 Thanks to the organizers and Finn and everyone for a really enjoyable time. It's great here at wider, isn't it? Really a lot of fun. Anyway, today I'm going to talk about a framework and it really is just that, a framework that's being developed because a lot of people have thought about things like inclusive growth measures or pro-poor growth measures. But I really think there's a chance right now of consolidating and making sense of a lot of different literatures in one framework. So this is a first attempt at trying to get some of these pieces together. It is crude. Okay, so we're right at the beginning stages of it. This is part of a project actually with Guanghua Wan from, used to be at Wider years ago, and now is at the ADB, and it is sponsored by the ADB. And it's trying to make sense of how we might actually bring conceptual ideas in very practical ways to monitor progress, not just the type of progress that is measured through growth in incomes, but growth in other things that may be corresponding with incomes. So why measure inclusive growth? I always like to put questions as sort of justifying what it is I'm thinking about. Well, growth really can improve lives. It can be one of the main things that can help people, but it's also possible that it may not be realized, its potential may not be realized when it comes time to seeing what's happened over time. You may have high growth with very little advance in many other dimensions. So let's consider the following scenarios. First, you have growth but with growing inequality, or growth with modest or no improvements in poverty. Growth that just leaves out certain groups in society, ethnic groups, groups in certain regions, or even leaves out entire sectors of the economy. Growth that has no improvements at all in other dimensions of well-being, and growth that leads to choking pollution. These are cases where people would argue as to whether growth is good. There are trade-offs. If you value more the other things that are not being realized, then you might say that that particular round of growth was not good. Alternatively, suppose that we had all the things moving at the same direction, growth with falling inequality, improving poverty. It, in fact, was spread out across all groups in society. Strong improvements across dimensions, and with lower pollution levels, imagine that. This would be a case where all policy makers, whether you happen to be interested in poverty or inequality, or whether you happen to be interested in particular subgroups of society, would say, look, we agree that these particular episodes of growth were good. So I take the idea from really Sen's book, The Idea of Justice, that it's a sense where the, it's an intersection approach. That was the basis of his book. When is it that all people with different points of view may agree that something is indeed an appropriate course of action? When is it that there doesn't have to be trade-offs among them, because all of them are in the same direction? And that's the case where you have inclusive growth. But how do you define it? That's the question. So I'm trying to get a very broad definition of inclusive growth. It goes like this, growth that simultaneously achieves other important ends. Now notice how vague that is. Other important ends are left for you to fill in, left for the researcher and for the policy maker to fill in. Obviously, someone must specify what ends are crucial, what ends that are policy relevant. So once you have however these ends, then it's hopefully it could be plug-and-chug. It would be a way of implementing immediately measures, different types of measures of inclusive growth. So the idea of this is to have a practical methodology. Not something that you could just think about in econometrica, but actually is able to be used again and again in empirical settings to monitor progress and guide policy. How are we to understand and measure the extent to which growth is inclusive? Now we have a definition in some sense, a vague definition. How can we use that in a sense to measure or to monitor the extent to which growth is inclusive? Well, the goal is to take into account the other outcomes and objectives besides the growth in mean income so that you can have broader policy traction. We're at the beginning stages. We'll appreciate any input you have. In fact, that's part of the reason I wanted to come to the conference, is I've been asking people, you may have been asked already, what's the conception of inclusive growth that you came to this conference with? What is it about inclusive growth that you think needs to be measured? Okay, so let's have a generic framework with generic definitions. We start out with a mu to denote the mean income, it's the average income, usually associated with the growth rates that you read about, and that's called the means. It's a play on words. Second E is some other outcome or the ends. It has to be cardinally measured, otherwise you really can't do this exercise. You have two observations, let's say a year apart to make it simple. Of course, if you have more than one year, you have to annualize it. Period one observation, period two observations. You could have many more ends, you're just going to focus on one at a time, and here are the growth rates of the ends and means. These are the kind of basic data that we're starting with here. An absolute measure will have three forms of measures, three, if you will, structures of how one might measure inclusive growth. First one throws out growth altogether, growth and income, and just focuses on the ends themselves. And many people feel this is the way we should be proceeding anyway. If you're interested in poverty, let's look at poverty and forget the other side. You're taking the end looking at its percentage change and judging the situation by that. You measure the extent to which E grows, the end grows. Percentage increase if it's something good, percentage decrease you hope if it's something bad. And just basically throw out the growth and means. Lower growth and means or higher growth and mean income doesn't affect things at all. All that matters are the ends. So, immediately you think what could this be applied to? Obviously, when you're evaluating the new World Bank shared prosperity measure, the mean of the lowest 40%, that's what they're doing. They're saying what's the percentage growth in the mean income of the lowest 40%. Likewise, you could focus on poverty and take a poverty gap measure and measure what's happening over time directly. That would be the percentage change, it would be a way of looking at it. Mean incomes of women or MPI poverty, multi-dimensional poverty index poverty. You see, you're able to take each one of these, look at it separately, and get another measure which is a strange way of looking at inclusive growth but it's one way of looking at it instead of looking at mean growth itself. Alternatively, there may be a comparison between growth rates in the ends and the means. A relative measure, as I've defined it, of inclusive growth would take a percentage change of ends over percentage change of means, and so therefore there would be an impact of how big the growth rate was in income on the ends. You're trying to make that comparison. If there's a lot of growth rate in income without much of an increase in ends, then the quality of growth wasn't quite as good. The inclusivity wasn't as much. That's what this type of measure would look at. So it's an elasticity, growth elasticity, growth elasticity of ends with respect to means. So if you did have lower growth in income with the same outcomes in terms of ends, you'd say, well, yes, it's a more effective growth rate. So there's more inclusivity per unit if you will of growth. And you could see that you'd have a very different idea of what inclusive growth was about here because you'd be comparing it, comparing two growth rates. So you'd compare the mean income of the lowest 40% its growth rate with the regular growth rate of a country, or the poverty gap growth rate with the growth rate of the company. And these elasticities are well known. Finally, many of those numbers that you would obtain in the previous two methods don't necessarily have absolute sense. So I've defined a third approach which is benchmark measures of inclusive growth. There's really two benchmarking processes. One is by empirically benchmarking looking at a league of countries or a particular country and saying, well, this comparable set of countries had this track record and we'll compare the track record of the country in question with the track record of the benchmark group. Alternatively, you might imagine a thought experiment, and this is often done in poverty analysis, where you say, if growth had been equally accruing to all parts of society, every single income went up by the same percent, then what would have happened to poverty? So you have a growth rate of 4% in the economy. If every income had grown at 4%, what would have happened to poverty? So you take this counterfactual and you imagine, what would have happened to poverty in that case, and that's your benchmark. It's kind of a best possible situation. And you say, that's what I'll use as my benchmark when I compare what number I saw. So you apply this to an absolute or relative measure of inclusive growth. You apply that to the counterfactual and you see, here's my benchmark. Take that as a ratio against the actual and that becomes the third type of inclusive growth measure. Now, the question I have for you as you're sitting here is, have you in your own work used other ways of measuring inclusive growth, other ways that you've thought of that are different to these and would not fit within them? If so, talk to me and ask questions. Right. In addition to those three structures of measures, there's different variety or orientations of measures of inclusive growth. The first is vertical, by which I mean within income distribution, you're looking at poverty, changes in poverty, inequality, or the size of the distribution. Likewise, there's horizontal issues. We mentioned it before with different groups, right? It's perhaps capturing the differential effect of growth across different groups in society. Men and women, ethnic groups, regional analyses, sectors of economy. Finally, dimensional measures of inclusive growth, which capture the impact of income changes across other dimensions that are of interest to people. Now, these other dimensions can actually be the things that would help growth more. And so you might be trying to track what kinds of effects are happening, which would in fact encourage future growth. But dimensional goes beyond the dimension of income. Of course, which one of these you would classify the particular measure as being would depend on what ends you consider to be important. So if you're interested in inequality or poverty, it's okay, vertical perhaps. If you're looking at multi-dimensional poverty, you might be capturing two of the three, vertical and dimensional. And if you look at multi-dimensional poverty across subgroups, you might capture all three. The focus here today, I'll give you a couple of examples of this, will be on two forms of variables. First, what I call income standards. It's a new way of thinking about income inequality and poverty and the size of income distributions. And so I'll spend a little bit of time getting into that. And then the second is an application to multi-dimensional poverty. So, yeah, the obligatory front of the book. The income standard approach has been brought forth in a book that I've written with Misha Locchin, a student of mine, Suman Seth. And we're showing how income standards unify the way of measuring income poverty inequality and the size or welfare. It's available online free, you don't have to pay a cent for it. An income standard summarizes the entire distribution in a single number, the representative income. And examples include the mean, the median income of the 90th percentile, the mean of the top 40% behind the World Bank's new shared prosperity index, sends measure of welfare, which is associated with the Genie inequality measure, and Atkinson's measure. It measures the size of the distribution. It can have a normative interpretation as being relevant for welfare. And as I claim, are the basis for measures of inequality and poverty. So let me give a, here's a picture of a distribution. Size is measured moving in that direction. Inequality is dependent on the size of the regions B and A. The base or the poverty is down in the left hand side. So those three aspects of the distribution are of interest to different people. Size is what income standards is trying to capture. The properties of income standards, I give them here just so you know what the hell is being measured here because sometimes people give definitions that don't give the properties. Symmetry. You flip incomes, an income standard won't change. Replication invariance. Replication, then, that means that you clone everyone's income. It has no impact on it. Linear homogeneity, double all incomes, you double the income standard. Normalization, if everyone has the same income, that's the standard. Continuity. Weak monotonicity. Those are self-explanatory. I have a ton of examples here of income standards. I'll just briefly roll through them. 10% income, mean of the bottom 40%, mean of the top 40%. The send mean is basically the expected value of the minimum of two incomes drawn at random from society. So it's kind of Rawlsian, but not fully Rawlsian. Geometric mean, notice the mean symbol is gone. Apologies. Geometric mean, as you well know. There's the indifference curve of the geometric mean, and it's from an entire family of things with little boxes for Greek letters, the general means. The general means is a class of income standards that can express preference for different sides, parts of the distribution with alpha less than one, the missing alpha for those boxes, then it emphasizes the lower incomes. Inequality can be measured in a plethora of ways, right? Plethora of ways. But what do they all have in common? Well, what they all have in common is that they fit within this kind of framework. That you have one income distribution, two different functional forms of income standards, one that focuses like odd of your left eye on the lower incomes, one that focuses on your right eye at higher incomes. The right one is always bigger or no smaller than the left one. And inequality is seen as kind of a difference between those representative incomes, measured either with this ratio on the left or some function of the ratio on the right. This encompasses every single inequality measure. So it turns out when you use an income standard, you're actually talking about inequality. When you use an inequality measure, you're using income standards. The measures plus their relevant income standards in this table. For instance, the Palmer or Kuznets would have the bottom 40%, the top 10%. Each of the first types of inclusive growth, the vertical and horizontal, is fundamentally related to income standards. The geometric mean, for instance, could be used. The absolute measure of inclusive growth would be the percentage growth in the geometric as opposed to the arithmetic mean. It'd be kind of focusing on growth of what, as Sen said, in inequality of what. You specify some other objective in maxima. It'd be really interesting to run a growth report exercise with other types of objectives, like the geometric mean or other income distribution-sensitive measures of size. You could construct a relative measure of inclusive growth, but what's interesting about this is that if you use the geometric mean and use then the arithmetic mean, by comparing the growth rates, you're tracking whether the mean log deviation is rising or falling. It so happens that by using an income standard, such as the geometric mean, you're already taking into account and giving information on inequality when you measure a relative measure of inclusive growth. And different standards yield different measures of inclusive growth. This is simply a picture showing how transparent the types of growth, the quality of growth is between two countries, as you have means that favor the lower incomes to the left, favor higher incomes to the right, the growth rates, which are on the vertical axis, are expanding at the lower levels for Costa Rica and depressing for lower levels. Mexico. So Mexico is not growing in a way which would be favoring the poor people. Okay, let me pass that by. If we have single-dimensional non-income variables, you can use exactly the same technology on them. But if you have many variables that you're interested in, how would you aggregate? Well, you could use an HDI or an IHDI as an end and do the same type of analysis, but there are all kinds of assumptions that are needed for that. And I would expect it would be very difficult for you to construct HDI because the definition changes on a yearly basis. But for poverty, there are all kinds of new technologies available. Here, I'll just talk about the adjusted headcount ratio. Or I won't. I'll just show you one number, Ethiopia. This is the growth elasticity of the income, the poverty measure, the multi-dimensional poverty measure that Sabina Alkar and I have come up with. Look at Ethiopia. You can see tremendously low elasticity there for the multi-dimensional inequality measure. This indicates the kind of lack of inclusiveness across dimensions of the growth patterns that we're seeing in Ethiopia. Other countries are given here. That's it. Summary. I had a framework for measuring inclusive growth based on ends and means, different types of measures. What's your conception of inclusivity? What does this framework miss? Help me out. Thank you very much.