National Statistics Methodology Advisory Committee
Third meeting, 23 September 2002

The following work in progress papers were considered.

Paper 1: Internet Price Indices - The challenges involved in including non-traditional retail channels in the RPI
The ONS is seeking to widen the scope of price collection to include goods sold over the internet in the calculation of the Retail Prices Index (RPI) and its European equivalent the Harmonised Index of Consumer Prices. This paper describes some challenges arising in the work to develop an internet price index (IPI), seeking views on the general strategy from the Committee.

Committee Conclusions:

• Clarifying the objectives of the RPI/IPI is an important factor in making the decisions needed;
• A survey, perhaps official, asking outlets for details on location was suggested;
• Delivery charges should be included as a separate item;
• Some possible identifiers for use in excluding trade sales are prices excluding VAT, and prices on large quantities;
• The use of internet search engines, the Inter-Departmental Business Register, and the telephone directory as a basis for a sampling frame should be considered.

Paper 2: Quality Measurement and Reporting
Users of statistics need to have access to a range of relevant quality measures and indicators so that they understand the strengths and limitations of official statistics. Within ONS comprehensive quality measurement and reporting is needed to monitor and achieve continuous improvement. This paper outlines the framework that is being developed as part of the quality measurement and reporting project underway within the ONS.

Committee Conclusions:

• A package of quality measures and indicators should be used for reporting, to avoid behaviour distortion;
• Appropriate consideration should be given to educating users;
• There is a need to recognise and address differences between different types of outputs;
• It is important to report on non-sampling as well as sampling errors.

Paper 3: Grossing and Constraining sample data
As a result of devolution, the demand for more detailed Scottish and other sub-UK/GB analysis of major surveys has increased significantly. As a result, some major problems have been identified, which in some cases make the Scottish totals meaningless. To illustrate this problem, two examples for Scotland (the Family Resources Survey and the Annual Business Inquiry) are presented.

Committee Conclusions:

• The Committee agreed that obtaining high quality regional estimates is important, and that the related issues should be considered at the outset of a survey;
• A common language should be used when discussing these issues;
• The use of alternative initial weights in CALMAR should be assessed;
• Methods (eg ridge regression) for approximate calibration should be evaluated;
• The prioritisation of variables and constraints should be considered.

Paper 4: Integrated Social Survey - weighting approaches and gains in precision
ONS is looking into the feasibility of amalgamating the four ONS social surveys (the Labour Force Survey (LFS), Expenditure and Food Survey, General Household Survey and Omnibus Survey) into an Integrated Social Survey (ISS). This has advantages in several areas, including work flow for interviewers, and expected precision of estimates. This paper concentrates on precision, and in particular explains the possibilities for weighting the question modules which are planned using information from a "core module" which will be asked of all respondents.

Committee Conclusions:

• The LFS rotating panel should be given appropriate consideration in the design of the ISS;
• Interpenetration of interviewers/modules was suggested;
• Principal component techniques could be used to select core variables;
• Methods for avoiding negative weights were suggested;
• Time-trend data, and multilevel models could be used in the weighting procedure.

Next meeting due 20 March 2003

This page last revised: Thursday, 19 December 2002