 Okay. Thank you Carol. Let me just start out setting the scene, so to say. I've read a lot of some papers on learning by exporting and I can see some of the authors of these papers are actually here, present here, so we should have a nice discussion. And while working on this, I got to look a little bit deeper into all of these databases we have on learning by exporting. And the first thing I will try to highlight today is are we actually telling the correct story? Are we analyzing the question where the firms learned by exporting? Are we analyzing this correctly? Are we asking the correct question? This is what I will try to end up at. So in 2008 I went to Mozambique. I'm not only doing this kind of studies in Mozambique but today I'm only going to talk about this. And in 2008 I went there to study industrial development and industrial policy and how to improve the productivity levels of manufacturing firms in Mozambique. And when I arrived I learned that Mozambique actually performs as the worst country in the static region in terms of productivity. So if we look at the work done by the World Bank and the results that they get out, Mozambique is the worst performer in terms of productivity growth in the 2000s. We also learned that the data availability in Mozambique was not as we were promised when we arrived. We were supposed to get a kind of a census when we arrived and when we turned up I could actually identify five larger scale firms with over 100 employees that were not in the census and it was not due to the age of the firm so it was not a newly established firm. So we had no frame to actually sample from when we arrived. So what to do then? What we decided to do is actually combine all the data sets that were available, quantitative data in Mozambique and try to carry out learning by exporting analysis. So standing here five years later I would like to say that we know a little bit more about learning by exporting in Mozambique. But the main conclusion that I'm going to reach today is basically that we have more questions now after analyzing data for five years than answers to the question whether firms actually learn from exporting. I will come back to this. So the hurdle or the problem that we face often in quantitative studies is that a positive association between firm level productivity and export participation does not necessarily mean that learning by exporting is taking place. So we may have a self-selection issue that more productive firms they choose to export or have the ability to export and therefore they start to export and we find this positive correlation. Moreover there was also the question that learning by exporting and so the self-selection issue and learning by exporting they are not mutually exclusive. So you can actually have both things taking place at the same time. Within this literature of learning by exporting we borrow a little bit from the terminology of the technology diffusion literature. So when I talk about learning by exporting it's not all authors that do this but I put it in this framework. We actually talk about these horizontal and vertical spillover effects that may come from entering export markets. So we have these knowledge flows from competitors what we call the horizontal spillovers and then we have the knowledge flows from customers potential customers abroad which I will label vertical spillovers in the following. We have the competition effect the horizontal spillovers that come directly from entering the international markets. So exporters must improve efficiency faster than firms only selling domestically to survive in foreign markets. This is the one reason that is often cited in the learning by exporting literature. Then we also have the vertical spillovers effects that occurs as foreign buyers may wish to improve the process technology by providing product specifications and technical assistance. The famous of Seminole paper by Claveritas and co-authors from 1998 they actually highlight this as one of the main issues in the learning by exporting hypothesis. So basically what we say in this paper is that learning by exporting is a learning process. So the decision to export is complicated by some cost startup or some startup costs. So you have to research, you have to study the markets that you enter. You have to establish marketing channels and you have an idea of what your productivity level will be in those markets but you do not know exactly how you will perform when you enter. That is basically the idea and it comes in some sense from the learning model by Jovanovic in 82. So the main learning by exporting question in most literature is how does the decision to export affect the productivity trajectories of exporters and plants that switch markets relative to those of non-exporters. So you can see to frame it in the new literature, the impact evaluation literature, you can see it is like a matching exercise that we have a firm that is on the borderline to start exporting and then we would like to compare with a similar firm that does not start to export. And then we follow them over time and see how they perform differently over time in terms of productivity. So we test whether exporting history enters significantly in the standard cost equation at least in the Gliridis setup and so we perform a greater causality type of test. That is the typical way of doing it. We have seen a lot of GMM studies on this issue in Africa on limited data sizes I have to say. So a number of firms that we actually have in the databases are fairly limited whereas in Asia, the studies done in Asia and in Latin America they actually have larger data sets and more switches. I will come back to this. So when we started our work we were actually thinking that we would do something similar to the previous literature. But in order to actually follow this idea, we need to identify those that actually change startup operation and then change to foreign markets. What we saw in the dataset in Mozambique is that we had no firms that actually were operating domestically and then changed to export markets. Most of the firms or all of the firms that we identify in the quantitative data, when they start up, they start up as exporters. They were not producing and selling domestically when they started up. So this puts a limitation to the methodology in some sense. That we do not have these switches. We cannot identify on those changing export behavior. So they are, as we call it in the management literature, I'm not that familiar with that literature but it's called the born global in the management literature. I will go quickly through this literature review. Most of you know it. But the main thing I would like to highlight here is that there is a strong correlation that we find very different results in different regions. But if you go into detail with the data, there is a strong correlation between quality and size of the data and the results obtained. So we have a tendency that the positive results regarding learning by exporting are often with the relatively small sample sizes. And where we have very little identification due to a very few switches. If we look at the variation, which is important for the identification of the effect, we see that we have very limited switching occurring in Sub-Saharan Africa. In some countries, or at least the countries I've been studying, I took the, there's some studies published on Zimbabwe. And it turns out that the effect in the case of Zimbabwe, the total effect is identified based on four observations. So four switches in that paper. Mozambique, we have even fewer. We only had over a five, six year span. Only one firm switching export status. So we needed to pursue another avenue. And I can say I'm not going to present this today, but our new 2012 data, we only have from 1999 until 2012, we have only four firms changing export status during that entire period out of the entire population. They are born global, more or less. So what is the reason for these few switches? Why do we not see more domestic firms starting to export? Why are they set up as limited liability companies from the start? It turns out that they switch managers often. It's actually that the manager of the firm that is set up is actually very familiar with running an exporting company. And they come from a firm that also sold goods items domestically. So maybe one of the things that we conclude in this paper, there is a paper uploaded to the wider web page, is that maybe when we analyze the learning by exporting, we should instead follow the manager rather than the firm when we talk about the learning by exporting hypothesis. So in this paper, given the data we have available, how do we actually disentangle the effect? This is the question. Being an accommodation, we always seek new estimators to actually try to identify this problem. What we do here is I would say second best in terms of identification. What we do here is taking traditional matching techniques and try to follow the productivity growth of the firms exporting and see if they grow faster in terms of productivity than domestic firms. This is not ideal. We would have liked to identify it based on the switches. What we also resort to is the generalized blind or hacker approach to see whether it's the characteristics of the firms or it's some kind of discrimination of exporting firms or non-exporting firms that is actually driving the result. For those of you not familiar with the estimator, what we try to do is explain the productivity gap and divide it into characteristics effect studying the observable characteristics of both firms. It's more like a matching technique and see whether they differ along those dimensions and the second component measures the importance of the differences in parameters for the two groups and that is called the coefficient effect. Moving to the data. We combine five different data sources here and this has been a lot of work because we have taken all the names and addresses of each firm and tried to do a consolidated database. Then we have also make sure that each firm is observed at least twice in the database to make sure that the time invariant characteristics are the same for the firms because we are not certain of the quality of the data in general. Finally we have checked all the financial data against accounting data from the KPMG, the largest accounting firm in Mozambique. So we have checked it. We have tried to, you can say, triangulate the information that we get. We end up with the 755 observations for 275 firms when doing this. Moreover, the new data from 2012 is compared, the results are compared. Unfortunately, we do not have a panel because we had to start all over again collecting new data. A lot of the firms that were in the previous data could not be identified anymore. A lot of managers have moved between firms. We see a lot of manager movements in these firms. So it was very difficult to identify firms in the new database with the old database. When we talk about exporting firms, a lot of us, we often think that these are big, large-scale firms with a nice front when we come to the office. But these two firms are examples of exporters in Mozambique. You can see the first one here in this column is producing furniture for South Africa. So they are exporting to South Africa. The second one is a bakery, also exporting to South Africa. So it is not per se larger-scale firms that export. But both of these firms were set up only for export purposes. They are exporting from 100% of their output, even though they look like this. So this is just what we are facing, not only the larger-scale firms. I just wanted to show you these firms. I will skip this very, very quickly. As mentioned, what we actually follow, we follow some of our months here. So we follow some of our months here. So we follow some of our months here. As mentioned, what we actually follow, we follow some of our months here. So we follow actually some of the work that he did somewhere with some of the co-authors. But there is a little trick compared to the things that months published that instead of the levels of TFP, we actually assume that the changes in TFP is dependent on export status and export participation. This is the only fundamental difference from the existing literature in this case, and it is due to the Bonn global phenomenon that we observe. So basically, we combine the blind-ohaka decomposition approach with traditional matching techniques to identify the learning by exporting hypothesis. Basically, we use the recent work by Klein in the American Economic Review stating that the blind-ohaka estimator enjoys the status of a double robust estimator of counterfactuals, as estimation is consistent if either the propensity score assumption or the model for outcomes is correct. This estimator is especially suitable when we have very few treated observations as we have in this case. What are the results quickly? We have a situation where the manufacturing sector is finding it difficult to keep up with the growth pace of the rest of the Mozambican economy. It remains relatively small, employs fewer than 3% of the late performance, diversification is low, even when defined at the three digit Isaac level. So production is carried out in very few sectors, especially bakeries, we see a lot of those. Very specialized production. So the within country competition in those sectors are actually fairly large. Almost no entry over a 10-year period. This is a long period. We see very little entry. We are trying to compare the census data now from 2002 with recent census data from 2012. And we have almost the same amount of manufacturing firms over a 10-year period. Have they grown in size in terms of employment? No, they haven't. So we are not seeing a lot of movement in the manufacturing sector. And we still see very few exporters. We are counting it on very few hands. I think it is 24 out of 761 firms. And they are even biased towards being large scale firms that are exporting in the manufacturing. So we also still see the born global phenomenon. When we focus on the models I just described and focus on labour productivity growth. We actually find that the firms they are actually in terms of labour productivity growth are moving ahead. 40% unconditional exporter premium and when it turns to a 20% conditional exporter premium with a negative characteristics effect. We also find a significant firm size export status interactions. But these negative learning effects they only start to hit in when we have approximately 1,000 employees and it is more or less 4 to 5 firms. So it is not really economically relevant the significance of the interaction term. So very few firms export and participation is highly persistent. There is an issue of lack of entry into export markets. And it seems as if it is the firms that lack entry and the new firms that start operating and enter export markets they actually substitute for previously exporting firms that exit. And the managers just switch to the new exporting firm. There is evidence supporting the learning by exporting hypothesis. So there is something to gain. So if we start moving or if we can move manufacturing firms into export markets we may improve productivity of the firms. At least in terms of labour productivity. The qualitative information we also gathered in this data is that none exporters when we ask them directly why do you not start exporting? Why do you not move into foreign markets? It is not lack of credit. It is not lack of finance, lack of suitable labour. So quality labour. It is more lack of knowledge of potential markets. That is what they state. Whether this is true now there has been set up based on these results. There is information system set up to actually enhance exporting in Mozambique. And we hope that this may solve some of the problems that we are facing in Mozambique manufacturing. Thank you.