Tariff Pass-Through Decomposition

Problem solved

When a tariff lands, three accounting identities have to balance: the importer's landed-cost margin, the exporter's FOB price, and the consumer's shelf price. Operating committees and Section 232 or 301 advocacy teams need the share each channel actually absorbs, by HS6 line, with confidence intervals. The framework uses Cavallo, Gopinath, and Itskhoki style microdata identification.

Inputs

• HS6 import flows by partner, monthly, with both unit value and quantity (Census, BACI, USITC DataWeb)
• Retail price scanner data or BLS CPI item-level series with matching to HS6 (Nielsen, IRI, BLS Research Microdata where licensed; BLS CPI public series otherwise)
• Tariff schedule with effective dates and HTS-to-HS6 crosswalk (USITC HTS, USTR proclamations)
• Bilateral exchange rate panel (Fed H.10) and producer price indices for substitute origin
• Importer-of-record concentration at HS10 to control for buyer power
• Section 301 List 1 to 4A and Section 232 designations as treatment indicators

Outputs

• Three-channel decomposition: percent absorbed by importer, exporter, consumer, with 95 percent CIs
• Heterogeneity by importer concentration quintile and by HS chapter
• Time path: contemporaneous, three month, six month, twelve month pass-through
• Counterfactual: pass-through under a plus 10 percent or minus 10 percent shift in tariff level
• Replication package with code and data sufficient for Section 232 or USTR comment filings

Method

1. Step 1. Build the panel. Stack HS6 by month by partner, 2017 January through current period. Merge tariff effective-date variables, exchange rates, and a quantity-weighted unit value. Drop HS6 with fewer than 24 monthly observations or fewer than 1 million dollars in baseline annual trade.
2. Step 2. Run the first-stage exporter-side regression. Regress log FOB unit value on the tariff dummy, partner fixed effects, HS6 fixed effects, year-month fixed effects, and the bilateral exchange rate. The tariff coefficient is the exporter share of pass-through.
3. Step 3. Run the importer-stage regression. Regress log landed cost net of tariff on the tariff dummy with the same fixed-effects structure plus importer concentration. Compare to the exporter coefficient to back out importer margin compression.
4. Step 4. Run the consumer-stage regression. Match HS6 to BLS CPI item indices via the published BLS-Census crosswalk. Regress log retail price on the tariff dummy, controlling for input substitution and store fixed effects where scanner data is available.
5. Step 5. Reconcile. The three channel shares must sum to one within sampling error. If they do not, the residual is reported as the gap and attributed to inventory drawdown, mode shift, or rerouting. Bootstrap confidence intervals with HS6-cluster resampling.

Assumptions

• Tariffs are exogenous to HS6-month variation conditional on fixed effects. This is defensible for List 3 and List 4A, less defensible for List 1 where lobbying selection is a known concern.
• Quantity-weighted unit values are an acceptable proxy for the FOB price in HS6 lines that pool heterogeneous goods. The assumption fails worst in HS chapters 84, 85, and 90, where the framework reports a hedonic adjustment as a robustness case.
• The BLS CPI item to HS6 crosswalk is one-to-many. The framework uses the BLS-Census published mapping and weights items within the cell by base-period expenditure share.
• Exchange-rate pass-through is allowed to differ across partners but is assumed stable within the estimation window. A subsample test on the post-2022 dollar episode confirms or rejects this.

Limitations

• Sourcing reroute through third countries is observed in the data but cannot be fully separated from genuine substitution without detailed rules-of-origin documentation.
• Scanner data licensing makes the consumer-stage regression expensive to run at scale. The BLS CPI public series version is robust but coarser.
• The framework does not estimate general-equilibrium reallocation across HS chapters. A separate gravity model is required for that.
• Confidence intervals widen rapidly for HS6 lines with thin trade. Cells under 5 million dollars in annual flow should be aggregated to HS4 before reporting.

Example application

Applied to the 2026 US tariff regime: List 1 to 4A baselines, the 2024 Section 301 add-ons, the Section 232 metals overlay, and the IEEPA reciprocal regime. The framework runs the five-step decomposition across roughly 1,800 HS6 lines and reports channel shares by industry. Apparel pass-through to consumer averages 65 percent within twelve months; industrial inputs average closer to 40 percent at the same horizon. See The 2026 tariff playbook: layered overlays, real exposures.