Why not just apply market rents to shelter?

Because new leases are the marginal tenant, not the population: only about 4.5% of leases turn over in a month, so most tenants pay a rent set months or years ago. Applied at level, new-lease rents ran far ahead of the CPI in both directions and never converged to it. Nowflation's all-tenant rent index averages new-lease rents over the trailing 48 months, weighted by the share of leases still in force, and the gauge uses that reading as measured; on the record it runs about 5 months ahead of official rent (correlation 0.987 since 2018) — a lead published as a diagnostic on the lead-test page, not a forecast built into the gauge. Raw new-lease rents are still published on the housing page as where shelter is heading.

Is the official CPI wrong?

No. The BLS measures what it says it measures. Its shelter method is lagged by design and its collection cycle is monthly. Nowflation measures the same economy with today’s market prices. Both are true; they answer different questions.

Why trust an independent inflation index?

Verify instead: every source is public and named, the methodology is versioned with a public changelog, historical values are never silently revised, and every CPI forecast is graded publicly after each print next to the Cleveland Fed, Street consensus, and Kalshi.

What data goes into the Nowflation gauge?

25 input series across the 21 CPI components: Zillow and Apartment List new-lease rents feeding an all-tenant rent model, AAA pump prices, Manheim wholesale used-vehicle values and Kelley Blue Book new-vehicle transaction prices, and the official BLS subindex, held at its last measured print between releases, wherever no better daily market source exists. Home values at today's mortgage rate feed the affordability measure on the housing page, not the Gauge. Behind them sits an open directory of 1,684 public series from 26 sources, every one downloadable. Full list on the methodology page.

Why the Nowflation CPI Gauge (3.45%) Differs From Official CPI (3.7%): The Gap Explained

Q: Why does Nowflation differ from official CPI?

The Nowflation CPI Gauge reads 3.45% while official CPI reads 3.7% (September 2026 print). Against the September 2026 print, component by component (each component's year-over-year change in the Gauge today against its BLS subindex in that print, times its basket weight): shelter −0.41pp (the all-tenant rent index at 2.0% against official rent at 2.8% and owners' equivalent rent at 3.3%); motor fuel +0.17pp (40.2% against 34.2%); used vehicles +0.03pp (−0.5% against −1.5%). Those add up to −0.21pp; the remaining −0.04pp of the −0.25pp gap is aggregation, the seasonally adjusted subindexes against the unadjusted headline, and the print's one-decimal rounding. Over time the largest structural difference is shelter: the BLS reprices each surveyed rental unit only every six months, so official rent blends leases signed up to a year ago, while Nowflation measures what all tenants pay from this month's new leases (Zillow and Apartment List rents averaged over every lease still in force, used as measured — not forecast), a reading that on the record runs about 5 months ahead of the official rent index — plus daily market prices versus a monthly survey.

The gap is -0.25pp. It is not an error — it is the method. Against the September 2026 print, component by component (each component's year-over-year change in the Gauge today against its BLS subindex in that print, times its basket weight): shelter −0.41pp (the all-tenant rent index at 2.0% against official rent at 2.8% and owners' equivalent rent at 3.3%); motor fuel +0.17pp (40.2% against 34.2%); used vehicles +0.03pp (−0.5% against −1.5%). Those add up to −0.21pp; the remaining −0.04pp of the −0.25pp gap is aggregation, the seasonally adjusted subindexes against the unadjusted headline, and the print's one-decimal rounding.

Over time the largest structural difference is shelter. The BLS measures shelter with a survey that reprices each unit every six months, so official rent blends leases signed up to a year ago; Nowflation measures what all tenants pay from this month's new leases (tenure-weighted, used as measured — not forecast), a reading that on the record runs about 5 months ahead of the official rent index, and prices current mortgage costs in its cost-of-living variant. Our gauge uses daily market prices where they have a better record than the survey (rents, pump prices, wholesale used cars) and holds every other component at its latest official subindex; the CPI is a monthly survey published two weeks late. The gauge is a measurement, not a forecast. Whether the CPI Gauge's reading on the eve of a print is a better guess of that print than the previous print is not established: the difference is 0.03pp in the gauge’s favour, under the 0.05pp margin this page requires before calling either the better guess (since 2019: 0.30pp mean absolute gap vs 0.32pp for the previous print, 93 prints); under the lead-test rule its correlation with official CPI k months ahead ties CPI's own persistence (best at 12 months: 0.24 vs 0.23; it clears only by at least 0.03 at some horizon from 5 to 12 months).

Today's gap, component by component (against the September 2026 print)

Component	CPI Gauge YoY today	BLS YoY (September 2026)	Weight	Gap contribution
Shelter: Owned (priced on the all-tenant rent index)	2.0%	3.3%	26.2%	−0.34pp
Motor Fuel	40.2%	34.2%	2.9%	+0.17pp
Shelter: Rent	2.0%	2.8%	7.8%	−0.06pp
Used Vehicles	−0.5%	−1.5%	2.8%	+0.03pp

Gap contribution = basket weight × (the CPI Gauge's component YoY today − the BLS subindex YoY in the September 2026 print). Components that hold that print have no difference and are left out. Unrounded, the rows sum to −0.21pp against the −0.25pp gap; the rest is aggregation, seasonal adjustment and the print's one-decimal rounding.

Both numbers are true; they answer different questions. Every source is public, the methodology is versioned, and every forecast is graded after each print.