 This morning we're going to be talking about affirmative action in South African higher education. This is joined work with Patrizio, who's over there, and also Vimal Ranchaud, who's also a colleague at the University of Cape Town in Seldru. And before I start, I'd just like to make a few thank yous. One to the university for giving us the data, the applicant data, which I'll be talking about. And the second to the Department of Higher Education and Training in South Africa, who gave us access to their HEMIS database. And lastly to Tim Brophy for some help with matching some of the census data that we're going to be using to our applicant data. That's the outline. So affirmative action is seen in South Africa as a way to make redress for a long history of racial discrimination. Returns to education, particularly at the upper end, are extremely high. And distributions of both race, of income and education amongst many, many other sort of many other concerns that we might have about people's lives are heavily skewed by race. And so there's been a recent discussion about actually whether we should have a class or socioeconomic status based policy rather than a race based policy, which is currently the case. And in fact for 2016 the university has moved to a policy that takes an income into account in making decisions about who gets into university and who doesn't rather than just race as is currently the case. So this kind of matters. The university has gone ahead and done this. But there hasn't been a huge amount of research, or in fact any, looking at actually the kinds of impacts that affirmative action is having on who gets in and who doesn't, how well targeted it is and how the beneficiaries and those who are displaced by affirmative action do. So we're trying to answer some of those questions in this paper today. So who gains and who loses? It would be interesting to know about some labour market implications if there's anyone here, South African policy makers. It would be useful to have data for example in South African revenue services or the tax data to find out how people are doing. We can say something a little bit at the moment only unfortunately about the possibility of mismatch that affirmative action is allowing into the university people who are not as well prepared and as a result they might not do as well when they're in the university. And aggregate social welfare effects, well that's a little bit beyond what we're trying to do in this paper today. So there's a bunch of literature from the US looking at the impact of affirmative action usually the bands that have been in place in a number of states fairly recently Texas, California, Michigan showing that if you outlaw the use of race and admissions that can lead to quite dramatic decreases in minority enrollments and a reduction in applications from minority applicants. Although there's some evidence that targeted recruitment programs can offset those kind of effects. Outside the US there's some recent work on both Brazil and India. In India there's a cost based affirmative action and there's a paper by Bertrand Edel in the Journal of Public Economics saying that basically affirmative action is successfully targeting the financially disadvantaged rather than high income members of disadvantaged groups and that low cost entrants obtain a positive return to admission. Although they do find that enrollment of women is much reduced as a result of this cost based affirmative action. And then there's some also quite interesting work from Brazil. The University of Brasília introduced affirmative action, I think it was a 20% quota for people who self-identify as black and that raised the proportion of black students and the displacing students were from quite lower socioeconomic status backgrounds than those who were displaced. So this is kind of specific to the university that I work for because there is a big debate in South Africa about affirmative action. I haven't checked the sort of polls but presumably the majority of South Africans think that affirmative action is a good idea. That being said that's not who the university gets pressure from and in fact I think this move despite the fact that most people think a sort of race based affirmative action policy in South Africa is probably a good idea that the university has felt enough pressure that it's decided to change its admissions policy in 2016 as I mentioned a bit earlier. So we're trying to answer a number of questions. First of all what would the distribution of offers to applicants look like if there was a race blind admissions policy. And so we're focusing on one particular part of the process that ends up with a bunch of people enrolling in the university. We're only focusing on the applications and offers that are made to the students not whether they, for the most part, not whether they end up taking up those offers. And how the admissions policy works at the University of Cape Town is that black students and to a lesser extent colored and Indian students we have these racial categorizations which we are still using in South Africa the stats SA still uses them and people still talk about them like this so they're still very salient in South Africa generally. And black Indian and colored students can get into mainstream programs through lower points requirements so they need less good secondary education results to get in. And then there are also these other programs which are for students who are a bit lower down even than most students they get into extended programs, academic development programs and only black colored Chinese and Indian students qualify for those kinds of programs. There's usually an extra year instead of taking three years for a bachelor's degree they would take four and they would get some extra tuition in years one and two to help them adjust to the university. So in practice how this policy works is really a system of quotas the university or targets the university calls them let me read one of them to you. The following example applies to applicants for medicine it relates to applicants who categorize themselves as black South African we set a target number which we hope to give qualified black applicants and this will be a proportion of the 200 places we have for the medicine class we set this target because we aim for a diverse medicine class and in order to give redress to black South Africans and then from 2011 we set overall enrollment targets and equity targets per program these are aspirational targets not quotas that's sort of seen as a bad word so I think the university wants to sort of avoid saying that it's doing a quota system and all faculties will aim to admit specified minimum numbers of eligible South African black Chinese colored and Indian students sounds like a quota to me but okay so we have applicant data for two academic years 2007 and 2013 and we have financial aid data only for 2013 and so we have data on all undergraduate first time applicants and which programs they applied for and really we're doing something quite simple we just want to simulate what would happen if the university used a race blind selection mechanism when it decided who to make offers to and who not to make offers to and our work is really focusing on applications almost all applicants in the data that we're looking at make two applications but we're looking at applications not applicants and foreign students are treated a bit differently they have their sort of own basket when the university is trying to allocate places to the different groups and it's much more difficult for us to do anything with the foreign student data because they don't have the same sort of score that the university uses to rank the person who's in charge of admissions at the university described it to me as kind of like a decision by committee rather than just the sort of mechanical ranking of people by their scores that the university does for everyone else so for the moment we're just going to ignore the foreign students that's not a sort of value free decision but for the moment that's really all we can do given the data that we have and the data that the university has and I think quite a novel and interesting part of our research is that we have two outside sources of data which we can use to I think say a lot more interesting stuff about what's going on and these sort of questions that I'm trying to answer and so the first is that we have a database of everyone who's registered at public high education institutions and we try and link that to the 2007 applicants to work out what happened to them we know what happened to the University of Cape Town students we have that data but what we're also interested in what happened to the students who didn't get made an offer or chose not to enroll even though they were made an offer and then we also are interested in a question of targeting does affirmative action successfully target low income students and to do that we have the financial aid data from the university that's a partial answer and another sort of partial answer is that we can link the addresses that the university students give to the university to census small area data from 2011 and using Google Maps so this is the data that we have for 2007 and 2013 you can see first of all that there's a big jump in the number of applications from 2013 that's because we had a change in our secondary school leaving certificate that made it easier for people to get the university entrance certificate and so that meant that a whole lot more people applied they obviously are not guaranteed to get in so the number of offers didn't change that much but the number of applications changed quite a lot and what you can see is in both fields black applicants make up about 45 to 50% of applicants so we're ignoring the foreign students here and then white applicants are around applications around 30%, 23% in 2013 and then in the next column you can see the actual offers that the university made to people so it made 31% of its offers to black South Africans 40% to white South Africans so there's a sort of change there white students generally have higher point scores when they arrive at the university the schooling system is very unequal and that's a key part of the explanation so you can see that there's sort of a bit of a reversal there and then what we try and do is simulate what would happen if the university made offers just based on the points and not made offers to certain groups and minimum numbers of offers to people in certain groups and what you can see is that for example in 2007 as a total fraction of offers the offers made to black South Africans would decrease from about 31% to about 25% so it's not massive but it's also not small for example the white students their fraction of offers would increase from about 40% of total offers to 47%, 48% and you can see similar kinds of numbers in 2013 maybe the jump up for white students would be a bit higher in 2013 and the jump down for black students would be a bit lower so that's our sort of answer to the first question what's proportion of... how significant is affirmative action in changing the distribution of offers that happens that are made by the university okay so I think what I'm just sort of repeating perhaps an important point to make those second from bottom is that most applications are either always rejected or always accepted and in 2007 about 17% of applications are affected by this change in policy that we sort of simulate between affirmative action race-based policy to a race-blind policy and that number decreases to 11.5% in 2013 mostly because there are a much larger number of low quality offers who are just always rejected whether or not there's affirmative action so the next question is is the admissions policy well targeted are black, colored and Indian students who benefit of lower socio-economic status so the first one measure we can use to assess this is the financial aid application and eligibility not everyone applies for financial aid so this is not going to be a very good measure and then the second one is per capita income in the census small areas in which applicants live and this is a bit of a difficult matching process Google only allows you to match a certain number of things per day and we made certain agreements with the university about who could use the data and who couldn't so in the end what we've done is just do this measure of the census per capita income measure for displaced and displacing students only and we check how well correlated these two measures are so this is, we only have this for 2030 how are we doing for time? five minutes, thanks so what you can see is that so these are our row percentages so the black applications around 40% did not apply of those who, of the total 7 or 8% were ineligible and 52% were eligible if you look down at the white applicants or applications 88% didn't apply 4% that did apply were found to be eligible and 6.5% ineligible and 6.5% were found to be eligible and if we look at say for example displacing black and displaced white students you can see that again a large fraction of the white students didn't apply a very small fraction were found to be ineligible and again a very small fraction were found to be eligible for the displacing black students 50% were found to be eligible and only 36% didn't apply you might think for example that the white students feel that they're going to be discriminated against in financial aid and so they don't bother to apply even if they might be eligible so you might think this is not a good measure and so the second thing that we do is to match the census match the census small area data these are about 250 households so we have a very consuming quite nicely on the areas where people gave their physical addresses and try and work out what's the per capita income in that small area obviously we don't know this for the individuals status A is not giving us the micro data with location so this comes from a kind of histogram that status A gives us of income in a small area so it's not household per capita it's small area per capita income and what you can see there is the CDFs from the black displacing and the white displaced students it's in South African rands per month and the status A like is that the 80th percentile for the black students is about the same as the 20th percentile for the white students so there's some overlap but it seems to us that the policy is very well targeted this is just because in South Africa race is really extremely highly correlated with income much more so than in the US or other places that you might be more familiar with and then this is just a sort of correlation this is for the displacing and the displaced students only this is why the numbers are a bit smaller here only 2013 and we're just trying to assess whether or not there's a good correlation between our measure of financial aid and the quintiles for small area per capita income and what you can see is that it's pretty well correlated most of those who don't apply are in the sort of higher quintiles and most of those who are eligible are in the lower quintiles so it suggests that's just perhaps interesting in and of itself are people who don't apply for financial aid sort of richer or poorer this suggests that they're generally speaking quite a lot richer although there are some people in the low quintiles for example who didn't apply I mentioned the possibility of mismatch this is something that we haven't done enough work on yet perhaps like to hear some suggestions from the audience later on at the moment all we've done is for the 2007 applicants we've matched them to the HEMIS data to public university data set to find out what actually happened to the people who were either displaced or were displacing as a result of the affirmative action policy in the university and what you can see is that enrollment rates are extremely high so private universities in South Africa not as common as there are in many other developing countries so most people end up, most of the applicants who applied to UCT whether they got in or not most of them, whether they made an offer or not most of them end up in a university in South Africa the graduation rates however suggest that there might be some possibility of mismatch displacing black students have quite low graduation rates relative to the displaced white students it turns out in a table that I'm not going to show you that this phenomenon of cascading where, so UCT is probably one of the leading universities in South Africa and displaced white students end up in kind of slightly lower ranked universities than the University of Cape Town or kind of similar types of universities so we're thinking about some sort of regression discontinuity design type stuff but at the moment we haven't got there so just to conclude the admissions policy does have important effects on the offer distribution 17% of applications are affected that's only 11% in 2013 and our results suggest that beneficiaries of the policy are generally a much lower income level than those disadvantaged by the policy and there's some suggestion that there's some cost to this and the fact that there's lower graduation rates for the displaced black students displacing black students relative to the displaced white students but that definitely needs more work thanks