1.1. Contact organisation
Statistics Austria
1.2. Contact organisation unit
Directorate Social Statistics: Unit Living Condtions, Social Protection
1.3. Contact name
Confidential because of GDPR
1.4. Contact person function
Confidential because of GDPR
1.5. Contact mail address
Guglgasse 13
Vienna 1110
Austria
1.6. Contact email address
Confidential because of GDPR
1.7. Contact phone number
Confidential because of GDPR
1.8. Contact fax number
Confidential because of GDPR
2.1. Metadata last certified
8 June 2026
2.2. Metadata last posted
8 June 2026
2.3. Metadata last update
8 June 2026
3.1. Data description
- Survey name(s) in the national language(s): Zeitverwendungserhebung
- Survey name in English: Time use survey
- Year(s) of (data collection) of the survey: 2021/2022
- Link to the survey website: Time use (in English) and Zeitverwendung (in German)
- National questionnaire: please refer to the annex
Annexes:
questionnaire
3.2. Classification system
See below.
3.2.1. Versions and breakdowns (level) of the classifications used for the data collection
| Acronym | Version | Level |
|---|---|---|
| NACE | 2008 | 2-digit level |
| ISCO | 08 | 2-digit level |
| ISCED | 2011 | 1-8 |
| DEGURBA | 2022 | 1-3 |
| ACL (Activity coding list) | 2018 | lowest |
3.2.2. Deviations from ESS or international standards
3.3. Coverage - sector
Not requested.
3.4. Statistical concepts and definitions
The survey was strictly aligned with the HETUS guidelines, without any deviations.
3.5. Statistical unit
The survey units of the Time Use Survey are private households in Austria.
This includes all individuals aged 10 and over living in private households at addresses where at least one person has their main residence registered in the Central Population Register (ZMR).
3.6. Statistical population
Austrian residential population in private households aged 10 and over; approx. 8 million people.
3.6.1. Main characteristics of the survey population
People living in institutional households are not included in the survey. This group includes, among others, residents of nursing homes, prisons, or collective accommodations such as refugee camps. Persons without a fixed residence are also excluded.
Guests are not counted as household members and are therefore also excluded from the survey.
These exclusions mean that certain population groups are not represented in the results, which should be taken into account when interpreting the data.
3.7. Reference area
Austria
3.8. Coverage - Time
The survey covers five quarters, starting with the fourth quarter of 2021 and ending with the fourth quarter of 2022.
3.9. Base period
Not applicable.
The units of observation are 144 ten-minute time intervals per day, also referred to as "timeslots", over the course of two days.
In these intervals, respondents recorded the activities they engaged in, using their own words.
So the units of measure are minutes per day;
- minutes per day spending time with a specific activity
- minutes per day spending time with a specific person (partner, child under 18 living in the same household, parent living in the same household, etc.)
- minutes per day spending time at a specific location (home, work, train, car, etc.)
HETUS should cover a full 12 months period, i.e. 365 consecutive days. Each respondent should fill in the diary for two days, one weekday (Monday to Friday) and one weekend day (Saturday, Sunday).
There were no deviations from the guidelines - except that weekdays were without public holidays and weekend days included weekdays which were public holidays. The survey covered all 12 months (5 quartals; 4th quartal 2021 - 4th quartal 2022; the 4th quartal is in the period twice because the response at the beginning was too low).
6.1. Institutional Mandate - legal acts and other agreements
Information on this concept is provided in the sub-concepts 6.1.1 and 6.1.2.
6.1.1. At European level
The current round of the harmonised European time use survey (HETUS) data collection has been carried out in the spirit of multi-national and multi-agency cooperation across Europe, in the form of an informal ’gentlemen's agreement’.
6.1.2. At National level
There is no national statistical act that includes the TUS.
6.2. Institutional Mandate - data sharing
No other institutions are involved in collecting the TUS data, and there are no set agreements with other institutions to carry out the analysis.
7.1. Confidentiality - policy
European legislation is followed closely. Information on confidentialty of data for respondents is attached in the annec (available only in German).
Annexes:
confidentiality information
7.2. Confidentiality - data treatment
Microdata will not be published on a level of detail that would compromize confidentiality (e.g. detailed information about region).
8.1. Release calendar
A release calender is available to the general public on the Statistics Austria webpage.
8.2. Release calendar access
Please access here the release calendar.
8.3. Release policy - user access
The data has been available since the day the publication was released, and the public was informed, among other channels, through a press release.
No regularity of publication on TUS is agreed upon. Previous surveys were conducted in 1981, 1992, 2008/09. The survey 2021/2022 was the fourth national implementation.
10.1. Dissemination format - News release
The final national report on the TUS was published together with a press release.
Link to the TUS page (in English)
Link to the national report (only available in German)
10.2. Dissemination format - Publications
The final national report on the TUS was published together with a press release.
Link to the TUS page (in English)
Link to the national report (only available in German)
10.3. Dissemination format - online database
On the TUS page (in German), under the section 'Charts, Tables, Maps' (Grafiken, Tabellen, Karten), highlights are presented, and more detailed tables can also be downloaded.
10.3.1. Data tables - consultations
Not requested.
10.4. Dissemination format - microdata access
Microdata files are available to users upon request. The structure of the data file is an adapted version of the HETUS variables.
Users can request access in the form of a Scientific Use File. The Data Use Agreement is available at the SUF page of Statistics Austria.
Once the Data Use Agreement has been completed, the data can be requested.
10.5. Dissemination format - other
Multiple workshops were organised, including one at the Care Day conference in Graz, and several internal meetings with different stakeholders - such as, trade unions, the Chamber of Labour, and equality officers at the federal level, the City of Vienna, and the Public Employment Service (AMS) - as part of targeted policy communication and stakeholder engagement.
The Time Use Survey (TUS) was also used in special chapters of publications by Statistics Austria. For example, in the publication “Wohnen”, a dedicated section presented data on how long and which groups of people stay at home.
10.5.1. Metadata - consultations
Not requested.
10.6. Documentation on methodology
A standard documentation detailing the employed methodology for the TUS survey is published. It was was revised in February 2025.
The standard documentation is available on the TUS webpage (in German, and an executive summary is available also in English).
10.6.1. Metadata completeness - rate
Not requested.
10.7. Quality management - documentation
The quality management documentation is part of the standard documentation that is published on the Statistics Austria webpage.
11.1. Quality assurance
The standard documentation (mentioned under 10.7) is part of the quality management system, and served as the basis for the Feedback-Talk with a subcommittee of the Statistical Council and Useres of the TUS data.
11.2. Quality management - assessment
In the feedback-talk some points were discussed:
- High relevance of the Time Use Survey (ZVE):
The survey is considered highly relevant by a wide range of user groups, and there is strong interest in using the data for various purposes. Accordingly, the discussion included suggestions for improving the survey design in future iterations. - Potential for over-sampling:
The current sample size appears to limit the analytical possibilities. Participants expressed interest in implementing over-sampling strategies in future surveys to better represent hard-to-reach groups of high analytical interest (e.g., migrants, individuals with low educational attainment). - Quality considerations:
Given the importance of the ZVE as a data source, maintaining comparability across survey waves through consistent concepts is crucial. There was also a suggestion to increase the frequency of the survey, which has so far occurred less than once every 10 years. - Issues with activity-related questions:
Some participants noted difficulties in distinguishing between certain activities. Additionally, the handling of new gender categories should be reconsidered.
12.1. Relevance - User Needs
Untill now, the data was requested (and granted) by 42 users. 29 times by researchers and students based at universites (national and international), while the other 13 requastes came from regional statisitcal institutes, other NSIs and public institutions in Austria (Austrian chamber of labour, Austrian Trade Union Federation, Environment Agency Austira, ...).
The data is used for very different research projects. Amoungst others, also professors at Universities requsted a small sample of the data to use it in the lectures - so students learn to handle TUS-Data.
12.2. Relevance - User Satisfaction
There are no standardized user consultations conducted. However, the feedback meeting mentioned above (as part of quality management, section 11) serves as a tool where users are also invited to share their experiences and provide feedback.
For example, users requested that the ACL and other documentation be made available in Word format, as this makes it easier to create variable labels in statistical software.
12.3. Completeness
All required variables are transmitted.
The national microdata (scientific use file) and the microdata transmitted to EUROSTAT differ in certain aspects.
12.3.1. Data completeness - rate
Not requested.
13.1. Accuracy - overall
95.4% [94.5%; 96.3%] of women and girls aged 10 and older spend time on housework.
– The estimated value is 95.4%, with a 95% confidence interval of 94.5% to 96.3%.
86.2% [84.7%; 87.8%] of men and boys aged 10 and older spend time on housework.
– The estimated value is 86.2%, with a 95% confidence interval of 84.7% to 87.8%.
Women and girls aged 10 and older spend on average 3h 7min [3h 2min; 3h 12min] per day on housework.
– The estimated value is 3h 7min, with a 95% confidence interval of 3h 2min to 3h 12min.
Men and boys aged 10 and older spend on average 1h 54min [1h 49min; 1h 58min] per day on housework.
– The estimated value is 1h 54min, with a 95% confidence interval of 1h 49min to 1h 58min.
36.7% [34.0%; 39.3%] of employed women almost always or always feel time pressure.
– The estimated value is 36.7%, with a 95% confidence interval of 34.0% to 39.3%.
31.9% [29.3%; 34.5%] of employed men almost always or always feel time pressure.
– The estimated value is 31.9%, with a 95% confidence interval of 29.3% to 34.5%.
13.2. Sampling error
The total sample consists of 4,342 households. Every person in the household older than 10 years was requested to fill out a diary. This included 4,244 women and girls, and 3,619 men and boys.
The weighting procedure for the 2021/22 Time Use Survey consisted primarily of determining the design weights, adjusting for non-response, and performing a final calibration.
13.2.1. Sampling error - indicators
The number of observations (n) was used in the calculation of the indicator, and the standard error was also taken into account to reflect the precision of the estimate. Additionally the Confidence Interval was calculated.
95.4% [94.5%; 96.3%] of women and girls aged 10 and older spend time on housework.
– The estimated value is 95.4%, with a 95% confidence interval of 94.5% to 96.3%.
86.2% [84.7%; 87.8%] of men and boys aged 10 and older spend time on housework.
– The estimated value is 86.2%, with a 95% confidence interval of 84.7% to 87.8%.
Women and girls aged 10 and older spend on average 3h 7min [3h 2min; 3h 12min] per day on housework.
– The estimated value is 3h 7min, with a 95% confidence interval of 3h 2min to 3h 12min.
Men and boys aged 10 and older spend on average 1h 54min [1h 49min; 1h 58min] per day on housework.
– The estimated value is 1h 54min, with a 95% confidence interval of 1h 49min to 1h 58min.
36.7% [34.0%; 39.3%] of employed women almost always or always feel time pressure.
– The estimated value is 36.7%, with a 95% confidence interval of 34.0% to 39.3%.
31.9% [29.3%; 34.5%] of employed men almost always or always feel time pressure.
– The estimated value is 31.9%, with a 95% confidence interval of 29.3% to 34.5%.
13.2.2. Sampling error - proportion and confidence interval
| Parameter of interest (p̂) | Number of respondents - n | Standard error for p̂ | 95% confidence interval for p̂: Lower | 95% confidence interval for p̂: Upper |
|---|---|---|---|---|
| Percentage of population aged 15 and over spending daily on average more than 10 % of time working in paid work (weighted) |
Number of individuals completed at least 1 diary day (unweighted) |
|||
| 43.30579 | 7406 (>= 15 years), 7863 (>=10 years) | 0.7278858 | 41.87913 | 44.73245 |
Calculation of (p̂)
- Calculate for all individuals the sum of time spent working (ACL activity 1 employment) considering both diary days. (i.e 48 hours or 2*1440 minutes)
- NUMERATOR: Count of the individuals whose cumulative time spent in ACL 2018 activity 1 in both diary days was equal or higher than 10% i.e. 4.8 hours in total. If only one diary was completed then 10% = 2.4 hours
- DENOMINATOR: Total number of individuals that completed at least 1 diary day.
- To obtain (p̂) calculate the proportion: Numerator/ Denominator
13.2.3. Sampling error - method used for the variance (SE) estimation
The standard errors and variances were estimated using the Taylor linearization method.
The data were systematically checked for plausibility and imputed:
- A portion of the plausibility checking was implemented directly during data collection within the electronic questionnaire. Built-in filter logic ensured that respondents were only presented with questions that could be answered logically and consistently.
- The second phase of the plausibility checks was conducted using the statistical software R. In this stage, all variables were systematically examined for validity, and implausible or missing values were flagged with error codes. In cases of logical inconsistencies, attempts were made to infer consistent values from other respondent information. When this was not possible, or when values were missing altogether, imputation was applied. Initially, the verification focused on the assignment of codes to free-text diary entries, followed by consistency checks between these codes and the corresponding supplementary diary information.
- In a third step, responses from the final questionnaire were cross-checked against the diary entries. Implausible or flagged diary entries were subsequently corrected where possible, using contextual information from the individual and household questionnaires. In cases of discrepancy, the information from the questionnaires was given priority over the diary records.
In the case of item non-response, missing values were imputed using a k-nearest neighbor algorithm. The imputation rates range between 0.01% and 1.9%. To ensure the consistency of the imputed values, dependencies between responses were taken into account during the imputation process. Consequently, follow-up questions were also imputed when their corresponding answers were imputed, and logically incompatible response combinations were not allowed to be imputed.
13.3. Non-sampling error
Information on this concept is provided in the sub-concepts 13.3.1-13.3.5.
13.3.1. Coverage error
The de facto population of Austria living in non-institutional households (N) and the weighted poputlation in TUS survey (N_weighted) is available in the annex.
Annexes:
key figures_weights_TUS
13.3.1.1. Over-coverage - rate
The sampling frame only includes people living in non-institutional households in Austria. People living in student quarters, pension homes, homeless people and illegal immigrants are not included.
13.3.1.2. Common units - proportion
Not applicable.
13.3.2. Measurement error
Causes of measurement error:
- Unclear or misleading explanations provided by interviewers or in the survey materials
- Third-party information / proxy responses
- Typing or data entry errors
13.3.2.1. Questionnaire design and testing
A friendly user test was implemented before the field-phase.
13.3.2.2. Interviewer training
A handout was prepared, and also two videos - one tutorial for the interviewers, and one for the participants. Aditionally, there were two briefings where the survey was explained.
13.3.2.3. Proxy interview rates
13.3.3. Non response error
Since participation in the 2021/22 Time Use Survey was voluntary, a certain proportion of households in the gross sample were either unavailable for the survey or refused to participate. As a result, some household groups relevant to the survey may be underrepresented, which can potentially lead to bias in the results.
The purpose of the non-response weighting is to counteract such selective non-participation. To determine the non-response weights, the participation probability of each household (i) in the sample was estimated using a LASSO regression model. This model used all relevant explanatory variables Xₖ available in the sampling frame (the register of all eligible units). The regularization parameter λ for the regression was estimated using cross-validation.
The non-response weights were then derived as the product of the design weights and a factor based on the estimated participation probability.
13.3.3.1. Unit non-response - rate
Please see the Annex 1, sheet "13.3.3.1.Unit non-response".
13.3.3.1.1. Reasons for non-response
In household surveys conducted on a voluntary basis, non-response can be expected. Reasons for this may include refusal to participate or the inability to reach a knowledgeable respondent within the household. For example, employed individuals living in single-person households are generally more difficult to contact than larger families with children. Households in which the knowledgeable respondents do not speak German, or speak it only poorly, are also likely to be underrepresented. In surveys where participation is voluntary, systematic non-response can therefore be expected.
13.3.3.1.2. Number of households in the gross sample according to the final results of the survey
Please see the Annex 1, sheet "13.3.3.1.2. Gross sample".
13.3.3.1.3. Characteristics of non-respondents
In household surveys conducted on a voluntary basis, non-response can be expected. Reasons for this may include refusal to participate or the inability to reach a knowledgeable respondent within the household. For example, employed individuals living in single-person households are generally more difficult to contact than larger families with children. Households in which the knowledgeable respondents do not speak German, or speak it only poorly, are also likely to be underrepresented. In surveys where participation is voluntary, systematic non-response can therefore be expected.
13.3.3.1.4. Efforts to reduce non-response
- Telephone support
- Additional remuneration for interviewers
- Improvement of communication tools
- Elimination of pre-incentives and increase of post-incentives for households in the ZE sample
- Promotion of paper diaries
- Active telephone contact
13.3.3.1.5. Adjustment of weights in order to reduce non-response
Please refer to the Annex attached to 13.3.1.
13.3.3.1.6. Other comments regarding non-response errors
Not applicable.
13.3.3.1.7. Replacement of non-responding households (substitution)
| Substitution rate | Rate |
|---|---|
| i.e. Number of substitute households successfully interviewed/Achieved sample size | Not applicable |
13.3.3.1.8. Description of the substitution
Not applicable.
13.3.3.1.9. Qualitative assessment of the bias associated with unit non-response
Since participation in the 2021/22 Time Use Survey was voluntary, a certain proportion of households in the gross sample were either unavailable for the survey or refused to participate. As a result, some household groups relevant to the survey may be underrepresented, which can potentially lead to bias in the results.
The purpose of the non-response weighting is to counteract such selective non-participation. To determine the non-response weights, the participation probability of each household (i) in the sample was estimated using a LASSO regression model. This model used all relevant explanatory variables Xₖ available in the sampling frame (the register of all eligible units). The regularization parameter λ for the regression was estimated using cross-validation.
The non-response weights were then derived as the product of the design weights and a factor based on the estimated participation probability.
13.3.3.2. Item non-response - rate
Times that could not be assigned to any activity, as well as small gaps in the daily sequence, were manually completed. If no other classification was possible, they were filled with the activity “Other unspecified time use” (999).
13.3.3.2.1. Variables most subject to item non-response
Missing values in the questionnaire were imputed.
A total of 69 variables had at least one missing value. The questions with the highest item non-response were those on income and share of income. Income was imputed 134 times, and share of income 104 times.
13.3.4. Processing error
To identify and correct potential processing errors, comprehensive plausibility checks were applied. Staff members received continuous training, documentation was regularly updated, and processing issues were discussed in team meetings to identify and subsequently minimize potential errors.
Possible sources of error lay in the manual assignment of plain text entries to the corresponding activity codes. Processing the diaries and assigning free-text entries to codes required a high level of accuracy and consistency to ensure high-quality and valid results. Therefore, it was essential to develop a set of rules and train the staff involved in the processing accordingly. Ambiguous cases were regularly discussed and resolved within the project group, and the solutions were documented. Furthermore, by providing the activity nomenclature simultaneously and uniformly with current examples, coding was standardized to reduce errors to an absolute minimum. As a result of the subsequent comprehensive plausibility checks, processing errors could largely be identified and corrected.
A particular challenge in coding was posed by entries where the general nature of the term made it impossible to assign the activity to a specific activity code—for example, “looking at the phone,” “household,” “housework,” etc. In such cases, the time was split across multiple activity codes (e.g., “looking at the phone”: 1/3 reading online news, 1/3 texting, 1/3 social media). For activities that could not be clearly assigned to a specific activity code, most activity groups included a category labeled “Not further specified …”.
13.3.5. Model assumption error
None.
14.1. Timeliness
Data collection took place until the end of December 2022. The data became available at national level with the publication of the Report in December 2023. Data was delivered to Eurostat in March 2024.
14.1.1. Time lag - first result
Not applicable.
14.1.2. Time lag - final result
Not applicable.
14.2. Punctuality
Not requested.
14.2.1. Punctuality - delivery and publication
Not requested.
15.1. Comparability - geographical
At the international level, the 2021/22 Time Use Survey was based on the HETUS Guidelines 2018, which were updated for the HETUS 2020 survey wave. This allows for international comparisons, taking into account various national differences.
Regional comparability is very limited due to the sample size. Results were published by degree of urbanisation. A breakdown by federal states (NUTS2) is only meaningful to a limited extent and only for very specific activity groups due to the sample size.
Additionally, a second set of weights was used for the analysis by federal states, in which the federal states were also considered during calibration.
15.1.1. Asymmetry for mirror flow statistics - coefficient
Not applicable.
15.2. Comparability - over time
Due to the differing designs of the surveys conducted in 1981, 1992, and 2008/09, direct comparability with the results of the 2021/22 survey is only possible to a limited extent. It is important to note that presenting time series data is challenging, as the long intervals between the individual surveys have led to significant conceptual differences, and the methodology has been continuously improved over time. The main differences include:
Reporting unit: In the 1981 survey, individuals aged 19 and older were interviewed, whereas in the 1992, 2008/09, and 2021/22 surveys, individuals aged 10 and older were included. Unlike the 2008/09 Time Use Survey, which was a person-based survey, the 2021/22 Time Use Survey was conducted at the household level.
Number of survey days and survey period: The timing and duration of data collection varied between the two surveys before 2000 and the two most recent ones. In 2021/22 and 2008/09, data were collected throughout the entire year, while in 1981, only September was surveyed, and in 1992, only March and September served as data collection periods. Year-round data collection allows for better observation of seasonal effects, such as variations in the average duration of different activities throughout the year. The number of survey days also differs: in surveys up to 2008/09, each respondent was surveyed for only one day, whereas in 2021/22, all respondents completed the diary for two days. Therefore, comparing participation rates is difficult.
Changes in activity code assignment: Compared to the 2008/09 survey, new activities were added, and some activities were reassigned to different overarching categories. The activity codes in the HETUS Guidelines are regularly adapted to reflect changing living conditions and are updated for each HETUS wave. The 2021/22 Time Use Survey followed the HETUS 2018 Guidelines provided by Eurostat for the HETUS 2020 wave.
Change in number of timeslots: In the 2008/09 Time Use Survey, all activities were recorded in 15-minute timeslots, meaning each respondent filled out 84 timeslots per day. In the 2021/22 survey, the interval was reduced to 10 minutes, resulting in 144 timeslots per day per respondent.
Change in survey modes: In the 2008/09 Time Use Survey, household and personal questionnaires were administered directly by interviewers (CAPI). In the 2021/22 survey, depending on the subsample, data collection was conducted either by interviewers or via a web questionnaire (CAPI & CAWI). For the first time in Austria, a web application was also used for time recording in the 2021/22 survey. The 1981 and 1992 surveys had additional differences. For example, in the 1981 study, a retrospective interview was conducted with an interviewer about the previous day. Coding of responses was done by the interviewer, who made decisions in case of uncertainty. Since the 1992 survey, respondents have recorded their activities in their own words in the diary sheet, and coding of activities has been carried out afterward.
15.2.1. Length of comparable time series
Not requested.
15.3. Coherence - cross domain
As a quality criterion, coherence refers to the comparability of the survey results with those from other data sources.
To assess coherence, the results were compared with (international) studies and surveys, such as the 2022 Survey on Volunteering. However, comparisons with other surveys are only possible to a limited extent, as the methodological design of the surveys differs significantly, allowing only for approximate alignment.
The 2022 Survey on Volunteering was selected as an example of coherence with other data sources. A direct comparison with the volunteering survey is only partially possible, as it asked about voluntary activities in a non-time-specific manner—for example: “Are you volunteering for an organization or association in the field of disaster relief or emergency services?” The time aspect, which is central to the Time Use Survey, was not considered. In the Time Use Survey, voluntary activity had to be performed on one of the two randomly selected survey days to be recorded. This explains the difference in the share of participants between the two surveys.
Nevertheless, both surveys show similar trends: men are more active in formal volunteering, while women are more active in informal volunteering.
15.3.1. Coherence - sub annual and annual statistics
Not applicable.
15.3.2. Coherence - National Accounts
Not requested.
15.4. Coherence - internal
The data were systematically checked for plausibility and imputed.
A portion of the plausibility checking was implemented directly during data collection within the electronic questionnaire. Built-in filter logic ensured that respondents were only presented with questions that could be answered logically and consistently.
The second phase of the plausibility checks was conducted using the statistical software R. In this stage, all variables were systematically examined for validity, and implausible or missing values were flagged with error codes. In cases of logical inconsistencies, attempts were made to infer consistent values from other respondent information. When this was not possible, or when values were missing altogether, imputation was applied.
Initially, the verification focused on the assignment of codes to free-text diary entries, followed by consistency checks between these codes and the corresponding supplementary diary information. In a third step, responses from the final questionnaire were cross-checked against the diary entries. Implausible or flagged diary entries were subsequently corrected where possible, using contextual information from the individual and household questionnaires. In cases of discrepancy, the information from the questionnaires was given priority over the diary records.
In the case of item non-response, missing values were imputed using a k-nearest neighbor algorithm. The imputation rates range between 0.01% and 1.9%. To ensure the consistency of the imputed values, dependencies between responses were taken into account during the imputation process. Consequently, follow-up questions were also imputed when their corresponding answers were imputed, and logically incompatible response combinations were not allowed to be imputed.
Information on this concept is provided in the sub-concepts 16.1-16.4.
16.1. Costs of the survey
Not available.
16.2. Average time used for answering the survey questionnaires (in minutes)
15-20 minutes - unfortunately not recorded in detail
16.3. Average time used to fill in the diary (in minutes)
Not recorded.
16.4. Measures taken to reduce the cost and burden of the survey
The survey sample was divided into two parts. The Microcensus-based sample had a disproportionately higher representation of individuals who are typically harder to reach through online surveys. This part of the sample was conducted using CAPI (Computer-Assisted Personal Interviewing). The integration with the Microcensus aimed to reduce costs, as the interviewers were primarily financed through the Microcensus framework.
Participant burden was reduced through the development of an application, which made it easier and more flexible for respondents to fill out the diary throughout the day.
17.1. Data revision - policy
The standard documentation (mentioned under 10.7) is part of the quality management system and served as the basis for the Feedback-Talk with a subcommittee of the Austrian Statistical Council and Useres of the TUS data. This process serves as Data revision.
17.2. Data revision - practice
The standard documentation (mentioned under 10.7) is part of the quality management system and served as the basis for the Feedback-Talk with a subcommittee of the Austrian Statistical Council and Useres of the TUS data. This process serves as Data revision.
17.2.1. Data revision - average size
Not applicable.
18.1. Source data
Information on sampling design, sampling frame and size is available under the full metadata concepts 18.1.1-18.1.3.
18.1.1. Sampling frame
The sampling frame was derived from the so called “Rich Frame”, a pseudonymized micro data set containing every person registered in the Austrian housing and living register as main residence within private (non-institutional) households. It includes both personal and household specific attribute. The sampling frame and data basis for the sample selection is the Central Residence Register (ZMR). The ZMR was first populated after the last population census in 2001 by merging the municipal residence registers. It is continuously updated based on residence registrations from municipalities. It contains the address data of all registered main and secondary residences for individuals living in Austria.
When merging address data from different individuals within a household, discrepancies may occur due to variations in how addresses are written, which can lead to unrecognized household connections. Typically, there is only one household at a given address. However, in rare cases, an address may contain multiple households, understood as separate economic units. Whether an address includes multiple households can only be clearly determined during the data collection process.
It should also be noted that the so-called 'registration reality' does not always match the actual 'living reality' of individuals. This means that the household composition recorded during the survey may differ from that in the ZMR.
The sociodemographic characteristics of the sub-sample that was also surveyed as part of the Microcensus are collected during the Microcensus interviews.
| Type | Name of data source used for building the sampling frame |
|---|---|
| Population census | Register-based Census (Registerzählung) |
| Population register | Central Residence Register (Zentrales Melderegister - ZMR) |
| Household register | derives from Register-based Census and Central Residence Register |
| Dwelling register | Buildings and Dwellings Register (Gebäude- und Wohnungsregister - GWR) |
18.1.2. Sampling design of the survey
See below.
18.1.2.1. Sampling design(s)
A particular feature of the survey is that the sample consists of two subsamples. This design was chosen for reasons of cost efficiency. The two subsamples are:
- MZ Sample: This subsample was drawn from the Microcensus sample for the 4th quarter of 2021 and the 1st to 4th quarters of 2022 (only from first-time interviews in each quarter). The first step of data collection – completing the questionnaire section – was conducted in this subsample by interviewers visiting households in person, using Computer-Assisted Personal Interviews (CAPI). The gross sample included 7,778 households. Based on the assumption and experience that older individuals are more likely to participate in surveys when an interviewer goes through the questionnaire with them personally, this subsample included a disproportionately higher number of households with older persons. Oversampling occurred when at least one person in the household was aged 75 or older.
- ZE Sample: This subsample was drawn independently from the Central Residence Register (ZMR), included 11,000 private households, and was selected as a stratified random sample. Stratification variables included federal state, household type (single-parent household, multi-child household, no person aged 75 or older, other), and education level (low, other). If strata were too small, the two education categories were merged. The selection rate for households in the strata of single-parent households, multi-child households, and low education was increased. Households already surveyed in the Microcensus were subtracted from the target numbers per stratum before drawing the independent ZE sample. Five subsamples – one per quarter – were drawn. The first step of data collection – completing the questionnaire section – was conducted online using Computer-Assisted Web Interviews (CAWI).
| Type | Description of Sampling Design |
|---|---|
| Simple random sampling | |
| Systematic sampling | |
| Stratified sampling | |
| Cluster sampling | |
| Other |
See description above |
18.1.2.2. Ultimate sampling unit(s)
Households18.1.2.3. Oversampling of specific populations
Oversampling was different in the two sub-samples (as specified in 18.1.2.1).
18.1.2.4. Assumptions used for determining the sample size
The sample size for the Time Use Survey was determined based on cost-efficiency and representativeness. The total sample was divided into two subsamples:
The MZ sample was drawn from the Microcensus sample, leveraging existing infrastructure and interviewer resources to reduce costs. Oversampling of households with individuals aged 75 and older was applied, based on the assumption that older persons are more likely to participate when interviewed in person (CAPI).
The ZE sample was drawn independently from the Central Residence Register (ZMR) as a stratified random sample. Stratification variables included federal state, household type, and education level. To ensure adequate representation of specific groups (e.g. single-parent households, multi-child households, and households with low education), the selection rate for these strata was increased.
The overall sample size was thus guided by:
- Budgetary constraints
- Expected response rates
- Need for subgroup analysis
- Use of existing survey infrastructure (Microcensus)
18.1.3. Sample size
| Sample size |
Number |
| Gross sample size Formula: initial sample size= responding units + non-responding units (these two groups make up the group of eligible units) + units with unknown eligibility (e.g. because they are not able to be reached/ contacted), + ineligible units. |
18778 |
|---|---|
| Number of eligible units Formula: net sample size= Gross sample size - units with unknown eligibility – ineligible units |
Unknown |
| Achieved sample size = Total number of households which were successfully surveyed (interviews+diary) |
4342 adjusted net sample / 4528 net sample |
18.2. Frequency of data collection
No fixed frequency.
18.3. Data collection
See below.
18.3.1. Data collection method used
Mixed mode18.3.2. Mode(s) and instruments for data collection: household and individual questionnaire
| Interview |
Specify Yes or No if used |
% of completed household interviews |
|---|---|---|
| Paper assisted personal interview (PAPI) |
No |
|
| Computer-assisted personal interview (CAPI) |
Yes |
58% of the households of the adjusted net sample completed the questionnaire in CAPI. |
| Computer assisted telephone interview (CATI) |
No |
|
| Computer assisted web-interview (CAWI) |
Yes |
42% of the households of the adjusted net sample completed the questionnaire in CAWI. For the standalone subsample, only CAWI was available. Gross sample of this specific sub-sample was 11 000 households, adjusted net sample 1 825 households: 17% |
| Smart mode (e.g. smartphone app) |
No |
|
| Other (e.g. administrative data). |
No |
|
18.3.3. Mode(s) and instruments for data collection: time use diary
| Diary |
Specify Yes or No if used |
% of completed diaries |
|---|---|---|
| Paper diary |
Yes |
22,7% (1 781 of the total 7 863) |
| Computer based non online diary |
No |
|
| Online web diary |
Yes |
77,3% (6 082 of the total 7 863) |
| Mobile based diaries (Smartphone apps, etc.) |
No |
|
18.3.4. Variables completed from an external source
The sample was drawn from the Central Residence Register (ZMR), meaning that all variables used for sample selection and stratification—such as age, sex, postal code, and address—were based on register data. With the exception of the postal code and address, variables from the sampling frame (e.g. age) may differ from the actual living situation of respondents. Therefore, these variables were used solely for sampling purposes and not for further analysis.
After data collection, only the information provided directly by the respondents was used (and for calculation of regional variables the postal code). Moreover, no register data was used in the calculation of indicators or variables beyond the sampling process.
18.4. Data validation
The validation process ensured the accuracy, consistency, and plausibility of both diary and questionnaire data. It was carried out in multiple stages:
Diary Validation
Manual Review & Coding Checks: All diary entries were reviewed for correct assignment of codes (main/secondary activity, internet use, location, presence of others).
Correction of Errors: Multiple entries in activity fields were cleaned, and missing codes were added.
Plausibility Checks
Implausible combinations (e.g., traveling without a transport mode, playing games alone) were flagged or corrected.
Activities like 'sleeping' were checked for inappropriate secondary activities.
'Internet use' and 'presence of others' were validated against the nature of the activity.
Cross-Referencing with Final Questionnaire: Entries were checked for consistency with end-of-day responses (e.g., type of day, most pleasant/unpleasant activity).
Contextual Corrections: If inconsistencies were found, corrections were made using contextual data from household and personal questionnaires.
Questionnaire Validation
Built-in Logic Checks: The electronic questionnaire included filters and alerts for unlikely responses (e.g., 80+ work hours/week).
Statistical Validation in R: Variables were systematically checked for valid values.
Implausible or missing data were flagged and either corrected using related responses, or imputed.
Diary vs. Questionnaire Consistency
Diary entries were adjusted to match questionnaire data, which was treated as the authoritative source.
For example, if a diary indicated childcare but the questionnaire showed no children in the household, the diary was corrected.
Presence of household members in diary entries had to align with household composition.
18.5. Data compilation
Data compilation includes information related to the imputation rate, methods applied to correct for item non-response, if applicable, and information on the calculation of weighting factors and weight adjustments. For more information, see concepts 18.5.1 - 18.5.3 in the Full metadata.
18.5.1. Imputation - rate
In 54 variables, at least one value was imputed. The ratio of imputed to total numer of values (incl. filter) lies between 1.90% and 0.01%.
- How many hours is your child visiting the Kindergarden per week?; 1,9% imputed
- Is your child visiting public or private Kindergarden?; 1,81% imputed
- Is you child visisting the Kindergarden?; 1,25% imputed
- What percentage of the household income do you contribute?; 1,15%
- How many days of vacation entitlement do you have?; 0,98% imputed
- Is your child visiting infant care?; 0,73% imputed
- Self-employment income; 0,64% imputed
- All other less than 0,5%
18.5.2. Method applied to correct for 'item non-response'
| Methods | Specify Yes or No if used |
|---|---|
| Simple imputation (deterministic) method | Yes, Hot Deck and KNN |
| Simple imputation (stochastic) method | No |
| Multiple imputation approach | No |
| Other | No |
18.5.3. Calculation of weighting factors and weight adjustments
The extrapolation of the Time Use Survey results was based on a multi-step weighting procedure designed to ensure representativeness and correct for sampling and response biases. The process included:
1. Design Weights
The survey was conducted over five quarters, each with a differently sized subsample. Although the same stratification variables were used across quarters, the selection rates varied. Design weights were calculated to adjust for the probability of selection within each stratum and quarter. Due to low sample sizes in certain strata, stratification was aggregated across federal states. Since each subsample was drawn independently from the full population, the combined design weights initially represented five times the population size. To correct for overlapping quarters (subsamples 1 and 5), their weights were halved.
2. Non-Response Adjustment
Participation in the survey was voluntary, resulting in potential non-response bias. To address this, non-response weights were calculated by estimating each household’s probability of participation using a LASSO regression model based on available auxiliary variables from the sampling frame. The regularization parameter was selected via cross-validation. The resulting non-response weights were derived by multiplying the design weights by a factor based on the estimated participation probability.
3. Final Calibration
A final calibration step was performed to align the weighted sample with known population margins from official statistics (Microcensus Q4 2021–Q4 2022). This raking procedure was applied at both the household and individual level.
Calibration Variables:
Individual level: Age (8 groups), employment type (7 categories crossed with gender), education level (7 categories), and equal distribution of individuals aged 10+ across survey months.
Household level: Urbanization level (3 categories) crossed with household size (5 categories) and household type (e.g., single-person, families with/without children).
The calibration was implemented using a raking ratio method in R, developed by Statistics Austria.
4. Weekday Adjustment
Respondents typically completed diaries for both a weekday and a weekend or holiday. After calibration, weights were distributed across weekdays to reflect the actual distribution of days in a calendar year.
5. Alternative Weighting for Regional Analysis
For special evaluations, an alternative weighting variant was created that included population counts per federal state as an additional calibration margin. In this version, distributions for employment type, gender, urbanization, household size, and household type were used without cross-tabulation.
18.6. Adjustment
Not applicable.
18.6.1. Seasonal adjustment
Not applicable.
- Survey name(s) in the national language(s): Zeitverwendungserhebung
- Survey name in English: Time use survey
- Year(s) of (data collection) of the survey: 2021/2022
- Link to the survey website: Time use (in English) and Zeitverwendung (in German)
- National questionnaire: please refer to the annex
Annexes:
questionnaire
8 June 2026
The survey was strictly aligned with the HETUS guidelines, without any deviations.
The survey units of the Time Use Survey are private households in Austria.
This includes all individuals aged 10 and over living in private households at addresses where at least one person has their main residence registered in the Central Population Register (ZMR).
Austrian residential population in private households aged 10 and over; approx. 8 million people.
Austria
HETUS should cover a full 12 months period, i.e. 365 consecutive days. Each respondent should fill in the diary for two days, one weekday (Monday to Friday) and one weekend day (Saturday, Sunday).
There were no deviations from the guidelines - except that weekdays were without public holidays and weekend days included weekdays which were public holidays. The survey covered all 12 months (5 quartals; 4th quartal 2021 - 4th quartal 2022; the 4th quartal is in the period twice because the response at the beginning was too low).
95.4% [94.5%; 96.3%] of women and girls aged 10 and older spend time on housework.
– The estimated value is 95.4%, with a 95% confidence interval of 94.5% to 96.3%.
86.2% [84.7%; 87.8%] of men and boys aged 10 and older spend time on housework.
– The estimated value is 86.2%, with a 95% confidence interval of 84.7% to 87.8%.
Women and girls aged 10 and older spend on average 3h 7min [3h 2min; 3h 12min] per day on housework.
– The estimated value is 3h 7min, with a 95% confidence interval of 3h 2min to 3h 12min.
Men and boys aged 10 and older spend on average 1h 54min [1h 49min; 1h 58min] per day on housework.
– The estimated value is 1h 54min, with a 95% confidence interval of 1h 49min to 1h 58min.
36.7% [34.0%; 39.3%] of employed women almost always or always feel time pressure.
– The estimated value is 36.7%, with a 95% confidence interval of 34.0% to 39.3%.
31.9% [29.3%; 34.5%] of employed men almost always or always feel time pressure.
– The estimated value is 31.9%, with a 95% confidence interval of 29.3% to 34.5%.
The units of observation are 144 ten-minute time intervals per day, also referred to as "timeslots", over the course of two days.
In these intervals, respondents recorded the activities they engaged in, using their own words.
So the units of measure are minutes per day;
- minutes per day spending time with a specific activity
- minutes per day spending time with a specific person (partner, child under 18 living in the same household, parent living in the same household, etc.)
- minutes per day spending time at a specific location (home, work, train, car, etc.)
Data compilation includes information related to the imputation rate, methods applied to correct for item non-response, if applicable, and information on the calculation of weighting factors and weight adjustments. For more information, see concepts 18.5.1 - 18.5.3 in the Full metadata.
Information on sampling design, sampling frame and size is available under the full metadata concepts 18.1.1-18.1.3.
No regularity of publication on TUS is agreed upon. Previous surveys were conducted in 1981, 1992, 2008/09. The survey 2021/2022 was the fourth national implementation.
Data collection took place until the end of December 2022. The data became available at national level with the publication of the Report in December 2023. Data was delivered to Eurostat in March 2024.
At the international level, the 2021/22 Time Use Survey was based on the HETUS Guidelines 2018, which were updated for the HETUS 2020 survey wave. This allows for international comparisons, taking into account various national differences.
Regional comparability is very limited due to the sample size. Results were published by degree of urbanisation. A breakdown by federal states (NUTS2) is only meaningful to a limited extent and only for very specific activity groups due to the sample size.
Additionally, a second set of weights was used for the analysis by federal states, in which the federal states were also considered during calibration.
Due to the differing designs of the surveys conducted in 1981, 1992, and 2008/09, direct comparability with the results of the 2021/22 survey is only possible to a limited extent. It is important to note that presenting time series data is challenging, as the long intervals between the individual surveys have led to significant conceptual differences, and the methodology has been continuously improved over time. The main differences include:
Reporting unit: In the 1981 survey, individuals aged 19 and older were interviewed, whereas in the 1992, 2008/09, and 2021/22 surveys, individuals aged 10 and older were included. Unlike the 2008/09 Time Use Survey, which was a person-based survey, the 2021/22 Time Use Survey was conducted at the household level.
Number of survey days and survey period: The timing and duration of data collection varied between the two surveys before 2000 and the two most recent ones. In 2021/22 and 2008/09, data were collected throughout the entire year, while in 1981, only September was surveyed, and in 1992, only March and September served as data collection periods. Year-round data collection allows for better observation of seasonal effects, such as variations in the average duration of different activities throughout the year. The number of survey days also differs: in surveys up to 2008/09, each respondent was surveyed for only one day, whereas in 2021/22, all respondents completed the diary for two days. Therefore, comparing participation rates is difficult.
Changes in activity code assignment: Compared to the 2008/09 survey, new activities were added, and some activities were reassigned to different overarching categories. The activity codes in the HETUS Guidelines are regularly adapted to reflect changing living conditions and are updated for each HETUS wave. The 2021/22 Time Use Survey followed the HETUS 2018 Guidelines provided by Eurostat for the HETUS 2020 wave.
Change in number of timeslots: In the 2008/09 Time Use Survey, all activities were recorded in 15-minute timeslots, meaning each respondent filled out 84 timeslots per day. In the 2021/22 survey, the interval was reduced to 10 minutes, resulting in 144 timeslots per day per respondent.
Change in survey modes: In the 2008/09 Time Use Survey, household and personal questionnaires were administered directly by interviewers (CAPI). In the 2021/22 survey, depending on the subsample, data collection was conducted either by interviewers or via a web questionnaire (CAPI & CAWI). For the first time in Austria, a web application was also used for time recording in the 2021/22 survey. The 1981 and 1992 surveys had additional differences. For example, in the 1981 study, a retrospective interview was conducted with an interviewer about the previous day. Coding of responses was done by the interviewer, who made decisions in case of uncertainty. Since the 1992 survey, respondents have recorded their activities in their own words in the diary sheet, and coding of activities has been carried out afterward.
