Scottish economics statistics: revisions policy

Introduction

This note describes the Revisions' Policy for the Scottish Government's economic statistics releases produced by the Office of the Chief Economic Adviser (OCEA). These include the quarterly Gross Domestic Product, Index of Manufactured Exports and Retail Sales Index for Scotland; the annual Government Expenditure and Revenue Scotland publication; the annual Global Connections Survey; and a number of other experimental releases.

This note augments the corporate revisions policy for all Scottish Government Official Statistics.

This note is a requirement of the Code of Practice for Official Statistics (January 2009). In particular, Practice 2.6 states "Publish a Revisions Policy for those outputs that are subject to scheduled revisions. Provide a statement explaining the nature and extent of revisions at the same time that they are released."

Revisions - explained

Statistics are generally subject to revisions. Revisions result from a reassessment of the past values of statistical variables. In general terms, revisions are mainly due to new data that could not be incorporated at the time of their previous release.

These new data may stem from either the inclusion of genuinely new information (due, for example, to delays in survey responses), or the correction or updating of information from data suppliers at a later date than the published earlier releases.

However, in addition revisions may also be determined by other factors such as:

• Conceptual changes (e.g. changes in nomenclatures and definitions);
• Improved methodology (e.g. changes in classification and stratification and weighting of sample survey data);
• Change in statistical data sources (e.g. replacement of survey data with data from administrative sources);
• Inclusion of additional observations (e.g. which may also affect time series estimates in different time periods, for example if corrections are made to remove seasonal factors);
• Errors in the computation of the given statistics.

Some revisions may be determined by a combination of a number of the above factors.

OCEA Statistics Revisions Policy

Planned Revisions

As part of continuous improvement in statistical methods, or due to the availability of more and fuller data, planned revisions are part of the production of statistical information. These can come in two main forms: ad-hoc revisions to methods or classifications, or scheduled revisions which are likely to happen regularly (e.g. each time a time series is updated). Unplanned revisions are rare and are covered for procedural consistency.

For ad-hoc planned revisions:

• We will consult users before making changes that affect statistics or publications.
• We will announce expected changes to methods or classifications well in advance of the release of the changed statistics.
• We will provide a statement explaining the nature and extent of the revisions at the time of publication.

For scheduled regular revisions:

• We will provide a statement explaining the nature and extent of the revisions each time the output is published with revisions. The first release may sometime be labelled "provisional estimates".

Unplanned Revisions

From time to time, revisions will need to be made outside the planned timetables. If revisions arising through improvements to methodology or changes to source data are found to be insignificant, they will be introduced in the next planned set of revisions according to the timetable above. However, if these revisions are thought to affect analysis or are sufficiently large, they will be introduced more quickly.

If incorrect data are discovered after publication, these too will be examined for their impact. Where the changes are significant, a correction will be issued as soon is practicable, while minor corrections will be included in the next planned release. In all cases a full explanation will be included as part of the release.

The course of action will be determined by the professional statistician and not influenced by any political pressure.

For unplanned revisions:

• When necessary, the same procedures apply as for planned revisions.

Examples of current practices

Quarterly Gross Domestic Product (GVA)

The quarterly Statistical Publication Notice (SPN) contains 7 tables of statistical analyses. These tables are replicated in each release showing the size of the revisions since the last quarter's release. These revisions will be both planned and ad-hoc. Revisions due to methodological changes are usually "stalled" and introduced once a year at the time of rebasing and re-weighting.

The SPN contains a summary of the most significant series changes in that publication. Additionally, it provides a brief statement about expected forthcoming revisions in the next release of the publication.

Quarterly Index of Manufactured Exports

The quarterly SPN contains 2 main tables of statistical analyses. These tables are reproduced to show the revisions since the previous quarterly release.

A graph showing the impact of the revisions is also provided.

Annual Global Connections Survey

This annual survey provides both official (not National) statistics for exports from Scotland to the Rest of the World, and statistics in development (experimental statistics) for exports to the Rest of the UK.

As the survey methodology has developed and more data have become available, there have been some significant revisions from one year to the next.

The size of the revisions is not mentioned specifically in the SPN, but may be deduced by comparing successive publications. Once the data are more stable, revisions analysis will be provided with the publication, and the scope of the survey may be revisited.

Annual Government Expenditure and Revenue Scotland (GERS)

This annual publication is compiled using numerous data sources and a range of apportionment methods to derive Scottish estimates from UK data. Both the UK data and the apportionment data are subject to revisions, many of which are made by other Government Departments.

High level revisions are detailed in the publication - details about revenue revisions are found in Annex A, and details about expenditure revisions are found in Annex B.

Quarterly Retail Sales Index for Scotland

This publication was first released on 10 February 2010. A table is provided in subsequent releases showing the revisions to the index numbers since the previous publication.

Annual Input-Output Tables

The Input-Output tables for Scotland are now compiled annually and as a consistent series of tables. As these are benchmarked to the annual Regional Accounts and other constraining variable, which are subject to annual revisions, the basic table and the analytical tables are subject to annual updates.

The tables, generally speaking, are in the form of huge matrices (123 rows x 123 columns). It is not practical to provide detailed revisions analysis for each of these tables. Users may, if required, derive revisions by comparing successive publications.

The tables are not used to analyse recent changes in the economy, so revisions are perhaps not so critical. The tables are however used to provide the weights for the short term indicators and to form the backbone of many economic models.

Scottish National Accounts Programme (SNAP) - Experimental Outputs

The Scottish National Accounts Programme (SNAP) products are, at this stage, experimental or statistics in development. Nevertheless, they are presented on the website as if they were official statistics and the revisions have been provided to show users how the estimates are converging to reliable series as the data sources and methodologies are developed.

SCOTTISH GOVERNMENT
OFFICE OF THE CHIEF ECONOMIC ADVISER