Lies, damn lies, and ABS statistics

Working for Treasury used to be the goal of the crème de la crème of economics students, while the Australian Bureau of Statistics (ABS) provided potential government employment for also-rans. But back then Ken Henry would have been in short pants and Treasury still had its reputation intact for providing intellectually-driven, impartial, advice. Its disintegration and final dissolution marked by the magic pudding mining tax, which Barnaby Joyce so relished ridiculing, was decades away.

How the worm has turned! Now everyone hangs onto the latest headline statistic coming out of the ABS to see how it might move the market and/or shift the political pendulum. Treasury, well, it appears to be run by a sloppy and compromised crew who can’t seem to get within cooee of forecasting revenue accurately, and who are forever trying to help out their spendthrift mate Wayne Swan by foraging about to see how much extra tax they can raise from elderly, infirm, superannuants to make up for the revenue shortfall they consistently fail to predict. Perhaps that’s a bit unfair. No, on reflection, perhaps it isn’t.

The impact of whatever is the latest statistic depends on how accurately private sector economists have guessed its order of magnitude in advance. It’s all a bit of a yawn if it falls within expectations. For example, something like, “retail sales rose by 1% last month, exactly in line with expectations”, isn’t the kind of thing to get the market excited. Fortunately for the ABS, many of its statistics are based on sampling. Sampling error is hard to guess. This doesn’t stop private sector economist from trying, of course, because that is what they’re paid to do; however silly that might seem to the common man.

Sampling error means that the ABS is often prone to issuing statistics which lie outside of expectations and thus create a stir within which the ABS can bask. However, the size of the fish matters. Accolades might flow from stories about catching a rainbow trout 2 to 3 foot long but once it gets much bigger it becomes a “fishy story”. And so it was with the latest employment statistics released last week.

The ABS reported that seasonally adjusted employment grew by over 71,000 in February. According to the Prime Minister, who took the opportunity to berate the Opposition for not exuberantly celebrating such good numbers, this was the biggest monthly growth in employment for over 12 years. The raw figures were even more impressive at plus 128,000.

The problem is that no-one believed the numbers. Are we to really believe that with manufacturing and tourism struggling as a result of the high dollar, not to mention the carbon tax, mining investment peaking, non-dwelling construction investment going backwards in the December quarter and dwelling investment growing only modestly, that 128,000 Australians (0.7% of the entire civilian population 15 years and older), previously unemployed, gained new employment in February? It simply didn’t happen.

Ms Gillard was in fact berating the Opposition for not hallucinating on a fictional account of employment growth as wholeheartedly as she did. If Ms Gillard is still around as PM when the next employment numbers come out (for March) on 11 April she might find that “tripping” has its down side.

Reliable point estimates of employment growth and most other statistics are non-existent. The employment statistics produced by the ABS are based on a household survey and have a large sampling standard error. The ABS reported that it was 95% confident that total seasonally adjusted employment grew by between 16,500 and 126,500 in February. 

This way of putting things just doesn’t do it for economists and politicians. They prefer false precision to the vagaries of the genuine article.

The ABS produces trend estimates which uses a moving average process to iron out untoward monthly movements. These show that employment grew by 16,000 in February, which is probably close to the mark given everything else we know about how the economy is travelling. Presumably, the monthly survey produced a positively-skewed sample outcome for employment growth.

If this proves to be right we can expect employment to regress to the norm next month and Ms Gillard – no, with bad news, probably Mr Swan – will need to construct a story about how Tony Abbott is destroying confidence to explain away the fictitious monthly fall in employment.

Finally, the Reserve Bank will not take too much notice of monthly employment figures in setting interest rates – neither the reported rise in February nor, of itself, the likely report in April of a fall in March. There is not much on the horizon to suggest the economy will do, at best, much more than continue to mark time. The next interest movement is therefore likely to be down and this would not be the case if employment, in fact, grew by anything close to 70,000 last month.

Peter Smith, a frequent Quadrant Online contributor, is the author of Bad Economics