A novel procedure to survey travel behaviour, time use, and goods consumption from the same individuals
- Open Access
- 21.10.2025
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by (Link öffnet in neuem Fenster)
Abstract
Introduction
The data requirements for transport research have changed in an increasingly connected world. The connectedness has weakened the link between time use and locations as well as the strict distinction between activity time use and travel. Those trends brought forth several new forms of time use and travel such as ‘work from home’ (Elldér 2020; Hensher et al. 2021; Stiles and Smart 2021) and ‘multitasking while travelling’ (Hartwig et al. 2024; Lyons and Urry 2005; Malokin et al. 2021). At the same time, transport models have also advanced, driven by these trends as well as by improved computing power and data sources. A growing number of models require disaggregated information on time use and travel behaviour, sometimes also on goods consumption, at the person or household level. This applies in particular to activity-based models (see e.g., Bhat and Koppelman (1999); Rasouli and Timmermans (2014), agent-based models (see e.g., Bastarianto et al. (2023); Huang et al. (2022) and value-of-time models (see e.g., Munizaga et al. (2008); Hössinger et al. (2020).
In response to these challenges and thanks to emerging technologies, the methods of (travel) data collection have also evolved and branched out in various directions. One prominent direction is the collection of data from mobile devices or internet use. These techniques can gather large amounts of data partly in real-time, but they cannot (yet) provide all relevant information on consumers’ choices and their determinants in a broad sense. Travel diaries still provide a fairly comprehensive insight into travel behaviour at the individual level (International Transport Forum 2021), but limited information on non-travel activities and expenditures. These latter aspects are covered by other survey types, namely, time use and household budget surveys. One option to obtain a dataset covering all this information connected at the individual level is to merge data from different surveys. However, this procedure yields only probabilistic connections, which blur the consumers’ trade-offs between travel, time use, and budget assignment (e.g., working from home vs. driving to the office; going to a shop vs. paying for delivery).
Anzeige
In this paper, we report on a combined household budget and mobility-activity survey (HBS-MAS), which addresses the aforementioned gap to obtain cross-linked information on all respondents’ trips and activities (over one week, i.e. one work-leisure cycle) as well as all expenditures of the same individuals (collected over two weeks). The resulting dataset covers all relevant aspects of household production from the same individual, especially the choices on (1) the time assigned to different activities, (2) the locations where these activities are performed, (3) the trips made to travel between these locations, and (4) the consumer goods to which the budget is assigned. This information can be used in the aforementioned models to better understand the ongoing changes in consumers’ (travel) choices, to successfully plan mobility services, and to implement suitable transport policies.
We conducted a mobility activity survey (MAS) from a subsample of the official household budget survey (HBS) of the National Statistics Office. Although, this means a gap of a few weeks between HBS and MAS, it allows to collect both time use and expenditures from the same individual as representations of their long-term equilibrium without overburdening respondents. The gains from the decoupled data collection are twofold: (i) expenditure data were collected at high quality according to the state of the art, and (ii) the lower respondent burden of the MAS (a few weeks after the HBS) allowed us to collect more information on travel and health, as specified in Sect. Previous approaches for obtaining combined data sets from surveys.
The remainder of this paper is organized as follows: Sect. Previous approaches for obtaining combined data sets from surveys gives an overview of the prevailing approaches for obtaining suitable data for different modelling purposes and the issues that go along with each. Section Survey process and preparation describes the survey process, including design considerations, questionnaire content, survey conduct, and ex-ante estimation of the response rate. Section Response compares the observed response rates with ex-ante estimates. Section Results describes the sample and evaluates its representativeness by comparing various indicators to the relevant benchmark surveys. Finally, in Sect. Conclusion and outlook we draw conclusions on the merits of the survey method and elaborate on future improvements and opportunities for analysis of the data.
Previous approaches for obtaining combined data sets from surveys
The lack of data sets containing comprehensive information on time used for activities in and out of the home as well as choices of activity locations, transport modes and consumer goods, led to the development of several auxiliary approaches in the past. Three different methods can be distinguished in principle: (a) inferring missing information from surveys, (b) merging information from separate surveys, and (c) combining different survey types into one.
Anzeige
A common approach for a) is the use of data from travel diaries, in which activities are inferred from the trip purposes (Aschauer et al. 2018; Astroza et al. 2018; Axhausen et al. 2002). This approach has its limitations because in-home activities are usually not recorded in travel diaries, although the substitution of in-home and out-of-home activities is a key determinant of travel demand (work from home vs. office work, shopping at the shop vs. online etc.). Another option is the inference of trips from the activity “travel” or from location changes for stationary activities in time use surveys (Harms et al. 2018; Hertkorn and Kracht 2002), but detailed information on the activity location, changes in locations (i.e., trips), and chosen travel modes is missing and has to be imputed. Furthermore, the fixed time interval in time use surveys makes it difficult to obtain precise travel times, especially for short trips (Gerike et al. 2015). More data imputations will be necessary when trip or time use data for periods shorter than one week (e.g., one or two days) is used (Jara-Díaz and Candia 2020). The same problem arises with the combination of data on travel and activities with expenditure data. A commonly used data set that combines time use and expenditure data is the LISS panel (Longitudinal Internet Studies for the Social Sciences; see Hays et al. (2015), which is a long-running registry-based online survey in the Netherlands that contains average activity durations and expenditures on a monthly basis and of which the data is freely available. Trip details are, however, not reported and the averaging questions raise a validity issue as they are well known to produce a larger bias than diary-based surveys (Bonke 2005; Browning and Gørtz 2012). Another option is to use pre-existing data sets from specific contexts, such as leisure-focused time use and expenditure data (Dane et al. 2015). Such data usually also don’t include detailed trip information, and the specific context limits the extent to which the results can be generalised.
An option for b) is the probabilistic merger of participants from time use and household budget surveys (Konduri et al. 2011), which can be based on an elaborate data fusion concept (Hawkins and Habib 2022). Although both the time use and expenditure data are presumably of high quality in themselves, merging may be questionable for estimating values of time given that these values are based on the trade-offs between time use and expenditure allocation at the individual level. Those trade-offs are blurred if data from different individuals is being merged (Hössinger et al. 2020). Castro et al. (2012) concede in this context that “one can certainly debate the merits and appropriateness of such a synthetic data generation procedure, suffice it to say that the authors were not able to obtain any data set which collected both time use and expenditure data”.
Most core-satellite surveys belong to this category. They are a combination of a core survey that focusses on key data, thereby keeping response burden low, with one or more satellite surveys. These satellite surveys target specific sub-populations or behaviours (e.g., long-distance trips, attitudinal questions, multi-day travel diaries) that would be impractical or not cost-effective to collect in the large-scale core survey (Loa et al. 2017). This approach can also be used to correct biases in either the core survey or one of the satellite surveys, particularly when leveraging data from sources perceived as less biased, such as cordon count, smart card, mobile phone or census data (Bwambale et al. 2021; Li et al. 2024). While allowing for a flexible design of the satellite surveys and expanding survey coverage to populations generally not reached by traditional survey formats, the core-satellite approach is not without its weaknesses. Most often, the satellite is not integrated with the core survey and data from individuals needs to be merged using probabilistic models, requiring careful harmonisation of definitions and classifications between surveys.
A straightforward although expensive option c) are combined surveys that collect data on all four choices (activities, locations, mode, goods) from the same decision maker. They combine three types of surveys that are usually conducted separately: (i) travel surveys collecting information about a person’s trips along with the trip attributes such as origin and destination, used transport modes etc.; (ii) time use surveys, which capture the time assigned to different types of activities, and (iii) household budget surveys, which report the expenses allocated to different goods.
Alho et al. (2022) use a smartphone app-based tool to conduct such a combined travel, time use, and expenditure survey. The survey tool is capable of automatically sensing trip characteristics and also allows to uploading photos of shopping receipts as well as forwarding mails pertaining to online purchases. These technologies were applied to reduce the high response burden of such surveys. In a test phase with a limited number of respondents recruited through convenience sampling they obtained promising data and could prove their concept. During the test phase, a non-negligible amount of direct support and instructional materials had to be given to participants to ensure that they used the app properly and gave it the necessary permissions to allow it to run permanently in the background. The authors also remarked how the elevated response burden from reporting expenditures might incentivise underreporting unless addressed through a suitable process that either detects missing entries or incentivises completeness.
Winkler et al. (2022) describe another smartphone app-based study that was used for a large-scale survey with more than 1 300 participants who tracked their time use, trips, and expenditures for 28 days. Participants filled out a first questionnaire with person-related questions before beginning the tracking. After the tracking period, they were asked to provide monthly and long-term expenses as well as attitudinal questions in a second questionnaire. A high-quality, comprehensive data set could be collected comprising a sufficiently large sample, characteristic of the German-speaking Swiss population. The authors mention as limitations of the survey format that the development of the mobile app and keeping it up to date and functional proved somewhat costly and time-consuming, as no suitable ready-made solutions were available. Additionally, assistance had to be provided to participants to solve technical problems and ensure continuous participation over the rather long study period. In terms of data quality, the authors state that the frequent choice of the time use category “other” limits the comparability with conventional time use surveys. Data on secondary activities was not collected, and expenditures were collected from individuals rather than households, how it is typically done in household budget surveys. It is therefore probable that not all of the total household spending was captured with this approach which limits comparability with other surveys.
The Mobility-Activity-Expenditure Diary (MAED; Aschauer et al. (2018b) is the direct precursor of the HBS-MAS reported in this paper. It is a self-administered mail-back survey, which collects trips, activities, and expenditures simultaneously from the same individuals in a diary-based format over a one-week period. It has been adapted to fit different contexts and study purposes (Lizana et al. 2020). With respect to data quality, we found that travel behaviour and non-travel activities were reported at high quality. The expenditure data however required extensive preparation before it was suitable for modelling (Hössinger et al. 2020). Increasing its accuracy would require longer observation periods and preferably personal support of respondents but extending the observation period of the MAED would increase the respondent burden beyond the level tolerated by many participants. The HBS-MAS incorporates the lessons learned from the MAED. The result is a combined but temporally decoupled household budget and mobility activity survey with several new features:
-
The temporal decoupling of the two surveys allowed us to gather more information, in particular (i) secondary activities conducted alongside the main activity (multitasking); (ii) the use of digital tools for travel planning and ticketing, and (iii) questions related to physical activity, health, and well-being both in the travel and activity diary and in the person section of the questionnaire.
-
Alongside written-postal participation, we also offered online participation to save cost and to better involve the younger generation in particular.
-
We conducted 3 survey waves, which by chance cover the effects of the COVID lockdown in Austria: wave 1 was conducted in the fall of 2019 (before the pandemic), wave 2 in the spring of 2020 (during 1st lockdown), and wave 3 (with participants of wave 1 and 2) in the fall of 2021 (relaxation and recovery phase).
Secondary activities or multitasking are commonplace in time use surveys; their inclusion is standard according to the Harmonised European Time Use Survey (HETUS) guidelines (Eurostat 2019). In travel-related surveys, it is not yet common to ask for secondary activities. Some of them did so, e.g. Malokin et al. (2021) as well as Varghese and Jana (2018), and subsequent analysis of the collected data has shown that multitasking has a significant influence on the Value of Travel Time Savings (VTTS) which is a central variable in, e.g., cost-benefit analysis in transport planning. Yet, many travel surveys do not ask for multitasking, among them the most recent Austrian National Travel Survey 2013/2014 (ANTS) (Tomschy et al. 2014) and the MAED survey. An update of the MAED format with multitasking questions is thus a worthwhile addition.
Survey process and preparation
General considerations
To obtain the expenditure data in better quality, we established a cooperation with the Austrian national statistical office Statistics Austria in their nationwide household budget survey (HBS) conducted from May 2019 to June 2020, which provides a detailed record of all expenditures of private households in Austria (Kronsteiner-Mann and Braun 2022a). A sub-sample of HBS participants took part in a subsequent mobility activity survey (MAS) conducted by the authors. The reporting weeks of the HBS and MAS were some weeks apart, which we consider a minor weakness, as the allocation of expenses and the use of purchased goods do in most cases not occur simultaneously (unlike time allocation and activities). This is particularly true for long-term expenses such as clothes or furniture and regular payments such as rental costs. But also short-term goods like groceries are usually not consumed during the shopping activity and often not by the individual who bought them. Splitting the response burden over two separate periods (HBS and MAS) enabled us, on the other hand, to gather more detailed information on activities in the MAS than in a combined survey.
Survey process
The HBS of Statistics Austria lasted one year from the end of May 2019 to June 2020. The accompanying MAS started in September 2019 (see Fig. 1). The COVID-19 pandemic reached Austria during the survey period. The first lockdown was put in place from 16 March to 30 April 2020. It coincides with the beginning of the second survey wave, which was carried out following official safety precautions and a strong push towards online participation. Wave 2 began in March 2020 and lasted until August 2020. In order to capture long-term behaviour shifts due to the COVID-19 pandemic and subsequent lockdowns, the participants of waves 1 and 2 were re-contacted and asked to participate in a third wave, which began in September 2021 and lasted until December 2021. During some parts of this period, special restrictions for unvaccinated persons were in place.
Fig. 1
Time scale of the three survey waves and the Covid-19 lockdowns in Austria. The histograms illustrate the share of participants whose reporting weeks began in the respective week
The participants filled out the MAS diary for seven consecutive days, so it represents a continuous record of 168 h, which represents a complete work-leisure cycle. An incentive in the form of a 25 € voucher was provided for each person on completion of both the survey and validation process.
Figure 2 shows an overview of the support and validation process. Participants were recruited after completion of the two-week HBS of Statistics Austria. In wave 1, they were invited to participate in the same mode (pen-and-paper, or online) in which they had already participated in the HBS. In wave 2, they were encouraged to participate online mainly to limit in-person contact with the HBS interviewers due to COVID-19. Depending on their participation mode, participants either logged in to the online questionnaire or received a paper questionnaire (see Annexes 1 and 2) at their home address after requesting it by sending back a postcard handed to them by the interviewers of Statistics Austria. All persons in a participating household aged 16 years and up were invited to participate, but it was also possible that only a subset of these household members took part. For online participation, all household members shared the same ID to log in to the survey platform. There they could add household members as participants which created a new person questionnaire with a corresponding time use and mobility diary for each participant.
Fig. 2
Support and validation process. The tasks carried out by the survey participants are printed in italics
Each household was asked to provide a telephone number. It was used for a support call by a survey team member after the first reporting day, to provide assistance in case of difficulties with reporting, or to motivate those participants who were hesitant to start filling in the diary. Upon completion, participants were again called to validate implausible, incorrect, or incomplete items that emerged at data entry (pen-and-paper) or didn’t pass second-level plausibility checks (online). Finally, households received a postal thank you letter with a voucher for each member who had passed the validation process.
Fig. 3
Histograms of response time by wave and participation mode
Figure 3 shows the response time by wave and participation mode. For waves 1 and 2, response time was calculated as the time difference between the last HBS reporting day – when respondents were invited to participate in the MAS by HBS interviewers – and the day the first activity was reported in the MAS. For pen-and-paper participants, 5 days were subtracted to account for the time needed to send back the postcard and the duration of the postal delivery of the questionnaire. For wave 3 where participants were recontacted by our survey team, we took the time difference of the day they were successfully reached via telephone and the day the first activity was reported in the MAS. For pen-and-paper participants, 2 days were subtracted to account for the duration of the postal delivery time of the questionnaire. The median response time for waves 1 and 2 was 13 days for pen-and-paper and 7 days for online participants. For wave 3, it was 8 days for pen-and-paper and 5 days for online participants.
Content of the MAS questionnaires
Since the HBS is well documented (Kronsteiner-Mann and Braun 2022a) and not part of our innovative development, we concentrate on the content of the MAS. The pen-and-paper version was based on the questionnaires from the Mobility, Activity, and Expenditure Diary (MAED) format (Aschauer et al. 2018b). It starts with several introductory pages, including a brief set of instructions, a list of activity categories, a page for recording frequently visited locations (such as home, work, and grocery stores) along with a keyword to facilitate reporting of trips to these destinations, and a few prefilled sample diary pages. The actual questionnaire consists of two parts:
-
The household questionnaire (see Annex 1) was to be filled in only once per household, thereby lowering the marginal burden for each additional participating member. It contains questions on socio-demographic attributes, employment and income of all household members, as well as available mobility tools (vehicles, parking facilities, PT discount cards and season tickets).
-
The person questionnaire (see Annex 2) was to be filled in by each participating person. It includes a few additional one-off questions relating to health and work satisfaction. But the main content is a mobility and activity diary in terms of a sequence of identical pages to report trips and ensuing activities at the destination. The trip section resembles the proven KONTIV design (Bundesanstalt für Straßenwesen and Socialdata Institut für Verkehrs- und Infrastrukturforschung 1987). But instead of a single trip purpose, all performed activities at the destination are reported along with their duration in open time intervals and predefined activity categories, similar to a traditional time use survey. An open field is provided to report secondary activities for both trips and stationary activities.
The online version comprises the same questions but some completion aids and plausibility checks: (i) an auto-completion field for car brands, models and engine types, (ii) automatic insertion of the destination of the previous trip as the start address of the next one, and (iii) mouse-over pop-ups with additional explanations instead of the introductory instructions. After logging in, participants landed on an overview page where they could see the status of all questionnaires and access them.
All addresses were recorded in text form and verified for validity. During data processing, we queried the geocoordinates of all locations, the distances between them, and the mode-specific travel times using online route planners (Google Maps Web Services 2022; Verkehrsauskunft Österreich GmbH 2020). We chose not to use GPS trackers for logging the routes and destinations, although they could automatically record high-resolution coordinates and remedy issues like forgotten trips and error-prone distance reporting. The drawback, however, is that many participants still have reservations against such tools because of usability issues and they have little control over the data that it reveals. Moreover, extensive data processing is still required which increases the complexity of the survey process (Allström et al. 2017; Harrison et al. 2020).
The body of the main questions stayed the same throughout all survey waves. In wave 2, questions about COVID-19-related changes were added. In wave 3, questions about the performance of online activities and online shopping as well as retrospective questions about the situation during the third lockdown were added. The annex provides the paper questionnaires in their most extensive form used in wave 3. The changes, compared to previous waves, are marked.
Ex-ante response rate estimation
During the preparation of the MAS, we conducted an ex-ante estimation of the response rate using the respective tool developed by Axhausen and Weis (2010) and further updated by Schmid and Axhausen (2019). It is based on a logit model, which was trained with the response burden and observed response rates of 65 travel surveys.
To estimate the response burden of the MAS, we used the same point system as the authors of the model, but we added point scores for some specific questions in our survey, e.g., complex numerical answers such as income and text addresses that might require a look-up. The burden of items which were relevant only for employed persons was multiplied by their share in the population (0.72). Trip items were multiplied with the average number of trips per week (23.10) and – for car-specific questions – with the average number of car trips per week (10.86) from the Austrian National Travel Survey (Tomschy et al. 2014). The burden of activity-related questions was multiplied by 70, which is the average weekly number of activities according to previous experience.
Table 1
Response burden and ex-ante estimated response rate
Wave | Mode | Response burden of | ||||
|---|---|---|---|---|---|---|
Household questionnaire | Person questionnaire | Trip & activity diary | Total | Estimated response rate | ||
1 | paper | 60.0 | 26.0 | 651.4 | 737.5 | 24.5% |
online | 39.0 | 40.5 | 613.3 | 692.9 | 24.8% | |
2 | paper | 61.5 | 28.9 | 651.4 | 741.8 | 24.5% |
online | 39.0 | 46.3 | 613.3 | 698.7 | 24.7% | |
3 | paper | 61.9 | 123.6 | 641.4 | 826.9 | 24.0% |
online | 43.5 | 142.7 | 615.3 | 801.6 | 24.1% | |
1/2 + 3* | paper | 1544.6 | 20.0% | |||
online | 1519.3 | 20.2% | ||||
The response burden for the different questionnaires and the estimated response rates are presented in Table 1. The shifts between household and person questionnaires in the paper and online versions are due to the shift of some items between these questionnaires for layout reasons. The burden of the online version is slightly lower due to some auto-completion features. Regarding the survey waves, the burden increased slightly from wave 1 to wave 2 and again to wave 3. This results from the inclusion of questions about changes due to the COVID-19 pandemic and the addition of retrospective questions in wave 3. Only participants who had finished either wave 1 or 2 could participate in wave 3. The row 1/2 + 3 describes the total response burden across all waves. It is worth noting that the paper diary consists of 56 pages, 48 of which are generic copies for the reporting of trips and activities. The large stock of pages prevented participants from running out of pages, but the thickness of the diary might have scared away some potential participants.
The response rates were estimated from the total response burdens using the logit model by Schmid and Axhausen (2019) which is described in Eq. 1.
$$ \log \left( {\frac{{responserate_{i} }}{{100 - ~responserate_{i} }}} \right) = ~\alpha _{i} + \beta _{i} \frac{{responseburden}}{{1000}} + \varepsilon _{i} $$
(1)
Equation 1: Logit model for the ex-ante estimation of the response rate based on the response burden.
They estimated a constant αi and a coefficient βi for surveys with both prior recruitment and incentive, and for surveys without both components, but not for our case of no prior recruitment but incentive. As αi captures the initial motivation through recruitment (1.26 vs. − 0.89 with and without recruitment) and βi the decrease in response rate with increasing response burden for participants that receive (− 0.32) or don’t receive an incentive (− 1.17 vs. − 1.55 without and with recruitment), we took those parameters that represent a survey without recruitment and with an incentive (αi: − 0.89; βi: − 0.32). The resulting response rates hover around 24–25% (20%, for participants of wave 3 where the response burden of wave 1/2 and wave 3 are added). These estimates are compared to the observed response rates in Sect. 4.1.
Pre-test and adaptations
As both the mobility and activity diary are proven designs based on conventional travel and time use surveys and their combination had also been tested previously in the MAED survey, we conducted merely cognitive pre-tests with members of the institute and students. The focus was on newly added questions. As a result of the test, we adjusted the wording of some questions and the instructions on how to classify activities to reduce ambiguity, and we added some pre-filled sample pages.
Response
Observed response rates
The response rates of the MAS were calculated at the household level and are presented in Table 2. The gross sample of each wave consists of those households who finished the HBS in the respective period. In waves 1 and 2 about one third of those households who started the MAS dropped out during the process. The share of usable interviews among those who completed participation is quite high; this is a result of the intensive support and validation process. The share of households with usable interviews corresponds to the AAPOR (American Association for Public Opinion Research) Response Rate 1 based on the response rate calculator Version 4.1 (AAPOR, 2016). All other AAPOR response rates are reported in Annex 3: AAPOR response, cooperation, refusal and contact rates.
Table 2
Response rates of households
Wave | Absolute | Percent | ||||||
|---|---|---|---|---|---|---|---|---|
1 | 2 | 3* | 1/2 + 3** | 1 | 2 | 3 | 1/2 + 3* | |
Gross sample | 1 541 | 1 790 | 553 | 3 331 | 100.0 | 100.0 | 100.0 | 100.0 |
Participation started | 414 | 440 | 389 | 389 | 26.9 | 24.6 | 70.3 | 11.7 |
Participation completed | 266 | 303 | 383 | 383 | 17.3 | 16.9 | 69.3 | 11.5 |
Usable net sample | 258 | 295 | 375 | 375 | 16.7 | 16.6 | 67.8 | 11.3 |
Comparison with response rates from the estimation and from other studies
A comparison of the expected response rates from the ex-ante estimation with the achieved rates (Table 3) shows that we fell below expectations with the MAS1. An exception is wave 3, in which we could win two thirds of waves 1 or 2 households to fill in the diary for another week, but the total response rate compared to the initial gross (wave 1/2 + 3) is also low. Figure 4 shows our response rates in the region of surveys without recruitment and incentive, although we provided an incentive.
Table 3
Estimated and observed response rates
Wave | 1 | 2 | 3 | 1/2 + 3* |
|---|---|---|---|---|
Ex-ante estimated response rate | 24.6 | 24.6 | 24.1 | 20.1 |
AAPOR response rate 1 | 16.7 | 16.5 | 67.8 | 11.3 |
Difference | − 7.9 | − 8.1 | + 43.7 | − 8.8 |
The reasons for the under-achievement can be manifold since the entire HBS-MAS process involves several unique aspects, which make it hardly comparable to other travel surveys in the model of Schmid and Axhausen (2019). One factor even argues for a higher response rate: habitual survey refusers had already been screened out in the HBS and are not included in our gross. But there are other factors as well that argue for the opposite: the MAS recruits had just finished a burdensome two-week HBS and were perhaps exhausted. Even more importantly, we were not allowed for privacy reasons to contact the finishers of the HBS directly to motivate them to participate in the MAS; instead, they were informed passively by HBS interviewers about the MAS or just saw a link in the HBS online form. Schmid and Axhausen (2019) note in this context that prior recruitment is most important for high response rates, whereas incentives add only about 3% in response but help to keep participants motivated in burdensome surveys, which is absolutely true for the MAS.
Fig. 4
Response burden and response rates of comparable mobility surveys from (Schmid and Axhausen 2019)
The main problem with low response rates is the higher risk of a self-selection bias, which is why we find it important to compare the HBS-MAS outcome variables with population characteristics and relevant benchmark surveys. This is done in the following section.
Results
In this section, we assess the representation bias by comparing the characteristics of the HBS-MAS sample with those of relevant benchmark surveys, which are representative of the Austrian population. In detail, we check the HBS-MAS sample against (i) socio-demographic characteristics from the 2020 census data of the National Statistics Office Statistics Austria (Statistics Austria 2021), (ii) mobility indicators from the most recent Austrian National Travel Survey (ANTS) Österreich Unterwegs 2013/2014 (Tomschy et al. 2014), and (iii) time use patterns from the Austrian National Time Use Survey (ATUS) Zeitverwendungserhebung 2008/2009 (Ghassemi-Bönisch 2011). The expenditure data are compared in two respects: (i) against the full sample of the Austrian Household Budget Survey (HBS) Konsumerhebung 2019/2020 (Kronsteiner-Mann and Braun 2022a), of which the MAS sample is a subset, and (ii) against the predecessor dataset MAED, as a main motivation for the cooperation with the HBS was to improve the quality of the expenditure data.
Socio-demographics
Table 4 shows the comparison of household characteristics from national census data to the HBS-MAS sample. All in all, the sample follows the characteristics of households in the total Austrian population reasonably well and the observed differences are likely negligible except for some specific cases. Some minor to moderate deviations can however be observed: while the average household size is nearly identical there is some underrepresentation of both single-person and large households while those with 2 or 4 persons are slightly overrepresented. Similarly, households with no children are slightly underrepresented while the share of households with two children is higher in the HBS-MAS sample than in the population. Regarding the federal state where the households are located, there are fewer participants from Vienna and Salzburg in the HBS-MAS sample than could be expected given their shares in the population. This is accompanied by an overrepresentation of participants from Lower Austria.
Table 4
Household characteristics of the HBS-MAS survey sample compared to 2020 census data from statistics Austria (Statistics Austria 2021)
Household characteristic | HBS-MAS | Census | Deviation |
|---|---|---|---|
Household size [%] | |||
1 person | 30.5 | 37.8 | − 7.3 |
2 persons | 36.7 | 30.4 | + 6.3 |
3 persons | 15.6 | 14.6 | + 1.0 |
4 persons | 15.1 | 11.3 | + 3.8 |
5 or more persons | 2.2 | 6.0 | − 3.8 |
Average household size | 2.22 | 2.20 | + 0.02 |
Household members aged < 15 [%] | |||
0 persons | 76.2 | 81.0 | − 4.8 |
1 person | 11.3 | 10.0 | + 1.3 |
2 persons | 10.9 | 7.0 | + 3.9 |
3 or more persons | 1.6 | 2.0 | − 0.4 |
Federal state [%] | |||
Burgenland | 4.2 | 3.2 | + 1.0 |
Lower Austria | 24.5 | 18.5 | + 6.0 |
Vienna | 20.9 | 23.0 | − 2.1 |
Carinthia | 6.2 | 6.4 | − 0.2 |
Styria | 13.6 | 13.9 | − 0.3 |
Upper Austria | 15.1 | 16.2 | − 1.1 |
Salzburg | 3.6 | 6.1 | − 2.5 |
Tyrol | 8.3 | 8.4 | − 0.1 |
Vorarlberg | 3.6 | 4.3 | − 0.7 |
The personal characteristics of the HBS-MAS sample in Table 5 show an overrepresentation of female participants. As is typical for mobility surveys, very young, very old, unemployed, and less educated persons participated below average, whereas employed persons and those with high education are overrepresented. Regarding the spatial distribution, there is a bias towards rural dwellers at the expense of urban ones. A comparison of available mobility tools with the Austrian National Travel Survey (ANTS) from 2013/2014 (Table 6) shows a surplus of PT permanent tickets and a shortfall of car and motorbike owners. It should be noted that the ANTS originates from 2013 to 2014 and the availability of PT tickets has increased since then. The Klimaticket, a flat-rate season ticket for all public transport in Austria that was introduced on October 26 2021 is not the only reason for that. While 4.4% of wave 3 participants owned this ticket, compared to 1.8% ownership of the preceding, more expensive Österreichcard in waves 1 and 2, there was no significant change in PT ticket ownership between the waves, so the difference to the ANTS may be due to a shift to PT over a longer period of time since that survey was conducted in 2013/2014. Moreover, motorcycle availability has many missing values in the ANTS, which reflect problems with data quality in this variable.
Table 5
Personal characteristics of the HBS-MAS survey sample compared to benchmark values from census data from statistics Austria (Statistics Austria 2021) (in percent)
Personal characteristic | HBS-MAS | Benchmark (Census) | Deviation |
|---|---|---|---|
Gender | |||
Male | 44.5 | 49.2 | − 4.7 |
Female | 55.5 | 50.8 | + 4.7 |
Age | |||
16–19 | 2.3 | 5.7 | − 3.4 |
20–29 | 13.9 | 14.4 | − 0.5 |
30–39 | 18.5 | 16.1 | + 2.4 |
40–49 | 17.3 | 15.4 | + 1.9 |
50–59 | 20.9 | 18.3 | + 2.6 |
60+ | 27.1 | 30.1 | − 3.0 |
Employment | |||
employed | 53.4 | 43.1 | + 10.3 |
self− employed | 6.8 | 6.0 | + 0.8 |
not employed | 39.8 | 50.9 | − 11.1 |
Education | |||
Compulsory schooling | 8.5 | 17.6 | − 9.1 |
Apprenticeship, training | 42.1 | 49.9 | − 7.8 |
High school | 22.2 | 16.0 | + 6.2 |
Tertiary education | 27.2 | 16.5 | + 10.7 |
Urbanity* | |||
Urban | 27.3 | 31.4 | − 4.1 |
Intermediate | 31.4 | 30.8 | + 0.6 |
Rural | 40.9 | 37.8 | + 3.1 |
Table 6
Mobility tools of the HBS-MAS survey sample compared to benchmark values from the Austrian National travel survey 2013/2014 (ANTS) (Tomschy et al. 2014) (in percent)
Mobility tool | HBS-MAS | Benchmark (ANTS) | Deviation |
|---|---|---|---|
Driver’s license | |||
Car | 91.7 | 91.9 | − 0.2 |
Moped/Motorbike | 30.9 | 31.9 | − 1.0 |
Vehicle availability | |||
Bicycle | 79.4 | 81.2 | − 1.8 |
Moped/Motorbike | 7.4 | 11.9 | − 4.5 |
Car always | 73.7 | 77.2 | − 3.5 |
Car sometimes | 14.2 | 14.1 | + 0.1 |
Carsharing | 4.4 | 2.5 | + 1.9 |
PT season tickets | |||
PT season or zone ticket | 25.2 | 20.9 | + 4.3 |
PT discount card | 23.3 | 19.6 | + 3.7 |
Mobility
Table 7 compares the mobility indicators of the HBS-MAS with the unweighted ANTS sample. Wave 1 of the HBS-MAS is shown separately, as it was unaffected by COVID-19 and can thus be more meaningfully compared to the older ANTS. The share of mobile persons is 6.4% higher in HBS-MAS than in the ANTS, probably due to the higher share of employed persons or the higher motivation of the HBS-MAS participants.
Trip rates are very similar whereas daily travel durations and distances are lower for mobile persons, but slightly higher if all participants are considered. While the higher indicators for all persons can be explained by the presence of more mobile persons in the HBS-MAS sample, the lower indicators for mobile persons might be due to methodological differences in counting travel time and distance. In ANTS, the self-reported travel durations were used which can potentially include time spent buying tickets or waiting in the car or are just rounding errors (Rietveld 2001). Distances were calculated or estimated by the respondents themselves which is known to lead to errors which vary with socio-demographic and trip characteristics (Witlox 2007). In HBS-MAS, however, travel times and distances were used that were queried from route planners based on the reported geocoded addresses, with some time added for access to and egress from the chosen mode, dependent on the urbanity of the trip’s origin and destination (see Table A3 in Annex 5). These values were calculated as the mean differences, for the respective urbanity categories, between the travel times reported by the participants and the travel time queried from route planners.
Table 7
Mobility indicators for all days of the week of the HBS-MAS survey sample compared to the unweighted sample of the Austrian National travel survey 2013/2014 (ANTS) with participants aged > 16 years (Tomschy et al. 2014)
Mobility indicator | HBS-MAS | ANTS | Difference to wave 1 | |||
|---|---|---|---|---|---|---|
All waves | Wave 1* | Wave 2 | Wave 3 | |||
Share of mobile persons [%] | 75.9 | 85.6 | 62.3 | 79.4 | 79.2 | + 6.4 |
No. of trips per day | ||||||
… for mobile persons | 3.3 | 3.4 | 3.0 | 3.5 | 3.4 | ± 0.0 |
… for all persons | 2.5 | 2.9 | 1.9 | 2.8 | 2.7 | + 0.2 |
Average daily travel duration [mins] | ||||||
… for mobile persons | 71.5 | 79.6 | 57.1 | 73.8 | 89.7 | − 10.1 |
… for all persons | 59.8 | 72.0 | 41.7 | 64.7 | 71.0 | + 1.0 |
Average daily travel distance [km] | ||||||
… for mobile persons | 40.0 | 47.3 | 31.0 | 39.8 | 49.9 | − 1.6 |
… for all persons | 30.4 | 40.5 | 19.3 | 31.6 | 39.5 | + 1.0 |
Mode shares [%] | ||||||
Walk | 19.8 | 17.7 | 19.1 | 21.7 | 17.0 | + 0.7 |
Bike | 6.8 | 5.8 | 10.5 | 5.7 | 5.9 | − 0.1 |
Car driver | 52.6 | 51.5 | 52.2 | 53.6 | 52.0 | − 0.5 |
Car passenger | 10.2 | 11.2 | 10.8 | 9.2 | 12.6 | − 1.4 |
PT | 10.0 | 13.1 | 7.1 | 9.2 | 11.7 | + 1.7 |
Other | 0.5 | 0.7 | 0.4 | 0.5 | 0.8 | − 0.1 |
Percentage of trips with app use for | ||||||
… routing: Car driver | 2.7 | 1.7 | 2.4 | 3.4 | – | |
Car passenger | 3.4 | 2.4 | 5.2 | 2.9 | – | |
PT | 4.4 | 1.7 | 8.2 | 5.6 | – | |
… ticketing: PT | 5.4 | 2.6 | 10.0 | 6.3 | – | |
Despite a slightly lower car availability for participants in the HBS-MAS, the mode car as driver is higher for them, while the modes car as passenger and public transport are lower. Public transport is especially affected in the survey waves that took place during the COVID-19 pandemic. The share of trips where apps were used for routing or buying tickets is generally very low. It is lower before the COVID-19 pandemic than in the later waves. The reasons for this are not quite clear but might have to do with the mobility changes that led to more trips to unfamiliar locations and a tendency to use digital contactless alternatives to buying tickets at booths or ticket machines.
We are not aware of any studies reporting app usage at trip level. Looking at the level of individuals, 5.74% of our participants reported using an app on routine trips (defined as same origin-destination more than once), 21.73% on non-routine trips, and 74.81% never during the survey week. This share is considerably lower than e.g. in a study by Shaheen et al. (2016) who reported a non-use of multimodal smartphone apps for routine trips (such as commuting to work, going to the gym, grocery shopping) of about 42% − 53%, however, among existing users of a smartphone app. Other studies who asked participants whether or not they used smartphone apps for travel purposes (e.g. Jamal and Habib (2019), Bian et al. (2024) also reported higher usage shares, however including occasional use.
Time use
The MAS time use patterns were assessed using the Austrian Time Use Survey 2008/2009 (ATUS) as a benchmark. It provides a fixed time grid of 15 to 30 min and open text fields for the description of activities, which were assigned to very detailed categories during data processing. The MAS diary, on the contrary, provides open time intervals and a fixed grid of ten main activity categories; only secondary activities were recorded using open text fields and later classified into the main activity categories, with more detailed categories retained for leisure activities. This simplification for the main activities served to reduce the respondent burden, as MAS participants filled in the diary for a longer period (7 days instead of 2) and provided more details on trips and locations than ATUS participants. Table A2 in the annex shows how the detailed ATUS categories were aggregated to match the ten MAS categories to compare the activity durations.
Table 8 shows the average daily activity duration by category of HBS-MAS and ATUS participants. The largest difference is that the HBS-MAS reports almost 30 min more leisure time. This has a conceptual reason: HBS-MAS participants were instructed to code “leisure” whenever an activity is perceived as free time, regardless of what kind. This instruction should satisfy the definition of “leisure” in time use theory, namely, any activity which is assigned more time than the required minimum (DeSerpa 1971). There is a broad overlap of activities that can either be perceived as duty or free time, e.g., cooking or childcare. The leisure time of ATUS participants had to be inferred from activity types that sound unambiguously like free time, which results in a narrower definition. Table 8 shows furthermore that the broad and more ambiguous categories “personal care” and “domestic work” come with larger deviations than the unambiguous categories “sleep”, “travel”, and “eating”, which indicates a greater vagueness in the process of coding and alignment of categories.
Table 8
Weekly averages of daily activity durations in HBS-MAS compared to the Austrian time use survey 2008/2009 (ATUS) (Ghassemi-Bönisch 2011)
Time use category | HBS-MAS | ATUS | Difference to wave 1 | |||
|---|---|---|---|---|---|---|
All waves | Wave 1 | Wave 2 | Wave 3 | |||
Personal | 01:20 | 01:35 | 01:11 | 01:15 | 01:00 | + 00:35 |
Sleep | 08:34 | 08:31 | 08:39 | 08:33 | 08:43 | − 00:12 |
Eating | 01:29 | 01:21 | 01:37 | 01:28 | 01:34 | − 00:13 |
Paid work | 02:48 | 02:52 | 02:24 | 03:04 | 03:02 | − 00:10 |
Education | 00:22 | 00:26 | 00:22 | 00:19 | 00:32 | − 00:06 |
Domestic (incl. childcare) | 02:18 | 02:18 | 02:30 | 02:11 | 02:37 | − 00:19 |
Shopping | 00:15 | 00:18 | 00:15 | 00:14 | 00:20 | − 00:02 |
Leisure | 05:31 | 05:25 | 06:14 | 05:03 | 05:02 | + 00:23 |
Travel | 01:08 | 01:12 | 00:41 | 01:27 | 01:01 | + 00:11 |
Other | 00:10 | 00:02 | 00:01 | 00:22 | 00:07 | − 00:05 |
Expenditures
Since the poor quality of the expenditure data in the preceding MAED survey was a main motivation for us to cooperate with the HBS, we assessed the expenditure data in two respects: (i) we compared both the MAED and HBS-MAS sample against the respective benchmark2, and (ii) we compared the HBS-MAS against the MAED concerning the deviations from the benchmark in order to assess the improvement of the HBS-MAS over the MAED.
Table 9 shows the comparison of the MAED and HBS-MAS against the respective benchmark household budget surveys 2014 and 2019 for the main expenditure categories including savings, i.e. the difference between household income and expenditures. The deviations from the benchmark are shown in separate columns. They were tested for statistical significance using a series of t-tests with Bonferroni correction; the resulting significance levels are indicated with asterisks. Note that imputed rent, calculated by Statistics Austria for home owning households, was removed from the aggregated expenditures for this comparison, as MAED did not include it. This is why the mean expenditures for housing may seem low.
The MAED sample shows strong and significant deviations from the benchmark in most expenditure categories. Deviations occur in both directions for both short-term (more for food, less for leisure) and long-term expenditure categories (more for housing and insurance, less for furniture). We do not see a systematic pattern in these deviations, but rather random fluctuations, which we attribute to the short observation period and the lack of control for double counting in short- and long-term expenditures. This shows that the challenges of surveying consumer expenditures were underestimated and the simplified procedure in the MAED was unable to replace a full-blown consumer expenditure survey.
The HBS-MAS, in contrast, only shows one moderately significant deviation in the category ‘leisure & sports’. The good agreement with the benchmark is not surprising, given that the HBS-MAS is a sub-sample of the HBS, which was selected more or less at random. This was precisely the desired outcome of combining the MAS with the nationwide HBS of the Austrian national statistical office. The improvement of the HBS-MAS over the MAED is also confirmed by the root mean square error of the deviations (bottom row in Table 9): the RMSE of the HBS-MAS is only one fifth of that of the MAED 2015. It underlines the improvement in the capture of expenditure data using the HBS-MAS approach.
Table 9
Mean expenditure per household and expenditure category from the MAED 15 survey compared to the household budget survey 2014/2015 (HBS 14) and the HBS-MAS sample compared to the full sample of the HBS 2019/2020 (HBS 19) in € per household per week excluding imputed rent for homeowning households
Expenditure category | MAED 15 | HBS 14 | MAED 15–HBS 14 | HBS-MAS | HBS 19 | HBS-MAS–HBS 19 | |||
|---|---|---|---|---|---|---|---|---|---|
Food & drink | 125.24 | 101.06 | + 24.18*** | 112.23 | 108.20 | + | 4.03 | ||
Clothes & shoes | 41.44 | 36.90 | + 4.54 | 30.89 | 32.31 | – | 1.42 | ||
Housing & energy | 168.29 | 108.34 | + 59.95 *** | 110.31 | 110.64 | – | 0.33 | ||
Furniture | 17.58 | 47.97 | − 30.39 *** | 47.56 | 47.90 | – | 0.34 | ||
Health | 17.65 | 27.78 | − 10.13 *** | 33.89 | 33.76 | + | 0.13 | ||
Transport | 91.57 | 86.06 | + 5.51 | 94.46 | 88.92 | + | 5.54 | ||
Electronic goods | 26.06 | 22.52 | + 3.54 ** | 28.05 | 26.66 | + | 1.39 | ||
Leisure & sports | 56.76 | 77.31 | − 20.55 *** | 113.82 | 94.84 | + | 18.98* | ||
Education | 14.69 | 7.79 | + 6.90 *** | 11.11 | 9.13 | + | 1.98 | ||
Cafés & restaurants | 41.9 | 49.51 | − 7.61 * | 42.62 | 47.41 | – | 4.79 | ||
Services | 31.47 | 43.68 | − 12.21 *** | 43.02 | 45.81 | – | 2.79 | ||
Insurance | 59.68 | 40.5 | + 19.18 *** | 49.55 | 44.33 | + | 5.22 | ||
Other | 33.55 | 15.02 | + 18.53 *** | 13.13 | 14.93 | – | 1.80 | ||
Savings | 37.89 | 131.28 | − 93.39 *** | 207.58 | 193.97 | + | 13.61 | ||
Total | 725.89 | 664.45 | + 61.44 ** | 730.64 | 704.83 | + | 25.81 | ||
RMSE | 33.14 | 6.88 | |||||||
Participation mode
One of the innovations of the HBS-MAS survey compared to the MAED was that we allowed for online participation in addition to paper participation for a variety of reasons: compatibility with the HBS which also offered both participation modes, higher inclusiveness in view of young people, as well as the saving of time and cost for mailing, data entry etc. In wave 2 emerged another important reason in terms of avoiding person-to-person contact during the COVID-19 pandemic, which is why we promoted online participation more actively, resulting in a higher online rate of 72 and 85% in wave 2 and 3, respectively, compared to only 28% in wave 1. The online rate is thus confounded with the wave and, in turn, also with the occurrence of COVID-19.
To assess the potential influence of the participation mode, we tested its effect on those two mobility indicators that were most affected by COVID-19: trip frequency and the share of public transport use, both of which dropped massively (Hartwig et al. 2022). Table 10 shows the result of two linear models, in which we regressed these indicators on the participation mode. Both models account for confounding variables, which were identified beforehand using semi-partial correlations; these are “Wave 2” in both cases, employment status in the case of trip frequency, and PT season ticket ownership in the case of PT use. The participation mode has no significant effect on the mobility indicator in both cases.
Table 10
Linear regression results for the models for trip frequency per person and week and share of public transport (PT)
Explanatory variables | Trip frequency | Share of PT use | ||||
|---|---|---|---|---|---|---|
Coeff | t-value | Coeff | t-value | |||
Intercept | 13.89 | 31.23 | *** | 0.05 | 5.51 | *** |
Person participated online | 0.55 | 1.05 | 0.01 | 0.87 | ||
Wave 2 (affected by Covid) | − 6.16 | − 12.08 | *** | − 0.07 | − 5.32 | *** |
Person is employed | 3.16 | 6.44 | *** | – | ||
Person owns PT season ticket | – | 0.32 | 22.92 | *** | ||
Model diagnostics | ||||||
# of observations | 906 | 906 | ||||
Multiple R2 | 0.186 | 0.390 | ||||
Conclusion and outlook
This paper reports on a significant update of the recently developed mobility-activity-expenditure diary (MAED). It comes with several innovations based on the lessons learned from the MAED. Instead of collecting all the information at once, we cooperated with Statistics Austria’s official household budget survey (HSB) and conducted a mobility-activities survey (MAS) with a subsample of participants. This cooperation was a unique opportunity that cannot be repeated easily, it enabled us to obtain the expenditure data in much better quality and in this way overcome a major weakness of the MAED. However, the split into two separate reporting periods would also work without this cooperation; it enabled us to gather more information than a one-off survey could, making the HBS-MAS sample even richer than the MAED.
The anticipated advantages largely materialised. The ‘backbone’ of the HBS-MAS is the same as in the MAED: it collects a seamless 168-hour sequence of activities, trips, and geocoded locations, along with the income and expenditures of the same individuals. But the sample is enriched by additional information at various levels: (i) for each activity the occurrence of secondary activities and the physical activity level, (ii) for each trip the use of digital travel planning and ticketing tools, and (iii) at the person level self-reported health and wellbeing, spatial flexibility of work, as well as the engagement in virtual vs. physical activities. Furthermore, we offered online participation in addition to pen-and-paper participation, which comes with many benefits, most of all a higher inclusiveness as well as the saving of time and cost for mailing and data entry. In the MAED, the time needed for data entry was 117 min on average; in the HBS-MAS, it decreased to 61/44/12 minutes in wave 1/2/3 (with an online rate of 28/72/85%, respectively). The drastic decrease for wave 3 stems also from the fact that participants were already familiar with the survey format and filled it out more correctly and complete than before. One learning we can take for future surveys is that the time needed for support during and for validation after data collection should not be underestimated, though it will pay off in the form of data quality.
A specific aspect emerged by coincidence: the 2nd survey wave started together with the first COVID-19 lockdown in Austria, such that it captures the effects of the lockdown and subsequent recovery period (in comparison to wave 1); the 3rd wave was launched later to capture longer-term changes. Advocating to continue with the survey despite the unforeseen events and great uncertainty and quickly adapting the questionnaire to capture the effects of these events, is one major lesson learned.
The low response rate, which only comes to around 17 per cent, is certainly a limitation of this study. A particular obstacle was the non-allowance to actively recruit MAS participants among finishers of the HBS for privacy reasons. To assess the (lacking) representativeness of the HBS-MAS sample, we compared it comprehensively against several benchmark samples. It revealed no deviations beyond the expected level in a conventional survey with a higher response rate, but this does not prove the absence of self-selection bias and is thus no substitute for a high response rate. It would be an advantage if the same organisation conducted both survey parts and had full control over the recruiting.
Furthermore, an app-based survey could reduce the respondent burden, but it is not clear to what extent this would increase the response rate, as app-based surveys come with other problems, in particular technical issues and data protection concerns (Greaves et al. 2025). More importantly, we are not aware of any app that can break down time spent at home into different activities, such as leisure, work, shopping and errands. This is a central motivation for our survey method because it creates the prerequisite for modelling the trade-off between activities conducted at home (without travel) and outside the home (with travel). Generating this information would require the respondents to manually post-process the data recorded by the app, which increases the respondent burden. Even if only the travel survey part was conducted with the support of smartphone apps, considerable effort would be needed for recruitment, retention and data cleaning and processing. Further research is needed into whether the potential reduction in respondent burden justifies this additional effort.
With increasing digitalisation of activities including travel, it would have been interesting to include further questions on the use of digital devices for mobility and in-home activities. However, this needs careful weighing-up with the additional response workload.
Overall, the HBS-MAS is another step forward towards a survey design that collects from the same individuals all the information required to model consumers’ choices on time use, locations, travel modes, and goods consumption. In doing so, it meets the requirements of many advanced transport models mentioned in the introduction, which aim to understand travel behaviour in an increasingly connected world, where spatial flexibility and parallel time use are gaining ever more importance. This includes models for which the MAED data is unsuitable, e.g. the effect of multitasking while travelling (Hartwig et al. 2024; Malokin et al. 2021). The increase in the amount and quality of information was achieved by splitting the survey into two periods. This makes reaching a high response rate a bigger challenge than in other surveys, which is something to keep in mind and work towards improving.
The mobility-activity data described in this paper and the accompanying materials are archived at BOKU and can be requested for research purposes from the authors in the framework of research co-operations.
Acknowledgements
We would like to acknowledge the outstanding work of further colleagues who were involved in the realisation of the survey, especially Gregor Husner, Martin Hinteregger, and Michael Skok, as well as a diligent and skilled team of interviewers who were indispensable for the data collection.
Declarations
Competing interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
