Coming full circle – introducing the Multinational Time Use Study Simple File

At time of writing, more than 170 publications use the Multinational Time Use Study (MTUS) archive, assembled by the Centre for Time Use Research at the University of Oxford. The MTUS includes more than three-quarters of a million diary-days from 68 surveys collected in 22 countries and spanning 60 years. In July 2013, CTUR released a new MTUS Simple File, representing a service not previously available to users as well as a return to principals that shaped the original MTUS file.

The MTUS arose during Jonathan Gershuny’s early career research into “post-industrial society”. Gershuny (1978, 1983) used expenditure diary surveys to test Daniel Bell’s 1976 thesis that as societal wealth increases, economies shift from the production of goods to the production of services. Gershuny noticed that households decreased spending on services while increasing their service use by producing many final services for themselves outside the economic activity typically measured at the time (1983). Gershuny characterised post-industrial society as the growth in “knowledge work” service jobs alongside increasingly sophisticated manufacturing technology that permitted the expanding service employment while household spending on service consumption declined.

Gershuny’s research built on ideas previously suggested by Robert Giffen (summarised in Gershuny 2000), and Hildegarde Kneeland (1929), who based her work on time diaries collected from samples of women in the United States from the 1920s through 1930s. Gershuny embarked on a search for available time diaries to explore the behaviour component of household self-servicing. Consistent with many researchers at the time, Gershuny compiled summary files containing total minutes per day devoted to various activities by individuals in national-level population surveys to assess alongside aggregate expenditure data.

Gershuny found the 1965 Multinational Comparative Time-Budget Research surveys (Szalai 1972), which restricted samples to working-aged people (18–65, though much of the analysis concentrated on the more limited age range of 20–59). Through extensive legwork, as well as personal interaction with John Robinson, staff of the BBC Audience Research Department, Peter Willmott and others, (Fisher and Gershuny 2013), Gershuny amassed a collection of British time diary surveys to harmonise with the Szalai surveys. Thus the basis for the MTUS was born.

The European Foundation for the Improvement of Living and Working Conditions funded the first release of the MTUS, covering seven countries (Gershuny 2000). The first MTUS file only included diaries from people aged 20–59, summary time in 41 activities, and 10 survey and demographic variables.

The initial MTUS harmonisation progressed rapidly. The Szalai project already offered harmonised classifications for activities, diary context, and background variables. Gershuny only needed to condense the Szalai activity codes for his post-industrial research, and adopt a limited range of background variables consistent with the variables available from the household expenditure surveys. The age of youngest diarists varies markedly from survey to survey, and quality of data collection also varies considerably among older diarists in some surveys. As the early MTUS only included diaries from people aged 20–59, the early MTUS sidestepped the problem of how to deal with differential youngest ages and variations in response rates and quality of diaries from the oldest diarists. The initial harmonisation process also included no effort to improve data or address inconsistencies. Gershuny and his main assistant, Sally Jones, created limited documentation. Complex research questions inspired the creation of an MTUS file reflecting the simple needs from the time diary data.

As more users have worked with the MTUS, and as the general sophistication of time use research has progressed over the decades, user interest in greater detail has expanded at a pace beyond the capacity to incorporate new files and new features in the MTUS. At the time of writing, CTUR included a tiny staff of one archivist, four researchers, one further researcher with substantial teaching commitments, and two PhD students. As this team undertakes a range of projects, of which MTUS is only one, and as many members do not contribute to the production of MTUS materials, CTUR has restricted capacity to expand these resources. In spite of these limitations, CTUR nevertheless has created a harmonised episode-level file (now covering 35 surveys from 12 countries), and an expanded range of household and diary variables. CTUR also has expanded the documentation for the MTUS. The main user guide now covers more than 150 pages of main text, with survey level supplementary documentation for each individual survey, separate documentation for child diaries and supplementary files, packages to add variable and value labels in two languages, and metadata summaries of all surveys included in the MTUS - as well as of over 400 other time use surveys (Fisher et. al. 2013).

Additionally, the MTUS adds value through data cleaning and enhancement which makes use of the narrative properties of diary data (Fisher and Gershuny 2013). As some information spans column categories in diary instruments (for instance, the word “train” simultaneously conveys an activity, a mode of transport, and a location; or an activity description “shopping on-line” both reflects the activity of shopping and the context that this activity involves use of the internet), MTUS codes all detail provided by diarists or collected by the survey into all relevant domains. CTUR takes advantage of contemporary computing power not available to survey designers in the 1950s through the 1980s to undertake a variety of data cleaning tasks skipped as too time consuming and expensive in the past. In some cases, the MTUS team has recovered corrupted files and worked from some stored paper materials to recover information not stored in the electronic files to which most researchers have had access. As a result, the best available version for many of the surveys – particularly some of older surveys, included in the MTUS is the MTUS version.

Future developments will improve the delivery of the data and documentation. The US National Institutes of Health fund collaboration between CTUR, the Maryland Population Research Center and the Minnesota Population Center, to add all years of, first, USA-based time use surveys, then surveys from a limited range of other countries included in the MTUS to the Time Use Survey-X distribution mechanism (Hofferth, Flood and Fisher 2012). The ATUS-X system, developed by the two population centres, presently enables registered users to select only required sets of variables for sub-samples of surveys. The ATUS-X system additionally assists users to create customised time use variables that map together elements of the diary, as well as to map in additional variables from earlier waves of the longitudinal Current Population Survey (from which the American Time Use Survey (ATUS) sample is drawn). The USA component of the MTUS should be available to the time use community through the TUS-X access system by late 2014 or early 2015. The collaboration will add some further countries from the main MTUS files into the TUS-X distribution mechanism over subsequent years, and enhance the accessibility of information in the MTUS documentation (Hofferth, Flood and Fisher 2012).

In parallel, CTUR has raised grants to overhaul the MTUS website. The site already links users to publications using MTUS data (available in a searchable database). Current development work will transform the current pdf documentation files into searchable databases of survey, harmonisation process, and variable-level metadata. CTUR also will be amending the MTUS file distribution mechanisms to facilitating customised downloads of documentation.

An unfortunate by-product of the size of the CTUR team and the range of services supplied through the MTUS, however, is that the addition of new surveys and the upgrade of surveys only harmonised to the standards of a previous version, is slow. The process of converting original files into the full range of current MTUS outputs presently takes a minimum of five weeks. Developing the harmonised episode file, new data distribution systems, data enhancement, and improved documentation facilities continue to require considerable effort and time resources.

As more surveys become available, more users want access to a wider range of recent data. At the same time, more users also hope to make use of the cross-time and historical change possibilities offered by the MTUS and eagerly await the upgrade of the older data. The lack of capacity to produce the range of desired outputs on our present resources has created a degree of frustration within CTUR.

This situation gave rise to the solution of the new MTUS Simple File. Until the release of this file, MTUS offered a wide range of files. The most used and best documented of these files offer the current best practice aggregate time use, demographic and episode data for some surveys. Older variations of the MTUS including only those surveys not yet upgraded to current best practice add to the number of total surveys available. Nevertheless, the standard of harmonisation in the older files is not as high quality as in the newer files. Some MTUS variables have changed, and the range of information has expanded. This had meant that matching the older and the newer files was not straight-forward, and required post-mapping effort. Less experienced users encountered some difficulties undertaking this task.

CTUR now documents the mapping of the older and newer files together (Appendix 3 of the MTUS User Guide), and highlights the shortcomings entailed. Algorithms now combine the older and all current best practice files into the single Simple File, where each row case represents one 24-hour time diary, including:

• □the seven essential MTUS survey to diary level identifiers;
• □three diary date variables;
• □a limited range of twelve household- and person-level variables;
• □summary time in twenty five activity categories (with four additional variables separating eating out or eating at school or work from other eating time; playing computer games from other computer and internet use; and child and adult care travel from other travel);
• □for diarists in couples, the total minutes in the diary day that the diarist reported that her or his partner was present.

Thus, users now can directly download a single file with a maximum number of countries immediately ready for use.

The twenty-five activities reflect recent amendments to the MTUS (covering new activities, like using computers and the internet, as well as improvements to older time use categories, including splitting out interactive child care (such as playing with, reading to or teaching children) from the more routine physical and supervisory forms of care. The limited activity range nevertheless returns to an earlier principal of emphasising the activities with the highest number of research applications.

This Simple File, like the original MTUS, offers a limited range of demographic variables. In so doing, the MTUS Simple File provides an easily-used file, suitable for rapid production of figures for experienced time use researchers, as well as a less daunting file for use by people who have recently begun working with time use data. Offering a readily comprehensible file to novice users particularly matters with the current dearth of time use research text books (of which Michelson 2005 and Durán 2007 are the most recent) and the limited number of time use training courses.

The one complexity in the Simple File not present in the earliest MTUS versions is that the Simple File covers all diaries from all age groups. As a consequence, users of the Simple File will need to consider potential diary quality issues in relation to the diaries from the oldest respondents (as suggested in the survey documentation). Users also will have to account for the differential age of the youngest diarists across the surveys. The MTUS documentation on using the child dairies (http://www.timeuse.org/sites/ctur/files/1796/youth-supplement.pdf) will assist users in consideration of how to plan analysis of the daily activities of the younger contributors to the MTUS archives.

As well as being simple to use, the Simple File also is simple to create. CTUR chose not to include complex variables that can be time consuming to produce (such as the identifiers mapping diaries between household members) – thereby significantly reducing the staff training and total time required to add new surveys (or to upgrade older surveys). In future, CTUR will follow two data inclusion strategies:

• □a precursor to full inclusion conversion that enables some surveys to be released in the simple format more rapidly, but which also facilitates full episode level conversion later (this process involves making only those episode and full aggregate file variables required for the Simple File – a single algorithm produces the Simple File format from the full MTUS files. This conversion procedure produces just enough of the full files to permit this algorithm to create the Simple File version of a survey); and
• □a Simple File only-conversion to include some form of datasets we are not likely to otherwise have resources to fully convert.

In consequence, the Simple File maximises the range of available countries and time periods for researchers exploring changes in daily behaviours.

The time use research field has grown considerably in recent years, as evidenced by the rising volume of new time use publications, numbers of people joining the International Association for Time Use Research e-mail list, and the expanding range of countries conducting national time use surveys. The MTUS archive offers a basis against which researchers can monitor changes in behaviour trends over time, as well as assess the potential impacts of changes in the way time use surveys are collected over time. The MTUS Simple File both provides a straight-forward entry point to time use research for people new to the field, while also facilitating ready access to production of basic statistics for experienced users. This new file structure also will facilitate the more rapid release of more surveys in the MTUS format. If you are not already an MTUS user and wish to explore the archive, access is free for all academic and policy researchers, and can be arranged following the registration process on the MTUS website (http://www.timeuse.org/mtus/register), With the release of the Simple File, the MTUS project has come full circle in finding a way to make the most of early efforts and principals while also facilitating the future expansion of MTUS resources.