 I would have used my roll-off share to ask the first question to Natalie, and I'm sorry I haven't read the paper, so please pardon my ignorance, but I was really interested in this issue of, I think it's an important contribution of different approach whether we are talking about chronic poverty or more those that move around the poverty lines. My question is about the separability issue, whether it truly holds that you can say for these two cases that you study that there is a sort of similar trajectory of well-being and then how that affects your final results. In the case of chronic and transient poverty, we can say that yes we can observe the similar trajectories of well-being and that holds, in the case when you are looking at the variance whether that really holds in both cases. So I'm not quite sure what you're asking holds in, but you said we observe similar, what do you mean? That's the trajectories. When you are specifying your group, right? So you show different characteristics and well you have the issue that you have the neutrality, I mean the neutrality of population, but there were other characteristics that changes and other that did not change. When you mentioned, when you specified your well-being indicator, if I understood well when you went through the foster or the separability issues, it said that there is this, do you mean, the properties where I said constant, for constant well-being? Yes. Ah, so obviously I was going through quite quickly and maybe I didn't explain clearly enough exactly what I was doing. I hope in the paper I explained that a bit more clearly. Well, I can't promise. So I'm not at all saying that I believe that my data looks like this. So at this stage, this is a stage in helping me to construct a measure with properties that make sense. So I do like a thought experiment. I say, suppose everybody in my sample was having constant well-being. There's no fluctuations. In that case, what would I want my poverty measure to look like? Now, by going through that thought experiment, I can't fully pin down my poverty measures, but I can restrict them quite a lot and pin them down quite a lot. So having kind of gone through that, I then come up with measures that make sense in the case where everyone's constant. And I say, if everyone's constant, then everything must be chronic. And then I can start to think about spreading out from there and what if I have fluctuations. Now, I still want to stay consistent with what I've done so far, but I have to look further. Does that explain a little bit clearly? It makes sense, because also when you look beyond at the moral level, the issues of vulnerability and vulnerability to poverty, these are the things I do. Like how chocks may change this whole setup. Absolutely. So I'm not imposing that assumption on the data at all. I'm letting the data tell me what is going on in the data. And I'm saying if I, like I'm just using that as a thought experiment along the way to constructing a nice measure. Perfect. Thank you so much. Okay. So I give the floor to two questions. Miguel. Can I have three questions? Well, to the three. Yes. Okay. One for each. Four. Yeah, four. Yes. Thanks for Natalie. Can you explain why, what is it that makes your method different? I mean, in terms of, yeah, so many different methods of chronic poverty. So why should, why do we need another one? So one. Very good question. Yeah. So the second one is to Tomoki. So when you present the residuals, of course, and you say when the residuals are too large, then the decomposability of your method is not efficient. So it's, so at what level of the residuals, we can say that the method may not be necessarily adequate to decompose poverty. And then for Bill, I think it's a very interesting point that you're making. And what do we need to do as a development economist? So it's my question. I should gather more questions and then we'll come back to them. I would like to ask from Tomoki, you know, how did you, you selected in decomposing some variables. So how did you come up selecting those? Or are they something you, you had a large number or, or, and then you just came down with some, you know, explaining in the best way. So, and to Bill, I would like to ask, you spoke about nutrition and how it should be included in millennium development goals. I wonder, isn't there any, any indicator now among those at present? And are you already doing something to include one additional one just to make a difference? Or what you explained, it's massive if it's based on all these studies, et cetera. Okay. Any other questions? I am Yong Fu-Huang from UNVIDER. I have three questions. And first question goes to the first speaker. And basically, I think I find your framework is very attractive. We decompose the poverty into chronic poverty and transient poverty. And I wonder how you, can you talk about how you differentiate between chronic poverty and transient poverty, especially the transient poverty, because this is highly related to economic fluctuation, economic shocks, because the poor are more vulnerable to external shocks. But you define the other transient poverty. And also, could you also give some details on how you, you solve the maximization problem in your paper? You probably using some kinds of dynamic programming or some kinds of, could you give some details? Because I'd like to know more about that. My second question goes to the second speaker. And in your paper, you're talking about the poverty decomposition, by using regression approach, especially the integration. You basically look at the linear regression. I wonder if you have ever considered the long linear regression case. Long linear will be different story, I suppose. Okay. My third question goes to the third speaker. Okay. And I find your, your finding is very interesting. And especially when you talk about, you mentioned that the worsening nutrition outcome, right, are related to some kinds of social and economic factors, right? I wonder if you could discuss the implications of funding for the post-2015 development agenda. Thank you. Okay. I will give the opportunity to the, to the speakers to respond and then later we can go to another round of questions. So the first question I was asked was basically, well, what does my method add to the several methods that have been proposed in the literature? So my argument is that none of these many methods that have been proposed actually combine a set of sensible properties for chronic poverty measurement. There are several, several of the measures that have been proposed are kind of nice, well-behaved measures. And if we think about what their properties are, then intuitively that makes an awful lot of sense for measuring something like the total burden of poverty, but in a way that's not sensitive to chronicity or persistence as such. In fact, that goes in the other direction. And I've actually done some work with Catherine Porter and it's still at working paper stage, but hopefully coming out soon where we actually kind of look into this that actually if you, you know, we're interested in vulnerability and depth and so on. And it's actually impossible to combine that in a measure with, with kind of picking up cronicity. So there are several measures which kind of nice total poverty measures which just don't, they kind of, they go in the opposite direction from what you'd expect for chronic poverty measure under certain transformations. Now there are other measures which have been mainly proposed just in the last couple of years or kind of published in the last couple of years where this has been recognised and the functional form has been constructed very much in order to pick up this persistence to kind of like sensitivity to persistence or duration or cont, contiguity. The problem with those measures is that the way that those functional forms have been built have had the nasty side effect of actually some pretty dodgy discontinuities which when you then start thinking about, okay, how does this measure order different alternative trajectories of well-being? You get really weird things going on that just aren't consistent. So like there's a, it's not that it's impossible to do this, it's just that none of the functions that have been proposed so far kind of do it without unintended consequences. So I'm kind of arguing, I want to combine the best from these measures and the best from these measures and actually come up with a measure that combines and makes sense as a chronic poverty measure. Sorry, that was a bit long-winded. Hopefully a quicker response to the next question I was asked. So I think you're interested in transient poverty. We're thinking about people who are vulnerable to shocks and so on and so forth. So I kind of start with hopefully a fairly sensible measure of total poverty or total burden of poverty. I didn't go into that in any detail but for this paper I follow Jelana Mavallian. They're just simply adding up foster degree of Thorbecker poverty gap squared over time. Seems a reasonable enough way to say, well that's the total burden of poverty. So that's total poverty. And then I come up with my chronic poverty measure which is perfectly consistent with that total poverty measure if, this is going back to your point, if everybody's experiencing constant levels of well-being. And then I say, whereas the transient poverty, that's the difference between the two. So in this decomposition, the transient poverty is kind of the residual when I compare what's hopefully a sensible measure for total poverty and a sensible measure for chronic poverty, it makes sense to see transient poverty as the residual. And if you know this data set, you'll know there's a lot of shocks, there are lots of vulnerability. And it makes a lot of sense that we saw what 60, 80% of the poverty in most of those villages actually being categorized as transient poverty. And very, very quickly, your second question, you asked about an optimization problem. There is no optimization problem in my paper. It's a normative, like, so I'm not doing any positive modeling here. I'm not thinking about behavior responses to anything. It's simply a kind of a welfare evaluation, poverty measurement kind of thing. So no optimization problem. Okay, so let me first address Miguel's question. So what would be considered adequate decomposition? So, well, first let me explain to you how you can interpret the residual term. So the residual term is the part of poverty that cannot be explained by covariates. So when I say that that component is large, poverty can be explained by something that cannot be explained by the model that you have. So if that's too large, well, poverty changes due to something that we don't observe pretty much, which is not particularly informative. How big should it be for the decomposition method to be useless? Well, I guess, you know, if that component explains, say, 70, 80% of the total change, I would say generally it's completely useless. If it's less, you know, depends on how other factors that you are looking at are explaining, you know, if it's still, you know, the systematic part explains the substantial part, especially those variables you're really interested in, explain the substantial part of poverty change, I would say that it may be still informative. I wouldn't dare to, you know, make a general statement, but I guess the judgment has to be done on a case-by-case basis, but that's how I think about how to judge the usefulness of this decomposition method. The second question about the selection of the model, so in a way, this is completely arbitrary, you know, you can include whatever, you know, covariates you're interested in, but I try to include the covariates that's reasonably sensible and that varies over time. If, you know, you know that, you know, if it doesn't change over time, you know by definition, x component is going to be zero and at least for the purpose of illustration it's not particularly interesting. I also don't want to have, you know, variable that doesn't have much explanatory power, then it means that, you know, the estimated coefficient is quite unstable. So I, you know, take into consideration these factors, but there's no single recipe for model selection. At least I haven't found one. Yeah, sure, sure. They have done lots of work in order to find good indicators in the country. In different regions, you know, selecting poverty indicators, they call it kukuta. So I think it's interesting to read a book and there are lots and lots of variables, you know, how equally in participatory process they have determined, you know, how they see what are the most important factors. It would be interesting for you just to look through how your own variables now... Is it in the national panel survey data? Yeah, you have been using... I don't know how they are now measuring it. Is there any Tanzania here? So my method is constrained by the fact that I have to use panel data, but I appreciate your input, yeah. There could be possibilities. You were looking for also some other data. There could be. Yeah, yeah, yeah. It's interesting at least me, you know, how to find the best possible measures to see reduction of poverty in some cases. Yeah. I discuss with you the region. Yeah, yeah, thank you. And the third one from Yong Fu about this non-linear model, right? So actually there is an application that I didn't really discuss properly, but when I use P0, the poverty rate or head count index, and it's discrete and it doesn't really apply very well. So instead I first run poverty regression and then I use that to decompose poverty. So there I have a non-linear model instead of having y is equal to, you know, x. Although, you know, I have a single index model. So this model can be easily extended, although in the paper I didn't really discuss non-linearity much, but this model can be easily extended to the non-linear case, provided that it's, you know, it's a differentiable function, basically, yeah. Thank you. Yeah, I think my, probably my main point here is one is to urge people to exercise caution in selecting variables for their models. I've looked at a couple of the poster papers that attempt to use, for example, to estimate the effect of cash transfer programs on nutritional status. And I think the weakness of the results is very much explained by the phenomenon that I've been discussing this afternoon. The outcomes that you may get a partial catch up, but you will not get a complete one. So the conclusion is foregone. I think there's a great danger. And we as economists tend to think a little bit about nutrition models like production functions. You've got consumption leads to an output, and that output is good nutrition. And that simply is not true. The ability of the body to process and utilize food is not so straightforward. The machinery, the technology is flawed. I think it also has some interesting ramifications, perhaps in some human capital approaches as well. But I'm not going in that direction. So I would urge people to be very cautious about their selection of nutrition-related variables, particularly for children. Adults suffer from this same condition. But they've passed the critical growth stage. So the adult indicators are perhaps more reliable. One thing I don't know if I didn't mention it, I don't think I did. There's no cure for this condition, but if you remove people from this environment, they recover it naturally. There's a famous medical study done of Peace Corps volunteers in Pakistan living in rural areas, and they all came down with environmental interopathy. When they went back to the United States, they, with no medical treatment whatsoever, they all slowly recovered. I also mentioned it's largely a rural phenomenon in urban areas with access to hygiene and sewage and what have you. It doesn't occur to any great extent. I would not urge this kind of nutritional indicator to be included in Millennium Development Goals. Why? Because there's enormous pressure being placed on low-income countries to monitor the indicators of Millennium Development Goals. And I think this would be a huge wasted expenditure to attempt to do this, except in a very controlled, very precise manner. I just think it would be a huge waste of funding. And it, as I say, it doesn't indicate anything about economic or social welfare. It simply tells you something about nutritional status. So if you identify poor chronic under nutrition, you still have a question, what the heck do you do about it? Any other question or comments? Yes, I think it's been a very rich discussion and particularly the last reflection about exercise caution in selective variables. I don't think it applies only to the health-related indicators but to other indicators as well. And I think it's a lesson. And I would love to see this panel and how we can use it for studying other issues in Africa. I have used the health surveys myself in the absence of other households or survey daytime in other countries. And I think they have a very good technique and that could be a good comparator between countries. So I really appreciate your effort to see the paper and the final panel. Hopefully it can be used. And going back to the point of measurements and indicators, that was a huge discussion when framing the new post-2015 development agenda because the question is, okay, we spent 10-plus years plus resources in donors, countries themselves trying to measure what are targets towards poverty and eradication. But what about the policy environment? What about the policy design? What is about the framework? And now we are moving from targets, indicators to what we call paradigms. And again, when we look at issues like chronic poverty, we're talking about, okay, poverty rate is reducing but what about what is going under that line that we draw. So I think this has been a very rich session from the technical point of view but also from the messages that emanate. And as I said, I look forward to reading the papers. If you have any questions, you have the email so you can contact the authors directly. So please join me in thanking the authors.