Writing about statistics

This content is currently under review. Please bear with us while we make updates.

Policy details

Metadata item Details
Publication date:27 November 2019
Author:Good Practice Team
Approver:Good Practice Team
Who this is for:Producers of statistics
Type:Guidance
Contact:

gsshelp@statistics.gov.uk

Review frequency:

Annual

Updates:

25 September 2020 - Updates related to accessibility and publishing in HTML

  • If you would like us to get in touch with you then please leave your contact details or email gsshelp@statistics.gov.uk directly.
  • This field is for validation purposes and should be left unchanged.

Things to be aware of

Your department may have standards that you have to work to when it comes to writing about statistics, particularly when it involves writing content for webpages.

This guidance is a general set of advice and guidelines for any and all statistical commentary. In the future, producers from across the Government Statistical Service (GSS) may produce different products for different users. If this happens, each different product will have its own unique approach to statistical commentary to fit the intended user that it is aimed at.

This guidance replaces the following:

 

Back to top of page

Introduction

Statistical commentary is required to bring numbers to life.

Commentary should do much more than just describe the statistics in words. It should help the user to understand the meaning of patterns, trends and limitations, and build on any factual and public information already known about the subject matter.

Clear, insightful and professionally sound commentary supports informed decision-making and democratic debate.

Statistics and data should be presented clearly, explained meaningfully and provide authoritative insights that serve the public good.

Code of Practice for Statistics, United Kingdom (UK) Statistics Authority, 2018

What is good commentary?

Good commentary draws attention to important findings, puts them in context and provides a clear take away message for users. It supports and enables the appropriate use of statistics. It clearly explains issues of quality and reliability, how these impact on the use of the statistics and the conclusions that can be drawn from them.

Good commentary opens up the statistics for re-use.  It ensures users fully understand the nature of the statistics and the top-line results they should be able to reproduce if undertaking further analysis.

Who is this guidance for?

This guidance has been developed for producers of commentary about official statistics. The guidance may also be helpful for others who produce and report on statistics.

The guidance has been developed by the Government Statistical Service Good Practice Team.

 What is the aim of this guidance?

The aim of this guidance is to help producers to write statistical commentary that provides insight, and is impartial, helpful and accessible to a range of audiences.

What does this guidance cover?

This guidance is not a set of standards, but rather provides a common approach for writing about statistics, drawing on recognised good practice.

We look at how to present a full picture of the subject, and how factors like structure and language can impact upon the messages that readers take away. We discuss the importance of considering the users of the statistics when writing commentary. We also explore how to convey to users what the statistics mean in practice, whilst keeping the commentary objective and impartial.

Back to top of page

Top tips for writing about statistics

  • Understand the users and uses of your statistics – find out who uses your statistics and what the statistics are used for
  • Use plain language – balance the need for technically exact but complex terminology and clarity
  • Put the statistics into context – provide neutral and impartial information about users and uses, strengths and limitations, other relevant statistics, long-term trends and changes, geographical comparisons, and why the statistics have been collected.
  • Present main messages clearly and concisely – don’t try to summarise all of the findings in the publication: focus on the main points of interest
  • Tell users about quality and methods – be upfront and specific about caveats that are vital to interpretation. Provide links to more detailed information that some users may want.
  • Explore patterns, relationships, causes and effects.
  • Help users find the information they need – a contents page can be helpful for longer releases
  • Write clear and informative titles
  • Use structure to tell the statistical story
  • Consider the online experience – think about how users access information online
  • Think beyond statistical bulletins
Back to top of page

Understand the users and uses of your statistics

A sound understanding of the users and uses of your statistics is essential to delivering effective commentary. Effective commentary caters well for different audiences.

Find out who uses your statistics and what the statistics are used for

The audience for statistics is diverse. Your commentary will be most useful if you have a clear understanding of your users and how they draw value from the statistics. What decisions are taken and arguments made? Are the statistics re-used in further analysis or publications? How do different results and levels of quality affect users’ actions?

Research might be required to identify the wide range of users and uses of the statistics. There is guidance for working with users on the GSS website.

Ensure commentary is accessible to all

Government bodies have a legal obligation to make publications accessible to all.  Avoid barriers to accessibility such as small fonts, colour contrasts that are hard to distinguish and complex walls of text.

Think about the requirements of different audiences, including people with disabilities. The Government Digital Service blog post “Writing content for everyone” is very helpful.

Engage with users online

StatsUserNet is an interactive website for communication between users and producers of official statistics, hosted by the Office for National Statistics. With over 3,000 individual members, it is a well-established forum for online user engagement. Make an effort to monitor the site regularly and engage with users’ posts and discussions.

Get feedback from users

When writing about statistics, it is easy to become too close to the process and unable to judge whether content is accessible, understandable and valuable. A second opinion is usually helpful.

Possible approaches:

Ask a colleague or non-specialist in your department to peer review your writing, placing themselves as a lay reader without your expert knowledge.

Consider inviting wider peer review, either through a group inside your department, or initiatives like the Government Statistical Service’s ‘scrum’ programme. Getting a perspective from outside your department can be very valuable.

Ask your users for feedback. Do they find the commentary easy to understand? Are the main messages clear?

Think about what users are trying to achieve

A ‘jobs to be done’ approach can provide a useful, simple and quick way of gaining user insight. Think about what users are trying to achieve with your statistics, not simply their inherent characteristics.

The Office for National Statistics (ONS) developed a set of user personas, based on research done with users of the ONS website.  Personas can help us think about how to present and tailor commentary for different types of user. The ONS research identified five user personas:

  • Expert analysts
  • Information foragers
  • Inquiring citizens
  • Technical users
  • Policy influencers

Example

This NHS digital publication: “Health and Care of People with Learning Disabilities” has a summary report, an easy read version and data files to address these different user personas.

Back to top of page

Put the statistics into context

Help users understand the statistics in the context of the wider world: the economy, society, or the environment.

Be clear on what your statistics can and can’t be used for

Depending on the type of publication you are working on it might be useful to discuss the users and uses of the statistics. You might also describe the types of decisions that can be made using the statistics. This information demonstrates the relevance and public value of the statistics.

Mention both the known and likely uses. It is acceptable to make assumptions about what the statistics might be used for.

Example

“The Summary Hospital-level Mortality Indicator (SHMI) is not a measure of quality of care. A higher than expected number of deaths should not immediately be interpreted as indicating poor performance and instead should be viewed as a ‘smoke alarm’ which requires further investigation. Similarly, an ‘as expected’ or ‘lower than expected’ SHMI should not immediately be interpreted as indicating satisfactory or good performance.”

NHS Digital, March 2018, Summary Hospital-level Mortality Indicator (SHMI)

Discuss the findings in the context of long-term trends and changes

Don’t focus solely on the latest numbers, or on point-to-point comparisons in isolation. Instead, give the overall picture, drawing attention to individual movements only where they add value to the story.

Do not report on changes without discussing the context

For example, if you report a 2% rise, help the user understand whether this is typical or unusual in comparison to previous statistics, to other countries, or in respect to policy targets.

Example

“The Creative Industries accounted for 9.1% of all UK services imports in 2016, the highest proportion contributed since 2010. The proportion of the UK total contributed by the Creative Industries has generally been growing since 2010, with the change in the contribution between 2015 and 2016 (up 2.0 percentage points) being higher than usual.”

Department for Digital, Culture, Media and Sport, June 2018, DCMS Sectors Economic Estimates 2016: Trade

Example

“In 2018 the UK farmland bird index was 45% of its 1970 value. The majority of this decline occurred between the late 1970s and the 1980s largely due to the negative impact of rapid changes in farmland management during this period. The decline has continued at a slower rate more recently; the smoothed index significantly decreased by 6% between 2012 and 2017.”

Department for Environment, Food and Rural Affairs, November 2017, Wild Bird Population in the UK, 1970 to 2016

 

Make, or enable users to make, geographical comparisons

Comparisons may be made between regions, countries in the UK or internationally. Establish where equivalent data and publications are held. Comment on these and include links to the relevant websites, where appropriate. If there are differences in methods or definitions, provide appropriate caveats to avoid misleading comparisons.

Explain the strengths and limitations of the statistics in relation to likely uses

If there are key issues that affect how the statistics should be used or interpreted, mention them up front to support appropriate use.

Don’t bury important limitations in the supporting information. Avoid any implication that the statistics are free from error. Include descriptions of the main likely errors, their potential impact on the statistics, and the implications for use.

Further information about quality and methods can be found in other sections of this guidance and in our Communicating quality, uncertainty and change guidance.

Example

Pages 4, 5 and 6 of the Welsh Index of Multiple Deprivation 2019 report outline what the index can and can’t be used for.

Be neutral and impartial

Describe policies and targets in factual terms. Don’t endorse or comment on the effectiveness of current or past policies, or comment on the appropriateness of targets.

Departmental logos are helpful for orientation but be cautious before using the logo or branding of a government programme to which the statistics relate. This can carry the risk of perceived endorsement.

Explain why the statistics have been collected

Include relevant, factual information about the policy and operational context. If the statistics are used to measure policies or targets, list or provide links to them.

Discuss:

  • what is measured
  • what the statistics show in relation to the policies or targets
  • any relevant frameworks or indicators
  • any relevant previous targets
  • why the policy is being monitored

Example

“Indicators are useful tools for summarising and communicating broad trends. They are not intended to incorporate all the relevant information available in the UK. They are best seen, as their name suggests, as indicative of wider changes.

The UK biodiversity indicators formed a major part of the UK’s 5th National Report to the CBD in 2014, supplemented with other information relating to UK biodiversity and implementation of the Strategic Plan for Biodiversity 2011-2020.

It is expected that the indicators will be amongst the information used to produce the 6th National Report to the CBD (due to be submitted in December 2018). In 2015, JNCC produced an updated mapping of the indicators against both global and European biodiversity targets.”

Department for Environment and Rural Affairs, August 2017, UK Biodiversity Indicators 2017

 

But, remember the message!

Remember to think about the type of context that adds the most to the messages you are trying to communicate.  Don’t give users the statistics in every type of context.

Back to top of page

Provide interpretation for the statistics

Good commentary should help users to understand and interpret the messages from the statistics, and should be insightful and objective.

Explore relationships, causes and effects

Explore relationships, causes and effects to the extent that they can be supported by evidence. Include possible reasons, appropriately justified, to explain what the statistics show.

It can be challenging to provide insightful commentary without straying into opinion and conjecture, but you have an obligation to explain how any contextual information has been used to validate your statistics.

Explore potential reasons for the patterns that you see

Do research and keep up to date with the latest developments in your subject area. Sound knowledge of your topic and its theoretical context will help you to interpret the statistics and add value through your commentary.

Example

“The decline in the number of certificates in Functional Skills is likely due to the changes in funding rules by the Education and Skills Funding Agency and revised guidance from DfE that post 16 students who have a grade D/grade 3 in English or maths must now be entered for GCSE resits rather than Functional Skills. In addition, colleges are also incentivised to enter students with grade E for GCSE as they gain more credit for distance travelled by improving a GCSE grade than for Functional Skills attainment.”

Office of Qualifications and Examinations Regulation (Ofqual), June 2018, ‘Vocational and other qualifications quarterly: January to March 2018

Example

“It is reductions from the energy production and manufacturing sectors that have been the strongest drivers for the long term trend of decreasing emissions, by switching fuel use from coal to gas and the fitting of flue gas desulphurisation in the remaining coal fired plants in the power sector. The decrease in SO2 emissions in recent years, with UK emissions falling by 61 per cent between 2012 and 2016, was largely due to the closure of a number of coal-fired power stations that had reached the end of their working lifetime. These closures, together with the conversion of a few other coal-fired units to burn biomass instead, have significantly reduced the overall coal-burning capacity.”

Department for Environment and Rural Affairs, February 2018, “Emissions of air pollutants in the UK, 1970 to 2016

Example

“The total Net Ingredient Cost (NIC) for items prescribed for alcohol dependence in 2017 was £4.42 million. This is 9% lower than in 2016 and breaks the recent trend of successive year on year increases. The decrease in cost has been mainly driven by reduced prescriptions items for Disulfiram.”

NHS Digital, May 2018, ‘Statistics on Alcohol: England 2018

Example

“Likewise, the decrease in passenger journeys on some systems (for example, Docklands Light Railway and Sheffield Supertram) are likely to be a result of planned work closure.”

Department for Transport, June 2018, “Light Rail and Tram Statistics England, 2017/18”

Provide insights into any trends

Mention relevant special events or circumstances that may have affected the statistics. Don’t start time series at a point that could be perceived as not being impartial. Similarly, avoid comparisons of two points that could be perceived as not being impartial.

Avoid ‘elevator’ commentary that describes every rise and fall in the numbers. Graphing the series and pointing out important features will help when examining trends.

Consult with policy teams and other specialists

Establish if there have been policy, societal or economic changes or new initiatives that may have caused the results observed and reflect this information in the commentary. Providing the analysis is evidence-based and impartial, this can legitimately be done in compliance with the Code of Practice as part of the quality assurance process.

Be mindful that what is relevant or important may change between releases

Don’t just update the numbers into the narrative of a previous release. Provide a relevant and insightful story behind the latest figures, particularly for topics that become of high national interest or feature in political debate.

Example

Figure 2.2 in the Home Office report on Hate Crime in England and Wales 2016/17 has annotation outlining high profile incidents that aid interpretation of the statistics.

Describe the extent of the uncertainty in the statistics

Good commentary will help the reader to understand the extent of uncertainty in the statistics. It should draw attention to and make clear the nature and implications of the uncertainty associated with the statistics.

See the “Communicating Uncertainty and Change Guidance” on the GSS Policy Store
for more information [20].

Back to top of page

Present main messages clearly and concisely

Conveying the main messages from the statistics is essential to maximise public value. Focus on the most important, useful and relevant messages and present these up front.

“Statistics should be accompanied by a clear description of the main statistical messages that explains the relevance and meaning of the statistics in a way that is not materially misleading.”

Code of Practice for Statistics, UK Statistics Authority, 2018

Focus on the main points of interest

Take into account your users’ requirements and the current context. If your statistics say something important about a current debate, try to incorporate this information to add public value.

Write accessible and easy to understand main messages

Try to write the main messages so that any user can understand them. Peer review can really help here.

Update the messages as well as the numbers

Are the messages from the last reporting period still the most relevant and newsworthy, or should you revisit them? Remember that the biggest change may not be the most important one. Take account of the current context.

Don’t try to summarise all of the findings in the publication

Main points should include up to six bullet points each no longer than one sentence.

Ensure messages can stand alone

Journalists and press offices often use main messages verbatim. Well drafted messaging increases the chance of the media identifying and re-presenting appropriately. Consider whether the messages can stand alone in a newspaper article without additional explanation. If not, they may be taken out of context.

 

Back to top of page

Use structure to tell the statistical story

The structure of the publication should help users understand the story behind the statistics.

Summarise the main messages at the start of the release

These should be the points that are most relevant or interesting to your users and for public debate. This ensures users come away with the main messages even if they don’t read the whole publication.

Use the inverted pyramid structure

The inverted pyramid structure is used by journalists and differs from the traditional style of academic reporting. The most important information is presented first. Further detail and less critical information can be provided afterwards.

The inverted pyramid structure breaks content up into three parts:

  • Crucial
  • Helpful
  • Nice to know

Don’t start with lengthy background information or technical definitions

Include a short paragraph explaining what the publication is about. More detailed information can be placed in a separate section.

Descriptive subheadings help users

Active headings outline the main message making them more memorable for users.

Only include information which adds to the statistical story

Consider each sentence and whether it adds to the story. If not, the information can be presented elsehwhere without disrupting the commentary.

Back to top of page

Write clear and informative titles

Users need clear and informative titles to help them to identify whether the statistics are of interest and relevant to them. They also help users scan online content.

Titles should stand alone

Titles should include:

  • a concise description of the statistics
  • the time period covered
  • the geographical coverage
  • how often the statistics are released
  • whether the statistics are provisional or final, if applicable

Avoid ‘producer-focused’ titles

Some titles betray the author’s understandable desire to publicise the work they have done on a data collection. This may also be a legacy title used for many years, but changing titles can (and has been) done.

Aim to convey a user’s perspective of the output. The data source can be included in a subtitle.

Don’t overload the title with too much information

If necessary, provide a short paragraph of additional detail on the front page.

Back to top of page

Use plain language

The language used in commentary should be simple, clear and appropriate for all audiences.

Use plain English

Avoid technical language, jargon and acronyms.

Example

“Gross weekly pay in the bottom income decile was below £276 for full-time employees” is much easier to understand when written like this:

“One in ten full-time employees earned less than £276 per week”.

The plain English campaign gives more information on this area.

The Government Digital Service have put together plain English guidance and a list of words to avoid.

Be impartial and objective

Avoid sensationalism. Do not use terms that reflect a value judgement such as ‘relatively strong rate’, ‘very few’, and ‘only’.

Avoid suggestions of partiality to government, e.g. by referring to government as ‘our’ or ‘we’.

Balance the need for technically exact but complex terminology and clarity

Users will understand that even with a well understood term like ‘unemployment’ there are detailed decisions taken about classifications. These do not need to be spelt out in the main messages.

There will be times where more detail is necessary to avoid risk of confusion between related concepts.

If technical terms and definitions are unavoidable, explain them on first use

Some well-known abbreviations and acronyms may need no explanation, but it is best to be cautious and to explain any terms that may be unfamiliar to most readers.

Embedding complex definitions into the main story makes the language complex and hard to follow. Hyperlinks can take users to definitions they need.

Include a glossary of specialist terms

Signpost users to a glossary, but don’t force them to rely on one. Do not place glossaries at the start of a release.

Be cautious if using words with specific meanings in the context of statistics

Take care not to misuse words like “significant”.

In some cases, there may be plausible but uncertain explanations for patterns in the statistics. It is important to apply sound professional judgement. With careful wording, less certain explanations can also be included.

Words which suggest causality: affect, cause, consequence, effect, impact

Words which suggest relationship but not causality: association, correlation,
corresponding, equivalent, parallel

Words which suggest a more provisional explanation: expect, believe, think, predict, envisage, forecast

Be consistent

Use the same terms, abbreviations and units throughout to help the reader understand and draw comparisons. For example, don’t switch between “0.3 million” and “300 thousand”.

Round numbers appropriately

Make sure that the level of numerical detail is appropriate given the precision of the numbers you are reporting. Figures with lots of detail give an impression of high accuracy that may be unwarranted.

Users find it difficult to process long, complex numbers. For example, use “3.5 million” instead of “3,546,882”. Use commas to separate out thousands when writing numbers.

Be concise

Write short sentences and paragraphs. Aim for 15 to 20 words per sentence and one concept per paragraph. Don’t overload sentences with lots of numbers.

Use tools to improve readability

Most word processors include tools to check readability. For example, Microsoft Word can give you a “Flesch-Kincaid Grade Level”. How to find and understand your readability score on Word.

Alternatively, if your content does not contain any sensitive unpublished material, paste it into the online Hemingway App. This will also give you a reading grade and it can help to improve your content by identifying complex sentence structures, phrasing and words.

Back to top of page

Help users find the information they need

Help users quickly and easily navigate the publication and identify points of relevance.

A contents page can be helpful for longer releases

A contents list should to be used to give a broad overview of the structure of the publication. A long, detailed list of tables and figures is off-putting and detracts from the main messages.

Clearly state whether the statistics are National Statistics or official statistics

Include clear labelling in the release and supporting documents.

Where the statistics are designated as such, always use the National Statistics logo. Never use the logo on outputs that are not National Statistics.

Also include:

  • timing of the next release
  • copyright terms
  • contact details for the producer

Provide or direct users to relevant supporting information

Supporting information helps users to understand and use the statistics correctly. Information should be readily available from the website landing page of a release.

Provide the underlying and any related data to enable further analysis

Where possible, include links to supplementary tables and datasets (e.g. lower geographies, time series) in a convenient format to allow for the reuse of the data.

Consider providing the data in machine readable open data formats. The Connected Open Government Statistics project has more information on this.

Outline the disclosure controls in place. Consider providing links to any related datasets.

Using a standard template can ensure a consistent structure

Standard templates can be useful for regular users who will be able to locate the information they need quickly and easily. Templates also demonstrate to users that publications are from a group of similar or related statistics.

Back to top of page

Consider the online experience

It is best practice to publish statistical releases in a HTML format (i.e. text on a webpage), as opposed to a PDF. This means thinking about the online experience is very important as it is a different user experience to reading a printed document.

Why HTML is better

Reasons for moving away from publishing content in documents (e.g. PDF, Word, Excel):

  • they are not best practice in terms of accessibility
  • search engines cannot look inside document formats meaning content is harder to find
  • we cannot use analytical tools to assess how users read documents
  • internet search engines sometimes take users directly to documents, but when this happens the user will often not be able to find out where on a website that document lives, this means it is difficult for users to tell if a document is out of date.

Read more about why website content should be published in HTML and not PDF.

Reading speed

Reading tends to be slower online, but people expect quicker results and spend little time on a page. Clear identification of the main messages and being able to easily scan content is even more important.

Essential information and key words should be on the top left

Our eyes move across web pages from left to right, top to bottom, in an F-pattern. This places most attention on the top left of the page. Use the right and lower part of the page for supporting information that is not essential for the main story.

Style.ONS has further guidance on how we read on the web. 

Use web analytics to gain user insight

Publishing content in HTML allows producers to use website data to gain insight into how users access and navigate in publications.

Analysis of visitors to the ONS website found that only 20% of users scroll a quarter of the way down a bulletin, and 53% of people who land on a bulletin page leave the site immediately. Bulletins take 9.5 times longer to read than people actually spend on the page.

Back to top of page

Tell users about quality and methods

Commentary should be supported by information that describes the quality of the statistics and the methods used to derive them.

“The quality of the statistics and data, including their accuracy and reliability, coherence and comparability, and timeliness and punctuality, should be monitored and reported regularly.”

Code of Practice for Statistics, UK Statistics Authority, 2018

Be upfront about any important caveats

Any caveats that arise because of the quality of the statistics or the methods used should be presented early on in the publication. However, ensure that these details do not dilute or obscure the main messages.

Use progressive disclosure

Adopt a tiered approach with different levels of information available for different users.

Think about user personas, In general, nontechnical users will not need to know the detailed methods involved to use the numbers with confidence.

Detailed quality and methods information should be provided and users should be signposted to it.

Be specific

Avoid general statements about the quality of the statistics. Instead, focus on how quality and methods impact on use.

Explain complex concepts

When discussing confidence intervals and other quality measures use a plain English explanation. Make sure that such explanations are as easy as possible to understand and not overly detailed. Ask a colleague or non-expert to peer review the explanation.

Example

Explanation of confidence intervals from Department for Work and Pensions on page 2 of ‘Fraud and error in the benefit system: final 2016 to 2017 estimates first release’

Explain the statistics are initial estimates if normally subject to later revision

“Scheduled revisions, or unscheduled corrections that result from errors, should be explained alongside the statistics, being clear on the scale, nature, cause and impact.”

Code of Practice for Statistics, UK Statistics Authority, 2018

Include a revisions statement which outlines:

  • when the statistics are likely to be revised
  • the extent and direction of any likely revision (take care to avoid conjecture)
  • a link to a published Revisions Policy relating to the statistics

Smaller revisions are a measure of reliability. However, small revisions do not necessarily mean that the statistics are accurate.

To prevent confusion or the use of incorrect figures, ensure only the latest version of a revised dataset is available. Explain the nature and extent of revisions, and how these revisions affect the interpretation of the statistics.

Report quality against the European Statistical System’s quality dimensions

Relevance is the degree to which a statistical product meets user needs in terms of content and coverage.

Accuracy and Reliability is how close the estimated value in the initial and final outputs are to the true result.

Timeliness and Punctuality describes the time between the date of publication and the date to which the data refers, and the time between the actual publication and the planned publication of a statistic.

Accessibility and Clarity is the quality and sufficiency of metadata, illustrations and accompanying advice, and the quality and sufficiency of metadata, illustrations and accompanying advice.

Coherence and Comparability is the degree to which data derived from different sources or methods, but that refers to the same topic, is similar, and the degree to which data can be compared over time and domain, for example, geographic level.

Back to top of page

Think beyond bulletins

All of the statistical outputs made available to users should include appropriate and accessible commentary.

Adapt commentary for different users and outputs

Commentary in statistical bulletins may not be appropriate for all types of users. Good commentary can be adapted from bulletins and used elsewhere.

Use main messages from your commentary for policy colleagues, social media outputs, infographics and board reports—but adapt to suit these different users.

Use social media alongside statistical releases

Social media can help reach a wide audience and convey headline messages quickly. The Government Digital Service’s ‘Social Media Playbook’ provides comprehensive guidance on using social media in government.

Back to top of page