Chapter 2: Information Asymmetry, Credit Rationing, and the Rationale for Integrated Finance
In the early 1980s, the United States Congress passed the Comprehensive Environmental Response, Compensation, and Liability Act—known in the trade as Superfund. The statute made lenders potentially liable for the environmental cleanup costs of their borrowers. A bank holding a mortgage on a contaminated site could, under certain conditions, find itself named as a responsible party. Credit departments at American banks responded within months. They began asking borrowers about underground storage tanks, asbestos, and past industrial use of the land. Credit officers who had spent their careers reading balance sheets were suddenly reading Phase I environmental site assessments.
That episode is usually told as the origin story of green finance. It is a useful story, often misread. The lesson bankers drew from Superfund was not that environmental performance mattered for its own sake. The lesson was that environmental information was credit information. A borrower with a hidden contamination problem was a borrower whose default probability had been systematically underestimated. The banks that adapted built environmental screening into their credit processes out of the same instinct that makes them check a borrower's tax records: they wanted to price risk accurately, and they had discovered that a relevant piece of information was missing.
Why is it so hard to act on that instinct? The green transition requires the financial system to allocate capital toward projects whose environmental benefits are uncertain, whose technologies are immature, and whose cash flows are back-loaded. These are exactly the conditions under which information problems bite hardest. The information asymmetries in green lending have a specific structure that makes conventional credit assessment systematically mispriced. That mispricing produces credit rationing, and the rationing falls disproportionately on green firms in ways that are both quantitatively large and analytically predictable. The three instruments this book develops—rating, insurance, and sci-tech finance catalysis—are not a menu of alternatives. Each addresses a different failure, and none works well alone.
The specific informational structure of green projects creates a specific set of frictions, and those frictions have specific institutional remedies. The chapter is organized around that logic.
2.1 Information Asymmetry in Green Lending
2.1.1 The Basic Problem
The economics of lending under asymmetric information is one of the most well-developed areas of financial theory. The essential structure is familiar. A borrower knows more about the quality of a project than a lender does. The lender can screen, but screening is costly and imperfect. The borrower can act in ways the lender cannot observe. The resulting equilibrium may involve prices that do not clear the market, and quantities allocated by non-price mechanisms.
Green projects have a particular configuration of these problems, and the policy implications depend on which part of the information problem you are trying to solve.
Consider a steel mill in Hebei province. The mill wants to finance a transition from blast furnace to electric arc furnace capacity. The technology is commercially proven but not at this scale in this location. The environmental benefit depends on the carbon intensity of the electricity the arc furnace will consume, which depends on the grid mix, which depends on policy decisions the mill does not control. The capital expenditure is large and front-loaded, while operating cost savings and carbon value accrue over a decade or more. The mill's management knows more about all three of these things than any lender can learn from a due diligence visit.
Now consider a forestry company in Yunnan. It wants to finance a rotation extension—letting trees grow longer before harvest to produce higher-value timber and increase carbon storage. The project's returns depend on growth rates that vary with rainfall and temperature, on fire risk, on the eventual timber price, and on whether the carbon storage will be recognized and compensated under a future policy regime. The company knows its own forest better than the bank does. It cannot fully convey that knowledge in a loan application. And it cannot commit to managing the forest in a particular way for twenty years when the bank is not watching.
These two cases share a structure. In both, the lender faces a project whose returns depend on a technology or practice whose outcomes are uncertain, an environmental benefit that is hard to verify, and a long time horizon over which the borrower's behavior and the policy environment can change. The conventional tools of credit assessment—historical financial statements, collateral valuation, cash flow projections—were built for a world where these three things were either absent or stable. They do not handle this configuration well.
2.1.2 Adverse Selection: The Green Lemons Problem
Adverse selection arises before the loan is made, when the lender cannot distinguish good green projects from bad ones.
George Akerlof's lemons model is the canonical treatment. When buyers cannot observe quality, the market price reflects average quality. Sellers of above-average quality withdraw, average quality falls, and the market may unravel entirely. The application to green lending is direct. A lender who cannot distinguish a genuinely low-carbon steel project from a project that claims to be low-carbon but is not will price both at the same rate. The genuinely green borrower, who has invested in better technology and faces higher upfront costs, finds the rate too high and either exits or cross-subsidizes the pretender. The pretender, who has lower costs and no real environmental benefit, finds the rate attractive. The pool deteriorates.
The Chinese green bond market offers an instructive case. From 2016 onward, the definition of what counted as a "green" bond was not fully harmonized across regulators. A bond labeled green by one standard might not qualify under another. Issuers quickly learned that the green label attracted investors willing to accept a slightly lower yield—the so-called greenium—and some issuers applied the label to projects whose environmental credentials were thin. The People's Bank of China and other regulators responded by tightening the catalog of eligible projects and requiring third-party verification. The verification requirement is a screening device: it raises the cost of false labeling and makes it easier for investors to distinguish quality.
The adverse selection problem in green lending has a feature that distinguishes it from the classic lemons case. The quality dimension that matters is multidimensional. A project can be good on carbon but bad on water, good on emissions but bad on biodiversity, good on current performance but bad on transition trajectory. A lender who wants to screen on environmental quality faces not one signal but a vector of signals, each noisy. Screening is more expensive and less reliable than in a single-dimension market. A single summary measure—a rating—becomes more valuable, not less, because it aggregates information that would otherwise be too costly to process project by project.
We do not have clean estimates of the magnitude of adverse selection in green credit markets, in the way we have estimates for used car markets. What we have is circumstantial: the persistence of yield differentials not explained by fundamental risk, the prevalence of third-party verification requirements, and the observed behavior of issuers who re-label projects in response to standard changes. The direction of the effect is clear; the size is not. I treat the magnitude as an open empirical question and return to it in Chapter 4, where the rating model provides a way to estimate it.
2.1.3 Moral Hazard: Behavior After the Loan
Moral hazard arises after the loan is made, when the borrower can take actions that affect the probability of repayment and that the lender cannot observe or enforce costlessly.
Green projects are unusually exposed to moral hazard for a structural reason: the environmental benefit is often a byproduct of how the project is operated, not a fixed attribute of the asset. A forestry company that borrows to extend a rotation can, after receiving the loan, shorten the rotation to harvest earlier and improve near-term cash flow, sacrificing carbon storage and long-term timber value. A steel mill that borrows to install emissions control equipment can operate the equipment intermittently to save on energy costs. A renewable energy developer can defer maintenance to boost distributable cash flow in the early years, reducing the asset's output and value later.
In each case, the borrower's private incentive diverges from the lender's interest. The lender wants the project operated in a way that preserves the cash flow stream and the collateral value. The borrower may prefer a different operating profile. The divergence is not necessarily large in any single period, but it compounds over a multi-year loan.
Conventional loan covenants address some of this. A covenant requiring the borrower to maintain a minimum debt service coverage ratio will catch a borrower who is harvesting too aggressively. But covenants are coarse instruments. They trigger on financial outcomes, not on environmental practices, and by the time a financial covenant is breached the environmental damage may already be done. A covenant requiring the emissions control equipment to operate at a specified utilization rate would be more targeted, but monitoring it requires data the lender does not routinely collect.
Better intentions do not solve this problem. Monitoring technologies and contractual structures that make the borrower's behavior observable at low cost do. Weather index insurance, which I take up in Chapter 6, is one such structure: it does not monitor the borrower's behavior directly, but it transfers a specific risk (weather) to a third party, which changes the borrower's incentive to take offsetting actions. Credit rating that incorporates environmental performance, which I take up in Chapter 3, is another: it creates a periodic, standardized assessment that makes drift visible.
2.1.4 The Verification Problem
A third asymmetry is less discussed in the standard literature but central to green finance: the verification problem. Even if a borrower is honest and a lender is diligent, the environmental outcome may be unobservable or unverifiable to both.
Carbon emissions are the obvious case. A steel mill's emissions depend on the carbon content of its inputs, the efficiency of its furnaces, and the grid mix of its electricity. Measuring them requires either continuous monitoring equipment—expensive and tamperable—or periodic audits, which are snapshots. The measurement error is not random; it is correlated with the borrower's incentive to underreport. A mill that wants to look greener than it is will find ways to make its reported emissions lower than its actual emissions.
The verification problem is worse for outcomes than for actions. A lender can verify that a borrower installed a piece of equipment. Verifying that the equipment reduced emissions by the claimed amount requires knowing the counterfactual—what emissions would have been without the equipment—which is unobservable. This is the fundamental evaluation problem in environmental finance, and it does not go away with better technology. It can only be managed by choosing contractual structures where the verifiable quantity is closely tied to the outcome of interest.
Weather index insurance is instructive here. It does not insure the farmer's actual loss; it insures a weather index correlated with loss. The index is verifiable—it comes from a government weather station—and it cannot be manipulated by the farmer. The cost of this verifiability is basis risk: the index may not match the farmer's actual loss in any given year. The design trade-off is between verifiability and accuracy, and it runs through every green financial instrument. I return to it repeatedly.
2.1.5 Why the Asymmetries Are Not Generic
The information asymmetries in green lending are not just the standard asymmetries with a green label attached. They have a specific structure: they are concentrated in the tail of the return distribution, they are correlated across borrowers, and they are sensitive to policy.
Concentration in the tail matters because credit risk is a tail phenomenon. A lender's expected loss is the product of probability of default, exposure at default, and loss given default. The information problems I have described affect the probability of default mainly in adverse scenarios—the scenarios where the technology fails, the policy support is withdrawn, or the environmental liability materializes. In normal scenarios, the information problems are latent. Conventional credit models, calibrated on average outcomes, will systematically underestimate green credit risk. The underestimation is small in good times and large in bad times, which is exactly the procyclicality problem that macroprudential regulators worry about.
Correlation across borrowers matters because it defeats diversification. If green projects fail independently, a diversified portfolio can absorb the failures. If they fail together—because they depend on the same policy support, the same technology trajectory, or the same weather system—then diversification does not help. The correlation is not a market failure in the conventional sense, but it is a reason why the information problems are harder to manage at the portfolio level than at the project level. I take this up in Chapter 5.
Sensitivity to policy matters because it makes the information environment non-stationary. A change in subsidy rules, a revision of green bond standards, or a shift in carbon pricing can change the value of a green project overnight. Lenders who have built their models on historical data will be wrong after the policy change, and they know it. This is a source of Knightian uncertainty, not just risk, and it is one reason why lenders ration credit rather than simply repricing it. When you cannot assign a probability to an outcome, you cannot price it. You can only limit your exposure.
These three features—tail concentration, correlation, and policy sensitivity—are what make green lending a distinct problem. They are also what make the integrated approach necessary, because no single instrument addresses all three. A rating addresses tail concentration by making the adverse scenario explicit. Insurance addresses correlation by transferring a defined risk to a third party with a different risk tolerance. Catalysis addresses policy sensitivity by providing a public backstop that reduces the borrower's exposure to policy shifts. I develop this argument in section 2.3.
2.2 Credit Rationing and Green Investment Gaps
2.2.1 The Theory of Credit Rationing
The observation that credit markets do not clear is old. It was formalized in the 1980s by Joseph Stiglitz and Andrew Weiss, whose 1981 paper showed that adverse selection and moral hazard can produce an equilibrium in which some borrowers are denied credit even though they are willing to pay the market rate, and even though the lender has funds to lend.
The mechanism is simple and worth restating because it is the foundation for everything that follows. Suppose a lender raises the interest rate to clear excess demand. Higher rates attract borrowers with riskier projects, because safer borrowers find the higher rate unattractive. The lender's expected return may fall as the rate rises, because the increase in default probability outweighs the increase in the contractual rate. Beyond some point, the lender's expected return is decreasing in the interest rate. The lender therefore chooses to ration credit rather than raise the rate further. The equilibrium is characterized by excess demand and non-price allocation.
Stiglitz and Weiss also showed that collateral does not solve the problem. Raising collateral requirements screens out poor borrowers, but it also discourages risk-averse borrowers and may increase the riskiness of the pool by attracting borrowers who are willing to pledge collateral because they have nothing to lose. Credit rationing can persist even with collateral.
The theory has been extended in many directions. What matters for green finance is the extension to projects with heterogeneous environmental quality. If the lender cannot observe environmental quality, and if environmental quality is correlated with default risk, then the rationing mechanism operates on the green dimension as well as the risk dimension. Green borrowers are rationed not because they are risky in the conventional sense but because the lender cannot distinguish them from non-green borrowers who are risky.
2.2.2 Why Rationing Falls Disproportionately on Green Firms
Green firms face more severe rationing than comparable non-green firms for four reasons, and the policy response depends on which one dominates.
The first is the novelty premium. Lenders have long histories with conventional technologies and short histories with green ones. The absence of historical default data means that the lender's estimate of default probability is imprecise. Under uncertainty, lenders behave conservatively: they assume the worst, or they decline to lend. A steel mill with a conventional blast furnace has a track record that a lender can model. A steel mill with a new electric arc furnace does not. The mill may be a better credit risk in reality, but the lender cannot know that, and the safe move is to ration.
Green assets are often poor collateral. A solar farm's panels have limited resale value and degrade over time. An emissions control system is bolted to a specific plant and cannot be removed without destroying its value. A forestry project's collateral is the standing timber, which is illiquid and subject to fire and disease risk. Conventional lenders rely on collateral to reduce loss given default. When collateral is weak, the lender's expected loss rises even if the default probability is unchanged, and the lender rations.
Cash flow timing makes the third reason. Green projects often have back-loaded cash flows: high upfront capital expenditure, then a long period of modest operating cash flow, then a terminal value that depends on policy and technology. Lenders prefer projects with front-loaded or stable cash flows, because those cash flows service the debt. A project whose cash flows are back-loaded requires the lender to take a view on the distant future, which increases the effective risk. The lender rations.
The fourth is the policy dependence. Green projects are more exposed to policy than conventional projects, because their returns depend on subsidies, carbon prices, and regulatory standards that can change. A lender that cannot predict policy cannot predict cash flows. The rational response is to shorten the tenor of the loan, to require more equity, or to decline. All three are forms of rationing.
These four reasons are not independent. They reinforce each other. A novel technology with weak collateral, back-loaded cash flows, and policy dependence is a lender's nightmare, even if the underlying project is sound. The mill in Hebei that I mentioned earlier faces all four. It is not surprising that it struggles to finance its transition, and it is not a sign that the lender is irrational. It is a sign that the lender's tools are mismatched to the project's characteristics.
2.2.3 The Size of the Gap
How large is the green investment gap? The estimates vary widely, and I want to present them with appropriate caution.
The most cited figures come from the International Energy Agency and the Climate Policy Initiative. The IEA's World Energy Investment reports put global energy investment at roughly $2.8 trillion in 2023, of which about $1.7 trillion was clean energy. The IEA's net-zero scenario requires clean energy investment to rise to about $4.5 trillion annually by 2030. The gap is on the order of $2.8 trillion per year.
For China specifically, the estimates are similarly large. A 2022 estimate by the People's Bank of China put the additional green investment needed through 2030 at 2 to 3 trillion RMB per year to meet the country's carbon peaking and carbon neutrality targets. The exact figure depends on assumptions about technology costs and policy ambition, but the order of magnitude is not in dispute.
These aggregate figures are useful for motivating the problem but not for diagnosing it. The gap is not a single number; it is a distribution across sectors, technologies, and firm sizes. Some segments of the green economy are well-financed—utility-scale solar and wind in China have attracted substantial capital, and the cost of capital for these projects has fallen. Other segments are chronically underfinanced—energy efficiency in small and medium enterprises, agricultural transition, forestry, and early-stage green technology. The rationing problem is concentrated in the underfinanced segments, and that is where the integrated approach matters most.
The existence of an investment gap does not by itself prove that credit rationing is the cause. The gap could reflect genuine uncertainty about returns, or a mismatch between the risk preferences of savers and the risk profile of green projects, or simply the early stage of the transition. The evidence that rationing is a distinct cause comes from three sources: the persistence of the gap despite falling technology costs, the observed behavior of lenders who decline creditworthy green projects, and the fact that the gap is largest in segments where information problems are most severe. None of these is conclusive on its own. Together they make a strong case.
2.2.4 The Consequences of Rationing
The consequences of credit rationing for green firms go beyond the direct loss of financing. Four transmission channels explain why rationing has persistent effects.
The first is the scale effect. A firm that cannot borrow cannot invest at the scale needed to achieve unit cost reductions. Green technologies have learning curves: costs fall with cumulative production. A firm that is rationed produces less, learns less, and remains high-cost. The rationing is self-reinforcing.
When credit is rationed, firms substitute toward projects that are easier to finance, which are often the less ambitious ones. A steel mill that cannot finance a full transition may finance a marginal efficiency improvement instead. The marginal improvement is better than nothing, but it does not achieve the transition. The rationing steers investment toward incrementalism.
The third is the selection effect. Rationing favors firms with strong balance sheets and established relationships with lenders. These are often the larger, older, more carbon-intensive firms. New entrants with better technology but no track record are excluded. The rationing protects incumbents and slows the entry of innovators.
The fourth is the expectations effect. Firms that expect to be rationed invest less in project preparation, because the expected return on preparation is lower. The reduction in preparation makes the projects less bankable, which reinforces the rationing. This is a coordination failure: each firm's response is rational given the others' behavior, but the aggregate outcome is worse for everyone.
These four channels are why the green investment gap is not self-correcting. In a well-functioning market, the gap would attract capital and close. In a market with rationing, the gap persists because the mechanisms that would close it are themselves impaired.
2.2.5 What Rationing Does Not Explain
I want to be explicit about the limits of the rationing story, because overclaiming would undermine the argument.
Rationing does not explain why some green projects with strong sponsors and proven technologies still fail to attract capital. Those failures are often due to genuine risk—the project may be a bad bet—or to factors outside the financial system, such as permitting delays or grid connection bottlenecks. The financial system can only do so much.
Rationing also does not explain the variation in financing costs across countries. The cost of capital for solar projects in China is lower than in many other developing countries, and the difference is not primarily due to information problems. It reflects macroeconomic factors, currency risk, and the depth of local capital markets. The integrated approach I develop in this book is not a substitute for macroeconomic stability.
Finally, rationing does not imply that the solution is to force lenders to lend more. Directed lending has a poor track record, and for good reason: it overrides the information that lenders have, which is often valuable even when incomplete. The solution is to improve the information environment, not to ignore it. That is the logic of the three instruments.
2.3 The Case for Integrated Finance
2.3.1 Three Problems Wearing the Same Coat
The argument so far has identified a set of information problems and a set of rationing consequences. The temptation is to look for a single instrument that solves them all. That temptation should be resisted.
It is at least three problems wearing the same coat. Lenders cannot distinguish good green projects from bad ones before lending. They cannot observe or enforce the borrower's behavior after lending. And even a well-screened, well-monitored project may fail because of risks that neither borrower nor lender can bear efficiently.
Each problem has a different structure and calls for a different remedy. Screening is solved by information production—by a rating that aggregates and standardizes the relevant signals. Monitoring is solved by contractual structures and data systems that make behavior observable—by insurance that ties payouts to verifiable events, and by rating systems that require periodic reassessment. Risk-sharing is solved by transferring specific risks to parties with a comparative advantage in bearing them—by insurance, and by public catalysis that absorbs the policy risk that private parties cannot.
No single instrument does all three. A rating without insurance leaves the lender exposed to weather and policy risk. Insurance without a rating leaves the insurer unable to price the risk. Catalysis without either leaves the public money exposed to the same information problems that afflict private lenders. The instruments are complements, and the complementarity is the core of the book's argument.
2.3.2 Rating as Screening Infrastructure
The first pillar is credit rating that incorporates environmental performance. I develop the model in Chapters 3 and 4; here I want to establish why it addresses the screening problem.
A rating is a standardized, forward-looking assessment of creditworthiness. Its value comes from three properties. It is comparative: it places a borrower on a scale relative to other borrowers, which allows the lender to price risk without building a bespoke model for every project. It is public: it can be shared among lenders, which reduces duplication and creates a common language. It is periodic: it is updated as new information arrives, which makes it a monitoring device as well as a screening device.
The extension to environmental performance is not a matter of adding a green score to a conventional rating. It is a matter of incorporating environmental variables into the assessment of default probability. A steel mill's transition risk—the risk that its carbon-intensive assets become stranded—is a credit risk. A forestry company's exposure to fire and disease is a credit risk. A renewable developer's dependence on subsidy policy is a credit risk. A rating that ignores these variables is not a complete rating; it is a rating that systematically misprices a class of risks that have become material.
The screening function is the key. By making environmental risk comparable across borrowers, the rating allows lenders to distinguish green projects from pretenders. The genuinely green borrower benefits from a lower rating-adjusted risk premium; the pretender does not. The adverse selection problem is mitigated because the signal is credible.
A rating is only as good as the data behind it. In markets where environmental data is scarce or unreliable, the rating will be noisy, and its screening value will be limited. This is why the digital infrastructure I discuss in Chapter 12 is a precondition for the rating pillar, not an afterthought.
2.3.3 Insurance as Risk Buffering
The second pillar is insurance, specifically weather index insurance and forest insurance. I develop the instruments in Chapters 6 and 7; here I want to establish why they address the monitoring and risk-sharing problems.
Insurance does something that a loan cannot do: it transfers a defined risk to a party that is better able to bear it. A weather index insurance contract transfers the risk of a specified weather event—drought, excess rainfall, low temperature—to an insurer. The farmer retains the risk of everything else. The transfer is valuable because the insurer can diversify the weather risk across a portfolio of policies, while the farmer cannot.
The monitoring function is subtler. A conventional loan covenant requires the lender to monitor the borrower's behavior. An insurance contract requires the insurer to monitor the index, which is a public, verifiable quantity. The monitoring cost is lower because the index is not under the borrower's control. This is the verifiability advantage I mentioned earlier. The index does not measure the farmer's loss perfectly, but it measures it in a way that cannot be manipulated.
The risk-sharing function is direct. By transferring weather risk, insurance reduces the variance of the farmer's income, which reduces the probability that the farmer defaults on a loan in a bad year. The reduction in default probability is the mechanism by which insurance supports credit. I will quantify this in Chapter 8, but the direction is clear.
The complementarity with rating is important. Insurance makes a borrower's cash flow more predictable, which improves the borrower's rating. A better rating lowers the cost of credit, which makes the insurance more affordable. Chapter 8 quantifies this feedback loop.
2.3.4 Catalysis as Risk Absorption
The third pillar is sci-tech finance—the set of institutions and instruments that channel capital toward early-stage green technology. I develop this in Chapters 9 and 10; here I want to establish why it addresses the policy-sensitivity problem.
Early-stage green technology is the hardest case for private finance. The technology is unproven, the market is uncertain, and the returns are distant. Private lenders and investors can bear some of this risk, but not all of it, because the risk is correlated across projects and because the policy environment that determines the returns is itself uncertain. The public sector has a comparative advantage in bearing these risks because it can diversify across the whole portfolio of green technologies and because it can influence the policy environment that creates the risk.
Catalysis takes many forms: government-guided venture capital, public guarantee schemes, concessional loans, and blended finance structures. The common thread is that public capital absorbs a portion of the risk that private capital cannot bear, which makes the remaining risk acceptable to private investors. The public capital functions as a risk-sharing arrangement that corrects a market failure in risk allocation, not as a subsidy in the conventional sense.
The complementarity with rating and insurance is less direct but still important. Catalysis is most effective when it is targeted at projects that have passed a rating screen—that is, projects that are fundamentally sound but face a specific risk that the public sector can absorb. And catalysis is more effective when the risks it absorbs are well-defined, which is what insurance provides. A public guarantee that covers a defined risk is cheaper and more effective than a guarantee that covers everything.
2.3.5 The Complementarity Argument
The complementarity argument is the theoretical core of the book, and I want to make it precisely.
Consider a lender deciding whether to finance a green project. The lender's decision depends on the expected loss, which is the product of default probability, exposure at default, and loss given default. Each of the three instruments affects a different component of this product.
The rating affects the default probability by improving the lender's estimate of it. A better estimate does not change the true default probability, but it changes the lender's perception, which is what determines the lending decision. The effect is largest when the lender's prior is most uncertain—that is, for novel technologies and new borrowers.
Insurance affects the default probability by reducing the borrower's cash flow volatility. A borrower with insurance is less likely to default in a bad state of the world, because the insurance payout covers the loss. The effect is largest when the borrower's cash flow is most exposed to the insured risk—that is, for agricultural and forestry borrowers.
Catalysis affects the loss given default by providing a backstop that reduces the lender's loss in the event of default. A public guarantee that covers a portion of the loss reduces the lender's expected loss even if the default probability is unchanged. The effect is largest when the lender's recovery is most uncertain—that is, for projects with weak collateral.
The three effects are additive in the expected loss, but they are multiplicative in their impact on the lending decision, because the decision is a threshold function. A project that is just below the threshold on all three dimensions may be above the threshold if any one of them improves. This is why the integrated approach can finance projects that no single instrument can finance. The instruments do not just add up; they interact.
There is a second complementarity that runs through the information environment. Each instrument generates information that improves the others. The rating generates data on environmental performance that the insurer can use to price risk. The insurance generates data on loss experience that the rating agency can use to calibrate default probabilities. The catalysis generates data on technology trajectories that both the rating agency and the insurer can use. The information is a public good, and the instruments are more valuable together than apart because they share it.
2.3.6 What the Integrated Approach Does Not Claim
I want to be explicit about what the argument does not claim, because the temptation to overclaim is strong.
It does not claim that the integrated approach eliminates information problems. It reduces them. There will always be residual asymmetry, and there will always be projects that cannot be financed. The goal is to move the frontier, not to reach a world without friction.
It does not claim that the three instruments are the only ones that matter. Fiscal policy, industrial policy, and international governance are all essential, as I noted in Chapter 1. The financial system can do its part, but it cannot do everything.
It does not claim that the instruments work equally well in all institutional contexts. The rating pillar requires a data infrastructure that many developing countries lack. The insurance pillar requires an actuarial base and a regulatory framework that many markets have not built. The catalysis pillar requires a public sector with the capacity to select and monitor investments. The approach is a framework, not a recipe, and its implementation must be adapted to local conditions.
It does not claim that the complementarity is automatic. The instruments can be deployed in ways that undermine each other. A rating that is captured by issuers is worse than no rating. Insurance that is mispriced creates adverse selection. Catalysis that is politicized misallocates capital. The design details matter, and I will spend much of the book on them.
2.3.7 The Road Ahead
The rest of the book develops the three pillars in sequence. Part II (Chapters 3 through 5) develops the rating pillar: how to incorporate environmental performance into credit assessment, how to model the resulting default probability and loss given default, and how to manage the portfolio-level risks that emerge when green exposures are correlated. Part III (Chapters 6 through 8) develops the insurance pillar: the design of weather index and forest insurance, the empirical evidence on their effects, and the linkage between insurance and credit. Part IV (Chapters 9 through 11) develops the catalysis pillar: the structure of sci-tech finance, the role of government-guided venture capital, and the design of performance evaluation and incentives. Part V (Chapters 12 through 15) addresses the cross-cutting issues—digital infrastructure, international cooperation, policy sequencing—and concludes.
The logic of the sequence is the logic of the argument. The rating comes first because it is the foundational information infrastructure. Insurance comes second because it builds on the rating and addresses the risks that the rating reveals. Catalysis comes third because it addresses the residual risks that neither rating nor insurance can handle. The sequence is not a hierarchy of importance; it is a sequence of dependence.
I want to close this chapter with a concrete image. The steel mill in Hebei that I mentioned at the start has a choice. It can continue to operate its blast furnaces, accepting the transition risk and hoping that policy support continues. Or it can invest in electric arc capacity, accepting the technology risk and hoping that the financing materializes. The choice is not irrational on either side. What makes it difficult is that the information needed to make it—about the technology, the policy, and the market—is dispersed and uncertain.
The integrated approach does not eliminate that uncertainty. It makes it manageable. A rating tells the mill and its lenders where it stands relative to peers. Insurance covers the risks that can be transferred. Catalysis absorbs the risks that cannot. The mill still faces a choice, but it faces it with better information and better risk-sharing.
The next chapter begins the work of building the rating pillar. It asks a question that is deceptively simple: how should a credit rating incorporate environmental performance? The answer, as we will see, is not to add a green score but to rebuild the assessment of default probability around the risks that the transition creates.