 Today I'm going to tell you about, from productivity to exporting or vice versa, evidence from Tunisian manufacturing firms. This work has been done with the cooperation of Mohamed Ayedi, here present. Before starting this presentation, I'm going to motivate for it. Enhancing the competitiveness of a country's industry is a key issue for economic growth. Several empirical analysis and bulk of literature suggests that competitiveness is closely related to factors such as firm's productivity and global engagement. In that, more productive firms with the likely large scale of sales and production are likely to self-select into exporting markets. This is termed as the self-selection effect, and the exporting activity is one way to accumulate external knowledge through various channels which might be translated into productivity improvements. This is termed as the learning by exporting effect. Moreover, the large productivity premiums of new exporters as compared to non-exporters imply that the decision to start exporting is determined by factors affecting firms' productivity, namely the innovation. This suggests that there is a channel linking productivity to exporting, namely the innovation activity. Two chains of relationships are identified by the literature in the first, firms invest in product innovation, which might be translated into efficiency improvements to the firms, allowing firms to involve in exporting activity. In the second chain, an increase in exporting, which might also be translated into efficiency gains to the firms, allows firms to invest in process innovation. The papers outlines are as follows. First, we export the link between productivity and export decision. We, in particular, test for the self-selection and the learning by exporting hypothesis. Then, we look inside the box and explore the link between exporting and innovation activity. Notes here that we do not distinguish between process and product innovation. We then extend the basic analysis to deal with sectoral studies. We tackle the same issues mentioned above for four main sectors, the textile and clothing sector, the electric, electronic and mechanical sector, the agro-food sector and the remaining sectors pull together. And before concluding, we give some policy recommendations tailored to our empirical findings. The empirical analysis will be based on firm-level data on 1,323 Tunisian manufacturing firms from 2004-2006. The data are compiled from an accounting, industrial and export flow surveys which are annually conducted by the Institut National de la Statistique of Tunisia. The empirical analysis will be based on two clusters of firms. In the first one, we distinguish exporters from non-exporters. In the second one, we distinguish fully exporting firms from others. The rationale for this distinction is driven by the particularity of the Tunisian manufacturing sector that almost 70% of exports come from the offshore sector which firms are mainly subcontractors benefitting from several advantages including the technological advancement and the fact that exporting is guaranteed. This means that pulling partially and fully exporting firms may well mask more than reveal some features of the real behavior of fully exporting firms. The estimation procedure is as follows. First, we model the self-selection effect as the probability of exporting a firm during period T, regressed on luggage exporting status, luggage sales, and we use other firm's characteristics including the firm's age, the firm's size, the capital intensity and the capital owning status. The key variable here is luggage sales which coefficient is a sufficient statistic for efficiency for self-selection whenever it's positive and significant. Second, we model the learning by exporting effect as a simple linear regression of sales of firm I during period T, regressed on luggage exporting, luggage sales, and we use the same vector of control variables used previously. The key variable here is luggage sales. We use one period lag because learning is unlikely to be instantaneous so it needs time to take its full effect. The results for the self-selection effect for both clusters of firms are as follows. For the first cluster of firms, there is no evidence about self-selection. Previous exporting increases current exporting which indicates the sunk cost of entry into exporting markets. Foreign-owned firms have high abilities to export because of their better knowledge of foreign markets' characteristics, the use of better governance strategies, among other issues. For the second cluster of firms, there is a strong evidence about self-selection. Moreover, the marginal effect on luggage sales is even stronger, meaning that fully exporting firms may exhibit superior productivity, allowing them to self-select much more often into the exporting markets. Luggage exporting increases current exporting, the marginal effect decreases slightly. Here, we provide two interpretations. Either the sunk cost of entry into exporting markets or the fixed cost of engaging in exporting might be reduced as compared to the previous involvement in exporting. Foreign-owned firms have high ability to export. The marginal effect is lower. This may be due to the higher rate at which foreign capital exhibits decreasing returns to scales. Based on the stylized fact that these firms or fully exporting firms are likely to have a higher foreign involvement than others. The firm size increases current exporting. Indeed, larger firms may have a large scale of production and sales or may enjoy low fixed costs associated with exporting compared to smaller ones. Now, I present the results for the learning by exporting effect for both clusters of firms. For the first one, there is an evidence about learning by exporting. Luggage sales increases current sales, which may indicate the persistence of the firm efficiency over time. For the second cluster of firms, there is no evidence about learning by exporting. The explanation is likely to be related to the dynamics of learning in that fully exporting firms in Tunisia are mainly subcontractors with relatively longer previous experience in exporting. Moreover, these firms, which are already ahead of technological advances, it is as if these firms, it is as if exporting is made between countries with similar technological advancement limiting thereby the scope for learning for these firms. Innovation increases current sales. It is an expected result. Now, we look inside the box and explore the link between innovation and exporting. And before going further, we need to precise the measure on which we propose to measure the innovation activity. Before doing so, we need to give the rationale for such a choice. First, there are no available data about expenditure and research and development activity. And in addition, any type of actual innovation is not directly observable. Second, the availability of team of scientists, engineers and technicians with suitable qualifications in the research and development activities might be quite a plausible source for innovation, meaning that a measure of human capital is necessary to account for the skills embedded in the firms employed themselves. This is why you have chosen the following variable, which is defined as the proportion of engineers and technicians with different degrees of qualification in the total labour force of firm eye during period T. This variable is likely to capture labour displacement. Note that in the literature, these kinds of variables capturing labour displacement are usually used to account for actual innovation more than expenditure that may or may not lead to innovation. The estimation procedure is as follows. First, we model the question of whether innovation is indeed a prior decision to exporting what we call the exporting equation as the probability of exporting of firm eye during period T regressed on lagged innovation, lagged exporting status, and we use the same vector of control variables used previously. Second, we model the question of whether exporting triggers innovation activity, what we call the innovation equation, as linear regression of innovation of firm eye during period T regressed on lagged innovation, lagged exporting, and we use the same vector of control variables. Results for the exporting equation for the first class of firms, lagged innovation increases the likelihood of current exporting. As for the remaining results, they are almost similar to those for the estimation of self-selection. For the second class of firms, lagged innovation increases current exporting. The marginal effect is slightly lower. Fully exporting firms are mainly subcontractors for which exporting is already guaranteed. This may well mask most of the effect of previous innovation on triggering exporting. Second, lagged exporting increases current exporting as expected, indicating the sunk cost of previous exporting. The marginal effect declines slightly. Here we tried to provide two interpretations. The first one is that fixed costs associated with exporting might be reduced as compared to the previous involvement. The second interpretation is that the marginal cost of production is likely to reduce. Indeed, it's commonly known in the literature that firm's fixed cost is inversely related to its marginal cost of production. So fully exporting firms are known to be larger firms, so they are likely to have large fixed costs, which means that their marginal cost is likely to be lower, allowing them to involve a larger scale of production. And then this increases their involvement in exporting. Foreign capital increases current exporting. The firm's age affects negatively current exporting. This might well refer to the rigidity of all the managing systems, especially when these firms are run by old individuals who can be less receptive to innovative technologies or be more risk averse. Now I present the results for the innovation equation. For the first cluster of firms, lagged exporting increases current innovation, so exporting leads indeed to new knowledge and not just investment in new knowledge. Lagged innovation increases current innovation, indicating the sunk cost of innovation. The firm size is also good determinant for innovation as expected. This might be due to the importance of scale in research activity. It might be due to the high abilities to diversify risks or to have access to a larger pool of financial means. Or it might also be due to the high absorptive capacity of such a kind of firms. So even when these firms do not innovate, they can invest in innovation activities in order to enhance their absorptive capacity. The firm's age reduces current innovation. Indeed all the firms may be less innovative because all the individuals may be more risk averse. Moreover, foreign capital increases current innovation. These firms have better access to new technologies and might be endowed with more financial resources to invest in innovation activities. For the second cluster of firms, results do not give new insights as compared to the previous cluster, except that the coefficient on lagged exporting increases. Indeed fully exporting firms have high abilities to acquire new knowledge and more incentives to innovate as expected. Now we extend the basic analysis to deal with sectoral studies. So since several findings recur, we will only present the main findings or the most striking findings for each sector. For the textile sector, we have found that the main result is about learning by exporting. There is no evidence about learning by exporting for both clusters of firms, especially for fully exporting firms. Given that the percentage of fully exporting firms in our data here in this sample exceed almost 84% for the three waves data, the explanation is likely to relate to the dynamics of learning. Indeed the textile sector has adopted an export-oriented strategy since the beginning of the 70s in the country. So each firms have long previous experience in exporting. For the second sector, the electric sector, before going further, notes the small size of the sample. So this means that this can be a possible source of biases in our results. So we should be very careful in interpreting them. The first main result is about self-selection. There is no evidence about self-selection for partially exporting firms and in turn strong evidence for fully exporting ones. This may indicate the superior productivity of fully exporting firms, which is a quite intuitive result. The second result is about learning by exporting. First, there is no evidence about learning by exporting for partially exporting firms. Honestly, we didn't know how to interpret this result. It can be a biased result or one possible explanation is that this sector is known to be capital intensive. It is possible that an increase in efficiency is associated with a more utilization of capital in such a way to mask the direct effect of exporting on efficiency. Second, there is an evidence about learning by exporting for fully exporting firms. Here, the explanation is also likely to relate to the dynamics of learning. Indeed, the electric sector has emerged in the country no longer ago. It has emerged only during the last decade. So its firms have no long experience in exporting. The third main result is about innovation. Exporting increases the incentive to innovate for both classes of firms. These sector's firms are heavily dependent on foreign technologies, increasing their incentives to innovate. The main results for the agro-food sector, the first one is about self-selection. There is no evidence about self-selection for partially exporting firms. Before giving the explanation, note that results for fully exporting firms were meaningless. That's why we didn't report them. So export decision is not driven by efficiency here, but rather by exogenous factors, including the availability of first quality products such as dates, olive oil, and also industrial policies encouraging and promoting exporting, which have been implemented in the country, I think, since the beginning of the 60s. The second main result is about learning by exporting. There is no evidence about learning by exporting for partially exporting firms. In turn, the evidence was quite strong for fully exporting ones. So the explanation here is quite different from the explanation provided previously for the other sectors because of the specificity of this sector. Indeed, this result is likely to be related to the destination of exporting. Indeed, according to De Locker, exporting to high-income countries offers a higher score for learning than exporting to medium or low-income countries. And in the agro-food sector in Tunisia, partially exporting firms export mainly to medium or low-income countries such as Libya and Morocco. In turn, fully exporting firms export mainly to high-income countries such as Spain, France, USA, and Switzerland. Now I will try to give some policy recommendations based mainly on sectoral studies. The first policy recommendation is based on the following finding, the lowest score for learning for the textile sector compared to the electric sector for fully exporting firms characterized by subcontracting regimes. This means that subcontracting is likely to benefit more to the emerging economies in the short term, but in the long term, when reaching saturation, the score for learning or the benefits from exporting are likely to decline. So the recommendation we propose here is that industrial policies of emerging economies should consider subcontracting as an intermediary stage for the economic development in order to increase the firm's efficiency, their competitiveness, reduce their technological dependency, and so on, and then move to co-contracting and then entirely finished products with higher added value. The second recommendation is based on the following finding in the agro-food sector, exporting is not driven by efficiency. So the sector might gain much more if export promotion could be increased endogenously through efficiency improvements. One way to do this is to change the structure of agro-food products the sector should move from general quick, easy, profit products towards more sophisticated and industrialized products with higher added value such as food processing. The third recommendation is based on the likely higher productivity gains for firms exporting towards high income countries than to medium and to low income countries. If the agro-food firms aim to acquire the maximum gain from exporting, it could be through extending exports to high income countries. The last recommendation is based on the strong statistical support for the positive impact of the foreign direct investment on increasing firm's efficiency, its incentives to exports, its incentives to innovate in almost all sectors. So the recommendation is to extend incentives given to firms with high foreign involvement than to local firms. Now I will conclude. The first conclusion is the stronger evidence about self-selection for fully exporting firms in almost all sectors indicating the superior productivity of fully exporting firms. The second is the importance of productivity gains from exporting has two driving factors. The first is the dynamics of learning. The scope of learning decreases with the lack of experience exporting as it is the case of textile and electric sector. The second driving force is the export destination exporting to high income countries brings about larger productivity gains as compared to exporting to medium or to low income regions as it is the case of the agro-food sector. Fully exporting firms have high abilities and more incentives to innovate and finally foreign direct involvement generally increases firm's efficiency, its export incentives and also innovation activities in almost all sectors. Thank you for your attention.