The New York Times


June 23, 2002

How One Decision Affects Many Players

By WILLIAM J. HOLSTEIN

The observations of Albert-Laszlo Barabasi about networks have broad applications in business. In an interview, he explained a few of the implications.  

Q. You say that the Internet has Achilles' heels and that we have to protect it better. How do we do that?

A. The Internet is fundamentally decentralized. Each company decides on its own which routers to connect to. By placing the decisions of where to link in the hands of local users, you create a self-organized network. Lately, we've learned that such decentralized networks inevitably develop hubs, a few highly connected nodes. If you take a few of these hubs out, you can easily handicap the whole Internet.

Q. You write that Cisco Systems and other technology companies didn't truly understand their network of suppliers. Did they not understand their financial responsibility for goods in these suppliers' hands?

A. That's one theory. The other is Cisco chose to cut deals that were too restrictive with its suppliers, forcing them to run on very little profit. When one of those supplier nodes failed, that paralyzed the whole supply chain. Cisco wanted to cut the best deal for itself, and it handicapped the whole supplier network it was depending on.

Q. Do you really think we could map out the global economy as a network and prevent financial panics and recessions?

A. Certainly, if we had a map, we could see where the vulnerabilities are. In economics, network thinking is the least developed of any profession. It's partly the absence of data, which most companies regard as proprietary. In the next 5 to 10 years, the biggest impact of networks will be on biology. But in the longer run, economics and business will benefit most from network thinking.

Q. Could we have prevented, say, the Asian economic crisis of 1997?

A. I wouldn't go that far. But if the data were available, we could have identified the weakest links and nodes of the economy. At that time, economists said there was no fundamental underlying economic problem. Yet, at the urging of the International Monetary Fund, the central banks squeezed the money supply, forcing local banks to tighten and call back their loans. A set of cascading failures led to a worldwide economic crisis. A good map of the economic interdependencies among the banks and companies would give us a much better picture and help us make predictions.

Q. Why can't the I.M.F. predict economic events?

A. The I.M.F. is dealing with aggregate data, not network data. They can't predict what will happen if I change a particular node. What will be the cascading effect of that? They don't know.

Q. You say the market is a "directed" network and that we can understand it better than we currently do. What is a directed network?

A. In the economy, companies and banks are the nodes, and the links represent the transactions among them. Each link is directed, telling you in which direction the money flows.

Traditional economic thinking assumes that companies interact with something called "the market," which determines the price. In reality, the direct transactions between individual companies are very important. Some companies can buy a product at a lower cost than the "market" value, because the supplier perceives the relationship to be important. That goes against the traditional thinking about market equilibrium ?that companies always try to get the best price. The market concept ignores the bilateral relationships among companies, aggregating them. In reality, there is a network that is defining everything that is happening in the market.

Q. Wouldn't it be an enormous undertaking to map out the entire global economy?

A. If you wanted a complete map, with all the mom-and-pop stores in your neighborhood, that would be very difficult indeed. But the new science of networks tells us that there are big differences in the importance of different nodes. Some nodes play a much bigger role, and you should start a map with them. Including only a small percentage of the nodes ?the hubs ?would be a starting point. If you monitor the hubs, you have a pretty good feel for what the system is doing in general. 


Copyright 2002 The New York Times Company | Permissions | Privacy Policy