Influence IQ: Online shopping problems by age
Analysis date: October 17, 2026

This package reproduces an exploratory descriptive study using Eurostat's 2025
survey estimates. It contains aggregate statistics, not respondent records.

Run with Python 3.10 or later, using only the standard library:
  python3 verify.py
  python3 analyze.py
  python3 verify.py

The first command checks every packaged file against its SHA-256 manifest.
The second regenerates study.json, figures.json, summary.json,
source-and-conditional-results.csv and country-comparisons.csv. The final
verification confirms that those results reproduce exactly. No network or
credentials are required. Run in an extracted, writable copy of the package.

SOURCE AND POPULATION
Eurostat isoc_ec_iprb21, survey year 2025, PC_IND_BUY3. The source was retrieved
October 17, 2026 and was last updated April 17, 2026. The unit is the percentage
of people who bought online in the previous three months. Six age groups cover
ages 16 through 74; the standard survey concerns private households and private
purchases through websites or apps. The 28 geographies are the EU aggregate and
all 27 member countries. National interview and recall periods differ; the data
do not represent a single common calendar-year purchasing window.

CALCULATIONS
For each country, compare final-costs-higher-than-indicated (I_BCPR1) with
complaint/redress difficulties (I_BCR1) at ages 16-24 and ages 65-74. The headline
counts countries with costs above redress in the younger group and redress above
costs in the older group. Each country counts once, regardless of population.
All four possible orderings are retained, not only the headline's 15 countries.

Conditional percentages divide a specific problem percentage by I_BARR2Y for
the SAME country and age group, then multiply by 100. The model's aggregated
variable guide explicitly defines I_BARR2Y as E9a-h: eight specified problems.
Other-only responses (E9i) and the separate safety/health question E10 are not
included. Responses overlap; these percentages are not parts summing to 100.
The combined measure's own conditional field is blank, not 100 or zero.
Results count people, not orders, incidents, complaints or support workload.

Each input's +/-0.005 percentage-point rounding interval is checked for the
country ordering. This is only sensitivity to published decimal rounding, not
a sampling confidence interval. These aggregate data do not provide the joint
responses and survey design needed for an appropriate significance test here.
The analysis does not control for purchase frequency, product mix, seller,
income or other differences, and does not establish age or advertising effects.

Files preserve all 1,512 source cells (28 geographies x 6 ages x 9 measures),
including source and denominator flags. None is missing or flagged in this
frozen selection; that does not imply negligible sampling or nonsampling error.
The country file's pattern codes are:
  costs_then_redress: cost-first younger, redress-first older
  redress_both: redress-first at both endpoint ages
  costs_both: cost-first at both endpoint ages
  redress_then_costs: the opposite reversal
CSV numeric fields use a decimal point; empty conditional fields are intentional.

METHOD DOCUMENTS
questionnaire-2025.pdf: printed page 18, E9/E10 routing and multiple answers.
variables-2025.pdf: printed page 21, the positive E9a-h union definition.
metadata.html: population, reference periods, sampling and comparability.
hu.html and hu-annex.xlsx: Hungary's report acknowledges possible questionnaire
ordering effects; its 2025 annex lists no E9 deviation. This is not proof of zero
survey effects. Excluding Hungary and Ireland leaves 14 reversals among 25 countries.
ireland-2025-cso.html: current 2025 notes, used because the Eurostat-linked Irish
national report is older. The national Irish publication covers 16+, while this
analysis uses Eurostat's 16-74 data. Do not mix those populations.

National reports were reviewed for survey periods and questionnaire deviations.
The separate CIRCABC country-note links were unavailable; no claim is made that
all national questionnaires were independently recoded. The selected comparison
was chosen after exploring several problem categories, not preregistered.
Earlier research already examines shopping problems and demographic differences.
The contribution is this reproducible 2025 comparison and conditional mix, with
exceptions and all six age groups disclosed, not a world-first claim.

SOURCE RIGHTS AND ATTRIBUTION
Source: Eurostat, https://doi.org/10.2908/ISOC_EC_IPRB21 . Eurostat statistical
data may be reused with attribution; see https://ec.europa.eu/eurostat/en/help/copyright-notice .
Calculations and interpretation are by Influence IQ. Eurostat is not responsible
for the adaptations. Source documents retain their originating rights and
notices. Irish background notes are attributed to the Central Statistics Office.
The manifest records original public source URLs and retrieval hashes.

No Ahrefs data, company operating records, tracking identifiers, credentials,
respondent microdata or restricted research papers are included.
