 Rwy'n cael ei iechyd rydyn ni, yn gwirionedd nifer i fynd i ddweud, a rwy'n rwy'n dod i'w ddweud i yw iddydd ymdian. ydw i ddweud chi'n cael eu cyfrifredeb Fyglwedd. Mae'r Fyglwedd Ardorladd fel Oesri, y mae eisiau gyda'r cyfrifredeb fydd yn gael arall yn y rhaid i gyd. Dwi'n cael e Womeniadau, yn fwy gwaith ymlaen, yn y pwyfin i'r cyflwyno i'r mateis cyfrifredeb. Onw. Onw'n weithio. Rhyw i'w ddweud a'n mynd i'w wneud yw'r bod yn ddechrau'n tŷu'r tŷu i fynd a'n dweud. Mae David a'r gyfnodd yw'n rŵn i'w ymddechrau yn dweud yn gweithio. Mae'n ddweud a'n ddweud i'n ddweud i'w ddweud i'w ddweudio'n ddweud ymddangos. Ond oed yn ddau'r fflawedau, rydw i'n byw'n bod yn dweud i'w ddweud ymddangos o'r ddweud. Ond rydyn ni'n gwybod, maelodden ni'n gallwn gweld unrhyw gyntafol i chi, neu oddo i ddweud yn gweithio newydd. On iddyn nhw, y gallwn arwain, dau 10 ysgol, rydyn ni'n gweithio'r reu'r hyffordd. Rydyn ni'n gweithio'r reu'r hyffordd sydd yn han i'n ddiwedd. Rydyn ni'n cael ei wneud ar mwylo i'r fodfyn Cymraethol a hynny'n gweithio'r fodfyn cyfnodol ateb ar ei wneud boi ddim yn ddweud ydym yn gweithio'r ddweud. Rydyn ni'n gwneud i'n ddweud i'r cyfnodd, rydyn ni'n mynd i'n ddweud y cyflwyntau. Rydyn ni'n gweithio, ynghylch yn ymddi, rydyn ni'n ddweud i'n ddweud i'r ddyfnodd yng nghymoedd o'r hollwch, wychwydd Lyfyn Hwyl yn boblnewisio fel y dysgu yn y dyma sy'n gallu cyfweld. Maen nhw'n ôl i'ch bod Folden Davies a'r awrf y project ond yn y fawr o'r gwahod, yn y bapers yw'r byw penderfyniad o fath yn deithun o'i bydd y gallwn gwithio y dyma maen nhw, y gweithiau ar gyfer gweld yn eich gwelwyd. Yna chi'n cael ei gwael wneud ond mae'r gwelwyd awr Jegol yw yn eich gweld. Felly, mae'r ffordd yn bryd ifnol yn unrhyw fyddi, tentrell o Melanovic yw fyddi yn ymgyrch basediaff y gallwn felly gydag y dyfodol. Felly, yn ein dwylliant, ryw'n edrych o ffrind i fơ o'r ysgolion sydd hi'n symud bod ydych yn gweithio lle mae'n ymhyfyr, oherwydd yr ysgolion yn ymgyrch. Ond, mae'n gweithio chi'n gwybod rwy'n edrych, oherwydd yr ysgolion yn fwy ar iechyd, ac mae hi'n fyddych sy'n adlengwys ar y cyflodiadau. And it's turned out to be, as I'll explain now, I think rather better than we had envisag'd. But ten years ago we started a project which was with Jim Davies, Ed Wolff, who's very well known in the US wealth field. Susanna Sanström, who was working with us here as a research assistant. Felly, fel am yw'r ysgolwydd, ydym yn cerddodd ystod, a'r ffordd o'r hollwch chi ar hyn yn ysgolwyddu, ac rydyn ni'n dod o'r ffordd am y byddai, ac mae'n ddod o'r rhaid o'r rhaid o'r hollwch chi ar y byddai. Rydyn ni'n dod o'r hollwch chi ar y byddai, ac mae'r hollwch chi ar y byddai. ac y ddatganiadio ar hunain. O'r cyfnod o'r ffordd yn ffordd. Gwyddoedd yw'r llwag yma. Rwy'n ffordd iawn, ac mae'n gweithio'r ddifŵr oedd cyffredinol. Mae'r llwg arall, oherwydd mae'n gweithio'r ddifthoedd, rydyn ni wedi eu cydnog yma, ond mae'n gweithio'r ddifthoedd. Yn y rhan o'r ffordd, and as Tony just said I expect you might even be able or if you can. I don't know if there's any copies to buy here but if you do have a copy you can get the editor to sign that since he's here as well. But it's an interesting book and still lots of interesting material in there. serious trouble. When I was, I have to say, some of my work on wealth has led me to think about some advice that I might give to people when they're working in their income and wealth field. One is, I think, first of all, to be skeptical, and that is, I'm strax all the time how people just get a data set and just process it, and discuss the results without really thinking about all the problems with the data. Ie ddaeth y survey, yn ei ddweud o ddysgu ddater, yn ei ddysgu ddater a i'w ddod oedd ganddysgu'n ddater. Mae'r cyfnod wedi bod yn ddau'n ganddysgu'n ddater, byddai'n ganddysgu'n ddater, byddai'n ganddysgu'n ddater, byddai'n ddater sydd yn ei ddater i'r llwyffydd yn ffordd fawr. Mae'n cyfrifio'n gyfrifio'n cyfrifio'n cyfrifio ydw i ddater, byddai'n ddater i chi ddau. just test it. But I think there's a lot of other data which people do do just rather process it rather unthinkingly and you do have to, I think, not trust any data in this field unless you really have confidence in it. I would give that advice because I think that you can benefit a lot from taking that advice. The second thing I've learnt by doing this wealth is to try to be resourceful or inventive, and that is, you know, try to use data which is not the standard data and try and put things together and try to get an overall perspective from combining data from different sources and processing it in interesting ways. We've done this in the wealth field in a couple of ways. First of all, usually you get data when it's lagging behind several years. We actually use, we are just now processing data for the middle of this year. We are about to publish wealth estimates for the middle of 2014. How do we do that? Well, because we use a lot of changes year to year in wealth holdings are driven by what's happening to stock market prices. We're working for Credit Suisse so they give us the data on stock market indices to the end of June. We have house price indices, which we use as a proxy for real assets. We have exchange rates quite easily available. So we use all this and we get, I think, a reasonable estimate of middle 2014. If you wait for all the data to come available, we're talking about two or three years lag. Of course, things are changing quite rapidly in the world and it's much better to have up-to-date data. So that's the way in which we can use some interesting available data in useful ways to get up-to-date estimates. The other thing that we've been doing is to use data on top wealth holders. There's actually quite a lot of that available. Every year it's more and more available. The Forbes data, which is the best known on billionaires, that goes back now, I think in the 26 or 27 years of billionaire data. The UK has a rich list, which they publish every year. They've done 25 years now. There's lots of data for individual countries now with that sort of information. There's problems about it, which I will also mention, but in terms of using that data, I think we're using it extensively and I'll try and persuade you that it's, I think, a very valuable source of data. But also in the income field, I think we have now, as Francois was talking today, and the PSC database, there's lots of data on top wealth income holders in various countries, the top 1%. A lot of the data which you get in processing survey data is not going to really capture those top incomes. Can you somehow combine the two together to get a more realistic estimate of income inequality? And I would say yes, and I'm going to at least show you how we do it for wealth so that you can perhaps do it also for income. And that's one of our objectives as well. The last point here is to be humble and that is, I think in all of this field, there's a lot of uncertainties here. And anyone who simply starts to talk about the third decimal point of the genicoefficient and so on, I just wonder, you know, I think you're giving a misleading impression. I think we have to just sort of say, you know, I think this seems to be the sort of trend, you know, if you've got enough data here, put points for enough data years, you can have a little bit more confidence. But I think we're talking about here of quite large confidence intervals and I think everyone should at least, I say, be humble in terms of what conclusions you draw. Today I'm just going to say I'm going to talk about the methodology we were using for our global wealth distribution side. I'm going to talk about the various problems that we've encountered. I'm going to actually end up offering a sort of alternative to the Lorraine's Curve as a method for capturing and analysing inequality data. And if I'm probably, I won't have time to talk very much about the income side, but if there's some questions I'll handle that as well. This is briefly, then. This is our strategy for doing global wealth, for constructing a sort of global wealth estimates. It's first of all we try to estimate the average level of wealth in different countries. How do we do this? Well, some countries have good balance sheet data. Some other countries have some survey data. Where this is available, we use it. There's a growing number of countries, so we're now up. I forget exactly, but I think we're talking about 40-odd countries that would have some sort of data, certainly on the financial side, fewer countries with non-financial data. We also, we then extend this to other countries by using regression methods, because we think that there's a relationship between income and wealth, and so we can sort of estimate what we think the asset holdings are in various countries that way. Other countries where we just can't do it at all, we like to have a complete picture of the world, so for some other countries we just sort of impute a number, which is the sort of regional average. These tend to be quite small countries, or countries like North Korea, Myanmar, Sudan, these sort of countries which you can't really do much about, but rather than just miss them out, some other people that talk about global income inequality, they just miss these countries out, so we just give an imputed value. Then we look at the distribution within each country. We do that. Some countries have wealth distribution data, but when we started it was really quite small, I think less than 20. Now it's up to around about 30, I'm sort of ballpark figures, so we thought, well, what we can do here is if we haven't got the wealth distribution, we do have income distribution figures, because we've got the WID data, and maybe there is a sort of systematic relationship between wealth inequality and income inequality, wealth inequality is always higher, and there is a sort of pattern. So we would estimate the Lorenz curve for the wealth distribution from the income using the 2030 countries where we do have both wealth and income data. So this was our strategy, and then we would add the, again, impute values to countries that didn't have any income distribution data either. I should also have gone back and said, for the mean for the, as I said before, for the recent years we use, we update the figures using recent data on what's been happening to the market capitalisation in different countries and house price indices and exchange rates. We then, I should also mention that in order, one of the problems, I won't mention problems, we have a little, we then use the distribution and the level we generate a synthetic sample for each country which exactly matches, which is consistent with the evidence that we have, consistent with the mean, consistent with the distribution, we generate a sample, and then we throw this all together. This sample is actually now one point, it's a sample of 1.3 million, so it's, you know, it's, it's roughly in proportion to the size of the populations of the countries. And then we process this and generate and discuss the results. What are the sort of problems? Well, some of them are problems which you have to tackle in the income side as well. You have problems with the definition of the income concept or the wealth concept. Again, I won't dwell on this, I've just mentioned, because it's not particularly interesting and it's rather predictable, I suppose. We also have problems with the unit of analysis. This is game one which anyone who's been working in this field, even if you're working within a country, you have a problem of what distribution are you talking about. For the wealth side, we've, we adopted the strategy that we're going to talk about adults, which is a bit unusual because people usually talk about households or about the distribution across persons. We took the view that really minors don't hold very much wealth and so we've, and it's not really interesting to look at the distribution across the world's population, it's more interesting to look at the distribution across world adults. And that actually seems to have been adopted by other people that work on wealth reports since we started that. There's issues also about exchange rates, whether we use current exchange rate or PPP, that type of thing, which we are progressively now just keeping with current exchange rates. More interesting questions that perhaps you won't have come across or the solutions we want to come across. One of the problems, if you look at the width, you've got a lot of, it's in a particular form, some countries you've got quintile data, sometimes it's DSAR data, you've got top 5%, bottom 5% missing bits and pieces. We've taken the view that we've put it all in a common framework, which is to look just to actually construct the Lorenzco of where you get the missing points. Then we have a nice little utility which we wrote when we were here at Ryder, which is on the Ryder website, which generates a sort of synthetic distribution to exactly match the distribution. The one that's on the Ryder website is the first version one of this. We've since been developed a more advanced one, which gives you sample with different weights. The reason why that's quite important in the wealth field is that we're now using one sample point for each 10,000 observations, that's for each country. But when you get to the top 10% in each country, we have one for every thousand, and when you get to the top 1%, we have one for every 100. This really gives us much more detail at the top end, which is really where all the driving interest is coming. So that's how we generate this 1.3 million observations around the world applying this rule. Here's an interesting question. I think we have come up today a little bit, particularly with France, so I think about the question about residency. It's not so important perhaps with the income side, although this question about people owning assets and keeping income or having income from overseas and whether they remitted and so on is an issue. In the wealth side it's really quite important because billionaires move around the world. And if you look at the Sunday time rich list for the UK, there's a lot of people that you wouldn't immediately think of being British. In fact, in the top three or four people are all, I think there's a couple of Russian billionaires in there. But this issue does crop up quite a lot and I think this is going to be another issue for the future. People do move around and certainly if they don't, some of their assets move around and it's not quite sure whether you're measuring them in the right sort of way. I don't have a particular solution to this but I think it is one which could be increasingly an issue for the future. The one which is really quite important and I'm going to spend almost time is the top tail adjustments. We think in the case of wealth distribution that the original data that we get, which is essentially from surveys, isn't accurate. Even in the best quality data is not and in most countries it's way off. So how do you somehow adjust the data to correct it for these missing people? There's two problems. One is that you miss people because in surveys there's a lack of response by people of the top end. This is well recognised. The second issue is that they tend to under-report it. It seems to be an endemic problem of surveys that to some extent you can try to improve on it but you're still going to be limited. The last issue here, I don't think this is completely exhaustive but there's questions about methods of presentation. Again, our experience of just sitting down and trying to present this data to various people. It's quite, you know, to think about how you present inequality data certainly in a way that is interesting to the general audience should we say. I think it is quite a challenge and I think just listening this morning there's various ways in which one does that quite successfully. I'm just going to mention too that we've used quite a lot. When we started doing with Credit Suisse, they kept talking about a pyramid, the wealth pyramid. Nobody had really thought about it but then I thought well we could do it as a wealth pyramid and that's exactly what we do now. This is the sort of pyramid that we produce. We group people into those with assets less than $10,000, 10 to 100, 100 to a million and over a million dollars and the areas here are reflecting the size in the world. So it's really quite an interesting way and I think one could do that for income side as well quite happily. The other thing which we do which doesn't seem to be used, I've never seen this used elsewhere but again it's something that we use first time in the wider book and I think it's really quite interesting is to look at the distribution of people at different points in the distribution in the world. So this is just looking at the desile points in the world and saying where do these people live. You could do this of course in individual countries and you could look at where people live within the country or various sorts of characteristics. It's a rather interesting way I think of rather than a table of doing a visual presentation. This is the latest version I've just taken from this year's report. China you can see there, I mean you look at this immediately you think there's the dominance of India and China in the world and China taking this really middle position. Ten years ago when we first started China was much more in the middle and you can just see it year by year this great blob moving towards the right. And I think it's really a sort of interesting again way of presenting the data. Now the real problem is here reflected in this diagram where we, this is just taking the data for China which is given from a survey and plotting it. What I've done here is to plot what you would normally do if you were looking at Pareto tails. In other words you plot the logarithm of the number of people with wealth above a certain level and you use the log of wealth. If it's a Pareto distribution you get a straight line, most times you're not going to get a straight line all the way down. But the notion that somehow as it towards the end it straightens out that would be the characteristic of the Pareto distribution. The problem as you can see here is that, oh here we are, there's a little blob there. Where's that blob from? That is the billionaire data. Back in the year 2000 China I think had one billionaire and there is the billionaire. And you can see if you were predicting that from the distribution you are not going to get any billionaires. In fact you would hardly get anyone with I would think even if you extend that you get, what do you get to? You get maybe 10 million if you're lucky. You wouldn't even get, the biggest wealth in China would be 10 million or something. Clearly they've got a billionaire. What's going on there? We think it's to do with all the problems that you have with survey data in the top tail. What do you do for that? Well we've taken the view that we'll assume that we have a Pareto tail. And we'll assume that the billionaire observation is correct. And that we will somehow straighten out the line and use that as our adjusted distribution. The problem with this is you're adding a huge amount of wealth in. So you then have to adjust the mean of the rest of the distribution down to correct it so that you've adjusted for this extra wealth that you're adding in. So you just reiterate and we would then, our adjusted distribution would be this sort of dotted line here. And this is what we didn't do this with the wider study here. We just used the unadjusted data. When we started doing it for Credit Suisse we have made these adjustments. It makes I think a huge difference to the inequality in these countries. I mean you get big jumps. We're talking about genies now. If you're used to thinking genie of 40 or 50 or 60 years high we'll just forget it. We're talking about up in the 80s and the global genies are close to 90 I think. So we are in a whole different world here. The problem is what happens here? How do you, this year we've been sitting and saying can we get a series of inequality, wealth inequality back to the year 2000? And the trouble now is saying okay, here is the data for China of 2005. If you keep your eye on that sblob as you, this is again, this would be the survey data. That would be the billionaire data. As we go on, we're up here, China in the year 2010, close to 100 billionaires. And here is 2013, it's 150 or something billionaires. Again, it's totally inconsistent with the survey data so we've got to do something about this or really give up our exercise. We get the same, and the question is how can, at least the initial thought might be, we just adjust this year by year. The problem there is we think that the billionaire data can be jumping around for, it's really quite, we're talking still about quite small samples. For other countries it can be, here's China with 100 or so, but other countries could only have four, five, six, ten billionaires. These are quite small samples, they can be jumping up and down each year if someone dies or splits their wealth. You can have quite erratic movements in billionaires. We could be relying too much on data which seems to be unreliable. I'm just going to show you the US data because the US is just so wonderful really. This is again the same figure, I'm afraid when I was printing this I realised there's a slight glitch somewhere in the US data and there's obviously some grouping somewhere because of the way that the data is being generated. So just forget about that, I'll sort that out at some point. The point to make here is here's the US figure. This is just their survey data, the survey of consumer finances and you can see actually the US has got something like 100. It's actually got more billionaires, 10 to the power to it, just 100 and this is for the year 2000. Even so we were able to generate our sample, I told you we were having one for every 100 people so when they get to billionaires in the hundreds we actually have a data point in our data set which is going to represent that. Here's the figure for 2013, again we've got sort of 400, I forget now, 400, 500 billionaires in the US. So our data point is, we've probably got a couple of data points in our data set. But again you see just how good it is, there's hardly any adjustment needed for the US. The US I think claims to just be about mis of top, about 2%, it only misses really the billionaires from their data set and it's roughly 2% of wealth. So the US data, they claim it doesn't mean much adjustment and in fact that's very strong evidence here. The US, the data seems to be way above much better than any other country possibly. There may be one or two other countries that do something that's pretty close. I think Canada, I don't know, maybe Italy, certainly not the UK, you wouldn't get anything like that. So the problem is, how do you combine these in different sort of years? This is my good idea I have to say now. Instead of here we're plotting the logarithm of the number of people, here we're plotting the logarithm of wealth. Suppose that we do here we plot the percentage, the logarithm of the percentage of the population and here not wealth but wealth relative to the mean wealth of the country, what multiple of mean wealth it is. The great advantage of that is you're adjusting for the population size, you're adjusting for the mean changes in mean wealth. So in terms of simple things it is, if we make that adjustment we're now, if all that's happening is there's a change in population or there's a change in mean, the actual curve doesn't change. It is a replication invariant and whatever it is, mean independent type, I forget the terms I've used in the past. So that allows us to say if inequality is not changing this distribution shouldn't change. But of course the mean changes so the billionaire point changes and what you then get is a fixed curve if there's no change in the distribution and you can plot all the billionaire points not just for a single year but for every year or indeed any rich list data on the same diagram. So here we have now, this is tracking the US figures, this is the billionaire data back to year 2000 all plotted on the same line and you can see roughly speaking here that the billionaire data as mean wealth goes up as it has been in the US the billionaire points are moving towards there's more billionaires because there's more people being pushed over the line. If inequality is not changing it'll just track up that curve. And indeed that is roughly what's been happening to the US. The evidence here is that there's a very little inequality change for the last, since the year 2000. Here's the one for Australia which I think is even nicer because Australia's had rather bigger increases in wealth partly because its exchange rates been appreciating. And here is the distribution curve adjusted for one year but we're keeping that distribution fixed. Now you can see you're pletting the logarithm of the percentage here so there is always two at the top here. 100% of the population is going to be the top end. Here is Australia and except for the first couple points up here in the first early 2000, 2001 you can look at the data, you're just tracking up, it's just moving straight up here which I'm interpreting as being almost no change in inequality year on year and simply moving up there. So this is a country in which wealth inequality has not been changing over time. But what's exciting me here is instead of relying on having billionaire data for one year and we just could be jumping around, here we're plotting them all and we can smooth out the data and actually get something which avoids some of the year on year variations. So I've got lots of, I've thought I would, for the Minister I've got a little diagram for everybody here now. This is Brazil again. Here the billionaires are going up but they're tracking up pretty well along the line indicating that wealth inequality, I mean I look in more detail in a minute but it doesn't seem to be that it's huge change in wealth inequality in Brazil over the time. Who else have we got? This is for Francois. Francois, well it actually looks like here where if you interpret this, these are the early years, we're moving in this direction which actually tends to, I'm interpreting this as actually if we started with the line here and the line gradually moves in this direction it means there's falling wealth inequalities. That's roughly what's happening. I've got here for, here's the Italian data again, not a huge, doesn't seem to be huge change. I'll at least, there's not as much obvious change as there would be from other countries. Mexico, this is for Nora. This is my Nora diagram for Mexico. Again, I don't quite, since you can't really tell which is the first year or the later years I'm not sure about the trend there but it's not quite, it's not so evident as it is for Australia where it sort of tracks up the curve. So, here's the UK but again all these points, interesting because we can plot them all on the same and we can start processing in the way. Now here is some interesting one. Here is China. What's happening here? Well, here is China in the year 2000, the line would be right down here as it goes, as we're going over time, it's moving up here and what's that mean? It means wealth inequalities going up big time in China. And this is looking at this data that's giving you a big signal that's happening. Here's the one, here's India again going up, trending upwards, here's the early years and we're trending up here. Here's Russia again the same sort of trend, perhaps not quite as strong as it is in China but again increasing inequality. So, let me just move back. What this allows you to do now is to look at, to take a sort of one line. I've taken for each year, I've taken the sort of average, what I've done is taken all of the Forbes data, combine that into a sort of single representative point, computed the curve based on that average and then looked at deviations each year from that line. So, you look now and you see, to what extent is it deviating, is it above the line or below the line and by how much? And then you can plot those against each year. The T here refers to the time since the year 2000. So, this is France and I have to say what is quite evident in this data is that there seems to be reduced inequality up to the financial crisis and then the crisis is changing and it's going up. In fact, I've just made that assumption and assumed that there's a split, there's a change in 2008. So, here we have France, so we're trending, here the deviations it was above the line but the line is going down and then the deviations are going up again. So, we've got reduced inequality for France going up again in the period since the year 2008. Here's Italy, again you look at the dots, there seems to be a pretty clear pattern there that Andrea would perhaps recognise. I don't know whether sometimes it's evident but I mean the French thing is because there's so much interest, the Piquet is generating this discussion in France, it may be, it's a groundswell of opinion thinking that inequality is going, wealth inequality is going up in France and this data would certainly tend to suggest that. So, that would be consistent and in Italy and indeed here's a Mexico, well not so much. I mean the Mexican figures are starting declining again up to the year 2008, a little bit of a jump, whatever. Here's the US, the US is quite interesting because it's been the main focus of all the discussion this year about what's happening to wealth inequality in the US. The SCF data is very close to our numbers here so we're not differing very much. In fact, the US data is probably good enough that you could take the individual year on year variations. Here we are, we seem to be, I forget now, according to the SCF it looks like I think 2010 is their latest figure and they said wealth inequality was higher than 2007 which is correct, I mean according to these numbers here. Beyond that, before that it was fairly flat, again this is exactly what we're generating here with our numbers but the indication here is we've generated the last three years and we think that the numbers are falling. So when the SCF come out with their next data, which will be for the year 2013, then according, if I were betting I would bet that actually it would be going down and that's going to confound quite a few people because they think that there's a trend going up. I'm about to, I'm getting close to finishing but here is China, again just all the time just unrelentless wealth inequality going up, particularly fast up to the financial crisis, not so fast now but still rising. India, where it's got flat, was rising up to 2008, now seems to be flat. Europe in general moving up fairly slowly but was going down so there's not much change really since the year 2000. Africa, one of the problems of course is that a lot of this billionaire data is there's not enough observations really to do individual countries, in those cases I've just done it, I've sort of put all the data together for Africa, used the billionaire data for Africa, adjusted the distribution and then given it back to the individual countries sort of separated it out. But there you are in Africa looks like since 2008 it is that inequalities increasing quite strongly. Canada's is one of the unusual cases where it seems to have gone down, that's a pattern quite common to lots of countries up to the year 2008 but it's still at least flat or perhaps even slightly decreasing since then. Australia again I told you that it was tracking up and that's exactly what's happening here, there's almost no trend at all in the Australian data. And that is actually the last slide, so I'm probably perfect in my timing. I don't know what generally I say, I think one of the implications here I think for if you're not interested in wealth inequality, clearly I think there's lots to be done here. And this is not just with the Forbes data because there's lots of countries which do have rich list data and you can use this and perhaps get longer series and do some interesting work. And this is really quite easy to do and so I could see quite a lot of interesting little studies for individual countries using these sort of techniques. For incomes I think one could do the same sort of thing, we do have data now on the top 1% incomes in lots of countries. We do have income distribution data, my guess is it would have the same problem that the top 1% is not, that you get from the top 1% database is not the numbers that you get from the survey type data. Can you match them together in the same sort of way and look at trends over time? I think that would really be interesting and really an important contribution. So I'm really quite excited about this, I think it opens up huge research possibilities and I have to say this is, there is not enough researchers certainly working in this field and I think particularly looking at trying to look across countries because one of the things we do try to do is get comparable data for different countries, otherwise when you're trying to compare different countries I think you're always rather dependent on the way that the variations in the individual countries can be bigger than the differences between countries. So this allows you I think to make much more of a sort of consistent comparable view of the world and how individual countries rank in their positions. Thank you very much. OK, thank you very much.