 So in terms of the questions, please keep to the format by mentioning your name and affiliation before you ask your question. And please keep the questions very brief. So, yes, Jukka. Yes, thank you. A very brief question to Andy about the poverty elasticity of growth. That might also depend on the initial deepness of poverty. So have you looked at the poverty elasticity of poverty cap or the square poverty cap to growth? Should we take a few? My question is for Andrea, but it's really to do with researchers in general. When you talk about the genie coefficient, which genie? Is it an income genie? Is it a consumption genie? Is it pre-tax, post-tax? Is it per adult equivalent? Is it per household member? Is it per household? It does make a difference. And I always feel uncomfortable when in practically 100% of the cases people do not specify. And also I am very worried that in many instances they don't use consistent definitions of genie. But this is a general point, but I think it's an important point. Then on the substance, I think you're absolutely right that there is a somewhat of an institutional vacuum that comes to social protection schemes in Africa. But I don't think it's as bad as you implied. For instance, I would add a number of institutions that have been quite successful. BRAC in Uganda has been extremely successful. It's a Bangladesh program that covers something like, I think, 200,000 farmers. It trains entrepreneurs. It provides scales to young ladies and so on. The Ethiopian Safety Net and Public Works program has also been quite successful. So there are some institutions which might be mentioned in addition to the ones that you mentioned in South Africa. And then finally, in terms of the relevance of Latin American institutions for Africa, I feel very strongly that both the Bolsa Familia in Brazil as well as Progresso and Oportunidades in Mexico could successfully be transplanted to Sub-Saharan Africa. And we know that Oportunidades in Mexico has been responsible for a major reduction in genie. I don't know if anybody measured it exactly, but for the whole of Mexico in the last 10, 12 years, the genie coefficient has gone down by eight points. And of course for Bolsa Familia, you mentioned that it was responsible for something like a four-point reduction in genie. So these are my comments. Yes, please. Thank you. My name is Stephen from the University of Arsalaam, Tanzania. And my question goes to Kornia. You seem to mention that inequality remained constant over, especially in the Sub-Saharan Africa. And for this, I also note that in Tanzania, for example, inequality as measured by genie coefficient has remained almost constant over the last 10 or so years. But in one of the slides, you seem to suggest that some measures to reduce inequality. I would like to know why you are making these comments in the sense that is the inequality that you have observed? Is it too high maybe to have like significant impact on poverty reduction? Because as you mentioned that it has remained constant over the period of time. Why are you making some suggestion on how to reduce the inequality? Is it too high maybe for the growth to have significant impact on poverty reduction? Thank you. So, oh, okay. Yes, Moabo. My name is Jeremy Moabo from the University of Nairobi. My comment is on South African paper. What you show there is that you are able to explain all the differences in poverty across groups. So, in fact, your model can be used to eliminate those differences and even to eliminate poverty. But when you look at your specification, all the variables there are endogenous. For example, location. People are in slum areas because they are poor. Okay, it's not the slum which is making people poor. Although also that is true. Okay, so you have to recognize that problem. Then you have the number of children. Okay, that is a debatable issue. Okay, what the research shows is that the higher the family size, the higher the family size, the higher the probability of being poor. Okay, but children are also a choice variable. Okay, so you also cannot take that from granted. Then your key variable is education. Okay, you did not show the mechanism through which education affects poverty. Okay, and what we know from the standards on unemployment among the youth is that the highly educated youth are the ones who have high incidences of unemployment. So the more education you have, the more likely to be an employee and therefore you have no income. So everything else being the same, okay, education makes you in Africa, makes you poor. So it's a curious finding for South Africa where actually youth unemployment rate is very high among the educated. To find that education is the inducing poverty. So you need a better structural model to do that in the composition. Thank you very much. Okay, so if you want to respond. Okay, just one comment from Yuka about the poverty elasticity, the growth elasticity of poverty. Yes, those were elasticity just of head cons of poverty. African poverty we know is quite deep and that's part of the reason why the elasticity is probably quite low. Certainly we should look at it. Certainly I should look at it for other measures of poverty. Yes, I completely agree. Now on Eric's question. Now the data which we have, sorry, I went a little bit too fast, but I covered the huge area because the German is very tough. So I took cover huge area in a few minutes. So the genetic efficient which are reported in WID and POVCAL, these are average household income per capita. Nobody knows whether a pre-tax or post-tax. These are probably post-tax for those who pay taxes, which is probably 10% or 5% of the population. Because most of the people are not subject to direct taxation and indirect taxes are not accounted in that. Now the new data which the World Bank is giving us is our consumption data. So which will create an additional problem which normally you solve either by standardizing or by including dummy variables in regression. And so the problem of the heterogeneity is either you recompute. So you take the average difference and help. Or basically use the dummy variable in regression. And normally the dummy variable for the consumption variable, consumption gene, it would be negative. So it would lower. Now in terms of protection, I think I basically read two papers. One, a compilation from the World Bank which says now there is a sort of an inventory of all the new programs which are available. And then I read Armando Mariento's interesting paper. Which basically argues that the Middle Africa model is expanding and probably I agree with you that it has quite a scope for expanding. But the only one which is really available at the moment is the Southern African one, particularly the pension and children transfers. And that is not by mistake because the system was available to the whites. So when apartheid finished and the system has been extended to everybody else. So the institutions pre-existed. And then these are also state funded in Southern Africa. And then in the case of Middle Africa, they are mainly donor funded. Which raises the issue of sustainability. Now I agree with you. I think that nobody believed, including myself, that the opportunity that is in Bolsa Escola would have had such a large impact. But then we are quite convinced that they did. Now state capacity in Brazil or in Mexico is higher. But then we do see that in also in some of the African countries, like in Southern Africa, they just transfer the money without giving the money. Just straight to the bank accounts. So now on the calling from Dar es Salaam. Now I don't remember all because I have 28 trends in mind. But I think that Tanzania had rising inequality. So if it did add rising inequality, then one of the reasons is that I think is among the options discussed this morning. Well, first of all, if it rises, the poverty elasticity of growth falls. And that is not good for poverty. And then it may be more. Higher inequality probably discourages growth. I do tend to be uncaldorian in this respect. So why do we recommend policies to reduce inequality? Because it's not good for poverty. It's not good for growth. Now I don't know in the specific case of Tanzania. But this is what the literature tells us. And the measures, I wonder why. But one of the reasons is that Tanzania has gold. So perhaps in Tanzania there is an element of natural resources which the sources distribute. But that I don't know for sure. But the data, we will look at it now. It tends to show that Tanzania, which had a very good name in terms of distribution, now has shown during the last 10, 15 years rise. Well, thank you very much for the comments. In this type of approach, in general in this type of the composition, we analyze the statistical association between the households' characteristics and the probability of being poor. So there is no causality here. In fact, there are some in the United States. There might be also some reverse causality. So sometimes being poor is what makes you to have, for example, more children or live in certain areas, et cetera. But what is clear is that there is a high correlation between these characteristics. For example, in the case of the number of children, I think what I'm capturing here is the main effect is the fact that if you have more children, this increases the needs of the households, so the per capita income falls. And probably this is the main effect. But it could be also the effect that some households have for the fact that they are poor or maybe they have different cultural values and they prefer to have more children or maybe they have lower access to family planning and then for that reason they have more children. So I cannot distinguish between these effects. For that, I have to rely in general on the literature. So what is the literature? Because you have to make very specific analysis of each of these causes. You cannot treat them all of them in one single structural model. I mean, I don't know such a model. The problem here, and this is the difference with the white, black and white differential, there is not much literature looking at different ethnic differences in any of these aspects in South Africa, despite the fact I think they are very, very important. But basically for the economic explanation of what is going on, you have to rely on the empirical evidence that have analyzed each specific factor, the labour market, the fertility, et cetera. Regarding, for example, the specific case you mentioned of education, I mean, here the high contribution of education to explain the gap comes from two facts. One is that COSA and Sulu, they have lower levels of education. I didn't show the summary statistics, but you can see that they have lower years of education. And the second one is that the coefficient that I estimate of the association between education and poverty is large. So in South Africa, having higher education reduces the probability of being poor. If it were the other case that you mentioned that having higher education increases the probability of being poor, the contribution should be negative because equalizing education should improve this group. It should increase the poverty of this group, but it happens right the opposite. So this type of effect is captured by the coefficient. The coefficient is telling you how, because you can have a very, very, very large gap in characteristics, but if this is not important for poverty, so the coefficient is nearly zero, the contribution will be zero. Thank you. Thank you. So we have about 10 minutes before lunch, so we'll take another round of questions. Thank you, Stefan Leiderer from the German Development Institute. Just coming back to the last point, and I'm not really sure I'm yet able to formulate it into a question, but I still have a substantial worry about the exogenous or endogenous nature of these variables. Because basically if we assume all these controls in your regression are exogenous, then basically there wouldn't be a need to look at ethnic differentials because you're able to explain the differences in poverty by all these household characteristics. So if they were exogenous, then there's no need to look at ethnicity, I would say. Now if we assume they're endogenous, what do we learn? All we learn that the sulu's are poorer than the josa, but what do we find out beyond that? I think it is important whether we consider the controls as exogenous or endogenous in your regressions, and depending on either side we decide, I'm not sure what do we learn. In either case, I'm not sure whether we really need to look at the ethnic lines or whether there's other issues that are more important than... In either case, we learn nothing or in the other case, we don't need the ethnic differentiation, I think. Or I might get it wrong there, but maybe you can comment on that. Thank you. Again, addressing the last paper as well. I was wondering whether you could comment on the impact of direct poverty reduction strategies that are being implemented by government and whether that has had an effect and whether you picked that up. And then I think the point, the last speaker is making the person who raised the question. I think the case of the Su-2 and Su-1 and Su-2 is important to retain as a comparator because I think the location of the largest parts of the Su-1 and Su-2 populations would be in provincial areas that are much better and are in a whole number of economic activities, whether it's manufacturing, mining and so forth. And whether that may account for the large difference between those two groups and Coisa and the Zulus, given that the structure of the economies in the provinces they found in is very different from the others. Thank you. Any more questions or comments? Carlos, you have the pleasure of taking both questions. Thank you very much again for the question. Here, I mean, the ethnicity is relevant and if you find that one group is much more poorer than the other, I think this means it's relevant. The question is why it's relevant. It's because they have lower characteristics, so they live in the poorest areas, for example, of the country by chance or by history or whatever. It's always because even when they have the same characteristics, the returns of these characteristics in terms of poverty through the labor market, through the social programs, et cetera, are worse than the other group. So this is basically the idea. I mean, the differential is large, so if I found a small differential, I would say looking at ethnicity is irrelevant. You find that there is a huge differential because 50% of higher poverty with these high levels of poverty I think is very, very important. I think in a characteristic that shouldn't be associated with poverty because probably we will agree ethnicity should not be one of these characteristics that justifies that some people are poorer than the other. I think it's important to have a look at that. Then the question is what is the nature of this differential? And this is what the papers also trying to disentangle with the limitations, obviously, of the approach. But I think it's the first attempt to look at these differential. And so the difference can come because they have lower or poorer characteristics because of the impact of these characteristics. What I found is that they mostly can, with some exceptions, they mostly come from differences in characteristics, so it's most a compositional effect. And I think it's interesting to identify which characteristics and this is relevant for political analysis, even if you cannot extract from here because of these indo-united problems, et cetera, and the assumptions in general of this type of counterfactual analysis. I think that the information is relevant for political analysis because, for example, if you identify education is the main issue here, clearly it points out that policies reducing the gap in education will be probably more effective than policies addressing other issues. If you find that it's location what is relevant may be these claims for local community or regional policies trying to increase the development of those areas that are left behind and that have ethnic bias. So I think you can extract some general action for policies, not very specific and cannot evaluate the importance of one specific program on these differential. For that you need a different framework, a different methodology, probably different data. But I think you can learn about what is going on. For example, comparing what I found using this approach in different countries, you see that in some countries geography is not very much important, so it's not that ethnic minorities are being concentrated in the poorest or remote areas of the country. In some countries it's very, very important. Obviously education tends to be important, but most in developing countries that for example in developed countries etc. So I think you have some lessons. Obviously you don't explain the whole gap here and as I said before you have to complement this for example with analysis of the wage gap in the labour market to see if there is a premium of belonging to one ethnicity or the other. You need to compare this with the analysis of the achievement in education to see if some groups are receiving lower quality education or not etc. Regarding the specific case of the location, these groups are highly concentrated in certain areas. In fact I didn't use the province as a variable because one of the assumptions of this counterfactual analysis is the common support. So you need enough people of all groups in all areas to have this kind of analysis otherwise it makes no sense. I used the degree of urbanization, reality etc. But what is important is that the effect of this concentration specific for example in mining provinces or in provinces with more industry or more agriculture etc. I would expect that to be important through the unexplained effect because it means that you have a certain level of education living in this province you will have higher returns to education and living in this one you will have lower returns of education. But the unexplained effect is not much important to explain the gap in poverty and in persistent poverty so probably that is not the important fact. But maybe it could be to explain for example higher temporary poverty of COSA for example. I can't explain because most is unexplained some part of this can come for example for returns in education or returns to other values. There may be room that this might be but when you explain most of them some of this effect can be captured because also these provinces have higher education they have more urban areas etc. And this I capture in this. Well once again thank you very much to all of us who have been able to stay but I think we all agree that we've had three very stimulating presentations so I'd like you all to join us in thanking our three presenters.