 Thank you so much and again, it's a delight to be here where a lot of the people who shaped and taught to me have done a lot of their Interesting research. I simply have seven suggestions and they are focused primarily on multidimensional poverty I think that's why I would be here among the presenters and by multidimensional poverty. We mean where for the same person or household a selection of variables are Identified that in some way describe pertinent aspects of poverty for that country or that region The attainments or deprivations for that person or household are Dicotomized and so you have a profile of zero or ones depending whether they're deprived of nutrition or education or water sanitation employment Violence, then those are weighted and aggregated. So for each person or household There is a deprivation score or it could be an attainment score with a percentage of different relevant Dimensions and indicators where they are deprived or where they have sufficient attainments and that vector of dispute Deprivation scores then can be used to make inequality measures or can be made to you make Poverty measures and that's based on a counting approach which has been used in Europe for a great many years Conchita de Embrosio and others have used it in this community James Foster and I have worked with it to use to build other multidimensional measures and So the question is where is next for the community both in methodology in policy applications and in measurement What is next for this group and? I begin in a sense very obviously with a concern for data and I have a couple cheat slides here But when inequality or sorry adjustment with a human face was published And Andrea Francis Stewart are here and there was a call at that time for much better statistics for malnutrition And child mortality much more regular much more easy to interface with the policymakers And so the first recognition is that there has been a radical shift in data availability Since that time if you look at Povkal net where surveys from 1980 to 2016 number 11 went 1189 Then we have a base Now part of the problem with multidimensional or multi-topic surveys is that there is no Povkal net So you have to individually count you have to define this work is not yet done It might be an interesting topic for for wider which has done so much on inequality But there are 980 some surveys 68 more coming through this year So it's roughly about like there were five years ago and because multi-topic surveys we start counting in 1985 So there's a much bigger mass of microdata, but first of all it's More poorly documented so to give a very concrete example the Atkinson Commission on monitoring global poverty that released its report in 2016 called for a measure of multiple overlapping deprivations that included work and personal security With UNDP we make a measure which does include the other four dimensions under consideration How could one find out the data availability to add those others again? There's no easy archiving of data So we had to individually open a hundred surveys for five point five billion people look at the response code Look at who is interviewed in the household look at the the wording of the question in order to obtain information on what was available So the first observation is that just in terms of the existing data Archiving it and sharing it in a way that's relevant to the research community is not going need we did find You know that 54 countries home to 4.6 billion people have data on whether the woman has worked in the past week or the past month year self-employed type of application, but not whether she is in the labor force and So we also couldn't use this and many are in that position So there's also in a sense a need for a citation index of survey questions Which survey questions are actually used by the research community by the policy community so that they can be awarded and Unused questions that waste time could perhaps be discontinued The other big data need is for data on the joint distribution of deprivations So I mentioned that for a multi-dimensional measure need data for the same person or the same household and when you look at the big data movements or the merging of administrative data sources often you're not able to return the joint distribution and That has some value added that other Data sources than household surveys at present are hard-pushed to deliver But if you look at the world data forum in Dubai There's one session on a household survey and the title is innovate or perish So where is the cutting edge pushing forward household surveys with the same energy as big data? Making them faster making them easier. It's an interesting question Second topical indices So there is a call now for individual child measures that reflect child's well-being from aged 0 to 17 or for gendered measures Which compare women and men and for each of these you need to select dimensions indicators weights and cut-offs and Any index like these Including the current widely used indices at the national and regional levels Have big questions about the indicators that are in use if you look at the world development report You know that schooling is a bad use for education But every single multi-dimensional poverty measure which is officially put out by countries uses years of schooling as an indicator And similarly for a new novel measure like children. What is the equivalent of attending school for a one-year-old? There are there are lots of conventions that still need to be developed And I think that's an exciting conversation between sectoral experts and measurement people that by necessity need a few very strong indicators that proxy others and and Again, there is a need for better data information So if you look across all of them demographic and health surveys and multiple indicator cluster surveys It's hard to find the right Combination of indicators present to make a good child poverty measure just with child labor and some cognitive development dimensions We can do for just over 20 countries and under 2 billion people for a gendered measure No for a women's measure less than 30 countries 2.5 billion people So we also need systems of looking at what is jointly available and what are better indicators the Third area is that a number of governments now use Multidimensional measures as official national poverty statistics and they are running ahead of research in terms of basing policy on them so The government for example of Bhutan or uses the MPI to allocate its Budget across the 20 districts of the country They also is also done in Costa Rica and Colombia and Mexico But the budget simulation exercises the public expenditure modeling is very much the first generation and it needs a great deal more research Similarly and at the other end of the spectrum there are interesting innovations within countries not only in targeting not only in Monitoring but also in management So President Santos of Colombia at the beginning of his term engaged McKenzie to take the Multidimensional poverty index of that country which reflects their national plan and turn it into a tool for management Which could track change and Realize results within a given time period and so that interface between measurement and management might also be explored bearing in mind the data limitations a Fourth area is very much more academic and it is about methodologies of analysis If you think of the sustainable development goals focusing much more on inter linkages across Indicators on the need for multisectoral or integrated policies and on recognizing Multidimensionality not just of poverty but also of well-being and other things and you look at our bag of econometric techniques of Structural equation models or indogeneity and instrumental variables They really are not keeping up with the needs and the recognition of the deep inter linkages And so I think it would be interesting to really try to come up with quite a radically different bag of analytical tools that could look at some of these interlinked phenomena You can still do standard things max for fixed-effect models with public expenditure institutions growth on the right-hand side But there's a need to go beyond that Particularly when you're looking at sequencing and also the the cost savings from a integrated programming There's another area which is very very different and where I think we are missing something and that's at the interface with the Respondents to the household questionnaires if you think of poverty measurement, it's extractive You take the data you analyze it you give it to the policymakers, but what about the protagonists of poverty? There's an interesting experiment. There are many of these but one of them is foundation Paraguay, where at the end of the Enumeration Interview they try to have which is done on a little computer They try to then feed back to the respondent their profile and have a conversation of what could be done By that person or in that community now That's a tough thing to do most survey enumerators don't have the people skills to have a closing conversation Most of our technology right now can't give the respondent their profile But I think it would be if we are trying to advance an empowering agenda that gives agency to the poor It would be an interface worth exploring There's also a strong set of topics in measurement methodology for example at the moment the counting based measures dichotomize and So that is very rigorous and can be used with ordinal data But if you have a cardinal variable like income you're missing something if you simply dichotomize And so how can we have hybrid measures that really do look at depth of poverty and cardinal variables? And yet blend them with ordinal and binary qualitative data There's there's certainly room for innovation there Similarly, we as came up yesterday with James Foster and Eric Klobbeck. There are many many robustness and sensitivity tests But when a country has regions with a lot of an equality considering standard errors Or if you have regions with very different population shares the robustness tests in a standard sense do not apply and so you need to go beyond them and and Also, there's a big call for imputation for years when not all variables are present and yet at present the multiple Imputation models don't return the same joint distribution that we obtain from primary data and I think the last area is really of Recognizing that multidimensional tools are just one of many sets of tools that people are using and perhaps here's a need to step back and Perhaps develop what James Foster was calling proto axioms if you think of the head count ratio It's not a beautiful measure, but it's used because it's intuitive. It can be understood similarly with the HDI Which is being launched today in New York. These are not Wonderful measures sophisticated measures, but they have an ability to communicate Another example is Sir Tony Ackinson's recommendation that dimensions be roughly equally weighted Chosen so that they were equally weighted. Why so they could be communicated to policy makers And so there is a need to think about the poverty and inequality measures that are developed In terms of their ability to be communicated to people who don't wake up in the morning wondering How to make a good poverty measure whose primary consideration is something else I happen to but it's okay and So let me give you a couple examples as I mentioned There's a big industry to make very much better child measures and one would hope very soon better gendered measures Individual measures of deprivation But if a country has an official national multi-dimensional poverty measure and they've finally gotten their mind round its dimensions and indicators It's used in public policy. It's used in budgeting and targeting and then there's a child poverty measure with completely different indicators Dimensions and weights and then there's a gendered measure and then a workers measure and then an elderly person's measure Then quite soon. There is simply inertia. There's overload. There's data deluge And so a question which may be on the borderline of research But I think is one we should not ignore is what are the ways of making a user-friendly, but more rigorous package of analyses both unidimensional and multi-dimensional poverty measures that can be understood and used together and Which don't create the deer in the headlights kind of fright from the amount of information That has to be digested in order to be used to guide policy and yet does better than any single measure Household measure a child measure a single monetary or multi-dimensional poverty measure and Related to that. There's a big need for better work on the integration between monetary and non-monetary multidimensional measures to give a closing example in Let's say Chile 14.5% of people were income poor 20.4% were multi-dimensionally poor only 5.5% were poor by both measures in Bhutan similarly it was about a quarter in 2012 and then also a quarter with new poverty numbers in 2017 so there's a large mismatch between monetary and multidimensional which may be partly because of the volatility of data But it also may be a big difference and because most countries now Have both measures side-by-side. There's a need to figure out how to operate both of those together And so I think Recognizing not everybody wakes up wanting to think about measurement that the purpose of measuring poverty and inequality is not to please ourselves But to be useful to those who can be more active than we can in eradicating it These could be some of the the questions for the way ahead. Thank you