Since I was not present at the DPC for the 2001 Census, I am not in a position to compare the functioning of that DPC with the 2006 DPC. It is possible, however, to assess whether the procedures put in place had strengths and weaknesses in their own right. My overwhelming impression—from my own observations and from conversations with IPT and other DPC staff—was that the IPT was a worthwhile innovation and that it should be retained for future censuses. It is needed for the same reason that the IES itself is needed—the ‘difference’ of Indigenous people in remote areas of Australia—and indeed it should be considered from now on as part and parcel of the IES.

It was found in 2001 that automatic coding (AC) was not successful in coding certain questions on the Indigenous forms, and for certain questions (as noted above) AC was switched off and the IPT coders were instructed to code manually. In this situation, there are advantages to the DPC in having to train only a small number of coders to work with the IHF, and having a small cohort allows for efficient quality assessment and feedback to the coders.

The IPT followed a policy of attempting to make minimal use of the option ‘Not adequately described’, and this meant the coders had sometimes to be quite lateral thinking in their coding solutions. In order to maintain consistency across coders, it was necessary to give feedback constantly to individuals and the group, and to monitor for patterns in the solutions adopted, particularly for less adequate solutions. When such patterns were noted, ad hoc tutorials were held to help the coders achieve better and more consistent solutions. All this would have been much more difficult to maintain rigorously with a larger number of coders.

The other major advantage of a specialist unit is that it can serve as a repository of specialist knowledge. A great deal of background research had been done to assist the coders in such matters as identifying the Indigenous languages that appeared on the forms and in compiling exhaustive lists of community names with their variants, and the family coding and the coding for occupation and qualifications required extra knowledge over and above what was required for the mainstream forms. Again, it is more efficient to train, monitor and assist a relatively small number of specialist coders rather than attempting a more general exercise involving all coders.

In assisting the coders to do their work, the DPC has to maintain a delicate balance. It must provide them with enough information to code accurately and quickly, while avoiding the pitfall of providing information that might bias them towards particular interpretations of the data. It is a cast-iron rule of coding that the coder must work with whatever information they have—and only that information—and this rule is in creative tension with the imperative to avoid coding a response as ‘not adequately described’ wherever possible.

The additional sources of information that were available to IPT coders included the community forms from the Discrete Indigenous Community Database (DICD) and the CFO and CMU checklists described in the previous chapter. They had access also—on request to the data-analyst on the floor—to extensive materials on Indigenous languages and on localities. These last two were invaluable and, for the Northern Territory at least, the CMU and CFO checklists often provided useful additional information, such as how the PTAs for a particular CD had been treated. The DICD forms were less useful, except as a check on the number of people and dwellings in a CD. As noted in earlier chapters, the CFOs had not, by and large, completed these in any great detail. Coders were encouraged to add to the forms, for example, by recording variant spellings of language names that appeared on the forms. If these forms are attended to and updated during the inter-censual period, they will potentially be an invaluable resource for the coders in 2011. This database is to be maintained by the National Centre for Aboriginal and Torres Strait Islander Statistics (NCATSIS), and we have suggested in Chapter 9 that the updating should be the responsibility of Indigenous Liaison Units—rather than a solitary State Indigenous Manager—located in the State and Territory offices.

Another procedure followed by the IPT that differed from the mainstream processing was to base processing on CDs. In the mainstream processing, coding was topic based, with different teams of coders working on different topics rather than on whole forms. The IPT procedures allowed the coders to become familiar with patterns of naming and other reoccurring information for particular communities, leading to more consistency in coding. From my observations, this generally worked very well. For example, once a coder became familiar with all the language names for a particular community, this speeded up the coding process considerably. It also helped the team managers in their quality-assessment work, since if a coder made a particular error it was likely to be repeated for the whole CD, showing up as a clear pattern.

In order to allow for flexibility in coding when faced with variety in the written responses to questions, the coders worked with colour-coded ‘pick lists’. Some choices had to be exact matches, others allowed for close or approximate matches. As a last resort, the coder could choose to bypass the pick lists and go into a ‘best-fit’ process. In best fit, the coder was required to state their reason for choosing that path and this will presumably allow the ABS data-analysts to refine the pick lists for 2011. This seems to me to be a good idea, because it allows the coders some latitude with aberrant answers, and also allows them to signal ‘gaps’ in the options offered. It is unclear to me, however, how the subtleties of these procedures will be translated into the output data. I noted that coders differed in their propensity to opt for going into best fit. Some tried several pathways within the pick-list system and went to best fit only as a last resort, while others were quicker to opt for the best-fit solution.