Aside from the supposition that the raw data is not in fact copyrightable in the first place, which is probably true, even if never tested, the law clearly provides for grant of permission for data to be represented in another form, such as sound or visual forms. It seems clear that the provisions of copyright (the sections are too tediously long and legally worded to reproduce here) are precisely applicable only to the form in which the information is presented by the author(s). Moreover, if the presentation of census data as published by the CA is in fact taken to be a design form as defined by the Design Act 1911 (but to be frank I haven't looked at what that creature is), then the copyright ceases as soon as 50 copies are circulated, which has obviously already happened if the data is online.  

I believe that mapped data represents precisely such alternate forms, especially if it is dynamically presented (but even if it is not). Making it dynamic is of course a highly useful form, one that I do not believe the census authority has yet conceived. The census authority cannot refuse permission for such presentation. If they do not publish the information as is planned by our colleagues, then their copyright effectively lapses in any case, for which proof an advertisement saying that (ie that no mapped data as has been proposed has been published) must be published in a popular newspaper (in English newspaper, for English language mapping, vernacular for vernacular mapping). Unfortunately, it says nothing (that I can find) about a public announcement on the Net, so maybe this advertisement stuff in newspapers is the only path.    

It seems that one must apply in the prescribed form for licensing permission, but also note that it is not possible to refuse permission for such applications, if the end use is scientific research or educational, and also for non-commercial purposes, provided the end use is in the form of a translation. However, this permission is only automatic after 3 and 7 years (subject to relevant conditions) from the date of first publication. Even here, I put it that the date of first publication is the date when the first Census was published, and not the current census. I think that would take it back to the early 20th century, and perhaps that might also mean that the government does not (heh, heh) in fact have the right to exclusive copyright of census data (even for the 'upgraded' 60 year copyright). 

The relevant clauses are:

1. Specificity: Sec 14

2. Design: Sec 15(2)

3. Government ownership: Sec 17(d) and (dd)

4. Compulsory licensing: Sec 31 (note that the RoC may assign some copyright fee payable to the government, but prima facie it is unlikely they will do so in this case)

5. Automatic permission for translations etc: Sec 32 (sec 5(b) specifically provides for 'broadcasting')

6. Automatic permission for technical stuff: Sec 32A

7. Right to broadcast: Sec 37 (worth checking!)

8. Automatic visual recording for teaching: Sec 39

9. Possible challenge to government copyright of census data: Sec 44 (register of copyrights: quite possible that the census information has not been registered under the Act, and if so makes it impossible for the government to take action against any form of infringement - sec 50A provides for publication of registrations in the Gazette)

10. Fair use: Sec 52 a(i) etc