What ‘harmonised’ actually covers
The Harmonised European Time Use Survey (HETUS) is the only instrument that produces cross-national diary data on unpaid domestic and care work in Europe. Its name invites a specific reading: that the resulting estimates are comparable across countries. The documentation supports a narrower reading. HETUS harmonises the activity coding list and the definitions of its three indicators. It does not harmonise who participates, when fieldwork runs, how samples are drawn, or how weights are constructed. Those four dimensions are where cross-country comparison of unpaid care work becomes fragile, and Eurostat’s own reference metadata says so.
This matters for gender analysis because unpaid care work is the category where the stakes are highest. Eurostat’s metadata lists among the uses of time-use data the questions of how much time is spent in paid compared with unpaid work and how this differs for women and men, and whether housework is fairly distributed between women and men (Time use survey – 2020 (tus_20), ESMS). A comparison that carries unquantified methodological differences cannot answer those questions at Member State level, however precisely the activity codes are defined.
The legal form: a gentlemen’s agreement, not a regulation
HETUS data are collected on the basis of a gentlemen’s agreement between participating countries and Eurostat (Eurostat Glossary: HETUS). The ESMS repeats the point under institutional mandate: the data were collected on the basis of a gentlemen’s agreement between participating countries and Eurostat. A Eurostat presentation to the January 2024 UN Expert Group on Innovative and Effective Ways to Collect Time-Use Statistics places this alongside Regulation (EU) 2019/1700, under which time use surveys remain voluntary (Tchipeva, Miceli and Ninka, ‘Time use surveys in Europe: State of play and a glimpse into the future’, 30 January 2024).
The legal form is not a procedural footnote. It determines the population from which an EU-level statement can be made. A voluntary instrument produces a picture of the countries that chose to participate, not a picture of the Union. The 2020 round involved 15 EU Member States, 1 EFTA country and 4 EU candidate countries, according to the same Eurostat presentation. The glossary records that in round 2, 15 EU countries and 3 non-EU countries collected and transmitted data to Eurostat, and that 13 EU countries did not conduct a time-use survey during that round. The glossary also notes that many countries had not yet decided whether they would carry out a survey in round 3.
For unpaid care work specifically, this means that any statement of the form ‘in the EU, women spend X more time on care than men’ rests on a participating subset whose composition changes between rounds. The subset is not a random sample of Member States, and nothing in the published documentation treats it as one.
What is harmonised: the coding list and the indicators
The harmonisation that does exist is real and should be stated precisely. The main classification system is the activity coding list (ACL), which assigns a three-digit code to primary and secondary activities. The 24 hours of the day are split into 144 ten-minute slots, and respondents’ descriptions are coded by professional coders using a classification with over 100 three-digit categories (Eurostat Glossary: HETUS). The ESMS specifies that activities are coded according to the Classification of Activities for Time Use 2018 (ACL 2018), and that during harmonisation Eurostat aggregates from 116 activities to 41 categories.
The three indicators are also harmonised by definition: time spent (mean time spent by all individuals), participation time (mean time spent by those who took part in the activity), and participation rate (the proportion of individuals who spent some time on the activity). These are compiled by sex, age group, household composition, highest level of education, self-declared labour status, professional status, day of the week and month of the year, among other breakdowns (ESMS, section 3.4).
So the harmonised layer is the vocabulary and the arithmetic. The question is whether the same vocabulary applied to differently produced data yields comparable numbers. The ESMS answers that question in the negative on several specific points.
The timing layer: a ‘2020 round’ without a common reference year
The ESMS lists national data-collection periods for the 2020 wave as follows: Estonia 2019–2021; Finland 2020–2021; Austria and Serbia 2021–2022; the Netherlands 2021–2023; Germany 2022; Bulgaria, Croatia and Norway 2022–2023; Poland 2023; Hungary 2024–2025. The HETUS guidelines require that the survey cover a full twelve-month period, with each respondent completing two diaries, one weekday and one weekend day (ESMS, section 5).
Each national dataset therefore covers a full year, which controls for seasonality within a country. It does not control for period effects between countries. A comparison that places Estonian diaries collected in 2019–2021 beside Hungarian diaries collected in 2024–2025 spans a period that includes the pandemic and the subsequent reorganisation of paid work, school attendance and care arrangements across Europe. The published metadata do not provide a basis for decomposing a cross-country difference in unpaid care hours into behavioural change and period effect. The ‘2020 round’ label describes a survey generation, not a shared reference year.
The sampling and weighting layer
The ESMS is explicit on sampling design: national statistical institutes decided on the sampling design from their own precision needs, which is why sampling designs and sample sizes differ from one country to another (section 13.1). It is equally explicit on weighting: weight calculation methods differ between countries; in the past, calibration techniques or post-stratification were used; and in general, calibration vectors were not identified or post-stratification variables were not reported.
This is the most consequential sentence in the metadata for anyone constructing a cross-country comparison. If the calibration vectors are not reported, a reader cannot establish whether two national estimates are weighted to the same population margins. A difference in mean daily care time between two Member States could reflect a difference in behaviour, a difference in sample composition, a difference in weighting targets, or some combination. The published metadata do not permit the decomposition. The ESMS also notes that the sampling error will be presented once the metadata from all participating countries are available, and that only very few national statistical institutes reported item non-response rates.
There is a further asymmetry in the design. The ESMS records that the Netherlands collected seven diary days per respondent instead of two, and that the reliability thresholds applied to Dutch figures are correspondingly 70 and 175 observations rather than 50 and 20. A country with seven diary days per respondent produces a different kind of estimate from a country with two, and the harmonised indicator definitions do not erase that difference.
What has actually been published
The ESMS lists the countries published on Eurobase so far as Bulgaria, Germany, Estonia, Croatia, Hungary, Netherlands, Austria, Poland, Finland, Norway and Serbia. It states that HETUS 2020 microdata are not yet available, and that further indicators will be published during 2026. The metadata was last updated on 10 August 2026.
Two consequences follow. First, the set of published countries is smaller than the set of participating countries, which is itself smaller than the Union. Second, the absence of microdata means that independent re-weighting or re-estimation to test the sensitivity of cross-country comparisons is not currently possible through the standard access route. The ESMS states that microdata will be made available through Eurostat’s microdata access service.
The classification revision and the time-series problem
The ACL 2018 revision changes what counts as unpaid domestic and care work relative to earlier rounds. According to the Eurostat presentation to the UN Expert Group, division 3 was renamed from ‘Household and Family Care’ to ‘Unpaid Domestic Services and Caregiving Services for Household and Family Members’, and now includes only domestic and care services, that is, unpaid, non-productive work. Division 4 was renamed from ‘Voluntary Work and Meetings’ to ‘Unpaid Voluntary Work’, with sub-division 41 renamed ‘Community and Organisation-based Work’ and 42 renamed ‘Voluntary Work for Other Households’. A new one-digit division, ‘Production of Goods for Own Final Use’, aligned to ICATUS major division 2, collects activities previously spread across several ACL divisions.
The revision improves conceptual alignment with the System of National Accounts and with ICATUS, and the presentation frames it as a step toward better distinguishing paid from unpaid and productive from non-productive work. It also means that round-3 unpaid care figures are not straightforwardly continuous with round-2 figures. A time series that splices the two rounds without a documented bridge is comparing two different classifications. The presentation describes a proposed ‘light HETUS’ with a fixed set of around 25 activities, intended to allow more timely monitoring of unpaid work; it is presented as a proposal with objectives, not as an adopted instrument.
What a defensible comparison requires
The practical implication is presentational, and it can be stated as a procedure. Any cross-country figure on unpaid care work drawn from HETUS 2020 should carry three caveats attached at the point of use: the set of participating and published countries, the national fieldwork window, and the fact that sampling and weighting designs differ and calibration vectors are generally unreported. Where the metadata do not support a cross-country ranking, within-country trends over time are the more defensible use of the data, subject to the classification break described above.
This is not an argument that HETUS is unfit for purpose. It is an argument that the word ‘harmonised’ should be read with the scope the documentation gives it. The coding list and the indicator definitions are harmonised; the conditions of production are not. For an analysis of gendered outcomes, the distinction determines whether a gap between two Member States can be interpreted as a difference in how women and men allocate time, or whether it may be an artefact of when the diaries were filled in and how the sample was weighted. The published metadata do not, at present, allow the reader to tell which.
Frequently asked questions
Is HETUS legally binding on Member States?
No. Eurostat’s glossary and ESMS both state that HETUS data are collected on the basis of a gentlemen’s agreement between participating countries and Eurostat, and the Eurostat presentation to the UN Expert Group notes that under Regulation (EU) 2019/1700 time use surveys remain voluntary.
Does the ‘2020 round’ mean all countries collected data in 2020?
No. The ESMS lists national collection periods running from 2019–2021 in Estonia to 2024–2025 in Hungary. Each national dataset covers a full twelve-month period, but the periods do not coincide across countries.
Are the unpaid care indicators themselves defined differently across countries?
The three indicators — time spent, participation time and participation rate — are harmonised by definition, and activities are coded to ACL 2018 and then aggregated from 116 activities to 41 categories. What differs is the sampling design, the weighting method, the fieldwork period and, for the Netherlands, the number of diary days collected per respondent.
Can I compare HETUS 2020 unpaid care figures with the previous round?
Only with care. The ACL 2018 revision renamed division 3 and created a new division for production of goods for own final use, so the categories are not identical between rounds. A spliced series should document the classification change rather than assume continuity.
When will the full HETUS 2020 data be available?
The ESMS states that HETUS 2020 microdata are not yet available and that further indicators will be published during 2026. The countries published on Eurobase so far are Bulgaria, Germany, Estonia, Croatia, Hungary, Netherlands, Austria, Poland, Finland, Norway and Serbia.
