Skip to Content

Questioning Japan’s Miracle

Divided Sun: MITI and the Breakdown of Japanese High-Tech Industrial Policy
January 1, 1997

For over two decades, economists and political scientists have cited Japan to show just how effective a strong, well-designed industrial policy can be. Chalmers Johnson, president of the Japan Policy Research Institute, ably represented this point of view in his 1982 book MITI and the Japanese Miracle: The Growth of Industrial Policy, writing that “Japan’s postwar economic triumph-that is, the unprecedented economic growth that has made Japan the second most productive open economy that has ever existed-is the best example of a state-guided market system currently available.”

But more recently the Japanese system has also met with a fair amount of criticism, and in Divided Sun, Scott Callon, the chief market strategist of Bankers Trust Asia Securities in Tokyo, weighs in on that side of the argument. He notes that in the case of Japan, government’s ability to combine cooperation with competition has seriously disintegrated-that in fact the process of disintegration was already under way when glowing assessments like Johnson’s appeared. “The elaborate structures to promote cooperation that appear in Japanese high-tech consortia are often nothing but a public show: seemingly cooperative institutions mask an underlying reality of fierce competition and conflict,” he says. And the evidence he uses to support this charge comes from four different Ministry of International Trade and Industry (MITI) case studies extending from 1975 to the present time: the Supercomputer Consortium; the VLSI (Very Large-Scale Integrated Circuits) Consortium, which concerns itself with advanced semiconductor technologies; the Fifth Generation Consortium, which is designed to make artificial intelligence breakthroughs; and TRON (The Realtime Operating System Nucleus), an ambitious bid to revolutionize personal computers.

Callon concludes that cooperation absolutely cannot be forced upon companies that compete in the industrial sector. One of the most vivid demonstrations of what can happen if it is occurred when engineers from NEC and Hitachi came to a Fujitsu research facility to work together on the MITI Supercomputer Consortium, he reports. The NEC and Hitachi engineers found that they were forbidden to use the bus that connects the research facility with the train station lest they overhear important Fujitsu trade secrets. For similar reasons, they were instructed to stay in their respective rooms, unless they had to go to the bathroom. More significantly, the high degree of cooperation the project required meant that Fujitsu had to provide detailed, highly proprietary information on an existing product to NEC and Hitachi, which was something “it ultimately could not bring itself to do.”

The author points out, furthermore, that when cooperation does work for companies in a MITI-sponsored consortium, it is because competition among them is not a major difficulty anyway. Of all the high-tech case studies in Divided Sun, the VLSI Consortium is arguably the only one that can be considered even a partial success, and the main reason for this, he says, is that MITI provided the Japanese firms in it with large-scale financial subsidies at a time when they were in a financial crisis and faced increased technological competition from IBM. Under the circumstances, the companies were willing to overlook their corporate differences and work together. Such a situation is unlikely to occur today: ever since the 1980s, when Japan became a leading technology-driven economy on par with the United States, it has been clear to Japanese high-tech firms that their biggest competitors are one another.

A Conflict of Goals

One more lesson Callon draws from his observations is that a conflict of goals and institutional priorities is to be expected when private firms and government agencies become partners. The difficulty is particularly evident in the experience of the Fifth Generation Consortium. Callon notes that the consortium placed a huge bet that general-purpose machines were too slow ever to be good engines for artificial intelligence (AI), and that it would be essential to build specialized computers to operate particular AI applications. As it happens, this was the opinion of the AI community in the United States as well, and a number of companies like Symbolics  were founded to develop and build specialized AI computers. By the early 1980s, these companies had booming sales and profits.

Then, just a couple of years later, general-purpose Unix workstations arrived on the scene, built first by Apollo Computer and later by Sun Microsystems. Unlike specialized AI computers, Unix workstations are relatively inexpensive to maintain, thanks chiefly to standardized parts, and their standardized operating system, AT&T’s Unix, allows for ease of programming. As they rapidly gained market share, Japanese companies in the Fifth Generation Consortium, whose goal is to develop products that would actually sell in the marketplace, wanted to shift the focus of their research to these machines.

Unfortunately, however, government bureaucrats felt more comfortable sticking to the original technological road-map, even if the route it designated contained no clear signs and was full of potholes. Why? Largely because it made better sense politically. As Callon puts it, “the [Japanese] taxpayers and Ministry of Finance, the guardian of the national purse, want to see something in metal in return for their yen.” Specialized AI computers would satisfy that requirement; general-purpose workstations, whose technical sophistication is largely hidden away in their software, would not.

Moreover, MITI was feeling constrained by international trade pressures. The United States had traditionally restricted its complaints about Japan’s industrial policy to the “elaborate web of tariffs, quotas, and regulatory approvals” that made selling products in that country so hard, but by the mid-1970s, American policymakers realized that Japanese imports into the United States were posing a problem as well. The value of the Japanese goods sold in the United States that was not balanced by U.S. goods sold in Japan took off around that time, doubling about every two years through the mid-1980s until it flattened out at $40 to 50 billion annually. A project like the Fifth Generation, which emphasized more futuristic research and therefore posed no immediate threat to U.S. trade, tended not to heighten tensions. This made it attractive to MITI, which, as a government agency, had to worry about such issues in a way that Japanese firms did not.

In fact, when MITI did decide to back a project that the United States would have found threatening, it made sure to cover its tracks. Unlike the VLSI, Supercomputer, and Fifth Generation projects, which are funded, organized, and managed primarily by MITI, the TRON Consortium is privately funded and organized, with limited MITI involvement. “TRON was deliberately aimed at creating something new, something Japanese, that would erode Microsoft’s and Intel’s lock on the world personal computer market,” Callon explains. “As an official government effort, it would have been perceived as yet another example of unfair Japanese business/ government alliance fostering competitive advantage and would have inflamed trade tensions with the United States, so it could absolutely not take the shape of the other three, official MITI consortia.”

While Callon is an invaluable source of insight and historical perspective, his habit of equating the failure of high-tech consortia with the failure of Japanese industrial policy as a whole is annoying. Because of it, the most profound effect of Divided Sun may lie in providing ammunition for those who contrast the experience of the consortia with the success of U.S. firms like Microsoft and Intel, and go on from there to conclude that industrial policy does not work in general and certainly will not work for the United States.

The trouble with this line of reasoning is that it overlooks some significant and worthwhile components of Japanese industrial policy-particularly in “sunset industries” like shipbuilding, coal, and textiles and in the relatively low-tech consumer electronics sector, which encompasses such widely used items as faxes and VCRs. In these areas, where little, if any, new ground is being broken, cutting-edge research is not nearly as important as it is in the realms that have spawned the consortia Callon describes. As a result, policy initiatives can be relatively inexpensive, and success is far less likely to depend on a single product or technology.

Put simply, the stakes are lower, so that cooperation among firms is more feasible, and MITI and related agencies like the Japan External Trade Organization (JETRO) have worked well to exploit that possibility. Their efforts are particularly visible in emerging Asian economies like Thailand, Malaysia, and Indonesia, where JETRO offices, in cooperation with Mitsui, Mitsubishi, and others, often serve as one-stop information centers, providing technical assistance, investment advice, and joint-venture leads to fellow Japanese firms.

A mere recognition of this success could help make the industrial policy debate much more meaningful and substantive, yet as observers continue to swing from the notion that MITI can do nothing wrong to the similarly simple-minded notion that MITI can do nothing right, no such recognition seems imminent. U.S. economists appear not to have even grasped the key fact that MITI, while hardly omnipotent, nevertheless has much more clout in Japan than, say, the U.S. Department of Commerce and the U.S. trade representative have in the United States. The Sun, in other words, may indeed be divided-and Callon may have done a commendable job of showing why and how it is divided-but there are also some ways in which it is whole. And until analysts take a look at them, our understanding of what industrial policy can and cannot do will remain sadly limited.

Keep Reading

Most Popular

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

This baby with a head camera helped teach an AI how kids learn language

A neural network trained on the experiences of a single young child managed to learn one of the core components of language: how to match words to the objects they represent.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.