Abstract: This paper assesses the efficiency and welfare costs of retrenchment from international suppliers under tariff and geopolitical risk. Using a sector-level (Census data) and firm-level (Panjiva firm input-output data) exposure-weighted event-study analysis, I find marked retrenchment and reallocation that went beyond goods subject to a tariff and, through an anticipation design using surprises in trade-policy uncertainty and geopolitical risk, can be attributed to news about trade policy rather than tariff levels. I build an input-output network model in which firms, facing uninsurable productivity and currency risk, optimally choose suppliers (severing very risky ones, even if productive) and embed it in a dynamic macro model, where consumer-workers share the risk of the firms to which they are attached through a limited-commitment wage contract. I derive a risk-misallocation wedge that emerges only under an endogenous network and incomplete markets: it vanishes in the fixed-network benchmark (Baqaee and Farhi) or under complete markets. Calibrated to United States input-output tables and customs records, the model reproduces the gradual retreat from China after a surprise in trade-policy uncertainty and, without being targeted, the long-run retreat measured in the event study. Measured tariffs and tariff risk account for little of the retreat, which is driven by the firm owners' beliefs about trade risk. Workers gain about 0.5 percent of consumption from the tariff rebate and the appreciation of the dollar, but workers of employers that rely on Chinese inputs gain less, because the tariffs shift income risk onto them.

for information patrizia.pellizzari@unibocconi.it