Chapter 15: Toward an Integrated Green Finance Ecosystem: Conclusions and Policy Directions

A green credit line and a green bond can both be labelled correctly and still leave a steel plant's decarbonization unfinanced. The label is not the problem. The problem is that the three functions a financial system must perform for a transition—identifying which firms and projects are genuinely green and how green, buffering the risks that green investment carries, and catalyzing capital into early-stage green technology—are performed by three different sets of institutions that in most jurisdictions do not talk to one another. This book has argued that the transition requires those functions to operate as one mechanism. I will not restate the argument in full. I want to close it by showing what the mechanism looks like when assembled, what each actor must do to assemble it, and where the evidence still runs out.

15.1 Synthesis: What the Three Pillars Actually Establish

The three lines of analysis in this book were chosen because each addresses a failure the others cannot. Credit rating answers the question "how green, and how risky." Insurance answers the question "who absorbs the loss when the weather or the fire or the pest arrives." Sci-tech finance intermediation answers the question "who funds the ten-year development program that no loan officer will underwrite." A financial system that performs only the first identifies well and finances poorly. One that performs only the second buffers risk without directing capital. One that performs only the third catalyzes projects that no one can evaluate and no one can insure. Integration is not coordination for its own sake. Each pillar supplies the precondition the next one needs.

15.1.1 Rating as the identification layer

The argument of Chapters 3 and 4 can be compressed into a single claim: environmental performance is measurable at the firm level with the same instruments credit analysts already use, and the measurement improves rather than degrades default prediction when it is done properly. The technology-firm rating work that preceded this book established that intangible-heavy firms can be scored on a structured grid of technical capacity, market position, and management quality, and that the resulting scores discriminate among default outcomes. Extending that grid with environmental variables—emissions intensity relative to sector benchmark, energy sourcing, compliance history, transition-plan credibility—does not require a new rating methodology. It requires additional rows in an existing scorecard and a discipline about which rows carry weight.

The discipline matters more than the rows. A rating that folds environmental performance into a single composite score with financial variables lets a strong balance sheet mask a brown technology, and lets a weak balance sheet punish a firm that is doing the right thing. The design that works separates the environmental assessment into a modifier that adjusts the financial score within bounded limits, and discloses the modifier separately. A firm can then be told: your financial score is BB, your environmental modifier is negative, your effective rating is B. That sentence is actionable in a way that a composite grade is not.

Chapter 5 examined whether the modifier should be symmetric—whether superior environmental performance should lift a rating as well as poor performance depress it. The evidence suggests the lift is real but small, and that the depression is larger and better documented. Asymmetric modifiers are defensible on prudential grounds. They are also a political choice, and regulators should say so.

15.1.2 Insurance as the buffering layer

Chapters 6 through 8 developed the risk chain from identification to mitigation to transfer, and the empirical core of that development is the weather index insurance and forest insurance work. The finding that matters for the ecosystem is not that index insurance pays out when rainfall falls below a trigger. It is that index insurance changes the borrowing behavior of the insured. An agricultural producer with a weather index policy can carry debt through a bad season without liquidating assets, which means the lender's expected loss on that producer falls, which means the lender can lend at a lower spread or to a producer it would otherwise decline. The insurance does not merely transfer risk. It changes the credit assessment upstream.

The same logic applies to forest insurance, with a complication. Forest risk is correlated across a region in a way that crop risk is not always. A single fire season can trigger every policy in a province. The reinsurance layer therefore carries more weight in forestry than in agriculture, and the design of the public reinsurance backstop determines whether the primary insurer can offer the product at all. Chapter 8 worked through the economics of that backstop. The conclusion is that a forest insurance market without a defined public catastrophe layer is a market that will retreat after the first bad year.

15.1.3 Sci-tech finance as the catalysis layer

Chapters 10 through 12 described the intermediary institutions—venture funds, guiding funds, technology banks, intellectual property exchanges—that move capital into firms too early for credit and too risky for public markets. The distinguishing feature of these institutions is that they underwrite technology, not collateral. A guiding fund that invests only in firms with three years of audited revenue is a private equity fund with a public mandate and a lower return. A guiding fund that invests in pre-revenue technology firms is performing the catalysis function, and it will have losses. The design question is how to budget for those losses without turning the fund into a subsidy channel.

Chapter 11's analysis of market failure in early-stage green capital is the relevant reference. The gap is not a shortage of capital as such. It is a shortage of capital willing to accept the duration and the technology risk of green R&D. Guiding funds close that gap when they take the first-loss tranche or the longest-duration tranche of a fund structure, and when their investment decisions are made by people who can read a technology roadmap. They widen the gap when they are structured as grant programs with an equity label.

15.1.4 The integration

Figure 15.1 sets out the mechanism as a single flow. The point of the diagram is the feedback loops, not the boxes.

flowchart TD
    A[Firm or project] --> B[Rating: environmental modifier]
    B --> C{Credit decision}
    C -->|Pass| D[Loan or bond with green covenant]
    C -->|Fail| E[Referral to sci-tech finance]
    E --> F[Guiding fund / venture tranche]
    F --> B
    D --> G[Insurance: weather, forest, performance]
    G --> H[Risk-adjusted spread]
    H --> C
    G --> I[Reinsurance / catastrophe layer]
    I --> G
    F --> J[Exit: IPO, trade sale, buyback]
    J --> K[Recycled capital]
    K --> F
    B --> L[Disclosure to regulator]
    L --> M[Portfolio supervision]
    M --> C

The loop from insurance back to the credit decision is the one most systems lack. A lender that prices a green loan without asking whether the underlying risk is insured is pricing an unhedged exposure and calling it green. The loop from the guiding fund back to the rating is the one most systems lack in the other direction. A technology firm that fails a credit test today may pass it in three years if its development program succeeds, and the fund that carries it through those three years needs a way to hand it back to the banking system. Without that handoff, the fund is a permanent owner of firms that should have graduated to conventional finance.

The diagram compresses a set of decisions that deserve to be walked through in prose, because the arrows do not show who decides what. Take the insurance-to-credit loop first. A lender receives a loan application from an agricultural processor. The lender's credit analyst computes a financial score and an environmental modifier. Before the loan is priced, a risk officer asks whether the processor's input supply is insured against weather. If it is not, the risk officer has three options: require the insurance as a condition of the loan, price the loan as an unhedged exposure, or decline. The choice is not a technical one. It depends on whether the lender has an insurance product available to require, which depends on whether the insurer has a reinsurance backstop, which depends on whether the fiscal authority has pre-funded the catastrophe layer. A single pricing decision at a single loan desk traces back to a budget appropriation made years earlier.

Now take the fund-to-rating loop. A technology firm fails the credit test. The loan officer refers it to the bank's technology finance arm. The fund's investment committee decides whether to invest. If it invests, it sets a holding period and a set of milestones—certification, pilot production, first commercial contract. At each milestone, the fund can ask the credit arm to re-run the rating. The re-run is not automatic. Someone at the fund has to decide that the firm is ready for a second look, and someone at the credit arm has to agree to spend the analyst time. The handoff is a decision point, not an arrow. In the systems that work, the decision point has a name, a schedule, and a person responsible for it. In the systems that do not, the firm stays in the fund's portfolio until the fund's holding period expires, and then it is sold to whoever will buy it.

The regulator's loop is the one that closes the system. The rating modifier is disclosed to the supervisor, the supervisor compares modifiers across banks, and the comparison feeds back into the methodology. This loop is the weakest in practice, because supervisors rarely have the data to compare and rarely have the mandate to require the comparison. Where the loop exists—the European Central Bank's supervisory review of climate risk practices is the closest example—it has changed bank behavior, but slowly and unevenly.

15.2 The Evidence Base: What Fourteen Chapters Established and What They Did Not

Before turning to policy, I owe the reader an accounting of the strength of the claims. The book draws on three research programs, and they are not equally well evidenced.

The technology-firm rating work rests on a dataset of rated firms with observed outcomes. The environmental extension rests on a smaller dataset and on proxy variables for environmental performance that are less reliable than the financial variables. The direction of the findings is consistent across specifications. The magnitude is not precisely estimated. A policy that depends on the modifier moving a rating by a specific number of notches is a policy built on a point estimate that the data cannot support.

The weather index insurance work is the strongest of the three empirically. It has randomized or quasi-randomized variation in insurance access, observed income and borrowing outcomes, and a replication record. The finding that insurance stabilizes income and improves credit access is robust. The finding that it induces a shift toward higher-return production is robust in some settings and not in others, and the difference appears to depend on whether the insurance is bundled with input credit or extension advice.

The forest insurance work is the weakest. The data are administrative, the sample is small, and the counterfactual is constructed rather than observed. I have reported the findings because they are the best available and because the policy question is live. I would not want a reader to treat the elasticity estimates in Chapter 8 as settled.

The sci-tech finance material is largely institutional and case-based. It describes how guiding funds, technology banks, and IP exchanges are structured and what they appear to have achieved. It does not support causal claims about their effect on innovation output. The cross-jurisdiction comparison in Chapter 10 is suggestive, not identified.

This accounting is not a disclaimer. It is the basis for the policy recommendations that follow. Recommendations are calibrated to the strength of the evidence behind them: where the evidence is strong, the recommendation is specific; where it is weak, the recommendation is to build the measurement capacity that would make a specific recommendation possible.

15.3 Policy Recommendations

The recommendations below are organized by actor. They are not independent. A rating modifier without a disclosure requirement is unenforceable. An insurance product without a reinsurance backstop is unavailable after the first catastrophe. A guiding fund without an exit mechanism is a fiscal liability with a green label. I have tried to mark the dependencies where they matter.

15.3.1 For government and regulators

Mandate the separate disclosure of environmental modifiers in credit ratings used for regulatory capital purposes. The single most consequential regulatory change available is also the least dramatic. If a bank's internal rating of a corporate borrower must disclose the environmental modifier separately from the financial score, the modifier becomes auditable. Auditors will ask how it was computed. Supervisors will compare modifiers across banks. The modifier will converge toward a defensible methodology, not because a regulator specified the methodology but because disclosure creates the conditions for comparison.

The European Union's experience with the Taxonomy Regulation is instructive here, and not entirely encouraging. The Taxonomy created a common language for what counts as sustainable, and in doing so it created a compliance industry organized around classification rather than around measurement of transition progress. A firm can be taxonomy-aligned on the basis of its current revenue mix and still have no plan for the next decade. The disclosure mandate I am proposing is narrower and more useful: it asks not "is this activity sustainable" but "how did the lender assess this borrower's environmental performance, and what did the assessment change."

Define the public reinsurance layer for forest and catastrophe-exposed agricultural risk. Chapter 8's analysis leads to a specific design. The public layer should attach above a retention that the primary market can bear, should be triggered by an index rather than by loss adjustment, and should be priced at a level that does not crowd out private reinsurance for the layers below it. The trigger design matters more than the price. A loss-adjusted public layer requires an assessment infrastructure that most forest jurisdictions do not have, and the assessment lag defeats the purpose of the backstop.

How the public layer should be financed is a question I will return to. The short answer is that it should be pre-funded through a levy on the insured exposure, not appropriated after the event. Post-event appropriation is cheaper in the year it is not spent and more expensive in every year it is.

Structure guiding funds with a defined first-loss tranche and a defined exit obligation. The failure mode of public venture capital is well documented and does not need restating. The design that avoids it has three features. The public capital takes the first-loss position in a fund structure that includes private capital, so that the private partner has an incentive to select and monitor. The fund has a stated holding period after which it must exit, by trade sale or by sale into a secondary market, rather than an indefinite mandate. And the fund's managers are compensated on realized returns, not on assets under management.

The Chinese guiding fund experience described in Chapter 11 is the largest natural experiment in this design question. The funds that performed the catalysis function were the ones with private co-investors and defined exit horizons. The funds that became fiscal channels were the ones with neither.

Require that green credit programs report a transfer ledger. Chapter 14 made the case for evaluation built into program design. The transfer ledger is the simplest version of that case. Every green credit program should report, at the level of the individual loan, the environmental performance of the borrower before and after the loan, the insurance status of the underlying risk, and the source of the funds. The ledger is cheap to build at origination and expensive to reconstruct. The European Investment Bank's climate lending reporting is the reference standard, and it is not a high bar. Most programs do not meet it.

15.3.2 For financial institutions

Separate the environmental assessment from the credit score, and staff the assessment separately. A bank that assigns environmental assessment to the credit analyst who also assigns the financial score will produce an environmental assessment that is correlated with the financial score. That correlation is the problem. The assessment needs to be performed by someone whose job is to understand the technology and the transition plan, and the result needs to enter the credit decision as a bounded modifier. The organizational separation is more important than the methodology. A good methodology applied by the wrong person produces the wrong answer.

Price the insurance status of the exposure. A green loan to an agricultural processor whose input supply is uninsured is a loan with an embedded weather exposure. The lender can require the insurance as a condition of the loan, or it can price the exposure. Requiring the insurance is usually better for both parties, because the lender's requirement creates the demand that makes the insurance market viable. The bundling finding from the index insurance literature applies here: insurance that is required as a condition of credit is insurance that gets purchased.

Build the handoff from the venture portfolio to the credit portfolio. A bank with a technology finance arm and a corporate banking arm should have a defined process by which a firm that has been financed by the venture arm and has reached commercial traction is transferred to the corporate arm for conventional credit. Without that process, the venture arm's successful firms leave the bank when they refinance elsewhere, and the bank captures none of the upside of the risk it took. The handoff is a rating problem: the firm that was unrateable at inception should be rateable at graduation, and the rating methodology should be the same one the corporate arm uses.

15.3.3 For firms

Treat the environmental modifier as a financing variable, not a compliance obligation. The firm that computes its environmental performance the way its lenders will compute it has an advantage in negotiation. The firm that waits for the lender's assessment has no way to contest an unfavorable one. The specific action is to build the emissions and energy data at the facility level, in the format the lender's methodology requires, before the loan application rather than after.

Insure the risk that the balance sheet cannot absorb. The small and mid-sized firm's balance sheet cannot absorb a bad season. The insurance decision is a financing decision, not a risk-avoidance decision. The firms that carry weather or forest insurance borrow more and borrow cheaper, and the difference in borrowing cost exceeds the premium in the settings where the data exist.

15.4 The Design of Incentives Without Distortion

Chapter 13 worked through the incentive design problem in detail. The conclusion I want to carry forward is that the distortion risk is concentrated in one place: the point at which a green label changes the price of capital without changing the environmental outcome.

The mechanism is straightforward. If a green loan carries a subsidized spread, a firm has an incentive to obtain the label. If the label is granted on the basis of a plan rather than a performance, the firm has an incentive to write a plan. If the plan is not verified against outcomes, the label becomes a document rather than a fact. The subsidy is then transferred to firms that would have made the investment anyway, and the transition that the subsidy was meant to accelerate does not accelerate.

The design that avoids this has three features, and they are the same features that appear in the evaluation discussion. The subsidy should attach to the outcome, not the label. The outcome should be measured against a counterfactual, not against an absolute threshold. And the measurement should be built into the program at origination.

The counterfactual requirement is the one that gets dropped in practice, because it is the one that makes the program look less successful. A green credit program that reports the emissions reduction of its borrowers without reporting what those borrowers would have done without the program is reporting a number that cannot be interpreted. The borrowers may have reduced emissions because the program paid them to, or because the price of energy rose, or because the equipment they would have bought anyway became cheaper. The program's evaluator cannot distinguish among these without a comparison group.

The political economy of this is worth naming. The reason counterfactual evaluation gets dropped is not that it is technically difficult. It is that it produces findings that are less useful to the program's sponsors. The jurisdictions that have built it in have done so because a central budget authority required it, not because the program agency wanted it.

15.5 Future Research Directions

The research agenda that follows from this book has four parts. I list them in order of how soon the answers are needed, not in order of intellectual interest.

15.5.1 The environmental modifier and default prediction

The central open question is whether the environmental modifier improves default prediction out of sample, and by how much. The in-sample results in Chapter 5 are encouraging. The out-of-sample test requires a dataset that spans a credit cycle long enough to include defaults among firms with strong and weak environmental performance, and that dataset does not yet exist in most markets. The test is buildable. It requires a lender with a long rating history to release anonymized rating and outcome data with the environmental modifier included. The first lender to do so will produce the paper that settles the question.

A related question concerns the direction of causality. A firm with weak environmental performance may default because the environmental weakness is a symptom of poor management, or because the environmental weakness exposes it to regulatory or physical risk. The two mechanisms imply different policies. The first implies that the modifier is a management-quality proxy and adds little to a well-built financial score. The second implies that the modifier captures a risk the financial score misses. The data to distinguish them exist in jurisdictions with a carbon price, because the price creates the exposure that the second mechanism requires.

15.5.2 Index insurance in the presence of basis risk

The index insurance literature has a basis risk problem that the income stabilization findings do not resolve. A producer whose loss is not correlated with the index receives no payout in the year of the loss. The income stabilization result in Chapter 7 is an average across producers with different basis risk exposures. The question is whether the stabilization holds for the producers with the highest basis risk, who are the ones the product most needs to reach.

The research design that would answer this requires a product with a measurable basis risk distribution and a sample large enough to estimate the effect separately for high-basis-risk and low-basis-risk producers. The data exist in the weather index programs that have been running for a decade. The analysis has not been done at the necessary granularity.

15.5.3 The causal effect of sci-tech finance on innovation

The guiding fund literature is descriptive. The causal question—do guiding funds increase green patenting, and by how much per dollar—has not been answered with a credible identification strategy. The obstacle is selection: funds invest in firms that are already more likely to innovate. The identification strategies available are the ones used in the venture capital literature: discontinuities in fund eligibility, variation in fund size across otherwise similar regions, and the timing of fund establishment relative to firm formation. None of these has been applied to green guiding funds at scale.

The question is worth answering because the policy stakes are large. If guiding funds have no causal effect on innovation, the case for them rests on their effect on firm survival or on the supply of risk capital, which are different and smaller claims. If they have a large effect, the case for expanding them is strong and the design question is how to expand without losing the features that produce the effect.

15.5.4 The interaction effects

The most interesting questions are about interactions, and they are the least studied. Does insurance access increase the take-up of green credit? Does a rating modifier change firm behavior before the loan is priced? Does a guiding fund investment increase a firm's subsequent credit rating? Each of these is a question about the mechanism as a whole rather than about one pillar, and each requires data that spans institutions.

The rating-and-insurance interaction is the most tractable. A lender that requires insurance as a condition of a green loan generates a dataset in which the insured and uninsured borrowers are observable, and the difference in default rates between them is estimable. The first lender to run that comparison will have an answer within a credit cycle.

15.6 The Institutional Preconditions

The recommendations above assume a set of institutional capacities that not every jurisdiction has. I want to be explicit about them, because a policy recommendation that ignores institutional capacity is a recommendation that will be implemented as a form.

The first precondition is a credit registry with loan-level outcomes. Without it, the rating modifier cannot be validated and the green credit program cannot be evaluated. The registry does not need to be public. It needs to exist and to be accessible to the supervisor and to researchers under confidentiality.

The second precondition is an actuarial capacity in the insurance supervisor. Index insurance and forest insurance are not conventional products, and a supervisor that regulates them with the tools designed for motor insurance will either prohibit them or fail to notice when they are insolvent. The capacity is a small team of people who understand the products. It is buildable in two years.

The third precondition is a securities market that can absorb exits. A guiding fund without an exit is a fund that holds its portfolio indefinitely. The exits available are trade sales to strategic buyers, sales to secondary funds, and public listings. The first is available in most jurisdictions. The second requires a secondary market that most emerging jurisdictions lack. The third requires a listing venue with a technology-friendly listing standard. The Chinese experience with the STAR Market is the most relevant recent example, and it is too early to judge.

The fourth precondition is a fiscal authority willing to pre-fund the catastrophe layer. This is the hardest one, because it requires a budget appropriation today for a loss that may not occur for a decade. The jurisdictions that have done it have done so after a catastrophe that made the cost of not doing it visible. The argument for doing it before the catastrophe is an argument that has to be made in budget terms: the expected annual cost of the pre-funded layer is lower than the expected cost of post-event appropriation, because the post-event appropriation is made under duress and at a higher price.

15.7 What the Integrated Ecosystem Looks Like in Operation

I want to close with a description of the mechanism in operation, because the abstraction of three pillars can obscure what the integration means for a single firm. Consider a mid-sized manufacturer of building insulation materials in a jurisdiction with a green finance pilot program. The firm wants to convert a production line to a bio-based input. The conversion costs 40 million, takes three years, and the output sells at a premium that depends on certification the firm does not yet have.

In a single-channel system, the firm applies for a green loan. The loan officer has no way to assess the bio-based process, the certification is uncertain, and the collateral is the existing line. The loan is declined, or approved at a spread that reflects the uncertainty. The conversion does not happen, or happens three years later when the certification is standard.

In the integrated system, the sequence is different. The firm's rating is computed with the environmental modifier, which credits the existing line's energy performance and penalizes the emissions intensity of the current process. The modifier is negative, and the firm's effective rating is below the threshold for the green credit line. The loan officer refers the firm to the bank's technology finance arm, which has a guiding fund co-investment mandate. The fund invests 12 million in the conversion, taking a first-loss position alongside 8 million of private capital. The bank's credit arm lends 20 million against the existing line, secured, with a covenant that the loan converts to a green loan at a lower spread if the certification is obtained within three years. The bank requires the firm to carry business interruption insurance on the conversion period and performance insurance on the bio-based output. The insurance is priced off an index of certification outcomes for similar processes, which the insurer has built from the pilot program's data.

Three years later, the certification is obtained. The firm's environmental modifier turns positive. The effective rating crosses the threshold. The loan converts to a green loan at the lower spread. The guiding fund exits by selling its stake to a strategic buyer at a multiple. The bank's credit arm now holds a green loan to a firm it would have declined three years earlier. The insurer has a profitable book of performance policies. The pilot program has a transfer ledger that shows the emissions reduction, the insurance claims, and the fund's realized return.

None of the individual transactions in this sequence is novel. Each is a standard instrument. What is novel is the sequence, and the sequence is what the ecosystem means. The rating identified the firm as not-yet-green. The fund carried the technology risk. The insurance carried the performance risk. The credit converted when the identification changed. The exit recycled the capital. The ledger recorded the whole thing.

The sequence requires the four institutional preconditions, and it requires the three actors to have a reason to cooperate. The bank's reason is the conversion of a declined applicant into a green loan. The fund's reason is the deal flow. The insurer's reason is the data and the premium. The regulator's reason is the ledger. Every one of those reasons is a private return, not a public one. An ecosystem that depends on altruism is a program. An ecosystem that depends on the participants' ordinary incentives is a system.

15.8 The Limits of What Finance Can Do

I want to be careful about one thing before closing. The argument of this book is that the financial system can be organized to support the green transition more effectively than it currently is. Finance allocates capital among feasible options; it does not create them. The transition requires a carbon price that makes the brown process expensive, a regulatory framework that makes the green process legal and certifiable, and a technology base that makes the green process feasible.

The distinction matters for policy because the failure mode of green finance programs is to substitute for the carbon price rather than to complement it. A subsidized green loan is a small carbon price applied to a small set of investments. It is a less efficient instrument than a carbon price applied to the whole economy, and it is more politically achievable. The two are not substitutes. A jurisdiction that uses green credit subsidies as a reason to delay the carbon price is using the instrument to avoid the policy.

The same caution applies to the rating modifier. A rating that penalizes emissions intensity is a private carbon price applied to the cost of debt. It is a useful signal and a weak instrument. The firms that respond to it are the firms that are already close to the margin. The firms that do not respond are the firms whose cost of capital is insensitive to their environmental performance, which in most markets means the large firms with access to bond markets and the state-owned firms with implicit guarantees. The modifier reaches the middle of the distribution. The carbon price reaches the whole of it.

15.9 The Next Step

The book's central claim is that the three functions—rating, insurance, catalysis—must operate as one mechanism, and that the mechanism is buildable with existing instruments. The claim is a claim about design, not about resources. The instruments exist. The institutions exist in most jurisdictions in some form. What is missing is the sequence and the ledger.

Sequence is a matter of institutional arrangement, and the arrangements are specific. A bank that separates environmental assessment from credit scoring. An insurer that prices an index. A fund that takes a first-loss position and exits on a schedule. A regulator that requires the modifier to be disclosed. A budget authority that pre-funds the catastrophe layer. Each of these is a decision that someone can make this year.

The ledger is a matter of accounting, and it is the cheapest of the recommendations. Every green finance program should be able to report, at the level of the individual transaction, what was financed, what the environmental performance was before and after, whether the risk was insured, and what the counterfactual was. The programs that cannot report these things are not programs that can be evaluated, and programs that cannot be evaluated are programs that cannot be defended when the fiscal pressure arrives.

The next step is not a new instrument. It is the ledger, and the sequence, and the willingness to report what the ledger shows. The jurisdictions that build them will have a financial system that supports the transition. The jurisdictions that do not will have a financial system that supports the label. The difference will show up in two places: the default rate on green loans, and the budget line for the next catastrophe.