Peter Smith

More imaginary employment numbers

Pick a number, any number — that seems to be the guiding philosophy when the Australian Bureau of Statistics announces its latest employment estimates. Ill-served by such a ramshackle process, the only beneficiary is confusion

Apparently, seasonally adjusted employment grew by an extraordinary 74,000 in February only to plummet by more than 36,000 in March. And the moon is made of cream cheese.

The PM was on a visible high about the February numbers a month ago and now reportedly blames Wayne for the employments woes. “Here I am single-handedly, and with my winning personality," you can almost hear her saying, "forging new relationships with India, Japan, China and Indonesia and you can’t even get employment up!”

More jobs shed in March than in years, according to the Sky News business channel. More jobs added in February than for 12 years, according to the PM when commenting on the job numbers a month ago. Then we had economists talking about the possibility of employment in February being revised down before the March numbers came out; showing how little they knew about how the data is collected.

The data is gathered by interviewing a sample of households, mainly by telephone, asking: are you employed; are you unemployed and, if so, are you looking for work; or some such similar questions. They are what they are. Revisions can only be around the edges (and, in fact, the February increase was revised up by 2,500), unless we assume that the ABS is staffed by idiots who can’t clearly record answers to their questions and then add up the responses. The problem is sampling error. The ABS would do better, and narrow sampling error, by sampling employers, as they do in the United States. Maybe the ABS is short of money.

The ABS puts wide confidence intervals around its sample estimates. I don’t know why it bothers. This cautionary exercise is routinely ignored by numbers of private-sector economists who insist on acting, both in their predicting and in subsequently commenting, as though the ABS estimates represent reality.

Employment did not increase by 74,000 in February nor did it fall by 36,000 in March. What then did happen? No-one knows. Trend estimates, which even out sampling error, show employment rising by 15,500 in February and by 12,600 in March. This is the best guide we have. If employment growth of this order of magnitude were to continue, month in and month out, unemployment would rise gradually and progressively because the civilian population (of over 15 years of age) is currently growing by around 27,000 per month; that’s around 18,000 joining the workforce based on two-thirds participation. Do the math as the Americans might say.

Sluggish growth in employment is consistent with the rise in the seasonally adjusted unemployment rate from 5.2% in March, 2012, to 5.6% in March, 2013; during a period too when the participation rate has reportedly fallen from 65.4% to 65.1%. Without that fall in participation, the unemployment rate would now be about 6%. And well over 6%, if the participation rate in March 2011 of 65.7% still applied.

Job numbers, if we could get reliable ones, are the best guide to the state of the economy and central to community cohesion. GDP numbers, and the various components of GDP, are fudgy at best, and they can hide growing unemployment. Apparently GDP in Australia is growing at around 3%, if we believe the figures. That gives a false sense of economic well-being when people in growingly uncompetitive manufacturing industries are being thrown out of work.

I assume the Reserve Bank treats the monthly job numbers with the disdain they deserve but, really, it would have been brave of the Board to have reduced interest rates again earlier this month in the face of reported employment growth of 70,000 plus in February. It is quite ridiculous to have monetary policy potentially hostage to such dodgy numbers.

We need more reliable employment numbers and a properly resourced ABS to obtain them by sampling employers, and enough of them, to reduce sampling error to a relatively small margin. This would also provide a reliable and valuable reading on monthly employment growth on an industry-by-industry basis.

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

Post a comment