THE scene is Bethlehem 1AD. A Roman soldier, Perspicuus, is muttering: “Almost there. MMMDCCLVII, MMMDCCLVIII.” Then loudly to assembled motely mass of Palestinians: “Alright, you lot, you can clear off home now.” He turns to another soldier, Errorpronius, and says, “Ok, add this to the MMDCLVI you got at Nazereth and take away one for the bloke you crucified for mutinously suggesting there was an easier way of counting.
“Then we can get these figures off to Rome so they can do something about the appalling inequality in this Province. Can you believe that Herod fellow with his VII-bedroom and IV-stable house, and he is only in the XXV-XXXV age group?”
Errorpronius grunts and says: “Five Ms, two Ds, three Cs, two Ls, two Vs and four Is equals . . . . Aaagghh.”
Perspicuus: “Maybe there is an easier way. Why didn’t we ask every tenth person for the information we wanted? It works in the battlefield when we execute every tenth person. The message gets through. Decimation — that’s the answer.
“Next stop, Ramullah. It’s a village of about VIII people. ”
This scene popped into my mind on Tuesday night, Census night. I wondered was this $440 million exercise really necessary? Would it be value for money? Could the money be better spent?
In these days of deadly accurate sampling can the 2000-year-old Roman headcount be dispensed with.
The Australian Bureau of Statistics information sheet on this census tells us, “Australia has a population of 22,493,120 people.” Well why are we spending so much money counting them?
The acid question is whether the cost of a head count is worth extra accuracy you might get over surveys and other methods of gathering information, such as registries of births and deaths and counts of net immigration.
It will be interesting to see how close the population count from the census comes to the ABS’s population clock reading at midnight on census night of 22,6742,370. If it is not far out, the need for a headcount to get an overall population estimate loses value.
Bear in mind, too, that the headcount is not going to be totally accurate. It will miss some of the homeless and people who are in hiding or are here illegally – the very people who would not show in official registries and immigration figures. Further the population estimates might make an educated guess at the number of these people and so be more accurate than the headcount.
The census form says the headcount “is the only practical way to get information on how many people there are in each part of Australia, what they do and how they live”.
The census divides Australia into units of roughly 200 household and the data is categorised for each unit. It is about the smallest practical unit without compromising privacy. These units can be aggregated with the data to provide snapshots of larger units, such as postcodes, local government areas or whole states.
Then the data can be divided in other ways – by gender, income, rural/regional/suburban/city, or other attributes. Again these can be aggregated again other criteria.
Again, is the expense worth the level of accuracy? It would be impossible to do surveys of samples in a group as small as 220 with any degree of confidence. So the ABS is right to say the census “is the only practical way to get information . . . in each part of Australia” but only if you require those “parts of Australia” to be quite small.
But if you accepted a larger statistical unit – say a federal electorate of 100,000 people – you could be 95% confident that a random survey of 1200 would get you within a few percent of the true picture.
Australia-wide that might cost you a couple of million, against $440 million for the census. Is the extra accuracy worth it? What policy or service-delivery questions require detail greater than a federal electorate?
And again, bear in mind that the census is not going to be totally accurate on lot of questions.
Some biases can throw the data out – household forms for a start. Partners in the presence of each other are bound to exaggerate the amount of unpaid household work they do, for example.
Equally, they might lie over religion, pretending to go along with the family’s religious tradition when in truth they are agnostic or atheist.
There is another bias in the religion question. I see the Catholics get the donkey vote at the top of the box, followed by eight other options. Then there is “other please specify” and a box of four rows of nine boxes giving 36 characters for someone to specify another religion. And tucked away under that box and easy to miss is the choice “no religion”. My guess is that that layout will artificially depress the “no religion” count.
The questions asked in the census also bear some scrutiny. This one marks 100 years of counting. It seems some of the questions have an historical grounding in White Australia paranoia – let’s make sure we remain Anglican and of British stock. Are the questions on religion and ancestry really necessary? What policy considerations require the responses to these questions at the level of the 200-household statistical unit?
If you are spending $440 million you want to ask only questions which require that level of detail. Indigenous backgound would be one because of the high level of indigenous disadvantage and the need to address it at a micro level. A military base or a mining community might be another reason for detail because they would skew the results from surrounding areas.
But policies related to ancestry or religion hardly require that detail. You could get a good result with a national survey.
Maybe there is some community comfort in retaining questions asked in previous censuses and some comfort in the headcount given the widespread ignorance of the statistical theory behind the confidence you can have in good sample surveying. But is it worth $440 million?
On the whole, the ABS does a terrific job and I have no problem with data security. It is likely that the level of community expectation for the headcount is too high to permit its the replacement with surveys, about which many are suspicious or ignorant.
Perhaps the real advance on the Roman method of counting will come with the gradual replacement of paper forms and human collectors with the internet. Surely not two millennia away?
CRISPIN HULL
This article first appeared in The Canberra Times on 13 August 2011.