The Bureau announced yesterday that in the coming months it will be shrinking the size of its monthly employment survey from around 65,000 Australians to 55,500 – its smallest sample to date...
The head of the Bureau’s labour and demography branch, Paul Sullivan said that recent improvements in the way the data was interpreted meant that it should be possible to cut the sample size without compromising the accuracy of the employment estimate.
But the unemployment estimate – a smaller, harder to obtain figure – should become “slightly” less accurate.
Mr Sullivan said the smaller sample should save the Bureau around $100,000 each year, some of which it would need because the booming labour market itself had made the survey more expensive.
The Bureau tries to collect data from each household it surveys for eight consecutive months. But with fewer people at home during the day that is becoming more expensive, requiring more repeat phone calls and return visits.
Even at the reduced sample size of 55,500, the Bureau’s labour force survey will remain the biggest apart from the census. Each political opinion poll quoted during the election campaign surveyed no more than around 2,000 people.
The Bureau has been making its estimates more accurate in recent months by under-weighting the answers given by people who are responding for the first time. These answers are thought to be less reliable and they don’t reflect changes in employment status in the same way that the subsequent answers do.
Under the new procedures the Bureau will much more intensively sample residents of the Northern Territory who were previously only as intensively sampled as residents of the ACT.
After the changes one in every 54 Northern Territorians will be asked about their employment status, up from around one in every 100 as in the ACT - greatly improving the accuracy of the Northern Territory’s figures.
Paul Sullivan said that the increased inaccuracy of the Australia’s unemployment total would not usually be noticeable by the time it had been turned into an unemployment rate. The extra uncertainty should account for less than one tenth of one percentage point, meaning that what would have been a national unemployment rate of 4.3 per cent would most probably remain an unemployment rate of 4.3 per cent.