Friday, February 13, 2015

New ABS chief David Kalisch: Don't take our numbers literally

Literally.

The new head of the Bureau of Statistics has a disarming reply to people who complain that the bureau's unemployment data is unreliable. It's to not rely on it. 

The January unemployment rate was 6.4 per cent. David Kalisch says the media and markets should focus instead on what the bureau calls its 95 per cent confidence interval, reporting that the bureau is confident the true rate is between 6 per cent and 6.8 per cent.

It would mean reporting that the total number of Australians in jobs did something between sliding 72,200 in January and climbing 45,800. It would mean reporting that the number of Australians unemployed did something between climbing 74,900 and falling 5900. It would mean acknowledging that the bureau's employment estimates are even less accurate than is widely believed.

And he says even those broad ranges should not be treated as gospel. "You've got to look at whether the numbers are in line with other economic conditions. You've got to ask: does this number have a clear alignment with other economic indicators and expectations?"

Kalisch comes to the ABS from the Australian Institute of Health and Welfare where he was chief executive. An economics graduate from the University of Adelaide, he has specialised in labour markets for the past three decades, working in Canberra for the Australian government and in Paris for the Organisation for Economic Co-operation and Development.

He says there's a good case for merging the ABS with the organisation he used to head,  a move that would bring the bureau more money, more staff and a more externally focused culture. It's a "live issue" currently before the government.

His immediate priority is more money, a lot of it... When his predecessor as Australian Statistician Brian Pink left in January 2014, he wrote that the bureau had barely enough cash to "keep the lights on".

Instead of replacing him promptly, the Treasurer and the Prime Minister's offices tossed around options and deferred the decision until December when they finally gave the job to Kalisch, one of the original applicants from earlier in the year.

Without a chief executive for the best part of a year, the ABS faced a crisis over the arrest of a staff member who later pleaded guilty to insider trading in the labour force figures and a convulsion in the figures themselves, which in August reported an unlikely jump of 121,000 in employment followed in September by what would have been an extraordinary slide of 172,000 had the bureau not disowned the figure and substituted it with something less volatile.

Kalisch says he believes those problems are behind the survey, but he says people need to recognise that it is just that – a survey, of  about 26,000 households conducted once a month. "Any sense that the number is exactly whatever we report to the second decimal point is not an accurate use of those numbers," he says.

But when asked whether the bureau's numbers were accurate to even the first decimal place, he deflects the question and says it is "wisest to look at  the published confidence intervals". They show the bureau is not always confident to first decimal place. They put the true unemployment rate at anywhere from  0.4 of a percentage point lower to 0.4 of a percentage point higher than the published rate. Kalisch says there's nothing new in this. The range has always been wide.

The bigger problem with the figures is that they are still being prepared on outdated stand-alone computer systems. Some are up to 40 years old and can't easily talk to each other. They are kept going by the idiosyncratic knowledge of the ABS staff who have nursed them for years. "To make things work requires a number of people with insights into that particular way of, operating," Kalisch says, acknowledging that "knowledge retention is an issue".

The calcified operating systems can't quickly accede to demands to do things differently. They are built to produce things in the way they have always been produced, not to experiment.

Kalisch wants several hundred million dollars from the government for an entire technology refresh – almost as much, but not quite, as the ABS annual $312 million budget.

The project will take at least four years – Kalisch has been appointed for five – and will draw from the experience of other organisations, such as banks, that have had to retool while continuing to deliver their core service.

It will also provide an opportunity to change the nature of the bureau, from an organisation that primarily conducts its own surveys to one that makes greater use of outside data, curating it and turning it into useful products.

The bureau is already using supermarket scanner data to help compile the consumer price index, relying less on its own staff scanning shelves with clipboards.
His vision is for the bureau to mine so-called administrative data held in places such as the tax office, Medicare and education systems to deliver products that tell us things we don't yet know, such as how a child progresses through the school system, and perhaps how that child gets ill along the way.  

These days the ABS is in an environment awash with information, he says. In the 1980s it was something of a monopoly provider.

Kalisch says he has found the staff up for change. "I am finding a workforce that understands we need to engage with a changing world and wants to be part of it. I am seeing more enthusiasm for change than I probably expected."

Coincident with his arrival, the government is considering merging the ABS with his old employer, the Institute of Health and Welfare. The 2014 Commission of Audit recommended merging the Institute with smaller health organisations. Kalisch thinks the ABS could be a better fit.

"Both organisations have a strong adherence to statistical standards," he says. "The institute has a strong reputation for engaging with other stakeholders, and it is that dimension that I would like to see more pervasive in the bureau. If government was to accept that as an operating model then I think there would be some synergies in the way both organisations operate. I think they would benefit each other."

He gives the impression the ABS he will leave in five years will be an organisation different from the one he has just joined.

In BusinessDay


Related Posts

. 2012. Lunch with Australian Statistician Brian Pink

. Dumbing down the ABS. It's like driving at night dimming the lights

. Wow. Would Australia's ABS tick off our prime minister like this?