Market Data

Seasonally Adjusted or Not: The Same Month Can Be a Gain and a Loss at Once

A single month of data can show a gain seasonally adjusted and a loss unadjusted, and both are true. Knowing which one answers your question is a small piece of literacy that prevents a lot of bad decisions.

Most manufacturing series are published two ways: seasonally adjusted and not seasonally adjusted. The two can point in opposite directions for the same month, which sounds like a contradiction and is not. Seasonal adjustment removes the predictable calendar swings, holiday shutdowns, summer slowdowns, year-end pushes, so that what remains is the underlying trend. The unadjusted number is what literally happened. Both are correct; they answer different questions, and quoting the wrong one for your question is a common, avoidable error.

What the adjustment actually does

Seasonal adjustment uses a statistical model (the X-13 method is the standard) to estimate the regular calendar pattern in a series from its history, then removes it. If manufacturing employment reliably dips every July for retooling shutdowns, the adjustment expects that dip and strips it out, so an adjusted July number that is flat actually reflects a stronger-than-normal month, because the usual seasonal drop did not fully happen. Manufacturing employment currently reads 12,611 thousands of employees (Jul 2026) on the adjusted basis, down about 0.1% from a year ago, and that adjusted figure is the one built for month-to-month trend comparison.

When to use which

• Seasonally adjusted: Trend questions (Is the sector growing or shrinking this month versus last? Use adjusted, so the summer shutdown does not masquerade as a downturn.)
• Not seasonally adjusted: Actual-level questions (How many people were actually on payroll, or how much was actually shipped? Use unadjusted, the literal count.)
• Year-over-year: Either, consistently (Comparing the same month across years cancels most of the seasonal pattern by itself, which is why year-over-year is a robust cross-check.)

Seasonally adjusted answers is it getting better. Unadjusted answers what actually happened. Mixing them up turns a normal July into a false alarm.

The trap to avoid

The classic mistake is comparing an adjusted number in one breath with an unadjusted number in the next, or reading an unadjusted seasonal swing as if it were a trend. A housing starts figure, currently 1,427 thousands (SAAR) (Jun 2026) adjusted, will always look dramatic in raw winter-versus-summer terms, but that drama is the season, not the economy. The safe habits: state which basis you are quoting, keep the basis consistent across a comparison, and when in doubt fall back on year-over-year, which quietly cancels most of the seasonal effect without any model at all.

Read how seasonal adjustment is applied and verified across the series on the methodology page. See the adjustment

Published 2026-08-05.