## wpiea2019238-print-pdf - 2.3 log points in magnitude and focus only on market transactions.

## Source details

**Canonical URL:** [wpiea2019238-print-pdf - 2.3 log points in magnitude and focus only on market transactions.](https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019238-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2019/wpiea2019238-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2019/wpiea2019238-print-pdf.pdf.json)

---

### Data and scope
- Prices of trades invoiced in US dollars, representing over 94 percent of US trade in the sample.
- Petroleum products excluded.
- Partner countries restricted to those with data on aggregate prices and exchange rates (macro variables cover 182 countries).
- BLS import price data at the individual good level, concorded to HS codes; over one-third of BLS import prices are non-market transactions (intrafirm or related-party shipments) and such non-market transactions are excluded from market-focused analyses.
- Goods matched to 6-digit HS codes; associated tariff taken as the highest value among corresponding 8-digit HS codes (this assumption holds exactly for over 95 percent of 6-digit codes).
- Exclusions: goods impacted both by a China tariff and another product-based tariff (e.g., steel and aluminum products, lumber, washing machines, solar panels) and data on imports from India for a subset of analyses.

### US imports from China — descriptive patterns
- Tariff waves on China in 2018–2019:
  - July 2018: 25 percent ad-valorem tariff on roughly $34 billion of imports.
  - August 2018: 25 percent tariff extended to roughly $16 billion of shipments.
  - September 2018: 10 percent tariff applied to roughly $200 billion in goods.
  - May 2019: tariff on the September wave increased from 10 to 25 percent.
- Price-index evidence (indices normalized to 1 in June 2018):
  - Affected goods imported from China saw an immediate jump in their price inclusive of tariffs in the month the policy was implemented; the scale of the jumps is only slightly below the scale of the tariff rates.
  - Ex-tariff price indices for affected Chinese goods did not exhibit meaningful breaks from prior trends.
- Frequency-of-price-change patterns:
  - No obvious differences in the share of monthly price decreases or increases across groups (affected/unaffected × China/Not-China).
  - Prices of products affected and imported from China appear at least as stable as others; little evidence that tariffs changed price stickiness.

### Regression framework and identification
- Two main specifications:
  1. Monthly-data specification including zero price-change months (Equation (2)):
     - Dependent variable: ∆ ln(P_I_{i,j,k,t}) where P_I is ex-tariff price of item i from country j in sector k at month t.
     - Controls: sector fixed effects δ^I_k, China-sector fixed effects φ^I,Ω_CN and φ^I,−Ω_CN, lags of log gross additional tariff ∆τ_CN,k,t−l (up to 11 lags), lags of ∆ ln(S_{j,t−l}) (exchange rate) and ∆ ln(X_{j,t−l}) (producer price index).
     - Cumulative 12-month tariff passthrough estimated as ∑_{l=0}^{11} γ^I_CN,l.
     - 12-month ERPT estimated as ∑_{l=0}^{11} β^I,S_l.
  2. Conditional-on-price-change (spell) specification (Equation (3)):
     - Uses only non-zero price changes, scales changes to monthly frequency, reports 12×γ^I (annualized tariff-associated change) and β^I,S (ERPT) on the scaled changes.

### Monthly regression estimates (Table 1)
- Sample: monthly data January 2005 to August 2019; Obs. = 820,318.
- Key coefficient estimates (robust standard errors in parentheses; ***, **, * denote 1, 5, 10 percent significance):
  - Tariffs 1 yr. (∑_{l=0}^{11} γ^I_CN,l):
    - Column (1): -0.079*** (0.026)
    - Column (2): -0.076*** (0.028)
    - Column (3): -0.018 (0.030)
  - ERPT 1 yr. (∑_{l=0}^{11} β^I,S_l):
    - Column (3): 0.219*** (0.027)
    - Column (4): 0.221*** (0.027)
  - PPI PT 1 yr. (∑_{l=0}^{11} β^I,X_l):
    - Column (3): 0.019 (0.070)
    - Column (4): 0.012 (0.073)
  - China φ^I,Ω_CN (Affected):
    - Column (2): 0.000 (0.000)
  - China φ^I,−Ω_CN (Not-Affected):
    - Column (2): -0.000 (0.000)
    - Column (3): -0.001 (0.001)
  - Adjusted R^2:
    - Column (1): 0.000
    - Column (2): 0.003
    - Column (3): 0.004
    - Column (4): 0.004
- Interpretation example from Column (1): the estimated coefficient of -0.079 implies that a 10 percent tariff would be associated with a 0.8 percent lower ex-tariff price and a 9.2 percent higher overall price faced by the importer.

### Conditional-on-price-change regression estimates (Table 2)
- Sample: non-zero price-change spells; Obs. = 99,406.
- Key coefficient estimates (robust standard errors in parentheses):
  - Tariffs annualized (12×γ^I):
    - Column (1): -0.228 (0.171)
    - Column (2): -0.109 (0.187)
    - Column (3): 0.006 (0.188)
  - ERPT β^I,S:
    - Column (1): 0.382*** (0.052)
    - Column (2): 0.381*** (0.052)
  - PPI passthrough β^I,X:
    - Column (1): 0.757*** (0.110)
    - Column (2): 0.766*** (0.111)
  - China φ^I,Ω_CN (Affected):
    - Column (2): 0.004*** (0.002)
  - China φ^I,−Ω_CN (Not-Affected):
    - Column (2): 0.002* (0.001)
    - Column (3): 0.003** (0.001)
  - Adjusted R^2:
    - Column (1): 0.000
    - Column (2): 0.006
    - Column (3): 0.017
    - Column (4): 0.018
- Summary: estimates are qualitatively consistent with monthly specifications — tariffs are associated with economically or statistically insignificant changes in ex-tariff import prices in many specifications, while exchange rate passthrough rises to roughly 38 percent in these conditional regressions.

### Tariffs on steel imports (descriptive)
- Background:
  - March 2018: US placed a 25 percent tariff on steel imports from all countries; exemptions initially for Argentina, Australia, Brazil, Canada, Mexico, EU, South Korea.
  - June 2018: exemptions lifted for Canada, the EU, and Mexico (second wave); exemptions for other countries made permanent.
- Price-index evidence for steel:
  - Steel prices were volatile in the preceding four years.
  - After steel tariffs were introduced, prices on imports from affected countries rose; imports from affected countries jumped to roughly 20 percent above those from unaffected countries.
  - Regression analyses for steel give similar qualitative conclusions, but estimates are imprecise due to the small number of imported steel products.

### Summary of results on US import tariffs
- Across analyses (China imports, steel imports, aggregated price indices, and product-level regressions) the 12-month price response to US import tariffs imposed in 2018–2019 shows:
  - Ex-tariff prices do not obviously behave differently for goods affected by trade policy versus those not affected.
  - Tariffs exhibited nearly complete passthrough into the total import cost; incidence of tariffs lies largely with the United States (importers/consumers).
- Exchange rate passthrough:
  - Using the same data/time period, estimated ERPT into import prices is in the range of 25 to 35 percent after one year (conditional regressions indicate up to roughly 38 percent), consistent with existing literature and substantially lower than tariff passthrough into total import prices.
- Implications:
  - Standard trade and international macroeconomic models that assume symmetric responses to tariffs and exchange rate shocks may be misleading in the short run; observed asymmetry suggests models might better apply to longer-run outcomes or be amended to allow for more uncertainty or mean-reversion in shocks.
  - The recent depreciation of the Chinese renminbi did not offset the impact of the tariffs for US importers during the sample period.

### Downstream (retail) pricing — preview of findings
- Preview conclusions:
  - Some evidence that tariffs have passed through into higher retail prices, but effects are much more muted than for total import prices.
  - Retailers appear to have absorbed much of the higher import costs via lower margins.
  - Additional retailer adjustments observed: front-running of tariffs (inventory build-up before higher tariffs took effect) and partial diversion of orders to non-Chinese suppliers — adjustments that could flatten observed retail price responses so far.

### Washing machines and retail-level evidence (2018)
- Washing machine retail prices:
  - Prices for about 700 washing machines from PriceStats and the Billion Prices Project (BPP), scraped at a daily or weekly frequency from 16 large multi-channel US retailers.
  - Adjacent retail price observations that differ by more than 2.3 log points in absolute value are excluded.
- Sample sizes and coverage:
  - 300 handbags from 12 retailers; 400 tires from 7 retailers; 5,000 refrigerators from 18 retailers; 200 bicycles from 11 retailers.
  - Two large US retailers datasets: Retailer 1 about 38,000 products; Retailer 2 about 55,000 products; combined more than 90,000 products covering nearly 1,991 different 6-digit HS categories.
  - Imported products: about 60,000; products imported from China: about 43,000.
  - Products in categories affected by tariffs: about 59,460; products from China & in affected categories: about 30,101.
- HS6 automated matching: roughly three-quarters classified automatically; remainder manually.

### Washing machines: high and rapid passthrough
- BPP and CPI washing machine price indices normalized to equal 1 in February 2018.
- Pre-tariff trend: BPP and CPI indices declined by about 5 percent per year prior to tariffs.
- Post-tariff dynamics:
  - Within a few months of the import tariffs, both BPP and CPI series exhibit a break, with inflation rates switching from negative to positive.
  - In the second half of 2018:
    - Washing machine inflation in BPP data typically between 5 and 10 percent.
    - Washing machine inflation in CPI data typically between 10 and 15 percent.
- Interpretation: evidence strongly suggests moderate to high passthrough of washing machine tariffs to retail prices; heterogeneity across brands observed.

### Other affected goods: heterogeneous responses
- Handbags, tires, refrigerators, bicycles (indices normalized to 1 in September 2018):
  - Three months after the first tariffs: none exhibited sharp price increases relative to trend.
  - By the time tariffs were increased to 25 percent:
    - Handbags and tires experienced unusually rapid price increases.
    - Refrigerators exhibited a mild increase in inflation relative to pre-tariff trend, but the increase appears to have started before tariffs.
    - Bicycles showed decreased price inflation in the post-tariff regime.
- Summary: tariff passthrough varies by sector — rapid/high for washing machines; slower/ultimately high for tires; low for bicycles.

### Two-retailer analysis with country-of-origin information
- Retail-level indices for four product groups (imported from China & affected; imported from China & unaffected; not imported from China & affected; not imported from China & unaffected) show similar inflation behavior across groups, with an exception: unaffected products sold by China exhibited the largest increase in inflation rates over the sample.
- Pricing moments:
  - Median price spell durations: Retailer 1 median 9.7 months; Retailer 2 median 8.5 months.
  - Percent of products never experiencing a price change: Retailer 1 49 percent; Retailer 2 37 percent.
- Identification: HS6 association via 3CE with manual checks for about one-quarter of products.

### Regression evidence on retail passthrough
- Estimation: monthly regression of change in retail prices on current and lagged tariff changes with sector and China-source fixed effects; Equation (4).
- Key estimated cumulative effects (sum over 1 year; ∑_{l=0}^{11} γ^R_CN,l) — monthly data Jan 2017–Jul 2019:
  - All products (both retailers): 0.044*** (0.009)
  - Retailer 1 only: 0.049*** (0.013)
  - Retailer 2 only: 0.046*** (0.011)
  - Imported products only: 0.046*** (0.009)
  - Household products: 0.045*** (0.010)
  - Electronics products: 0.070*** (0.025)
- Interpretation: coefficient 0.044 means that after one year, a 10 percentage point tariff increase on a good is associated with a 0.44 percent increase in that good’s price relative to other goods in the same sector.
- Robustness: inclusion of month dummies raises one-year estimates to 0.057 for all products, 0.063 for household products, and 0.073 for electronics.
- China φ^R,Ω_CN (Affected): small negative values (e.g., -0.001) with mixed significance.
- Adjusted R^2 values: near zero, reflecting low explanatory power for cross-product variation.
- Subsample checks: manual HS matching or retailer-imported subsets show no large differences but have less power.

### International comparison: US vs Canada
- Rationale: compare affected-sector dynamics in the US (tariffs imposed) versus Canada (no tariffs) to assess whether retailers absorb costs.
- CPI-sector comparisons:
  - Unaffected sectors: Canada’s unaffected sectors had higher inflation before mid-2018; both countries’ unaffected-sector indices flat after tariffs.
  - Affected sectors: moderate increase in inflation among affected categories in the United States after tariffs; a similar but lesser increase also present in Canada.
  - Caveats: imperfect sector matching, differences in sector definitions and consumption baskets, and lack of distinction between China and other trade partners.
- Matched retail-product comparisons:
  - Retailer 2: 2,436 products sold both in the United States and Canada identified by exact model-number matches; NO obvious unusual dynamics in US prices relative to Canadian prices.
- Multi-retailer matched comparisons:
  - Data from six other retailers across 43 3-digit product categories; US and Canadian price indices show no clear evidence that retailers raised prices for US customers relative to Canadian customers.
- Conclusion: retailer profit margins absorbed at least a moderate amount of the adjustment to the import tariffs.

### Overall conclusions and implications
- Tariff passthrough is sectorally heterogeneous:
  - High and rapid passthrough for washing machines.
  - Slower but ultimately high passthrough for some sectors (e.g., tires).
  - Little or negative passthrough in other sectors (e.g., bicycles).
- Retail-level evidence: regression estimates imply modest relative retail price effects — a 10 percentage point tariff increase associated with ~0.44 percent higher price after one year for the affected good relative to other goods in the same sector (estimate 0.044); electronics estimate 0.070.
- Margin adjustment and general equilibrium considerations:
  - Total import cost increased roughly one-for-one with tariffs.
  - If imported input cost accounts for half of a retailer’s marginal cost, a 20 percent tariff would imply a larger required retail price hike to maintain margins — observed retail changes are smaller, implying retailers/producers altered margins or absorbed costs.
  - International comparisons suggest retailers absorbed at least a moderate share of tariff-induced cost increases rather than fully passing them through to US consumers.

### Other adjustment margins: front-running and trade diversion (Section 4.4)
- Front-running and inventory buildup by two large US retailers:
  - Maritime bills of lading from Datamyne used to sum monthly tonnage; plotted as 3-month moving averages.
  - Tonnage imported from China was around 55,000 tons and relatively flat from Q3 2016 through Q2 2017, then jumped in August 2017.
  - Imports increased roughly 20 percent at the August 2017 event, consistent with front-running to import supplies prior to tariff imposition.
  - When tariffs were announced, imports jumped further, then declined a bit thereafter but remained elevated by early 2019.
  - Many goods likely affected by the 10 percent tariff rate and importers may have stockpiled before announced 25 percent tariffs.
- Trade diversion to non-China suppliers:
  - From near-zero prior to tariffs, imports from countries other than China rose above 20,000 tons per month after tariffs were introduced.
  - China’s share of these two firms’ total imports was about 97 percent prior to the tariffs, then declined to about 80 percent since the late summer of 2018.
  - Results robust to plotting containers or value instead of tons.
  - Interpretation: front-running and partial supplier substitution away from China by these large importers; generalizability to the rest of the US retail sector is an open question.

### US export-price response to retaliatory tariffs
- Visual evidence:
  - Post-tariff, prices of unaffected US export goods were highly stable while prices of affected goods dropped by about 7 percent.
  - Differentiated goods account for more than 90 percent of the affected imports to the United States from China but less than half of the US exports to countries that imposed retaliatory tariffs; agricultural goods account for roughly 10 percent of affected US exports.
  - Accounting decomposition shows undifferentiated goods and agricultural goods drive the decline in US export prices.
- Monthly regression evidence (Table 5; Obs. = 433,664):
  - Tariffs 1 yr. (∑_{l=0}^{11} γ^E_l):
    - Column (1): -0.541*** (0.107)
    - Column (2): -0.525*** (0.111)
    - Column (3): -0.481*** (0.111)
  - China Tariffs 1 yr. (∑_{l=0}^{11} γ^E,CN_l):
    - Column (5): -0.628*** (0.152)
  - Non-China Tariffs 1 yr. (∑_{l=0}^{11} γ^E,−CN_l):
    - Column (5): 0.064 (0.115)
  - ERPT 1 yr. (∑_{l=0}^{11} β^E,S_l):
    - 0.188*** (0.018) where reported
  - PPI PT 1 yr. (∑_{l=0}^{11} β^E,X_l):
    - 0.239*** (0.040) and similar in other columns
  - Adj. R^2 ranges from 0.000 to 0.002.
- Interpretation:
  - About a 54 percent passthrough of retaliatory tariffs into ex-tariff US export prices after 12 months (column (1)).
  - A 10 percent tariff imposed on US exports reduces US ex-tariff export prices by about 5.4 percent per column (1); estimate reduces to 4.8 percent when controlling for other factors (column (4)).
  - Retaliation from China drives strong negative effects; shipments to countries other than China show no statistically significant decline.

### Conditional export regressions (Table 6; Obs. = 66,104)
- Tariffs 12×γ^E (Annualized):
  - Column (1): -0.876*** (0.164)
  - Column (2): -0.919*** (0.170)
  - Column (3): -0.793*** (0.168)
- China Tariffs 12×γ^E,CN (Annualized):
  - Column (5): -0.993* (0.201)
- Non-China Tariffs 12×γ^E,−CN (Annualized):
  - Column (5): 0.405 (0.320)
- ERPT β^E,S:
  - 0.362*** (0.029)
- PPI PT β^E,X:
  - 1.028*** (0.079), 1.023*** (0.079), 1.021*** (0.079)
- Adj. R^2 ranges from 0.000 to 0.014.
- Interpretation: conditional on price change, exchange rate passthrough estimates rise to about 36 percent; annualized tariff coefficients large and significant, driven largely by exports to China.

### Mechanism, asymmetric incidence, and time horizons
- Product composition matters:
  - US imports targeted by US tariffs tended to be highly differentiated (harder to substitute), while many US exports targeted by retaliatory tariffs were less differentiated and included roughly 10 percent agricultural goods.
- Consequence:
  - Foreign exporters facing US tariffs showed little or no ex-tariff price declines, while US exporters facing retaliatory tariffs substantially reduced ex-tariff export prices, particularly on shipments to China.
- Short-run vs medium/long-run considerations:
  - Non-price margins (front-running, trade diversion) documented indicate important short-run adjustments; medium- or longer-term responses could differ if tariffs remain in place.

*Source — wpiea2019238-print-pdf - 2.3 log points in magnitude and focus only on market transactions.*

### 2.3 log points in magnitude and focus only on market transactions.

### wpiea2019238-print-pdf - 2.3 log points in magnitude and focus only on market transactions.

### Data and scope
- Analysis uses prices of trades invoiced in US dollars, representing over 94 percent of US trade in the sample.
- Petroleum products are excluded.
- Partner countries restricted to those with data on aggregate prices and exchange rates (macro variables cover 182 countries).
- BLS import price data are at the individual good level and can be concorded to HS codes; over one-third of BLS import prices are non-market transactions (intrafirm or related-party shipments) and such non-market transactions are excluded from market-focused analyses.
- Goods are matched to 6-digit HS codes and the associated tariff is taken as the highest value among corresponding 8-digit HS codes (this assumption holds exactly for over 95 percent of 6-digit codes).
- Exclusions: goods impacted both by a China tariff and another product-based tariff (e.g., steel and aluminum products, lumber, washing machines, solar panels) and data on imports from India for a subset of analyses.

### US imports from China — descriptive patterns
- Tariff waves on China in 2018–2019:
  - July 2018: 25 percent ad-valorem tariff on roughly $34 billion of imports.
  - August 2018: 25 percent tariff extended to roughly $16 billion of shipments.
  - September 2018: 10 percent tariff applied to roughly $200 billion in goods.
  - May 2019: tariff on the September wave increased from 10 to 25 percent.
- Price-index evidence (indices normalized to 1 in June 2018):
  - Affected goods imported from China saw an immediate jump in their price inclusive of tariffs in the month the policy was implemented; the scale of the jumps is only slightly below the scale of the tariff rates.
  - Ex-tariff price indices for affected Chinese goods did not exhibit meaningful breaks from prior trends.
- Frequency-of-price-change patterns (Figure 2):
  - No obvious differences in the share of monthly price decreases or increases across groups (affected/unaffected × China/Not-China).
  - Prices of products affected and imported from China appear at least as stable as others; little evidence that tariffs changed price stickiness.

### Regression framework and identification
- Two main specifications:
  1. Monthly-data specification including zero price-change months (Equation (2)):
     - Dependent variable: ∆ ln(P_I_{i,j,k,t}) where P_I is ex-tariff price of item i from country j in sector k at month t.
     - Controls: sector fixed effects δ^I_k, China-sector fixed effects φ^I,Ω_CN and φ^I,−Ω_CN, lags of log gross additional tariff ∆τ_CN,k,t−l (up to 11 lags), lags of ∆ ln(S_{j,t−l}) (exchange rate) and ∆ ln(X_{j,t−l}) (producer price index).
     - Cumulative 12-month tariff passthrough estimated as ∑_{l=0}^{11} γ^I_CN,l.
     - 12-month ERPT estimated as ∑_{l=0}^{11} β^I,S_l.
  2. Conditional-on-price-change (spell) specification (Equation (3)):
     - Uses only non-zero price changes, scales changes to monthly frequency, reports 12×γ^I (annualized tariff-associated change) and β^I,S (ERPT) on the scaled changes.

### Monthly regression estimates (Table 1)
- Sample: monthly data January 2005 to August 2019; Obs. = 820,318 (columns vary by included fixed effects).
- Key coefficient estimates (robust standard errors in parentheses; ***, **, * denote 1, 5, 10 percent significance):
  - Tariffs 1 yr. (∑_{l=0}^{11} γ^I_CN,l):
    - Column (1): -0.079*** (0.026)
    - Column (2): -0.076*** (0.028)
    - Column (3): -0.018 (0.030)
  - ERPT 1 yr. (∑_{l=0}^{11} β^I,S_l):
    - Column (3): 0.219*** (0.027)
    - Column (4): 0.221*** (0.027)
  - PPI PT 1 yr. (∑_{l=0}^{11} β^I,X_l):
    - Column (3): 0.019 (0.070)
    - Column (4): 0.012 (0.073)
  - China φ^I,Ω_CN (Affected):
    - Column (2): 0.000 (0.000)
  - China φ^I,−Ω_CN (Not-Affected):
    - Column (2): -0.000 (0.000)
    - Column (3): -0.001 (0.001)
  - Adjusted R^2:
    - Column (1): 0.000
    - Column (2): 0.003
    - Column (3): 0.004
    - Column (4): 0.004
- Interpretation example from Column (1): the estimated coefficient of -0.079 implies that a 10 percent tariff would be associated with a 0.8 percent lower ex-tariff price and a 9.2 percent higher overall price faced by the importer.

### Conditional-on-price-change regression estimates (Table 2)
- Sample: non-zero price-change spells; Obs. = 99,406.
- Key coefficient estimates (robust standard errors in parentheses):
  - Tariffs annualized (12×γ^I):
    - Column (1): -0.228 (0.171)
    - Column (2): -0.109 (0.187)
    - Column (3): 0.006 (0.188)
  - ERPT β^I,S:
    - Column (1): 0.382*** (0.052)
    - Column (2): 0.381*** (0.052)
  - PPI passthrough β^I,X:
    - Column (1): 0.757*** (0.110)
    - Column (2): 0.766*** (0.111)
  - China φ^I,Ω_CN (Affected):
    - Column (2): 0.004*** (0.002)
  - China φ^I,−Ω_CN (Not-Affected):
    - Column (2): 0.002* (0.001)
    - Column (3): 0.003** (0.001)
  - Adjusted R^2:
    - Column (1): 0.000
    - Column (2): 0.006
    - Column (3): 0.017
    - Column (4): 0.018
- Summary: estimates are qualitatively consistent with monthly specifications — tariffs are associated with economically or statistically insignificant changes in ex-tariff import prices in many specifications, while exchange rate passthrough rises to roughly 38 percent in these conditional regressions.

### Tariffs on steel imports (descriptive)
- Background:
  - March 2018: US placed a 25 percent tariff on steel imports from all countries; exemptions initially for Argentina, Australia, Brazil, Canada, Mexico, EU, South Korea.
  - June 2018: exemptions lifted for Canada, the EU, and Mexico (second wave); exemptions for other countries made permanent.
- Price-index evidence for steel (Figure 3):
  - Steel prices were volatile in the preceding four years.
  - After steel tariffs were introduced, prices on imports from affected countries rose; imports from affected countries jumped to roughly 20 percent above those from unaffected countries.
  - Regression analyses for steel give similar qualitative conclusions, but estimates are imprecise due to the small number of imported steel products.

### Summary of results on US import tariffs
- Across analyses (China imports, steel imports, aggregated price indices, and product-level regressions) the 12-month price response to US import tariffs imposed in 2018–2019 shows:
  - Ex-tariff prices do not obviously behave differently for goods affected by trade policy versus those not affected.
  - Tariffs exhibited nearly complete passthrough into the total import cost; incidence of tariffs lies largely with the United States (importers/consumers).
- Exchange rate passthrough:
  - Using the same data/time period, estimated ERPT into import prices is in the range of 25 to 35 percent after one year (conditional regressions indicate up to roughly 38 percent), consistent with existing literature and substantially lower than tariff passthrough into total import prices.
- Implications:
  - Standard trade and international macroeconomic models that assume symmetric responses to tariffs and exchange rate shocks may be misleading in the short run; observed asymmetry suggests models might better apply to longer-run outcomes or be amended to allow for more uncertainty or mean-reversion in shocks.
  - The recent depreciation of the Chinese renminbi did not offset the impact of the tariffs for US importers during the sample period.

### Downstream (retail) pricing — preview of findings
- Preview: retail-level evidence (next section) uses millions of online prices from major US retailers.
- Main preview conclusions:
  - Some evidence that tariffs have passed through into higher retail prices, but effects are much more muted than for total import prices.
  - Retailers appear to have absorbed much of the higher import costs via lower margins.
  - Additional retailer adjustments observed: front-running of tariffs (inventory build-up before higher tariffs took effect) and partial diversion of orders to non-Chinese suppliers — adjustments that could flatten observed retail price responses so far.

*Italic: Source — wpiea2019238-print-pdf - 2.3 log points in magnitude and focus only on market transactions.*

### 2018.  This sector has received significant attention from academics, and is the focus of Flaaen,

### wpiea2019238-print-pdf - 2018

### Data and measurement
- Washing machine retail prices:
  - Prices for about 700 washing machines from PriceStats and the Billion Prices Project (BPP), scraped at a daily or weekly frequency from the online pages of 16 large multi-channel US retailers.
  - Washing machines defined as goods appearing in the data for at least one year, with product descriptions including “washing machine” or “washer”, and excluding disqualifying words such as “washer fluid”.
  - Adjacent retail price observations that differ by more than 2.3 log points in absolute value are excluded from retail price analyses.
- Additional product samples:
  - 300 handbags from 12 retailers.
  - 400 tires from 7 retailers.
  - 5,000 refrigerators from 18 retailers.
  - 200 bicycles from 11 retailers.
- Two large US retailers dataset:
  - Retailer 1: about 38,000 products (scraped webpages including country of origin).
  - Retailer 2: about 55,000 products (scraped prices matched to top 100,000 products by sales rank; country of origin, product sales rank, and a text product description provided directly by retailer).
  - Combined: more than 90,000 products covering nearly 1,991 different 6-digit HS categories.
  - Imported products: about 60,000 products imported from more than 80 countries.
  - Products imported from China: about 43,000.
  - Products in categories affected by tariffs: about 59,460 (sample-level number reported).
  - Products from China & in affected categories: about 30,101.
- Country-of-origin and HS classification:
  - Automated HS6 matching performed using 3CE technologies; roughly three-quarters of products classified automatically and the remainder manually with research-assistant input.

### Washing machines: high and rapid passthrough
- Price series construction and normalization:
  - BPP and CPI washing machine price indices normalized to equal 1 in February 2018 (the month tariffs were imposed).
- Pre-tariff trend:
  - BPP and CPI price indices for washing machines declined by about 5 percent per year prior to tariffs.
- Post-tariff dynamics:
  - Within a few months of the import tariffs, both BPP and CPI series exhibit a break, with inflation rates switching from negative to positive.
  - In the second half of 2018:
    - Washing machine inflation in BPP data typically between 5 and 10 percent.
    - Washing machine inflation in CPI data typically between 10 and 15 percent.
- Interpretation:
  - Evidence strongly suggests moderate to high passthrough of washing machine tariffs to retail prices.
  - Heterogeneity across brands: Samsung prices clearly increased; Haier showed little change. Similar pricing patterns observed for US brands and imported brands, indicating tariffs led to general price hikes including on products unaffected by tariffs.

### Other affected goods: heterogeneous responses
- Handbags, tires, refrigerators, bicycles:
  - Price indices normalized to equal 1 in September 2018 (the 10 percent tariffs introduced on that date were most relevant to this group).
  - Three months after the first tariffs: none of these four goods exhibited sharp price increases relative to trend.
  - By the time tariffs were increased to 25 percent:
    - Handbags and tires experienced unusually rapid price increases.
    - Refrigerators exhibited a mild increase in inflation relative to pre-tariff trend, but the increase appears to have started before tariffs.
    - Bicycles showed, if anything, decreased price inflation in the post-tariff regime.
- Summary:
  - Tariff passthrough varies by sector: rapid/high for washing machines; slower/ultimately high for tires; low for bicycles.

### Two-retailer analysis with country-of-origin information
- Retail-level price indices:
  - Plots of daily retail price indices and annual retail inflation rates for four product groups (imported from China & affected; imported from China & unaffected; not imported from China & affected; not imported from China & unaffected) show similar inflation behavior across groups.
  - Exception: unaffected products sold by China exhibited the largest increase in inflation rates over the sample period.
- Pricing dynamics summary (selected moments):
  - Median price spell durations: Retailer 1 median 9.7 months; Retailer 2 median 8.5 months.
  - Percent of products never experiencing a price change: Retailer 1 49 percent; Retailer 2 37 percent.
- Identification strategy:
  - Products associated to HS6 via 3CE; manual checks for a subset (about one-quarter) to reduce misclassification concerns.

### Regression evidence on retail passthrough
- Estimation framework:
  - Monthly regression of change in retail prices on current and lagged tariff changes, with fixed effects allowing different price trends per 3-digit COICOP sector and different trends for China-sourced products that are and are not affected by tariffs.
  - Equation (4) specified as:
    ∆ ln(P_R i,j,k,t) = δ^R_k + φ^R,Ω_CN + φ^R,−Ω_CN + ∑_{l=0}^9 γ^R_CN,l ∆τ_CN,k,t−l + ϵ_i,j,k,t
- Key estimated cumulative effects (sum over 1 year; ∑_{l=0}^{11} γ^R_CN,l):
  - Column estimates (monthly data Jan 2017–Jul 2019):
    - All products (both retailers): 0.044*** (standard error (0.009))
    - Retailer 1 only: 0.049*** (0.013)
    - Retailer 2 only: 0.046*** (0.011)
    - Imported products only: 0.046*** (0.009)
    - Household products: 0.045*** (0.010)
    - Electronics products: 0.070*** (0.025)
  - Interpretation provided in text: the coefficient of 0.044 means that after one year, a 10 percentage point tariff increase on a good is associated with a 0.44 percent increase in that good’s price relative to other goods in the same sector.
  - Robustness note: If month dummies are additionally included, estimates for price increases after one year increase to 0.057 for all products, 0.063 for household products, and 0.073 for electronics products.
- Additional coefficient estimates:
  - China φ^R,Ω_CN (Affected): reported small negative values (e.g., -0.001) with limited magnitude and mixed significance.
  - Adjusted R^2 values: near zero (e.g., 0.000, 0.002) reflecting low explanatory power of specifications for cross-product variation in retail price changes.
- Subsample checks:
  - Regressions on subsets with manual HS matching or products directly imported by Retailer 2 do not expose large differences between affected and unaffected groups, though they have less power.

### International comparison: US vs Canada
- Rationale:
  - If retailers absorb tariff cost increases (e.g., by increasing margins on unaffected goods or by absorbing margin reductions), affected US sector prices might not rise relative to foreign comparators; conversely, if tariffs directly increase consumer prices in the US, affected US sectors should rise relative to similar sectors in countries that did not impose tariffs (Canada).
- CPI-sector comparisons:
  - Affected CPI sectors designation (nine CPI sectors identified as “affected” based on tariff coverage; sectors listed in source text) vs unaffected sectors.
  - Figure 8 observations:
    - Unaffected sectors: Canada’s unaffected sectors had higher inflation before mid-2018; both US and Canada unaffected-sector indices are essentially flat after tariffs.
    - Affected sectors: moderate increase in inflation among affected categories in the United States after tariffs; a similar but lesser increase also present in Canada (which did not impose tariffs on Chinese imports).
  - Caveats noted: imperfect sector matching across countries, differences in sector definitions and consumption baskets, and lack of distinction between trade from China and from other countries.
- Matched retail-product comparisons:
  - Retailer 2: 2,436 products sold both in the United States and Canada identified via exact model-number matches (model numbers required to have at least five characters; near-identical model numbers differing only by country suffix excluded).
  - NO obvious unusual dynamics in US prices for these matched goods relative to Canadian goods over the tariff period.
- Multi-retailer matched comparisons:
  - Additional pricing data for six other retailers selling in both countries; 43 3-digit product categories selected and price indices constructed per category, country, and retailer.
  - US and Canadian price indices constructed using equal weights for each retailer and same average sectoral expenditure weights for both countries.
  - Results: no clear evidence that retailers raised prices for US customers relative to Canadian customers for the same set of goods.
- Conclusion from international comparisons:
  - Retailer profit margins absorbed at least a moderate amount of the adjustment to the import tariffs.

### Overall conclusions and implications
- Tariff passthrough is sectorally heterogeneous:
  - High and rapid passthrough for washing machines (clear, large retail price increases).
  - Slower but ultimately high passthrough for some sectors (e.g., tires).
  - Little or negative passthrough in other sectors (e.g., bicycles; refrigerators show mild pre-existing inflation).
- Retail-level evidence:
  - Regression estimates imply modest relative retail price effects: a 10 percentage point tariff increase associated with ~0.44 percent higher price after one year for the affected good relative to other goods in the same sector (estimate 0.044).
  - Electronics and household products exhibit somewhat larger responses (electronics estimate 0.070).
  - Inclusion of month dummies raises the one-year cumulative estimates slightly (e.g., 0.057 overall).
- Margin adjustment and general equilibrium considerations:
  - The total cost of imports increased roughly one-for-one with tariffs (earlier section reference).
  - If imported input cost accounts for half of a retailer’s marginal cost, a 20 percent tariff would imply a larger required retail price hike to maintain margins — yet observed retail price changes are smaller, implying retailers and/or producers altered margins or absorbed costs.
  - International comparisons (Canada vs US) and matched-retailer analyses suggest retailers absorbed at least a moderate share of tariff-induced cost increases rather than fully passing them through to US consumers.

*Italic source: wpiea2019238-print-pdf - 2018 (https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019238-print-pdf.pdf)*

### 4.4    Other Adjustment Margins:  Front-Running and Trade Diversion

### 4.4    Other Adjustment Margins:  Front-Running and Trade Diversion

### Front-running and inventory buildup by two large US retailers
- Data source and measurement:
  - Maritime bills of lading from Datamyne used to sum monthly tonnage imported by two US retailers; plotted as 3-month moving averages.
- Key observations:
  - Tonnage imported from China (solid blue line) was around 55,000 tons and relatively flat from Q3 2016 through Q2 2017, but appears to jump in August 2017 (date indicated with dashed vertical line).
  - Imports increased roughly 20 percent at the point of the August 2017 event, consistent with front-running to import supplies prior to tariff imposition.
  - When tariffs were announced, imports jumped further and then declined a bit thereafter, but remained elevated by early 2019.
  - Many of these goods were likely affected by the 10 percent tariff rate and importers may have stockpiled them before the announced 25 percent tariffs on those same goods.

### Trade diversion to non-China suppliers
- Measured behavior:
  - From a near-zero level prior to tariffs, imports from countries other than China rose above 20,000 tons per month after tariffs were introduced (red dashed line).
- Market share changes:
  - China’s share of these two firms’ total imports was about 97 percent prior to the tariffs, then declined to about 80 percent since the late summer of 2018.
- Robustness:
  - Results for Figure 11 (tons) would be very similar if shipping containers or value were plotted instead of tons.
- Interpretation:
  - These two importers engaged in front-running and partially adjusted through supplier substitution away from China.
  - Note: These two retailers are large firms that might have more working capital and an easier time building inventories; the extent to which this pattern applies to the rest of the US retail sector remains an open question.

### US export-price response to retaliatory tariffs
- Context:
  - Many countries (Canada, China, the EU, Mexico, among others) imposed retaliatory measures on the United States in response to US trade policies enacted in 2018.
  - The paper studies ex-tariff prices of US exports affected and unaffected by recently imposed foreign tariffs.
- Visual evidence (Figure 12 and Figure 13):
  - Post-tariff, prices of unaffected US export goods were highly stable while prices of affected goods dropped by about 7 percent.
  - Using Rauch (1999) classification:
    - Differentiated goods account for more than 90 percent of the affected imports to the United States from China but less than half of the US exports to countries that imposed retaliatory tariffs.
    - Agricultural goods accounted for roughly 10 percent of affected US exports in the sample.
  - Accounting decomposition (Figures 13(a) and 13(b)) shows undifferentiated goods and agricultural goods drive the decline in US export prices.

### Regression evidence from monthly data (Table 5)
- Specification (equation (5)) estimates of retaliatory tariff passthrough into ex-tariff US export prices:
  - Tariffs 1 yr. (∑_{l=0}^{11} γ^E_l):
    - Column (1): -0.541*** (0.107)
    - Column (2): -0.525*** (0.111)
    - Column (3): -0.481*** (0.111)
  - China Tariffs 1 yr. (∑_{l=0}^{11} γ^E,CN_l):
    - Column (5): -0.628*** (0.152)
  - Non-China Tariffs 1 yr. (∑_{l=0}^{11} γ^E,−CN_l):
    - Column (5): 0.064 (0.115)
  - ERPT 1 yr. (∑_{l=0}^{11} β^E,S_l):
    - 0.188*** (0.018) in columns reporting it
  - PPI PT 1 yr. (∑_{l=0}^{11} β^E,X_l):
    - 0.239*** (0.040) and similar in other columns
  - Model fit and sample:
    - Adj. R^2 ranges from 0.000 to 0.002 across reported columns.
    - Obs. = 433,664 in all columns.
- Interpretation:
  - About a 54 percent passthrough of retaliatory tariffs into ex-tariff US export prices after 12 months (column (1) result).
  - A 10 percent tariff imposed on US exports reduces US ex-tariff export prices by about 5.4 percent per column (1).
  - The estimate reduces to 4.8 percent when controlling for other price-determining factors (column (4)).
  - Retaliation from China accounts for about three-quarters of observations; shipments to China show a strong negative effect, while shipments to countries other than China show no statistically significant decline.

### Regression evidence conditional on non-zero price changes (Table 6)
- Specification (equation (6)) results (annualized where indicated):
  - Tariffs 12×γ^E (Annualized):
    - Column (1): -0.876*** (0.164)
    - Column (2): -0.919*** (0.170)
    - Column (3): -0.793*** (0.168)
  - China Tariffs 12×γ^E,CN (Annualized):
    - Column (5): -0.993* (0.201)
  - Non-China Tariffs 12×γ^E,−CN (Annualized):
    - Column (5): 0.405 (0.320)
  - ERPT β^E,S:
    - 0.362*** (0.029) and similar in reported columns
  - PPI PT β^E,X:
    - 1.028*** (0.079), 1.023*** (0.079), 1.021*** (0.079) in reported columns
  - Model fit and sample:
    - Adj. R^2 ranges from 0.000 to 0.014 across reported columns.
    - Obs. = 66,104 in all columns.
- Interpretation:
  - Conditional on price change, exchange rate passthrough estimates rise to about 36 percent (ERPT ~0.362).
  - Annualized tariff coefficients are large and statistically significant when conditioning on price changes, driven largely by exports to China.

### Mechanism and asymmetric incidence
- Product composition matters:
  - US imports targeted by US tariffs tended to be highly differentiated goods (making substitution more difficult), while many US exports targeted by retaliatory tariffs were less differentiated and included roughly 10 percent agricultural goods.
- Consequence:
  - Foreign exporters facing US tariffs showed little or no ex-tariff price declines, while US exporters facing retaliatory tariffs significantly reduced their ex-tariff export prices, particularly on shipments to China.
- Aggregate implication:
  - The retaliatory tariffs applied to US exports exhibited significantly lower passthrough than the US tariffs on imports, driven in large part by differences in differentiation of goods.

### Short-run vs. medium/long-run considerations
- Non-price margins documented:
  - Evidence of front-running (inventory buildup) and trade diversion away from China suggests important non-price margins of adjustment.
- Implication:
  - These non-price margins indicate that observed adjustments may reflect short-run behavior; medium- or longer-term responses could differ if tariffs remain in place.

*Source: 4.4 Other Adjustment Margins: Front-Running and Trade Diversion (wpiea2019238-print-pdf)*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2019/wpiea2019238-print-pdf.pdf_