Yoshiyoshi Fujiwara, Hiroyasu Inoue, Takayuki Yamaguchi, Hideaki Aoyama, Takuma Tanaka, Kentaro Kikuchi August 9, 2021
From the flow of funds between companies, you can learn about the economic activities of companies and how to respond to economic shocks such as Covid-19. In this column, using data on remittances from local banks in Japan, how three parts of the network structure of the flow of funds (upstream, downstream, and flow circulation) reflect the characteristics of the supplier-customer relationship. Indicates whether you are doing it. In addition to helping to predict what will happen after an economic shock, the findings will also impact the management of banks’ credit risk.
Real-time monitoring of the real economy provides a useful tool for understanding the present and predicting the future. Monitoring tools that utilize large amounts of data, especially daily or even shorter timescales, have recently become widely available (eg Diebold2021). This setup leverages the big data of a company’s bank account to understand the flow of funds in the local economy, which consists of the relationship between the company and its suppliers and customers. This opens up possibilities for a variety of applications, including capturing regional economic booms and dynamics under the bust, detecting sudden changes due to natural disasters and pandemics, and estimating the impact of corporate defaults (some). Take an example). As far as we know, no such research has been done so far. This is because such data is not available, even for academic purposes, due to privacy concerns. Here, we report the first steps to realize such research in the entire region in cooperation with local banks in Japan.
We show that a bank’s corporate account can provide an ideal tool for understanding the structure of the flow of funds behind economic activity (Fujiwara et al. 2021). Our approach is based on a complex network of enterprises as nodes and funding flows as links (see Aoyama et al. 2017 for a broader perspective).
Data: A unique opportunity to study the flow of money
Our dataset consists of all bank transfers sent to and from the bank accounts of local bank companies. The regional bank is “Shiga Bank”, which is the largest bank in the target prefecture, and is medium-sized in terms of population (more than one million people) and economic activity. All account data is anonymously encrypted, but some attributes, such as geographic location, are anonymously given to corporate-owned accounts. The target period of this survey is from March 2017 to July 2019 (29 months). The number of accounts is 30,000, the number of remittances is 2.4 million, and the total amount is 2.1 billion yen.
From the data, we mathematically constructed a network that is a graph consisting of the nodes and links of a company’s account and the flow of funds from one account to another. Obviously, the flow data includes the direction and the amount. Such a graph is called a directed weighted network. Processing the dataset produces 280,000 directed links, which greatly distorts the distribution of weights (amount of flow). In other words, there were “a few giants and many dwarves.” (Note that money flows from the customer to the supplier in the opposite direction of the goods / services).
Identify upstream, downstream, and core flows of funds
One of the key features of networks is the way money flows from upstream to downstream and circulates in the local economy. Here we define the direction of customer-to-supplier flow, or payment, as opposed to the flow of goods and services. Social networks are often known to have a so-called “bow tie” structure. In the center is a huge core called the “Giant Strongly Connected Component” (GSCC), where any pair of nodes is reachable to each other by at least one path of connected links and is essentially circular. Means the overall pattern of. There are nodes that are unreachable, but the core called the “IN component” is reachable. The opposite is true for “OUT components”. The IN and OUT components are upstream and downstream of the cycle, serving customers and suppliers, respectively. What remains on the remaining nodes occupies only a small portion called the “Tendrill” (or TE).
Figure 1 The flow of funds between corporate accounts forms the structure of walnuts
The overall shape looks like a “walnut” rather than a bow tie. In other words, IN and OUT are two thin skins that are separated from each other, rather than two wings extending, wrapping around a huge connecting component core. Center. This feature is very similar to what you see on a national production network in Japan, but unlike many social networks such as the World Wide Web.
Figure 1 A schematic diagram is shown. The results show that the OUT component is relatively large. This means that suppliers dominate the region.
Figure 2 Walnut structure and upstream, downstream, core location
You can use commonly used mathematical tools in physics to quantify the location of each account to identify the upstream, downstream, and core flows of individual corporate accounts. The position of each node can be measured by a kind of height in the stream (“hodge potential”). Figure 2 shows a histogram of the heights of all the nodes. The OUT component (supplier) is clearly separated from the other components, but there is a great deal of overlap between the huge strongly connected component and the IN component (customer). This result means that the companies that produce intermediate production goods (OUT) are relatively independent of the companies involved in the core and final consumer goods (IN). This is very different from previous surveys of national production networks. The results of this survey suggest that the surveyed areas have a different industrial structure from the urban areas of Tokyo and Osaka, which dominate the national production network. A systematic analysis of different regions of Japan can provide valuable information about the heterogeneous characteristics of the region.
Clarify the “main components” of flow and community activities
The flow of funds occurs in localized regions and interconnected geographic areas within the region. A company sends and receives money to and from other companies using only a small number of specific destinations and sources. For example, businesses in suburban towns often do business with suppliers and customers in the same area or in neighboring big cities, but rarely with other areas farther away. Because bank accounts are associated with geographic areas by address, you can build a remittance matrix by summing them in geographically distant areas, depending on the area of the source and destination locations.
Figure 3 Publish Components: Source (Left) and Destination (Right)
Such a matrix is large, but you can expect to be able to break it down into a relatively small number of components. Each component represents a major frequent flow from one area to another. We used a mathematical method (called non-negative matrix factorization) exactly to clarify these principal components. We found that there are about 12 components that can accurately explain the matrix. Figure 3 shows two such selected components depicted on the geographic map. Each component provides a source-to-destination pair as an important sender-receiver pair. It is clear that both the left side (source) and the right side (destination) are concentrated in a particular city, indicating that this component is primarily a flow within the city. It was found that other factors corresponded to the flow of funds within and between cities (Fujiwara et al. 2021).
The network structure revealed in our research serves several purposes. Since financial information of small and medium-sized enterprises is often difficult to obtain, we will improve the credit risk management of banks by utilizing the information obtained from the network. Information about the network structure can also help promote the local economy, monitor the overall flow of the network, and assess changes over time in the network. Finally, by studying the network of money flows, we can predict what will happen after a pandemic or other disaster after an economic shock. Studies in the case of Covid-19 prove that this direction is very promising for real-time monitoring of the real economy (Yamaguchi et al. 2020).
Editor’s Note: The main research on which this column is based is first Discussion paper Research Institute of Economy, Trade and Industry (RIETI)
Hideki Aoyama, Yuichi Fujiwara, Yuichi Ikeda, Hideki Ietomi, Nishiichi Soma, Hideki Yoshikawa (2017), Macroeconomics: New research on economic networks and synchronization, Cambridge, United Kingdom: Cambridge University Press & Assessment.
Diebold, F (2021), “Measure real-time activity: get out of the Great Recession and enter a pandemic recession, VoxEU.org, January 22nd.
Kyoko Fujiwara, Hideki Inoue, Toru Yamaguchi, Hideki Aoyama, Toru Tanaka, Kenichi Kikuchi (2021), “Money Flow Network Between Corporate Accounts in Japanese Regional Banks”, EPJ Data Science 10 (19).
Toru Yamaguchi, Kenichi Tsuji, Yuichi Nakagawa, Toru Tanaka, Kenichi Kikuchi (2020), “COVID-19 Pandemic Sectoral Impact on Business Transactions: Real-Time Analysis of Financial Big Data,” Shiga University, Economic and Business Research Institute Discussion Paper Series J-1.
Money Flow Network: Evidence from Japan
Source link Money Flow Network: Evidence from Japan