Five Questions to Ask Before You Believe a Tariff-Inflation Claim

A reader forwarded us two Federal Reserve publications from the same week in April and May 2026. One, from the Minneapolis Fed, was titled "Tariffs can't explain rising goods inflation." The other, from the Dallas Fed, concluded that tariff cost increases had reached "complete pass-through" to consumers and were driving most of core goods inflation's rise above target. The question was simple: which one is right?

The honest answer is that both are doing careful, publicly documented work, and they are not really answering the same question the same way. That gap is worth understanding, because "tariffs are/aren't causing inflation" is one of the most repeated claims in 2026 economic commentary, and most versions of it online are far more confident than the Fed's own researchers are willing to be.

Three regional Fed banks, three different reads

None of these is a fringe outlet. All three are Reserve Bank research departments publishing under their own name, with named economists and public data. And they land in different places.

Why this isn't really a contradiction — it's two different questions

Look closely at the Minneapolis and Dallas approaches and the disagreement narrows. Dallas asked, in effect: given the tariffs that were actually collected, how much cost should have passed through to consumer prices, in aggregate? Their answer — about 0.8 percentage points on 12-month core PCE inflation by March 2026 — is a top-down estimate built from a pass-through model.

Minneapolis asked a different question: if tariffs are really driving this, shouldn't the goods with the highest tariff exposure be the ones with the biggest price increases? Their answer was no — new vehicles carried heavy tariff exposure but showed comparatively little inflation, while some lower-tariff categories ran hotter. That's a bottom-up pattern-matching test, and it's a genuine methodological critique of the pass-through framework, not just a different opinion about the same number.

Both things can be true at once: tariffs may be adding real, measurable cost to the economy in aggregate, while the specific category-by-category pattern of price increases doesn't cleanly match a tariffs-did-it story. That's a more useful — and more honest — takeaway than picking whichever headline confirms what you already believed.

The St. Louis Fed's August 2026 research adds a time dimension to both: whatever tariffs were contributing, that contribution appears to have peaked and stabilized by February 2026, coinciding with the Supreme Court's ruling against tariffs imposed under the International Emergency Economic Powers Act (IEEPA) and the subsequent drop in the effective tariff rate. Since March 2026, their model attributes more of the inflation still running above the Fed's 2% target to other causes, particularly energy prices — a finding that lines up with the 14.7% year-over-year jump in energy CPI reported for July 2026.

What isn't contested

A few facts hold steady across all three research shops, whatever they conclude about the inflation math:

• The effective U.S. tariff rate rose through 2025, then fell meaningfully after the Supreme Court's February 2026 IEEPA ruling — from around 11% in late 2025 to below 7% by May 2026.
• Core goods inflation in 2026 has run well above its pre-pandemic norm (Minneapolis Fed cited 1.9% year-over-year through January 2026, against a roughly -0.6% pre-pandemic average).
• The Tax Foundation, whose estimate has moved with policy changes throughout the year, put the average 2026 household tariff cost at $840 as of its August 20, 2026 update — down from about $1,000 in 2025, reflecting the IEEPA ruling not being fully offset by other tariff actions.

Those are the load-bearing facts. The contested part is how much of the inflation — as opposed to the tariff cost itself — actually reached the prices households pay, and how that compares to everything else moving prices at the same time.

Five questions before you believe a tariff-inflation claim

1. Is this citing a specific, dated study, or just asserting a conclusion? "Tariffs are driving inflation" and "the Dallas Fed's May 2026 pass-through model estimated a 0.8-point contribution to core PCE inflation" are very different claims. Only one of them can be checked.
2. Is it measuring an aggregate pass-through estimate, or a category-level pattern? As shown above, these can disagree even when both are done well. A claim that doesn't specify which one it's making usually hasn't grappled with the distinction.
3. Does it separate goods inflation from services, shelter, and energy? Tariffs act mainly on traded goods. A claim about "inflation" broadly that doesn't isolate core goods is conflating categories that respond to very different forces — including, per the St. Louis Fed, energy price swings that have little to do with tariff policy.
4. Is the number a precise-sounding single figure, or a range with a stated counterfactual? Dallas Fed's 0.8-point estimate came with an explicit counterfactual (2.3% core inflation without tariffs). Minneapolis Fed's 0.5-point figure was stated as the low end of a range. A claim that drops the uncertainty is usually dropping information, not adding precision.
5. Would a comparably rigorous economist on the other side dispute the method, not just the number? The real Minneapolis–Dallas disagreement is a methods dispute — pattern-matching versus aggregate accounting — not two people shouting past each other. If a claim can't survive that kind of scrutiny, treat it as a talking point, not a finding.

None of this means tariffs are irrelevant to your grocery bill or your next car purchase — the Tax Foundation's $840 household estimate is real money, tariff-driven, and separate from the inflation-attribution debate entirely. It means that "tariffs caused X% of inflation" is, as of August 2026, a genuinely contested empirical question inside the Federal Reserve System itself — and a claim that sounds more certain than the Fed's own regional banks is a signal to ask which of the five questions above it would fail.