Political uncertainty, bank loans, and corporate behavior: New investigation with machine learning

Abstract

This paper investigates how uncertain term length, a novel source of political uncertainty, affects the behaviors of banks and firms using a machine-learning approach. China's local authorities do not have a fixed term, creating an ideal environment for studying how economic agents react to their perception of political uncertainty without an actual political turnover. We implement a machine-learning method to predict the term length of city leaders by observing others with similar backgrounds. Combining this new measurement of political uncertainty and bank- and firm-level data, we find an inverted U-shaped relationship between city leaders' predicted remaining term length and bank loans, corporate liabilities, and investment, which matches the change of political uncertainty over the term. We also record the potential adverse consequences of the politically motivated loan and investment expansion, such as a loss of corporate efficiency and a disruption in market order.