Seasonal adjustment: compare compatible series variants
Seasonally adjusted and unadjusted series answer different comparisons. Match the adjustment status, reference period and vintage before interpreting a difference as conflicting evidence. A month-to-month move in one variant should not be directly compared with a year-over-year move in another and described as an economic contradiction.
Understand what adjustment is doing
Recurring calendar patterns can obscure short-term changes. Seasonal adjustment estimates and removes a seasonal component to help analyze movements beyond that recurring pattern. It does not remove every temporary shock or prove the cause of a move. The unadjusted series remains useful for questions tied to actual observed price levels or specified contractual conventions.
The US Bureau of Labor Statistics CPI questions and answers explains the distinction between seasonally adjusted and unadjusted CPI and discusses revisions to seasonal factors. Use the exact official series definition for the question at hand. This article’s numerical example is hypothetical and does not report current inflation data.
A hypothetical apparent disagreement
Imagine a price index has an unadjusted January level of 110 and February level of 112.2, a 2% increase. The seasonally adjusted levels are 111 and 111.555, a 0.5% increase. Both can be valid variants of the same underlying information. The difference reflects the adjustment process rather than an arithmetic error or proof that one source is misleading.
| Field | Unadjusted example | Adjusted example |
|---|---|---|
| January | 110 | 111 |
| February | 112.2 | 111.555 |
| Monthly change | 2% | 0.5% |
| Comparison answered | Observed index movement | Movement after estimated seasonal adjustment |
Do not subtract the two percentage changes and label the result a precisely measured seasonal cause. Adjustment methods and index construction matter, and the variants are related estimates rather than independent experiments. The example demonstrates compatible definitions, not a method for attributing inflation to individual mechanisms.
Build a compatibility record
For each series, record its identifier, population, geography, basket or measure, adjustment status, frequency, index reference base and release vintage. Check whether the statistic is a level, month-over-month change, annualized rate or year-over-year change. Similar headlines can refer to different transformations of different series.
Rebasing an index changes its scale but should not by itself change percentage movements within the same coherent series. Mixing levels from differently based indexes can create a spurious break. Use one consistent series for the calculation rather than joining whichever values are easiest to find in separate articles.
Treat revisions as part of the record
An adjusted historical value may change when seasonal factors are updated. Preserve the vintage used for a memo and distinguish the originally available number from a later revised estimate. A chart built today may therefore differ from a memo written at the time without either having fabricated data.
When assessing a historical decision, use the information that was actually available at the decision date if the research design requires that perspective. For a current economic description, the latest consistent revised series may be preferable. State the choice; using different vintages selectively across periods can produce a comparison that nobody could reproduce from one release.
Match the series to the claim
For short-term momentum, an adjusted series can help reduce predictable seasonal variation. For a contract, follow the specified index and adjustment convention rather than substituting your preferred variant. For annual comparisons, still check definitions and revisions instead of assuming a twelve-month interval solves every seasonal issue.
An AI-generated summary should repeat the adjustment label and transformation alongside the figure. If two sources disagree, first reconstruct those fields. Only then investigate genuine numerical differences within an otherwise compatible definition. The analytical gain is a precise comparison, not an automatic preference for whichever number supports the thesis.
FAQ
Is the adjusted series more accurate?
It serves a different analytical purpose and includes estimated seasonal treatment. Suitability depends on the claim, not a universal accuracy ranking.
Can I compare an adjusted monthly rate with an unadjusted annual rate?
You may discuss both with explicit labels, but they are different statistics. Their different movements do not alone establish a conflict.