Despite the ubiquity of the Internet, innovation still happens mainly in hubs, where face-to-face contact matters more than ever.
Damascus had steel. In Venice, it was glass. Switzerland made watches. For hundreds of years, regions developed specialities that often arose from access to a natural resource, but then intensified as people moved to the regions to be among the expertise.
The Internet was supposed to change all that. Around-the-clock connectivity that allowed researchers and entrepreneurs to collaborate from anywhere at any time meant that distance would no longer be an issue, predicted popular economic theory of the early 2000s. A decade later, it hasn't panned out that way.
Clusters of related and interconnected companies are stronger and growing more quickly than ever, innovation experts say — a trend that seems to be, paradoxically, fuelled by the Internet. Innovators and PhD students are now clumped together in fewer places, often in big cities. And collaborations are more likely to happen between researchers who live, or have lived, close to each other. “Lots of people want to write this story that clusters used to matter in the past, but they don't matter anymore,” says Scott Stern, an economist of innovation and entrepreneurship at the Massachusetts Institute of Technology (MIT) in Cambridge. “For advanced economies and advanced science, location still seems to play a tremendously important role.”
The classic example is Silicon Valley, “the mother of all clusters”, says Martin Kenney, a geographer at the University of California, Davis. The region continues to attract technology workers and entrepreneurs at higher rates than other cities. But as data accumulate that show the importance of geography for spurring economic growth and signs of innovation (such as patent filings and scientific publications), economists have also uncovered examples of how distance, both physical and cultural, inspires discovery too.
Questions about how, when and why location matters are fuelling an active area of inquiry, with plenty at stake. And policymakers who want to stimulate economic development by attracting talent, boosting innovation and encouraging discovery are watching closely.
The first economic treatise on the cluster phenomenon emerged in the late nineteenth century, when British economist Alfred Marshall described how concentrations of related businesses could be beneficial for the regions that host them. More recently, specialists such as economic geographers have taken a data-driven approach to try to understand the value of clusters, which form around all sorts of industries, from medical technology in Minnesota to mechanical engineering in Germany. Overall, Stern says, dozens of studies show that clusters are good for both economies and for the generation of new ideas.
On the economic side, regions that host clusters have more jobs with salaries that grow more quickly compared with regions that don't host clusters — and not just within the speciality at the heart of the region. An analysis of data from 9-million workers across the United States found that every new high-tech and innovation job leads to the creation of 5 other jobs in the region, including lawyers, baristas, teachers and therapists. In places with lots of high-tech jobs, the result is many other jobs1.
Regions with strong clusters are also more resilient in tough times. Stern and his colleagues found that US industries grew more during the 2007–09 recession if they belonged to an established cluster2. Within a strong cluster of medical-device manufacturers around Salt Lake City, Utah, for example, recession years saw annual employment grow by 5%, compared with an average decline of 14% across the United States and a drop of 31% around Madison, Wisconsin, which had few complementary industries. “During downturns,” Stern says, “the financial guillotines hit hardest in regions where clusters are weakest.”
New companies are more likely to form and start-ups are more likely to survive within clusters. When it comes to research advancements in science and technology, firms are more likely to file patents compared with more isolated companies. Writing in Science, Stern and Jorge Guzman, also at MIT, visualized this in a new way3. By focusing not on how many entrepreneurs there are in California, but on how likely those entrepreneurs are to be successful, they mapped where economic growth is likely to be the greatest — with a hot spot over Silicon Valley.
An analysis of the scientific literature also shows how success occurs in clusters. Patent filings and academic papers are more likely to cite other patents or publications that were produced nearby. When researchers looked at the citations of about 9,500 elite (frequently cited or highly funded, for example) life scientists, they found that when scholars moved to a new location, their work was cited less in patents by researchers in the place of departure, but cited more in articles by researchers living near the new destination, and face time seemed more important in industry than in academia4. “All indications are that proximity matters,” says economist Paul Romer at New York University. “And it's possible it matters more now than it did in the past.”
Serendipitous interactions with other researchers who might influence their work may be one reason why proximity remains so important, despite the ubiquity of the Internet, says Ajay Agrawal, a visiting economist at Stanford University in California. Agrawal studies collaborations. These have become increasingly important over the past few decades, as shown by the steady increase in the number of authors on papers — a trend that spans disciplines, from engineering to the life sciences.
When Agrawal analysed the geography of these ballooning federations of researchers, he found surprising clues about the foundations needed to encourage new discoveries, particularly in the context of the Internet age. Along with Avi Goldfarb at the University of Toronto, Canada, Agrawal found that between 1981 and 1991 (when many researchers began to use an early form of the Internet called Bitnet) the rate of collaborations between all institutions whose researchers published in top electrical engineering journals increased — but that collaborations were several times more likely to happen if researchers from each institution lived in the same US city5. That pattern persisted, he says, even as Internet use became universal.
And when scientists from different institutions do collaborate, there is a high chance that they once shared a physical space. For example, Agrawal and his colleagues found that typical collaborations among authors of articles in evolutionary-biology journals included teams of professors with former graduate students and postdocs who have since moved away6 (see 'Research at a distance'). “The lesson we learned from that is that in scientific discovery, a lot depends on relationships,” Agrawal says. “Science is a social process.”
That is not to say that online networks have no role in building relationships. Agrawal also found that the Internet has accentuated and accelerated the productivity that comes from face-to-face interactions. Researchers might meet for lunch or bump into each other in a hallway and end up discussing ideas, for example. But once they're back in their offices, they can easily follow up with e-mails and file-sharing. That kind of interplay between online and in-person communication, “means that it's very critical to people's careers that they spend some time in clusters to connect”, he says. “When those relationships are established, they can go anywhere in the world.” (see 'The power of face time').
And yet, the ease by which in-person meetings can be arranged still matters for long-distance collaborations. Christian Catalini, who studies the economics of innovation, entrepreneurship, and scientific productivity at MIT, and his colleagues found that the introduction of flights by low-cost US airline Southwest Airlines led to a 50% increase in paper collaborations by scientists at universities linked by those flights (see go.nature.com/fivxsr).
In a related analysis, MIT financial economist Xavier Giroud found that when airlines introduced new routes that decreased travel time from a company's headquarters to its remote plants, companies invested 8% more in those plants, and productivity increased7. And last year, Giroud and his colleagues found that direct flights make venture capitalists more likely to interact with their portfolio companies. The introduction of faster airline routes led to an increase of around 3% in the number of patents produced by the portfolio company and an increase of nearly 6% in the number of citations each patent received8.
In one of the most illuminating and creative studies to illustrate the value of proximity, Catalini took advantage of a long-term clean-up project to remove asbestos from the largest medical and scientific complex in France — Paris's Pierre-and-Marie-Curie University. During the 15-year clean-up, which started in 1997, the university staged 5 major waves of relocations that involved moving lab groups around, essentially at random. Researchers had no control over where they ended up, and they were given little notice of the moves.
Catalini found that these relocations had a major effect on both collaborations and publications. After getting shuffled around, lab groups became 3.5 times more likely to collaborate with their new neighbours and 5 times more likely to publish in a journal that was new for at least one of the collaborators9. The collaborations were also more likely to publish in higher quality journals. And although the study couldn't explain exactly why this happens, Catalini suspects that proximity simply reduces the “opportunity cost” of meeting up, in turn increasing the potential for more interactions and conversations that might lead to new ideas for research. “You could imagine that once people get co-located, grabbing coffee or having a conversation is less costly,” he says. “They're more likely to engage in this exploratory behaviour.”
And even for papers that did not involve neighbouring groups, the influence of these new relationships is evident. Catalini analysed author keywords of around 39,000 papers published by the complex's labs over 30 years. Compared with papers published before relocations, he found a 44% increase in keyword overlap among papers published 5 years or more after labs were placed near each other.
These collaborations can have big impacts. For instance, articles with four or fewer authors that were published by Harvard researchers in the same building were cited 45% more than were papers by authors working in different buildings10. Stories abound of research collaborations that formed because two people happened to connect and hit it off. Molecular biologist Herbert Boyer and geneticist Stanley Cohen teamed up to create the first recombinant organism after they ended up talking over a late-night snack at a deli in Hawaii, where they first met at a conference. Robert Solow, an economics Nobel laureate who was one of the researchers shuffled around in Catalini's Paris study, said at the time, “The truth is, it may have changed my whole life.” His office relocation led to a friendship with fellow laureate Paul Samuelson, a relationship that steered him away from statistics and towards straight economics. “The location of that office and the fact that we liked each other so much had a major influence on the direction my career took.”
Although clusters persist as important drivers of economic growth and innovation, the Internet means that distance has an important role in scientific discovery, too. After researchers move, many collaborations between separated colleagues drop off. But the most worthwhile relationships continue, Catalini says. And web-based programs, including e-mail, Slack and Twitter, are essential to making those relationships work.
Clusters with relatively few companies may benefit in particular from connections beyond their borders, says Rune Dahl Fitjar, an economic geographer at the University of Stavanger in Norway. To test the idea that proximity accelerates innovation, Fitjar and Andrés Rodríguez-Pose of the London School of Economics surveyed chief executives of more than 500 Norwegian companies in 2013 (ref. 11). The executives answered questions about their firms' levels of innovation, including the kinds of collaborations that they engaged in and the numbers of new products that they had introduced.
Fitjar and Rodríguez-Pose's findings were unexpected. For this group of Norwegian companies, which included hotels and manufacturing, construction and communication firms, regionally clustered collaborations failed to spark innovation. Instead, innovation was much more likely when firms collaborated with companies in other countries. The chief executives also said that meetings with colleagues are usually purposeful and planned, not random and accidental, suggesting that innovation is often as deliberate as it is serendipitous.
Physical geography isn't the only type of distance worth considering, Fitjar adds. He and colleagues asked Norwegian firms about their main partners in innovation, and uncovered what he calls a Goldilocks principle12. The most successful collaborations occurred between partners that were neither too alike nor too different in their values, attitudes, social structures and ways of thinking.
“What we see is that firms and innovators depend on these long-distance connections to innovate,” Fitjar says “It's kind of a new story that hasn't been told in the literature before.”