Amid escalating US-China trade conflicts, American firms are compelled to reconfigure their global sourcing strategies. However, the classification and heterogeneity of these reconfiguration choices among U.S. firms still require further exploration. In this study, we propose three modes of China-sourcing strategy: Stay in China, Increase of China, and Withdraw from China. Among those withdrawing from China, we further distinguish substantive reconfiguration (i.e., Friend-shoring) from symbolic reconfiguration (i.e., Springboard). Using both classical regression and machine learning methods, along with a unique U.S. Customs bill of lading dataset from 2014 to 2023, we find that the trade conflict has significantly reduced the U.S. firms’ overall proportion of China import, and increased that from friend-shoring and springboard countries. Utilizing Random Forest algorithms, we further uncover the non-linear effects of firm’ size, profitability, and the extent of tariff shock. Specifically, we found (a) an inverted U-shaped relationship between firm size and the probability of withdraw from China; (b) as firms’ profitability increases, they shift from withdraw from China to stay in China, and ultimately increase of China; and (c) a U-shaped relationship between the extent of tariff shock and the probability of withdraw from China. A closer examination of firms withdrawing from China shows (a) an inverted U-shaped relationship between firm size and the probability of friend-shoring or springboard, (b) an S-shaped relationship between firm profitability and the probability of friend-shoring or springboard; and (c) firms prefer springboard over friend-shoring when the extent of tariff shock they suffered is small.
