CSI methodology versions.
csi-v1.1.0, beside the preserved csi-v1.0.0.
Two changes define csi-v1.1.0: the trade sub-index is re-anchored so real change registers again instead of pinning at its rail, and the consumer score is redefined on official price statistics. The version was built and certified as a side-by-side candidate with its own proof chain, and activated only after every point of its gate passed; the csi-v1.0.0 series is preserved untouched as the historical record.
csi-v1.1.0 was certified on a side-by-side candidate track under the ADR-009 eleven-point activation gate, scoped by ADR-011 to the trade recalibration and the consumer redefinition. The comparison below replays the full series over 27 epochs from the declared candidate material, with direct-versus-composed parity at E8 at every epoch, a separate proof chain, the candidate-track certification gates, and the recorded activation. All values remain provisional until authenticated production lineage and the benchmark trusted gate pass; a missing or fail-closed artefact renders as an explicit unavailable state, never a fabricated value.
Methodology status
activated candidate-track certification: passed
ADR-011 (Reading A): csi-v1.1.0 = trade recalibration + consumer redefinition. csi-v1.1.0 is the published methodology, activated through the full ADR-009 eleven-point gate with the recorded owner pre-authorisation and a fully green adversarial confirmation audit. The csi-v1.0.0 series is preserved untouched as the historical record: its locked proof chain is unedited and its final publication is archived verbatim. Parity over 27 epochs: maximum direct-versus-composed difference 0 E8 (all passed).
| Period | csi-v1.0.0 (locked) | csi-v1.1.0 candidate | Difference | Trade sub-index, v1.0.0 → candidate | Consumer sub-index, v1.0.0 → candidate |
|---|---|---|---|---|---|
| 2000 | 101.517 | 102.387 | +0.870 | 101.25 → 110.00 | 100.32 → 100.42 |
| 2001 | 101.424 | 103.118 | +1.694 | 102.34 → 120.00 | 100.46 → 99.09 |
| 2002 | 100.867 | 103.532 | +2.665 | 101.40 → 130.00 | 100.28 → 96.09 |
| 2003 | 99.498 | 103.355 | +3.857 | 98.91 → 140.00 | 99.96 → 94.60 |
| 2004 | 98.165 | 102.532 | +4.367 | 95.89 → 141.30 | 99.64 → 96.41 |
| 2005 | 96.430 | 101.249 | +4.819 | 90.99 → 140.17 | 99.25 → 97.96 |
| 2006 | 94.966 | 100.024 | +5.058 | 86.23 → 138.54 | 98.82 → 95.77 |
| 2007 | 93.951 | 99.219 | +5.268 | 81.23 → 136.09 | 98.37 → 94.24 |
| 2008 | 92.621 | 97.970 | +5.349 | 76.23 → 132.60 | 97.98 → 92.13 |
| 2009 | 91.480 | 97.007 | +5.527 | 71.23 → 129.79 | 97.67 → 90.84 |
| 2010 | 90.165 | 95.998 | +5.833 | 66.23 → 127.03 | 97.50 → 92.80 |
| 2011 | 88.843 | 95.034 | +6.191 | 61.23 → 125.30 | 97.34 → 93.50 |
| 2012 | 87.623 | 94.088 | +6.465 | 56.23 → 122.51 | 97.06 → 94.60 |
| 2013 | 86.265 | 92.825 | +6.560 | 51.23 → 118.96 | 97.02 → 93.36 |
| 2014 | 84.902 | 91.599 | +6.697 | 46.23 → 115.26 | 96.88 → 93.43 |
| 2015 | 83.397 | 90.415 | +7.018 | 41.23 → 113.41 | 96.78 → 93.53 |
| 2016 | 81.837 | 89.204 | +7.367 | 36.23 → 112.13 | 96.74 → 93.02 |
| 2017 | 80.058 | 87.661 | +7.603 | 31.23 → 109.61 | 96.79 → 92.83 |
| 2018 | 78.729 | 86.528 | +7.799 | 26.23 → 106.57 | 96.48 → 92.55 |
| 2019 | 77.577 | 85.520 | +7.943 | 21.23 → 103.74 | 95.97 → 90.27 |
| 2020 | 75.523 | 83.836 | +8.313 | 16.23 → 102.24 | 95.87 → 90.76 |
| 2021 | 73.190 | 81.858 | +8.668 | 11.69 → 100.45 | 95.49 → 92.46 |
| 2022 | 71.299 | 80.045 | +8.746 | 11.58 → 100.39 | 95.13 → 93.93 |
| 2023 | 69.371 | 78.361 | +8.990 | 9.17 → 99.12 | 94.80 → 96.93 |
| 2024 | 69.143 | 77.814 | +8.671 | 9.17 → 95.00 | 94.80 → 98.57 |
| 2025 | 68.532 | 76.887 | +8.355 | 9.17 → 91.61 | 94.80 → 98.33 |
| 2026 | 68.099 | 76.146 | +8.047 | 9.17 → 89.58 | 94.80 → 97.55 |
What changes, and why
Re-anchored: baseline window 2000-2004 -> 2018-2022 and per-epoch cap 0.050 -> 0.100. The locked anchor sits a 2.9x volume era ago, which pins the sub-index at its rail (9.17 by 2023, 0 by 2025) regardless of real change; against the modern anchor the same data reads 99.12 (2023) and stays rate-sensitive. The current factors come from the quantity-consistent UN Comtrade bridge (HS6 leaf net weight, 18 of 19 G20 reporters).
Redefined from a stalled plant-protein-share construction to an official-statistics relative-price score: -1 * ln(plant basket / animal basket) * expenditure scaler, from Eurostat HICP and US BLS CPI/CE, pooled over the non-overlapping blocs EU27, Türkiye and the United States, replayed from real price history over all 27 epochs. The v1.0.0 commercial sources are replaced by official ones.
Proof anchors
Candidate proof chain: 31 blocks, head sha256:9badc001f0085db73d77c59549c92e83af2db5f9f884626bcfd588fd244f285e. Certification certificate sha256:7e7625988c164a05a064827ad7d2ffe71cfa7e8139beb15e4525a4fba6a3c826.
Both anchors are re-derived from the committed candidate evidence at build time: the chain head is the hash of the proof chain’s head block, and the certificate hash is the hash of the certification artefact’s own body. A reader can recompute either from the committed files; neither is copied from a publication field.
config hash: sha256:5e20924ef908fcefa5c1ba617ae6040b6a63f9701086b3f6b62a621f5355395b baseline hash: sha256:6db8d8a9a464f10822dfe1653c1ab94a14c7d7eaf7886a34eb8e6719f208118f source registry hash: sha256:31c3c2f3ecda6633f79d7d817673c003bc7917c140e74019433c80b204308099 metric unit registry hash: sha256:bfff78e5327a5e1be11cb9b9d74a6c923532fe99ed61544ffd932e96dda76711
candidate-track certification: the four release gates of unpt_csi.sub_indexes.audit.certify_composition_release (E8 parity, eight canonical members, canonical-only, determinism-by-re-derivation) evaluated over the committed candidate artefacts. This does NOT mint any public verified-live status; the production trusted gate and the locked v1.0.0 chain are untouched.
A candidate is a labelled successor, never an edit.
The locked csi-v1.0.0 methodology bytes are hashed into every published artefact and are never edited; the candidate carries its own config, baseline, source and metric-unit registries, its own replay evidence and its own proof chain.
Activation required every point of the ADR-009 candidate gate to pass against the ADR-011 scope, with the owner's recorded approval. The served static dashboard, proof ledger and provisional reference value use csi-v1.1.0 candidate artefacts; this does not establish a running production service. The csi-v1.0.0 series is preserved untouched as the historical record.