Protecting confidentiality with statistical disclosure control

UK policy position agreed

The Registrars General of Scotland, England and Wales and Northern Ireland have agreed to aim for a common statistical disclosure control (SDC) methodology for 2011 Census outputs. An important step towards achieving this aim was the agreement by the Census Offices of a UK SDC policy position setting out what constitutes a disclosive risk in a census context.

This will help achieve the aim of harmonising the three censuses where that is in the interest of census users. A position on SDC policy has been agreed by the Registrars General of the three census offices. This is an agreement about what constitutes a disclosive risk in a census context. More information is provided in Annex B of the Registrars General's agreement.

The aim of the SDC project is to design a UK SDC strategy in accordance with the agreed policy which protects against disclosure in the 2011 Census outputs whilst maintaining maximum utility of the data. The strategy will need to cover (pre-defined) tabular outputs, microdata samples and possibly flexible user defined tabular outputs whilst taking into account the impact of interactions between these types of output. The strategy will also be designed to address the concerns of users whilst adequately protecting the data.

Governance

An SDC Working Group with representatives from each UK Census Office has been formally established. The purpose of this group is to provide advice, review and to steer work associated with developing the disclosure control methodology for the 2011 Census. A Disclosure Control Subgroup of UK Census Design and Methodology Advisory Committee (UKCDMAC) has also been set up. This group is responsible for providing advice on methodological issues and acts as a formal quality review panel for the SDC project.

The work to develop the 2011 Census UK SDC strategy takes into account the Downstream Processing schedule and the work on 2011 Census Outputs and 2011 Census Geography. The UK Census Offices have agreed to work closely throughout the development of the SDC strategy. Within ONS the Statistical Policy Committee (SPC) will be responsible for signing off the recommended UK 2011 SDC strategy. Census Project Boards within the UK Census Offices will also be responsible for signing off the strategy prior to final sign off by the UK Census Committee (UKCC).

Methodological research

The work programme to decide the SDC strategy consists of three stages. The first stage of methodological research involved producing a review of SDC in a census context which addressed:

• research conducted prior to 2001

• 2001 SDC Methodology (tabular outputs and microdata)

• lessons learnt from SDC in 2001

• work conducted in lead up to 2011

The second stage of research involved producing a high level review of the advantages and disadvantages of a wide range of SDC techniques. This review resulted in the SDC Working Group agreeing a shortlist of methods to be taken forward for quantitative evaluation using a disclosure risk - data utility framework. The shortlist was drawn by assessing a wide range of methods against a set of qualitative criteria that were in line with the SDC policy statement previously agreed by the RsG. The shortlist was agreed by the UK Census Offices in October 2007. Formal quality assurance was provided by the SDC Subgroup of UKCDMAC. A summary of the approach used to derive the short-list of methods going forward for quantitative evaluation is presented in the Census Advisory Group paper AG(07)10. The approach taken is presented in detail in the SDC UKCDMAC Subgroup Paper 1 which is available to download in PDF format. The paper presented to GSS Methodology Conference in June 2007 is also available.

The third stage of research is a quantitative evaluation of the short-listed SDC methods using a disclosure risk - data utility framework. It is focused on how much protection is provided by each method as well as the impact they have on the utility of the statistics. The quantitative evaluation is being used to inform the recommended SDC method(s) for the 2011 Census for all types of census outputs, although the focus is on tabular outputs.

The short-listed methods:

• record swapping

• over-imputation

• Australian Bureau of Statistics (ABS) cell perturbation method

are being evaluated by running tables based on data from 2001 Census, assessing disclosure risk and utility loss.

Initial results were reported to the SDC UKCDMAC sub group in May 2008 and a response given in July 2008. A presentation at GSS Methodology Conference in June 2008 gives a flavour of the work. All are available to download.

Findings from user consultations have been incorporated into the evaluation to ensure we take into account users’ views of desirable features of a disclosure control strategy. The quantitative evaluation is due to be completed by September 2008.

Recommendations will then be made to governing bodies before a three month period of communication and user consultation. The SDC strategy for Census tabular outputs will be finalised in Spring/Summer 2009.

A further stage of research will be conducted to develop methods for the protection of 2011 Census microdata samples in accordance with the 2011 microdata strategy developed by the Census Microdata Working Group. This phase of research will commence in February 2009 and will run concurrently with user consultation and sign-off phase of the SDC strategy for tabular outputs.

Timetable for development of UK SDC strategy

• September 2008 - End of methodological research phase for the 2011 Census UK SDC Strategy

• October 2008 - January 2009 - Recommendations made to governing bodies

• February 2009 to April 2009 - User consultation, communication, education and feedback

• February 2009 – Commence research into methods for the protection of 2011 Census microdata samples

• Spring/ Summer 2009 - Finalise UK SDC strategy for tabular outputs