National Statistics Methodology Advisory Committee
Fifth meeting, 29 September 2003

The following work in progress papers were considered.

Paper 1: Disclosure Risk Assessment and Management, associated with Outputs from Research Data Centres
Associated paper
Access to confidential microdata is sometimes allowed within a 'safe setting', such as a Research Data Centre (RDC). Research Data Centres provide controlled computing environments where confidential microdata can be accessed and analysed. This paper describes the problems associated with disclosure risk assessment and management in outputs from Research Data Centres, and suggests approaches to solutions.

Committee Conclusions:

• We should work with a combination of safe data and safe settings.
• Time lags can complicate the issue, but can also help to protect data.
• There may be a problem with fuzzy estimation - being able to identify a household or business with high probability.
• It would be useful to train researchers and develop guidelines to manage a safe centre for ONS.
• We have to be careful to avoid disclosure from regression analyses.
• It may be useful to perturb cell frequencies.
• We need to be very careful when using synthetic data.
The need for caution when releasing user-defined tables was endorsed, because of the risk of disclosure-by-differencing. It might be possible to reduce the risk by presenting only interval estimates, rather than exact values, of frequencies at cell level.

Paper 2: Methodology for the Average Earnings Ratio
The Average Earnings Ratio (AER) has been developed by ONS as a result of a recommendation by Turnbull/King to find an earnings measure that reflects the movements of wages and salaries more closely than the Average Earnings Index (AEI). The AER will be published as an experimental series towards the end of 2003. Although the input data for the AER comes from the same source as for the AEI, the methodology used differs considerably. This paper describes the AER methodology for estimation, imputation for non-response and outlier detection and treatment.

Committee Conclusions:

• It is important to focus on change if this is the target variable.
It is worth looking at longer term rates of change.
• We should investigate different methods of estimation for average earnings.
• It is worth re-evaluating the imputation method and the outlier procedure.
• We should produce standard errors in order to interpret changes in pay.
• We could review the sample rotation rate of 20% and look for an optimal rotation rate.
• We should seasonally adjust the series when enough data is available.
• It may be worth investigating the use of medians in estimation of average earnings.

Paper 3: The NS 2001 Census-based Area Classifications
The 2001 Census-based area classifications were due to be released at Local Authority level during September 2003, with classifications at smaller levels to follow in the later part of the year. These general-purpose classifications aid further analysis of Census data by summarising socio-economic and demographic data within geographical areas. This paper describes the research that has been undertaken for the LAD classification, gives the results, and discusses the issues that will need to be dealt with for the classifications of smaller geographical areas.

Committee Conclusions:

• he methodology used for the Local Authority classification and those being proposed for the lower levels of classifications are sound.
• Consideration should be given to producing a number of more specific classifications, perhaps using fewer than 42 variables for each.

Paper 4: Effectiveness of alternative methodologies for surveying students in higher education
The Department for Education and Skills (DfES) is planning to commission a survey of income and expenditure of Higher Education students in the academic year 2004/5. The Department has traditionally used face-to-face interviewing for surveying students in Higher Education. However, other researchers commonly use postal, telephone and web surveys to collect data from students. The DfES is keen to consider these methodologies in the light of survey evidence and seeks advice on the suitability of different methods for ensuring a successful 2004/5 Student Income and Expenditure Survey.

Committee Conclusions:

• A comprehensive response was given, describing the issues associated with selecting an appropriate sample design.
• DfES were encouraged to assess the relative advantages and disadvantages of each method.
• Probability sampling was favoured over quota sampling, and face-to-face interviews were favoured as the means of data collection.
• It may be worth considering making the survey more attractive to students by using incentives. However, care should be taken not to introduce bias by doing this.

Paper 5: MAC support for National Statistics: Possible future developments
The National Statistics Methodology Advisory Committee was set up in October 2001. Two years later, and at the end of the Committee's fifth meeting, it seems appropriate to review the method of working. This paper provides an opportunity to do that, asking MAC members to suggest changes to the way in which the Committee works.

Committee Conclusions:

• Early access to the agenda is important. It would also be very helpful to have references.
• There was some support for themed meetings. It may be better to have a two day meeting with one day based on a theme.
• We could have 'beginning to end' meetings for important outputs.
• We should encourage greater involvement from the wider GSS, as presenters and observers.
• We should look at broadening the membership of the MAC.
• There are various possibilities for fellowships. We should look at the experience of the US Bureau of the Census.
• Fellowships should be prestigious and have appropriate remuneration.
• Fellowships should be at different levels - post doc and experienced academics.
• There could be fellowships for ONS staff at universities, possibly set up as exchanges with university staff.
• We should keep the MAC better informed about MG developments.

Next meeting: 26th March 2004

This page last revised: Friday, 4 June 2004