Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

In The review of socionetwork strategies

We quantify the magnitude of shock that propagates individual firms through direct supplier-customer relationships. First, we construct machine learning models that predict a firm's sales growth rate based on corporate attributes and sales information of the firm and its suppliers/customers. The prediction models indicate that not only macroeconomic factors, such as the year and country, but also sales fluctuation of suppliers/customers are important predictors of the firm's sales growth rate. Second, we plot the change in the predicted sales growth rates in accordance with those of suppliers/customers using a partial dependence plot. Thus, we quantify how much a firm's sales growth rate changes in accordance with the changes of its suppliers/customers, namely, the magnitude of shock propagation. Finally, we verify the magnitude of shock propagation by comparing it with the sales growth rate of firms that have suppliers/customers negatively impacted by Hurricane Sandy in the U.S. in 2012. The comparison indicates that there is no significant difference between them and further demonstrates that we can simulate how much the shock that occurred in the disaster-affected firms propagates to their transaction firms.

Sato Ryoji, Mizuno Takayuki


Machine learning, Propagation of shock, Sales growth rate, Supply chain