Chapter 1: Preface: Why Single-Channel Green Credit Fails

In the first quarter of 2021, a provincial commercial bank in eastern China approved a ¥400 million green loan to a mid-sized specialty chemical producer. The borrower had a clean environmental record, a valid pollution discharge permit, and a project that appeared on the local green project catalogue: a solvent recovery unit that would cut volatile organic compound emissions by roughly 60 percent. The loan was classified as green under the bank's internal taxonomy. Eighteen months later the borrower missed an interest payment. The recovery unit had been installed, but the downstream market for the recovered solvent collapsed, the firm's working capital cycle stretched, and a second, older production line—never part of the green project—became the source of a regulatory penalty that froze the firm's credit line at two other banks. The green loan itself was technically performing. The borrower was not.

Every green lending officer can produce a version of this case. The borrower passed the environmental screen. The project passed the technical screen. What failed was everything the screen did not look at: the borrower's full exposure to environmental liabilities across its asset base, the market risk in the output the green project produced, and the absence of any buffer between a single adverse event and default. The loan was green. The credit was not safe.

This book is about that gap. Its argument, stated plainly at the outset, is that the financial system's support for industrial green transition cannot run through a single channel—green credit—without failing at scale. Identification of which firms genuinely merit capital, buffering of the risks those firms carry, and catalysis of the technologies that do not yet have a bankable cash flow are three distinct functions. They require three distinct instruments and a mechanism that links them. I call this the identification-buffering-catalysis chain.

1.1 The Green Transition Imperative

1.1.1 What transition actually demands

Start with the physical scale of the problem, because the financial scale follows from it. The International Energy Agency's 2021 net-zero scenario put cumulative investment needs in clean energy and related infrastructure at roughly $4 trillion per year by 2030, against approximately $1.5 trillion then being deployed. The gap is not a rounding error. It is larger than the annual GDP of most countries.

Those aggregate numbers hide the harder structural fact: the transition is not a single investment project. It is a re-engineering of existing industrial capital stock. A cement plant does not become low-carbon by building a new wing. Its kiln, its fuel supply, its logistics, and its product mix all change, and the change is sequenced over a decade or more. A steel mill's transition depends on whether hydrogen arrives at the right price, whether scrap steel is available in the right grade, and whether its customers will pay a green premium for the output. These are not project risks in the conventional lending sense. They are transition-path risks, and they are poorly suited to a loan structured around a single asset.

Consider what a bank actually evaluates when it makes a green loan. The typical green credit framework, whether the People's Bank of China's green bond catalogue, the EU Taxonomy, or a bank's internal green list, asks one question: does this project meet an environmental standard? The answer is binary. A project either qualifies or it does not. This binary structure is useful for labelling and for reporting. It says nothing about whether the borrower will repay.

The two questions are not the same. A project can meet the highest environmental standard and still be a bad credit. A firm can be a good credit and have no green project at all. The green credit channel conflates environmental eligibility with creditworthiness, and in doing so it misprices both.

1.1.2 The financing gap is not uniform

It helps to disaggregate the gap by the maturity of the technology and the size of the firm, because different segments fail for different reasons.

At one end sit mature, commercially proven technologies—onshore wind, utility-scale solar, efficient lighting, waste heat recovery. These are bankable. Their costs are known, their output is contracted or merchant-market-traded, and their risks can be modelled with tolerable error. Green credit works here, and it has worked: global renewable capacity additions have grown at double-digit rates for a decade, financed overwhelmingly by debt.

In the middle sit technologies that are technically proven but not yet cost-competitive at scale—green hydrogen electrolysers, carbon capture on cement, advanced biofuels. Their problem is not creditworthiness in the conventional sense. Their problem is that the revenue line depends on a carbon price, a subsidy, or a customer's willingness to pay a premium, all of which are uncertain over the life of a loan. A bank lending to a green hydrogen project is underwriting a policy bet. Most banks will not do it at scale, and the ones that do demand credit enhancement.

At the far end sit technologies still in development—next-generation storage chemistries, direct air capture, novel low-carbon process routes. These have no bankable cash flow. They are the domain of equity, venture capital, and government grant. No amount of green credit reform will make them lendable, because the problem is not the label. It is the absence of a revenue stream to service debt.

The financing gap, then, is not one gap. It is a set of gaps along a maturity spectrum, and each requires a different financial instrument. A framework that treats all of them as "green finance" and routes them all through credit will fail at the two ends of the spectrum: it will over-lend to marginal projects in the mature segment, where the green label substitutes for credit analysis, and it will under-lend to everything in the middle and far segments, where no amount of labelling can overcome the missing revenue line.

1.1.3 The Chinese context and why it matters beyond China

China is the largest laboratory for green credit in the world, and the results are instructive. By the end of 2023, China's green loan balance exceeded ¥30 trillion, the largest such stock in any economy. The growth was driven by a policy architecture: green bond catalogues issued by the People's Bank of China, green credit statistics and evaluation rules from the banking regulator, and a system of relending facilities that gives banks cheaper funding for qualifying green loans.

The architecture worked in the sense that it moved money. It moved money toward projects that could be labelled green. Whether it moved money toward the transition—toward the firms and technologies that most need capital to change—is a different question. A large share of Chinese green loans has gone to mature segments: clean energy generation, green transport infrastructure, energy efficiency in large state-owned enterprises. These are exactly the segments that are most bankable and least in need of a policy subsidy to attract capital. The segments that are hardest to finance—transitioning heavy industry, small and medium-sized green technology firms, agricultural and forestry projects with long payback periods—remain under-served.

The same pattern appears in the EU's sustainable finance framework. The Taxonomy has grown more detailed year by year, and the volume of labelled issuance has risen, yet the financing of hard-to-abate industrial transition still depends on public support. The framework labels. It does not intermediate.

The reason is structural. It is the subject of the next section.

1.1.4 The financial gap in numbers

Three figures frame the problem. The investment gap: roughly $2.5 trillion per year of additional clean investment is needed by 2030. The composition gap: private finance accounts for the majority of clean investment in mature technologies but a minority in early-stage and hard-to-abate segments, where public and concessional capital dominate. The instrument gap: debt dominates green finance, while equity and insurance are marginal. In most green finance statistics, bank loans and bonds account for 80 to 90 percent of the total. Insurance and risk-transfer products barely appear.

That third figure is the one this book takes most seriously. A financial system that can only lend has only one way to handle risk: it can decline to lend. A financial system that can lend, insure, and take equity has three ways. The transition requires all three, because the risks along the maturity spectrum are different in kind, not just in degree.

1.2 The Single-Channel Fallacy

1.2.1 The logic of the single channel

Banks are the dominant source of external finance for firms in most economies, particularly in emerging markets and particularly for small and medium enterprises. If you want to change corporate behavior at scale, you work through banks. Green credit does this by changing the price and availability of bank funding for environmentally eligible projects. A subsidized green loan rate lowers the cost of capital for a qualifying project. A green credit quota raises the supply of funding. The mechanism is direct.

There is a second, subtler argument. Green credit changes the information environment. A firm that wants a green loan must document its environmental performance, and that documentation enters the bank's credit file. Over time, the argument goes, environmental performance becomes part of the credit assessment, and firms internalize the cost of poor environmental performance. The label becomes a governance device.

Both arguments have merit. Neither justifies relying on green credit alone, and the reasons are not about implementation. They are about the structure of credit as a financial instrument.

1.2.2 Credit is a one-sided bet

A loan is a contract in which the lender bears downside risk and forgoes upside. The borrower keeps any return above the interest rate. This asymmetry is the defining feature of debt, and it shapes what a lender can finance. A lender will finance a project whose downside is bounded and whose cash flow is predictable. A lender will not finance a project whose upside is large but whose downside is a total loss, because the lender does not share the upside.

Green transition projects in the middle of the maturity spectrum have exactly the payoff profile that debt cannot finance: large upside if the technology scales and the carbon price holds, total loss if it does not. Equity can absorb that profile. Debt cannot. A green credit program that targets these projects will either fail to lend or will lend with credit enhancement that shifts the risk to someone else—a government guarantee, a multilateral development bank, a parent company. The enhancement is doing the work, not the green label.

This is not a criticism of banks. It is a description of the instrument. The single-channel fallacy begins when the instrument's limits are treated as a policy failure to be corrected by more of the same instrument.

1.2.3 Identification failure: the label substitutes for analysis

Picture the loan committee on a Tuesday afternoon. The credit officer has two files in front of her. The first is a specialty chemical firm seeking ¥200 million for a solvent recovery unit; the project sits on the provincial green catalogue, which means the loan will count toward the bank's green credit target and qualify for a relending line at 1.75 percent. The second is a packaging manufacturer with fifteen years of stable cash flow and no green project at all. The green file is thinner on financials, and the borrower's compliance history shows a warning letter from two years ago. The committee approves the green loan and defers the packaging firm. The green target gets met. The credit analysis gets skipped.

When a green label is attached to a loan, it changes the incentives inside the lending institution. A green loan may count toward a regulatory target, qualify for a subsidized funding line, or improve the bank's own sustainability reporting. These are legitimate policy goals, but they create a path of least resistance: the loan officer's job becomes to find projects that qualify for the label, not to find borrowers who will repay.

The result is a selection effect. A project that is both green and highly profitable does not need a green loan at a subsidized rate; it can borrow on commercial terms. The subsidy therefore selects for the marginal project, and the marginal project is the one most likely to default. This is structural. It does not go away with better due diligence at the individual loan level. It goes away only when the identification mechanism distinguishes environmental performance from creditworthiness and prices them separately. A rating that incorporates environmental performance into a credit assessment—the subject of Chapters 5 and 6—does this. A label that certifies environmental eligibility does not.

The distinction matters for another reason. The green label is binary. A firm either qualifies or does not. But environmental performance is continuous: a firm can reduce emissions by 10 percent or 60 percent, can have a clean current record and a contaminated legacy site, can be improving or deteriorating. A binary label cannot represent any of this. It flattens a multidimensional attribute into a yes-or-no, and in doing so it throws away exactly the information a credit assessment needs.

1.2.4 Risk buffering failure: the missing instrument

Ask a Chinese agricultural insurer what share of the country's smallholders carry weather index cover, and the answer comes back in single digits. Ask how many forestry projects with a twenty-year rotation have fire and drought insurance attached to their financing, and the number is close to zero. These are not obscure corners of the market. They are the segments where the transition's physical risk is most concentrated.

Consider the risks that a green transition project carries and ask which financial instrument absorbs each.

Technology risk—the risk that the chosen technology underperforms—is borne by equity. Construction risk is borne by the contractor and, during the construction period, by the lender. Market risk—the risk that the output price falls—is borne by the offtaker if there is a long-term contract, and by the project if there is not. Policy risk—the risk that a subsidy or carbon price changes—is borne by whoever holds the residual claim. Weather risk—the risk that a drought, a flood, or a temperature swing destroys output—is borne by the project unless it is transferred.

For a renewable energy project with a long-term power purchase agreement, most of these risks are allocated by contract, and the residual is small enough for debt to carry. For an eco-agriculture project, the weather risk is dominant and there is no offtaker to absorb it. For a forestry project, the risk of fire, pest, and drought is correlated across the entire portfolio, which means it cannot be diversified away by holding many projects. These are precisely the segments where green credit fails, and the reason is not that the projects are unworthy. The reason is that the risk they carry has no instrument to absorb it.

Insurance is that instrument, and it is largely absent from green finance frameworks. Weather index insurance pays out when a measurable weather variable—rainfall, temperature, a vegetation index—crosses a threshold, without requiring the insurer to verify individual losses. It is well suited to smallholder agriculture, where loss verification is expensive and moral hazard is high. Forest insurance covers fire, disease, and increasingly drought, and it is the precondition for any lender to finance a forestry project with a 20-year rotation. Neither product appears in the green bond catalogue. Neither is counted in green finance statistics. Both are essential.

The absence is not accidental. Insurance is a different business from lending, with different regulation, different capital requirements, and different distribution channels. A bank cannot simply add insurance to its green loan product. The integration has to be built, and building it is one of the tasks this book describes.

Chapters 8, 9, and 10 take up this task in detail. The point to register here is the structural one: a financial system that offers only credit to a risk-laden project is offering a single instrument against a bundle of risks. The instrument will be declined or mispriced. The project will not be financed, or it will be financed badly.

1.2.5 Catalysis failure: the early-stage gap

Here is a counter-intuitive fact about green finance programmes. Many of them list "green technology" as an eligible category and then route it through bank lending. The result is predictable: low uptake, or uptake by firms that are not actually early-stage but have learned to describe themselves as such. A loan to a pre-revenue company is not a loan; it is an equity investment with a fixed claim, which is the worst of both instruments.

Early-stage green technology firms have no bankable cash flow and no insurable risk in the conventional sense. They have an idea, a prototype, and a team. The financial instrument that matches this profile is equity, specifically venture capital, and the public instrument that supports it is the government-guided fund.

The catalysis function requires a different institutional architecture: venture capital funds with the patience to hold a position for seven to ten years, government guiding funds that provide first-loss capital or matching capital to crowd in private investors, and exit mechanisms that allow the equity to be realized. Chapters 11 and 12 describe this architecture. The point for the preface is that it is a separate link in the chain, not a variant of the credit link.

1.2.6 Why the three failures are linked

The three failures could be treated as three separate policy problems, each with its own fix. They are not separate. They are linked by a common structure: the transition requires capital to flow to firms and projects whose risk profile does not match the instrument being used. Green credit fails at identification because the label does not measure creditworthiness. It fails at buffering because credit cannot absorb the risks that insurance should carry. It fails at catalysis because debt cannot finance what equity should finance.

Each failure creates pressure on the others. When insurance is absent, lenders demand collateral, and firms without collateral are excluded—which is why small green technology firms and smallholder agricultural projects are under-financed. When catalysis is absent, early-stage firms cannot reach commercial scale, and the pipeline of bankable projects dries up. When identification is weak, the loans that are made go to the wrong borrowers, and the losses that result make banks more reluctant to lend to the next green project, which widens the financing gap.

The chain is not a metaphor. It is a description of how the three functions interact. Identification without buffering produces loans that default when a shock arrives. Buffering without identification insures the wrong risks. Catalysis without identification funds the wrong technologies, and catalysis without a buffering mechanism for the resulting firms leaves them exposed when they try to scale. The functions have to be designed together.

flowchart TD
    A[Green transition financing need] --> B{Single-channel green credit}
    B --> C[Identification failure: label substitutes for credit analysis]
    B --> D[Buffering failure: no instrument absorbs weather, price, policy risk]
    B --> E[Catalysis failure: debt cannot finance pre-revenue technology]
    C --> F[Marginal projects selected, defaults rise]
    D --> G[Risk-laden projects declined or mispriced]
    E --> H[Early-stage firms starve; bankable pipeline shrinks]
    F --> I[Bank appetite for green lending falls]
    G --> I
    H --> I
    I --> J[Financing gap widens]
    J --> B

The loop in the figure is the reason single-channel green credit does not merely underperform. It is self-limiting. Each failure reinforces the others, and the system settles into an equilibrium in which a large volume of green-labelled credit coexists with a persistent gap in transition finance.

1.2.7 A note on what "failure" means here

The claim is not that green credit has failed in the sense of producing no results. It has produced results. Renewable energy costs have fallen, capacity has grown, and a large volume of capital has moved toward environmentally beneficial projects. The claim is narrower and more precise: single-channel green credit fails at the margin where the transition is hardest—the transition of existing industries, the financing of early-stage technology, and the management of correlated environmental risk. These are the margins that determine whether the transition is completed or stalls.

There is a second sense in which the word "failure" is apt. A single channel is fragile. When the channel is blocked—by a credit crunch, by a regulatory change, by a rise in defaults—there is no alternative route for capital to reach the transition. A multi-channel system is resilient because the channels can substitute for one another at the margin. A firm that cannot get a green loan because its credit is weak can still get insurance that makes its project bankable, or equity that does not require debt service. The chain is a redundancy mechanism as much as it is a division of labor.

1.3 Thesis and Structure

1.3.1 The three-link chain

The thesis of this book can be stated in one sentence: financial support for green transition must move beyond single-channel credit lending toward an integrated mechanism of rating-based identification, insurance-based risk buffering, and sci-tech finance catalysis.

Each link has a specific function and a specific institutional form.

Rating-based identification means incorporating environmental performance into a credit rating or scoring model, so that the lender's assessment of default probability reflects the borrower's environmental risk exposure and environmental performance, not just its financial ratios. This is more than adding an ESG score to a credit file. It means building environmental variables into the model that predicts default, testing whether they carry information, and calibrating their weights. Chapters 5 and 6 develop this, drawing on prior work on credit rating for technology enterprises.

Insurance-based risk buffering means using insurance instruments—weather index insurance for eco-agriculture, forest insurance for the forestry industry, mutual and cooperative risk-sharing for rural green transition—to absorb the risks that credit cannot carry. The buffer does not eliminate risk. It transfers it to an institution that can hold it more efficiently, which is what makes the underlying project bankable. Chapters 8, 9, and 10 develop this.

Sci-tech finance catalysis means using equity, venture capital, and government guiding funds to finance the early-stage and scale-up phases that debt cannot reach, and designing the intermediation system so that capital reaches green technology firms. Chapters 11 and 12 develop this.

The three links are not a sequence in the sense that identification must precede buffering, which must precede catalysis. They operate in parallel, on different segments of the market. The chain is a portfolio of functions, not a pipeline. What makes it a chain is the linkage: identification determines which risks need buffering, buffering determines which projects become bankable, and catalysis determines which technologies enter the bankable pipeline. The links feed each other.

1.3.2 The diagnostic metric that connects the links

A chain needs a common unit of account, a metric that lets a lender, an insurer, and an equity investor compare notes on the same project. The book proposes the green premium as that metric. The green premium is the difference between the cost of a green alternative and the cost of its carbon-intensive substitute, expressed per unit of output or per unit of emissions avoided. It is a measure of how far a technology or a project is from being competitive without policy support.

The green premium is useful because it is comparable across the three links. A high green premium tells the lender that the project's cash flow depends on policy support, which is a credit risk. It tells the insurer that the project's viability is sensitive to the policy environment, which affects the correlation of losses. It tells the equity investor that the technology needs to move down the cost curve before it can be financed with debt, which sets the investment horizon. Chapter 4 develops the metric.

The metric also connects the book to the policy debate. A carbon price, a subsidy, or a green credit subsidy all work by changing the green premium. The question for policy is not whether to subsidize but where, along the maturity spectrum, the subsidy has the most effect per unit of public money. The green premium gives a way to answer that question.

1.3.3 Why the chain is a better framework than a taxonomy

The dominant approach to green finance policy is the taxonomy: a list of eligible activities, with technical screening criteria. The EU Taxonomy is the most elaborate example; China's green bond catalogue is another. Taxonomies are useful. They create a common language, they reduce greenwashing, and they give regulators a basis for reporting.

They are not a substitute for the chain. A taxonomy answers the question "is this activity green?" It does not answer "will this borrower repay?" or "who bears the weather risk?" or "is this technology ready for debt?" Those are the questions that determine whether capital actually flows. A taxonomy can label a project green and still leave it unfinanceable, because the label does not change the project's risk profile.

The chain framework treats the taxonomy as an input, not an answer. The taxonomy defines what counts as green. The rating model uses the taxonomy to identify environmental variables. The insurance product uses it to define the insured activity. The venture fund uses it to define the investment universe. The chain is the machinery that turns the taxonomy into capital flows.

1.3.4 The evidence base

The book draws on three strands of prior research. The first is credit rating for technology enterprises, which provides the modelling framework for incorporating non-financial variables into default prediction. The second is weather index insurance, which provides the design framework for buffering agricultural risk. The third is the impact of agricultural insurance on farmer income, which provides the empirical evidence that insurance changes behavior and not just risk transfer.

These strands are supplemented by the policy experience of China's green finance system, the EU's sustainable finance framework, and the international insurance and reinsurance markets. The book is not a country study. It uses the Chinese experience because it is the largest and most instructive, but the framework is general.

The evidence is imperfect, and the book says so where it matters. Green credit data are reported by lenders with an incentive to classify favorably. Insurance penetration in rural areas is measured with error. Venture returns in green technology are too recent to evaluate over a full cycle. The framework is offered as a way to organize what is known and to identify what is not.

1.3.5 A concrete starting point

Return to the chemical producer from the opening. What would the chain have done differently?

Identification: a rating model that incorporated the firm's environmental liabilities across its full asset base, not just the green project, would have flagged the older production line as a contingent liability. The penalty that froze the credit line at two other banks was foreseeable from the firm's compliance history. A rating that weighted compliance history would have priced the loan higher or required a covenant.

Buffering: the market risk in the recovered solvent—the risk that the output price would fall—was not insurable in the conventional sense, but a price hedge or a tolling arrangement with the downstream buyer could have transferred it. The point is that the risk had a potential home, and the loan structure did not use it.

Catalysis: the solvent recovery technology was mature, so catalysis was not the binding constraint. But the firm's next project, a catalytic conversion unit, was not. A venture or scale-up equity investment, structured alongside the loan, could have financed the pilot without loading the debt service onto the firm's balance sheet.

None of these fixes is exotic. Each is standard practice in some part of the financial system. What was missing was the integration: the lender did not have a rating model that looked at the whole firm, did not have an insurance or hedging product that could absorb the market risk, and did not have an equity channel for the next project. The single channel—green credit—was the only instrument in the room, and it was asked to do three jobs it could not do.

1.3.6 What the reader should take away

The single-channel approach fails for structural reasons, not implementation reasons. Credit is a one-sided instrument, and the transition's risk profile does not match it. More green credit, better targeted, will not close the gap at the margins where the transition is hardest.

The three functions—identification, buffering, catalysis—are distinct and require distinct instruments. A rating model is not an insurance product. An insurance product is not a venture fund. The policy mistake is to treat them as variants of a single green finance program.

The functions have to be linked. A rating without a buffer produces loans that default. A buffer without a rating insures the wrong risks. Catalysis without either funds technologies that cannot scale. The chain is the unit of design.

Chapter 2 begins with the problem that makes identification hard in the first place. It opens with a loan officer in Zhejiang rejecting a green application she cannot price, and works outward from her decision to the information asymmetry that produced it—why credit rationing is worse when the risk being assessed is environmental, and why the standard remedies of collateral, covenants, and monitoring lose their grip on a risk that has no historical loss data behind it.