Adcorp’s employment and unemployment figures are not taken seriously by researchers – yet they can do much harm

Reads 96,799
Servaas van der Berg, 12 February 2013
Adcorp’s unemployment figures are derived from weak research and is repeated too often by gullible journalists. Based on a flawed methodology and dubious assumptions, the Adcorp figures imply that only about a million people are unemployed and that the total unemployment rate is 5%. At the same time, Adcorp has published an inflated figure for graduate unemployment (600 000) – a grave inconsistency. Whilst serious researchers will not touch Adcorp data, it can harm decision-making by policymakers and potential university students and their parents.


Academic researchers in the field of economics sometimes find the research done in the private sector rather superficial and therefore do not engage with it much. Adcorp’s labour market data and analysis are prime examples. Their employment and unemployment figures illustrate weak research that gets far too much media attention from gullible journalists who do not necessarily understand the implications of the statistics they report.

There are important exceptions, though. For instance, the South African Advertising Research Foundation (SAARF) undertakes excellent surveys and engages with all the important technical issues that academic researchers take very seriously. The SAARF has to provide accurate figures because the statistics affect their core business. Inaccuracies in the sampling of their All Media and Product Survey (AMPS), for instance, would soon show up in discrepancies between AMPS-reported sales and the actual sales of newspapers and magazines, and would lead to strong pressure from the media for them to correct their data – after all, the media’s advertising rates and attractiveness to advertisers are determined by their readership, as reported by the AMPS.

Other private sector researchers also do a good job. Some, like Mike Schüssler, may not always dig as deeply into labour data as academic researchers would prefer but one has to give him credit for seriously getting to grips with an important part of labour statistics and perhaps for understanding some underlying trends better than many an academic researcher.

Adcorp’s data is so clearly wrong that it does not even warrant paying their methodologies serious attention. Two detailed papers by Wittenberg and Kerr (2012a; 2012b) tried precisely that – to no avail. I will not enter that debate, as it appears futile – Adcorp simply persists with its wrong methodology.

In addition, Adcorp publishes inconsistent unemployment numbers. I want to discuss two surprising figures which go against all logic, cannot be internally consistent and which are used by Adcorp. I refer to their total unemployment figure and their graduate unemployment figure.

If one considers Adcorp’s statistics at face value, they imply that total unemployment amounts to barely one million. At the same time, they say that graduate unemployment is 600 000. Thus, more than half of the total of the unemployed are unemployed university graduates. Unemployment is mainly a graduate problem – if Adcorp is to be believed!

Both these numbers are used by Adcorp but, because they refer to them in different media releases, the obvious contradiction seems to go unnoticed by our media.

Recap: Adcorp’s flawed figures on total employment and unemployment

Adcorp argues that the informal sector in South Africa is extremely large. They apparently base this on (a crude version of) the currency demand method, which has severe limitations as a methodology, to infer the size of the informal sector, particularly in a developing economy (see the Wittenberg and Kerr articles). It is even more dubious as a way of determining monthly trends in the size of the informal sector. Furthermore, estimating the level of employment in the informal sector requires assumptions about labour intensity in the informal sector and how this evolves over time. There is no dependable source of data to determine such a trend. I appears that Adcorp is happy to guess the labour intensity of the informal sector.

This means that the method is highly flawed. Yet, from it Adcorp deduces that unemployment in South Africa is not really a major issue. Estimating total employment (informal and formal) to be approximately 19 million, and given a labour force of roughly 20 million, Adcorp’s figures imply only a 5% unemployment rate (Wittenberg and Kerr 2012a).

By contrast, StatsSA estimates narrow unemployment as a quarter of the labour force, and broad unemployment as a third. The independent National Income Dynamics Survey (NiDS) by SALDRU (at the University of Cape Town) corroborates this result, even though they find a slightly lower unemployment rate than StatsSA (about four percentage points lower) – but nowhere near 5%.

Thus Adcorp’s total employment and unemployment statistics are deeply flawed. (An unemployment rate of 5% is a very low figure and would make most OECD policy makers jealous. In the US, unemployment stands at 8.3%, in the Eurozone at 10.9%, and in the UK at 8.1%. Only Japan, at 4.5%, has a rate lower than Adcorp’s guesstimate for SA.)

Adcorp’s inconsistency on graduate unemployment

The other figure that gains Adcorp a lot of publicity relates to graduate unemployment, which they estimate to be 600 000 and rising (Adcorp 2012). It is not clear how they arrive at this alarming figure but it has not been taken from any published survey. StatsSA’s Labour Force Survey shows broad graduate unemployment to be about 55 000. Similarly, the NiDS 2010 survey puts it at 51 000.

Adcorp’s higher figure cannot be attributed to confusion about what is meant by ‘a graduate’. In its reports, Adcorp explicitly indicates that this number refers to university graduates –  therefore people with diplomas or certificates are not included in their number. (If one wants to include the holders of diplomas and certificates, the relevant measure will be broad unemployment amongst all people with tertiary education people; this has been estimated by StatsSA to be only 340 000 at the end of 2011 – still little more than half of the number of graduates that Adcorp deems unemployed.)

If the Adcorp numbers were correct, it would imply graduate unemployment of about 54%, given that there are approximately 1.1 million graduates in the labour force. These numbers would imply that unemployment for the whole labour force was only 5%, but above 50% for graduates. That is, the unemployment rate amongst graduates would be more than 10 times as much as the overall unemployment rate. (Annual graduation numbers from universities provide us with a solid basis for believing that the figure of 1.1 million graduates in the labour force is pretty accurate and that it cannot simply be regarded as a gross underestimate by StatsSA.)

Clearly Adcorp’s numbers go against all logic and cannot be internally consistent. 

Do Adcorp’s silly unemployment numbers matter?

Now, does all this matter all that much? Usually, silly figures are best ignored for bad research tends to disappear in the course of time because the underlying reality has a way of reasserting itself. However, the Adcorp figures should not go unanswered, and the reason is simple: by repeated repetition they are creating erroneous pictures of reality (and of the quality and reliability of their data). And the news media dutifully report the monthly Adcorp statistics and analysis as if they had validity.

This could have dangerous consequences.

For one, they may make policy makers believe that unemployment is an issue they can safely ignore (and many would be only too glad to do so). How strong an argument would there be for a youth wage subsidy if the unemployment rate were only 5%? Why would labour-intensive economic growth be pursued? Why would a National Development Plan spend time on unemployment?

More specifically, and dangerously for the youth, such figures could lead parents and young people to believe that university studies bring little benefit or improvement of employment prospects. (In fact, the two inconsistent sets of Adcorp figures may be understood to imply that those pursuing degree studies are far less likely to be employed afterwards, incredible as that may sound.)

Consider a potential first-generation university student who has to make the important choice of whether to enter the labour market or to study for a degree. The latter is likely to be costly and to require much sacrifice for a considerable period from the whole extended family. Many people in this position waver about the right option. Add to that the exaggerated picture about graduate unemployment that Adcorp so readily presents, and the waverer becomes even less likely to choose further studies. The country and the individual concerned are all likely to be the poorer for it.


The Adcorp employment and unemployment figures are dangerous fictions – and need to be exposed as such. It is unfortunate that the news media publish them as if they come from a credible source. That feeds at least two contradictory and mistaken beliefs: that unemployment is not a problem of great consequence, and that graduate unemployment is a problem of grandiose proportions. Neither of these two conclusions could be further from the truth. I do not know any serious researcher in this field who would agree with those conclusions. Perhaps feedback in these pages from other researchers about how they interpret these numbers would bear me out.     


Adcorp (2012): Adcorp Employment Index, December 2011, released 10 January 2012.

Wittenberg, M & Kerr, A. 2012a: Criticisms of the Adcorp Employment Index, DataFirst, University of Cape Town, February.

Wittenberg, M & Kerr, A. 2012b: Science and nonsense: Further criticisms of Adcorp, DataFirst, University of Cape Town, March.

Both of these papers can be downloaded from:

Download article

PDF icon Vd Berg Feb 2013 Adcorp data dangers FINAL.pdf
Post a commentary

This comment facility is intended for considered commentaries to stimulate substantive debate. Comments may be screened by an editor before they appear online. To comment one must be registered and logged in.

This comment facility is intended for considered commentaries to stimulate substantive debate. Comments may be screened by an editor before they appear online. Please view "Submitting a commentary" for more information.