Why It Matters
Most federal economic data you see comes in two flavors: seasonally adjusted (SA) and not seasonally adjusted (NSA). The difference matters because raw monthly or quarterly data is wildly seasonal. Unemployment always rises in January (post-holiday layoffs). Home sales always peak in June. Construction permits always crash in December. Without seasonal adjustment, comparing January to June would tell you less about the economy and more about the calendar. The BLS and Census publish SA versions of most series for month-to-month and quarter-to-quarter comparisons. They also publish NSA versions for year-over-year analysis and academic work. The critical trap for investors: BLS LAUS (metro-level unemployment) is NSA by default — so you can't compare metro Nov to metro Dec directly. Use year-over-year instead.
At a Glance
- What it is: Removing calendar-driven patterns from a time series so you can compare across months or quarters.
- Who does it: BLS, Census, FHFA, and most federal statistical agencies. Uses a standard statistical tool called X-13ARIMA-SEATS.
- SA = Seasonally Adjusted. Month-to-month changes are meaningful.
- NSA = Not Seasonally Adjusted. Month-to-month changes mix trend and seasonality — use YoY instead.
- Key trap: BLS CES employment is SA at metro. BLS LAUS unemployment is NSA at metro. Check which series you're pulling.
- How to spot it: Federal data release footnotes always specify "Seasonally Adjusted" or "Not Seasonally Adjusted." FRED tags series with "SA" or "NSA" in the name.
How It Works
What the adjustment actually does. Seasonal adjustment uses a statistical method (X-13ARIMA-SEATS is the Census Bureau's standard) that decomposes a time series into three components: trend (long-term direction), seasonal (predictable calendar pattern), and residual (everything else). The seasonal component gets subtracted, leaving trend plus residual. What you see on the published SA series is "what the economy is doing after stripping what the calendar would do anyway." For most federal series (housing starts, unemployment national, CPI, retail sales, GDP), SA is the published version and NSA is archived. For some series (metro-level unemployment via LAUS, some local economic data), NSA is the published version and SA doesn't exist at that grain.
Why BLS LAUS at metro grain is NSA. The short answer: the sample size at metro level isn't large enough to reliably identify the seasonal component. Seasonal adjustment requires 5+ years of stable seasonal patterns to calibrate. National unemployment has that. Metro unemployment — especially for smaller metros — has thinner data and shifting demographics that make the seasonal pattern less stable. Rather than publishing a poorly-adjusted metro series, BLS publishes NSA and leaves seasonal adjustment to the user. The implication for investors: if you pull metro LAUS for November vs December, you'll see unemployment rising — because that's the seasonal pattern, not the underlying trend. To strip seasonality, compare to the same month a year ago (YoY) instead.
SA vs NSA affects your analysis differently by use case. Three scenarios. First, month-over-month trend monitoring: you MUST use SA. NSA MoM is dominated by seasonal swings. Second, year-over-year level comparison: either SA or NSA works — same-month comparisons cancel out seasonality. Third, academic or research analysis: NSA is the raw truth and lets you apply your own adjustment method. For real estate investors, the practical rule: use SA for monthly tracking, use YoY on NSA when SA isn't available, and always check which version you're pulling.
When seasonal adjustment fails. Three scenarios where SA can mislead. First, structural breaks in the seasonal pattern — the pandemic fundamentally changed seasonal hiring patterns in hospitality and retail. The SA models have adjusted but the pre-pandemic vs post-pandemic SA series aren't strictly comparable. Second, small samples with unstable patterns — very small metros or specific industries where the seasonal component isn't reliably identifiable. Third, data revisions — SA series are revised periodically as new data improves the seasonal model. A November 2024 SA value you saw last month might read differently today. For backtests, pull fresh data rather than using archived snapshots.
Real-World Example
Carlos Medina reads metro unemployment correctly.
Carlos is tracking unemployment in the Austin metro before acquiring a multifamily property. He pulls BLS LAUS data from FRED.
His first pull shows:
- Austin metro unemployment, November 2025: 3.2%
- Austin metro unemployment, December 2025: 3.6%
- MoM change: +0.4 percentage points
First instinct: "Austin unemployment jumped 12% in a month — something's wrong in the labor market." But then he remembers that metro LAUS is NSA. The December holiday-season layoffs in retail, hospitality, and warehousing produce a predictable seasonal rise every year — it's not a signal about Austin specifically.
He pulls the comparison that actually answers his question:
- Austin metro unemployment, December 2025: 3.6%
- Austin metro unemployment, December 2024: 3.8%
- YoY change: -0.2 percentage points
Now the reading makes sense. Austin's December 2025 is 0.2 points LOWER than December 2024. The labor market is actually tightening year-over-year, not weakening. Stripping seasonality by comparing same-month-to-same-month shows Austin is in better shape now than a year ago — even though the raw MoM move in the SAAR-adjacent NSA data looked alarming.
Had Carlos used national BLS CES employment data (which IS seasonally adjusted at the national level), he could have done MoM directly. But for metro-level questions, NSA + YoY is the correct methodology.
Pros & Cons
- Makes month-over-month and quarter-over-quarter comparisons meaningful
- Standard across federal statistical agencies — consistent convention
- Statistical method (X-13ARIMA-SEATS) is well-documented and peer-reviewed
- Revisions improve the seasonal model over time
- Lets you separate underlying trend from calendar noise
- Not available at every geographic grain — metro LAUS is NSA, metro CES is SA
- Revisions can retroactively change historical SA values
- Structural breaks (pandemic, major policy changes) can confuse the seasonal model for years
- Small-sample series can't reliably calculate seasonal components
- Lay readers often assume every series is SA and get confused by NSA data
Watch Out
- Metro LAUS is NSA, metro CES is SA: BLS publishes two metro-level employment series. CES counts payroll jobs and is seasonally adjusted. LAUS estimates the unemployment rate and is NSA. If you're comparing metro unemployment month-over-month, you're reading calendar noise, not economic signal.
- Always check the footnote: Federal data releases say "Seasonally Adjusted Annual Rate" or "Not Seasonally Adjusted" in the release notes. FRED series names include "SA" or "NSA" for clarity. When in doubt, check.
- Revisions can move history: An SA series you pulled in March might read differently in May after the seasonal model updates. For published work, cite the date of the pull along with the data.
- Structural breaks: Pandemic-era labor and housing seasonality was disrupted. Pre-2020 SA series and post-2022 SA series use different seasonal factors. Long-horizon comparisons need to acknowledge this.
- Don't compound SA: If a series is already published as SAAR (e.g., existing home sales), that's already SA plus annualized. Don't layer another adjustment on top.
Ask an Investor
The Takeaway
Seasonal adjustment is the statistical backbone that makes month-over-month economic data readable. For federal releases, always pull SA when you want to track short-term direction and know which version you have. The classic investor trap is comparing NSA metro unemployment across months — the move is mostly calendar, not economy. When SA isn't available (metro LAUS being the canonical case), use year-over-year comparisons to strip seasonality through same-month matching. Distributed through FRED with every series tagged SA or NSA in the name. Methodology documentation at bls.gov and census.gov.