[LINK] Census stuff(ed)

[David Lochrin]

I wonder whether the whole census concept should be re-architected?

Much of the information could be supplied from existing government databases without going near the public Internet (ATO, Immigration, Health, state databases such as births, deaths, and marriages, etc.).  Some of it, such as personal income, would probably be more reliable.

There's a theorem in statistics known as the Central Limit Theorem if my memory is correct, and this states that the accuracy of a "well behaved" sample depends on the absolute size of that sample, not on the proportion of the total population which it represents.  (Politicians often fail to understand that.)  So once the internally assembled data was analysed, the rest of the "census" process could be conducted on a sampling basis by face-to-face interviews.  It could be treated as an audit of the assembled data and a fleshing-out of aspects of interest identified thereform, such as English language proficiency.

David L.