Chapter 13: Designing Incentives Without Distortion

Spain's solar feed-in tariff was set at €0.44 per kWh in 2007 and guaranteed for 25 years. By 2010 the cost of a solar panel had fallen by more than half, and the tariff had become a transfer of roughly €6 billion to €8 billion a year from electricity consumers to solar developers. The Spanish government tried to claw the money back, first in 2010 and again in 2013. Investors sued. Lending to Spanish renewable projects dried up for the better part of a decade.

That sequence—a subsidy set above the cost gap, locked in for a generation, then reversed under fiscal pressure—is the most common failure mode in green finance. It is not a story about generosity or stinginess. It is a story about whether an instrument changes behavior at the margin or merely pays for behavior that would have happened anyway. The design question is always the same: does the euro of public money buy an environmental outcome that private capital would not have produced on its own?

Answering that question requires more than good intentions. It requires a contract that survives the recipient's incentive to game it, the intermediary's incentive to relax underwriting, and the politician's incentive to renege when the bill comes due. This chapter works through the mechanics of such contracts, the cases where they have failed, and the institutional rules that separate a working incentive from a rent.

13.1 The Distortion Risk of Green Subsidies

13.1.1 What a rent looks like

A rent is a return above what is needed to bring a factor of production into a given use. Competitive markets compete rents away. Subsidized markets do not, because the subsidy itself is the barrier to entry. The firm that captures the subsidy earns more than it would without the program, and the difference is the rent.

Green subsidies create rents whenever the payment exceeds the incremental cost of the green activity relative to what the firm would have done anyway. Suppose a solar developer would have built the plant regardless, because solar was already the cheapest source of new generation in that market. A production subsidy paid to that developer changes nothing about emissions. It moves income from taxpayers to the developer.

This is not a hypothetical. The International Energy Agency's World Energy Outlook 2020 estimated that support policies for solar PV and wind in OECD countries during the 2010s paid out on the order of $100 billion more than the cost gap those technologies faced at the time of investment. Learning curves were steeper than the subsidy designers assumed. The gap closed faster than the tariffs did.

The calibration rule follows directly. A subsidy should be set to the gap between the private return and the social return, not to the total cost of the project. The gap is the only thing the subsidy needs to close. Pay more than the gap and you fund a rent. Pay less and you fail to induce the activity. The gap moves as technology costs fall and carbon prices rise, which means the subsidy has to move with it.

13.1.2 Additionality is the first test

Additionality is the property that an incentive causes an outcome that would not have occurred otherwise. Every green finance instrument should be tested against it, and most are not.

Carbon offset markets show the failure clearly. A forestry project that protects a stand of trees generates credits only if the trees were genuinely at risk of being logged. If the land was already protected by law, or if the timber was not economically accessible, the credits represent no avoided emissions. The purchaser has paid for an environmental benefit that did not occur.

The Clean Development Mechanism, established under the Kyoto Protocol in 1997, produced a large volume of credits whose additionality was questionable. A 2016 report by the EU Court of Auditors found that about 85% of the CDM projects it reviewed had a low likelihood of additionality. The mechanism had created a market in the appearance of emissions reduction. The environmental outcome was close to zero, and the institutional credibility of carbon markets took a decade to repair.

Additionality is hard to establish because it requires a counterfactual. We cannot observe what would have happened without the subsidy. We can only estimate it, and the estimate is contestable. The workable practice is to use conservative baselines, require independent verification, and build in periodic reassessment. The failure mode is to let the project developer set the baseline. The developer has an incentive to inflate it, because a higher baseline means more credits for the same physical outcome. This is a principal-agent problem with a known solution: separate the baseline-setter from the credit-recipient.

13.1.3 The ratchet

Subsidies create constituencies. Once a firm has organized its business around a production subsidy, it will lobby to keep the subsidy after the underlying cost gap has closed. The subsidy becomes a property right in the eyes of the recipient, and removing it is politically expensive.

The Spanish tariff is the canonical case. It was set well above the cost of solar generation at the time and guaranteed for 25 years. When the cost of solar fell faster than expected, the tariff became a large rent. The government's attempts to reduce it retroactively triggered a wave of investor lawsuits and a collapse in confidence in Spanish renewable policy. The legal uncertainty deterred investment for years afterward.

Two design implications follow. First, the subsidy should decline automatically as the cost gap closes. A tariff fixed for 25 years is a bet that the cost curve will not move. The cost curve always moves. Second, the subsidy should be tied to a measurable outcome, not to a technology. A production subsidy for solar pays for solar electricity regardless of whether solar is the cheapest way to reduce emissions at that moment. A subsidy for avoided emissions pays for the outcome and lets the market choose the technology.

13.1.4 Guarantees and moral hazard

Loan guarantees are common in green finance. A public entity guarantees a portion of a green loan, reducing the lender's credit risk and allowing the borrower to access capital at a lower rate. The logic is that green projects are capital-intensive and have long payback periods, and the guarantee bridges the maturity mismatch.

The guarantee also creates moral hazard. If the lender knows the public entity will absorb the first loss, the lender has less incentive to screen the borrower carefully or to monitor the project after disbursement. Underwriting standards relax. The guarantee attracts the projects that most need it, which are often the riskiest. The result is adverse selection: the guarantee pool is worse than the average green loan pool, and the public entity's losses exceed its actuarial estimate.

The U.S. Department of Energy's loan guarantee program, established under the Energy Policy Act of 2005, illustrates the dynamic. Solyndra received a $535 million guarantee and failed in 2011. The failure was not primarily a design failure—Solyndra's cylindrical panel technology was overtaken by cheaper flat silicon panels from Chinese manufacturers operating at scale—but the guarantee structure meant the lender, a syndicate led by a commercial bank, had limited incentive to monitor the project's commercial viability once the guarantee was in place. The public absorbed the loss.

A guarantee that leaves the lender with no capital at risk removes the lender's incentive to underwrite carefully. The standard remedy is a first-loss tranche. The European Investment Fund's InnovFin program guarantees a portion of the loan and requires the intermediary bank to retain the remainder on its balance sheet. The bank's own capital is the first loss, and the bank's underwriting standards stay intact. InnovFin's realized loss rates have run below the Department of Energy's guarantee program, though the comparison is complicated by differences in the underlying portfolios.

13.1.5 Directed credit and the selection problem

Directed credit programs allocate capital to specific sectors. The green finance version directs credit to renewable energy, energy efficiency, and clean technology. The logic is that these sectors produce positive externalities the market does not price, and directed credit corrects the underpricing.

The selection problem is that the program administrator does not know which firms in the target sector are creditworthy. The administrator can screen on observable characteristics—firm size, collateral, project type—but the unobservable characteristics, such as management quality and technical feasibility, determine repayment. The firms that self-select into the program are those that cannot access credit on commercial terms. The program becomes a magnet for the marginal borrower.

This is the same adverse selection problem that afflicts any credit program with a subsidized rate. The solution is not to abandon directed credit but to price it correctly. The subsidized rate should reflect the social value of the externality, not the credit risk of the borrower. If the borrower's credit risk is high, the borrower should pay a risk premium on top of the subsidized rate. A fixed subsidy per ton of CO2 avoided is transparent and does not distort the lender's risk assessment. A below-market interest rate is opaque and invites the lender to underprice risk.

13.1.6 Measuring the outcome

Every green incentive depends on a measurement. The subsidy is paid per ton of CO2 avoided, per megawatt-hour of renewable electricity, per hectare of forest protected. The measurement determines the payment, and the measurement is contestable.

Two dimensions matter. Accuracy asks whether the measurement reflects the physical outcome. Gaming asks whether the recipient can manipulate the measurement to increase the payment without increasing the outcome.

Accuracy is a technical problem. Emissions factors, baseline methodologies, and monitoring protocols are all approximations, and the error can be large. A study by the Stockholm Environment Institute found that the emissions factors used in the CDM for industrial gas projects overstated baseline emissions by a factor of two to three in some cases. The credits represented emissions reductions that did not occur.

Gaming is an incentive problem. If the payment is per hectare of forest, the recipient has an incentive to claim more hectares than are actually protected. If the payment is per ton of CO2 avoided, the recipient has an incentive to inflate the baseline. The standard remedy is independent verification by a third party with no financial interest in the outcome. The verifier's fee should not depend on the verification result, and the verifier should be liable for errors. The CDM's verification system had both flaws: verifiers were paid by the project developer, and their liability was limited. The result was a verification process that approved projects that should have been rejected.

13.1.7 Who gets paid

Green subsidies are not distributionally neutral. They transfer income from taxpayers or electricity consumers to the recipients of the subsidy. The recipients are typically firms with the capacity to navigate the application process, which means larger firms with dedicated grant-writing staff. Smaller firms, cooperatives, and community projects are underrepresented in the subsidy pool.

This is a design problem, not an inevitability. The German feed-in tariff, for all its cost, was designed to be accessible to small producers. Any household or cooperative could install a solar panel and receive the tariff. About half of Germany's renewable capacity in the 2010s was owned by individuals and cooperatives, not by large utilities. The distributional outcome was broader than in countries where the subsidy was channeled through large developers.

Access rules determine the distribution. A subsidy that requires a complex application, a large minimum project size, or a credit rating is a subsidy for large firms. A subsidy available on standard terms to any qualifying producer can reach smaller actors. The design choice is explicit, and it should be made deliberately.

13.2 Incentive-Compatible Contract Design

13.2.1 The principal-agent frame

Designing green finance incentives is an application of principal-agent theory. The principal is the public entity that wants a green outcome. The agent is the firm or financial institution that can produce the outcome. The principal cannot observe the agent's effort or the agent's private information. The principal can observe the outcome, and the outcome is a noisy function of effort and exogenous factors.

The principal's problem is to design a contract that induces the agent to exert effort and to reveal private information, subject to the agent's participation constraint (the agent must be no worse off than its outside option) and the agent's incentive compatibility constraint (the agent must prefer to exert effort rather than shirk).

Let \(e\) be the agent's effort, \(a\) be the agent's private information (for example, the true cost of the green project), and \(\theta\) be an exogenous shock. The outcome is \(y = f(e, a, \theta)\). The principal offers a payment \(t(y)\). The agent's utility is \(u(t(y)) - c(e)\), where \(c(e)\) is the cost of effort. The principal's problem is:

subject to:

where \(v(y)\) is the principal's valuation of the outcome and \(\bar{u}\) is the agent's reservation utility.

The solution has a familiar shape. The optimal contract pays the agent a share of the outcome that increases with the precision of the signal. If the outcome is a precise measure of effort, the agent should bear most of the risk. If the outcome is noisy, the agent should bear less risk, and the contract should rely more on fixed payments.

The application to green finance is direct. If the green outcome is easy to measure—the amount of renewable electricity produced, which can be metered—the contract can pay per unit of output. If the green outcome is hard to measure—the reduction in emissions from a complex industrial process—the contract should rely more on fixed payments or on proxies correlated with the outcome.

13.2.2 The informativeness principle

The informativeness principle, formalized by Holmström in 1979, states that any signal correlated with the agent's effort should be included in the contract, even if the signal is not itself the outcome the principal cares about. The contract should use all available information about the agent's behavior, not just the final outcome.

In green finance, the relevant signals include the firm's environmental management system, its emissions monitoring data, its supply chain disclosures, and its compliance history. A contract that pays only on the final emissions outcome ignores these signals and exposes the agent to more risk than necessary. A contract that uses the signals to adjust the payment can reduce the risk premium the agent demands and increase the efficiency of the contract.

The EU Taxonomy Regulation is an information infrastructure of this kind. It defines which economic activities qualify as environmentally sustainable and requires large companies to report the share of turnover and capital expenditure aligned with the taxonomy. The taxonomy does not itself pay anyone. It produces a signal that investors and lenders can use to condition their own contracts. A bank lending to a firm with high taxonomy alignment can offer a lower rate because the signal reduces the bank's uncertainty about the firm's environmental performance. The taxonomy's reporting requirement gives evaluators and investors the same data, which reduces information asymmetry.

13.2.3 Screening versus monitoring

The principal-agent problem has two dimensions: screening (the principal does not know the agent's type) and monitoring (the principal does not observe the agent's action). Green finance incentives must address both.

Screening separates good green projects from bad ones. The standard solution is a menu of contracts that induces self-selection. A firm with a genuinely green project will choose a contract with a high-powered incentive—a large share of the outcome—because it is confident in the outcome. A firm with a marginal project will choose a contract with a low-powered incentive, a fixed payment, because it wants to avoid the risk. The menu separates the types.

The green bond market uses a version of this mechanism. A green bond is certified by an independent reviewer, and the certification signals that the proceeds will be used for green projects. The issuer pays for the certification, and the cost is a signal of commitment. A firm that intends to use the proceeds for non-green purposes will not pay for certification. The certification is a screening device.

Monitoring ensures the agent carries out the project as promised. The standard solution combines reporting requirements, site visits, and financial covenants. The Green Loan Principles, developed by the Loan Market Association in 2018, require the borrower to report on the use of proceeds and to maintain a tracking system. The reporting requirement is a monitoring device.

The two dimensions interact. A screening device that works well can reduce the need for monitoring. A certification that is credible at the screening stage reduces the probability that the firm will deviate at the monitoring stage, because the firm has invested in its reputation. A firm that has paid for certification and publicly committed to a green use of proceeds has more to lose from deviation than a firm that has not. The reputation is a bond the firm posts.

13.2.4 Repeated interaction

The principal-agent model is typically static. The contract is signed once, and the agent's action is unobservable. In practice, green finance contracts are repeated. The firm borrows repeatedly, and the lender observes the firm's behavior over time. The repeated interaction changes the incentive structure.

In a repeated relationship, the agent's incentive to shirk is tempered by the threat of losing future business. The lender can terminate the relationship if the firm deviates. The value of the future relationship is the bond that sustains cooperation. This is the logic of relational contracting, and it explains why relationship banking is often more effective than arm's-length lending in markets where information is scarce.

Green finance incentives should be designed to support long-term relationships, not one-off transactions. A subsidy available only for the first green loan encourages a one-off transaction. A subsidy available for a series of green loans, conditional on continued compliance, encourages a relationship. The relationship is the enforcement mechanism.

13.2.5 Aligning the intermediary

Most green finance flows through intermediaries. The bank lends, the fund invests, the insurer underwrites. The intermediary's incentives determine whether the green finance reaches the intended borrower.

The intermediary's incentive is typically volume-based. The bank earns a fee on the loan, the fund earns a management fee on assets under management, the insurer earns a premium on the policy. The volume-based incentive is not aligned with the green outcome. The intermediary earns the same fee whether the project is green or not.

The alignment problem has a standard solution: pay the intermediary for the green outcome, not for the volume. A bank that originates a green loan could receive a payment contingent on the verified emissions reduction of the project. A fund that invests in green ventures could receive a carried interest contingent on the environmental performance of the portfolio. The contingent payment aligns the intermediary's incentive with the principal's objective.

Green securitization offers a working example. In a green securitization, the originator pools green loans and sells them to investors. The originator retains a portion of the pool, typically 5% under the EU's risk retention rules, and the retained portion is the first loss. The originator's own capital is at risk, and the originator has an incentive to screen the loans carefully. The risk retention rule is an alignment device, and it is the reason green securitization has maintained higher underwriting standards than the unsecuritized green loan market in some jurisdictions.

13.2.6 The counterfactual problem

Every incentive-compatible contract requires a baseline. The payment is the difference between the observed outcome and the baseline. If the baseline is wrong, the payment is wrong.

The baseline problem is acute in green finance because the counterfactual is unobservable. We do not know what the firm would have done without the subsidy. We can estimate the counterfactual using a control group, a trend line, or an engineering model, but the estimate is always contestable.

The workable practice is to use multiple baselines and to pay on the most conservative one. A project that reduces emissions by 100,000 tons under a conservative baseline and 150,000 tons under an optimistic baseline should be paid on the 100,000-ton figure. The conservative baseline reduces the risk of overpayment and preserves the credibility of the program. The cost is that some genuinely additional reductions are underpaid. The trade-off is deliberate: it is better to underpay a genuine reduction than to overpay a fictitious one.

The baseline should also be updated as technology and policy change. A baseline appropriate in 2015 may be obsolete in 2025. Periodic reassessment is a design feature, not a bug. The CDM's failure to update its baselines was one reason its additionality record was poor.

13.2.7 Contract design for technology firms

The credit rating framework developed in Chapters 3 and 4 is a screening device for technology firms. The rating incorporates environmental performance into the assessment of the firm's creditworthiness. The rating is a signal the lender can use to price the loan.

The incentive-compatible contract that builds on the rating has a specific structure. The firm's interest rate is a function of its rating. A firm with a high environmental rating pays a lower rate. The rating is updated annually, and the rate is reset. The firm has an incentive to maintain or improve its rating, because the rating determines the cost of capital.

The contract is incentive-compatible if the cost of improving the rating is less than the benefit of the lower rate. The firm will invest in environmental performance up to the point where the marginal cost of improvement equals the marginal benefit of the rate reduction. The design problem is to set the rate schedule so that the firm's optimal investment is the socially optimal investment.

The rate schedule should be steep enough to induce investment but not so steep that it creates a cliff. A cliff occurs when a small change in the rating produces a large change in the rate. The firm near the cliff has an incentive to manipulate the rating rather than to improve the underlying performance. A smooth schedule, in which the rate declines gradually as the rating improves, avoids the cliff and preserves the incentive to invest.

13.2.8 Penalties

Incentives are not only rewards. Penalties are the other side of the contract. A green finance instrument that rewards good performance must also penalize bad performance, or the agent has no incentive to avoid the bad outcome.

The penalty should be proportional to the harm. A firm that fails to meet its emissions target should pay a penalty that reflects the social cost of the excess emissions. The U.S. Environmental Protection Agency's 2020 estimate of the social cost of carbon was about $51 per ton (and higher under more recent estimates). A penalty below the social cost under-deters. A penalty above the social cost over-deters and invites litigation.

The penalty should also be credible. A penalty that is never enforced is not a penalty. The enforcement mechanism must be independent of the agent and must have the resources to detect and punish violations. The EU's emissions trading system has a penalty for non-compliance of €100 per ton of CO2, plus the obligation to surrender the missing allowances. The penalty is credible because the EU has the administrative capacity to enforce it. The credibility is why the price of allowances has risen over time.

13.2.9 Time consistency

A government that announces a green subsidy today may have an incentive to renege tomorrow. The reneging can take the form of reducing the subsidy, changing the eligibility rules, or failing to enforce the penalty. If firms anticipate reneging, they will not invest.

The time-consistency problem is the reason green finance incentives must be embedded in institutions, not in discretionary policy. A feed-in tariff written into law is more credible than a subsidy renewed annually. A green bond standard administered by an independent regulator is more credible than a certification granted by a government ministry. The institution is the commitment device.

The European Union's taxonomy is an institutional commitment. The regulation is binding, and the technical screening criteria are developed by an independent platform. The commitment is credible because the cost of changing the regulation is high. A firm that invests in taxonomy-aligned activities can expect the alignment to be recognized in future years, which lowers its cost of capital.

13.3 Cases in Incentive Failure

13.3.1 Spain: a tariff that could not fall

Spain introduced a feed-in tariff for renewable electricity in 2007. The tariff was set at €0.44 per kWh for solar photovoltaic, roughly four times the average retail price of electricity at the time, and guaranteed for 25 years. Solar capacity grew from 700 MW in 2007 to 4,000 MW in 2008, far exceeding the government's target. The cost of the subsidy program rose to €6 billion to €8 billion per year, creating a tariff deficit the Spanish government financed by securitizing the debt.

The design failure was the combination of a high tariff and a long guarantee. The high tariff reflected the cost of solar in 2007, when solar was expensive. The cost fell rapidly, and by 2010 the tariff was a large rent. The long guarantee meant the rent persisted. The government's attempts to reduce the tariff retroactively in 2010 and 2013 triggered investor lawsuits and a collapse in confidence.

Spain's tariff could have declined automatically; the government chose a fixed 25-year rate instead. A tariff that declines by a fixed percentage each year, or that is set by auction, would have captured the cost reduction and avoided the rent. The Spanish experience is why subsequent feed-in tariff designs in Germany and elsewhere included automatic degression.

13.3.2 The CDM: when the baseline-setter is the beneficiary

The Clean Development Mechanism was established under the Kyoto Protocol to allow developed countries to meet their emissions targets by investing in emissions reduction projects in developing countries. The mechanism issued Certified Emissions Reductions (CERs) for verified reductions, and the credits could be used for compliance with the Kyoto targets.

The additionality problem was built into the mechanism. The project developer proposed the baseline, and the baseline was approved by a designated operational entity (DOE) paid by the developer. The developer had an incentive to propose a high baseline, and the DOE had an incentive to approve it.

The most notorious category was industrial gas projects. Projects that captured and destroyed HFC-23, a potent greenhouse gas produced as a byproduct of refrigerant manufacturing, generated large volumes of CERs at low cost. The credits were worth more than the refrigerant, so manufacturers had an incentive to increase production of the refrigerant in order to generate more credits. The mechanism paid firms to produce the very gas it was trying to eliminate.

The EU eventually banned the use of HFC-23 credits in its emissions trading system, and the CDM's credibility never recovered. The baseline must be set by an entity that does not benefit from a high baseline. The CDM violated that rule at its foundation.

13.3.3 Solyndra: a guarantee without a first loss

Solyndra was a California-based manufacturer of cylindrical solar photovoltaic panels. The company received a $535 million loan guarantee from the U.S. Department of Energy in 2009. The guarantee was intended to support the commercialization of an innovative technology—cylindrical tubes coated with copper indium gallium diselenide—expected to be cheaper than flat panels.

The bet failed. The price of flat silicon panels fell sharply between 2009 and 2011, driven by Chinese manufacturers operating at scale. Solyndra's cylindrical panels could not compete. The company filed for bankruptcy in 2011, and the U.S. government lost the guaranteed amount.

The design failure was not the guarantee itself but the absence of a first-loss tranche. The guarantee covered the entire loan, and the lender had no capital at risk after the guarantee was in place. The lender's incentive to monitor the company's commercial viability was weak. The government, as guarantor, did not have the capacity to monitor the company's technology and market position. The result was a guarantee that transferred risk to the public without preserving the private sector's screening and monitoring function.

A guarantee that leaves the lender with no capital at risk removes the lender's incentive to underwrite carefully. The European Investment Fund's guarantee programs use a first-loss structure, and their loss rates have been lower than the Department of Energy's.

13.3.4 Verra: verification paid by the seller

The voluntary carbon offset market has grown rapidly since 2015, driven by corporate net-zero commitments. The Taskforce on Scaling Voluntary Carbon Markets, convened by the Institute of International Finance in 2020, projected that demand for carbon credits could reach $50 billion by 2030. The growth has been accompanied by a series of verification failures.

A January 2023 investigation by The Guardian and Die Zeit found that more than 90% of the rainforest offset credits issued by Verra, the largest certifier, were "phantom credits" that did not represent genuine emissions reductions. The investigation focused on projects in the Brazilian Amazon and found that the projects had overestimated the threat of deforestation and had issued credits for forest that was not actually at risk. The credits were sold to companies including Shell, Gucci, and Salesforce.

The verification failure had the same structure as the CDM failure. The project developer proposed the baseline, and the certifier approved it. The certifier's fee depended on the volume of credits issued, which created an incentive to approve. The independent verification that was supposed to catch the problem was not independent enough.

Verification must be separated from the commercial interest in the credits. The verifier should be paid a fixed fee, not a fee that depends on the volume of credits. The verifier should be liable for errors, and the liability should be large enough to deter careless approval. The verification standard should be set by a public body, not by the certifier.

13.3.5 Green bonds: a small greenium

The green bond market has grown from $11 billion in 2013 to over $500 billion in 2023. The growth has been supported by certification standards, including the Climate Bonds Standard and the Green Bond Principles. The certification is a screening device: it signals that the bond's proceeds will be used for green projects.

The certification has produced a "greenium," a small premium in the price of green bonds relative to comparable conventional bonds. A 2022 study by the Bank for International Settlements estimated the greenium at 1 to 3 basis points. The greenium is the market's valuation of the certification signal. It is small because the certification is imperfect: investors know the proceeds may not be used as promised, and they discount the signal accordingly.

Certification is only as valuable as its enforcement. A certification not backed by monitoring and penalties is a label, not a commitment. The greenium is small because the label is cheap to obtain and the enforcement is weak. A certification that included independent monitoring and financial penalties for misuse would command a larger greenium, and the larger greenium would attract more capital to green projects.

13.3.6 Germany: success with a regressive bill

Germany's Renewable Energy Sources Act (Erneuerbare-Energien-Gesetz, EEG), introduced in 2000, is often cited as a success. The tariff supported the growth of renewable capacity from 6% of electricity generation in 2000 to 46% in 2020. The cost was borne by electricity consumers through a surcharge on their bills.

The success came at a cost. The surcharge rose to about €0.07 per kWh by 2020, adding roughly €200 per year to the average household's electricity bill. The cost was regressive: it fell disproportionately on low-income households, who spend a larger share of their income on electricity. The environmental outcome was a success; the distributional impact was a design failure.

The financing mechanism matters as much as the tariff level. A tariff financed by a broad-based surcharge is regressive. A tariff financed by general taxation is progressive, because the tax system is progressive. The choice of financing mechanism determines who bears the cost, and the cost should be borne by those with the ability to pay.

13.3.7 The EU ETS: a cap that had to be rebuilt

The EU Emissions Trading System, launched in 2005, is the world's largest carbon market. It covers about 40% of the EU's emissions, including power generation, industry, and aviation. The system caps total emissions and allows firms to trade allowances.

The ETS had a difficult start. The first phase (2005–2007) allocated allowances for free, based on historical emissions. The free allocation created windfall profits for firms that passed the cost of allowances through to consumers while receiving the allowances for free. The price of allowances collapsed to near zero in 2007 because the cap was too loose.

The system was reformed in subsequent phases. The cap was tightened, the free allocation was reduced, and a market stability reserve was introduced to absorb excess allowances. The price of allowances rose from €5 per ton in 2017 to over €80 per ton in 2023. The reform was a design correction: it aligned the cap with the emissions target and introduced a mechanism to manage the price.

A cap-and-trade system requires a credible cap and a price floor. A cap set above actual emissions is not a cap. A price that can fall to zero is not a price. The market stability reserve is a design innovation that has made the ETS more credible, and it is a model for other carbon markets.

13.3.8 The UK CfD: auctions and risk allocation

The UK introduced Contracts for Difference in 2014 as a replacement for the Renewables Obligation. A CfD is a long-term contract between a renewable generator and a government-owned counterparty, the Low Carbon Contracts Company. The generator receives a fixed strike price for its electricity, and the difference between the strike price and the market price is settled through the contract.

The CfD is an incentive-compatible design. The strike price is set by auction, so the subsidy is calibrated to the cost of the technology. The generator bears the risk of production—if the wind does not blow, the generator does not get paid—which preserves the incentive to build a reliable project. The generator does not bear the risk of the electricity price, which the contract transfers to the government. The allocation of risk is deliberate: the generator is best placed to manage production risk, and the government is best placed to manage price risk.

The auction mechanism is the key design feature. The auction reveals the cost of the technology and sets the subsidy at the minimum level needed to induce investment. The strike price for offshore wind fell from £119 per MWh in the first auction in 2015 to £37 per MWh in the fourth auction in 2022. The decline reflects the cost reduction in the technology and the competitive pressure of the auction.

Auctions are a superior mechanism for setting subsidies. An auction reveals the cost and sets the subsidy at the margin. A fixed tariff, by contrast, is set by an administrator who does not know the cost and is vulnerable to rent-seeking. The CfD is the design the Spanish feed-in tariff should have been.

13.3.9 India's PAT scheme: trading where measurement is cheap

India's Perform, Achieve and Trade (PAT) scheme, launched in 2012, targets energy-intensive industries: cement, steel, aluminum, fertilizer, pulp and paper, textiles, and chlor-alkali. The regulator sets a specific energy consumption target for each designated plant, expressed in tons of oil equivalent per unit of output. Plants that beat their target earn tradable energy-saving certificates. Plants that miss their target must buy certificates or pay a penalty.

PAT is instructive because it applies an emissions-trading logic to a measurement problem that is easier than carbon. Specific energy consumption is metered at the plant gate. The baseline is the plant's own historical consumption, verified by an accredited energy auditor. The regulator does not need to estimate a counterfactual about what the plant would have done; it needs only to measure what the plant actually consumed.

The first cycle (2012–2015) covered 478 plants and produced energy savings of about 8.7 million tons of oil equivalent, according to the Bureau of Energy Efficiency. The scheme's design has weaknesses. The targets were set plant by plant, which invited negotiation; some plants received lenient targets. The certificate market was thin, and prices were volatile. The penalty for non-compliance was set at the cost of the certificates, which meant a plant could simply pay and continue.

Still, PAT shows that a trading scheme can work when the measured quantity is observable at low cost. The lesson for green finance is that the choice of metric matters. A metric that requires an elaborate counterfactual—avoided deforestation, avoided emissions from a baseline—invites gaming. A metric that is metered at the boundary of the firm—energy consumed, electricity produced—is harder to game. Where the physics is measurable, the incentive is easier to design.

13.3.10 China's green bond taxonomy: a standard with a gap

China's green bond market became the world's second largest within five years of its 2015 launch. The People's Bank of China published the Green Bond Endorsed Project Catalogue in 2015, and the National Development and Reform Commission issued its own Green Bond Issuance Guidelines the same year. The two standards differed, and both differed from international practice.

The most consequential difference was the inclusion of "clean use of coal" and "clean coal" in the early Chinese catalogue. Projects that improved the efficiency of coal-fired power plants qualified as green. International investors and the Climate Bonds Initiative did not accept coal efficiency as green, because a plant that burns coal more efficiently still burns coal. The result was a split market: Chinese green bonds that would not qualify under international standards, and a discount on those bonds when sold to foreign investors.

China revised the catalogue in 2021, removing clean coal and aligning more closely with the EU taxonomy. The revision was prompted in part by the market's own signal: issuers wanted access to international capital, and international capital would not accept the coal definition. The episode shows that a taxonomy is not a neutral technical document. It defines what counts, and the definition determines who gets financed.

The design lesson is about the boundary of the standard. A taxonomy that is too broad—that includes activities with marginal environmental benefit—dilutes the signal and invites the greenium to shrink. A taxonomy that is too narrow excludes activities that could be improved. The 2021 revision was a correction toward the international boundary, and the greenium on Chinese green bonds sold to foreign investors narrowed after the revision.

13.4 Building the Incentive Architecture

13.4.1 Three layers, one loop

The analysis points to a three-layer incentive architecture for green finance. The first layer is the price signal: a carbon price that reflects the social cost of emissions. The second layer is the risk-sharing instrument: a guarantee, an insurance policy, or a securitization that transfers the risks the private sector cannot bear. The third layer is the outcome payment: a subsidy or a contract that pays for the green outcome the price signal does not fully capture.

The three layers are complementary, not substitutes. A carbon price alone is insufficient because it does not address the risk and maturity mismatches that afflict green projects. A guarantee alone is insufficient because it does not internalize the environmental externality. An outcome payment alone is insufficient because it does not provide the price signal that guides the allocation of capital across technologies.

This is the "rating-based identification—insurance-based risk buffering—sci-tech finance catalysis" mechanism the book has developed. The rating identifies the green firms. The insurance buffers the risk. The sci-tech finance intermediary channels the capital. The incentive design is the glue that holds the three together.

flowchart TD
    A[Carbon price signal] --> B[Green project economics]
    B --> C[Credit rating with environmental performance]
    C --> D[Risk-sharing instrument]
    D --> E[Outcome payment]
    E --> F[Green investment decision]
    F --> G[Verified environmental outcome]
    G --> A
    G --> C

The loop matters more than any single layer. The carbon price sets the economics. The rating identifies the green firms. The risk-sharing instrument buffers the risk. The outcome payment rewards the verified outcome. The verified outcome feeds back into the carbon price and the rating. A program that breaks the loop—that pays the outcome but never verifies it, or verifies it but never feeds the result back—will drift toward rent.

13.4.2 Calibrating the payment

The calibration rule for the outcome payment is the gap between the private return and the social return. Let \(r_p\) be the private return on the green project and \(r_s\) be the social return. The gap is \(\Delta = r_s - r_p\). The outcome payment should be set at \(\Delta\), no more and no less.

The rule is simple to state and hard to implement. The private return is observable in the project's financial statements. The social return is not. It must be estimated from the social cost of carbon, the local air quality benefits, the employment effects, and the technology learning spillovers. Each estimate carries an error band, and the errors compound. The calibration is therefore always approximate.

The practical discipline is to calibrate conservatively and revise on a schedule. A conservative calibration under-pays the early adopters and preserves the credibility of the program. A revision every three years, announced in advance, allows the payment to track the cost curve without inviting the ratchet. The announcement matters: an unannounced revision is a policy shock, an announced revision is a known feature of the contract that investors can price.

Consider what this means for a solar program. If the gap in 2020 is €40 per MWh and the learning curve suggests it will fall to €20 by 2023, the announced schedule should show the payment declining. A developer who builds in 2020 gets the 2020 rate for the life of the contract, but the contract's terms make clear that new projects in 2023 will receive less. The developer's return is protected; the taxpayer's exposure shrinks with the technology's cost.

13.4.3 Separating the functions

The most important design principle in this chapter is the separation of functions. The entity that sets the baseline must not be the entity that receives the payment. The entity that verifies the outcome must not be the entity that sells the credits. The entity that bears the first loss must not be the entity that approves the loan.

The separation is the institutional expression of the incentive-compatibility constraint. When two functions sit in one entity, the entity faces a conflict of interest, and the conflict is resolved in favor of the function that generates the revenue. The CDM put baseline-setting and credit-receiving in the project developer, and the baseline was inflated. The voluntary offset market put verification and credit sales in the certifier, and verification was weak. Solyndra put the guarantee and the monitoring in a public entity that lacked the capacity to monitor, and monitoring failed.

The separation costs something. It requires more institutions, more contracts, and higher transaction costs. That cost is the price of credibility. A green finance system that is credible commands a greenium; a system that is not commands nothing. The transaction cost is worth paying, and the evidence is the greenium itself: markets pay for certification they trust and ignore certification they do not.

There is a temptation, especially in countries with weak administrative capacity, to collapse the functions into a single agency because building three institutions is harder than building one. The temptation should be resisted. A single agency that sets the baseline, verifies the outcome, and disburses the payment is a rent-generating machine, however honest its staff. The honest version of the single agency is still a single point of failure; the corrupt version is worse.

13.4.4 What the public balance sheet is for

The public balance sheet has a role in the incentive architecture that the private balance sheet cannot fill. The public entity can bear risks the private sector cannot—the risk of a new technology, the risk of a long maturity, the risk of a policy change. The public entity can also set the rules the private sector must follow.

The public balance sheet should be used for the risks that are genuinely public: technological failure at the pre-commercial stage, a market that does not yet exist, a policy that is not yet credible. It should not be used for the risks the private sector can bear: a commercial project, a known technology, a market that is already established.

The boundary is the line between the public and private roles. The public entity takes the early-stage risk and exits when the technology is commercial. The private sector takes the commercial risk and scales the technology. The exit is the mechanism that transfers risk from the public to the private balance sheet. A public guarantee that never exits is not a bridge; it is a permanent subsidy to the lender, who collects the fee without bearing the risk.

The distinction is not always easy to draw in practice. A technology that looks pre-commercial to one analyst looks commercial to another. The discipline is to write the exit condition into the program at the start: the guarantee expires when the technology's levelized cost falls below a stated threshold, or when private lenders fund the technology without the guarantee for two consecutive years. An exit condition written after the fact is a political negotiation; an exit condition written before is a contract.

13.4.5 Giving the intermediary a stake

The intermediary's incentive is the weakest link in the architecture. The bank, the fund, and the insurer are paid for volume, not for outcome. The outcome payment reaches the project, not the intermediary. The intermediary has no direct stake in the environmental performance of the project.

The fix is to give the intermediary a stake. A bank that originates a green loan could receive a portion of the outcome payment, conditional on the verified performance of the loan. A fund that invests in green ventures could receive a carried interest tied to the environmental performance of the portfolio. The stake aligns the intermediary's incentive with the principal's objective.

The stake must be large enough to matter and small enough not to create a new distortion. A stake that is too large gives the intermediary an incentive to originate projects with high environmental performance regardless of their financial viability—a green version of the originate-to-distribute problem. A stake that is too small has no effect. The calibration is empirical, and it should be tested in pilot programs before it is scaled.

There is a second, cheaper lever: disclosure. An intermediary that must report the environmental performance of its green portfolio to its regulator and its investors has a reputational stake even without a financial one. The disclosure requirement is not as strong as a contingent fee, but it is easier to administer and harder to game than a fee that depends on a metric the intermediary itself reports. The two levers can be combined: a modest contingent fee plus a mandatory disclosure.

13.4.6 Reaching the small borrower

The small borrower—the farmer, the cooperative, the community project—faces a different incentive problem. The transaction cost of participating in a green finance program is high relative to the size of the loan. Certification, verification, and reporting are fixed costs, and the fixed costs are prohibitive for a small project.

The fix is aggregation. A cooperative of farmers can participate in a weather index insurance program as a single entity, with the insurance contract covering the aggregate risk. A community energy project can participate in a green bond issuance through a pooled vehicle. Aggregation spreads the fixed cost across many borrowers and makes the program accessible.

This is the logic of the cooperative and mutual finance institutions discussed in Chapter 9. The cooperative is the vehicle that allows the small borrower to access the green finance system. Its governance structure—one member, one vote—ensures the benefits of aggregation are shared among the members rather than captured by the management. The design of the cooperative's governance is therefore part of the design of the green finance incentive, not a separate question.

Weather index insurance, discussed in Chapter 7, is a case in point. The index is measured at a weather station, not at the individual farm, which removes the need to verify each farmer's loss. The fixed cost of the weather station is shared across all the farmers in the catchment. A single farmer cannot afford the station; a cooperative of two hundred farmers can. The aggregation is what makes the index insurance viable, and the index insurance is what makes the small farmer insurable.

13.4.7 Evaluation as a design feature

The evaluation of a green finance incentive tests whether it changed behavior. The test is the counterfactual: what would have happened without the incentive? The counterfactual is estimated using a control group, a difference-in-differences design, or a regression discontinuity. The estimate is the additionality of the incentive.

Evaluation should be built into the program from the start. A program that is not evaluated cannot be improved. The evaluation should be independent of the program administrator, and the results should be public. Public disclosure is the accountability mechanism that keeps the program from becoming a rent.

The evaluation should also measure the distributional impact. Who received the subsidy? Who bore the cost? The answers determine whether the program is progressive or regressive. The German feed-in tariff was environmentally successful and distributionally regressive. An evaluation that measured only the emissions outcome would have missed the regressivity and would not have prompted a redesign of the financing mechanism.

The evaluation is not free. It costs money, and it takes time. A program that is evaluated in its third year cannot correct its first two years. The practical compromise is to build a data infrastructure from the start—the reporting requirement that generates the data the evaluation will use—and to commission the evaluation once the data are available. The EU Taxonomy's reporting requirement is a data infrastructure of this kind. Large companies report the share of turnover and capital expenditure aligned with the taxonomy. The data allow evaluators to measure the taxonomy's effect on capital allocation, and they allow investors to condition their contracts on taxonomy alignment.

13.4.8 The political economy

The design of a green finance incentive is a political act. The incentive creates winners and losers, and the losers will resist. The winners will organize to defend the incentive. The political equilibrium determines whether the incentive is designed well or poorly.

The political economy has a predictable pattern. A subsidy concentrated on a few firms is easier to organize and defend than a subsidy spread across many households. The concentrated subsidy is more likely to become a rent. The dispersed subsidy is more likely to be designed for the public interest, because the beneficiaries are too numerous to organize.

The design implication is to favor dispersed subsidies. A carbon price that falls on all emitters is more politically sustainable than a subsidy to a few favored firms. A green bond standard that applies to all issuers is more sustainable than a certification granted to a select group. Dispersion is the political safeguard.

There is a counter-argument, and it deserves a hearing. Dispersed subsidies are harder to target, and a poorly targeted subsidy can be more wasteful than a well-targeted concentrated one. A carbon price that falls on all emitters is efficient only if the carbon price is high enough to change behavior; a low carbon price is a dispersed subsidy to the status quo. The choice between dispersion and concentration is therefore a choice between two failure modes—rent capture and dilution—and the right answer depends on the administrative capacity of the jurisdiction and the political strength of the affected interests.

13.4.9 The international dimension

Green finance incentives are set at the national level, but the environmental outcome is global. A subsidy in one country can be undermined by the absence of a carbon price in another. A firm can relocate its emissions-intensive production to a jurisdiction with weaker rules, and global emissions are unchanged.

The international dimension requires coordination. The EU's Carbon Border Adjustment Mechanism, introduced in 2023, is an attempt. The CBAM imposes a carbon price on imports of emissions-intensive goods from countries without a comparable carbon price. The mechanism extends the EU's carbon price to the global supply chain and prevents carbon leakage.

The CBAM is an incentive design with cross-border reach. It is also a source of friction: exporting countries view it as protectionist. The design challenge is to make the mechanism compatible with WTO rules and with the development needs of the exporting countries. The revenue from the CBAM could be recycled to support the green transition in the exporting countries, which would address the equity concern and improve the political acceptability. Whether the EU will do so is an open question; the regulation as adopted phases in the mechanism over 2026–2034 and leaves the revenue question to the member states.

13.5 The Design Checklist

The analysis in this chapter condenses into ten questions. The checklist is not a substitute for judgment, but it is a discipline that catches the most common failures.

Walk through the table with a concrete program. Take a hypothetical green guarantee fund for small-scale solar in a middle-income country. The fund guarantees 70% of each loan and requires the originating bank to retain 30% as first loss. That answers the fourth question. The subsidy component—a top-up payment per verified MWh—is set at the estimated gap and scheduled to decline by 10% a year. That answers the first. The MWh are metered at the inverter, not estimated from a baseline, which answers the second and third questions by making the baseline the meter reading. The penalty for a bank that misreports the MWh is a clawback of the guarantee fee plus a fine, enforced by the central bank, which answers the fifth. The program is written into the banking law rather than renewed by ministerial decree, which answers the sixth. Cooperatives can apply as a single borrower, which answers the seventh. The bank receives a small contingent fee on verified MWh, which answers the eighth. An independent evaluation is commissioned in year three, which answers the ninth. The program is financed from the general budget rather than a surcharge on electricity, which answers the tenth.

A program that passes all ten tests is likely to be incentive-compatible. A program that fails one or more is likely to produce a rent, a moral hazard, or a distributional injustice. The checklist does not guarantee success—no checklist does—but it makes the failure modes visible before the money is spent.

13.6 What the Cases Add Up To

Four cases recur through this chapter, and each isolates one failure. Spain shows what happens when a tariff cannot fall: the rent persists until the fiscal cost forces a reversal, and the reversal destroys the credibility that the next program needs. The CDM shows what happens when the baseline-setter is the beneficiary: the baseline inflates, the credits are phantom, and the market's credibility takes a decade to rebuild. Solyndra shows what happens when the guarantee removes all downside from the lender: the lender stops monitoring, and the public absorbs the loss. Verra shows what happens when the verifier is paid by the seller: the verification approves what it should reject.

The four failures share a structure. In each, a single entity held two functions that should have been separated. The Spanish government held both the tariff-setting and the fiscal cost, and it resolved the conflict by reneging. The CDM developer held both the baseline and the credit, and it resolved the conflict by inflating. The Solyndra lender held both the origination fee and the credit risk, and after the guarantee it held only the fee. The Verra certifier held both the verification and the volume-based fee, and it resolved the conflict by approving.

The successes share the opposite structure. The UK CfD separates the auctioneer (the government) from the generator (the bidder) from the settlement counterparty (the Low Carbon Contracts Company). The EU ETS separates the cap-setter (the Commission) from the allowance-holder (the firm) from the verifier (the accredited auditor). Germany's EEG separated the tariff-setter from the grid operator, which is why the tariff was administered consistently even as its cost became politically contentious. India's PAT separated the target-setter (the Bureau of Energy Efficiency) from the auditor (the accredited energy auditor) from the certificate market (the power exchanges), which is why the measurement held up even as the target-setting was negotiated.

The separation of functions is the single most important rule in this chapter. It is also the rule most often violated, because building separate institutions is expensive and because the conflict of interest is invisible until the money is gone. The next chapter tests whether the policy-guided green finance programs followed the rule. The ones that did will have a record of additionality; the ones that did not will have a record of rents. The test is the same one this chapter has applied to the cases: who set the baseline, who verified the outcome, and who bore the first loss.