At the beginning of the training session on IHF family-coding procedures, the coders were reintroduced to the concept of the ‘statistical family’ that had been explained to them in the training for mainstream family coding. They were reminded: ‘Statistical families are governed by a strict set of coding rules that conform to international standards. Often these statistical families do not reflect the true status of family members as seen by other family members and their community.’

They were told also that this ‘applies even more’ to some Indigenous households. The definition of the statistical family is: ‘Two or more persons, one of whom is at least 15 years of age, who are related by blood, marriage (registered or de facto), adoption, step or fostering, and who are usually resident in the same household.’

Indigenous households pose particular problems for coding according to this definition, for several reasons. The ABS allows for only three statistical families in any household, and it allows for only three generations in a family. If there are more families, or more generations, additional families have to be merged with the ‘primary’ family in the household. In some cases, individuals who are closely related to someone in a household are classified simply as ‘other relative’ to the ‘head’ of the primary family once this process has dismantled their own family. It was interesting that the trainers anticipated that coders would find this unsettling—they were told to be ‘unemotional’ about following this procedure.

Other stray comments alerted me to the emotions that family-coding procedures could engender—in stark contrast with coding for occupation and qualifications, which were equally subject to a set of formal definitions and procedures. One visitor from the Darwin CMU advised a colleague, also visiting from Darwin, not to watch what was being done in family coding because of what it did to the data, and one of the coders who had lived in a discrete Aboriginal community also commented to me in passing how aware she was that family coding did not capture ‘what is really out there’. It seems at first glance that matters to do with family strike an emotional chord that is at odds with a ‘statistical’ approach; it seems to make people uneasy to see relationships being objectified in this way. That is not the whole story. People would probably feel less ambivalent if they felt that what was being captured was somewhat closer to what is ‘really out there’.

About halfway through the family-coding procedures, at a feedback session on the floor, the coders discussed difficulties they were having with family coding. They mentioned the three-generation and the three-family rules as problematic. One person also thought part of the problem was having to relate everyone to a single person, and suggested a matrix approach, in which everyone in the household was related to everyone else.

In my view, this last suggestion would be unworkable, given the size of many Indigenous households. For example, with a household of 20 people, according to my calculations, 190 different kinship dyads would have to be recorded. Even with a household of 10, 45 different kinship dyads would result. The chances of recording, in every case, the ‘correct’ Anglo-Celtic relationship would approach zero. The result would be more incoherent data, on a grander scale, and a coder’s nightmare. There is little point in asking for further relationship details for other relatives, since the data that result have a good chance of being incoherent and misleading (see Morphy 2004, 2006). I have written in detail elsewhere (Morphy 2004, 2006, 2007) about the inherent difficulties of translating between incommensurable kinship systems and I will not repeat that discussion here. Nothing of what I observed in 2006 has, however, persuaded me against the recommendation that I made as a result of observing the count in 2001: that, for coding purposes, a new type of household should be added to the ABS list of definitions—the ‘extended family household’. This type of household could potentially be useful for capturing information about more than just Indigenous households.

The default assumption, if evidence is missing or conflicting, would be that everyone in such a household is related to everyone else. Only where a person is stated explicitly to be something else—for example, a ‘friend’—would they be classified as unrelated. In the 2006 Census, the large number of PTA who were moved back inside the form—at least in the Northern Territory—had to be classified as unrelated because there was no information on their relationships to other members of their households, whereas in fact it is much more probable that they were related to other members of the household (invariably so in the cases that I observed). In 2006, the large numbers of PTAs with no relationship data who were put back into the forms will generate a lot of spurious ‘unrelatedness’ in Indigenous households.

The sub-units within these households should not be called ‘families’; they are, rather, sub-units of the extended family, and should be distinguished terminologically as conjugal units—with or without children—or single parent–child units. The terms ‘couple family’ and ‘lone-parent family’ should be reserved for households consisting of such unit types, and applied to them only.

I also see no useful purpose in maintaining the arbitrary cap on the number of such units that occur in a household. Nor do I see any useful purpose in assigning members of the grandparental generation arbitrarily to one unit or another, when they are equally closely related to members of several units. If this convention is abandoned, the necessity for the arbitrary cap on the number of generations within a ‘family’ also disappears.

Questions 31 and 32 (see Appendix A) should stay for 2011 as a means of identifying the sub-units within the extended family household. The collectors just need to be trained to implement them better. There is more chance of a ‘match’ between Anglo-Celtic and Indigenous systems when dealing with very close core kin. Indigenous CIs can in general cope with translation between Indigenous terms and the Anglo-Celtic terms ‘father’, ‘mother’, ‘son’ and ‘daughter’.

During family coding, the coders were asked to document on paper the solutions they reached for the more complex households. This was invaluable for me in my analysis of what happens in the translation process. In my opinion, however, they would be of limited use as a basis for attempting a new classification of the internal structure of Indigenous households. They do not record what is really ‘out there’. They are not neutral representations of Indigenous family structures, but artefacts of the current coding system (see Morphy 2007).^[6]

In conclusion to this section, I pose the same question that I posed in 2001: what is family coding for? In their training, the coders were told that ‘we do not try to capture relationships in terms of carers and finances’. Perhaps this is precisely what the census should be trying to capture and compare across different sections of the population. In socioeconomic terms, the family—however it is constituted—is the bedrock institution of any society, the site where children are raised, supported and socialised, and where those not in the workforce are supported by those who are. Family policy is almost invariably directed to these aspects of the family as a social institution. The question of precisely which categories of kin do or do not live in the same household is—or should be—a secondary consideration at most. The terms ‘couple family’ and ‘lone-parent family’ carry with them cultural assumptions about the nature of such families that cannot be projected simply from one cultural setting to another. For example, the child of a single mother who is living in an extended family household—where other adults besides the mother play a large part in their care—is in a very different situation to a child whose mother lives without such a supportive kin-based network.