In-depth practical guides

Read and verify a public report

Find a figure, understand its scope and cite the right passage: a method for reading reports, annexes and tables while retaining their qualifications.

Updated :

Document verification desk with a ledger and company file.

AI-generated illustration

Which conclusion does the document actually support?

A report has several layers: summary, method, findings, annexes and sometimes an associated dataset. The same expression can refer to different populations or periods across chapters. Start with a precise question to select useful passages and avoid treating a summary sentence as a universal measure. Connect each conclusion to evidence, a definition and a limitation.

  1. Identify the correct edition

    Record the complete title, issuing body, reference, publication date and any revision notice. Open the document from its official publication page rather than an isolated copy. A new edition may change a method or correct a table without retaining pagination.

  2. Define the question before extracting

    Write the proposition to check: which indicator, population, territory and period? Then find the corresponding definition. A national total, respondent average and subgroup estimate answer different questions even when their labels look similar.

  3. Read methods and qualifications

    Check collection rules, exclusions, estimates and missing values. Read notes beneath the table and references to annexes. Distinguish observed result, estimate, target and scenario; preserve that qualification in your written summary.

  4. Check tables in context

    Connect every value to its header, unit and denominator. Test a total’s consistency without assuming all categories are mutually exclusive. If PDF extraction mixes columns, compare copied text with the visible layout; seek an official HTML or tabular version.

  5. Build a retrievable reference

    Record the printed page and, if different, its position in the PDF. Add table, paragraph or annex number. Keep a short contextual excerpt and document link; linking only to the issuing organisation’s homepage does not locate the figure.

  6. Write a proportionate conclusion

    State what this version establishes within its scope. Separate cited data from interpretation. If editions differ, explain a verified change or leave its cause undetermined; do not combine figures based on incompatible methods.

Reading workshop

A half depends on which population is counted

The DEMO report and its pagination are fictional. The exercise links a percentage to its base, table note and retrievable passage.

Fictional data and records for learning: none of these results comes from a live service query.

51 positive answers, two denominators

The table lists 120 people approached: 18 with no response and 102 respondents. Of the latter, 51 give a positive answer. The note states that the table percentage excludes nonresponses.

If the table exceeds the screen, scroll horizontally; arrow keys work when the region is focused.

51 positive answers, two denominators
ItemCalculation or locatorPermitted wordingWhat to retain
Table counts120 − 18 = 102 respondentsAnswers cover 102 peopleThe note excludes 18 nonresponses from the denominator.
Share among respondents51 / 102 × 100 = 50%Half the respondents answered positivelyRetain question, population and base of 102.
Share among people approached51 / 120 × 100 = 42.5%42.5% of people approached have a recorded positive answerThis does not reveal what absent respondents would have answered.
Passage locationPrinted page 12; PDF position 16; appendix ACite the table and note in this editionCheck that the appendix uses the same population before comparing.

Which sentence respects the table without widening its scope?

Open each proposition to compare its diagnosis. More checks may still be needed.

Half of the 120 people are satisfied

No. The 50% concerns 102 respondents, not all 120 people approached. “Positive” means “satisfied” only if the question defines it that way.

51 of 102 respondents give a positive answer, or 50%

Yes. Add the 18 nonresponses and the table locator. Keep the scope of the question asked.

The 18 nonresponses count as negative answers

No. A missing answer does not document an opinion. This imputation changes the result’s meaning.

Fictional reading record

Document: DEMO report, exercise edition
Locator: printed page 12, PDF position 16, appendix A
Approached: 120; no response: 18; respondents: 102
Positive answers: 51
Reading: 50% of respondents; 42.5% of people approached
Reservation: nonrespondents’ opinions unknown

PDF position does not replace printed pagination. These fractions describe only the fictional table; they are not a standardized AAPOR response-rate calculation.

Documentation of the rules, separate from fictional examples

Compare other situations and exercises

A summary percentage has no stated base

Find the table and count of valid responses. If the base cannot be identified, quote the percentage with that qualification and avoid deriving a headcount from an assumed population.

An annex changes an initial reading

Check whether passages concern the same year, unit and scope. A note may exclude a category from the total. Describe the scope difference before calling it a contradiction.

The PDF and dataset disagree

Compare versions and dates, then a small set of rows or cells. Preserve both documents and record the discrepancy. The easiest file to process is not automatically the newest.

Worked exercise

More responses than people

Fictional example: the values and documents teach the method and do not describe a real organisation.

The report counts 100 respondents. Instructions permit multiple choices per person; no table of individual selections is supplied.

If the table exceeds the screen, scroll horizontally; arrow keys work when the region is focused.

Starting information
Reason selectedNumber of selections
A40
B35
C30

Your checks

  1. Sum choices rather than people.
  2. Read instructions and each percentage's base.
  3. Determine whether overlap can be calculated.
Read the answer

Answer and decision

There are 105 choices from 100 people. This is consistent with a multiple-choice question. Marginal counts do not reveal how many people chose several reasons or the exact intersections.

If the case changes

If instructions required one choice, investigate consistency instead of assuming multiple responses.

Record to retain

Keep table, instructions, rate bases and the intersection limitation.

Record to retain

  • Complete reference, official URL, edition and consultation date.
  • Question, scope, indicator definition and unit.
  • Page, table or annex, contextual excerpt and relevant qualification.
  • Transformations, version differences and unresolved points.

Official references

Related protocols

Checks before reaching a conclusion

  • Can the cited value be located with its cell, unit, base and qualifying note?
  • Are the PDF edition and the associated dataset edition identified separately when their dates differ?
  • Does the conclusion distinguish observation, estimate and interpretation without expanding the studied population?

Frequently asked questions

Must the entire report be read?

Start with the question, then read sections defining its meaning: method, findings and associated notes. Scan the overall structure to locate decisive annexes. Targeted reading is valid when its scope is explicit.

Can a screenshot be the only citation?

It can help show a table, but add the reference, version and context. It does not replace notes explaining scope, units or exclusions.

How should two years be compared?

First check compatibility of definitions, populations, collection and units. Record breaks in the series. If compatibility is unestablished, present results separately instead of calculating a misleading change.

What if the document disappears?

Keep the reference and last-access date. A dated copy or archive can support historical reading with provenance stated; it does not itself prove that the content remains current.

Should one read all the annexes?

Priorize those that define the data and the method of which the conclusion depends. The relevant annexes may contain the units, exclusions and changes necessary for a proper reading of the used passage.

DOCUMENTED REFERENCE

Choose a digital measure for the question being asked

Similar names do not make indicators interchangeable. The selector compares six questions with the extract’s three telecommunications series. It also identifies questions none of them answers, to avoid turning subscriptions into a measure of equipment or behaviour.

IT.CEL.SETS.P2people / households≠127.47302579722 / 100 · FIN · 2023
Choose a digital measure for the question being asked
ReferenceValue or formWhat it identifies
Internet useIT.NET.USER.ZSTo measure the share of people who recently used the Internet, use the people-based series and its three-month window, retaining collection limits.
Fixed subscriptionsIT.NET.BBND.P2To describe fixed subscriptions relative to population, use the fixed series; subscription counts are not connected-household counts.
Voice cellular subscriptionsIT.CEL.SETS.P2To describe cellular subscriptions offering voice, use the mobile series. Multiple lines per person make a result above 100 possible.
Households / speed / AINot measured by these three seriesHousehold access at home, experienced speed and application use require other measures. Arithmetic on these three codes cannot create missing observations.

Which series answers your question?

Local selections and calculations; no World Bank request during use.

Frozen example displayed; tables remain readable without JavaScript.

IT.NET.USER.ZS

Use IT.NET.USER.ZS and its three-month window. The series concerns people; also read collection notes.

Six questions and available measures
Question to documentAnswer according to the definition
What share of people used the Internet in the last three months?Use IT.NET.USER.ZS and its three-month window. The series concerns people; also read collection notes.
How many fixed broadband subscriptions per 100 people?Use IT.NET.BBND.P2. The ratio includes organisational subscriptions and does not give a household share.
How many cellular voice subscriptions per 100 people?Use IT.CEL.SETS.P2. Multiple subscriptions may belong to one person; 100 is not a percentage ceiling.
What share of households has Internet access at home?None of the three: seek a household-access measure with its reference population.
What speed does a user actually receive?None of the three: a subscription-definition threshold is not an experienced-speed test.
What share of the population uses a particular AI application?None of the three: Internet use does not describe use of a particular application.

Telecommunications indicators originate with the International Telecommunication Union (ITU). Distribution: World Bank, World Development Indicators. Population: WDI sources detailed in the documentation. Extract and presentation calculations prepared for this site.

Update reported by the API : · Extract accessed on :

Data reused under CC BY 4.0.

Implication for the record

The appropriate series depends on population, unit and question. Selector answers concern only these three indicator definitions; they do not rank countries or establish network quality, geographic coverage or a commercial audience.

Primary documentation

Documentation consulted on .

DEEPER READING / PRACTICE

Separate results, method and comment

A public report is being made to support a conclusion. The reading must find the result in its context, examine the method and identify the reserves that accompany its presentation.

Updated on .

Priority checks

Report the organization, version, date, scope and location of the passage. Compare the summary to the table and methodological annexes when the conclusion depends on a figure.

Interpretation pitfalls

A chapter title can simplify a more nuanced situation. A sample or period result does not automatically describe the entire domain.

Working record to retain

The reading sheet retains question, passage, data, method and limits. The final summary distinguishes what the report establishes from the proposed interpretation.