Here is the contradiction that should not exist.
The effective tariff rate—total duties collected divided by the total value of imports—quadrupled from roughly 2.5% to over 11% in about a year. Then, per the St. Louis Fed, it "appear[ed] to have peaked in late 2025 and [has] since begun declining."
The legal foundation cracked too. Per WWD/Sourcing Journal, the Supreme Court invalidated the IEEPA tariffs, and Customs and Border Protection opened a refund portal to 330,000 importers who paid duties on 53 million entries worth around $166 billion.
Every top-line signal says the shock is draining out. Rate falling, refunds flowing, duties on some entries reportedly no longer even being collected.
And yet: 47% of service firms and 44% of manufacturers that paid tariffs directly told the New York Fed they still plan more price increases. Sixteen percent of service firms and 7% of manufacturers plan hikes more than six months out.
So the tax is receding while the pricing pipeline keeps filling. The question this piece answers is why—and the answer lives entirely inside firm-level pricing mechanics, not the tariff schedule.
Start with what "pass-through" actually looked like in the durable-goods aisle. Appliances, electronics, and furniture had been in outright deflation—prices falling as much as 3% year-over-year from mid-2023 into early 2025. Then the effective rate spiked, and durable-goods prices flipped to +2% to 3% by early 2026.
That flip is the aggregate result. What produced it was thousands of individual pricing decisions, staged on schedules that have nothing to do with when a tariff was announced.
The first mechanism is the most mechanical: contracts. Some firms sell under fixed-price agreements and simply cannot reprice until the contract expires. Until then they absorb the cost, and research cited by the Fed confirms long-term contracts impede pass-through.
This is not lagged catch-up in the usual sense. It's a scheduling artifact. The cost hit landed in 2025; the price adjustment lands whenever the contract cycle turns over—which could be well into 2026 or beyond, long after the rate that triggered it has fallen.
The second mechanism is deliberate, and it's where the compounding is manufactured. The Fed economists call it "trickle up": firms "gradually raise prices over time rather than immediately raising prices to fully cover tariffs."
The stated purpose is twofold. It avoids "shocking their customers with sharp price increases"—and, critically, it lets firms retain "the ability to accelerate price increases if input costs continue to rise."
Read that second clause closely, because it's the whole game. Trickle-up isn't just slow pass-through. It's an option structure.
A firm that raises prices in small increments has built itself a ratchet. If costs stabilize or fall, it can stop where it is. If costs rise again, it accelerates from an already-elevated base. The downside is capped; the upside is open.
A firm that raises prices in small increments has built itself a ratchet. The downside is capped; the upside is open.
That asymmetry is, in this reading, what makes the pass-through compound rather than fade. On the way up, increases are staged slowly to keep customers calm. But nothing forces a symmetric glide back down when the effective rate declines.
Policy uncertainty pours fuel on this. The Fed notes that uncertainty over "potential rate changes, exemptions, or tariff responses from other countries" pushes firms toward "cautious, incremental pricing strategies rather than making large, discrete adjustments."
And the environment keeps validating that caution. The administration declined to extend the USMCA pact and threatened new barriers with Spain. Each fresh threat resets the firm's planning horizon and keeps the incremental strategy switched on.
Now the distributional layer, which determines whether the pipeline drains or refills. This is where the small-business data matters.
The largest 1% of U.S. firms account for 90% of international trade by value, but exposure to imported inputs runs deep and small. About 70% of goods firms and 80% of retail firms reported using foreign-sourced inputs in 2024—versus only about 30% and 20%, respectively, that sell to international customers.
So small goods and retail firms are input-exposed without being export-exposed. They eat the cost side of tariffs with almost none of the offset. In 2025, about 80% of them reported their imported input prices rose.
Their response was overwhelmingly to pass it on. Per the SBCS data, about 80% of national and regional goods and retail firms passed on at least some of the higher costs, while about 60% also absorbed some—36% of goods firms and 43% of retail firms doing both.
That "both" is the tell. Partial pass-through plus partial absorption is precisely the profile of a firm that hasn't finished adjusting—it has taken a margin hit it intends to recover, and the recovery is still queued.
Who can afford to absorb, and who can't, is not random. The Fed's regressions found that older and more profitable small firms were more inclined to pass costs on and less inclined to absorb them, because survival and profitability confer pricing power.
The corollary is the dangerous part. Smaller, less profitable firms "with fewer resources are less able" to maintain markups or mitigate through legal means—yet they're the ones with no margin buffer to sit on the cost while it works out.
So the pipeline stays loaded from both ends. Firms with pricing power raise deliberately, in trickle-up increments they can accelerate. Firms without it have no cushion to absorb, so the cost must eventually surface as price whenever they can manage it.
This is why "one-time price-level adjustment" is a category error at the firm level. The phrase assumes a clean impulse: tariff lands, prices step up once, done.
But as the Fed puts it, "what 'one-time' means in practice may be a drawn-out affair, especially when the tariffs change frequently." Every policy change resets each firm's adjustment horizon, and every contract expiry releases another batch of deferred increases.
Stack enough of those staggered, resetting horizons on top of each other and the aggregate stops looking like a step. It arguably looks like persistent supply-side inflation—reportedly refilling faster than it drains, even as the headline rate falls and refund checks go out.
The timing quietly confirms it. The durable-goods flip from deflation to +2–3% didn't peak with the tariff rate and reverse; it arrived after, and the Fed's own survey says some firms plan increases more than six months in the future.
One handoff, then I'll stay in my lane. The persistence documented here arguably reflects the same supply-side pressure that shows up as a constraint on the Fed—but note what it is and isn't. In this reading, it is a firm-behavior artifact: contract cycles, option-value pricing, and small-firm margin math. It is arguably not a demand signal.
That distinction matters because, in this reading, the mechanism is self-limiting in a way a demand-driven spiral is not. Trickle-up would compound only as long as firms expect cost risk to stay live. Give them a stable schedule and the option-to-accelerate arguably loses its value; the ratchet stops clicking.
Until then, the counterintuitive fact pattern holds. The tax is receding and the price hikes are still coming, because the compounding was never in the tariff rate.
It was built into the pipeline—one contract, one trickle, one reset at a time.