Chapter 9: Institutions, Clusters, and Multi-Level Governance: Enabling Conditions for Green Capability Building

The Shunde sofa factory that closed Chapter 8 sits inside a dense institutional web that its managers rarely see and never draw. A single export container of sofas to Rotterdam passes through at least four regulatory regimes before it clears customs: the Guangdong provincial environmental inspection that licenses the foam line, the national GB standard that governs formaldehyde emissions from the fabric, the European Union's REACH chemical registry, and the carbon accounting that will, from 2026, attach a price to the embedded emissions in the steel springs and the polyurethane. None of these regimes was designed with the other in mind. The factory's green governance capability consists, in large part, of the managerial skill to hold all four in view at once and to translate their overlapping demands into a single production plan. That skill is not taught in the factory. It is acquired from the cluster.

This chapter turns from the firm to the enabling conditions that surround it. The argument runs against the grain of much policy writing on green upgrading. Capability is not produced by regulation alone, nor by clusters alone, nor by the elegant coordination of governance levels that appears in policy documents. It is produced at the intersection of three institutional layers that operate on different logics and different time horizons: formal rules that set the floor, informal norms that determine whether firms treat the floor as a ceiling, and the cluster-level mechanisms that transmit both. Where these three align, green capability accumulates. Where they diverge—where formal rules tighten while informal norms lag, or where clusters transmit competitive pressure without transmitting knowledge—firms comply and do not climb. The compliance trap described in Chapter 4 is, at bottom, an institutional misalignment.

The causal claim needs care. Institutions are enabling conditions, not sufficient causes. A well-designed emissions trading scheme does not manufacture a green product; it changes the relative price of doing so. A dense cluster does not guarantee knowledge diffusion; it lowers the cost of the conversations through which diffusion happens. The chapter therefore treats institutions as a production function for capability, with diminishing returns and complementarities that are easy to mismanage. The evidence is drawn from Chinese cluster experience, from comparative work on East Asian industrial districts, and from the multi-level governance literature that has grown up around the EU emissions trading system and its Chinese counterparts.

9.1 Formal and Informal Institutions: The Floor and the Ceiling

9.1.1 What Formal Rules Can and Cannot Do

Begin with a distinction that the institutional economics literature has made familiar but that green governance debates routinely blur. Formal institutions are the codified rules—laws, standards, permits, tax schedules, emissions caps—enforced by a third party with the authority to sanction. Informal institutions are the unwritten norms, conventions, and expectations that govern behavior when no one is watching and no sanction is available. Douglass North's (1990) formulation treats both as constraints on behavior, but the enforcement mechanisms differ in ways that matter enormously for green upgrading. A formal rule changes the payoff matrix at the margin; an informal norm changes what firms consider a reasonable thing to do.

China's formal green governance architecture has expanded at a pace that outstrips most comparable economies. The 2014 GB/T 50378 standard for green buildings, issued by the Ministry of Housing and Urban-Rural Development, evaluates energy, water, and material conservation across the whole building life cycle, and it is backed by a procurement regime that gives compliant projects preferential access to land and credit. The green credit balance reached RMB 11.95 trillion by the end of 2020, roughly 6.9 percent of total credit, the largest such book in the world at the time. Green bond issuance over 2015–2020 totaled RMB 1.2 trillion. On paper, the incentive structure for green investment is formidable.

Yet the same period produced a compliance pattern that should give pause to anyone who believes formal rules do the work by themselves. Provincial environmental inspections in Guangdong during 2017–2018 shuttered tens of thousands of small factories in the Pearl River Delta, many of them furniture and textile firms, for violations that had been tolerated for years. The rules had not changed. Enforcement had. The episode reveals something important about formal institutions in a decentralized authoritarian system: the rule on the books is a necessary but not sufficient condition for the rule in practice, and the gap between them is where firms learn whether to invest in capability or to wait out the campaign.

The Shunde factory's experience is instructive. Its managers had known for at least five years that their foam line would eventually have to meet a tighter volatile organic compound (VOC) standard. They had not invested, because the probability of enforcement in any given quarter seemed low and the capital cost of retrofitting was high relative to the margin on a sofa. When the inspection campaign arrived, they scrambled. The retrofit they eventually installed was the same one they could have installed five years earlier at a lower cost and with a longer payback period. The formal rule had been stable; the expectation of enforcement had not. Capability investment requires a credible horizon, and credible horizons are an institutional product.

Formal rules enable capability when they are predictable, when the enforcement trajectory is legible to firms, and when the sanction for non-compliance is proportional to the cost of compliance. They disable capability when they arrive as campaigns, because campaigns reward the appearance of compliance over the substance of it. A factory that installs a scrubber to survive an inspection and then turns it off when the inspectors leave has complied. It has not climbed.

9.1.2 The Compliance Trap as an Institutional Equilibrium

I introduced the compliance trap in Chapter 4 as a firm-level phenomenon: the condition in which a firm meets the letter of environmental regulation while its underlying production process remains unchanged, so that each new regulation extracts a fresh round of cost without generating a corresponding capability. It is worth restating the concept here in institutional terms, because the trap is better understood as an equilibrium than as a failure.

Consider two firms in the same cluster facing the same formal rule. Firm A invests in process redesign that reduces emissions and, as a byproduct, reduces energy and material costs. Firm B installs end-of-pipe abatement that reduces emissions and adds cost. Both are compliant. If the cluster's informal norms reward visible compliance—the scrubber on the roof, the certificate on the wall—and do not distinguish between A and B, then the marginal return to A's investment is lower than it should be, and over time B's strategy will dominate. The trap is sustained not by the formal rule but by the informal norm that treats all compliance as equivalent.

Breaking the trap requires an informal institution that distinguishes between compliance modes. In the German Mittelstand, the distinction is carried by the engineering culture of the trade associations and by the banks that finance process investment on the strength of a firm's technical reputation. In the Italian industrial districts, it is carried by the dense web of subcontracting relationships in which a firm's process choices are visible to its customers and its peers. In the Chinese cluster, the distinction is weaker. The trade associations are often vehicles for lobbying on standards rather than for diffusing process knowledge, and the banks lend against collateral rather than against technical reputation. The result is that Firm A and Firm B look alike from the outside, and the cluster's informal norms do not push firms toward A.

The policy implication is uncomfortable for those who favor formal instruments. Tightening the standard does not break the trap; it may deepen it, because it raises the cost of B's compliance without changing the relative return to A's investment. What breaks the trap is a change in the informal institution—a shift in what firms in the cluster consider a respectable way to comply. That shift is slow, it is local, and it is not amenable to central design. It is, however, amenable to facilitation, which is where clusters and multi-level governance enter the argument.

9.1.3 Informal Norms and the Social Construction of Green Worth

The informal institution that matters most for green upgrading is the cluster's shared understanding of what counts as a good firm. That understanding determines which investments are legible as status-enhancing and which are legible as cost. In clusters where green process investment is read as technical sophistication, firms compete to be early adopters. In clusters where it is read as regulatory burden, firms compete to be late adopters and to share the cost of delay through collective lobbying.

The sources of this shared understanding are varied. Trade associations matter, but so do the technical schools that supply the cluster's engineers, the local banks that finance its investment, and the lead firms that anchor its supply chains. In the Shunde cluster, the furniture association has historically functioned as a bargaining agent with the provincial environmental bureau rather than as a knowledge-diffusion body. Its annual meetings are occasions for exchanging information about enforcement timing, not about process technology. The informal norm that results is defensive: compliance is something to be managed, not something to be improved.

Contrast this with the Ningbo cluster in Zhejiang, where the garment and home-textile associations run joint training programs with the local polytechnic and where the lead export firms publish their supplier environmental requirements to the cluster. The norm there is competitive: a supplier that cannot meet the lead firm's standard is visibly behind. The difference is not in the formal rules, which are national, nor in the enforcement intensity, which is comparable. It is in the associational infrastructure that mediates between the two.

Institutional complementarity is the relevant concept here. A formal rule works better when it is embedded in an informal norm that reinforces it, and an informal norm is more durable when it is backed by a formal rule that gives it teeth. The two are complements, not substitutes. Chinese green governance has invested heavily in the formal component and comparatively little in the associational infrastructure that would make the informal component reinforce rather than undermine it. The result is a system that can produce compliance on demand but struggles to produce capability on a schedule.

9.1.4 The Temporal Mismatch

There is a temporal dimension to the formal-informal gap that deserves separate treatment, because it explains why green capability building is slower than the policy calendar assumes. Formal rules can change in a legislative session. Informal norms change over a generation. A firm that responds to a new formal rule by investing in capability is making a bet on the durability of the rule and on the eventual alignment of the informal norm. If the rule is reversed or the enforcement relaxed, the bet is lost. If the informal norm does not shift, the firm bears a cost its competitors do not.

The Chinese dual carbon framework, announced in September 2020 with a 2030 peak and a 2060 neutrality target, is a formal commitment with a thirty-year horizon. The firms that must invest against it operate on horizons of three to five years, which is the typical payback period for process investment in labor-intensive manufacturing. The mismatch is structural. It can be bridged by instruments that shorten the effective horizon of the formal commitment—carbon pricing with a rising trajectory, long-term offtake agreements for green products, preferential financing for verified process investment—but each of these instruments depends on the credibility of the underlying commitment, which in turn depends on the informal norm that the commitment will be honored across political cycles.

I do not want to overstate the fragility. The dual carbon commitment has been embedded in the 14th Five-Year Plan, in provincial targets, and in the lending guidelines of the major state banks. Its credibility is higher than that of a single administrative campaign. But the gap between the thirty-year formal horizon and the three-year firm horizon is real, and it is the gap in which the compliance trap lives. Closing it is a governance problem, not a technology problem.

9.1.5 Vignette: The Foam Line and the Two Standards

Return to the Shunde factory's foam line, because the details illustrate the interaction of formal and informal institutions more concretely than any abstraction. The line produces polyurethane foam for sofa cushions. It emits VOCs, principally from the blowing agents and the curing process. Two formal standards apply: a national emission standard that sets a concentration limit at the stack, and a provincial standard that sets a limit on the total mass emitted per unit of output. The national standard is easy to meet with end-of-pipe abatement. The provincial standard is harder, because it constrains the process rather than the exhaust.

The factory's managers understood the national standard well. They had installed a carbon adsorber in 2016 and passed every inspection since. They understood the provincial standard poorly. It had been on the books since 2018 but had been enforced only against the largest plants in the cluster. When the 2021 inspection campaign extended it to mid-sized plants, the managers discovered that their adsorber did not address the process emissions that the mass-based standard captured. The retrofit they needed was a change in the blowing agent, which required reformulating the foam and re-qualifying it with their European customers—a twelve-month process.

What is striking about the episode is not the technical difficulty. It is that the factory's informal network had given it no warning. The trade association had circulated the text of the provincial standard but not its implications. The equipment suppliers had continued to sell carbon adsorbers because that was what their customers asked for. The lead customers had not asked about process emissions because their own procurement standards were keyed to the national standard. Every node in the informal network was transmitting the old equilibrium. The formal rule had changed; the network had not.

The factory eventually reformulated the foam, and the reformulation turned out to reduce material costs by about 4 percent, which paid back the qualification cost in eighteen months. The capability it built—the ability to read a mass-based standard, to trace it to the process, and to reformulate—is now a selling point with its European customers. But it built the capability two years later than it could have, and it built it under duress. A cluster whose informal network had transmitted the implications of the provincial standard would have built it earlier and at lower cost.

9.2 Cluster-Based Green Upgrading: How Proximity Becomes Capability

9.2.1 The Cluster as a Knowledge Institution

The cluster literature has long argued that geographic proximity produces economic benefits that are not reducible to the sum of firm-level advantages. Alfred Marshall's (1890) original account emphasized labor market pooling, input sharing, and knowledge spillovers. The third of these is the one that matters for green upgrading, and it is the one that is least well understood.

Knowledge spillovers in clusters are not automatic. They flow through specific channels: the movement of engineers between firms, the shared use of suppliers and testing facilities, the informal exchange of information at trade association meetings, and the observation of competitors' visible choices. Each channel has a cost of use, and the cost is lower when the cluster has institutions that reduce it—a technical school that trains engineers in the relevant process, a testing lab that is affordable for small firms, an association that convenes the right people. The cluster's contribution to green capability is the density of these channels, not the mere fact of co-location.

The distinction matters for policy. A cluster that has co-location but no channels will not diffuse green knowledge, no matter how many green firms it contains. A cluster that has channels but no green firms will diffuse the wrong knowledge. The Chinese experience includes both failure modes. Some of the export-processing zones in the Pearl River Delta are dense with firms but thin on associational infrastructure; green knowledge enters through the lead firms' supplier requirements and does not diffuse laterally. Some of the inland clusters have active associations but few firms with the technical depth to generate knowledge worth diffusing.

The clusters that have upgraded most successfully combine density with channels and a critical mass of technically capable firms. The Ningbo garment cluster is one. The Foshan ceramics cluster is another, though its path has been more contested. The Shunde furniture cluster is a third, and its trajectory over the past decade illustrates the mechanisms at work.

9.2.2 Three Mechanisms of Cluster Green Upgrading

Three mechanisms convert proximity into green capability. They operate at different scales and with different policy implications, and conflating them is a common source of policy error.

The first is supplier upgrading through lead-firm requirements. When a lead firm in the cluster—typically an exporter or a brand-owner—sets an environmental requirement for its suppliers, it transmits a standard through the supply chain faster than any regulator can. The lead firm has the commercial leverage to enforce the requirement, the technical staff to specify it, and the commercial interest in avoiding a reputational or regulatory shock. In the Shunde cluster, the arrival of a small number of European buyers with REACH compliance requirements in the mid-2010s did more to diffuse chemical management knowledge than a decade of provincial inspection. The suppliers that met the requirement kept the business; the suppliers that did not lost it. The knowledge that diffused was specific and commercially validated, which is precisely the kind that firms retain.

The mechanism has limits. It diffuses the requirements of the lead firm, not the frontier of green technology. A supplier that meets REACH has met REACH; it has not necessarily built the capability to reduce its carbon footprint, which is the next requirement on the horizon. Lead-firm upgrading also concentrates capability in the first tier of suppliers and can leave the second and third tiers untouched, which is a problem when the carbon border adjustment begins to reach embedded emissions several tiers down.

Collective service provision is the second mechanism. Clusters can support shared facilities—testing labs, training centers, common effluent treatment plants—that individual firms cannot afford. The classic case is the shared wastewater treatment plant in the textile dyeing clusters of Zhejiang, which allows small firms to meet effluent standards that would be uneconomic to meet individually. The green upgrading literature has documented similar arrangements for electroplating, ceramics, and furniture finishing. The shared facility reduces the fixed cost of compliance and, more importantly, creates a node at which process knowledge concentrates. The technicians who run the shared plant become a cluster-level resource, and the firms that use the plant learn from them.

Collective service provision is fragile in ways that are worth naming. It requires a governance structure that can allocate costs and resolve disputes among firms with different needs and different willingness to pay. It requires a minimum scale of demand, which means it works in dense clusters and fails in thin ones. And it requires a local government willing to subsidize the fixed cost, which means it depends on the fiscal capacity and the political priorities of the locality. The Zhejiang plants work because the provincial government made them a priority and because the clusters are dense enough to support them. The model does not transfer automatically to the inland clusters that are now receiving relocated furniture capacity.

The third mechanism is labor mobility and the circulation of engineers. The most underrated channel of green knowledge diffusion in Chinese clusters is the movement of technical staff between firms. An engineer who has commissioned a VOC abatement system at one factory carries that knowledge to the next employer. A quality manager who has learned to document chemical compliance for a European buyer carries the documentation system. The cluster's labor market is a knowledge market, and its efficiency depends on the thickness of the technical labor pool and the willingness of firms to hire from competitors.

The Shunde cluster's labor market is thick at the level of production workers and thin at the level of process engineers. The factories compete for the same small pool of engineers who understand foam chemistry and emission control, and they have responded by hoarding rather than sharing. The result is that the knowledge circulates slowly and the cluster as a whole lags the frontier. The Ningbo cluster, by contrast, has a polytechnic that produces a steady flow of process engineers, and the flow has lowered the cost of hiring and increased the rate of circulation. The difference is an institutional one: the presence of a training institution that produces the relevant skill at scale.

9.2.3 The Cluster Life Cycle and the Green Window

Clusters are not static. They move through a life cycle that runs from emergence through growth and maturity to decline or reinvention, and the phase of the cycle shapes the opportunities for green upgrading. The green window—the period during which a cluster can leap to a new process trajectory rather than retrofit an old one—opens at particular moments and closes at others.

In the emergence phase, the cluster's process choices are still fluid, and a green trajectory can be locked in at low cost. The Chinese photovoltaic cluster is the clearest example: it was built green from the start because the technology it produces is green, and the cluster's process standards were set by export requirements from the beginning. In the growth phase, the cluster is investing heavily and the marginal cost of incorporating green process is low relative to the total investment. In maturity, the capital stock is sunk, the process is standardized, and green upgrading becomes a retrofit problem with a higher cost and a longer payback. In decline, the cluster lacks the resources to invest and the green window closes entirely.

The Shunde furniture cluster is in late maturity. Its capital stock is sunk, its process is standardized, and its margins are thin. The green window is narrow. The cluster can retrofit, and it is retrofitting, but it cannot leap. The clusters that are now receiving relocated furniture capacity—in Jiangxi, in Guangxi, in inland Guangdong—are in the growth phase, and their green windows are wider. Whether they use the window depends on the institutional infrastructure they inherit or build. A relocated cluster that arrives with the Shunde model of weak associational infrastructure will reproduce the Shunde trajectory, including its late and costly retrofit. A relocated cluster that arrives with a shared treatment plant, a technical school, and an active association will lock in a greener trajectory at lower cost.

Cluster policy is therefore green policy, whether or not it is labeled as such. The decisions that determine a cluster's green trajectory are made at the moment of formation, and they are made by the local government, the lead firms, and the associations that set the initial conditions. The industrial relocation policy that has driven the inland furniture clusters is, in effect, a green policy with a long lag. It deserves to be evaluated as such.

9.2.4 Vignette: The Shared Finishing Line

A concrete case from the Shunde cluster illustrates both the promise and the limits of collective service provision. In 2019, a group of eleven mid-sized furniture firms in the Lecong sub-cluster pooled capital to build a shared finishing line—a computer-controlled spray booth with solvent recovery and a common quality lab. The line was expensive, roughly RMB 18 million, and no single firm could justify it. The eleven firms each held a share, and each committed to route a minimum volume through the line.

The economics worked. The line reduced solvent consumption per unit by about 30 percent, recovered solvent worth roughly RMB 2 million a year, and cut VOC emissions by enough to meet the provincial mass-based standard without end-of-pipe abatement. The quality lab allowed the firms to certify their finishes to European standards, which opened a customer segment that had been closed to them. The payback was four years on the capital and immediate on the operating cost.

The governance was harder than the economics. The eleven firms had different product mixes and different quality requirements, and the line had to be scheduled across them. Two firms dropped out in the first year, one because it was acquired by a competitor and one because its volumes fell. The remaining nine had to absorb the fixed cost of the departed shares. The dispute resolution mechanism—a monthly meeting of the shareholder firms—was informal and worked because the firms had known each other for two decades. It would not have worked among strangers.

The case is instructive for what it shows about the conditions under which collective service provision succeeds. It requires a minimum scale of demand, which the eleven firms provided. It requires a governance structure that can allocate costs and resolve disputes, which the firms' prior relationships provided. It requires a local government willing to facilitate the land and the permits, which the Lecong town government provided because the line supported its own air quality targets. And it requires a technical champion—in this case, the retired chief engineer of one of the firms—who could specify the equipment and manage the commissioning. Remove any of these and the line does not happen.

The case also shows the limits. The line serves eleven firms out of several hundred in the sub-cluster. The firms that are not shareholders continue to use conventional booths and continue to emit. The shared line has raised the frontier for the firms that use it without raising the floor for the cluster as a whole. Scaling the model would require either a much larger consortium or a regulatory requirement that forces non-participants to use the line, and neither is imminent. The cluster's green upgrading is proceeding at two speeds, and the gap between them is widening.

9.2.5 When Clusters Fail: The Lock-In of Dirty Specialization

Clusters can also lock in dirty specialization. The mechanism is well documented in the economic geography literature on path dependence, and it operates through the same channels that diffuse green knowledge, running in reverse. A cluster that specializes in a dirty process attracts firms that are good at the dirty process, trains workers in the dirty process, and builds supplier networks around the dirty process. The agglomeration economies reinforce the specialization, and the cluster becomes resistant to the clean alternative even when the clean alternative is technically superior.

The Chinese ceramics cluster in Foshan is a partial example. The cluster's kilns are fired on natural gas, which is cleaner than the coal they replaced but still carbon-intensive. The cluster has the density and the supplier network to convert to electric kilns, and the technology exists. The conversion has been slow because the cluster's workforce is trained on gas kilns, its suppliers sell gas burners, and its firms have sunk capital in gas infrastructure. The agglomeration economies that made Foshan a ceramics center now work against its decarbonization.

The lock-in is not permanent, but it is costly to break. Breaking it requires an external shock—a carbon price high enough to change the economics, a regulation that forces conversion, or a technological breakthrough that makes the clean alternative dramatically cheaper. The cluster's own institutions will not break it, because the cluster's institutions are organized around the incumbent process. This is the strongest argument for multi-level governance: the levels above the cluster are the only ones with the perspective and the leverage to break a local lock-in.

9.3 Multi-Level Governance and Policy Integration: Coordinating the Levels

9.3.1 The Multi-Level Problem

The governance of green upgrading involves at least four levels, each with its own authority, its own information, and its own incentives. The international level sets the carbon accounting rules, the border adjustment regimes, and the trade disciplines that shape what firms must disclose and what they will pay. The national level sets the emissions targets, the carbon market, the industrial policy, and the standards. The provincial level allocates the targets, enforces the standards, and controls the land and credit that firms need to invest. The local level operates the inspection regime, the shared infrastructure, and the associational infrastructure that determines whether knowledge diffuses.

The multi-level governance literature, following Hooghe and Marks (2001, 2003) and the earlier work of Keohane (1984) on international institutions, distinguishes between two types of coordination: a nested hierarchy in which the levels are ordered and the higher level constrains the lower, and a network in which the levels interact without a clear hierarchy. Chinese green governance is a hybrid. The national level sets binding targets that are decomposed to the provinces, which is a nested hierarchy. But the provinces have substantial discretion in how they meet the targets, and the localities have substantial discretion in enforcement, which is a network. The hybrid produces both the system's capacity to mobilize and its tendency to produce compliance without capability.

The coordination failures are predictable and they are documented. A national carbon market that prices emissions at a level too low to change investment decisions transmits a weak signal to firms that are making investment decisions on a three-to-five-year horizon. A provincial target that is met by shuttering small firms rather than by upgrading them produces compliance and destroys capability. A local inspection regime that is tightened during campaigns and relaxed between them produces the temporal mismatch described in Section 9.1. Each failure is a coordination problem, and each has a governance solution that is not a matter of more central control.

9.3.2 The Carbon Market as a Coordination Device

The national emissions trading system, launched in 2021 and initially covering the power sector, is the most important coordination device in the Chinese green governance architecture. Its design determines the price signal that reaches firms, and the price signal determines whether green investment is economic. The early evidence is not encouraging on this dimension. The carbon price in the national market has traded in a range that is low by international standards, well below the level at which most process investments in manufacturing become economic. The low price reflects a generous allocation of free permits and a compliance regime that allows firms to meet their obligations with offsets and with banked allowances.

The consequence for green capability is direct. A furniture firm considering a process investment that would reduce its embedded carbon by 20 percent faces a carbon price that values the reduction at a small fraction of the investment cost. The investment is not economic on the carbon price alone; it becomes economic only if the firm can also capture a price premium from customers, a financing advantage from green credit, or a regulatory advantage from being an early mover. The carbon market, in its current design, is not doing the coordination work that the dual carbon target requires.

This is not a counsel of despair. Carbon market design is a policy variable, and the trajectory matters more than the level. A market with a rising cap and a credible long-term trajectory can transmit a signal that is weak today and strong tomorrow, and firms that understand the trajectory can invest against it. The EU emissions trading system went through a similar evolution: its first phase (2005–2007) crashed the price to near zero, its second phase (2008–2012) was undermined by the financial crisis, and its third phase (2013–2020) introduced the market stability reserve that has since produced a price in the range that changes investment decisions. The Chinese market is in its first phase. The question is whether its trajectory will be credible enough, soon enough, to reach the furniture firms before their green window closes.

9.3.3 Policy Integration: The Whole-of-Government Problem

Policy integration—the coordination of climate policy with industrial, trade, fiscal, and social policy—is the hardest part of multi-level governance and the part most often declared rather than practiced. The Chinese institutional architecture has a coordination body, the National Development and Reform Commission, and a set of sectoral ministries with their own mandates. The dual carbon target is a whole-of-government commitment, but the instruments that deliver it are sectoral, and the sectoral ministries have their own priorities.

The furniture case illustrates the integration problem. The Ministry of Industry and Information Technology is responsible for the industry's upgrading and has a mandate to promote advanced manufacturing. The Ministry of Ecology and Environment is responsible for emissions and has a mandate to enforce standards. The Ministry of Commerce is responsible for exports and has a mandate to defend market access. The Ministry of Finance is responsible for the fiscal cost of the transition. Each ministry has a legitimate interest in the furniture industry, and the interests are not always aligned. A policy that tightens emissions standards may raise the industry's costs and reduce its export competitiveness, which the commerce ministry will resist. A policy that subsidizes relocation to inland provinces may reduce the industry's exposure to the carbon border adjustment, which the environment ministry will support and the industry ministry will question.

The integration problem is not unique to China. The EU has struggled with the same coordination issues in its own Green Deal, and the tension between the emissions trading system and the carbon border adjustment mechanism is a live policy debate. What is distinctive about the Chinese case is the scale of the coordination required and the speed at which it must happen. A continent-sized economy with a continent-sized manufacturing sector is trying to integrate climate policy across a dozen ministries and thirty-one provincial jurisdictions in a decade. The coordination failures are not surprising; the surprising thing is that the system coordinates as well as it does.

9.3.4 The Local State as the Frontline

The local state—the county and township governments that operate the inspection regime, allocate the land, and convene the associations—is the frontline of green governance. It is also the level with the weakest capacity and the most ambiguous incentives. The local government wants economic growth, tax revenue, and employment, and it wants to avoid being sanctioned by the province for environmental failures. The two sets of wants are in tension, and the resolution of the tension varies across localities and over time.

The variation is visible in the inspection data. Some counties enforce the provincial standards consistently and have done so for years. Others enforce them only during campaigns. The difference is not explained by the formal rules, which are the same, or by the fiscal capacity, which is comparable. It is explained by the local leadership's priorities and by the associational infrastructure that mediates between the local state and the firms. A county with an active trade association that has internalized the environmental standard will enforce it more consistently than a county where the association is a lobbying vehicle, because the association does part of the enforcement work.

The local state's capacity is therefore an enabling condition for green capability. The local state does not build the capability; it creates the conditions under which firms build it. It provides the credible horizon, the shared infrastructure, and the convening power that the cluster needs. Where the local state is capable and aligned, the cluster upgrades. Where it is not, the cluster complies.

The Shunde case shows the local state's role in its most constructive form. The Lecong town government facilitated the shared finishing line by providing the land, expediting the permits, and brokering the initial conversations among the eleven firms. It did not finance the line, and it did not manage it. It created the conditions, and the firms did the rest. The model is not scalable by decree—it depends on a particular local leadership and a particular set of firm relationships—but it is replicable in the sense that other localities can create similar conditions if they have the capacity and the will.

9.3.5 The International Level: Carbon Border Adjustment and the Governance of Embedded Emissions

The international level enters the furniture story through the EU's carbon border adjustment mechanism (CBAM), which will apply to imports of carbon-intensive goods and, in its eventual scope, to the embedded emissions in complex products. The mechanism is a governance instrument with a specific logic: it prices the carbon embedded in imports at the border, so that the EU's own climate policy is not undermined by imports from jurisdictions with weaker carbon prices. Its effect on Chinese furniture exporters is indirect but significant. The steel in the sofa springs, the aluminum in the frames, and the chemicals in the foam all carry embedded emissions, and the mechanism will eventually require exporters to account for them.

The governance challenge is that the accounting rules are set at the international level, the production processes are at the firm level, and the data that connects them is at the supply-chain level. A furniture exporter that wants to comply with CBAM must be able to trace the embedded emissions in its inputs back to the suppliers of those inputs, and the suppliers must be able to document their emissions using a methodology that the EU recognizes. The chain of documentation runs through several tiers of suppliers, most of which are small firms with no carbon accounting capacity.

This is where the cluster and the multi-level governance architecture intersect most consequentially. A cluster that has a shared carbon accounting facility, or an association that provides a common methodology, can document embedded emissions at a cost that small firms can bear. A cluster that does not have these institutions will see its exporters excluded from the European market, not because their products are dirty but because they cannot prove that they are clean. The compliance cost is a documentation cost, and documentation is a collective good.

The Chinese response has been to build the infrastructure at the national level. The national carbon market, the provincial carbon accounting platforms, and the green finance taxonomy are all attempts to create the data infrastructure that CBAM will require. The gap is at the cluster level, where the data is generated and where the capacity to generate it is thinnest. The multi-level governance problem is to connect the national infrastructure to the cluster-level capacity, and the connection is not automatic.

9.3.6 Vignette: The Documentation Chain

A Shunde exporter's experience with CBAM documentation illustrates the coordination problem in concrete terms. The firm exports sofas to Germany and was asked by its buyer in 2023 to provide an estimate of the embedded emissions in its products. The firm could calculate its own direct emissions from its gas and electricity consumption. It could not calculate the embedded emissions in its steel springs, because its spring supplier—a small firm in Foshan—had no carbon accounting. It could not calculate the emissions in its foam, because its chemical supplier—a larger firm with a sustainability report—used a methodology that the EU did not recognize.

The firm spent four months and roughly RMB 300,000 assembling an estimate. It hired a consultant, who traced the spring supplier's electricity consumption from its utility bills and applied a grid emission factor. It persuaded the chemical supplier to share its process data, which the supplier had never disclosed to a customer. It submitted the estimate to the buyer, who forwarded it to the EU importer, who accepted it provisionally. The estimate is rough, and it will not survive a formal audit.

The episode reveals the governance gap. The firm had the commercial incentive to document, the resources to hire a consultant, and a buyer who demanded the documentation. Most of its competitors in the cluster had none of these. The documentation chain stopped at the first tier of suppliers, and the emissions in the second and third tiers were invisible. The national carbon accounting infrastructure did not reach the firm, and the cluster had no shared facility to fill the gap. The firm's capability to document is now a competitive advantage, but it is an advantage that depends on a consultant and a cooperative supplier, neither of which is a durable institutional arrangement.

9.3.7 Designing Coordination: Principles Rather Than Blueprints

The coordination failure is easiest to see in a single episode. In 2022 a mid-sized Foshan furniture firm received three signals in the same quarter. The provincial environmental bureau extended the mass-based VOC standard to mid-sized plants. The provincial development and reform commission published a subsidy schedule for firms that relocated capacity to an inland industrial park. The local bank, following national green credit guidelines, offered preferential financing for verified process investment. The firm's managers read the three signals and did nothing for eight months, because the relocation subsidy made the retrofit uneconomic, the green credit required collateral the firm did not have, and the VOC standard's enforcement date was not published. Each instrument was defensible on its own. Together they cancelled out.

That episode is the raw material for the design principles that Hooghe and Marks (2003) and, in the climate context, the OECD's (2021) work on multi-level governance have tried to formalize. The first principle is subsidiarity with a floor: decisions should be made at thelowest level that has the capacity and the information to make them, subject to a floor set by the higher level. The floor prevents a race to the bottom; local discretion allows adaptation to local conditions. The Chinese system already operates on this principle in some domains—the provincial allocation of national targets is a form of subsidiarity with a floor—but the floor is set by administrative target rather than by a price, which weakens the signal. In the Foshan episode, the floor existed (the VOC standard) but the local level had no incentive to enforce it on the schedule the province intended, and the firm had no way to know when enforcement would arrive.

Policy coherence across instruments is the second principle, and the Foshan episode is a textbook violation of it. The carbon price, the standards, the subsidies, and the trade measures should point in the same direction, so that a firm that responds to one is not penalized by another. In the Chinese furniture case the coherence is imperfect: the carbon price is low, the standards are tightening, the green credit is available but collateral-constrained, and the relocation subsidies reward moving rather than upgrading. A firm that reads all four signals at once receives a mixed message, and the rational response to a mixed message is to wait. The firm waited eight months, and the wait cost it the early-mover premium it could have captured from its German buyer.

Information symmetry across levels is the third principle. The national level cannot design effective instruments without cluster-level data, and the cluster cannot comply with national instruments without national-level methodology. The documentation chain described in 9.3.6 is a symptom of asymmetry: the data exists at the firm level, the methodology exists at the international level, and the connection between them is missing. The OECD's 2021 report on climate governance across levels of government makes the same point in general terms—that the weakest link in multi-level climate policy is usually the data interface between the level that generates information and the level that sets rules—and the Chinese furniture case confirms it at the scale of a single export order.

Credible commitment over time is the fourth principle, and it is the one the Foshan firm's managers named explicitly when asked why they had not invested earlier. A multi-level system coordinates effectively only when the levels can make commitments that outlast the political cycle at any one level. The dual carbon target is a commitment of this kind at the national level. The provincial targets are commitments of a shorter kind. Local enforcement is a commitment of a shorter kind still. The hierarchy of credibility is inverted relative to the hierarchy of authority: the level with the most authority has the most credible commitment, and the level with the least authority has the least. Firms respond to the least credible level, because that is the level they deal with. Closing the credibility gap is the central coordination problem, and it is not solved by announcing more targets. It is solved by publishing enforcement schedules and holding to them, by legislating a carbon cap trajectory rather than setting it annually, and by giving the local state incentives that align with the province's timetable rather than with the campaign calendar.

These four principles are not a blueprint. They are the conditions that the Foshan episode shows to be missing, stated in the abstract language that the comparative literature uses. The value of stating them is diagnostic: a locality or a province can ask which of the four is absent and design the remedy accordingly. The value of the Foshan episode is that it shows what the absence looks like from the firm's side—eight months of waiting, a missed premium, and a retrofit that will now be done under duress rather than on schedule.

9.3.8 The Role of Intermediary Organizations

Between the firm and the state, and between the cluster and the international regime, sits a layer of intermediary organizations that the governance literature has only recently begun to theorize. Trade associations, technical standards bodies, testing and certification firms, industry foundations, and public-private platforms all perform functions that neither the state nor the firm can perform alone. They translate between the languages of regulation and production. They aggregate the interests of small firms that cannot represent themselves. They carry knowledge across cluster boundaries.

The Chinese intermediary layer is uneven. The national-level associations are well resourced and well connected to the ministries, but they represent the largest firms and their interests are not always the interests of the cluster. The provincial associations are closer to the firms but thinner in capacity. The local associations are closest of all but often function as social clubs or lobbying vehicles rather than as technical service providers. The testing and certification firms are growing rapidly, driven by the demand for compliance documentation, but their services are expensive for small firms and their methodologies are not always aligned with international requirements.

The intermediation gap is most acute precisely where the green upgrading challenge is most acute: among the small and medium firms that make up the bulk of the labor-intensive clusters. These firms need translation, aggregation, and knowledge carriage, and the organizations that could provide these services are weakest at the local level. The policy response has been to build platforms—industry innovation centers, cluster service stations, public testing labs—and the platforms have helped where they have been staffed with people who understand the firms' problems. They have not helped where they have been staffed with administrators who understand the reporting requirements.

Intermediation is a skill, not a structure. A platform with the right mandate and the wrong staff will not intermediate. A cluster with a skilled intermediary—a retired engineer, a university extension agent, a trade association secretary who knows the firms—will intermediate even without a formal platform. The institutional design question is how to reproduce the skilled intermediary at scale, and the answer is not obvious. It involves training, career paths, and the recognition of intermediation as a profession, none of which is currently on the policy agenda.

9.4 Institutional Complementarity and the Limits of Transfer

9.4.1 Why the German and Italian Models Do Not Travel Intact

The comparative literature on cluster green upgrading returns repeatedly to the German and Italian cases, and the Chinese policy conversation has drawn heavily on them. The German Mittelstand combines firm-level technical depth with a dense associational infrastructure, a dual training system, and a banking sector that lends against technical reputation. The Italian industrial districts combine small-firm specialization with cooperative institutions, shared service centers, and a culture of inter-firm collaboration that goes back a century. Both systems produce green capability as a byproduct of their general institutional endowment, not as the result of a targeted green policy.

The temptation is to import the visible institutions—the training system, the shared service centers, the associations—and to expect the outcomes to follow. They do not, because the visible institutions are embedded in a set of complementary arrangements that are not visible and not transferable. The German dual training system works because German firms are willing to bear the cost of training apprentices whom they may lose to competitors, and they are willing because the chamber system and the labor market institutions make poaching costly. The Italian cooperative institutions work because Italian small firms have a history of collaboration that predates the institutions and sustains them. Remove the complements and the institutions become shells.

The Chinese experience with imported cluster institutions is instructive. The shared service centers that were built in the 2000s in several coastal clusters have had mixed results. Where they were staffed with engineers and connected to the firms' actual problems, they became nodes of knowledge diffusion. Where they were built as physical facilities with administrative staff, they became real estate. The difference was not in the design of the center but in the institutional complementarity: the centers that worked were embedded in a cluster that had the associational density to use them, and the centers that failed were built in clusters that did not.

This is not an argument against policy borrowing. It is an argument for borrowing with an understanding of the complements. A locality that wants to build a shared testing lab should first ask whether it has the associational density to generate demand for the lab, the technical labor pool to staff it, and the local state capacity to convene the firms. If it does not, the lab will be a building. If it does, the lab will be an institution.

9.4.2 The Chinese Institutional Endowment: What Is Distinctive

The Chinese institutional endowment has distinctive features that shape the green upgrading path, and they are not all disadvantages. The first is the capacity for rapid infrastructure provision. The Chinese state can build a shared treatment plant, a testing lab, or a carbon accounting platform faster than most comparable states, because it controls the land, the credit, and the permitting. The speed is an advantage when the infrastructure is the binding constraint, and it is a disadvantage when the binding constraint is the associational density to use the infrastructure. The Chinese system is good at building the hardware and weaker at building the software.

The second is the experimentalist policy process. The Chinese system runs pilots—emissions trading pilots in seven provinces, green finance pilot zones, low-carbon industrial park pilots—and scales the ones that work. The experimentalist process is well suited to the green upgrading problem, because the problem is heterogeneous across clusters and the solutions are not known in advance. The pilots generate the local knowledge that the national policy needs, and the scaling mechanism transmits it. The weakness of the process is that the pilots are often evaluated on their formal outputs rather than on their capability outcomes, and the scaling can reproduce the form of a successful pilot without its substance.

The third is the integration of the local state into the cluster. The Chinese local state is not an external regulator; it is an actor within the cluster, with ownership stakes, land interests, and personnel ties. The integration can be a source of coordination and a source of capture. It coordinates when the local state's interests align with the cluster's upgrading, as in the Lecong shared finishing line. It captures when the local state's interests align with the incumbent firms' resistance to upgrading, as in the clusters where inspection is relaxed for the largest taxpayers. The balance depends on the local leadership and on the higher-level incentives that shape it.

The fourth is the scale of the domestic market. Chinese firms can upgrade for the domestic market as well as for export, and the domestic market is large enough to support green product segments that would be too small in a smaller economy. The green building standard, the new energy vehicle mandate, and the dual carbon procurement rules all create domestic demand for green products. The domestic demand is a buffer against the volatility of export requirements, and it is a source of the scale economies that make green investment economic. The buffer is not unlimited—the domestic green premium is thinner than the European one—but it is a distinctive advantage.

9.4.3 The Complementarity Between Carbon Pricing and Industrial Policy

The most important institutional complementarity for green upgrading is the one between carbon pricing and industrial policy. Carbon pricing changes the relative price of emissions; industrial policy changes the capacity to respond to the price. Neither works well alone. A carbon price without the capacity to respond produces stranded assets and relocation. Industrial policy without a carbon price produces green investment that is not economically sustainable and a fiscal burden that grows without bound.

The Chinese system has invested heavily in industrial policy and lightly in carbon pricing, and the imbalance has consequences. The green credit, the subsidies, and the pilot zones have built capacity in the firms that received them. The carbon price has been too low to reward the capacity once built. The result is a system that can produce green products but cannot yet sell them at a green premium that reflects their carbon content. The firms that have upgraded are waiting for the price signal that would make their investment pay.

The complementarity runs in the other direction as well. A carbon price without the industrial policy capacity to respond would produce the compliance trap at a higher level: firms would pay the price, pass it through to customers, and change nothing. The industrial policy is what allows firms to convert the price signal into process change. The two instruments are complements, and the policy challenge is to sequence them so that the capacity is in place when the price arrives.

The sequencing question is live in the furniture case. The carbon price that CBAM will eventually transmit through the supply chain is a price on embedded emissions, and the firms that can document and reduce their embedded emissions will capture the premium. The firms that cannot will lose the market. The industrial policy that builds the documentation and reduction capacity is therefore a precondition for the carbon price to work as an upgrading instrument rather than as an exclusion instrument. The current policy trajectory builds the capacity slowly and raises the price slowly, and the two trajectories may or may not intersect before the green window closes.

9.4.4 Institutional Learning and the Diffusion of Governance Innovations

Institutions learn, and the learning is itself a governance process. The Chinese green governance system has learned from its own pilots and from international experience, and the learning has changed the instruments. The emissions trading pilots informed the design of the national market. The green finance pilot zones informed the design of the national taxonomy. The low-carbon park pilots informed the design of the cluster service platforms. The learning is real, and it is faster than the learning in most comparable systems.

The learning has limits. It is better at learning about instruments than about the institutional complements that make instruments work. The pilots that succeeded are studied for their formal design, and the associational and local-state conditions that made them succeed are noted but not reproduced. The scaling therefore reproduces the instrument without the complement, and the scaled instrument underperforms. This is the mechanism behind the mixed record of the shared service centers described above, and it is a general pattern in Chinese policy diffusion.

The remedy is to treat the complements as part of the innovation. A pilot that succeeds because of a particular local leadership, a particular association, or a particular set of firm relationships should be documented with those conditions, and the scaling should include the conditions or acknowledge their absence. The documentation is a governance task, and it is not currently performed systematically. The result is that the system learns less than it could from its own successes.

A concrete illustration comes from the low-carbon industrial park pilots that the Ministry of Industry and Information Technology ran between 2015 and 2020. The pilots that reported the largest emissions reductions were concentrated in parks where the park management company also operated the shared utilities and therefore had both the authority and the commercial incentive to invest in efficiency. The parks that reported smaller reductions had the same formal pilot design—the same reporting requirements, the same subsidy schedule, the same technical guidance—but the management company was a landlord rather than a utility operator, and it had no stake in the energy performance of its tenants. When the ministry scaled the pilot to a second tranche of parks in 2021, the scaling documents reproduced the reporting template and the subsidy schedule and said nothing about the ownership structure of the park utility. The second tranche has performed unevenly, and the unevenness tracks the ownership variable that the scaling documents did not mention. The lesson is not that the ministry made an error of judgment. It is that the documentation of the pilot captured the instrument and not the condition, and the condition was the thing that made the instrument work.

The same pattern appears in the green finance pilot zones. The zones that succeeded in directing credit to verified process investment were the ones where the local branch of the central bank had built a data-sharing arrangement with the environmental bureau, so that a firm's environmental compliance record was visible to its lender. The zones that did not build the arrangement had the same green bond guidelines and the same preferential reserve ratios, and the credit did not move. The data-sharing arrangement was an institutional complement, and it was not part of the formal pilot design. The scaling of the green finance model to the national taxonomy reproduced the guidelines and did not reproduce the data-sharing arrangement, because the arrangement was a local administrative practice rather than a policy instrument. Whether the national taxonomy will produce the same credit redirection without the data-sharing complement is an open question, and the early evidence is that it will not.

These two cases point to a general weakness in how the Chinese system learns. The learning apparatus is organized around instruments, because instruments are what the ministries control and what the evaluation metrics measure. The complements are organized around local relationships, because relationships are what the localities control and what the evaluation metrics do not measure. The result is a learning system that is excellent at codifying the visible and poor at codifying the invisible. The remedy is not a new evaluation metric. It is a documentation practice—case studies that record the conditions alongside the instrument, and scaling guidelines that name the conditions that must be present for the instrument to work. Some provincial governments have begun to do this, and the practice is worth watching.

9.5 Toward an Enabling Environment: What the State Can and Cannot Do

9.5.1 The Division of Labor Between State and Cluster

The argument of this chapter converges on a division of labor that is neither the market-fundamentalist nor the state-centric position. The state can set the floor, provide the credible horizon, build the shared infrastructure, and convene the actors. The cluster can generate the knowledge, diffuse the norms, and hold the firms to a standard that the state cannot enforce. Neither can do the other's job. A state that tries to build capability by decree produces compliance. A cluster that tries to upgrade without a credible floor produces a race to the bottom.

The division of labor is not a fixed allocation. It shifts with the phase of the cluster's life cycle, with the maturity of the technology, and with the intensity of the external pressure. In the early phase of a cluster's green trajectory, the state's role is larger: it sets the standard, provides the infrastructure, and absorbs the coordination cost. In the mature phase, the cluster's role is larger: the firms know their processes, the associations know their members, and the state's job is to maintain the floor and stay out of the way. The policy error is to apply the early-phase role in the mature phase, which produces the compliance trap, or the mature-phase role in the early phase, which produces a cluster that never forms.

The Shunde cluster is in a phase where the state's role should be shrinking and the cluster's role should be growing. The provincial standards are in place, the infrastructure is built, and the firms have the technical capacity to respond. What the cluster lacks is the associational density to diffuse the response across the firms that have not yet upgraded. The state cannot build that density by fiat. It can support it by recognizing the associations as partners, by funding the intermediation, and by giving the cluster the time and the space to develop its own norms. The current policy does some of this and not enough of it.

9.5.2 The Credible Horizon as the Central Public Good

If there is a single institutional good that the state can provide and that the cluster cannot, it is the credible horizon. Firms invest in capability when they believe the return will materialize. The return depends on the durability of the demand for green products, the trajectory of the carbon price, and the stability of the standards. Each of these is a public good that only the state can provide, and each is currently provided imperfectly.

The credible horizon is not a matter of announcing a target. It is a matter of building the institutions that make the target self-enforcing. A carbon market with a legislated cap trajectory, an independent oversight body, and a compliance regime that does not bend to political pressure is a credible horizon. A target announced in a speech and implemented through administrative campaigns is not. The difference is the institutional depth behind the commitment, and the depth is what the Chinese system is still building.

The furniture case shows why the horizon matters. The Shunde factory's managers understood the direction of policy. They did not know the pace. They deferred the foam reformulation because the payback depended on the pace, and the pace depended on enforcement, and enforcement depended on the campaign calendar. A credible horizon would have told them that the provincial standard would be enforced against mid-sized plants by a certain date, and they would have reformulated earlier. The state could have provided that horizon by publishing an enforcement schedule and sticking to it. It did not, and the factory paid the cost of the uncertainty.

9.5.3 What the Cluster Can Do That the State Cannot

The cluster can do three things that the state cannot, and they are the things that convert compliance into capability. It can generate process-specific knowledge, because the knowledge is tacit and resides in the firms and the engineers who run them. It can diffuse norms of what counts as a good firm, because the norms are social and are transmitted through the cluster's relationships. It can hold firms to a standard that is higher than the legal floor, because the standard is enforced by the peer group rather than by the inspector.

Each of these is a collective action problem, and each is solved by the cluster's institutions rather than by the state. The state can support the institutions—by funding the associations, by recognizing the testing labs, by convening the firms—but it cannot substitute for them. The temptation to substitute is strong, because the state has the resources and the cluster does not, and the substitution produces the compliance trap. The firms comply with the state's standard and do not internalize the cluster's norm, because the cluster's norm was never allowed to develop.

The Ningbo garment cluster is the counterexample. The cluster's associations were allowed to develop their own standards, and the standards are higher than the national ones. The lead firms enforce them through their supplier requirements, the associations enforce them through their membership rules, and the local state enforces them through its procurement. The three enforcement mechanisms reinforce each other, and the result is a cluster where green process investment is the norm rather than the exception. The state's role was to allow the cluster's institutions to develop and to reinforce them, not to substitute for them.

9.5.4 The Limits of Institutional Design

I want to close the analytical argument with a caution about the limits of institutional design. The chapter has treated institutions as enabling conditions that can be designed, and the treatment is warranted up to a point. But institutions are also historical products, and the ones that work are often the ones that grew rather than the ones that were built. The German and Italian institutions grew over a century. The Chinese institutions are being built in a decade. The difference in tempo has consequences for the kind of institution that can be built.

Fast-built institutions tend to be formal, because formal institutions can be legislated. Slow-grown institutions tend to be informal, because informal institutions require the accumulation of shared experience. The Chinese system has built the formal institutions quickly and is now discovering that the informal ones are the binding constraint. The discovery is not a failure of design; it is a feature of the tempo. The policy implication is that the informal institutions need time, and the time needs to be protected from the pressure to produce visible results.

The protection is itself a governance task. It means resisting the temptation to evaluate the cluster's green upgrading by the number of certificates issued or the number of inspections passed, and evaluating it instead by the depth of the process change and the durability of the norms. The evaluation is harder, and it is slower, and it is the only one that measures capability rather than compliance. The Chinese system's evaluation culture is moving in this direction—the shift from input targets to outcome targets in the 14th Five-Year Plan is a step—but the movement is incomplete.

9.5.5 Vignette: The Association That Learned to Teach

A final case illustrates what the cluster can do when its institutions are allowed to develop. The furniture association in a Foshan sub-cluster was, for most of its history, a lobbying vehicle. Its meetings were occasions for exchanging information about enforcement timing, and its relationship with the local environmental bureau was adversarial. In 2018, a new secretary general took over. She had been a process engineer at one of the larger firms, and she understood the firms' technical problems. She changed the association's program from lobbying to training, and she persuaded the local polytechnic to run a course on VOC abatement for the member firms' engineers.

The course was oversubscribed. The engineers who attended learned the mass-based standard, the process implications, and the reformulation options. They took the knowledge back to their firms, and the firms began to reformulate. The association's role changed from representing the firms to the state to representing the state's standard to the firms. The local environmental bureau noticed, and it began to route its training and its compliance guidance through the association. The association became an intermediary, and the intermediation lowered the cost of compliance for the member firms and raised the rate of process change.

The case is small, and it is not a model that can be scaled by decree. It depended on a particular person, a particular set of relationships, and a particular moment. But it shows the mechanism. The cluster's institution learned to teach, and the teaching diffused the capability. The state's role was to allow the association to develop and to use it, not to substitute for it. By 2023 the association was running four courses a year, had hired a full-time technical director, and was charging member firms a fee that covered about 60 percent of its training budget. The remaining 40 percent came from a provincial program that funds industry associations providing technical services—a program that had existed since 2016 but had gone unused in this sub-cluster until the new secretary general applied for it. The program's existence is a policy fact. Its use depended on an intermediary who knew it existed and could write the application.

9.6 Conclusion: The Enabling Environment as a Configuration

The chapter's argument can be restated as a configuration rather than a list. Green capability building requires a configuration of formal rules, informal norms, cluster institutions, and multi-level coordination in which each element reinforces the others. The configuration is not a blueprint, and it is not the same across clusters. It is a set of conditions that, when they align, allow firms to treat compliance as a floor and to invest in the capability that turns the floor into a platform.

The configuration has four elements. A credible formal floor sets the standard and the trajectory, and it is enforced consistently enough that firms believe it. An informal norm that distinguishes compliance modes makes process investment respectable and end-of-pipe abatement a mark of lag. A cluster infrastructure that diffuses knowledge—associations, testing labs, shared facilities, technical schools—lowers the cost of the conversations through which capability is transmitted. A multi-level coordination that connects the levels ensures that the price signal, the standard, the finance, and the trade regime point in the same direction and that the documentation chain reaches the firm.

Where the configuration is present, the compliance trap is escapable. Where it is absent, the trap is an equilibrium, and no amount of firm-level effort will break it. The Shunde factory that reformulated its foam did so because the configuration was partially present: the formal standard existed, the cluster had the technical knowledge, and the buyer demanded the documentation. It did so late and at high cost because the configuration was incomplete: the informal norm did not distinguish its investment from its competitors' abatement, and the multi-level coordination did not give it a credible horizon. The factory's experience is the chapter's evidence in miniature. The capability was built, and it could have been built earlier.

The policy implication is that the enabling environment is the object of policy, not the firm. The Chinese green governance system has invested heavily in the formal floor and the national coordination, and it has underinvested in the informal norms and the cluster infrastructure. The underinvestment is the binding constraint on green capability building in the labor-intensive clusters, and it is the constraint that the next phase of policy must address. The instruments are available—support for associations, funding for intermediation, recognition for shared facilities, patience for the norms to develop—and they are cheaper than the formal instruments that have dominated the policy conversation.

The next chapter turns from the enabling environment to the synthesis. It asks what the book's cases, taken together, imply for the theory of green capability and for the practice of value chain climbing. The answer will return to the Shunde factory, because the factory is where the theory meets the ground. The factory's managers did not read the institutional literature. They built the capability anyway, because the configuration around them, incomplete as it was, made the investment possible. The task of policy is to complete the configuration, so that the next factory does not have to wait for a campaign to do what it could have done earlier.