Features
Part of Statistics and polls: a reporting guide to estimates and uncertainty
Seasonal data desk case: explaining a revised jobs headline
Seasonal data desk case showing how a fictional newsroom separates raw and adjusted jobs figures, calculates changes, checks revisions, and fixes a headline.
What to take away
- This fictional case demonstrates a time-series editing method.
- Raw and seasonally adjusted values answer different questions.
- A month-to-month headline needs a consistent series.
- Later seasonal revisions can change earlier adjusted values.
- The correction explains why the first comparison failed.
The fictional North County Brief receives a local labor table. Employment rises from 48,000 in December to 50,400 in January before seasonal adjustment. The seasonally adjusted series moves from 49,700 to 49,900. A draft headline says, "County jobs jump 5 percent in January."
The arithmetic for the raw series is correct: 2,400 divided by 48,000 equals 5 percent. The interpretation is not ready. It is the one-year-trend problem from statistics reporting problems wearing a seasonal coat.
Label both series
| Series | December | January | Change | Primary use |
|---|---|---|---|---|
| Not seasonally adjusted | 48,000 | 50,400 | +5.0% | Levels before removing recurring calendar patterns |
| Seasonally adjusted | 49,700 | 49,900 | +0.4% | Short-term movement after estimated seasonal effects |
The county routinely adds temporary workers in January for an annual event. Comparing adjacent raw months mixes that recurring pattern with other movement.
Check the source's guidance
The U.S. Bureau of Labor Statistics explains that seasonally adjusted CPI changes are generally preferred for short-term trend analysis because the adjustment removes recurring calendar effects, while unadjusted data matter for what consumers paid and for some escalation uses. The BLS guide to adjusted and unadjusted data supports choosing a series by question. It discusses CPI, so the fictional desk also checks the documentation for its local jobs series rather than transferring every detail.
The method principle is simple: use a consistently adjusted series for adjacent-period trend language when the publisher supplies one for that purpose. Use raw levels when describing observed totals, and compare the same calendar month across years if that fits the question. Series selection questions belong in the statistical claim checklist pass.
Rewrite the headline and lead
The desk changes the headline to:
Adjusted county employment edges up 0.4 percent in January
The lead says:
North County employment rose by an estimated 200 jobs from December after seasonal adjustment, a 0.4 percent increase. The unadjusted count rose by 2,400, including recurring January hiring that the adjustment is designed to account for.
The text does not say seasonal adjustment identifies the exact number of temporary event workers. It describes the purpose of the published series.
Inspect uncertainty and precision
The values are estimates. The newsroom seeks the program's sampling-error and revision documentation before calling a 200-job change meaningful. If the source does not support a claim of statistical difference, the article avoids "surge," "drop," or "turnaround." Restraint with directional verbs is a rule carried over from reporting statistics and polls.
It rounds percentages consistently and gives counts so readers can see scale.
Add the revision rule
Seasonal factors can be recalculated when more observations become available. BLS describes a Current Population Survey process in which seasonal factors are reestimated annually and several years of adjusted history can be revised, while noting that the underlying original sample data normally are not revised. The BLS labor-force seasonal adjustment methodology supports distinguishing revision to adjustment from revision to raw observations. The fictional local program may use another schedule, which the desk must report from its own documentation.
The article adds the data vintage and next scheduled revision date.
A later revision
Two months later, the source revises the adjusted December value from 49,700 to 49,760 and January from 49,900 to 49,880. The new change is 120 jobs, about 0.2 percent.
The desk recalculates the lead and chart. Because the original 0.4 percent figure was accurate for its stated vintage and the conclusion remains a small rise, the update note says the source revised seasonal factors and provides the new values. If the direction had reversed, the desk would publish a correction explaining the earlier and revised findings. Vintage-aware updates are also how the record version desk case handles a moving total.
Desk protocol
- Name whether each figure is adjusted or unadjusted.
- Compare values from the same series and vintage.
- Use short-term trend language only with suitable treatment of seasonality.
- Check estimate uncertainty before using directional verbs.
- Record revision schedules and download dates.
- Recalculate all dependent prose and graphics after revision.
Common questions
Does seasonal adjustment change the observed raw count?
No. It creates a separate series designed to remove estimated recurring seasonal effects. Keep the two labels distinct.
Is the adjusted series always better?
No. It is often suited to short-term trend analysis. Raw values may be the right answer for observed levels, contracts, budgets, or same-month comparisons.
Is a data revision an admission of error?
Not necessarily. Many statistical programs schedule revisions as new reports, benchmarks, or seasonal information arrive.
Why state the data vintage?
It tells readers which published version supported the article and makes later changes traceable.
