In the first part of this mini-series, I surveyed how robots, automation and other productivity-enhancing technology is affecting industries at the heart of the economy as well as some more exotic science fiction examples. In the second part, I considered the “Solow paradox¹” — the strange combination of breathless innovation with stagnant² labour productivity (though not so strange when you realise there is not that much investment in capital embodying³ the new technologies).

在这几篇迷你系列文章的第一部分，我调查了现今机器人、自动化以及其他提高生产率的技术是如何影响经济核心产业以及一些更奇特的科幻小说例子。在第二部分，我考察了“索洛悖论”(Solow paradox)——令人喘不过气的创新与停滞不前的劳动生产率的奇怪组合（不过当你意识到并没有那么多体现新技术的投资时，这就没那么奇怪了）。

Today, we focus on the question of greatest political consequence⁴. Who stole the jobs — was it robots or foreigners? Or less tendentiously, was the falling number of manufacturing jobs in rich countries caused by trade liberalisation or by automation and other productivity-enhancing technological⁵ change?

现在，我们把重点放在这带来的最大的政治影响上。谁偷走了工作——是机器人还是外国人？或者不那么带有倾向性地来看，富裕国家制造业就业数量不断下滑，是由贸易自由化或自动化和其他提高生产率的技术变革造成的吗？

For it is largely manufacturing jobs we are talking about. The US is special in that overall employment rates (for all jobs taken together) have fallen since the turn of the century, and for longer than that among men. As Jason Furman and his colleagues have documented extensively, the US faces an especially aggravating⁶ version of a more common problem in which manufacturing jobs have not only disappeared but failed to be replaced by anything at all.

我们主要讨论制造业就业。美国的特殊之处在于，自从进入21世纪以来，总体就业率（所有工作都算在一起）一直下滑，并且下滑时间长于男性就业率下滑的时间。正如杰森?福尔曼(Jason Furman)及其同事记录的大量文档所表明的那样，美国正面临着一个比较普遍的问题的严重版——制造业岗位不仅消失了，而且没有得到任何替代。

So are robots or trade to blame? The simple and largely true answer is: both. But there is still a question of their relative importance, and of what exactly the blame entails⁷.

那么到底该怪机器人还是贸易？一个简单而大致上正确的回答是：两者都怪。但现在还存在另一个问题——两者相对的重要性、以及它们到底罪在哪里。

All industrialised countries have been shedding labour in manufacturing for decades, a process that started before the wave of globalisation in the 1990s. It is clear that the balance of trade has little to do with it: the similarities between structural⁸ change in employment in perennial-surplus Germany and permanent-deficit US are much greater than the differences.

数十年来，所有工业化国家的制造业都在减员，该过程始于上世纪90年代全球化浪潮开始前。贸易差额明显与该问题无关：常年处于贸易顺差的德国与始终处于贸易逆差的美国，在就业结构变化方面的相似点远远超过差异。

But the growth in overall trade that accompanied the regional and global trade liberalisation during the three decades before the global financial crisis will have the effect of changing the employment and production structure of the opening economies — indeed that is part of the point of lowering trade barriers. Standard theory predicts that with more open trade, countries will specialise more intensively in production that makes most use of their relative endowment of labour, skill, capital and natural resources.

但全球金融危机发生前的30年里的区域和全球贸易自由化带来的总体贸易增长，将会改变开放型经济体的就业和生产结构——实际上这正是减少贸易壁垒的部分意义。标准理论预测，随着进一步开放贸易，各国将更集中地专攻最大化地利用它们在劳动力、技能、资本和自然资源上的相对禀赋的生产。

Recent research by Adrian Wood measures to what extent this has indeed happened. As the table below shows (for more detail, look up the background paper) the share of manufacturing in global production and employment fell noticeably in the three decades from 1985. But different regions went through dramatically different changes. In particular, in most land-scarce regions (particularly prone⁹ to specialise in manufacturing, according to theory) the share of manufacturing in the economy expanded, while it shrank in all land-abundant ones.

艾德里安?伍德(Adrian Wood)最近的研究衡量了这种情况发生的程度。如下表所示（若需要更多细节，请查阅相关论文），在1985年之后的30年里，制造业在全球生产和就业中所占比例明显下滑。但不同地区经历了截然不同的变化。尤其是，在大多数土地稀缺的地区（根据理论，这些是尤其倾向于专攻制造业的地区），制造业在经济中的占比有所扩大，而在所有土地充裕的地区，制造业在经济中的占比缩小了。

Wood suggests this shows that the dramatic changes in manufacturing employment can be laid at the door of economic globalisation. But the story is not as simple as that. Look where the biggest changes in employment shares happened. Among rich countries (OECD members), the manufacturing employment share fell just as much in land-rich and land-scarce economies. The output share increased in land-scarce ones — but in conjunction with the loss of manufacturing jobs, this is surely an effect of automation and technology. Meanwhile the two other regions with particularly large structural changes were the Soviet¹⁰ sphere, which in 1985 had an overgrown and inefficient¹¹ manufacturing sector¹² that collapsed¹³ under its own weight once the economy was liberalised, and China, whose liberalisation and trade integration¹⁴ surely contributed to its industrial revolution.

伍德认为，这表明制造业就业的急剧变化可以归咎于经济全球化。但事实没那么简单。看看哪些地区的制造业就业占比变化最大。在富裕国家中（经合组织(OECD)成员国），土地充裕经济体与土地稀缺经济体的制造业就业占比降幅相当。制造业在土地稀缺经济体的产出中的占比增加——但同时制造业就业流失，这肯定是自动化和技术的影响。与此同时，另外两个结构变化特别大的地区是前苏联地区和中国。前者在1985年制造业发展过快、效率低下，在经济进行自由化时被自身重量压垮，而后者的自由化和贸易一体化显然促进了工业变革。

What all this points to, then, is a process in which many poor countries went from a pre-industrial employment structure to an industrial one (but some stagnated¹⁵, in particular in Africa), and in which all rich countries largely went from an industrial to a service-based employment structure. The poor country transition or lack of it no doubt owes a lot to the ability to enter the world trading system. In the rich country transition this looks more like what you would expect from the continued growth in manufacturing productivity, echoing what had earlier happened in agriculture.

那么，这一切都指向了很多贫穷国家从工业化前就业结构到工业化就业结构的过程（但有些国家的这一过程陷入了停滞，特别是在非洲），以及所有富裕国家基本上从工业化就业结构到以服务业为基础的就业结构的过程。贫穷国家实现转型或未能转型，无疑与它们进入世界贸易体系的能力有很大关系。而富裕国家的转型，看起来更像是制造业生产率持续增长所必然带来的，与之前农业出现的情况相似。

And conversely, there is direct evidence for the automation thesis. A new study by Daron Acemoglu and Pascual Restrepo tries to measure the effect on US jobs (and wages) of the increased use of industrial robots. An interview with Acemoglu about the findings puts the number of manufacturing jobs lost because of robots at 670,000 between 1990 and 2007.

相反，有直接证据支撑了自动化造成制造业失业的论点。达龙?阿西莫格鲁(Daron Acemoglu)和帕斯夸尔?雷斯特雷波(Pascual Restrepo)的新研究试图衡量，加大使用工业机器人对美国就业（和工资水平）的影响。在一篇就研究发现对阿西莫格鲁的采访中，1990年至2007年因机器人而造成的制造业失业数字为67万。

What lessons can we draw? First, that both trade and automation play a role. But second and more importantly, that the two cannot be neatly¹⁶ separated — automation-driven productivity growth and attendant job loss may be both an ultimately unavoidable part of economic change and be accelerated by trade liberalisation. Rich economies with highly-skilled labour forces are well placed to respond to greater trade by specialising in higher-value added products — just those where automation can do the most to increase productivity. Hence, for example, the success so far of the US car industry, which produces more vehicles than ever and exports finished cars to China.

我们能得出什么结论呢？首先，贸易和自动化都发挥了作用。但第二点也是更为重要的一点是，二者无法彻底分开，由自动化驱动的生产率增长以及随之而来的就业流失可能都是经济变革最终不可避免的一部分，并且被贸易自由化加速。拥有高技能劳动力的富裕经济体有能力通过专门生产更高附加值的产品，应对贸易扩大——在更高附加值产品领域，自动化最能提高生产率。例如迄今美国汽车业的成功，其生产的汽车数量超过任何时候，并将整车出口到中国。

But third, that a protectionist trade policy may not do much good even if trade was part of what eroded¹⁷ certain jobs in the past. For if trade helped automation along the way, it is not as if restricting it will wind automation back. At most it may delay further automation, but that will come at a cost. In particular, it will make it harder to export manufacturing goods into a global market that uses the most cost-effective techniques. Trade sceptics who aim to protect manufacturing jobs should be alert to the distinction of protectionism and mercantilism. While the latter aims to boost exports, the former, by restricting imports, may well hold back exports, too.

但第三，尽管贸易曾经造成过去某些就业的消失，但保护主义贸易政策可能也不会有大好处。因为，即使贸易曾经在这个过程中帮助过自动化，限制贸易也不会让自动化倒退。它充其量可能会推迟进一步自动化，但这是要付出代价的。特别的，它将加大把制造业产品出口到利用最节约成本技术的全球市场的难度。试图保护制造业就业的贸易怀疑论者应注意保护主义和重商主义的区别。后者旨在扩大出口，而前者由于限制进口很有可能同样阻碍出口。