Media framing
Walk one claim from strong frame to supportable sentence.
Organization byline: Measurement Press editorial desk. This page does not list an individual author or reviewer. Read the editorial method and corrections policy.
About 11 min read
Checklists are useful, but readers also need to see the habit in motion. This worked example takes one neutral claim through the source trail, denominator, comparison class, and final wording a careful reader can defend.
Picture an editor, analyst, or communications lead reviewing a public-program update minutes before it is cited in a report. The number looks useful, the timing looks plausible, and the headline sounds like progress. The risk is that one true count gets asked to prove reach, cause, and access before the source trail is strong enough.
The example below uses a local-program claim because the pattern is common across public records, surveys, research reports, campaign readouts, and dashboards: a true number is made to sound like a stronger conclusion than the evidence can yet support.
The starting claim
Strong frame"Applications jumped 42% after the agency launched a new online form, showing the portal expanded access."
The sentence may point toward a real improvement, but it asks readers to accept more than one claim at once. It contains a descriptive claim, a timing claim, a causal claim, and an access claim. Each needs a different source standard.
Split the sentence before checking sources
| Part of the sentence | Claim type | Evidence needed | First downgrade risk |
|---|---|---|---|
| "Applications jumped 42%" | Descriptive count or rate change. | Original intake table, dates, revision status, and count definition. | The numerator may be true while the base period is selective. |
| "after the agency launched" | Timeline claim. | Launch date, reporting window, intake cutoff, and prior trend. | Sequence may be mistaken for cause. |
| "showing the portal expanded access" | Causal and distributional claim. | Eligible population, applicant mix, completion rates, outreach, and comparison group. | Higher volume may reflect a deadline, outreach, duplicate starts, or easier counting. |
Run five gates before choosing the verb
The audit is easier when each gate has a pass condition and a downgrade rule. A downgrade does not mean the story is false. It means the final sentence should stop at the strongest claim the visible record can carry.
| Gate | Question to answer | Pass condition | If it does not pass |
|---|---|---|---|
| Claim split | Which parts are descriptive, comparative, causal, or distributional? | Each part can be checked against a specific record. | Rewrite the headline as separate evidence jobs before continuing. |
| Source origin | Do the citations come from independent records or one repeated announcement? | At least one close record confirms the count and at least one source checks context or limits. | Use "reported" or "said," not "showed" or "proved." |
| Denominator | Which base matches the public conclusion? | The base is visible, stable, and tied to the claim being made. | Name the narrow base, such as recorded applications, starts, or completed forms. |
| Comparison class | Is the observed change unusual compared with a fair reference point? | The comparison uses a prior period, seasonal pattern, similar route, or explicit no-comparison note. | Keep the sentence descriptive and avoid "jumped because" language. |
| Final verb | What verb would still be fair if a skeptical reader opened the records? | The verb matches the weakest unresolved gate. | Downgrade to "is consistent with," "followed," "rose," or "the agency reported." |
Triangulate the source trail
The first pass is not about collecting more links. It is about finding sources with different origins and roles, then deciding which part of the sentence each source can actually support.
Source boundary: The linked AAPOR and Census standards were checked for this edition. They support the inspection steps below—method disclosure, assumptions, uncertainty, comparison, and review—but they do not turn this fictional agency example into evidence about a real program.
| Source role | What it can show | What it cannot show alone | How it changes confidence |
|---|---|---|---|
| Agency announcement | The launch date, intended goal, and headline count the agency wants readers to notice. | Whether the count is independently verified or causally explained. | Useful for the claim origin, weak as proof. |
| Monthly intake table | Applications recorded by month, accepted definitions, and whether March-to-April changed from 1,000 to 1,420. | Whether the new form caused the change or improved access for the intended population. | Strong for the descriptive count if definitions and revisions are stable. |
| Program operations note | Form launch date, eligibility changes, staffing changes, duplicate handling, and deadline effects. | Population-level access impact. | Can narrow alternate explanations before the final verb is chosen. |
| Community partner or applicant support group | Whether people had practical trouble applying, whether outreach changed, and which groups still faced barriers. | Full volume, causal lift, or representative applicant mix without supporting data. | Adds counter-context and prevents the agency source from carrying the access claim alone. |
| Prior-year or neighboring-period record | Seasonality, deadline cycles, and whether April normally rises. | Individual-level access improvement. | Helps decide whether the 42% change is exceptional or expected. |
Choose the denominator
A 42% increase can be accurate and still ambiguous. The denominator decides whether the story is about processing volume, public reach, eligible-population access, or completed applications.
| Denominator | Better question | What the answer would support |
|---|---|---|
| Prior-month applications | Did recorded applications rise from one reporting month to the next? | "Applications recorded by the agency rose 42% month over month." |
| Same month last year | Is this increase larger than the normal seasonal pattern? | "Applications were above, below, or near the prior-year April level." |
| Eligible population | Did a larger share of eligible people apply? | "Application reach among eligible residents increased," if eligibility and population counts are visible. |
| Started versus completed forms | Did the portal reduce drop-off or only increase starts? | "The form improved completion," only if completion rate and duplicate handling are available. |
| Applicant mix | Did underserved groups make up a larger share of completed applications? | "Access improved for specific groups," only with subgroup bases and stable definitions. |
Test three fairer comparisons
Comparison 1: prior periodMarch to April shows the immediate change, but it is the weakest comparison if March was unusually low or April usually rises.
Comparison 2: historical seasonalityApril versus prior Aprils checks whether the observed increase is larger than the normal cycle for the program.
Comparison 3: operational counterfactualIf similar programs, similar offices, or offline application routes did not rise in the same way, the portal claim gets stronger. If they rose too, the portal claim should be narrower.
Rewrite by evidence level
| Visible evidence | Do not write | Supportable wording |
|---|---|---|
| Only the announcement and the monthly count are visible. | "The portal expanded access." | "The agency reported 1,420 applications in April, up from 1,000 in March, after the portal launched." |
| The intake table confirms stable definitions and no major revisions. | "The portal caused applications to rise." | "Recorded applications rose 42% from March to April under the agency's published intake definition." |
| Prior-year data shows April usually rises by 15% to 20%. | "Applications surged because of the portal." | "Applications rose more than the usual April increase, though the record does not isolate the portal from outreach, deadlines, or other changes." |
| Operations notes show no eligibility change, no deadline shift, and lower form drop-off. | "The portal solved access barriers." | "The available records are consistent with the portal making completed applications easier, but subgroup access still depends on applicant-mix evidence." |
| Eligible-population and applicant-mix data show higher completion among previously underrepresented groups. | "Access expanded for everyone." | "The strongest visible evidence supports a narrower access claim for the groups whose eligibility and completion rates are documented." |
The final reader note
A clean version can keep the useful news without overstating the conclusion:
Careful sentence"The agency's published intake table shows applications rose from 1,000 in March to 1,420 in April, the first month after the online form launched. That is consistent with easier application flow, but access claims still need eligible-population, completion-rate, and applicant-mix evidence."
This wording does three things at once: it preserves the observed increase, names the timing, and blocks the causal and access claims from becoming stronger than the current source trail.
Method boundary: The Census Bureau E1 analysis standard calls for an analysis plan covering the question, data, methods, assumptions, limitations, and computation checks. The National Academies and Federal Judicial Center Reference Manual on Scientific Evidence is a broader reference point for statistics and research methods. These method sources help define what a reader should request; they are not sources for the fictional applications, rates, or portal claim.
Simulated Worked Example: expose the arithmetic
This is a simulated worked example, not an agency result or Measurement Press case study. Assume the records below are fictional and definitions stay constant. The point is to show how the same count can support one narrow statement while failing a broader access claim.
| Input | March | April | Calculation or decision meaning |
|---|---|---|---|
| Recorded applications | 1,000 | 1,420 | (1,420 - 1,000) / 1,000 = 420 / 1,000 = 42% growth. |
| Completed applications | 680 | 852 | Completed count rose 172, but the completion rate moved from 680 / 1,000 = 68% to 852 / 1,420 = 60%. |
| Eligible population | 10,000 | 10,000 | Recorded reach moved from 1,000 / 10,000 = 10.0% to 1,420 / 10,000 = 14.2%; this still does not identify who gained access or why. |
The simulated arithmetic supports “recorded applications rose 42%.” It does not, by itself, support “the portal expanded access”: the completion share fell in the example, and no applicant-mix, deadline, outreach, duplicate, or comparison evidence has been supplied. A reader should ask for those missing records before accepting a causal or distributional verb.
Source notes: AAPOR’s disclosure standards identify the sponsor, data-collection strategy, population, sample, field dates, processing, precision, and limitations as useful minimum context for publicly released survey findings; apply those items when a claim relies on survey evidence, not as a claim that this fictional program used a survey. The U.S. Census Bureau’s S2 documentation standard provides a practical benchmark for retaining definitions, data, analysis plans, code, and quality measures needed to evaluate or reproduce results. Its E3 review standard emphasizes logical presentation, source statements, accurate calculations, uncertainty, and conclusions that follow from the data. The National Academies distinguishes computational reproducibility using the same data and code from replication using new data and similar methods; neither turns this simulated arithmetic into evidence about a real agency or portal. See AAPOR disclosure standards and the National Academies reproducibility highlights. These references are method sources, not sources for the fictional numbers.
Claim decision packet: publish the count, hold the explanation
Simulated release decision: publish the narrow descriptive change only; hold causal and population-level access wording until the missing records are available. This packet applies the same fictional inputs above one gate at a time, so a reader can see exactly why the wording stops where it does.
| Gate | Input in this simulation | Status | Reader-facing consequence |
|---|---|---|---|
| Definition stability | March and April use the same recorded-application definition. | Pass, assumed | "Recorded applications rose 42%" is permitted, subject to checking version history and revisions. |
| Count to completion | Completion moved from 680 / 1,000 = 68% to 852 / 1,420 = 60%. | Hold | Do not translate more starts into better access; request the full funnel and duplicate-handling rule. |
| Fair comparison | Only March and April are supplied; a prior-year seasonal pattern is not actually supplied. | Hold | Name the timing, but do not call the increase unusual or attribute it to the portal. |
| Distributional access | The eligible base is held at 10,000, but applicant mix and subgroup completion are absent. | Hold | Do not say access expanded "for everyone" or for any subgroup. |
Stress-test the conclusion before publishing it
These are deliberately transparent sensitivity cases, not observed facts. Change one assumption at a time and keep the original inputs visible. The decision should remain descriptive whenever a plausible change leaves the causal or access explanation unresolved.
| Sensitivity case | Reproducible calculation | Result | What changes |
|---|---|---|---|
| April normally rises 20% | Expected April = 1,000 x 1.20; observed excess = 1,420 - 1,200; excess rate = 220 / 1,200. | 1,200 expected; 18.3% above expected | The increase may be larger than a seasonal baseline, but the comparison still does not isolate the portal. |
| 5% of April records are duplicates | Adjusted April = 1,420 x 0.95; adjusted growth = (1,349 - 1,000) / 1,000. | 1,349; 34.9% growth | The count remains higher, but a duplicate audit is required before using it as reach or access evidence. |
| Completion held at March's 68% | Expected completed April = 1,420 x 0.68 = 965.6, or about 966; observed shortfall = 966 - 852. | About 114 fewer completed applications | The higher start count could coexist with a weaker completion outcome, so the access claim becomes less supportable. |
Reproduction rule: copy the three fictional inputs from the arithmetic table, apply each displayed formula without rounding intermediate values, and replace the illustrative 20% and 5% assumptions with observed records when available. A real release would also retain the definition version, duplicate logic, comparison window, eligible-population frame, and subgroup denominators. Until those artifacts are supplied, the defensible sentence remains the narrow one above.
Reusable audit pattern
- Write the headline claim as one sentence, then split it into descriptive, comparative, causal, and distributional parts.
- Find the closest record for the number before weighing quotes, summaries, or reactions.
- Ask whether several citations are independent or only repeating one announcement, report, or dashboard.
- Choose the denominator that matches the strongest public conclusion, not the one that makes the number easiest to notice.
- Pick the final verb after the source trail is complete: reported, rose, suggests, is consistent with, or shows.
Use with
Pair this example with the source triangulation checklist to separate same-origin citations from independent confirmation, the denominator framing examples to choose the right base, the causal claim review protocol when the sentence implies cause, the evidence-to-claim language matrix to choose the final verb, and the media claim audit worksheet when a full article needs review.