RIETI Policy Symposium

Frontier of Inter-firm Network Analysis: Power of network and geographical friction (Summary)



The inter-firm network is becoming increasingly important both politically and academically. The availability of micro level data encourages empirical and quantitative analysis, which forces us to construct a theory to explain the observed facts. Politically, we expect the "power of network" works to improve the productivity of firms, but at the same time there are concerns about its side effects. For the development of the network, what kinds of policies should we perform? What kinds of results can we expect from those policies? In this symposium, we invited the world's leading researchers in the fields of international trade and spatial economics to report on their latest research from theoretical, empirical, and quantitative aspects. We then collected questions from the audience and, based on them, held a panel discussion on the prospect of both research and policy.

Opening Remarks and Introduction

  • FUJITA Masahisa (President and CRO, RIETI /Professor, Konan University / Adjunct Professor, Institute of Economic Research, Kyoto University)

Improvement in transport technology and information and communications technology (ICT) has reduced transportation costs significantly. This has induced globalization of production and localization into attractive regions and also has created an open brain power society, which has resulted in a complex networked world. Not only academically, but also politically, the "power of network" becomes more important. This encouraged us to conduct the Geospatial Networks and Spillover Effects in Inter-organizational Economic Activities project and hold this symposium. In this symposium, we would like to focus on the inter-firm network and discuss its implications for economic societies and policies.

The development of networks may make the world efficient under normal conditions, but also make it vulnerable so that local shocks can affect the aggregate economy. The automobile industry is an example. The production of each key part in this industry became localized to achieve scale economies under low transport costs. This resulted in the propagation of the shock of the Great East Japan Earthquake over the entire Japanese economy, other Asian countries, and the United States. Most firms, including those in other industries, have direct or indirect partners that suffered from the earthquake, which suggests the broad inter-firm network in Japan. It is very surprising that the local region of Tohoku, which accounts for only 4% of the gross domestic product (GDP) in Japan, affected the entire economy of Japan and also of the world. We cannot explain this with the simple framework of perfect competition in textbooks, thus we have become more interested in the inter-firm network both politically and academically.

We attempted and performed policies for the reduction of geographical frictions because we generally can expect improvement in firm productivity or an increase in welfare. But the question is whether the frictionless world is also suitable for knowledge creation in the long run. Innovation occurs most when common and differential knowledge is in balance. We need some common knowledge to share ideas, but the synergy effects should lessen with little differential knowledge. Concentration in urban areas may generate the synergy effect in the short run. In the long run, however, it may induce expansion of common knowledge and prevent innovation. To avoid this problem, we need to enhance the flow of knowledge and people among diverse organizations, cities, regions, and countries.

Presentation 1 "Implications of Inter-firm Networks for Theories of Production and Trade"

  • Samuel KORTUM (James Burrows Moffatt Professor of Economics, Yale University)

Before discussing networks, I would like to consider about international trade, because the theory of networks fits into the theory of international trade. For the last 20 years, new evidence has continually forced us to construct better models for understanding the world. There was back and forth between the observable data and theory we used. At the same period, the volume of trade increased vastly relative to other activities. A study shows that gains from trade are inversely related to the ratio of what is sold domestically to what is purchased by a country, which implies that the world experienced increased gains from trade during this period.

Progress in modeling, however, is not due to a changing world. We just try to model the world more closely to how it actually is, measure the features that we previously ignored, and embed those features into general equilibrium systems so that we can discuss welfare improvement around the world. We still have many issues that we do not understand. For example, we progressed by incorporating more heterogeneity in the supplier side, but not in the demand/buyer side. Similar to this, theories of importing behavior have progressed less than those of exporting behavior, though the measure for the gains is based more on importing rather than exporting.

About buyer-supplier networks, they are also just a feature of the world and are not a trend either. We need to, and are forced to, incorporate a firm-to-firm network into our theory of trade because data and measures of networks at the micro level are now available and in need of explanation. Incorporating such a network will enable us to give more appropriate policy advice. Modeling networks may solve other puzzles or shortcomings in the theory, similar to what incorporating granularity and heterogeneity did, on the demand/importer side. It also forces us to think about the boundaries of a firm, e.g., what tasks get done in-house and which are outsourced to other firms.

There is a nice analogy between international trade and the firm-to-firm network. Countries may stop producing a certain good and instead purchase it from other countries. We would expect the same things for firms. Firms may stop performing a certain task and instead purchase an intermediate from another firm in the network. Then, intermediates may increasingly dominate production costs. We should also focus on the latter, which may give us a more subtle view of the impacts of international trade.

Ideas also could move from one firm to another, and this flow of ideas may be more important than the flow of goods. This is another challenge for economics. It happens through the network and holds the possibility of huge gains as firms share knowledge with each other.

Presentation 2 "Empirical Evidence of Firm-to-firm Network in Trade and its Implications"

  • Andrew BERNARD (Jack Byrne Professor, Tuck School of Business, Dartmouth College)

Up to now, we have worried a lot about the production side and almost not at all about the customers. We need to look for the customers and see how they matter. We also need to think about how firms form production networks and how the pairing of firms affects prices, quantities, and welfare. We need to learn more about them because the trade costs or policy costs that prevent those connections may have big welfare implications.

We tend to think manufacturing exporters make everything they export, but a study shows that they export more products than they actually produce. They source goods from other firms and ship them in their distribution networks. Current models cannot explain this well, but the most likely explanation is that customers value the bundle more than the individual components. This is also important for policy because barriers on foreign made products may limit those goods that could be put into the bundle.

About connections of buyers and sellers, studies on several countries show that most firms have very few partners in a market, but most matches have a well-connected big firm on one side. Geography also matters for connections. For instance, a study on Japanese data shows that the majority of connections is formed in the same area, though big firms can find suppliers at greater distances. Besides, big firms connect with every type of firm, but small firms tend not to match up with other small firms and tend to connect with big firms because there is a cost in making those connections.

Actually, a study shows that exogenous reduction in search costs increases the connections for firms and improves their performance.

There also are many things that we do not know about the connections, e.g., what costs are important for matching and how production networks evolve. We tend to presume that there is knowledge passing through these networks, but we do not have any direct evidence on that.

Contract choices also matter in a model with two-sided heterogeneity in which exporters and importers have market power. A study shows that, as trade costs fall, importers and exporters are able to change their contracts in some products and might benefit at the expense of domestic customers.

Research on the network has just started so we do not have much to say right now, but I think the welfare implications of this type of research are enormous.

Presentation 3 "Trade, Sectoral Linkages, and Labor Market Dynamics: Quantitative implications"

  • Lorenzo CALIENDO (Associate Professor of Economics, Yale University)

Fluctuations in aggregate economic activity are the result of a variety of disaggregated changes which can be classified into three types of shocks: sectoral shocks, regional shocks, and sectoral and regional shocks. These affect the aggregate economy through the four important mechanisms: sectoral linkages, geographic factors, inter-regional trade, and migration.

There is a huge network of connections across sectors, but how much a sector is connected to the others is different among sectors. For geographic factors, some sectors concentrates in a few regions while others are uniformly dispersed across regions. Therefore, depending on a sector, a shock may affect some regions more than others or may affect almost the entire economy. Inter-regional trade is also important; especially for a large country, it can be more important than international trade. Besides, some resources move across a space after a shock, especially laborers.

I try to build a model to quantify different disaggregated shocks, taking into account these four mechanisms. I will present three examples: productivity boom in one region in one sector, reduction in inter-regional shipping costs, and some results about the Japanese economy.

From 2002-2007, California experienced a boom in computers and electronics. California gained the most, but many other regions also gained from the shock because the products are important inputs in the production of other sectors and are shipped across regions. However, a region close to California loses because productive firms and laborers leave the region for California after the shock. We also find this local shock actually generated large aggregated gains in the United States even compared to the North American Free Trade Agreement (NAFTA).

If the inter-regional shipping costs disappeared in the United States, we find that the aggregate productivity increases by 3.62%, aggregate gross domestic product (GDP) increases by 10.54%, and aggregate welfare also increases hugely. Reducing inter-regional trade distortions can potentially have considerable aggregate effects, especially for large countries.

About aggregate effects to the Japanese economy, NAFTA affected the economy negatively because of trade diversion effects, though the level was mild. Preferential trade agreements in the world from 1995-2010 increased its real income by 4%, and the economy also gains from China's productivity boom because of the cheaper intermediate goods from the country.

I think there is much more to be done in this area. So far, our limitation is access to data. The more data we have, the more we can learn with these tools, and we can construct even better tools to understand how disaggregated shocks actually affect the aggregate real economy.

Presentation 4 "The Implications of Agglomeration and Regional Spillover Effects"

  • Robert DEKLE (Professor of Economics, University of Southern California)

Why do economic activities concentrate in some regions? One of the most important reasons is agglomeration effects: more concentration leads to higher productivity. However they do not concentrate in one place, because dispersion forces also exist, e.g., congestions costs and land prices. These effects depend on industries. A study shows that finance is concentrated relative to manufacturing in Japan because agglomeration effects do not spill over very far for financial sectors, whereas they do for manufacturing.

The Krugman model explains other reasons for agglomeration. Firms want to be located where there is strong demand for their products. Then firms can save transportation costs and have more money to pay their workers, which attracts more workers. Large manufacturing supply there lowers prices, which also attracts workers. Some firms however choose to locate in rural areas because immobile agricultural labor is there, so multiple cities exist in equilibrium. This model predicts that a fall in transport costs leads to more agglomeration as firms agglomerate and export goods to the agricultural labor. The Helpman model, in contrast, has an opposite prediction that a fall in transport costs promotes dispersion. It considers fixed land and mobile labor, so reducing costs may spread activities around the regions for low land prices. Thus, the effects of transportation costs on agglomeration really depend on which models we believe in.

An interesting example in practice is the raw silk industry in Japan. After the Meiji restoration, transport costs exogenously declined rapidly, both domestically and internationally. Raw silk became a huge export industry, so the Krugman model suggests that the industry moves to coastal areas that are close to international export markets. However, the raw silk industry stayed in the countryside because of certain important fixed factors: mulberry plants that are necessary to feed the butterflies and to spin cocoons. The government also picked Tomioka because it has the tradition of the raw silk industry in the region. Interestingly, workers were not fixed factors, since the workers for the silk factories came from all over Japan. For the raw silk industry, the Helpman model seems to fit better than the Krugman model.

Another study shows that industry concentration helps Japanese total factor productivity (TFP) growth in non-manufacturing, though it does not help in manufacturing. These studies suggest that manufacturing as a whole may not be characterized by these dynamic externalities as much as non-manufacturing. As another aspect for a city to become a place that people with desirable skills grow and gather, it should have great universities, and also should have a diversity of cultural activities because those people who are in demand elsewhere will leave if the place is not attractive.

Panel Discussion

Introduction to panel: Yukiko Saito

Researchers in economics have become more interested in inter-firm networks, especially regarding how idiosyncratic shocks affect the aggregate macro economy through networks and how we can incorporate the inter-firm network into trade theory as an extension of the new new trade theory.

Geographical frictions are important for the development of networks, and their effect on mobility seems to be different among goods and people as the evidence shows that knowledge-creating activities are localized compared to overall economic activities in Japan. Also, frictions can be classified into financial costs and time costs. Reduction of these costs should have both positive and negative effects and different effects depending on periods, i.e., the long-run effect or the short-run effect. These effects also should be heterogeneous among regions, firms, and people.

In order to reduce geographical frictions, we have attempted and performed policies such as cluster policy, liberalization of trade, and the development of the transportation infrastructure. The maglev train will connect Tokyo and Osaka and create a super mega-region. What effect can we expect as a result of these policies? What policies should we perform additionally to compensate for the negative impacts?

  • HAMAGUCHI Nobuaki (Program Director, RIETI / Professor, Research Institute for Economics and Business Administration (RIEB), Kobe University)

We will have a question and answer session first. I will ask a couple of questions collected from the audience to each panelist.

Q1. Is it okay to work on a specific industry in network analysis?

There is a tradeoff: studying a particular industry enables us to understand its details, whereas studying the network more broadly enables us to make generalization. I think there is room for both types of analysis.

Q2. The most natural way to model the formation of inter-firm transaction networks seems to be based on two-sided matching game. What difficulty do we have if we use this?

We might be able to start working with the job search model in labor economics, which I think is a nice analogy. However, a single supplier can supply many different customers, unlike a worker who can only supply labor to one employer. We need to think about applying the model in the context we are using and need to move the model in the direction that makes it more appropriate for the application.

In the typical two-sided matching game, it is one-to-one matching. The best attributes in each side match, the second ones match, and so on. For firms, it works differently. Most matches and most activities are many to many. We have to make sure that a model is designed correctly, otherwise we may reach incorrect conclusions.

Q3. Is it possible to introduce heterogeneity in consumers into buyer-seller matching?

Modelling if all aspects are heterogeneous is not necessarily useful. Differentiation in consumers might matter for firms and consumers, but it is hard to work even with two-sided heterogeneity. At this point, adding a third heterogeneity is beyond my capacity.

Q4. Can new new trade theory be extended or incorporated into new economic geography?

In my work with Dr. Saito, we argue that the ability to find more and better suppliers lowers firms' costs. When firms agglomerate, they can have a richer range of suppliers and that lowers the marginal costs, which induces a kind of agglomeration effect. The heterogeneity of firms in trade theory is going to naturally get us to understand what distance does, and I think this is the link between geography and trade theory.

Q5. What kind of firm-to-firm trade data are available in the United States?

Not much, but we do not necessarily need the U.S. data to answer questions.

Q6. What are differences and similarities between your analysis and the Acemoglu model?

Our research is complementary to what Acemoglu does. He mostly focuses on the effects of sectoral shocks on the aggregate economy in characterizing the properties of the input-output matrix of the network. He does not have geography and regional shocks. He does not consider selection channels either. The most productive firms are the ones that are able to survive a shock and export in my model. These are important channels which affect an inter-regional network.

Q7. What is a typical example of inter-regional distortions?

We find that 90% of the costs of shipping goods across the United States are due to distance. It attributes to some regulations, but mostly to transportation costs. Improving transportation costs would allow us to get gains and spillover effects.

Q8. Many people with high productivity and who work in Tokyo actually reside in suburban areas. How do you control this fact?

Even with broader commuting regions, we still find much higher productivity in Tokyo. It is not just driven by the idiosyncrasy that commuters live in the outlying regions of Tokyo and that all of the production is in Tokyo.

Q9. Whether a fall in transport costs induces agglomeration or dispersion seems to depend on the industries of firms. Do you have any opinion on that?

There must be industrial heterogeneity in the characteristics of industries in the mixture of industries. Actually, there are many case studies and empirical studies on the characteristics of industries. To digress a bit, I also want to stress an aspect that people do have much mobility both domestically and globally. Compared to the United States, what Japan needs to do is to make areas that the government wants to develop into more attractive places for talented people to live.

I would like to discuss about policies. Many policies are attempted and made to reduce the friction of distance, not only inside a country but also internationally. I would like to ask questions about that.

Q10. If construction of the maglev will be completed between Tokyo and Osaka, this will create a huge mega-region in which we can exchange goods and ideas. What kind of effect will we have from this?

In Professor Dekle and Dr. Saito's presentations, the finance sector seems to crowd out manufacturing from Tokyo, and knowledge production is more concentrated in Japan. Ironically, we think ideas can move any distance with the same costs, yet concentration seems more important for industries that rely on ideas, and they seem to benefit the most from the maglev and agglomeration.

We think ideas can flow easily, but there is something about being present to exchange and understand complicated ideas. The maglev may induce clustering, but it may allow firms to stay where they are. My work with Dr. Saito suggests that firms can find suppliers in greater distances by the maglev, which means they do not have to move closer. I think forces work in different ways in different industries. However, knowledge workers want to live and agglomerate in attractive places, which may induce another political problem of regional inequality.

A city in the United States, which was prominent in the past because of a canal, has lost people these days because we do not need the canal to move goods anymore. Improving technology may generate booms in some places, but it may generate busts in others. Reducing frictions inside a country in general generates welfare gains, but we need to keep in mind that some areas actually lose from this.

Q11. What we can expect as a result of co-location of business and people driven by cluster policy? Should the agglomeration go more intense in a particular location?

I worry that we tend to think we know which industries should cluster and where they should locate. I believe that locating close to suppliers is potentially a big benefit for firms, but I am not sure that the benefits are big enough to outweigh the costs, which are hard to measure.

Q12. What can you say about promoting liberalizing trade policies such as the Trans-Pacific Partnership (TPP)?

That sort of policy, in general, is good for countries. The only worry is that specific businesses can capture the negotiations and get special deals, which may not serve the general public.

In a recent study, we tried to think about the hypothetical scenario of Japan moving to a world with zero tariff. We find little positive aggregate gains, because tariffs are already low in Japan. I think what was discussed in the TPP are other types of trade costs and other types of ways to facilitate trade between countries, which should generate larger gains. For example, in my home country Uruguay, exports are subject to inspections which take from a day to a month. Reducing those types of regulations should benefit exporters.

Q13. Is there any policy measure to handle with inequality?

We can quantify who gains and who loses, and potentially think about a redistribution scheme that actually makes all of us better off. In the United States, one key reason for inequality is huge mobility costs. I do not know what the costs are for now, but reducing these costs potentially can reduce the potential losses and allows the gains to spill over.

In a study on Denmark, we find that displaced workers have higher wages five years later than the workers who maintained their jobs, partly because of the flexibility of the Danish labor market and the support in the short term. It means that it is possible to have shocks that can seem to have big negative consequences on inequality turn out to have almost no negative effects on inequality.

Interestingly, a survey shows that there is no increase in asset income inequality in Japan from 2000-2013, because asset prices have not gone up in Japan. So if there is an inequality problem, it is on the wage side.

We talked about the matching within a country that enhances the performance of firms. It could also enhance the competitiveness of firms in international competition. I would like you to speak about this.

I think a nice way to think about the macroeconomic benefits of improving the network is as an improvement in productivity. We actually think about international trade as if it gave us access to a new technology. Viewing the network as a productivity booster is a good first cut at the problem.

We find that firms with richer and deeper supplier bases seem to succeed in international markets as well, though causality has not been proved yet. My guess is that the domestic success of firms in establishing their connections leads to their international success.

Infrastructure helps to enhance the network and to improve productivity of companies, but this is not necessarily considered when we build infrastructures. We may be able to incorporate such a benefit into the cost-benefit analysis. Could you discuss this?

Our research shows that opening up the shinkansen in Kyushu lowers the costs of finding suppliers or the costs of regularly going to the suppliers to exchange ideas, and this increases firm productivity and sales. We are almost sure that it might improve productivity, but we need to worry about the magnitude. It is also true for international relations. How can we encourage the international supply chains to deepen and become more robust between Japanese firms and foreign suppliers and customers? The shinkansen and maglev should reduce search costs, but they are incredibly expensive to do. There should be other ways to reduce search costs. Thinking about this in terms of supply side cost-benefit is something we have not done, and we should do it for both international and domestic infrastructure.

In a course about old old trade models that I took decades ago, the professor said that infrastructure projects have to be self-sustaining and make at least zero profits. I think infrastructure projects at some point have to meet the profit requirement. I think each project should really be measured on a cost-benefit analysis for the sake of raising Japanese aggregate GDP.