A smaller frequent-reading share, with a measurement break
The share of Switzerland's resident population reporting thirteen or more privately read books was about 19.2% in 2014, 19.7% in 2019 and 14.2% in 2024. Among people reporting at least one privately read book, the corresponding shares were approximately 24.5%, 25.2% and 19.2%. These are two different perspectives on the same published estimates: prevalence in the population and composition within the reported reader category. Both matter when describing what changed in the survey distributions.
The central qualification belongs beside these numbers, not in a distant footnote. The Federal Statistical Office, known by its German abbreviation BFS, changed the survey's collection design in 2024. Online self-completion became the main mode, the fieldwork window shortened, and weighting and questionnaire arrangements were revised. Consequently, the differences described here combine whatever changes occurred in the measured population with possible effects of measurement and survey design. They cannot be attributed wholly to changes in reading behaviour.
This analysis adds a conditional reading share and an explicit arithmetic decomposition to the BFS's already published category chart. It asks whether the lower prevalence of frequent private reading is associated numerically more with a smaller reported reader category or with a smaller frequent-reading fraction inside that category. It does not follow individuals, identify readers who stopped reading, or show frequent readers moving into another group.
What counts as reading in this comparison
The comparison concerns people aged fifteen and over in Switzerland's permanent resident population living in private households. It is not a survey of children, institutions, book buyers, library members or Swiss citizens alone. The national estimates must not be labelled as results for German-speaking Switzerland merely because this account uses the German-language source tables. Language-region comparisons require their own estimates and are not part of the present calculation.
The activity is reading books for private purposes during the twelve months before the interview. Ebooks are included and comics are excluded. Books read for work or education are outside this particular measure. This restriction matters because the source workbook also contains a total-books series covering private and work or education purposes, but that total series ends in 2019. Joining its earlier values to the private-purpose 2024 value would create a change in definition rather than a consistent three-wave comparison.
The recall period is rolling: each person's answer covers the previous twelve months relative to their interview. A survey labelled 2024 therefore does not measure exactly the calendar year's reading for every respondent. Nor does it count purchases, publications, pages or time spent reading. A book can be borrowed, bought, received or accessed electronically; these tables do not distinguish those routes. The private-purpose restriction is an activity definition, not a statement about the ownership of the books.
Keeping the entire distribution visible
The published reader categories are one to three books, four to twelve books, and thirteen or more books. In 2014 their population shares were about 26.0%, 33.2% and 19.2%. In 2019 they were 25.1%, 33.6% and 19.7%. In 2024 they were 29.8%, 29.9% and 14.2%. The categories collectively reproduce the published share reporting at least one book, to the precision of the source cells.
The first interval is relatively stable at this descriptive level. Between 2014 and 2019, the any-book estimate increases by only about 0.06 percentage points, while the thirteen-plus estimate increases by approximately 0.52 points. The one-to-three category decreases by around 0.84 points and the four-to-twelve category increases by 0.37 points. These small point differences are not presented as new statistical tests or as proof of stable behaviour among the same people.
The later interval has a different observed pattern. From 2019 to 2024, one-to-three books rises by about 4.67 percentage points, four-to-twelve falls by 3.70 points, and thirteen-plus falls by 5.55 points. Showing all three prevents the fall in frequent reading from obscuring the simultaneous rise in the lowest positive category. However, simultaneous movements in aggregate categories do not connect their members across surveys. No flow diagram or individual transition count can be recovered from these margins.
Why nonreaders are not simply everyone else
The workbook separately reports people who read no books for private purposes. Their estimated shares are 21.0% in 2014, 21.2% in 2019 and 25.0% in 2024, rounded to one decimal. These numbers are not exactly one hundred minus the corresponding any-book percentages. Treating the complement as a published nonreader estimate would assign unclassified answers to a substantive reading category without evidence.
The source explains that unknown and no-answer responses are not displayed and that rounding can also prevent totals from reaching one hundred. The residual obtained by subtracting both published reading and no-reading shares from one hundred is about 0.65%, 0.40% and 1.10%, respectively. It is retained as an unclassified residual in the distribution figure and tables. It is not interpreted as an additional reading-intensity category, nor is it assumed to contain people who did not read.
This distinction changes how the later interval should be described. The any-book estimate falls by around 4.58 percentage points from 2019 to 2024, while the explicitly reported no-book share rises by approximately 3.87 points. The residual increases by about 0.71 points. Those movements reconcile the published distribution without forcing missing answers into zero. They also show why a statement that every decline in reported reading is an equal increase in nonreading would be arithmetically incorrect here.
The conditional share answers a narrower question
Dividing the population share reporting thirteen or more books by the population share reporting any books gives the frequent-reading share among reported readers. Multiplying by one hundred expresses it as a percentage. This denominator excludes the published no-book category and the unclassified residual. It does not describe the share of all residents who read frequently, and it must not be confused with a percentage calculated among everyone who gave a nonmissing response.
On this basis, approximately one quarter of reported private-purpose readers were in the thirteen-plus category in both 2014 and 2019. In 2024 the corresponding point estimate was closer to one fifth. The decrease is about 5.30 percentage points over the decade and 5.95 points between 2019 and 2024. These are percentage-point changes in a conditional share, not the percentage of earlier readers who subsequently changed their reading volume.
The calculation helps separate the size of a reported activity group from its internal composition. If fewer residents reported any books but the same fraction of that smaller reader category reported thirteen or more, the population prevalence of frequent reading would still decline. Conversely, frequent-reader prevalence can decline even when overall reading prevalence is unchanged, if the conditional frequent share falls. The published margins show both ingredients for the later interval, but they do not identify the personal histories behind either ingredient.
An exact identity, not an explanatory model
Let R denote the percentage of the population reporting any privately read books, Q the percentage of reported readers who read thirteen or more, and H the population percentage reporting thirteen or more. By construction, H equals R multiplied by Q and divided by one hundred. This is an identity between percentages, not a fitted model, a theory of reading, or a causal relationship estimated from individuals.
The change in H can be written exactly as two components. The reader-prevalence-associated component is the change in R multiplied by the average of the two Q values, divided by one hundred. The conditional-composition-associated component is the change in Q multiplied by the average of the two R values, also divided by one hundred. Their sum equals the observed change in H, apart from negligible numerical representation error.
Using endpoint averages makes the calculation symmetric: neither the earlier nor the later distribution is privileged as the reference. Alternative reference-based decompositions can allocate the interaction between the two changing quantities differently. The midpoint rule is stated explicitly so that the reported components are reproducible rather than implied to be the only possible analytical allocation. Words such as associated refer to arithmetic placement in this identity, not to an identified mechanism. No component represents a count of people or a measured effect of an intervention.
What the decomposition shows across the three intervals
For 2014 to 2024, frequent private-reading prevalence decreases by about 5.03 percentage points. The midpoint identity allocates approximately minus 0.99 points to the reader-prevalence-associated component and minus 4.04 points to the conditional-composition-associated component. The latter is larger in absolute magnitude. This is the precise descriptive sense in which the conditional distribution contributes more to the arithmetic change than the size of the reported reader category.
For 2019 to 2024, the total decrease is around 5.55 points, split into minus 1.02 points for reader prevalence and minus 4.53 points for conditional composition. For 2014 to 2019, the frequent-reader estimate instead increases by around 0.52 points. Its two components are approximately plus 0.01 and plus 0.51 points. Keeping all three comparisons visible avoids describing the decade as though it were a continuous, uniformly declining annual trajectory.
Only three survey waves are observed. The data do not establish the intervening year's values or identify when any underlying change occurred. In particular, the larger observed movement falls in the interval that also contains the 2024 survey redesign. The arithmetic remains correct for the published margins, but its substantive interpretation must remain narrower than a claim that reading habits changed by the calculated components. It does not isolate demographic composition, individual behaviour, measurement effects or their combination.
The 2024 redesign is central to interpretation
The earlier survey design combined a telephone interview with a subsequent online or paper questionnaire. Since 2024, respondents can complete the survey online or participate by telephone, with online completion accounting for the great majority of interviews. For the culture and leisure survey, the reported 2024 total is 14,361 respondents: 13,080 online and 1,281 by telephone. These are overall survey sample counts, not the weighted population denominators of the individual book-reading estimates.
The redesign also separated the culture survey from the language and religion survey, revised the sequence and form of questions, shortened collection to August through October, and incorporated income information into nonresponse modelling and weighting. The source explains that the latter change improves representation of lower-income residents. A different interview context and different representation can affect reported distributions even if the broad conceptual definition of private-purpose reading remains similar.
The relevant instruments confirm continuity of subject matter, but not identical presentation. The 2014 and 2019 book-count questions ask about private-purpose books excluding comics, with an ebook follow-up. The 2024 main question explicitly mentions ebooks and is asked of all respondents, with mode-specific estimation guidance. These checks support comparison of the published concepts while also explaining why unchanged category labels cannot guarantee that every observed difference reflects behavioural change. There is no bridge experiment in the aggregate evidence used here that quantifies the redesign's contribution.
Uncertainty is retained without inventing new intervals
The source workbook provides marginal ninety-five-percent confidence-interval half-widths for each published population estimate. They are preserved alongside the selected inputs in the downloadable data. For example, the frequent-reading estimates have half-widths of approximately 0.70 percentage points in 2014, 0.77 in 2019 and 0.65 in 2024. These describe uncertainty associated with each source estimate under the survey's methods; they are not uncertainty limits for this analysis's conditional ratios or decomposition components.
Calculating a ratio from two estimates derived from the same respondents requires their covariance to characterize its sampling uncertainty properly. The same problem extends to products, differences and decomposition components. The inspected aggregate tables do not provide the required joint design information. Therefore no independent-errors assumption, fabricated ratio interval or newly asserted significance test accompanies the derived results. The figures show point estimates and descriptive components, not statistical certainty about their ordering.
Sampling uncertainty is also not the only issue. Recall error, response patterns, the survey-mode change and altered weighting can affect comparability in ways that a marginal confidence interval does not summarize. A narrow interval around an annual estimate would not remove those limitations. Conversely, declining to invent a derived interval does not mean ignoring uncertainty: it means distinguishing the published uncertainty actually available from additional quantities that would require more information to estimate responsibly.
Rounding checks and reproducibility
The main calculations use the source workbook's underlying numerical values rather than the one-decimal labels in the report's chart. This prevents intermediate rounding from becoming part of the definition of the conditional shares. Exact source values and high-precision derived values are retained in the tables and CSV files, while the narrative deliberately uses fewer digits. Numerical precision in storage should not be mistaken for equally precise knowledge of population behaviour.
A separate computation read the underlying spreadsheet XML and reproduced the ratios and midpoint components independently. Its disagreement with the primary decimal calculation was negligible, below two quadrillionths of a percentage point. Recomputing from the displayed one-decimal inputs gave the same qualitative allocation: the conditional-composition component remains larger than the reader-prevalence component for the decade and the later interval. Endpoint checks allowing half a tenth around the displayed input values retained that ordering too.
Those checks address rounding sensitivity, not sampling error or survey comparability, and their ranges are not confidence intervals. The public reproduction script requires only the Python standard library and the compact selected-input CSV. It writes the conditional shares, residuals and all three interval decompositions without downloading microdata or accessing a private workspace. A clean-folder test reproduced the supplied result files byte for byte. Readers can inspect the formulas and change the output location without needing the original research environment.
What this adds to the official publication
The BFS publication already reports the national reading distribution and places it alongside other cultural activities. This article does not claim the discovery of its underlying percentages. Its additional contribution is to state the denominator of the frequent-reading ratio, retain the unclassified residual, and provide a transparent arithmetic allocation of the frequent-reader prevalence difference. These additions address a more specific question than simply whether the share reading books was lower in 2024.
The distinction also prevents several tempting but unsupported interpretations. The thirteen-plus group is not a measure of literary expertise, reading quality or preference for a genre. The book-count categories do not show pages read, completion rates, comprehension or time spent reading. A lower count category need not represent less engagement on any of those other dimensions. Nor does the series identify whether print books were replaced by ebooks, because the private-purpose book measure already includes both formats.
The most defensible conclusion is therefore limited but informative. In the published national private-purpose reading distributions, the lower frequent-reader prevalence in 2024 is arithmetically associated more with a lower frequent share among reported readers than with a lower overall reader share. The population scope, missing-response residual and survey redesign remain integral to that conclusion. Establishing who changed their behaviour, why they did so, or how much of the observed difference reflects measurement would require evidence beyond these repeated cross-sectional margins.
Swiss private reading: the published distribution
Population shares by number of books read in the previous 12 months; 2014, 2019 and 2024.
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Category | 2014 (%) | 2019 (%) | 2024 (%) |
|---|---|---|---|
| 1–3 books | 25.9862913832338 | 25.146593397202 | 29.8138337481359 |
| 4–12 books | 33.1867724310901 | 33.5613513608022 | 29.8623091678164 |
| 13+ books | 19.2244010209512 | 19.744958501578 | 14.1979674508964 |
| No books reported | 20.9521140908944 | 21.1513962092791 | 25.0218195075963 |
| Unclassified residual | 0.6504210738305 | 0.3957005311386 | 1.1040701255550 |
Population percentages; permanent residents aged 15+ in private households. Ebooks included, comics excluded; preceding twelve months. Source precision retained. Residual is unclassified, not no books. 2024 method break.
An arithmetic decomposition, not reader transitions
Change in the population share reading 13+ private-purpose books (percentage points).
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Interval | Change in 13+ prevalence (pp) | Reader-prevalence component (pp) | Conditional-composition component (pp) |
|---|---|---|---|
| 2014–2019 | 0.5205574806268 | 0.01377357276369041874601196732109406611034 | 0.506783907863109581253988032678905933887 |
| 2019–2024 | −5.5469910506816 | −1.016195577313836830685537638083439890994 | −4.530795473367763169314462361916560109001 |
| 2014–2024 | −5.0264335700548 | −0.9892768674526354216197592420089944767375 | −4.037156702602164578380240757991005523261 |
Percentage points; all values use unrounded inputs. Each pair of components sums to total change. Arithmetic allocation, not effects or people moving; 2024 design break. No ratio/component CIs.
Data & downloads
Scroll horizontally to see all columns.
| Category | 2014 (%) | 2019 (%) | 2024 (%) |
|---|---|---|---|
| 1–3 books | 25.9862913832338 | 25.146593397202 | 29.8138337481359 |
| 4–12 books | 33.1867724310901 | 33.5613513608022 | 29.8623091678164 |
| 13+ books | 19.2244010209512 | 19.744958501578 | 14.1979674508964 |
| No books reported | 20.9521140908944 | 21.1513962092791 | 25.0218195075963 |
| Unclassified residual | 0.6504210738305 | 0.3957005311386 | 1.1040701255550 |
Population percentages; permanent residents aged 15+ in private households. Ebooks included, comics excluded; preceding twelve months. Source precision retained. Residual is unclassified, not no books. 2024 method break.
Scroll horizontally to see all columns.
| Survey | Any books (% of population) | 13+ books (% of population) | 13+ among reported readers (%) |
|---|---|---|---|
| 2014 | 78.3974648352751 | 19.2244010209512 | 24.52171261066230467554132102929701301463 |
| 2019 | 78.4529032595823 | 19.744958501578 | 25.16791308059888155238891477608311075398 |
| 2024 | 73.8741103668487 | 14.1979674508964 | 19.21913831569847804279230894318858249789 |
Any-book denominator: population. Conditional denominator: published any-book category only. Derived decimals reflect computational precision, not measurement accuracy; no derived CIs.
Scroll horizontally to see all columns.
| Interval | Change in 13+ prevalence (pp) | Reader-prevalence component (pp) | Conditional-composition component (pp) |
|---|---|---|---|
| 2014–2019 | 0.5205574806268 | 0.01377357276369041874601196732109406611034 | 0.506783907863109581253988032678905933887 |
| 2019–2024 | −5.5469910506816 | −1.016195577313836830685537638083439890994 | −4.530795473367763169314462361916560109001 |
| 2014–2024 | −5.0264335700548 | −0.9892768674526354216197592420089944767375 | −4.037156702602164578380240757991005523261 |
Percentage points; all values use unrounded inputs. Each pair of components sums to total change. Arithmetic allocation, not effects or people moving; 2024 design break. No ratio/component CIs.
Scroll horizontally to see all columns.
| Source estimate | 2014 CI ± | 2019 CI ± | 2024 CI ± |
|---|---|---|---|
| Any books | 0.77201474520776 | 0.84330775790649 | 0.85575058051805 |
| 1–3 books | 0.80648742602656 | 0.88724226015688 | 0.87657213761788 |
| 4–12 books | 0.85246182295169 | 0.93722198918486 | 0.86513628751799 |
| 13+ books | 0.69940689118001 | 0.77383875420512 | 0.64579847149388 |
| No books | 0.76324755572243 | 0.83805046075519 | 0.84405217150721 |
Source 95% confidence-interval half-widths, percentage points. These are not CIs for ratios, changes, residuals or decomposition components.
Evidence & method
How we calculated it
- Secondary descriptive analysis of BFS table T 16.02.04.17, Bücher_Privat, national 2014/2019/2024 rows; all-purpose rows excluded.
- Population: permanent residents aged 15+ in private households. Private-purpose books in preceding twelve months; ebooks included, comics excluded.
- Source category estimates and published marginal 95% CI half-widths retained. Unknown/no-answer residual kept separate from no books.
- Q=100H/R; midpoint components ΔR×mean(Q)/100 and ΔQ×mean(R)/100. Components sum to ΔH. All three wave-pair comparisons reported.
- Source-informed specification followed inspection of source cells; independent direct-XML computation and rounding sensitivity verified. Clean-folder standard-library reproduction is byte-identical.
What the data cannot tell us
- 2024 collection-mode, questionnaire, fieldwork and weighting changes prevent attributing observed differences wholly to reading behaviour.
- Repeated cross-sectional margins identify no individual transitions, causal mechanisms or trajectories between the three waves.
- No covariance/design information for new ratio/component CIs; marginal source CIs are retained only for their original estimates. Rounding sensitivity is not uncertainty estimation.
- Unknown/no-answer are not reassigned to nonreading. Conditional reader shares exclude both no-book responses and the unclassified residual.
- National private-purpose book counts do not measure German-speaking Switzerland alone, children under 15, work/education reading, pages, time, reading quality, purchases or genre.
Original sources
- BFS, Books, ebooks, comics and libraries: 2014, 2019 and 2024 tables
BFS, Books, ebooks, comics and libraries: 2014, 2019 and 2024 tables: Bücher_Privat, national rows 99, 63 and 5; categories, marginal 95% CI half-widths and missing-response note. Updated 14 November 2025. Private-purpose books include ebooks; all-purpose series ends in 2019.
Accessed October 30, 2026
- BFS, Culture and leisure behaviour in Switzerland, 2024
BFS, Culture and leisure behaviour in Switzerland, 2024: Printed page 21 already displays national reading categories; page 28 explains the 2024 collection-mode, fieldwork and weighting changes. Derived conditional shares and decomposition are additional calculations, not BFS findings.
Accessed October 30, 2026
- BFS, Culture and leisure statistics: survey profile
BFS, Culture and leisure statistics: survey profile: Permanent residents aged 15+ in private households; previous-twelve-month recall; sample totals and response modes. Overall respondent counts are not reading-category weighted denominators.
Accessed October 30, 2026
- BFS, Culture survey 2024 questionnaire
BFS, Culture survey 2024 questionnaire: Q52000 FILTER ALL: private-purpose book count, excluding comics and explicitly including ebooks; unknown/no-answer codes and mode-specific estimation guidance.
Accessed October 30, 2026
- BFS, Language, religion and culture survey 2019 CATI flow
BFS, Language, religion and culture survey 2019 CATI flow: Printed page 23, CNBRLIVR: private-purpose books in previous twelve months, excluding comics; subsequent ebook question.
Accessed October 30, 2026
- BFS, Language, religion and culture survey 2014 questionnaire
BFS, Language, religion and culture survey 2014 questionnaire: Printed page 52: Q508 CNBRLIVR defines private-purpose books excluding comics; Q510 identifies ebooks as part of that book count. Documented from an earlier cached retrieval of the official source, not from a fresh local check of the original file.
Accessed October 30, 2026
- Federal Council, removal of statistical reuse permissions and fees
Federal Council, removal of statistical reuse permissions and fees: Official announcement dated 5 June 2026: commercial reuse permission requirements and fees for federally published statistical results abolished effective 15 July 2026. Supports bounded attributed statistical-result reuse, not a blanket license to report graphics.
Accessed October 30, 2026
- Swiss Confederation, enacted ordinance AS 2026 326
Swiss Confederation, enacted ordinance AS 2026 326: Articles 9 and 12 corroborate repeal of the 2003 statistical fee ordinance and the 15 July 2026 effective date. No agency logos or copied source graphics are included.
Accessed October 30, 2026
Cite & reuse this study
Alisson Bento. Swiss private reading: participation and the frequent-reader share, 2014–2024. Secondary analysis of Swiss Federal Statistical Office, Culture and leisure behaviour / ESRK, table T 16.02.04.17. Sources accessed 30 October 2026. https://alissonbento.com/research/swiss-private-reading-distribution-2014-2024/
Underlying statistical results: Swiss Federal Statistical Office (BFS), Culture and leisure behaviour / ESRK. Bounded attributed aggregate reuse supported by the 2026 statistical-reuse revision; see rights.md and official sources. Selection, calculations, original figures and explanation: Alisson Bento. No agency endorsement, logos, copied source graphics or blanket source-publication license claimed.
Corrections or editorial questions