 Okay, thank you. I'm Jukka Birtele here from ULU-Vaider. Thank you first of all for the very nice panel. I don't know what happened in the competing sessions, but I certainly don't regret that I came to this panel, so thank you so much. A couple of questions related to the papers by Mare, first of all regarding the first. Can you elaborate on the policy conclusions you can derive from that analysis regarding the South African education system? How should the education system be reformed to make it more prone to foster egalitarian distribution of earnings? A second question related to the paper on the incidence of the grants in South Africa, which was presented by your cohort. I was wondering that we know that these grants must be financed by taxis, and you didn't seem to have the incidence of taxis in the papers. I was wondering what was the reason for that, and if you did include those, what would happen to the full incidence of the combined tax and benefit system? It's a very quick question. I echo Jukka's comments. It was a great session. It's just a clarification question on the second paper on the differences between Brazil and South Africa. We heard this morning that inequality came back big time. I think structural changes is next. Then it's a very wise choice of topic. I just didn't hear much about one of the most fascinating aspects of structural change in Brazil. I think that throws a lot of light in your comparison, which is the very, very rapid growth in productivity in the agricultural sector. If you look at any measure of productivity of Brazilian agriculture between 1960 and 2010, less than 50 years, the astonishing bit is to grow faster than Chinese productivity. I just would like to hear what are your thoughts and how much it illustrates your point on unusual patterns of structural change, but I didn't see much thoughts on that. I'd like to hear more. Question from Murray. I'm curious whether your independent variable, the number of years of education, is measuring the same thing at the beginning of the period and at the end. The metric results rather imply, or at least the ones I've seen, rather imply that the variable changes fairly substantially over the period of time. I wonder if you could comment on that, please. Thank you, Christophe Muller from the University of Ex Marseille in France. I'm impressed by the breadth, the quality, the interest of all this work coming from South Africa. So I'm happy to say that. Murray, your presentation was very nice and very clear and it helps a lot to understand the relationship between inequality and community equality and the dispersion of education by using this regression, this usual income regression or wage regression. Naturally, you know the main problems with this regression, with the measurement of return to education, the two main problems for a long time in the econometric literature. The probable undogeneity of the education variable because of an observability, an observ motivation, measurement errors and so on. There is a huge econometric literature on that. And also the question of selectivity, especially over such a long period. It's likely that while people come in the labor market, they leave, they finish school, they migrate, etc. So it's quite complicated. And naturally you may want to correct the estimation to try to confirm for that with typical method as usual. But also I found that because your setting is so clear in the beginning, it's interesting perhaps to take on board these difficulties and to see how this would be modified by acknowledging this difficulty with the education on the selection. And I think probably it's not going to be too complicated and you may be able to find proxies to measure the impact of this problem. On the second paper, Joshua, at the end of the day, you gave us a lot of interesting information on South Africa and Brazil. I have a little bit of difficulty to understand what do we learn about this comparison. Could you perhaps summarize in one sentence what is really the main thing that we learn by comparing Brazil and South Africa from your point of view. On the third paper, it's very tempting to try to think about the issue you are looking at by using treatment effect estimation instead of decomposition. So have you thought about the connection of these two kinds of technologies? There are different ways of looking at the same thing. More and more people are going to expect treatment effect estimates when you talk about the impact of any program. So you need at least to position yourself in comparison to this literature. On the final paper, this asset index, the CPA or any other multivariate summary method, they are really useful and I think in the last few years we have seen a lot of work using this data which was not used before. And we learn a lot by trying to mobilize this information by having these aggregating indices. One problem with these indices is that they are data dependent. When you estimate a score, you change the data set, you have a different score. And because you summarize dispersion of information, a lot of what you summarize is about dispersion. When the inequality changes, the formula of the score changes too. So I think there is probably a good reason when you work with these kind of indices to also partly monitor how does the score change with the change in equality over time. Because you want to control from that. Thank you. My name is Ali Rashid. I'm from Gayron University, Egypt. My question is for paper number three, which titled a ceasing impact of social grants. Actually my question is about the over-employed person's impact on the inequality. You showed that in your simulation that it increases the inequality by five times. But I'm not sure about that result if you would desegregate this by quantizers of population. For example, maybe the over-employed person's decrease in equality. And maybe for the fifth quantizer, the fifth quantizer, maybe you are right, it has decreased the inequality. So I'm not sure about the final result, the total amount of a change should be different. Thank you. I think there is one question in front here. Thanks, Joel Neshtenger from South Africa. Two questions based on the last presentation. The one is simply about the definition of assets. They are private assets, but they are also public assets privately accessed by households. So do we include those in assessing the asset index and inequality? This would apply for instance to water, sanitation, refuse removal and so on and so forth. That's the first issue. The second issue would be about assessment of assets in rural areas and the impact of communal land tenure system. Disorder would argue that many assets of the poor are dead. They are dead assets, but this is even aggravated by the communal land tenure system. So that in the rural areas those assets maybe are not only dead but also buried. I'll do it with that challenge. Thanks. Okay, so Marie, can you start with answering the questions and then we can go in the order of the presentations so Josh can go next. Okay, so on the education paper the policy conclusions, yeah, so that's a hard question because hard in the following way that part of the point of the paper is to show that if you're translating education change into earnings change, there's both the x's, the years of schooling and what's in the realm of government and what government can do and then there's these betas and they squared when it comes to inequality. And those are returns and those come out of the labour market in general. I would say though that understanding that the actual cross point with the mean earnings and the mean, the years of schooling corresponding to mean earnings is quite useful in understanding how the labour market is working at any point in time. It gives you some benchmark because otherwise you just have a furrowed brow. You have the Brazilian guys bemoaning the fact that they did a lot for education and in South Africa we bemoan the fact that we worked quite hard, put a lot of money into increasing years of schooling and were successful at that. But what happened? We didn't get a big return but then suddenly in Brazil come 1999, 2000 they start getting a huge return. For sort of the same thing, it wasn't though that policy changed dramatically. So I don't think there's a sharp policy implication in a sense but hopefully understanding these things and understanding the zones where you're operating. So we are, like we do sit in South Africa and worry a lot now about people that we've successfully moved up into grades 10 in 11. But we haven't given them much actually as citizens. One of the issues that does be devilish of course is the focus on years of schooling as was said when the question that was really asking about metric. It's asking to some extent about quality concerns. The return that you get on schooling depends upon quality and that's valid and that's correct. So I'm not disputing that. It's a very important issue in South Africa and everywhere. But don't forget that there's huge debates about skill bias, technical change and the impacts. At the end of the day the literature is coming back to years of schooling and the Brazilian story that was told. These issues of quality are fundamental but I wouldn't want to turn the South African explanation purely on a quality explanation as well. When there is the demand side for labour in the mix as well. So obviously quality concerns and then Christoph raised a bunch of concerns about the actual earnings function itself and the endogeniative education and selection etc. That's true. I mean some of the results in the paper in a sense we're trying to use some mapping between education and earnings. And we're just using the earnings function as that mapping. But nonetheless I think you're almost making the same point about the quality but in a different way. There's many concerns about years of education and whether they are real and therefore whether you get the true estimate of the returns even. And looking hard at that I think is a very important part of this research programme. That's quite right. Shall I stop there on the schooling and then other people can talk. Okay thanks so agricultural sector in Brazil. Yes this is a great point and there is a little bit about that in the paper. And that's one of these things that got lost somewhere between the paper and rushing through the presentation here. But yeah so Brazil in the sample that I'm looking at Brazil there is a higher portion of the population that are working in agriculture than much higher than there is in South Africa. And when I did this decomposition of change within group improvement and shifting there was a big portion of Brazil's improvement was explained by within agricultural worker groups. So I agree that's a great point and I'm glad that you brought it up. And it also goes towards this example this question of you know what do we learn by looking at the comparison where here's something where there's a big improvement in one country that didn't really happen in the other country. So maybe we can learn something about that as far as answering you know what other routes to improvement with these different structural change might take place. What do we learn from the comparison can I summarize it in one sentence probably not but I'll try in a four or five let's say. So I mean one thing is sort of the example that the other questioner just brought up of okay so we can look at okay here's one way that there was improvement in Brazil that there wasn't in South Africa maybe there's something interesting here. Another thing we could take away is something that I brought up actually in the conclusion of how okay so we saw in South Africa there was a big improvement in one of the big sources of improvement was a shift towards low productivity urban workers. And in Brazil there was a major source of improvement from within that group of low productivity urban workers. So maybe there's something hopeful here that this will work and we can hope that South Africa could look at how these workers have improved in Brazil and try to do the same thing maybe. But I think a bigger takeaway perhaps related to the original question of what is this these new patterns of structural change matter for well being. I think in both countries we could see that these urban high productivity workers which is basically manufacturing are doing better than these other groups of urban workers that I'm talking about the low productivity workers. And also in both cases there's almost no contribution of change of a shift towards those types of work right. So at the very least I think we can look at that and say maybe these new patterns of employment really do matter in a way that are affecting people's well being and that might be an interesting thing to learn more about. Thanks. Hello. So there were two questions the first related to estimation and the one to the over employed. So just on the estimation that the purpose of the paper was specifically to compare income decomposition techniques and using the South African case study of the expansion of social grants. The main reason for that focus was pretty much based on the last technique the Barros approach which originally is published in Portuguese and probably because of that hasn't received much attention in the literature. While the others have and so the purpose of the focus of this paper was to pretty much bring that to light and draw attention to that while comparing it to other techniques and show that there's been great strides made in this literature. Because of that we haven't focused on any estimation approach as you suggested but I think that would be the natural next step is to compare these techniques to an estimation type approach. And then finally I believe there was some confusion over the over employed. Maybe this is just a result of rushing through the presentation. It is not the over employed but one over the employed so as a fraction and so changes in the proportion of individuals in a household of adults. Changes in the proportion of adults that are employed in the household has resulted in changes in inequality but not necessarily the over employed. That's a different subject matter altogether. So quick commercial on Jukka's point about the taxes, the tax side of it. There is a big study that's looking at the sort of taxes and the incidents and the redistributive impact of these social transfers. That's not really the literature that we locked into here but it is happening. It is obviously very, very important component to all of these studies. So there was a comment about the volatility of these index scores on different surveys. And I think that's a fair comment and I think that one thing which I didn't stress enough towards the end of my presentation. I had it right up front is that it goes back to the point that you actually really want to interrogate the indices before you use them. I mean I guess there are too many people who are just quite happy to smash things together and run with it and I think that's a problem. So they have their use and the last thing that I want to encourage is an equally uncritical smash things together and hope for the best. And now you can also calculate genies on them that could lead to a lot of really dubious papers unfortunately. So I don't want to encourage that. I just want to say that there is actually a possibility of thinking through some issues which we might not have been able to do beforehand. On the question of the rural stuff or more generally the public versus private assets. That is actually a really interesting question and I think that it really goes to the heart of what you think of an asset. So I think that if one wants to do what I advocate here then you really have to be quite clear in your own head as to what is an asset for a household. And I think that the film and Pritchett approach actually was quite agnostic. You see what works whether it's infrastructure, whether it's anything effectively is game. You can basically building materials and I think that it's not clear to me that that kind of completely agnostic approach is right if you're wanting to think about who's got stuff and who hasn't got stuff. If that's ultimately what you're interested in. So then I would say kind of like which types of infrastructure if you're talking about public assets really behave like assets to households. If you have a tap inside your house is that really an asset and that probably is. If you've got a connection, electricity connection in your house is that really an asset for that household? Yes or no probably, but there's a whole bunch of other things which are probably not assets in that kind of context. The stuff about land tenure I think that really I'm going to cry off there because I think that was really not in my stuff at the moment. But I think that you may have a point, but this is really not something I'm looking at at the moment. We have about five minutes left so if somebody has a last one to questions or comments we can take that now. So it's to Martin. So now you said that everyone should use it carefully and all that which I think is great, but my question is like you were cautioning about the potential problems of your approach. But then you mentioned them, I don't know if I got it wrong, but you mentioned them with respect to inequality, things that would seem a bit like it's just the distance between those that have and those that have not. But like to use it as a proxy for wealth, what is the problem in principle? If it satisfies the actions that you said which to me seem sensible, what is the problem? Yeah, so that's something I actually had to cut out in my rush to try and finish at least within a minute of the closing time. So the real problem is that the normalization that that banergy procedure does, if an asset is scarce it gets a much higher waiting, which for some things seems right. But I mean if you do it on certain assets for example motorcycles are very scarce. So they actually get an insanely high score if you actually put a motorcycle in such a way that actually you'd say that can't possibly be the right score because it still works at the end of the day. It's not a standard correlation, it's a correlation on the uncentred variables. But if you wait the motorcycle up to that extent that effectively somebody has a motorcycle there will be at the top of the thing pretty much regardless of anything else they have and then that seems wrong, which is. And I guess that that's the problem with any automated procedure. If you're essentially letting the computer make certain decisions for you there's going to be some context in which it's going to create garbage. And that's I guess why I do want to caution people before that this thing takes off. Yeah just please keep it short. My question is don't you think finally that the score you define can give the very high weight to asset which final is not very important for the household. So that's true even of the Pritchard and Filmer stuff. So I mean I guess the problem with any summary measure when you're effectively looking for a correlation structure inside the data is some things may be an indicator that somebody is doing well even if you think that particular asset shouldn't have that weight. I mean I guess there's sort of two ways of thinking about it. The best way of taking assets and converting it into wealth scores if you had prices on them. Nobody would quibble if you actually had the price of the asset to use the prices the weight. If you don't have prices then you're using something else. You're using essentially something internal to this structure that the way these things correlate with each other. And then you can find things that may be a sign that these people are well off whereas the thing that's actually picking it up itself is not very expensive. So it may be a reliable sign of the fact that the household is doing well but the weight it's getting may not be right in terms of the price of that particular asset. So in the case that I looked at on the DHS the asset that gave me after motorcycle the biggest kick was computers. And there I had no qualms in saying well that probably in terms of prices much higher than cars. So you'd say in terms of the relative prices those are the wrong exchange rates. But in terms of what it symbolizes probably everybody who had a computer almost definitely in South Africa would have been right at the top end of the income distribution. So that's the problem with these procedures is that they're actually kind of really relying on the fact that overall it gives you the right kind of ranking and the right kind of scores even if in the detail it sometimes looks odd. The thing that I guess I just have a problem with is when people put in things that are real assets like horses with negative weights which actually does mean in South Africa that the poorest of the poor in the rural areas if nothing on anything actually looked better off than people who have those assets. So we've come to the end of the session. Thank you to all four speakers for very interesting presentations and thank you all for being here. So we now break for coffee and then reconvene in half an hour for the next and the final session of the day.