VETUS – Visual Exploration of Time Use Data to Support Environmental Assessment of Lifestyles Jan C. T. Bieser David Haas Lorenz M. Hilty Department of Informatics, University Department of Informatics, University Department of Informatics, University of Zurich of Zurich of Zurich Zurich, Switzerland Zurich, Switzerland Technology and Society Lab, Empa jan.bieser@ifi.uzh.ch Materials Science and Technology St. Gallen, Switzerland Abstract— The time-use (or activity) patterns individuals per- different environmental impacts). In that sense, goods and form on a typical day – their individual lifestyles – fundamentally services are “best perceived not as ends in themselves [...], but shape our society and the environment we live in. Not only are as instrumental to the performance of an activity” [4, p. 825]. lifestyles evolving over time, driven by societal and technological Building on these premises, time use of individuals has been change, they also significantly contribute to the achievement of the subject of interest in various disciplines yielding scientific Sustainable Development Goal 12 “responsible consumption and production”, namely through the resource use and emissions theories such as the theory of time allocation [5], the time-use associated with goods and services consumed to perform approach [2], social practice theory [6], [7], time geography activities. We created an interactive, browser-based tool to visu- [8], wealth in time [9], [10], or activity-based models of alize and intuitively explore statistical time-use data. The visuali- transport demand [11]. zation helps to gain an overview about the available data, identify At the same time, individual lifestyles are subject to contin- and compare common time-use patterns and draw up hypotheses uous change driven by societal and technological developments about the relationship between changes in lifestyles and their [12]. For example, as people are increasingly moving to urban social and environmental consequences. We use the tool to com- environments, the commuting patterns – and thus the time pare time-use data from different regions, time periods as well as socio-economic and demographic backgrounds and estimate the spent in transport – can change. Also, the increasing use of associated energy consumption. From a time-use perspective, any information and communication technology (ICT) leads to a technological change which triggers changes in time allocation relaxation of some time and space constraints of activities [13]. can only be environmentally sustainable if the environmental For example, “virtual mobility” solutions, such as impact of the total of the activities performed after the change is telecommuting or videoconferencing, can have direct impact lower than before. on the time spent in transport [14]–[16]. They can even Index Terms— Time use, time-use data, lifestyles, activities, eradicate the need to live close to the employer and thus change energy intensity of activities, visualization, sustainability. land-use patterns (e.g. the attractiveness of living in urban or rural environments) and commuting patterns in the long run I. INTRODUCTION [17], [18]. To summarize, individual lifestyles (i) are a major Achieving “responsible consumption and production” determinant of environmental impact, (ii) are subject to patterns has been manifested as Sustainable Development Goal continuous change, and, (iii) for these reasons, have been of 12 by the United Nations [1]. Individual lifestyles, for this interest in many academic disciplines. study defined as “dynamic pattern[s] of consumption activities” Today, large collections of time-use data – diaries of the [2, p. 111] directly impact the environment through the time individuals spend on activities – from various countries resource use and emissions associated with goods and services and time frames is available [19]–[21]. In this paper, we pre- consumed to perform the activities. sent a tool for visual exploration of time-use data (VETUS), Lifestyles can be analyzed from various perspectives, e.g. developed to process the data provided by the Multinational from a functional perspective (products fulfilling stable needs), Time Use Study (MTUS) of the Centre for Time Use Research from a neo-classical budget constraint perspective (products at the University of Oxford [20]. The tool can be used to com- fulfilling individual needs with a budget-constraint on pare individual time-use patterns (the time individuals spend on consumption) or from a time-use perspective (individual needs various activities on a 24-hour day) from different regions, time and utility with a time constraint on consumption) [2]. Time frames as well as socio-economic and demographic back- use is a suitable perspective for the analysis of lifestyles, grounds, and to draw up hypotheses on environmental impacts. because time budget is naturally limited and constant (24 h per As humans are good at visual perception [22], visualization of day) and the activities to which people assign their time can be time-use data can help researchers to explore time-use data in related to environmental impacts [2], [3]. For example, an intuitive way [23]. someone can spend an evening reading a book at home or We analyzed existing work in the field of time-use re- taking a trip with a private car (activities with significantly search, environmental impact assessment of everyday activities and data visualization, developed the tool considering visuali- Many researchers followed this approach, e.g. Aal et al. zation trade-offs and appropriate visualization idioms, and used [25] estimated the energy intensity of leisure activities in Nor- it for environmental assessments of lifestyles extracted from way in 2001, Minx and Baiocchi [4] estimated activity material time-use data. intensities in West Germany in 1990, Yu et al. [26] activity CO2 intensities in China in 2008 and Druckmann et al. [3] ac- II. TIME-USE DATA, ACTIVITIES AND ENVIRONMENTAL tivity greenhouse gas intensities in Great Britain in 2005. IMPACTS The time-use approach can also be used to explain indirect The time-use approach is a perspective to analyze lifestyles environmental effects of technological change. For example, from a consumption perspective focusing on temporal con- telecommuting allows employees to work from home, save straints (as opposed to financial budget constraints). A time-use commuting time and the related energy consumption. However, pattern is an observable set of activities and the time spent on net energy savings depend on how the time saved is spent. De- these activities, in our case by an individual in 24-hours. Time- pending on the energy intensity of the substitute activities, the use data provided by the MTUS describes the time (in minutes) environmental benefits can be partially compensated or even individuals spend on distinct activities on a specific day and overcompensated for – a phenomenon called time rebound combines over a million diary days from 23 countries from the effect [2]. The time-use approach is especially useful to inves- 1960s to the 2010s [20]. tigate such rebound effects because of the hard 24-hour con- Jalas describes sustainable lifestyles as “the requirement of straint, which provides a natural system boundary to behavior. no increase in the materials-intensity of everyday life” [2, p. Exemplary research questions that can be investigated with the 113]. By applying decomposition analysis on household ex- time-use approach are: Does a given ICT use case increase or penditure, energy consumption, time-use and input-output data, decrease the environmental impact? Does a given ICT use case he estimates the energy intensities of activities for Finnish increase or decrease the time individuals spend in transport? households considering direct energy use (e.g. the fuel con- Does a given ICT use case increase the pace of life (“the speed sumption of a car) and indirect energy use (“energy use of pro- and compression of actions and experiences” [27, p. 8/9])? Do ducing the goods and services that are needed in the activity” people who live in urban environments spend less or more time (p. 114) – Tab. I). traveling than people who live in rural environments? Due to the high energy intensity of transportation, outside- of-home activities, even if not very energy-intensive as such, III. VISUALIZATION can cause relatively high energy consumption if transportation A. Data Visualization is included. Sleeping has an energy intensity of zero since domestic heating is not allocated to any activitiy. Work has an Visualization “transforms the symbolic into the geometric, [..] energy intensity of zero since no final consumption is allocated offers a method for seeing the unseen” and “enriches the pro- to it. cess of scientific discovery and fosters profound and unex- pected insights” [23, p. 3]. Specifically, as the volume of avail- TABLE I. ENERGY INTENSITIES OF ACTIVITIES IN FINISH HOUSEHOLDS able data is increasing at a tremendous pace, it becomes more 1998-2000 BASED ON [24]. ACTIVITY CATEGORIES ARE BASED ON challenging to derive meaningful insights from the data without DRUCKMANN ET AL [3] AND ARE USED LATER IN THE STUDY. adequate visualization [28]. Visualization helps especially re- Activity Energy Activity category Avg. energy searchers who want to explore data to find interesting hypothe- intensity intensity ses. Visualization methods are suitable where human pattern [MJ/hr] [MJ/hr] Leisure-time travel 83 Private travel (PT) 83 recognition capabilities are to be supported, rather than re- Work- and education-related 73 Work travel and 73 placed, in our case for the exploratory analysis of time-use pat- trips commute (WTC) terns to support environmental assessment of lifestyles [29]. Having meals 41 Food and drink (FD) 41 Services and civic matters 46 Personal, household 30 B. Used data Personal hygiene, dressing up 36 and family care Phone calls 27 (PHF) For developing the application, we focus on the ‘adult’ ag- Shopping, family business 24 gregate dataset using the 69-activity typology. In this dataset, Housework 19 each record represents a 24-hour observation day, providing the Culture and amusement events 8 Leisure and recrea- 4 Hobbies 6 tion (LR) time spent on 69 activities plus socio-economic and demo- Reading 3 graphic variables of the diary person. We compared the varia- Sports and recreation 2 bles with socio-economic and demographic indicators com- TV viewing 1 monly used to describe populations (e.g. by federal statistical Sleeping 0 Sleep and rest (SR) 0 offices) and selected 96 variables (all 69 activity plus 27 demo- Paid work 0 Paid and voluntary 0 work (PVW) graphic and socio-economic variables, Tab. II) to be used as a core set for visualization We did not include energy intensities of activities directly into the visualization because such data is only available for few time frames and regions and has to be considered in a later step of the process. TABLE II. VARIABLES INCLUDED IN THE ANALYSIS. is limited by the size of the page, which is bound to (normal) Variable Description display size. COUNTRYA Country where the study was conducted 3) Limitation in computing power: The number of included DAY Day of the week the diary was kept variables, the size of the dataset and the used visual elements YEAR Year the diary was kept BADCASE Marker of low quality observations impact the performance of the tool with respect to response HHTYPE Household type (e.g. couple) time in displaying data. HHLDSIZE Household size (number of people) NCHILD Number of children under the age of 18 D. Selected visualization idioms OWNHOME Does the diarist own or rent the home URBAN Does diarist live in an urban or rural area “A vis idiom is a distinct approach to creating and manipu- COHAB Are household members married or cohabiting lating visual representations” [30, p. 10], i.e. “any specific se- COMPUTER Does the household have a computer and/or internet quence of data enrichment and enhancement transformations, access at home Type and number of private vehicles in the household visualization mappings, and rendering transformations that VEHICLE produce an abstract display of a scientific dataset” and are usu- (e.g. non-motorized, motorized) SEX Sex of diarist ally based on “intuitive analogies between familiar objects and AGE Age of diarist […] physical abstractions” (e.g. bar, scatterplot or line charts) EMP Is the diarist in paid work [31, p. 77]. UNEMP Is the diarist unemployed Number of paid working hours incl. overtime in the week We first created a prototype to test different visualization WORKHRS prior to the survey idioms and then developed the final version, which is described OCCUP Diarist’s (most recent) occupation (e.g. medical, legal) in the following. SECTOR Sector of employment of diarist (public or private) STUDENT Whether diarist is a student 1) Time spent on 69 activities by day of the week: Time-use RETIRED Whether diarist has retired patterns can significantly change from day to day, especially EDCAT Harmonized highest level of education between working and non-working days. Therefore, we visual- CITIZEN Is the diarist citizen of the country he lives in CIVSTAT Is diarist in a couple and lives with the spouse/partner ize the average time spent by individuals on 69 activities in EMPSP Employment of spouse (e.g. full-time, part-time) minutes by day of the week. This yields a matrix of 69 activi- FAMSTAT Age of diarist and age of co-resident children (if any) SINGPAR Is the diarist a single parent ties by seven days. Displaying such a large amount of infor- MAIN1– mation is challenging and can best be done with heat maps Time spent on 69 distinct activities MAIN69 (Fig. 1), an intuitive way to display matrix alignment of two C. Visualization requirements and trade-offs key attributes. Each matrix cell holds an area mark denoting a quantitative value attribute encoded with color (time spent on The visualization tool should enable the user to browse through available time-use data in an exploratory, tentative way activities). Additionally, when hovering over a field, the aver- and allow to derive initial interpretations of differences in time- age time spent on the activity on the respective day will be dis- use patterns among regions or time-frames or among groups played. defined by socio-economic and demographic properties of in- 2) Time spent on activity categories: Visually comparing 69 dividuals. Therefore, the tool needs to display the time spent on distinct activities is cognitively challenging, which is why we activities in an intelligible and comprehensible way and allow show the average time spent on eight activity categories as the researcher to set filters on geographic, temporal, socio- described in column 3 of Tab. 1. For displaying this variable, economic and demographic variables. After having applied we use a pie chart, to visualize how the single parts (activity filters to the dataset, visualized the data and derived an inter- categories) contribute to the whole (24 hours) [30]. pretation, the user should be well prepared for applying statis- 3) Day of the week, age group, family status, working hours: tics software, e.g. to test a hypothesis1. Days of the week and family status are categorical variables, To meet these requirements, we needed to address several whereas age and working hours are continuous variables trade-offs caused by three limitations of resources (humans, which are often transformed into categorical variables by cre- computers, displays) [30]: ating bins (e.g. age groups “18-30” or “30-40”). These varia- 1) Cognitive limitations of humans: The dataset contains in bles are mainly used to filter the data set and compare time- total 69 activity variables and 27 demographic and socio- use patterns among individuals with different demographic economic variables. This choice could be criticized for induc- and socio-economic backgrounds. Also, the number of obser- ing a bias by limiting the flexibility for the researcher. On the vations for each category of a filter variable can be displayed other hand, including a high number of variables in the core set to provide information on the distribution of the socio- can harm the simplicity and usability of the tool. economic and demographic variables. We used bar charts (Fig. 2) Limitations in displays: To increase usability, we decided 2) to visualize the distributions of these variables because they that the tool should be accessible through a standard web are useful to compare quantitative values of different catego- browser and show all required information on one single page, ries of a variable [30]. without the need to scroll. Therefore, space for visual elements 4) Occupation: The occupation of the diarist is also a cate- gorical variable, however with significantly more categories 1 For detailed investigations of MTUS data users should also refer to the than the variables described above (MTUS distinguishes 14 MTUS User Guide: https://www.timeuse.org/MTUS-User-Guide occupation categories such as “farming, forestry and fishing”). We used a pie chart (Fig. 3) because bar charts require much 6) Year the survey was conducted: For visualizing the year space as the number of categories increases. The limitation of the diary was kept, we created a timeline using a vertical bar display size then pose a harder constraint than the fact that the chart (Fig. 5). The vertical axis denotes the number of observa- legibility of pie charts suffers with increasing numbers of cat- tions and the horizontal axis shows the years. Users can filter egories. the dataset by selecting a time frame using a draggable selector 5) Country where the survey was conducted: The most nat- frame. ural way to display the country where the survey was conduct- 7) Further demographic and socio-economic variables: Fi- ed is a choropleth map (Fig. 4). This is a geographic map of nally, we wanted to improve the filter options for the user, regions which displays a quantitative attribute (i.e. the number while staying within the display limitation of one single page. of observations from each country) encoded as color over the For this purpose, we added additional select lists for variables different regions [30]. In our case, the more color-intense a with few filter options at the bottom of the page (Fig. 6). The country, the more observations for that particular country are number of observations by category for these variables is dis- contained in the dataset. played as a number in the end of each category name. At startup of the tool, the whole data set is loaded and the visualization idioms are created showing the average time spent on activities and the number of observations by category for the described variables. In order to compare time spent on different activities by regions, daytimes and other variables, users can filter the data set by clicking on variable categories in the visu- alization idioms (e.g. the bar representing a specific age group) and select/deselect it. When deselected, all observations of the respective category are filtered and the displayed values for each other variable are recalculated and updated in all visuali- Fig. 1. Heat map for time spent on activities by day of the week (only 29 of zation idioms. 69 activities selected for visualization in this example). Fig. 4. Choropleth map, using geo data to encode an attribute (number of observations) with color. Fig. 2. Bar chart showing the number of observations (value attribute) for each age group (key attribute). Fig. 5. Bar chart showing the number of observations (value attribute) for each year (key attribute) and a draggable selector frame. Fig. 3. Pie chart showing distribution of observations across exemplary Fig. 6. Select lists for additional demographic and socio-economic variables occupations. (excerpt). Additional tests would help to improve the tool, especially be- cause of the many degrees of freedom in visualization design. V. EXEMPLARY APPLICATION OF THE TOOL TO ASSESS LIFESTYLES AND THEIR ENVIRONMENTAL IMPACTS We used the visualization tool for an initial analysis of dif- ferences in 24-hour time-use patterns across regions, time frames, socio-economic and demographic backgrounds. For each time-use pattern we also estimated the total energy con- sumption associated with the activities performed on the day using average energy intensities of activity categories (see Tab. I; energy intensities are based on an analysis of finish house- Fig. 7. Final dashboard. holds in 1998-2000 and need to be interpreted with care be- Finally, we show the number of currently selected observa- cause of their age). Tab. III shows the result of the analysis and tions at the top center of the page, and a menu for options in the potential interpretations of differences in time-use patterns. The sidebar. The whole dashboard (Fig. 7) can be considered a vis- table illustrates one example how the visualization tool can be ualization idiom itself, combining the idioms described above. applied to investigate time-use data and environmental impacts. All charts are interconnected and changes in one chart trigger Due to methodological differences in surveys across countries, changes in the other charts. different numbers of observations for each time frame and country, and high numbers of missing values for some varia- IV. IMPLEMENTATION bles the results need to be interpreted with caution. They do not imply causality and only have value as a starting point for more A. Software technologies detailed investigations. In the following we describe the main For building the tool, we needed three main components: an results by variable to demonstrate the approach and the tool. output panel which displays the visual representation, a visuali- A. Age, gender, number of children zation engine which transforms the data into the visual repre- sentation and a database storing the data. Younger people spend more time on pvw and wtc than older We developed the tool as a web application to make it ac- people, who spend more time on lr and fd. In this analysis, cessible to anyone with a standard web browser (output panel). spending few time on pvw reduces environmental impacts as no As database system, we used MongoDB and as a visualization energy consumption is allocated to pvw (0 MJ/hr), however wtc engine the JavaScript libraries D3.js and dc.js, which together seems to be related to pvw and is energy intensive (73 MJ/hr). can be used to create and render charts providing instant feed- Women seem to cause high energy consumption by spend- back on user input [32], [33], [34]. To layout the charts, we ing more time on phf (30 MJ/hr) and less time on pvw than used the frontend framework Bootstrap, as it is particularly men. However, this energy consumption should be allocated to user-friendly and easy to implement [35]. A repository on all members of a household, as the activity phf commonly GitHub was used for version control and documentation: serves all of them, not just the person who performs the activi- https://github.com/Sonnenstrahl/datavis [36]. The dashboard ty. Gerushny et al. [37] showed that time women spent on phf can be accessed at: https://files.ifi.uzh.ch/datavis continuously decreases since the 1960s, and increases for men. Unsurprisingly, people without children seem to spend B. Performance and testing more time on lr and less on phf. In a first step, we created the dashboard without the heat B. Education, motorized vehicle computer/Internet access map. The performance was exceptionally good and had no in- put lag when displaying all observations for Europe. In a se- People with higher education, a motorized vehicle, or a com- cond step, we added the heat map, which significantly lowered puter and/or Internet access tend to spend more time on pvw performance, as it is multidimensional and requires two key and travel (pt + wtc)2, which increases their energy consump- attributes (day of the week and activities). Therefore, we creat- tion. One possible explanation is that individuals with these ed custom launch parameters which enable the user to launch characteristics have a higher-than-average income which is the application without the heat map or grouped activities (this related with time spent on pvw and wtc. functionality is not available in the online version of the tool). C. Working hours and employment status To inform the user that the system is busy while loading data, a Compared to the average, people who spend more time on loading wheel was added. pvw and wtc (see variables employment status and working The prototype and the final dashboard were tested by two hours in Tab. III) mainly sacrifice time spent on phf, followed researchers and used for environmental assessment of lifestyles by lr. Sacrifice of time spent on sr, fd and pt for pvw and wtc is in a pilot use case (see section V). The researchers reported that lower. they successfully used the tool to compare time-use patterns. A list of further potential improvements can be found on GitHub. 2 We have to consider that diary years span from 1974-2010. Having a com- puter and Internet access was not always common in this time frame. TABLE III. TIME SPENT ON ACTIVITY CATEGORIES ON A 24-HOUR DAY FILTERED ACROSS DIFFERENT REGIONS, TIME FRAMES AND DIFFERENT SOCIO- ECONOMIC AND DEMOGRAPHIC BACKGROUNDS USING THE VISUALIZATION TOOL. THE VALUES REPRESENT THE RELATIVE DEVIATION OF THE SPECIFIC FILTERED DATA FROM THE AVERAGE ACROSS ALL OBSERVATIONS ([tfiltered/tall]-1). FOR UNDEFINED USE OF TIME WE USED AN AVERAGE ENERGY INTENSITY OF 11 MJ/HR. Variable Filter SR LR PHF PVW FD PT WTC En- #records Possible interpretation ergy cons. Average of all observa- No filter 507 346 250 181 91 44 20 297 343’107 n/a tions [min] resp. [MJ] Age <40 years -2% -9% -4% 27% -9% 9% 30% 0% 191’741 Younger people spend more time on pvw and >= 40 years 1% 7% 4% -21% 7% -7% -25% 0% 151’366 wtc than older people. Gender Women 0% -5% 31% -30% -2% 2% -35% 10% 188’457 Women spend more time on phf and less time Men -1% 7% -38% 36% 2% 0% 40% -12% 154’649 on pvw and wtc than men Number of children None 2% 9% -9% -9% 0% -2% -10% -4% 201’607 Adults living without children spend more <18y in household >=1 -3% -13% 13% 12% -1% 5% 10% 6% 141’500 time on lr and less time on phf, pvw and wtc. Single parent/number Yes/>=1 -1% -10% 29% -12% -16% 18% -15% 11% 7’433 Single parents spend more time on phf and less of children >18y in No/>=1 -3% -13% 12% 13% -1% 5% 15% 6% 134’040 time on pvw and wtc. household Cohabiting In a couple -1% -4% 6% -1% 8% -2% -10% 3% 163’936 People who are in a couple spend more time Not in a couple 2% 9% -18% 2% -3% 5% -5% -7% 71’293 on phf. Living area Urban/suburban -1% 1% -1% 0% -3% 5% 10% 1% 184’785 People in urban environments spend slightly Rural/semi-rural 0% -1% 3% -1% 5% -9% 5% 1% 77’723 more time on travel (pt + wtc) than people living in rural environments. Diary year 1974-1980 -2% 2% 4% -2% -7% -11% 15% 0% 39’566 In the 2000s, people spend more time on travel 1983-1987 -6% 9% 4% -4% -11% -2% 5% 0% 40’759 than earlier. 1989-1995 -2% -1% 4% 3% 2% -5% -20% -1% 124’490 1997-2003 4% -4% -6% -2% 5% 5% 5% 0% 96’042 2005-2010 2% 0% -5% -2% -3% 16% 10% 1% 42’250 Occupation Management -3% -12% -22% 52% -7% 16% 65% -3% 9’611 Managers work and travel more than non- Not management -4% -5% -11% 31% 3% -11% 25% -4% 119’335 managers. Completed secondary Yes -1% -5% -7% 19% -3% 11% 25% 0% 191’937 The higher the education the more people education No 1% 7% 9% -25% 5% -16% -35% 0% 142’348 work and travel. Private motorized >=1 -2% -4% 0% 13% -3% 2% 20% 1% 214’527 People who have a motorized vehicle work vehicle in household No 2% 14% 3% -29% -8% -7% -20% -2% 54’813 and travel more than people who do not have a motorized vehicle. Computer/Internet Yes -1% -10% -7% 25% -8% 16% 45% 2% 88’576 People with computer/Internet work and travel access in household No 1% 2% 3% -10% -2% 2% -5% 1% 134’902 more than people without computer/Internet. Country Austria -14% 15% 8% 1% 31% -45% -45% -2% 22’306 People from Southern European countries tend France 6% -8% -9% 5% 21% -20% 10% -4% 14’631 to sleep more than people from Norther Euro- pean countries. Germany -4% -15% 4% 39% -14% 5% 20% 0% 22’554 Italy 6% -3% 6% -17% 16% -2% -90% -2% 29’973 Netherlands 0% 0% 3% -2% -12% 11% 20% 3% 113’351 Spain 5% -2% -5% -6% 10% 2% 0% 0% 81’347 United Kingdom -4% 8% -4% 2% -12% 2% 10% -2% 58’945 Employment status Full-time -5% -15% -22% 68% -4% -5% 65% -7% 136’905 People who spend much time on pvw spend Part-time -2% -8% 16% 0% -11% 9% 20% 7% 45’120 less time on phf and lr. Not in paid work 5% 18% 20% -72% 8% 0% -80% 5% 138’272 Working hours the >=40 hours -6% -15% -22% 72% -7% -9% 80% -7% 83’790 People who spend much time on pvw spend week bevor the survey less time on phf and lr. D. Urban/rural living environment until 1990s it seems that in the 2000s people travel slightly It seems that living in an urban or rural environment has no more (see also V.F). strong impact on time-use patterns. People in urban environ- Comparing results across countries and time periods has to ments spend slightly more time on travel. For assessing the be done with caution because the data from different years or environmental consequences, differences in the modal split in countries usually stems from different studies which might rural and urban environments need to be considered. differ in survey methodology. E.g., time spent on wtc in Italy E. Country and year and, on pt and wtc in Austria, seems to be implausibly low. In Southern European countries people spend more time on sr than in Northern European countries. Compared to the 1970s F. Energy consumption formance. We encourage researchers interested in time-use data Highest (private) energy consumption is found for women, to use this visualization and even add further functionality. people with children in the same household and part-time em- ACKNOWLEDGMENT ployees. These effects occur as we are not considering energy consumption at the workplace and thus people who work less We thank the Centre for Time Use Research of the Univer- (0 MJ/hr), spend the time on more energy intensive activities sity of Oxford for collecting and standardizing time-use data (e.g. phf, lr). It is an interesting question how to include energy from various countries and providing the data free of charge. consumed during the time spent on pvw in such analyses. REFERENCES Traveling should be treated with special attention, because it is highly energy-intensive. Time spent on traveling in the [1] United Nations, “Sustainable Development Goals,” 2017. 2000s seems to be higher than in the 1970s, a phenomenon [Online]. Available: http://www.un.org/sustainabledevelopment /sustainable-development-goals/. [Accessed: 07-Nov-2017]. which increases energy consumption (however, this also de- [2] M. Jalas, “A time use perspective on the materials intensity of pends on development of passenger miles, modal split and consumption,” Ecological Economics, vol. 41, no. 1, pp. 109– transport energy intensity). 123, Apr. 2002. This result potentially contradicts results of other studies [3] A. Druckman, I. Buck, B. Hayward, and T. Jackson, “Time, which find that time spent on travel did not change in the past gender and carbon: A study of the carbon implications of Brit- 25 to 30 years (based on Hungarian time-use survey [38]). This ish adults’ use of time,” Ecological Economics, vol. 84, no. aspect needs to be further investigated. Also, full-time employ- Supplement C, pp. 153–163, Dec. 2012. ees/people with high-working hours and people who have a [4] J. C. Minx and G. Baiocchi, “Time Use and Sustainability: An computer and/or Internet access travel more than others. Input-Output Approach in Mixed Units,” in Handbook of Input- Output Economics in Industrial Ecology, Springer, Dordrecht, Finally, it is unclear if spending more time on an activity 2009, pp. 819–846. really increases the energy consumption for that activity. For [5] G. S. Becker, “A Theory of the Allocation of Time,” The Eco- example, in Southern European countries people spend more nomic Journal, vol. 75, no. 299, pp. 493–517, 1965. time on fd, but does this imply they eat more? In fact, if people [6] U. Birsl, “Anthony Giddens: The Constitution of Society. Out- just eat slower and therefore spend less time on other energy line of the Theory of Structuration, Cambridge: Polity Press intensive activities, total energy consumption might decrease. 1984, 402 S. (dt. Die Konstitution der Gesellschaft. Grundzüge einer Theorie der Strukturierung, Campus: Frankfurt/New York VI. DISCUSSION 1988, 460 S.),” in Klassiker der Sozialwissenschaften, Springer VS, Wiesbaden, 2016, pp. 346–349. The application of the visualization tool shows that it can be [7] C. Katzeff and J. Wangel, “Social Practices, Households, and used to compare lifestyles and associated environmental im- Design in the Smart Grid,” in ICT Innovations for Sustainabil- pacts. The chosen visualization idioms display the data in a ity, Springer, Cham, 2015, pp. 351–365. meaningful way that is easy to interpret by an end user; con- [8] T. Hägerstrand, “Time-Geography. Man, Society and Environ- text-dependent guidance is directly provided. However, the set ment,” The United Nations Newsletters, 8, 1985. of visualization idioms provided is not exhaustive and receiv- [9] L. A. Reisch, “Time and Wealth,” Time & Society, vol. 10, no. ing feedback from more users could yield valuable information 2–3, pp. 367–385, Sep. 2001. for further refinement and extensions. [10] J. P. Rinderspacher, “Zeitwohlstand in der Moderne,” WZB Directly enhancing the tool with environmental data (in this Discussion Paper, Working Paper P 00-502, 2000. [11] J. Castiglione, M. Bradley, and J. Gliebe, Activity-Based Travel case energy intensities or emission factors of activities) would Demand Models: A Primer. Washington D.C., 2015. allow users to immediately compare the environmental impacts [12] J. C. T. Bieser and L. M. Hilty, “Assessing Indirect Environ- of various lifestyles with the tool. However, this is also subject mental Effects of Information and Communication Technology to availability of such data, which so far is only available for (ICT): A Systematic Literature Review,” Sustainability, vol. 10, specific countries and time frames. A full list of potential im- no. 8, p. 2662, Jul. 2018. provements can be found on GitHub. [13] I. Røpke and T. H. Christensen, “Energy impacts of ICT – In- sights from an everyday life perspective,” Telematics and In- VII. CONCLUSION formatics, vol. 29, no. 4, pp. 348–361, Nov. 2012. [14] J. C. T. Bieser and L. M. Hilty, “An approach to assess indirect We created a tool to visually explore time-use data and de- environmental effects of digitalization based on a time-use per- rive initial hypotheses regarding changes in lifestyles which spective,” in Proceedings of EnviroInfo 2018, Munich, 2018. can have relevant environmental impacts. [15] J. C. T. Bieser and L. M. Hilty, “Indirect Effects of the Digital In our pilot application of the tool, we found initial evi- Transformation on Environmental Sustainability Methodologi- dence that increased use of ICT does not necessarily reduce cal Challenges in Assessing the Greenhouse Gas Abatement energy consumption of individual lifestyle. From a time-use Potential of ICT,” in ICT4S2018. 5th International Conference perspective, any technological change which triggers changes on Information and Communication Technology for Sustaina- in time allocation can only be environmentally sustainable if bility, Toronto, Canada, 2018, vol. 52, p. 14. total environmental impacts of activities performed after the [16] L. M. Hilty and J. C. T. Bieser, “Opportunities and Risks of Digitalization for Climate Protection in Switzerland,” Universi- change is lower than of the activities performed before. ty of Zurich, Zurich, 2017. There is much potential to improve the tool, i.e. directly in- [17] P. L. Mokhtarian, “A Typology of Relationships Between Tele- cluding environmental data in the tool or improving the per- communications And Transportation,” University of California Transportation Center, Jun. 1990. [29] T. Munzner, “Visualization Analysis & Design,” University of [18] I. Salomon, “Telecommunications and travel relationships: a British Columbia, Vancouver, BC, 2017. review,” Transportation Research Part A: General, vol. 20, no. [30] T. Munzner, Visualization Analysis and Design. A K Pe- 3, pp. 223–238, May 1986. ters/CRC Press, 2015. [19] European Commission, “Harmonised European time use sur- [31] R. B. Haber and D. A. McNabb, “Visualization idioms: A con- veys,” 2009. ceptual model for scientific visualization systems,” Visualiza- [20] J. Gershuny and K. Fisher, “Multinational Time Use Study,” tion in Scientific Computing, vol. 74, 1990. Centre for Time Use Research, University of Oxford, 2013. [32] MongoDB, “Open Source Document Database | MongoDB,” [21] Statistics Sweden, “HETUS - Harmonised European Time Use 01-Dec-2018. [Online]. Available: https://www.mongodb.com/. Survey,” 2018. [Online]. Available: https://www.h6.scb.se/tus/ [Accessed: 01-Dec-2018]. tus/default.htm. [Accessed: 11-Nov-2018]. [33] M. Bostock, “D3.js - Data-Driven Documents.” [Online]. [22] K. W. Brodlie et al., Scientific Visualization: Techniques and Available: https://d3js.org/. [Accessed: 01-Dec-2018]. Applications. Springer Science & Business Media, 2012. [34] dc.js, “dc.js - Dimensional Charting Javascript Library,” 01- [23] B. H. McCormick, T. DeFanit, and M. D. Brown, “Visualiza- Dec-2018. [Online]. Available: https://dc-js.github.io/dc.js/. tion in Scientific Computing,” 21, 1987. [Accessed: 01-Dec-2018]. [24] M. Jalas, “The Everyday Life Context of Increasing Energy [35] Bootstrap, “Bootstrap,” 01-Dec-2018. [Online]. Available: Demands: Time Use Survey Data in a Decomposition Analy- https://getbootstrap.com/. [Accessed: 01-Dec-2018]. sis,” Journal of Industrial Ecology, vol. 9, no. 1–2, pp. 129– [36] GitHub, “Build software better, together,” GitHub, 01-Dec- 145, Jan. 2005. 2018. [Online]. Available: https://github.com. [Accessed: 01- [25] C. Aall, I. G. Klepp, A. B. Engeset, S. E. Skuland, and E. Støa, Dec-2018]. “Leisure and sustainable development in Norway: part of the [37] J. Gerushni, E. Jarosz, J. Suh, and M. Vega, “The Multinational solution and the problem,” Leisure Studies, vol. 30, no. 4, pp. Time Use Study (MTUS). Poster.” Centre for Time Use Re- 453–476, Oct. 2011. search. [26] B. Yu, J. Zhang, and Y.-M. Wei, “Time use and carbon dioxide [38] T. Fleischer and M. Tir, “The transport in our time-budget,” emissions accounting: An empirical analysis from China,” University Library of Munich, Germany, 76851, Feb. 2017. Journal of Cleaner Production, vol. 215, pp. 582–599, Apr. 2019. [27] H. Rosa, “Social Acceleration: Ethical and Political Conse- quences of a Desynchronized High–Speed Society,” Constella- tions, vol. 10, no. 1, pp. 3–33, 2003. [28] R. M. Friedhoff and T. Kiely, “The Eye of the Beholder,” Computer Graphics World, vol. 13, no. 8, p. 46, 1990.