Europe’s book-reading statistics: reach and frequency need different denominators

Which reading question are you trying to answer?

A percentage about reading needs an audience as well as a number. A library planning to reach people who completed no books faces a different question from a reading group asking how often its existing readers finish books. The first question concerns participation across the source population. The second concerns the composition of the group who reported reading. A statistic can be useful for one purpose and misleading for the other if its denominator is left unstated.

This analysis separates those questions using Eurostat’s 2022 book-reading estimates. It reports the share of people who completed at least one book, the share who completed ten or more, and the derived ten-plus share among those who completed at least one. The country table and Portugal-focused appendix preserve the underlying categories and source flags. The aim is practical: help readers recognize which group a percentage describes before comparing countries, demographic groups or a quoted headline.

What the survey counts as a book

The source population is persons aged 16 or over living in private households. It includes 16- and 17-year-olds, so an unqualified reference to “adults” would narrow the stated population incorrectly. Institutions and other collective households are generally outside the EU-SILC household coverage. The country comparisons describe this survey population rather than every resident, every child, every library member or every customer in the book market.

The intended question counts completed books during the preceding twelve months. The official 2022 module assessment includes ebooks and audiobooks, as well as other book formats, and covers all book genres. It excludes podcasts, schoolbooks, work manuals, magazines and journals. A person who started a book without completing it is not thereby counted as having completed that book. Keeping the completion requirement visible prevents the results from becoming a broader claim about any contact with reading.

The categories are no books, one to four, five to nine, and ten or more. The source code LT5 labels the positive one-to-four category in this mutually exclusive scheme; it must not be interpreted as also containing the separately recorded zero category. Counts do not standardize page length or audiobook duration. Completing ten books is a count threshold, not a measure of hours, attention, difficulty, literary quality, reading speed or the number of novels read.

One EU numerator, two useful percentages

The supplied estimated EU aggregate offers a clear worked example. Its positive categories are 26.8% for one to four books, 12.2% for five to nine and 13.8% for ten or more. Adding those categories gives 52.8% reporting at least one completed book. This is the reach measure: the share of the source population who reported completing any book in the preceding twelve months. It does not distinguish purchasing from borrowing or identify a preferred format.

The 13.8% ten-plus figure uses that same population denominator. To ask what share of readers completed ten or more books, the analysis divides 13.8 by 52.8 and multiplies by 100. The result is approximately 26.1%, with the unrounded derived value retained in the download. The numerator has not changed. The denominator now consists of people reporting at least one completed book. The larger percentage therefore does not mean that more people have been discovered or that reading increased.

The EU row carries the source flag e, meaning estimated. It is a supplied Eurostat aggregate, not a mean calculated from the countries in our chart. The separate euro-area row is likewise retained as a supplied reference. Averaging the displayed country percentages would answer another question and would not reproduce an EU population statistic. The reference rows remain distinct from the 32 eligible country entities shown in the primary figures.

A country can look different when the denominator changes

The scatter figure places population reach on one axis and the ten-plus share among readers on the other. These are deliberately different denominators, with both named on the axes. The paired-bar figure asks a separate question: it places population reach beside population ten-plus prevalence, using the same population denominator for both bars. Reading the figures together helps distinguish how widely book completion is reported from how the ten-plus category is represented within the reader group.

Switzerland and Ireland illustrate the distinction without requiring a league table. Switzerland’s primary values imply reach of 80.6% and a ten-plus share among readers of approximately 35.5%. Ireland’s imply reach of 66.6% and a ten-plus share among readers of approximately 39.5%. A higher reach percentage therefore does not automatically accompany a higher conditional percentage. These are descriptive point estimates, not evidence that one population reads better or that the difference between the countries is statistically established.

The main table contains 35 rows: 33 listed country entities and two supplied aggregates. Thirty-two country entities meet the primary completeness and flag rules. Germany remains in the table with missing primary values and the raw u flag, which the source defines as low reliability. It is excluded from the figures, not assigned zero. A country absent from the source dictionary is not added with an invented value. The coverage is the available source set, not a census of all Europe or a worldwide ranking.

Portugal shows why a sex difference needs a named audience

The approved appendix compares female and male estimates within compatible age and education categories for Portugal, Spain and the supplied EU aggregate. For Portugal’s overall persons-16-plus reference row, female reach is 49.4% and male reach is 33.3%. The female-minus-male difference is 16.1 percentage points. Each percentage uses its corresponding female or male source population as the denominator, rather than all people in Portugal pooled together.

Among those reporting at least one completed book, Portugal’s derived ten-plus shares are approximately 12.1457% for females and 11.7117% for males. The difference is approximately 0.4340 percentage points, displayed as 0.4 points. This contrasts two reader groups, whereas the 16.1-point reach difference contrasts participation across the two source populations. Both calculations can be correct because they answer different questions. The conditional comparison does not overturn or erase the reach comparison.

It would be misleading to describe the Portuguese difference as “shrinking” from 16.1 to 0.4 points without explaining that the denominator changed. These are not two dates, a before-and-after intervention or an adjusted estimate of the same effect. Nor does the smaller point difference establish statistical equality between female and male readers. The aggregate extraction does not supply the joint sampling uncertainty needed for such a conclusion. The result is a reason to name the audience, not a causal explanation for reading patterns.

Spain and the full appendix keep the example in context

Spain’s corresponding female-minus-male reach difference is 14.3 percentage points. Its derived ten-plus difference among readers is approximately 2.2413 points, displayed as 2.2. This provides a second example of how the descriptive contrast changes when the denominator changes. It does not make Spain a control group for Portugal, establish a statistically significant cross-country difference, or isolate an effect of education, age, policy or access to books.

The appendix retains 204 geography-by-age-by-education combinations and 1,632 source category cells. Portugal contributes 67 unflagged rows and one flagged numeric row. Spain contributes 68 unflagged rows. The supplied EU aggregate contributes 66 flagged numeric rows and two missing or invalid rows excluded from calculation. These status distinctions remain visible. A row that is numerically computable is not automatically included in the primary unflagged comparison, and a missing cell is never changed into a zero.

The age groups overlap. For example, people aged 16–29 also fall within the broader 16–44 category, so the two rows do not represent independent sets of people. Education totals likewise sit alongside more detailed education categories. The 204 rows cannot be added together to produce a sample size, pooled as independent observations or used to create an age-adjusted sex difference without the appropriate population weights and methods. The overall examples introduce the appendix; they do not summarize every stratum or replace its source flags.

Why Ireland has 66.6% reach in this table

The source publishes category estimates in steps of one tenth of a percentage point. Once categories are rounded separately, adding the three positive categories can differ slightly from subtracting the no-book category from 100. The analysis declares one rule and uses it consistently: reach is the sum of one-to-four, five-to-nine and ten-plus. It does not switch between formulas by country or alter a category to make the displayed total equal exactly 100.

Ireland is a useful check. Its four published categories sum to 100.1%. Adding the positive categories gives 66.6% reach, while subtracting its 33.5% no-book value from 100 gives 66.5%. The difference is disclosed rather than silently corrected. It is a consequence of the rounded inputs in these calculations, not evidence of two incompatible populations. The conditional ratio in the country table uses the declared 66.6% positive-category denominator.

The machine-readable files preserve the calculated ratios, while reader-facing percentages are normally displayed to one decimal place. Additional decimal digits in a derived value do not create additional survey precision. A ratio calculated from rounded estimates is still a derived point estimate, and its uncertainty cannot be recovered merely by printing more digits. When quoting a small percentage-point contrast, retain the calculation rule and avoid using apparent numerical detail to imply a formal statistical result.

Read the flags as part of the evidence

The source dictionary defines u as low reliability, e as estimated and n as not significant. These codes belong to the original data and are preserved independently of whether a numeric value is present. In particular, Germany’s missing primary cells carry u, not n. A missing number with a flag is not an observed zero, and a blank flag field does not prove that the associated estimate is error-free or measured with unusually high precision.

The primary country rules require all four category values, each between zero and 100, with their sum within 0.2 percentage points of 100. They exclude cells carrying u or n. Supplied estimated aggregates are kept as explicitly labeled context rather than country observations. The appendix applies its separate declared rule: any nonempty source flag excludes that stratum from the primary unflagged comparison, while eligible numeric calculations can remain in a clearly marked flagged category.

The label “not significant” is the source’s description of n, not a test conducted in this article. Neither that code nor the absence of a code supplies the covariance needed for new country or sex comparisons. The analysis does not construct confidence intervals for the conditional ratios or claim that a contrast differs from zero. Treating a source flag as a complete answer to a different statistical question would go beyond the evidence.

Compatible arithmetic does not make every questionnaire identical

Eurostat’s frozen 2022 module assessment documents national wording and collection differences on page 92. The Netherlands emphasized leisure time, while Sweden used separate reading and audiobook questions. Other implementations combined or separated question steps, and the assessment notes differences or unclear specifications in some collection-unit filters. Those details matter even when the final published table uses common category labels.

The country matrix therefore offers compatible denominator arithmetic, not proof of identical measurement in every country. An observed difference can be influenced by population composition, the survey process and national implementation as well as the behavior the question intends to capture. The analysis does not separate those contributions. It would be too strong to attribute an entire country gap to enthusiasm for books, a national policy, library provision or a cultural preference.

The dates also need separation. The observations refer to 2022; the frozen dataset’s revision is dated 24 April 2026; retrieval took place on 28 October 2026. These are not measurements of 2026 reading habits. A linked October 2025 note describes revisions to Hungarian EU-SILC demographics and income imputation, but it does not establish that a particular reading estimate changed for that reason. The article uses the frozen data version without inventing a history for individual cells.

Use the measure that matches the decision

For a question about the breadth of book completion across the survey population, use reach and state that it means at least one completed book in the preceding twelve months. For a question about the population prevalence of the ten-plus category, use its original population percentage. For a question about the composition of people who completed any book, use the derived conditional share and say “among readers.” Choosing between these measures should follow the intended audience of the question rather than whichever percentage looks more striking.

A reading organizer can use this distinction to describe the population a project hopes to engage. A journalist can check whether two headlines use the same denominator before presenting them as contradictory. A librarian can distinguish a general participation statistic from a description of existing readers. These are uses of the evidence as context; the data do not estimate the effect of a particular campaign, predict future participation or tell an organization which intervention will work.

The study also offers no direct measure of book purchasing, borrowing, author demand or romance readership. Completing a book does not disclose how it was obtained, what it cost or which genre it belonged to. The combined treatment of formats cannot establish a preference for print, ebooks or audio. A careful summary can explain the population and reader denominators without importing an unrelated commercial or format-specific interpretation.

The contribution is a transparent comparison, with credit to the source

Eurostat has already published book-count summaries and country examples. The frozen 2022 module assessment provides such coverage on page 43, defines the measure on page 14 and discusses implementation on page 92. This article does not claim to discover European reading rates or to be the first to discuss frequent reading. Its contribution is a source-linked comparison that keeps the three measures and their denominators, rounding choices, flags and exclusions together, with a detailed Portugal, Spain and EU appendix.

The reproducible country matrix connects each original category to its source cell index. The accompanying code records the selection and arithmetic, and the semantic tables make the values available without extracting them from an image. The scatter and paired bars retain their different denominator meanings. Source data and definitions are credited to Eurostat; the transformations, interpretation and presentation are original work by Alisson Bento. Eurostat is not responsible for this modified analysis or its conclusions, and no endorsement is implied.

A takeaway with the denominator still visible

In Eurostat’s 2022 estimates for persons aged 16 or over in private households, the supplied estimated EU aggregate is 52.8% reporting at least one completed book and 13.8% reporting ten or more in the preceding twelve months. The latter is approximately 26.1% among people who reported at least one completed book. This is a derived ratio of rounded estimates, not a new survey rate with independently established uncertainty. The data were revised in April 2026 and retrieved in October 2026; the behavior described remains 2022 behavior.

Portugal’s female-minus-male difference is 16.1 percentage points for reach and approximately 0.4 points for the derived ten-plus share among readers; Spain’s corresponding differences are 14.3 and approximately 2.2 points. These are different denominators within the same observation year, not changes over time or tests of equality. National questionnaire differences, missing and flagged cells, and overlapping appendix strata remain part of the interpretation. Retain those qualifications when reusing the numbers.

Data & downloads

Scroll horizontally to see all columns. ↔

2022, persons aged 16+ in private households. The 35 rows include 33 country entities and two separate supplied aggregates; 32 country entities are eligible for primary figures. Germany remains missing/excluded with u. Reach sums the positive categories; Ireland’s 66.6% is not silently changed to 66.5%. Preserve all source flags and indices.

Scroll horizontally to see all columns. ↔

204 overlapping strata, not independent people or an age adjustment. Numerically computable flagged rows remain separate from primary unflagged comparisons. Differences use percentage points; conditional shares use each sex’s reader denominator. No significance or equality claim.

Evidence & method

How we calculated it

1. The recorded analysis plan fixes annual 2022 observations for total sex, age 16+, total education and percent units in Eurostat dataset ilc_scp27. It was written after source acquisition and dimension inspection but before the derived country results; it is not described as prospective preregistration. The source response includes its dimension order, category dictionary and flags.
2. For each geography, reach is LT 5 + 5–9 + GE 10, with LT 5 interpreted as the positive 1–4 category after questionnaire confirmation. Population ten-plus prevalence is GE 10. The conditional ten-plus percentage is 100 × GE 10/reach, calculated only with compatible present values and a positive denominator. The declared positive-bin sum is used even where rounding makes it differ from 100 minus the no-book estimate.
3. Primary country eligibility requires four present categories within 0–100, a total within 0.2 percentage points of 100 and no u or n flag. Supplied EU and euro-area aggregates remain separate contextual references. The Portugal appendix compares female-minus-male estimates within each compatible geography/age/education stratum and separates unflagged, flagged numeric and missing/invalid rows.
4. Independent review decoded all 27,524 published observations and source flag positions, checked all 35 country/aggregate rows and 204 appendix strata, and verified the 32 scatter points and 64 paired-bar lengths. No new weighting, age adjustment, microdata reconstruction, confidence interval or significance test was added. The downloads preserve source-cell references so that the arithmetic can be checked against the frozen inputs.

Original sources

1. Eurostat ilc_scp27 — frozen JSON-stat data and dictionary

   Eurostat dataset ilc_scp27, Persons reading books in the last 12 months by sex, age, educational attainment and number of books; DOI 10.2908/ILC_SCP27. Observations 2022, revision 24 April 2026, retrieved 28 October 2026. The source dictionary supplies categories, geography codes and raw u/e/n flags.

   Accessed October 28, 2026

2. Eurostat EU-SILC metadata

   Eurostat EU-SILC metadata defines the population as persons 16+ in private households; people aged 16–17 are included and institutional coverage is not implied.

   Accessed October 28, 2026

3. Eurostat EU-SILC 2022 modules Assessment

   The 2022 module assessment page 14 defines completed books and formats, page 92 documents national wording and implementation differences. The same official 2022 module assessment page 43 already reports book-count summaries and country examples. These are credited prior source analyses; older published figures are not substituted for the frozen dataset used here.

   Accessed October 28, 2026

4. Eurostat dataset information note, October 2025

   The linked October 2025 dataset information note concerns revised Hungarian demographics and income imputation. It is not evidence that a particular reading estimate changed for a specified reason.

   Accessed October 28, 2026

5. Eurostat Data Browser API documentation

   Eurostat Data Browser API documentation describes access and response structure. The derived downloads retain source indices and flags; this citation is not a new data observation.

   Accessed October 28, 2026

6. Eurostat Copyright notice and free re-use of data

   Eurostat Copyright notice and free re-use of data permits reuse with attribution subject to its stated exceptions. Original modified calculations and Eurostat non-responsibility are identified in the article and figure notes; no logos or third-party illustrations are reused.

   Accessed October 28, 2026

7. Eurostat prior coverage — Younger people and women in the EU read more books

   Eurostat, Younger people and women in the EU read more books, 9 August 2024: prior coverage of age, sex and country reading-reach and book-count headlines is credited. No figure or photograph is reused, and the earlier news values do not replace the frozen 2022 data revised in 2026.

   Accessed October 28, 2026