NEWS FROM THE ROYAL STATISTICAL SOCIETY

Developments in business survey methodology in the Office for National Statistics, 1994-2000

Meeting of 15 January 2003

Paul Smith, Mark Pont and Tim Jones (ONS) gave an informative paper to the RSS, which was a 'must' for anyone concerned with contributing information to business surveys or using the results. Although concerned with statistical methodology, the conclusions are of the utmost importance to users, as much of the explanation directly concerns the quality of data output.

The Newport-based Prices and Business Group [formerly the Business Statistics Group (BSG)] is responsible for the majority of the 100 or so business surveys. The main ones are:

Inquiry Sample  size
Annual Register Inquiry 68000
Annual Business Inquiry 75000
Monthly Production Inquiry 9000
Retail Sales Inquiry 5000
Monthly and Quarterly Inquiries into Distribution & Services Sectors 30000
Monthly Wages & Salaries Survey 8000
Quarterly Capital Expenditure & Stocks Inquiries 32000
Quarterly Profits Inquiry 1600
Products of the European Community (PRODCOM) 28000
Producer Price Indices (PPI) 9000

Following (1) an introductory section, the remaining five sections of the paper describe and discuss (2) sampling frame and sampling procedures; (3) the search for efficiencies and standard methods; (4) merging the collection of employment and business data; (5) quality and its place in National Statistics; and (6) future developments.

Section 2 discusses the characteristics of business surveys, the Inter-Departmental Business Register (IDBR), statistical units, industrial classification the Standard Industrial Classification (SIC92), size variables, sampling procedures, burdens on business, stratification and coverage.

Section 3 is concerned with reorganisation, common software and the short-period inquiries, data collection initiatives and questionnaire design, statistical computing and the Annual Census of Production, sample designs and allocations, estimation methods, measuring change, PRODCOM and PPI, outliers, editing and imputation.

Section 4 covers what has been done on some of the inquiries the wages and salary survey, monthly and quarterly inquiries, and the annual business inquiry.

Section 5 describes the work on quality undertaken at the request of Eurostat, including model quality reports, sampling errors in the Index of Production, examples of quality indicators, and quality reviews.

Section 6 is of interest to users in mentioning the ONS customer focus. Though ONS has a "reasonably welldeveloped system for consulting users", it states that the process by which it evaluates users' requirements needs to be more systematic. "The requirements that are expressed during a user consultation process should be balanced against those that are less clearly articulated as well as against a more effective evaluation of the costs. Reasonable users, involved in these processes, are likely to accept that some requirements cannot properly be met without jeopardising the whole exercise." Topics addressed include auxiliary size variables and stratification, industrial classification, matched pairs, outliers, and monthly business statistics.

It is perhaps disappointing that the paper's horizon ends in the year 2000 but statisticians often have to be satisfied with 'yesterday's Bradshaw'.

The paper will be published in The Statistician (2003), 52, part 3. Preprints are available from Anna Mair at the RSS ([email protected]) and online at www.rss.org.uk/publications/pre-prints.html in 'pdf' version.