CBP Adds 100 HTS Codes to Canada's 50% Section 338 Duty, Drops 22, From 30 October

The Analysis

CBP sent CSMS message number 70314289 to importers, brokers and filers on 29 October, and the punchline landed at 12:01 a.m. eastern time on 30 October. That is roughly eighteen hours of working notice for something that changes the duty rate on real cargo. The Section 338 rate itself does not move: still 50 per cent ad valorem on goods of Canada. What moves is the population of goods carrying it.

Under heading 9903.03.14 of the Harmonized Tariff Schedule of the United States, twenty-two classifications come off the list, 3923.21.00 and 8433.60.00 among them, and one hundred new ones go on, marked in bold on the annex attached to the message. Headings 9903.03.12 and 9903.03.13 are untouched. The Section 338 duty stacks on top of whatever Section 232 already applies, so anybody doing mental arithmetic on 50 per cent alone is understating the bill.

My inbox filled up yesterday with the same question: are we in or are we out. It is the wrong place to start. The notice names two triggering events, goods entered for consumption and goods withdrawn from warehouse for consumption, and those two events sit at very different points in a shipment's life. If you habitually think of duty as decided at the arrival port, this message quietly says otherwise.

Start with the paperwork, because that is where every customs decision finally has to live. Release comes off the CBP Form 3461; the money lines live on the entry summary, CBP Form 7501. A Chapter 99 heading like 9903.03.14 does not replace your ordinary eight-digit classification. The cargo is reported under its regular HTSUS number, and then the same entered value is reported a second time under the Chapter 99 heading that carries the additional duty. One shipment, two lines on the summary, one rate on each, and the total is the sum.

Put that in plainer language: somebody has to type a second line, or stop typing one. Broker software keeps a cross-reference table between ordinary tariff numbers and Chapter 99 provisions, and a human being has to load CBP's new annex into that table. A hundred additions and twenty-two deletions mean that table needs edits, testing and a release into production. When the load is late or wrong, the failure is silent at release. Nothing stops the container, nobody telephones the importer, the yard gate opens exactly as it did last week. The mistake surfaces later, either as a demand from CBP or as money quietly donated on classifications that had already been taken off.

Moving both edges of a list at once is nastier than moving one edge. When everything runs in a single direction, "probably unchanged" is a reasonable default. With additions and deletions landing together there is no safe default, and the population that genuinely needs re-confirmation is everybody whose goods sit anywhere near the boundary, not merely the hundred and twenty-two numbers that were named.

Kill one hopeful reading before it costs somebody money. People see a code come off the list and assume last week's payment comes back with it. Read the effective language again. The new scope applies to goods entered for consumption or withdrawn from warehouse for consumption at or after 12:01 a.m. eastern time on 30 October. Entries filed on 29 October were assessed against the rules that existed on 29 October, and a change of scope does not reopen them the following morning. Duty attaches at entry, and nothing in that message says old entries get relitigated.

Whether some other instrument could deliver relief is a question for somebody who reads statutes for a living, and here I mean Dr. Ingrid Voss, who would take that sentence apart clause by clause and name the instrument that could move it. My lane stops much earlier: brokers will apply the old list to old entries and the new list to new ones, and anything beyond that has to be fought for deliberately rather than assumed.

The part that did not make the headline is the part I sat down to write about. Every write-up leads with one hundred in and twenty-two out. Few mention the second trigger, and almost none follow it to its conclusion: goods withdrawn from warehouse for consumption.

A bonded warehouse is duty-deferred storage. You land the goods, pay nothing that day, and duty becomes due the moment they come out for consumption. That withdrawal date, not the arrival date, is the one the notice points at. A container entered into bonded storage in Buffalo, Detroit, Champlain or Seattle back in July, under a classification nobody had reason to worry about, has its answer decided by the morning somebody decides to pull it out. Every carton of Canadian origin sitting in bonded storage got re-priced at one minute past midnight on 30 October.

There are therefore two inventories holding two different answers. Cargo waiting in a container yard for consumption entry is governed by the 30 October cut-off in the ordinary way. Cargo inside a bonded warehouse is governed by the withdrawal instruction, and that instruction is written by whoever runs the warehouse, often against release paperwork cut weeks earlier. This is exactly where notices like this leak. The broker read the message at his desk. The person who actually has to know is standing in an aisle with a handheld terminal.

Foreign-trade zones deserve their own sentence. Unless privileged status is elected at admission, merchandise in most zones is judged when it transfers into the customs territory rather than when it enters the zone, which puts the withdrawal date back in charge. If you have been treating zone admission as locking in yesterday's rate, go and read your operator's status election before you rely on it.

Practically speaking, the pre-midnight escape hatch was already shut. Eighteen hours of notice will not pull meaningful volume out of a bonded warehouse: no truck capacity, no labour booked, no customs staffing at that scale. Nobody should build a rescue around beating the clock. Whatever you build has to work on goods that are already caught.

Sort affected houses into piles and the picture sharpens. The hundred additions take the extra 50 per cent from the first consumption entry after the cut, with no transition period and no grace. The twenty-two removals recover nothing on past entries but bring real relief going forward, and a price carried back to a customer this week is worth considerably more than one renegotiated a quarter from now. Everybody else, including anything sitting under 9903.03.12 or 9903.03.13, gets hurt by carelessness rather than by the rule: either a stale cross-reference table keeps writing Chapter 99 lines that no longer apply, or a re-determination of the underlying eight-digit code walks them into a listed classification without anybody noticing.

Who figures this out first depends less on the importer's size than on who writes the entry. Houses with their own filing team and their own table are usually corrected within a day. Houses relying on a third-party broker depend on that broker's data queue, and the honest ones will tell you how deep theirs is. The asymmetry that bites is that the shipper knows what is in the box and the broker knows which line to write, and neither can read the other's file unless somebody asks. Nobody asks. That is the entire failure mode, laid out in one sentence.

Nobody asks about the surety either, and after a jump of this size somebody should. A continuous customs bond is sized against the duties, taxes and fees you paid over the preceding twelve months, so the import bill is what pays for the guarantee, and the guarantee is what keeps entries flowing. Multiply one product line's duty by three and the number your surety used twelve months ago stops describing your business. Whether that triggers a review, a request for collateral or nothing at all depends on your specific bond and your specific surety, which is exactly why the question has to be asked of them rather than answered in an article. My team puts that call in the same week as any exposure step change, and we have more than once been glad we did. An entry that cannot be filed because the bond went thin is a much uglier problem than a duty you had budgeted for, because it stops every shipment, not just the affected line.

One further thing gets misread in boardrooms and truck stops alike: this duty attaches to goods of Canada, not to shipments that happen to come from Canada. A container trucked across a Canadian border carrying third-country goods is not a Canadian-origin shipment. Run it in reverse and the logic still holds: goods of Canada routed through somewhere else do not lose their origin. Origin follows where the goods were made or substantially transformed, and routing is evidence of nothing. Anyone who translated "duties on Canadian goods" into "duties on goods shipped from Canada" has already made a mistake worth 50 per cent of entered value.

Timing matters because compliance doors close in sequence and a few of them are wide. The rate changed one minute past midnight on 30 October. Broker systems follow within hours to a couple of days, and entries filed in that gap can bounce. The entry summary is normally due within ten working days of release, which is roughly when the cash actually leaves, so anyone modelling cash flow should use that date rather than the arrival date. Liquidation follows about 314 days after the entry date in the ordinary run, and straightforward corrections are comfortably available up to that point. After liquidation you argue through a different mechanism with a clock attached to it. The window for a cheap fix is long, the window for a free fix is longer, and both require somebody to look. Write the dates down.

None of that helps if the premises move. Three things could flip this analysis and one of them eventually will. Classification is the first: CBP can re-determine your eight-digit number and pull you into a listed code regardless of what you believe about your own product, and a marketing brochure has no standing against a laboratory report. The next revision is the second: this list has moved before and will move again, possibly within weeks, which means any one-off fix you engineer is wasted effort, so build a check you can run monthly rather than a rescue you run once. The 232 layer underneath is the third. Section 338 stacks on Section 232, so a change in that underlying layer moves your total exposure even if 338 freezes exactly where it is. Model them together or you will be surprised by arithmetic you already knew.

There is one more trap in this notice, and it lives in the formatting rather than in the tariff language. The annex marks additions in bold and lists removals as plain text, which is a fine convention on a printed page and a poor one on a mobile screen, where bold is frequently invisible. Work from the numeric list and never from the weight of the type. Cross-check every line against the eight-digit classification you actually declare, not against the product description in your catalogue. A single digit changed somewhere between your invoice narrative and your entry summary can move you from exempt to taxable, and the duty is the whole 50 per cent either way.

Worth understanding too is why this instrument keeps moving. A list-based mechanism lets the government change exposure without legislation, without comment periods and without a publication cycle the trade can plan around, which means the correct way to treat that annex is the way you treat a price index: something that gets refreshed on a schedule nobody owes you advance warning about, and something your business should re-read at fixed intervals rather than when a headline appears. Teams that built a monthly tariff-review habit during the earlier rounds of tariff expansion are barely noticing this one. Teams that treated each notice as a project are rebuilding the wheel every time, and paying service fees each time they do.

The duty layers deserve a sentence about how they combine. Section 338 sits on top of Section 232 rather than replacing it, and when every layer is ad valorem applied to the same entered value the order of the lines changes nothing about the total. It stops being indifferent when one layer is a specific duty, or when a later provision applies to a base that already carries an earlier one, the sequence genuinely changes the total. Tell your broker to confirm in writing which basis each additional Chapter 99 line uses before you accept a landed-cost calculation built on assumptions about stacking.

Then the numbers, because the abstract version always sounds survivable and the arithmetic usually is not. Assume entered value of USD 200,000 on a 40-foot container, and be clear that this is entered value rather than retail. Assume the regular tariff line is duty-free, which you must replace with your own actual rate. Assume the cargo already carries a Section 232 additional duty of 25 per cent ad valorem. Ignore merchandise processing and harbour maintenance fees, which are fractions of a percent with caps attached and do not move the story.

Before the change, with the code off the list, the bill is 200,000 times 25 per cent, so USD 50,000. After the change, with the code added under 9903.03.14, it is the same USD 50,000 plus 200,000 times 50 per cent, so USD 100,000 more, for a total of USD 150,000. The delta is USD 100,000 on a single box. Ten boxes a month is USD 1,000,000 a month and roughly USD 12,000,000 a year. Compare that with something honest: if your contribution margin on that line is 15 per cent of a USD 200,000 invoice, which is USD 30,000 a container, the new duty is about three and a third times every dollar of margin you had. No freight negotiation touches a gap like that. The available answers are re-price, re-source within the rules, or stop shipping the line, and pretending otherwise only moves the bill into next quarter.

Run the same arithmetic on bonded inventory, because the exposure sitting in a warehouse is the part nobody has counted. Assume forty containers resting in a bonded warehouse at USD 200,000 entered value each, so USD 8,000,000 in total. Assume half of them now fall inside newly added codes: twenty containers times USD 100,000, or USD 2,000,000 of liability that did not exist the day before. Assume your warehouse charges USD 25 per container per day all-in; holding those twenty containers another sixty days costs 20 times 25 times 60, which is USD 30,000. Thirty thousand against two million. Nobody should be having the rent conversation. Leave it parked, pay the storage, and spend the time on the classification instead.

Leaving it parked is not the only option worth pricing. A withdrawal for export rather than for consumption does not create the duty, which means finding a buyer outside the United States is sometimes the cheapest available exit. It is an awkward conversation to have with a sales team, but awkward beats expensive when seven figures are on the table. What does not work is any of the clever ideas. Transshipping through a third port changes nothing, because the duty keys off origin. Repacking or relabelling changes nothing either, because substantial transformation is the standard, and a false statement about origin on an entry is a categorically worse problem than an underpayment. If the arithmetic says the product cannot carry the duty, then say so out loud this week and let the sourcing team do their job.

Drawback deserves one honest paragraph rather than a confident one. Imported goods later exported can recover most of the duties paid. Whether duties imposed under these additional provisions are recoverable, and under which drawback procedure, is not something I would bet a client's money on without written confirmation. If drawback sits anywhere in your plan, make the broker put it in writing and cite the authority before you book the credit. An unverified refund assumption inside a proposal is how a bad month turns into a bad year.

There is a version of this where nothing bad happens, and it is boring. Somebody downloads the annex, compares it against the classifications your company actually declares, gets written confirmation from the filer, re-screens the bonded inventory line by line, and puts a monthly re-check on the calendar because this list will move again. Everything else follows from that being done properly.

↑ Back to the key points

— By Derek Xu