The median versus the mean

Cosatu’s Worker’s Day event celebrated labourers’ rights and progress, yet 75% of them earn less than R6?500 a month. (Delwyn Verasamy)

Cosatu’s Worker’s Day event celebrated labourers’ rights and progress, yet 75% of them earn less than R6?500 a month. (Delwyn Verasamy)

In “Wages up as unemployed queues grow” (Mail & Guardian, June 22 to 28), Teigue Payne writes that “both [Mike] Schussler’s and his critics’ analyses suffer from the problem of averages, although they fight about ‘median’ (or typical) wages”. In a section titled “What ‘workers’ are paid”, Payne states that “Schussler’s critics are deliberately falsifying his position to discredit his case”, apparently because “the R13200 figure is taken directly from the ‘Quarterly Employment Statistics’ and is not a derivative figure”. He then attempts some sector-level analysis that, frankly, does not contribute much to the debate.

First let me clarify what my view is in relation to the debate.
The main point I raised with Schussler was that he used an inappropriate statistic, the average, when he should have used the median. Payne knows this but decided to ignore it, yet this is a major pillar of my argument. Payne does not venture to report the median earnings for his sectors. In fact, he does not report any new figures. The only figures he mentions are minimum wages, whose enforceability is known to be weak, and the average wage, which is an inappropriate statistic.

Payne confuses the debate by raising irrelevant issues. For example, he said “much depends on what you mean by workers”. This has never been an issue in this debate. What is at issue is: Does the average reflect the reality of the South African economy? Is it valid to say, on the basis of an average of R13200 a month, that “the working class can now be called a comfortable class” in South Africa? My argument said no. Payne seeks to salvage whatever remains of the working class as a comfortable class view by “canvassing employers in four major South African industries”. What is the result of this canvassing?  

It came to nothing because he does not give us the precise earnings figures from his research. For example, in the sectors he canvassed, Payne does not provide median earnings. In mining he said the entry-level wage for a sweeper was R4400 “but the employee can more than double that amount with merit and performance bonuses”. This does not solve Payne’s problem, because it does not tell us what the median earnings are in mining. The same applies to the clothing, motoring and engineering industries. Without providing median earnings, we are bound to get a distorted picture of South African workplaces.

General conclusion

Surprisingly, Payne ignores the “Monthly Earnings of South Africans (2010)” report, which details the earnings by occupation, education, age, race and gender. I am sure Statistics South Africa could qualify these earnings by sector on request. But what emerges from this report is that median earnings in South Africa in 2010 were R2800. Inflating this by 10% over one year, we can approximate the median earnings to be R3080. This does not come close to the R13200 from which a general conclusion can be drawn that “the working class can be called the comfortable class”. As Payne himself admits, “the truth, using other statistics, is probably that [workers] earn less than half that amount”. In fact, 75% of employees in South Africa earn less than R6500 and the top 10% earners earn above R12000 a month. These figures are based on a more scientific survey than Payne’s “canvassing”.

Payne tries hard to link employment performance between 2008 and last year to the growth of wages, but he is also unscientific in this. He presents a table drawn from the “Quarterly Employment Statistics”, which reports that wages (I think he forgot to write three more zeroes for gross earnings). He writes that we now have 36000 fewer employees whereas wages have increased by 55%. Let us suppose that between 2008 and 2011 union-employer wage settlements were 10% a year on average. We arrive at 33% as the average wage growth, not 55%. It means that, among employees, there are some who got more than a 20% earnings increase. Who are these people? Payne should also have used the increase in median earnings to control for inequality even in the growth of earnings.

The more problematic issue is that Payne seeks to draw a link between earnings growth and employment contraction. This is misleading because employment is not determined solely by earnings. Payne has not provided controls for the impact of the global crisis on job losses. He might be outraged to learn about a view that says in a recession, when earnings drop, demand drops and this leads to more job losses. That is to say, over certain phases of the business cycle the increase in workers’ earnings boosts employment and a cut in earnings destroys jobs.

The last point I want to raise is the claim that “Schussler’s critics are deliberately falsifying his position to discredit his case”. As far as I am aware there is no critic of Schussler who questioned the R13200 figure. What was being questioned is the use of an average in the context of earnings analysis. Why would Payne want to insist on the R13200 average when it is known that there is massive inequality in income distribution? This is not done in the scientific community, because it is accepted that earnings distribution is skewed. Even the “Annual Industrial Action Report (2010)” reports median wage settlements, because its authors are fully aware that wage increases are unequal.

Christopher Malikane is an associate professor of economics at the University of the Witwatersrand and an economic adviser to union federation Cosatu

Client Media Releases

Fedgroup drives industry reform in unclaimed benefits sector
Hardworking students win big at architecture awards
VUT presents 2019 registration introduction
Vocational training: good start to great career