 Good morning ladies and gentlemen. Thank you for attending this session. I'll be talking about affirmative action in South Africa particularly. It's an interesting case study particularly because South Africa has a long history of racial segregation but also racial segregation particularly by a minority group towards a majority group. And this historical legacy as I'll show to you hasn't necessarily been reduced despite an aggressive affirmative action policy which hasn't seemed to narrow the gaps and the question is is why and why not. So what this paper will do is is we'll really monitor the effects of the the racial wage gap over time and it'll monitor particular points and junctures where we find that laws have been updated to try and achieve racial equality. So we know that discrimination hurts incomes of various groups. If the majority discriminates against the minority such as in the USA then it's typically the minority whose incomes have get affected. But as in the case of South Africa where you typically have a minority discriminator particularly the white population that discriminated against the black African population then all incomes are affected. It's not really good for the situation in the long run and Gary Becker in his Nobel Prize lecture actually suggested that this is why apartheid policy faced its demise. By conjecture what should have happened is is that once apartheid was formally dismantled where job reservation was removed by conjecture all of this should have all the differences should have imploded and we should have seen a new trajectory of equality if that is true. But this is really not the case that we've seen. We haven't seen racial wage gaps disappear and despite these targeted policies which as I'm going to outline for you in just a second have been very very targeted but not necessarily had the effect. So we'll look at some empirical estimates from server data to try and show these changes. Just to give you a background of institutionalised ways discrimination in South Africa most of you would be aware of the apartheid, grand apartheid which basically instituted job reservation particularly favouring white individuals. However what many people don't know is that this predated apartheid and job reservation actually started much earlier than 1948 when the apartheid laws were instituted. The other important aspect here is that also apartheid laws banned African unions which really removed bargaining power of Africans vis-a-vis their white counterparts. So the early evidence suggests that by the 1970s when there was a significant history of racial segregation white men earned about five and a half times more than black men particularly due to job reservation. Some skilled jobs were reserved for white people and unskilled jobs were reserved for black people. In the 1980s there was a narrowing of the wage gap this was partially due to increases in attributes but also due to the unbanning of black unions which led to a convergence of wages across the groups but it didn't necessarily remove the discriminatory component within the labour market. Unexpectedly in 1994 when democracy appeared and when a lot of these laws were dismantled we actually witnessed a widening wage gap and especially also the discrimination component of it. So this wasn't due to necessarily greater gaps in productive characteristics such as education in actual fact discrimination worsened. One of the reasons for this might be is that there were actually no formal laws yet in place to counter discrimination effectively this only came in after 1998 as I'll discuss in just a second. So the the laws are really important in this section. From 2000 onwards there was an active affirmative action policy but the evidence that is available at this stage suggests that there was a negligible effect at the mean incomes. There was however a positive effect at the top of the income distribution where wage gaps narrowed at the top of the income distribution. This suggests that affirmative action has had the effect of removing discrimination within the elite but not necessarily for the average worker in the labour market. This led to actually the term which we call broad-based black economic empowerment which means that it was supposed to reach the entire income distribution not just the top of the income distribution. So what I'm going to do in this paper is I'm going to show you that we're going to look at monitor the effects of these targeted affirmative action policies and to see whether this changed a lot of the trajectories. We because the evidence has been mixed what we're also going to look as we're going to look at unobserved generational heterogeneity. It's not something which we can look at the individual level because in typically in these studies there are a lot of unobservables which we really do worry about. So in 1993 the interim constitution laid the foundation for affirmative action but it didn't necessarily start implementing it. This only started happening in 1998 with the Employment Equity Act which as it suggests it eliminated any unfair discrimination on paper. There were positive development objectives towards black individuals colored which in South Africa means mixed race it's the official terminology Indian but also towards women and disabled. The interesting thing here is that this also included white women because they were classified along the gender dimension. The Employment Equity particularly targeted firms above the age above the size of 50 in other words firms that were large and this is particularly with reference to the fact that South Africa has a very small small firm sector a large doth of small firms and that industry is quite concentrated. So the exception is in place to allow small firms to not be burdened by over regulation. But in 2003 this policy was stepped up a bit as I've suggested the BBEE and as you'll see by the title of this act it particularly focuses on race and removes this to some degree the gender dimension because it focuses particularly on black individuals okay and black in this sense is broadly defined not just black Africans for instance Asian individuals are also counted as black under this law. So it went beyond workplace regulation it went beyond Employment Equity but towards positive opportunity in all facets of life. So it really looked at the empowerment of black people including women workers youth people with disabilities and people living in rural areas and the important part here is this at the bottom here is this balance scorecard which firms have to require. They basically they develop a scorecard which each firm gets a point which works as some sort of accreditation and this accreditation allows firms to find government contracts it also creates the opportunity for firms to interact with each other if they want to work with credible firms okay. It also helps in public-private partnership and also when licenses are applied for. You'll see that a lot of firms would advertise and they'd tell you your BBEE certification okay this is an important component and as you'll see here this is how the BEE scorecard worked it looked at direct empowerment through also not just employment as the 1998 Employment Equity Act did it also looked at the demographic constituents of ownership it also looked at the management of the firms who's in charge of directing the firms but also what was previously covered by the Employment Equity Act only counts 15% of this BEE score. There's also a skills development component to actively promote the workforce but also indirect empowerment and it's this preferential procurement and socio-economic development. This part here is very key to understanding how the policy worked as is this part because it doesn't just focus on what happens within the labour market it focuses on what happens outside the labour market and also firms interactions with each other. In South Africa we there's this term which we call tender preneurs it's basically entrepreneurs who have been able to leverage their positions their BEE positions to be able to get preferential procurement which has led to a lot of corruption and to the creation of an elite and this is an important thing. So typically when people look at server data they turn to the classic Wachika Blinder decomposition and they decompose gaps wage gaps into observable components and unobservable components and the unobservable component is then just seen as discrimination but the important part here which we often don't think about is the fact that unobservables are not randomly distributed across racial groups and in South Africa this is particularly true because some of those unobservables are importantly for instance the quality of schooling which we know is very highly unequal and very polarized the distribution where the top 20 percent of schools is performing very well but the bottom 20 40 80 percent of schools are dysfunctional and this is still distributed across racial groups and this is an important unobservable which is a pre-labor market discrimination which we often cannot account for in just simple decompositions like this so what we really want to address here is this unobservable I'll show you we won't perfectly address it but we start to arrive at a solution. Typically these unobservables produce upward estimates of this unobservable discrimination component and as I've suggested quality of schooling is the main one there is a study that gives an indicator of evidence that school quality presents about 75 percent of the unobserved component so it really what we would call labor market discrimination is actually largely pre-labor market discrimination which suggests that there's a barrier to entry into good jobs into the labor market as a result of the hindrance of poor school quality now the one thing which we can note here is that school quality definitely is distinctly varies by generation people entering through the same schooling system also because a lot of the the laws surrounding schooling affected particular generations in South Africa such as for instance the Bantu Education Act which was also an apartheid error law which limited black African individuals to the type of schools they could attend and these were typically poorer schools so particular generations could attend particular type of schools and therefore the way we're going to tackle the problem of unobservables is by taking the unobservable up from the individual to the generation it won't account for absolutely all the unobservables but it will start arriving at better estimates to what we used to have we also have additional problems here which I won't discuss in much detail so our approach attempts to address these shortcomings because of problems of common support we limit ourselves to individuals who have eight years or more education this is quite simply because very few white individuals have primary school education or lower in South Africa and that that address that suggests this whole problem of varying schooling experiences which then have an impact on the labor market we also restrict our sample to formal sector workers because we assume that a lot of the laws that have been implemented do not necessarily affect informal sector workers directly we use the post-apartheid labor market series we we use it from 1997 to 2011 there was no earnings data in 2008 to 2009 that was formally published so we do have some gaps in those time series which I'll show you what is our approach we we decompose the changes in the wage gaps to try and arrive at the unobservables but what we do so is we specifically look at this within the same birth cohorts by looking at variation within particular birth cohorts to try and account for the fact that different birth cohorts have different unobservables okay so we we rely on a less stringent assumption even though it doesn't necessarily hold perfectly is that once we control for a look within the cohort that the unobservables across race would be more likely to be randomly distributed across race even if not completely so to look at this cohort type of analysis we borrow elements from age period cohort tradition particularly the the name here that economists would know most is Deaton who looked at this most effectively within our context what we do is we look at the change in the wage gap relative to a base period and we're going to start with 1997 just before the employment equity act was implemented we avoid looking at data before that because there's survey idiosyncrasies and essentially what we do end up with is an initial component which relies basically on changes in coefficients which is typically what we would understand from a Wachow-Blinder decomposition then we have a second component which is just a change in productivity because we know for instance education has been strongly converging across race groups in South Africa we then have an interaction term which needs a bit of unpacking and what you'll see is is that there's two components and I'm going to emphasize the first term because essentially what it suggests it's the increase in white attributes but valued at the initial discrimination so this does measure some form of discrimination but it's a persistent form of discrimination what we're measuring in this green bit is the changes in discrimination that has happened what we're measuring with this bit is a persistent part of the discrimination there is also a small second term but because what we're going to see is is that this blue term increases this is not as important even though for instance what we observe is that the returns to education for black African individuals particularly the returns to higher education have increased this should be important but we don't find that necessarily to be the dominant change in this identity just to give you an idea this is the evolution of the average white wage for formal sector above 8 education the dotted line is for black individuals and what we have here is the wage gap as I suggested the wage gap increased even beyond 1998 when employment equity was introduced but only started declining at a later stage we also there could be problems here with changes in server design the instrument changed along this period but it could also be differential responses to business cycle phases as this came shortly after the financial crisis and what we generally do see is is that the yeah we also can just recast this by cohort and what we see is is that wages are declining across cohort but I don't want to emphasize this because there's also an age effect in here what we see is is that across cohort the this top green line is is that the wage gap is actually remaining quite constant if we just do this with the usual Wachika blinder decomposition in each time period what we see is is that this wage this increasing wage gap only declining much later towards 2011 we actually don't see this any decline in this discrimination component not systematically except perhaps towards the end of the period so if you consider this 2003 as a threshold point when this aggressive black economic empowerment came into play it didn't seem to have any effect whatsoever but if you apply this this new approach we can look at this green term which is the change in discrimination relative to 1997 we actually see that this started picking off after 2004 recall however that the this blue component is also a part of component which represents the increase in white characteristics as if discrimination had remained constant across time that is increasing okay so what does it suggest to us it suggests to us that there's two components of discrimination that are moving in opposite directions this is the one that has been affected by the implementation of black economic empowerment it represents the pure change in discrimination which we observe as a result of the change in the laws but this component here is not not changing whatsoever so what does it suggest to us it suggests to us that the if we go back to exactly what that means this bit over here it suggests that the reduction in the for instance the the wage the attributes of white individuals it's it's not being the value it's not being revalued in that sense and we therefore do not see that reduction in the in the discrimination component so what do we know well we know that racial wage gaps declined before the end of apartheid it increased in the period shortly after the transition we do not see a massive effect for the 1998 Employment Equity Act even though it's something one would expect even despite the fact that tertiary education the returns to tertiary education increased for the black population this gives us an idea that perhaps these reforms weren't broad-based enough they weren't affecting the entire distribution and this is something which you would find being spoken about in the discourse quite extensively it's only the 2003 BE Act that does start seeing a change in one component of discrimination but in not all components of discrimination which suggests to us that even this hasn't had the desired effect to change the face of the labour market in a big way i'll leave it there thank you very much