Apartheid was declared a crime against humanity by the United Nations (UN) in 1973, for the systematic oppression and domination of Black South Africans by the apartheid government. The effects of those policies are still very much with us in contemporary South Africa, as evidenced by the racial disparities in any number of a multitude of social welfare measures. To date, however, we do not have any credible measures of the economic harm done to Black individuals, nor economic benefits that accrued to White individuals, who lived under the apartheid regime.
Economists conventionally define labour market discrimination as the differential treatment in the labour market experienced by individuals of otherwise equal productivity, due to their group membership.
Thus, observed differences in labour market outcomes that are due to underlying differences in productivity would not conventionally be recognized as labour market discrimination.
In the South African case, however, this definition is inadequate. In addition to labour market discrimination as conventionally defined, the apartheid policies were explicitly designed to induce productivity advantages and disadvantages for White people and Black people respectively.
The number of ways and ruthlessness with which this was achieved is quite overwhelming. Direct labour market discrimination included racial occupational restrictions, and wage differentials for people of different races who were doing the same job. This was complemented by restrictions on property ownership, the right to start businesses, geographic mobility, access to healthcare, access to public transport, residential segregation, and access to educational opportunities.
In a new SALDRU working paper, Prof. Vimal Ranchhod and Dr Miquel Pellicer introduce a new concept; the cumulative effect of discrimination.
The key idea here is that the apartheid system had many discriminatory layers to it, and that the effects of this multi-dimensional discrimination would accumulate and interact over time.
For example, the Bantu Education Act of 1953 led to differences in school funding levels and differences in curricula. These would impact on educational attainment and subsequent labour market outcomes, thus reinforcing the effects of any direct discrimination in the labour market.
A valid question to ask, if one wants to measure the relative advantage of being White under apartheid, is why not simply compute the average difference in economic outcomes across racial groups using data from that period? After all, measures of outcomes such as employment, wages and poverty levels could be relatively easily produced.
While these are all valid measures of welfare differences, they would probably not accurately measure the causal effect of being White under apartheid.
Why not? Imagine a world where, in 1948, South Africa had somehow transitioned to a full democracy with no forms of discrimination whatsoever. At that moment in time, the people who were subsequently classified as ‘White’ would already have had substantial advantages in terms of economic opportunities. They were much more likely to have been based in urban centres, had substantially greater land holdings and financial capital, and would have had greater levels of formal education. These advantages would be transmitted to their children, such that some component of these differences would have persisted over time, as is the case in present day South Africa. While all of these advantages are relevant for understanding differences in outcomes across racial groups, they are not particular to either South Africa or the apartheid era. Thus, a simple comparison of mean outcomes across racial groups would probably over-estimate the magnitude of the special privileges that apartheid conferred on Whites.
How do the authors try to resolve this empirical challenge? They make use of a change in how people were racially classified in order to estimate the changes in outcomes that can be attributed purely to an individual’s racial classification.
Legally, racial classification was determined according to the Population Registration Act of 1950. The original act required that every individual be classified into one of three mutually exclusive and exhaustive racial groups: White, Native and Coloured.*
The implementation of this policy was undertaken by Race Classification Boards, who used some combination of three different criteria to determine a person’s race: appearance, social acceptability, and ancestry. However, when the classifications were first introduced, there was no systemic way to verify a person’s ancestry, so the people born before the 1951 Census were generally classified based only on their appearance and social acceptance, as reported in the 1951 Census. As time progressed, more and more people obtained an official racial classification. Thus, for children born after the 1951 Census, there was an increased ability to use the classification of the parents to determine the classification of the child. Thus, two officially White parents would have children who would be classified as White, two officially Native parents would have children who would be classified as Native, and all other children would be classified as Coloured; generally without regard to the appearance of the child.
To illustrate how this change in the classification process can help to identify the effects of racial classification, consider the following hypothetical example. Imagine a couple, both of whom are racially ambiguous. They have two very fair-skinned daughters, one born in 1950 and one in 1952. The child born in 1950 would be classified based on her appearance, and would thus be White. Now imagine that, based on the 1951 Census, at least one parent was classified as Coloured before the younger sibling was born. The younger daughter would then be classified based on ancestry, not on appearance, and would be classified as Coloured. Conceptually, if we could follow a number of such sibling-pairs through the course of their lives, we could measure the average effect of classification on several different outcomes.
The example above develops an intuition for how the change in implementing the policy can be used for identification, but no such dataset exists for that specific analysis. However, a similar approach can be used at the population level. The authors use Census data from the 1980, 1991 and 1996 Censuses, and employ a statistical method known as the instrumental variables approach. They estimate the effects of being classified as White, as compared to being Coloured, on educational attainment, employment and income.
The main findings are that one’s racial classification had a very large impact on one’s educational attainment, and on income for men.
For the subset of people who were affected by the change in the classification process, the authors estimate that being classified as White led to an increase of 3.7 and 3.9 years of schooling, on average, for men and women respectively. The most striking finding, however, is that for a male to be classified as White would have more than tripled his income.
This is an astoundingly large number, when we consider the difference over the course of a lifetime. For example, assume that a hypothetical Coloured man earned R100,000 per annum, and that if the same person were classified as White then he would have earned R300,000 per annum. If we then assume a working lifespan of forty years, this would result in a lifetime earnings of R4,000,000 if he were classified as Coloured, or R12,000,000 if the very same person were classified as White!
Differences of these magnitudes make it easy to understand why racial disparities remain so prevalent in contemporary South Africa.
As with all research, there are limitations to the findings. The major weakness in this work is that the African race group was excluded from the analysis. This was unavoidable because apartheid era Censuses excluded the former Homeland areas. Africans make up the substantial majority of the population, and were undoubtedly the most oppressed and excluded by the apartheid policies. One might conjecture that the effects of reclassification for Africans, if that could be estimated, would be considerably larger than those estimated for Coloureds.
* The Indian/Asian group was only created in 1959 through a different piece of legislature.