I want to conclude this analysis by briefly addressing three related questions about the IES. How special, how successful and how necessary is it?

At one level, the IES does indeed seem quite special. It relies on interview rather than self-completion, on a separate two-form structure rather than the standard single household form- 92 - and it is time extended, over a month rather than focused on a single day. However, the questions on the special Indigenous forms suggest a reluctance to move away from the standard questions and information sought on the standard household form. The standard household form and its standard questions seeking standard information have simply been broken up into a series of forms and questions with slightly more personalised language, asking in the SIPF about 'you' rather than 'the person' and dealing with each person individually for each question rather than all people in the household together for each question as in the standard household form.

This is arguably a 'worst of both worlds' solution to the challenge of devising an appropriate census collection methodology for Indigenous Australians in discrete communities in sparsely settled northern and central Australia. The content of the census remains standard, but the collection process actually becomes more complex and elaborate. Interviewers are asked to coordinate SIHFs and SIPFs through person numbers, as well as record numbers, and to transfer items of information between SIHFs and SIPFs. Interviewees are asked to hang around during a two-form, two-stage drawn out interview process and to revisit bits of information already elicited. The 'you' language of the SIPF also actually discourages people from filling out forms for others, even though the push to enumerate people on a usual residents basis in the Northern Territory in fact relies on lots of people filling out SIPFs for absent others.

So I am not, in fact, convinced that in the Alice Springs town camps in 2001 at least, the IES was very successful at all. Indeed, it had to be modified half way through just in order to get a basic enumeration completed. In many ways it seemed that the special enumeration only improved on the standard enumeration in two respects—in being time-extended and interview-based. Another way of putting this would be to say that the 2001 Census might have been just as easily and successfully carried out in the Alice Springs town camps by sending in the interviewers armed with standard census household forms with two slight modifying instructions: enumerate all people who spent the night at the dwelling the previous night (rather than on 7 August), and complete multiple household forms for the one dwelling if there are more than six people present. What would have been 'special' about the Indigenous enumeration would have then simply been that it was time-extended and interview-based. But it would have been a much more standard census procedure in being an enumeration of people present, first and foremost, combined with the subsequent identification of absent usual residents. And it would have also been easier for people to fill in forms for others, as there would not have been identifiable SIPFs, with their personalised 'you' language.

This is perhaps a harsh judgment on the success of and need for the IES. But it is, to my mind, worth considering, if only in thinking seriously about how best to proceed in the future and what elements of the IES are indeed necessary, and which are unnecessary complications.

The alternative to reverting to a far more standardised census collection procedure, adapted only by time extension and interview processes, would seem to be to go for a far more thoroughly thought-through Indigenous enumeration process, with as much emphasis on changes in the content of what is asked as on changes in the process through which people- 93 - are enumerated and information is collected. But even if such a more thoroughly thought-through strategy were developed, I would argue strongly against any shift away from the standard census procedures of enumerating people where they are presently sleeping, rather than where they are usual residents. Enumerating those present stands as something of a bedrock principle for census administration. It is surely the most reliable way to count people and to find out from them, directly, their basic demographic and socioeconomic characteristics. When combined with sensible questioning about absent usual residents and, because of time extension, questioning about whether a person has been counted before recently, it surely provides still the best way to enumerate any population, no matter how mobile or how different its circumstances from broader community norms.