standard and the bank's scoring standard are different, the firm faces a re-underwriting at the graduation point, and the graduation may fail. The coordination link described in Section 12.2.2—the catalytic-to-scoring link—exists to prevent exactly this outcome.

12.1.5 Price and Quantity Signals: Carbon Markets and Regulatory Differentials

The fourth family operates on the underlying economics rather than on the financial system directly. A carbon price, whether delivered through a trading system or a tax, changes the relative profitability of green and brown production, and that change propagates into credit demand and credit quality. China's national emissions trading system, launched in 2021, initially covered the power sector and has expanded gradually; the price signal it delivers is modest relative to the investment decisions it is meant to influence, which is why the financial instruments in the other three families remain necessary.

Differentiated reserve requirements and refinancing windows are the central-bank instruments that operate on the liability side of bank balance sheets. A lower reserve requirement ratio for green loans, or a dedicated refinancing facility that funds green lending at a concessionary rate, changes the funding cost of green credit without directing individual lending decisions. The People's Bank of China has used both. The design lesson from the first years of operation is that the instrument works when the green classification is unambiguous and the reporting is verifiable. A refinancing window that accepts self-reported green classifications will be used to refinance loans that were going to be made anyway.

The interaction between the price signals and the other three layers is worth stating precisely, because it is often misunderstood. A carbon price does not substitute for a green taxonomy; it changes the profitability of activities that the taxonomy classifies. A firm that is brown under the taxonomy but faces a rising carbon price has an incentive to transition, and the transition path—retrofitting a plant, switching a fuel, changing a process—is precisely the activity that the scoring layer should be able to recognize and the risk-transfer layer should be able to support. If the scoring layer cannot recognize a transition path, the carbon price changes behavior without changing access to finance, and the firm that most needs credit to retrofit is the firm that cannot get it. The design requirement, then, is that the scoring standard include a transition dimension, not only a static green/brown classification.

The refinancing window has a parallel requirement. A concessionary refinancing facility that is available only for loans classified as green under a static taxonomy will not finance transition. A facility that is available for loans to firms with a credible transition plan, scored against the same standard the lender uses, can finance transition. The difference is administrative—the central bank has to accept the lender's transition score as the eligibility criterion—and it is the difference between a facility that moves the 5 percent and one that can move part of the 95 percent.

12.1.6 The Instrument Mix in Practice

Figure 12-1 arranges the four families against the stage of a green project's development. The arrangement is not a rigid sequence; it is a diagnostic map. A project at the demonstration stage needs catalytic capital and risk transfer, not a green bond. A mature cooperative with a certified forest management plan needs scoring and insurance, not venture money. Policy that applies the whole toolkit uniformly to every project wastes fiscal resources and produces instruments that no one uses.

flowchart TD
    A[Green project pipeline] --> B{Stage of development}
    B -->|Demonstration / pre-commercial| C[Catalytic capital<br/>Guiding funds, venture co-investment]
    B -->|Early commercial| D[Risk transfer<br/>Guarantees, first-loss, weather index]
    B -->|Mature, cash-generating| E[Information + credit<br/>Scoring, taxonomy, green loans]
    C --> F[Commercial viability]
    D --> F
    E --> F
    F --> G{Environmental outcome verified?}
    G -->|Yes| H[Outcome-linked refinancing<br/>and reporting]
    G -->|No| I[Re-scoring and<br/>corrective action]
    I --> E

The map also shows the feedback loop that most policy frameworks omit. Environmental outcomes, once verified, should feed back into the scoring layer. A cooperative that maintains its certification and its buffer strips for five years should see its score improve, and with it its access to credit at better terms. Without this loop, the information layer is a one-time gate rather than a continuing incentive.

12.2 Coordinating Rating, Insurance, and Sci-Tech Finance

12.2.1 The Coordination Problem

Suppose a jurisdiction has built all three layers. It has a taxonomy and a scoring standard. It has a subsidized weather index program and a guarantee fund. It has a guiding fund that co-invests with private venture capital in green technology. Each layer functions. The question that determines whether the system delivers the integrated mechanism described in Chapter 9 is whether the layers talk to each other.

Three coordination failures recur across jurisdictions.

Data fragmentation is the most common. The scoring layer collects environmental and financial data on borrowers. The insurance layer collects loss and weather data. The venture layer collects technology and commercial milestones. In most systems these datasets sit in different institutions, under different confidentiality rules, and are never linked. The consequence is that a cooperative's five-year record of certified forest management, which is exactly the evidence an insurer needs to price a multi-year policy and exactly the evidence a lender needs to justify a lower rate, is invisible to both.

Incentive misalignment is the second. An insurer that pays out on a weather index trigger has no direct interest in whether the insured cooperative adopts climate-resilient practices, because the payout is triggered by weather, not by loss. A lender that holds a guarantee has reduced incentive to monitor the borrower, because part of the loss is borne by the guarantor. A guiding fund that is evaluated on the number of companies financed has an incentive to finance many small tickets rather than to concentrate on the firms with the highest probability of commercial success. Each of these is a standard principal-agent problem, and each requires a contract design that restores the missing incentive.

Sequencing error is the third and the most damaging in practice. Building a scoring standard before there is any demand for scored lending produces a standard that no one uses. Building a guarantee fund before there is a pipeline of green projects produces a fund that guarantees loans to whatever is available. The diagnostic process described in the previous chapter—identifying the constraint that actually binds before prescribing an instrument—is the remedy, but it requires an institution with the authority to make the diagnosis.

12.2.2 A Coordination Architecture

The architecture that addresses these failures has four components: a shared data layer, a set of cross-layer contracts, a sequencing rule, and a coordinating authority.

The shared data layer is the least glamorous and the most consequential. It requires a common borrower identifier across the scoring, insurance, and credit systems; a legal basis for sharing data among the participating institutions; and a set of standard data fields that each layer commits to populate. The fields do not need to be exhaustive. A workable minimum for the agriculture and forestry segment includes: the borrower's environmental certification status and date, the weather index trigger and payout history for the relevant grid cell, the loan performance history, and the insurance claims history. For the technology segment: the technology readiness level at each financing round, the intellectual property filings, the revenue milestones, and the environmental outcome metrics agreed at the time of investment.

The cross-layer contracts are the mechanisms that convert shared data into changed behavior. Three are worth specifying.

Scoring-to-insurance link. The insurance premium is adjusted according to the borrower's environmental score, with the adjustment justified by the score's correlation with loss experience. A cooperative that maintains buffer strips and certified management has a lower expected loss on forest fire and a lower basis risk on the weather index, and the premium should reflect it. This link gives the borrower a reason to maintain the practices that the score rewards, and it gives the insurer a reason to use the score.

Insurance-to-credit link. The lender treats the insurance contract as a credit enhancement. A loan to a cooperative whose weather risk is covered by an index policy has a different loss distribution from a loan to an uncovered cooperative, and the difference should show up in the risk weight, the pricing, and the loan-to-value ratio. This link is what converts the insurance layer from a standalone social program into a credit-expansion instrument, which is the subject of Chapter 6.

Catalytic-to-scoring link. The guiding fund requires its portfolio companies to report against the same environmental metrics used in the scoring standard, and it makes subsequent tranches conditional on meeting them. This link prevents the venture layer from becoming a green-labeled generalist fund, and it generates the outcome data that the scoring layer needs to validate its environmental weights.

Figure 12-2 shows the links as a directed graph. The arrows are the contracts; the nodes are the institutions.

flowchart LR
    S[Scoring / rating layer<br/>Taxonomy, environmental score] -->|Premium adjustment| I[Insurance layer<br/>Weather index, forest cover]
    I -->|Credit enhancement<br/>Risk weight, LTV| C[Credit institutions]
    C -->|Loan performance data| S
    G[Guiding fund /<br/>sci-tech finance] -->|Outcome reporting<br/>against scoring metrics| S
    S -->|Eligibility screen<br/>for portfolio companies| G
    I -->|Loss and weather data| D["(Shared data layer)"]
    S -->|Score and certification data| D
    C -->|Repayment data| D
    D --> S
    D --> I
    D --> G

The sequencing rule is the diagnostic process itself. Before an instrument is deployed, the coordinating authority asks which constraint binds: is the problem that lenders cannot identify green borrowers (information), that they can identify them but cannot bear the risk (risk transfer), or that the projects cannot attract any finance at the early stage (catalytic)? The answer determines which layer receives the next increment of policy resources. In practice the diagnosis is rarely clean—most systems have all three constraints to some degree—but the discipline of ranking them prevents the common failure of building the most visible instrument rather than the one that binds.

The coordinating authority is the institutional answer to the question of who makes the diagnosis. It does not need to be a new agency. In many jurisdictions the central bank, the financial regulator, or a dedicated green finance committee within the finance ministry can perform the function, provided it has three attributes: authority to convene the institutions in all three layers, access to the shared data, and a mandate that spans credit, insurance, and venture finance rather than sitting within a single sectoral silo. The last attribute is the hardest to create, because insurance supervision and banking supervision are typically separate, and venture capital policy often sits in an industry ministry rather than a financial one.

12.2.3 Coordination in the Agriculture and Forestry Segment

The agriculture and forestry segment shows the coordination architecture at work, and it also shows where it breaks.

Consider the sequence that a well-coordinated system would follow for a tea cooperative seeking transition finance. The cooperative applies for a loan. The lender's scoring template, drawing on the shared data layer, returns a score that reflects the cooperative's certification status, its input practices, and its yield history. The score qualifies the cooperative for a weather index policy at a discounted premium, because the scoring model has established that certified cooperatives in this grid cell have lower basis risk. The lender, holding the policy as credit enhancement, approves a loan at a loan-to-value ratio that would be unavailable to an uncovered borrower. The cooperative's repayment history flows back into the shared data layer, where it improves the score for the next cycle. If the cooperative later seeks equity-like capital for a processing facility, the guiding fund's eligibility screen draws on the same score.

Each step in that sequence is a policy choice. The scoring template has to include the certification variable. The insurer has to accept the score as a rating factor. The lender's regulator has to recognize the insurance as a credit risk mitigant. The shared data layer has to permit the repayment history to flow back. The guiding fund has to accept the score as an eligibility screen. A jurisdiction that gets four of the five right will see a partial version of the sequence; a jurisdiction that gets the insurance recognition wrong will see the cooperative approved for a loan it cannot afford to insure.

The forestry segment adds a dimension that agriculture does not have: the time horizon. A forest management plan runs for decades, and the insurance contract that covers it should run for more than one year. Multi-year forest insurance exists in several jurisdictions, but it is typically priced on an annual basis with renewal risk, which means the cooperative or forest enterprise bears the risk that the premium rises or the product is withdrawn before the rotation is complete. A coordination policy that addresses this is to have the public reinsurance layer—the government-backed catastrophe program that sits behind the commercial insurer—offer a multi-year commitment, so that the primary insurer can offer a multi-year policy. The cost is a contingent liability; the benefit is that the forest enterprise can plan against a known insurance cost.

There is a further asymmetry in the forestry segment that policy has to confront. The forest enterprise's revenue arrives at harvest, which may be twenty or thirty years after planting, while the insurance premium is paid annually. A multi-year policy with a level premium is, in effect, a financing instrument: the insurer is lending the enterprise the difference between the level premium and the risk-adjusted premium in the early years. If the insurer cannot finance that difference, the level premium is not available, and the enterprise faces the renewal risk. The public reinsurance commitment is what makes the level premium financeable, because it caps the insurer's tail risk and lets the insurer hold the policy on its books without reserving against a catastrophic year.

12.2.4 Coordination in the Green Technology Segment

The technology segment has a different coordination problem. The scoring layer for technology firms is less about environmental certification and more about the credibility of the technology and the management team, because the environmental outcome is prospective rather than current. Chapter 4 treated the scoring model construction; the coordination question is how the score interacts with the other two layers.

The insurance layer is thinner for technology firms, because the risks are not weather or fire but technology failure and market adoption. The instruments that exist are credit guarantees and first-loss tranches, which transfer default risk rather than technology risk. The coordination link is that the guarantee provider should use the same scoring standard as the lender, so that a firm that scores well on technology readiness and management quality receives a guarantee at a lower fee. Without the link, the guarantee is priced on collateral, and the firm with intangible assets is back where it started.

The catalytic layer is where the technology segment's coordination is most developed. A guiding fund that co-invests with private venture capital in green technology needs a screen that identifies which firms are genuinely green and which are attaching a green narrative to a general technology. The screen is the same scoring standard, applied to prospective environmental outcomes. The fund also needs an exit path, and the coordination question is whether the credit layer can provide it: a firm that has reached commercial revenue and a validated environmental outcome should be able to graduate from venture financing to bank credit, using its score and its revenue history as the basis. A system in which the venture layer and the credit layer use different standards has no graduation path, and the venture fund is forced to hold the company longer than its mandate allows.

The graduation path deserves a concrete illustration, because it is the point at which the technology segment's coordination most often fails. A green technology firm that has completed demonstration and reached first commercial revenue typically has a revenue base in the range of RMB 20 to 50 million, a customer concentration problem, and a balance sheet dominated by intangibles. A bank applying its standard small-business scorecard will see the concentration and the intangibles and decline. A bank applying the green scoring standard, with the firm's technology readiness level and validated environmental outcome as inputs, can see a different risk profile. The difference between the two outcomes is whether the bank's credit policy recognizes the green score as a valid input to the small-business scorecard. That recognition is a policy decision, made by the bank's risk committee in consultation with the regulator, and it is the technology segment's equivalent of the insurance recognition question in the agriculture segment.

12.2.5 The Coordination Failure That Matters Most

Of the three coordination failures, the one that does the most damage is the incentive misalignment between the insurer and the lender. The reason is arithmetic. The insurance layer's purpose, in the integrated mechanism, is to convert uninsurable risk into a known cost that the lender can underwrite. If the insurance contract is priced without reference to the borrower's environmental practices, the lender has no reason to prefer the insured borrower over the uninsured one except for the coverage itself, and the coverage is expensive. If the insurance contract is priced with reference to the score, the lender has a reason to use the score, and the borrower has a reason to maintain the practices. The premium adjustment is the hinge.

The magnitude of the adjustment needed to change behavior is an empirical question, and it varies by segment. For the tea cooperative, a premium discount in the range of 10 to 20 percent is enough to make the policy attractive relative to self-insurance, provided the basis risk is explained. For the forest enterprise, the discount matters less than the multi-year commitment, because the enterprise's planning horizon is long and the renewal risk is the dominant concern. For the technology firm, the guarantee fee reduction is the relevant lever, and the reduction needs to be large enough to change the lender's expected loss calculation, which in practice means a fee in the range of 1 to 3 percent of the guaranteed amount rather than the 5 to 8 percent that a collateral-based guarantee would command.

These numbers are not universal. They are the kind of parameter that a coordinating authority should calibrate from the shared data layer rather than import from another jurisdiction. The principle is what travels: the cross-layer contract must change a price that the decision-maker actually faces.

12.3 Implementation and Governance

12.3.1 From Framework to Administration

A policy framework that exists only on paper is a common artifact. The governance question is what makes the difference between a framework that is implemented and one that is announced.

Three governance functions determine the answer: the allocation of authority, the measurement of performance, and the management of the political economy. Each has a design choice that is easy to get wrong.

12.3.2 Allocating Authority

The integrated mechanism spans at least four regulatory domains: banking supervision, insurance supervision, securities regulation (for green bonds and funds), and industrial or agricultural policy. No single existing agency typically holds all four. The allocation choice is between creating a new coordinating body, designating an existing body as the lead, and relying on inter-agency agreements.

The designated-lead model has been the practical choice in most jurisdictions that have built green finance frameworks. The UK's Green Finance Institute, established in 2019 as a joint venture between the City of London Corporation and the UK government, was set up to convene the financial sector and the government without creating a new regulator; its mandate is coordination, and the regulatory authority remains with the Bank of England and the Financial Conduct Authority. Singapore's Green Finance Action Plan, launched by the Monetary Authority of Singapore in 2019, placed the coordination mandate inside the central bank and used inter-agency agreements to bring in the other ministries. Both arrangements avoided the multi-year lag that a new agency would have imposed, and both kept the data and the supervisory tools inside the institution that already held them.

The risk of the designated-lead model is that the lead agency's own mandate dominates: a central bank will tend to see the coordination problem as a credit problem, and an insurance regulator will see it as a solvency problem. The mitigation is a formal coordination mandate, written into the lead agency's terms of reference, with reporting obligations to the finance ministry or the cabinet. The mandate should specify the layers the lead agency is responsible for coordinating, the data it is entitled to receive, and the reporting it owes.

The allocation of authority also has a sub-national dimension that is often overlooked. In China, the provincial branches of the central bank and the provincial financial supervision bureaus are the institutions that actually implement green finance policy at the level where the tea cooperatives are. A framework that is designed in Beijing and not adapted at the provincial level will produce the familiar pattern of pilot programs that report success and do not scale. The governance design needs to specify what the provincial units are authorized to adapt—the scoring weights, the insurance triggers, the guarantee coverage ratios—and what they are not.

12.3.3 Measuring Performance

The measurement question is where most green finance frameworks are weakest. The standard metrics—green loan balances, green bond issuance, the number of firms financed—measure activity, not outcome. A framework that reports a growing green loan balance and a falling environmental score in the loan portfolio has failed, but the standard metrics will not show it.

The measurement architecture that avoids this failure has three tiers.

Tier 1: Activity metrics. Volume of green credit, insurance coverage, and catalytic capital deployed, disaggregated by segment and by layer. These are necessary for budget accountability and insufficient for policy evaluation.

Tier 2: Intermediate outcome metrics. The share of scored borrowers whose environmental score improved over the reporting period; the loss ratio on insured green loans relative to uninsured; the private capital mobilized per unit of public catalytic capital; the graduation rate of venture-financed firms to commercial credit. These metrics connect the instrument to the mechanism.

Tier 3: Environmental outcome metrics. The measured environmental performance of the financed activities: emissions avoided, forest area under certified management, water quality indicators in the relevant watershed, soil organic matter. These are the metrics that justify the framework, and they are the hardest to collect. The shared data layer is what makes them collectible, because it links the environmental measurement to the borrower identifier that the financial system already tracks.

A concrete example of the three tiers in the tea cooperative segment: Tier 1 reports the number of cooperatives receiving scored loans. Tier 2 reports the share of those cooperatives that maintained certification and buffer strips over three years, and the default rate on the scored segment relative to the union's book. Tier 3 reports the change in nitrogen runoff in the watershed, measured at monitoring points, attributable to the cooperatives in the program. A framework that reports only Tier 1 will be unable to defend its budget when the fiscal pressure arrives, because it will have no evidence that the money changed anything.

The tier structure also has a diagnostic use that is easy to miss. When Tier 1 rises and Tier 2 does not, the instrument is being deployed but the mechanism is not working—the scoring is not changing behavior, or the insurance is not changing lending. When Tier 2 rises and Tier 3 does not, the mechanism is working but the environmental proxy is wrong—the practices the score rewards are not producing the outcome the score assumes. Each pattern points to a different corrective action, and the corrective action is more precise than anything a single aggregate metric could support.

12.3.4 The Political Economy of Implementation

The governance question that receives the least attention in policy documents and the most attention in practice is who bears the cost of the transition and who captures the benefit.

The costs are concentrated and near-term. A brown firm that must retrofit its plant bears the capital cost. A bank that must reclassify its portfolio bears the provisioning cost. An insurer that must underwrite a new product bears the development cost and the early loss experience. A local government that must enforce environmental standards bears the political cost of the firms that leave or contract.

The benefits are diffuse and delayed. Cleaner air and water accrue to the whole population over years. The reduced physical risk from climate change accrues to future cohorts. The productivity gains from green technology accrue to the firms that adopt it and to their customers, over a horizon that exceeds the electoral cycle.

This asymmetry is the reason that green finance frameworks are vulnerable to two failure modes. Capture is the first: the framework is designed so that the costs fall on firms without political voice and the benefits flow to firms with political access. Reversal is the second: the framework is announced with fanfare and quietly defunded when the near-term costs become visible and the delayed benefits have not yet arrived.

The governance design that addresses capture is transparency in the allocation of subsidies and guarantees, with published criteria and published recipient lists. The design that addresses reversal is institutional anchoring: writing the framework's core elements into the mandate of an independent agency, so that defunding requires a legislative act rather than an administrative decision. Neither design is foolproof. Both are better than the alternative of relying on the good intentions of the incumbent administration.

12.3.5 A Worked Governance Sequence

A jurisdiction starting from a low base would first adopt a green classification consistent with international standards but calibrated to the domestic economy, and build a scoring template for the dominant borrower segments—small manufacturers, agricultural cooperatives, forest enterprises—with weights that can be validated against existing loan performance data. Publishing the template matters as much as building it, because a template that stays inside the regulator cannot be used by the lenders who need it.

With the template in place, the next move is the shared data layer: designate the borrower identifier, the legal basis for data sharing, and the minimum data fields, assign data stewardship to the lead agency, and require the participating institutions to populate the fields for new loans from a specified date. This step is unglamorous and it is the one most often deferred. A jurisdiction that defers it will find, two years later, that the pilot cannot be evaluated because the data was never linked.

The cross-layer contracts should be piloted in one segment and one region, selected for the cleanliness of the data and the willingness of the institutions. In the Chinese case, the agricultural segment in a province with an established weather index program is the natural candidate. The pilot implements the premium adjustment, the credit enhancement recognition, and the outcome reporting, and it runs for at least two full cycles so that the loss experience is observable. Two cycles is the minimum because a single cycle cannot distinguish a good year from a good design.

Calibration follows the pilot. The premium discounts, the guarantee fees, and the scoring weights are adjusted on thebasis of the observed loss experience, and the calibration methodology is published so that other regions can see what was changed and why. This is the point at which the parameters stop being imported from another jurisdiction and start being domestic.

Extending to the technology segment comes next. The scoring standard is applied to the venture pipeline, the eligibility screen for the guiding fund is established, and the graduation path to bank credit is negotiated with the credit institutions. The negotiation is the hard part: the bank's small-business scorecard was not designed to accept a technology readiness level as an input, and the risk committee has to be persuaded that the green score adds information rather than noise. A pilot cohort of ten to twenty firms, tracked through one financing round, is usually enough to settle the question.

Institutionalizing the coordination is the step that determines whether the framework outlasts its sponsors. The coordination mandate is written into the lead agency's terms of reference, the three-tier measurement system is established, and an annual report against the Tier 2 and Tier 3 metrics becomes a statutory obligation rather than a discretionary publication. A public review at the end of the third year, with the explicit purpose of identifying instruments that are not working and reallocating resources to the constraint that binds, closes the loop.

The sequence is deliberately slow at the start and fast at the end. The temptation in practice is to reverse it: announce the framework, launch the guiding fund, and defer the data layer. That sequence produces visible activity and no measurable outcome.

12.3.6 The Governance of the Guiding Fund

The catalytic layer deserves separate governance treatment because it involves the largest fiscal exposure and the most complex principal-agent relationships.

The first governance choice is the fund's legal form. A fund established as a government department is subject to civil service compensation rules that make it difficult to hire investment professionals. A fund established as a state-owned enterprise is subject to state-owned asset supervision rules that may prioritize capital preservation over risk-taking. A fund established as a market-based limited partnership with the government as a limited partner is the form most commonly used, because it allows professional management and market-based compensation while keeping the government at arm's length from individual investment decisions. The arm's-length principle is the governance content of the form: the government sets the mandate, the fund manager makes the investments, and the government does not intervene in individual deals.

The second choice is the performance evaluation. Chapter 8 set out the evaluation framework; the governance point is that the evaluation must be conducted over a horizon that matches the fund's investment horizon. A venture fund with a seven-to-ten-year life cannot be evaluated on its three-year returns. The evaluation should measure the fund against its mandate—private capital mobilized, green technology firms financed, environmental outcomes achieved—rather than against a short-term financial return benchmark that the fund was never designed to meet.

The third choice is the exit and the recycling of proceeds. A guiding fund that returns its proceeds to the general budget has no capital to reinvest and must be recapitalized each cycle, which exposes it to the annual budget process. A fund that recycles its proceeds into new investments has a sustainable capital base but requires a legal structure that permits retention. The choice matters for the fund's ability to commit to long-horizon investments, which is precisely what the catalytic layer exists to provide.

12.3.7 Regulatory Recognition as the Binding Governance Constraint

One governance decision more than any other determines whether the integrated mechanism functions: whether the banking regulator recognizes insurance contracts and guarantees as credit risk mitigants, and under what conditions.

The Basel framework allows banks to recognize certain credit risk mitigants in the calculation of capital requirements, subject to supervisory approval of the mitigation technique and the provider. A weather index insurance contract is not a standard credit risk mitigant, because the payout is triggered by weather rather than by default, and the bank bears the basis risk. A guarantee is a standard mitigant, but its recognition depends on the guarantor's creditworthiness and on the legal enforceability of the guarantee.

The practical consequence is that a bank holding an insurance contract as credit enhancement may receive no capital relief for it, which means the contract does not change the bank's cost of capital, which means the bank has no reason to require it. The coordination link between the insurance layer and the credit layer is severed at the regulatory level, regardless of what the policy framework says.

The governance remedy is for the banking regulator to issue explicit guidance on the recognition of green insurance and guarantee products, specifying the conditions under which they qualify for capital relief. The conditions should include the insurer's solvency rating, the legal enforceability of the contract, the treatment of basis risk, and the disclosure requirements. The guidance does not need to be generous; it needs to be clear. A bank that knows the conditions under which a weather index policy reduces its risk weight can price the policy into the loan, and the borrower can see the benefit.

This is the point at which the framework's layers are most likely to disconnect. The insurance supervisor approves the product. The agriculture ministry subsidizes the premium. The bank wants to use the product. The banking regulator has not issued the recognition guidance, and so the product sits on the shelf. The coordinating authority's most valuable intervention is to close this gap, and it is a gap that can be closed with a supervisory circular rather than a new law.

12.3.8 International Coordination and the Limits of the Framework

The framework described in this chapter operates within a national jurisdiction, and it is subject to two international constraints that the governance design should acknowledge.

The first is the green bond and sustainability-linked loan standards that are set by international market associations and increasingly by foreign regulators. A domestic taxonomy that is not interoperable with the International Capital Market Association's Green Bond Principles or the EU Taxonomy will limit the access of domestic issuers to international investors. The governance response is mutual recognition and mapping rather than harmonization: a domestic taxonomy can differ in its thresholds while mapping to the international standard at the level of categories, so that an issuer can report against both without maintaining two sets of books. The mapping exercise is technical and unglamorous, and it is what determines whether the domestic green bond market is a closed circuit or a channel into global capital.

The second constraint is the emerging set of cross-border disclosure and due diligence requirements, of which the EU's Corporate Sustainability Due Diligence Directive is the most developed example. A domestic exporter that supplies a European manufacturer will face disclosure demands that originate outside the domestic framework, and the domestic framework's proportionality rules will not protect it. The governance response is to build the shared data layer so that it can produce the disclosures that foreign customers require, which in practice means aligning the environmental data fields with the international reporting standards rather than inventing a domestic vocabulary. A jurisdiction that does this gives its firms a compliance asset; a jurisdiction that does not leaves each firm to build its own compliance capability, which the small firms cannot afford.

Neither constraint changes the core architecture. They change the specifications of the information layer, and they add a reporting burden that the governance design should allocate to the data layer rather than to individual borrowers.

12.4 The Policy Package as a Whole

12.4.1 What the Package Looks Like

It is useful to state the package in one place, without the qualifications that the preceding sections have attached to each component. Table 12-2 sets out the instruments, the institutions that administer them, and the coordination links that connect them.

Table 12-2: The integrated policy package

The table is a summary, and like all summaries it hides the sequencing. The information layer must exist before the other layers can use it. The risk-transfer layer must be recognized by the banking regulator before it can change credit pricing. The catalytic layer must have an eligibility screen before it can avoid becoming a generalist fund. The price signals operate continuously and are the background against which the other three layers are calibrated.

12.4.2 The Fiscal Arithmetic

A finance ministry will ask what the package costs. The honest answer is that the cost depends on the coverage and the subsidy rates, and that the largest cost is the catalytic layer, but that the layers have different cost profiles over time.

The information layer is cheap to operate and expensive to build. The taxonomy and the scoring templates require technical work and consultation, which is a one-time cost of a few million dollars in a middle-income jurisdiction. The shared data layer requires investment in the institutions' information systems, which is larger but is largely a reallocation of existing IT budgets. The ongoing cost is the verification and the standard maintenance.

The risk-transfer layer is a recurring subsidy. A weather index program that covers a defined share of planted area at a 40 to 50 percent premium subsidy costs the public budget the subsidized portion of the premium each year, plus the administrative cost and the cost of the public reinsurance layer that sits behind the commercial insurer. The guarantee fund's expected cost is the expected loss on the guaranteed portfolio, which in a well-designed program is a fraction of the guaranteed amount and is recovered in part from the guarantee fees.

The catalytic layer is the largest and the most recoverable. A guiding fund capitalizes at a scale that is one or two orders of magnitude larger than the annual subsidy of the risk-transfer layer, and it recovers its capital over the fund's life, subject to investment performance. The fiscal risk is concentrated in the early years, and the governance design should ensure that the fund's capitalization is a multi-year commitment rather than an annual appropriation.

The arithmetic that matters for the ministry is the leverage ratio: how much private capital and how much additional lending does each unit of public money mobilize? A guarantee that mobilizes ten units of lending per unit of expected loss is a better use of fiscal space than a direct subsidy that mobilizes one. The measurement architecture in Section 12.3.3 is what produces this ratio, and the governance design should require it to be reported.

12.4.3 Objections and Responses

Three objections to the framework are predictable, and each deserves a direct answer.

The first objection is that the framework is too complex for the administrative capacity available. The response is that the complexity is in the coordination, not in the individual instruments. A jurisdiction can begin with one segment and one region, as the worked sequence in Section 12.3.5 specifies, and expand as the capacity builds. The alternative—a simple framework that ignores the interactions—is the single-channel approach that Chapter 1 showed to be insufficient. The complexity is the price of addressing the failure that the single channel leaves unaddressed.

The second objection is that the framework subsidizes what the market should finance. The response distinguishes between the layers. The information layer is a public good and is appropriately publicly provided. The risk-transfer layer subsidizes risks that private markets cannot price at a level borrowers can afford, and the subsidy is targeted at the segment where the market failure is documented. The catalytic layer fills a financing gap at the early stage, and the design requirement is that it co-invests with private capital rather than substituting for it. A framework that subsidizes mature, bankable projects has overreached; a framework that subsidizes the information and the early-stage risk has not.

The third objection is that the framework will be captured by incumbents. The response is the transparency requirement in Section 12.3.4: published criteria, published recipient lists, and an independent evaluation. Capture is a real risk, and it is managed by disclosure rather than by the assumption that it will not occur.

12.4.4 What the Framework Does Not Do

It is worth being explicit about the limits. The framework does not create demand for green finance where the underlying project is not viable. A tea cooperative with no market for its tea will not become viable because it has a score and an insurance policy. The framework improves the allocation of capital among viable projects; it does not manufacture viability.

The framework also does not resolve the distributional question of who pays for the transition. The instruments shift some costs to the public balance sheet and some to the borrowers, and the incidence depends on the subsidy rates and the pricing. A framework that is efficient in the aggregate can still leave the poorest borrowers facing higher costs, and the governance design should include an explicit assessment of the incidence. The assessment is not a substitute for the distributional decision, which is political; it is the information the decision requires.

And the framework does not guarantee that the environmental outcomes are achieved. The scoring, the insurance, and the catalytic capital all operate through the behavior of the financed firms, and the firms may adopt the practices that the score rewards without achieving the environmental outcome that the score proxies for. The Tier 3 measurement in Section 12.3.3 is the check, and it is the reason the shared data layer matters: without it, the framework measures its own activity and not the world.

12.5 Conclusion: The Framework as a Diagnostic Discipline

The tea cooperative's default rate stayed below the union's existing book. That is the fact to hold onto, because it is the one number in this chapter that came from the world rather than from a design document. The cooperative was not approved because a new institution was created. It was approved because a scoring template, an insurance contract, and a decision to treat environmental performance as information were combined in a way that changed the lender's calculation.

Whether the same combination scales is the open question. The provincial union had three advantages that a national framework does not automatically confer: a single decision-maker, a homogeneous borrower population, and a two-year horizon over which the loss experience could be observed. A national framework has none of these. It has multiple regulators with separate mandates, a borrower population that ranges from listed manufacturers to tea cooperatives, and a political cycle that rewards announcement over evaluation. The governance design in Section 12.3 is an attempt to manufacture, administratively, the conditions that the provincial union had by accident.

The most important element of the framework is not any single instrument. It is the diagnostic discipline: identify the constraint that binds before prescribing the instrument, and calibrate the instrument from observed data rather than from another jurisdiction's parameters. That discipline is what distinguishes a framework that moves the 95 percent from one that expands the 5 percent. It is also the element most likely to be lost in implementation, because diagnosis takes time and announcement takes a press release.

The next chapter tests the framework against jurisdictions that have attempted the same combination under different institutional conditions. The comparison is not a beauty contest between national models. It is a test of which governance choices travel and which are artifacts of a particular administrative system. The tea cooperative will appear again in that comparison, as a reminder that the test is whether the lender's calculation changes, not whether the framework reads well.