ABS accuracy is all in the mind

Increasingly the health of the domestic economy is whatever you think it is -- literally.

To state that the ABS’s recent data problems are embarrassing is a fairly serious understatement. It’s not that the previously reported spike in employment should have been taken as truth. As I noted at the time it should have been taken with a grain of salt (A dose of reality with the jobs data, September 12). But then, so should every month’s result --  August wasn’t an exception.

The bigger problem is that the veracity of the ABS’s entire methodology, their competence, has now been brought into question. Can Australians rely on the figures the Bureau produces? Clearly we cannot. The labour market has gone from looking quite healthy, to now looking quite weak -- in one month. Noting this, it would be unreasonable to suggest that such errors are isolated to the labour force survey alone. What’s to say, for instance, that the seasonal factors applied to the national accounts aren’t wrong as well? Or the monthly retail sales survey? You don’t need errors of the magnitude that we’ve seen in these recent employment results to produce an entirely different economy.

As one, not so small example, the seasonally adjusted (real) measure of consumer spending shows consumption rising only 0.5 per cent per quarter, on average, over the last couple of years. That annualises to about 2 per cent -- a pace well below trend and clearly soft. On the original figures, however, that jumps to 0.8 per cent per quarter. It’s an average growth rate over two years, so no seasonality should be present. Yet on this basis, consumer spending is in line with normal growth. All of a sudden consumers aren’t so cautious.  Economic growth more broadly has been running at an above-trend pace on the original figures -- a brighter picture than has been painted at times.

Don’t get me wrong. Seasonally adjusting data makes sense. Yet it clearly is more art than science and highly subjective. In a way, I guess that’s the big positive out of this debacle. It highlights the simple fact that even the most robust statistics are, ultimately, subjective. A work of fiction even.

Noting this, it should be the aim of statisticians to reduce to the subjectivity of the data, not increase it -- which is why the ABS decision to apply seasonal factors randomly is simply unacceptable. The issue isn’t whether the data is volatile or that we get rogue results. It is and we do. The decision would, however, introduce much more uncertainty than currently exists.

Citizens, business and market participants would be far less certain, at any given time, of the extent to which bias has been introduced into the figures. That this occurs anyway I don’t doubt -- even subconsciously -- and small changes to seasonal factors can produce an entirely different economic picture. Indeed, I recall a conversation I had some years ago with an ABS employee: Having questioned some data the reply came “don’t you read the papers?” Only one employee, sure, and a person who may not even work there anymore. Yet that conversation highlights a point. Statisticians are people too -- and the media is all-powerful!

Having been bruised, battered and humiliated beyond question, the risk is that this bias becomes amplified. The ABS is going to want to fly under the radar for a while which means producing figures that don’t stir up much controversy. A policy of no surprises. This makes the ABS a little more amenable in producing statistics that simply reflect the current consensus (of weakness). I’m not suggesting this would be a deliberate policy, but the impulse would be hard to resist -- and as we know, applying seasonal factors is artful at best.

So how healthy is the economy? Increasingly it’s whatever you think -- literally. For what it’s worth, and I realise it’s probably not much given the incredibly low quality of the data, I still think the labour market shows quite healthy outcomes. Dismiss these last two results -- original or otherwise. The pattern in place, prior to these months, was that jobs growth had picked up markedly (from flat growth last year to 17,000 on average this year). I see no reason for that to have changed. Economic growth is above trend and momentum is accelerating. At least it was, we’ll see what pressure the ABS comes under. Housing construction is otherwise surging and that sector is a big employer. Finally, the unemployment rate has been constant at 5.9 per cent. All up not a bad set of metrics.