The 1991 CVS was a relatively small survey conducted in 1,200 census enumeration districts, selected randomly to be nationally representative. Specifically its aim was to check that all the persons present on census night were correctly counted in the census, and also to check that enumerators had made correct decisions around the identification of vacant dwellings and absent households. The CVS results were then used to derive an estimate of the national undercount.
Although the 1991 CVS detected an undercount of just under 300,000 people in households, this was well short of the estimated 1 million people that were presumed missed by the census, based on comparison with the Registrar General's estimates. Moreover, confidence in the CVS was further undermined by the observation that the adjusted Census counts indicated an implausibly high ratio of young women to men. As a result, it was decided that the Registrar Generals population estimates based on the 1981 Census were a more reliable indicator of the population total in 1991. For more information on the 1991 CVS see Heady P, Smith S, Avery V (1994) 1991 Census Validation Survey, Coverage Report, HMSO, London.
While a coverage of 98% (96% enumerated + 2% imported) is high in comparison with international standards, there is evidence that the level of undercount varied considerably between subgroups in the population and from area to area (so-called 'differential underenumeration')