Census 1961, Great Britain : general report.

  • Great Britain. General Register Office.
Date:
1968
Catalogue details

Licence: Public Domain Mark

Credit: Census 1961, Great Britain : general report. Source: Wellcome Collection.

    28/238 (page 12)
    The expenditure includes the costs of the Post-Enumeration Survey, which were not estimated for and amounted to £26,000 of which £10,000 went to the Census Officers and the remainder to selected Enumerators. Delay in producing results With the use of a computer it was estimated that all results would be pub- lished within 23 years of census day. Actually, although the preliminary report was published within 2 months of census day, the first county report was not published until March 1963, and the national reports were published from September 1964 onwards. It was 44 years before computer running was finished and 54 years before the last tables were published. The delay arose mainly from the combination of two factors, namely lack of experience in processing a large statistical exercise like the census on a computer, and on increasing demand for census statistics throughout the planning and processing stages. This increase in demand was part of a general expansion in the use of statistics for planning and could not properly be resisted; but, coupled with inexperience of computers, it resulted in a serious under-estimation of the number of programmers and the overall computer capacity required. Programming began in the autumn of 1959 with the object of completing all programmes within 4 years. Eight programmers were trained initially but in 1961 a further 10 were recruited and trained. Eventually 72 man years were spent in programming over a period of six years compared with the original estimate of 32 man years over four years. In addition programming for all the economic activity tabulations was put out to contract. No additional programmers were recruited after 1961 because the long training period necessary before they would become effective would have diverted trained programmers from the main task, Also the computer capacity available was only sufficient to justify the employ- ment of the existing staff. The amount of computer running time required to process the census was seri- ously under-estimated, So much so that it was considered that the census could be easily processed by sharing the I.B.M. 705 installed for the Royal Army Pay Corps at Worthy Down. However by 1961 the R.A.P.C.'s own requirements had so far expanded as to exceed the capacity of one machine, and an additional machine was obtained in 1962, Both machines were then worked three shifts a day five days a week for the next eighteen months with occasional weekend working. In the final twelve months a large proportion of the census work was done by contin- uous shift work at weekends. Eventually some 15,000 hours of computer running time were used for census processing, equivalent to almost 4 years work on a double shift, not including the time spent in processing the main economic activity tabulations, To have completed census processing in 24 years as planned would have required the use of an I.B.M. 705 for three shifts a day throughout the entire period, and a programming team capable of keeping it provided with work, The position was aggravated by the fact that census processing was the junior partner in a shared arrangement, so that computer running time was not always readily available when required and the most effective use could not always be made of the time when it did become available. These unforeseen difficulties of using a computer contributed about 16 months delay to the publication programme, The remaining delay was due to quite different factors, First, the discovery that the ten per cent sample included within the census was biased led to extensive calculations to estimate the effect of the bias. This contributed an additional 3 to 4 months delay. Secondly the time needed t« programme and investigate the results of calculations of empirical sampling variances, in order to assess the precision of the results of the ten per cent sample, was much longer than expected, contributing a further 3 months delay to the publication programme. The experience gained in the course of processing the 1961 Census was, how- ever, invaluable, The lessons learned have been applied to the 1966 Census, with the result that the basic aim of publishing the main results within two years of census day is likely to be met.