Browse by subject:
System Asset Pricing Model · Public Method

Measure the missing coordinate.

SAPM estimates the annual system-welfare loss associated with an activity and compares it with annual revenue from that same activity. This page provides the route from a decision question to a reviewable welfare signal.
Choose the correct route

A domain name alone does not establish a match.

Before importing a published welfare beta, compare six facts: the activity, harm mechanism, geography, period, system boundary, and revenue denominator.
Route 1

Existing-domain match

Use a versioned registry estimate when all six facts match. Import its uncertainty and evidence status with the estimate. A point value without its boundary is incomplete.
Route 2

Adjacent-domain recalibration

Use the registry as a starting benchmark when the mechanism is related but one or more boundaries differ. Replace the affected inputs and test whether the result survives.
Route 3

New-domain study

Build a new source ledger, revenue boundary, uncertainty model, overlap review, and evidence card when no defensible match exists.
Important measurement boundary

An industry average is not automatically a transaction estimate.

Multiplying an industry-average βW by the revenue from one deal assumes that the deal produces welfare loss in the same proportion as the wider industry. State and test that assumption. Use a marginal or decision-specific estimate when the evidence permits one.
The average measure answers an accounting question: how much annual system-welfare loss is associated with each dollar of annual revenue across the defined domain? A marginal measure answers a decision question: how much additional welfare loss is expected from one more unit of activity? They may differ materially.
Open cookbook

Fourteen steps from question to evidence card.

  1. 01
    Define the decision
    State the activity, proposed change, affected system, geography, period, and decision-maker before collecting numbers.
  2. 02
    Search the registry
    Check the 61-domain table for a candidate match. A familiar industry name is not enough.
  3. 03
    Test the boundary
    Compare activity, harm mechanism, geography, time period, system boundary, and revenue denominator.
  4. 04
    Choose the route
    Record the study as an exact match, an adjacent recalibration, or a new-domain study.
  5. 05
    Build the harm ledger
    List annual welfare-loss channels separately. Retain the affected population, units, source period, and valuation rule.
  6. 06
    Separate losses from transfers
    A payment moving from one party to another is not automatically a net welfare loss. Record both when they differ.
  7. 07
    Match annual revenue
    Use annual revenue from the same activity and boundary. Profit, market value, transaction volume, and unrelated revenue are different quantities.
  8. 08
    Record every source
    Preserve the URL or DOI, page or table, quoted support, geography, population, unit, date, extraction method, and transformation.
  9. 09
    Represent uncertainty
    Derive ranges and distributions from the evidence. Do not invent narrow bounds because the calculation needs them.
  10. 10
    Correct overlap
    Identify shared causal pathways, duplicated populations, correlations, and cumulative stocks before adding channels.
  11. 11
    Run and challenge the model
    Report the central estimate and interval, then test alternative defensible assumptions and distributions.
  12. 12
    Apply admission gates
    Withhold decision-grade status when sources, boundaries, denominator, overlap, reproducibility, or independent review remain incomplete.
  13. 13
    Export the evidence card
    Publish the estimate with its version, scope, uncertainty, source bundle, review status, expiry date, and reconsideration triggers.
  14. 14
    Use it in the decision
    Show the system-welfare estimate beside the parties’ gains and record how it affected approval, redesign, pricing, or review.
Evidence status

The label tells the reader what the estimate can support.

Every output should carry one of these statuses. Movement between them requires evidence, review, and recorded gates.
Level 1

Screening

A fast, source-linked indication that the decision may carry material system-welfare consequences. It is suitable for triage.
Level 2

Provisional

A reproducible calculation with stated boundaries and uncertainty. Material source or review work remains open.
Level 3

Decision-grade

Sources, annualization, denominator, overlap, sensitivity, and reproducibility have passed the declared review gates.
Level 4

Independently audited

A separate reviewer has reconstructed the calculation from the evidence receipts and recorded the result.
AI-assisted research

What the agent does, and what the evidence must still establish.

An AI system can assemble a first-pass evidence packet, compare domains, write source receipts, run simulations, and identify missing fields. Consequential use requires review of the evidence and the boundaries.

Agent-readable inputs

  • Structured decision and boundary description
  • Versioned domain-registry record
  • Channel-level evidence receipts
  • Declared distributions and dependence assumptions
  • Admission and falsification tests

Required review record

  • Source support and extraction checked
  • Stocks, flows, transfers, and overlap resolved
  • Revenue matched to the activity
  • Sensitivity and alternate assumptions reported
  • Independent rerun recorded when required
GDSS integration

A decision system receives the estimate and its conditions.

The portable output includes the domain and version, match type, system and revenue boundaries, central estimate and interval, evidence status, source-bundle identifier, review status, validity period, and reconsideration triggers.
{
  "domain_id": "canonical-domain-id",
  "estimate_version": "1.0.0",
  "match_type": "exact | adjacent | new",
  "system_boundary": {},
  "revenue_boundary": {},
  "beta_w": { "median": null, "lower": null, "upper": null },
  "decision_specific_delta_w": {},
  "evidence_status": "screening | provisional | decision-grade | audited",
  "source_bundle": {},
  "independent_review": {},
  "valid_until": null,
  "reconsideration_triggers": []
}
Open method · maintained evidence

The research method remains public. Applied work is available as a service.

The public materials are intended to support independent use, criticism, replication, and software integration. Organizations can engage the Center for new measurements, decision-specific calibration, independent review, and implementation.
Open methods make the work testable. Maintained evidence, decision-specific measurement, independent review, and implementation make it usable in consequential decisions.