Research standards
Data & Methods
Inspect the route from source to conclusion. Find the data behind a reading, how it was analysed, and where the evidence stops.
Evidence quality
Identify the original source, when it was retrieved, what it measures and whether it was verified. State rights and reuse limits alongside the citation.
Reproducibility
Record inputs, units, periods, transformations and analytical methods. Share supporting materials only where permission allows. A reproducible calculation does not establish causation.
Independent review
Before release, an independent reader checks the evidence, interpretation and confidentiality of the exact report and its derivatives. Uncertain material is excluded.
Uncertainty
Label observation, interpretation and forecast. Write Unknown with a reason. Describe scope, missing evidence and limitations where they affect the conclusion.
A global signal, a local decision
Version 0.1
Method
- Data: New York Fed GSCPI workbook independently acquired 10 October 2026; latest released observation September 2026. Acquired workbook: 345 ordered consecutive monthly observations, January 1998 to September 2026. Exhibit 1 and Table 2 show the latest 60, October 2021 to September 2026. The separately acquired release-CSV also includes September to December 1997, outside this bounded workbook analysis.
- Historical context: monthly levels are compared with the full-precision September 2026 value (1.284014886203148), inclusive of equality and of September itself. 33/345=9.57% describes months in the acquired data, not independent events or a future probability. April’s rank is 1+the number of strictly larger adjacent changes among 344; two are larger, no other change ties. Its rank within the displayed window uses 59 adjacent changes.
- Window test: each start compares its first twelve months with the same final twelve months, October 2025 to September 2026. Starts are October 2021 (60 months), October 2023 (36 months) and January 1998 (345 months). These selected starts test dependence; they are not an exhaustive sensitivity or revision study.
- Reproduction: two separately written calculation programs checked all 345 date/value pairs, 344 changes, counts, ranks and means; results agree within 1e−9 point. Both use the same xlrd workbook decoder and issuer source. Their agreement cannot rule out a shared decoding or source error or validate the issuer’s underlying estimates. Workbook versus release-CSV agreement remains checked only for the previously selected 60 rows at two decimals.
- Precision: display index levels, differences, means and historical shares to two decimals using round half up; calculate on original released values before rounding. Counts and ordinal ranks are exact integers. No percentage change of this zero-centred index is calculated.
- Provenance: the chart, table and calculations are original Chain of Insight derivatives of issuer data, not copied issuer exhibits. No original workbook, source paper, dataset or replication download is offered; conditional issuer permissions and notices apply to the published composite; no rights to proprietary component inputs are claimed.
- Source: Federal Reserve Bank of New York, Global Supply Chain Pressure Index. Issuer component credits: Bureau of Labor Statistics; Harper Petersen Holding GmbH; Baltic Exchange; IHS Markit; Institute for Supply Management; Haver Analytics; Refinitiv; issuer authors’ calculations. The GSCPI is not an official estimate of the Federal Reserve Bank of New York, its President, the Federal Reserve System or the Federal Open Market Committee. The issuer did not prepare or endorse this analysis. Source citation: https://www.newyorkfed.org/research/policy/gscpi. Issuer terms apply; the source ledger and public reuse notice retain applicable conditions and restrictions.
Limits
- One acquisition: recent estimates and earlier observations may be revised; no revision-size analysis was performed.
- The three starts are a bounded comparison, not evidence of a general trend, cause or forecast. Historical shares and ranks are not local decision thresholds.
- Both calculations share the original issuer data and xlrd decoder. Component estimates, current normalization weights and sample were not independently reproduced.
- No practitioner evidence, individual supplier or route outcome, operational recommendation or issuer endorsement. Conditional issuer permissions and notices apply; no proprietary input rights or issuer endorsement is claimed.
Supporting data
No separately cleared data downloads are available for this version.
Inspect sources and reuse limits