Inequality: Why Isaacs is half right and half wrong

Gilad Isaacs, drawing on his work with Andrew Bowman, has shed important light on the economics of the platinum strike. (Here, here and here.)

During the platinum boom of 2000 to 2008, executives and shareholders benefitted massively but semi-skilled underground workers did not share in the bonanza.

This year, despite tight post-boom economic constraints in the industry, the underground rock-drillers demanded very large wage increases. A long and bitter strike culminated in a compromise in which the mines paid more than they offered originally but the mineworkers got much less than they demanded.

Isaacs combines this analysis with a simple narrative, in which workers are exploited by excessively-paid executives and rapacious shareholders. In his account, the conflict in the platinum industry is an expression of conditions across South Africa.

South Africa is deeply unequal. When workers like the platinum mineworkers win real wage increases, according to Isaacs, inequality is reduced.

In response to Isaacs, business economist Mike Schussler makes two main arguments. First, he suggests, South Africa is not as unequal as Isaacs claims. Secondly, according to Schussler, it is unemployment rather than the ‘wage gap’ (between well- and low-paid workers) that is the major cause of inequality.

Isaacs and others rightly challenge the first of Schussler’s arguments. They point out that Schussler misuses and misinterprets his ‘evidence’. His comparison between the distribution of ‘income’ inclusive of the social wage in South Africa and the distribution of income exclusive of the social wage in other countries is ‘absolute nonsense’, as Morne Oosthuizen pointed out in a comment on GroundUp (and Isaacs discussed further in his rejoinder to Schussler).

Schussler’s ‘evidence’ for his second argument – concerning the relationship between wages and inequality – also involves the misuse of data (as Isaacs and others show), but the underlying concerns cannot be so quickly dismissed.

Isaacs seems to think that higher wages for platinum miners and similar workers will reduce overall inequality. Schussler, in contrast, concludes that higher wages for workers such as platinum miners will not reduce overall inequality in South Africa.

In his rejoinder to Schussler, Isaacs rebuts Schussler’s crude and flawed version of this argument.

But his rejoinder is itself incomplete and flawed, and does not address the arguments that Nicoli Nattrass and I have made for more than 20 years, including in our 2005 book on Class, Race and Inequality in South Africa.

Wage increases have both direct and indirect effects on overall inequality. Both direct and indirect effects depend on whose wages are increased. The direct effects of wage increases on inequality are much larger if the wages of poor people rise than if the wages of richer people rise. The indirect or dynamic effects depend, especially, on how changing relative labour costs affect the demand for different kinds of labour in affected sectors. Let us consider each of these in turn.

The direct effects on overall inequality of changes in people’s wages – or any other income – depend on whose incomes are changing. More precisely, the effects of increased earnings on inequality (measured, for example, using a Gini coefficient) depend on where the beneficiaries are in the overall distribution of income.

If executive pay is increased – as happened in the 2000s – then both the wage gap and overall inequality worsen. If the earnings of very poor people rise, then inequality declines dramatically.

Prior to this year’s strike, the typical underground mineworker formally employed in the platinum industry was very poorly paid relative to either mine executives or the dangers of underground work, but was nonetheless close to the mean household income, and well above the median, in contemporary South Africa.

When discussing income distribution, we often refer to ‘deciles’, that is, tenths of the population. The typical platinum rock-driller was probably in the seventh or eighth income decile in terms of household incomes, that is, better off than the poorest six deciles (or 60 percent of the population) while worse off than the richest two deciles (or 20 percent of the population).

The redistribution of any amount from the rich (such as the top income decile) to the poor (such as the bottom five income deciles, or bottom half of the population) reduces inequality by much more than the redistribution of the same amount to households in the seventh and eighth income deciles.

If you want to reduce inequality, redistribute income to the poor, not to mineworkers.

This can be demonstrated with reference to the Gini coefficient, which is a standard measure of inequality. If the distribution of income is totally equal, such that everyone has the same income, then the Gini coefficient will be zero. If the distribution is totally unequal, such that one person earns everything, then the Gini coefficient will be one. A high Gini coefficient indicates high inequality. The Gini coefficient for income in South Africa is about 0.67, which is very, very high.

If we took 5 percent of national income away from the rich and paid it to households in the seventh and eighth deciles, then the Gini coefficient would drop to 0.65. If, however, we took the same 5 percent of national income away from the rich and paid it to the poorest half of the population (that is, people in the poorest five income deciles or below the median), then the Gini coefficient would drop to less than 0.62.

If you really want to reduce inequality, redistribute to the poor through either job creation or pensions and grants that are targeted on the poor. Don’t take money from the rich and pay it to people who are already in formal employment.

The real world is dynamic, not static, and changes in wages have real dynamic effects on employment across the economy. Crucially, changes in relative wages and hence labour costs will affect the demand for different kinds of labour.

Over the past 40 years, rising wages for less skilled workers in formal employment have pushed employers to switch to more capital- and skill-intensive production and even, in some sectors where South African producers face global competition, to close down entirely (consider the clothing industry).

There are now very few unskilled workers in formal employment in South Africa because the labour costs of unskilled workers in formal employment have increased. The reduced demand for less skilled labour is the major cause of South Africa’s unemployment crisis.

Isaacs writes that employment depends solely on levels of investment, and ignores entirely employers’ choices between capital/skill-intensive and labour-intensive technologies. High investment in capital/skill-intensive production results in rising demand for highly-skilled workers, rising labour productivity, and reduced demand for less skilled workers - ‘jobless growth’.

Conversely, investment in labour-intensive production results in a raised demand for less-skilled workers, a stable or reduced wage gap, and reduced unemployment.

In a dynamic economy with full or nearly full employment, the priority should be increased productivity, replacing low-skilled and low-paid jobs with more skilled and better-paid ones. The outcome would be reduced inequality, as wages at the bottom catch up wages at the top.

But in an economy with very high employment among unskilled workers, a capital- and skill-intensive economic growth path results in higher wage gaps, higher unemployment and higher inequality.

In sector after sector of the South African economy, wage increases for less skilled workers have contributed to employers choosing to invest in more capital- and skill-intensive production, resulting in reduced employment, rising productivity and a rising profit share. If increased wages in the platinum industry have similar effects on the technological choices of platinum mines, then the overall effects of wage increases on inequality become much more complex than Isaacs recognises.

Isaacs is dismissive of concerns that wage increases contribute to unemployment and thus worsen poverty and inequality. His reasons for doing so are based on misunderstanding of the economic evidence.

His first argument is based on studies that try to separate out the contributions to overall inequality of (1) unemployment and (2) the wage gap(s) between high and low-paid workers.

Isaacs cites a 2012 study by Leibbrandt, Finn and Woolard that supposedly found that ‘62% of inequality is accounted for by differences in wages, 38% by unemployment’. In fact, this is not precisely what the study found. The study found that 62% of inequality was accounted for by the differences between households where someone worked, and 38% by the differences between these and households where no one worked.

Isaacs misreports this finding, because households where someone worked included many unemployed people. To estimate the effect of unemployment on inequality, one would need to take into account also the effects of unemployment on income differences between households where someone works.

In short, the fraction of inequality accounted for by unemployment is much larger than 38%.

The evidence he cites does not support his conclusion that ‘unemployment plays an important role in inequality but not the leading one.’

This is not the main flaw in Isaacs’ account, however. Even if it were the case that the statistical decomposition of inequality revealed that unemployment accounted for a smaller share of total inequality than the wage gap between low and high-earners, it does not follow that wage increases are more important than job creation for the reduction of inequality, as Isaacs implies.

As we have shown above, overall inequality (as measured by the Gini coefficient) is reduced much more if poor households earn an additional R1,000 per month (through low-wage job creation or social grants) than it is by wage increases of R1,000 for the same number of people who already have jobs.

Put simply, total inequality is reduced more by creating a low-wage job for someone from a poor household, even if that job pays only R1,000 per month, than it is by paying platinum mineworkers an additional R1,000 per month. The ‘wage gap’ might even increase at the same time as total inequality declines. This is a simple arithmetic truth.

Statistical decompositions of inequality tell us nothing about the relative effects of wage increases (for people in the top half of the income distribution) and job creation (for people in the bottom half of the income distribution) on overall inequality. Isaacs is simply wrong to imply that wage increases reduce inequality more than job creation.

Insofar as wage increases result in job destruction, they can worsen inequality just as surely as rising executive pay.

‘Are greedy unions that make ever increasing wage demands retarding employment?’, asks Isaacs. His answer is that ‘the evidence suggests otherwise’.

Isaacs is entirely wrong here. Every study in South Africa finds that real wage increases reduce the demand for labour (see studies by Conradie, Murray and van Walbeek, Bhorat, Kanbur and Mayet, Bhorat, Kanbur and Stanwix, Pauw and Leibbrandt, as well as our 2005 book on Class, Race and Inequality in South Africa and our more recent work on the clothing industry).

The question is not whether wage increases destroy jobs, but how many jobs are destroyed, and how to weigh up the costs of job destruction with the benefits of higher wages.

Isaacs’ second reason for dismissing anxieties about job destruction is his claim that, since the end of apartheid, real wages have declined for everyone except the top two income deciles. This finding he attributes to a 2010 paper by Leibbrandt, Woolard, McEwen and Koep. We should bear in mind that a majority of trade union members in South Africa are in the top two income deciles, and the platinum mineworkers are not far below this.

More importantly, the analysis by Leibbrandt and the others is flawed because it is based on the comparison of data that were collected using very different methodologies and samples. Recent work by Wittenberg, using the Post-Apartheid Labour Market Series (PALMS), demonstrates convincingly that the kind of claim made by Leibbrandt is due to methodological changes between datasets, not real trends.

Wittenberg’s work confirms that the rich have indeed benefitted disproportionately since the end of apartheid. But, when methodological changes are taken into account, everyone with earnings above the median has benefitted to some extent.

Secondly, it is likely that recent survey datasets seriously and increasingly underestimate the earnings of most people with earnings above the median (probably because surveys have not kept up with rising deductions from gross earnings, by either employers or banks). It is likely that that above-median earners have benefitted even more than the PALMS data suggests.

What this means, in plain English, is that real earnings have risen since 1994 for almost all formal employment.

Isaacs makes many valid criticisms of Schussler. But he makes a number of mistakes of his own.

Contrary to his assertion, most formally-employed workers have benefitted since the end of apartheid. Contrary to his assertion, rising wage increases for most formally-employed workers make a minor difference to overall inequality because these workers are not in the poorest half or more of the population. Insofar as rising wages reduce the demand for labour, especially less skilled labour in tradable sectors, they are likely to worsen inequality substantially at the same time as improving it marginally.

Reducing inequality requires raising the incomes of the poor, through either redistribution or job creation.

Seekings is the Director of the Centre for Social Science Research at the University of Cape Town.