# Methods: Concentration of US reading expenditure by income quintile

This study describes income concentration in U.S. consumer-unit Reading expenditure. The primary comparison is 2014–2024; 2025 is reported separately as a newly available extension. Sources retrieved 2026-10-30 UTC.

## Research question and evidence
This study asks whether the share of reading expenditure concentrated in the highest income quintile changed between 2014 and 2024. The metrics are quintile aggregate shares, top/bottom mean-spending ratios, reading budget shares, and multi-year smoothing. The source panel now contains all 12 years 2014–2025, five quintiles (60 observations), based on 24 BLS Table 1101 workbooks. The historical index now lists 2025 as latest; BLS released it October 29, 2026, after the primary comparison endpoint of 2024.

## Main finding: substantial endpoint difference, much smaller smoothed difference
| Measure | 2014 | 2024 | 2025 extension |
|---|---:|---:|---:|
| Highest quintile's share of all reading expenditure |39.0%|48.1%|43.4%|
| Lowest quintile's share of all reading expenditure |9.2%|7.5%|5.9%|
| Highest two quintiles' combined reading share |60.6%|67.5%|64.3%|
| Highest-quintile annual mean reading spending per CU |$201|$300|$248|
| Lowest-quintile annual mean reading spending per CU |$48|$47|$33|
| Highest/lowest mean-spending ratio |4.19|6.38|7.52|
| Highest quintile's share of all expenditures |39.0%|38.4%|38.8%|

2014–2024 highest-quintile reading share rose **9.1 percentage points**, while the lowest-quintile share fell 1.7 points. Do not describe a steady increase: annual highest-quintile shares fell as low as 31.7% in 2018, then rose to 48.1% in 2024. The simple average of annual highest-quintile shares was **38.23% in 2014–2016 versus 40.33% in 2022–2024**, a difference of **2.10 percentage points**. The 2023–2025 average is 42.33%, shown only as a latest-data extension. These are arithmetic averages of annual published shares, NOT pooled survey estimates or a ratio of total dollars across years. The sensitivity is central, not a footnote: endpoint selection materially affects the apparent magnitude.

2024 reading spending was more concentrated at the top than all expenditure: 48.1% versus 38.4%, a 9.7-point difference. The corresponding differences are zero in 2014 and 4.6 points in 2025. These cross-category descriptive gaps are not significance-tested.

## Denominators and units
- Aggregate reading share: reading dollars attributable to a quintile / reading dollars attributable to all CUs, times 100. Use the published aggregate-share table directly, not an unweighted average of quintile means.
- Annual mean reading spending: annual US-dollar expenditure per CU in that quintile, including nonpurchasers. NOT spending per reader or purchaser.
- Highest/lowest ratio: highest-quintile mean divided by lowest-quintile mean, calculated within the same year. No dollar inflation adjustment required for this within-year ratio.
- Reading budget share: mean reading spending / mean all expenditure within a quintile, times 100. This differs from the quintile's share of aggregate reading spending. Published budget shares are rounded to tenths of a percent; panel also gives the ratio calculated from published rounded dollar means. For example, in 2024 the latter is 0.1995% for the top quintile and 0.1341% for the bottom. Extra decimals are computational, not extra measurement precision.
- Aggregate dollar amounts are millions; CU counts are thousands. They are not sample sizes. National reading aggregate: $13,086m (2014), $17,002m (2024), $15,644m (2025). These are nominal spending estimates, not publisher revenue or book-unit sales.
- CUs include families, financially independent individuals, and people sharing major expenses. They are not people, individual readers or strictly identical to households. Quintiles are ranked by income BEFORE taxes and constructed from survey weights; membership changes over time. Counts are near, but not exactly, 20% each because construction is quarterly. The study is repeated cross-sectional, not a panel of the same families.

## Scope of Reading and coverage
Use the BLS category name **Reading**, never book market, novel buying or ebook sales as a substitute. BLS glossary includes newspapers, magazines, books, reference works and digital reading materials; its table guide also lists digital book readers. Textbooks are classified under Education. No format split is available in these summary rows. Expenditures include gifts and exclude business-assigned purchases. Free reading, libraries, reading frequency and volumes consumed are not measured. No claim about willingness to read or causality from income is justified.

## Sampling uncertainty and comparability
Retain standard errors and RSE/CV from each mean table. For the 2024 top-quintile mean $300, SE=$45.78 and RSE=15.24%; bottom mean $47, SE=$7.28 and RSE=15.42%. For 2014 top/bottom RSEs are 4.33%/7.29%. For 2025 they are 12.15%/15.07%. BLS values are based on unrounded underlying means, so recalculating RSE from rounded displayed means can differ.

These SEs describe mean expenditure, NOT the aggregate shares or derived ratios. Do not attach them as share error bars, infer significance of the 9.1-point change, or manufacture confidence intervals for multi-year averages. Valid share/ratio tests require covariance-aware survey methods or replicate-weight microdata work beyond this brief. Small budget shares and annual subgroup volatility warrant caution. Neither smoothing nor low RSE proves a trend. Any causal attribution to inflation, income, digitization or other events is outside this design. Cross-year dollar comparisons remain nominal unless separately deflated.

**2025 qualification:** BLS could not collect October/November survey data during the federal funding lapse. It adjusted Interview expenditure weights and doubled September/December Diary weights to represent missing months. BLS expects increased variances; the true missing-data effect is not observable. Keep 2025 supplementary rather than silently replacing the original 2024 endpoint. The BLS annual release marks the all-CU 2024–25 reading-mean change as not statistically significant; that does not establish anything about quintile share significance.

## Reproducibility and quality checks
Download `reproduce.py` and `quintile_panel_2014_2025.csv` into the same directory. With Python 3, run from that directory:

```sh
python reproduce.py
```

The public script uses only the Python standard library. It reads the supplied frozen source-cell panel, asserts 60 rows, five quintiles in each of the 12 years, and aggregate reading-share sums within 0.3 percentage points of 100. It recomputes and overwrites `annual_metrics.csv` and `three_year_share_averages.csv` alongside the script. It does not download or extract XLSX workbooks, regenerate the source-cell panel, or calculate sampling uncertainty. The panel, both result CSVs, `reproduce.py`, `sources.json`, and this `methods.md` are the six public downloads. CSV files are UTF-8 with LF line endings.

The original source extraction read stored XLSX XML without modifying the workbooks and checked headers, years and share sums. Exact source row/column locations are retained in the panel. The first main-budget Reading block was selected deliberately: pre-2020 workbooks also contain a later Reading row under gift expenditures, which must NOT replace the primary reading measure. No missing values were imputed. CV(%) in older files and RSE in newer files are retained as the mean's relative standard error in percent. Source values retain published rounding; quintile shares need not sum to exactly 100 in every year.

`sources.json` records the official source URLs and retrieval date; it does not include source byte lengths or hashes. Raw workbooks and the original XML extraction and source-integration scripts are not bundled. The public script reproduces the derived results from the supplied panel, not the full workbook-to-panel extraction. Readers checking the extraction against official workbooks should note that later downloads may contain revised tables; this bundle does not provide the archived raw files or their checksums. The two unversioned initial 2025 files duplicate the subsequently year-labelled sources and are not extra observations.

## Primary sources and reuse
- Aggregate index: https://www.bls.gov/cex/tables/calendar-year/aggregate-group-share.htm
- Mean/SE index: https://www.bls.gov/cex/tables/calendar-year/mean-item-share-average-standard-error.htm
- Main 2024 aggregate: https://www.bls.gov/cex/tables/calendar-year/aggregate-group-share/cu-income-quintiles-before-taxes-2024.xlsx
- Main 2024 means/SE: https://www.bls.gov/cex/tables/calendar-year/mean-item-share-average-standard-error/cu-income-quintiles-before-taxes-2024.xlsx
- Latest aggregate: https://www.bls.gov/cex/tables/calendar-year/aggregate-group-share/cu-income-quintiles-before-taxes-2025.xlsx
- Latest means/SE: https://www.bls.gov/cex/tables/calendar-year/mean-item-share-average-standard-error/cu-income-quintiles-before-taxes-2025.xlsx
- Definitions: https://www.bls.gov/cex/csxgloss.htm
- Table methods: https://www.bls.gov/cex/tables-getting-started-guide.htm
- Latest release: https://www.bls.gov/news.release/archives/cesan_10292026.htm
- 2025 methodological break: https://www.bls.gov/cex/2025-federal-government-shutdown-impact-ce.htm
- Public-domain reuse: https://www.bls.gov/bls/linksite.htm
BLS data are public domain with attribution requested; do not reuse its emblem.

This study measures consumer expenditure, not all market funding.

## Source integration and endpoint interpretation
The official source-integration workbook https://www.bls.gov/cex/ce_source_integrate.xlsx was inspected at Reading rows 1047–1060, year columns D–O (2025–2014). Newspapers/magazines switch from Interview to Diary in 2015; encyclopedia/reference-books also switch in 2015; books-through-book-clubs switch in 2017. The latter and reference-book component codes end after Q20244, with no active source in 2025. This does not establish that purchases vanished rather than were reorganized. Digital book readers are included throughout. The published Reading label is therefore NOT a verified fixed measurement basket across 2014–2025. Neither endpoints nor early smoothing isolate behavioral change. General guide explains that BLS selects a survey source for overlapping items and changes categories/sources to improve measurement and reduce respondent burden.

All five 2024 main Reading-mean RSEs: Q1 15.42%, Q2 12.54%, Q3 12.61%, Q4 9.35%, Q5 15.24%. None meets 25% caution threshold; this is not evidence that aggregate-share differences are significant. Differing CU sizes/composition mean ratios are not per-capita or behavioral effects.