Chapter 3: Carbon Accounting, MRV, and the Data Infrastructure of Green Governance

In the autumn of 2021, a mid-sized furniture exporter in Foshan received a fourteen-page document from its largest European buyer. It was not a purchase order. It was a carbon disclosure questionnaire. The buyer, a flat-pack retailer with stores across Scandinavia and the Benelux, wanted the embodied emissions of every particleboard panel, every melamine veneer, every container of flat-packed bookcases the Foshan factory shipped. It wanted the electricity mix of Guangdong Province, the diesel consumption of the factory's forklift fleet, the refrigerant charge in the workshop's air-conditioning units, and the emissions of the upstream sawmills that supplied the raw timber. The factory manager, who had run the plant for eleven years, told me he could produce a bill of materials within an hour. The carbon number took his team six weeks and, by his own admission, was still "maybe sixty percent right."

The gap between the material accounting a factory knows and the carbon accounting it is now asked to do is where this chapter begins. Chapter 2 argued that green governance capability is built, not gifted, and that firms which measure well navigate the double-edged sword of GVC embedding better than firms which measure badly. This chapter takes the measurement question seriously. It examines the standards that define what counts as a carbon emission, the monitoring, reporting, and verification (MRV) systems that turn a standard into an auditable number, and the data infrastructure that allows those numbers to travel up and down a value chain without losing their meaning. Carbon accounting is not bookkeeping. It is the epistemic foundation of green governance, and the quality of that foundation determines which firms can climb and which firms get pushed out.

3.1 Carbon Accounting Standards and Scope Definitions

3.1.1 The Three Scopes and the Architecture of Corporate Carbon Accounting

The vocabulary of corporate carbon accounting was largely settled by the time the World Resources Institute (WRI) and the World Business Council for Sustainable Development (WBCSD) released the first version of The Greenhouse Gas Protocol: A Corporate Accounting and Reporting Standard in 2001. That document, revised in 2013, remains the reference framework for almost every national and sectoral standard in use today, including ISO 14064-1. Its central move was to divide a firm's emissions into three scopes, a division that sounds technical but is really a statement about where the firm's boundary of responsibility lies.

Scope 1 covers direct emissions from sources the firm owns or controls: fuel burned in boilers and furnaces, process emissions from chemical reactions (the calcination of limestone in cement, for instance), refrigerant leakage, and the company's own vehicle fleet. Scope 2 covers indirect emissions from purchased electricity, steam, heat, and cooling. Scope 3 covers everything else in the value chain: purchased goods and services, upstream and downstream transport, business travel, employee commuting, waste disposal, and the use-phase and end-of-life emissions of sold products.

Walk into a typical Guangdong furniture plant and the architecture reveals its seams. The drying ovens burn natural gas through a meter the gas utility reads once a month. That is Scope 1, and the number is knowable to within a few percent if someone bothers to read the meter. The electricity that runs the dust extraction system arrives through a single main breaker; no one has ever sub-metered the individual machines. That is Scope 2, and the number depends on whether the firm uses the location-based method (the average emissions intensity of the grid where it operates) or the market-based method (the emissions of the specific electricity contracts it has signed, including renewable energy certificates). The two methods can produce dramatically different numbers for the same factory. A Guangdong furniture plant purchasing green certificates might report Scope 2 emissions near zero on a market basis while its location-based figure remains tied to a grid that is still roughly 55 percent coal-fired. Both numbers are "true" under the Protocol. Neither is the truth.

Scope 3 is where the architecture becomes genuinely difficult, and where the stakes for Chinese manufacturing are highest. For most manufacturers, Scope 3 is the largest share of the footprint, often 70 to 90 percent of total emissions, because it includes the embodied carbon of purchased materials. A furniture factory's Scope 3 includes the emissions of the sawmills and board mills that supply it, the trucks that move the boards, and the disposal of the product at end of life. The factory controls none of these directly. It can influence them through procurement, but it cannot measure them without data from suppliers who may have no measurement capability at all.

Tier-1 suppliers, the factories that ship directly to the buyer, are increasingly asked to produce Scope 1 and 2 data and to estimate their own Scope 3. Tier-2 suppliers, one step upstream, are often asked nothing, which means their emissions are estimated by the Tier-1 firm using industry averages. Tier-3 suppliers, the sawmills and chemical plants at the base of the chain, are invisible. The carbon number that arrives at the European buyer's sustainability office is a composite of measured data and estimates whose proportions are rarely disclosed.

This is not a failure of the standard. It is a structural feature of value chain accounting, and it has a predictable consequence: the further a firm sits from the final buyer, the more likely its emissions are to be estimated rather than measured, and the more likely those estimates are to be wrong in ways that disadvantage the firm. A Chinese board mill whose emissions are estimated using a provincial average may be penalized for the inefficiency of its neighbors. A factory that has invested in efficient boilers but cannot document the improvement will see no benefit in the buyer's scorecard. Measurement capability is not a neutral technical asset. It is a source of bargaining power.

3.1.2 Methodologies: From Spend-Based Estimates to Process-Level Measurement

The GHG Protocol defines what to count. It is less prescriptive about how to count it, and the methodological choices available to a firm span a wide range of cost and accuracy.

At the low-cost end sits spend-based estimation, which multiplies the monetary value of a purchase by an emissions factor derived from economic input-output tables. The method is cheap, fast, and almost useless for operational decisions. It assumes that a yuan spent on steel always carries the same emissions, regardless of whether the steel came from a modern electric-arc furnace or an aging blast furnace. For a first-pass screening of a supply chain, spend-based estimation is better than nothing. For a supplier negotiation, it is a liability.

At the high-cost end sits process-level measurement: direct metering of fuel and electricity at the facility, combined with material flow analysis and site-specific emissions factors. This is what a cement plant does when it installs continuous emissions monitoring systems on its kiln stacks, and what a furniture factory does when it sub-meters its drying ovens and computes emissions from actual gas consumption. Process-level data is expensive to collect and expensive to verify, but it is the only kind of data that supports the operational decisions—which kiln to upgrade, which supplier to switch, which product line to redesign—that green governance actually requires.

Between these poles sits a range of hybrid methods: proxy factors from industry databases, extrapolation from a sample of facilities, and the use of default factors published by the Intergovernmental Panel on Climate Change (IPCC) or by national authorities. The choice among them is rarely made on accuracy grounds alone. It is made on cost, on the availability of data, and on the requirements of the buyer or regulator demanding the number. A firm selling into the European Union's Carbon Border Adjustment Mechanism (CBAM) will eventually need installation-level verified data. A firm selling into a domestic voluntary market may need only a plausible estimate.

The methodological landscape is further complicated by the proliferation of product-level standards. The GHG Protocol's Product Life Cycle Accounting and Reporting Standard, published in 2011, provides guidance for calculating the carbon footprint of an individual product, and it competes with ISO 14067, with the European Commission's Product Environmental Footprint (PEF) method, and with a growing number of sector-specific rules. For a Chinese manufacturer exporting to multiple markets, the practical consequence is that the same product may need to be measured three or four different ways, with different system boundaries and different allocation rules, producing different numbers for different customers. The compliance burden is real, but it is also an opportunity: firms that build a single, well-documented measurement system capable of generating multiple reporting outputs at low marginal cost will find themselves at an advantage over firms that rebuild their accounting from scratch for every new request.

3.1.3 The Boundary Problem: Organizational and Operational Control

Every carbon inventory must answer a deceptively simple question: what belongs inside the firm? The GHG Protocol offers two approaches. Under the equity share approach, a firm accounts for emissions in proportion to its ownership share in a facility. Under the control approach—either financial or operational control—a firm accounts for 100 percent of the emissions of facilities it controls, regardless of ownership share.

The choice matters enormously for Chinese manufacturers, and especially for the cluster-based firms that are the subject of this book. Consider a furniture cluster in Shunde where a single export order is fulfilled by a network of specialized workshops: one does cutting, one does edge-banding, one does finishing, one does packing. The workshops may share ownership, or they may be independent firms linked by subcontracting relationships. Under operational control, the exporting firm accounts only for the emissions of the workshops it operates directly. Under equity share, it accounts for the share it owns. Under a value chain approach, it accounts for everything, because the buyer's Scope 3 accounting will capture the entire network whether the exporting firm does or not.

This is the boundary problem in its practical form. The buyer's carbon number is calculated on a value chain basis, which means it includes the emissions of every subcontractor in the cluster. The exporting firm's own inventory, calculated on an operational control basis, excludes most of them. The two numbers do not match, and the gap is not a rounding error. It is the emissions of the cluster's small workshops, which are precisely the facilities least likely to have measurement capability and most likely to be using inefficient equipment.

The consequence is a form of accounting asymmetry that shapes upgrading trajectories. A large exporting firm that wants to reduce its reported footprint has two options: it can invest in efficiency within its own boundary, or it can shift emissions outside its boundary by subcontracting more work to independent workshops. The second option improves the firm's reported number without improving the cluster's actual emissions. Under a value chain accounting regime, the buyer can see through this maneuver if the buyer has the data. Under a weaker regime, the incentive to offshore emissions to the informal sector is real.

The Chinese national emissions trading system (ETS), which began with power generation and is expanding to other sectors, uses a different boundary convention again: it regulates installations above a size threshold, measured at the point of combustion. A furniture factory below the threshold is not regulated at all, even if its aggregate emissions across a cluster are substantial. The mismatch between the ETS boundary, the GHG Protocol boundary, and the buyer's value chain boundary means that the same physical emissions can be counted three different ways for three different purposes, and a firm may be optimizing against one boundary while being penalized against another.

3.1.4 Carbon Accounting in the Chinese Institutional Context

China's carbon accounting infrastructure has developed along a trajectory distinct from the European one, and the differences matter for understanding how Chinese firms experience the measurement burden.

The domestic architecture rests on a set of national standards for greenhouse gas emissions accounting and reporting across industrial sectors, issued by the National Development and Reform Commission (NDRC) and later administered by the Ministry of Ecology and Environment (MEE). These sectoral standards—for power generation, steel, cement, chemicals, and others—define the calculation methods, the emissions factors, and the reporting templates that regulated enterprises must use. They are, by design, simpler than the GHG Protocol: they focus on Scope 1 combustion and process emissions, treat purchased electricity through a grid-average factor, and largely exclude Scope 3. The simplicity is deliberate. It reflects the administrative capacity of local environmental bureaus and the need to produce comparable numbers across thousands of enterprises.

The voluntary offset market, the China Certified Emission Reduction (CCER) program, draws heavily on the Clean Development Mechanism (CDM) methodologies developed under the UNFCCC. The CDM was the original laboratory for project-level carbon accounting in China, and the methodological toolkit it built—baseline setting, additionality testing, monitoring plans—remains the reference point for domestic offset projects. The CCER program was suspended in 2017 and relaunched in 2023 with a narrower scope, but the methodological lineage is visible in its project categories and its verification requirements.

Two accounting tracks now run in parallel. Regulated enterprises in the national ETS report according to domestic sectoral standards, which are audited by designated verification bodies but are not directly comparable to the GHG Protocol inventories that foreign buyers request. Export-oriented firms, meanwhile, must produce GHG Protocol-compliant inventories for their customers, often using third-party consultants and international software platforms. The two tracks use different emissions factors, different boundary conventions, and different verification protocols. A firm operating in both tracks may find itself maintaining two parallel carbon ledgers, reconciled only by a footnote.

This dual-track system is not unique to China. But its scale is. China has more enterprises subject to mandatory carbon reporting than any other country, and the vast majority of them are small or medium-sized by international standards. The measurement burden falls disproportionately on firms with the least capacity to bear it, which is precisely the pattern that the capability argument in this book predicts: measurement capability becomes a source of differentiation, and firms without it are pushed toward the low-margin, low-visibility end of the value chain.

3.1.5 What the Numbers Reveal About Firm Capability

Firms that export to European markets are more likely to have a dedicated sustainability function, more likely to have commissioned a third-party inventory, and more likely to have invested in sub-metering and process-level data collection. Firms that sell only domestically, or that sell into markets without carbon disclosure requirements, are more likely to rely on spend-based estimates or on default factors. This pattern shows up consistently in the supplier data that international buyers collect, though the firms themselves rarely publish it.

The correlation is not causation, and the direction of causality is contested. It is possible that firms with better measurement capability self-select into export markets. It is also possible that export market requirements drive the investment. Both mechanisms appear to be at work, with buyer pressure acting as the proximate trigger and prior organizational capability determining which firms can respond.

What is clearer is that the measurement investment has spillovers. A furniture factory that installs sub-metering to satisfy a European buyer discovers that it can identify energy waste in its drying operations. A board mill that commissions a process-level inventory discovers that its resin formulation is a larger source of emissions than its boiler. The carbon number, once produced, becomes a management tool. The act of measurement creates the information that makes improvement possible.

The counterfactual is equally instructive. Firms that produce carbon numbers only under duress, using the cheapest available method, learn nothing from the exercise. Their numbers satisfy the buyer's form and change nothing in the factory. The difference between these two outcomes is not the standard. It is the organizational capability that surrounds the standard, and that capability is what the next section examines.

3.2 MRV Systems: Design and Implementation

3.2.1 What MRV Is For

Monitoring, reporting, and verification (MRV) is the institutional machinery that turns a carbon number into a credible one. Monitoring is the ongoing collection of activity data and emissions factors. Reporting is the compilation of that data into a standardized disclosure. Verification is the independent check that the reported number is accurate, complete, and consistent with the applicable standard.

The three functions are often discussed as a single system, but they serve different purposes and are subject to different failure modes. Monitoring is an operational function, embedded in the factory's routines. Reporting is a disclosure function, shaped by the requirements of the audience. Verification is an assurance function, performed by a third party with no stake in the number.

The credibility of the entire system depends on the weakest link. A rigorous verification of a poorly monitored number produces a certified fiction. A rigorous monitoring effort reported in a format that obscures the underlying data produces a number that cannot be compared. And a well-monitored, well-reported number that is verified by a body with a conflict of interest produces a number that no one should trust.

MRV systems are therefore best understood not as a compliance checklist but as an information architecture. The architecture determines what data flows where, who can see it, and what can be done with it. A well-designed architecture supports both compliance and improvement. A poorly designed one supports compliance and nothing else.

3.2.2 The Monitoring Layer: Activity Data and Emissions Factors

Every emissions calculation is a product of two quantities: activity data (how much fuel was burned, how much electricity was consumed, how much clinker was produced) and an emissions factor (how much carbon dioxide equivalent is released per unit of activity). The accuracy of the result is bounded by the accuracy of both.

Activity data is the easier half, in principle. A factory that meters its gas consumption knows how much gas it burned. A factory that does not meter its gas consumption knows only how much it paid for. The gap between these two states is the gap between measured and estimated emissions, and it is wider than most managers assume. In many Chinese industrial facilities, fuel consumption is tracked at the level of the monthly utility bill, not at the level of the individual process. The bill tells the firm how much energy it bought. It does not tell the firm where the energy went.

The monitoring layer that MRV requires is more granular. It requires sub-metering at the level of significant energy users, continuous or periodic measurement of process emissions, and record-keeping that allows a verifier to trace a reported number back to a meter reading. For a cement plant, this means continuous emissions monitoring on the kiln stack or, where that is not feasible, a documented calculation based on clinker production and raw material composition. For a furniture factory, it means sub-metering the drying ovens, the presses, and the dust extraction system, and tracking the refrigerant charge in the air-conditioning units.

The emissions factor half is harder, because emissions factors are not properties of the firm. They are properties of the fuel, the grid, or the process, and they are assigned by an authority. The choice of factor can change the reported number substantially. A factory that uses a default factor for grid electricity will report a different Scope 2 number than a factory that uses a market-based factor reflecting its renewable energy purchases. A cement plant that uses a default clinker factor will report differently from one that uses a factor derived from its own raw meal composition.

The Chinese domestic system handles this by prescribing factors. The MEE publishes grid emissions factors by region and by year, and the sectoral accounting standards specify the default factors to be used for common fuels and processes. The prescription ensures comparability across firms, which is essential for the ETS. It also creates a lag: the grid factor for a given year is typically published two or three years later, which means firms are reporting against a factor that may not reflect the current grid mix. As the grid decarbonizes, this lag systematically overstates Scope 2 emissions for firms purchasing grid electricity.

International buyers, by contrast, often require firms to use the most recent factors available, or to disclose the factor vintage. The result is another dual-track problem: the same factory may report one Scope 2 number to the domestic regulator and a different one to the foreign buyer, with the difference attributable entirely to the choice of factor.

3.2.3 The Reporting Layer: Templates, Platforms, and the Burden of Translation

Reporting is where the measurement burden becomes visible to the firm. A reporting requirement arrives as a template—a spreadsheet, a questionnaire, a platform—and the firm must map its internal data onto the template's categories.

The templates are not standardized. The CDP climate change questionnaire, the GHG Protocol corporate standard, the ISO 14064-1 verification protocol, the European Sustainability Reporting Standards (ESRS), the CBAM transitional reporting format, and the individual questionnaires of large buyers all use different categories, different boundary conventions, and different levels of granularity. A firm reporting to all of them must maintain a master dataset and a set of translation rules, or it must rebuild its inventory for each audience.

The translation burden is heaviest for Scope 3, because Scope 3 categories are defined by the GHG Protocol but interpreted differently by different buyers. A furniture factory's "purchased goods and services" category may be interpreted to include the emissions of the board mill, the emissions of the timber harvest, or both, depending on the buyer's methodology. The factory must therefore know not only its own emissions but the methodology of its customer, which is a form of supply chain literacy that few firms have developed.

Digital reporting platforms have reduced some of this burden and created new ones. Platforms such as CDP, EcoVadis, and the various buyer-specific supplier portals automate the collection and comparison of data, but they also impose their own categories and their own scoring rules. A firm that scores poorly on a platform may lose business without ever understanding why, because the scoring methodology is proprietary and the feedback is limited. The platform becomes a gatekeeper, and the gatekeeping function is exercised through a data architecture that the firm cannot see.

The Chinese response to this burden has been mixed. Some large exporters have built internal sustainability teams capable of managing multiple reporting formats. Others have outsourced the work to consultants, who produce the numbers but leave no capability behind. The cluster-based small firms, which constitute the majority of Chinese manufacturers by number, have done neither. They report what they are asked to report, when they are asked, using whatever method is cheapest, and they treat the exercise as a cost of doing business rather than as a source of information.

3.2.4 The Verification Layer: Accreditation, Independence, and the Chinese Verification Market

Verification is the assurance that the reported number is true. It is performed by accredited third-party bodies, who review the firm's monitoring data, check the calculation methodology, and issue a verification statement. The credibility of the entire MRV system rests on the independence and competence of these bodies.

China has built a substantial verification industry. The MEE accredits verification bodies for the national ETS, and the China National Accreditation Service for Conformity Assessment (CNAS) accredits bodies for voluntary and international work. The largest verification bodies are affiliated with research institutes, universities, and state-owned enterprises, and they employ engineers with sector-specific expertise. The system is younger than its European counterpart but has scaled rapidly, and the quality of verification has improved as the ETS has matured.

Independence is the persistent concern. Verification bodies are paid by the firms they verify, which creates a structural conflict of interest. The conflict is managed through accreditation requirements, rotation rules, and penalties for deficient verification, but it is not eliminated. Rigor varies across provinces and across verification bodies, with some bodies consistently producing more conservative findings than others. The variation is not necessarily a sign of corruption. It may reflect differences in the competence of the verification teams, the quality of the data they are given, and the ambiguity of the standards they apply.

For export-oriented firms, the relevant verification is often performed by international bodies—Bureau Veritas, SGS, TÜV, and their peers—because the buyer requires it. These verifications are more expensive and more rigorous than domestic ones, and they produce numbers that are accepted in European and North American markets. The cost is a barrier for small firms, and the barrier is highest precisely for the firms that most need access to the markets where verification is required.

The verification layer also determines what can be done with a number. A verified number can be used in a carbon market, in a border adjustment calculation, or in a buyer's supplier scorecard. An unverified number is a management estimate. The gap between the two is the gap between a firm that can participate in green governance and a firm that can only be governed by it.

3.2.5 MRV in the Chinese ETS: Design Choices and Their Consequences

The Chinese national ETS, launched in 2021 with the power sector and expanded to other sectors since, is the largest carbon market in the world by covered emissions. Its MRV system is the largest in the world by number of reporting entities, and its design choices have consequences that extend beyond the ETS itself.

The system covers enterprises above an emissions threshold, currently around 26,000 tonnes of carbon dioxide equivalent per year for the power sector and similar thresholds for other sectors. Covered enterprises must prepare an annual emissions report using the prescribed methodology, have it verified by an accredited body, and surrender allowances equal to their verified emissions. The allowances are allocated largely for free, based on benchmarks or on historical intensity, and the market price has been modest by international standards—in the range of 60 to 90 yuan per tonne in recent years, compared with 80 to 100 euros in the EU ETS.

The MRV design reflects the constraints of the system. The use of prescribed factors and simplified methodologies reduces the measurement burden on enterprises and the verification burden on regulators, at the cost of accuracy. The exclusion of Scope 3 keeps the system focused on direct emissions, at the cost of ignoring the value chain. The free allocation and the modest price mean that the system creates a reporting obligation more than a strong abatement incentive, at least in its early years.

These choices are defensible as a first phase. They have built a reporting infrastructure and a verification industry that did not exist a decade ago, and they have given thousands of enterprises their first experience of carbon accounting. The question is whether the infrastructure can be upgraded as the system matures. The expansion of the ETS to new sectors—cement, steel, aluminum, and eventually others—will bring in enterprises with more complex processes and more heterogeneous emissions profiles. The transition from free allocation to auctioning, which is on the policy agenda, will raise the price and make the MRV system's accuracy more consequential. The inclusion of Scope 2 at a finer granularity, and eventually Scope 3, would align the domestic system more closely with the international requirements that export-oriented firms already face.

The evidence from the pilot ETS programs, which operated in seven provinces and cities before the national system, suggests that MRV quality improves with experience. The pilots that invested in verification capacity and in enterprise training produced more reliable data than those that did not. MRV is a learning system, and its quality is a function of the investment put into it.

3.2.6 The Verification of Avoided Emissions: Additionality and Its Discontents

A separate MRV challenge arises in the voluntary offset market, where the commodity being verified is not an emission but an emission reduction. The CCER program, like the CDM before it, credits projects that reduce emissions relative to a baseline. The credibility of the credit depends on the credibility of the baseline, and the baseline is a counterfactual that cannot be observed.

The additionality problem—whether the project would have happened without the carbon revenue—is the classic difficulty of offset markets. A wind farm that would have been built anyway does not represent an additional reduction. A forestry project that protects a forest that was not under threat does not either. The methodologies used to test additionality are necessarily approximate, and the history of the CDM is littered with projects whose additionality was questionable.

China's CCER program has learned from the CDM's experience and has narrowed its scope to project types with clearer additionality, including certain forestry, renewable energy, and methane recovery projects. The methodological toolkit is more conservative than the CDM's, and the verification requirements are stricter. The program remains small relative to the compliance market, and its future role depends on how the ETS evolves and whether offsets are allowed for compliance.

For the firms in this book, the offset market is less important than the compliance market, but the additionality problem is instructive. It shows that MRV involves a judgment about what counts, and that judgment is technical only in part. The methodology carries it, and the methodology is a political artifact as much as a scientific one. Firms that understand this can navigate the system more effectively. Firms that treat the methodology as a black box will find their projects rejected or their credits discounted.

3.2.7 MRV as a Capability, Not a Cost

The conventional view of MRV is that it is a compliance cost: something a firm must do because a regulator or a buyer requires it. This view is accurate as far as it goes, but it misses the capability dimension.

A firm that has built a robust MRV system has, by that fact, built a set of organizational capabilities: data collection routines, metering infrastructure, analytical staff, and relationships with verifiers. These capabilities can be redeployed. The same data that supports an ETS report can support an energy management system, a supplier scorecard, a product design decision, or a green marketing claim. The marginal cost of the second use is far lower than the cost of the first.

The evidence from Chinese firms that have invested in MRV capability suggests that the returns are real but uneven. Firms with strong internal engineering functions find it easier to integrate carbon data into operational decisions. Firms that treat MRV as a reporting exercise, handled by the finance or public relations department, find that the data sits in a spreadsheet and changes nothing. The difference is organizational, not technical, and it is the same difference that the previous chapter identified between firms that navigate the double-edged sword and firms that are cut by it.

3.3 Data Infrastructure for Value Chain Decarbonization

3.3.1 The Value Chain Data Problem

The previous two sections examined carbon accounting and MRV at the level of the individual firm. This section scales up to the value chain, where the data problem becomes qualitatively different.

A value chain is a sequence of firms, each with its own boundary, its own accounting method, and its own data quality. The carbon footprint of a final product is the sum of emissions across the sequence, plus the emissions of transport and distribution between the links. To calculate that sum, each firm must be able to receive data from its suppliers, combine it with its own data, and pass the result to its customers. The data must be commensurable: the same units, the same boundary conventions, the same emissions factors, the same level of granularity. Without commensurability, the sum is meaningless.

Commensurability is rare. Most reported Scope 3 emissions are calculated using secondary data—industry averages, spend-based factors, or supplier estimates—rather than primary data from the suppliers themselves. The secondary data may be adequate for a rough estimate, but it cannot support the kind of targeted intervention that decarbonization requires. A buyer that knows only the average emissions intensity of Chinese furniture cannot tell which of its suppliers is efficient and which is not, and therefore cannot reward the efficient ones or help the inefficient ones improve.

The data problem is compounded by the asymmetry of capability along the chain. The final buyer, typically a large retailer or brand, has the most sophisticated sustainability function and the greatest data capability. Its Tier-1 suppliers have less. Its Tier-2 and Tier-3 suppliers have less still. The data flows upward in a form that the buyer can use, but the capability to produce that form does not flow downward. The result is a system in which the buyer's knowledge of the chain is only as good as the weakest link's measurement, and the weakest link is often the one with the most emissions.

3.3.2 Traceability: From Facility-Level to Product-Level Data

The frontier of value chain data infrastructure is traceability: the ability to associate a specific product with the specific emissions of the specific facilities that made it. Traceability is what allows a buyer to distinguish between two suppliers of the same component, and what allows a firm to make a credible product-level carbon claim.

Traceability requires more than carbon accounting. It requires a system of identifiers that links materials to processes to products, and a data architecture that preserves the link as the product moves through the chain. In the furniture sector, this means tracking a batch of particleboard from the board mill through the cutting, edging, and assembly stages to the finished flat-pack, and associating each stage with its energy and material inputs. In the cement sector, it means tracking clinker from the kiln to the grinding mill to the concrete batching plant, and associating each stage with its process emissions.

The technologies that enable traceability are available: RFID tags, QR codes, blockchain-based registries, and digital product passports. The European Union's Ecodesign for Sustainable Products Regulation (ESPR) is moving toward a digital product passport requirement that would make traceability mandatory for a wide range of products sold in the EU market. China's own industrial internet initiatives have built traceability infrastructure in some sectors, particularly in consumer goods and electronics.

The adoption of traceability in the carbon-intensive sectors that matter most for this book is uneven. Large cement producers have the capability but limited incentive, because their products are commodities and their customers do not yet demand product-level carbon data. Furniture exporters have the incentive, because their buyers demand it, but limited capability, because their supply chains are fragmented and their suppliers are small. The gap between incentive and capability is where the data infrastructure is being built, slowly and unevenly.

3.3.3 Interoperability: Standards, Protocols, and the Politics of Data Exchange

Even where traceability exists, the data must be exchangeable. Interoperability—the ability of different systems to exchange and use data—is the technical condition for value chain decarbonization, and it is as much a political problem as a technical one.

The politics arise because standards embody interests. A data standard that defines "product carbon footprint" in one way will favor firms whose processes fit that definition and disadvantage firms whose processes do not. A standard that requires facility-level data will favor large firms with metering infrastructure and disadvantage small firms without it. A standard that allows estimated data will favor firms with complex supply chains and disadvantage firms that have invested in measurement. The choice of standard is therefore a choice about which firms can participate in the low-carbon market and on what terms.

The current landscape is fragmented. The GHG Protocol provides a corporate-level framework but not a product-level data exchange protocol. ISO 14067 provides a product-level accounting standard but not an interoperability specification. The various industry initiatives—the automotive sector's Catena-X, the consumer goods sector's Sustainability Consortium, the apparel sector's Higg Index—have built sector-specific data architectures that are not interoperable with each other. The European Commission's PEF method is an attempt to build a harmonized framework, but its adoption outside Europe is limited.

China's approach to interoperability has emphasized domestic standards and domestic platforms. The national carbon market's reporting system is a closed architecture, designed for regulatory compliance rather than for value chain data exchange. The industrial internet platforms built by large Chinese firms, such as those in the consumer electronics and automotive sectors, have their own data architectures, which are often proprietary. The result is a set of domestic data silos that are not easily connected to the international systems that buyers use.

The silos create a risk for Chinese exporters. If the international systems become the de facto standard for value chain data exchange, Chinese firms will need to adapt their data architectures to connect to those systems, or they will be excluded from the value chains that use them. Adaptation is possible—the technical requirements are not prohibitive—but it requires investment and a willingness to adopt standards that were not designed with Chinese firms in mind.

3.3.4 The Role of Digital Platforms and Third-Party Data Providers

The gap between the data that buyers need and the data that suppliers can produce has created a market for intermediaries. Digital platforms and third-party data providers aggregate supplier data, fill gaps with estimates, and sell the result to buyers who want a value chain footprint without building the infrastructure themselves.

The platforms vary in their methods. Some collect primary data from suppliers through standardized questionnaires and verify it against documentation. Others use economic input-output models to estimate emissions from spend data. Others combine satellite data, utility records, and machine learning to estimate facility-level emissions without direct supplier engagement. Each method has a different accuracy profile, and the platforms are not always transparent about which method they use for which supplier.

The intermediary market has advantages. It lowers the cost of value chain data collection for buyers, and it gives suppliers a single channel through which to report to multiple customers. It also has risks. The platforms become gatekeepers, and their scoring methodologies become the de facto standard for supplier evaluation. A supplier that scores poorly on a platform may lose business without understanding why, and the platform's methodology may embed assumptions that disadvantage certain types of firms.

The Chinese response to the intermediary market has been mixed. Some large exporters have engaged directly with international platforms, treating the engagement as a marketing investment. Others have built their own data systems and refused to participate in platforms they do not control. The cluster-based small firms have largely been excluded from the platform economy, because they lack the data to participate and the capacity to respond to platform requests.

3.3.5 Data Infrastructure as a Public Good

The final question for this section is whether value chain data infrastructure should be treated as a private asset or a public good. The current trajectory is toward private provision: platforms built by firms, standards set by industry consortia, data held in proprietary systems. The private model has driven investment and innovation, but it has also created fragmentation, gatekeeping, and exclusion.

The public-good model would treat the basic infrastructure—identifiers, protocols, registries, and verification standards—as a shared resource, funded and governed collectively. Elements of this model exist: the GHG Protocol is a multi-stakeholder standard, the UNFCCC's CDM methodologies are public, and the Chinese ETS reporting system is a public platform. But the model has not been extended to the value chain data that decarbonization requires.

The case for public provision rests on the same logic as the case for public provision of other infrastructure. The value of a data network increases with the number of participants, and the private incentive to build a network is weaker than the social benefit. Small firms cannot afford to build their own data systems, and they cannot afford to participate in every platform their customers use. A public infrastructure would lower the barrier to participation and reduce the fragmentation that currently limits the usefulness of value chain data.

The case against public provision rests on the pace of technological change and the difficulty of governing a shared resource. Public infrastructure can lag behind private innovation, and the governance of a shared data resource raises questions about privacy, commercial confidentiality, and control that are not easily resolved.

The evidence from other infrastructure sectors suggests that a hybrid model is most likely to emerge. The basic protocols and registries will be public or multi-stakeholder, while the applications and analytics built on top of them will be private. The Chinese experience with industrial internet platforms suggests that this hybrid model can work, provided that the public layer is genuinely open and the private layer is genuinely competitive.

3.3.6 Building Data Infrastructure in Chinese Clusters

The cluster context, which is central to this book's argument, adds a specific dimension to the data infrastructure question. Clusters are dense networks of firms that share suppliers, customers, and workers. They are also, in many cases, data-poor: the firms are small, the equipment is old, and the measurement culture is thin. The question is whether the cluster itself can become a unit of data infrastructure, rather than each firm building its own.

There are precedents. The industrial symbiosis networks inTianjin and in the Suzhou Industrial Park built shared data platforms for energy and material flows, allowing firms to identify waste heat, byproduct, and shared utility opportunities that no single firm could see. The "blue map" style environmental management platforms in some Zhejiang clusters aggregate emissions data across firms for regulatory purposes. The cluster-level carbon accounting initiatives in Guangdong's ceramics and furniture clusters, supported by local industry associations and provincial research institutes, have begun to produce sectoral benchmarks that individual firms can use to locate themselves.

These initiatives share a feature: they treat the cluster, not the firm, as the unit of measurement. A cluster-level benchmark allows a small firm to compare its emissions intensity against its peers without building a full inventory. A shared metering platform allows a cluster to aggregate its emissions for a collective bargaining position with buyers or with the grid. A shared verification arrangement allows small firms to access third-party assurance at a cost they can afford.

The cluster model is not a substitute for firm-level MRV, and it does not resolve the boundary problem. It is a complement, and it addresses the specific failure mode of cluster-based manufacturing: the inability of small firms to produce the data that value chain decarbonization requires. Whether it scales depends on the governance capacity of the cluster associations, which varies widely across regions and sectors.

3.3.7 The Data Gap and the Relocation Decision

The data infrastructure question connects directly to the relocation dynamics that Chapter 4 will examine in the furniture sector. A firm that cannot produce credible carbon data faces a specific set of pressures. Its European buyer may downgrade it in the supplier scorecard, shift orders to a competitor with better data, or require it to use the buyer's own measurement platform at its own cost. The firm's domestic customers may not care about carbon at all. The firm's regulator may require an ETS report that is unrelated to the buyer's request.

The rational response for a firm in this position is to consider relocation. Moving production to a region with lower energy costs, or to a country with weaker carbon disclosure requirements, reduces the compliance burden in the short term. It does not reduce the underlying emissions, and it does not build capability. But it may preserve the firm's margin for a few more years.

The data infrastructure argument suggests that relocation is a symptom, not a solution. The firms that relocate to escape measurement requirements are the firms that have not built measurement capability, and they will face the same requirements wherever they go, as carbon disclosure regimes spread. The firms that stay and invest in data infrastructure are the firms that can meet the requirements and use them to differentiate. The divergence between these two trajectories is the mechanism by which green governance capability becomes a determinant of industrial location.

3.3.8 Data Quality, Uncertainty, and the Limits of Measurement

A chapter on data infrastructure would be incomplete without acknowledging the limits of the data. Carbon accounting is not a precise science. The emissions factors are estimates, the activity data is sampled, the boundary conventions are conventions, and the Scope 3 estimates are, in many cases, educated guesses. The uncertainty in a corporate carbon inventory can be substantial: plus or minus 20 percent for a well-measured Scope 1 and 2 inventory, and far more for a Scope 3 estimate built on secondary data.

The uncertainty matters for governance. A buyer that uses a supplier's carbon number to make sourcing decisions is making those decisions on the basis of a number that may be wrong by a wide margin. A regulator that sets an allowance allocation on the basis of reported emissions is allocating on the basis of a number that may not reflect actual emissions. A firm that invests in abatement on the basis of its inventory may find that the abatement it achieves is smaller or larger than the inventory predicted.

The response to uncertainty is not to abandon measurement but to disclose it. The GHG Protocol requires firms to report on the uncertainty of their inventories, and the better inventories include a quantitative uncertainty assessment. The disclosure allows users of the data to weigh it appropriately, and it allows firms to improve their measurement over time by targeting the largest sources of uncertainty.

The Chinese domestic system has been slower to adopt uncertainty disclosure. The sectoral accounting standards prescribe methods but do not require firms to quantify the uncertainty of their results. The verification process checks whether the methods were followed, not whether the resulting number is close to the true value. The gap is a weakness in the system, and it is one that the expansion of the ETS and the tightening of international requirements will likely close.

3.3.9 The Data Infrastructure of the Future: What Would It Take?

If the argument of this chapter is correct—that data infrastructure is the foundation of green governance, and that the quality of the foundation determines which firms can climb—then the policy question is what it would take to build a foundation adequate to the task.

Four elements seem necessary. The first is a set of interoperable standards for product-level and facility-level carbon data, developed through a process that includes Chinese firms and Chinese regulators as participants rather than as adopters. The second is a public or multi-stakeholder registry that allows firms to store and share verified data without building their own platforms. The third is a verification capacity that is independent, competent, and affordable for small firms, possibly through cluster-level shared services. The fourth is a measurement capability program that trains firm-level staff in carbon accounting, in the same way that quality management programs trained a generation of Chinese factory managers in statistical process control.

None of these elements is unprecedented. Each exists in some form, in some jurisdiction, for some sector. What does not exist is an integrated system that connects them and makes them accessible to the firms that need them most. Building that system is the infrastructure task of the next decade, and it is as much a matter of industrial policy as of environmental policy.

3.3.10 The Capability Ladder, Revisited

Chapter 2 introduced the image of a capability ladder, with firms at different rungs and different trajectories. This chapter has given the ladder a specific content. The bottom rung is the firm that cannot produce a carbon number at all, and that relies on its buyer or its consultant to produce one for it. The next rung is the firm that can produce a number using default factors and spend-based estimates, without understanding its own emissions. The next is the firm that has built a facility-level inventory with sub-metering and process-level data, and that uses the inventory to identify improvement opportunities. The top rung is the firm that has integrated carbon data into its product design, its supplier management, and its customer relationships, and that can generate verified product-level data on demand.

The ladder is a hierarchy of capability, and the rungs are defined by the data infrastructure a firm can access and operate. A firm's position on the ladder is determined by its own investment, by the requirements of its customers, and by the public infrastructure available to it. The policy implication is that moving firms up the ladder requires action at all three levels, and that the public infrastructure level is the one most amenable to deliberate design.

The Chinese manufacturing system has a distinctive advantage in this regard: its scale and its cluster structure allow for collective solutions that would be difficult in more fragmented industrial systems. A cluster-level data platform, a shared verification service, a sectoral benchmark—these are feasible in a system where thousands of firms are concentrated in a few hundred clusters, and where industry associations and local governments have experience in organizing collective action. The advantage is not automatic. It requires investment and governance, and it requires that the collective solutions be designed to serve the firms rather than the platforms.

3.3.11 What the Foshan Factory Learned

Return to the Foshan furniture exporter and its fourteen-page questionnaire. Six weeks after receiving it, the firm's team produced a carbon number that was, by the manager's estimate, sixty percent right. The buyer accepted it, with a note asking for improvement in the next reporting cycle.

The manager's response was not to hire a consultant. It was to install sub-meters on the drying ovens and the presses, and to assign one engineer to maintain the data. The investment was modest: a few thousand yuan for the meters, a few weeks of the engineer's time. The first result was a surprise. The drying ovens, which the manager had assumed were the plant's largest energy user, were less significant than the dust extraction system, which ran continuously regardless of production schedule. The second result was a change in the production schedule, which reduced the extraction system's runtime and cut the plant's electricity consumption by seven percent.

The carbon number did not cause the change. The measurement did. The buyer's questionnaire was the trigger, but the capability that the firm built in response was the mechanism. That is the pattern this chapter has described, and it is the pattern that the rest of the book will trace through the labor-intensive, technology-intensive, and capital-intensive sectors.

Chapter 4 takes up the furniture sector in detail, examining how the measurement capability described here interacts with the relocation and upgrading dynamics that are reshaping the cluster. The data infrastructure argument provides the link: firms that measure well can upgrade in place, while firms that measure badly face a choice between relocation and marginalization. The choice is not determined by the carbon number itself. It is determined by the capability that produces the number, and by the infrastructure that makes the capability accessible.

Carbon accounting, MRV, and data infrastructure are not peripheral administrative functions. They are the epistemic core of green governance. The three-scope architecture defines what counts; the MRV system determines whether the count is credible; the data infrastructure determines whether the count can travel. Each layer has its own failure modes, and the failure modes are distributed unevenly across firms and across value chain positions.

The Chinese institutional context adds a dual-track structure: a domestic system designed for ETS compliance and an international system designed for value chain disclosure. The two tracks use different boundaries, different factors, and different verification protocols, and firms that operate in both bear a translation burden that falls most heavily on those with the least capacity. The dual-track structure is not a temporary anomaly. It reflects the different purposes of the two systems, and it will persist as long as the purposes differ.

The capability argument, which is the thread running through this book, applies to data infrastructure as much as to any other dimension of green governance. Firms that build measurement capability can use it for multiple purposes, can respond to multiple audiences, and can convert the data into operational improvement. Firms that do not build it are governed by data they did not produce, using methods they do not understand, for purposes they did not choose. The difference between these two states is the difference between a firm that participates in green governance and a firm that is subject to it.

The policy agenda that follows is not complicated, though it is demanding. It requires interoperable standards, accessible registries, affordable verification, and a training infrastructure for firm-level staff. It requires that these elements be designed with the participation of the firms that will use them, and that they be funded as public goods rather than as private platforms. Treating measurement as an investment in capability, rather than a cost to minimize, changes which firms can climb.