 That's my question today is global inequality really falling cut to the chase. This is the stylized fact that I'm going to question today This is one of many possible graphs showing showing exactly the same thing one construct construct from the data. What is it showing? Overall falls since about 1990 in global inequality global inequality is defined as pooling all people in the world forgetting where they live and constructing inequality measure on that distribution Falling since 1990 now important to note it was rising Back to a 1820 for a long-run trend increase. So that's a significant change What is driven at driving at a proper approximate reason was was Falling inequality between countries as you see here. I've given a decomposition. This is a decomposable inequality index Into between country and within country. So this is all well known in the literature But declining inequality between countries inequality within countries. It's not not Obvious entirely obvious there, but it is drifting upwards In the developing world alone. So taking out the rich world the picture is somewhat similar But it's got this interesting feature that it's starting to tick up That's the developing world only So clearly that's been driven by China passing through the mean For the developing world as a whole China for a long time was bringing down global inequality through the between-country component And now it's Increasing global inequality because of its its mean is above the mean for the developing world for the real thing Well, but you can kind of ignore that picture because I want to focus on the world as a whole Within some countries. We're seeing rising inequality somewhat dramatically It may be half of the developing world people think it's rising everywhere. It's not maybe a bit more than half an hour Indonesia is an interesting case because it was on a downward trend or stable for a long period Indonesia is now seeing rising inequality on a on a trend that will basically reach Brazil within 10 years. I'm not claiming that will happen There will be mitigating forces coming into play and there are some sign that they're already doing that but So you see the picture But the question today is this stylized fact robust to how we measure poverty inequality. That's a almost Freudian slip on my part I'm going to look at three things One thing I'm not going to do I'm going to stick to what I call it what others call the cosmopolitan approach I'm not going to say that foreigners are worth less than people within your country Right, I'm going to stick to a cosmopolitan approach which treats people as people wherever they may live And I'm going to do that as an ethical starting point I'm going to look at three things an easy one Lorentz dominance. Is this where we say inequality is falling in the world Do we have Lorentz dominance? Second what about relative absolute versus relative inequality and third and this is going to be the harder one What about relative deprivation? Lorentz dominance easy answer. We don't have Lorentz dominance We have intersection Lorentz curves over the long run and quite a most sub periods too In other words, if you have a sufficiently strong ethical aversion to inequality, you will say inequality is increasing What does sufficiently strong mean? If you know the Atkinson inequality measure, it means an Atkinson parameter over five. I think most people would say Answered in answering the question is the rising inequality of falling inequality robust They probably say reasonably robust in that case. It requires quite a high Atkinson parameter, but but duly noted It's not that we don't have Lorentz dominance. There are inequality measures, which will show an increase Absolute versus relative now. This is slightly more subtle we have an inequality measurement the scale independence axiom Multiplier all incomes by constant doesn't change the inequality measure The problem is most people don't seem to agree with that At least most students don't and here's a survey I did of nearly 400 students at Georgetown. It's a selected sample It's all the students who come to my course And it corroborates things that Frank Karl and me and others have found in the past A majority of people do not accept scale independence. In other words, if I show them distributions like this one two three two four six Majority a bit over 50 percent majority think the inequality is increased Actually, most the other axioms look fine the standard inequality axioms like the Pagood Dalton principle Where's that? That's this one here making a transfer from the Rich to the poor they they agree that's inequality Decreasing in that case, but it's really the scale independence axiom. That's contentious So let's see what happens if you relax it bang if instead I use the comms absolute genie index here, I've used numbers from Atkinson and Brandolini I've had constructed similar and more recent numbers same pattern Absolute genie index is rising globally Relax scale independence. What does that mean? It means that I'm going to measure inequality by the absolute gaps between the rich and the poor not the proportionate gaps Here's another way of looking at it the lack number Milanovic elephant graph This is a quote growth incidence curve for the world percentiles on the horizontal axis Proportionate changes on the vertical axis and you see this pattern of the much discussion of this has got a lot of Plane of literature in the media as well That's all relative. This is what the growth of what the elephant curve looks like in absolute terms Those are the absolute gains on the vertical axis against percentiles of the distribution It's massively absolute inequality increasing And we don't really know how how bad it is up here I'll come back to that at one aspect of this that's that's really striking to me is what's happening at the bottom Where you saw already in the developing world? It's not my very little progress right at the bottom of the distribution. We're seeing near zero gains at the bottom Rising absolute inequality another way of looking at it. This is a somewhat difficult measurement issue how to measure the lowest level of living in the world These are my estimates the blue line This is my estimate of the it's a weighted mean of all incomes below the poverty line with highest weight on the poorest people That that weighted mean is is pretty much constant. It's very small Increase and yet we've seen this dramatic increase in the growth rates in the developing world from roughly 2% to 4% Turning with a turning point around to the year 2000 Okay And this is the similar picture for Indonesia a little bit more progress for Indonesia and raising the floor Then for the developing world as a whole I want to go to this third issue, which is a bit more difficult When we think about when we measure in Income inequality from surveys we're making pretty strong assumptions about how we measure real income And they're obviously inadequacies and one of the inadequacies is is about the coverage and the time period a second is the time period The coverage of surveys in terms of time periods are typically relatively short So we're not capturing or getting a longer term view of living standards And and another aspect is is is the consumption of non-market goods if you get benefits in kind like Food stamps in the United States Those benefits in kind and not included in your income measure, so a number of well-known inadequacies in how we measure incomes I want to address those in a particular way Just to get a an in road into this question. It's not ideal, but it gives us a Quick answer to the robustness question. Well, the first point you could make is okay I may be cosmopolitan, but I can be nationalistic What does that mean that I care about relative deprivation within the country? I live in in the extreme case suppose I my real income or my welfare economic welfare depends only on my relative income That's an extremely nationalistic position if you take that position then obviously global inequality It's just average inequality across countries and we've seen already is rising Okay, so there we've got a non robustness obviously But wait a sec, who's now how could you think of own welfare depending only on relative income? It surely depends also on absolute level of income so I want to compass this within a Perspective that allows both negative and positive effects of mean income on your standard of living given your own income Now we always allow that mean income depend influences your standard of living through your own income But that if we assume it only matters through your own income. We're making quite a strong exclusion restriction Why because of the reasons they mentioned survey data are incomplete We've got missing aspects of income like benefits from in-kind benefits There are many other reasons to living in a richer country benefits you in many ways So relative deprivation the idea that I'm poorer if I'm living in a welfare sense I'm poorer if I'm living in a rich country at a given level of income because of relative Relative considerations working against that we have a lot of arguments which are just quite positive strong positive external effects Benefits of living in a richer country that aren't evident in the income measures we're using So let's encompass that within a very simple in a very simple way Log adjusted income is going to be log own income plus some weight attached to the mean income of the country you live in Now I want to ask I want to ask how robust are the statements we make about global inequality to the choice of alpha Now if you're a relative deprivation person some people argue this a If you're a relative deprivation person, you think alpha is minus one All that matters is your relative income in this case Logged it but you see the see that if I believe that living in a richer country gives me higher welfare for any of those reasons I mentioned you think alpha is positive If we look at subjective welfare, which is one of the few clues we've got and unfortunately most of the literature on subjective welfare is useless for this point of view Good for many points of view But useless this point of view because people either look at social at subjective welfare across countries Or they look at subjective welfare within countries if you do it within countries almost every time you'll find evidence of relative deprivation In other words people subjective welfare responds to their relative position That appears to be uncontentious now But what about national income? We can't with in-country good studies not going to help between country studies not going to help either We need studies which poor all people the world looking at their subjective welfare And then we look at whether national income matters to your own subjective welfare given your own income You do that you get a positive effect the alpha I'm going to just claim that you're going to have to believe me the alpha is Somewhere between point three and point five way way higher than minus one in other words with a National income matters to your welfare with an elasticity of About point three to point five But don't you don't have to believe me on that because I think you know we don't have a lot of evidence here So again, I'm going to be fairly encompassing Measures of inequality. Thanks Clive measures of inequality I'm going to naturally use if you see my my log specification. It's going to be obvious I'm going to want to use MLD mean log deviation We also know it has a lot of nice properties There's a beautiful paper by Frank Carl and Emmanuel Flecher on the properties MLD Very cute. Okay, so MLD. We've got standard properties MLD, but this time MLD is going to have a parameter alpha This thing somewhere between minus one and plus one Here's my step just within my ordinary MLD It's decomposable the same way and the MLD that you're probably used to is alpha equals zero The properties of the measure Well one obvious thing is that the within since we're essentially what we're doing is scaling up all incomes by a constant within a given country Obviously the within country component of inequality is not going to change Everything's going to be on the between country component The other thing is that you've got a strong monetistic property You can prove that this measure that I've shown you is increasing in alpha the higher level alpha the higher the MLD As long as the means vary across countries What do we find? Data galore Cut to the chase. This is the picture I've done it for between 1983 in 2012 This is alpha. This is global inequality So the answer to my question. Well, it's not robust Well, we knew that because I could always go to relative deprivation I could always argue that all I care about is relative deprivation which means it's non robust But the answer to this is saying you have to be essentially very very nationalistic You have to actually have an alpha less than minus point six Before you claim that the to global inequality is rising very robust the interval I talked about Was point three to point five now the other thing to note about this the magnitude of the effect One way of looking at it is we've got a lot of literature now on getting better better measures of incomes of the rich Suppose that I'm underestimating incomes of the rich That really they're double One percent. Let's take the top one percent and let's suppose that their incomes are actually double What I see in the data that would add point one to the global MLD That's like adding a tiny little weight on national income small positive weight. We're seeing We're seeing Really large impacts with the way you see that in the picture look at the zero for either year and look at the impact on MLD of Allowing for even small positive values of alpha This swamps the problems. We've been all worried about Just putting a small positive weight On national income in your welfare metric at given own income you get substantially larger global inequality, but still falling okay, conclusions The claim that global relative inequality to be falling is is is not robust, but the degree of Non robustness is important here. I think you need very high ethical aversion to inequality I would judge, but you know, that's we I don't have a any way of saying that your inequality aversion Isn't over point five over five in the Atkinson parameter. So that's entirely possible It's clearly not robust to relaxing scale independence big time not robust. So there it's clear Not robust to to allowing for a relative deprivation. That's clear but it requires an implausible degree of nationalism nationalistic measure that's we I believe quite Remarkably high to reverse the conclusion. Thank you