 So in a sense, this is I think looking at the the spectrum of papers in the conference This is probably a slightly different one Maybe my colleague my colleague that will follow me is also in that category of trying to get beneath if you like into the guts of What happens in the labor market when you think about inclusive growth? And it's a paper essentially that tries to tell three types of stories For a country that you all hopefully will know for pretty well But just in case you're not sure South Africa's sort of an upper-middle-income country Population about 50 million similar to South Korea The one that's important in this context is probably one of the highest unemployment rates in the world If you use the ILO definition at about 25% And a growth if you like poverty dynamic That's very clear since apartheid ended in 94 which has shown Modest depending on the time period and depending on the data set and the variables you use Either a reduction marginal reduction in poverty or absolutely no change in poverty or the poverty gap Independent of the measure as well, but then without a doubt an increase and a significant increase in inequality again Invariant to the measure that you use But what interests us in this paper was what it is that explains this change if you like at the level of the labor market and in particular We wanted to look if and see if you could update our sort of earlier work on 1970 to 94 and see if there was a skills bias story in terms of our employment trajectory And then secondly whether that was explained by specific factors So we start out it's sort of a three-part paper in search of some unity which is not quite there yet It'll be very clear to you and the three parts involves a descriptive overview and particularly Interested in looking at the changing structure of the economy in terms of employment shares by sector and skill We use that decomposition technique then to ask the question around the nature of these shifts Shifts whether they've been skills biased in nature and also whether within or between sector shifts explain those changes And that's the standard sort of cats and Murphy decomposition and then we go one step further and Try and link up to the author literature now, which has switched away ironically from this occupation focus to a task-based measure looking rather at different tasks Link to occupations With the notion that these tasks and the wage returns to these tasks may actually explain Dynamics related to things such as outsourcing declines in the middle of the distribution in terms of wages and so on And I guess underlying all of this is our continuous lessons for inclusive growth So we used our labor force surveys for South Africa, which were by annual until 2007 and in the quarterly labor force surveys we switch between them depending on whether we have and need wage data I'm happy to talk about the details of which data set we used and why but essentially we don't have wages for every single year and And when we when you do the returns analysis we can only go up to 2011 So some of the quick early descriptive statistics the first part of the presentation Just a standard story the lines are more interesting You can see the collapse in employment and GDP it's South Africa's first three-quarter recession since 94 So we had you know over 70 quarters of positive economic growth, and then you see the recession striking And it's very clear What is not clear from the figure is that the jobs lost in that period in the sort of third quarter of 2008 to Second up to the second quarter of 2009 those jobs lost have you not yet been regained in pure quantity terms following the recession The simple elasticity before the recession the output elasticity was point six four And after the recession it's minus point one six, and that's indicative of this inability to recover We thought we'd stick this isn't it's actually not in the paper But just to give you a sense of some of the structural changes you're seeing and you've seen of a very short period That's 1993 over there and 2012 using just simple shares of GDP Real GDP over the post-apartheid period and the big story here is financial and business services From 17 percent to 24 percent of GDP and that's a that's that's a big change over a short period And partly explained by mining share which goes from 11 to 6 percent Agriculture is important and we'll see for another reason, but it's not a GDP story But you see constant shares of GDP for agriculture of the period. It's important in our employment story so if you did What we call our bubble graphs within the DPR you these sort of average value added changes by employment The size of the the circle tells you the share of employment and total employment. It's very clear. What's happening? You've seen job destruction and output destruction in the mining industry right, so you've seen a decline in in both average annual employment and gross value added in agriculture you've seen a decline in employment but amidst growing value added over the period and We'll talk about the exogenous policy shock But it's essentially the minimum wage in agriculture that causes that that Employment loss for the rest what you see is fairly tempered employment growth over the period across all other sectors With financial services probably being the the big performer So a little bit more into the guts of the employment shifts. You've got those relative changes which is simply The percentage change in employment by that sector Relative to the percentage change in employment and total say any number above one Shows a faster relative growth in employment and you can see the green rectangle tells you what we're most of the growth has been right so within financial and business services and community services, which is 80% the public sector is where you see most of the growth in employment Losses in employment have been in agriculture Which we've spoken about and then mining Right and a sort of inclusive growth story, which I'm sure Eric and my chicken are very well about is in the East Asian sense For example would be a fast-growing light manufacturing sector, and it's very clear in the South African context You haven't seen this so you see a manufacturing employment growth a relative employment growth of point three percent and and the lots of other interesting stories in there, but essentially One which is of employment shifts of a rapid growth in the public sector Financial and business services, which has a statistical anomaly Embedded in there are temporary employment service providers again We don't have time to get into it a large proportion of whom are security workers that are driving This because of business services and essentially job destruction in the primary sectors At the skills level You've got medium skilled employment Declines in the primary sector. So if you look if we do the rather do the sectors Occupation or the sector skills sells you see job destruction in the primary sector For medium skilled workers same with the secondary sector, right and in a sense That's where we start thinking well if it's medium skilled workers in Manufacturing for example, is there something going on about outsourcing of these jobs? Is there something about automation foreign competition and so on that's causing this deterioration in employment? in in in that particular skills category so we then We then Try and add a bit more substance to it in the second part of the paper by by doing the standard decomposition Which in the literature divides employment shifts or relative labor demand shifts Devise them into sort of being driven by two forces one is a within sector shift So you'll see a technological change the relative price of capital and labor will shift And so on causing employment choices to change within the firm, right? There's also one which is not mentioned ever in in any of this literature because it's developed country focused is Exogenous policy shocks So a sectoral minimum wage, right? Is a within sector shift that can cause substantive changes to employment and we actually see that happening in in in in agriculture? Between sector shifts are those driven by changing preferences tasteful products and also Crucially global competition and trade flows and so the question then is in this methodology is which of the two forces are actually more important in Identifying your changing preferences for certain occupations. So the cats and Murphy decomposition is very simple one. That's the aggregate Shift you get the between sector shift when you subscript by sector and then the within is the difference between the two sort of fairly standard sort of Excel spreadsheet driven estimation when you do it for South Africa for the 2001 to 2012 period which we've done as I say for the long run from 1970 to 95 we then updated we can do it from 94 to 2012 but we thought we'd we'd control for data changes from 2000 onwards and Working with the LFS as you see I guess the probably the most interesting result is the last column Which is the share of the relative employment shift in those skills categories explained by Technology or within sector shifts or as the case may be an exogenous policy shock and you see overwhelmingly our relative preference for skilled versus unskilled or medium skilled workers has been influenced by within sector forces at At the firm level so In the case of managers and professionals that relative increase in the demand which is this Let me just get the cursor which is that 13 and 17 percent, right has been overwhelmingly driven by Changes within firms changes related to technology related to changes of capital and labor prices And so on here's the if I could just hit that is that minus 20.47 is the rapid reduction We've done sort of classic card and Kruger work Ben and I have done it on the sort of impact for the modern wage in Agriculture and you see the massive reduction in employment in agriculture That's that this is the other side of the story saying that this is the exogenous policy shock that caused this relative employment destruction amongst farm workers Over the period the minimum wage is instituted I think in 2003 and so it's a nice period to pick up that employment loss There's some evidence of between sector shifts particularly in the case of the mining industry. So you get it you get Rather than a dominant within sector Share you get some action if you like on the between sector side when you look at unskilled workers And that's partly this changing shares in output for the mining industry So Part of this part of this literature then then then stops if you like in the 2000s and and and they tell a nice story about skills bias technical change And we think we can tell that story for South Africa that part of the reason you haven't seen inclusive growth part of the reason If you like that you haven't seen these rising part of the reason you have seen these rising genies is because of what's happened to the skills profile of this economy So you've got just to close it off outside of the intestines of the estimates. You've got these You've got a high unemployment excess labor supply economy, and you've got a skills biased or a labor demand trajectory. That's heavily Biased in favor of skilled workers. And so all that does is stretch that wage distribution Because of the shortages and because of the nature of the labor demand that you see and The majority of which is being driven by within sector within sector ships How much time do I lose So in a sense, that's the that's that's where the sort of the in terms of the the techniques that you use in the methodology That's where the literature sort of stops There was now since then been very recent work by auto I see Moglu on the theory side and then goose and Manning and Then the paper that we actually used but we didn't quote is the Lemieux paper Fortin and Lemieux in which they They starting point is to say the following that they've observed if you like in in in North American wage distributions the sort of nonlinear, right But but but a decline in wages in the middle of the distribution. So you have something That's very specific across. I think it's between 1980 and 2000s They then propose and partly driven by the data that have access to that Routine based tasks. So tasks that are easily automated that can easily be Routined are easily off shoreable are easily exported to other countries And so the demand for and if you think of software engineers in India, right? The demand for those occupations with those types of tasks actually has fallen in the US And so you see a hollowing out of the middle and there's this broader discussion about the missing middle in the wage distribution and so on But it's it's driven by this auto and levy view. I think essentially that you have Wages that are responding because of the nature of tasks that are easily outsourced to other countries And and that's their starting point that in the threat of sort of international cheap competitors has forced Production lines, and I'm sure there's an industrial policy equivalent of that But parts of the production process can actually be outsourced But you can't get it that production process without looking or by looking at occupations but you can get it by looking at tasks, right and what they have which unfortunately we don't quite have is a Ben's looked at this in more detail a data set which which lines up occupation against specific tasks So if you think of a job jobs would have different types of Sorry, it's a lot to read but rather read the ball They would have different types of tasks in an occupation so an occupation will have an ICT type component It will have a routine or automation type component face-to-face Components or tasks on site tasks and decision-making analytic tasks So in a way the argument they make is that some jobs, right? So if you think of if you think of an ICT type job, right? That's clearly a computer program or software engineer or even if you think of routine type jobs like textile weavers Right machine operators assemblers those kinds of occupations which predominate Predominantly have those forms of tasks are easily offshoreable and if they're offshoreable They're gonna have specific impacts on the distribution of income labor market outcomes on demand and supply and so on, right? And so the search is really for trying to see or estimate whether you see these kinds of outcomes They have really detailed data They they have more than these five categories and they have it by I think it's at a very granular Occupational level we don't have that kind of data And so we all we did in the in the paper is actually to to accrue and create tasks categories by occupation and so we went through the four-digit level of the ISOC That's in the label for survey So and then assigned what we thought based on you know if it was a software program and so on What basically what we thought would be the predominant tasks and and that's available for everybody to look at and critique But in a sense our question is to what extent do you see changing returns? Right to these tasks across the time period. Do we find for example evidence that routine based tasks or automated tasks? Have been facing global competition and therefore you see reduction in in returns Any five minutes but in a sense these are the tasks and across the occupations As you can see how you can align different tasks to different occupations note that they're more tasks than there are Occupations because each occupation will have more than one task aligned to it Just a little bit about what you can see going on say for 2001 you'll see that in in agriculture right the majority of tasks involved in agriculture would be automated type tasks or in manufacturing the majority of Tasks in manufacturing will be automated and so in a sense we take the automotive kind of tasks those are the ones according to At least the global literature or the North American literature would suggest that that's the task That's most prone to global competition because it can be easily so it can be easily outsourced So all we do is we take a quanta regression approach So we go across the distribution and we estimate the the coefficient and our regression is not there So the regression should be below It's disappeared, but essentially it's a quanta regression with our standard controls So we have the you know age gender union controls. No, we don't have union, but Experience and so on and then instead of the occupation dummy what we have are these these tasks Dummies right so an individual will be signed a task and then we measure the coefficient on the basis of the The distribution that you're measuring at the particular quanta and so this is the third last slide But this is essentially a key results. It's I haven't seen it done for any developing country and I'm sure It'd be great to sort of see if we can tell similar stories for other parts of the world But let's take the automation one. So these are your quanta So these are the coefficients on the automation dummy across the quanta distribution, right? So from the first quanta to the 90th quanta and these are the wage returns, right? But you'll see that between 2001 and 2011. There's actually been a decline in the return I'll come back to the bottom end of the distribution But from sort of the 20th percentile right upwards You've seen a decline across all the percentiles in the returns to Automative tasks In a sense for us that tells us that you've seen this because of foreign competition You've seen this off-shoring of jobs, right that are easily automated two more minutes, okay, and Just very quickly. So for example in the face-to-face that would be Predominantly jobs that are either sort of in financial services or in the public sector We have to engage with individuals you seeing an accident increase across the quanta Just very quickly for ICT We don't see a result that you see in North America because most of your ICT jobs in South Africa Or at least in the developing country context like South Africa are in financial services and the public sector And in fact for various reasons you've actually seen an increase in the returns to wages across the distribution there Just very quickly at the bottom end what's going on and it's an important part of the story is strong trade union involvement in collective bargaining that pushes the bottom end up as well as As well as minimum wage intervention in certain sectors. So very quickly our conclusions on employment Most of the employment growth has been driven by Pre-crisis employment growth. We see agriculture the impact of minimum wages is key to the employment collapse There's another story another paper there, but mining loses jobs at a rapid rate for various reasons which I can discuss Very little growth in employment in manufacturing and that's that's an lack That's an exclusive or lack of an inclusive growth story, which I think is important to inject into into broad discussions And then finally that's those are the conclusions on wages which you can Read yourself. Okay, I'll stop there. Thanks very much Thanks very much