Measuring the Wealth of Nations - Replication Package

Complete replication and extension of every empirical claim in Shaikh & Tonak (1994), Measuring the Wealth of Nations: The Political Economy of National Accounts, plus eight follow-up studies. All 64 data series span 1947-2025 (book-period replication 1948-1989 plus extension where applicable) and are reproducible from the included loaders, processors, and validators.

Reconstruct, verify, and extend the empirical core of Shaikh & Tonak (1994) under modern provenance discipline:

• 100% of values trace to a published source, cached API response, or documented derivation;
• every series has a Data Provenance Record (DPR) and, where extended past the original book window, an Extension Provenance Record (EPR);
• every output passes V03 validators against book benchmarks and cross-source identity checks before publication.

The XS prefix supersedes the retired AS (analytical) and ES (external follow-up) prefixes; xs_class distinguishes the former analytical aggregates (appendix) from the follow-up-study series. Readers who referenced the legacy AS#### / ES#### IDs can map them to current XS#### IDs via MIGRATION/crosswalk.csv (see MIGRATION/PREFIX_SCHEME.md).

Follow-up studies covered (XS series):

• Tonak (1984) - labor share, net tax
• Shaikh & Tonak (1987, 2002) - productive labor refinements
• Moos (2017) - accumulation
• Mohun (2005, 2013) - exploitation rate, productive / unproductive labor
• Karabacak & Tonak (2022) - Turkey
• Cronin (2001) - New Zealand

# 1. Create a clean virtual environment
python -m venv .venv
source .venv/bin/activate         # Linux / macOS
# .venv\Scripts\activate          # Windows

# 2. Install dependencies
pip install -r requirements.txt

# 3. Run the full replication pipeline (Makefile and Docker variants
#    are also provided; see INSTALL.md)
python build.py status

# 4. Package-integrity check (registry + every shipped data CSV load)
python tests/ci_smoke.py

The tests/ directory carries the development regression + identity suite (95 pass / 2 skip / 2 xfail) as run against the internal build tree; tests/ci_smoke.py is the self-contained check that validates this published bundle on its own.

The orchestrator discovers and runs scripts in canonical phase order: S -> L -> P -> V -> M -> A -> O -> E (Setup -> Load -> Process -> Validate -> Manual -> Analysis -> Output -> Extend).

See INSTALL.md for environment setup details and (optional) API key configuration for fresh data fetches.

Publish/
|-- README.md                  # this file
|-- VERSION.txt                # bundle version string (v2.1)
|-- LICENSE                    # MIT (Code) + CC-BY-4.0 (Data)
|-- CITATION.cff               # machine-readable citation
|-- INSTALL.md                 # setup details
|-- requirements.txt           # Python dependencies
|-- pyproject.toml             # project metadata
|-- Makefile                   # convenience targets (replicate, validate, viz)
|-- Dockerfile                 # reproducible build container
|-- build.py                   # build orchestrator + status reporter
|-- series_registry.json       # single source of truth (64 series)
|-- DIVERGENCE_REGISTER.json   # intentional methodology divergences (73 entries)
|-- data/                      # 64 chopped CSVs (one per series, wide format)
|-- extenbooks/                # 64 Excel workbooks (4 sheets per series)
|-- docs/
|   |-- series/                # 149 DPR / EPR markdown docs
|   |-- decisions/             # framework decisions 0007, 0008
|   |-- methodology/           # methodology.md (replicator-facing narrative)
|   `-- precommit/             # pre-commit hook policy
|-- code/                      # L01 loaders, P02 processors, V03 validators, M04, A05, O06
|-- viz/                       # Plotly Dash / Shiny visualization apps
`-- tests/                     # 95 pass / 2 skip / 2 xfail regression + identity suite

• 64 chopped CSVs (data/) - canonical wide format (Row 1 metadata, Row 2 column IDs, Row 3+ data).
• 64 extenbooks (extenbooks/) - human-readable Excel workbooks with 4 sheets per series (Data / Provenance / Research / Construction).
• 149 per-series docs (docs/) - DPRs for every series, EPRs where extension applies, decomposition notes for compound series.
• 185 pipeline scripts (code/) - L01/P02/V03/M04/A05/O06 phase scripts plus shared lib/ helpers.
• Visualization app (viz/) - Dash + Shiny apps for interactive exploration.
• DIVERGENCE_REGISTER.json - every intentional deviation from upstream sources and predecessor methodology (73 entries).

Code is released under the MIT License (see LICENSE). Data outputs in data/ and extenbooks/ are released under Creative Commons Attribution 4.0 International (CC BY 4.0) - the data file itself, not the upstream sources, which retain their own terms (BEA, BLS, FRED, etc.).

If you use this replication package, please cite both this package and the original book:

Shaikh, A., & Tonak, E. A. (1994). Measuring the Wealth of Nations: The Political Economy of National Accounts. Cambridge University Press.

A machine-readable citation is available in CITATION.cff.

Continuous Integration installs this bundle into a clean Python 3.12 virtual environment on every push to main and on every pull request (see .github/workflows/replicate.yml) and runs a package-integrity check: the dependencies install, series_registry.json and the pipeline state load, and every shipped data/*.csv parses and is non-empty (tests/ci_smoke.py). A green build means the published package installs cleanly and its data layer is intact.

The full V03 validation against book benchmarks and cross-source identities (95 pass / 2 skip / 2 justified xfail; anu-doctor 0/0) is run in the canonical build tree, where the intermediate build inputs live — see DIVERGENCE_REGISTER.json, PIPELINE_STATE.json, and the per-series docs/.

This package is a published-outputs + provenance bundle, not a from-raw build tree. Concretely:

You can, offline, with only pip install -r requirements.txt:

• Install cleanly into a fresh Python 3.11+ venv (CI does this on every push).
• python build.py status — inspect the 9-stage pipeline state and gate marks.
• python tests/ci_smoke.py — verify the registry loads and every shipped data/*.csv parses and is non-empty. A green run means the package installs cleanly and its published data layer is intact.
• Read every value's lineage: series_registry.json (single source of truth), data/PROVENANCE_DICTIONARY.csv (per-observation), data/COMPONENT_CHAINS.{csv,json}, the 149 per-series docs/series/*_DPR.md / *_EPR.md, and DIVERGENCE_REGISTER.json (73 intentional divergences).

You cannot, from this bundle alone, regenerate the numbers from raw sources. The intermediate build inputs (data/final/, cached agency responses, the project checker) live in the maintainer build tree and are not shipped. Therefore:

• build.py chopped / build.py extenbooks and make build skip all series when data/final/ is absent — they are maintainer-side regenerators, not clean-room builders.
• The full V03 validation against book benchmarks (95 pass / 2 skip / 2 xfail; anu-doctor 0/0) is run in the maintainer tree, not here.
• Fresh BEA/BLS/FRED fetches require your own free API keys and network access (see INSTALL.md).

In short: this bundle lets you audit and trust the published data and its provenance; it does not let you rebuild the data from scratch. That is an intentional scope, and it is what a green CI badge attests to.

See methodology/methodology.md for the conceptual narrative. Per-series details live in docs/ (one DPR and, where applicable, one EPR per series).

• Chapter 7 labor-value series (S701, S702, S703) are now first-class implementations (proxy: false in the registry), built from BLS CES sectoral employment, BEA Benchmark I-O matrices, and the Appendix F productive-share filter. They replace the v1.0 matrix-structure proxies. Sector coverage is bounded by the eight Shaikh-Tonak productive sectors (Appendix G Table G.2); benchmark years without I-O coverage are reported as nan rather than interpolated. See DIVERGENCE_REGISTER.json entry DIV-011.
• Some series with status: book_period_validated cover only the book's original window (typically 1947-1989); extension to the present is documented in their EPR when applicable.
• All limitations are documented in per-series DPRs / EPRs and in DIVERGENCE_REGISTER.json.

• S517 (Productive Capital Stock K*) - net vs. gross. Shaikh & Tonak (1994) Table 5.8 labels K* as the "gross" private nonresidential capital stock, but the published book values match BEA Fixed Assets Table 4.1 Net Stock exactly at all 5 benchmark years (1948 = 292, 1958 = 551, 1967 = 871, 1980 = 3 800, 1989 = 6 700 $B). The net-capital column is retained as the headline series. New in v2.1: a book-faithful gross-capital variant S517-GROSS-A (with S513-GROSS-A / S514-GROSS-A) is now shipped for the book period 1948–1989 only — it reproduces the book's printed gross K* / r* directly (r* at MAE 0.0025) and gives the profit-rate family its first non-tautological book anchor. Beyond the mid-1990s BEA discontinued the current-cost gross stock for private nonresidential fixed assets, so the gross variant is nan after the book period (not synthetically extended). See DIVERGENCE_REGISTER.json entries DIV-008 / DIV-070 / DIV-058 and docs/series/S517_EPR.md Section 5 for the full divergence record.

Current: v2.1 (see VERSION.txt and CHANGELOG.md). An additive release on top of v2.0 — every pre-existing published cell is byte-identical; new content appends as new columns, new series arms, and new sidecars.

New in v2.1 — book-faithful profit rate, reconstructed exploitation, truth fixes:

• Book-faithful gross-capital profit-rate variants (S513-GROSS-A, S514-GROSS-A, S517-GROSS-A, book period 1948–1989): the profit-rate family's first non-tautological book anchor — the printed book r* is reproduced at MAE 0.0025 (42/42 years exact at 2 dp). The v2.0 net-capital headline series are retained unchanged (see DIVERGENCE_REGISTER.json DIV-070, DIV-058).
• Reconstructed time-varying I-O-uplift (kIO) rate-of-exploitation arm (S506-EXT-MARX-KIO): the frozen kIO = 1.5714 is replaced by an officially-sourced, backward-validated, uncertainty-banded annual series (kIO₁₉₉₂ ≈ 1.5696 → 1.731 by 2024; e ≈ 2.43 → 4.46). The frozen arm stays published as the conservative lower bound (DIV-071, DIV-028).
• Uncertainty-band sidecars under data/bands/: S506_KIO_BAND.csv, S701_LAMBDA_BAND.csv, S702_LAMBDA_BAND.csv.
• λ / p* labor-value series (S701, S702) publish at 3 significant figures (inside the tightest F3 sensitivity bound) instead of false 15–17-digit precision.
• Restored per-series author quotes for 34 published series (loader-visible quotes 110 → 348; zero invented or dropped), guarded by a new hard test.
• Truth fixes: zero unexplained book cells (S601/S603 1964 → DIV-072/073); validator honesty (5 tautological anchors re-anchored to independent sources → 0); XS002 de-hardcoded; the appendix_F/ filter dir renamed ch7_productive_filter/.
• DIVERGENCE_REGISTER.json: 73 entries (was 69); gates anu-doctor 0/0, pytest 95 pass / 2 skip / 2 xfail, per-series validators 60 PASS / 4 honest registered FAIL.
• New public artifact ANSWERS.md — an answers-first brief at the repo root.

Carried from v2.0 — a comprehensive series-by-series review (~50 series re-examined) and the D-batch of book-fidelity decisions:

• Comprehensive review corrections across ~50 series (provenance, units, content labels, and reference-value fixes), with the full audit trail recorded in series_registry.json and DIVERGENCE_REGISTER.json.
• D-batch decisions encoded: S514 rebuilt on the book's own division into r*/u components; the Chapter-7 K* series carry a replication-first primary column plus labelled variant columns; S515/S516 adopt the seam candidate (d); the net-social-wage extension uses the book-faithful Candidate A for 1952–1989; and XS1501–XS1504 are rebuilt on Mohun (2014) published shares.
• Series-ID migration (AS/ES → XS): the 4 analytical and 25 follow-up-study series adopt the canonical XS prefix (xs_class sectioning); legacy IDs map via MIGRATION/crosswalk.csv.
• Chapter 7 proxy retired: S701/S702/S703 are first-class labor-value series (proxy: false), built from BLS CES, BEA Benchmark I-O, and the Appendix F productive-share filter.
• Provenance deliverables shipped: per-observation data/PROVENANCE_DICTIONARY.csv (+ README and PROVENANCE_COVERAGE.csv), data/COMPONENT_CHAINS.{csv,json}, data/S506_STEP_TABLE.csv, and the master I-O concordance (data/concordances/).
• DIVERGENCE_REGISTER.json: 73 entries capturing every intentional deviation from upstream methodology.
• 95 passed / 2 skipped / 2 justified xfail test suite (tests/) plus V03 validators driven off year-keyed reference_values; anu-doctor reports 0 errors / 0 warnings.
• Reproducibility infrastructure: Dockerfile, Makefile, tests/, CI, and docs/precommit/ ship with the bundle so a clean replicator can docker build or make the entire pipeline.