The Workshop is a chance to hear about developments and future plans for the FES, and to be able to discuss these with the people who run the survey. It is also a chance to hear about research using the FES. Participants were not only from universities, but also from other bodies such as the Civil Service College, the Bank of England, and the Isle of Man Government, plus a few from computing and market research companies.
Denise Lievesley from the Data Archive introduced the Workshop. Two papers then gave an overview of the survey, and possible developments. The first of these papers was by John King of the Central Statistical Office (CSO), which is responsible for the survey. He outlined the processing and validation of the data, which is done using the database management system Ingres, and the production of tables for use within CSO using the tabulation package QuickTab. The fieldwork for the survey is carried out for CSO by the Office of Population Censuses and Surveys (OPCS). The next paper by Tony Manners of OPCS described the data collection. This is now done using computer-assisted interviewing with Blaise software, each interview taking about one and a half hours. Editing and checking is carried out during the interview (using Blaise), at the interviewer's home (check programs), and at OPCS. If the interviewer finds an anomalous answer, it can be checked when the interviewer returns to collect the completed expenditure diaries. The expenditure diaries are still collected on paper and coded up at OPCS.
Developments being considered include expenditure diaries for children, for which pilot studies have already been carried out.
After this overview, Louise Pickering of the CSO gave some details of the Ingres database structure. There is a raw database containing the data values largely as supplied by OPCS, and a derived database containing all the variables derived from this by CSO. The derived database has been designed to be similar to the 1993 SIR database - the last year for which the Department of Employment processed the dataset, using the SIR database management system. Following discussions with the Data Archive, and suggestions at the FES Workshop last year, CSO will continue to make the FES available as a SIR database for academic researchers. (The FES datasets for 1961 onwards, except 1964-1967, are available as SIR databases, Data Archive reference GN:33057, from the Archive or on-line at Manchester Computing Centre).
The documentation to be released will include the Ingres SQL programs used for calculating derived variables. Eloise Critchley, also from CSO, discussed the changeover to the financial year, and some of the methodological research at CSO. The 1993 FES followed the calendar year. However the "1994" FES will be for the financial year April 1994 to March 1995. Data have been collected for the first quarter of 1993, and a financial year 1993/94 dataset will be released for comparison with 1993 and 1994/95 datasets - this will be of great value for academic researchers. The 1993/94 dataset is due for release shortly, and the 1994/95 dataset around October 1995. Methodological issues being investigated include grossing-up factors to give national figures (possibly to be included with the 1994/95 dataset); self-employment income questions; and the effectiveness of incentives on response rate (in FES there are currently payments for members of responding households).
The Family Resources Survey (FRS), run by the Department of Social Security, was described by Mike McDowell from that Department. This was not strictly to do with the FES - but there is an overlap in that both surveys collect income data, and the FRS is of interest to many FES users. The Department is particularly interested in benefit payments, since the breakdown of payments by type is not recorded centrally for analysis. In the FRS, the data are collected using computer-assisted interviewing (Blaise software), and processed using SIR. The survey achieved some fame for its use of neural networks for imputation - tests of neural networks versus multiple regression were described briefly. The first year of data to be released will be for the 1993/94 financial year, which has an achived sample of about 25,000 people and a response rate of 68%. Data will be released sometime in 1995.
Returning to the FES, Kate Foster of OPCS described the Census check on non-responders in the FES. For the 1991 FES, the sample addresses for which there was no response have been checked against the 1991 Census data, so that the characteristics of non-responders can be found and compared with responders. Other surveys are also compared with the Census - General Household Survey, Labour Force Survey, and National Food Survey. It was done for the FES in 1971 and 1981. New for the 1991 FES comparison is an examination of non-response to parts of the questionnaire, and also multivariate analysis of non-response using more than one Census variable. It is possible to calculate "Correction factors" (weights) for various types of household, and OPCS are looking at ways of weighting the data for general use. The results of the Census check have not yet been published.
The FES in Northern Ireland was described by Kevin Sweeney of the Policy Planning and Research Unit (PPRU), the body responsible for the FES in that part of the UK. A sample of about 1200 households is taken, but only one in five of these are included in the amalgamated UK FES - the full Northern Ireland data are available from the Data Archive as separate annual datasets (GN:33240). Differences in questions between Northern Ireland and the rest of the UK are small. Some questions have different wording, e.g. DHSS instead of DSS. Some questions reflect different circumstances, e.g. rates instead of council tax. And finally some questions are extra, e.g. fuel use, religion. The sampling method is also somewhat different. PPRU changed to computer-assisted interviewing at the same time as OPCS (April 1994).
Richard Dorsett spoke as an analyst of the FES for research purposes. He described a project underway to analyze the demand for meat. This is being carried out in the School of Economic Studies, Manchester University, and is funded by the ESRC. The FES is being used to examine the household characteristics which influence two decisions: (i) the decision whether to eat meat at all, and (ii) the spending on meat, given that meat is being eaten. Characteristics can influence these two decisions independently. To look at changes over time, FESs for 1973, 78, 83, 88, and 93 are being used. So far most of the analysis has been with single-adult households, as this avoids complications such as meat-eating and vegetarian adults in the same household. Since the FES does not have the price information needed for some analyses, the National Food Survey will also be used.
An interesting perspective on the FES was given by Eleanor Emberson of CSO. One of the main purposes of the FES is to see what people spend money on - it provides "weights" for the Retail Prices Index. The actual prices are not taken from the FES however, but are collected from shops and service-providers (e.g. plumbers) - this is done every month at around 200 places in the UK. Prices are collected for about 500 products and services, which are generally the same from month to month but have to change occassionally to reflect changes in spending patterns. Until now the collection of prices has been done by unemployment benefit office staff, but this is now to be done by a private company. Eleanor described some of the problems - for example it is easy to compare the price of a famous chocolate bar, but what is the price of a "pair of men's shoes"? Work is being done to control the price collection more tightly to improve comparisons across places and over time. A training video for fieldworkers was shown, and demonstrated that collecting prices is not as easy as you might have thought!
The Workshop ended with a panel session for questions and discussion, rounding off a very useful day.
I recommend the next FES Workshop for all users of FES data. The date will be publicized in the ESRC Data Archive Bulletin, and through other routes such as the FES emailing list (firstname.lastname@example.org).
Manchester Computing Centre
ESRC Data Archive, University of Essex, Colchester, CO4 3SQ Tel: +44 (0)1206 872001 Fax: +44 (0)1206 872003