On 10-Jan-04 Ray Thomas wrote:
> There are a number of important matters covered in Ludi's judicious
> paper.  But what caught my attention is the way 1s and 2s have been
> eliminated (to protect privacy/confidentiality etc) that must surely
> conflict with several Code of Practice principles?    Ludi says that
> the Census output does not identify the 0s and 3s that are genuine for
> those that are the results of elimation of 1s an 2s.
>
> Can it really be that the ONS ( for reasons of confidentiality) is
> saying that ANY census statistic of 3 or less could be 0, 1, 2, or 3?

It looks very much as though this is the case!

This kind of thing is reminiscent of the corruption of flora
distribution maps (in which  the country is divided into 1km squares,
a square being marked with a colour if a given species is present),
whereby for rare flora the mark may be displaced to a neighbouring
square. The idea is to make it more difficult for collectors of
rare specimens to locate them (though it strikes me that a knowledgeable
specimen-hunter, who knew what sort of niche to look for, would probably
not be much hindered by this device).

In either case, information is being adulterated, with the result that
genuine use of such data may be corrupted.

Likewise, while it can hardly be of great concern to the global overview
of the nation whether there are 0, 1, 2 or 3 unemployed in Grassington,
N Yorks, corrupting the true value could distort, and in particular
render less precise, and inconsistent, the estimated distribution of
unemployment in small rural communities. As such, in my view it is to be
strongly deprecated, unless it serves an over-riding purpose.

However, is this purpose (the protection of anonymity) well served here?

If you see a figure 0 or 3, then you know that it could be 0 or 1
or 2 or 3 and, if you are the sort of person who wants to actually
identify who are the unemployed in Grassington (presuming that you
would be equipped to do so if you knew exactly the number, e.g. 2),
then I suggest that you would not have much more difficulty in
the identification if you go there on the basis that it might be
1, or 2, or 3; and presumably it would not take you that much
trouble to suss out that it was 0 if that really were the case.

So I can't see that this trick of "adjusting" to 0/3 offers much
protection anyway, so the "over-riding purpose" would not be well
served. And how much difference is it going to make to this if the
result is 4? This figure would be reported accurately (one hopes),
yet the task of identification would be similar.

For such reasons, I therefore side with the Scottish view, according
to Ludi:

  "The adjustment procedure has not been adopted by the Census
   agency in Scotland, where the Registrar General feels that the
   risk of disclosing personally identifiable information is not
   sufficient to warrant the cost and impact of adjusting the census
   data in this way."

[Ludi's RSS/ONS paper, p.4]

Best wishes to all,
Ted.


--------------------------------------------------------------------
E-Mail: (Ted Harding) <[log in to unmask]>
Fax-to-email: +44 (0)870 167 1972
Date: 10-Jan-04                                       Time: 13:06:36
------------------------------ XFMail ------------------------------

******************************************************
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]
*******************************************************