Government statistics can strengthen a news story, but only when the number is understood in context. This guide offers a repeatable process for finding the original release, checking what a figure measures, distinguishing counts from rates and estimates, reviewing uncertainty and revisions, and documenting the result before publication.
Overview
An official statistic is not automatically self-explanatory or final. A number may describe a specific population, geography, time period, or definition that differs from the claim being made. It may also be preliminary, revised, seasonally adjusted, survey-based, administratively recorded, or produced through a statistical model.
The central question is not simply, “Did a government office publish this?” It is, “What exactly does this number measure, how was it produced, and does it support the wording of my story?” That distinction matters in local reporting, public safety coverage, election news updates, economic news for families, and current events explained for a general audience.
Before using a figure, trace it to the original source whenever possible. A news release, dashboard, social post, or secondary article may summarize the data accurately, but the underlying table or methodology usually contains the definitions and limitations needed for responsible reporting. Save the source URL, publication date, table name, download date, and relevant notes in the story file.
For broader reporting habits, see How to Build a Weekly Local News Briefing You Can Trust and How to Verify Breaking News Before You Share It.
How to estimate
Use this six-step workflow when checking government statistics before publication.
- Locate the original release or dataset. Search for the publishing office, release title, table number, methodology page, and release date. Prefer a primary source over a screenshot or unattributed graphic. If a figure came from a public meeting, ask where the presenter obtained it.
- Record the exact figure and unit. Copy the number as displayed, including its unit: people, households, dollars, percentage points, incidents, permits, or another measure. Note whether it is a total, average, median, index value, rate, ratio, or percentage.
- Check the population and geography. Identify who or what is included. A city figure may exclude unincorporated areas; a county figure may include several municipalities. A household survey may not cover institutions, people without a fixed address, or other groups. Never assume that a labeled “population” includes everyone readers may imagine.
- Check the time period. Establish whether the figure covers a day, month, quarter, calendar year, fiscal year, rolling period, or date of publication. Compare like periods. A year-to-date number should not be presented as a full-year result, and a monthly change should not automatically be described as a long-term trend.
- Review the status of the data. Look for labels such as preliminary, revised, final, estimated, provisional, modeled, or seasonally adjusted. Read the notes about revisions and missing submissions. A figure can be official while still subject to later change.
- Test the claim against related measures. Compare the number with an earlier period, a neighboring geography, a denominator, or a related series. This is not a search for a preferred conclusion. It is a check for mismatched definitions, unit errors, sudden breaks in reporting, or a numerator that lacks context.
When calculating a rate, write the formula before entering values: rate = events ÷ population at risk × multiplier. The multiplier might be 100, 1,000, 10,000, or another stated base. Confirm that the numerator and denominator describe the same place and period. A percentage change also requires care: percentage change = (new value − old value) ÷ old value × 100. Do not confuse a percentage-point change with a percent change.
Inputs and assumptions
A short source audit can prevent most avoidable errors. Before publication, answer these questions:
- What office or public body produced the data?
- What is the original release date, reference period, and latest revision date?
- What population, geography, and category definitions apply?
- Is the figure a count, rate, percentage, estimate, index, or modeled result?
- Are there exclusions, missing records, reporting delays, or changes in methodology?
- Does the dataset provide a margin of error, confidence interval, or reliability warning?
- Does the headline wording make a stronger claim than the data supports?
- Can a reader open the source and reproduce the basic comparison?
| Data type | What it means | What to check |
|---|---|---|
| Preliminary data | An early release that may be incomplete or changed later. | Revision schedule, missing reports, and whether the story labels it preliminary. |
| Revised data | An updated value incorporating corrections or additional records. | What changed, when it changed, and whether earlier stories need correction. |
| Survey estimate | An estimate drawn from responses by a sample of the target population. | Sample design, response issues, margin of error, and population coverage. |
| Administrative count | Records collected through a program, transaction, filing, or service. | Who must report, who may be missing, duplicate records, and definition changes. |
| Modeled figure | A calculated estimate based on observed data and stated assumptions. | Model inputs, uncertainty range, assumptions, and sensitivity to new data. |
Several terms deserve plain-language treatment. A count is a number of events or entities. A rate relates events to a population or exposure base. A percentage expresses a part of a whole, while a percentage point describes the difference between two percentages. A margin of error describes sampling uncertainty under stated conditions; it does not account for every possible error in collection or definition. A revision is a later change to a previously published value, not necessarily evidence that the original release was improper.
Worked examples
Example 1: A local public safety count. Suppose a dashboard lists incidents for a county and a story says incidents “rose sharply.” First determine whether the dashboard counts reports, arrests, victims, or separate events. Check whether the latest month is complete and whether the county boundary matches the comparison period. Then compare the current count with the same month in the prior year if seasonality may matter. If the number of reports increased but the population or reporting practices also changed, the story should explain that context rather than treating the raw count as a direct measure of risk.
Example 2: A household survey estimate. Suppose a survey reports that a share of households experienced a condition. Confirm the survey population, reference period, sample size, and margin of error. If two estimates are close and their uncertainty ranges overlap, avoid declaring a meaningful difference without support from the source methodology. Use wording such as “the estimates were similar” when the data does not establish a clear change.
Example 3: A cost comparison. If a story compares average prices or income across places, identify whether the figures are nominal or adjusted, monthly or annual, and average or median. Averages can be influenced by a small number of unusually high or low values. Pair the statistic with its unit and period, and avoid converting a broad index into a claim about every household. For related context, see How to Compare Cost-of-Living News With Your Household Budget.
When to recalculate
Revisit a statistic whenever the source publishes a revision, changes its methodology, replaces a dashboard, updates a release calendar, or moves data to a new reporting system. Recheck recurring stories on a set schedule rather than assuming an old link remains complete. Annual audits are useful for evergreen explainers, while fast-moving beats may require checks with every new release.
Recalculate a comparison when the underlying period changes, a denominator is updated, a geographic boundary is redrawn, a category definition is revised, or a benchmark or rate moves. If a source labels data provisional, add a reminder to review it after the expected revision date. Keep both the original and updated values when a change affects a published conclusion.
Before pressing publish, complete this final action list: open the primary source, save the relevant table, write down the definitions, verify the time period, identify uncertainty or revision status, reproduce any calculation, and make the story’s wording no stronger than the evidence. Link to the source and state important limitations in plain language. This approach makes statistics easier for readers to trust and gives editors a clear record when the numbers change.
