Skip to content

Portugal · Library programmes · 2024

Library event counts in Portugal: why provision, sessions and weights matter

Among 257 RNBP respondent municipalities, 93 have valid count pairs for all four selected reading-related activity types. Comparing those records shows why programme provision, session volume and the choice of average need separate explanations.

Published
Share
Share this page
EmailCite & reuse this study
Key findings as text
Reusable imagesLibrary event counts in Portugal: why provision, sessions and weights matter — 93 / 257: Municipalities in common support. All 93 report all four types with valid session/participant pairs; a selected multi-activity group, not the whole network.Download sharing image (JPG) ↓

Key findings

93 / 257

Municipalities in common support

All 93 report all four types with valid session/participant pairs; a selected multi-activity group, not the whole network.

133 / 257

Municipalities reporting no reading clubs

Distinct from only 2 club providers missing a complete pair; blank fields do not all mean nonresponse.

31.19 vs 22.50

Workshop pairwise pooled ratio versus municipal median

195 valid pairs; reported participant-count/session quotients, not verified audience sizes.

0.41

Median equal-type-index difference

92 common-support municipalities with all session counts positive; descriptive weighting index, not a counterfactual.

A busy calendar and a large recorded total are different things

Story sessions account for more than half of the selected sessions in the matched municipal records examined here. Author meetings and book presentations account for less than a tenth. Yet the latter category has a larger share of the summed recorded participant counts than of sessions. That is not evidence that one kind of event is better. It shows why a calendar of activities and a total labeled “participants” cannot answer the same question without additional context.

This study compares four reading-related activity categories in Portugal's Rede Nacional de Bibliotecas Públicas, or RNBP, for 2024: story sessions, author meetings and book presentations, reading clubs, and reading/literacy workshops. It uses municipal records published by the Direção-Geral do Livro, dos Arquivos e das Bibliotecas, DGLAB. The focus is how the picture changes when the analysis distinguishes provision, usable paired records, session volume and the weight given to each municipality. It does not estimate the effect of programmes on reading habits.

A central limitation must come first. The questionnaire asks for numbers of sessions and participants but the inspected material does not specify how repeat participation is counted. Someone who attends several sessions might contribute more than once, or local reporting could use another convention. Throughout this article, a reported participant-count-to-session ratio means exactly the quotient of those two recorded fields. It is not a verified average audience size, a count of unique readers or a measure of programme effectiveness.

Start with the municipalities that offered each activity

The file contains responses from 257 municipalities in the surveyed RNBP frame. It is not a census of every Portuguese municipality, every library building or every event. A municipal response can describe a service containing several locations. The official report also notes that some services undergoing works or changes of premises continued other activities. A low or high total cannot, by itself, reveal the conditions under which a programme operated.

Provision differs sharply across the four categories. Story sessions and author/book presentations are each reported by 241 municipalities. Reading/literacy workshops are reported by 205, while reading clubs are reported by 124. The remaining municipalities mark the relevant activity as not provided. These flags matter because a missing pair of numerical fields does not always mean that a provider failed to supply a number. Often the municipality says it did not offer the category at all.

For clubs, 133 municipalities report no provision, while only two of the 124 reporting provision lack a complete numerical pair. Calling all 135 missing pairs “nonresponse” would make a largely substantive difference in programme availability look like a data-quality problem. The same distinction applies elsewhere: story sessions have 16 nonproviders and 15 providers with an incomplete pair; author/book presentations have 16 and 16; workshops have 52 and 10. The analysis retains these distinctions instead of converting every blank into zero.

Matching the two counts changes the denominator

To calculate a quotient for a municipality, the session count and recorded participant count must belong to the same municipal record. Comparing two separately published national totals is not necessarily equivalent. A municipality can report its number of sessions while leaving the participant field blank. If its sessions remain in the denominator but its unknown participant count disappears from the numerator, the resulting quotient mixes different response sets.

Story sessions illustrate the issue. Across all available session fields, the source records 20,422 sessions. The 226 municipalities with valid pairs account for 19,943 sessions and 483,398 recorded participants. The paired quotient is therefore 24.24, using those matched records. The excluded sessions are not assumed to have had nobody participating. Their corresponding participant counts are unavailable in this frozen file. The calculation asks what the complete pairs show, not how many participants the missing records might have contained.

The valid paired sets contain 226 municipalities for stories, 225 for author/book presentations, 122 for clubs and 195 for workshops. Requiring valid pairs for all four categories leaves 93 municipalities. All 93 report providing all four activities. This matched four-category set is useful for comparing the composition of their records, but it is a selected group of multi-activity providers. It cannot represent the programme mix of the entire network, especially municipalities without reading clubs. The stricter denominator improves comparability within the set while narrowing what the result describes.

Two averages answer two different questions

A ratio of sums divides the combined recorded participant count by combined sessions. Municipalities with more reported sessions have more influence on that result. A median municipality ratio instead finds the middle of the individually calculated municipal quotients. Neither is automatically the correct “average.” They describe different aspects of the records, and their separation can be informative even when both calculations are arithmetically sound.

On each category's valid paired set, the story-session pooled quotient is 24.24 and its municipal median is 22.58. For author/book presentations the corresponding values are 50.01 and 42.92. Clubs have a pooled quotient of 12.92 and a median of 12.00. Workshops show a wider separation: 31.19 when records are pooled, compared with 22.50 for the median municipality. A reader looking only at the pooled workshop figure would miss the substantially lower middle municipal value.

These differences do not identify the reason for the gap. Larger-session providers may have different recorded ratios, and unusually large participant totals can influence a pooled result. Event duration, capacity, format and local recording conventions are not standardized by the calculation. Nor should a club's smaller quotient be treated as a failure next to an author presentation. A recurring discussion group and a large public presentation can serve different purposes. The arithmetic helps describe the record distribution; it cannot rank their educational, cultural or social value.

Comparing the same 93 municipalities

The common set makes one important change: every activity category is now drawn from the same 93 municipal records. Within it, story sessions total 11,551 and recorded participants total 280,359. Author/book presentations total 1,774 sessions and 79,267 recorded participants. Clubs contribute 1,968 sessions and 24,876 recorded participants, while workshops contribute 5,122 sessions and 152,367 recorded participants. These are sums of reported category fields, not an unduplicated count of people across programmes.

The pooled quotients on this support are 24.27 for stories, 44.68 for author/book presentations, 12.64 for clubs and 29.75 for workshops. The municipal medians are 22.79, 40.86, 11.81 and 24.00, respectively. The comparison with the larger pairwise sets shows that changing which municipalities qualify can matter independently of the choice between pooling and taking a median. Author/book presentations, for example, have a pooled value of 50.01 on their own paired set but 44.68 on the common set.

The common set is not a more accurate estimate of the wider category. It answers a narrower question about municipalities with usable records for all four activities. The pairwise set keeps more information about a particular category; the common set keeps the municipal membership constant across categories. Reporting both prevents a change in analytical support from being mistaken for a change in the underlying service. All of these observations refer to the same year, so none is a trend.

Why session shares and recorded-count shares differ

Within the common set, stories represent 56.58% of selected sessions and 52.22% of the summed recorded participant counts. Author/book presentations represent 8.69% of sessions and 14.76% of recorded counts. Clubs represent 9.64% and 4.63%, and workshops 25.09% and 28.38%. These four pairs show why a headline based on one total can give a different impression of the activity mix from a headline based on the other.

The relationship is an exact accounting identity. A category's share of summed recorded participant counts equals its share of sessions multiplied by its reported count-to-session quotient relative to the overall pooled quotient. A category can therefore have a large share of recorded counts because many sessions are reported, because its quotient is higher, or because both are true. No assumption about reader preference is needed to produce this pattern. It follows from the recorded quantities and their denominators.

The identity does not turn the summed counts into unique people. A person could appear in more than one category, and the repeat-count rule within a category is unverified. It also does not show unmet demand for a category with a smaller share. The appropriate use is explanatory: distinguish the volume of sessions from the relationship between the two reporting fields. It is inappropriate to infer that more people would read, learn or feel included if a municipality replaced one event type with another.

An equal-type index is not a new programme forecast

A second descriptive comparison gives each of the four category quotients equal weight within a municipality. The ordinary pooled quotient weights categories by their reported session numbers. The equal-type index instead averages four quotients, regardless of how many sessions each category supplied. This comparison is defined only for municipalities with positive session counts in every selected category. There are 92 such municipalities in the common set.

The median difference between the equal-type index and each municipality's own pooled quotient is approximately 0.41. The index is higher in 51 municipalities and lower in 41; none has an exactly equal value under the calculation. This is a modest and mixed central result. It does not support a story that equal weighting produces a large, uniform increase. The purpose of showing it is to reveal what changes when the weights change, including when the typical change is small.

One club record reports provision but has zero sessions and zero recorded participants. Its zeros can remain in sums, but zero divided by zero is not a defined category quotient. That record is excluded from the equal-four-type index rather than assigned an invented ratio. More generally, the index is not a prediction of what would happen if all categories received equal calendar space. Reallocating sessions could change content, repeat participation, reporting or capacity. The index holds the recorded quotients fixed for arithmetic comparison only.

Checking how much one large record influences the pool

Pooled quantities can be sensitive to a large contributor. A predeclared check removes, for each activity type separately, the municipality with the largest recorded participant count and recalculates the pooled quotient. This is a sensitivity calculation alongside the main result. It is not a correction, an outlier deletion rule or an allegation that a large reported value is wrong. The complete result continues to include the record.

The largest contrast occurs for workshops on the pairwise support. The full quotient is 31.19; after the largest recorded participant contributor is left out, it is 27.57. The remaining set has 8,314 sessions and 229,211 recorded participants. The removed record had 54 sessions and 31,752 recorded participants. Those figures show the arithmetic source of the sensitivity, but they do not explain its real-world cause. The file cannot establish whether a reporting convention, event arrangement or another feature accounts for it.

The same check gives 24.42 for stories compared with the full 24.24, 49.09 for author/book presentations compared with 50.01, and 12.55 for clubs compared with 12.92. On the common municipal support, the workshop quotient moves from 29.75 to 28.59 when that set's largest contributor is removed. Thus the visible sensitivity depends both on the type of pooling and on the municipalities retained. Reporting these checks is more transparent than choosing whichever average produces the strongest contrast.

A small source discrepancy remains visible

The calculations were reproduced from the frozen official CSV, and the 2,056 cells in the eight selected numeric fields (257 municipalities × 8 fields, 1,572 of them non-blank) were checked against the corresponding XLSX cells. Observed counts and sums reconcile with the workbook's summary rows. That agreement does not mean that every display in every official source is identical. In particular, the printed statistical report contains a response-count discrepancy for the club participant field.

On page 27, table 16.3.3, the report gives 123 responses for club participants and a sum of 29,488. The machine-readable CSV and XLSX contain 122 numerical observations for that participant field, with the same sum. The analysis uses the actual frozen rows and explicitly retains the discrepancy. It does not create a missing observation to force agreement, and it does not claim to know whether the difference reflects a revision, a display convention or something else.

Several other checks found no negative or fractional counts in the selected fields and no municipality marked as a nonprovider with a positive selected count. Those checks support internal consistency for the operations performed, but they cannot validate the underlying recording practice. A clean numerical file does not establish that every municipality counted repeat participation in the same way. The distinction between computational verification and measurement interpretation is particularly important when a seemingly familiar label such as “participants” appears beside a precise number.

Reading an activity statistic without asking it to do too much

The results suggest a practical sequence for interpreting this kind of library statistic. First ask whether the activity was offered. Then ask whether the session and participant fields are available for the same records. Next decide whether the question concerns pooled recorded quantities or a typical municipality. Only after those choices should different categories be compared. Skipping one step can change the denominator without making that change visible to the reader.

For someone considering a library's programme, these statistics provide context about recorded activity types, not a guide to the experience of attending. A regular club, a story session and an author presentation are different opportunities. Choosing among them requires information absent here, such as the programme's subject, schedule, accessibility and intended participants. For an analyst writing about the network, the useful lesson is equally concrete: explain the support set and the weights before turning a quotient into a narrative.

The clearest findings are therefore about interpretation rather than winners. Club-related exclusions mostly identify nonproviders, not unanswered questions. Pooling and the municipal median produce notably different workshop quotients. Common support changes some category comparisons, while the equal-type index produces only a modest median difference. Together, these results show how much information a single activity total leaves out. They also establish the boundary of this analysis: recorded programme quantities can be compared carefully without claiming to have measured unique readers, programme quality or changes in reading behavior.

Different event types, different reporting sets

Counts are reported fields, not verified unique people or audience sizes. Blank counts are not converted to zero. No claim about event quality.

Full chart and source
Different event types, different reporting sets. Counts are reported fields, not verified unique people or audience sizes.
Blank counts are not converted to zero. No claim about event quality.

Scroll horizontally to inspect the chart. ↔

Chart and values

Scroll horizontally to see all columns.

Provision and complete count pairs among 257 municipalities
Activity typeValid count pairsOffered but pair incompleteNot offered, pair blank
Story sessions2261516
Author meetings / book presentations2251616
Reading clubs1222133
Reading / literacy workshops1951052

Valid pairs require both count fields; no blank is recoded as zero. Nonproviders are distinct from providers missing a pair. One club zero/zero remains a valid pair for sums.

Counts are reported fields, not verified unique people or audience sizes. Blank counts are not converted to zero. No claim about event quality. DGLAB RNBP 2024 · CC BY 4.0 data · Original calculations: Alisson Bento

Download chart (SVG) ↓ · Download chart (PNG) ↓

Two summaries of the same recorded counts

Common support: 93 municipalities offering all four types. Club median: 92 positive-session municipalities; one zero/zero remains in totals. Counts are reported fields, not verified unique people or audience sizes.

Full chart and source
Two summaries of the same recorded counts. Common support: 93 municipalities offering all four types.
Club median: 92 positive-session municipalities; one zero/zero remains in totals.
Counts are reported fields, not verified unique people or audience sizes.

Scroll horizontally to inspect the chart. ↔

Chart and values

Scroll horizontally to see all columns.

Two quotients on the same 93 municipal records
Activity typePooled recorded-count/session ratioMedian municipal ratioPositive-session municipalities
Story sessions24.2722.7993
Author meetings / book presentations44.6840.8693
Reading clubs12.6411.8192
Reading / literacy workshops29.7524.0093

Reported participant-count-to-session ratios, not verified audience sizes. All 93 provide all four activity types; club median excludes one zero/zero record and uses 92 positive-session records. Other medians use 93.

Common support: 93 municipalities offering all four types. Club median: 92 positive-session municipalities; one zero/zero remains in totals. Counts are reported fields, not verified unique people or audience sizes. DGLAB RNBP 2024 · CC BY 4.0 data · Original calculations: Alisson Bento

Download chart (SVG) ↓ · Download chart (PNG) ↓

Data & downloads

Scroll horizontally to see all columns.

Provision and complete count pairs among 257 municipalities
Activity typeValid count pairsOffered but pair incompleteNot offered, pair blank
Story sessions2261516
Author meetings / book presentations2251616
Reading clubs1222133
Reading / literacy workshops1951052

Valid pairs require both count fields; no blank is recoded as zero. Nonproviders are distinct from providers missing a pair. One club zero/zero remains a valid pair for sums.

Scroll horizontally to see all columns.

Two quotients on the same 93 municipal records
Activity typePooled recorded-count/session ratioMedian municipal ratioPositive-session municipalities
Story sessions24.2722.7993
Author meetings / book presentations44.6840.8693
Reading clubs12.6411.8192
Reading / literacy workshops29.7524.0093

Reported participant-count-to-session ratios, not verified audience sizes. All 93 provide all four activity types; club median excludes one zero/zero record and uses 92 positive-session records. Other medians use 93.

Scroll horizontally to see all columns.

Pairwise versus common-support pooling
Activity typePairwise municipalitiesPairwise pooled ratioPairwise median ratioCommon-support pooled ratio
Story sessions22624.2422.5824.27
Author meetings / book presentations22550.0142.9244.68
Reading clubs12212.9212.0012.64
Reading / literacy workshops19531.1922.5029.75

Each pairwise column uses that category's own valid set; common support uses the same 93 municipalities for every category. Cross-sectional comparisons, not trends.

Scroll horizontally to see all columns.

Activity mix in the common 93 municipalities
Activity typeSessionsRecorded participant countSession share (%)Recorded-count share (%)
Story sessions11,551280,35956.5852.22
Author meetings / book presentations1,77479,2678.6914.76
Reading clubs1,96824,8769.644.63
Reading / literacy workshops5,122152,36725.0928.38

Shares of selected sessions and summed recorded category participant counts; not shares of unique readers. Percentages rounded independently.

Scroll horizontally to see all columns.

Sensitivity to the largest recorded participant contributor
Activity typeFull pairwise pooled ratioLeave-largest-out pooled ratio
Story sessions24.2424.42
Author meetings / book presentations50.0149.09
Reading clubs12.9212.55
Reading / literacy workshops31.1927.57

For each type on its pairwise support, omit one largest participant-count contributor only for sensitivity. Main results retain all records; no error or outlier allegation.

Evidence & method

How we calculated it

  1. Select actual 2024 municipality rows and four activity types: stories, author/book presentations, reading clubs and reading/literacy workshops. Exclude source footer/summary rows.
  2. Preserve provision flags, blanks and zeros. Valid pairs require nonnegative integer counts, excluding zero sessions with positive participant counts. Zero/zero may enter sums but has no defined individual ratio.
  3. Report category-specific paired support and common valid support. Pooled ratio is summed recorded participant counts divided by summed sessions. Median and inclusive quartiles use municipalities with positive sessions.
  4. Compute shares of selected session totals and summed recorded participant counts on common support. Their relation is an exact accounting identity, not causal evidence or unique-reader shares.
  5. For municipalities positive in all four session fields, compare their session-weighted pooled quotient with an equal-four-type arithmetic index. Report median difference and sign counts; no counterfactual interpretation.
  6. For each type and support, separately omit the largest recorded participant contributor as a sensitivity only. Main results retain every valid record.
  7. Original CSV results were independently reproduced using Decimal/Fraction arithmetic; 147 checked result paths agree. Source CSV/XLSX selected numeric cells agree; PDF club response-count discrepancy is explicitly retained.

What the data cannot tell us

  • Repeat-participant counting is not defined in the inspected questionnaire. Ratios are recorded-field quotients, not audience sizes, attendance rates or unique-reader measures.
  • 257 respondent municipalities are not all Portuguese municipalities or library buildings. The common 93 form a selected set of four-type providers.
  • Most missing club pairs concern explicit non-provision. Provider item nonresponse, absent activities and recorded zero are different states; no missing values are imputed.
  • Four category participant counts can overlap in persons. Their sum is not a deduplicated total. No programme quality, efficiency, unmet demand, causal effect or reading change is estimated.
  • Printed report p. 27/table 16.3.3 says 123 club-participant responses; CSV/XLSX contain 122 numeric values and the same 29,488 sum. Cause unknown; calculations use frozen machine-readable rows.
  • Cross-sectional 2024 analysis only. The equal-mix index and leave-largest checks are descriptive arithmetic sensitivities, not forecasts or data-error diagnoses.

Original sources

  1. DGLAB RNBP 2024 open dataset

    DGLAB RNBP 2024 open dataset: Municipal administrative records for reference year 2024; official metadata identifies Creative Commons Attribution 4.0. CSV and XLSX resources updated April 14, 2026.

    Accessed October 30, 2026

  2. Official RNBP 2024 CSV

    Official RNBP 2024 CSV: Frozen source rows for 257 municipalities; provision flags and four session/participant pairs. Original source hash retained in the reproducibility materials.

    Accessed October 30, 2026

  3. Official RNBP 2024 XLSX

    Official RNBP 2024 XLSX: All 2,056 cells in the eight selected numeric fields (257 municipalities × 8 fields; 1,572 of them non-blank) agree with the CSV. Observed counts and sums reconcile with workbook summary rows; this does not establish agreement with every PDF display.

    Accessed October 30, 2026

  4. DGLAB Statistical Report RNBP 2024

    DGLAB Statistical Report RNBP 2024: Report pp. 25–28 and questionnaire pp. 86–87: paired sessions/participants fields, without a verified repeat-person rule. Page 27/table 16.3.3 gives 123 club-participant responses, versus 122 numeric observations in CSV/XLSX with the same 29,488 sum; discrepancy retained without explanation.

    Accessed October 30, 2026

  5. Creative Commons Attribution 4.0

    Creative Commons Attribution 4.0: The official dataset metadata assigns CC BY 4.0. Attribute DGLAB, identify transformations and imply no endorsement. No claim that report artwork shares the dataset licence.

    Accessed October 30, 2026

Cite & reuse this study

Alisson Bento. Library event counts in Portugal: why provision, sessions and weights matter. Secondary analysis of DGLAB, Estatística RNBP 2024. Sources accessed October 30, 2026. https://alissonbento.com/research/portugal-library-programme-counts-and-event-mix-2024/

Source data: Direção-Geral do Livro, dos Arquivos e das Bibliotecas, Estatística RNBP 2024, CC BY 4.0, https://creativecommons.org/licenses/by/4.0/. Selected fields, matched-support calculations, English labels and original figures by Alisson Bento. No DGLAB endorsement implied. Report artwork is not reproduced.

Corrections or editorial questions

Pass it on.

Library event counts in Portugal: why provision, sessions and weights matter | Alisson Bentoalissonbento.com

WhatsAppFacebookXTelegramRedditThreadsEmailSMS