1991 Census Appendix 33

The development of a 1981-1991 migration indicator

As current LS users will be aware, the LS data for 1981 includes a migration indicator showing whether the LS member moved between censuses. This was the result of a huge clerical exercise involving a comparison of the addresses found on LS members' 1971 index cards with those for 1981. To repeat the exercise would be prohibitively expensive and take up to 2 years, so a new method of calculating a migration indicator has had to be devised.

As the LS contains ED (enumeration district) centroids for both 1981 and 1991 it was decided to base the new 1981-1991 migration indicator on a comparison of EDs. (An ED centroid is defined as the central point of the most built-up area within an ED.) This method will maximise the speed of the operation while minimising costs. However, there are a number of problems with this method:

i. ED centroids can move between censuses - usually this will only be a few metres, but in some cases the difference between the two can be substantial.

ii. There appear to be a large number of errors in the centroids held on the OPCS computer.

iii. A number of LS members in 1991 have complex multiple enumerations. (A complex multiple enumeration occurs where an LS member is enumerated in two or more locations and where it is not possible to establish the person's usual residence or where s/he spent census night.)

The above problems have been given a great deal of consideration and will be dealt with in the following way:

Firstly, it is absolutely crucial that non-moves are coded as such. For this reason any move under 0.5km will be treated as a non-move. This can be augmented by using a "distance moved" variable held on the LS database and coded to the nearest kilometre, which indicates whether an ED centroid has moved at all. In the case of non moving centroids, this highlights whether there has been a move within the ED or not. The second problem has partly been resolved by plotting the centroids. Obvious errors, such as centroids in the North Sea have been ironed out and a further investigation of this problem will be carried out when the new variable is tested. Finally, a decision has been taken to disregard those LS members with complex multiple enumerations as the difficulty of establishing their correct address of enumeration would have led to a substantial delay in the release of the variable. OPCS will be beginning this work during February or March and the new variables should be ready for use by Summer 1995.

Simon Gleave