Consider an image: hubs and spokes sprawling across a map. At the Bank, we work in many fields that could be portrayed this way – finance, trade, transportation, infrastructure or urban and regional development. The position of a country, a city or a bank in its network is vital to explaining its individual economic performance. The property of the network as a whole is also important to understanding the resilience of the system and its parts to shocks or contagion effects.
In May, the Bank’s Trade Department and the IMF’s research department brought together, for the first time, a group of experts on various types of networks. The objective was to find out what the "science of networks" can tell us about the practice of international and development economics. The group included network planners from the private sectors, regulators, economists and physicists.
A new science for economics
In economics, we often talk about “networks,” “connectivity” or “connectedness” in our work, but rarely are the concepts clearly defined. They are too often intuitive or based on vague assumptions; they refer to different ideas in different sectors. A banking network offers a connectivity that is different from the connectivity of a road network. We use the terms loosely, and in various publications on transportation and trade, I count myself among the worst offenders.
Fortunately, we can look to other disciplines to bring clarity, a common understanding and a unifying language. The science of networks began in condensed matter physics. It branched out to find applications in social science, computer science and biology. Applications to the internet or social networks are well-known and advertised: Google, Facebook and Twitter all use some variation of networks theory.
Yet the applications to international and development economics are just emerging.
At the conference, Professor Alessandro Vespignani, a physicist from Northeastern University in Boston, gave us some insight into the current state of the art and applications of networks theory. He described how networks operate in biology, in communications and in the financial system. The internet, he pointed out, is an example of a typical “complex system:” it has a large number of interacting units and no “mastermind,” or central control; it is self-organized and there is no “map” of the internet.
In essence, a network describes a complex system in a graph form that portrays nodes (hubs) connected by edges or links (spokes). The nodes and links can differ based on the application – for example, countries would be the nodes in international trade and the links would be bilateral trade flows or direct shipping connections. But the concepts and tools are generic. In fact, there are several definitions of connectivity or centrality that describe the different ways to assess the position or importance of a node relative to all other nodes on a network.
This is important, not only to explain the relative position of elements described as nodes, but to depict the global properties (and risks) of networks as a whole: a pandemic spread by air transport; a financial crisis spread by electronic transactions. New tools and simulations of networks are critical to understanding these phenomena. These tools can help measure contagion effects, and identify what type of intervention may contain them before a certain tipping point; they can give enough lead time to react effectively. Understanding these non-linear properties and “small world” effects is essential to stemming a wide range of crises in the modern world.
The language of big data?
In the conference, we saw presentations explaining networks’ applications to trade, finance, air and shipping networks and regional development. One of the most striking multidisciplinary aspects of networks science is that it is a practical way to make sense of large amount of data. This is increasingly important in a wide range of sectors, where automation of activities and processes means large amount of detailed data is available -- including trade, finance, international supply chains, health, and others. In fact, networks science is behing a tool many of us use daily: the eigenvalue centrality algorithm helps the search engine of Google make sense of (rank) the millions of pages that apply to a single search term.
The potential for cross-fertilization is huge. In fact, some of this collaboration happened in real time during the workshop. Our Bank aviation colleagues agreed to develop a partnership provide higher-quality aviation statistics to a research group to improve a model describing the spread of pandemic disease.
However, as noted by the IMF’s Stijn Claessens, not every tool applies to every sector. The fact is, network structure varies from one application to the other. Global transport networks, such as aviation and shipping, have a “spikey ring” topology. Container shipping is like a conveyer belt, according to an industry expert. In contrast, trade and finance have a hierarchical network structure, with few central cores. Indeed, I am confident that we have many new images to look forward to in our future of partnerships and collaboration. Stay tuned…