Tariff Imbroglio — The Surface and the Floor
What the aggregate numbers conceal — and where the actual trade is
Tariffs are showing up exactly where macro models are least sensitive — and disappearing exactly where investors are looking. The result is a market reading the surface while mispricing the floor.
What looks like a goods-price shock is masking a deeper shift in the equilibrium real rate — and that’s where the Duration trade lives.
EXECUTIVE SUMMARY
Tariffs are imposing a regressive real-income shock that weakens demand where aggregate statistics are least sensitive, allowing headline PCE to remain resilient even as localized economic damage accumulates. At the same time, technology — particularly AI — is compressing the marginal cost of cognitive work, exerting a structural disinflationary force that is pulling the neutral real rate lower. The market’s focus on the visible goods-price shock is obscuring this deeper shift, creating a mispricing in long-Duration assets. The trade is not about imminent rate cuts but about where the equilibrium rate ultimately clears once the surface noise fades.
MACRO ANALYSIS
The current tariff regime is doing two things simultaneously, and most commentary is only seeing one of them. On the surface, it is imposing a goods price shock — visible, measurable, and giving the Federal Reserve exactly the kind of ambiguous inflation signal that argues for inaction. Underneath, a structural deflationary force driven by technology is pulling the neutral rate steadily lower. Markets reading only the surface risk being systematically wrong about where rates ultimately settle — and that is the trade this piece is about.
The aggregate demand arithmetic is where the analysis must start, because it is where the most consequential misdirection lives.
The analysis operates across three horizons: cyclical demand effects, policy mechanics, and a secular cost shock.
I. The arithmetic — and what it conceals
The Budget Lab at Yale estimates the average US household loss from current tariff policy at $800 per year in real purchasing power. Multiply by approximately 134 million households and the implied aggregate demand drag is roughly $100 billion — call it 0.47% of the roughly $21–22 trillion annualized PCE base as of late 2025. That sounds manageable. It is supposed to sound manageable. Averages are how damage gets buried.
The $800 figure is computed across a distribution where the number does almost no analytical work. The Budget Lab’s own distributional tables show the bottom income decile carrying a burden of 1.1% of post-tax income — three times the 0.4% borne by the top decile. In dollars: $400 extracted from a household running on fumes, versus $1,800 from a household whose consumption is underwritten by asset returns. These are not symmetrical losses. The behavioral consequences diverge completely.
The $800 average is how damage gets buried. What matters is not the mean loss but whose loss it is — and those two questions have opposite answers.
II. Who drives PCE — the 60/35 reality
The precise number matters less than the direction: consumption power is more concentrated than the headline surveys imply.
Consumer spending in the United States is heavily concentrated at the top of the income distribution, though the degree of concentration is genuinely disputed — and the gap between available measures is itself worth examining.
The Bureau of Labor Statistics Consumer Expenditure Survey attributes roughly 35% of total consumption to the top income quintile, a figure unchanged for two decades. That stability should invite immediate skepticism. The same two decades produced the most dramatic run-up in income and wealth concentration since the Gilded Age. A consumption measure showing no corresponding shift is almost certainly missing the upper tail — which the survey’s own income data confirms: it puts the top quintile’s income share at 47%, against the 60% recorded by the Federal Reserve’s Survey of Consumer Finances, a tool built precisely to capture wealthy households that surveys routinely undercount.
Dallas Fed research published in November 2025, using SCF-based methodology, finds the top 20% of earners responsible for 57% of overall consumption — up four percentage points over thirty years. Some private estimates place the top decile’s share alone near 49%.
A working estimate grounded in the higher-quality measures puts roughly 60% of PCE with the top 30% of earners. Call it the 60/35 reality. The implication for tariff analysis runs directly: the households bearing the lightest burden as a share of income are the same households whose spending decisions dominate the PCE print. When analysts cite aggregate PCE resilience as evidence that tariff damage is contained, they are observing that wealthy households are still spending. That is not reassurance. It is a measurement artifact of inequality.
Aggregate PCE resilience is not evidence that tariff damage is contained. It is evidence that the households driving PCE are not the ones being damaged.
III. The K-shape transmission — where the damage actually lands
The lower two income quintiles will not absorb a price shock the way the upper half does. They substitute — store brands for name brands, deferred medical appointments, cut cable, no restaurant meals. This is not conjecture; it is the documented response of lower-income households in every inflationary episode of the past twenty years. The spending that disappears does not diffuse broadly. It comes out of the specific sectors that serve lower-income consumers: mass-market retail, fast food, dollar stores, community services.
The workers in those sectors are themselves lower-income. A demand contraction at the bottom propagates into employment at the bottom, producing a second-order labor market effect that headline unemployment figures will catch only partially and with a lag. The Budget Lab projects a model-estimated 0.3 percentage point rise in the unemployment rate by end-2026 — a floor, not a ceiling, for communities where the income concentration of the damage is highest.
The K-shape is not just a distributional observation. It explains why the headline numbers will look acceptable while the structural damage compounds. PCE holds because the top half keeps spending. Unemployment ticks up modestly because the job losses concentrate in sectors that carry low statistical weight. The aggregate economy and the lived economy diverge — slowly in the data, immediately in reality.
The SCOTUS ruling on IEEPA tariffs adds a further twist that sharpens this picture rather than softening it. The $133 billion in IEEPA duties collected through December 2025 will ultimately flow back — but to whom? The Supreme Court left the refund mechanism entirely unaddressed. Justice Kavanaugh’s dissent, borrowing a word Justice Barrett used at oral argument, warned that the process is likely to be a “mess” — with the path running through the Court of International Trade, involving over 301,000 importers across 34 million separate entries, taking an estimated 12 to 18 months to clear according to TD Securities.
Critically, the refunds flow to importers of record — the companies that wrote the checks to Customs. Not to consumers. The households who absorbed the tariff costs through higher retail prices have no standing to claim refunds; that money, to the extent it is recovered at all, returns to corporate balance sheets. Trump has also signaled active resistance to the refund process, suggesting litigation could stretch the timeline further. Penn Wharton projects up to $175 billion in total refunds owed — a genuine fiscal impulse when it arrives, but one that accrues to companies and shareholders, not to the lower-income households who bore the real purchasing power loss. The K-shape extends into the refund mechanics. The damage was distributed down; the recovery will be distributed up.
The tariff pain was passed through to consumers. The refunds will not pass back. The K-shape extends into the recovery mechanics.
IV. The offsets — real, but differently located
Three forces push against the tariff drag, each genuine, each operating on a different part of the national accounts.
AI capital expenditure is the cleanest near-term offset. The major hyperscalers collectively guided over $300 billion in capex for 2025, flowing directly into the fixed investment component of GDP with high domestic labor content in construction and deployment. This demand driver is independent of trade policy entirely. The productivity payoff — the secular cost deflation that AI will eventually impose on the knowledge economy — is a medium-term story addressed in the next section.
Net exports provide a second offset that standard tariff models underweight. Import compression is mechanical: a 15% tariff raises import prices, volume falls through substitution and demand destruction, and the import subtraction in the GDP identity shrinks. Against a 2024 goods import base of $3.3 trillion, even a 3–4% volume decline represents roughly $100 billion of positive GDP arithmetic — approximately matching the consumer-side demand destruction in real terms.
The dollar complicates this further. Conventional theory predicts tariff-driven dollar appreciation — reduced import demand raises relative demand for the domestic currency. The 2025 data went the other way: the dollar weakened roughly 6% against its December 2024 average, driven by institutional credibility erosion, sovereign portfolio diversification away from dollar assets, and long-run fiscal anxiety. A structurally weaker dollar improves export competitiveness on top of import compression. It also amplifies the goods price shock on imported inputs, feeding back into exactly the inflation reading that is keeping the Fed pinned. The currency channel cuts in both directions simultaneously, which is what makes it so analytically treacherous. The drivers of that weakness — credibility erosion, sovereign diversification away from dollar assets, and structural fiscal deterioration — are not cyclical; the 2025 depreciation is more plausibly the opening move of a further repricing than a transient overshoot, given the pillars of US exceptionalism are eroding.
V. The surface and the floor — two price regimes, one confused market
The goods price shock from tariffs and dollar weakness is real. It is also transitory by construction — tariffs impose a one-time price level shift, not a self-sustaining inflation impulse. The Fed knows this. Its communication has been consistent: supply shock, look through it, watch for second-round effects. The resulting hamstrung posture — unable to cut because of surface inflation, unable to hike without further crushing lower-income households — is not a policy error. It is the only defensible response to a price structure whose layers are pointing in opposite directions.
The deeper layer is deflationary, and it predates the tariff regime by years.
AI is compressing the marginal cost of cognitive work sharply downward across the knowledge economy. Legal drafting, financial analysis, software development, diagnostic support, customer service: the unit cost of each is in structural decline, accelerating as deployment scales and model costs fall. This is not forecast. Firm-level data in professional services labor markets already shows it.
The late-1990s US productivity acceleration is the right historical parallel. Measured TFP growth ran above 3% annually for several years, wrong-footing a Fed whose models assumed a stable Phillips Curve. Greenspan held off on tightening longer than his models said he should because he read the productivity story correctly. The difference in the current episode is distributional: the 1990s gains were broadly shared, lifting wages across the income spectrum. AI’s deflationary force accrues to capital owners and the highest-skilled knowledge workers, while the displacement costs fall hardest on the cognitive professional middle — the $60–120K tier that survived earlier automation rounds by moving up the skill ladder. Aggregate productivity rises. The K-shape deepens. Both are true simultaneously.
Goods prices rise on the surface. The marginal cost of cognitive work trends toward zero underneath. The Fed is reading the surface. The bond market will eventually price the floor.
The yield curve call follows from this directly. Markets price policy rates, but Duration prices the equilibrium real rate. If the structural deflationary force is real — and the evidence from labor markets, AI deployment cost trajectories, and the secular decline in r-star all point that way — the Fed’s terminal rate in this cycle is lower than market pricing implies. Not because the economy is weak, but because the neutral real rate is being pulled down by a technology-driven secular force that tariff-generated goods inflation is temporarily masking.
Near term, the curve steepens: the front-end stays anchored as the Fed correctly reads the goods shock as transitory and declines to tighten into it, while the long end remains hostage to fiscal supply and dollar uncertainty. As the deflationary floor asserts itself in the data — and it will, because the productivity gains are real and will eventually be unmistakable — the curve bull-flattens. Duration is a position on that floor, not a bet on near-term cuts. The entry point improves the longer the surface noise keeps other investors away.
The argument, stated plainly
Tariff damage is real and it is concentrated exactly where the aggregate statistics are least sensitive: in the lower half of the income distribution, in the sectors that serve them, and now — through the refund mechanics — in the asymmetric recovery that will return $130-plus billion to corporate balance sheets while the households who actually absorbed the price increases see nothing back. Aggregate PCE will hold up. That is not health. It is what an unequal economy looks like when you tax it from the bottom.
Underneath the tariff-driven goods price shock, technology is compressing the cost of cognitive work with the same secular force that containerization brought to logistics and cloud computing brought to storage. Each of those transitions was disinflationary in aggregate while redistributing income sharply upward. This one is the same, faster, and broader.
The Section 122 clock — 150 days from February 24 — is the next forcing event. Whether the administration rebuilds the tariff wall under Section 232 and 301 authority, negotiates extensions, or lets the duties lapse will determine the timing of the trade described here, not its direction. Trading partners watching that clock are making the same calculation: hold existing deals, seek quiet renegotiation, or risk provoking a White House that has already demonstrated it will respond to judicial constraint with immediate executive escalation.
The counter-view is that fiscal dominance overwhelms the deflationary impulse; the trade here is that technology outruns policy.
Markets that price only the surface — the tariff-driven goods inflation that is real but temporary — will be wrong about the terminal rate, wrong about r-star, and wrong about the long-run equilibrium for fixed income. The confusion between transitory goods inflation and structural deflation is not a minor analytical error. It is the central macro mispricing of this moment.
Sources: Budget Lab at Yale, State of Tariffs February 21, 2026; BLS Consumer Expenditure Survey 2024; Dallas Federal Reserve, Consumption Concentration May Be Up, November 2025; Federal Reserve Survey of Consumer Finances; BEA Personal Consumption Expenditures through December 2025; Census Bureau Housing Vacancy Survey; U.S. Customs and Border Protection, IEEPA Tariff Collections Data, December 2025; Penn Wharton Budget Model, IEEPA Revenue and Potential Refunds, February 2026; NPR / Fortune refund coverage February 21–22 2026; TD Securities timeline estimate; SCOTUSblog, A Breakdown of the Court’s Tariff Decision, February 2026; Morgan Stanley capex estimates; company earnings guidance, 2025.
macrofireside.com · The Macro Fireside · February 2026 · For professional enquiries: gs@macrofireside.com