assembly-economics/checkpoints/IndicatorCatalog.md

Checkpoint indicator catalog

Status: v0.2 (2026-10-13). v0.1 held only the light and full lists. v0.2 adds a sensor bundle for every GADM stock (§ Stock sensor bundles), following metrics/MetricsCatalog.md. IDs L1–L7 and F1–F8 are unchanged because recorded checkpoints refer to them. Only indicators with an entry in sources.json are fetched automatically. All others are collected by hand at reviews.

Jurisdictions: US, EU (with Germany where available), Argentina (the owner's hypotheses: the EU lags, Argentina may thrive). Add others only with a reason.

Mapping caveat: these indicators are proxies. None of them equals the model quantity. Band positions are judged from several proxies together, never from one.

Light checkpoint: adoption, investment, sentiment, legislation

#	Indicator	Jurisdiction	Primary source	Model quantity (proxy for)
L1	Share of firms using AI to produce goods or services	US	U.S. Census Bureau, Business Trends and Outlook Survey (BTOS)	K_Ω adoption breadth
L2	Share of enterprises using AI technologies	EU, DE	Eurostat, ICT usage in enterprises (AI module)	K_Ω adoption breadth
L3	AI use by firms or workers	AR	INDEC / national ICT surveys, or OECD / IDB regional data	K_Ω adoption breadth
L4	Share of tasks or conversations by occupation; automation vs. augmentation	global, per country where published	Anthropic Economic Index; comparable usage indices	K_Ω task share, automation share
L5	AI-related investment (data-center capex, information-processing equipment) as share of GDP or of GDP growth	US, EU	BEA; Fed / central bank notes; company capex guidance	organism investment aO
L6	Public sentiment toward AI (benefit vs. harm, trust)	US, EU, AR	Pew Research; Eurobarometer; Edelman Trust Barometer; Latinobarómetro	sentiment Σ
L7	AI legislation milestones since the last checkpoint	US (federal and state), EU (AI Act implementation), AR	official journals; legislative trackers	legislation levers (permissiveness, m, expl)

Full checkpoint: additions (annual)

#	Indicator	Jurisdiction	Primary source	Model quantity
F1	Labour share of income	US, EU, AR	BLS / BEA, Eurostat, INDEC	claims breadth B, wage share
F2	Market concentration (top-firm sales shares, HHI) in digital sectors	US, EU	competition authorities, academic trackers	domination P (market, infrastructure channels)
F3	Leading-firm displacement / persistence	US	Bessen et al. series and updates	self-reinforcement r_k
F4	Time use: leisure, TV, social media	US (ATUS), EU (HETUS where available)	BLS ATUS; DataReportal	attention headroom A
F5	Informal care hours	US, EU	AARP/NAC; Eurofound	care load
F6	Displaced-worker re-employment by age	US	BLS Displaced Worker Survey (biennial)	senior agency modifiers
F7	Public R&D share of GDP; open-source investment initiatives	US, EU, DE	OECD MSTI; national data	commons investment IX
F8	Energy use of data centers	US, EU	IEA; national energy agencies	ecological draw of K_Ω (e2)

Stock sensor bundles

One bundle per GADM stock, with the four measure kinds from metrics/MetricsCatalog.md: O objective outcome, St structural, B behavioral, Su subjective. Most bundles reuse L and F indicators where they apply.

Column Access: api = machine-readable source that can get a sources.json entry; file = downloadable tables without a stable API; manual = report, paywall or registration, collected by hand.

Fetched automatically (sources.json, kind: full, verified 2026-10-13): H1 (US, DE, AR), H3 (US, DE, AR), H4 (EU, DE), K1 (US, EU, DE; AR as GDP per person employed), F1 (US; EU, DE via unit labour cost as a proxy), F-R1 (US), F-R2 (EU, DE), G4 (US quits rate), X2 (AR, open-access share = two counts), E1 (US, DE, AR), E2 (EU, DE), E4 (EU, DE). Everything else is collected by hand at reviews, even where a source is marked api. Probes failed for N1 (no FRED series; use the Census BDS tables) and Q1 (the AI Incident Database has no simple REST endpoint). X1 is reachable, but the API gives package counts, not reuse, so it is not wired.

Cadence: a annual, q quarterly, m monthly, irr irregular or multi-year survey waves. Bundles are collected at full (annual) checkpoints unless an L indicator already covers them.

H — Human capability

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
H1	O	Healthy life expectancy (HALE) at birth	US, EU, DE, AR	WHO Global Health Observatory	api	a
H2	O	Adult literacy, numeracy, adaptive problem solving	US, DE (no AR)	OECD Survey of Adult Skills (PIAAC cycle 2)	file	irr
H3	St	Public expenditure on education, % of GDP	US, EU, DE, AR	World Bank WDI; UNESCO UIS	api	a
H4	B	Adult participation in learning (last 4 weeks)	EU, DE	Eurostat trng_lfse_01	api	a
H5	Su	Perceived freedom of choice and control over one's life	US, DE, AR	World Values Survey; Gallup World Poll (via World Happiness Report)	file	irr

Gaps: no unassisted-capability measure (MetricsCatalog link #40). Argentina is not in PIAAC; PISA covers only 15-year-olds.

X — Assembly commons

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
X1	O	Reuse of open packages (dependents per package, top ecosystems)	global	ecosyste.ms; deps.dev	api	q
X2	St	Open-access share of research output	US, EU, DE, AR	OpenAlex	api	a
X3	B	Developers and git pushes per economy	US, DE, AR, EU members	GitHub Innovation Graph	file	q
X4	B	Maintainer continuity in critical projects	global	OpenSSF Scorecard; CHAOSS metrics	api	irr
X5	Su	Developers' ease of building on existing components	global, per country	Stack Overflow Developer Survey	file	a
F7	St	Public R&D share; open-source investment	US, EU, DE	OECD MSTI	api	a

K_phys / K_Ω — Productive capacity and organism capability

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
L4	O	AI task coverage by occupation; automation vs. augmentation	global	Anthropic Economic Index	manual	irr
K1	O	Labour productivity growth	US, EU, DE, AR	BLS; Eurostat nama_10_lp_ulc; ILOSTAT	api	q/a
K2	St	Installed supercomputing capacity per country	US, EU, DE	TOP500	file	semiannual
K3	St	Number of independent developers of notable or frontier models	global, per country	Epoch AI data; Stanford AI Index	file	irr
L5	St	Investment in information-processing equipment	US	BEA via FRED	api	q
L1–L3	B	Firm adoption of AI	US, EU, DE, AR	BTOS; Eurostat; manual for AR	api / manual	2-weekly / a
K4	Su	Workers' assessment of AI usefulness at work	US, EU	Pew Research; Eurofound	manual	irr

Report K together with B (F1) and P, as the MetricsCatalog requires.

F — Risk capacity

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
F-R1	O	Business applications (high-propensity)	US	Census Business Formation Statistics via FRED	api	m
F-R2	O	Business registrations and bankruptcy declarations	EU, DE	Eurostat sts_rb_q	api	q
F-R3	O	Employer firm births and deaths	AR	Ministerio de Trabajo, OEDE (Observatorio de Empleo y Dinámica Empresarial)	file	q
F-R4	St	Insolvency and business-entry regime	US, DE, AR	World Bank B-READY	file	a
F-R5	B	Total early-stage entrepreneurial activity (TEA)	US, DE, AR	Global Entrepreneurship Monitor	file	a
F-R6	Su	Fear-of-failure rate	US, DE, AR	Global Entrepreneurship Monitor	file	a

Gap: no source separates founders without inherited wealth.

G — Effective agency

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
G1	O	Intergenerational income mobility	US	Opportunity Insights	file	irr
G2	O	Re-employment of displaced workers by age	US	BLS Displaced Worker Survey (= F6)	file	biennial
G3	St	Data and service portability rules in force	EU, US, AR	EU Data Act, DMA; legislative trackers (with L7)	manual	irr
G4	B	Voluntary job switching: quits rate (US); job-to-job transitions (EU)	US, EU	BLS JOLTS via FRED; Eurostat LFS	api (US) / file (EU)	m / a
H5	Su	Perceived freedom of choice (shared with H)	US, DE, AR	WVS; Gallup	file	irr

Gap: restartability (time × cost × restrictions) has no direct source; it needs a composite built from F-R4, G2 and G4.

N — Diversity & composability

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
N1	O	Firm entry and exit rates (business dynamism)	US	Census Business Dynamics Statistics	file	a
N2	O	Diversity of model publishers (entropy of downloads by organization)	global	Hugging Face Hub API	api	q
F2	St	Concentration per component class	US, EU	competition authorities, trackers	manual	a
N3	B	Multi-vendor use of AI models and clouds by firms	US, EU	vendor and analyst surveys	manual	irr
N4	Su	Perceived lock-in	global	Stack Overflow Developer Survey; Eurostat cloud module	file	a

Gap: interoperability test pass rates have no public series.

Q — Epistemic capacity

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
Q1	O	Reported AI incidents and time to response	global	AI Incident Database	file	irr
Q2	St	Press freedom; freedom of academic and cultural expression	US, EU, DE, AR	RSF World Press Freedom Index; V-Dem	file	a
Q3	St	Concentration of news and information channels	US, DE, AR	Reuters Institute Digital News Report; StatCounter (search share)	file	a / m
Q4	B	Share of AI-mediated decisions with retrievable provenance	EU	AI Act transparency obligations (with L7)	manual	irr
Q5	Su	Trust in news and in AI-generated information	US, DE, AR	Reuters Institute Digital News Report; Edelman	file	a

Caveat: Q5 measures trust levels. Calibration (trust against reliability) needs Q1 or Q2 alongside. Disagreement between them is the finding.

P — Domination concentration (per channel)

ID	Kind	Channel	Indicator	Jurisdiction	Source	Access	Cadence
F2	O	market	Top-firm shares, HHI in digital sectors	US, EU	competition authorities	manual	a
P1	St	infrastructure	Cloud infrastructure market shares	global, regional	Synergy Research; Canalys (press summaries)	manual	q
P2	St	discovery	Search engine market share	US, DE, AR	StatCounter Global Stats	file	m
P3	St	membrane	Mobile OS and browser shares	US, DE, AR	StatCounter Global Stats	file	m
P4	St	lineage	Share of notable models from the top three developers	global	Epoch AI; Stanford AI Index (with K3)	file	irr
P5	St	political	Lobbying spend of large technology firms	US, EU	OpenSecrets; EU Transparency Register	api	q / a
P6	B	—	Self-preferencing and exclusion decisions	US, EU	FTC, DOJ, European Commission DMA decisions	manual	irr
P7	Su	—	Perceived ability to contest decisions of platforms	EU	Eurobarometer (when asked)	manual	irr

E — Ecological headroom

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
E1	O	Territorial CO₂ emissions	US, EU, DE, AR	Global Carbon Budget; Our World in Data	api	a
E2	O	Domestic material consumption per capita	US, EU, DE, AR	Eurostat env_ac_mfa (EU, DE); UNEP IRP Global Material Flows (US, AR)	api / file	a
E3	St	Electricity mix (low-carbon share)	US, EU, DE, AR	Ember	api	m
E4	St	Circular material use rate	EU, DE	Eurostat env_ac_cur	api	a
F8	B	Data-center energy use	US, EU	IEA; national energy agencies	manual	a

Su is empty by design (MetricsCatalog).

S — Social cohesion

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
S1	O	Political violence and protest events	US, AR	ACLED (registration)	manual	m
S2	St	Democracy and deliberation indices	US, EU, DE, AR	V-Dem	file	a
S3	B	Voter turnout in national elections	US, DE, AR	International IDEA	file	per election
S4	Su	Generalized interpersonal trust	US, DE, AR	World Values Survey; Latinobarómetro (AR)	file	irr / a
S5	Su	Trust in national institutions	US, EU, DE, AR	Pew; Eurobarometer; Latinobarómetro	file	a

A — Attention headroom

ID	Kind	Indicator	Jurisdiction	Source	Access	Cadence
F4	O	Time use: leisure, TV, social media	US, EU	BLS ATUS; Eurostat HETUS	api / file	a / irr
A1	O	Time use including unpaid care	AR	INDEC Encuesta Nacional de Uso del Tiempo (2021)	file	irr
A2	St	Rules on engagement design and default notifications	EU, US	DSA enforcement; state laws (with L7)	manual	irr
A3	B	Daily time on social media and on the internet	US, DE, AR	DataReportal (GWI survey)	manual	a
A4	Su	Reported time scarcity; perceived control over attention	US	Pew Research	manual	irr

Caveat: A3 measures captured attention. It is the inverse of A, not A itself (MetricsCatalog, Goodhart risk). F4 and A1 give self-directed time.

Coverage summary

Stock	O	St	B	Su	AR covered
H	H1, H2	H3	H4 (EU only)	H5	partly
X	X1	X2, F7	X3, X4	X5	X2, X3
K	L4, K1	K2, K3, L5	L1–L3	K4	L3 manual
F	F-R1–3	F-R4	F-R5	F-R6	yes
G	G1, G2 (US only)	G3	G4	H5	weak
N	N1, N2	F2	N3	N4	no
Q	Q1	Q2, Q3	Q4	Q5	yes
P	F2	P1–P5	P6	P7	P2, P3
E	E1, E2	E3, E4	F8	—	E1–E3
S	S1	S2	S3	S4, S5	yes
A	F4, A1	A2	A3	A4	A1 only

Weakest bundles: G (restartability), N (composability) and Q (calibration). Their readings should carry wide uncertainty until better sources exist.

Coupling indicators (ASSEMBLY-WP-0011-T06)

For the cross-border model (models/CrossBorder.md): foreign shares f per channel, the strength of host access rules, and frontier access.

ID	Indicator	Jurisdiction	Source	Access	Cadence
CB1	Search and mobile-OS shares by operator, with operator home	US, EU members, AR	StatCounter monthly CSV (chart.php?...&granularity=monthly&csv=1)	file	m
CB2	DMA effect: EU-minus-US change of the leading search share since March 2024	EU members vs. US	derived from CB1 (cross_border.py, DMA_EXCESS)	derived	q
CB3	Origin of deployed models: enterprise spend by developer; open-weight downloads by developer home	global, US, EU	Menlo Ventures, Ramp AI Index; Hugging Face Hub API	manual / api	irr / q
CB4	Cloud infrastructure share of non-domestic providers by region	EU, AR	Synergy Research regional releases	manual	q
CB5	Frontier access: share of firms using frontier models from abroad; export controls on chips and models	EU, AR	surveys; official journals (with L7)	manual	irr

CB2 is the observable for the access-rule strength a. At the 2026-10-13 review it gave a ≈ 0 (at most 0.006 per year).

Investment indicator for the regime filter (ASSEMBLY-WP-0017-T06)

ID	Indicator	Jurisdiction	Source	Access	Cadence
L5b	Private construction put in place, Office (includes data centers), SAAR	US	Census Construction Spending, https://www.census.gov/construction/c30/xlsx/release.xlsx, Table 1 (second "Office" row = private)	file (fixed layout; last 5 months + year-ago month)	m

L5b is closer to organism investment than L5, which covers all information-processing equipment. The regime filter's reliability rests on an investment indicator (research/ModelMatching.md § 4.2). Hyperscaler capital expenditure (company filings) is a second candidate, but the 2026 guidance figures conflict across sources and are not audited quarterly values.