Contact

GlossaryPublic records, privacy, and data quality

Revised data

Revised data incorporates changes after an earlier release. Learn how corrections, late records, and method changes affect historical comparisons.

Revised data is data that has changed since an earlier release. A revision may add late records, correct an error, change a classification, or incorporate a new method. The revised version should identify the affected information and reason for the change so readers can distinguish updated knowledge from newly occurring activity.

Historical figures can have newer versions

Revisions are part of many published data systems. For example, the Census Bureau explains that each new Population Estimates Program vintage can revise earlier estimates using updated inputs or methods. A newer figure for an earlier date is still an estimate of that earlier population. Census Bureau population-estimates FAQ.

The same version distinction is useful for public-safety summaries. An updated January count published in April is not automatically evidence of April incidents. Its meaning depends on what changed in the January records or calculation.

Example

Hypothetical: A monthly incident total is initially 300. A later review adds eight previously unavailable records and removes three duplicate imports. The revised total becomes 305.

The net change is five, but that number alone hides two different processes. A revision note explaining both additions and removals gives a clearer account than saying only that the total increased. It also helps a reader reconcile an earlier download with the current release.

Common confusion

Revised and provisional describe different things. Revised means a change has occurred relative to an earlier version. Provisional means the current version remains subject to a stated unfinished process. Data can be both revised and provisional.

A revision is not necessarily an error correction. Later records can expand coverage, and a changed method can produce a different result from the same observations. Conversely, a corrected label may leave the numerical value unchanged.

A higher revised figure also does not by itself demonstrate a worsening trend. The comparison must separate changes in recorded activity from changes in coverage, classification, or calculation.

Comparing and citing versions

Use the source's preferred version for the intended analysis, while retaining enough version information to reproduce the result. If a trend is recalculated, apply compatible methods and sources across its periods where possible. Mixing old and new versions without explanation can introduce artificial differences.

A useful revision note states what changed, the periods affected, why it changed, and whether earlier conclusions are affected. When the reason is unknown, say that the values differ between releases instead of assigning an unsupported explanation.

Keep the original release and revision dates distinct. That preserves both when readers first received the information and when its current form became available.

Sources