How demand shock propagates through an input-output network

How demand shocks are transmitted through an input-output network: the global financial crisis and the events of the COVID-19 epidemic

Promoting shocks through input-output connections between companies (such as supply chains) has attracted the attention of economists. The ongoing COVID-19 epidemic and recent geopolitical events require further study on the theme. In this context, recent theoretical analysis (e.g. Acemoglu et al. 2012) shows that the microeconomic push for hub firms in an input-output network has a major impact on the overall output. Furthermore, due to the increasing availability of solid-level data, recent empirical studies have directly analyzed such solid-level input-output networks and detected propagation phenomena. For example, Barrot and Sauvagnat (2016) use natural disasters in the United States as a source of negative shocks for organizations and examine the spread of these shocks through input-output connections. Bohem et al. (2019) and Carvalho et al. (2021) Use the Tohoku earthquake in Japan in 2011 and show that this negative shock spreads to other consumers in the unaffected region.

These previous studies focus on propaganda driven by supply shock; That is, Shock promotes ‘downstream’ from customer supply. In contrast, the empirical literature on ‘upstream’ promotion from demand-driven customers to suppliers is rarely tested. The two exceptions are Acemoglu et al. (2016) and Kisat and Phan (2020), but their analysis relies on a sector-level input-output network. To the best of our knowledge, there is no empirical study that directly monitors demand-shock propagation through a solid-level input-output network.

In a recent study (Arata and Miyakawa 2022), we aim to fill this gap in the literature by using Japanese firm-level input-output data. We focus on demand-shock propaganda arising from the sharp decline in exports during the global crisis and the change in consumer behavior during the COVID-19 epidemic. By looking at these phenomena as external to Japanese companies, we examine how the growth rate of sales is affected by the presence or absence of customers who are transacted by negative shocks. We use heterogeneous treatment effect models developed by Athey et al. (2019) and Wager and Athey (2018), which enable us to analyze how the promotion effect depends on strong features, such as firm size. Focusing on the variety of propagation effects, we examine the route of demand-shock propagation through the input-output network.

Our analysis shows that during the global financial crisis, the impact of publicity is sufficient for large suppliers but not for small suppliers (see Figure 1). That is, the negative pushes of the exporting companies are not transmitted to their small suppliers, especially when the exporting companies are large. We see that while exporters are experiencing declining exports they are reducing inventories, their small supplier sales growth rate does not respond to the negative growth rate of these exporting companies.

Figure 1 In response to the global financial crisis, the response to the growth rate of sales of suppliers to the growth rate of sales of their customers

Comments: The effect of the promotion depends on the log size of the provider and their customers. Customer sizes are divided into four groups: Small (Sales ≤ First Quantile), Middle 1 (First Quantile ≤ Sales ≤ Median), Middle 2 (Medium ≤ Cell ≤ Third Quantile), and Large (Sales ≥ Third Quantile). The left panel and the right panel account for the manufacturing industry and the suppliers in the wholesale and retail industries, respectively.

This is because the effect of the promotion is not homogeneous, and in particular, demand shocks are transmitted from customers to suppliers only when the latter are the main suppliers (see Figure 2). We see that even in cases where consumers are experiencing a major decline in exports they see the supplier as the main customer, if this supplier does not see the customer as the main supplier then negative shocks are not transmitted to the supplier. In addition, we see that large suppliers are more likely to choose as major suppliers, especially large customers. Since most of the exporters and their major suppliers are large corporations, demand-shock promotion occurs mainly among these large corporations during the global financial crisis.

Figure 2 The route of demand-shock propaganda

Comments: The shape of the circle represents the solid shape. Demand shocks are only propagated to suppliers when they are the main supplier for the customer. In particular, there is no demand-driven promotion between large customers and small suppliers.

Finding differences in the effects of reproduction raises another question. If demand shocks hit smaller major suppliers, do negative shocks propagate to smaller suppliers? This is what happened during the COVID-19 epidemic (see Figure 3). Most companies in the covid-affected sector, such as restaurants and hotels, are small and their main suppliers are also small. We see that there is no significant difference in the effect of promotions across supplier sizes. That is, the negative demand push spreads to smaller suppliers as well. This result is consistent with our interpretation that promotion occurs only through relevant links for both supplier and customer.

Figure 3 COVID-19 Responses to Sales Growth Rates from Suppliers during Epidemics

Note:: Customer sizes are classified into three groups: small, medium and large.

The overall fluctuations of our research have a significant impact on the literature of micro-origins. The above studies emphasize the importance of network structure in the context of overall fluctuations driven by microeconomic shock. However, it is indirectly assumed that the pushes are homogeneous throughout the propaganda agencies. Because of this assumption, the effect of the microeconomic push is proportional to the number of links (i.e. transaction relationships) in the firm. Since large corporations deal with many smaller suppliers and customers, or a combination of negative dimensions (e.g. Bernard and Maxness 2018, Bernard et al. 2014, Lim 2018, Bernard et al. 2019), the model predicts that these will push larger corporations. Promoting across an economy. In contrast, our research suggests that the role of large corporations in demand-shock promotion is limited because demand-shock promotion does not occur through these links between large customers and small suppliers. The demand push of large corporations is only transmitted to their large suppliers and does not spread further in an economy.

An additional important message from our analysis is that demand shocks are propagated through links and demand shocks through links stop promoting coexistence. In other words, connecting the former links to an input-output network enables us to identify the route of demand-shock propagation. Identifying the path of publicity is also relevant for policy makers because we can more accurately evaluate the effectiveness of policy measures such as subsidies in targeted organizations. Our search is the first step for this purpose.

Editor’s note: This column is based on a major study (Arata and Miyakawa 2022) that was first published as a dissertation by the Research Institute of Economy, Trade and Industry (RIETI) in Japan.


Acemoglu, D, U Akcigit and W Kerr (2016), “Networks and the Macro Economy: An Experimental Exploration”, NBER Annual of macroeconomics 30 (1): 273–335.

Acemoglu, D, VM Carvalho, AE Ozdaglar and A Tahbaz-Salehi (2012), “The Network Origins of Aggregate Fluctuations”, Econometrics 80 (5): 1977–2016.

Arata, Y and D Miyakawa (2022), “The demand for shock propagation in Japan through an input-output network,” RIETI Discussion Paper Series 22-E-027.

Athey, S, J Tibshirani and S Wager (2019), “Generalized Random Forest”, The Annals of Statistics 47 (2): 1148-1178.

Barrot, J and J Sauvagnat (2016), “Promoting Idiosyncratic Shocks in Input Specification and Production Networks”, Quarterly Journal of Economics 131 (3): 1543–1592.

Bernard, AB, E Dhyne, G Magerman, K Manova and A Moxnes (2019), “Sources of Strong Inequality: A Production Network Method”, Technical Report, National Bureau of Economic Research.

Bernard, AB and A Maxness (2018), “Networks and Commerce”, Annual review of the economy 10: 65-85.

Bernard, AB, A Moxnes and YU Saito (2014), “Geography and Strong Performance in the Japanese Production Network”, Technical Report, RIETI.

Boehm, CE, A Flaaen and N Pandalai-Nyar (2019), “Input Linkage and the Transmission of Shock: Farm-Level Evidence from the 2011 Tohoku Earthquake”, Economics and Statistics Review 101 (1): 60-75.

Carvalho, VM, M Nirei, YU Saito and A Tahbaz-Salehi (2021), “Supply Chain Disruption: Evidence of Great East Japan Earthquake”, Quarterly Journal of Economics 136 (2): 1255–1321.

Kisat, F and M Phan (2020), “Consumer demand shocks and strong connections: evidence from demonetization in India”, available at SSRN 3698258.

Lim, K (2018), “Endogenous Production Networks and Business Cycles”, Working Paper.

Wager, S and S Athey (2018), “Estimation and Estimation of Different Therapeutic Effects Using Random Forests”, Journal of the American Statistical Association 113 (523): 1228-1242.

Leave a Reply

Your email address will not be published.