At 18:22 29/10/03 +0000, Paul Spicker wrote:

>There seems to be a confusion about the purpose of a census; it's about
>much more than the total population count.  The census is also about where
>people are, what circumstances they live in and so what the demand for
>local services is likely to be.  It provides a basis for checking
>estimates of population change and movement, denominators for  future
>survey work an neighbourhood statistics,  and, not least important,
>resource allocation.

Indeed - that's exactly the point I made in my message yesterday.

Combined with an effectively-maintained register (if such a thing is
possible/ practical), one could go a fair bit further than just a total
count, since there would be at least age/sex/ethnic origin information tied
to location, but there would certainly be a lot of other information (e.g.
about 'living circumstances', employment etc.) that would be missed.

>There is no sense in which ID cards as currently used in Europe can
>replace the functions of a census.   Civil  status, which identity cards
>record, records very few of the things that matter.  We also know from the
>experience of the poll tax that people move far more frequently than we
>used to think and that, quite apart from the problem of compliance, the
>effect of any procedure which fails to record location will vitiate any
>attempt to produce small area statistics.

Again, I largely agree.  One can envisage situations in which attempts were
made to 'automate' notifications of change of address when people bought
homes or entered into tenancy agreements, but I imagine that would be
almost unworkable, and would almost certainly miss out on a lot of people
who were not actually 'signatories' to the purchase or tenancy.  Can one
imagine, for example, that any system would be likely to successfully track
the movements of babies as they moved from one rented property to another?

>There is a separate question, though: do we need a census to be
>national?  In France, they're moving to a rolling census with additional
>sampling.  That's much more in keeping with the purposes of a modern
>census, and it has the advantage of providing triangulation on population
>estimates elsewhere.

Indeed - there are many ways in which it can be done.  Apart from anything
else, most of those reading this will be approaching the situation from a
viewpoint of a statistician.  With such a viewpoint, it seems that it is
almost certainly wasteful of resources to attempt a 100% census of such a
large population, given that even the smallest sub-units of real interest
probably have populations of many thousands, more usually tens of
thousands.  It goes without saying that a tiny random sample of the
population would be adequate for 'national statistics' purposes, and I also
imagine that relatively small (under 10%) random samples would be adequate
for almost all local/regional purposes.  If one is going to contemplate
attempting a 100% census, this implies the existence of an adequate
'sampling frame' for sample surveys.

Kind Regards


John

----------------------------------------------------------------
Dr John Whittington,       Voice:    +44 (0) 1296 730225
Mediscience Services       Fax:      +44 (0) 1296 738893
Twyford Manor, Twyford,    E-mail:   [log in to unmask]
Buckingham  MK18 4EL, UK             [log in to unmask]
----------------------------------------------------------------

******************************************************
Please note that if you press the 'Reply' button your
message will go only to the sender of this message.
If you want to reply to the whole list, use your mailer's
'Reply-to-All' button to send your message automatically
to [log in to unmask]
*******************************************************