Skip to main content

Data issues

Likely errors as found in published source files

General limitations of the data

  • Short histories. Earnings growth, vacancies, public sector net borrowing, CPIH, core and services CPI, and the levels of employment and unemployment are held only from about September 2015. Nothing earlier has been found for them.
  • GDP vintages from 1992 to 1998. Published growth rates in the IHYQ table often differ from growth worked out from GDP levels at the same date. Only 40% agree to within rounding and 77% to within 0.15 points, with gaps of up to 0.5 points. This is not yet explained, so treat that stretch with caution.
  • Missing releases. Some releases are in no source collected. For example, the labour market levels have no vintages between September 2023 and February 2024. Check a series for gaps like this before relying on it.
  • Business surveys. Survey results come from samples of firms and are not comparable with official figures or with each other. Their histories are also short in real time: BICS results start in December 2021 and mostly May 2022, Credit Conditions Survey spreadsheets in January 2016, and Decision Maker Panel headline worksheets in late 2021 or 2022. Earlier survey periods first appear late, often as history.
  • BICS release dates. 20 of the 160 ONS editions carry the previous wave’s release date. These use the next release instead, so they can be about two weeks late.
  • Agents’ scores. Past versions of the file cannot be seen, so dates are estimates that can be up to about three months late, and any later correction by the Bank would be hidden.
  • Sources not yet used. ONS pages saved by the National Archives and the ONS API could fill in more of the years before September 2015.

Specific issues

While building this archive, we have found 7 errors in published files. They are listed here so that publishers can correct them and other users can avoid them.

Each entry says where the error is, what goes wrong if the values are used as published and what this archive does about it. The archive normally keeps numbers as published, so it corrects how a file is read but does not change the published values themselves.

Report errors or updates on GitHub Issues.

Bank of England

The year 2024 is missing from the corporate sheet of the 2024 Q1 annex

  • Dataset: Credit Conditions Survey
  • File published: 2024-04-11
  • Where: Corporate annex sheet, row of years, column BR
  • Problem: The newest column has the quarter label Q1 but no year above it, so it appears to belong to 2023.
  • Correct content: Column BR should be headed 2024.
  • Effect if used as published: Reading the years as written replaces the 2023 Q1 results with those for 2024 Q1, and delays the first appearance of 2024 Q1 by a quarter.
  • What this archive does: The archive works out each column’s year from the order of the quarter labels, so these values are stored against 2024 Q1.
  • File: ccs-2024-q1-annex.xlsx
  • Publisher’s page: Go to the page
  • Logged: 2026-11-02
  • Reference: ccs-2024q1-corporate-year

The year 2016 is two columns late in the 2019 Q2 annex

  • Dataset: Credit Conditions Survey
  • File published: 2019-07-18
  • Where: Secured annex, Unsecured annex and Corporate annex sheets, row of years, columns AL and AN
  • Problem: The label 2016 sits above the third quarter of 2016 instead of the first, so the first two quarters of 2016 appear under 2015.
  • Correct content: The label 2016 should be in column AL, above the first quarter of 2016.
  • Effect if used as published: Reading the years as written replaces the 2015 Q1 and Q2 results with those for 2016 Q1 and Q2, in all three sheets.
  • What this archive does: The archive works out each column’s year from the order of the quarter labels, so these values are stored against 2016.
  • File: 2019-q2-annex.xlsx
  • Publisher’s page: Go to the page
  • Logged: 2026-11-02
  • Reference: ccs-2019q2-misplaced-year

A year is mistyped in the corporate sheet of the 2015 Q4 annex

  • Dataset: Credit Conditions Survey
  • File published: 2016-01-15
  • Where: Corporate annex sheet, cell in column AD of the row of years
  • Problem: The row of years reads 2013 twice, so the four columns for 2014 are headed 2013.
  • Correct content: Column AD should read 2014.
  • Effect if used as published: Reading the years as written puts the 2014 results against 2013, and leaves 2014 with no results.
  • What this archive does: The archive works out each column’s year from the order of the quarter labels, so these values are stored against 2014.
  • File: 2015-annex-q4.xlsx
  • Publisher’s page: Go to the page
  • Logged: 2026-11-02
  • Reference: ccs-2015q4-corporate-year

Office for National Statistics

Annual total weekly hours worked are about 1,000 times too large

  • Dataset: Labour market statistics
  • File published: 2019-02-19
  • Where: Series YBUS, annual values, version 40 of the time-series page
  • Problem: Annual values are about 1,000 times too large, for example 1,037,545 for 2018. The series is measured in millions of hours per week.
  • Correct content: Version 41 of the page, which the archive dates 20 February 2019, gives 1,037.6 for 2018.
  • Effect if used as published: The annual values appear to be revised by about 1,000 times, twice, within two days.
  • What this archive does: Kept as published. Leave out annual YBUS rows dated 19 February 2019 when studying revisions.
  • Publisher’s page: Go to the page
  • Logged: 2026-11-02
  • Reference: ons-ybus-annual-scale

Growth in gross fixed capital formation for 2002 Q2 looks mistyped

  • Dataset: Revisions triangle for gross fixed capital formation
  • File published: 2002-08-31
  • Where: Series NPQT_QOQ in the archive, August 2002 vintage, value for 2002 Q2
  • Problem: The triangle gives growth of -5.5% for 2002 Q2. The next vintage gives -0.1%. This is our assessment, not a statement from ONS.
  • Effect if used as published: The figure looks like a large revision, and it is one of the largest gaps between published growth and growth worked out from levels.
  • What this archive does: Kept as published.
  • File: expenditure.zip
  • Publisher’s page: Go to the page
  • Logged: 2026-11-02
  • Reference: ons-npqt-2002q2-growth

Some older files hold mistyped survey dates

  • Dataset: Business Insights and Conditions Survey
  • Where: Dates worksheets in older files, waves 20, 55 and 135
  • Problem: The dates given for these waves are mistyped, for example the wrong year for wave 55.
  • Effect if used as published: Using the dates as written would place these waves’ results on the wrong days.
  • What this archive does: Every wave uses the date given in the newest file.
  • Logged: 2026-11-02
  • Reference: ons-bics-mistyped-wave-dates

Some editions carry the previous wave’s release date

  • Dataset: Business Insights and Conditions Survey
  • Where: Release date of 20 of the 160 editions read
  • Problem: The release date shown falls before that edition’s own survey had closed, so it cannot be right.
  • Effect if used as published: Using the date as written would show results as public before they could have existed.
  • What this archive does: These editions use the next ONS release date instead, with precision upper_bound, so they can be about two weeks late.
  • Logged: 2026-11-02
  • Reference: ons-bics-stale-release-dates