The incident that occurred in the autumn three years ago on a corner of Wall Street shook the whole world. Besides the magnitude of the shock itself delivered by the collapse of Lehman Brothers, what also surprised people were its repercussions. Its damage spread beyond entities that had direct business with Lehman Brothers to those that seemed to have no relationship with the collapsed financial institution, causing a storm that affected financial institutions all over the world.
It is not only the financial institutions that are closely connected. When production by the companies affected by the earthquake and tsunami this past March halted, it impacted their partner companies located away from the affected areas as well as those overseas. It should be noted that this incident also impacted companies that had no direct business with the affected companies and seemingly no relationships.
Financial institutions and companies are closely connected through business transactions, leading to the formation of networks. When a part of the network is damaged, the effect spreads to the other constituent members, indicating the risk involved in being connected.
Being connected, however, is not always negative. The bonds between the players determine the result of a soccer match. In the economic world as well, being connected is a source of productivity. In fact, during the Great Depression in the U.S. as companies lost both mutual trust and normal, stable business transactions and relations, they experienced significant decline in productivity.
How can we minimize the risk of "being connected" while also taking advantage of it? This was the question that motivated me to found the Research Center for Interfirm Network at Hitotsubashi University in the summer of 2008 while employed by the university. There are few accounts in economics textbooks or other material that address the inter-firm network.
♦ ♦ ♦
Even in the world of business, the significance of being connected is not fully understood. Companies naturally know the vendors from whom they purchase goods and may even know the source from whom those suppliers purchase materials. However, few companies understand the supply chain beyond these suppliers.
This fact was validated when management of companies distant from the areas affected by the Great East Japan Earthquake recognized for the first time that they had actually been connected with affected companies.
I conducted joint research with Drs. Takayuki Mizuno (Tsukuba University) and Wataru Souma (Nihon University) to measure how closely companies are connected, using a database jointly built by Hitotsubashi University and Teikoku Data Bank which lists 500,000 Japanese companies and the names and the number of partner companies for each of them. Whereas Wassily Leonief's Input Output Table is an economic research work analyzing the connections between industries, our data constitute an input output table that analyzes companies (rather than for industries).
The figure shows the number of partner companies of a company (Company A) chosen randomly from the 500,000 companies in the database. Link 1 on the horizontal axis refers to the number of companies, less than 200, to which Company A directly sells its products. While Company A is a medium-sized company, the number of its partner companies increases drastically to over 10,000 when companies conducting business with their direct clients or those in Link 2 are included. Furthermore, the number of indirect clients in Link 3 reaches 230,000, comprising almost half of all of the Japanese companies in the database. It is therefore understandable that the Company A's management does not know the members of the Link 3 group. In Link 4, the company is connected with most of the companies.
A study of two companies randomly selected and the number of links in which they become connected indicates that they do so in approximately four links, confirming that Company A is not an exceptional case.
It is beyond the imagination of the management of most companies to find out that any pair of randomly selected companies are connected in only four links. In the area of network science, this kind of situation in which constituent members are closely connected is referred to as a "small world." A well-known example of this phenomenon is a human network, in which complete strangers are surprisingly close to each other. Our analysis indicates that corporate society, similar to human society, is also a small world.
Companies try to save on costs by decreasing the number of suppliers. However, in our data, each company has an average of only 50 suppliers. In other words, the links between each of the 500,000 companies are not very tight but instead are rather loose. Why, then, are they connected in only four links? The answer is that there are companies out there that have a huge number of clients to whom they sell.
♦ ♦ ♦
Our research demonstrates that the number of clients for each company matches the regularity principle known as Zipf's Law. While approximately 30% of the companies have 100 or more clients, 3% have 1,000 or more clients and 0.3% have 10,000 or more. Thus, even small- and medium-sized companies with few clients can suddenly have indirect relations with many clients once they are connected with these "hub companies."
Why do hub companies exist? Each company competes to gain as many clients as possible in order to increase sales. Companies could, of course, do so by increasing sales volume to existing clients, but our research indicates that doing this through existing links is limited and that development of new clients would be pivotal for sales increase. As a result of the competition for clients, companies that win clients grow into hub companies. Unsuccessful companies, on the other hand, lose clients. The significant gap in the number of clients between the winners and losers causes the development of a small world.
The fact that companies are closely connected is an important implication on this system that creates changes in the whole economy, including economic conditions. The traditional argument identifies a shock that commonly affects many companies as the reason for economic fluctuations. It argues, on the other hand, that an uncommon shock or a shock to only a specific company affects that company alone and its surrounding entities and does not spill over to the whole economy. In the background of this argument was the underestimated gap in the number of clients between companies.
Daron Acemoglu and his research team at the Massachusetts Institute of Technology (MIT) discussed the possibility of specific shocks spilling over to the whole economy when inter-firm business networks fulfill certain conditions. These conditions are companies with a large number of clients and obvious gaps existing in the number of clients between companies. The result of our research certainly supports the argument concerning these conditions.
If economic changes were caused by shocks common to a multitude of companies, policies that equally reach all of the companies would be effective. However, since company-specific shocks spilling over to the entirety cause economic changes, such policies would not be effective. Rather, policies focused on the company at the origin of the shock and its surrounding entities would be effective.
♦ ♦ ♦
During the credit easing period immediately following the financial crisis caused by Lehman Brothers' collapse, the Federal Reserve Board (FRB) implemented a policy to focus on and supply liquidity to specific industries, markets, and financial institutions. There was substantial criticism against this policy, claiming that it was too selective and industry-oriented. While the traditional financial policy argument believes that it is preferable for the effects of policies to reach all of the companies equally, the FRB's policy went against this belief. Nevertheless, if the source of economic changes is a specific shock to a specific company, a selective approach in policy-making and management is reasonable and ideal.
Also, if specific shocks cause economic changes, the hub companies with many clients are the key entities in the process of spilling over of the shock. Thus, it is important to break the cycle of shocks by supporting such companies.
Policies supporting hub companies have already been implemented for financial institutions. Following the expression "too big to fail," the idea is referred to as "too interconnected to fail"—saving financial institutions with many clients before the damage of the cycle becomes extensive.
Policies with strong selectivity are accompanied with moral hazard. Therefore, it is important that companies and financial institutions continue to make effort such as prudently selecting sound clients in order to prepare properly for shocks that may reach them through their business networks. Yet, given that the management of companies cannot see their entire networks, such response is limited. Macro policy management should assimilate the idea of breaking the cycle of shocks transmitted through business networks. As part of the infrastructure to deal with the issue, information and statistics for monitoring the situations of business networks of financial institutions and companies should be developed urgently.
* Translated by RIETI.
October 7, 2011 Nihon Keizai Shimbun
