Losing Our Lead in Innovative R&D

A new Duke and Harvard study shows that research in China and India is getting more sophisticated much faster than the numbers let on

By Vivek Wadhwa

To hear the National Academy of Sciences tell it, the U.S. faces little risk of losing its technology dominance. In a new study, the NAS says industries such as software, semiconductors, pharmaceuticals, and biotech have adapted well to the opportunities presented by globalization and have kept innovation close to home. After analyzing patent filings, economic data, and prior academic research, NAS researchers report that work being done in India and China is mainly low-level, confined to such tasks as manufacturing and clinical trials. The U.S., the study concludes, will maintain its global lead in research and development if it keeps investing in its research infrastructure. This would be great news for the U.S.—if it presented the full picture.

Our research at Duke and Harvard shows that the reality on the ground is much different. Indian and Chinese companies are performing the most advanced types of R&D for multinational corporations. As a result, scientists from those corporations are rapidly developing the ability to innovate and create their own intellectual property. Our take is that the global technology landscape has changed dramatically over the last decade and that we're at the beginning of a new wave of globalization.

The U.S. can continue to lead, but we need to be clear about our means and the enormity of our challenges. Industries are changing at an unprecedented rate, and we must take the rise of India and China seriously. We also must understand our new competition and work hard to stay ahead—or risk losing dominance in key industries.

Missed Clues

We agree with the NAS that basic supply chains and distributor networks in most industries are becoming fragmented—that they are now more like LEGO blocks. Decades ago, industries tended to be monolithic, with most R&D occurring within national boundaries. Today, in industries such as personal computers, components are designed anywhere in the world, with manufacturing performed mainly in China. American companies such as Microsoft (MSFT) and Google (GOOG) still dominate in operating systems and Web applications, but they are increasingly developing important elements of their software in India.

That's where the similarities between the research efforts of the NAS and and Duke and Harvard end. The NAS research relied largely on patent and R&D investment data. But in our view, the scientists looked in the wrong places and relied on information that was incomplete or out of date.

Take their report on the pharmaceuticals industry. It analyzed U.S. Patent & Trademark Office filings in 2000. They determined that an insignificant 1.1% of inventors filing pharma patents were based in India, and 0.2% were in China. We believe that to understand what is happening globally, you need to look outside the U.S. So in our recent research into outsourced R&D, released on June 11, we analyzed patent applications filed with the World International Property Organization, based in Switzerland. And we used 2006 data, which showed a fourfold rise in patent filings by Chinese and Indian inventors since 1995. We found that 5.5% of these patent applications listed one or more inventors located in India, and 8.4% listed one or more in China.

But patents only tell part of the story. Patents are a good indicator of R&D, but they don't always translate into innovation. And it usually takes 18 months to five years to get a patent. So these are indicators of research investments made years earlier.

Go, Meet, Ask Questions

Another way to chart R&D, of course, is to follow the money. Yet R&D spending data available through the National Science Foundation and other agencies are very limited; they don't include activities such as product design or technology acquisition, for example. And there are hardly any data on international R&D investment by U.S. and non-U.S. companies. The NAS report noted this and acknowledged the limitations of their analysis, which used these data.