 Gwyddo yn dod o'i edrych ar gyffredinol gyda ni'n dweud yma'r gweithio gyda'r hanfodol yn ymgyrch gyda'r hanfodol i gael fy colleg i Jon Rand a Fintarp, ac i'r tyn o'r Cyfu Llyfrgell yn Hanoi. Gweithio'r gweithio ar y cyffredinol, yma'r gweithio yma, rydyn ni'n gwybod i'r cyffredinol ynddau chi'n gweithio'r cyfnodol i'r cyffredinol i'r cyffredinol. We're focusing on the difference in that relationship and the pre-accession to WTO versus post-accession to WTO. We're using a firm level panel data set to do this which is a rich panel data source from Vietnam for the period two thousand and one, to two thousand and ten and we're focused on separating out the productivity effects due to self-selection and compared with the effects that are trying to isolate the effects that are due to learning and learning by exporting. And we also focus on and try and disentangle some of the underlying meccanism here, looking in particular at the trade regime and how that impacts on this relationship. We look at trade costs and protection as two forces that might be driving these relationships. Just to give you a preview of our results, our results suggest that protecting sectors in order to help firms prepare for export markets seems to be a good strategy in promoting participation in export markets but learning is much less likely in these environments and I'll show some results to suggest this. We also have an additional part of the paper which looks at the mechanisms through which learning actually happens in the post WTO period. We explore in particular technology transfers and we find some evidence to suggest that this is one of the mechanisms through which firms learn. So to provide some motivation, this is the second session on learning by exporting at the conference so a lot of this motivation has been provided already and there is quite a large body of literature that looks at learning from exporting and developing country context and one of the big findings here is that it's not clear whether it's selection into exporting that explains the fact that export firms are more productive or whether they start to learn from exporting and this is a productivity spillover that they gain as a result of being involved in export markets. So there's mixed evidence of this across a variety of different African countries and we've seen that today and yesterday also in these sessions. One of the gaps in the knowledge is just looking at the mechanisms through which learning happens and we try and address some of these gaps by looking at trade barriers and protection and seeing you know how that impacts on this relationship as I've said and also try and contribute to the knowledge of how firms actually learn by exporting. So our empirical approach, as we follow the standard in the literature, we first of all try to detect self-selection. So the seminal paper here, the clearly digital paper in 98, proposes two testable hypotheses that are consistent with self-selection. So entry exporters should experience positive productivity shocks in the period prior to entry into foreign markets and also firms experiencing negative productivity shocks should cease exporting in the subsequent period. If we observe this then this provides us with some indication that there's a self-selection process at work. To detect self-selection we just use a standard measure of total factor productivity using an index number of approach and we use this measure to then construct some binary indicators of whether a firm has a positive or a negative productivity shock between any two given years. We then estimate a fixed effects regression model where we look at the decision to enter into export markets and the first instance are the decision to exit export markets. And we regressed this on the TFP shock controlling for a number of factors such as the size of the firm to capture the fact that larger firms may be better able to incur that sunk cost of entry and also factors that might be related to the underlying technology such as labour productivity, capital labour ratios and so on. So in this regression we control for sector fixed effects and all of the usual controls, time fixed effects, problems fixed effects. So if we find a positive coefficient here then we have evidence for self-selection. So to detect learning by exporting we use a one-step approach where we estimate a production function and within that production function we control for the various selection variables that we have included in the previous model at two lags and what we're interested in here is the coefficient on beta one which is the indicator of export status in the previous period and we want to check and see whether that is related to productivity within this model. Of course one of the problems with this dynamic model is that we have the lag of the output in this equation so this complicates the econometric estimation if we want to use our fixed effects estimator. For now we've not gone through all of the steps we're trying to address this because it's an early enough paper. To begin with we just exclude the lag. We then include it and see if our results change and we also estimate the model using this random effects estimator with the one lack adjustment and we find the same result. Our preliminary estimates using the blended bond approach suggests that our results hold but it's not completely crucial. So the data set or the context that we're looking at is the Vietnamese context and of course the Vietnamese economy began to open up back in the 1980s with a range of policy measures that involved trade liberalisation but also liberalisation of investment laws, promotion of foreign direct investment. So there's this gradual process particularly throughout the 2000s that led to the removal of export taxes, non tariff barriers and through the negotiation of various trade agreements which ultimately led to accession to the World Trade Organization in 2007. We see significant growth in exports and imports over the period, besides the dip period in 2009. These are aggregate figures that we have taken from the national accounts data just to illustrate that trend. The data set that we use is the Vietnamese Enterprise Survey collected by the General Statistics Office. The data cover of the population of all registered enterprises in Vietnam with 30 employees or more and we have then a random sample of the smaller firms below 30 employees. What we do here is we focus on a select group from that Enterprise Survey. We focus on trade intensive sectors and this is consistent with what Clerides does so we consider a sector, a trade intensive sector if they're a two digit sector where exports account for more than 10% of output in any one given time period within our data set. We combine our data with the export and import data at the four digit level from the Comtrade database and we also use a panel data set to look at these issues because we want to abstract in this paper from reallocation effects to the typical firm turnover that you see when you have trade liberalisation. We also run robustness checks using the unbalanced panel and we get the same result. In addition we combine our data with a new specially designed module that's part of the GSO survey that we included in the GSO survey in 2010 and 2011 where we interview, add an extra set of interviews to around 8,000 firms looking specifically at technology transfers and asking firms how technology is transferred within their enterprise so we have some additional results that we are analysis that we do using these data. Just to give you an indicator of the export firms in our data set, we don't have an actual variable on whether a firm exports so we go to the tax data so we have a variable on whether they pay export taxes and you will see for all firms this is increasing over the time period and in particular for the trade intensive sectors in the balanced panel so there is a lot of these firms that are actually starting to export over this decade in Vietnam. In terms of the characteristics of the firms that export what we see is that the characteristics are different in the pre WTO period and the post WTO period this is just a simple regression fixed effects linear probability model of whether a firm exports and we see that the firms that export, if we look at the balance panel they tend to be more productive firms that export and that's the case in the pre and the post period and there's no difference pre and post WTO on the balanced panel they tend to be more productive on the TFP measure post WTO but we also find that they're more likely to be far known enterprises so it's just to give kind of a picture of the characteristics of the export firms that are in our data set so the first set of empirical results we present are detecting the self selection and learning by exporting effects so we estimate the model that has whether you're an entry exporter regressed on all of the various characteristics all of the various fixed effects and we're looking at the impact of whether we're looking at whether or not these firms experience a positive productivity shock before they entered the export markets and we see throughout for the entry exporters that this is the case and we see that it is a greater effect in the post WTO period we also do the same for exit exporters and we find that only in the post WTO period is there's this selection back out of the export market following a negative productivity shock so I mean one possible explanation for this is that perhaps in the pre WTO period the sunk cost of entry into export markets is much higher than in the post WTO period where it's more easy to to go in and out of export markets perhaps so as a result even after a negative productivity shock maybe firms are less likely to exit the export markets in the pre accession period. In terms of detecting or learning by exporting we've a couple of various models here so the first two columns are just so we include this WTO interaction in the second column here in this third and fourth column we have the just the level variable and then including a WTO interaction but these controls for selection and here we have the lag of the dependent variable included and in all cases we find this positive relationship between the lag of export and productivity but it's entirely driven by the post WTO period as you can see with the interaction term here with the WTO in all cases so what are our key findings from these results well our productivity differences between exporting and non exporting firms appears to depend somewhat on this prevent trade regime so under the more strict regime pre WTO export firms are less productive and are less likely to self-select in and out they do self-select but they're less likely to do so not to the same extent as the post WTO period under the liberalized more liberalized regime export firms are more productive and self-selection is more obvious so in terms of trying to disentangle this a bit further we're going to look at the the role of the the trade regime so what kind of mechanisms might be in place here that are preventing firms from from being from learning what kind of mechanisms might be in place that are imposed by the trade regime that might make self-selection less likely so we consider two things we consider first of all and the fact that trade restrictions might make exporting a lot more costly than it would be without these trade restrictions and so if this is the case we should observe less selection into exporting in sectors where costs are lower her costs are higher and we proxy trade costs in this analysis using an indicator for low versus high export sectors constructed using the aggregate data on the basis that if you're in a high export sector it's kind of a signal that it's a their lower costs associated with exporting in that sector. We also look at whether protection through the form of barriers to imports in the pre-trade regime might impact on the selection into export markets for some firms so if you have higher levels of import tariffs or if you're in a more concentrated sector that's protected through some other means this could actually give firms more protection and allow them to enter or start exporting as a result and if this is the case we should observe more selection into exporting in protected sectors in the pre WTO period when the costs of exporting are higher. Post WTO selection is probably less likely in the concentrated sectors where they need the competitive pressures that within the domestic market in order to be able to do well on export markets so we should observe a different result there. So in terms of how we do this as I said we construct a variable for a high cost of exports and we interact it with this positive shock to productivity and the negative shock to productivity down here and what we do see is that in the pre WTO period there is less selection into exporting if you're in a high cost export sector and this happens in both periods and so this kind of suggests that this high cost of exporting is one of the reasons that's inhibiting selection. We do the same for a measure of a high tariff sector versus a low tariff sector and what we see in the pre WTO period is that it doesn't make any difference really to our results in terms of entry looking at entry exporters but we see this positive shock on the high tariff sector is associated in the high tariff sector with them so there's a positive association between being in a protected sector and then entering into the export market. So this kind of suggests that selection to a certain extent into export markets is higher where the firms are a little bit more protected from competition. We don't find any effect of another measure of sector concentration that we use just as typical Hirshman Harfin Dell index. So in terms of learning by exporting we'll also look at the role of the trade regime and you know this is just some literature that is looking at the fact that there is heterogeneity and who learns from exporting. So what we want, the question we want to ask ourselves here is why do firms not learn in the pre WTO period but do once the trade is then liberalised. So again we look at the costs imposed by the protectionist regime that may make it more difficult for firms to learn so it's more difficult, they incur much more costs so it's harder for them to learn in the pre WTO regime. We also think about the fact that in protected sectors if you're in a protected sector through tariffs or through just other measures inefficient firms can survive for longer so this might be a reason why we don't see them any learning on average because the inefficient firms are not being pushed out. And also firms in protected sectors might be less efficient due to the fact that they're not exposed to competition and so learning can't happen. So what we do is we do a similar type of model to what we have estimated before but we use the interactions between these other measures that we have that we have been talking about. So first of all in terms of the high-cost export sector doesn't seem to make any difference on our results here. What does seem to make a difference is if you're in a protected sector and this is a very strong result in fact so any of the benefits or any of the learning effects that we observe in the post WTO period are just don't appear in any of the high-tariff sectors. So our result here I guess is that you know learning is not is much less likely where you've got other types of protection in place it's much more likely where sectors are competitive. Okay so let's skip that bit and go to the third part of our analysis which is looking at technology transfers and here we're exploring one of the channels you know there's very little known about learning effects in the literature it's kind of more in more recent years people have been trying to disentangler how learning actually happens so you know what you export matters it seems who you export to matters it also seems and also it seems that export participation is associated with investment. What we do here is use this module that in our data we only have two years of data for this so we're slightly limited in terms of what kind of econometric analysis we can do but what we're looking at is looking at the same relationship as before this learning by exporting relationship and we estimate the same model as before except we disaggregate our export variable into different types of exporting and this is based on a question that we that we ask the firm. So we can disaggregate whether you export final goods or intermediate goods and we can disaggregate this further by a question which asks the firm did you experience a technology transfer as a result of this and when we do this analysis the same analysis as before and we find okay the positive export and relationship between your export and productivity we find that it's holds for both final goods and intermediate goods there's no real distinction there and it holds for firms that have technology transfers and those that don't have technology transfers so it doesn't appear to be that there's you know there is technology transfers happening but it's not the only part of the story it's one part of the story but what we do see is that and when we disaggregate by final goods and intermediate goods it's firms that export intermediate goods and say that they experience a technology transfer that experience this greater learning effect. So this provides some evidence of backward linkages through the international supply chain. So just to summarize then we find that productive firms self-selecting to export markets in both the pre and post WTO regime but learning effects are only observed in the more liberalized trade environment. This led us to kind of investigate this a bit further and there's three key findings regarding the other mechanisms that we've investigated. In terms of self-selection it seems that lowering trade costs will assist in the self-selection process but selection of productive firms into export markets is more likely in sectors that are themselves protected from imports. In terms of learning by exporting we don't find any evidence that the cost of exporting impacts on the learning process but firms in protected sectors are actually much less likely to experience these learning effects. In terms of technology transfers we do find learning by exporting effects are the greatest for exporters of intermediate goods and this is most likely as we saw on the previous table attributed to technology transfers.