Adcorp’s employment index has reached the office of the Presidency. Speaking at the Wits Business School, Deputy President Kgalema Motlanthe reportedly claimed that labour productivity in South Africa had “declined by 41.2 percent since 1993” (polity.org.za, May 14). What a coup for the merchants of such nonsense. Take a bow, Adcorp.
Leading experts in econometrics and statistics, such as Andrew Kerr and Martin Wittenberg of UCT and Servaas van der Berg of Stellenbosch University, have exposed Adcorp’s methods. As academics, they usually refrain from emotive language, yet Kerr and Wittenberg were so outraged that they headlined one of their articles “Science and Nonsense”.
Why is labour broker Adcorp taken seriously? Is it difficult to see that something is wrong with its figures? The latest Adcorp employment index (May 13) reports that 262 000 people are employed in mining. Statistics SA and the Department of Mineral Resources report double that number: 519 000 for December 2012.
The Adcorp employment index of October 2012 stated on page 1 that “employment in the mining sector has declined to 523 000” (obviously using Stats SA data), yet three pages later in the same press statement 291 000 are allegedly employed in the mining industry. This is 90 000 less than in the Quarterly Labour Force Survey household surveys from Stats SA, which grossly underreport mining employment because of its geographical concentration, as Stats SA has admitted.
Calculating productivity development by dividing gross domestic product (GDP) by the number of workers employed, or by the number of hours worked (L), does not suit Adcorp and its lead economist, Loane Sharp. This method, universally applied by economists, is too friendly to labour in their view. Every year output per worker increases – but might it be that workers are just using more advanced machinery when instead they should be working harder and more efficiently? Are we not cheating about our “true” labour productivity when we use tools?
Machines, equipment and infrastructure are invented, constructed, built, used, maintained and repaired by employees. There is no clear separation of “labour” from “capital” in the real world of production. Capital is dead without labour. Labour power in pure isolation has a limit. Better organisation, management and work morale certainly matter when rowing a boat, but strip the team naked, take away paddles and boat, and we are left with a crowd of exhausted swimmers. “Capital”, in its fundamental sense, is not stockpiles of money owned by a few.
Undeterred by this, Sharp believes he can torture out a confession of “labour unique productivity” from the statistics. In August 2011, he made a first effort in a booklet published by the Centre for Development and Enterprise. He divided the established productivity measure (GDP/L) by the real value of the capital stock (C) for each year, using his novelty formula (GDP/L)/C.
The result was an alarming diagram, misleadingly referenced “SARB”. Interestingly, if we switch the places of L and C in his new formula, keeping their values, we get the same numerical result and exactly the same diagram, arriving at an unintended unity of capital and labour, not their separation. Indeed, the capitalists are also cheating – by using workers to handle “their” machines. And now we have a catastrophic fall in “capital productivity” too (if using Sharp’s data, which is another story).
From November 2011, Adcorp’s alarming reports are instead based on a statistical method called “regression”. Since then, labour productivity in South Africa has turned “negative”. Basically, this would mean that more employment causes a fall in production. Tell that to the National Planning Commission.
The November 2011 employment index contained a bizarre warning that GDP was growing faster than employment (which for the rest of us means more output per worker and good news). But for the core alarm message, Adcorp had used a model of this type: GDP = aL + bC + c.
If the model was of any use, “a” would tell by what percentage GDP would change on average given a 1 percent change in the number of employees (L), assuming that the capital stock is fixed. Similarly, “b” would tell us the percentage change in GDP given a 1 percent change in the capital stock (C), given that the number of employees stays exactly the same. “c”, finally, is a term known as total factor productivity. In this formula, it is “everything” that affects GDP beside changes in the number of employees and value of the capital stock.
For the 2000s, Adcorp reported a positive value for the “b”, reported no value for “c” and a minus value for “a”: -0.08. This suggests that if 137 000 people are added to the 13.7 million employees in today’s economy it would lower total output by 0.08 percent and South Africa would be a little worse off.
To be useful, an econometric model must have a minimal resemblance to actual reality. “What is causing what” – this has to be thought through. If, for example, a historical drop in South Africa’s GDP occurs in 2008/09 because of a world crisis and 900 000 workers are rapidly retrenched, but slower than investors are halting investments, this doesn’t mean that South African workers are becoming less productive.
Secondly, to get accurate predictions the variables in this type of model must change as independently as possible from each other also in real life. But employment does not change independently of changes in investment! The lack of private investment is at the centre of the unemployment debate.
On the other hand, private investments can lead to retrenchments and increase unemployment (if the government or the trade unions do not intervene). The model is so riven with basic errors, these and more, that it should lead us to question its honesty. It is no surprise, therefore, that “a” was found “statistically insignificant”, something mentioned in a footnote by Adcorp. In plain language this means No Clear Result. That message alone from the computer should have stopped the report.
But these points and more are superfluous. Sharp’s model contains a fundamental conceptual error. Positive value or not, “a” shows how GDP would change if we add workers without giving them more tools. It says nothing about work efficiency and so on. If there is such a thing as “labour unique productivity”, it is a part of “c”, total factor productivity: a part of the complicated interaction between machines, technology and employees and “everything” intangible that promotes economic growth; “c” quantifies the unquantifiable.
And it is because of that difficulty, that the measure, output per worker, has gained prominence as the main productivity measure in economic policy debates; and because of employees being breadwinners: it is the growth in GDP per person in the country that is our real interest.
Increases in output per worker makes us all better off, if the increased production per worker and per person in the economy is the least equally distributed.
A fall in output per employee occurred last time during the turbulent years before 1994. Today, deputy president, output per employee is close to 70 percent higher than in 1994 (diagram). In another South Africa, this had been the basis for eradicating poverty.
The last two years, labour productivity has grown at a slower pace than before, by about 2 percent per year. This is related to the revolt on the labour market. Still, output per employee continues to grow. People always continue to invent and take on board new tools to learn and to improve the use of them. To secure that this continues to happen: pay living wages to all employees. Let the mass of workers and their families share in the ever growing new wealth per person produced in the country.
Dick Forslund is an economist and researcher at the Alternative Information Development Centre in Cape Town.