Krugman on Cluster's Last Stand

Paul Krugman has an interesting op-ed in the NY Times on industrial clusters, jobs, and economic policy (specifically the appropriateness of the US auto bailout, and it role in the US electoral debate). So I'll use the hiatus in the Euro crisis (a brief one, I expect) to talk about one of the other great applications of complexity theory to economics, the relationship between increasing returns, economic geography, and trade, and what it might mean for current issues like manufacturing job creation/destruction, the rise of China and the decline of the rest...

I needn't mention that Paul got the Nobel Prize for his work on trade, economic geography and agglomeration economies (see his Nobel Prize speech  and 2010 New Economic Geography paper for very readable summaries). And very fine work it was, especially since it made the case in a way academic, neoclassical economists could finally take seriously, using general equilibrium and 'micro-founded' optimizing behavior, which says more about the state of American academic economics than how plausible the argument is. For it is actually a very old argument (and even more, implemented national policy), going back to Alexander Hamilton and Friedrich List (as well as Alfred Marshall), and known as the infant industry and the industrial district. Industrial activity is characterized by a variety of forms of increasing returns - learning by doing, innovation performance curves (e.g. Moore's Law), scale economies, network externalities, rich labor markets, informational spillovers, agglomeration economies due to the presence of clusters of suppliers... The list of possible sources is long and I won't go into details, but the results are similar. Under certain circumstances geographical concentrations of activity will emerge, patterns of specialization arise, a core-periphery structure becomes dominant. It's why cities exist, as well as Silicon Valleys, Wall Streets. It's an old phenomenon: in the 19th century a 50 square mile district of Lancashire, England, produced 80% of world cotton exports, although it was located halfway around the world from its raw cotton sources. It exists at many scales: from small market towns to globally dominant industrial districts or even nations. And whether it comes about is more or less a function of the strength of the increasing returns factors compared to the classical determinants of location: transport costs, factor prices, comparative advantage. And where it comes about can be almost accidental: early adventitious  advantages can grow into insuperable barriers to change.

Now we come to the jumping-off point of Paul's op-ed. An extensive feature in the NY Times on "How the U.S. Lost Out on iPhone Work" examines the question of why Apple moved its production of iPhones to China, and whether these jobs will ever come back to the US. The authors come to the conclusion that Apple's principal contractor there, the Taiwanese company Foxconn (the largest electronics contract manufacturer in the world) and its labor force are so proficient, flexible, and dedicated in what they do, that no American firm could ever hope to compete with them. (The article fails to mention the other reasons Foxconn's Chinese plants have been so much in the news: an epidemic of overworked employees jumping off their rooftops, and the casualties from aluminum dust explosions due to poor workplace ventilation in the iPhone plants - the unsung costs of globalization). It is not so much wages that drove this outsourcing, according to the Times authors, but the Chinese complex of skilled engineers, disciplined workers, and suppliers. In Paul's terminology, in other words, an industrial cluster!

I would not dispute the importance of industrial clusters in explaining much of economic geography. However, it does not seem to be the case that Apple outsources the final assembly of products like the iPhone to China because so much of its supply chain is already located there, as Paul claims.

Quite the contrary: Xing (2011) shows that only 3.7% of the iPhone's final cost of production is due to Chinese value added, based on 2009 data for the iPhone 3G S (original data from iSuppli teardown). The rest comes from imported high-valued electronic components from Korea (12.8%), Japan (33.8%), Germany (16.8%), and yes, the USA (6.2%), with 26.8% of unspecified (non-Chinese?) origin, including Corning's acclaimed Gorilla Glass for the screen, manufactured in Harrodsburg, Kentucky and Shizuoka, Japan . Thus the fact that the entire cost of the finished iPhone shows up as an export from China to the USA is highly misleading, since 96.3% of its manufacturing value added was produced elsewhere, including a substantial amount in the USA (not to mention Apple's 64% markup on unit costs!). And much of that huge Apple margin goes to software development, R&D, and marketing, the bulk of which I presume is still being done in the US.

While there are certainly agglomeration (and scale!) economies in final assembly of electronic goods (after all, Foxconn has huge assembly plants with campuses of highly disciplined and docile Chinese workers, when they are not jumping off rooftops or getting killed by exploding aluminum dust), one shouldn't underestimate how much wage and transport costs are still driving factors. Another recent Times article reports that GE is repatriating water heater assembly from China to the US because the fall of entry-level wages in the US by $10-15 due to the economic crisis now makes it profitable to do so. Thus wage and transport costs still play a substantial role in location decisions, and can even result in manufacturing jobs returning to the US (or migrating even further to lower-wage countries like Vietnam, Bangladesh and Indonesia).

Now come the caveats. The data are from 2009. The Chinese value added may have grown since then. Second, Xing assumes that none of the unidentified 26.8% of iSuppli's teardown represents Chinese production (and even some of the identified Korean, Japanese, Germany and US input may indirectly represent some Chinese production).

Nevertheless. the fact that "Designed in California by Apple, Assembled in China" is printed on iPhones, and that the direct bilateral trade gap between the US and China is so wide, create a misleading picture. Very little of the iPhone is made in China. From a technological and value-added standpoint, one would be more justified in saying "Made in Japan, Germany, Korea, and the USA", in that order, before adding "And then Assembled in China." And in many ways the Chinese contribution is still the least industrial cluster-like part, namely final, labor-intensive assembly. So, no, Apple doesn't assemble the iPhone in China because so much of its supply chain is already there. It really does seem to be just about low wages.

Where the Chinese electronics industrial cluster (not to mention the Chinese automobile cluster) will really be flexing its muscle in the near future is in the rise of domestic Chinese producers of smartphones like Huawei and ZTE . They have been growing very rapidly in the Chinese domestic market but increasing are targeting the export market, either as OEM suppliers or under their own brand names. I have no doubt that they are or will become focal points of true industrial clusters,  while Foxconn is still just the end node of the value-added chain, however proficient it is at what it does.

So what does this imply for a theory of economic geography and international trade?. In a world in which supply chains now twist and turn through a large number of countries, a network or international input-output approach is needed to understand what is happening and who is contributing what, both in monetary and technological terms. Johnson and Noguera (2011) have applied this value-added approach to recalculate the net international flows (of course, making a number of simplifying assumptions in order to do the calculations with the available data). The result is that, on a value-added basis, China's trade surplus with the US is 40% smaller and Japan's is 33% larger than when measured in bilateral terms (all those Japanese components in iPhones assembled in China again - see figure below).

US trade deficit with six Asian countries, measured bilaterally and using value-added concept (Johnson and Noguera, 2011).

And what do the seemingly contradictory stories in the Times about Apple and GE manufacturing in China tell us? To the extent that these companies subcontract work to China rather than invest in manufacturing capacity themselves. they are  completely footloose. Apple can decide, on the basis of shifting currency exchange rates, wage and transport costs, to pack up and move its production tomorrow to another country, conceivably even back to the US, as GE, for much more transport-intensive products like water heaters and refrigerators, already seems to be doing. A purported Chinese industrial cluster and existing supply chain is no constraint (unique high-technology components from Japan, Korea and Germany are more critical, but these can easily be shipped anywhere). It's Foxconn that has made the sunk investments in plant and trained workers in China. Of course, Apple might have a hard time at first finding the massive assembly capacity somewhere else that Foxconn already provides. But remember, an industrial district north of Bangkok manufactures 40% of the world's hard drives (it was badly hit by the recent flooding), so this kind of electronics capacity could and has been recreated in another low-wage country when costs warranted it.

That said, Paul's main point, that the Obama administration acted correctly in bailing out the auto industry and preserving American industrial capacity in the whole value-added chain of auto manufacturing, is well taken. But the iPhone is not really a comparable case, and Apple and Corning have actually done a fine job of at least keeping the highest paid and most innovative work at home (it could have gone entirely to Samsung and Schott).

While industrial clusters are a fact of economic history, the long-term process of creative destruction means that eventually even the most successful clusters decline. Would it have made sense for England to pour money open-endedly into preserving her textile industries through the 20th century against Japanese and other low-wage competition (the same applied even to Japan 50 years later)? Time and the product cycle march on, and wait for no man. The real problem in high-wage countries is to find other employment opportunities in time to maintain international competitiveness (much as Germany has done) before existing clusters collapse. And the growth of domestic and financial services have not done the job in the US.

This tale has one epilogue I will come back to in a later post: China's refusal to let the RMB appreciate to balance out real exchange rates has unleashed a race to the bottom (not even exempting Germany), forcing the rest of the world to lower their real and even nominal wages to stay in the game. This deflationary spiral will prevent the industrialized countries from returning to vigorous growth (more on that later).