Tariff Summary Table

Section 301 tariff lists: coverage, rates, dates and share of US imports from China, 2017–18.
List HS8 codes in tariff file HS8 codes with no trade, 2017–21 Initial rate Effective date Later change Date of change Imports 2017–18 (USD bn) Share of 2017–18 total
List 1 818 26 25.0% Jul 06, 2018
62.2 5.9%
List 2 279 4 25.0% Aug 23, 2018
28.6 2.7%
List 3 5,745 538 10.0% Sep 24, 2018 Raised to 25% May 10, 2019 400.8 38.3%
List 4A 3,233 905 15.0% Sep 01, 2019 Cut to 7.5% Feb 14, 2020 209.2 20.0%
List 4B 550 0 15.0% Dec 15, 2019 Suspended Dec 2019 310.3 29.6%
Never listed
35.8 3.4%
Total
1,473
1,046.9 100.0%
Note: HS8 codes with no imports are tariff-file codes with no recorded imports from China in any month of 2017–21. Imports and shares cover 2017–18. Source: USITC DataWeb; USTR Section 301 notices.
## share_with_pregrowth 
##            0.8064443

Model Trial

Lists 1, 2 and 3 are pooled, with List 4B products as the control group.

Notation.

Symbol Meaning
\(p\) Product: HS8 for \(\text{CHN}\to\text{US}\) from USITC, HS6 for the Comtrade legs
\(l\) Tariff list, \(l\in\{1,2,3\}\), first tariffed in month \(T_l\) (Jan 2017 \(=1\)): \(T_1=19\), \(T_2=20\), \(T_3=21\)
\(j\) Third country (rerouting candidate), e.g. Vietnam
\(\tau\) Trade leg: \(\tau\in\{\text{CHN}\to\text{US},\ \text{CHN}\to j,\ j\to\text{US}\}\)
\(Y^{\tau}_{p,t}\) Imports of product \(p\) on leg \(\tau\) in month \(t\)
\(h\) Horizon in months relative to \(T_l\)
\(\text{Treated}_p\) In the sub-experiment for list \(l\): 1 if \(p\) is first tariffed on list \(l\), 0 if \(p\) is a List 4B control

Treatment comes from the US tariff on the Chinese version of product \(p\), so \(\text{Treated}_p\) is the same on every leg. Each leg \(\tau\), and each country \(j\), is estimated separately, so every coefficient below carries an implicit \(\tau\) (and \(j\)) superscript.

Outcome and baselines. The outcome is a long difference from a pre-tariff baseline \(\bar Y^{\tau}_{pl}\): \[ \Delta_h \ln Y^{\tau}_{pl}=\ln Y^{\tau}_{p,\,T_l+h}-\ln \bar Y^{\tau}_{pl}, \qquad \bar Y^{\tau,(1)}_{pl}=Y^{\tau}_{p,\,T_l-1}, \qquad \bar Y^{\tau,\mathrm{B}}_{p}=\frac{1}{6}\sum_{m\in\mathrm{B}}Y^{\tau}_{p,\,m},\quad \mathrm{B}=\{\text{Jan 2018},\dots,\text{Jun 2018}\} \] The 2018 pre-tariff baseline \(\bar Y^{\tau,\mathrm{B}}_{p}\) is the same calendar window for every list, so outcome months inside \(\mathrm{B}\) are skipped.

PreGrowth. The 2017 pre-trend of product \(p\) on the same leg: the log change in average monthly imports from the first half of 2017 to the second half. Averaging six months, rather than comparing December with January alone, keeps more products (81% vs 64% of tariffed HS8 codes for \(\text{CHN}\to\text{US}\)) and is less sensitive to single-month swings. \[ \text{PreGrowth}^{\tau}_p=\ln\left(\frac{1}{6}\sum_{m=7}^{12}Y^{\tau}_{p,\,m}\right)-\ln\left(\frac{1}{6}\sum_{m=1}^{6}Y^{\tau}_{p,\,m}\right) \]

  1. List FE only \[ \ln Y^{\tau}_{p,\,T_l+h}-\ln Y^{\tau}_{p,\,T_l-1}=\beta^{\tau}_h\,\text{Treated}_p+\gamma_l+\varepsilon_{pl} \]

  2. Product FE \[ \ln Y^{\tau}_{p,t}=\alpha_{pl}+\lambda_{lt}+\sum_{k\neq -1}\beta^{\tau}_k\,\mathbf 1\{t-T_l=k\}\,\text{Treated}_p+\varepsilon_{plt} \]

  3. Industry FE \[ \ln Y^{\tau}_{p,\,T_l+h}-\ln Y^{\tau}_{p,\,T_l-1}=\beta^{\tau}_h\,\text{Treated}_p+\gamma_{l,\,\text{HS2}(p)}+\varepsilon_{pl} \]

  4. PreGrowth \[ \ln Y^{\tau}_{p,\,T_l+h}-\ln Y^{\tau}_{p,\,T_l-1}=\beta^{\tau}_h\,\text{Treated}_p+\delta^{\tau}_h\,\text{PreGrowth}^{\tau}_p+\gamma_{l,\,\text{HS2}(p)}+\varepsilon_{pl} \]

  5. WLS \[ \hat\beta^{\tau}_h=\arg\min_{\beta_h,\,\gamma}\sum_{l\in\{1,2,3\}}\sum_{p} w^{\tau}_{pl}\Bigl[\ln Y^{\tau}_{p,\,T_l+h}-\ln Y^{\tau}_{p,\,T_l-1}-\beta_h\,\text{Treated}_p-\gamma_{l,\,\text{HS2}(p)}\Bigr]^2, \qquad w^{\tau}_{pl}=Y^{\tau}_{p,\,T_l-1} \]

  6. 2018 pre-tariff average baseline \[ \ln Y^{\tau}_{p,\,T_l+h}-\ln\bar Y^{\tau,\mathrm{B}}_{p}=\beta^{\tau}_h\,\text{Treated}_p+\gamma_{l,\,\text{HS2}(p)}+\varepsilon_{pl} \]

  7. Combined HS2 FE, PreGrowth, and WLS \[ \hat\beta^{\tau}_h=\arg\min_{\beta_h,\,\delta_h,\,\gamma}\sum_{l\in\{1,2,3\}}\sum_{p} w^{\tau}_{pl}\Bigl[\ln Y^{\tau}_{p,\,T_l+h}-\ln Y^{\tau}_{p,\,T_l-1}-\beta_h\,\text{Treated}_p-\delta_h\,\text{PreGrowth}^{\tau}_p-\gamma_{l,\,\text{HS2}(p)}\Bigr]^2, \qquad w^{\tau}_{pl}=Y^{\tau}_{p,\,T_l-1} \]

  8. Combined HS2 FE, 2018 pre-tariff average baseline, and WLS \[ \hat\beta^{\tau}_h=\arg\min_{\beta_h,\,\gamma}\sum_{l\in\{1,2,3\}}\sum_{p} w^{\tau}_{p}\Bigl[\ln Y^{\tau}_{p,\,T_l+h}-\ln\bar Y^{\tau,\mathrm{B}}_{p}-\beta_h\,\text{Treated}_p-\gamma_{l,\,\text{HS2}(p)}\Bigr]^2, \qquad w^{\tau}_{p}=\bar Y^{\tau,\mathrm{B}}_{p} \]

  9. Combined HS2 FE, 2018 pre-tariff average baseline, 2017 PreGrowth, and WLS \[ \hat\beta^{\tau}_h=\arg\min_{\beta_h,\,\delta_h,\,\gamma}\sum_{l\in\{1,2,3\}}\sum_{p} w^{\tau}_{p}\Bigl[\ln Y^{\tau}_{p,\,T_l+h}-\ln\bar Y^{\tau,\mathrm{B}}_{p}-\beta_h\,\text{Treated}_p-\delta_h\,\text{PreGrowth}^{\tau}_p-\gamma_{l,\,\text{HS2}(p)}\Bigr]^2, \qquad w^{\tau}_{p}=\bar Y^{\tau,\mathrm{B}}_{p} \]

Model Estimates

Each model is estimated separately for every horizon \(h\) and stacked over Lists 1–3, with List 4B products as the control group. Standard errors are clustered by HS4. A product enters a horizon only if it has positive imports in both the baseline and the outcome month.

China to US at HS8 (USITC)

Width of the 95% confidence band by model, China to US imports (HS8), Lists 1–3 vs List 4B.
Model Band width, leads Band width, h = 0–27 Relative to (2) Mean beta, h = 0–27 Implied change Avg treated obs Avg control obs
(0) List FE only 0.199 0.237 0.80 -0.504 -39.6% 3,987 1,393
(1) Product FE (TWFE) 0.200 0.240 0.81 -0.496 -39.1%
(2) HS2 FE 0.210 0.295 1.00 -0.608 -45.6% 3,987 1,393
(3) HS2 FE + PreGrowth 0.199 0.296 1.00 -0.625 -46.5% 3,955 1,391
(4) HS2 FE + WLS 0.236 0.487 1.65 -0.852 -57.4% 3,987 1,393
(5) HS2 FE + 2018 base 0.229 0.325 1.10 -0.608 -45.6% 4,228 1,439
(6) HS2 FE + PreGrowth + WLS 0.277 0.487 1.65 -0.917 -60.0% 3,955 1,391
(7) HS2 FE + 2018 base + WLS 0.230 0.594 2.02 -0.793 -54.7% 4,228 1,439
(8) HS2 FE + 2018 base + PreGrowth + WLS 0.210 0.602 2.04 -0.765 -53.5% 4,152 1,429
Note: Band width is the average of 2 x 1.96 x SE across horizons. Leads are h = -18 to -2 (for the 2018 baseline, months in Jan-Jun 2018 are skipped). Implied change is exp(beta) - 1. Observations are product x list pairs per horizon; model (1) is one pooled regression.

Three trade legs

China to US uses the HS8 USITC data from above. The two legs through the third country use Comtrade HS6 data, where an HS6 product takes the earliest implemented list among its HS8 lines. It is a control if none of its lines was implemented and at least one is on List 4B. A month with no recorded trade counts as zero.

HS8 lines imported from China, 2017–21, by their own tariff status and the status of their HS6 product.
Status of the HS6 product
HS8 line status List 1 List 2 List 3 List 4A List 4B only Never listed Total
List 1 792 0 0 0 0 0 792
List 2 40 235 0 0 0 0 275
List 3 85 31 5,103 0 0 0 5,219
List 4A 39 5 175 2,124 0 0 2,343
List 4B only 6 0 111 270 155 0 542
Never listed 8 0 42 4 0 453 507
Total 970 271 5,431 2,398 155 453 9,678
Note: Diagonal cells keep their status when aggregated to HS6; off-diagonal cells change. An HS6 product takes the earliest implemented list among its HS8 lines; it is List 4B only if no line was implemented and at least one is on List 4B.
How aggregating to HS6 changes each HS8 line’s role (control group: List 4B).
Role at HS8 → role at HS6 HS8 lines Share of HS8 lines HS6 products involved
No change 8,862 91.6% 4,900
Control → Excluded (List 4A) 270 2.8% 148
Excluded (List 4A) → Treated 219 2.3% 112
Treated: moves to an earlier list 156 1.6% 91
Control → Treated 117 1.2% 65
Not used → Treated 50 0.5% 31
Not used → Excluded (List 4A) 4 0.0% 4
Note: Treated = first tariffed on List 1, 2 or 3. Control = List 4B only (never-listed lines are not used). Excluded = first tariffed on List 4A.

Vietnam

Band width and mean effect by model and trade leg, h = 0–27, Lists 1–3 vs List 4B (Viet Nam).
China → US
China → Viet Nam
Viet Nam → US
Model Band width Mean beta Band width Mean beta Band width Mean beta
(0) List FE only 0.237 -0.504 0.415 -0.012 0.767 0.033
(1) Product FE (TWFE) 0.240 -0.496 0.428 0.040 0.769 -0.060
(2) HS2 FE 0.295 -0.608 0.522 0.117 1.039 0.088
(3) HS2 FE + PreGrowth 0.296 -0.625 0.522 0.093 0.931 0.181
(4) HS2 FE + WLS 0.487 -0.852 0.440 0.054 1.042 0.179
(5) HS2 FE + 2018 base 0.325 -0.608 0.510 0.096 1.147 0.015
(6) HS2 FE + PreGrowth + WLS 0.487 -0.917 0.520 0.041 1.081 0.179
(7) HS2 FE + 2018 base + WLS 0.594 -0.793 0.633 0.343 0.862 -0.029
(8) HS2 FE + 2018 base + PreGrowth + WLS 0.602 -0.765 0.668 0.404 0.880 -0.039
Note: Band width is the average of 2 x 1.96 x SE over h = 0 to 27. Mean beta is the average estimate over the same horizons. China to US uses HS8 products (USITC); the other legs use HS6 products (Comtrade).

Mexico

Band width and mean effect by model and trade leg, h = 0–27, Lists 1–3 vs List 4B (Mexico).
China → US
China → Mexico
Mexico → US
Model Band width Mean beta Band width Mean beta Band width Mean beta
(0) List FE only 0.237 -0.504 0.369 0.096 0.511 0.146
(1) Product FE (TWFE) 0.240 -0.496 0.383 0.085 0.511 0.157
(2) HS2 FE 0.295 -0.608 0.430 0.031 0.527 0.108
(3) HS2 FE + PreGrowth 0.296 -0.625 0.425 0.047 0.521 0.093
(4) HS2 FE + WLS 0.487 -0.852 0.789 -0.305 0.952 0.309
(5) HS2 FE + 2018 base 0.325 -0.608 0.455 0.068 0.666 0.055
(6) HS2 FE + PreGrowth + WLS 0.487 -0.917 0.540 -0.245 0.981 0.309
(7) HS2 FE + 2018 base + WLS 0.594 -0.793 0.463 -0.076 1.147 0.320
(8) HS2 FE + 2018 base + PreGrowth + WLS 0.602 -0.765 0.459 -0.077 1.163 0.319
Note: Band width is the average of 2 x 1.96 x SE over h = 0 to 27. Mean beta is the average estimate over the same horizons. China to US uses HS8 products (USITC); the other legs use HS6 products (Comtrade).

Taiwan

Comtrade records Taiwan as “Other Asia, nes” (code S19). These flows are therefore mirror data, as reported by China and the US.

Band width and mean effect by model and trade leg, h = 0–27, Lists 1–3 vs List 4B (Taiwan).
China → US
China → Taiwan
Taiwan → US
Model Band width Mean beta Band width Mean beta Band width Mean beta
(0) List FE only 0.237 -0.504 0.405 0.096 0.456 0.040
(1) Product FE (TWFE) 0.240 -0.496 0.414 0.080 0.458 0.011
(2) HS2 FE 0.295 -0.608 0.590 -0.023 0.496 0.053
(3) HS2 FE + PreGrowth 0.296 -0.625 0.580 -0.015 0.489 0.032
(4) HS2 FE + WLS 0.487 -0.852 0.330 0.029 0.751 0.164
(5) HS2 FE + 2018 base 0.325 -0.608 0.694 -0.123 0.638 0.114
(6) HS2 FE + PreGrowth + WLS 0.487 -0.917 0.332 0.031 1.122 0.046
(7) HS2 FE + 2018 base + WLS 0.594 -0.793 0.395 0.078 1.898 0.563
(8) HS2 FE + 2018 base + PreGrowth + WLS 0.602 -0.765 0.393 0.078 2.288 0.641
Note: Band width is the average of 2 x 1.96 x SE over h = 0 to 27. Mean beta is the average estimate over the same horizons. China to US uses HS8 products (USITC); the other legs use HS6 products (Comtrade).

FDI Response

The FDI regressions are run only for countries that the trade legs identify as rerouting countries, meaning those whose exports to the US rise in tariffed products. So far that is Mexico. The unit is a 6-digit NAICS industry \(n\) in candidate country \(j\), observed yearly, with \(T = 2018\) as the tariff year.

Symbol Meaning
\(\text{FDI}^{CHN\to j}_{n,t}\) Completed Chinese acquisitions into industry \(n\) of country \(j\) in year \(t\): number of deals, or deal value in USD millions
\(\text{Treated}_n\) 1 if NAICS \(n\) contains at least one HS6 product on Lists 1–3, 0 otherwise
\(s(n)\) 3-digit NAICS sector of industry \(n\)
\(\overline{\text{FDI}}_{n,\text{pre}}\) Average yearly FDI into industry \(n\) over 2013–2017

FDI model, horizon \(h\): \[\text{FDI}^{CHN\to j}_{n,\,T+h} - \text{FDI}^{CHN\to j}_{n,\,\text{base}} = \beta_h\,\text{Treated}_n + \gamma_{j\,s(n)} + \varepsilon_{n,h}\]

The baseline is either the 2017 level (\(h = -1\)), which leaves the earlier years as leads, or the 2013–2017 average. The outcome is a level change, not a log change, because most industry-years have no deals. Differencing already removes each industry’s own level, so a NAICS fixed effect would absorb the outcome; the sector fixed effect \(\gamma_{j\,s(n)}\) compares industries within the same 3-digit sector. Standard errors are clustered by sector. \(\beta_h\) should be near zero before 2018 and turn positive as exports from \(j\) to the US take off.

6-digit NAICS goods industries by treatment status, and Chinese acquisitions into Mexico.
Status Industries Deals 2013–2017 Deals 2018–2024
List 4A only 52 0 0
No tariffed HS6 4 0 0
Treated (Lists 1-3) 481 0 2
Note: Goods industries are NAICS 11, 21 and 31–33. Of 1,138 Chinese goods acquisitions abroad in 2013–2024, 894 are coded at a 6-digit NAICS that links to HS6; the rest are coded at 4 or 5 digits. List 4A only industries count as controls unless FDI_DROP_4A is TRUE.
NAICS codes in the M&A data and their link to HS6, Chinese acquisitions abroad, 2013–2024.
NAICS codes Deals
NAICS titles in the M&A data 534 2,582
    matched to a NAICS 2007 code 510 2,481
        goods industries (NAICS 11, 21, 31-33) 274 1,138
            mapped to at least one HS6 237 1,052
                of which 6-digit codes 196 894
            not mapped to any HS6 37 86
Note: Titles are matched to NAICS 2007 codes by name. A code is mapped to HS6 through the concordance package (hs5_naics) at its own level, so a 4-digit code covers every HS6 under that 4-digit industry. The FDI regressions use the 6-digit codes only.
HS6 products (HS 2017) covered by the M&A NAICS codes, by tariff status.
HS6 status All HS6 Covered by any mapped code Covered by 6-digit codes
Tariffed, Lists 1-3 (2018) 3,703 2,605 (70%) 2,127 (57%)
Tariffed, List 4A (Sep 2019) 1,301 862 (66%) 755 (58%)
Not tariffed: List 4B only 115 72 (63%) 59 (51%)
Not tariffed: never listed 267 201 (75%) 199 (75%)
All HS6 5,386 3,740 (69%) 3,140 (58%)
Note: An HS6 is covered if at least one goods NAICS code in the M&A data contains it. Percentages are shares of all HS6 with that status. An HS6 takes the status of its earliest implemented HS8 line.
Goods industries by tariff list, at each NAICS level: no industry is List 4B only.
Industry status
Industries with a List 4B product
NAICS level Industries Treated (Lists 1-3) List 4A only List 4B only Never listed Total Also has Lists 1-3 Also has List 4A only
6-digit 537 481 (90%) 52 0 4 77 60 17
5-digit 220 209 (95%) 11 0 0 47 42 5
4-digit 103 101 (98%) 2 0 0 35 34 1
3-digit 27 27 (100%) 0 0 0 21 21 0
Note: An industry is treated if at least one of its HS6 products is on Lists 1-3; otherwise List 4A if any is on List 4A; otherwise List 4B if any is on List 4B only. Only 115 HS6 products are List 4B only, and every industry that contains one also contains a product on Lists 1-3 or List 4A. The List 4B control used in the trade models therefore has no industry counterpart. Industries are NAICS 11, 21 and 31-33 in the Pierce-Schott concordance (hs5_naics).

FDI response of tariffed industries, average over 2018–2024.
Mexico
Mexico, Vietnam, Taiwan (comparison)
Model Years estimated Mean beta Mean SE Years estimated Mean beta Mean SE
(A) Deals, 2017 base 2 0.000 0.000 7 0.008 0.008
(B) Deals, 2013-17 avg base 2 0.000 0.000 7 0.001 0.002
(C) Value, 2017 base 2 0.000 0.000 7 -0.025 0.026
(D) Value, 2013-17 avg base 2 0.000 0.000 7 -0.013 0.013
(E) Deals, 2017 base, no sector FE 2 0.002 0.002 7 0.004 0.006
Note: Years estimated counts the years 2018–2024 in which at least one industry’s FDI changed; in the other years there is nothing to estimate. Deals with no reported value add zero to the value outcome.