 A joint work by my colleague Sabina Alkaya and Adriana Konkoni. So, this year, I mean, every year we try to do analysis based on the multidimensional poverty index that OFI has developed based on the methodology proposed by Professor James Foster and Sabina Alkaya. And this year, our objective was to explore the distributional issues among the multidimensionally poor. And this is one way of looking at those multidimensionally poor who are facing much more severe form of deprivations. And to explore where we stand at this moment based on our empirical application on nearly 50 developing countries. So, indeed, understanding different degrees and kinds of poverty helps in their removal. It is quite well known that if we just use headcount ratios, then we can give deliberate incentive for the policymakers to ignore the situation of those that are poorest because you get maximum gain just by reducing the number of poor. And that can happen just by reducing the number of marginally poor even ignoring the poorest of the poor. It has also been argued in the literature that poorest of the poor may be catastrophically quite different and may require different kind of policies and different kind of assistance. There are literature proposed by Michael Lipton, Devereux, Barbara Harris White. Also, deprivations among the poorest may reflect more chronic form of deprivation. It has also been argued in the literature. Recent debates and goals. Of course, World Bank has mentioned that the aim is to eliminate poverty. The interest element reduced drastically, $1.2 per day poverty by 2030. There has also been the discussion of understanding shared prosperity, looking at the growth of the bottom 40% of the population. At the same time, the high level panel on post-2015 development agenda has mentioned that the poorest did not, we did not, or MDGs did not focus enough on the poorest. So amidst all these goals and objectives, certain concerns remain. Of course, the first one is does reducing $1.25 a day automatically reduces deprivations in other dimensions? Maybe not. And there are empirical findings. It has not been the case. And $1.25 a day actually is included in the first goal of MDGs. Is it sufficient to look at deprivations in different dimensions separately? We have MDGs, looking at them separately. Does it make sense? Or does it need to look at them jointly? What method is appropriate that respects the ordinal nature of dimensions? Some dimensions that we have, they are cardinal in nature, like income. But there are dimensions and indicators that are more ordinal in nature, where you only know the orders, the categories. But you have no idea what is the difference between these categories are. You may have, suppose, categories of sanitation facilities. You have piped water to include it in the house. You have, sorry, for the sanitation, you have flush toilets. You have open defecations and other categories. What are the distance between categories? Is it the same as we assume in some other measures? Maybe not. And then finally, does the overall improvement ensure improvement among the situation of the poorest? So all these things are valid concerns. And into this presentation, I will mostly be focusing on the final part, of course, based on the other three concerns. So our approach is going to be a multi-dimensional approach, which will take into account joint distribution of deprivations. They will try to look at the deprivations together rather than separately. We will apply accounting approach, which means that it respects the ordinal nature of dimensions. And then we will try to see if we can understand how we can understand destitution. So the main concerns for this paper is how do we legitimately use ordinal information to identify the destitute? And of course, our approach is based on Sabin Al-Qaeda and James Foster's accounting approach on multi-dimensional poverty measurement. And the distributional concern is how has poverty reduced among destitutes in comparison to the multi-dimensionally poor? How are the poorest of the poor referred in the literature? Various terms has been used, ultra-poor, destitute, extreme poor. No agreement or hierarchy on this hierarchy has meant. We choose the term destitute because it has been used more as a multi-dimensional concept, whereas ultra-poor, as you will see here, has been used more in terms of the monetary poverty literature. Lipton, for example, in 1983, identified the destitute based on those eating below 80% of dietary energy requirement and spending 80% or more of total income on food. Similar kind of approaches has been followed by Kakwani and Ellis. Other monetary approaches, Kornia studied ultra-poverty in Eastern European countries during transition, Stephen Klassen and others have studied ultra-poverty in South Africa. If pre-tried to identify the destutes as those earning 0.5 or 0.5 cents a day or lower and so on. There has also been some applications by the non-governmental organizations like BRAC in Bangladesh, another NGO, Burden in a district of Murshidabad in West Bengal in India. They are, however, used, along with income, some other multiple criteria to identify the ultra-poor. So of course it looks like there has been interest in understanding the situation of those who are the poorest rather than just being poor. But here, in ultra-poverty literature, the focus has remained mostly in terms of income, at least in academia. Literature identification of destitute, although, as I said, it has not been categorized systematically. Devereux proposed identifying destitute using inability to meet subsistence need, assetlessness and dependence on transfers. Devereux is, I find, the only person who actually acknowledges that the destitution, and also Barbara Harris White, that destitution should be understood as a multi-dimensional concept rather than just in terms of monetary dimension. Elise, however, took an interesting approach, identifies destitute as those who are ultra-poor and have a labor dependency ratio of three or more. So there is a clear hierarchy for the first time in this particular paper where destitute is represented as a subset of ultra-poor. In our paper, however, we use the counting approach framework to identify the destitute using a multi-dimensional concept. Just to present very briefly the counting approach as developed by James Foster and Seben Alkaya. So we have a matrix, and we denote the achievement of a person in particular dimension by x, i, j. So we have i persons, and we have j dimensions. Each dimension has a deprivation cutoff based on which a person is identified as deprived or not in that particular dimension. And a deprivation status value, which we denote by g, i, j, is assigned. If the person is deprived, a value of one is assigned. If a person is not deprived, a value of zero is assigned. And so this is the deprivation matrix. This is called the deprivation matrix, which contains values of zero and once. Deprived means zero, one, non-deprived means zero. Once a relative weight, I'm not going to discuss very broadly how the weights are determined. That's another discussion, another session that can be spent. But some of the weights has been determined. Once the weight has been determined for each person, a deprivation score is obtained, and the person is identified as poor if the deprivation score is larger than a cutoff, than the poverty cutoff k. And we denote the set of poor by capital Z. So the identification here is denoted by rho, which depends on the achievement vector of a person. Deprivation cutoff, poverty cutoff, and weight. This identification function gives a value of one if the person is included in the set of poor. That means monthly dimensionally poor. Otherwise, the identification function is equal to zero. So we have three sets of parameters here generally. One is the deprivation cutoff vector, which is denoted by Z. The poverty cutoff, this dual cutoff, and the weight vector. Now, how to identify destitute, which are going to be a subset of the poor? So the way we consider here, the destitute, or those who are the poorest, has to be a subset of those who are identified as multi-dimensionally poor, which is denoted by Z. So suppose we use different deprivation cutoff, a deeper deprivation cutoff, a more stringent poverty cutoff, or another weight vector to identify those destitudes. Now, it turns out that if we follow a non-union approach, that means if we set an assumption that somebody has to be identified as multi-dimensionally poor, who is deprived in at least two indicators. That's when you are deprived multi-dimensionally. Then it turns out the weight vector must remain same, the deprivation cutoff vector has to be deeper, and the poverty cutoff has to be higher. That is the condition to identify a subset of multi-dimensionally poor. Now, within that framework, there could be different approaches. One approach that we refer as the intensity approach. What is this approach? This approach looks at the extent of multiple deprivations. Suppose you say if somebody is identified in three out of ten indicators, we identify them as multi-dimensionally poor, but somebody who is deprived in five or more of indicators, then we identify that person as much poorer. You look at the intensity of multiple deprivations to identify those that are poorest. This actually has been applied in Human Development Report to construct the proportion of severely poor. This is how they term it. The other one is actually depth approach, which we try to explore in this paper and show an empirical example using that. When we have ordinal variables, it prevents us to compute the gap or squared gap that has been used to construct the well-known Foster-Gear-Thorwek method, because we do not know the distance. Even if we try to compute that, we do not know how accurate it is. So what would we do? One option that we could do here, we use a different set of deprivation cutoff. Instead of taking the gap, we identify the deprivations with respect to deeper deprivation cutoff. That's why we call it a depth approach. And then similarly, the rest of the approach follows similarly as the identification of the MPI poor. We identify the deprivations, we take the weighted sum, compute the deprivation score of those people, and then set a poverty cutoff to identify those that are much poorer. So this is the depth approach. And this is what we will be using in this particular paper. And the other approach could be a mixed approach, because you may argue, well, some may argue that multiple deprivations is more important rather than if you are deprived in smaller number of dimensions but you are severely deprived. We do not know which one is more severe or not. So one option could be to identify the severely poor and poor by depth approach, and then take an intersection of that. This is an approach we try to take in one of our paper, joint paper on India. So let me show you an example how we have implemented it. So just to give you a brief slide on what is multidimensional poverty index, it was developed by Sabin Alkar and Maria Masantos based on the methodology proposed by James Foster and Sabin Alkar. And it had three dimensions, it consists of three dimensions and ten indicators. And somebody's identified as multi-dimensionally poor if they're deprived in 33% or one-third of the weighted indicators. That's k equal to one-third in this case. And these are the indicators and the deprivation cut-offs that are used to identify the deprivations and then the multidimensionally poor. So what we do here in the depth approach, as we cannot take into account the gaps and square gap, we set deeper deprivation cut-offs for each indicator where possible. So if you see here, sorry, if you see here for schooling, we had earlier, for the original MPI case, no household member has completed five years of schooling. We used no one completed at least one year of schooling. For attendance, any school-aged child in the household not attending schools up to class eight, yet no child attending school up to the age which they should finish class six. We try to link where possible these deprivation cut-offs to world health organizations cut-off of severe malnutrition or MDGs as much as possible. As you can see for two indicators, electricity and floor, we could not take deeper deprivation cut-off because we only have information on access, yes or no. So it did not allow us to say deeper deprivation cut-off for these two indicators. So indicators are same. Deprivation cut-offs are deeper. Weights are same. We also keep the poverty cut-offs same. That means you have to be identified, destitutes are identified if they're deprived in one third of the deeper, weighted deeper deprivation cut-off. So 49 countries that we cover, they're mostly demographic health survey or multiple indicator cluster service. Perfect, thank you. So we include populist countries such as India, Indonesia, Pakistan, Nigeria and Bangladesh. These 49 countries contain 1.2 billion MPI pool and our expectation was, based on the severe deprivation cut-offs, there will be a bigger number of destitutes in those countries. However, we find half of the 1.2 billion MPI pool people as destitute. So of course we had the methodology, we had the selection, we did not select the indicators based on how many destitutes we are expecting. We selected the deprivation cut-off independent of that. And then we computed the results and still it was showing nearly half of those that identified as MPI pool in those countries were destitute. And of course of those destitutes, this is not a surprise, probably 97.3% live in sub-Saharan Africa and South Asia and over half of them live in India, but India data is old, one may argue. And what kind of progress has taken place in India, we cannot say until the data that we required to compute MPI is out. So they had a lot of claims since 2004 that poverty has gone down, but every time the poverty estimates are revised, we get a higher and higher number of poor. So in 2004-5 initially they were claimed to be 28% of the population as income poor. Then it has been revised in 2009, increased to 37%. And now they have been revised and so historical poverty at now is nearly 45 to 48% instead of 28%. So they're approaching the number we got in 2005-6 for MPI pool. Now suppose we denote the proportion of population were destitute as h bar and the proportion of population destitute and deprived in an indicator by this notation of small h bar. Then this particular thing which is here, h bar k bar over h gives us the proportion of destitute deprived in each indicator. So this gives us an idea in which indicators the destitudes are deprived in. So 46% don't have anyone in their home. Of course here you can see 67% have someone at home at least with severe malnutrition. 71% don't have electricity. 90% practice open defecation and so on. It's quite high. So this gives you some idea where the source of destitution, where they're coming from. We look at, we plot for countries where we could compare the $1.25 a day poverty and the destitution rate and although there appears to be a positive relationship, it's quite volatile. We cannot say it's a strong positive relationship. This graph plots the percentage of population destitute versus the percentage of population MPI poor. Again, appears to be a positive relationship. But little more volatility is added when we present the percentage of MPI poor who are destitute as against percentage of population who are MPI poor. These are the next two slides. I'm skipping quickly. We look at the sub-national regions and we come to similar kind of results. Before we conclude, and this is what we are trying to study at this moment. How we can use these concepts to find poverty inter-temporarily. Here you have three examples of Malawi, Ethiopia, Pakistan. We could not find a better example. Ideally, we should have examples where we have similar level of poverty and see similar reduction and so on. We did not find that. Initial poverty, they were different. However, the reduction in the proportion of multi-dimensionally poor percentage points were same or very similar across these three countries. So here, for per year we have a reduction of 0.9 percentage point in Ethiopia, 0.8 percentage point per year and for Pakistan 0.7 percentage point per year. So, similar reduction in multi-dimensional poverty. However, if we look inside and break it down, the reduction in the change in the proportion of multi-dimensionally poor into change in the proportion of destitutes and proportion of non-destitutes, it unfolds the story. Here, if you look at here, there has been reduction among destitutes and some destitutes probably graduated to being non-destitute. As a result, there has been a slight increase in the share of non-destitutes. But overall poverty came down. For Pakistan, there has been a reduction in the number of non-destitutes and as well as reduction, it's very modest reduction. But, sorry, for the destitutes actually there is almost no reduction. So the reduction of poverty actually came by reducing the number of or share of non-destitutes and not the destitutes. Unlike in Ethiopia which also had very modest reduction in the proportion of MPI poor, however a large reduction came by reducing the or improving the condition of destitutes. As a result, many destitutes graduated to being non-destitutes. As a result, their proportion increased. So although they had similar reduction in overall multi-dimensional poverty it was quite different when you break it down the situation of destitutes and non-destitutes. So go to the conclusion without further delay. The counting approach based on the counting approach framework which has not been studied. Most of the study, you know, understanding the situation of the poorest of the poor has been based on the income approach. There are very few studies that looked at multi-dimensional destitution. I mean, they operationalized the concept of destitution but they stopped short of measuring it. We tried to go ahead and use this particular counting approach to identify the destitutes and try to analyze their situation globally. We are working on country-specific example to explore more how it can be applied on specific countries. And our application still shows that at least based on the data, our data sets ranged between 2006 and 2013 and most of the data sets were on or after 2010. So we cannot accept India's, of course, some big chunk of the population. Still, if you look at that particular year, the share of destitutes as a share of MPA poor is still quite high. It's soberingly high. So it requires focusing on the poorest of the poor as the high-level panel pointed out that poorest of the poor has not been benefited. Although we have had a reduction in extreme poverty in terms of income, other dimensions and indicators, they remained still quite bad. I don't have a picture which we try to put in our upcoming book by Oxford University Press on Multidimensional Poverty which clearly shows that many countries have met or very close to meeting the goal of extreme poverty reduction. But when you look beside in that graph other indicators, child mortality, maternal health, sanitation, most of the countries, they are not even close to meeting it. So this reduction in income poverty has not met or has not actually translated to reduction or improvement in other indicators. So it requires to understand the joint distribution, the joint deprivations and understanding the poorest of the poor and then shape the policy accordingly. Thank you so much.