Tuesday, 8 September 2026

Section 338 was written into the Smoot–Hawley Tariff Act of 1930 as a bilateral retaliation weapon for a world with no GATT and no WTO. Its logic is reciprocity: if a foreign country discriminates against American commerce in a way that burdens or disadvantages it, the President may impose new or additional duties of up to 50 per cent on that country's goods, and in the limit may exclude them entirely. The 50 per cent ceiling in the Canada proclamations is not a coincidence — it is the statutory maximum, which is why the number is what it is.

The provision has sat unused for the whole postwar period because the architecture that replaced it was better suited to the job. GATT and then the WTO gave the United States a dispute mechanism; Section 301 gave it a domestic one with a defined investigative process; Section 232 gave it a national-security route. Section 338 asks the International Trade Commission to ascertain and report on the discrimination before the President acts, and that step is where the legal exposure sits. Trade scholars have made three arguments against the Canada proclamations, and they are worth holding separately because they fail or succeed independently: that the ITC never conducted the contemplated investigation; that the goods selected — vehicles, alcohol, dairy — bear little relation to the Canadian measures the proclamations complain of, which matters because the statute is framed as an offset rather than as general leverage; and that later trade legislation occupied the field and impliedly repealed it. The first is procedural and the easiest to prove. The third is the most ambitious and the one that would kill the authority permanently.

The absence of a lawsuit five weeks in is not evidence that importers think the tariffs are lawful. It is a function of how tariff litigation works. The importer pays the duty at entry and sues to recover, which means the plaintiff finances the case out of cash flow and waits, sometimes years, for a refund it may share with everyone else in the industry once the ruling issues. That is a classic collective-action problem, and it resolves slowly: somebody with enough exposure and enough patience eventually files, usually a mid-sized importer rather than an automaker with a lobbying relationship to protect. The IEEPA litigation followed exactly that pattern, and the refund machinery that followed the February ruling is the template for what a successful Section 338 challenge would look like.

There is a second-order effect that belongs in the macro story rather than this one. The Federal Reserve's July minutes record several participants judging that the pass-through of past tariff increases into the price level was "largely complete." A new 50 per cent duty on imported vehicles, plus a Canadian counter-tariff on American steel and agricultural equipment that raises input costs for American manufacturers who buy back their own supply chains, is a live test of that judgment. If the August CPI on Friday shows goods prices reaccelerating, the "largely complete" line is the sentence that will get re-read.

Core inflation exists because food and energy prices are volatile and mean-reverting, so stripping them out gives a cleaner read on the trend a central bank can actually influence. The usual configuration in an inflationary episode is core at or above headline: demand pressure shows up across services and shelter, and the volatile components add noise around it. The current configuration is the other way round, and by nearly a full point. Headline at 3.4 with core at 2.5 says the price level is being pushed by something outside the basket the Fed's instrument reaches.

That "something" is not mysterious. The Strait of Hormuz has been effectively closed for six months, transits are running at a small fraction of pre-war volumes, and diesel is up roughly half from where it began the year (see the fuel story below). Diesel is the input that converts an oil shock into a consumer price index, because almost everything in the goods basket moves by truck at some point, and freight cost passes through to core goods with a lag of one to two quarters. So the two numbers are not independent: a sustained energy shock eventually stops being a headline story and becomes a core one. Part of what the Committee is arguing about is how much of that transmission has already happened and how much is still in the pipeline.

This is why the July dissents are more interesting than the July decision. Hammack, Kashkari and Logan are not obviously a bloc — they do not sit together on the usual hawk-dove axis, and Kashkari in particular has spent much of the past decade on the other side of it. Three dissents of that composition, all in the same direction, is a signal that the argument inside the Committee has moved off the standard axis and onto a different question: whether a central bank facing a supply shock that is now bleeding into inflation expectations has to demonstrate resolve even when its instrument is poorly matched to the shock. That is the Volcker argument, and it is the argument Chair Warsh has been making in public since Jackson Hole.

Against it sits the fiscal channel, which is where this story meets the two below it. The federal government is financing a large deficit at a 30-year yield above 5 per cent, Treasury is intervening in the long end to support liquidity, and every quarter-point of policy tightening raises the cost of the bill stock that is being used to term out the debt. A Fed that hikes into an energy shock while Treasury is buying back long bonds is a government working against itself in one direction and with itself in another, and nobody at either institution will describe it that way. Watch whether any question at the 16 September press conference is about Treasury's buyback programme. The answer, whatever it is, will be carefully constructed.

Every buyback has to be funded. When Treasury retires a 30-year bond and issues bills to pay for it, it has swapped a long, fixed, known cost for a short, floating, repriced-every-few-weeks one. In a normal cycle that is cheap: the yield curve slopes up, bills cost less than bonds, and shortening the debt saves money in the near term at the price of rollover risk later. The current curve is upward-sloping, so the arithmetic still works in the narrow sense — but the front end is anchored at a policy rate of 3.50–3.75 per cent that a majority of the market now expects to go higher next week. Financing at the short end into an expected hiking cycle is the one configuration in which the bills-for-bonds trade turns expensive quickly, and it is the configuration the government is in.

There is a second mechanism worth understanding, because it is where the money physically comes from. Treasury's cash balance at the Fed, the Treasury General Account, is the buffer between issuance and spending. A buyback can be funded out of the TGA rather than out of matched new issuance, which is faster and does not require telling the market in advance how much bill supply is coming. Drawing the TGA down adds reserves to the banking system, which is a monetary effect produced by a fiscal agent, and it is one of the few things Treasury can do that moves in the opposite direction to a Fed tightening. Nobody at Treasury will characterise it that way and the amounts here are too small to matter for reserves in aggregate. But if the operations run at the cap for the full quarter and the TGA is visibly the source, that is a fact worth having.

The debt-service arithmetic is the reason any of this is in a brief rather than a rates newsletter. Interest is now among the largest single lines in the federal budget, it is compounding, and it responds to the long end with a lag as old low-coupon debt matures and is replaced at current rates. A 30-year at 5.25 per cent rather than 4.25 is, on a rough marginal basis, an extra $10m a year in interest for every billion of new 30-year issuance, forever. That is why a nine-basis-point move on an announcement day is a real number rather than a decoration, and why Treasury is willing to spend balance sheet to get it.

The thing this operation cannot do is change why the long end is where it is. Buybacks address liquidity — the ease of transacting in size in off-the-run issues — and liquidity problems and term-premium problems look similar on a screen and are not the same thing. If the 30-year is at 5.25 because investors want more compensation for holding 30 years of American inflation and issuance risk, a $4bn operation twice a month does not touch that. It buys an orderly market, which is worth having, and it buys time. The test is simple and arrives quickly: if yields drift back up through 5.33 while the enlarged operations are running, the diagnosis was wrong and everyone will know it by November.

The Entity List is a list of named parties. That is its strength — it is precise, legally durable, and gives exporters a checkable obligation — and it is also the whole of its weakness, because a list of names is defeated by a new name. Adding a subsidiary requires an affirmative decision by Commerce's Bureau of Industry and Security about that specific entity, supported by a record. There is a doctrine for reaching subsidiaries, and BIS has used it, but it is discretionary and it consumes enforcement attention that a resource-constrained bureau has to allocate. A renamed subsidiary is therefore not a loophole in the sense of a drafting error. It is the predictable consequence of a list-based regime meeting a corporate structure that can be reorganised in an afternoon, and it recurs because it works.

The cloud gap is a different species and considerably harder. Export controls are built on the physical movement of a controlled item across a border. Remote access does not move the item; the customer moves to the item, virtually, and the compute happens where the hardware already legally sits. Nothing is exported. There is a defensible case that this is fine — the chips remain under the jurisdiction of a partner government, the data centre is auditable, and denying access pushes demand toward domestic Chinese silicon and accelerates exactly the substitution the controls exist to prevent. There is an equally defensible case that it renders the whole regime decorative for the one use that matters most, which is training frontier models. The Remote Access Security Act picks the second view. Its problem is enforcement: verifying who is renting compute in a third country requires either cooperation from the host government or intrusive know-your-customer obligations on foreign cloud operators, and the United States has limited leverage over either.

Set against that, the detail in the Times account that deserves the most weight is the institutional one, not the financial one. Ten months without an Entity List addition is a measurable fact about capacity and will, and it is the variable that determines whether any of the legal architecture above matters. A control regime is only as good as the frequency with which it is updated, because the counterparty's adaptation cycle — rename, restructure, re-route — runs in months. If the enforcement cycle runs in years, the regime is a tax on the honest rather than a barrier to the determined.

This connects directly to the thread this brief opened on 5 September about gated cyber-capability tiers. The gating story and the export story are the same story viewed from two ends. Frontier labs are being asked to restrict who can use the most dangerous capabilities of models running on American hardware; the export regime is meant to restrict who can obtain the hardware. If the hardware restriction leaks through renamed subsidiaries and leased capacity, the capability gating at the model layer becomes the only real control, which is to say the control has been privatised to the labs. That is a large policy shift and nobody has argued for it out loud.

The underlying rule comes out of Executive Order 14,399 and the Postal Service's Ballot Mail for Federal Elections final rule. Its operative provisions put the Postal Service into the ballot-handling chain in ways it has not previously occupied: review of ballot envelopes, a federal ballot-mail portal that election jurisdictions must register with, enrolment of voters in that portal, and associated processing requirements. The objection from the states is structural rather than partisan in form — elections are administered by states, and a federal mail carrier conditioning the handling of ballots on registration with a federal system is a transfer of administrative control that no statute plainly authorises. Judge Talwani's injunction blocked the mandatory provisions.

The three-trip history matters because emergency applications are supposed to be exceptional. An applicant who returns repeatedly is either encountering a lower court that keeps getting it wrong, or is shopping for a procedural posture the Court will accept. The Court's silence so far — no administrative stay on the first application, none yet on the third — is itself a signal, though a soft one. Justices who thought the injunction plainly wrong and the election calendar urgent had an easy, low-cost move available: freeze the injunction for a week while briefing completes. Declining to do that three times running is not a ruling, and should not be read as one, but it is not nothing either.

This is why the administrative stay, not the eventual order, is the thing to watch. An administrative stay is a housekeeping device with no merits content: it preserves the status quo the applicant wants while the Court reads the papers. In ordinary litigation it is close to costless. In election litigation, decided against a calendar of statutory deadlines, it is frequently decisive, because the status quo it preserves is the one that will still be in place when the deadlines pass. A rule that is administratively stayed into effect in mid-September is a rule that governs November, whatever the Court later concludes about its lawfulness. A rule that is not is a rule for the next cycle.

The countervailing consideration the Court has to weigh is the Purcell principle — the idea that federal courts should not change election rules close to an election, because late changes confuse voters and administrators. Purcell is usually invoked against injunctions that disrupt a settled process. Here the shape is inverted: the injunction preserves the process the states have been running for decades, and the stay would introduce a new federal layer eight weeks out. An applicant asking for a stay in September is asking the Court to make the late change, which is an awkward posture, and the government's briefing has to work around it. Whether the justices see it that way is the question Wednesday's filings will be arguing about.

The awkwardness for the applicant is that Missouri is asking a federal court to overturn a state supreme court's interpretation of the state's own constitution. That is close to the least comfortable posture in which to arrive on the emergency docket. Federal courts do not ordinarily review state courts on questions of state law at all; the state's highest court is the final word on what its constitution means. To get around that, an applicant has to locate a federal question — typically by arguing that the state court's construction was so far outside the range of plausible readings that it amounts to a federal constitutional violation in itself, usually framed through the Elections Clause and the argument that state legislatures hold a distinct federal function in setting the rules for congressional elections.

That argument has been available for a long time and the Court has been notably cool toward its strong form. Its most recent significant engagement rejected the maximal version while leaving room for the proposition that a state court can go too far. The result is a standard with no clear content, and applicants have to argue their case is the extreme one. Election-law commentators reading Missouri's papers have raised the question of whether the state conceded too much on this point in its own framing — whether, in trying to sound moderate about the scope of the federal question, it argued itself out of the only theory that would let the Court act. I could not read that analysis directly; the site returned a 403 to this brief. I flag the argument as existing without endorsing it, and if it is right it would explain a silence that is otherwise starting to look conspicuous.

The referendum dimension makes it harder still. The Missouri Supreme Court did not simply strike the map down; it held that citizens may put HB 1 to a popular vote and ordered the previous map used in the meantime. A federal stay would therefore not merely reinstate a legislature's map, it would override a state court's ruling about the availability of a state constitutional mechanism of direct democracy. Justices who have spent careers writing about federalism and about the limits of federal judicial power over states have to reach past both to grant this.

Which leaves the practical question of what the silence means. Two readings fit the facts equally well, and it is worth holding both. One is that the Court is drafting something — that a case this consequential, arriving with two dozen states on one side and a state supreme court on the other, will not be resolved in an unexplained one-line order, and writing takes days. The other is that the calendar is doing the work for them: if no order issues before the ballots are mailed, the case becomes moot for 2026 without anyone having to decide anything, and the Court is spared a ruling on legislature-versus-state-court in an election year. The second is cheaper and the Court has taken cheaper roads before. Both readings predict the same thing this week, which is nothing, and they diverge only after the 19th.

Gasoline is the price the public notices, because it is posted in six-foot numerals on every corner and most households buy it weekly. Diesel is the price that moves the economy. Almost every physical good in the consumer basket travels part of its journey in a diesel truck, and much of the rest moves by rail or ship burning closely related distillates. Diesel is also the input to agricultural machinery, to construction equipment, and to the backup generation that data centres and hospitals run on. A move in gasoline is a direct hit to household budgets that shows up immediately in the energy component of CPI and then stops. A move in diesel is a cost shock that propagates into the price of everything else over the following two to three quarters, and shows up in core goods long after the energy line has stabilised.

That lag is what makes the current configuration awkward for the Federal Reserve, and it is the mechanical link between this story and the Fed story above. July CPI ran 3.4 per cent headline against 2.5 per cent core — energy pushing the total up while underlying prices sat lower. The natural reading is that this is a supply shock and will pass. The diesel channel is the reason to be less relaxed: if freight costs have been elevated since the spring, the pass-through into core goods is arriving now and over the next two quarters, which means core drifts up while headline may already be flattening. A committee looking at a falling headline and a rising core in November would face a much harder decision than the one it faces next week, and the members who dissented in July for a hike are the ones most likely to be reasoning this way.

The supply side of it is worth stating plainly, because it explains why this is not self-correcting on the usual timetable. Distillate is the tight part of the barrel. Refineries have limited flexibility to shift yield from gasoline toward diesel, the closure of refining capacity over the past decade removed slack precisely in that product, and the Hormuz disruption has hit crude grades that yield distillate well. Adding crude supply does not straightforwardly fix a distillate problem. That is also why the strategic petroleum reserve is a weaker instrument here than the headlines imply: the SPR holds crude, not diesel, and releasing crude into a refining bottleneck moves the price of the thing that is not scarce.

One practical note on reading tomorrow's number. The print covers the week ending 7 September, a holiday week, and holiday weeks distort retail fuel surveys in both directions — travel demand lifts gasoline while freight volumes fall, which softens diesel. If diesel comes in flat or slightly down, that is not necessarily a turn; check the following week before concluding anything. If it comes in above $5.652 despite a holiday week, that is a stronger signal than the number itself suggests.

Aedes aegypti is a difficult target and understanding why explains the shape of the response. It is a container breeder: it lays eggs in artificial water-holding vessels around houses — plant saucers, blocked gutters, tarpaulins, bromeliads, discarded tyres, bin lids — rather than in marshes or ditches. It bites during the day, mostly around dawn and dusk, rests indoors and in shaded vegetation close to the ground, and rarely travels more than a few hundred metres in its life. Almost every mosquito biting someone in a Tampa neighbourhood hatched within a block or two of where it bit them.

That biology sets the limits of aerial application. Ultra-low-volume adulticide dropped from a helicopter kills adult mosquitoes that are flying in the open air when the droplets fall. It has little effect on adults resting under eaves or inside garages, and essentially none on eggs and larvae in containers, which hatch on the next cycle. An aerial pass therefore knocks the adult population down sharply for a few days and buys time; it does not end an outbreak. The measures that actually end one are unglamorous and slow: door-to-door container removal, larvicide in standing water, and case-based response, where each confirmed infection triggers intensive treatment of the surrounding few hundred metres. The helicopters are a visible signal that the counties think the adult population is high enough to justify the cost and the public objections, which is itself information about how the departments read the trend.

The reason all of this is worth a local slot: dengue is one of the few infectious diseases where a household's own actions materially change its exposure, because the breeding sites are on the property. Emptying anything holding water, weekly, is a genuinely effective intervention at the individual level in a way that is rare in public health. Standing water in a saucer is enough. There is also a clinical point worth carrying: second infections with a different dengue serotype carry a substantially higher risk of severe disease than first infections, through antibody-dependent enhancement. In a region that has not had endemic transmission for over a decade, almost everyone is a first infection, which is the one piece of good news in the picture — and it is also the reason a second season would be worse than this one.

Finally, the connection to a story this brief has run twice at state level. Florida's Department of Health is simultaneously running an aggressive vector-control operation in Tampa Bay and moving by rule to make four childhood vaccines optional for school entry, with the comment period closing on 14 September. Those are not the same programme or the same officials, and it would be cheap to imply hypocrisy. But they are the same department making public statements about infectious disease control within weeks of each other, and a reader tracking how the state reasons about communicable disease should hold both.