Application Series 17

LABOUR FORCE SURVEY CERTAINLY SIR

Mike Staley of the UK Office of Population Censuses and Surveys (OPCS) talks about SIR's use in Labour Force Surveys.

Background

The Labour Force Survey (LFS) is conducted by OPCS on behalf of the Employment Department and has existed in various forms since the UK entry into the European Community. It collects data relating to employment and, using the processed results, monitors employment patterns on a national scale. The results of the LFS are extremely accurate and constitute an increasingly useful tool for formulating policy decisions.

In the early years LFS was a biennial affair processed on an ICL mainframe. After evolving to an annual survey, it has become a quarterly, panel-based survey sampling over 60,000 addresses each quarter and using a PC based interactive questionnaire interviewing and collection package, Blaise. "Gentlemen, we have seen the future and it works..."

We provide the Employment Department with primary data, derived variables and abstract the sample to the total population (a process known as grossing). This work is done in SIR. An efficient Blaise-SIR interface has proved vital and we have devised a method of generating SIR database structure in Blaise metadata files. This has enabled us to maintain the same data shape and simply load the data into SIR - no fuss, no mess.

Analysis

The LFS derives three sets of variables: at the person, family and household levels. Those for the household and family have proved very stable; by and large, they are simple in concept. Those at the person level are a bit more interesting; besides which, we are here down to the basic unit, the data level which is of most interest to the Employment Department.

There are around seventy person derived variables, many of them involving searches of look-up tables in order to classify a respondent's occupation or industry code, or place of work. Some of these tables are small, and accommodated as temporary arrays held in SIR's workspace when being calculated. Larger ones have been designated as separate record- types to reduce the memory demands of the process.

Grossing, the process of extending the sample to the total population, is simple in concept, but can be bewildering in terms of the scale and number of the processing operations which have to be gone through. The whole operation is performed in tabfiles, which lends itself to this sort of high-speed data manipulation. tablespace limitations have forced some programs to be divided into modules.

Output

For data processing, the hierarchical structure of the data - persons within households within quota - has to be adhered to. Once processing has finished, however, this structure is no longer necessary. Various outputs are provided, such as SPSS, SAS, SIR export and ASCII files, as required by Employment Department clients. These different types of outputs are all available through SIR and frequently produce a flat file, consisting of one record type only.

If we write all of the variables which we need for these outputs to a single file, we can cut out the need to wander hither and thither through a hierarchical database. When there are several outputs to be run off the same database, this is definitely worthwhile.

SIR SAVE FILE

The use of the SIR SAVE FILE utility has proved invaluable in the generation of LFS output. It produces a flat-file database, consisting of one record type only. It can be used to efficiently generate a collection of outputs as mentioned above.

The Employment Department requirements increase continually as we continue to deliver the goods. Much of these extra requirements involve tabulating the data by other area definitions - for instance, in terms of travel to work areas, or rural district cooperation, or counties. This is easy enough, given the definitions of the new areas in terms of our standard areas.

We output a SIR SAVE FILE database consisting only of those variables which we need for this particular work, and run the tables in that. This provides significantly quicker access. For example, a job which takes three hours or more in the full hierarchical database, may take a mere ten minutes in the subset.

Flexibility

Although the SIR SAVE FILE produces a flat-file database, the structure can be augmented if necessary. For purposes of grossing on the LFS, we use only four variables. Recently it became necessary to restructure our grossing suite to meet revised targets; the option used involved regrouping the person records, and introducing a dummy household record to control the processing. We generated our flat-file database via SIR SAVE FILE and introduced a new record type, running the necessary record schema. This has worked splendidly, as indeed it should.

We are able therefore to combine the speed advantage of SIR SAVE FILE with the structural advantage of a modicum of hierarchy. The resulting hybrid is far from being a Frankenstein's Monster, and runs considerably faster than the original hierarchical database.

Conclusion

Essentially, the Labour Force Survey is a fast moving survey. Being so much larger than formerly, the results are increasingly useful to the Employment Department, and their vision of its possible application are - like the universe - in a state of perpetual expansion. Our processing techniques have to keep changing, as a consequence. It is an invigorating challenge to keep up with the demands - to regard the present methods of processing as being in a permanent state of review. SIR has been used to process the data for a good many years now, throughout the various recensions of the Survey. In that time, though, SIR has changed a good deal - but more importantly, we have been making more and more optimum use of SIR.

For more information on the use of SIR at the OPCS contact:

Michael Staley
OPCS
St Catherine's House
10 Kingsway
London WC2B 6JP UK

Back Sir Home