Fix the numbers – Loane Sharp

2013-02-17 10:00

Unemployment stats are a point of contention. Loane Sharp and Pali Lehohla debate the issue.

Last week, Stats SA indicated that in the fourth quarter of last year, the unemployment rate dropped from 25.5% to 24.9%.

Analysts and the general public were amazed by the announcement, since in the final quarter the economy experienced the greatest level of labour disruption in the post-apartheid era, occurring most notably in mining and agriculture.

Two facts stand out about the announcement.

First, Stats SA continues to emphasise the narrow unemployment rate of 24.9%, rather than the broad unemployment rate of 37.4%.

The difference between the two is that discouraged work seekers – those who have stopped looking for work – are not counted by the statisticians as unemployed.

The most important table in its quarterly labour force survey and key labour market indicators does ­­­ not even report the broad unemployment rate.

This is easy to fix: Stats SA should publish both the narrow and broad rates and allow commentators to choose which measure to cite.

For example, the broad rate increased from 37.3% to 37.4% in the fourth quarter, which no doubt better matches what people understand when they use the word ‘unemployment’.

Secondly, when Stats SA releases its labour force estimates, it should emphasise that these are subject to a margin of error.

In its announcement last week, it indicated that employment dropped by 68?000.

This is an estimate and, based on the calculated margin of error, employment may have fallen by 222?000 or increased by 86?000 during the quarter, or really any figure in between.

In some cases – such as agriculture, the informal sector and private households, which account for nearly one-third of total employment – the margin of error is very large. The statistical “confidence” of estimates for these sectors (a number from 0% to 100%, with higher values being better) is as low as 13%.

Thus, the reported unemployment rate of 24.9% could vary anywhere between 24.4% and 26.1%.

If people knew the margin of error, they would know how much weight to give to Stats SA’s estimates.

To some extent, the margin of error is unavoidable and difficult to remedy.

It is in the nature of a sample, being a subset of the total population, that estimates derived from the sample cannot be certain.

For example, Stats SA surveys a handful of dwellings each quarter and makes calculated guesses from these about the entire labour force.

Extrapolating the results from 30?000 survey respondents to 18.1?million people in the labour force is subject to error.

But there are reasons to believe Stats SA’s margin of error is, to a great extent, avoidable.

For example, it publishes quarterly estimates of a host of figures, including the number of taxpayers, medical aid principal members, business owners, bargaining council members, unemployment insurance contributors and so on.

What is unique about these figures is that we can attempt to verify them against the actual data provided by other data sources.

For example, the SA Revenue Service (Sars) publishes the actual, known and verified number of income taxpayers, which is currently 8?million.

By contrast, Stats SA estimates the figure to be 7.2?million – an undercount of about 10%.

Or, to give another example, the Unemployment Insurance Fund reports there are 7.9?million contributors.

By contrast, Stats SA estimates the figure to be 6.6?million – an undercount of 17%.

To give yet another example, the National Association of Bargaining Councils reports that 1.8?million workers are covered by bargaining council agreements, whereas Stats SA estimates that only one?million are covered – an undercount of 41%.

In every case where we can cross-verify Stats SA data against so-called administrative data sources, the former undercounts the true figure.

The most likely reason it is undercounting employment is that survey respondents are not reporting all their economic and labour market activities to the authorities.

In this month’s Adcorp Employment Index, it clearly shows that households hide up to 30% of their income from Sars and Stats SA. We need to get to grips with these discrepancies.

»?Sharp is a chief economist at Adcorp

Join the conversation! encourages commentary submitted via MyNews24. Contributions of 200 words or more will be considered for publication.

We reserve editorial discretion to decide what will be published.
Read our comments policy for guidelines on contributions. publishes all comments posted on articles provided that they adhere to our Comments Policy. Should you wish to report a comment for editorial review, please do so by clicking the 'Report Comment' button to the right of each comment.

Comment on this story
Comments have been closed for this article.

Inside News24

Traffic Alerts
There are new stories on the homepage. Click here to see them.


Create Profile

Creating your profile will enable you to submit photos and stories to get published on News24.

Please provide a username for your profile page:

This username must be unique, cannot be edited and will be used in the URL to your profile page across the entire network.


Location Settings

News24 allows you to edit the display of certain components based on a location. If you wish to personalise the page based on your preferences, please select a location for each component and click "Submit" in order for the changes to take affect.

Facebook Sign-In

Hi News addict,

Join the News24 Community to be involved in breaking the news.

Log in with Facebook to comment and personalise news, weather and listings.