The 1991 Census is estimated to have directly counted only 96% of the normally resident population. A further 2% were identified in the form of absent households and 'counted in' by a process of 'imputation', leaving around 2% (around 1 million people) of the population who were thought to be missed entirely: the so-called 'missing million'.
Imputation: One of the tasks of enumerators is to record households that are thought to exist even though no census form has been returned. In 1991, these 'wholly absent households' were added into the census through a process of imputation. In 1991 about 2% of the published census counts were imputed people. For further information about the imputation process in the 1991 Census see Dale and Marsh.
The 'missing' million: Even after imputation of missing households, counts from the 1991 census appeared considerably lower than had been expected, based on the most recent estimates of the national population from the Registrar General (RG). These estimates were based on the previous 1981 Census, annually updated for registered births and deaths and an estimate of migration. For more information about the RG official estimates see: Making a population estimate in England and Wales [Note: This link opens in a new window]
The main method for checking how well a census counts the population is through some form of post-enumeration survey. This involves a repeat population count for a small sample of areas, which is then compared to the census count to inform the extent of underenumeration and the accuracy of responses. In 1991 this survey was called the Census Validation Survey (CVS). It was hoped that the 1991 CVS would identify an undercount to explain the shortfall between the census count and the Registrar Generals official estimates.