Labelling official statistics
|Publication date:||15 October 2019|
|Author:||Good Practice Team|
|Approver:||Statistical Policy and Standards Committee|
|Who this is for:||Producers of statistics|
This guidance is for producers of statistical outputs across government who need to decide whether to label data as official statistics. The guidance covers:
- the legal background
- points to consider when deciding how data should be labelled
- good practice for releasing data which are not designated as official statistics.
The guidance has been drawn up to support data producers in government who need to decide whether to label their data as official statistics and may need to advise Ministers or senior officials about this.
Defining official statistics
The 2007 Statistics and Registration Service Act defines official statistics as:
statistics produced by the United Kingdom (UK) Statistics Authority, government departments (including executive agencies), the devolved administrations in Scotland, Wales and Northern Ireland, any other person acting on behalf of the Crown or any other organisation named on an Official Statistics Order.
Under the Act, official statistics should comply with the Code of Practice for Statistics. They also fall within the remit of the Office for Statistics Regulation. While the Act identifies bodies that can produce official statistics in Section 6, it does not set out which data should be classed as official statistics.
The UK Statistics Authority takes a pragmatic approach to this question. Put simply, if the Authority believes that Parliament would regard a body of information as “statistical”, it should be labelled and treated as such. Producers should be similarly pragmatic. Users outside government are very unlikely to see any meaningful distinction between official statistics and other numbers published by government – rather, they are likely to be confused and frustrated by such distinctions, and their trust in the numbers undermined, unless there are very clear reasons for them.
The most important reason for treating numbers as official statistics is to support and enhance public value. Whether the numbers support major decisions on policy or resource allocation, underpin accountability, offer insight into national debate, or simply because there is widespread public interest in the topic they address, official statistics that comply as far as possible with the Code of Practice for Statistics, are the most robust pathway to ensure trustworthiness and sufficient quality to deliver public value.
We recommend that data published by government bodies should be classified as official statistics unless there are very good reasons not to do so – for example because of serious concerns about quality or lack of transparency that mean compliance with the Code of Practice is problematic. When official statistics labelling is not appropriate, we recommend that producers consider voluntary adoption of the code.
Role of the Head of Profession for Statistics
For brevity, we use the term “Head of Profession” here to include statistical Heads of Profession in UK government departments, agencies and non-ministerial bodies, the Chief Statisticians of the devolved administrations and lead officials in arms-length bodies affiliated to government departments.
The Code of Practice for Statistics sets out the responsibilities of the Head of Profession in regard to labelling official statistics in Practices T2.1 and T2.3:
T2.1 The Chief Statistician or Head of Profession for Statistics should have sole authority for deciding on methods, standards and procedures, and on the content and timing of the release of regular and ad hoc official statistics. This should include determining the need for new official statistics, ceasing the release of official statistics, and the development of experimental statistics.
T2.3 As the principal adviser and accountable officer on statistical matters within the organisation, the views of the Chief Statistician or Head of Profession for Statistics should be considered in all matters relating to statistics and data.
General considerations when labelling official statistics
Official statistics can be derived from surveys and censuses, administrative or commercial sources (including management information), research and derived or modelled outputs. Modelled outputs, in turn, might include nowcasts, projections or forecasts.
For producers able to demonstrate fully that statistics meet the high standards of the code, it is appropriate to seek National Statistics designation through formal assessment by the Office for Statistics Regulation.
For forecasts and projections, management information or statistical research, producers should pay particular attention to how well the estimates comply with the quality pillar of the code and how limitations and assumptions feed through into trustworthiness, appropriate use and public value. Data like this are discussed in more detail below. For social research, producers should consult with Government Social Research (GSR) colleagues, including the departmental head of profession.
Producers should assess the following practical considerations in deciding when to use the official statistics label. While the first two are mandatory, it is not necessary for all of the others to be met. This is not a “pass or fail” checklist, and the importance placed on each consideration may change depending on the specific context.
While the list is designed to cover most eventualities, this is a matter of professional judgement for the Head of Profession, Chief Statistician or Lead Official. They can bring in other relevant considerations that have a bearing on the trustworthiness, quality and value of the statistics.
1. Mandatory: the producer is covered by the Statistics and Registration Service Act
The organisation that produced the statistics is within a Crown body, or named on an Official Statistics Order and so the numbers that they produce are within the scope of the Statistics Act. For producers that fall outside the scope of the Act, official statistics classification is not an option. Here, voluntary adoption of the code is a means to demonstrate commitment to trustworthiness, quality and value and so ensure that the statistics are presented in a clear, robust way to support appropriate use.
2. Mandatory: the statistics are put in the public domain
This could be through a regular output or an ad hoc release. Examples of ad hoc releases include statistics to support a ministerial statement or policy announcement, part of an answer to a Parliamentary Question or Freedom of Information request, statistical evidence for a Select Committee or included in a Ministerial speech, or a one-off report, in-depth study, supplementary analysis or other infrequent output.
It is not a requirement for data to be made available on a regular or recurring basis before they can be labelled as official statistics. When publishing for the first time, or in the case of an ad hoc release, the official statistics label may still be entirely appropriate. If the data are not going to be published, they cannot be classified as official statistics; however, the principles of open data should still be considered and voluntary adoption of the Code of Practice may be helpful.
The data are used publicly in support of major decisions on policy, resource allocation or other topics of public interest. This might include:
- monitoring government performance
- promoting and enabling accountability and transparency
- supporting the formulation, implementation or evaluation of policy
- serving the public good by helping citizens, businesses, academics, charities and the public sector, supporting effective decisions or supporting public insight
- information that is market sensitive
4. Public Interest
The data have a high public profile, attract controversy or debate when published and/or public debate would be better informed if they were classified as official statistics. The independent review of UK economic statistics recommended that when published data are likely to attract media or market attention they should be treated as official statistics and published in a manner that is compliant with the Code of Practice for Statistics.
5. Methods used
The data are collected and results compiled using widely accepted statistical methods that are appropriate for the intended use. Methods and assumptions (along with how and why they were decided) are fully documented and transparent. Where there is a standard, widely accepted methodology, this is adhered to unless there are good statistical reasons to depart from it. Where appropriate, independent experts advise on model assumptions and methods and their input is publicly available.
When improvements to official statistics are made, such as revisions to methods, they can be labelled as experimental statistics as they go through development and evaluation. This is the preferred approach when enhancements are made to existing official statistics, or new official statistics are produced, because such outputs are official statistics by definition and within the scope of the Code of Practice.
For statistical research and management information, producers should use their professional judgement when deciding whether to apply the official or experimental statistics label. Consider the nature of the outputs, maturity of the work, methods used and considerations of statistical quality, including how the data complement existing official statistics. Guidance on publishing management information and experimental statistics may be helpful, as well as regulatory guidance on experimental statistics from the Office for Statistics Regulation (OSR).
The data were produced by appropriately skilled analysts who are free from conflicts of interest, including political and commercial pressures. There is a commitment to transparency about methods, data and quality to allow for independent scrutiny, including the publication of assumptions and the justification for them.
Data inputs and statistical outputs are of sufficient quality to support the intended uses. Relevant limitations are assessed and addressed where possible, while assumptions and uncertainty are explained. Uncertainty is understood and presented, including details of the past performance of forecasts and projections when appropriate.
Some outputs require special consideration of quality and its implications. For example, forecasts and projections may be less than perfect, but the best that is achievable given the data and methods available. They may underpin important decisions despite having clear limitations.
Official statistics classification here ensures that estimates are presented with full transparency about methods and assumptions, are released in an orderly manner, that limitations are explained and that their implications for decision-makers are set out rigorously. This supports the public good.
At the same time, official statistics must not be materially misleading. Labelling data as official statistics when there is low confidence in their accuracy could lead users to assume an unwarranted level of certainty and lead to over reliance on the findings. Producers should take account of the level of uncertainty and how this impacts on use when balancing this risk. Consider voluntary application of the code , including full discussion of limitations and appropriate use when quality is a significant issue. When the quality of the data and methods are insufficient for official statistics classification, voluntary adoption will provide reassurance that the statistics are being produced in a trustworthy manner and are of sufficient quality to meet the intended uses, being clear about limitations.
The statistics are broadly representative of the population that they are designed to measure. Official statistics can focus on specific sub-groups, themes or geographical areas, but limitations in coverage and known biases should be explained, including how this affects responsible use.
Consider voluntary adoption of the code
For producers outside the scope of the Statistics Act, and for government data that are outside the scope of official statistics or do not meet the considerations in this guidance, we recommend voluntary adoption of the code. This demonstrates a commitment to trustworthiness, quality and value and supports appropriate use.
Guidance about voluntary adoption, including case studies, is available from the OSR website. Voluntary application comes in different shades, from near-full adoption and an in-depth statement of compliance to a basic summary of issues and considerations for users.
Producers should use their professional judgement to decide what is required to support users effectively in a given context. The Department for Work and Pensions, in its guidance about benefit expenditure and caseload statistics, Cabinet Office Race Disparity Unit in its statement of code compliance for the Ethnicity Facts and Figures website and Her Majesty’s Courts and Tribunals Service in a statement about voluntary application all use voluntary adoption of the code to build public confidence in their published management information and research. Her Majesty’s Treasury has committed to applying the code voluntarily in producing and publishing its budget data sources.
The Office for National Statistics (ONS) publishes International Passenger Survey quarterly estimates in response to a specific request from users. ONS supports the figures with a qualifying statement about why they are not of sufficient quality or coverage to be labelled as official statistics and what this means for effective use.
The Scottish Fiscal Commission (SFC) provides a helpful example of a forecast producer voluntarily adopting the code. Under current circumstances, SFC cannot become an official statistics producer. It is not named on the Official Statistics (Scotland) Order, and, as an independent fiscal institution, does not wish to seek such status from the Scottish Government. However, it has made an active choice to apply the code voluntarily wherever possible. It has published a statement of voluntary compliance setting out how it applies the Code of Practice.
Are experimental statistics also official statistics?
Yes. Experimental statistics are a subset of official statistics.
Can National Statistics also be labelled as experimental statistics?
No. National Statistics, by definition, are fully compliant with the Code of Practice for Statistics, while experimental statistics are not. If National Statistics require substantial revision, you must consult the OSR. The statistics may need to be reclassified as official statistics with the agreement of the regulator and, if appropriate, labelled as experimental statistics. OSR will also advise on how National Statistics status can be regained.
If a data output is not suitable to be labelled as official statistics, how should we label it?
Because outputs vary widely across departments, there is no definitive list of terms for data products that do not meet the requirements for official statistics labelling. Some departments have local policies on what to do, and you should check with your Head of Profession.
From a public value perspective, the main concern should be to support users so that they can work with these outputs effectively. The following terms are quite common:
Management Information (MI)
This term describes aggregated data collated and used in the normal course of business to inform operational delivery, policy development or the management of organisational performance. Several departments publish MI alongside their official statistics (making a clear distinction because of quality concerns) and use voluntary adoption of the code to support users. For more information can be found in the Publishing management information guidance.
Projections and forecasts
These products may be considered inappropriate for official statistics labelling, usually on grounds of quality and transparency, but may still be published. We recommend voluntary adoption of the code to draw out the issues and appropriate use cases for data that fall into this group.
Social research publications may fall outside the scope of official statistics – this will depend on context. We advise that producers consult with colleagues in the Government Social Research service, including the departmental head of profession, to agree the most sensible way forward for outputs of this type.
Statistical research is published to ensure transparency and enable users to feed into development. ONS statistical research is not classified as official statistics. For example ONS administrative population estimates use very new methods and data. They are too preliminary to meet the requirements for official or experimental statistics because of methodological and quality concerns. There is also a risk that they could confuse users or undermine existing official statistics on the same topic. Voluntary adoption of the code can be used in cases like this to draw out the issues and set out appropriate use cases.
Appendix – official statistics labels and what they mean
National Statistics are official statistics that have been assessed by the Office for Statistics Regulation as fully compliant with the Code of Practice for Statistics (that is, they meet the highest standards of trustworthiness, quality and value). A list of National Statistics is published by the Office for Statistics Regulation.
Section 6.1 of the 2007 Statistics and Registration Service Act defines official statistics as statistics produced by the UK Statistics Authority, government departments (including executive agencies), the Devolved Administrations in Scotland, Wales and Northern Ireland, any other person acting on behalf of the Crown or any other organisation named on an Official Statistics Order.
Under the Act, official statistics should comply with the Code of Practice for Statistics and fall within the scope of the Office for Statistics Regulation, which assesses their compliance against the Code of Practice.
Experimental statistics are newly developed or innovative official statistics that are undergoing evaluation. They are published to involve users and stakeholders in the assessment of their suitability and quality at an early stage.
Experimental statistics can fall into two categories:
- a new statistical product
- an existing statistical product that is undergoing transformation, for example through the introduction of new data sources or substantial changes to methods
When National Statistics are undergoing transformation, you should consult the Office for Statistics Regulation about whether National Statistics status should be removed so that the statistics can be labelled as experimental. They will also be able to advise on whether a compliance check or further assessment will be needed to reinstate it.
This guidance is reviewed anually.