Statistical Officer / Government Statistical Service (GSS) Assessment Official Statistics: Governance, Ethics and Quality Syllabus

What does the 'Orderly release' principle require, including the pre-release rule?

Statistics must be released to all users at the same time, normally at 9:30am, with a pre-announced publication date. Access before publication (pre-release access) must be limited, controlled and time-restricted to maintain equal access and public confidence.

Under the Code of Practice, how far in advance should the release date of regular statistics be announced?

Release dates should be pre-announced, normally at least four weeks (one calendar month) in advance, and any change to a pre-announced date must be approved by the chief statistician and explained publicly.

What is the standard daily release time for UK official statistics, and why is it fixed?

9:30am on a working day. A fixed, equal-access time prevents any user (including Ministers or markets) from gaining unfair early advantage, supporting the orderly release principle.

What is 'pre-release access' (PRA) and what are its key controls?

PRA is the limited, controlled access given to a named list of eligible people (e.g. relevant Ministers and officials) to statistics before public release. It is restricted to a maximum of 24 hours in advance, must be logged, and the list of recipients must be published.

What is the purpose of the 'pre-announcement' and a published release calendar?

To guarantee orderly, predictable release; demonstrate that timing is not manipulated for political advantage; and give all users equal opportunity to plan for and access new statistics simultaneously.

What should a producer do if they discover an error in published official statistics?

Be open and transparent: correct the error promptly, clearly mark the corrected version, publish a correction notice explaining the nature and cause of the error, and follow the organisation's revisions and corrections policy.

What is the difference between a 'scheduled revision' and an 'unscheduled correction'?

A scheduled revision is a planned update to figures as more/better data become available (e.g. provisional to final GDP), managed under a published revisions policy. An unscheduled correction fixes an error and requires a transparent correction notice.

How does the OSR typically respond to a suspected breach of the Code of Practice?

It can investigate, issue public correspondence/casework letters, require remedial action, publish findings, and — for accredited statistics that no longer comply — suspend or remove the accreditation until standards are restored.

What is the rule on Ministers and politicians making statements about statistics before official release?

Ministers and officials must not comment publicly on, or reference, the substance of statistics before they are officially published; doing so undermines orderly release and equal access and is treated as a breach by the OSR.

Define statistical disclosure control (SDC).

The set of methods used to protect the confidentiality of individuals or businesses in published statistics by reducing the risk that a data subject can be identified or that their attributes can be disclosed, while keeping the data as useful as possible.

Distinguish 'identity disclosure' from 'attribute disclosure' in SDC.

Identity disclosure occurs when an individual/unit can be recognised in released data. Attribute disclosure occurs when new confidential information about an individual/unit is revealed (e.g. learning a value), even without certain identification.

What is the difference between primary and secondary disclosure in tabular data?

Primary disclosure is a cell that is itself risky (e.g. a small count or a dominated total). Secondary disclosure is when a safe cell could be deduced from suppressed primary cells via row/column totals; secondary suppression is applied to prevent this.

Name common statistical disclosure control techniques for tables.

Cell suppression (primary and secondary), rounding (e.g. to base 5 or controlled/random rounding), aggregation, recoding/banding categories, the small-cell threshold rule, and adding noise/perturbation.

What is the 'threshold rule' (small cell rule) in disclosure control?

A rule that any cell with a frequency below a set threshold (e.g. fewer than 3, or fewer than 10 individuals depending on policy) is considered unsafe and is suppressed or modified to prevent identification of individuals.

What is the 'dominance rule' (n,k rule) used for in business statistics?

A rule for magnitude tables: a cell is unsafe if the largest n contributors account for more than k% of the cell total (e.g. (1,k) or (2,k) rule), because a dominant contributor could estimate another's value. Such cells are suppressed.

What are the 'Five Safes' framework dimensions for safe data access?

Safe Projects, Safe People, Safe Settings, Safe Data, and Safe Outputs — a framework used (e.g. in the ONS Secure Research Service) to manage the risk of granting access to sensitive microdata.

What is the difference between 'p% rule' (perturbation/noise) and 'suppression' as SDC approaches?

Suppression removes/hides risky cell values entirely (reducing completeness). Perturbation methods (rounding, noise addition, cell-key method) alter values slightly to mask exact figures while preserving overall patterns and table additivity where possible.

Which legislation principally governs the processing of personal data in UK official statistics today?

The UK General Data Protection Regulation (UK GDPR) together with the Data Protection Act 2018, which set out the lawful bases, principles and safeguards for processing personal data, including special provisions for research and statistical purposes.

What special status do 'statistical purposes' have under UK GDPR/DPA 2018?

Processing for archiving in the public interest, scientific/historical research, and statistical purposes benefits from exemptions and safeguards (Article 89): such processing is not regarded as incompatible with the original purpose, provided appropriate safeguards (e.g. data minimisation, no decisions about individuals) are in place.

What data-sharing powers did the Digital Economy Act 2017 give for statistics and research?

It created gateways enabling public authorities to share data with the UK Statistics Authority/ONS for statistics and research purposes, strengthening access to administrative and other data sources for producing official statistics.

What confidentiality duty does the Statistics and Registration Service Act 2007 place on personal information held by the Statistics Board?

Section 39 makes it a criminal offence to unlawfully disclose personal information held by the Authority that identifies a person, subject to limited exemptions — a strong legal safeguard underpinning public confidence in providing data.

List the core data protection principles a government statistician must apply when handling personal data.

Lawfulness, fairness and transparency; purpose limitation; data minimisation; accuracy; storage limitation; integrity and confidentiality (security); and accountability.

What is the main purpose and content of the National Statistician's Data Ethics Self-Assessment / the UK Statistics Authority Data Ethics framework?

To ensure data use is ethical by applying principles such as: clear public benefit, transparency, legal compliance, proportionality and minimisation of intrusion, sound methods, robust data security, and ensuring the analysis is necessary and not harmful.

State the six principles of the UK Statistics Authority's ethical use of data (data ethics) framework.

(1) Public good/benefit; (2) sound methods and data quality; (3) legal compliance; (4) confidentiality and security of data; (5) proportionate and necessary use minimising intrusion; (6) transparency and accountability about how data are used.

In data ethics, what does 'public good' or 'public benefit' require an analyst to demonstrate?

That the use of data delivers clear value to society or improves public services/policy, and that the benefits justify any use of personal data and any intrusion into privacy — use of data must serve the public good, not private interest.

Name the standard dimensions of statistical quality (the European Statistical System quality dimensions).

Relevance, Accuracy and reliability, Timeliness and punctuality, Accessibility and clarity, Coherence and comparability, and (often) Completeness — sometimes summarised with the underlying concept of fitness for purpose.

Distinguish 'timeliness' from 'punctuality' as quality dimensions.

Timeliness is the gap between the reference period of the data and the date the statistics are published (how soon after the event). Punctuality is whether the statistics are released on the pre-announced/scheduled date.

Distinguish 'coherence' from 'comparability' as quality dimensions.

Coherence is the degree to which statistics from different sources or domains can be reliably combined/reconciled. Comparability is the degree to which statistics can be validly compared over time, across regions, or across populations using consistent concepts and methods.

What is the difference between sampling error and non-sampling error?

Sampling error arises because a statistic is based on a sample rather than the whole population, and can be quantified (e.g. via standard errors/confidence intervals). Non-sampling error arises from other sources — coverage, non-response, measurement, processing — and is harder to quantify.

What does QAAD stand for, and what is its purpose?

Quality Assurance of Administrative Data — the OSR's toolkit/framework for assuring the quality of administrative data sources used to produce official statistics, ensuring producers understand and communicate the strengths and limitations of data collected for non-statistical purposes.

What four areas of practice does the OSR QAAD toolkit assess, and what risk-rating does it use?

(1) Operational context and data collection; (2) Communication with data supply partners; (3) Quality assurance principles, standards and checks; (4) Producer's audit and assurance of QA arrangements. Each combines a level of risk of quality concerns (low/medium/high) with a level of assurance (basic/enhanced/comprehensive).

Why must administrative data be quality-assured differently from survey data, and how should producers communicate uncertainty?

Administrative data are collected for operational (not statistical) purposes, so coverage, definitions and accuracy may not match statistical needs and there is no standard error to quote. Producers should be transparent about strengths and limitations, explain potential biases, and convey uncertainty using ranges/quality notes, confidence intervals where applicable, clear caveats, and revisions information.