Chapter 10: Policy Architecture and Research Agenda for Green Value Chain Upgrading

The first compliance deadline under the European Union's Carbon Border Adjustment Mechanism arrived on 31 January 2024. Every Chinese exporter of iron, steel, aluminium, cement, fertiliser, hydrogen and electricity shipping into the EU had to file an emissions report covering the fourth quarter of 2023. In Foshan, a mid-sized aluminium extrusion plant supplying window frames to German fabricators found that its reported embedded emissions exceeded the default values its European buyer had used for preliminary cost estimates. The buyer did not cancel the order. It sent a spreadsheet and a request: document electricity sourcing at the facility level, and commit to a reduction trajectory over the next three years. The plant manager, who had spent two decades competing on price per tonne, spent the following month learning the difference between a location-based and a market-based grid emission factor.

That episode contains the architecture problem in miniature. CBAM is a trade instrument. The reporting it requires depends on accounting standards still being harmonised. The reduction trajectory the buyer wants depends on access to green electricity, which depends on grid infrastructure and power market design. The plant's ability to finance the retrofit depends on whether its bank treats the investment as a cost or as a capability upgrade. Four policy domains—trade, environment, energy, finance—converge on a single decision made by a single manager in a single factory. Misalign any one of them and the chain of incentives collapses. The manager files the report, pays a consultant, and changes nothing.

This chapter takes that convergence as its organising problem. The preceding nine chapters have argued from different angles that green governance capability is endogenous to value chain climbing: firms that build the capacity to measure, manage and improve environmental performance are also building the capacity to move up the value chain. The policy question is how to build an architecture in which the reward for that capability is legible, consistent and available to firms at different positions in different industries. The research question is what we still do not know about how the architecture works, and where the evidence is thin enough that policy is running ahead of it.

10.1 Designing a Coherent Policy Architecture

10.1.1 The integration problem

Industrial policy, environmental policy and trade policy are made by different ministries, in different legal instruments, on different time horizons. In China, the Ministry of Industry and Information Technology (MIIT) administers the industrial upgrading catalogue and the green factory certification programme. The Ministry of Ecology and Environment (MEE) sets emissions standards and runs the national carbon market. The Ministry of Commerce (MOFCOM) handles export promotion and trade remedy. The National Development and Reform Commission (NDRC) coordinates the dual carbon framework and the energy transition. Each has its own targets, reporting lines and relationship with firms. The Foshan aluminium plant deals with at least three of these ministries, plus a provincial development and reform commission, a municipal environmental bureau, and its European buyer's sustainability team.

The integration problem is not bureaucratic turf. It is a problem of incentive compatibility across instruments. Consider a stylised case. MIIT designates a firm as a "green factory" and awards it a subsidy for energy-efficient equipment. MEE's carbon market allocates free allowances to the same firm based on historical emissions. MOFCOM's export tax rebate structure does not distinguish between green and non-green products. The net effect may be that the firm receives a subsidy for the equipment, faces a carbon price too low to change its investment calculus, and exports at a rebate that offsets the cost of the equipment entirely. The instruments cancel each other. The firm is no worse off, but it has not climbed.

Coherence, then, is not about having a single policy. It is about ensuring that the marginal signal from one instrument is not neutralised by the marginal signal from another. The literature on policy mix—Borrás and Edquist (2013) on innovation policy instruments, Arfaoui, Brouillat and Saint Jean (2014) on the REACH regulation—has established that instrument interaction is not additive. Two instruments can be complementary, redundant or contradictory. The design task is to identify which combinations produce a marginal incentive to invest in green capability that survives the interaction.

Three principles follow from the evidence reviewed in this book.

The instrument that measures must be the instrument that rewards. The CBAM reporting requirement is a measurement instrument. If it is not connected to a cost or a market advantage, firms will treat it as compliance overhead. The EU's decision to phase in the financial obligation from 2026 gives firms a three-year window in which measurement is a preparation cost, not a penalty. That window is a design choice, and it is the right one—but only if firms use it to build measurement capability rather than to defer investment.

The instrument that rewards must be available at the scale at which firms actually invest. A carbon price of 52.78 yuan per tonne—the opening price on the national carbon market in July 2021—is not a marginal signal for a furniture factory deciding whether to replace a solvent-based coating line. The capital cost of the line is in the millions; the carbon saving is in the hundreds of tonnes. The price would need to be an order of magnitude higher, or the instrument would need to be a capital subsidy, or the reward would need to come through the buyer relationship rather than the carbon market. In the furniture case, the reward came through the buyer: the European retailer that required low-VOC coatings also paid a premium for the certified product. The carbon market was irrelevant to the decision. The buyer's specification was the instrument.

The architecture must distinguish between industries at different factor intensities. The preceding chapters have shown that labour-intensive, technology-intensive and capital-intensive industries face different green governance problems. A furniture cluster in Longjiang needs supplier finance and shared testing infrastructure. A battery manufacturer in Ningde needs grid access and chemical recycling capacity. A steel mill in Hebei needs carbon capture and a market for green hydrogen. A single policy architecture cannot address all three with the same instrument. The design task is a portfolio: differentiated instruments, coordinated across domains, with a common measurement standard that allows the portfolio to be evaluated.

10.1.2 The measurement layer: standards, accounting and the infrastructure of comparison

Before any instrument can reward green capability, the capability must be measurable in a way that is comparable across firms, sectors and borders. This is the measurement layer, and it is the least glamorous part of the policy architecture. It is also the part where the most consequential decisions are being made.

Carbon accounting for value chains is fragmented. The Greenhouse Gas Protocol provides three scopes: Scope 1 (direct emissions), Scope 2 (purchased electricity) and Scope 3 (value chain emissions). Most Chinese exporters report Scope 1 and Scope 2. Scope 3—which for a furniture manufacturer includes the emissions embedded in purchased wood, hardware, coatings and transport—is reported by a minority. CBAM initially covers only Scope 1 for most covered goods, with Scope 2 included for some and Scope 3 deferred. The EU's Corporate Sustainability Reporting Directive (CSRD) requires Scope 3 disclosure from large companies, which cascades the requirement down to their suppliers. A Chinese supplier to a European original equipment manufacturer may therefore face a Scope 3 data request from its buyer before it faces a CBAM obligation from the EU.

The measurement problem is not only technical. It is institutional. Who certifies the data? Who accredits the certifier? What happens when a Chinese provincial emission factor differs from an EU default value? The CBAM regulation allows the use of actual emissions data if it is verified by an accredited verifier, but the accreditation infrastructure for Chinese verifiers is still being built. In the interim, default values apply, and default values are conservative—they assume higher emissions than a well-managed plant actually produces. The Foshan aluminium plant that could document its electricity sourcing would be rewarded with a lower CBAM liability. The plant that cannot is penalised by the default.

This is where the measurement layer becomes a competitive instrument. Dechezleprêtre and Sato (2017) reviewed the evidence on environmental regulation and competitiveness and found that well-designed regulation can spur innovation that offsets compliance costs. The operative phrase is "well-designed." A measurement regime that rewards documented performance and penalises undocumented performance is well-designed if the documentation infrastructure is accessible. If it is not—if only large firms can afford the verifiers, the consultants and the software—then the measurement layer becomes a barrier to entry that protects incumbents and excludes the small and medium enterprises that make up the bulk of China's manufacturing employment.

The policy response has three components. The first is public investment in measurement infrastructure: shared testing laboratories at the cluster level, as discussed in Chapter 9, where small firms can access verified emissions data without building their own metering systems. The second is standard harmonisation: mutual recognition of verification between Chinese and EU accreditors, which is under negotiation but not concluded. The third is tiered reporting: a simplified reporting regime for SMEs that captures the major emissions sources without requiring the full Scope 3 inventory. The EU's own approach to SME reporting under CSRD—a "limited" reporting standard for listed SMEs and a voluntary standard for non-listed ones—offers a template. China's national carbon market, which initially covered only the power sector and is expanding to other sectors, could adopt a similar tiering.

Table 10-1 summarises the measurement instruments currently in play and their coverage gaps.

The table makes the integration problem visible. Five instruments, five different scopes, five different verification regimes. A firm that wants to comply with all of them maintains five datasets. A firm that wants to climb the value chain needs one dataset that is credible to all five. The architecture task is to build that one dataset—a "single source of truth" for emissions—and to make it the basis for every instrument in the portfolio.

10.1.3 The incentive layer: carbon pricing, subsidies and the buyer premium

If the measurement layer tells firms what their emissions are, the incentive layer tells them what those emissions are worth. Three incentive mechanisms are relevant to green value chain upgrading: carbon pricing, direct subsidies and the buyer premium.

Carbon pricing in China is currently weak. The national ETS opened in July 2021 with a price of 52.78 yuan per tonne and has traded in a range that is an order of magnitude below the EU ETS price. The low price reflects a generous allocation of free allowances, which in turn reflects a political decision to protect industrial competitiveness during the transition. The decision is defensible on short-run competitiveness grounds. It is not defensible as a long-run signal. A firm that invests in emissions reduction today receives a return of 50 yuan per tonne avoided. A firm that invests in the EU ETS receives a return of 80 euros per tonne avoided. The Chinese firm is being asked to climb the same value chain with a fraction of the reward.

The policy question is how to strengthen the price without stranding the firms it is meant to protect. Three options are on the table. The first is to tighten the cap and reduce free allocation, which raises the price but also raises the cost for covered firms. The second is to introduce a price floor or a price collar, which provides a predictable signal but requires a market design that can sustain it. The third is to link the Chinese ETS to other markets—the EU ETS, the Korean ETS, the emerging ASEAN markets—which would import a higher price but also expose Chinese firms to international competition for allowances. Joosung (2000) and Jan, Marliese and Michael (2008) on the dynamics of CO2 spot prices suggest that linkage is technically feasible but politically difficult, because it transfers price authority from the domestic regulator to the international market.

Subsidies are the second incentive mechanism. China's green manufacturing subsidy programmes—the green factory designation, the energy-saving equipment subsidy, the new energy vehicle (NEV) purchase subsidy—have been effective at building domestic demand for green products. They have been less effective at building export competitiveness, because they are not connected to the trade instruments that determine market access. A firm that receives a subsidy for energy-efficient equipment and then exports to the EU faces a CBAM liability calculated on the basis of its actual emissions, not on the basis of the subsidy. The subsidy lowers the firm's cost but does not lower its CBAM bill. The two instruments are not integrated.

The integration fix is to make the subsidy conditional on the measurement. A green factory that can document its emissions at the facility level should receive a larger subsidy than one that cannot, because the documented firm is closer to the standard that the export market requires. The EU's Innovation Fund for the ETS makes a similar link between documented innovation and subsidy, but the approach has not been adopted in China's green manufacturing programmes. The design change is administrative, not legislative: the subsidy criteria can be amended by ministerial regulation.

The third incentive mechanism is the buyer premium. This is the mechanism that has driven the most green upgrading in the industries this book has examined. The European furniture retailer that required low-VOC coatings paid a premium for the certified product. The German automotive OEM that required a carbon footprint declaration from its battery supplier paid a premium for the documented cell. The premium is not a carbon price in the textbook sense—it bundles a quality premium with a risk premium. But it functions as a price signal, and it is often larger than the carbon market price. Van Weele and Vivanco (2014) describe this as the shift "from compliance to value creation in value chain relationships": the buyer imposes a requirement, the supplier innovates to meet it, and the innovation creates value that the buyer shares. The policy implication is that the buyer premium should be supported, not supplanted, by public instruments. The public instrument that supports it is the measurement layer: if the supplier can document the requirement credibly, the buyer can pay the premium with confidence.

10.1.4 The trade layer: CBAM, export competitiveness and the risk of leakage

The trade layer is where the architecture is most contested. CBAM is the first border carbon adjustment to be implemented at scale, and its design choices will shape the terms on which Chinese manufacturers access the European market for decades.

CBAM's stated purpose is to prevent carbon leakage—the relocation of emissions-intensive production to jurisdictions with weaker climate policy. The mechanism is a charge on the embedded emissions of imported goods, calibrated to the EU ETS price, with a deduction for any carbon price already paid in the country of origin. The deduction is the crucial design feature for Chinese exporters. If a Chinese steel mill has paid the Chinese ETS price for its emissions, that payment is deducted from its CBAM liability. The effective CBAM charge is the difference between the EU price and the Chinese price. At current prices, that difference is large: a steel mill that pays 50 yuan per tonne in China and faces an 80 euro per tonne CBAM charge in Europe pays the difference, which is most of the CBAM bill.

The deduction mechanism creates an incentive for China to strengthen its own carbon price. Every yuan of Chinese carbon price reduces the CBAM liability euro for euro. This is the "climate club" logic that William Nordhaus proposed in 2015: a border adjustment that recognises domestic carbon pricing creates a reciprocal incentive for trading partners to price carbon. The logic is sound, but it depends on the Chinese price being credible to the EU. The EU's verification requirements for the deduction—the need to demonstrate that the carbon price was actually paid, that it was not rebated, and that it corresponds to the embedded emissions—are stringent. A firm that pays the Chinese ETS price but receives a free allocation that offsets it has not, in the EU's eyes, paid a carbon price. The deduction is reduced accordingly.

The trade layer also includes the export side. China's export tax rebates are a blunt instrument: they refund indirect taxes on exports to avoid double taxation. They do not distinguish between green and non-green products. A proposal that has circulated in policy circles is to differentiate the rebate by environmental performance: full rebate for products with documented low emissions, reduced rebate for products with high emissions. The proposal is attractive in principle—it would reward green capability at the border—but it faces a WTO problem. Differentiated rebates could be challenged as a subsidy or as a discriminatory measure. The legal analysis is unsettled. The policy design would need to be carefully calibrated to survive a dispute.

A less contested trade instrument is the mutual recognition of standards. China and the EU have been negotiating mutual recognition of conformity assessment for electrical and electronic equipment for years. Extending that negotiation to carbon accounting standards would reduce the measurement burden on exporters and increase the credibility of Chinese data in European markets. The negotiation is technical and slow, but it is the most direct route to reducing the CBAM's compliance cost without weakening its environmental integrity.

10.1.5 The cluster layer: place-based policy and the geography of upgrading

The instruments discussed so far are national or international. The cluster layer is local, and it is where the architecture meets the firm.

Chapter 9 argued that cluster governance is the meso-level institution that connects national policy to firm behaviour. The argument has a policy corollary: the cluster is the appropriate unit for certain kinds of public investment. Shared testing laboratories, supplier finance schemes, retraining programmes and industrial symbiosis infrastructure—the heat recovery networks, the shared wastewater treatment plants, the recycled material sorting facilities—are all more efficient at the cluster level than at the firm level. A single furniture factory cannot afford a low-VOC coating testing lab. A cluster of fifty furniture factories can.

The cluster layer is also where the regional differentiation problem is most acute. China's manufacturing clusters are not evenly distributed, and their green governance capacities are not evenly developed. The Pearl River Delta clusters—furniture in Longjiang, electronics in Shenzhen, textiles in Dongguan—are export-oriented and have been exposed to buyer-driven green requirements for longer. The Yangtze River Delta clusters—automotive in Shanghai, batteries in Ningde, chemicals in Nanjing—are more capital-intensive and more integrated into domestic supply chains. The inland clusters—steel in Hebei, coal chemicals in Shanxi, nonferrous metals in Yunnan—are more emissions-intensive and more exposed to the carbon market. A single cluster policy would be misallocated. The architecture needs a regional differentiation mechanism: different instruments, different intensities, different timelines, calibrated to the cluster's industrial composition and export orientation.

The regional differentiation mechanism also has a social dimension. The Longjiang worker whose sanding job moved to Haiphong is not the only worker affected by green upgrading. The coal miner in Shanxi, the steel worker in Hebei, the chemical plant operator in Nanjing—all face transition risks that are unevenly distributed across regions. The policy architecture needs a just transition component: retraining, redeployment and social protection for workers in declining green sectors. The component is not an add-on. It is a condition for the political sustainability of the transition. A carbon price that is not accompanied by a just transition mechanism will face resistance from the regions that bear the cost.

10.1.6 The finance layer: green credit, transition finance and the cost of capital

The final layer of the architecture is finance. A firm that wants to invest in green capability needs capital, and the cost of that capital depends on how the financial system prices green risk.

China's green bond market is the largest in the world by issuance, but the proceeds are concentrated in renewable energy, clean transport and green buildings. Green bonds for industrial process upgrading—the kind of investment a furniture factory or an aluminium extruder needs—are a small share of the market. The reason is partly definitional: the green bond taxonomy prioritises projects with clear environmental benefits, and process upgrades are harder to classify. It is partly risk-related: banks perceive process upgrades in traditional industries as higher-risk than renewable energy projects with guaranteed offtake.

The policy response has two components. The first is to expand the green finance taxonomy to include process upgrading in traditional industries, with clear eligibility criteria and measurement requirements. The second is to develop transition finance instruments—loans, bonds and guarantees that finance the transition path rather than the end state. The transition finance concept is newer and less standardised than green finance, but it is more relevant to the firms this book has examined. A steel mill that is reducing its emissions by 30 per cent over five years is not yet "green" by the taxonomy's definition, but it is on a transition path that deserves financing. A transition finance framework that recognises the path, not just the destination, would mobilise capital for the firms that need it most.

The cost of capital also depends on the carbon price. A higher carbon price raises the cost of brown assets and lowers the cost of green assets, which is the intended effect. But the effect is uneven across firms. A large firm with access to international capital markets can hedge its carbon exposure. A small firm cannot. The finance layer needs a mechanism to channel cheap capital to small firms that are making the transition, or the transition will be financed by the firms that can afford it, which are not always the firms that need it.

10.1.7 Putting the layers together: a stylised architecture

The five layers—measurement, incentive, trade, cluster, finance—are not a sequence. They are a system, and the system works when the layers are aligned. Table 10-2 presents a stylised architecture, with the alignment conditions marked as rows.

The alignment conditions are the arrows in the system. The measurement layer feeds the incentive layer: without a credible dataset, the carbon price and the subsidy cannot be calibrated. The incentive layer feeds the finance layer: a carbon price that is expected to rise makes green assets more bankable. The trade layer feeds back into the incentive layer: the CBAM deduction rewards a stronger domestic carbon price. The cluster layer supports all of the above: shared infrastructure lowers the cost of measurement, finance and compliance. And the whole system feeds the firm-level capability that is the subject of this book.

The architecture is stylised because the real system is messier. Instruments overlap, jurisdictions conflict, timelines slip. But the stylised version is useful because it makes the alignment conditions explicit. A policy reform can be evaluated by asking which alignment condition it strengthens or weakens. A reform that strengthens the measurement layer without connecting it to the incentive layer is incomplete. A reform that strengthens the incentive layer without addressing the finance layer is inequitable. A reform that strengthens the trade layer without the cluster layer is fragile.

10.2 Research Agenda for Green Governance

10.2.1 What we know and what we don't

The evidence reviewed in this book supports a set of claims with varying degrees of confidence. We know, with reasonable confidence, that green governance capability is correlated with value chain position: firms that have climbed the value chain tend to have better environmental performance, and firms with better environmental performance tend to climb. We know, with less confidence, that the relationship is causal and that the causality runs in both directions. We know, from case evidence, that the mechanisms differ by factor intensity: labour-intensive firms upgrade through buyer requirements and cluster infrastructure, technology-intensive firms through eco-innovation and the green premium, capital-intensive firms through energy transition and carbon markets. We do not know, with any precision, how large the effects are, how they vary across regions and firm sizes, or how they interact with the macro-level policy architecture.

The research agenda follows from these gaps. It is organised around four questions, each of which corresponds to a layer of the architecture and a body of evidence that is thin.

10.2.2 How does green governance capability actually diffuse within clusters?

The cluster is the unit of analysis that this book has emphasised, but the evidence on within-cluster diffusion is mostly case-based. We know that lead firms impose requirements on suppliers. We know that shared infrastructure lowers costs. We do not know how the requirements propagate through the supplier tier—whether a tier-one supplier's green standards reach the tier-two and tier-three suppliers, or whether they stop at the first tier. We do not know how much of the diffusion is driven by the buyer, how much by the local government, and how much by peer effects among firms in the same cluster.

The research design that would answer these questions is a network study with longitudinal data. Map the supplier relationships in a cluster, measure the green governance practices of each firm at two or three points in time, and trace the diffusion path. The study would need to distinguish between adoption driven by coercion (the buyer requires it), mimicry (the peer does it) and competition (the firm does it to win a contract). The distinction matters for policy: coercion is fast but fragile, mimicry is cheap but shallow, competition is slow but durable. A cluster policy that relies on coercion will produce compliance without capability. A policy that relies on competition will produce capability but take longer.

The cluster study would also need to address the measurement problem identified in section 10.1.2. The firms in the cluster have different levels of measurement capability, and the study's ability to detect diffusion depends on measuring practices consistently across firms. The study design should include a measurement audit as a baseline, so that the diffusion analysis is not confounded by differences in reporting quality.

10.2.3 What is the size and distribution of the green premium?

The green premium—the price advantage that a firm earns from documented environmental performance—is central to the book's argument, but its magnitude is poorly measured. The case evidence suggests that the premium varies widely: high in consumer-facing categories where brand value is at stake, low in commodity categories where price competition is intense, and variable in intermediate goods where the buyer's own green requirements determine the willingness to pay.

The research design that would measure the premium is a hedonic pricing study using transaction-level trade data. The data would need to link product-level prices to product-level environmental attributes—certified low-VOC content, recycled material share, documented carbon footprint—which requires matching customs records to certification databases. The matching is feasible in principle but has not been done at scale. The study would estimate the premium as the price differential between certified and uncertified products, controlling for quality, origin, buyer and time. The distributional analysis would ask whether the premium is higher for products sold to EU buyers than for products sold to domestic buyers, and whether it is higher for large firms than for small ones.

The premium study matters for policy because it determines the return on investment in green capability. If the premium is large, firms have a private incentive to invest and the policy task is to reduce the cost of measurement. If the premium is small, the private incentive is weak and the policy task is to create a public incentive—a subsidy, a carbon price, a procurement preference—that fills the gap. The current policy debate assumes the premium is large without evidence. The study would test the assumption.

10.2.4 Does the carbon market drive upgrading, and for whom?

The national ETS is the centrepiece of China's carbon pricing architecture, but its effect on industrial upgrading is uncertain. The price is low, the free allocation is generous, and the covered firms are large emitters in the power sector and a few industrial sectors. The theoretical prediction is that a carbon price induces abatement investment, but the empirical literature on the Chinese ETS is young and the results are mixed. Some studies find modest abatement effects; others find no effect on investment, only on operational adjustments.

The research design that would answer the question is a firm-level panel study exploiting the phased expansion of the ETS. The expansion from the power sector to other sectors creates a natural experiment: firms that are covered can be compared with firms that are not, before and after coverage, on investment, emissions and output. The study would need to address the selection problem—coverage is not random—by using the eligibility thresholds (energy consumption, emissions volume) as instruments. It would also need to distinguish between abatement that is driven by the carbon price and abatement that is driven by the co-benefits of energy efficiency, which are often the dominant motive in Chinese firms.

The distributional question is equally important. Does the ETS drive upgrading in large firms and exit in small firms? The evidence from the EU ETS suggests that the effect is regressive: large firms with the resources to invest in abatement capture the rents, while small firms bear the costs. If the Chinese ETS has the same effect, it will accelerate consolidation in the covered sectors, which may be desirable on efficiency grounds but has social consequences that the policy architecture needs to address. The just transition component discussed in section 10.1.5 is the policy response, but its design needs to be informed by evidence on who is actually affected.

10.2.5 How do trade instruments interact with domestic policy?

CBAM is the most studied trade instrument in the green governance literature, but the studies are mostly ex ante simulations. The ex post evidence is just beginning to accumulate, and it is limited to the reporting phase. We do not yet know how firms are responding to the reporting requirement—whether they are building measurement capability, hiring consultants, or deferring investment until the financial obligation begins. We do not know how the deduction mechanism is working in practice, whether Chinese firms are claiming the deduction, and whether the claims are being accepted. We do not know whether CBAM is causing leakage to other jurisdictions or whether it is inducing upgrading in the covered sectors.

The research design that would answer these questions is a mixed-methods study combining trade data analysis with firm-level interviews. The trade data would track export volumes and prices for covered goods, comparing Chinese exports to the EU with Chinese exports to non-EU markets, before and after CBAM's implementation. The interviews would probe the firm-level response: what the firm did when it received the reporting request, what it invested in, what it changed. The combination would allow the study to distinguish between the extensive margin (firms exiting the EU market) and the intensive margin (firms staying and upgrading), which is the distinction that matters for the upgrading argument.

The study would also need to examine the interaction between CBAM and the domestic carbon market. The deduction mechanism creates a link between the two, and the link is the channel through which CBAM could strengthen domestic carbon pricing. Whether it does depends on the credibility of the Chinese price to the EU verifier, which depends on the design of the Chinese ETS. The study would trace the link from the Chinese allocation rules to the EU verification decision, which is a chain of institutional choices that has not been documented.

10.2.6 What is the role of the buyer in the absence of a carbon price?

The buyer premium is the mechanism that has driven the most green upgrading in the cases this book has examined, but it is the least theorised. The global value chain literature has studied buyer-driven governance extensively—Gereffi's work on buyer-driven commodity chains is the foundational reference—but the green dimension of buyer governance is underdeveloped. We do not know when buyers impose green requirements, how they enforce them, and how they share the cost of compliance with suppliers.

The research design that would answer these questions is a buyer-supplier dyad study. The unit of analysis is the relationship, not the firm. The study would collect data from both sides of the relationship: the buyer's sourcing policy, the supplier's compliance cost, the price adjustment that reconciles them. The dyad design is expensive and difficult, but it is the only way to observe the bargaining that determines the premium. The alternative—surveying buyers and suppliers separately—cannot capture the interaction.

The dyad study would also address a question that the policy debate has not resolved: is the buyer premium a substitute for public policy or a complement? The case evidence suggests that it is a complement: buyers impose requirements that are stricter than public standards, and public standards provide the measurement infrastructure that makes the buyer's requirement credible. If the complementarity holds, the policy task is to strengthen the measurement infrastructure, not to regulate the buyer. If the relationship is substitutive—if buyers impose requirements only where public standards are weak—the policy task is different. The dyad study would test the relationship.

10.2.7 What are the limits of the compliance-to-capability transition?

The book's core argument is that green governance capability is endogenous to value chain climbing—that compliance can become capability. The argument is supported by the cases, but the cases are selected. They are firms that made the transition. The firms that did not make it—the firms that complied, paid the cost and did not upgrade—are not in the sample. The selection problem is the most serious limitation of the evidence base.

The research design that would address the problem is a panel study of firms that faced a green requirement and either upgraded or did not. The study would need to observe the firms before the requirement, at the moment of compliance, and after, and to compare the upgraders with the non-upgraders on a range of characteristics: size, ownership, prior capability, access to finance, position in the value chain. The comparison would identify the conditions under which compliance becomes capability, which is the question the book's argument implies but does not answer.

The panel design is feasible in the clusters where the book's cases are located, because the firms can be identified and tracked. It is less feasible in the aggregate, because the data do not exist. The study would be a cluster-level contribution, not a national one, but it would be the first to test the selection problem directly. The result would either strengthen the book's argument—if the upgraders are distinguishable ex ante by characteristics that policy can influence—or qualify it—if the upgraders are distinguished by luck, which policy cannot replicate.

10.2.8 How should the research agenda be sequenced?

The seven questions are not equally urgent. The sequencing should follow the policy timeline. CBAM's financial obligation begins in 2026, which makes question four the most urgent: the ex post evidence on the reporting phase needs to be collected now, before the financial phase begins and the reporting behaviour changes. The carbon market expansion is underway, which makes question three urgent: the panel data need to be assembled before the expansion is complete, or the natural experiment is lost. The cluster diffusion question (question one) is less time-sensitive but more foundational: it is the question that determines whether the cluster layer of the architecture is well-designed.

The premium question (question two) and the buyer question (question five) are complementary and should be sequenced together: the dyad study would generate the data that the hedonic pricing study needs. The limits question (question six) is the most ambitious and should be sequenced last, when the other studies have generated the measurement instruments and the panel infrastructure it requires.

Table 10-3 summarises the research agenda.

10.3 Implications for Firms and Policymakers

10.3.1 For firms: build the measurement capability first

The most actionable implication of this book for firms is that measurement capability is the foundation of green governance capability, and it should be built before it is required. The Foshan aluminium plant that discovered its CBAM exposure in January 2024 had a year to build the capability that its European buyer wanted. The plant that had built the capability earlier—that had installed facility-level metering, that had documented its electricity sourcing, that had trained a staff member in carbon accounting—was able to respond to the buyer's request in weeks rather than months. The difference is not the cost of the metering, which is modest. The difference is the decision to treat measurement as an investment rather than a compliance cost.

The decision has a second-order benefit that is not obvious ex ante. A firm that measures its emissions discovers inefficiencies that it did not know it had. The energy audit that accompanies a carbon inventory typically identifies savings that pay for the inventory within a year or two. The savings are not the point—the capability is—but they make the investment easier to justify to a sceptical manager or a cautious bank.

The firm-level implication extends to the choice of standards. A firm that adopts an internationally recognised standard—the GHG Protocol, ISO 14064, the Science Based Targets initiative—is building capability that is legible to multiple buyers in multiple markets. A firm that adopts a domestic standard that is not recognised internationally is building capability that is legible only at home. The former is more expensive and more useful. The choice is a strategic one, and it should be made with the export market in mind.

10.3.2 For firms: treat the buyer relationship as a capability channel

The cases in this book show that the buyer relationship is the most important channel through which green requirements reach Chinese firms. The relationship is often adversarial—the buyer imposes, the supplier complies—but it can be collaborative. The difference is whether the supplier treats the buyer's requirement as a cost to be minimised or as a problem to be solved jointly. The suppliers that climbed the value chainwere the ones that engaged the buyer in the specification process—that asked what the buyer was trying to achieve, that proposed alternatives, that negotiated the timeline. The suppliers that did not climb treated the requirement as a tax and passed it to the cheapest compliance vendor.

The practical implication is that the supplier's sustainability staff should be embedded in the commercial relationship, not sequestered in a compliance department. The buyer's sustainability team and the supplier's sustainability team should be in the same conversation as the buyer's sourcing team and the supplier's sales team. When they are, the requirement becomes a design parameter. When they are not, it becomes a line item. Van Weele and Van Tubergen's (2013) account of the Nespresso value chain makes the point: the shared-value model worked because the sustainability requirement was built into the sourcing contract, not bolted on after it.

10.3.3 For firms: position for the cluster, not just the firm

The cluster layer of the architecture rewards firms that cooperate. A furniture factory that shares a testing laboratory with its neighbours pays a fraction of the cost of a dedicated lab. A steel mill that connects to a shared heat recovery network captures waste heat that would otherwise be lost. An electronics firm that participates in a cluster-level recycled material sorting scheme secures feedstock that it could not source alone. The cooperation is not altruism—it is a cost-sharing arrangement that improves each firm's competitive position.

The implication for firm strategy is that the firm's location and its cluster relationships are part of its green governance capability. A firm that is isolated, even if it is well managed, faces higher measurement, compliance and infrastructure costs than a firm in a dense cluster. The firm that is considering relocation—the Longjiang furniture firm that moved sanding and assembly to Haiphong—should weigh the cluster's green infrastructure alongside the labour cost. The relocation that saves on wages but loses access to the cluster's testing lab and supplier finance scheme may be a net loss.

10.3.4 For policymakers: sequence the instruments to the firm's capability

The architecture is only coherent if the instruments arrive in an order that firms can absorb. A carbon price that arrives before the measurement infrastructure is in place penalises firms for not measuring. A CBAM obligation that arrives before the verification infrastructure is accredited penalises firms for not verifying. A subsidy that arrives before the eligibility criteria are clear rewards the firms that can navigate the ambiguity, which are the large firms.

The sequencing principle is that measurement comes first, incentives second, trade obligations third. CBAM's three-year reporting-only phase is an application of the principle: it gives firms time to build measurement before they face the financial obligation. The Chinese ETS's phased expansion is another: it brings firms into the market gradually, with free allocation that declines over time. The principle should be applied to the domestic instruments as well. The green factory certification, the energy-saving subsidy and the carbon market should be sequenced so that a firm can move from one to the next without rebuilding its data infrastructure at each step.

10.3.5 For policymakers: differentiate by industry and by region

The single most important design choice in the architecture is differentiation. A labour-intensive furniture cluster needs supplier finance and shared testing. A technology-intensive battery cluster needs grid access and chemical recycling. A capital-intensive steel cluster needs carbon capture and hydrogen infrastructure. A policy that treats them the same will be misallocated for all three.

The differentiation should be built into the instrument design, not added as an exception. The green factory certification could have sector-specific criteria, with the measurement requirements calibrated to the sector's emissions profile. The carbon market could have sector-specific benchmarks, with the free allocation calibrated to the sector's trade exposure. The cluster policy could have region-specific infrastructure priorities, with the investment calibrated to the cluster's industrial composition.

The regional dimension is equally important. The coastal clusters that are most exposed to CBAM have the most developed measurement infrastructure and the most access to finance. The inland clusters that are most exposed to the carbon market have the least. A policy that allocates support evenly across regions will under-serve the inland clusters, which need more support per unit of abatement. The allocation should be weighted by need, not by population or output.

10.3.6 For policymakers: build the just transition into the architecture

The transition has losers as well as winners. The coal miner in Shanxi, the steel worker in Hebei, the sanding operator in Longjiang—their jobs are affected by the green upgrading that the architecture is designed to promote. A policy architecture that ignores them will face resistance that undermines its own objectives.

The just transition component has three parts. The first is income support for workers in declining sectors, which is a social protection function that the existing system can absorb with modest expansion. The second is retraining and redeployment, which is the harder part: the Longjiang worker's skills are still needed, but not in Longjiang. The retraining programme needs to connect the worker to the design studio, the testing lab or the logistics function that the firm kept in Foshan. The connection is a labour market matching problem, and it requires the cluster governance institution to coordinate with the firm and the training provider. The third is regional investment, which is the hardest part: the inland cluster that loses its coal mine needs a new economic base, and the base is not created by a retraining programme alone.

The just transition is not a separate policy domain. It is a condition for the coherence of the architecture. A carbon price that is not accompanied by a just transition mechanism will be politically unsustainable in the regions that bear the cost. A CBAM that is not accompanied by support for the affected workers will be resisted by the firms and localities that depend on the exports. The architecture is coherent only if it is fair.

10.3.7 For policymakers: measure the architecture, not just the firms

The final implication is about evaluation. The architecture is a portfolio of instruments, and its performance cannot be assessed instrument by instrument. The relevant question is whether the portfolio as a whole is producing the capability and the climbing that it is meant to produce. That question requires portfolio-level indicators: the share of firms with verified emissions data, the share of exports with documented environmental attributes, the green premium earned by certified products, the carbon price differential between China and its trading partners, the employment transition rate in affected sectors.

The indicators should be published, and they should be used to adjust the portfolio. An architecture that is not evaluated is an architecture that drifts. The drift is not neutral: it favours the firms that can navigate the ambiguity, which are the large firms, and it disadvantages the firms that cannot, which are the small ones. The evaluation is the mechanism that keeps the architecture honest.

10.3.8 The closing argument: capability as the through-line

The book opened with a Foshan factory and a question about whether green governance is a cost or a capability. The intervening chapters have argued, from theory and from evidence, that it is both—that compliance is a cost and capability is an investment, and that the transition from one to the other is the central problem of green value chain upgrading. The policy architecture is the set of instruments that determines whether the transition happens, for which firms, and at what speed.

The architecture is a set of alignment conditions: measurement that feeds incentive, incentive that feeds finance, trade that reinforces domestic policy, cluster that supports the firm, and a just transition that sustains the political coalition. The conditions are demanding, and the current architecture meets some of them and not others. The measurement layer is being built, unevenly. The incentive layer is weak. The trade layer is contested. The cluster layer is under-resourced. The finance layer is concentrated. The just transition is nascent.

Each gap has a specific instrument and a specific deadline attached to it. The measurement gap closes with the accreditation of Chinese verifiers under the CBAM regulation and the conclusion of mutual recognition negotiations with the EU; the European Commission's implementing acts on verification are due before the financial obligation begins in January 2026. The incentive gap closes with the cap tightening already scheduled in the national ETS expansion to steel, cement and aluminium, which will raise the price from its current range toward a level that begins to register against retrofit costs. The trade gap closes with the deduction methodology the EU must publish before 2026, which determines whether a Chinese carbon payment is recognised euro for euro. The cluster gap closes with the place-based grants already piloted in Guangdong and Zhejiang, which need to be extended to the inland clusters before the carbon market expansion reaches them. The finance gap closes with the transition finance taxonomy the People's Bank of China has been drafting since 2022, which needs to recognise process upgrading in traditional industries as eligible. The just transition gap closes with the expansion of the social protection system to cover displaced workers in the covered sectors, which the 2024 State Council employment guidelines began to address.

The research agenda is the complement to the policy agenda. The questions in section 10.2 are the questions that the policy design cannot answer without evidence. How fast does capability diffuse? How large is the premium? Does the carbon market drive investment? How do firms respond to CBAM? What are the limits of the compliance-to-capability transition? The answers will not settle the policy debate, but they will discipline it. A policy architecture that is designed without evidence is a guess. A policy architecture that is designed with evidence is an experiment, and experiments can be corrected.

The Foshan plant manager who spent a month learning about grid emission factors is the test case. If the architecture gives her the measurement infrastructure, the incentive to invest, the finance to pay for it and the buyer who rewards it, she will climb. If the architecture gives her a reporting form and a deadline, she will comply, and the capability will not be built. The difference between the two outcomes is not the firm's ambition. It is the architecture's coherence. The first test of that coherence arrives on 1 January 2026, when the CBAM financial obligation begins and the reporting form becomes a bill.