 I have a question for Haroun and a comment for Carlos. So Haroun, you seem to pay a lot of emphasis on the top part of the distribution, on public sector unions pulling wages up, as I understood it. Can you sort of make some more comments of that vis-a-vis the technical change component? So if you took away public sector workers from your data, and you only showed private sector behavior as reflecting technical change and all that, would you think that you have the same pattern, or is it mostly the public sector union part? And that's sort of relevant to try to get a little bit to Kunal's point. And then on Carlos, I don't know whether you can speculate a little bit on the following. So on your data, you have countries in which the minimum wage was raised very aggressively in the period you were looking at Brazil, particularly on others, and countries in which the minimum wage was not raised at all. Mexico was the one in which that didn't happen. But you have the following returns to education, the relative returns in both. So can you sort of, why was it in Mexico where the minimum wage didn't go up that a behavior similar as you've seen in other countries in which you could explain part of the wage compression through a minimum wage going up? It seems like it was a further falling in returns in Mexico in the upper part of the skilled people. So, to which it's. Thanks, Santiago. So the explanation we have, and I'll deal with the difference between the public and the private sector, but this is the way we like to think about the public sector union story is that they've been in bargaining in other work that we've done, they have aggressively pursued bringing the bottom up, right? So in other words, strong, so there's an increase in the floor. So the minimum wage in the public sector has gone up, mimicking the other sector minimum wages and have also sought out, but the public sector has responded because there's a scarce skill effect, right? With large increases for the top end, and the state then is forced to manage the wage bill, which they effectively do by hollowing out the middle. And that's, you see that for teachers, you see that for firemen and women and so on. But if you abstract from that, and your question is, but is it a public sector effect? I think what definitely what's going on, and if you look at the, I won't try and do the panel B graph again, but if you look at it carefully, and I've just double checked that the 80th percentile and above coefficient effect is positive for both public employment and the analytic tasks. And so I think what's going on is both stories. There's this public sector union that's representing the elite and then a skilled biased technical change story that's talking about tasks and rewarding analytical tasks and it's Conal's point about structural change. So you've got this services dominant economy that's driven by financial and business services, telecoms and analytical jobs then get this high premium. So I think it's both effects, but what you're suggesting is a really interesting thing is to rerun the whole thing, taking public sector workers out. And I think that's a great suggestion actually. Carlos. Yes, again from the Brazil-Mexico comparison from the case studies for what I remember, Ferreira shows with this brief approach that both minimum wages and the education premium were important for the reduction on which inequality in the bottom of distribution. The point of Mexico is certainly something that has been not done in this project, but from what we have seen is Mexico has an economic structure that is quite different to the rest of the Latin America in terms of their size of the manufacturing sector, which by some means is almost equal to the total value of the manufacturing sectors of the rest of Latin America. Some of them, I mean this paper of Gordon Hanson of why Mexico is not rich, he makes the points of Mexico producing what China also produces rather than what China consumes. And in that sense, Mexico didn't benefit in terms of the distribution of wages and earnings from this commodity boom. And what happened in terms of your question of what happened in the top of the earnings distribution, some of the hypothesis is also because this last period is confounded with the effects of the global financial crisis is that some of those jobs for high skill actually were destructed and some people have to go into lower, high skilled workers have to go into lower productivity jobs as sort of an adjustment. And the other really has to do with some sort of skills mismatch because some of the literature on education in Mexico shows that even thought there has been a higher years of education with respect to the past in terms of the total amount of education that the workforce has. I mean there is some evidence also that the firms that are coming to Mexico are not finding the specific skills they need. So there is an issue also with the education system in terms of providing the skills that the job market needs. Plus we know there is a lot of important issues in terms of factor misallocation that really doesn't match the higher skill with the higher productivity jobs and really creates that a lot of firms as shown by your work really stay as small and low productivity. But I think that's more or less what happens in the top of the distribution. So Harun a great paper and there's a lot in it too so I need to read it more to wrap my head around it. But two questions you know one in this RIF I'm always confused with the interaction term the contribution of the interaction term and in your several of those it's sort of kind of flattish so it's not very clear to me how that can be interpreted that's always my confusion about the interaction term in RIF. But more substantively you so the wage so there's one wage distribution story which is that the middle is being hollowed out and then you've reclassified occupation by task types and from your description it seems like there are many kinds of many tiers of occupation within a certain task. So domestic workers and teachers etc like all the on-site people you know but if you think of the wage distribution within this task type it's not people at similar wage levels. So I'm not exactly sure how your task type distribution maps out to the wage distribution so nicely. I'm confused by that. Yeah the interaction term could I mean it's I think it gets dominated by the flat endowments effect right which is close to zero and that probably picks up what you see as the interaction effect. It's a little bit like these interaction terms generally if you get a lot of zeros you're going to see a flat right and that's effectively what it's what it's picking up. So you write that the the tasks are distributed across occupations and so effectively you're seeing though the returns to the task right and so what should happen as with any multivariate story is that if there's a dominant analytical component to the task the returns will be to the task right. So it's a slightly different way so all the modeling that's done in this literature does that where the occupation gets replaced with a composite set of tasks if you like. So all it's suggesting is that if your job is thinking this way in the individual cell right where you have the individual if your task is 80 percent analytical right on average you're seeing a higher return I mean that's you know what I'm getting it right. So in essence that's the real innovation I think in the way that you think about task content jobs rather than pure occupations and it should come out in the wash right. So in other words if you see on average that analytical jobs analytical tasks within occupations are seeing a declining return and that's what the results will show. I just have a question for Haram and it's related with my previous experience in a contribution with the Honduras case where the incidence growth curves showing in Honduras at the decline in the bottom part of the distribution. So my explanation in that work was related to the inability of rural workers to move to modern sectors okay and this is basically because of language so ethnicity and probably very bad basic education so they were not able to to make easy mathematical calculations. So now this is what I call workers heterogeneity and I want to listen a little bit what are your thoughts on heterogeneity in the South African context because what I can try to see in your graphs is that probably workers in the middle of the distribution are not able to progress in occupations with rising pace so with increasing prices. At the bottom part of the distribution it seems to be clear that they have to accept what is going on and probably is there a demand driven problem for the middle part so that they cannot shift to other occupations because unemployment is high and they are trapped in their occupation that probably because of demand lost their premium. So I think you are right that there is all sorts of segmentation going on that the middle of the distribution is our sense is that it is a structural change story and it is a demand driven one and workers are heterogeneous and so that gives you the outcome that you see. The bottom end workers tend to be homogenous and so you would expect what you would expect is maybe the Honduras case or a decline in earnings so in other words I would have almost not been surprised if the bottom end also saw negative growth rates and then the top end is where you see all the action because it services the structural change argument but I think what it shows is you have an exogenous shock and to be careful with the paper because we don't have minimum wages on the right-hand side I don't show you any of the minimum wage data but you have had very, very aggressive real wage increases for minimum wage workers and I think that exogenous shock plays itself out in that outcome that you see. Looking at your paper that is Gary, the increase in labour earnings and I thought that perhaps we may need to interrogate it further and add on the story of job creation. The increase in earnings is it because of jobs that are highly paying that are being created or it's a matter of price setting that they existing and then they are paid more and then with Harron as looking at the missing middle what could be the policy implication because you tend to emphasize that it's mainly an issue of payment not so much of endowment but could it be that this middle class do not have the skills required in the job market that perhaps as a policy implication that it could be to do with the endowment as well. I think but I may be wrong that the increase in labour earnings is caused not so much by wage setting institutions as by supply and demand in labour markets. Unemployment rates are fairly low and that workers are being reallocated as employers bid for them but Carlos you're the expert in Latin America, do you agree with that? It was mostly I mean from the evidence in other studies that were done it's mostly because of a higher returns to low-skill workers particularly rural areas or associated to agricultural activities particularly in South America a lot of the effect came to that price effect. Okay but in answer to her question is that that's not because of unions of farm workers or something, it's supply and demand? Yes supply and demand and it will be from this exogenous shock presented as commodity boom. So just over the top of my head three policy options one is to improve the quality of the schooling system specifically a lot of the collapse in the middle of the distribution is because employers don't trust the certification either incomplete or complete in the high school system. I think that's one difference that may be true in the Latin American case that there were massive improvements and I think in the quality of the schooling system. So in the South African case that would be one clear policy option. The second is the entire TVET system is dysfunctional and the TVET system is the bedrock for the middle of the distribution workers. So the TVET system is the system that provides those semi-skilled workers and so even if we wanted to be a hub for like manufacturing in the region we just don't have the semi-skilled workforce to do that and the TVET sector is critical. And then the third is a lot of the middle of the distribution jobs come from innovative, well in the policy sense come from innovative labor friendly if you like industrial policy. South Africa's most, the biggest subsidy spend by industrial policies on the motor industry which is incredibly capital intensive creates very few jobs. You can imagine if that was diverted to plastics, chemicals, other light manufacturing jobs there'd be a big kick for middle of the distribution jobs.