Skip to content
RecessionIndicatorEconomic Observatory
Observatory / Research roadmap

Economic observatory

Research roadmap

Last updated

The current release separates an assessment of economic activity, leading warnings and evidence quality. A proprietary recession probability remains a research objective, with the actual test results published on the validation page.

Implemented in this release

  • Corrected lending-survey periods and separate fieldwork, release and retrieval dates.
  • Direct industry expectations from ONS BICS and US Census BTOS: 26 separately labelled feeds, with missing observations and sampling errors retained.
  • A corrected US historical test using histories available at fitting and inputs available at prediction; earlier failed studies remain accessible.
  • Longer UK and euro-area historical archives with source files and revision dates retained, plus separate quarterly evaluation results.
  • A homepage that leads with dated activity evidence and explains the forward outlook separately.

Near-term research is not yet a validated forecast

The expanded catalogue includes new labour, housing, expectations and credit cross-checks. Some connections contain verified histories; others still require source integration, research replication or authorised provider access. A listed indicator is not a completed live feed. Inspect the near-term collection and its connection status.

Separate one-, three- and six-month vintage-aware forecasting studies for these additions remain unfinished. The unemployment recession clock and district Beige Book sentiment methods have not been implemented as verified numerical signals. Additional entries do not automatically change the headline concern index or the frozen model inputs.

What still limits the forecast

The new industry panels supply useful services evidence, but do not constitute a weighted whole-services aggregate or replace the definitions of inaccessible PMI feeds. UK manufacturing has a direct turnover-expectations cross-check; current-model coverage rules and the frozen forecasting feature set remain separate.

The corrected US candidates still face recent-period and calibration failures. The UK and euro-area quarterly studies now report their own target definitions, eligible samples and failure reasons; neither has publication approval. Completing a connector or passing software tests does not confer probability approval.

Inspect the actual studies, event counts and regional limitations.

Next acceptance evidence

  1. Complete remaining historical continuity and target-definition checks, including the UK post-2016 ONS bridge.
  2. Extend the recorded automated reproduction with attributable external review of calibration, false alarms and performance across economic regimes.
  3. Register a model only when its exact procedure and target have attributable publication approval.
  4. Demonstrate dependable operation, including missed-update detection, source recovery, restore tests and an as-issued forecast record.
Operational and public-release requirements

Monthly observations can remain unchanged across four daily source checks. A stalled source or overdue brief must remain visible. A manual health-check pass is not evidence of continuous external monitoring, and no external alert destination is configured by this edition.

  • Document permitted use for each source, including storage, AI processing, display, downloads and redistribution.
  • Appoint an accountable economic editor and publish actual authorship, review and correction records.
  • Verify the public domain, access policy, security controls, operational monitoring and recovery procedure.
  • Keep probabilities withheld until the exact model, region, target and horizon have independent vintage-aware validation.

International access to US, euro-area and UK evidence is distinct from a global recession measure. Additional countries and languages need their own measurement and validation.