 It's a great pleasure to be with you all this morning as another of the representing a very good team from the South African, that did the South African study. It's Vimal Ranchard who's here and myself are writing the summary paper. That's how come our names are up there. I'll tell you who the other players are as we go. Like the other studies there's been a lot of work on South African inequality and we know lots of things. It's a very, very unequal society, one of the most unequal societies in the world. Previous work has placed the income genie around 0.65 with lots of debates about the data and the etc etc but that doesn't shift the very high level that none of those debates change the fact and it's a fact. It has a history too and so history doesn't start when the surveys arrive and so one of the analytic issues that we confront in South Africa is to take some sort of stock of initial conditions or call it what you like because otherwise if you're doing the changes and similar to all the other country studies our focus is more on explaining some of the changes in inequality as the value add rather than just measuring the inequality. So our history doesn't go away. You have to somehow bring it into the surveys. So we produced four papers that they quite similar in the topics because these are areas that we had thought in the broader project would add value to our understanding of inequality in developing countries. And the first paper was a sort of a context setting paper in which we did the type of decomposition exercises that we've seen for the other countries where you take in this specific case you take income sources and you decompose the changes over time. I guess that was the value add in our exercise. We saw exactly that exercise in the Brazilian data and in fact we got the technique from our Brazilian collaborators. And we have done many decomposition exercises in South Africa. We've decomposed everything in site between and within race group between and within rural areas etc. And we redid the static decomposition exercises using slightly more updated taking the story up to 2014 just to make sure we were getting the same story. And it is the same story in a static sense of the dominance of wage income in dominating the household per capita income earnings distribution. And that is what the static decomposition shows. But it also hides quite a lot because it doesn't take account for example South Africa has implemented a huge cash transfer system over the post-apartheid period. Now if you take a cross-section and you do an income source decomposition you're never going to pick up the impact of that because it's embedded in the data. The dynamic decompositions help a bit in that regard. They also allow you to bring in demographic change because the denominator in these exercises is household size or something which you can also then decompose in our case into adults you know the share of adults in the household and the share of employed adults in the household. And so some of the changes that we've been attributing to income sources are actually due to substantial change in the demography greatly reduced household size in South Africa which increases the number of adults per household. But then you've got an underperforming labor market. So enough said, so what do we find from this exercise? It doesn't take away the key finding and you wouldn't expect it to. It's very hard to think of a society where the drivers of the dynamics of inequality aren't going to feed through the labor market. It's just hard to think about that. And so the labor market is still very important. But some of the demographic changes do lower the contribution then straight of the wage income coming into the household. So as I said earlier, the number of adults per household has increased but our labor market hasn't been great. And so the share of employed adults in the household hasn't increased so much but if you're in a smaller household they can make the contribution as complicated demographics there that have lowered inequality. So if you just look at the income source, the wage contribution, it looks as though the wage contribution has gone down, not become equalizing but become less disequalizing in the recent period. If you take out the demographic changes that goes away. The wage income itself hasn't dampened. And then the social grants then kick in. If you look at the densities in South Africa and you take the social, you include the social grants and you take them out, the whole bottom of the distribution changes. It's implausible that an income source decomposition can say wage, social grants are doing nothing. It doesn't make sense. If you do the dynamic decomposition you can see it has a big redistributive impact. It picks up in the changes. So and that's very interesting and very important I think. The aggregate trend seems to suggest that our inequality sort of topped out in about 2008 and has decreased a little bit since then. Other data shows it definitely shows a flattening but doesn't show the decrease. Now we've been using the data for this exercise was the National Income Dynamics Study which is a panel data survey. And one of the issues obviously with panel data is attrition. And who attrits out of panel data, the wealthier sections of the data set, they don't like telling you about their stuff, right? And so these issues about the data quality and whether that decline in inequality is really real and a need to explore the top end of the income distribution has an obvious resonance. It's not just because that's what everybody's doing. There's some logic to it. So the second paper had a look at the top end of the distribution beginning to merge tax data and these surveys. Following methodologies in the team, what we did in this particular exercise, the idea was to use the surveys for the bottom end of the distribution and the tax data for the top. But where is that threshold point? In this particular paper which was produced by Ingrid Willard and her student Janina Hundenborn and Jan Gelsamer from the World Bank, they wanted to use the thresholds in which you actually have to file tax returns and motivated for the use of those data, applied those thresholds and then so basically you use the NIDS data before the threshold and then you use the tax data above the threshold. And the estimated Gini coefficient of taxable income, be careful here. So the 0.66 I was talking about earlier on is the disposable income. This is income before tax is 0.83 which then does also decrease. I don't know if you can see this. It will certainly make sure you're still awake. Basically this is the difference in the mean income between the NIDS data minus the tax data. So you don't have to report at the bottom but some people do. So you do have tax data across the distribution. And anyway the graph just makes, it's quite intriguing because you can see at the bottom end the tax data, there's a huge difference in a positive NIDS gets much higher income levels than the tax data at the bottom end with all the problems of not many people actually reporting there. In the middle it's very flat, it's sort of the same. And then at the top you can see that the NIDS data is substantially below the tax data. And so it's likely, without getting too panicky about this percentage point we can explain it, et cetera, et cetera. But that decline in inequality that we see from 2008 to 2014 is probably not there to the extent that it's there. And the other data triangulating implies that it's not there. And if you then look into the labor market, which is what the third paper does, because the labor market still is the driver and it's the driver of the changes and it's going to have to be the driver of the transformative impacts on our society. So we had a very hard look at the labor market earnings inequality situation in South Africa. And we have a very good data set, it's called the post-apartate labor market series, it's publicly available if I can put in a little advertorial. It's basically statistics South Africa data, but one of the data units at the University of Cape Town has worked very hard to harmonize the data, give you harmonized weights for appropriate comparisons over time, et cetera. Actually definitionally, the income, the earnings variables measure the same stuff. And so we had a close look at this inequality thing, two key points came up. We did some complicated things, well they're complicated for us, we're learning as we go. So we did some recented influence function regressions we learned from our partners in the network. And to look across the distribution, not just to look at average effects. And I'll give you the findings of those. But first we found this huge break in the data. In 2011, the inequality, if you just plot it, you're not riffing anything, you're just plotting the genie. There's this jump in the genie coefficient in the data. At one point it's implausible, well it would be plausible, but you'd have to construct a story of something happening in South Africa, a meteor hit or something. And so we explore this data break and it makes a huge difference to everything we do. If you ignore the break, inequality goes up, right, it jumps up. And so basically we've got a problem. We've got a data problem that you have to confront and if you stretch across that band, what we do show in the paper is that the riff exercises you do give very, very different answers to if you do to stop just before the break. And so that's very important we think. And for example, there's a World Bank country report that was released this year that ignores the break. And so it makes too much of a big deal about the explosion in the earnings inequality in the labor market. It has gone up over time. It hasn't gone up like that. And that's really important. So what do we find in the riffing thing? Nothing fantastically interesting, but all important, I think. For example, the changes in the returns in the labor market, which you've even picked up in the average earnings function type thing, but become very, very important in allocating you across the distribution, for example. Going towards policy, our country study is then closed with a paper that, again, Ingrid Willard wrote with Masher Kwe Maboshe, one of our young researchers, PhD students in the School of Economics, that does a benefit incidence type exercise. These have been done before we updated to a more recent year using something called the South African Living Condition Survey. We've put good data in the country, and it's made publicly available by a stats agency. And so we extended it. So there are two contributions, really. One is the extension. But we also focused on a few issues that hadn't been explored before, tax exemptions in the distribution of disposable income, like we looked at their role. And then one can look at the impact by race, group or gender or age. We find if the red and the blue, the red and the blue reflect is the big story here. It tells you, let me start with the dotted line, which is market income. So that's the inequality of South African society. That's your market income. The fact that the Lorenz curves for concentration curves, really, for the tax income is so much below that, implies that the upper end of the distribution of being taxed, right? So it's a very progressive, well, it's a progressive tax structure. It's not quite as progressive as it looks, which is what the careful exercises in this paper show. But that's what we've got. We've got a progressive tax structure. And it's partly a product of our inequality, right? The very, very low incomes at the bottom, you don't pay much tax on that. That's not because we've got a, well, necessarily a progressively designed tax structure. So some of it's about the design, but some of it's about South African inequality. This graph then just shows you two of these sort of deductibles, if you like, in the tax system that we haven't looked at in the country before. You only start paying tax on interest income above a threshold of interest earnings, and you get a medical tax credit. You can claim back private expenditures that you put into your medical aid. You can claim back through the tax system, which is really important if you're trying to articulate the health equity in the country and the mesh between public and private. You can see here that they're very regressive. And there's other good work in this particular paper that shows, for example, how of evaluated tax, certain components of evaluated tax are very regressive too. And there's some stuff that we haven't yet got the data for. So there's work to be done. For example, the capital gains tax is a very important component of the South African system. Now you've got to dive into the tax data. It doesn't, we don't have it yet. It hasn't been articulated yet, but it's almost there, ready for us. And it's very, very important. And then the wealth data isn't well done. So quickly, the big picture coming out of this slide is that our cash transfers are very, very focused on the bottom end of the distribution. They're very pro-poor. And so that affirms the household level study that we started out with. But now we've dived into government, right? So this is now admin data combined with survey data, part of the richness of the study. So our main findings, our taxes are progressive. With the top three decels contributing 96% of personal taxes, our cash transfers are highly progressive. And they reduce the genie coefficient quite dramatically if you do it properly in the change. Tax exemptions, however, are regressive. We did lots of other work to, we've done lots of other work on the side of the project that resonates with what else you've been told. We've used our panel data in particular to tell a story of very regressive, this is the intergenerational mobility story that came up in a few other papers. And it's a dreadful story. So I keep on contentiously calling it intergenerational failure whenever I talk about this in South Africa. Because these are conditional upon some quite sophisticated modeling for these initial conditions, et cetera, et cetera. This is a sort of a father-son, a parent-child correlation, if you like, coming out of some modeling work. And you can see this is sitting at .9, right? So it's not the straight correlation, the .7 that was the Premier League in Latin America. That's not a fair comparison. It's not the straight correlation. But nonetheless, this is devastating, right? This is how South African society is reproducing itself. So not only do we have high inequality, but we're not really, it's persistent. We've done other work on the poverty dynamics and the middle class dynamics that also show a very slowly transforming South Africa. Leave that for now. So final point then. So the study, I think, our studies have made some contribution in pushing on a little bit in the knowledge base. But they've only then surfaced the next round of issues, which are then about the sort of social and economic and labor market processes that drive the inequality. These decompositions are proximate drivers, if you like. You've still got to work on how come education is better years of schooling, massive change in years of schooling, hasn't translated into better outcomes for South Africa, how's the labor market working, et cetera, et cetera. And that's then the next issue that we're dying to confront, hopefully with our lovely partners. Thank you.