Separate a Policy's Legal Vehicle From Its Economic Effect Before Pricing Durability

Track the instrument, not the announcement. Between February and August 2026 the United States changed the legal vehicle carrying its import duties at least four times, lost two of those vehicles in court, watched a third expire on a statutory clock, and still ended the period with an average effective tariff rate roughly four times what it was two years earlier. That sequence is the cleanest natural experiment available on a question markets have been pricing badly for years: how much of a policy's market relevance is attached to the legal mechanism delivering it, and how much is attached to the policy itself.

The short version of what the 2026 record shows is that these two things separate further than most framing allows. Every individual legal instrument in the chain proved fragile. The economic effect proved considerably less so. A framework that treated "executive action is reversible, legislation is durable" as the operative distinction would have produced the right call on each vehicle and the wrong call on the exposure.

The Data Flow: What Happened to Each Vehicle

The useful unit of analysis is not the policy direction but the specific statutory authority invoked, because each authority carries a different procedural cost and a different exit. The following sequence is drawn from the decisions, proclamations, and effective-rate estimates published across the period.

Date	Event	Vehicle affected	Effective tariff rate
Before Feb 20, 2026	Emergency-powers duties in force	Emergency economic powers statute	About 16%, described as the highest since 1936
Feb 20, 2026	Supreme Court holds the emergency statute does not authorize tariffs, 6-3	Emergency powers voided	Falls to about 9.1%
Feb 24, 2026	Flat 10% balance-of-payments surcharge takes effect by proclamation	Trade Act Sec. 122	Rises to about 13.7%
May 7, 2026	Court of International Trade holds the surcharge exceeded the statute	Sec. 122 challenged, temporary stay follows	Unchanged during appeal
Jul 24, 2026	150-day statutory maximum expires; no extension enacted	Sec. 122 lapses on its own terms	Surcharge removed
Jul 24, 2026	Conduct-based duties on roughly 60 economies take effect	Trade Act Sec. 301	10% to 12.5% by origin, no sunset
Through the period	Sectoral metal duties restructured into tiered rates	Trade Expansion Act Sec. 232	50% primary, 25% derivative, 15% tiers

Five verifiable anchors sit inside that table and each one is a boundary condition rather than a forecast. The emergency statute produced a rate near 16% and survived nothing. The surcharge statute caps at 15 percent ad valorem for a period not exceeding 150 days unless extended by Act of Congress. The sectoral authority requires the Commerce Secretary to report within 270 days of initiating an investigation, the President to determine within 90 days of receiving that report, implementation within 15 days of the determination, and a written statement to Congress within 30 days. The conduct-based authority carries no expiration date at all. And a national emergency, the foundation of the vehicle that failed first, terminates on its own anniversary unless the President publishes a continuation notice in the Federal Register within the 90-day window before that date, with each chamber of Congress obligated to meet on a termination resolution every six months.

Reading the Sequence: Procedural Cost Buys Persistence

Arrange those authorities by how much process the executive must complete before duties attach, and the survival pattern lines up almost exactly. The instrument requiring only a declaration was struck down outright. The instrument requiring a statutory finding but carrying a hard sunset both lost at trial and expired on schedule. The instruments requiring a formal agency investigation, a published record, and a defined reporting chain are the ones still standing and expanding.

This is a more useful ordering than the executive-versus-legislative binary, because all four instruments in the sequence were executive actions. They differed only in how much statutorily specified procedure stood behind them, and that variable alone explained the outcomes. The reasoning in the February decision made the mechanism explicit: where an action reaches into a power the Constitution assigns to Congress, the government must point to clear congressional authorization rather than infer it from general language. Broad statutory verbs did not supply it. In May, the trial court applied the same logic one level down, holding that a surcharge justified by trade-deficit and current-account figures did not match the balance-of-payments concepts the 1974 Congress actually wrote into the statute. Both rulings turned on the fit between the claimed authority and the specific statutory text, not on the merits of the policy.

The Substitution Channel Is What the Binary Misses

A framework that scores durability instrument by instrument would have registered three separate policy defeats across six months. What actually happened was substitution. Within hours of the February ruling, replacement proclamations issued under a different statute. When the 150-day clock ran out in July, conduct-based duties covering roughly 60 economies and something close to 99% of imports attached the same day. The direction of policy never went to zero for a full trading session, even though not one of the original legal vehicles survived the period intact.

The practical implication is that the reversal probability attached to any single instrument overstates the reversal probability attached to the underlying policy, by roughly the availability of substitutes. Where the statutory toolkit is deep, a successful legal challenge changes the mechanism, the rate structure, the country coverage, and the compliance burden without changing the direction. Where the toolkit is shallow, the same challenge ends the policy. That distinction is knowable in advance and it is not what most policy-risk commentary measures.

The Refund Channel: A Second, Slower Data Series

Voiding an instrument retroactively creates a cash-flow event that resolves on a completely different timetable from the policy question, and this is where the durability framework produces its most concrete number. Duties collected under the voided emergency authority formed a refundable pool estimated at roughly $166 billion. Customs began processing claims through a dedicated program in April 2026. By early August, on the agency's own court filings, about $100 billion had been certified and forwarded to Treasury for disbursement, covering an eligible population of roughly 330,000 importers across approximately 53 million entries.

Three things follow. First, roughly 60% recovery over about five and a half months is a reasonable base rate for how quickly a large retroactive unwind actually reaches balance sheets, and it is slower than the market reaction to the ruling implied. Second, the government appealed the refund order, meaning the remaining portion carries both timing and outcome uncertainty rather than timing alone. Third, the refund is a one-time working-capital event while the replacement duties are an ongoing cost, so the two do not offset in any given quarter even when they roughly offset in aggregate.

Where Legislation Turns Out to Be Less Durable Than Assumed

The reverse asymmetry deserves equal weight, because the standard framing overstates legislative permanence in one specific and increasingly used channel. Agency rules issued under statutory authority sit inside a review window: a joint resolution of disapproval can nullify a rule, and the Senate procedure for it is expedited past the ordinary threshold. That mechanism was used twenty times in total between 1996 and 2024. In the single year 2025 it was used more than that, with the current Congress recording 23 resolutions enacted on one running tracker.

The consequential detail is the aftermath. A disapproved rule may not be reissued in substantially the same form, and no substantially similar rule may be issued, unless specifically authorized by a law enacted after the disapproval. That is a stronger and more permanent effect than the reversal of a typical executive action, which a later officeholder can simply reinstate. So the durability ranking inverts in this corner: a disapproved regulation is harder to revive than a rescinded executive order, and the tool doing the disapproving runs on a compressed calendar tied to when the rule was reported to Congress.

The Reading That Cuts the Other Way

Several conditions would make this framework the wrong lens, and at least two of them are live.

One period is one observation. The 2026 sequence involved a single policy area with an unusually deep statutory toolkit. Trade has at minimum four independent tariff authorities with different triggers and different rate ceilings, one of which permits duties up to 50% on discriminatory commerce with a 30-day notice and has never been invoked. Immigration, environmental permitting, and health regulation do not offer comparable substitution depth. Generalizing a substitution rate from the most substitutable policy area available would be a category error.

Substitution is not free, and the friction may be the point. Moving from a flat surcharge to origin-and-classification-dependent duties with exemptions and partner caps changed the compliance calculation for every importer even where the headline burden was similar. For a firm concentrated in a carve-out, the vehicle change was the entire story. Aggregate effective rates smooth over exactly the dispersion that determines individual outcomes, so a framework built on the aggregate series is the wrong tool for firm-level exposure.

The procedural-cost ordering may be reading a trend into a small sample. The sectoral and conduct-based authorities have not been tested at the Supreme Court on the same major-questions grounds. A ruling that applied the February reasoning to a sectoral finding built on an expansive national-security definition would break the ladder. The correct statement is that procedurally grounded authorities have not yet been struck down, not that they cannot be.

Speed still has independent value. A duty in force for 150 days before expiring is not economically equivalent to no duty. It moved the effective rate by roughly 4.6 percentage points for five months, on published estimates, and produced price effects in the range of half a percentage point to a full point depending on persistence. Treating low durability as low relevance would have missed all of it.

Risk Factors Around the Framework

• Appellate reversal risk runs in both directions. The trial-level ruling against the surcharge is on appeal and the surcharge has already lapsed, so an appellate outcome now affects refunds and precedent rather than the duty itself. A reversal would restore an authority that is currently treated as unavailable.
• Codification is the tail that changes the distribution. Any statute converting current duties into legislated schedules would move the entire structure to the bottom of the ladder in one step. Nothing in the record makes this likely, but the payoff shape is asymmetric and it is cheap to monitor.
• The 270-day investigation clock is a leading indicator. Sectoral investigations initiated now produce reports up to nine months out, followed by a 90-day determination window and 15 days to implement. Initiations are therefore visible roughly a year before duties can attach, which is the longest reliable lead time in the sequence.
• Refund concentration is uneven. With about 330,000 eligible importers, recovery timing varies enormously by entry volume and filing sophistication. A 60% aggregate disbursement rate says little about any specific claimant.
• Emergency-renewal dates are calendar events. Where an underlying declaration still supports other measures, the anniversary and the 90-day continuation window are fixed, observable dates rather than discretionary ones.

What to Watch Next Week

• Federal Register notices initiating new sectoral investigations, and the initiation date on each, since that date starts a 270-day statutory clock that is fully observable from the first day.
• Docket movement at the appellate level on the surcharge and refund questions, specifically whether stays are extended, narrowed, or dissolved, which changes refund timing before it changes any legal conclusion.
• Weekly customs disbursement figures against the estimated $166 billion pool. The relevant question is whether the pace after the June surge holds or decays, since the shape of that curve is the observable input to any unwind assumption.
• Exclusion and exemption notices under the conduct-based duties. Coverage carve-outs move firm-level exposure more than headline rates do and they are published individually.
• Any introduced legislation that would extend a lapsed surcharge or codify existing schedules, watched at introduction rather than at passage, because introduction is when the tail risk becomes priceable.

Concrete Framework — How to Verify It Yourself

A monitoring checklist for policy delivered through executive channels, applied instrument by instrument rather than headline by headline.

1. Name the statute, not the announcement. Record the specific authority invoked and the section number. Two policies with identical economic content and different statutory bases carry different reversal probabilities, and only the statute tells you which.
2. Score procedural cost on four inputs. Is a formal agency investigation required? Is there a published report? Are there statutory deadlines with day counts? Is there a notification duty to Congress? Zero of four indicates a declaration-only instrument. Four of four indicates something considerably harder to unwind.
3. Find the sunset before anything else. Some authorities carry hard expiries measured in days. If a hard expiry exists, the expiry date is a higher-quality signal than any litigation forecast, because it requires no one to act.
4. Count the substitutes. List every alternative authority that could deliver similar content. Reversal risk for the policy direction is the joint probability that all of them fail, not the probability that this one does. In a four-authority policy area those are very different numbers.
5. Separate the retroactive channel from the forward channel. If an instrument is voided, model the refund or restitution flow on its own timetable, with its own appellate risk, and do not net it against ongoing costs within a quarter.
6. Check whether the disapproval channel applies. Where the action is an agency rule rather than a direct executive order, the compressed review window and the prohibition on reissuing substantially similar rules make disapproval more permanent than ordinary reversal.
7. Re-score on every judicial ruling. A decision on one authority resets the estimated durability of every neighboring authority resting on similar reasoning, whether or not that authority was before the court.

The framework does not forecast outcomes and is not built to. It converts a question that usually gets answered by tone into one answered by observable inputs: which statute, how much process, what sunset, how many substitutes. On the 2026 record those four inputs ordered the results correctly while the executive-versus-legislative binary did not.

Disclaimer: This is analysis of policy transmission mechanisms and is not investment, legal, or tax advice. Figures cited reflect published estimates and court records available at the time of writing and are subject to revision, appeal, and subsequent action.