 questions and then pass the microphone so you can just sit there and answer. Is that okay for everybody? Good. Who's going to start off? Thanks. I hate to do this, but I'm going to make a little bit of a comment as opposed to a question. Andy Berg from the IMF. I was mentioned a few times. I look forward to reading the 62-page paper, and the points are well taken. Many of them are familiar to us, but I have to say that the recommendation to use the WID and not the SWID reminds me of someone saying that they've done the pros and cons, and really bicycling is better than driving. Okay, you know, that's useful for some purposes that compares that suggestion, but for many others it's not. If what we've done is create demand with our work to create demand for better cross-country comparable databases, that's great for me. I have no particular vested interest in the SWID, and when that bike can cross the ocean, we'll take it. But I do think that, again, if I can promote myself, listening to the various presentations, both to just now and Tony Shorrocks before, I am struck with how much imputation and how much judgment goes into all these series. And I do think that if we sent Spanish statisticians to Mexico and Mexican statisticians to Spain to measure GDP, we would see huge differences in views about what GDP was across those countries. We know the GDP levels sometimes jump by 50% when people do revisions in many countries, not in rich countries, but in developing countries. And we know that when you go back and see they change the weights from 94 to 2000, and now the growth rate series looks totally different. You know, that doesn't stop us from even putting GDP growth on the left-hand side, let alone putting the level of GDP on the right in regressions. When we do growth, when we analyze growth, we measure institutions on the right-hand side. Imagine the measurement error and the variation across researchers in measurements of institutions. Openness. We always put openness. Not we, but we as a profession put openness on the right-hand side. What does that even mean? So I applaud, I really applaud these efforts to get better databases, but I do think there would be an element of sort of inequality if we users of these databases were told we kind of have to wait until we get the inequality data that can be put on the right-hand side. Okay. Let's try and have some questions. Are you managing there? I think just to respond, I will respond a bit to that, to say I think the trouble with the salt, the SWID data is that there is a gap in the market there, and this is what, I mean, the researchers want inequality data and they want it for every year, and there was a vacuum there, and if someone, our collective responsibility really is to try to provide the best data or simply make it quite clear that nobody should be trying to get data for years where we don't have any data, but you know, I think, I think there is a gap there, there is a vacuum, and it's going to be filled by somebody, and it would be better to have it filled by somebody who knows what they're doing. Jose Maria Lago from the University of San Paolo from Spain. A couple of days if my question is very easy, but I recently read some working papers from Pinkowski and Salai Martin, and they use exclusively data from national accounts instead of surveys. I would like if anybody of the commentators or presenters could give me some light why national accounts are more quality or lower quality than surveys, thank you. Okay, who else will we... This is a comment on the... I'd like questions rather than comments, but the comment can be short. The comment is that I can claim to be the father of WIDD, and actually the idea was that WIDD was supposed to be just a collection of information, and for every number there were like six fields which were being filled in, and in the end the data were rated, and there was a big warning at the beginning of the database saying don't use these for, say, time series, regressions, and so on and so forth. And if you want to do that, use data which do come exactly from the same type of surveys and so on and so forth. So that basically then, since the market demands regression support, I mean data support and regression analysis, that actually the users have been using this data wrongly in some cases. So we suggested that you can do regression analysis if you choose data which come from comparable surveys and same income concept, same coverage, and so on and so forth. Now people are making, perhaps you should lock up the data, I mean, I don't know, but the original idea was not that use this information provides you, you should use all the data, even the one rated one as consistent, because they were clearly indicated they were coming from different cities. Now a brief comment on Francois, I think that comparing the discrepancies that exist on African datas or Eastern European datas and Latin American datas, the Latin American datas are the one which on average, they show the least discrepancy. You're cut behind you. Yes, you're cut behind you wider and now working on trying to improve to be it. We agree. Some comments on the comments made by Professor Jenkins. First of all, we are very grateful for these very useful comments and in the very latest version, we have tried to actually take into account some of them or at least those which are more technically oriented. So for example, mistakes have been corrected. And we completely share the view that it's very good to ask researchers to be explicit about the data they use. And in fact, this is now our final conclusion in the revised user guide. One point regarding the multitude of series there are in the weed. One reason for that is that sometimes we also report in addition to the country average, we also report specific figures for rural and urban areas. And that's one of the reasons there are multiple theories. And we do give guidance to the researchers by giving this quality variable data. So not researchers to use those observations with the highest quality. But maybe as a further development, one option would be to provide in a sense two versions of the data. One is the full with everything we can find. And the second one, which is in a simplified with just one observation for each country, it's here for particular definition of genie, like whether that refers to before or after taxes. So that's just one possibility. And perhaps Stephen would like to comment on whether that would make sense. Okay, do we have any more questions? I would. Okay, then I'll come in. Yes, all right. Thank you. I hope it's okay that I make sort of a bit more like a comment, but I hope then that you can comment on that. The first one I want to say is that a few years ago, when I started at wider, just before I came to my horror, I saw that the weed was used for paper that was published in journal development economics. And it was a totally misuse of the database. Together with colleague, I prepared a review of that paper to the JDE and the review was rejected. The editor at the time was a colleague called Easterly and it was about aid and income inequality changes. It taught me a lesson. It taught me that the way it can be terribly misused. And it made me somewhat nervous to take over the heritage of the word. In addition to that, at the time, we had some transitional problems in terms of getting the word updated. Those problems are recognized up front. We were not quick in getting the updating process going. And I'm saying this in all honesty and I can give you lots of good excuses, but that's what happened. Then we were bashed very severely by AARC in a couple of meetings where we were told the weed is useless. So I went a little bit into my dark room and was reflecting a little bit. And then I consulted my academic board and they said, you have to continue to keep the weed moving. Okay. So we went back and we continued and we went on with the process. But then basically what happened, which sort of made us somewhat reflectful, was that then certainly there was this other database out there which had been put out and it wasn't quite clear that it was actually essentially just the weed. It took a bit of time for us to figure out what was going on. We were never contacted. So we have sort of, and I'm saying this in all respect, it is of course completely appropriate and okay that an outside researcher uses it, but internally we are sometimes jokingly referring to it as the stolen weed. Because we certainly saw that the decrease in website hits and so on went down considerably. And I'm saying this in all honesty and I don't mean that it's perfectly fine for somebody to continue to the work. But we were behind. We were late. But then I was also somewhat scared because then we certainly discovered the Dessert. This is sort of not exactly our equivalent, but within the UN headquarters is the one that sort of makes statements and studies about inequality. They came out and we were asked to comment on a number of papers saying that inequality in a number of African countries have gone down and that was strictly against what we had found. So then back to the drawing board again. Now we were in the process. We have been working hard to get it updated. And then by coincidence, and I'm saying this also just to say that I did not know of the work by the Journal of Inequality. I did not know about it until Nora four months ago. Something like that. When we then decided it was appropriate to have a special session. This session and I really wish and I hope it's okay that I just sort of make this plea. We are in a process of trying to figure out how we as an institution can help provide a public good and put the resources behind that, which is bigger than what an individual researcher can do or not do. But we had of course also been doing some of this homework that has now been referred to. And the big question for us, of course, is how do we take the next step? I mean, how can we try to add to the general environment of researchers in a good and productive and forward looking way? And that's why I would like to express my appreciation for the work and that's been done. And really this is for us. Just okay. This is now the next step. This is where do we move from here? And that's why we very much appreciate this interaction and this feedback between producers of data, users of data, so that we can figure out what it is that we as an institution can productively try to do in a helpful way. Thank you. Sorry, this became a bit long. So let's go back to the speakers and see if they have any final comments. Let's start with Nora. Sorry. All right. Can I just respond to Andrew Berg? First of all, I want to apologize if my remarks appeared ad hominem. It was certainly not intended that way. You just unfortunately happened to be here this morning and giving a paper about swid. And I like the paper in many respects. So I just want to point out that I appreciate you want answers. It's true that I'm a cyclist. I'm no longer a car owner. I live in central London. But beware of cars. They often go down wrong directions. And I really do appreciate that you want answers. But I think there are issues. This quality coverage conundrum really just has to be recognized a bit more. And essentially, I guess my beef, if I do have a beef with your paper, is about not taking data issues sufficiently seriously. And I like what you do in table four, for example, in the sensitivity analysis. But I would basically wanting rather more. And I'd also point out that, okay, I've grumbled about swid. But please note that I did talk about with use as well. The point about swid is that it promises to be a magic general all purpose solution. Okay, you don't have to provide the numbers. You're given the numbers. You use them with swid. It's the other way around. The responsibility is put on you, the user. And so my points were very much the same to with users. They have to address the same issues. It's just being done at different steps. And we have to we have to remember this. So the bottom line is essentially the things that Tony was saying before, everybody's been saying, take data, data issues very carefully, treat them seriously. Okay, I'm responsible behind you. Are you gonna? Not on a few points. The first thing that maybe commenting on the comments by Marcus, I think that when we discuss all these issues, I mean, there is one actor which was missing in this discussion, which is the National Statistical Office. Because all those people dealing with those databases, I mean, they cannot reinvent the surveys. I mean, they cannot introduce questions which were not there. And very much of the critiques that we can make and the hard feelings that we have about all these data are very much linked to those people responsible for the survey. So this means that in this journal of economic inequality issue, there must be something about this must be something urging more consultation between users and producers of those surveys. The second comment, maybe it's linked to what Stephen said, and also Marcus, is also linked to the discussion with Andy. I think that where we have a problem with this noise in the data on inequality, it is when we put inequality on the right hand side. Because here we have measurement error, and we know that all those coefficients will be very much biased. And this means that if we find that inequality has no impact on the left hand side variable, we don't know how to conclude maybe this is because there is too much measurement error on inequality. And the point made by Andy about why don't we have the same problem with institutions? We have the same problem with institutions. This is very, very badly measured. And we don't really know what we do with this. So I think that this is an invitation to take much more carefully all those cross country panel cross country work, looking more carefully to the way to the accuracy of the variables on the right hand side. The final point is about the question about Salah Martin and Mikhovsky. This is a different story. The debate there was about not so much a distribution data. It is about the mean income data. So some people would like to in estimating global distribution. Some people would like to use the mean income as given by the surveys. And some people say, okay, but this is not really accurate or this is not a good representation of the average welfare in the economy because public goods are missing. And so it is better to normalize all the distributional data by national account by GDP per capita or consumption per capita, whatever. So it is a different story because this is not something which has to do with distributional data, but simply with mean income, which is something different. And finally, a point on Andrea. I didn't say that what was going on in Latin America was bad. I simply said that there was a problem when you had two databases which are not giving you the same kind of information. But overall, I would say that this is what I'm saying in the paper. I think that CEDLAQ is doing a rather good job. And this is really the best practice at this stage. And the problem they have are coming from the surveys, not from the way in which they are using the surveys. So I think it is an important point. Okay, I just want to say a couple of things. I think you know that Andy, I mean, Andy and I have been talking about this now for quite some time. And that's why you said you were scared of me, but I never say things in public very much. The issue I think here is what I said at the end of my presentation. I don't have, you know, I am no expert of imputations. I just, the same way as you, Andy, want to use data, I'd like to use that data, but I want to be sanctioned by somebody who is an expert. Now I have Steven's opinion, and it makes me feel very uneasy. I think that what we're trying to do with this special issue also is to begin to apply the same scientific demands on databases as we do on articles. I mean, articles that are not published in reputable journals usually are not taken as seriously as us. It doesn't mean that they're always right, like Finn said. So I think that one of the things that we could have as a convention is that these databases get scrutinized, and maybe, you know, this time maybe not everybody got it right, but it's the beginning of a process. And I think that that is what I'm asking from the data users, because we know that this data, many of them, like you said, come from imputation, et cetera, but we're learning what the problems might be, and therefore we have to be more careful in terms of how we use them, maybe do much more robustness checks using, you know, replacing the values when there is hard data in the cases in which they're imputed, et cetera. And Finn, if I may, as I was listening to you, maybe wider is well positioned to start a process which is more formal in terms of how to do this interaction between users and producers and begin to generate certain conventions instead of everybody working in isolation so that we never come to a conclusion of what is acceptable from the different points of view, using science at the frontier as it is today, and, you know, in progress with progress as progress occurs. Thank you. Thank you. I think your sort of comments are very much in line with my own views about where the wind should be going. So we, I'm sure Finn would be very happy if you sat down before you go and just flesh that out a little bit. Andrea, you have any more comments to make, or you're the last speaker who hasn't re-spoken? Don't far too many things have been said already, but let me make very quickly two points. First one, you said at the end of the session that you are scared about the data issues. Let me instead stress one word that was used by Francois in his talk today, progress. I think that we have made huge progress in understanding and community on all sides, theoretical and empirical. We have much better data and especially we are much more aware of all the problems with data. So we are moving on. The second point is that there is no real tension between the national accounts and the sample surveys, tax-based data and so on. They tell us different stories with their own weaknesses and their own strengths. We have to live with multiplicity of sources. So there is no easy way to simplify problems. So that's the problem of salt. That seems to provide the right, the simple answer to a problem that is not possible to reduce to a single number. We have to live with different information and to understand why they are different. Okay, well thank you very much. I just realized we've run 25 minutes over which it's almost a record, I think, and just shows us how much interest there is in this topic and hardly anyone has been leaving the room. So thank you very much. Let's give a big hand to our and hopefully this will be a good input into doing something but to get the with back in its central role and to make sure it's updated regularly and we are aware of these issues. So thank you very much and we there's half an hour before the reception here.