 Welkom bij de volgende session, mijn naam is Gijs van de Vries, ik ben de professor in de Universiteit Groningen in Nederland. Dit is een joint-work met Marcel Timmer, maar zoals je kunt leren, is deze presentatie eigenlijk een partij van een brede team-effort. Dus Klaas de Vries en Anna Moreno, we zitten hier ook. We werken eigenlijk op een grote schilder-effort om de database te constructeren om productiviteit en grof in Afrika over de lange ronde te analyseren. Dus we werken op dat, het is een ander Differt ESAC-fonderd project en ik zal een van die papieren volgen van het project presenteren. Dus om te beginnen met de motivatie voor dit project of dit papier, als je het literatuur read, heb ik hier een paar boeken van Gijs van Groningen naar Nelson en Phelps, naar Abramovicen en een reden van landes. Als je deze boeken read, de meeste economische historieën, ze praten veel over technologievertrekking om in kapotel te worden geëmbodd, waarin het een belangrijke baas voor productiviteit en grof in landen ontwikkelen kan zijn. Dus als je dan kijkt naar de empirische literatie die dit effect probeert te testen, dan ga je naar firmlevelstudies en ga ik hier een reden in de economische grieftje van Amiti en Konings voor Indonesië en er werkt van Marc-Andrés Muntel in Brazil. Maar ik kan je nog meer dan 100 van deze papieren listen. Typisch op de firmlevel vind je zeer sterk effecten van de transfer van technologie op productiviteit en grof op grof. So firms that import intermediate inputs, import capital goods from leading countries, they typically have a high productivity growth. But the problem is that if you move to a more aggregate analysis, you tend to find much weaker support for this in the empirical literature. So one of the key references is here is the work by Rodriguez and Roderick. They argue that a lot of trade measures actually do not really find support for this aggregate effect which economic historians talk so much about. So what we do in this paper is actually we're going to take a step in between to analyze where this congruence is coming from. What we're going to do is we're analyzing technology transfer and explaining growth in developing countries. We take more a meso perspective, so we look at the level of sectors. So instead of this firm level, this micro level analysis that does not match with the aggregate macro analysis, our sector perspective is providing a bridge between these two. And we're using a new data set. So when I submitted the paper, I submitted specifically for Africa, I will talk more in depth about Africa, but basically the data set is broader. It includes 30 countries, 9 in Asia, 10 in Latin America and 11 in Africa. And there is annual data for all these countries from 1960 to 2010 for value added and for employment. So these two allow you to calculate labour productivity for 10 broad sectors of the total economy. So for each of these countries, for each of these years, for each of these sectors of the total economy, you observe value added and employment. And the value added, we put those into an international perspective using sector specific PVPs. So if you are aware of the literature and in most of the literature would use aggregate PVPs to calculate productivity levels also across sectors. But we have been using the latest round of the World Bank ICP project, to derive sector specific PVPs. So we move into the theoretical framework that I consider here. Basically what I think of is if you have labour productivity, so it's defined as value added divided by employment in a sector J of a country I, this may increase either because of innovation or because of technology transfer. So this is the distance to the frontier model introduced by Chimoglu in Agion in 2006, or at least that's when it was published. So you have the change in productivity here and you're saying this is related to this gamma ij which is basically innovation or there could be a technology transfer if you're removed from the productivity frontier. So the further you away from a technology frontier the larger this lambda ij can be, it can be a positive effect of technology transfer coming about. And then you can take this model and extend it a little further. The full model is not what's shown here, but this is basically the idea that I want to show to you here. So you have productivity growth and you're going to explain this taking the distance to the frontier so a kind of convergence effect. But I'm going to moderate this by the effect that the imports of intermediate inputs can take. And then this set ijt which could be the direct effect of imports of intermediate inputs and capital goods. And then this set ijt interacted with the distance to the frontier. So the further you are away from the frontier the larger the potential for technology transfer from international trade there could be. So this first effect the effect of changes in productivity to the distance to the frontier is something that's fairly well known. So just here a scatterplot on how this looks like in my dataset. So if you have here on the x-axis you have the distance to the frontier. So the further you are away from the frontier the larger the scope for productivity catch up. And then on the y-axis you see the average labour productivity growth. So over this 50 year period you have the average productivity growth rate and you see that on this this cloud of dots you see that it's positively up with sloping. So this is indicating that the further you are away from the frontier the higher your productivity growth is likely to be. So the frontier is measured by the US. So you have the sectors in the US the productivity level and you have the productivity level of each of these country sectors in these developing countries. So say agriculture in Tanzania or mining in India or any of these country sectors and you translate the value added to the US dollars, the international dollars using this sector specific PPPs. So then you can take those. So you have your productivity measure which is on a comparable US dollar measure and you can put those here. So this is the US frontier or the productivity level from the developing country. So if this is the US and this is the developing country it's the logarithm so the difference is the potential for technology transfer. The difference is the gap. So that you see here that the further you are on this x-axis the further you move to the right the further you are away from the frontier. That's how we measure it. So our main data source for this is an update, an extension of the 10 sector database that has value added employment data with broad sectors of the total economy. I worked on this for quite a long time. An early publication was in 2009 but more recently we have been extending this to 11 African countries. So Klas en Anna have been working on this by myself and part of a larger team. So if you are interested in learning more about this then feel free to discuss this with us. Two key points about this database I want to erase. First of all it's internally consistent. So you have value added and employment that both match with each other. So you take a national accounts concept that you cover the full economy that these are internally consistent they have to be internationally consistent so that you know you measure the same thing if you're talking about agriculture you're measuring agriculture in all of these countries and they have to be intertemporally consistent. So over time you follow this in the same you follow the same variables using the same sources and methods and then the second point which is important it's a full coverage of the economy so it tries to include both formal and informal activities. So the second speaker will discuss I just discussed paper by Roderik on unconditional convergence recently published in the quarterly June of Economics he only looks at informal activities so here this dataset includes both this one I just discussed with you also so these sector specific PPPs work for the World Bank where you derive these PPPs from the International Comparison Project 2005 round only for agriculture we thought it would be more accurate to use FAO data and then the trade data so far this is also not yet published but it's going to be published it's worked by Robert Feenstra the new version of the Penn World Table so we have detailed trade data so what you have typically is trade data that you see the imports but you don't know where the imports are going to so are they going to final consumption in a country or are they actually going to use by firms in these countries to produce something else so this trade data by broad economic categories allows you to distinguish between intermediate goods between capital goods and between consumption goods so I'm taking those data we're also working on that with Robert Feenstra so I have the advantage to already be using this data to test of this data and things will become available online starting in July so via our website the gddc.net and also the Penn World Tables website so here's an indication of how the productivity gap looks like across a set of developing countries so we take the United States as the frontier country as one I explained this before how I derived this level so we have a hundred then this means that for the total economy on average in these developing countries they're at 17% of the US if you look over time there's a slight catch up and I'll discuss a little bit more how this looks like across countries but there's substantial variation across sectors in agriculture then agriculture is further away from the frontier than a typical services sector so you have here if you look here the agriculture is about 10% with a couple of services sectors they are much closer to the frontier and of course you can argue with me about where particular services sectors you can well measure market services sectors are much more difficult to measure over time it's slightly improved but this was mainly related to improvements in services sectors so workers in developing countries moved to services sectors with somewhat higher productivity levels so that was causing this catch up for the total economy but if you look at particular sectors you do not see this trend yes in each year yes yes so this is the US is it a frontier indeed so if the US basically what you could have here so manufacturing is falling behind right but it's falling behind relative to the frontier so if you take productivity growth data of manufacturing this is positive for most of the countries but the US has been improving its productivity in a faster rate right so what you would expect is if there is a productivity level then you would expect a catch up over time so you would expect that these developing countries their productivity growth would be faster and if that were the case they would be actually coming closer to the frontier sorry exactly yes I mean you can think of some sectors some countries might be leading in terms of productivity but typically and also in my entire data set the US is leading in most this is for Africa for Africa you can see that the gap is typically larger right so so you see instead of about 17 to 20 percent you have about 9 percent agriculture there is a 50 percent difference in labor productivity between the US and Africa also in manufacturing it is falling behind what you can see is that services also have somewhat relatively higher levels so people in Africa have been moving to urban areas and largely to services sectors and this is causing a slight shift in the aggregate productivity level yes was that what you bring cause we are already running through your minutes and you have been able to answer some questions already oh yes oh I think I have 10 minutes left no but I very much appreciate these questions because they are quite clarifying so indeed one of the limitations that we still face here is that we do not observe capital so we have been putting a lot of effort into constructing this database of value added and labor but we are still missing an important input which is capital so you could argue that ok so to what extent is there other differences with capital and non-capital or less capital intensive sectors so we know that agriculture in the US has become extremely capital intensive and this might be related to this decline in this respect I do tend to believe so I for a more limited set of countries we also have capital data so I did these types of analysis in other studies I do tend to believe that even if you take capital into account you still observe these types of patterns and then what I am doing here in my empirical analysis I look within sectors over time if you take this within then you are less likely to right so if there is I mean mining if you take the labor productivity level of mining and compare it to agriculture there will be differences because mining is much more capital intensive to compare it to agriculture so I will look within sectors over time that is another way but it is indeed a relevant point that capital is not taken into account here so here is the relative productivity level in manufacturing over time from our database so I took manufacturing I took the unweighted average across three regions so the green dots you have Latin America in 1960 the blue line is Africa in 1960 and then the red line is Asia what you can see is that at the start of the period Asia and Africa had more or less the same productivity levels and Latin America was at about 40% but you very clearly see this pattern where you had a slight catch up until 1980 and then a long run decline all the way to now the same happened in Latin America so after 1980 it started to decline in Asia you see this opposite pattern so you have this this increase in relative productivity levels over time so that has been a divergence for these sectors so some of the econometrics of this paper so what I do here is I try to bridge these micro findings on the effect of imported inputs to the more aggregate analysis what you see here is if you focus first on this column 1 you see that if there is a productivity gap so the larger the gap the larger the effect will be on productivity growth this is the correlation I showed you in the beginning so that's the positive effect it's significant across all these specifications what I do here is for the total economy I include this role of imports of inputs so that's the direct effect it can have on innovation and then the imports of inputs interacted with the productivity gap so that's the kind of technology transfer potential so if I do that for the total economy I do not find any significant effect and this is also one in a lot of these cross country macro studies has been found there is a limited role for trade in effecting productivity growth but if I go to the more meso level then I find that if you look at goods producing sectors and within goods producing I could look at agriculture manufacturing there you do find a positive very positive and very strongly significant effect of inputs on productivity growth so this is in line with all these firm level studies and most of these firm level studies is for manufacturing firms so here I also find for these sectors I find this positive effect but if you go to particular services sectors and then I split this up into market and non market services so market services are typically somewhat better to measure because they are active on the market so you have retail or you have transport activities these types you measurement is always a matter of degree the biggest had put it but here you do not find any effect non market services if you non market services like government activities are much more difficult in health and education much more difficult to measure outputs put them separately but here I also do not find any effect so what they suggest is that for these sectors you do have a positive effect for a lot of services you don't and as a result you also do not observe an effect in the aggregate and this is to my in my opinion this is providing a bridge in this literature where these effects from the inputs of inputs are coming from so as said this is a pretty robust finding I tried a couple of other specifications further suggestions are welcome I used 5-10 year averages one could argue that there is some endogeneity issues because the different sectors of the economy so if other sectors of the economy are growing then they do not require these inputs to be imported anymore so I tried using internal instruments still it's it's a significant finding that it's only in in these goods producing sectors I included additional controls so growth at the US technology frontier controls for financial development so credit to GDP human capital human capital interacted with the technology the gap so potentials for technology transfer my main focus is here on the role of imported inputs which is robust and significant if I do this by regions however I tend to find a positive effect for Asia and Latin America but most of the results actually disappear once I do this for Africa so this is something to discuss perhaps that you in Africa I do not find any significant effects of the role of trade on productivity growth so I'm concluding here so what I did I used this new database to document trends, productivity and convergence in developing countries this allows you to analyze where this for a more general equilibrium perspective where these changes are coming from and what we find is that imported inputs enhance technology transfer in goods producing sectors but not for the total economy I've argued that this would provide a bridge in the current literature on growth in Africa so thanks for your attention