 Excellencies, distinguished guests, friends, it is very much an honor to be here and I'm grateful to Wider for this opportunity to fin for his kind words. And Wider also plays a very special role in my mind when the time came to found OFI, our research center. This was the place I came for advice to try to understand in a sense how research centers can continue to have a broad intellectual horizon, continue to do research to high standard, and yet also engage with the policy world and with those most directly engaged in the issues of development. And I've been very grateful to learn, although not perhaps very well to emulate some of these features of Wider, I admire so much. There's a lovely line in a ditty by Edgar Leslie that goes, Taint no sin to take off your skin and dance about in your bones. Taint no sin to take off your skin and dance about in your bones. Now, obviously, this is stuff and nonsense. We can hardly throw off our skin and muscle and fat and hair and dance around as grinning skeletons, as lovely and invigorating as it might be to do so. But just the thought of doing it, of being able to dance as naked bones, uninhibited and unencumbered feels intriguing, liberating, and somehow clarifying. I mentioned that because as unlikely as it may see, I hope we can take some of that freewheeling creative and curious sensation into the next hour, when we will talk about a rather serious subject, which is how people are acutely poor. Ending acute poverty is of ethical importance and human urgency. It is a collective priority within and across nations, and most of all, for the protagonists of poverty. And it is so because, despite fantastic gains, far too many continue to go about their lives in abject, dangerous, and harsh conditions, unable to shape lives with meaning and dignity. But in the next hour, I'd like to step back from the many complexities and isolate one aspect only, which is poverty measurement. The question is whether we are able to use multidimensional poverty measures to expose a relevant, if limited, skeletal structure of poverty in such a way that makes the task of addressing poverty somewhat clearer for all concerned. My own thought on this is not fixed. Our commitment, all of us, must be on the desired state of no poverty, and not on any particular skeletal arrangement. But for the next hour, I'll present a structure of poverty measurement that may, if well executed, be of some use. And I do so knowing that this work is not my own. The methodology was developed with James Foster, as Finn said, a leading, precise, and powerful mind across the field of poverty and inequality measurement. And the applications were all undertaken with students and co-workers in our research center, OFI, whose brilliance and care and steely determination are wonderful to work with. But let us begin as is natural for us here at wider with the Martia Sen's capability approach. And I'd actually like to start with a paragraph from the 1990 book edited with Martha Nussbaum, in which Sen had an article called Capability and Well-Bean. He wrote, turning to poverty analysis, identifying a minimum combination of basic capabilities can be a good way of setting up the problem of diagnosing and measuring poverty. It can lead to results quite different from those obtained by concentrating on inadequacy of income as the criterion of identifying the poor. The conversion of income into basic capabilities can go with varying levels of minimally adequate income. The income centered view on poverty, he argues, may be misleading in some respects. That paragraph now seems prescient for its claim as to the persistent relevance of paying attention to non-monetary dimensions of poverty stands empirically validated by extensive micro-dated analysis, which was not available in 1990, by participatory insights and voices of the poor and other such studies, longitudinal and ethnographic studies, social movements, and so on. This conceptual articulation by Martia Sen recognized implicitly the importance of numbers and measures of poverty, even crude measures, such as Human Development Index, as a guide to action. And more recently, of course, the statistics have gained public prominence in the Sen Siglitz Fatusi Commission and its successors in the data revolution and in many, many different fora on statistical capacity building and strengthening statistics. Miguel's Tsecoli's words well encapsulate the motivation for improving statistics and measurement because of their link to action. A number, he writes in Spanish, can awaken consciences. It can mobilize the reluctant. It can ignite action, generate debate. It can even, in the best of circumstances, end a pressing problem. And of course, that is our topic today. But until recently, as we know, I think Sen's insights and other work on multi-dimensional poverty was in a sense marginal. With the advent of the Sustainable Development Goals, multi-dimensional poverty appears to be graduating from the margins to take a more central place, but it's not yet certain how that space will evolve. But clearly, the development now is framed multi-dimensionally with the SDGs being an integrated and indivisible balance of three dimensions, economic, social, and environmental. And turning to poverty, the second sentence of the pivotal document of the SDGs transforming our world, the 2030 Agenda for Sustainable Development, reads, we recognize that eradicating poverty in all its forms and dimensions is the greatest global challenge and an indispensable requirement for sustainable development. Here, poverty is definitively recognized to have multiple forms and dimensions, and that those include extreme poverty, the measure of $1.90 a day, as one particularly high-profile component of a wider concept of poverty. And the shift of emphasis is sustained throughout that document, with the phrase many forms and dimensions of poverty recurring seven times. And the goal one has a number of targets of which just the first target focuses on income poverty. Complimenting it, target 1.2 is to reduce at least by half the proportion of men, women, and children of all ages living in poverty in all its dimensions, according to national definitions. So the discussions leading up to the SDGs also highlighted the importance of multi-dimensional measures, multi-dimensional understandings. So for example, in the UN Secretary's General's Report in December 2014, trying to set a positive and ambitious tone for the SDGs, he argued that poverty measures should reflect the multi-dimensional nature of poverty. And in the same December 2014, the UN General Assembly spoke of the need to develop complementary measurements that better reflect multi-dimensionality of poverty and of well-being. And turning to financing Addis Ababa Accord in 2015 also called on the United Nations and international financial institutions to develop measures of progress that recognize this multi-dimensionality. So the space is there, the concept from Amartya Sen is there, the recognition that numbers guide action. But what in a sense and how should that space for better measures be filled? So we might turn naturally to the indicators of the Sustainable Development Goals and these were approved in July of this year, the list of 232 distinctive indicators that were proposed by the UN Statistics Commission. And their proposal clearly echoed what has become a pivotal phrase within the SDGs of the need to leave no one behind, have measures that can be disaggregated, to break down the silos and have policies that are integrated, multi-sectoral and indivisible and to recognize the priority of eradicating poverty in all its forms and dimensions. But the list of indicators of the SDGs leaves the precise definition of multi-dimensional poverty measures open and we'll see that later. So at this point I just wanted to pause and say that we are here celebrating UN Day in Finland, but there are also other agendas that intersect and like the SDG agenda have created a space for wider measures of poverty. But these agendas might be regional or they might be national. And so if we think of the 2063 Africa Agenda we could also find space. And so there seems to be a turning of the tide and an interest in better measures and structures of poverty that also pay attention to the non-monetary aspects. So what would multi-dimensional measures do? Clearly they would look across different dimensions perhaps including perhaps supplementing monetary measures. And they would also somehow reflect the priority that we see of poverty reduction. So choosing different SDG indicators that reflect poverty and so merit the distinctive priority given to its alleviation. There's a big focus also on recognizing the interlinkages across indicators. And a multi-dimensional measure which draws on different indicators for the same household at the same time organically does this by its very structure. And clearly a measure would need to be disaggregated in order to assess over time if groups were being left behind. But I think beyond all of these more measurement focused desiderata there's also a very strong need for a measure to be able to engage very practically with policies and will come a little bit later to those policies of resource allocation, targeting, policy design, coordination, monitoring, evaluation. So if we think of what measures should fill this space there is a very obvious document to turn to. And that is the report monitoring global poverty led by the commission established by the World Bank and chaired by our late and beloved Sir Tony Atkinson. The opening lines of his report recognize the controversiality of monitoring global poverty however it is defined. He recognizes that some people will consider the exercise to be futile. It is simply too demanding to find measures of poverty that are relevant globally or that can be measured with adequate precision for the task. He however states that this is not a view he shares and not one that underlies the report that was released. And so the aim of the report was to say how can global poverty be measured. And clearly the focus therefore is not on precisely identifying the number of people, the level of poverty, but of accurately measuring its trend over time. And the second gloss is that that trend must be measured not just as a precise point estimate, but with standard errors, with a recognition not only of sampling errors, but of also the many other kinds of non-sampling errors that affect the data used to make these estimations. And so it's important to be able to both measure and convey publicly the margins of error. So the commission was concerned only with monitoring the extent of global poverty and its reduction. And it had three parts, monitoring extreme poverty, which was the monetary sections, and then beyond one goal 1.1 complementary indicators and multi-dimensionality and making it happen. And so the resource document, if you don't know it, that I would warmly commend to you is the second part of this report, which is looking at non-monetary indicators that the World Bank might use to monitor global poverty and what characteristics they should have by what principles they should be designed and precisely what dimensions and methods should be used to develop them. So, Sir Tony Atkinson of course was one of the progenitors of discussions about social indicators in Europe and recognized the need for a broad engagement in terms of the process of indicator design and the transparency by which the indicators that finally are launched were justified. And so in developing complementary measures to $1.90 a day global poverty measure, he advised a number of different principles. One was that a non-monetary measure or set of measures should be truly global, covering not only developing countries, but also high-income countries or vice versa. He specified later how that could be done, perhaps in ways that accord with the aspirations of different country blocks by using the same indicator or dimension but with different cutoffs. Also, the indicator itself had to be transparent. It had to identify the essence of a problem so we could understand if increases in the indicator were good or bad. And the definition had to be generally valid and have a clear normative interpretation about the changes over time and also robust and statistically validated so that there is a literature about its definition and construction. And also the global indicators needed to somehow be in accord with national data and national measures. And so there needs to be and we'll say more about this a conversation between the global and the national measures which is perhaps more active than has been in the past. Furthermore, if a dashboard of different indicators is to be used, the dashboard should somehow be balanced so that the component indicators, roughly equal in importance, would span the issues of interest. And finally, in order not to tax the data producers too much, he recommended insofar as possible to design complementary measures that use existing data or where they must be extended to do so using existing instruments or administrative data. So drawing on this process, there were two recommendations that I think must inform any efforts to think about the structure of multi-dimensional poverty measures. The first was recommendation 18 that the bank should establish its own requirements about measuring a small set, a parsimonious set of complementary indicators that would include an overlapping measure will come to next and make sure that these were fully represented in the activities of the international statistic system. So where the sustainable development goals have a great breadth with 232 indicators, this report also advocated parsimony to have some high visibility indicators that would get special attention. Indeed, the report gives a draft proposal of what the dimensions should be of complementary indicators, nutrition, health status, education, housing conditions, access to work, and personal security, for example, from violence. So that is already a step forward, a step which draws on a huge amount of literature, ranging from studies of the dimensions of poverty, from voices of the poor to studies of the dimensions of quality of life of the sans-diglitz and Fatouci report to many other data exercises across Europe about the appropriate social indicators. But the report continued to then think about what a dashboard with these six component indicators would be missing. And what it was missing was the joint distribution of deprivations, that is, being able to look across and see who is deprived in several dimensions at the same time. Do those with low levels of education also suffer from poor health? Do those who are at risk of violence also have a lack of access to work? And so this need for an understanding of the interrelationships, the interlinkages across dimensions, motivated the recommendation to have a measure of overlapping deprivations. And for those in Europe, this will be very familiar that if you think of the Venn diagram of three different deprivations, then you can see who is deprived only in one, the outer clear circle, who is deprived in two of the three indicators at the same time, and who is deprived in all three. And that reflects a counting approach where you're counting the number of deprivations that a person has. It's very simple, very intuitive, but it conveys information that a dashboard of six isolated indicators cannot convey. So the final recommendation of the report was that a multi-dimension poverty indicator based on the counting approach should also be developed and reported. The report specified it was not proposed that the indicator include monetary poverty dimension. There are a number of national multi-dimensional poverty measures that are official statistics will come to later, among them Chile, Costa Rica, Honduras, Pakistan, Mozambique, and so on, that do not include the income dimension. And these are the model in a sense of this recommendation and not, for example, Mexico, which does include income at 50% of the weight. Furthermore, in terms of methodology, the recommendation admisaged the counting approach being implemented in terms of the adjusted head count ratio, which is the methodology that extends the Foster-Greer-Thurbeck index into multi-dimensional space that we have had the privilege to develop and try to work on its use a little bit. And it has two components, the percentage of people who are poor or the incidence of poverty, and the average share of deprivation, the average percentage of possible deprivations that poor people experience, which is the intensity of poverty. So it's a kind of a poverty gap measure for those of you who know that in multi-dimensional space. And the product of those two is the adjusted head count ratio. So now what I would like to do is spend about 10 minutes and give you a briefing on a implementation of the adjusted head count ratio, which is inadequate, incomplete, does not have all of the dimensions that were recommended in this report, but shows what is possible using a consistent set of indicators for over 100 developing countries. Very briefly, the Global MPI was first developed, co-developed with the UNDP Human Development Report Office for a launch in 2010 at the 10-year anniversary of the Human Development Reports. And it therefore sought to measure directly acute multi-dimensional poverty for over 100 developing countries. And in doing so, it was clearly data constrained. But there were 10 indicators for which sufficient data were available for its construction. First of all, you are deprived of any member of your household as malnourished. You're deprived of a child has died. If nobody has completed five years of schooling in the household, if a child is not attending school up to the age at which it would complete grade eight, if you do not have clean cooking fuel, improved sanitation, safe drinking water, electricity, if your floor is dirt, sand, or natural, you're deprived. And if you do not own more than one of a small set of assets, radio, television, telephone, bicycle, and motorcycle, you're deprived in assets. So from these indicators, remembering the Venn diagram, what one does is construct the deprivation profile of each person or each household to see what indicators they are deprived in at the same time. These indicators naturally were designed in the time of the Millennium Development Goals, but they match or they very partially reflect SDGs 1, 2, 3, 4, 6, 7, and 11. And we'll talk at the end of some changes that might be possible to better align them. They also are not perfect. In some countries, not all of the assets are present. In some countries, we only have nutritional data for children and not for women or not for men. In some cases, we have only whether the woman said a child had died in a household and not whether the man also said so. So as in any global measure, there are some very strong differences between countries that are thoroughly documented, but that do exist. So how does this information go into a national measure? I was told not to put a lot of equations on the slides, so I'll try to do it very simply. Consider three-year-old Nahato from Uganda and the deprivations that she experiences. Her house has a dirt floor. Her house is made of poles and of mud, but there is a solar lamp, and so she has access to electricity and, indeed, to clean and sustainable electricity. She is one of 10 children of her mother, Namubi, who is 38 years old, and there are problems with school attendance because some of her elder siblings have not been able to attend school, though it's $2.75 US for four months of schooling. The global MPI leaves out work, and of course the working conditions are vital, but they were not available, and the working hours are long, but there is also food insecurity and undernourishment in the family. Any poverty measure also overlooks many riches and cultural or relational aspects of life. So, for example, Nahato and her family will be considered MPI poor, but there are also characteristics like being outgoing and self-confident that they enjoy, and at night they dance together to the music from a radio shared between neighbors. So clearly any poverty measure, particularly a global one, will overlook a number of aspects. It will not have data, for example, on work or on violence or on access to health care, but it does measure a subset of indicators that may be relevant to their lives. So having constructed, in a sense, the deprivation profile for Nahato and her family, the last step is to identify who is poor, and this is done by fixing a poverty cutoff in which a person by this measure is poor if they're deprived in one third or more of the indicators, weighted, and they're weighted equally across dimensions and across indicators within dimensions. Nahato is deprived in 50%, so she is poor. So the adjusted head count ratio then looks at the percentage of people who are poor because they are deprived in one third or more of the indicators, weighted, and it multiplies this by the average intensity. So Nahato intensity was 50%, and on average the intensity in Uganda is some number, and multiplying these two together gives the adjusted head count ratio. You're adjusting by intensity. It also gives a measure which has strong characteristics, like the ability to be able to break down by indicator, which is what it makes it, which is a vital use for policy, as we will see. So I've given this a very light touch, but for those with any interest, there is extensive documentation, and in particular, reflecting on the Atkinson Commission's requirements, there are standard errors and confidence intervals, statistical inference about the levels and the trends of poverty published. There are robustness tests, both to the poverty cutoff, to the weights, to the deprivation cutoffs, and there's clarification of the axiomatic properties that the measure respects. So what does this look like when it's implemented? So in the 2017 Global MPI, we used demographic and health surveys for 55 countries, multiple indicator cluster for 38, and PAPFAM for 3, and national surveys. These are quite familiar, so I'll just look at a couple results. But really, I'd like to focus again on the structure of multilateral poverty measures and not on these findings. But in 2017, we updated 25 countries. Each year, we update the countries for which new data are available. And the country coverage is such that we cover 5.4 billion people. Quite good coverage in Sub-Saharan Africa with 96% of the population covered, East Asian Pacific with 95, South Asia with 94. And low-income countries, it's 99% coverage, also in lower-middle-income countries, and 92% across all middle-income countries, population coverage. So it's very much a measure focused on developing countries. It's not a global measure, does not include high-income countries. But within its population, there's perhaps some interest. So what does this kind of measure give you at in one period of time? Clearly, you can see how many people are poor, 1.45 billion across 103 countries. You can see where they wake up, a billion of them wake up in middle-income countries, 72%, 48% in South Asia, 36% in Sub-Saharan Africa. Again, keeping in mind the sustainable development goal focused on leaving no one behind, it's necessary that the measure be disaggregated. And for a direct measure like an MPI, that's straightforward. So for example, within Afghanistan, we can see that poverty ranges from 25% in Kabul to 95% in Urozgan or 94% in Nur-Astan. So quite a range. But again, the policy traction comes not just at looking the level of poverty, but the composition. So in Myanmar, we see in the Rakhine district where the tragic Rohingya genocide is underway, we see that it is the poorest district of Myanmar. And we also, though, can see the composition of poverty for the Rakhine and how it is higher, but actually, in this case, not dissimilar to that of Myanmar nationally, except for higher levels of malnutrition in particular. The MPI, you might think, is useful when there's a large disparity of levels of poverty. But it's interesting then to look in Chad, where on average, 87% of people are poor. But there are two regions, Lakh and Wadi Firda, where 98, 99% of people are poor. So what's the value added of having a multi-dimensional measure? If you look at the composition, and that's the beige and green stripes, you see in the beige, which is Lakh, rates of child mortality and nutrition are much higher. And you see in Wadi Firda that deprivations in water are signally difficult. And so actually, the policies needed to confront multi-dimensional poverty, even in these very, very high poverty regions, can be somewhat distinct. And so trying to pull that out can add value. So the global MPI is broken down for 988 subnational regions, but of course, you want to look at other groups. So one of them is disability, and DHS now have incorporated disability status questions of the Washington group into their future surveys. And so you can see in Uganda that 76% of people who live with a person in the disability, with disability in their household are poor, and 69% of those without. And finally, on the global MPI, we desegregate by children, and we find that 48% of that 1.45 billion people who are multi-dimensionally poor are children. And furthermore that children are overrepresented, even given their presence demographically in the population. Clearly most of them, 84% live in South Asia and sub-Saharan Africa. In India, Pakistan, Nigeria, Ethiopia, more than half of the poor children live. And again, we can see that compared with adults, children are more deprived, not just in school attendance or malnutrition, but in absolutely every indicator of the MPI. So age desegregation even when it is not precise, and we might want an individual child poverty measure, which we are doing in some places. But even this kind of desegregation can bring useful information out. So that's a brief overview of the global MPI. The question is, is tracking such a structure, imperfect as it is over time, useful for the sustainable development goals, target of cutting by half or ending poverty in all its dimensions. So let me say a little bit about how the data are treated to look over time. First of all, they are very rigorously harmonized in every detail. So this is Cote d'Ivoire, and in 2005 the survey didn't have nutrition. So we drop it in order to have a comparable portfolio of indicators. In Sierra Leone, they didn't have male malnutrition in 2008, the first period. So we drop it to have rigorous computations. In Central African Republic, they didn't have a mobile phone in their asset basket back in 2000, so it is dropped. So there's a very rigorous harmonization across time, which is in the public tables. We are in the process of a big study we will launch in about six months, which will be then tracking trends in global MPI over time. But let me just show you a few results, and I know there's an interest here and wider in Sub-Saharan Africa, so I will share with you coverage of 35 Sub-Saharan African countries in 234 subnational regions. This is one of the papers that we've published. It was published this year. And across these countries, we can see the speed of reduction with Rwanda being the fastest followed by Ghana, Liberia, and Comoros. We can see the relative reduction even of countries that have low poverty with South Africa reducing its MPI the fastest relative to its starting level in 2008. We can also identify sub-nationally regions, and this is a map that shows the regions that reduced MPI even faster than Rwanda, which was our fastest country. They're runaways. They're really positive outliers, and understanding what went right in those regions is interesting. We can compare trends nationally with trends in $1.90 a day. The $1.90 a day trend is green, and the red is reduction in headcount ratio. It should have error bars. And so we can see that the trends are distinctive for the two indicators, and so there's a value in measuring the trends of each because they don't necessarily mirror each other. And we can also look a little bit more within countries. So, for example, Côte d'Avoire reduced multi-dimensional poverty statistically significantly at 1%, but the number of poor people increased from 10.7 to 10.9 million. There was a population growth in 18 of the 30 countries in Africa that significantly reduced poverty, and that wiped out the increase in that the number of poor people increased, although the level of poverty decreased. But the interesting thing for policy, again, is how that happened. And taking apart the indicators in Côte d'Avoire, we see that there were significant reductions in the number of children out of school, in child mortality, inadequate sanitation, unsafe drinking water, and a lack of assets. And it was changes in those five indicators that drove the national results. And we can also, of course, break it down sub-nationally to see for each component region what was driving the change. So this is how the the headline looks at trends and, in a sense, can track progress, but then can also be a tool to see how that progress happened and perhaps to incentivize change in the next period. In terms of leaving no one behind, when there are disaggregated statistics, then you can look at the poorest and see if it reduced poverty the fastest. So this is a happy story. In Côte d'Avoire, North Est is the poorest region. So on the horizontal axis, the poorest are on the right-hand side. And it's a race, and so it's reducing poverty the fastest. And so it's catching up. It's a positive story in terms of leaving no one behind. And in eight of the countries, we saw that happy story with the highest reduction in MPI occurring in the poorest sub-national region. There's also a question of whether the trends sub-nationally match the trends in national or in $1.90 a day poverty. So in India, for example, we see a positive pattern with monetary poverty reduction. These are the regions again with the poorest on the right-hand side and with the same axes and the lowest dots reducing poverty the fastest. And we see that on the left-hand plot, which is monetary poverty, that the poorest regions reduced monetary poverty the fastest. But on the right hand, you see that they reduced multi-dimensional poverty the slowest. So the biggest gains were in Tamil Nadu or Andhra Pradesh, Kerala, other states of India that were not the poorest to begin with. So this divergence of sub-national patterns is quite important and again calls for these indicators to be monitored separately. And this finding was taken up and republished in the Global Monitoring Report 2015. So that's a little bit of the stories that could come out of changes over time with a incomplete multi-dimensional poverty measure, but one that covers a good proportion of low-income and lower-middle-income countries. And of course, we can also see patterns by indicator statistical significant reductions in each MPI indicator, for example, for nine countries or reduction in every sub-national region. In terms of the SDG aim to cut by half poverty in all its dimensions, there were two countries that had a 12-year period between surveys, which is not advisable, but in their case they both more than have the MPI incidents during that period and other countries have also done so across several periods. So certainly it is possible within 15 years to have global MPI from very different starting points. So that's the end of really presenting to you a question, which is whether this skeletal structure of poverty measurement would be useful in tracking trends over time, useful both to see how we are doing, but much more importantly to give information to policy actors so that they might be able to catalyze and accelerate the reduction of poverty. Now, for those of you who need a little bit of a tea break, I'll do three minutes that will be of most interest to young people and students, which is where you can get all of this on the web, and the rest of you could step back if you need. So all of what I've presented, all of the tables are freely accessible as Excel files online. Not only do we report one poverty line, but we report three poverty lines. Not only do we report poverty, but we also have a destitution measure and the last 706 million people are destitute, meaning that they have severe malnutrition or open defecation or not even one year of schooling. We also have all of the indicator level detail and the desegregations that I've shown in the tables and things like standard errors or confidence intervals or retained sample if you're really a geek. And the strictly harmonized data that we have is online, and as I said, a lot more of such data will be forthcoming soon. If you're interested in a particular country, you can download the country briefing and look at graphics that visually display the information. And finally, if you want to make your own infographics, you can use the interactive data bank, choose the information that you would like to put into your PowerPoint or your presentation, and clip it out. So if you want to look at Cote d'Ivoire subnationally and Cote d'Ivoire in the region of its neighbors and see how its poverty is related to Ghana, Burkina Faso, Guinea, Amali, et cetera, you can do that. So that's a question and a tool, and I'll come back to that question later. But what I'd like to do first is ask whether information in the global MPI is already present in other indicators, whether they're $1.90 a day, whether it's an index like Human Development Index or Social Progress Index, Global Peace Index. Is there any value added to an index like the global MPI? Not considering its structure, but also its empirical results. So first, we turn to the $1.90 a day, and I very much see it as a necessary complement to the global MPI. The global MPI uses the demographic health surveys and multiple indicator cluster surveys that do not have consumption or income poverty, so it is impossible to include them in it. And there would also be questions as to whether the volatility of household level consumption data meant that their aggregate would be an accurate representation of their monetary poverty over the last year. So I do see them as essentially complementary. And of the 103 countries for which we have global MPI in this 2016 update, we have $1.90 a day for 86 of those countries. In 10, those measures come from the same year. In 24, the $1.90 are more recent, and in 52, the MPI data are more recent. There are MPIs for some countries without $1.90 measures like Afghanistan or Algeria or Egypt or Myanmar, South Sudan or Yemen. But there are also $1.90 estimations for countries where we lack global MPIs, including Chile, Costa Rica, Iran, Malaysia, and Venezuela. So in a sense, the measures complement each other, and there are also some differences in terms of their country coverage, though the number of countries is by and large similar for both of the measures. If we look at the message or the head count ratios, and that's what we can compare, this is a scatter chart with the level of $1.90 a day poverty on the bottom axis and MPI on the vertical axis. Another way of presenting it is this chart in which the height of the bar is the head count ratio of multidimensional poverty, and the black dot is the head count ratio of a $1.90 day poverty for the same measures. So what you see is that they agree on the low income countries, low poverty countries, but there's quite a bit of disagreement on the more poor countries. And because we are talking a little bit about the need for standard errors, we put confidence intervals for the MPI head count ratio. So you could see that even when we consider at 5% the confidence intervals, it doesn't really change the overall relationship between the two measures very much. Now let's look at some of the other composite indices, and then we will briefly talk about the structure of composite indices vis-a-vis the structure of a poverty measure. So one of the best documented and most rigorous indices is the Global Peace Index, which has 23 indicators of violence or fear of violence. And it's clearly of interest because the Global MPI lacks personal security, which was one of the dimensions of great interest. And it has qualitative and quantitative indicators, which are banded, weighted, and aggregated. And there are robustness tests and extensive documentation. So if you're looking for how to write a document on composite indicator structures, it's a very, very good one to look at. And the 23 components do not overlap with the components of the Global MPI. And so as we might expect, the measures diverge empirically, and they clearly complement each other, but the correlation is very low for the 95 countries for which we have data on both measures. Another index which is put forward or has put itself forward in the context of the Sustainable Development Goals is the Social Progress Index, which includes three dimensions, human needs, foundations of wellbeing, and opportunity, and four dimensions, four components within each dimension. And each component then itself includes a number of indicators of different kinds. For example, for nutrition, there's undernourishment, depth of food deficit, maternal mortality rate, child mortality rate, and death from infectious disease. So these are put together. It's sort of a massive undertaking in terms of the number of indicators, but it covers many of the same similar domains, as well as additional ones on opportunity and environmental quality and personal safety. The relationship to the Global MPI in terms of the country levels is higher for the 73 countries for which both indicators are available. And we could go on, but the relationships are not as strong, for example, with the Lugatam Prosperity Index or the Ease of Doing Business Index, because we might wonder whether countries with a strong business sector have a better environment in which also to redress poverty. The Fragile State Index is important because half of the poor children live in fragile states, but interestingly again, there's not a strong association. The strongest of all the indicators is with the Human Development Index, primarily through its education component, but also somewhat through its life expectancy. So a question, we talked about the conceptual need for multi-dimensional measures that complement monetary poverty measure, drawing on Amartya Sen's work. We then looked at Miguel Zekali, like Mahbub al-Haq, who recognized the importance of numbers to incite policy action and looked at the space in the Sustainable Development Goals for multi-dimensional measures of poverty and of well-being. So what would be the value added of having a multi-dimensional poverty measure alongside these other composite indicators? I would look a little bit at some of the features of the measures. First of all, remember the diagram of Sir Tony Atkinson of the overlapping deprivations. To do a counting-based measure requires aggregation first across dimensions for the same person or household, and then across households, and that's a different order of aggregation from every composite measure. So a composite measure like the Human Development Index, Social Progress Index, or the others, first aggregate across all people. This means they can draw from different survey instruments from different years that can combine data for different base populations, but it means that they do not reflect the joint distribution of deprivations. So if you have a society like this in which deprivations are spread across four people, these being the people, or a society like this in which one person has all four deprivations, if the marginal numbers are the same, the Human Development Index or any composite index would have the same value. So they're not going to reflect the jointness of deprivations. In contrast, counting-based measures will be able immediately to capture this, but they will be limited by requiring all the data to be from the same data source, and also by requiring a common unit of analysis, whether it's the individual or the household. A second observation is something that I very much learned from James Foster, which is the need to have very clear understandings of definitions between different kinds of indicators. So one might think of an indicator of well-being, which existentially essentially measures the size of a distribution, the income per capita, for example, and that would be different from an inequality measure, which essentially measures the spread, the ratio of two income standards. A poverty measure following CEN 1976, that is an absolute poverty measure, would first identify everybody who is poor, and then focusing only on them aggregate their information across the population in order to have a poverty measure. So a poverty measure reflects the base of the population, and it's quite useful to mentally distinguish them because what you find is that many composite indicators combine different kinds of measures, well-being, inequality, and poverty, and they also will include other statistics perhaps that do not relate to people, but relate to small arms or to the environment, or that relate to mortality statistics, like numbers of homicides. If you look at the sustainable development goals and goal indicators, and you just think of the definition of poverty, not the content of the indicator, but the structure of the indicator, then over 60 of the 232 SDG indicators are poverty in structure. They identify a condition, they identify the people who experience that condition, and then they aggregate across the population to reflect, to create the SDG indicator. So it might be interesting to really look at those SDG indicators and think how do we summarize at least a subset of them into an overall measure that is consistent, and perhaps how do we do so with accounting methodology. A couple other distinctions between a global MPI-based or accounting-based measure and the composite indicators is that because the composite measures are from different surveys, they often come from different years, and the frequency of updating those different surveys will be different, and so it may be that there will be a very long lag in updating some component indicators. Furthermore, the surveys used for composite indicators may not be desegregable by the same populations, and so it's generally more difficult to desegregate the broader indicators. And there may be less possibility again because of the combined data sources of having standard errors, of having harmonization in a strict sense over time. And also we could go into this more, the weights that would be affixed on a normalized 0-1 cardinal index component have a very different character than the weights that are fixed to a 0-1 deprivation of a woman like Nahato and whether or not she's deprived in each indicator. In particular, as Martin Ravallian said, the weights across normalized indicators in composite indices imply a marginal rate of substitutability between the indicators at different levels of achievement and across indicators. So there may be some value in again thinking of the structure of a poverty measure in also having a structure that is accounting-based structure as was recommended in the Atkinson Commission Report, even if some perhaps some of the indicators change. Finally, I'll mention in passing that in 2014 UNICEF released a counting-based measure for children, the Multiple Overlapping Deprivation Analysis, which covered 40 countries using data from 2008 to 2013. And this is again a very similar measure. It also makes an adjusted head count ratio for two populations for children 0 to 4 and children aged 5 to 17. It's a tool for advocacy, and so rather than a focus on policy that is the need to understand each indicator and how to change it, the focus is on really thinking of child rights as indivisible. And so they identify a child as deprived, for example, if they are not immunized in DPT or if they have not had skilled birth attendants or both. And so that gives a higher number, which is very useful in advocacy terms. It does lose information that would be relevant for policy. If a child is deprived, you want to know do they need immunization or did they need a skilled birth attendant? So now let's come to the practical part. We've seen the landscape and the openness to multidimensional measures that in a sense this is moving from the margins to have a space of its own, and it is invited to do so. But where does it fit in the SDG reporting? So if you look at the SDG report 2017 on Goal 1, it includes $1.90 a day measure updates and unemployment updates, as well as those on unemployment benefits and natural hazards. So those were the four indicators mentioned in this year's SDG report for Global 1. Sorry, these are out of order. I will come back to the SDG reporting, but mention that there is a bit of disquiet with the silence around multidimensional measures, and that's partly coming from many of the countries that use national multidimensional poverty measures as official statistics. And these countries are designing measures that are not comparable, like the global MPI, but that reflect their national development plan that are implemented using their national data sources that might include, for example, employment or internet access or violence or environmental conditions, such as are relevant in those countries. And they are comparable over time within the country for use and comparisons there. So the countries are using these national MPIs quite extensively to complement their monetary poverty measures and all but Mexico have distinct monetary and MPI measures. But also for budget allocation, for example, President Solis in Costa Rica passed a presidential decree by which allocations sub-nationally now must reflect the level of multidimensional poverty as well as monetary poverty and population density. It's used for targeting marginal regions and marginal groups in a number of countries with an associated census instrument. It's also used extensively for policy coordination because the MPI gives a headline, one number, that reflects the work of many sectors. And so it's in a sense a common goal politically. And so the different ministries can learn how they must work together to shift the dial on that common goal. And the Minister of Health also can realize, as he said in Colombia, how he needs the work of the Minister of Transport or of Water or of Nutrition to do their work so that he can do his. And again, there's a strong emphasis on desegregation. So when Mexico launched its MPI for the first time, it desegregated by indigenous status. Panama launched its national MPI in June and it desegregated and found that in the comarques of Panama over 90% of people are poor whereas in other regions it's 4%. So this wide disparity of levels of poverty. And there's also an interesting set of initiatives around engaging the private sector in doing a business MPI to where they identify who within their employees are poor but according to the National Multidimensional Poverty Index and have interventions on their behalf. So these countries in a sense are both learning about using the MPIs, not necessarily the global MPI although some countries use that directly but using it to prioritize the SDGs in their context that are poverty related and that need the most emphasis. And this community includes statisticians and the heads of statistics offices present in different areas. It also includes more policy leaders. So for example in the high level political forum, many countries have mentioned multidimensional poverty in their voluntary national reviews. And also there were three side events in the most recent general assembly on multidimensional poverty including the heads of states of Honduras, Colombia, Mexico, Chile, vice presidents of Costa Rica, Panama, the administrator of UNDP and two of the three and a number of high level people trying to think through how can multidimensional poverty measures actually be a tool for prioritization and for joined up planning in the SDG environment as well as for monitoring groups at risk of being left behind. But when we come to SDG reporting the question is what should be reported and that remains an unanswered question. In this table I've just put together some of the national official national MPIs of some of the countries and their corresponding global MPI and what you can see is that they're very different. In Armenia 29% of people are poor by their national MPI and less than 1% by their global MPI and the national MPIs tend to be higher except in the case of Mozambique and Pakistan. But if you go to the SDG indicator platform the space is filled for 1.1.1 the dollar 90 a day for 1.2.1 the national income poverty measures but it is blank for multidimensional poverty and that's quite interesting. It's because countries are custodian agencies of the 232 indicators only for two indicators 1.2.1 and 1.2.2 the multidimensional poverty measure and there is at this point not a way for countries to enter their data and so it's a pause because this will be corrected but in that pause there's a question of what should be reported. Should it be a comparable MPI like a dollar 90 a day there would be some arguments for saying that that would be better. For example the target is to reduce by half the proportion of men women and children living in poverty in all its dimensions and so a comparable measure would make sense like a dollar 90 a day because you are identifying an unacceptable degree level of poverty and wanting to bring it down but then we have the clause by national definitions which is so important in the SDG environment that indicators be nationally owned nationally constructed and as we have seen when countries come to design their national MPIs like their national income poverty measures they are different often they are more ambitious than the global MPI so there may be space for two indicators as in the case of monetary poverty a comparable one and a national one but in any case as the Atkinson commission report articulated there's need for a very strong conversation between national measures and comparable measures a greater understanding of how they relate to each other and of how the relative priorities that each embody can synergize with each other so in closing I just wanted to say that I have shared one potential structure skeletal structure if you will of multi-dimensional poverty but I've also very much pointed out its weaknesses and its weaknesses are largely because of data constraints it does not have empowerment or violence or employment or many of the other aspects that it might reasonably contain at the same time one wants to be very grateful for the data that exists in the demographic and health survey and multiple indicator cluster survey all of this disaggregation is impossible without those data that are free and publicly available but if we look at the more up-to-date data for 83 countries and 5 billion people we can ask what might be possible in the near future in terms of improving the global MPI and we looked across 31 indicators in quite some depth and to our interest and perhaps disappointment a little bit there there's not a huge margin for change at present given the data that exists if we want changes to involve at least 70 countries and 3 billion people but we can move to stunting or we can combine stunting with under nutrition we can have age specific body mass index we can add roof and wall to the flooring indicator and perhaps replace electricity with overcrowding as it becomes less relevant so there's some changes in the margin but there's also a very strong need to continue to explore better ways of merging geospatial data on violence or thinking about short work modules for inclusion in these surveys so I think my time is coming to a close and so I hope that we have shared a little bit of what are the possibilities in terms of developing a better global multi-dimensional poverty measure what value added it might have what kind of rigor it might offer that would in a sense add value to the landscape of existing measures each of which do important work of their own so we began by basking in the lovely image of edgard lesley the taint no sin to take off your skin and dance about in your bones and we've tried to take that freely freewheeling curiosity into the sober matter of poverty measurement we explored whether multi-dimensional poverty measures could describe the skeletal structure of poverty in a way that might better evoke understanding and ignite action and this exercise was very incomplete because it mainly left out policy analysis participatory and community engagement private sector interventions and many of the real work of reducing poverty but the question that I put before you and the one that we must ask is whether even this exercise as limited as it is might clarify the task and the priorities of reducing a small set of deprivations that are interlinked um catalytically and so I just want to close unapologetically with the words of the akinsen commission that did put before us the value of thinking about measurement alone um even as we know we must then move on to the next step of uh of policy and analysis the estimation of the extent of global poverty is an exercise in description as the martin sen has written description as an intellectual activity is typically not regarded as very challenging however as he goes on to say description isn't just observing and reporting it involves the exercise possibly difficult of selection description can be characterized as choosing from the set of possibly true statements on a subset a ground a subset of them on grounds of their relevance understanding the choices underlying the monitoring indicators and their full implications is challenging there will be differences of view but it is hope that the ensuing debate will bring together all those concerned and provide a basis for action to tackle one of the gravest problems thank you so much