Last updated 7 November 2026 · every figure recomputed from the sources listed at the foot of this page
Steel Antidumping Duty: 164 Cases and the 7.1× Gap Between the Two Rates They Publish
If you are searching for the anti dumping duty on steel, the honest answer is that there is no single figure — and the reason is not that the rules are complicated. It is that the same case file publishes two different numbers — a dumping margin and a cash-deposit rate — and in the 164 steel cases in this corpus the two differ by a factor of 7.1 at the median. This page prints both, says which one CBP actually collects at the port, and shows why a figure quoted from the wrong column is off by more than a rounding error.
What the 164 steel cases actually are
Everything below is computed from 920 Federal Register order-lifecycle records attached to 164 case numbers whose product scope names steel. 160 of those cases are active and 4 are revoked. The records split 637 antidumping, 201 countervailing-duty and 82 combined AD/CVD documents, which is the first thing to understand about “the steel duty”: for a great many steel products two duties stack, and a page that quotes one of them is describing half the bill.
The family is also wide. The cases name 132 distinct six-digit candidates, spread across 92 six-digit subheadings, and 163 of the 164 cases carry an explicit HTS linkage covering 328 eight-digit lines. Steel is not one line in the tariff schedule that you can look up once; it is a material that appears in a hundred product descriptions, and each description has its own order.
The two numbers that are not the same number
Of the 920 records, only 289 state a single unambiguous percentage (31.4 percent). Those 289 are the sample this page is about, and they are not one population: they divide by which quantity the document is stating.
This table is wider than the screen: scroll it sideways to reach every column.
| Rate caliber | Records | Median | Mean | Minimum | Maximum |
|---|---|---|---|---|---|
| Cash-deposit rate | 184 | 10.77% | 20.33% | 0.00% | 206.00% |
| Dumping margin | 105 | 76.45% | 79.70% | 0.00% | 237.91% |
Source: Federal Register API (federalregister.gov), AD/CVD order-lifecycle corpus built by scripts/build_adcvd.py · checked .
Read the two rows as two different questions. A cash-deposit rate is the percentage CBP applies to the entered value at the time of entry, collected as a deposit against the final liability. A dumping margin is the finding: the amount by which the exporter's US price fell below normal value. They are related but they are not the same quantity, and the table shows how far apart they sit in practice — a median of 10.77% against 76.45%, a ratio of 7.1×.
The dispersion matters as much as the medians. Cash-deposit rates run from 0.0% to 206.0%, margins from 0.0% to 237.91%. A zero on either column is a real finding, not a missing value: it means the determination produced no dumping or no deposit above zero for the exporter it covers. This is why a single “steel AD rate” quoted without naming the exporter is not a rate at all — the same order can carry a zero for one company and a triple-digit figure for another.
9 document types, and the ones that state no rate
The corpus carries 9 kinds of document. Counting them is the fastest way to see why 631 of 920 records print no percentage: most document types in an AD/CVD lifecycle do not state a rate. Continuations extend an order, revocations end it, and combined orders announce that two duties now apply — none of those is a rate document.
This table is wider than the screen: scroll it sideways to reach every column.
| Document type | Records | Cases | Rated records | Median of rated |
|---|---|---|---|---|
| Administrative review results | 656 | 99 | 233 | 11.04% |
| Sunset review results | 82 | 70 | 32 | 112.44% |
| Antidumping order | 22 | 15 | 7 | 66.70% |
| Amended final determination | 43 | 27 | 16 | 12.75% |
| Countervailing duty order | 25 | 20 | 0 | — |
| Combined AD/CVD order | 30 | 22 | 0 | — |
| Continuation | 52 | 45 | 0 | — |
| Revocation | 7 | 7 | 0 | — |
| Suspension agreement | 3 | 2 | 1 | 237.91% |
Source: Federal Register API (federalregister.gov), AD/CVD order-lifecycle corpus built by scripts/build_adcvd.py · checked .
The last column is the one to read carefully. Sunset review results carry a median of 112.44% across 27 cases; administrative review results carry 11.04% across 65. Taken at face value that is a ten-fold difference, and it is the single most common way to misread this dataset. The next section shows why it is an artefact.
The ten-fold gap that is not a ten-fold gap
Two review types, one product family, apparently very different rates. Before concluding that sunset reviews are harsher, check whether the two groups are describing the same cases. They are mostly not.
This table is wider than the screen: scroll it sideways to reach every column.
| Case group | Cases | Median latest rate |
|---|---|---|
| Carries a sunset-review rate | 27 | 112.44% |
| Carries an administrative-review rate | 65 | 11.04% |
| Carries only a sunset-review rate | 17 | 110.25% |
| Carries only an administrative-review rate | 55 | 7.10% |
| Carries both (the paired sample) | 10 | paired ratio median 1.108x |
Source: Federal Register API (federalregister.gov), AD/CVD order-lifecycle corpus built by scripts/build_adcvd.py · checked .
The first two rows reproduce the headline comparison. The next two rows split the cases by which review type they actually carry, and the picture inverts: cases that carry only a sunset-review rate have a median latest rate of 110.25% (17 cases), while cases that carry only an administrative-review rate sit at 7.1% (55 cases). The difference was never in what the review type does. It is in which cases end up in each bucket: the high-margin orders are the ones that reach a sunset review, because a sunset review exists to ask whether an order is still needed, and an order protecting a triple-digit margin is the one that gets litigated.
The paired sample settles it. Only 10 cases carry both review types, and on those 10 the ratio of the sunset figure to the administrative figure has a median of 1.108×, a minimum of 1.0× and a maximum of 42.886×. Half of them sit at exactly 1.0× — the two documents state the identical number, which is what happens when a sunset review restates the existing margin rather than recomputing it. With a paired sample of 10, no population claim is available and this page makes none; what the sample does establish is that the ten-fold headline is a composition effect, and that a reader who quotes it is quoting a property of the case mix rather than a property of the law.
Where the rates actually concentrate
Banding the 289 rated records by caliber shows the shape the medians hide. The two columns are not two versions of one distribution; they are two different distributions.
This table is wider than the screen: scroll it sideways to reach every column.
| Rate band | Cash-deposit records | Margin records |
|---|---|---|
| 0% | 2 | 1 |
| 0.01-9.99% | 89 | 27 |
| 10-24.99% | 66 | 10 |
| 25-49.99% | 14 | 7 |
| 50-99.99% | 2 | 15 |
| 100% and above | 11 | 45 |
Source: Federal Register API (federalregister.gov), AD/CVD order-lifecycle corpus built by scripts/build_adcvd.py · checked .
The cash-deposit column is bottom-heavy and the margin column is not. That asymmetry is the practical consequence of the distinction above, and it has a direct effect on how an importer should plan: the deposit is a cash-flow item that arrives on every entry, while the margin is the liability the deposit is provisional against. Budgeting on the margin overstates the cash you will pay this quarter; budgeting on the deposit understates the liability you may be assessed at liquidation.
Who the cases run against
The family is not one corridor. 23 origin countries appear in the rated records, and the ranking is worth seeing before assuming that a steel order means China.
This table is wider than the screen: scroll it sideways to reach every column.
| Origin | Rated records | Median of rated |
|---|---|---|
| China | 62 | 118.04% |
| Korea | 46 | 6.05% |
| India | 28 | 7.00% |
| Mexico | 25 | 20.11% |
| Thailand | 22 | 15.67% |
| Taiwan | 19 | 11.04% |
| Italy | 11 | 47.87% |
| Turkey | 11 | 14.74% |
| Japan | 9 | 45.42% |
| Spain | 9 | 18.81% |
| United Arab Emirates | 9 | 4.30% |
| Malaysia | 8 | 2.66% |
Source: Federal Register API (federalregister.gov), AD/CVD order-lifecycle corpus built by scripts/build_adcvd.py · checked .
Two cautions on that table. First, a case can name several countries, so the column counts records rather than cases and the numbers do not sum to 289. Second, a country's median is a median over the records that name it, across all document types — it is not the rate that applies to that country's steel today, and it is not the rate that applies to a specific exporter.
How to find the rate that applies to your own line
Four steps, in order, because doing them out of order is how people end up quoting the wrong column.
One. Classify the article to an HTS line, not to a material. “Steel” is not a classification; 328 eight-digit lines appear in this family alone, and the order that covers your line is the one whose product scope names your product.
Two. Find the case number. Search the product scope text rather than the tariff code — scope language is what decides whether your article is inside an order, and it is written to describe products, not codes. A line can appear in an order's HTS list and still be outside its scope if the scope language excludes it.
Three. Read the most recent document for that case and record which quantity it states. If it prints a cash-deposit rate, that is what you will pay at entry. If it prints a margin, you are looking at a finding and still need the deposit instruction.
Four. Check whether a second duty applies. 82 of the 920 records in this family are combined AD/CVD documents, and a countervailing duty on the same product is a separate percentage on top.
What this page cannot tell you
Three limits, stated plainly. The corpus is a snapshot dated 2026-10-29; determinations published after that date are not in it. The 631 records that state no percentage are not failures of the build — they are documents that genuinely print no rate, and treating them as zero would be wrong. And a median is not a rate: nothing on this page is a statement about what a specific shipment of a specific product from a specific exporter will pay. That figure requires the classification, the case, the exporter and the entry date.
Frequently asked questions
- What is the anti dumping duty on steel from China?
- There is no single rate. The steel cases in this corpus carry 637 antidumping records, and 164 cases name a product scope containing steel across 23 origin countries. A rate attaches to a case and an exporter, not to a country and a material. The cash-deposit median across all rated steel records is 10.77%, but individual records run from 0.0% to 206.0%.
- Why are there two different rates for the same steel product?
- Because they answer different questions. The dumping margin is the finding of how far the US price fell below normal value; the cash-deposit rate is the percentage collected at entry against the final liability. Across the rated steel records the margin median is 76.45% and the cash-deposit median is 10.77%, a ratio of 7.1×.
- Is a sunset review rate higher than an administrative review rate?
- Not per case. The headline medians differ ten-fold (112.44% against 11.04%), but that is a composition effect: cases carrying only a sunset rate have a median of 110.25% while cases carrying only an administrative rate sit at 7.1%. On the 10 cases that carry both, the paired ratio's median is 1.108×.
- What is the cash-deposit rate for steel?
- The median across the 184 steel records that state a cash-deposit rate is 10.77%, with a mean of 20.33% and a range from 0.0% to 206.0%. That is a distribution, not a rate you can apply to a shipment; the applicable figure depends on the case and the exporter.
- Does a countervailing duty apply on top of the antidumping duty on steel?
- Sometimes, and it is common in this family: 82 of the 920 records are combined AD/CVD documents. When both apply, the countervailing duty is a separate percentage on the same entered value, so the combined burden is the sum of the two plus the base tariff rate.
- How many US antidumping cases cover steel?
- 164 case numbers in this corpus have a product scope naming steel, of which 160 are active and 4 revoked. They generate 920 order-lifecycle records, 289 of which state a single unambiguous percentage.
Sources
- Federal Register API (federalregister.gov), AD/CVD order-lifecycle corpus built by scripts/build_adcvd.py — data/adcvd-orders.json -- 2,756 records, one per Federal Register order-lifecycle document x attributable case number (2,972 documents kept of 4,615 listed) (read 2026-10-29T08:46:20Z, sha256 7fc488813c2a2e49…)
Every figure above is recomputed from the datasets listed here at render time; the dataset vintage and each source's check timestamp are printed with it. This page extends the prototype that already earns search traffic on the same mechanism. Rates are informational, not a binding customs determination: confirm the classification and the entry with CBP before shipping.