Seasonally adjusted means that predictable, calendar-driven patterns have been stripped out of economic data so you can see what’s actually changing. Every year, certain things happen on schedule: holiday shopping spikes retail sales in December, tax season boosts demand for accountants in spring, and schools closing in June sends teen unemployment up. Seasonal adjustment removes these expected swings so that a month-to-month or quarter-to-quarter comparison reflects real shifts in the economy, not just the time of year.
Why Raw Numbers Can Be Misleading
Imagine retail sales drop 15% from December to January. That sounds alarming, but it happens every single year because the holiday shopping season ends. The drop tells you nothing about whether the economy is weakening. Raw (unadjusted) data includes all of these recurring patterns baked in, making it nearly impossible to spot genuine trends by comparing one month to the next.
Seasonal adjustment solves this by estimating how much of a change is just the calendar talking and how much is something new. After adjustment, if January retail sales still fall, that decline likely reflects an actual slowdown in consumer spending rather than the predictable post-holiday dip.
What Gets Removed
The “seasonal effects” that statisticians filter out go beyond just weather and holidays. They include any repeating pattern tied to the calendar that shows up in roughly the same quarter or month, with similar size and direction, year after year. Common examples include:
- Weather: Construction hiring surges in spring and drops in winter.
- Holidays: December retail sales, back-to-school spending in August.
- School calendars: Teen employment jumps when school lets out and falls when it resumes.
- Tax deadlines: Accounting and financial services see predictable demand around April.
- Trading days: Some months have more business days than others, which affects output and sales totals.
For retail sales specifically, the U.S. Census Bureau adjusts not only for broad seasonal swings but also for trading-day and holiday differences, since the number of weekdays in a month and the timing of holidays like Easter shift from year to year.
How the Adjustment Works
U.S. government agencies use software called X-13ARIMA-SEATS, developed by the Census Bureau, to perform seasonal adjustment. The basic idea is that the program looks at years of past data, identifies the recurring seasonal pattern, and calculates a “seasonal factor” for each month or quarter. That factor is then used to divide out (or subtract) the seasonal effect from the raw number.
The process relies on moving averages: the software slides a window across several years of data, averaging values to estimate what the seasonal pattern looks like at each point in time. This allows the seasonal factors to evolve gradually. A mild winter, for example, might slowly shift the seasonal pattern for construction employment over several years rather than changing it overnight.
For recent months, the math is trickier. In the middle of a long data series, the software can look at equal amounts of past and future data to calculate a balanced estimate. For the most recent months, there’s no future data yet, so shorter, less precise filters are used. As new months roll in, earlier estimates get revised. The Bureau of Labor Statistics notes that it can take up to five years of additional data before a given month’s seasonally adjusted figure is considered final.
Why Adjusted Numbers Get Revised
If you’ve noticed that jobs reports or GDP figures get quietly updated months later, seasonal adjustment is a big reason. Each time a new month of data is added, the statistical model recalculates its estimate of the seasonal pattern, and that can nudge previously published numbers. At the end of each calendar year, the BLS reestimates seasonal factors using the full additional year of data and revises the previous five years of seasonally adjusted figures.
This is normal and expected. It’s the tradeoff for getting timely data: the first estimate is the best available with incomplete information, and it improves as more data arrives.
Seasonal Adjustment and GDP
When you hear that GDP grew at a 2.4% annual rate last quarter, two separate adjustments have been applied. First, the quarterly data is seasonally adjusted to strip out calendar effects. Then it’s annualized, meaning the quarterly growth rate is scaled up to show what a full year of growth at that pace would look like. Annualizing is only done after seasonal effects are removed, because projecting a raw winter quarter forward would give a distorted picture. These are two distinct steps, but they’re almost always reported together, which is why GDP headlines typically say “seasonally adjusted annual rate.”
When Unadjusted Data Is More Useful
Seasonal adjustment isn’t always the right choice. If you want to know what you’re actually paying at the grocery store right now, the unadjusted Consumer Price Index is more relevant because it reflects real prices in real time, seasonal swings and all. The BLS specifically recommends unadjusted data for contracts and pension plans that tie payments to the CPI, because seasonally adjusted series get revised each year, which would create accounting headaches. Many collective bargaining agreements and pension plans use unadjusted CPI for exactly this reason.
Year-over-year comparisons (January 2025 vs. January 2024) are another case where you don’t need seasonal adjustment, since you’re already comparing the same season to itself. The seasonal effects cancel out naturally.
How the Pandemic Stressed the System
Seasonal adjustment models assume that patterns repeat in a fairly stable way. The COVID-19 pandemic broke that assumption. Employment didn’t just dip in spring 2020; it collapsed in ways no seasonal pattern had ever captured. These extreme values, called outliers, threatened to distort seasonal factors for years if left uncorrected.
Starting in 2020, BLS staff began monitoring household survey data in real time and modifying seasonal adjustment models to flag pandemic-related outliers so they wouldn’t warp the underlying seasonal estimates. The vast majority of data series required some form of correction. These modifications have been updated annually since, most recently in January 2024, as agencies continue refining how their models handle that unprecedented period.
Reading Seasonally Adjusted Data in the News
When a headline says the economy added 250,000 jobs last month, that’s a seasonally adjusted number. It doesn’t mean exactly 250,000 people started new positions. It means that after accounting for the hiring and layoff patterns typical for that month, the net change works out to 250,000 more than the seasonal pattern alone would explain. The actual raw number could be higher or lower depending on the time of year.
This is why economists focus on seasonally adjusted figures for spotting turning points. A rise in unemployment that shows up in adjusted data is more meaningful than one that appears only in raw data during a month when unemployment always rises. The adjustment acts as a filter, letting the signal through while muting the predictable noise of the calendar.