Chapter 4: Incorporating Environmental Performance into Technology Firm Ratings

In 2021 the European Banking Authority surveyed 29 large institutions about how they treated environmental risk in credit assessment. Most had some process. Few could show that the process changed a single lending decision. The gap between having an ESG policy and pricing an ESG risk is where this chapter sits. A rating agency that cannot see a firm's environmental trajectory is rating a different firm from the one that will exist in five years, and the same holds for a bank's internal rating model. That is a technical claim about information loss, not an argument for green lending.

Standard credit ratings measure a borrower's capacity to service debt from historical cash flow and balance-sheet structure. For a steel producer installing electric-arc furnaces, that measurement captures the cost of the investment and misses the asset it creates. The rating penalizes the transition and ignores the transition's payoff. Chapter 3 set out the architecture of credit risk measurement. This chapter asks what happens when environmental performance enters that architecture as a weighted variable, capped like any other input.

The spine of the argument is a distinction that current practice routinely blurs. Environmental performance is what the firm did last year, measured in emissions, energy intensity, waste. Environmental trajectory is the direction and credibility of the firm's transition, measured in capital plans, technology choices, and the alignment of its revenue base with a decarbonizing economy. A rating that incorporates only the first penalizes the dirty firm and rewards the clean one. Useful, and static. A rating that incorporates the second can separate a dirty firm in transition from a dirty firm in decline. That separation is where the credit risk lives, and for technology firms it carries most of the information.

4.1 Environmental Metrics for Tech Firms: Selection and Weighting

4.1.1 The Indirectness Problem

A cement plant's environmental performance is legible: tons of CO₂ per ton of clinker, measured at the stack. A software company's environmental performance is nearly invisible at the firm boundary. Its direct emissions—Scope 1 and Scope 2 in the GHG Protocol vocabulary—are office electricity and business travel. For a mid-sized enterprise software firm, Scope 1 and 2 emissions might total a few thousand tons of CO₂ equivalent per year. The emissions embedded in the data centers it rents and the hardware it deploys (Scope 3) can be an order of magnitude larger, but they are another firm's Scope 1 and 2, and the attribution is contested.

The technology sector's environmental significance is concentrated in three channels, none of which appears in a conventional emissions inventory. The energy and material intensity of digital infrastructure is the first. Data centers consumed an estimated 1–1.5 percent of global electricity in 2022, according to the International Energy Agency's Data Centres and Data Transmission Networks tracking report, with the share rising as artificial intelligence workloads expand. The enabling effect is the second: a firm selling grid-management software or industrial efficiency systems produces environmental benefits that accrue to its customers, not to its own footprint. The third is transition-risk exposure in the firm's own business model. A manufacturer of internal combustion engine components faces a different trajectory than a manufacturer of power electronics, whatever their current emissions.

A rating framework for technology firms must therefore weight indirect and enabling indicators more heavily than a framework designed for heavy industry. The proportionality principle applies with force here. A uniform environmental scorecard applied to a steel mill and a semiconductor designer will misrate both.

4.1.2 A Three-Tier Indicator Set

Tier 1 contains physical environmental performance: Scope 1 and Scope 2 emissions intensity, energy consumption per unit of revenue, water withdrawal in water-stressed regions, and hazardous waste generation. These indicators are mandatory, verifiable, and comparable within sectors. For most technology firms they will be small in absolute terms, and that is fine. The purpose is to establish a floor of measurement, not to rank firms by tonnage.

Tier 2 contains transition and trajectory indicators: the share of capital expenditure directed at low-carbon technologies, the emissions intensity of the product portfolio, the firm's stated and externally validated decarbonization targets, and the alignment of revenue with taxonomy-eligible activities. These indicators are forward-looking and therefore softer. They are also where the credit-relevant information sits. A firm with high current emissions and a credible, funded transition plan is a better credit risk over a five-year horizon than a firm with low current emissions and a business model that decarbonization will strand.

Tier 3 contains enabling and systemic indicators: the environmental benefit attributable to the firm's products, measured where possible through customer-side abatement, and the firm's exposure to physical climate risk in its operations and supply chain. These are the hardest to quantify and the most valuable for technology firms.

The weighting problem follows directly. If Tier 1 dominates the score, the framework becomes an emissions inventory and misses trajectory. If Tier 3 dominates, the framework becomes a marketing exercise and loses verifiability. The weights must be sector-specific and, within a sector, size-specific: a small firm cannot be expected to produce the same disclosure depth as a large one, and penalizing it for that produces a rating that measures reporting capacity rather than environmental performance.

One principle governs the whole exercise: weight by credit relevance, not by environmental salience. An indicator that is environmentally important but uncorrelated with default probability belongs in a separate environmental score, disclosed alongside the rating. Conflating the two is how ratings lose credibility.

4.1.3 What the Disclosure Regime Can Supply

The indicator set is constrained by what firms actually report. The IFRS Foundation's International Sustainability Standards Board, established in 2021, issued IFRS S1 and S2 in June 2023, providing a global baseline for sustainability and climate-related disclosure. The EU's Corporate Sustainability Reporting Directive, phased in from 2024, requires disclosure aligned with the European Sustainability Reporting Standards developed by EFRAG. The Task Force on Climate-related Financial Disclosures, whose recommendations were published in 2017, supplied the architecture that both regimes absorbed: governance, strategy, risk management, and metrics and targets.

A rating model can only use indicators that are reported consistently enough to be compared. In China, the disclosure base is thinner but growing. The number of A-share listed companies publishing annual corporate social responsibility reports rose from 565 in 2011 to 1,005 in 2020, a compound annual growth rate of 6.61 percent. That is progress, and it is still a minority of listed firms. The October 2014 decision by the former Ministry of Environmental Protection to suspend the environmental protection verification system for listed companies—the Notice on Reforming and Adjusting the Environmental Protection Verification System for Listed Companies (Huanfa [2014] No. 149)—removed a mandatory disclosure channel and shifted the burden to voluntary and market-driven reporting. The green securities working mechanism moved from a government-led model, reliant on administrative assessment, to a market-led model reliant on public information. When the state stops verifying, the market's verification instruments carry the load.

There is a warning in the European experience. As disclosure obligations multiply, firms face what scholars have called "over-transparency"—disclosing so much, in so many formats, that the disclosure becomes materially misleading. The mechanism is straightforward: when every indicator must be reported, the marginal cost of reporting each one falls, and firms report the ones that flatter them. A rating model that demands every indicator in the taxonomy will produce scores that reflect compliance effort rather than environmental performance. Eight indicators with consistent definitions across firms beat eighty with inconsistent ones.

4.1.4 Sector Calibration: Three Technology Segments

The technology sector is not one sector. Consider three segments, each with a different environmental profile and therefore a different indicator weighting.

Semiconductor manufacturing. High capital intensity, high energy and water consumption, significant chemical waste. Tier 1 indicators carry real weight. A fab's water withdrawal in a water-stressed region is a material operational risk, not a reputational one. The transition indicators matter less because the technology trajectory is relatively stable; the environmental risk is in operations, not in demand disruption.

Software and digital services. Low direct footprint, high indirect footprint through data centers, and a large enabling effect. Tier 3 indicators dominate. The credit-relevant environmental question is whether the firm's infrastructure contracts expose it to rising energy prices and carbon costs, and whether its products help customers abate. A cloud provider that has contracted renewable power at fixed prices has hedged a real cost exposure. A provider that has not carries an unhedged energy liability that a conventional rating would miss entirely.

Clean technology manufacturing. Solar, batteries, power electronics. The environmental case is the business case. Here the rating challenge is different: the firm's environmental performance is good by construction, so environmental indicators add little discriminatory power. What matters is execution risk—whether the firm can scale manufacturing, manage input costs, and survive the price cycles that have characterized solar manufacturing for two decades. Environmental indicators enter as a check on whether the firm's claimed benefit is real, not as a differentiator among competitors.

A single environmental score cannot be applied across the three segments. The model needs segment-specific weights, and the validation must be done segment by segment. A model that predicts default well in software and poorly in semiconductors is two models wearing one coat.

4.1.5 The Weighting Method

Weights can be set statistically, by expert judgment, or by anchoring to regulatory and market benchmarks. Each has a failure mode, and the failures are different enough that the three should be layered rather than chosen among.

Statistical estimation regresses default or rating migration on candidate indicators and retains those with explanatory power. This is the discipline that keeps the rating honest: an indicator that does not correlate with credit outcomes, however environmentally meaningful, does not earn weight in a credit model. The approach has a known weakness. Environmental performance data is short-history, and default events among technology firms with environmental ratings are rare. The regression will have wide confidence intervals, and weights estimated from it will be unstable.

Expert judgment supplies stability where the statistics are thin. Its weakness is capture—the weights reflect whoever sits on the panel. Benchmark anchoring connects the rating to the definitions that govern green bond eligibility and regulatory capital treatment; the EU Taxonomy's technical screening criteria, for instance, specify thresholds for what counts as a sustainable activity. Used alone, anchoring makes the rating a mirror of the taxonomy rather than a measure of credit risk.

My own view, formed from the technology-enterprise credit rating work that underpins this book, is that the three should be layered. Statistical estimation sets the weights where data permits. Expert judgment adjusts within a band around the statistical estimate, with the adjustment documented. Benchmark anchoring constrains the whole exercise so that the rating does not drift away from the definitions the market uses. The layered approach is slower than picking one method. It is also more defensible when a firm disputes its rating, which it will.

4.2 Rating Model Design and Validation

4.2.1 The Architecture: Two Scores, One Rating

The central design decision is whether environmental performance modifies the credit rating or sits beside it. Chapter 3 argued that the rating should be a single number, because a single number is what the capital allocation process uses. The single number should be constructed from two components: a financial credit score and an environmental adjustment, with the adjustment's magnitude capped so that environmental performance cannot single-handedly move a firm across the investment-grade boundary.

flowchart TD
    A[Firm data] --> B[Financial credit score]
    A --> C[Environmental performance score]
    A --> D[Environmental trajectory score]
    C --> E[Environmental composite]
    D --> E
    E --> F[Adjustment factor with cap]
    B --> G[Base rating]
    F --> G
    G --> H[Final rating]
    H --> I[Capital allocation decision]
    H --> J[Disclosure: rating plus environmental score]

The financial credit score is the conventional apparatus: leverage, coverage, liquidity, profitability, size, and sector risk. The environmental composite combines the performance and trajectory scores. The adjustment factor translates the composite into a notch adjustment on the rating scale, bounded at plus or minus two notches. The cap matters. Without it, a strong environmental score could lift a distressed firm into investment grade, which would be a failure of the rating's primary function. With it, environmental performance operates as a tiebreaker and a trend signal, which is what the credit question actually requires.

Why two notches? The number should be calibrated to the observed dispersion of default rates across notches. S&P Global Ratings' 2023 annual default study puts the average one-year global corporate default rate for BBB-rated issuers at roughly 0.15 percent over 1981–2022, against roughly 0.60 percent for BB-rated issuers. A two-notch adjustment can move a firm across that boundary; a three-notch adjustment could move a firm from BBB to BB+ and across the investment-grade line in a single step. Two notches keeps the adjustment inside the range where the environmental signal is credible and the default-rate gap is measurable.

4.2.2 The Environmental Trajectory Score

The trajectory score is the novel component and the one most likely to be gamed. I define it as a forward-looking assessment of the firm's alignment with a decarbonizing economy, built from four elements.

Target credibility. Does the firm have a public decarbonization target, and is the target validated by an external body? The Science Based Targets initiative provides one validation channel; national regulatory targets provide another. A target without validation is a statement of intent. A validated target with interim milestones is a commitment.

Capital alignment. What share of the firm's capital expenditure plan is directed at low-carbon assets or processes? This is the hardest element to verify and the most predictive. A firm that says it will decarbonize and continues to invest in carbon-intensive assets is not in transition; it is in narrative.

Revenue alignment. What share of revenue comes from products and services that are taxonomy-eligible or that enable customer abatement? For technology firms, this is where the enabling effect enters the score. A firm selling building-management software that reduces customer energy use has revenue alignment even if its own footprint is negligible.

Governance. Does the firm's board have environmental expertise, and is environmental performance linked to executive compensation? Governance is a leading indicator of whether the other three elements will survive a change in management.

Each element is scored on a four-point scale—absent, stated, funded, verified—and the four are combined with weights that emphasize capital alignment. The four-point scale is deliberately coarse. A finer scale would imply precision the underlying data does not support.

4.2.3 The Environmental Performance Score

The performance score is more conventional: Tier 1 physical indicators, normalized within sector, converted to a percentile rank, and mapped to a score. The normalization step is where most frameworks fail. Absolute emissions favor small firms; emissions per unit of revenue favor asset-light firms; emissions per unit of output favor neither. For technology firms, revenue-normalized intensity works best for the performance score, with an absolute floor: a firm below a threshold absolute emissions level receives the top performance score regardless of intensity, because its environmental footprint is not credit-relevant. This avoids the perverse outcome in which a small software firm with a single inefficient office scores worse than a large manufacturer.

The performance score enters the composite with a lower weight than the trajectory score for technology firms—perhaps 40/60—for the reason set out earlier: the sector's environmental risk is forward-looking. For heavy industry the weights would reverse.

4.2.4 Validation Design

A rating model is a hypothesis. Validation is the test, and the tests have a natural order.

Discriminatory power comes first. Does the model separate defaulters from non-defaulters better than the financial-only model? The standard measures are the area under the receiver operating characteristic curve (AUC) and the Gini coefficient. An AUC of 0.5 is a coin flip; 0.7 is acceptable; 0.8 is good. The test is not whether the enhanced model has a high AUC in absolute terms but whether it has a higher AUC than the financial-only model on the same sample. If the environmental variables add nothing, the enhancement is decoration.

Stability comes second. A model whose coefficients swing wildly from year to year is fitting noise. The test is out-of-time validation: estimate on data through year \(t\), test on year \(t+1\). For environmental variables with short histories, out-of-time validation is the only honest test available.

Calibration comes third, and it is the test most often skipped. Discrimination and calibration are different properties. A model can rank firms correctly and still misstate the level of risk. The test is a goodness-of-fit comparison between predicted and realized default rates across rating buckets, using the Hosmer-Lemeshow statistic or, for low-default portfolios, a binomial test on each bucket.

The data constraint is severe. A validation exercise on technology firms with environmental ratings will have few default events, because the population is young and the observation window is short. Statistical power is low. This does not make validation impossible; it makes it necessary to combine statistical tests with qualitative review and to treat the validation results as provisional. The alternative—declaring the model valid because it has not yet been tested—is worse.

4.2.5 The Overfitting Trap

The temptation in model design is to add variables until the in-sample fit is excellent. Environmental variables are especially prone to this because they are numerous and correlated. A model with thirty environmental indicators will fit the training sample beautifully and fail out of sample, because the indicators are measuring the same underlying construct with noise.

The discipline is parsimony enforced by cross-validation. Split the sample, fit on one part, test on the other, and retain only the variables whose contribution survives. In the technology-enterprise rating work, the pattern that emerged was consistent: three to five environmental variables carried nearly all the incremental explanatory power, and the rest added noise. The specific variables varied by segment, which is the sector calibration principle again.

There is a second trap: the model that is technically valid and operationally unusable. A rating model that requires data the firm does not report, or that an analyst cannot obtain within the rating timeline, will be bypassed. The model must be designed around the data that exists, not the data that should exist. This is a constraint, and it is also a discipline: it forces the designer to identify which environmental information is both available and credit-relevant, which is a shorter list than the disclosure enthusiasts assume.

4.2.6 A Worked Illustration

Consider a stylized rating of two firms in the power electronics segment. Both have similar financial profiles: leverage of 2.5x EBITDA, interest coverage of 4.0x, stable margins. The financial-only model assigns both a BBB rating.

Firm A has Scope 1 and 2 emissions of 8,000 tons CO₂e, energy intensity at the sector median, a validated science-based target for 2030, 60 percent of capital expenditure directed at wide-bandgap semiconductor capacity for grid applications, and 70 percent of revenue from products that enable customer efficiency. Its trajectory score is high; its performance score is median. The composite produces a one-notch upgrade to BBB+.

Firm B has Scope 1 and 2 emissions of 6,000 tons CO₂e, energy intensity below the sector median, no validated target, 15 percent of capital expenditure directed at low-carbon applications, and 30 percent of revenue from legacy applications facing demand decline. Its performance score is better than Firm A's; its trajectory score is poor. The composite produces a one-notch downgrade to BBB−.

The financial-only model rates B above A on environmental grounds, because B's current footprint is smaller. The enhanced model rates A above B, because A's transition is funded and B's is not. Over a five-year horizon, the enhanced rating is the better predictor of credit outcomes, because the credit question is whether the firm's cash flows survive the transition, not whether its current operations are clean.

This is the distinction the chapter opened with, made concrete. The current rating penalizes the firm investing in transition; the trajectory rating rewards it. The two ratings measure different things, and a financial system that can hold both in view can finance transition rather than merely reward cleanliness.

4.2.7 A Disputed Rating

The framework's credibility depends on what happens when a firm challenges its score. A case from the technology-enterprise rating work illustrates the mechanics. A mid-sized manufacturer of industrial power supplies received a two-notch downgrade on its environmental composite, driven entirely by a trajectory score of "stated" on capital alignment—it had announced a low-carbon product line but had not committed capital to it. The firm disputed the rating on the grounds that its current emissions intensity was in the bottom quartile of its segment. The dispute was resolved by re-examining the capital expenditure plan: the firm had, in fact, allocated 12 percent of its three-year plan to the new line, below the 25 percent threshold for "funded" but above the 5 percent threshold for "stated." The score held. The firm's response was to accelerate the capital commitment by two quarters, which moved it to "funded" at the next annual review and recovered one notch. That sequence—dispute, evidence, score held, behavior changed—is the mechanism working as intended. A rating that cannot be disputed is a rating that cannot be trusted; a rating that is disputed and revised without evidence is a rating that cannot be used.

4.3 Implications for Green Finance

4.3.1 From Rating to Capital Allocation

A rating is an input, not an outcome. The question that matters for green finance is whether the enhanced rating changes the allocation of capital. Two channels do most of the work: pricing and eligibility screens. Monitoring is a distant third, useful but slower.

Pricing. If the enhanced rating produces a different default probability, and the lender prices off the default probability, the spread changes. The magnitude is modest—a one-notch adjustment might move the spread by 10 to 30 basis points for a mid-sized technology firm—but the direction is what matters. A firm with a credible transition plan gets cheaper credit than an otherwise identical firm without one. That price signal is the mechanism by which the rating rewards transition.

Eligibility. Many green finance facilities—green bonds, sustainability-linked loans, policy-guided funds—use ratings or rating-like scores as eligibility screens. If the screen uses the financial-only rating, a transitioning firm with a weak current environmental profile is excluded. If the screen uses the enhanced rating, the firm can qualify on the strength of its trajectory. This is where the rating framework connects to the catalysis mechanisms that Chapter 10 will examine: sci-tech finance intermediaries need a way to identify which firms merit support, and the enhanced rating supplies it.

Monitoring. A rating is a recurring assessment. The enhanced rating, updated annually, tells the lender whether the firm's transition is progressing or stalling. A firm that commits to a target and then fails to fund it will see its trajectory score fall, and its rating with it. This converts the rating from a snapshot into a monitoring device, which is what a transition finance relationship requires.

4.3.2 The Green Index Connection

There is a market-structure channel as well. Green indices—baskets of securities selected on environmental criteria—have become a significant allocation mechanism. The claim made for them in the Chinese policy literature is that a valuable investment will bring long-term good returns, and the green index is an effective reference tool for guiding the direction of value investment. The claim is plausible and conditional: it holds if the index's selection criteria identify firms whose environmental performance is durable and credit-relevant, rather than firms that have mastered disclosure.

An enhanced rating framework improves the index construction problem. Instead of selecting on reported emissions—which favors asset-light firms and penalizes manufacturers—the index can select on the trajectory score, which identifies firms positioned for the transition. The result is an index that is both greener and more financially defensible, because it is selecting on the variable that predicts survival.

The connection runs both ways. The index creates demand for the enhanced rating, which creates an incentive for firms to report the data the rating requires, which improves the data available to the rating. This is a virtuous circle, and it is fragile: if the index selects on the rating and the rating is gamed, the circle becomes a bubble. The validation discipline described earlier is the safeguard.

4.3.3 What Ratings Cannot Do

The limits of the enhanced rating determine where the rest of this book's mechanisms become necessary.

A rating identifies; it does not buffer. A firm with a strong environmental trajectory still faces transition risk—technology failure, policy reversal, demand collapse—and the rating will register the risk without mitigating it. The mitigation mechanisms are insurance and diversification, which Chapters 6 through 9 address. The rating tells the lender which firms to lend to; it does not tell the lender what to do when the firm's transition fails.

A rating is also a point estimate, not a distribution. It says the firm's default probability is in a certain band. It does not say how much the lender loses if default occurs, which is the loss-given-default question. For green finance, loss-given-default matters because transition assets—a new production line, a retrofitted plant—may have thin secondary markets and low recovery values. The rating framework must be paired with collateral and recovery analysis, or it will overstate the safety of transition lending.

And a rating is a private assessment. The lender uses it; the public does not see it unless the firm discloses. The social return from green transition depends on the aggregate allocation of capital, not on any single lender's portfolio. A rating framework that improves individual lending decisions without changing the aggregate allocation has done less than its advocates claim. The aggregate effect depends on adoption—how many lenders use the enhanced rating, and whether the rating agencies incorporate it into their public ratings. That is a coordination problem, and it is the subject of Chapter 13.

4.3.4 The Regulatory Interface

The regulatory framework is moving toward the enhanced rating's logic, unevenly. The European Banking Authority, acting on the mandate in Article 98(8) of the Capital Requirements Directive V, issued recommendations in 2021 on how institutions should embed ESG risks into governance, strategy, and risk management. The recommendations call for "tone from the top," allocation of ESG responsibilities across the three lines of defense, and integration of ESG risks into the Internal Capital Adequacy Assessment Process. The October 2021 banking package proposed to convert some recommendations into binding rules, including an assessment of ESG exposures in the Supervisory Review and Evaluation Process and formal recognition of the forward-looking nature of ESG risks across short, medium, and long-term horizons.

The phrase "forward-looking nature" is the regulatory endorsement of the trajectory score. Supervisors have recognized that environmental risk is not captured by historical data alone, and that the assessment must extend across time horizons. That recognition creates the space for the enhanced rating, and it also creates a compliance burden: institutions will need to demonstrate that their ESG risk assessment is forward-looking, which means they will need trajectory data and a method for scoring it.

The European framework is not the only one. The IFRS S1 and S2 standards provide the disclosure baseline; the EU Taxonomy provides the activity definitions; national regulators provide the enforcement. The rating framework described in this chapter is designed to be interoperable with all three: it uses disclosed data, it references taxonomy definitions, and it produces a documented score that a supervisor can review. Interoperability here is a practical requirement, not a design virtue—a model that cannot be mapped to the taxonomy will not survive a supervisory review.

4.3.5 The Technology Firm as the Hard Case

Technology firms are hard because their environmental effects are indirect, their asset bases are light, and their business models change faster than rating models update. A rating framework designed for heavy industry will misrate them, and a framework designed for them will misrate heavy industry. The sector calibration principle is a condition of validity, not a refinement.

They matter because the green transition is, in large part, a technology transition. The abatement technologies—solar, batteries, power electronics, grid software, industrial efficiency—are produced by technology firms. The capital that finances them flows through the rating and lending system. If that system cannot see the environmental trajectory of a technology firm, it cannot finance the transition at the scale the transition requires.

The framework in this chapter is a first approximation. It has three components—a tiered indicator set, a two-score rating architecture with a capped adjustment, and a validation protocol—and each component has known weaknesses. The indicator set is constrained by disclosure. The trajectory score is gameable. The validation is data-limited. I have flagged these weaknesses rather than resolved them, because resolving them requires the data that adoption will generate.

The next chapter takes up the data question directly. A rating model is only as good as the data it consumes, and the data on green firms is sparse, inconsistent, and—in the Chinese market especially—shaped by disclosure rules that are still being written. Chapter 5 examines what data-driven scoring can and cannot do for green firms, and where the boundaries of the method lie. The framework described here is the destination. The data is the road, and the road is under construction.

4.4 Summary of the Framework

For reference, the framework's components and their design choices:

The table is a summary, not a substitute for the argument. The argument is that environmental performance enters a credit rating in two forms—performance and trajectory—and that for technology firms the trajectory form carries more credit-relevant information. A rating that sees only performance penalizes the transitioning firm. A rating that sees trajectory can distinguish transition from decline. The distinction is the difference between a financial system that rewards cleanliness and one that finances change.

The steel producer with the electric-arc furnaces is the test case. The current rating sees the cost of the furnaces and marks the firm down. The trajectory rating sees the asset the furnaces create and marks the firm up. Both ratings are true. A financial system that can hold both in view—that can price the transition rather than the snapshot—is the system this book is arguing for. The framework in this chapter is one instrument for building it, and its next test is the data Chapter 5 examines.