Sent: 12-05-2009 12:15:03
In this issue:
Return to full article list
HomeFree weekly newsletterSelf Managed Super Fund ArticlesContact usLogin
Second Guessing the Statistics
Why have a national statistician if we are going to second guess his data collections? The overwhelming disregard for the most recent Australian employment data highlights an ongoing risk for our judgments about the pace of economic and market recovery.
The results of the latest monthly labor force survey from the Australian statistician were at odds with the preconceived views held by most market commentators.
Rather than a sharp fall in employment in April, the official statistics showed a seasonally adjusted employment gain of 27,300 and a fall in the number of unemployed of 35,300.
This was hardly consistent with an economy lurching into recession and so contrary to the accepted wisdom about what is happening to the economy that even government spokespeople were squeamish about acknowledging the results.
One reaction by politicians and analysts alike was to highlight the sampling errors, an arcane aspect of official statistics which rarely gains much attention in the news. The 95% confidence interval surrounding the higher employment aggregate, for example, was also consistent with an employment decline of 33,300, an outcome more in keeping with a recession.
True enough but the other end of the same confidence interval is an employment gain of 87,900. Putting aside a community consensus that a recession is underway, there is no argument that one end of the range should be given a higher probability than the other.
The other factor supporting the quality of the data but which was given little currency was the use of overlapping samples. According to the statistician, in the labor force survey, one eighth of the dwellings sampled in the previous month are replaced in the current month by a new set of dwellings from the same geographic area. The majority of the dwellings surveyed each month (i.e. seven eighths) are the same as in the previous month. This makes for a more reliable measure of labor force change than if an entirely new sample was introduced each month.
The overlapping sample permits a calculation of employment changes within the same sample group as in the prior month. For example, within the matched sample in April, there were 116,000 persons employed who had been unemployed in March and 86,000 persons employed in March who were unemployed in April. The matched sample shows that there were more persons moving from unemployment to employment than in the other direction so that, at least among this group and contrary to widespread community perceptions, there was a real positive outcome.
And the employment numbers have not been the only positive news. Retail sales data have also thrown up a surprisingly strong outcome. Unfortunately, in the midst of a recession there is a tendency to place a pessimistic filter over any news.
In the recessions of the early 1980s and early 1990s, unemployment rates continued rising for over two years. So far, in the current cycle, the rise in unemployment has been under way for less than 12 months. It seems reasonable, from such a historical perspective, to assume a worsening rate of unemployment for another year or more and to toss out this most recent statistical aberration.
However, the April result is a timely reminder as we make our way through the cycle that momentum is a key ingredient in equity market outcomes. An equity investor should be less concerned about the level of employment or even, perhaps, the rate of change as much as the change in the rate of change.
Link that with the possibility that market adjustments could be occurring far more quickly than in past cycles. I raised this topic in the ATC e-mail on 25 November 2008 and will not repeat the arguments here other than to observe the tendency for global market reactions to become faster and more emphatic in response to changing economic conditions.
This prompts the question of how well we will eventually distinguish a real improvement from a statistical aberration and whether we will actually notice a turning point when it arrives.
We risk allowing pessimism about our current circumstances to overwhelm judgments about cyclical turning points. Concerns about short term conditions could easily lead us to discount relevant information about the timing or possibility of a turning point when that information sits uncomfortably with our preconceived views about what might be happening.
In the instance of the employment data, a view that unemployment will continue to rise for another year or more, reinforced by a growing consensus echoing such a single view, could significantly bias our investment decision making. It could prevent us recognizing signs of improvement when they happen more quickly than an existing.
This email is general in nature only and does not constitute or convey specific or professional advice. Legislation changes may occur quickly. Formal advice should be sought before acting in any of the areas discussed. Be aware that the information in these articles may become innaccurate with time. Responsibility is disclaimed for any inaccuracies, errors or omissions. Particular investments are neither invited nor recommended and hence this publication is not "financial product advice" as defined in Section 766B of the above legislation. All expressions of opinion by contributors are published on the basis that they are not to be regarded as expressing the official opinion of any other person or entity unless expressly stated. No responsibility for the accuracy of the opinions or information contained in the contributor's articles is accepted by any other person or entity. Copyright: This publication is copyright. If you wish to reproduce this article you require a license, which can be purchased here, to do so.