 Welcome to session 3B. This particular session is about a project that UNI wider did, the title of the project is changing nature of work and inequality. Apologies, the title of this project is different from what's on the program itself. The project was done in a situation with IBS, also in DPRU-UCT, Cape Town, South Africa. I'm going to just give you a bit of an overview about what the project is about before we get to each of the presentations. So moving on to the next slide. So to give a background and motivation of this project, we know in recent years we've seen a sharp increase in weight inequality in many developed countries. While early literature had focused on the role of skilled bias technological change as BTC, as is widely known now, recent literature has been highlighting the importance of routine bias technological change, RBTC in short. RBTC leads to a depressed demand for workers involved in routine tasks that can be now executed by computer technologies. What's really interesting about this literature is that it puts occupational change at the forefront understanding the effect of technological change in wages or earnings inequality. But we also know, and this is the big limitation of the literature, that much of the existing literature you've seen has been focusing on developed countries, US, continental Europe, and so on. And we know very little about the effects of technological change and routine task intensity are turning the inequality in developing countries. That's a big gap in the literature. This is the reason why last year Univider, in association with DPRU and IBS, launched a cross-country project with the focus in countries in the global south. So this project, let me now turn to the main research questions for each of the country case studies. So we have three, four research questions. First is such question. What do we know about earnings inequality and occupational change in developing countries? Second is such question. What has been the role of routine bio-technological change in explaining changes in occupational structure and earnings equality? Third research question. What role have country-specific factors, minimum wages, trade unions, structural change, and so on, depending on the country in question, are played? The thing that was important about this project that we made sure that each country case study, and I'm going to hear a saver of the country case studies today, had to address each of these three core questions. So that meant the project comparative, so we can learn lessons about what's happening in the global south. In terms of countries that we studied in this project, we had four from Latin America, Argentina, you're going to hear about the Virginia case study later today, Brazil, Chile, and Peru. From Africa, we had three. South Africa, and we're here at the South Africa presentation later today. Indonesia, which is also there in the session, and Ghana. Asia, we had four country case studies, Bangladesh, which is also in the session today, China, India, which is also in the session today, and Indonesia. So you had 11 countries. An important point about the country's election was first, as you can see, we covered the global south, Latin America, Africa, and Asia. The other point that's important to keep in mind that we chose countries where most of the workers are outside agriculture, or at least are moving out of agriculture, because this is not really applicable for countries where most of that workforce are still in agriculture. And that's important to keep in mind why we chose a set of countries. You also had a global case study, which is also going to be presented today, Pieter Lewrenski, Albert Park, and Simone Scott. Now, before I ask the presenters to present their six papers in sequence, couple of logical parts. One, well, you probably know this by now, it's the third day of the conference. Each visitor has math, eight minutes, and also questions should be sent to me, or sent to all the presenters and myself to the chat function. I will collect the questions and then we're going to ask. I will ask the questions on your behalf. But the other point, which is a difference from previous sessions is, you would have the question and other sessions after three papers are presented. Three papers, a Q&A session, three more papers and a Q&A session. The reason we're doing it this way is because the papers, as you're going to see, have similar methodologies. And in a sense, it's better to bunch the papers because one, there may be questions asked out of more than one paper. We're going to have this three by three, three model. However, if there are any questions regarding clarifications, then I will make sure that I'll answer this question. We answer this question right away because we don't have any confusion on the methodologies staying on for the rest of the session. So apart from clarificatory questions, other questions will be kept after three papers are presented. Hope that's clear. The other point is to make this clear if the session is being recorded. Okay, let's start with the first of the six papers. And that's by Simone Scoti on the UNU Ida. Thank you very much for the introduction, Konal. I hope you can all hear me well. I think that's some echo. So the papers that I'm going to present is joint work with Pieter Lewandowski from the Institute for Structure Research in Wasser and Albert Park from the Hong Kong University of Science and Technology. And as the title suggests, we're interested in estimating the global distribution of protein and non-routine work in this paper. So as Konal has mentioned in his introduction, economists have been interested in analyzing the past content of occupations over the past two decades, more or less. And specifically having an interest in how the twin forces of technological change and globalization, specifically of shoring are shaping the nature of work around the world. Where the most of the literature so far has focused on the web as well as other industrialized countries. So as a methodological level, the most widely used measures of routine task intensity of occupations draw on ONET data. So ONET is a great database that has very useful information about occupational demands, abilities, skills required for occupations which allow researchers to have a good idea of the past content of an occupation. However, the big drawback is that it's only available for the US. So it's an US expert survey, essentially. Which means that structural differences between countries, particularly between low and low income countries on the one hand and high income countries on the other in terms of technology adoption, skill supply, infrastructure, productivity, et cetera, will not be accounted for when just using this data to have an understanding of the task of occupation in this poorer country, so to say. And there's an important question as to whether the same occupation has the same task structure in various countries, given these important differences. And that's something that we're trying to tackle on this paper. What we're doing is we draw on a companion paper by Piotra Albert and co-authors who have used very data from three surveys, which is the PIAC collected by the OECD, mainly for developed countries, though not only. The steps survey collected by the World Bank, mainly for low and low income countries. And the ULSS, which is the China Urban Labor Survey, which has questions that are very similar to PIAC and SPAC. And what they do in this companion paper is to start with the US PIAC and find a combination of questions that are very close to the measures of routine task intensity based on OECD. And this table, you kind of see those elements that I used to get an idea of the non-routine and routine task content of occupations, which are then combined into this routine task intensity measure, the RTI, which is higher than more routine intensive and occupations. So with this data using this survey measure for the 46 countries, we are able to get already an idea of how different occupations are in their task content for the set of countries, which is nice and a good addition. However, there's a number of large low income countries on the income countries, such as India, Brazil, South Africa, for which this data is still not available. So they do not allow us to get a fully global picture. Therefore, what we do in this paper is to estimate the range of models which relate the occupation-specific RTIs to the countries' level of economic development as measured by GDP per capita, technology as approximated by the share of internet users, skills supply measured by average years of schooling, and globalization measured by how narrowly the country is specified in terms in their global value chain contributions. So we use this model then, these models, to make artisanal predictions for those countries where we don't have any survey data yet. And in the scatterplots that you're seeing, the red dots are our predicted values essentially and these gray dots are the actual survey measures. And what you're seeing is that generally less developed countries are characterized by a higher RTI of jobs within the same occupation in comparison to high-income countries. And interestingly, this gradient with GDP is quite steep, especially for the high-income, high-skill occupation, such as here in the example of fiscal one managers, but also it could be professionals, for example. So what this implies is that a manager in, say, Ghana has, especially more, a repeat intent South portfolio than a manager in, say, Germany. And given this finding, what we're trying to tackle a second research question, which is to use our calculated and predicted values to get an idea of the evolution of RTI across countries across the 2000s. And to tackle this second question, we merge our RTI measures, the country specific RTI measures with ILO employment data for 87 countries, which together accounts for about 75% of global employment. And what we're able to show with this is that when using ONET, essentially, all countries are shifting away from routine to more non-routine work. So this is kind of this gray line here, which you're seeing where you have this declining RTI across countries. However, importantly, once we account for the fact that especially there's high-skill occupation, a substantially more routine intent in the low-income countries compared to the high-income countries, we see that the routine tough intensity over the 2000s has essentially been almost flat in the low-income countries, which implies that this relocation of labor from less productive routine towards more productive non-routine work has occurred much slower there and has been quite far especially in the high-income countries, leading to a rising gap in the routine tough content of occupation. So it's actually not narrowing, but rather widening. And this would be essentially overlooked when using the ONET measure, assuming the tasks are the same across countries. And from this, we conclude that using the ONET essentially overestimates the removal opportunities, replacing technological change and kind of calls for increasing technology adoption, raising skills of buying these countries. So let me close here. And as Kunar was saying, the other countries' studies are going to dig deeper into some of the country's cases, also using our country-specific RTI measure as well as the ONET measures. Thank you. Thanks, Mouri. Let's move on to the first of the five country case studies that's on Tunisia and Spongleimid. Spongleimid, the floor is yours. Go ahead. Hi. So can you hear me? Okay. Hi, thank you for the introduction. So I'm Spongleimid and I'm glad to present our work on Earnings in Equality and Changing Job Mages in Tunisia. This paper is called, sorry, by Mohamed Ali Marwani, Mission Mashalian and me. And next slide, please. Thank you. So in overall, we observed a steady decline in earnings inequality in Tunisia and labor market over the last two decades. So at first glance, the aggregate data suggests that the potential determinants of inequality change work in opposite direction. But precisely, in one hand, the decline in education premium associated, especially education premium associated with high earnings jobs, with high earning jobs and the increase of the average RTI are in line with the reductions of inequality. But in another hand, the changes in occupation distribution appears to increase inequality before the revolution from 2000 to 2010, and then it shows an ambiguous impact after the revolution. So when we performed the polarization test, we failed to find a jeopardization but an L-shaped patterns of earnings revolution so that this is safe. Locked earnings increase at the lower end of the earnings distribution, but stagnated at the upper end of the distribution. Since the public sectors, including public administration and public enterprise has critical role in the Tunisian economies, we try removing the public sectors from our dataset. And interestingly, we find a job prioritization in the private sector and also which prioritization. So together with the increase in average RTI over time, we suggest that the other driving forces might have greater impact on the earning distribution than the routinization. So next slide, please. And here to measure the impact of other, like of all determinations on the earning distribution of Tunisia labor market, we decompose the changes in earnings distribution into the contributions of main determinants using the recenter influence functions. So you can see here during the first period from the 2000 to 2010, the public sectors play the main role in reducing inequality, especially in the 50 times wage gap. And also the decline of education premium relating to the high income jobs also helped to reduce inequality at the upper end. We can see here in the first period, occupations or routinization has a disequalizing effect, but it was counteracted by the education and public sectors. And during the second period, it's after the revolution, although the public sectors still absorbed around 20 to 22% of the active population and wages in the public sector were held at a higher level than the private sectors. It had no equalizing impact on the earning distribution and both routinization and education increased inequality of the 50 times wage gap. And here we can see that from 2010 and to 2017, most of the decline in inequality in Tunisia comes from the improvement of gender equality. So next slide please. Thank you. So our main findings here is that the dynamics of Tunisia labor market are characterized by the four main factors. The first one is the decreasing earnings inequality. The second is the L-shaped wage polarization and then the increase in the share of high skewed jobs until the revolution and then it's decreased. And finally the same pattern with the average RTI. And what we found is that this trend were mostly explained by the form of education premier employment and which policy in the public sector and vitrination. And finally, one thing we didn't highlight here is that gender equality improved in Tunisia. So that's all, thank you for listening. Thank you for mentioning that you're doing a PhD in the University of Sorbonne, in the Spanish on Sorbonne, thanks. You're very good. So we hope to let you move on to the next presentation. That's by Saima Pidisha, who's at University of Dhaka. Saima, go ahead. Good afternoon or good evening. My paper is about the changing nature of work and inequality, the case of Bangladesh. Basically, if you look at the cases of economic indicators for Bangladesh, Bangladesh has been doing quite well in terms of a number of indicators. Especially if you look at the GDP for the last five years or so, the country has been able to have a GDP of more than 6.5%, which was quite impressive. However, if you look at some other indicators and there remains concerns, and one concern area is that of labor market because it is often argued that the country's growth experience has not been translated into the labor market outcome. And there remains bottlenecks in the both supply side and the demand side. Based on this background, in our paper, actually we tried to analyze some of the selected labor market indicators, and for that, we have used 2005, 2010, and 2016-17 labor force survey of data and also combine it with the own aid data. The methodology is similar to the other paper, so I'm not going into detail. We have found in terms of the result, there has been some interesting results. For example, in terms of the inequality, although we have adopted a number of indicators, although not very consistent, but in terms of Gini indicator, what we have seen that from 2005 to 2010, it's remain almost same, but in the later years, there has been a falling shred. Now, if you look at the other indicators in the data set, then we have what we have seen that there has been the concentration of worker in the meat scale occupation and in the secondary occupation, and over 10, there has been an increase in the meat scale occupation of people as well. And the tertiary level occupation, although they are enjoying a high education premium, their proportion is very small. So from that point of view, the result seems to be more or less consistent with the other indicators. Now, in terms of the share of workers, as I have already mentioned, that for the tertiary and the secondary educated workers over time, in recent years, there has been a decline. In terms of the share of workers in the different scale component, in the first stage of our analysis, we have seen sort of an indication of job polarization because there has been a reduction in the meat scale occupation people. But in the latest stage of the analysis, it seems that there has been a shift from the low scale to the meat scale. So over time, on an average, we can say there has been an increase in the meat scale and the high scale people. When we look at the earnings, then we have seen that across more or less all of the education classes, we have seen there has been an increase in earning, in real earning, and when we look at the education premium, for almost all of the classes, they have enjoyed education premium, but the education premium was the cheapest for those who have tertiary level education. In terms of the polarization, if you look at it, then we've seen that in the first stage, we don't see employment polarization, but in the second part, we see some indication of polarization. So on an average, while combining our results with some of the graphical analysis which I am not showing here, we have seen that there is no, we cannot come to the conclusion that there is any polarization. So there is no, there has not been any employment polarization over time. In terms of country RTI, we have seen a fall. So we can say that over time, there has been a decline in the share of occupation which involves more routine stuff. In the next stage of our analysis, next slide please. We have applied to decomposition method, shafted decomposition method and RIS decomposition method to dig into more detail into the inequality. Shafted decomposition, which I'm not showing it here, that has in fact shown that in the first part of our analysis, so 2005 to 2010, there has been the within occupation inequality was more prominent. And in the second part, the between occupation became more stronger. RIS decomposition, which in fact looking at different factors, contribution to the inequality, that shows that the importance of routine task intensity, RTI and education, that has been the more prominent factors. But the interesting thing is in the first stage of our analysis, routine task intensity sort of having a pro-rich effect. On the other hand, the education, if you look at the different quintiles distribution, then we can see that education sort of have a pro-poor effect. But when we look in 2010 to 2016-17, then the results in fact, the effect of education is not that prominent, rather we see that RTI is playing the key role. And on an average, maybe we can say that the role of RTI is here, we see that it is more pro-poor. So combining these two things, the inequality can be, we can say in the context of Bangladesh, the role of RTI has become more and more important in explaining the unique differences across occupation. So if we go into the next slide, then we can combine our result and we can say that over time in the context of Bangladesh, we observed shifts towards more educated and better skilled workers. Returns to education seems to have increased over time and that is more prominent when we look into those who have tertiary education. There has been a shift towards education jobs which involve mid-skill and high-skill jobs and also those jobs which is less routine intensive and which is more analytical and more cognitive in nature. In terms of inequality, it has declined when we look at the labor earnings, but we must keep in mind that there are also some other various factors like the institutional factors and the issue of tax-to-GDP ratio. This has to also be taken into account when we think about the overall inequality in the country. As for the implication, there are like two points I would like to highlight is because we have seen the education premium for the tertiary educated worker is so high. That's why the education policies and the labor market policies, they should be strongly integrated and there is a skill mismatch and all of these factors which are concerning our labor market, that should be keeping into account while designing the labor market policy. And secondly, because the reduction of the decline in the share of routine intensive care, the more training and training program has to be designed involving more tasks which involve more cognitive skill and less routine intensive care. So I will stop here. Thank you very much. Thanks, Patricia. That is very nice because you're all very much in time. I want to see any questions in the chat yet. So please send the questions as you start thinking about them. Let me ask a question to Pongli. So Pongli, why do you think the revolution plays such an important role in the nature of what you see in terms of earnings and employment? What was the nature of the structural break that happened in the revolution? Yeah, actually, the public sector, they occupy a large part of the active population in Tunisia. Before the revolution, as our data suggests, in 2000, the public sector occupied like 45% of the active population and that's a very large part. And actually, in the public sector, the wage policy in the public sector is always higher than the wage policy in the private sector. So before the revolution, it really has a desequalizing effect. However, after the revolution, the public sector, actually, it's reduced like most of the reduction comes from the privatization. So the public sector occupies just 20, 25, and almost of them are the public administration. And also in the second period, also wage in the public sector will keep higher than in the private sector. It has no desequalizing, it has no effect on the earnings distribution. That is to say, the average wage in the public sector is higher than average wage in the private sector, but it has no effect on the distribution at all. So mostly, it comes from the wage policy of the public sector. Thank you, that was a very comprehensive answer. I don't see any chat question in the chat yet, but I think we can carry on to the next set of presentations. And for those who have questions that you're thinking about for the first three presenters, you can always keep on sending them, and we can take them at the end of the presentations. So let's move on. The next presentation is by Kanika Mahajan from Mashoka University on India. Thanks, Kunal. Good afternoon to everyone, and thanks for coming to this session. So the focus of this presentation is going to be India and the changing wage inequality in India and the effect of changing the occupation structure and the routine tasks intensity in the country in affecting the direction in which we see the wage inequality to be moving. Next slide, please. So just to set the context, here we give a brief overview of the Indian economy. We have the base values for 2004, since they're looking at changes in these macroeconomic variables since 2000 for this analysis. Overall, we find that GDP growth has been robust for the country since 2000. There has actually been a decline in wage inequality over time, and this decline has been larger. This red cells that we see reflect the decline in wage inequality over time, which has become steeper since 2011, the main variable of interest in this analysis. The urban workforce structure has changed towards more of construction-based activity and reduction in agricultural-based activity. There has been an increase in education and a steep increase, especially in tertiary education. However, this has been accompanied by a decline in the education premium. On the other hand, if one looks at job polarization, we find that there has been intense job polarization in urban labor markets of the country where the middle sector, where the mid-skill jobs have shown a sharp decline and the high-skill and the low-skill jobs have shown a sharp increase since 2000. There has also been a substantial decline in the routine intensity of the jobs, and we see that especially in the decade of 2004 to 2011, there was a steep decline in the routine task intensity, whether we look at the country-specific or the ONET measures to measure RTI, whereas between since 2011, these routine task intensities have been pretty stable. Next slide, please. Next slide, please. Thank you. Here we decompose the changing wage inequality which we have seen has fallen over the last two decades in the country. We find that there has been a steep decline since 2011, and here we break it down into the part which is explained by the changing factors like age, education, religion, past structure, as well as the routine task intensity of these jobs. We divide our analysis into two periods. Our main finding is that the changing composition of the workforce due to these factors has not really resulted in any change or decline in wage inequality that we see in our main analysis. What has really contributed to the decline in wage inequality especially since 2011 has been the changing return to these characteristics. So what we do is that we plot the changing returns by quantize, and we show for each quantize what has been the contribution of each of these characteristics towards the change in earnings for that particular quantize. If we see a negative effect here, it's basically saying that that factor has contributed towards a decline in wages for that particular quantize. So here we clearly see, for example, for 2004 to 2011, that education has led to a decline in wages for the mid-level jobs, middle-learning jobs, whereas it's increased the earnings at the high end and at the lowest end for once. And overall, we'll see that it has had a disequalizing effect because of this large negative effect in the middle part of the wage distribution. This pattern continues in 2011 to 2017 as well, with this large declining return, like the return to education playing a role in decrease in earnings in the middle part of the distribution, whereas there's an increase in earnings towards the higher end of the distribution. The routine task intensity, on the other hand, has had the opposite effect. So what's fun is that in the middle part of the earnings distribution, the routine task intensity has led to an increase in earnings in both the time periods, that's having an equalizing effect on the wage inequality structure. Next slide, please. So overall, we have seen that wage inequality has fallen in India and the major factor responsible for this has been a change in returns to attributes over time, especially returns to education, which has been so rich, whereas the returns to RTI, which has been so poor. However, one must note that economic factors are not the only factors which affect wage formation in the country. Institutional factors are equally important. And our primary findings are indicating that the changing structure of minimum wages in the country, especially since the thousands where the minimum wage increases have been large, have played an important role, are playing an important role in shaping the wage inequality in India. So I, so thank you. I would like to stop here. Thank you, Kanika. And I'll move to the fifth presentation and the fourth case study by Roxana Mauricio, who's at the University of Buenos Aires on Argentina. Can you hear me? Yes, you can. Okay, thank you, Kunal. Good afternoon, everybody, good morning. I will present the case of Argentina, where we study the changes in employment in the health tasks and composition and their impact on new quality. Next, please. The period under analysis is particularly important because over the new millennial, Argentina experienced two very different micro-economical labor market cycles. In particular, during the south side of the three, the south side of the four, a strong activity growth was accompanied by a reducing trend in the quality, resulting in an important improvement in the living condition of the population. In addition, following a long-standing trend, the arching time was forced and became more skilled, but alike in the 90s, we tend to education-friendly during this period. In contrast, with the job and polarization of health policies, there was a reallocation for a lot and to a lesser extent, for high-to-near skilled occupations. However, even when these patterns seemed to be consistent with an inverted and blue-shaped set of five, they were not a large enough to be reflected in statistically significant results. In contrast, the changes in earnings did not follow the same pattern on those in jobs. In the context of increasing real wages and strengthening of minimum wage and collecting bargaining, the unearning inverted use-shaped profiles were verified. Finally, as a consequence of the reduction in the share of a low-stakeholder, there was a decreasing trend in the proper and low-stakeholder content measured by using ONET or a country-specific RTI. During the second sub-period, several opposite trends were observed, in particular, in the context of the low-economic binarism and increasing macroeconomic stability and the quality increase. Education upgrades continued, but unlike the first period, now it was verified together with a moderate increase in the education period. Occupational changes were less clear than in the first period, but again, they were statistically non-significant. In addition, if the widespread reduction in real wages as a consequence of the acceleration of inflation was polarized in the sub-period. Overall, we can see that a dense stable macroeconomic performance over the new millennium was accompanied by a slight reduction in inequality, education upgrading, changes in occupations, apparently consisting with an inverted reshaped profile as we can see this year, but statistically non-significant. A weak inverted reshaped behavior in elements and a reducing trend in the top low-economic intensity. Next, please. Therefore, considering all these changes over the two different inequality trends, we wanted to evaluate to what extent the income distribution dynamic is being explained by changes in the social low-economic content among other factors. To do this, we carry out an aggregate and a decrease in human composition. As we can see in the table above, most parts of the changes in the human composition was explained by the aggregate return to different values. However, when we look at the individual contribution of different types of human attributes, we can observe that some changes in the composition of employment were also important here, in particular, during the first period, the reduction of the chemical efficient was driven by the equalizing impacts of the labor formalization process and by the shift of water towards less routine in terms of occupations. The reduction in the RTI continues to be in equalizing, although with a less intensity during the second half period. However, the increasing gender weight gap during the first half period and the widening in the formality weight gap during the second one were unequalizing. Next, please. To conclude, in Argentina, we did not find econometric results in occupations. We needed support in polarization and not invert it in U-shaped patterns. Nevertheless, we did find significant changes in the environment, in particular, where it increased in low-paying jobs where they labeled the months fell along the whole period. These results are different from those observed in several countries, especially developed countries. So the question that arises here is what are the potential parts of the time-explaning response? First, stroke microeconomic instability together with significant changes in the production structure that can difficult the adoption and spread of technology in these countries. Second, maybe the influence of labor institutions, such as minimum wage, collective bargaining, explaining the local relation between employment and illness, in particular, especially at the bottom part of the income distribution. Maybe we are observing a ongoing process which is full realization for the longer periods of time, especially considering the initial conditions regarding the composition of employment, the skill of the workforce, the position of occupations, different occupations along the wage distribution, the spread, the spread of the story of technology penetration, among other factors. So finally, after the reaction of inequality during the first decade of the human union, the current situation is lowering because inequality continues to be very high in Argentina and even more important, those factors that contributed to the improvement in income distribution during the first year, the first period of the new millennium weekend of reverse is during the last year. Thank you so much. Roxana, thank you so much. So we can move to the final presentation by Amy Thornton on South Africa. Hi. Hi, can you hear me? I don't know why my video isn't working. Yes, you can hear you well. Okay. Yeah, sorry about the video. Okay, hi. My name is Amy Thornton. I am a researcher at the Developmental Research Unit in the School of Economics at the University of Cape Town. And what I'm presenting to you is joint work with the Human Rights, Caregiver, and the team and the point where they've been also at the Developmental Research Unit. So thank you very much for the opportunity to do this. Yes. So Canal has been a certainly nice introduction to this project. So essentially what we're doing is investigating inequality and stuff, but by targeting these changes in the earnings and employment and using made market data for 2020 and our focus is on task application and gender. And next slide, please. Right, so South Africa is an interesting case study in terms of inequality because we have some of the highest levels of inequality in the world. And this is largely driven by the labor market. So our wage genie is at something like 0.55, which is extremely high at the beginning of the period of study. And it's really like state at that really high level throughout and even started to increase both end of the period where there are some questions about gender equality. And generally, a lot of work has been done on inequality in South Africa and there is some consensus that returns in terms of wages have really accrued to that top detail and investigation is about why that is the case. And a key characteristic of both job and wage growth over this period is that it has been sealed by us. So if you look at the figure here, what I'm showing here is an annual average growth rate for wages across the earnings distribution and between 2000 and 2015 for men and women. And you can see that it is quite U-shaped for women but a bit more monotonic for men and increasing for men. And you can also see employment change there or across skill categories for men and women. And you can see then both of these tasks for wages you can see returns accruing to the top end. And in terms of employment, a lot of the job growth has been for high-skilled occupations. And so for us, that is managers and professionals. Now, behind these important changes are these key changes that we are investigating there being some important changes in terms of the schooling sector and in terms of sector. So the main changes here in terms of the wage structure, what we think is the most important, one of the most important factors are is the changes in the schooling sector. So over this same period of time, there's been a large expansion in people with a high school certificate. But even so, people who have tertiary skills remain in relatively short supply. So even though I've just told you that job and wage growth has been pretty biased, skills are still relatively safe. Plus in addition, there is some work suggesting that the skills that people are graduating with from a high school, there's been a delinking or deterioration of how many skills are really signaled by that certificate. So that's an important wage structure effect. In terms of employment, a really key change has been the economy undergoing a process of premature deindustrialization. What this means is that the beginning of the period, agriculture, mining and manufacturing were really important contributors to GDP and really important for sourcing up a lot of mid and low-skilled labor. However, over the period, we have seen a decline in these primary secondary sectors over a variety of historical and context-based reasons. And instead, manufacturing has never sort of risen to be the engine of the economy that we thought it might be. And instead, the service sector has really risen in its place to be the engine of job and GDP growth. This is quite important. So over this period, we've seen something like 90% of new job growth has been accounted for by the services sector. So what we're trying to do in the paper is really understand this figure 1A. What is going on with these wages? So we set up an earnings model to look like all the other papers have and try and understand the relative contribution of different key explanations for wage growth in South Africa. And in doing this, we think about some key explanations. So there are four key explanations that we think about. These are skills by aesthetic change related to the student changes I've just told you about. These are sectoral shifts related to the financial deindustrialization I've just told you about. Some important labor market institutional effects. So South Africa has a large public sector. We have a strong high-level unionization. And we also have a schedule of minimum wages. So we also think about that. And then lastly, we also think about routine by aesthetic change and next slide please. So this figure constitutes the main result of that piece of analysis of what you're looking at is the detailed decomposition of the relative contribution of these different explanations to wage growth over the main periods. We have a pool based period 2000 to 2004 versus a pooled end period 2013 to 2015 for men and women. And what you can see is that skills or education as measured by education have remained relevant across that distribution, really important. We can see that returns to occupations for men has become quite important at the top end. I think that's something that requires more impacting in general. And for women, you can see that returns to industry are quite important at the bottom end. And what we actually think this is referring to is minimum wages. So in terms of that U-shape that you saw for women, we actually think that that bottom end of the U-shape is largely related to a large portion of domestic workers in minimum wage covered in South Africa. So those are sort of later stated real increases in that minimum wage. And then if you are struggling to see the contribution of our own at RTI, it is because it is really, really small. So our conclusion from this work is that we see this routine task intensity as contributing sort of a supporting role. We don't think it's been a major determinant of wage changes over the period, but it certainly is related to, for example, collapsed in manufacturing on a more global scale. But locally, there are some more compelling reasons of historic and context-intensive reasons about why this has happened. However, some closer analysis has suggested that routine work is probably quite an important factor for office clerks, which is a vocation mainly occupied by women. Next slide please. So a key change that I've described to you is that South Africa has really been shifting towards a service-led economy. Now, typically, people think about this type of movement is associated with economy being richer, but what we want to caution based on the analysis on this paper is that this might not necessarily be the case. So the move to services has certainly benefited the task because people at the task are the people who... The highly educated people who are in relative short supply have the types of skills that are required to perform the jobs in the financial and business services sector, in particular, which is really an important sector in our economy. At the same time, this type of move has undermined traditional jobs in the bottom and middle of the economy. Well, they've undermined traditional jobs that low and mid-skill people have relied on, in particular manufacturing, for example. And we've concluded what I've just told you is that we see routine work in the supporting role. However, we think going forward, it's no coincidence that a lot of job growth in middle and low-skill jobs have been jobs that are non-routine and which need to be performed on site. So, particularly, we see enormous growth for women in terms of care work and for men in terms of security guard work. These are jobs that are non-routine and which need to be on site. You can't offshore these jobs and get a computer to perform them. So going forward, we think an important insight that's come out of this work is that this move to the services sector might actually replicate and reinforce the existing patterns of inequality. Thank you very much. Amy, that's a conclusive presentation. So I have a couple of questions that have come through chat. Maybe others can keep on sending questions. So let me ask these two questions and they might actually be applying to more than one panelist. So one question is about the fact that I think some papers found that the education premier has been declining. And the question is that, what does, is that a good thing or not? So are we, should we be pleased that the education premier is declining or should we not be pleased? I think, Vidisha, do you want to answer that? Because you find that in your work. Tanika too. Vidisha, you want to go first? Hi, yeah. So let me first answer this question. So in the case of Tunisia, we find that the education premium reduced. But personally, I think that this is all, maybe education premium reductions, it's reduced inequality, but it's going to keep like the economy at the lower equilibrium. I mean, it's reduced the income of the upper end and it keeps the lower end still low. So it's back for, personally, I think the reductions of education premium, it's not a good thing to do. We should try to improve the equality in another way rather than reduce the overall education premium. Vidisha, do you want to reflect on this question? Okay. In case of Bangladesh, in fact, we have found the opposite results. So education premium has been increasing over time. And especially when we are looking at those involving tertiary education, they have the steepest rise over time. And secondary education also for both of the sectors we have seen a rise. So that seems to be more consistent with the result that we have found in our inequality story and wage premium story as well. So education premium increase, that's why we have highlighted that we should focus more on bringing people to more in the tertiary and high-skill position because there is a, because we have very few people who are in the tertiary level education. So although education premium has increased, the thing is like, because there are very few people in the tertiary education, unless we don't increase those people who are in the tertiary education, we will not be able to enjoy the benefit of education premium that the country is offering. So that's why focusing more on education it should be one of the priorities. Kanika, did you want to respond about the question of the education premium? Yeah. So I think that it depends on the factors that are leading to the change in education premium. So for example, if the education premium is declining because there are skill mismatches at the upper end, then I would say that the declining education premium is bad, but the education premium could also decline because there has been an increase at the lower end, which could be because of changing institutional factors like rising minimum wages. In that case, I don't think it's a bad sign. I think the fact that you're giving decent wages to people may not necessarily be a bad fact. It may just be an equality-enhancing factor, in my opinion. So that's all. Thanks Kanika. There is a very specific question for Amy from Pieter Lieronski. Amy, let me read the other question. It's a fairly long question. So the question that Pieter asked was that, I'm a bit puzzled that she controls for occupations and own it RTI at the same time because RTI is defined at the occupation level. The contribution of RTI is probably the contribution of changing the task content of occupations over time only. Occupation changes captured by occupation dummies. We don't know the character of this change. Why are you doing it this way? Hi, can you hear me? Yes, so I thought about this a lot and I'm glad that we can talk about it because I wasn't sure. So I find that occupations contribute a lot to any integrations in general in Alaska and that even when you run an OV and you get that total effect change over time, you actually can get quite a different effect if you don't include occupations. So I think excluding occupations is quite a big problem. At the same time, I understand what you're saying about RTI being defined at the occupational level, but in this instance, occupations here are at the one-digit level where they've been included in this regression, but the only RTI that we include here is at the four-digit level. So it is actually much more detailed. So I think that it is certainly fine that we can be including them at the same time because if the RTI is ready to make a contribution, it should be making a contribution over and above controlling for an occupation. But I'm happy to think further about that. It was my intuition to put it in because I could see how important those occupational dummies were. I don't know what you think about that. Peter, do you want to actually just follow up and you can ask your question, your comments live if you want to. Right, so let me unmute, okay? Well, right, occupations will be obviously important, but if you do it this way, then the RTI, the effect of RTI will only, it actually captures the differences in the task content between various narrowly defined occupations within a given one-digit group. And you average all the differences between the one-digit group as they are captured by the dummies. So let's say the effect of RTI is that the composition of narrowly defined types of managers among all the managers changes. So it's a very different, let's say, meaning of the routinization of jobs than what we do, what we find in the literature in general, because in the literature in general, what you capture is increasing share of managers and the declining share of plan and machine operators. So I think that changes the interpretation of the contribution of RTI quite substantial. So Amy, do you want to, I know that Harun is also listening in. Harun, if you wanted to jump in, please do. I can contribute, but let me have first have a go of, of what's that Amy, my views, so let me read this. Okay, all right, Amy, go ahead. Sure, no, I take Koda's point. I think he is raising a good point. I think at the time I was just disturbed by how my total change, my total OB effect was changing. And I was worried that it was a large amount of variable. I must say that I have run the regression without occupation and that I do get a larger effect for RTI. So maybe this is something we can think about. Let's move on to a couple of more questions. One is on the role of trade unions. Now some of the business should talk about it explicitly for example, South Africa and Argentina to some extent on collective bargaining, but it was not really discussed in the ones from Asia, Bangladesh and India in particular. Is it the case that this is because unions are not important in explaining earnings inequality or is it the omitted variable problem? I mean, the data problem, you don't have data units at the level you want. So maybe Kanika, you want to go first? So I think it's sort of both. A, we don't have data employment surveys, won't capture whether a person is a part of a trade union or not, but over time the importance of trade unions has been documented to be falling in India, especially post-19. So it's both, at least for the case of India. Yeah, and the problem we also don't know, separation formal and informal workers in your data set. If you were to result in formal workers are put in by, or have union representation, right? So that's another problem. Yes, there will be maybe overcoming it, but all of them are noisy ways. There is no cut identification between formal and workers. Okay. Okay. Disha, did you want to reflect on this? The thing is in the context of Bangladesh, a union is more or less concentrated in some sectors, like RMG and some of the relatively organized sector, but many of the sectors, like leather sectors and a lot of these sectors, which are quite important, but those are not unionized. They don't have proper unions. So that's why RMG and this type of sectors, those are concentrated in crafts and trades. That is co-digit. So other than that, in other sectors, like in the high-skill sectors, we have some, obviously we have some unions, but on an average, the effect of union, in order to capture the effect of the union, we should, it should be the case that union is more or less, many of the sectors are unionized, but we don't have that. That's why the effect of union at this moment, it is, we don't find, it is not possible for us to dig into greater detail, but if you look at the Shapley decomposition, union is one of the institutional factors. So that should become within group inequality. And if you think about from that point of view, over time we have seen the within group inequality was dominant in the first part of our analysis. So before 2010, but in the later part of analysis, in recent years, the importance of within group factors became less important and between group occupation differences became important. So from that point of view, I can say that probably the importance of union is still not that prominent in the context of Bangladesh. Thank you. I have a question from Timothy Schipp, which is, I think for all of you, maybe even for Simone, actually. The question that Timothy Schipp asked is that, given that the studies point to how different each country's experience have been, do these studies point to any common or shared experiences that we might think of as a general trend? Similarly, are there any implications that are more general in nature, such as a shared impact of globalization or technological change? So how would you reflect on what you've heard from each other's presentations and obviously from your own work, especially in a lot of the literature that we've heard about a lot for the West, that the RBTC literature you've seen has been applied a lot to developed countries? How would you see the findings that you've seen, both in your work and others' work, are they common trends, or if they are, are they different from what we have seen in the US and the UK and so on? Who wants to answer this question first? Simone, do you want to try it or give it a shot? It's starting from our global study. As I was showing in the last slide, what we see when accounting for the fact that occupations tend to be so different in their tough content in these low and low income countries, what we actually find is that the shift away from routine work has been way less than one would suspect. If we see the patterns in high income countries and just use this type of tough patterns on the other time, those to the low and middle income countries, we would think, okay, we have this kind of global shift away from routine work, from routine work towards non-routine work. And what we find in this paper is that this is not really actually happening. And one potential explanation is, in the sense globalization or occurring, which kind of what it gives support to the hypothesis that especially the poorer countries are actually specializing in this more type of routine work, increasing the gap, as I was saying. And also there's still way more technological change needed and technology adoption to actually close this gap between countries. So this is what this kind of global picture that we are painting is showing. And beyond that, also maybe talking about this country study on Ghana that we have been doing as you mentioned in the introduction, which hasn't been presented here, it generally finds a rather little role played by RTI. So kind of this more standard explanation, like for example, the change in the education premium played a much larger role than explaining inequality patterns in most of these countries in this kind of routine perspective. It doesn't necessarily mean that this globalization or technological change didn't play a role, but they rather work maybe through those channels. So it's still rather a good biotechnology change maybe than necessarily routine biotechnology change. Thanks. Does anybody else want to think about the big picture question? What are we learning from this project in terms of the overall lessons from the Global South for the literature on routine biotechnology change? Roksana, do you want to think about it? Yes, following Simon, I think that for instance, in the case of Argentina it's interesting to analyze the particular impacts of technological change because I already said the initial conditions and the characteristics of the labor market and the macroeconomic situation are different from, for instance, from European countries. For instance, when we look at the position, the ranking of different jobs, different occupations along the whole way to distribution are different. So even when we can observe similar changes in each of these occupations, the overall picture will be different. Maybe different from the polarization of hypothesis, maybe different from the state behavior. So I think that is interesting to analyze specific cases because we have especially in developing countries different conditions to the spread of and the penetration of the technological changes are different, but also the conditions in the labor market and the microeconomic situation that can shape a difference or can generate a difference in terms of the adoption of technology and the impact of this technology on the labor market and in the distribution. Thanks, Lux. I should also mention that we have several other countries to study. So we're not really seeing the complete picture in this set of presentations. So that should be also what keeping in mind. I was wondering, Amy, there's a question now that I think, do you want to again respond back to what Peter said, because I think it's important to get this methodological question right because it also has an impact for others in this project. Did you want to come back again on that? Sure, yeah, I've no, in my mind, I think that there are both important to be there or at least you should run specifications with both of them. Well, we should run both specifications with them with our occupations. I understand the point that RTIs have been operationalized at this occupational level, but essentially you can get the same RTI across different level, like across different one-digit occupations. Plus I don't think the one-digit occupations are that narrow, they're pretty chunky. So I think we should at least run both specifications because the RTI is like this fine-grained measurement whereas the occupational dummy is made. It should be capturing something a little bit different. I mean, occupation is measuring occupation whereas the RTI is measuring toss intensity. So I mean, I would, I think my position is that when you just think extra hard about it maybe and at least run both specifications and think about what it means when that RTI coefficient changes with and without that occupation variable. Yeah, thanks. And did you want to also reflect on this question of what are we learning from the work that you've done in South Africa on the big literature on RBTC and its implications for the global South? Do you see that literature has less relevance to South Africa? Do you think, as Simone was saying, that there are some parts which are relevant and perhaps some parts which are not so relevant? Sure. I definitely think it's an important tool in our toolbox. When I think about, for example, the premature de-industrialization that's happening in South Africa, for example, I think it's very tempting to be, oh, we're manufacturing very routine sector. It's obviously declined because this is very routine. But really when you take a closer look at South Africa and South Africa's history, that's not necessarily the case. So I think for me going forward, it potentially will become more important because of what I said in that final slide about how the job growth that we are seeing has a particular affiliation in terms of not being, in terms of being non-routine and if you can't offshore these kinds of jobs. I think it will become increasingly important even if it hasn't necessarily been important up till now. And I think it's not that it's unimportant, it's just that there are some other, well, there's a lot of diversity in developing markets in particular in terms of government intervention in the live market, which in South Africa has been really important. Yeah, thanks. I was just thinking that, since I think this is a very important question that Tim is asking us. Peter, did you want to also reflect on this because you're also part of this project, IBS is part of this project, and maybe Haroun also. So speaking through what you see from the case studies, not just South Africa, the other country case studies. Peter, do you want to just reflect a little bit on what you're seeing from the case studies so far on the big picture question? Yeah, I'll try. I think one finding which is here is that the findings are a little bit different than what you find for the OECD countries. And it seems to me that these drivers of inequality are less related to technological adoption and globalization a little bit more to what is happening with the skill structure. And that may mean that these low and middle income countries are lagging behind the OECD countries by let's say 20 or 30 years in that dynamic. That may also be related to the fact that the OECD countries, you don't really see that much change in the structure of skill supply anymore as the enrollment in education are quite high and stable in the last, let's say, 20 or 30 years, whereas these countries record the increase in skill supply that plays a larger role. And the other factor which probably explained the difference is that the OECD countries, let's say you have this auto mining salamones paper, who's mining salamones paper, I'm sorry, that argued that there is a polarization of labor market in the OECD countries that is driven partly by technology adoption, partly by offshoring. But here we're actually looking at countries that receive these offshored jobs. So they are on the other side of that equation. So global values chains are not really driving the inequality here to the same extent because this middle skill, the low skill jobs are not being sent somewhere, as is the case in the high income countries. So I think that is an important lesson because the literature on globalization and technological change tends to treat conclusions from studies done for the US or the UK as universal lessons. So I think that's something we can highlight as one of the findings of the project that it's maybe not the case. Very point, I think, Peter. Harun, did you want to also reflect whether you have over three minutes or so left in the session on what you see as an emerging findings from this case study that you heard today? Yeah, I mean, I'd probably need to look at that more, look at them more carefully, you know, but I mean, my sense is, this is, I mean, certainly feeding off what Piotre said, there is a story about emerging markets or developing countries in terms of the various drivers we've seen, whether we're thinking about the institutions or the institutional factors. And I do want to emphasize de-industrialization because, Kunal, you've done work in this area. How pervasive that is and the extent to which those different factors, apart from the standard sort of Mincerian factors we always look at, right, whether it's education, premier or skills and so on, they all, those three seem to be really important in thinking about inequality dynamics in middle income countries. Now, their weights may differ and the importance may change over time, but I think this already speaks to how we think about inequality within the distribution, if you like, far more carefully than we have in the past. Now, I think it's one of the first few sort of cross-country studies that does do that. The addition to start thinking about those right-hand side variables that look very different to the standard sort of earnings function literature, but more importantly, plug into the larger debates, global debates about de-industrialization, about wage colorization for me has been really useful in terms of the study results. Actually, that's the important point that I think we can end on, which is that all the case studies, especially the case, the country case studies, all use a similar methodologies as we all saw with the composition and so on. In a sense, one can compare life for life, we're not comparing apples and oranges here, and that's really important because from that, we can then get some general lessons for what's going on in the global south. So I think that's one of the strengths of this project that we're trying to do similar approaches methodologies across these different country case studies, and then we also have this work by Peter Simone and Albert, which takes us away from the standard way many people approach this literature, which is using purely on it, or the on it, mapping stuff, which is not really very relevant as you already saw from the first paper, Simone's presentation for many developing countries. So we are still, of course, sort of, I would say three ports in a project, and we're going to see slowly some of the papers appearing on our website, UNI Writers website, and working papers. And so hopefully for those who are listening in, you're going to see several of the papers coming out in the next few months, and it would be great to get your views on the papers and the work you're doing, and so feel free to write directly to me or to Peter Haroon or Simone who's also quite involved in this project from Writers side. We're very good to get your feedback on this, because you're still in a situation where we are being together, finding things and trying to find common lessons and common issues that are emerging issues are coming across these different countries studies. And so thank you so much. I'm going to stop here. Just pause, pause the hour. Simone, Vidisha, Kongli, Roxana, Tanika, and Amy, thanks so much, and look forward to seeing you at some point soon. Bye.