1. MID" the abbreviated foreign supplier name, and "date" the shipment date. In most of our analyses, we restrict the sample to "active" U.S. firms, with at least one imported shipment per quarter in quarters;Notes: Steps in the creation of analysis dataset
2. Step (7) forms our baseline analysis sample. The last two rows show the number of recalls and firms associated with single or multiple shipments. Firm shares in the last two rows do not add up to 100% because a single firm may experience multiple recalls;Source: Authors' calculations from CPSC, BR, and LFTTD data
3. Supplier-Based Controls Notes: Coefficient estimates and 95% confidence intervals showing the probability of trade between the treated trading pairs in the HS6 product category, for 12 quarters before and after a public recall announcement, relative to control trading pairs. The regression includes quarter and buyer-supplier-product fixed effects. Controls are selected based on supplier;Figure B-3. Extensive-Margin Dynamics: Buyer-Supplier-Product
4. Control relationships consist of the same U.S. firms as treated relationships, two years prior to the recall. of shipments Notes: Sample includes LFTTD shipments in 2007 aggregated to the buyersupplier-HS10-date level. The dependent variable is 100,000 times the assigned probability that the shipment was recalled. Standard errors are clustered by HS10. Each regression has 13,410,000 observations. All regressions control for HS10 and country-of-origin fixed effects;Figure B-4. Extensive-Margin Dynamics: Buyer-Supplier-Product
5. Table 1: Significant and non-significant models and their evaluation measures.