The Australian Bureau of Statistics (ABS) published two items last
month which suggested that its last five or six years' figures on
poverty and inequality are probably flawed.
The first sign that something was amiss came in an article published
in the April edition of Australian Economic Indicators. This is
not exactly bedtime reading, and the article went largely unnoticed
at first, but it made some rather worrying admissions.
For starters, it told us that since 1994/5, people interviewed
about their incomes in ABS surveys have not been asked for documentary
evidence like tax returns or pay slips. The result is that "the
quality of responses declined". In other words, people's reported
incomes got less accurate.
It also told us that "closer analysis" of the latest
batch of income distribution statistics raises "a significant
quality concern." Apparently, less well-off households have
been under-reporting their incomes from welfare payments. This means
that the gap between lower and higher income groups has probably
These difficulties might seem a bit technical, maybe even trivial.
But we began to see their full significance when, a week later,
the ABS published a major report called Measuring Australia's Progress.
This report identified 15 indicators of 'progress', one of which
concerned poverty and income distribution. Strangely, what the report
said about the pattern of income distribution in Australia did not
square with what previous studies have reported.
The report says that relative income inequality has remained roughly
constant since 1995: "There has been little change in the income
gap between households." But this is not what other researchers
who have used ABS data have been saying. According to them, the
relative income gap has been widening.
It turns out that one reason why the ABS findings are different
from other people's is that it has recalculated the income statistics
to try to remove some of the inaccuracies it has found in them.
Its new version of the data shows that many people towards the bottom
end of the income distribution are better off than had been thought.
Some of the most important inaccuracies in the data cannot, however,
be corrected. For example, most of the households in the bottom
10 per cent of reported incomes told interviewers that they received
less money each week than the income support system guarantees them.
Some even reported zero or negative incomes. It is possible that
a few of these claims may be accurate (small business people, for
example, might make a loss one year and have to live off savings).
But most clearly are not.
Faced with this problem, the ABS has taken drastic action. It has
removed from its analysis all those people in the bottom ten per
cent of reported incomes on the grounds that their answers are too
unreliable to be used: "The extremely low incomes…recorded
for some households in this group do not accurately reflect their
At the Centre for Independent Studies we have been warning for
some time that the data on the lower income bands are inaccurate,
and it is gratifying to find the ABS now agrees with us. Indeed,
ABS warns that using these figures will result in a "misleading
impression of the economic wellbeing of the most disadvantaged households."
So, official income data produced since 1995 are having to be recalculated,
and the data on the bottom 10 per cent of incomes are so inaccurate
that we are being advised to ignore them altogether. But where does
this leave all those reports on poverty and income inequality that
have been published in recent years using precisely these statistics?
Clearly, all this research will have to be re-examined. In particular,
'evidence' suggesting that the relative income gap is widening or
that poverty is worsening now looks less convincing than ever.
The lesson to be learned from this sorry tale is that Australia's
research community needs to be more vigilant and sceptical, even
when using the government's own statistics. All research is fallible,
and even the ABS can make mistakes.