Predicting the Supply Chain Impact of National Level Conflicts: a Recursive Neural Network Based Approach

• Activity Number: 320 - Statistical Approaches for Modeling Social Unrests
• Type: Invited
• Date/Time: Tuesday, July 30, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Defense and National Security
• Author(s): Ujjal Kumar Mukherjee* and Benjamin E. Bagozzi and Snigdhansu Chatterjee
• Companies: University of Illinois and University of Delaware and University of Minnesota
• Keywords: Recursive Networks; Conflict; Supply Chain; International Trade; Prediction Problem

Abstract

We investigate the impact of conflict on international trade and supply chains. First, we show that different types of conflicts such as material conflict and non-material conflict; and civil-government and civil-civil conflict has differential effects on international trade and cross-border supply chains. Further, we show that several economic, political, and social factors are associated with the different types of conflicts, and therefore in-turn, with international trade and cross-border supply chains. Finally, we develop a two-step prediction framework for first, predicting the different types of conflict, and finally, predicting abnormal changes in international trade. The prediction framework is developed using a Long Short-Term Memory (LSTM) based framework. Due to the spatio-temporal dependence of social, political and economic factors, LSTM based models which take into account the long-term dependence of relevant predictors for predicting conflict, are suitable. We compare the LSTM model with other standard prediction models and show that LSTMs are able to achieve better prediction of conflicts and their impact on international trade and cross-border supply chains.