ONS have published findings which show that, although the Census Coverage Survey largely met the important statistical requirement of independence from the actual Census enumeration, an element of dependence between the Census and the Survey was identified. As independence was a key factor in ensuring that the final Census estimates produced in the One Number Census process were of high quality, these findings describe how this dependence was measured and how the ONC estimates were adjusted to produce a more accurate estimate of the 2001 population.
The One Number Census (ONC) project was developed to adjust the results of the 2001 Census for under enumeration - that is the number of households and people not counted. This was achieved through carrying out an additional survey - the Census Coverage Survey (CCS) - after the Census which enumerated a sample of postcodes throughout the UK. By matching the results of the Census Coverage Survey with the Census, it was possible to estimate the number of households and people missed by both.
It is important to this method of estimation that the CCS was independent from the Census. That is, the possibility of someone being captured by the Census is not related to the possibility of being captured by the CCS. The estimation of persons missed in both surveys may be incorrect if there was significant dependence between the Census and the CCS. It was recognised throughout the ONC process that there might be some dependence, and the methodology was designed to keep the level of dependence to a minimum. However, a statistical method was developed which was used to adjust for dependence between the Census and the CCS. This adjustment has produced a more accurate estimate of the 2001 population.
There is now a paper on the NS website with an overview of the methods used to measure dependence between the Census and the CCS HERE. It also describes the adjustment that was made to account for it. There is also an earlier draft of this paper, published in December 2002 on the NS website, which has extra technical detail about the dependence adjustment HERE.
Census News
