Revisiting the reliability of ABS' GDP growth estimates - 2002 RBA research paper

Estimates of GDP growth have, at least for some periods, been subject to substantial alteration for quite some time after they were first released—sometimes for as much as 5-10 years. This issue was the subject of one of the more interesting past RBA research discusison papers—Real-time National Accounts Data by Andrew Stone and Sharon Wardrop (September 2002).

The often overlooked fact is that the first release by the ABS of the Australian National Accounts must be treated as these estimates are intended—subject to substantial revision. The mean absolute size of the contemporaneous estimation errors for through-the-year GDP growth were 0.9 percentage points for the 30 year period 1971 to 2001. Significantly, this average error had declined by only 20 per cent two years after the initial estimates. However, there appears to be little bias in these estimates, with the average error being only an over-estimate of 0.1 percentage points.

The following chart illustrates the often substantial and erratic revisions that can occur to GDP growth estimates.

The paper notes that "quite a number of the initial estimates of quarterly hybrid GDP growth have differed by more than 2 percentage points from the ‘true’ value for the quarter". The paper also notes that, "while remaining the best guide available, such data can in isolation sometimes lead to serious misperceptions about where the economy is and, in turn, where it is going".

It should be noted that the errors have declined over time, which indicates that the ABS is probably getting better in formulating its estimates. However:

• with a few exceptions, the last 20 years has been a relatively stable/predictable period of economic growth

• the estimates errors for the latest periods depend on the latest estimate, and consequently exhibit an explicit downward bias

I am not aware of this work being updated anywhere. It would make an interesting research project.