US Reciprocal Tariffs Hit About 100 Countries From Aug 1; China Window Expires Aug 12

The Analysis

Strictly speaking, this report contains two effective dates and only one of them is still moving. The first is 1 August 2026, the day Washington's reciprocal tariff schedule took effect across roughly 100 trading partners. That day has passed. You cannot plan around it, you can only plan inside it. The second is 12 August 2026, when China's lower reciprocal rate is due to expire. That one may not have passed yet, and it is the only fact in this report that your own behaviour can still change. Most of what follows is about keeping those two apart, and about a third thing mentioned alongside them, an appellate court review of the statutory authority underneath the whole programme, which is being read around the trade as though it were a refund scheme. It is not one, and reading it as one is probably the most expensive mistake available to an importer this month.

The sentence that will be quoted most often and understood least is the one about 1993. Reuters describes this as the largest US import tax increase relative to GDP since 1993. That is a claim about a ratio, and the ratio has three moving parts: the level of the duty, the breadth of imports it reaches, and the size of the denominator underneath it. Roughly 100 trading partners tells you most of what matters about the second part. This is a measure of extraordinary reach, and reach is not the same thing as height. For any one country's goods the height is a separate question, and it has to be answered line by line rather than by looking at an average. When somebody drops the 1993 comparison into a pricing meeting, the honest follow-up is not how large but measured how. I have not seen the calculation behind that sentence and I am not going to reconstruct one from memory.

Where does that leave the rest of this column. I am working from the Reuters summary and nothing else. I have not read the schedule itself, I have not read the instrument that put it into effect, and I have not read one page of the appellate briefing. So there will be no clause numbers here, no annex references, no citations and no holdings. Every figure that follows is either traceable to that summary, derived arithmetically from it, or labelled explicitly as my own assumption. Where the answer turns on text I have not seen, I will say so and point you at what you need to go and read. That is the method, and I would rather hand you something narrow and reliable than something broad and invented.

Let me deal with the effective date first, because it is where the most actionable mistake lives. When a tariff takes effect on a date, the instinct is to count backwards from that date: did the vessel sail before it, did it arrive before it, was the commercial invoice dated before it, was the purchase order signed before it. In most customs systems only one fact in that sequence is the operative one, and it is the entry made for consumption. Everything earlier in the chain is evidence about commercial intent and about who should bear the cost between buyer and seller; it is not the determination of what is owed to the state.

I am deliberately not going to tell you how entries are made or what a filing looks like, because that is operational work done daily by people whose profession it is. What I am telling you is which fact the law picks out of that sequence, because until you know it, the rest of your planning is aimed at the wrong target.

There is a second version of the same question for goods sitting in a bonded warehouse, and here I have to stop and say that I do not know the answer for this instrument. Whether the operative event for warehoused goods is the arrival or the withdrawal depends entirely on how the instrument is drafted. Some say one, some say the other, some are silent and leave it to agency interpretation. I have seen teams congratulate themselves for beating a deadline by three days and then pull the goods out of the warehouse three weeks later into the new rate, because nobody read the clause. If you have cargo in bond right now, the instruction is not assume. It is find out. Ask whoever handles your entries to point at the sentence that answers it, and read that sentence yourself rather than accepting a conclusion. Confident verbal answers from people who have not read the text either are the most expensive free things in this business.

Before you can apply any rate at all, you need three answers, and none of them can be extracted from a news summary. Does this programme reach goods of the origin you actually declare. Does it reach your product, taking into account whatever product exclusions or exceptions the instrument carries. And how does it sit alongside the duties already assessed on the same goods. The third one is the one people assume, and here is the exception worth circling: whether a new duty stacks on top of existing ones, replaces them, or applies only to part of a calculation, is decided by the text and not by logic. I have watched intelligent people get this wrong in both directions, once by assuming stacking when there was none and once by assuming displacement when there was none.

There is also an origin point buried in the word reciprocal that deserves a sentence. The country printed on your invoice is not necessarily the country the tariff law cares about. Origin for duty purposes is a legal determination with its own rules about transformation and assembly, and goods shipped from one place often carry the origin of another. If your supply chain touches more than one country before it reaches a US port, do not let the shipping paperwork answer a question the law answers differently. This is precisely the kind of point that stays invisible until an audit, and then it is not invisible at all.

Mexico is the part of this report I find most interesting precisely because nothing happened to it. Its 25 per cent rate was held unchanged for a further 90 days of negotiation. Held unchanged means there is no new legal obligation today and no new relief today either. Anyone budgeting on a Mexico reduction has nothing in law to stand on at this moment. The date that matters for Mexico is not 1 August; it is the day those 90 days run out. And here I have to say something unhelpful: I do not know when that is.

The word further in the summary is doing real work. Ninety days from 1 August is one calendar date. Ninety days from the instrument, from the letters sent to trading partners, or from some earlier understanding produces others. I am not going to guess between them, and you should not let anyone else do it casually. Establish the start date in writing, from someone who has read the source, count forward, and put a flag two weeks either side of the result. Treat it as contested until it is confirmed.

Note also what a freeze does to planning. A standstill is not a direction of travel. It tells you the answer today and nothing at all about the answer after the period. Companies routinely mistake a pause for a signal and then discover, a quarter late, that their planning assumed a trend nobody ever announced. 'Held unchanged' is a statement about today. That is all it is.

Now the date you can still do something about. China's lower reciprocal rate expires on 12 August according to this report, and there are two structurally different things that can happen at an expiry. One is that the higher scheduled rate applies automatically the following day with no further announcement. The other is that any continuation, extension or change arrives through a further instrument, so there is a document to wait for and a date on which someone publishes it. The gap between those two worlds is large. In the first, silence is itself the answer and you carry the entire risk of guessing wrong. In the second, silence is merely silence. I do not know which applies here. The summary does not say, and this is exactly the kind of point around which people improvise a confident view and then move money on it. Do not.

While we are on China, get your own dates straight, because they are three different facts and conflating them is routine. The sailing date, the arrival date and the entry date describe three separate events that can fall in three different weeks. If the change bites on entry, then a container that arrives on 10 August but is not entered until 14 August sits on the far side of the line, and one entered on 11 August sits on the near side even though it left the same factory in the same month. Write all three down, on one sheet, for every shipment in flight between now and mid-August, alongside the name of the person accountable for that row. It is an afternoon's work, and I will argue below that it is worth more than another week of reading commentary.

That brings me to the appellate review of IEEPA authority, which is where I would urge the most caution, because it is where the most confident predictions are being made by people who have not read the briefing. Let me separate three questions that are constantly collapsed into one. Does the statute authorise this exercise of power. That is the question the court is deciding. What happens to entries already made if the answer is no. That is a separate question, decided later, by different people, through a different process. And what happens to your entries specifically. That is a third question, and it is the only one most people reading the news actually care about.

Start with what an appellate decision binds. A court decides the case in front of it. It resolves the legal question, and it grants relief to the parties before it. Whether its reasoning reaches anybody else depends on the scope of the order and on how the executive branch and the administering agency then choose to act. So the first trap is assuming a ruling travels further than it does. Almost every importer reading about this case is not a party to it.

The second trap is timing. A decision adverse to the government is normally followed almost immediately by an application to stay it pending further appeal, and stays are frequently granted while the case goes further. During a stay, entries continue to be assessed essentially as they were. So the realistic path after a loss is not that duties stop next week. It is that duties continue for months, during which nothing at your dock changes at all, and that any remedial process begins long after the headlines have moved on. Anyone building cash flow around a near-term refund is betting on the least likely branch, and simultaneously failing to prepare for the most likely one.

Now the part with the most practical content, and it concerns your own entries rather than the court. Entries become final. There is a defined process and there are defined windows, and once an entry is final and has not been challenged in the manner the law requires, the assessment generally stands even where the legal basis for it is later called into question in litigation between other parties. I am deliberately not reciting that process or its deadlines here, because the detail varies by context and because I have not verified how this particular programme interacts with it. The structural point is what I want you to take away: whatever remedy eventually exists is a function of steps taken while an entry is still open, not arguments made after it has closed. If you want to preserve optionality, the work is this month, on entries being made now, not some unspecified later date when the case is decided.

Here is the applicability point I would write on a wall. The authority named for this schedule is IEEPA. A ruling about IEEPA resolves questions about measures taken under IEEPA. It does not by itself resolve anything about tariff measures resting on other statutory authorities, and US trade law contains several. I am not going to enumerate them, because I have not verified which are live for the products actually in front of you, and an unsourced list would be worse than no list at all. But do not let anyone tell you that a loss in this case dismantles the apparatus. It does not. Establish the authority behind each measure you genuinely face, because they are not interchangeable, and because a remedy in one does not travel to another.

So carry two cases and refuse to collapse them. In case one, the authority is sustained and these rates become an ordinary feature of your cost structure for years, in which case the useful work is contract drafting and pricing rather than litigation watching. In case two, it is struck down eventually, with some scope and some remedy attached, in which case the operative question is whether your entries qualify, and qualification will be decided by records built today. These two cases need largely the same preparation, which is convenient: documenting the basis on which duty was assessed and the dates on which entries were made serves you in both worlds. It is cheap insurance, and August 2026 is the moment it is cheapest.

Let me do some arithmetic, with every assumption visible so that you can replace each one with your own figure. Take Mexico first, because 25 per cent is a number that is actually in the report. Assume your monthly entered value of Mexico-origin goods is M dollars. Duty at the current rate is 0.25 times M every month, stable for the standstill period and no further. What you need for planning is M multiplied by the difference between 25 per cent and whatever follows. Simplify it: each one percentage point of change costs you M divided by 100 per month. Assume M is USD 1,200,000 a month, which is my assumption and not the report's, and one point of movement is USD 12,000 a month, roughly USD 144,000 on an annualised basis. The number I want you to notice is not the total. It is that the sensitivity per point is knowable right now, before anybody knows the destination rate.

Notice what that calculation also establishes. Establishing the start date of those 90 days has a cash value, even though it produces no cash on its own. Until you know it, you cannot tell which month's forecast absorbs the change, and nobody budgets well for a change that lands in the wrong quarter. Ninety days either side of a disputed date is the difference between a manageable repricing and an unplanned one, and it costs you one email to resolve.

Second calculation, this one about China, and the point of it is sensitivity rather than a total. Assume a shipment with entered value C. Assume the difference between the lower reciprocal rate now in force and whatever applies after 12 August is D percentage points. Assume you cannot move sourcing or renegotiate the customer price inside the window, so the whole delta lands somewhere in your accounts. The exposure is C times D. Put in my assumed figures: assume C is USD 500,000 and assume D is 20 points. Both are mine, neither comes from the report, and you must replace them with your own schedule. That is USD 100,000 on one shipment. The property of that number worth thinking about is that it does not depend on how clever anybody is in November. It depends entirely on which side of 12 August the entry falls, which is why building that three-date sheet this week outperforms another hour of reading about the case.

One more use of the 1993 sentence, and this is the only forward-looking observation I will allow myself. A measure described as the largest increase relative to GDP since 1993 is by definition fiscally large. Measures that are fiscally large have a habit of being built into the baseline against which subsequent budgets are drawn up, and things inside a baseline are much harder to remove than things announced on top of it. I am not predicting that, and I would distrust anyone who does. I am observing that the structural incentives around a measure of this size point towards duration, and that your contracts and your pricing should therefore be built for a multi-year horizon while staying ready for the alternative. Holding two cases is not indecision. It is what the situation requires.

What I would actually do this week, sitting in your chair, comes down to a few moves and none of them are clever. Name one person responsible for watching official publications and any agency guidance, with a rule that anything published is reviewed within five business days and circulated to whoever sets your prices. Obtain a written determination, from whoever handles your entries, of exactly which rate applies to each of your principal origins effective after 12 August, together with the sentence in the source document relied on for each. Build the three-date sheet for everything afloat and keep it live until every row has cleared.

Review every customer contract, quotation and price list running past the end of the year for a duty-change provision, because if you sold landed prices without one you own the delta by default, and if you sold them with a badly drafted one you may own part of it anyway. And start recording, entry by entry, the basis on which duty was assessed this month, because that record is what any later remedy runs on.

The traps are almost all versions of over-reading. Held unchanged is not progress. An expiry defaults to expiring, so plan on expiry and treat continuation as upside rather than a base case. An appellate ruling is not a refund programme and certainly not one you can borrow against. A decision about one statutory authority does not decide another. Dates that feel interchangeable in conversation, meaning sailed, arrived and entered, are not interchangeable where money changes hands. And the most expensive trap of all is assuming this whole thing is temporary, placing money on that assumption through sourcing commitments, customer pricing or inventory timing, while simultaneously failing to do the modest documentation that would preserve your position in the world where temporary turns out to be right but the remedy turns out to be narrow.

I have spent much of this column refusing to tell you things, and that is deliberate. Between now and 12 August there will be no shortage of confident readings of what happens next, most of them built on exactly the same summary you just read. My value here is narrower and considerably less satisfying: to say which facts in this story are established, which are assumed, and which are simply unknown. The date that will most affect your business this month is probably not 1 August, which has already passed, and maybe not even 12 August, about which nobody outside the process knows anything with certainty. It is the day each of your own entries becomes final. Find out when that is, write it down, and set a reminder that fires before it does.

↑ Back to the key points

— By Dr. Ingrid Voss