 Okay. Thanks very much, and especially to John and the other organizers for having me. So I'm going to give a fairly broad brush interpretation of Mexico's experience with trade liberalization and try to draw some general lessons, especially for Vietnam. I should say that this talk is out of the ordinary for me for a couple of reasons. One is it's much broader brush than talks I usually give. I'm usually interested in papers about the effect of X on Y. I were trying to be very, very precise about what X is doing to Y. So, and it's also out of the ordinary for me. I'm not making any claim to novelty as you'll see as I go. I'm relying heavily on other ideas of others and in some case, quite old ideas. Okay. So why is Mexico possibly an interesting example here, especially for Vietnam, but other countries? So it clearly differs in many ways, but it's illustrative arguably of issues that Vietnam and other countries will face in their future growth as they move into what I think was middle income status and the puzzle of how to keep moving after you reach middle income status. And this is my first time in Vietnam. I don't know so much, but Vietnam has, as other countries, has been signing a number of different trade agreements with the US. There was a WTO accession with Japan. There's agreements through ASEAN, and there's pending agreements with the EU and the TPP, the Trans-Pacific Partnership. So arguably the sort of issues we're going to be talking about are ones that are directly relevant to Vietnam now. With the advantage though that in Mexico, NAFTA is now 20 years old. Liberalization is essentially 30 years old in Mexico. And so we have a chance to take the sort of broad view of what's happened there. Okay. So again, sort of very, very broad view of what's happened. Between 85 and 94, so following macro crisis in the early 80s, Mexico undertook a quite ambitious program of liberalization. So in part that included trade liberalization. So Mexico unilaterally joined the GATT in 85. That continued in 1994 as often NAFTA's often seen as sort of a further commitment to this broader program of liberalization. There was private citizen of state owned enterprises. There was liberalization of investment regime. And there was a general reduction of the role of the state in the economy. So Mexico, I guess the point here in sort of motivating what I'm going to say later is that Mexico was trying to do this all by the book. Basically was doing liberalization, kind of a poster child for liberalization. And the people who are advocating these reforms are quite confident that they would lead to rising average incomes and growth. And the fact is that Mexico's growth over the last 30 years has been quite disappointing. So I'm borrowing some slides from, this is from Gordon Hansen has a piece in the Journal of Economic Literature in 2010. Rafael Bergoying this morning had a graph that looked something similar to this. So basically he's comparing just normalizing in 1980 GDP per capita to be zero. And then looking, comparing to a number of different middle income countries, there are going to be three of these slides. So here's comparing to Latin America. You can see that Chile is just vastly outperformed Mexico. Mexico is the green here. Mexico is closer to Argentina and Brazil. But I think the key point there is Argentina and Brazil have significantly more heterodox policy packages. And so this is not what the advocates of reform would have predicted back when Mexico was undertaking these reforms. The only country that Mexico is convincingly beating is Venezuela, which is not an enormous mark of distinction. If you look at Asian countries, again, so Mexico is down here, so Malaysia, Indonesia, and Thailand are vastly outperforming Mexico. Mexico is down here with the Philippines. Or also if you look at a number of countries in Eastern Europe, again, Mexico was green. You have Turkey, Hungary, Bulgaria are outperforming. And even Romania, which has had a mixed history, is sorry, doing better than Mexico. Okay, now there are obviously a number of reasons why Mexico's growth may have been disappointing, a number of that have been emphasized. So there's monopolies and inefficient regulations. So this is a paper, Jim Heckman is a co-author, underdeveloped credit markets. Santiago Levy is here in the conference, has emphasized informality and evasion and how the policy regime may foster those. Corruption is obviously a problem, and more is violence is obviously a problem. All these likely played a role, I don't want to discount them. But I do want to focus on trade and what I'm going to call the links between the pattern of specialization and innovation. Okay, so I'm going to talk about the intra, what I'm going to call inter and intersexual resource shifts. Looking at, at, at, at sectoral reallocation in Mexico. I'm going to try and relate that to, and then relate that to the sort of overall rate of innovation. Okay, now, I guess already a caveat about this. So evaluating the effect of trade policy is, is, is extremely challenging. It's challenging in many situations, but especially challenging in Mexico here, partly because NAFTA, which was undertaken and were implemented in January 1994, was followed very quickly by, by a major pace of devaluation and the pace of crisis, started in December 1994 and continued into 1995. As, and Kruger and others have noted that the, the, the nominal devaluation was just much, much larger than the tariff changes under, under, under NAFTA. There may also have been sort of lagging effects of this mid-80s liberalization of Mexico, which was in many ways probably larger than, than the, than the changes under NAFTA, both in, in terms of tariff changes, but also in the removal of non-tariff barriers, etc. So essentially, I'm going to lump all these things together. I'm going to look broadly at what we think this opening of the Mexican economy, which had been quite closed prior to 1905, is now quite open at broadly what that opening has, has done to, to, to sectoral shifts. Okay, so here, so I'm using data from the Mexican industrial census supplemented by data from a 1999 survey called the NST, sort of a special survey where we can also on manufacturing firms. So the, here on the X axis, you just have the share of workers with more than, greater than or equal to 12 years of education as a measure of skill intensity by, by industry. Each dot is a four digit NAICS industry, so that's the industrial classification used in North America. And then we have the change in log employment between 1988 and 1998, the, the census that are every five years, 88, 93, 98, etc. Okay, I'm focusing, you can see there's highlighting the size of these, of these symbols is, is just the size of the sector. I'm going to focus on a few. These, these red diamonds are in apparel and textile products. The, the blue triangles are transportation equipment, which is largely auto and auto parts. And there's also various electronics industries are the, are the green squares. And then we have other, other industries. Okay, so what message are we taking from this? This is basically just saying that from, at least from 88 to 98 was this period of initial period of liberalization. What you see is that there's a, there's a sort of roughly downward sloping relationship. That's saying that the industries that grew the most were the ones that were least skill intensive, which is basically exactly what we'd expect from a standard Hector, Hector-Olean type model. That Mexico should integrating with the United States. That was by far the major, major trade partner. You'd expect Mexico to specialize in, in unskilled labor intensive activities, and that's more or less, more or less what happened. I've also done work, other work, sort of looking at effects within industries. There's sort of differential effects within industries. There's some pressure to upgrade within an industry. But for the most part, I'd, I'd even argue, arguing against my previous work, the first order effect is just the sectoral shifts in the direction that we'd expect from the, from a standard Hector-Olean type trade model. Okay, if you look at capital intensity, so if you just take, this is just the capital labor ratio. This is also from the economic censuses. There you can see the relationship even more starkly. Okay, so it looks like the least capital intensive industries are the ones that between 1988 and 1998 are growing the fastest. Okay, then what happened in Mexico? So this is now from, graph is in every way similar, except now we're going from 1998 to 2008. And you can see basically those industries, so compared to this one, and especially that big square, that's, that's apparel, that big diamond, I mean, just tanked, just really didn't do well at all from 1998 to 2008. We'll talk a little bit about why that is. Okay, but you can see that, that these, here, these sectors in general, sort of at the low skill ends, had very underwhelming growth over the later period. If you look at the capital intensity, you see something, see something similar. Now part of what's going on there, which John alluded to, is what's happening in the McElodora sector. So the McElodora's are these sort of assembly sector, they are sometimes built in bond plants where they get relief from import duties. If goods are subsequently exported, you have to be a McElodora, you have to register with the government, and then you have to follow certain rules and provide data to the government on a monthly basis. So here, the main McElodora sectors are apparel, electrical and electronic equipment, and then transportation equipment, which is mainly auto parts. Here, I'm using two different sources of data. The top one would be all, it's from the economic censuses, where we only have data every five years, and that's employment, total employment in the sector. And then for the McElodora's, we actually have yearly data, or you could even get monthly data if you wanted. So that's the following. So that's saying, so here we have total employment apparel, and here we have McElodora in apparel. And the key thing to notice here is you have a big boom. So McElodora employment in general, in particular apparel employment, is often credited with sort of getting Mexico out of the pace of crisis and the crisis in the mid-90s, okay? So clearly, employment grew very quickly at sort of what people were expecting, and advocates of reform were quite excited about that, and there's no question that that helped cushion the Mexican economy during this big crisis. Here you can see, and then what happened is, so basically following 2000, you see a major decline, okay? So a big part of that was that big diamond that looks so low in the second graph when looking cross-sector was, in part, this apparel sector has just taken a major hit. Electrical and electronic equipment, this is slightly misleading in the sense that here I only have data every five years, and so it looks like McElodora employment exceeds total employment. That's obviously not what's going on. There would be a bump there if I could see what's happening in between years. But there again, you can see pretty strong growth between 95 and 2000. That's actually, sorry, 93 and 98, and then stagnation afterwards with a little bit of recovery, so not quite as severe as an apparel, but anyway, certainly in this period there's not very strong growth in employment. The one sector sort of McElodora that seems to be doing okay is transportation equipment, so auto parts. So many of those auto parts factories are integrated geographically into a North American production system for autos, and so that sector has been more, I'd say, more robust growth. What can we say about this, about the McElas relative to the non-McElas? So here this is comparing, so we're going to have McElodora sector versus non-McElas, and then within the non-McElodora sector, we're going to compare non-exporters and exporters. This is just sort of means, and this is from this 1999 NSTIC survey which has the advantage, unlike other Mexican datasets that it, or the sort of the high-frequency Mexican datasets that it combines McElas and non-McElas in the same dataset. So what do we see here? So the McElas tend to be large. They obviously export almost all of their output, although they don't have to. They're highly, high share of foreign ownership. Now, okay, if we're getting here, you can see for a capital labor ratio, which is here, so much, much lower than even the non-exporters in the non-McElas sector. So they're clearly focused in labor-intensive activities. When you look at the share with greater than or equal to 12 years of schooling, again, that share is much lower. The percentage blue collar is higher than in the non-McElas sector, the sort of non-exporters or exporters in the non-McElodora sector. Years of schooling is also lower. So what's clear is not surprising to anybody, but this tends to be the least skilled sort of phase of the production process that McElodoras are specializing in. Interestingly, I guess giving the full picture here, is their wages are not necessarily lower. You can see the wages are closer here to the exporter wages in both cases, which is interesting. Partly that's because McElas tend to be focused in the northern part of Mexico, where wages are a bit higher than in the center in the south, but it's also, I think even when you look within the north, wages are reasonably high relative to the non-exporting non-McElodora sectors. Okay, so what's my basic message so far in terms of just what happened? So from 88 to 98, manufacturing specialized in less capital, less skill-intensive activities, both across sectors and within sectors, in the sense that there was a big shift towards this McElodora type production. From 98 to 2008, these sectors and sub-sectors tended to stack, okay? And so then, now as I mentioned before, there's some sort of other effects within industry that partly, then in my own work I've tried to emphasize, but I think when you look at this broad story of what was Mexico's response, I think the first two points are really the first order facts. Now, okay, so the question is what happened? I'd say probably the most common explanation for what happened is that Mexico, well, that's China entered. And so people in Mexico often think of this as we had bad luck, right? This program would have worked, liberalization was a good idea. It just so happened that we liberalizing at the same time that China entered the world trading system. And in particular, this is a more severe problem for Mexico, but for other countries in part because the pattern of specialization, the goods that Mexico was exporting to the US are highly correlated, essentially with the goods that China is exporting to the US. Okay, so there's been various work on that. I have a graphic and show if you want. That's clearly true. And I don't want to say there's no truth to this story. There clearly is a lot of truth to this story. So there's a number of papers. This Uttara-Zuiz paper is looking at the effect on Mexican microdures of Chinese competition in the US market. They had an impact. This is a graduate student of mine who's just finishing, who's looking at using an approach of Otto Dornan Hansen to look at local labor market effects in Mexico of Chinese competition. He's finding strong effects that regions in Mexico that had a lot of firms that were competing directly with Chinese firms tend to see lower employment growth. There are a number of different papers in the trade literature. This Hansen and Robertson are saying, oh, so looking sort of slightly more structural approaches, looking at the effect of China on Mexico. Okay, what I'd like to argue though, and here slightly more speculative than the sorts of arguments I usually make. But what I'd like to argue is that I think that Mexico would have had problems even if China had not entered. So we don't observe that counterfactual, obviously. But I think there's something to support this story and I'm going to tell you why. Okay, and this is a very, again, here's this very old fashioned idea. Going back to prebish and then sort of formalized by Matsuyama, Hansen and Rodriguez Clair, sorry, and this is not Harrison and Rodriguez Clair in a handbook of development economics sort of literature view make this point and summarize it well, but basically different activities are associated with different inherent rates of innovation and productivity growth. And liberalization, which may bring about sort of reallocation to sectors that have a static, when I think of as a static comparative advantage, may affect the rate of long-term growth and innovation. Okay, so it's a very simple idea, but I happen to think that it applies quite directly to the Mexican experience. Okay, so let me show you a little bit of evidence of that. The, I'm going to use a measure of R&D from this NSDIC survey, this sort of, which I mentioned before. So the survey asked, this is in 1999, and asked, since 1997, has this established an undertaken R&D. And I like the survey a lot because it used a quite broad notion of what R&D is. It's not just patenting, not just something new to the world. But you can see it says, if yes, if they answered yes, they said, what did the R&D principally consist of, which could be design of new products, process improvements, product quality improvements, design, improvement in manufacture of machinery, equipment, or other. So it's quite broad view. Basically, anything new or innovative, sort of include what I think of as upgrading, not necessarily coming up with something patentable. So I like that, it's not perfect, but it's not bad as a first pass. We're just going to code that as zero one about whether the firm had any R&D or not, okay? And then I'm going to take an average at the sector level, this four to the sector level, and plot that using a plot similar to the ones I showed before, okay? So this is just a cross-section, I just have one year of data for this. I don't have it in the other years of the, or at least the previous years of the NSD. So I can't do it before and after. But anyway, in 1999, so what does this look like? Recall, this is this big red diamond. So recall what we said is that in that 88 to 98 period, the sectors that were growing fast were the ones down here, the low skill intensive, okay, and low capital intensive. Those sectors are sort of fairly systematically doing less upgrading than the more skill intensive or more, if you do more capital intensive, it's even stronger, more capital intensive sectors, okay? So again, the same thing, you see that the sectors out here are doing more of this R&D, the sectors down here, which were the ones that the Mexican economy was specializing in, were doing less, okay? So that, I mean, mechanically is going to lead to less R&D overall than would have taken place in the absence of those sectoral shifts. Another way of seeing this, you could see, so this is just R&D, this is looking at McElias, this is kind of, again, giving you, make sure you get the full view R&D intensity by sector. So if you look at all, I guess, all manufacturing for the non-exporters or the exporters of McElias, so McElias are sort of intermediate between the non-exporters and the exporters. It's not the case that McElias are worse than the non-exporters, but they're sort of intermediate. Now, in part, that's because, so in apparel, actually, McEliadores are sort of comparable to exporters. But the apparel sector overall has less R&D or less upgrading than other sectors, and that's partly what's bringing this number down. Okay, a couple of other ways of showing this, just to, if you don't like my, that R&D measure from the Anastique. This is from a book by Laderman, Maloney, and Cervéna at the World Bank. So if you look at, this is just patents per million workers. And here they've got Latin American, the Caribbean, sort of on average, versus Mexico, okay? And you can see there, so in the 60s, Mexico has actually outperformed on this alternative measure, outperforming the rest of Latin American Caribbean. By 1995 to 2000, it's actually significantly underperforming, because in terms of patents, okay? They have some other comparisons. Korea is obviously a star here, and here, notice that the Y-axis are changing, but the patents for workers are significantly below also other high income countries. Another way of looking at a simple way, and again, this is from downloading from the World Development Indicators, just the R&D spending as a fraction of GDP. You have Chile and China at about 0.65%. Korea obviously is a star up there with the US and Canada, but Mexico is significantly below, okay? Which doesn't say there are other countries, if you look at Argentina or look at Brazil, they're also down there with Mexico. But Mexico's being outperformed by Chile and China, okay? So summing up, what it seems to me is that integration led Mexico to specialize in less capital and skill-intensive activities, both across and within sectors. These sectors tend to be less innovative. This did not have to be true. I wasn't actually convinced that this would be true when I first did those plots of the innovation rate or the upgrading rate against capital-intensive or skill-intensity. But it turned out to be, the correlation appears to be quite robust. So these sector shifts tended to dampen the overall rate of innovation in the economy. What if China had not entered? So we don't, again, observe the counterfactual. But my sense is that even if there hadn't been China, there would have been other countries moving up the quality ladder and competing with Mexico. So that would have been Malaysia, Thailand, Indonesia, maybe even Vietnam, would have been entering. So you can't not innovate and not have productivity growth for long before you're out-competed by other countries, okay? So more research is needed and needless to say. If I'd been asked by Tony Addison yesterday what graduate students should work out, I'd say go out and figure out what works and what doesn't work in industrial policy. So we don't have a lot of evidence about that. Or we don't have a lot of what are called sort of rigorous econometric evidence about what works and doesn't work. But the pattern suggests that there may be this trade-off between what I call static, allocative efficiency, and long-term productivity growth. So clearly there was some benefit in a static sense of reallocation towards sectors in which Mexico had a competitive advantage, right? But it may lead to lower growth in the long run. And so now that we're in the long run, arguably, that may be what we're seeing. And so maybe that trade-off needs to be taken into account now when you start thinking about entering the liberalized trade regime. So liberalization alone may not be enough to bring about sustained growth. That's another way of saying that, saying that same point. I think that there's a strong case which Justin and John and many others have made also for interventions to promote the sorts of activities that generate innovation and productivity growth, both across and within sectors. Underlying that, there's an argument about externalities. Okay, so you somehow firms have to not be taking into account the effect of their own investments in innovation. I'm working on other, have other work, especially in Pakistan with some soccer ball producers, which I'm happy to tell you about at length during a break or something, about that working on that. Anyway, so some caveats. And this is why this relates to another point that John made about trade policy versus industrial policy. So these interventions do not have to be. So if we agree that there's a case for intervention to subsidize innovative activities, it's not clear these interventions have to be at the border. Okay, so there's a well-established principle that it's actually better to address a market failure or something like this at the source. In this case, you'd go to the innovative activities and subsidize them directly in some sense. And obviously, WTO accession limits the ability to conduct industrial policy through tariffs or other restrictions at the border. So we should, in that sense, I think my argument is really about industrial policy rather than trade policy. The other caveat, which is again not a new point to make, is that it's true that government officials have no special knowledge about which sectors or firms or ideas or technologies are likely to be successful in the future, okay? So many of the best-intentioned officials may be less well-informed than the business people in the sector, although they have different incentives and different objective functions arguably, but they may have less information. Okay, so I don't know how much I've got, a minute or two minutes or something, zero. Okay, let me do at least the next two slides, and then maybe we'll stop. So one common approach, which is that later Manamaloni have a new book, which is arguing is that what you do is you do broad-based policies which provide broad support for innovative activities without having to pick which particular firm or sector is going to be successful. That seems to me like a very good idea. So you've got infrastructure or technical vocational education, or another one would be removal of restrictions on imports of high quality or high tech imported inputs. That seems like no brainer, but that would be in the spirit of this argument of doing things that broadly will support innovation. Normally these policies along these lines are already in place in Vietnam, which we can talk more, but I don't have very much time, so I won't do that. Now, so the other alternative would be targeted policies. Now, targeted policies as we all know are inherently more subject to corruption or capture or plain old, you know, sort of mis-targeting, but if implemented well, they can be very effective. So again, this is, I was reading up on this in preparation for this talk, but so Vietnam has this law on corporate income tax and law 31 where we have a sense of preferential tax treatment for investments in a number of areas. And so this is all very much in the spirit of trying to promote these innovative activities. So in high technology, scientific research, technology development, I mean, set our high grade steel, energy saving products, et cetera, et cetera. Okay. So I think, you know, I formulated the interpretation of the mixing experience before finding out what Vietnam was doing, but I think from the mixing experience, you'd have to say these are the sorts of, we do want to think about these sorts of policies. Obviously, they're caveats. I'm sure they're ways that they could be implemented better, but there's, I'd say that there's a strong case for these sorts of policies. It's not a blanket case for it. Blanket endorsement of intervention. Obviously, there are many interventions that produce more distortions and rent seeking than growth, but I think that the fact that, so not that there should be trade policy to promote innovative activities, but the fact that Vietnam is liberalizing trade means that direct support for innovative activities may be more salient, or maybe even more necessary now than before. Okay. So go out, the graduate students either go out there and tell us what works and what doesn't work in trying to evaluate some of these policies. I have one idea that we could talk about, but maybe I'll stop and I'll let you ask me about that in the question and answer. Thanks.