Chapter 1: The Identification Problem in Green Finance

In March 2007 the European Investment Bank issued a bond whose proceeds were earmarked for renewable energy and energy efficiency projects, with an annual reporting commitment attached. The Climate Awareness Bond was the first environmental-themed bond from a multilateral institution, and its rulebook was strict for its time: eligible projects had to be defined in advance, reviewed independently, and reported on every year. Five years later the World Bank's green bond program adopted the same structure. By the middle of the following decade the market had grown from a curiosity into a segment measured in hundreds of billions of dollars a year.

That growth is usually told as a story about investor appetite. The more useful story is about paperwork. Every green bond rests on a claim about what the money will do, and the claim has to be checked. An issuer defines eligible projects, a reviewer confirms the definition, an auditor or second-party opinion examines whether the money went where the issuer said. The bond itself is a contract about cash flows. The green label is a contract about intent. Finance has always been comfortable pricing the first. It has never been entirely comfortable with the second.

That discomfort is the subject of this book. Industry's green transition requires enormous capital to move from one set of activities to another, and the financial system is the mechanism through which the movement happens. But the financial system is built to evaluate a particular kind of information: cash flows, collateral, the probability that a borrower will repay. Environmental performance is a different kind of information. It is often unmeasured, sometimes unmeasurable with current tools, and almost always separable from the borrower's ability to service a loan in the short run. A steel mill that has not yet installed a low-emissions furnace can repay its debt on time for years. A software company with a spotless carbon footprint can go bankrupt in a quarter.

The identification problem, stated plainly: the financial system cannot reliably tell which firms are genuinely making the transition and which are merely talking about it. A second problem follows. If you cannot identify the transition, you cannot finance it well. You can finance something and call it green, but you cannot be confident you have done what you set out to do. The gap between what capital markets claim to fund and what they actually fund is not a moral failing of bankers. It is a measurement problem, and measurement problems have architectural solutions.

This chapter sets out the shape of the problem and the outline of the solution the book develops. The solution has three parts: rating, buffering, and catalysis. Rating is the work of building an evaluation framework that places environmental performance alongside financial performance in a single assessment of a firm. Buffering is the use of insurance instruments to absorb the risks that make green projects hard to lend to. Catalysis is the design of intermediary systems that move capital toward green innovation earlier than ordinary credit markets reach.

Before any of that can be argued, we need to be clear about why the transition is a financial problem at all, and why the financial system's ordinary tools fail on it.

1.1 The Green Transition Imperative

1.1.1 What transition means, and why it is not a single event

The word "transition" suggests a passage from one state to another, with a beginning and an end. Industrial decarbonization does not work that way. A cement plant that replaces part of its fuel with biomass has made a transition, but it remains a cement plant with substantial emissions. A shipping company that orders dual-fuel vessels has moved toward lower emissions, but the vessels will burn fossil fuel for most of their service life. Transition is a direction of travel measured in increments, not a status a firm either has or lacks.

This matters for finance because financial instruments are designed around discrete states. A loan is performing or non-performing. A bond is investment grade or speculative. A firm is solvent or insolvent. Environmental performance does not fit these categories. It is continuous, multidimensional, and frequently improving without yet being good. A rating framework that treats it as binary will misclassify most of the firms it examines.

The European Union's taxonomy regulation, which took effect in 2020, tried to solve this by defining what counts as a sustainable economic activity. The regulation sets technical screening criteria for dozens of sectors and requires that an activity contribute substantially to one of six environmental objectives while doing no significant harm to the others. It is serious work, and it has been criticized from both directions: too strict for heavy industry, too loose for environmental groups. The criticism is itself instructive. The taxonomy is an attempt to draw a line through a continuum, and any line drawn through a continuum will be contested at the margin.

China's green bond catalog, first issued in 2015 and revised since, took a different approach. It listed eligible project categories—clean energy, pollution control, resource conservation—and initially included categories such as the "clean utilization of coal" that foreign investors found difficult to reconcile with international definitions. The divergence was a practical problem for issuers who wanted to sell to both audiences. A bond that qualified as green in Shanghai might not qualify in London. The same physical project could carry two different labels depending on which rulebook was applied.

These definitional differences are not bureaucratic noise. They reflect a real question: what is the unit of analysis? The EU taxonomy evaluates activities. The Chinese catalog evaluates projects. The Climate Bonds Initiative, a London-based nonprofit, evaluates assets and has developed sector-specific criteria for wind, solar, rail, and buildings. Each unit implies a different kind of evidence. An activity-level assessment asks what the firm does. A project-level assessment asks what the money funds. An asset-level assessment asks what the physical thing is.

For a lender deciding whether to extend credit to a mid-sized manufacturer, none of these units maps cleanly onto the decision. The lender needs to know whether the borrower will repay. The borrower's environmental performance matters to that question only through specific channels: regulatory risk, energy cost exposure, technology obsolescence, access to green-labeled capital. Those channels are what a credit-oriented rating has to capture. A taxonomy designed for bond labeling does not capture them, because it was not designed to.

1.1.2 The scale of capital required

The International Energy Agency's World Energy Investment 2021 put the additional annual investment needed to reach net-zero emissions by 2050 at roughly four trillion dollars, with a substantial share required in emerging and developing economies. The precise number depends on assumptions about technology costs and growth. The order of magnitude is not in dispute: the transition requires capital on a scale public budgets cannot supply alone.

Private capital has to do most of the work. The claim is repeated so often it has lost its edge, but the arithmetic behind it is straightforward. Government spending on energy transition in most countries runs in the low single digits as a share of GDP. The investment required runs higher. The gap has to be filled by firms investing their own funds, by banks lending to them, and by capital markets buying their securities. Whether that happens depends on whether returns are attractive relative to alternatives and whether the risks can be assessed and managed.

Here the picture becomes less encouraging. A 2020 study by the Bank for International Settlements examined the pricing of green bonds and found that they traded at a small premium, meaning issuers paid slightly less to borrow. That premium, the "greenium," is good news for issuers. It is small, typically a few basis points, and it suggests investors will pay something for the label. A few basis points does not close the funding gap for a steel decarbonization project with a fifteen-year payback and uncertain technology risk. The greenium helps at the margin. It does not solve the problem at the center.

The center of the problem is risk. Green projects in heavy industry carry technology risk (will the process work at scale?), policy risk (will carbon prices hold?), and market risk (will there be buyers for green steel at a premium?). Ordinary credit analysis can handle market risk. It struggles with technology and policy risk because those risks are not captured in historical financial statements. A firm with a strong balance sheet and a weak technology bet looks creditworthy right up until the technology fails.

1.1.3 Why the financial system is the right lever

One could argue that the transition should be funded by governments directly, through grants and public procurement, and that the financial system's role is secondary. There is a case for this. Public procurement of green cement or green steel can create the initial demand that makes private investment viable. But public funding has a limit private funding does not: it is subject to annual budget cycles, and it cannot easily absorb the risk of a single large project failure. A government that funds a failed demonstration plant faces a political cost. A diversified lender or investor faces a portfolio loss.

The financial system's advantage is its capacity to pool and diversify risk. A bank that lends to fifty industrial firms, each with a different transition pathway, holds a portfolio whose expected loss is more predictable than any single loan. An insurance company that underwrites weather risk across a region can price that risk more accurately than any individual farmer. A venture fund that invests in ten clean technology startups expects nine to fail and one to succeed, and prices the portfolio accordingly.

This capacity is not automatic. It requires that the risks be measurable, that the instruments be correctly designed, and that the institutions have the information they need to price and monitor. Where those conditions fail, the financial system does not diversify risk; it misprices it, and then discovers the error when losses arrive. The 2008 financial crisis was, among other things, a failure of risk measurement in mortgage markets. The green transition carries a similar hazard. If lenders classify loans as green without a way to verify the environmental performance of the borrower, they will accumulate exposure to firms that are not transitioning, and the resulting losses will discredit the entire category.

So the financial system is the right lever, but it is a lever that requires a fulcrum. The fulcrum is information. Without a way to identify which firms are genuinely transitioning, the financial system will fund the appearance of transition rather than the substance.

1.2 Defining the Identification Problem

1.2.1 Two firms, one loan officer

A loan officer at a regional commercial bank has two applicants in front of her. The first is a mid-sized textile manufacturer that has installed a wastewater recycling system, switched part of its boiler fuel to natural gas, and signed a contract to buy renewable electricity. Its emissions per unit of output have fallen by a third over five years. Its financial statements show thin margins, high leverage, and a customer base concentrated in two large buyers. The second applicant is a logistics company with a modern fleet, strong cash flow, low debt, and a carbon footprint that has not changed in a decade. It has no plans to electrify its trucks because the routes it serves are too long for current battery technology.

Which firm is the green borrower? The first, by any reasonable environmental measure. Which firm is the better credit risk? The second, by any reasonable financial measure. A loan officer evaluated on default rates will favor the second. A bank evaluated on the green share of its portfolio will favor the first. If the bank's green targets are not tied to credit performance, it will accumulate environmental progress and financial fragility at the same time. If the targets are tied to credit performance, the first firm may not qualify for the green label at all, because its financial weakness disqualifies it.

This is the identification problem in miniature. Environmental performance and creditworthiness are correlated, but they are not the same thing, and they can point in opposite directions. A framework that collapses them into a single score without being explicit about the trade-off will produce decisions that are defensible on one dimension and indefensible on the other. A framework that keeps them separate but does not integrate them will leave the loan officer with two assessments and no way to combine them.

The problem is not that banks lack environmental data. Large banks have been collecting data on their corporate clients for years, driven by disclosure requirements, by the recommendations of the Task Force on Climate-related Financial Disclosures, and by their own risk management. The problem is that the data are not organized around the lending decision. A bank may know a client's Scope 1 and Scope 2 emissions, its energy consumption, and its participation in an emissions trading scheme. It may not know how those numbers relate to the client's ability to repay a five-year loan. The gap between disclosure and decision is where identification fails.

1.2.2 The information gap, in three layers

The information gap has three layers, and they require different responses.

The first layer is measurement. Many firms, particularly small and medium-sized enterprises, do not measure their environmental performance at all. A firm with fewer than 250 employees may not have an energy manager, may not track its emissions, and may not know which of its processes are the most carbon-intensive. The European Commission's 2018 impact assessment for the non-financial reporting directive found that reporting burdens fell disproportionately on smaller firms, and that many of them responded by producing boilerplate disclosures that satisfied the letter of the requirement without conveying useful information. Measurement is a precondition for identification, and for a large share of the corporate sector, it is missing.

The second layer is verification. Where firms do report, the reports are not always reliable. A 2021 analysis by the financial data provider Refinitiv found that the quality of corporate emissions disclosures varied widely, with some firms reporting at the level of individual facilities and others reporting a single aggregate figure with no supporting detail. Third-party assurance of sustainability reports is less common than financial audit, and the standards for assurance are less developed. A lender who relies on self-reported data is relying on the firm's own assessment of its performance, which creates an obvious incentive to present favorably. The history of environmental disclosure includes several well-documented cases of firms overstating their progress, from the Volkswagen emissions scandal to the more mundane practice of greenwashing in marketing.

The third layer is translation. Even where data are measured and verified, they need to be translated into the terms financial decisions use. A firm's emissions intensity is a physical quantity. A loan officer needs to know what that quantity implies for default probability, collateral value, or regulatory exposure. The translation requires a model, and the model requires assumptions about how environmental performance affects financial outcomes. Those assumptions are not obvious. A firm with high emissions may face a carbon price that raises its costs, or it may operate in a jurisdiction without a carbon price and face no immediate cost at all. The same emissions figure implies different financial risks in different contexts.

The three layers interact. Better measurement makes verification easier, and better verification makes translation more reliable. But the layers also have different owners. Measurement is the firm's responsibility. Verification is the responsibility of auditors and standards bodies. Translation is the responsibility of the financial institution. A solution that addresses only one layer will fail, because the other two will remain bottlenecks. This is why the identification problem cannot be solved by disclosure regulation alone, and it is why the financial system has to build its own translation capacity.

1.2.3 Why existing tools fall short

The financial system already has tools for evaluating uncertain information. Credit ratings, credit scoring models, and risk-adjusted pricing all exist to handle the fact that lenders do not know everything about their borrowers. Why can't these tools be extended to environmental performance?

The first answer is that existing tools are built around historical financial data. A credit rating agency evaluates a firm's past revenues, profits, and debt service. A credit scoring model uses past payment behavior to predict future payment behavior. Environmental performance is different in kind: it is forward-looking, it depends on technology and policy that may change, and it is not reflected in past financial statements until the costs or benefits arrive. A firm that has invested heavily in decarbonization may show lower profits today and higher resilience tomorrow. A model trained on historical financials will see the lower profits and miss the resilience.

The second answer is that existing tools are built around a single bottom line. Credit analysis asks one question: will the borrower repay? Environmental performance is relevant to that question, but it is not the same question. A framework that tries to answer both at once will produce a score that is neither a good credit score nor a good environmental score. The credit rating agencies learned this lesson when they began issuing ESG scores. Their corporate ESG scores have been criticized for conflating environmental performance with governance quality and social factors, and for producing rankings that do not predict either credit outcomes or environmental outcomes particularly well. The problem is not that the agencies are incompetent. The problem is that they are trying to compress a multidimensional concept into a single number, and the compression loses the information that makes the concept useful.

The third answer is that existing tools are built for a world in which the relevant risks are stable. A credit model calibrated on twenty years of data assumes the future will resemble the past. The green transition is precisely a period in which the future will not resemble the past. Carbon prices, technology costs, and regulation are all changing, and the changes are not random. They are directional. A model that treats them as stationary will systematically misprice the risks. This is the same problem banks faced in the 1990s when they tried to model emerging market risk using developed market data. The solution was not to abandon models but to rebuild them around the specific features of the new environment.

1.2.4 The identification problem is not only about firms

There is a temptation to frame the identification problem as a firm-level issue: given a firm, determine whether it is green. That framing is too narrow. The financial system also has to identify green projects, green assets, and green portfolios. Each level has its own information problems.

At the project level, the question is whether a specific investment will deliver the environmental benefits claimed for it. A solar farm's output can be metered. A reforestation project's carbon sequestration is much harder to measure, because it depends on survival rates, soil carbon, and the counterfactual of what would have happened without the project. The integrity of carbon offsets has been a persistent problem, and the price of offsets reflects the uncertainty: high-quality removal credits trade at a large premium to avoidance credits of uncertain additionality.

At the asset level, the question is whether a physical asset is compatible with a low-carbon future. A building's energy performance can be rated, and the EU's Energy Performance of Buildings Directive does exactly that. A ship's emissions can be estimated from its fuel consumption and engine specifications. The asset-level data are often better than the firm-level data, because assets are physical and can be inspected. But asset-level data do not tell you whether the firm that owns the asset is creditworthy.

At the portfolio level, the question is whether a lender's or investor's overall exposure is consistent with a transition pathway. This is the hardest level, because it requires aggregating firm-level and asset-level data across a portfolio and projecting how the portfolio will evolve. The Partnership for Carbon Accounting Financials, founded in 2019, has developed standards for measuring the emissions associated with loans and investments. The standards are a significant step, but they measure current emissions, not transition trajectories. A portfolio can have low current emissions and high transition risk if it is concentrated in firms that will struggle to adapt.

The four levels—firm, project, asset, portfolio—require different data and different methods. A solution that works at one level will not automatically work at another. This is a reason for modesty about any single indicator. It is also a reason to build a framework that is explicit about which level it is addressing.

1.2.5 A working definition

For the purposes of this book, I will define the identification problem as follows: the difficulty of producing, at reasonable cost and with acceptable reliability, an assessment of a firm's environmental performance that can be integrated into financial decision-making without either distorting the financial assessment or reducing the environmental assessment to a label.

Three features of this definition deserve emphasis.

First, it is about assessment, not certification. Certification is a binary judgment: a bond is green or it is not. Assessment is a graded judgment: a firm is at this point on a transition pathway, with this degree of confidence. Financial decisions need graded judgments, because they involve trade-offs. A loan officer who can only distinguish green from non-green cannot price the difference between a firm that is 80 percent of the way to a low-carbon process and a firm that is 20 percent of the way. A graded assessment lets her price that difference.

Second, it is about integration, not substitution. The assessment has to sit alongside the financial assessment, not replace it. A firm with excellent environmental performance and poor creditworthiness is still a poor credit risk. The integration means that the environmental assessment informs the financial one, through specific channels, rather than standing apart from it.

Third, it is about cost and reliability. An assessment that is perfectly accurate but costs more than the loan is worth is not useful. An assessment that is cheap but unreliable is worse than useless, because it creates false confidence. The framework has to be designed for the economics of the lending decision, which means it has to be scalable and standardized enough to be applied to many firms at reasonable cost.

1.3 Why the Market Has Not Solved This on Its Own

1.3.1 The public good problem

Environmental information has the characteristics of a public good. If one lender invests in measuring a firm's emissions, other lenders can benefit from the information without paying for it, provided the firm discloses it. The firm has an incentive to disclose favorable information and withhold unfavorable information. The result is under-provision: less information is produced than would be socially optimal, and the information that is produced is skewed toward the favorable.

Markets have developed partial solutions. Credit rating agencies sell their assessments to investors, who pay for the information because it helps them price risk. Sustainability rating agencies, such as MSCI and Sustainalytics, do the same for environmental and social performance. But the sustainability ratings market is fragmented, the methodologies differ, and the correlation between raters is low. A 2019 study by the MIT Sloan School of Management found that the correlation between ESG ratings from different providers was around 0.5, compared with a correlation of over 0.9 for credit ratings. The low correlation means that a firm's ESG score depends substantially on which agency is doing the scoring. That is not a reliable basis for a lending decision.

The fragmentation has a structural cause. Credit ratings converge because the rating agencies are all trying to predict the same thing: default. ESG ratings diverge because the agencies are trying to measure different things, using different definitions and different data. One agency weights carbon emissions heavily; another weights governance; a third weights labor practices. The divergence is not a failure of the market. It is a reflection of the fact that "ESG" is not a single concept.

For the identification problem, the relevant concept is narrower: environmental performance as it affects the transition and the financial risks associated with it. A rating built around that concept would have a better chance of convergence than a general ESG rating, because it would be anchored to a specific decision. This is the logic behind the rating framework developed in Chapters 3 and 4.

1.3.2 The horizon problem

Financial markets are often criticized for short-termism, and the criticism has some force in the context of the green transition. A firm that invests in a low-carbon furnace today will see the benefits over fifteen or twenty years. A fund manager evaluated on quarterly returns has little incentive to hold the firm's bonds through the transition. The mismatch between the investment horizon and the evaluation horizon is a well-known feature of capital markets, and it is particularly acute for green investments, where the payback periods are long and the risks are back-loaded.

The horizon problem is not absolute. Infrastructure investors, pension funds, and insurance companies have long horizons and are natural holders of green assets. The green bond market grew partly because these investors wanted long-duration assets with an environmental label. But the horizon problem does affect the banking system, which is the primary source of external finance for mid-sized firms in most economies. A bank that lends for five years to a firm making a fifteen-year transition investment is taking on refinancing risk. If the transition is not complete when the loan matures, the bank may not be able to roll the loan over, and the firm may be forced to abandon the investment.

Insurance can help with this problem, which is one reason the buffering layer matters. A weather index insurance policy that pays out when rainfall falls below a threshold can protect a farmer's revenue in a bad year, making it easier for the farmer to service a loan. The insurance does not lengthen the loan, but it reduces the probability that a bad year turns into a default, which makes the bank more willing to lend. The same logic applies to other green risks: a guarantee or a risk-sharing arrangement can transfer part of the transition risk to a party better able to bear it.

1.3.3 The coordination problem

The transition requires coordination across firms, sectors, and regions. A green steel plant needs green hydrogen, which needs renewable electricity, which needs transmission capacity, which needs permits. Each of these is a separate investment with its own financier, and the investments are complementary: the steel plant is not viable without the hydrogen, and the hydrogen plant is not viable without the steel plant's demand. Coordinating complementary investments is a classic problem in development economics, and it is not one that individual financial institutions can solve on their own.

The coordination problem is one reason for the catalysis layer. A government-guided fund or a public-private partnership can make the first investment in a complementary chain, absorbing the coordination risk that private investors cannot. The China Development Bank's lending for renewable energy, the European Investment Bank's support for offshore wind, and the Green Climate Fund's co-financing arrangements are all attempts to solve the coordination problem by putting public capital at the point where it can unlock private capital.

The coordination problem also affects the identification problem. If a firm's transition depends on infrastructure that does not yet exist, its environmental performance cannot be assessed in isolation. A cement plant that could switch to green hydrogen if the hydrogen were available is in a different position from a plant that has no access to hydrogen at all. An identification framework that looks only at the firm will miss the infrastructure dependence. This is a reason to assess firms in their sectoral and regional context, not as isolated units.

1.4 Toward an Integrated Mechanism

1.4.1 The three functions

The argument of this book is that financial support for the green transition requires three distinct functions, performed in sequence, and that the failure to perform any one of them undermines the others.

Identification comes first. The financial system needs a way to assess a firm's environmental performance and to integrate that assessment into credit decisions. This is the rating function. It produces a judgment about where a firm stands on its transition pathway and how that position affects its financial risk.

Consider what happens when rating is absent. A bank in a mid-sized European economy wants to build a green loan book, so it designates a category of "sustainable" loans and instructs its relationship managers to fill it. The relationship managers, who are evaluated on volume and default rates, assign the label to the borrowers they already like best. The result is a green portfolio that looks identical to the rest of the book, with the same risk profile and the same emissions trajectory. The bank has spent money on a labeling exercise and bought nothing but a reporting line.

The second function is risk absorption. Green investments carry risks that individual firms and lenders cannot easily bear: weather risk, technology risk, policy risk. Insurance instruments can transfer some of these risks to parties with diversified portfolios. This is the buffering function. It makes green lending safer and therefore more abundant.

The third function is capital catalysis. Early-stage green innovation is underfunded because the risks are high and the information is scarce. Public and quasi-public intermediaries can provide the first loss capital, the technical assistance, and the patient funding that private markets will not supply on their own. This is the catalysis function. It moves capital toward the frontier of green technology.

The three functions are not independent. Rating without buffering produces accurate assessments of risky firms, which lenders then decline to finance. Buffering without rating produces insurance against risks that no one has measured, which is a recipe for mispricing. Catalysis without rating and buffering produces public money flowing into projects whose risks are not understood and whose performance is not tracked.

flowchart TD
    A[Green transition capital demand] --> B{Identification: can the firm's transition be assessed?}
    B -- No --> C[Capital stays with incumbents or exits]
    B -- Yes --> D[Rating: environmental performance integrated into credit assessment]
    D --> E{Are the risks absorbable?}
    E -- No --> F[Lender declines or prices prohibitively]
    E -- Yes --> G[Buffering: insurance and guarantees transfer weather, technology, policy risk]
    G --> H[Credit flows to transitioning firms]
    H --> I{Is early-stage innovation funded?}
    I -- No --> J[Transition slows at the technology frontier]
    I -- Yes --> K[Catalysis: public and quasi-public capital absorbs first-loss risk]
    K --> H
    C --> L[Transition delayed]
    F --> L
    J --> L

The diagram is a simplification, but it captures the logic. Identification is the gate. Without it, capital either stays with incumbents or leaves the sector. With it, the question becomes whether the risks can be absorbed. If they cannot, the lender declines or prices the loan so high that the borrower cannot afford it. If they can, credit flows. The catalysis function operates at the frontier, where the risks are highest and the private market is thinnest.

1.4.2 Rating: from credit scoring to environmental performance

The rating function builds on a long tradition of credit scoring. The logic of credit scoring is to identify the observable characteristics that predict repayment and to combine them into a score. The same logic can be applied to environmental performance, provided the characteristics are chosen carefully and the combination is transparent.

The framework developed in Chapters 3 and 4 has three components. The first is a set of environmental indicators that are relevant to the firm's sector and that can be measured at reasonable cost. For a manufacturer, these might include emissions intensity, energy source mix, and waste treatment. For an agricultural firm, they might include water use, soil management, and input intensity. The indicators are not universal; they are selected to match the sector's transition pathway.

The second component is a weighting scheme that reflects the indicators' relationship to financial risk. This is where the rating differs from a general ESG score. The weights are not chosen to reflect the importance of the indicator to society. They are chosen to reflect the indicator's predictive power for the financial outcomes that matter to the lender: default probability, recovery rate, and cost volatility. An indicator that is environmentally important but financially irrelevant receives a low weight, not because it is unimportant but because the rating is a financial instrument.

The third component is an aggregation rule that produces a score and a confidence band. The confidence band is essential. Environmental data are noisy, and a rating that presents a single number without an indication of its uncertainty invites overconfidence. The confidence band tells the lender how much to trust the score, and it allows the lender to adjust the loan terms accordingly.

The rating framework is developed in detail in Chapter 3, and its application to technology firms is the subject of Chapter 4. The key point for this chapter is that rating is not a labeling exercise. It is an assessment of a firm's position on a transition pathway, integrated into the financial decision.

1.4.3 Buffering: insurance as a risk layer

Many green investments fail not because the firm is badly run but because the risks are correlated and large. A drought hits every farm in a region at once. A technology that works in the laboratory fails at commercial scale. A carbon price that was expected to rise stalls or reverses.

Insurance is the financial instrument designed for this problem. The traditional insurance model indemnifies the policyholder for a verified loss. The policyholder files a claim, the insurer investigates, and the insurer pays. For small, frequent losses, this works well. For large, correlated losses, it works less well, because the investigation is slow and the insurer's own capacity may be strained.

Weather index insurance takes a different approach. Instead of indemnifying the loss, it pays out when an index crosses a threshold. A rainfall index insurance policy might pay if rainfall at a specified weather station falls below 80 percent of the long-run average during a defined window. The payout is triggered by the index, not by the farmer's actual loss. This reduces the verification cost and speeds up payment, which matters when farmers need cash to replant.

The index approach has a known weakness: basis risk. The index may trigger when the farmer has not suffered a loss, or it may fail to trigger when the farmer has. The basis risk is a function of how well the index represents the farmer's exposure. A weather station fifty kilometers away may record rain that did not fall on the farmer's field. The design of the index, and the density of the weather station network, determine how well the insurance works.

Weather index insurance is the subject of Chapter 7. Forest insurance, which faces a different set of risks and measurement problems, is the subject of Chapter 8. The link between insurance and credit, which is the mechanism through which buffering makes lending safer, is the subject of Chapter 9. For this chapter, the point is that buffering is not a substitute for rating. It is a complement. Insurance prices risk, and to price risk you need to measure it. The rating framework provides the measurement that the insurance market needs.

1.4.4 Catalysis: sci-tech finance and the innovation frontier

The catalysis function addresses the earliest stage of the transition, where the technology is unproven and the private market is thin. A venture capital fund will invest in a clean technology startup if the expected return compensates for the risk. But the expected return is hard to estimate when the technology has not been deployed at scale, and the risk is hard to diversify when the fund is small. The result is underinvestment in the innovation that the transition ultimately depends on.

Sci-tech finance is the set of institutions and instruments that fill this gap. The term is used in China to describe the financial system that supports technology-based firms, including venture capital, government-guided funds, science and technology banks, and intellectual property financing. The Chinese experience is instructive because the government has made a deliberate effort to build these institutions, and the results have been mixed: rapid growth in some sectors, persistent difficulties in others.

The catalysis function is the subject of Chapters 10 through 12. Chapter 10 sets out the logic of sci-tech finance and its role in green innovation. Chapter 11 examines government-guided funds and the market failure in early-stage green capital. Chapter 12 turns to performance evaluation and incentive design, which is where many public programs fail: the incentives are misaligned, the evaluation is superficial, and the capital is not recycled.

The connection between catalysis and the other two functions is the subject of Chapter 13, which sets out the integrated mechanism in sequence. The sequence matters. Catalysis without rating produces public money flowing into projects whose performance is not measured. Catalysis without buffering produces public money absorbing risks that could have been transferred to the insurance market. The three functions have to be designed together.

1.4.5 What the integrated mechanism looks like

The integrated mechanism can be described as a pipeline. A firm enters the pipeline at the rating stage. The rating produces an assessment of its environmental performance and its credit risk. If the assessment is favorable, the firm proceeds to the financing stage, where the lender considers the loan. If the risks are too high for the lender's appetite, the firm proceeds to the buffering stage, where insurance or a guarantee can transfer part of the risk. If the firm is at the innovation frontier and the private market will not finance it, it proceeds to the catalysis stage, where public or quasi-public capital provides the first loss.

The pipeline is not a rigid sequence. A firm can enter at any stage, and the stages can operate in parallel. A large firm with a strong balance sheet may need only the rating. A small firm with a promising technology may need all three. The point is that the three functions are available and that they are connected, so that a firm that fails at one stage can access the next.

The pipeline also has a feedback loop. The performance data collected at the rating stage inform the insurance pricing at the buffering stage, and the loss experience at the buffering stage informs the rating methodology. The catalysis stage generates the innovation that the rating stage eventually assesses. When the feedback loop is missing, the three instruments drift apart: insurers price weather risk without reference to the borrower's transition trajectory, lenders re-rate firms without access to the loss data the insurers have accumulated, and public funds underwrite the same risks the insurance market was built to absorb. The cost of that drift is paid twice, once in mispriced capital and once in slower deployment.

1.5 The Argument in Brief

The chapters that follow develop this argument in detail. Part I, comprising Chapters 2 through 5, develops the rating function. Chapter 2 examines the information asymmetry that underlies all credit, and the architecture that financial institutions have built to manage it. Chapter 3 turns to the specific problem of environmental performance rating, setting out the methodology and its relationship to credit scoring. Chapter 4 applies the methodology to technology firms, where the tension between cash flow, collateral, and carbon is sharpest. Chapter 5 examines the portfolio consequences: when many lenders use the same rating framework, they may concentrate their exposure in the same firms, creating a new source of systemic risk.

Part II, comprising Chapters 6 through 9, develops the buffering function. Chapter 6 sets out the general logic of insurance as a risk-buffering layer. Chapter 7 examines weather index insurance and its role in eco-agricultural resilience. Chapter 8 examines forest insurance and the forestry transition. Chapter 9 links insurance to credit, showing how risk transfer affects default probability and how the link can be priced.

Part III, comprising Chapters 10 through 12, develops the catalysis function. Chapter 10 sets out the logic of sci-tech finance and its application to green innovation. Chapter 11 examines government-guided funds and the market failure in early-stage green capital. Chapter 12 turns to performance evaluation and incentive design.

Part IV, comprising Chapters 13 through 15, integrates the three functions. Chapter 13 sets out the integrated mechanism in sequence. Chapter 14 examines the digital infrastructure that the mechanism requires. Chapter 15 draws out the policy implications and sets the research agenda.

The argument is not that the integrated mechanism will solve the identification problem completely. It will not. Environmental performance is too complex, and the data are too incomplete, for any framework to produce perfect assessments. The argument is that the integrated mechanism will produce assessments that are good enough to support financial decisions, and that the alternative—continuing to fund the appearance of transition rather than its substance—is worse.

1.6 A Note on Method and Scope

This book draws on several strands of research, and it is worth being explicit about what they are and what they are not.

The first strand is the author's prior work on credit rating for technology enterprises. That work developed a methodology for assessing the creditworthiness of firms whose main assets are intangible and whose cash flows are volatile. The methodology combines financial indicators with indicators of technological capability and market position. The environmental performance rating developed in this book extends that methodology by adding environmental indicators to the assessment.

The second strand is the author's work on weather index insurance. That work examined the design of index insurance products for agricultural risks and the conditions under which they improve farmer welfare. The environmental performance rating and the buffering layer both draw on that work: the rating framework incorporatesweather-related risk into the assessment of agricultural firms, and the buffering layer uses index insurance as the primary instrument for transferring that risk.

The third strand is the author's work on the impact of agricultural insurance on farmers' income. That work examined how insurance affects farmers' investment decisions and their resilience to shocks. Its relevance to this book is the link between insurance and credit: a farmer who is insured against weather risk is a better credit risk, because the insurance reduces the probability that a bad year turns into a default. The same logic applies to firms in other sectors, which is why the buffering layer is not limited to agriculture.

The book also draws on a broader literature: the credit risk literature that runs from the Merton model through the modern structural and reduced-form approaches; the development economics literature on index insurance and its limitations; the corporate governance literature on sustainable value creation; and the policy literature on green finance, including the work of the Network for Greening the Financial System and the various national green finance taxonomies.

The scope of the book is deliberately narrow in one respect and broad in another. It is narrow in that it focuses on the financial system's role in identifying and supporting the transition of existing industries, not on the transition of the energy system as a whole or on the decarbonization of households. It is broad in that it covers banking, insurance, and capital markets, because the identification problem cuts across all three. A book that examined only green bonds would miss the credit channel. A book that examined only bank lending would miss the insurance channel. The integrated mechanism requires all three.

The book is also explicit about its limits. It does not claim to have solved the measurement problem. It does not claim that the integrated mechanism will work in every institutional context. It does not claim that the private sector can finance the transition without public support. What it does claim is that the three functions—rating, buffering, and catalysis—are necessary components of any serious attempt to finance the transition, and that the failure to integrate them is a major reason why green finance has so far fallen short of its promise.

1.7 The Stakes

What happens if the identification problem is not solved? The answer is not abstract. It shows up in loan books, in bond spreads, and in the pace at which furnaces and boilers get replaced.

Start with capital allocation. If the financial system cannot distinguish genuine transition from green marketing, capital will flow to the firms that are best at presenting themselves as green, not to the firms that are actually reducing emissions. The result is a misallocation that is costly in two ways: the transition is slower than it could be, and the firms that are genuinely transitioning are penalized because they compete with firms that have lower costs and no real environmental investment. A textile manufacturer that spends on wastewater recycling and renewable power carries a cost its competitor avoids, and if both receive the same green label, the label rewards the cheaper firm. The mechanism is not hypothetical. It is the ordinary logic of any market in which quality is unobservable and the label is cheap to obtain.

The second stake is financial stability. If lenders classify loans as green without a way to verify the environmental performance of the borrower, they will accumulate exposure to firms that are not transitioning. When the transition accelerates—through a carbon price, a technology shift, or a regulatory change—those firms will face higher costs and lower revenues, and the loans will deteriorate. The losses will not be confined to the green-labeled portion of the portfolio, because the firms that are not transitioning are often the same firms that are exposed to transition risk across all their borrowing. The European Central Bank's 2021 economy-wide climate stress test found that under a disorderly transition scenario, cumulative credit losses for euro-area banks over thirty years could reach roughly 8 percent of total loans, with the losses concentrated in the carbon-intensive sectors that had been slowest to adjust. The identification problem is a precondition for managing that risk: a bank cannot hedge an exposure it has not measured.

The third stake is credibility. Green finance has attracted a great deal of attention and a great deal of skepticism. The skepticism is not unfounded. The history of green-labeled financial products includes several cases where the label was not backed by the substance. In 2021 the German asset manager DWS became the subject of a probe by the US Securities and Exchange Commission and the German regulator BaFin over allegations that its ESG funds held investments that did not meet the stated criteria; the company's share price fell sharply and its chief executive later announced his departure. Cases like this do damage beyond the firm involved. If the financial system cannot demonstrate that green finance is doing what it claims, the political support for green finance policies will erode, and the capital that has been mobilized will be redirected. The identification problem is, in this sense, a legitimacy problem.

The fourth stake is the transition itself. The firms that need to transition most are often the firms that are hardest to finance: heavy industry, agriculture, forestry, and small manufacturers. These are the firms where the identification problem is most acute, because their environmental performance is hard to measure and their financial position is often weak. If the financial system cannot reach them, the transition will be confined to the sectors that are already easy to finance—renewable energy, green buildings, and clean technology—and the harder sectors will lag. The result will be a transition that is real but incomplete. A power system that runs on renewables while the cement and steel that build it are made with coal is not a decarbonized economy. It is a decarbonized electricity sector with an emissions problem attached.

There is a fifth stake that is less often named: the cost of the delay itself. Every year a steel plant postpones relining its furnace for a lower-carbon process, the plant's remaining service life shortens and the case for retrofitting weakens. A furnace built in 2025 will still be operating in 2045. The investment decisions made in the next few years lock in emissions for decades, and the financial system's inability to identify which plants are worth retrofitting is one reason those decisions get deferred. The identification problem is not a bookkeeping inconvenience. It has a physical counterpart in equipment that gets built or not built, and the equipment lasts a long time.

1.8 The Road Ahead

The next chapter begins with a problem older than green finance: the lender never knows as much about the borrower as the borrower knows about itself. The financial system has spent two centuries building an architecture to manage that asymmetry—screening, monitoring, collateral, covenants, diversification. The architecture works well for the risks it was designed for. It works less well for environmental risk, because environmental risk has features the architecture does not handle: it is correlated across borrowers, it is forward-looking, and it is not reflected in the historical data the architecture relies on.

Consider how the architecture handles a conventional risk. A bank lending to a chain of restaurants can look at the chain's past revenue, its lease obligations, and the failure rate of comparable restaurants in comparable locations. The data are abundant, the comparison group is large, and the historical record is a reasonable guide to the future. Now consider the same bank lending to a cement plant that proposes to install a carbon capture system. There is no large comparison group of cement plants with carbon capture systems, because very few exist. The historical record says nothing about whether the technology will work at this plant's scale. The collateral—the plant itself—may be worth less in a carbon-constrained world, not more. Every input the conventional architecture relies on is either missing or misleading.

That gap is where the book's argument starts. Chapter 2 sets out the architecture and shows, in detail, which of its components fail on environmental risk and why. Chapters 3 through 5 build the rating function that fills the gap. Chapters 6 through 9 build the buffering layer. Chapters 10 through 12 build the catalysis layer. Chapter 13 puts the three together and specifies the sequence in which they have to operate. The rest of the book is about the design choices inside each layer and the evidence, from China and elsewhere, about which choices work.

One last point about the method. The book is written from the perspective of a researcher who has spent years inside the institutions being described—building credit models for technology firms, designing index insurance products, evaluating the income effects of agricultural insurance. That vantage point shapes what the book emphasizes. It is more interested in how a rating is actually computed than in how a taxonomy is drafted. It is more interested in the basis risk of a rainfall index than in the headline size of the green bond market. The premise is that the identification problem is solved or not solved at the level of the individual credit decision, and that is where the analysis has to start.