US 50% Section 338 Tariffs on Canadian Goods Take Effect Aug 19 With No USMCA Carve-Out

The Analysis

Three proclamations signed on 20 July put an additional 50 percent ad valorem duty on roughly 20 billion dollars of Canadian imports, effective 19 August. The scope is more than 550 tariff subheadings, concentrated in motor vehicles, dairy and alcoholic beverages. Energy, potash, fish and critical minerals are excluded. Goods that would otherwise qualify under USMCA receive no exemption.

That last sentence is the one to read twice. Strictly speaking, this is a Section 338 action, a separate statutory lane from the Section 301 and Section 232 measures most importers watch, and the reason the lane matters is that it does not inherit the assumptions built into the others. A preference programme and a duty action are different instruments, and eligibility under the first does not, by its own terms, disable the second.

I will be careful about what I assert. The proclamations are the instrument; where a detail is not in them in a form I can cite, I will say so rather than infer it from the coverage.

Begin with the shape of the package rather than the rate. Three documents were signed on one day, they add 50 percent ad valorem, they apply to about 20 billion dollars of imports, they take effect on 19 August, and they cover more than 550 subheadings in three clusters. Those are the only quantities in the action, and every one of them is a boundary rather than a burden. The burden is the 50 percent, and it is flat inside the boundary.

One word in the description does more work than it looks like it does. The duty is additional, which means it sits on top of whatever already applies rather than replacing a rate. That matters for the arithmetic and it matters for the argument. A line that was duty-free under the preference does not become 50 percent instead of something; it becomes 50 percent, while a line that already carried a rate carries that rate plus this one. The distinction is not academic when the exposure is being explained to a board, because the two versions of the number differ by the amount that was already there.

Section 338 is its own authority, not a variant of the two everyone watches. It has its own premise, its own record and its own procedure, and the compliance function in most companies does not monitor it with the attention given to the other two. That is precisely why an action under it arrives as a surprise. The practical consequence of a separate lane is that precedents do not transfer cleanly: an exclusion that exists under one authority is not an exclusion here by analogy, and a rate structure familiar from another is not the structure here.

The premise is stated in the proclamations, and I am not going to paraphrase a legal test I have not read in a citable form. What can be said without overreach is that this premise is not the same kind of finding as a labor-regime finding or a security finding, which means the record supporting it is a different record and the arguments that can be run against it are different arguments. Anyone importing the analysis from another tariff file is importing the wrong analysis.

The line that changes plans is the USMCA one. A group of importers built a sourcing book on the assumption that agreement-origin goods sit outside a duty action of this kind, and for these subheadings that assumption no longer holds. Preference and duty are separate instruments. A preference governs the rate you are entitled to when the rules are met; it does not, on its own terms, switch off another authority. The proclamations reach past the preference for the named list.

Which makes the absence of a carve-out a drafting decision rather than an oversight. A carve-out for agreement-origin goods is one of the easiest provisions to write, and drafting teams write them routinely when the intent is to preserve a preferential relationship. Its absence here is a statement about intent and should be read as one. Arguing that the agreement must impliedly protect the goods is arguing against the text rather than about it.

Twenty billion dollars is the number in the coverage and the least useful number in the file. It describes an aggregate, not an incidence. A duty that lands on 20 billion dollars of trade and a duty that lands on 550 subheadings are the same fact described two ways, and only the second description can be checked against anything you hold. You can test your lines against a list. You cannot test them against a total.

Twenty billion across more than 550 subheadings works out, on my arithmetic, to an average of roughly 36 million dollars a subheading. An average across a list like this is close to meaningless, because the distribution is not flat: a handful of lines will carry most of the value and a long tail will carry almost none. Which is why the subheading count and the dollar total have to be used for different jobs. The total sizes the political event. The count sizes the compliance job. Confusing the two is how a company arrives at the conclusion that it is too small to be affected.

Five hundred and fifty subheadings is a breadth statement about supply chains, not just about a product catalogue. Some of those lines will be inputs, and an input is where a duty is hardest to see, because it enters the cost of a product you price under a different classification and often a different supplier relationship. A component that becomes 50 percent more expensive does not announce itself. It shows up two quarters later as a margin that will not close.

The exclusions tell you more than the inclusions. Energy, potash, fish and critical minerals are spared. Those are the categories where dependence is hardest to substitute on a short clock and where a price increase transmits into domestic costs quickly and visibly. Sparing them is not generosity; it is a statement about which costs the action was designed not to create. By elimination, the cost is meant to land where a supply chain can absorb it rather than where a household bill can.

The four exclusions are also narrower than they sound. Whether a given line sits inside one of them is a question about the definition in the text, not about how the product is described in a catalogue or on a commercial invoice. A product that a sales team calls a fish product, or an input that a chemist would call critical, is not thereby inside the carve-out. This is the likeliest place for a good-faith error, and it is an error of reading rather than of classification.

Why 20 July to 19 August is worth dwelling on: thirty days between signature and effect is a short runway for a scope of this size and long enough that nobody can claim surprise. Compare it with a handover between two authorities, where the entire design problem is avoiding a gap. Here there is no handover and no gap to manage. There is a signature, a month, and then a duty on entries, and the month is calibrated to be inconvenient without being unfair.

Three documents rather than one is itself informative. A package assembled on a single day, covering three product clusters and a set of exclusions, is a coordinated instrument rather than a single notice, and coordination of that kind usually means the boundaries were drawn together, with the exclusions decided at the same table as the inclusions. That matters, because it makes the exclusion list part of the design rather than a technical annex, and designs are harder to attack than annexes.

Obligations change by position in the chain rather than by size. A large importer with an automotive book faces a line-level problem: the 550 subheadings have to be matched against a classification file that may run to thousands of lines, and the match has to be right at the subheading level, not at the chapter level where most exposure summaries live. The obligation is not to pay the duty. It is to know which of your lines is in scope, and to know it before the entry rather than after it.

A mid-sized or small importer with a single dairy or alcohol line faces the same percentage and a different problem. Fifty percent ad valorem is uniform inside the scope, so the rate does not reward size. What size changes is absorption: a distributor with one affected line and no substitute origin carries the whole increase on that line, while a larger book can spread it across categories where there is room. The obligation is identical, the capacity to meet it is not, and the proclamations are indifferent to the difference.

The three named clusters do not behave alike under the duty. Motor vehicles are a components story, where the exposure is often in a part several tiers from the assembly and the fix is an engineering decision running on a product-cycle clock. Dairy and alcoholic beverages are a finished-goods story with a slow substitution clock: origin is part of the specification, and a buyer who wants a Canadian product does not accept a substitute because the arithmetic changed. A duty on a component can be designed around. A duty on a differentiated finished good is carried or passed on.

Port of entry barely enters the analysis, and that is worth stating because it is usually the first question asked. The duty is ad valorem and attaches by entry date, so identical goods on the same day carry the same rate whether they cross at a land border or arrive at a seaport. What the port changes is the entry date itself, and with a fixed effective date the entry date is the variable that decides everything.

The effective date is 19 August and the scope is entries on or after it, which makes the entry date rather than the shipment date the operative fact. Goods shipped in early August and entered on 20 August are inside. Goods shipped in late July and entered on 18 August are outside. That distinction is routinely collapsed in coverage, and it is the distinction that decides the duty. It is also a legal fact about the scope of the measure rather than a logistics fact, which is why it belongs in the compliance file and not in the transport plan.

The window before the date is real but narrow. It is the gap between now and 19 August, and it is the only period in which anything moves from inside to outside. What can move is limited by what is already in transit and by whether the entry can be made before the date without distorting the transaction. A month is enough to re-sequence a shipment. It is not enough to re-source a product, and a plan that assumes otherwise is a plan about the next cycle rather than this one.

After the date, the question changes from incidence to consistency. Whether every line you believe is outside the scope actually is. The first place an error surfaces is usually not the duty calculation but the exclusion claim, because the four spared categories are broad words with narrow definitions, and the definition is the only thing that counts. That is a review you can run in advance and cannot run retroactively, which sets the lead time at roughly one entry cycle.

Run the arithmetic with assumptions stated. Assume a dairy or spirits line at a landed cost of 20 dollars a unit before duty, 100,000 units a year. Fifty percent ad valorem adds 10 dollars a unit, which is 1 million dollars a year. If that line had been treated as agreement-origin and duty-free, the whole million is new money rather than a substitution for an existing duty. The part that hurts is not the million. It is that 10 dollars against a 20 dollar cost base is a 50 percent increase in the cost of the goods, and nobody carries that on a single line without changing something else.

Take the component case, again with my assumptions. Assume an automotive part at 45 dollars a unit, 400,000 units a year, 18 million dollars of annual entered value. Fifty percent is 9 million a year. The interesting number is not 9 million. It is what happens to a product programme priced on the assumption that the part was preference-eligible, because there is no price on the finished good that visibly corresponds to the increase. A duty on a visible line gets argued about. A duty on an invisible one gets discovered.

What the coverage buries is the override itself. Summaries lead with the rate, the date and the twenty billion, and treat the agreement line as a footnote. For anyone whose sourcing book was built on the preference, it is the only sentence that matters, because it invalidates the assumption that governs everything else in the file. A footnote that changes the answer is not a footnote.

Reading the exclusion list as a targeting statement gives you more than the inclusions do. Sparing exactly four categories tells you where the action expects the cost to land and, by extension, which sectors it expects to absorb it. That is more useful than the aggregate value, because it says what the drafters thought was substitutable and what was not, and that judgment is the one that gets tested first if the list is ever revisited.

Breadth is the other thing the summaries flatten. Five hundred and fifty subheadings is a statement about how far into a bill of materials a duty can reach, not just how many products are covered. A duty that enters through a component is more consequential than one that enters through a finished good, because it does not appear in the landed cost of the product being priced until someone traces it, and by then the entry has been made and the assumption has hardened.

A proclamation is also amendable. The list can be added to, the exclusions can be narrowed, and neither requires a new statute or a long process. So the map built in the next two weeks is a snapshot rather than a boundary, and treating it as settled is the same error as treating any published list as closed. Instruments of this kind are administrative, and administrative instruments move without a headline.

There is a difference between a duty that gets negotiated away and one that gets withdrawn, and it is worth being precise about which this is. A negotiated outcome produces an agreement, and an agreement has a date and a text. A withdrawal is an administrative act, and it can happen quietly, partially, or not at all. Planning for a negotiated outcome when the instrument is administrative is how companies end up with three years of exposure they described to their board as temporary.

A challenge does not suspend a duty by existing. The proclamations are the operative text from 19 August, and unless something stays them, the obligation runs while the argument is heard. That is worth saying plainly, because the opposite is a common assumption in rooms where the legal question is still open: the sense that a filing has bought time. It has not. Time is bought by a stay, and a stay is a decision somebody has to make, on a timetable nobody in the importing business controls.

Where the conclusion flips. If the proclamation list is amended, subheadings added or removed, the map moves, and any re-sequencing done to beat the date may have been spent on goods that are no longer in scope or, worse, on goods that entered scope afterwards. The direction that gets less attention is the reverse: if Canada restricts its own exports in the four spared categories, the exclusions stop functioning as protection for the sectors that believed they were outside, because the constraint then arrives through supply rather than through duty.

The other flip is legal rather than political. If a challenge produces a stay, the relief is procedural and the date can be reinstated, so an importer who dismantles the exposure analysis on the strength of a stay will rebuild it at a worse moment. The condition that genuinely changes the answer is narrower: whether the four exclusions hold their meaning on their definitions, and whether the 550 list holds its breadth. Both are questions about text, and both are answered by reading the proclamations rather than the summaries of them.

There is no end date in what I can see, and its absence is a feature of the instrument rather than an omission. A temporary authority announces its own expiry. This kind of action does not, which means the planning horizon defaults to whatever you can defend internally rather than to a date printed on a document. That is a harder question than it sounds, because most cost models are built to run to a known end date. Strip the end date out and the model does not become wrong. It becomes unbounded, which is worse.

So the thing to hold onto is the difference between a preference and an immunity. A preference is a rate you are entitled to when the rules are met. It is not a wall around the transaction, and the moment a separate authority is used, the wall is not there. The twenty billion, the 550 subheadings and the 50 percent will be quoted at you for months. The sentence that should change the file is the one about the agreement, and it is the shortest sentence in the documents.

↑ Back to the key points

— By Dr. Ingrid Voss