We thank the Economic and Social Research Council (ESRC) for funding this research (grant numbers RES-000-23-TO704 and RES-000-23-TO704-A) from September 2003 until September 2008.
Substantive Research:
Historical changes in time use
This central focus of this research involved use the Multinational Time Use Study to explore the international picture of change in time use. Many of our findings confirmed previously observed trends (as for example the continuing, though the slowing, reduction of gender bias in the division of domestic labour – e.g. Gershuny 2000; Fisher et al 2007). Some results challenged previous expectations: Gershuny’s ‘Trends in the Work-life Balance: Veblen in Reverse” (2008), used the most recent release of MTUS data available during this project to show a reversal of the long-term reduction in the total of overall work time (both paid and unpaid) in four Anglophone countries (UK, US, Canada and Australia) but not in Nordic and continental European countries. Others results pointed to unexpected changes in the relative positions of social groups, as in Gershuny (2005) ”Busyness as the Badge of Honour for the New Superordinate Working Class”. At the end of this project, the MTUS provided evidence from 12 countries to test the hypothesis that the previous positive association between social status and leisure time reversed during the last third of the 20th century.
Social mobility
This project investigated the implications of time-use patterns for life-course mobility and life chances, and hence for “social justice”. It brought together the three time-use perspectives (the micro and macro cross-sectional, and the longitudinal) into an integrated—and innovative—account of social change. Mobility chances may be conceived as the joint outcome of the life-course accumulation of human capital, and the distribution of occupational employment opportunities at each point in the life-course. A society's time budget plays a key role in both of these and is therefore a crucial nexus in the determination of social structure and mobility. Key areas of interest were: the impact of women's changing employment circumstances on their husband's housework performance (showing that over time couples tend to adapt to changes in women's employment status - Gershuny, Bittman and Brice 2005); the debate over the significance of relative income against gender ideology in determining men's and women's housework (not supporting the contention that 'gender trumps money' - Kan 2008); the relationship between work orientation and women's lifecourse employment (indicating some support for, and also casting some doubt on, preference theory - Kan 2007); and how women's housework over the lifecourse impacts on wage rates (Bryan and Sevilla-Sanz 2007). This project extended this body of research to the wider assessment of social mobility over time, Gershuny and Kan (2006) and showed how an index of human capital (or 'social position') may be developed which provides a measure of an individual's potential wage in the labour market based on educational level, on degree of present and past attachment to the labour market as well as on present or previous occupational membership. Finally, Gershuny provided an example of the theoretical relationship between a society's time budget, the development of class interests and social change through a mathematical model based on the combinations of embodied capital and personal and family circumstances. Since this model derived from the society's time budget it incorporated inequalities of time as well as money, and therefore was able to address the, sometimes cross-cutting, interests of different social interests both in production and in consumption (Gershuny 2007).
Time use and technological change
Analyses of the relationship between technological and social change based on time use data which included information on domestic appliances and information and communication technologies (ICTs) challenged the simplistic assumptions underlying much previous research on the impact of technological change. For example, a debate between the investigator and a CTUR Research Associate in the British Journal of Sociology over the effect of the ownership of domestic appliances on women's unpaid work (based on cross-sectional Australian time use data) contrasted the argument that women's domestic work time was at best not affected by the ownership of domestic appliances (c.f. Vanek) with the refutation argument that this finding was a methodological trap ('unobserved heterogeneity'), reflecting differences in the sorts of households who are likely to have many domestic appliances and those who are not (Bittman et al 2004; Gershuny 2004). Another strand of research focused on assumptions regarding the effect of ICTs at the micro-social level. This research challenged the idea that the spread of mobile phone technologies irrevocably blurred the distinction between the personal and public spheres (Wacjman et al 2008), undercutting the stereotype of the 'net-nerd' who had no social contact outside of a computer. In an analysis of a special time-diary panel study collected by CTUR, Gershuny showed, contrary to previous results from cross-sectional diary materials, those who started to use the internet during the period of panel observation also exhibited increased non-computer-related sociable activity (Gershuny 2007). Gershuny also was joint editor of a book using time diary and other methods in qualitative and quantitative analyses of the inter-relationships between information and communication technologies and society in European countries, the overall conclusion of which was that the technological changes it identified were complex, widespread, and so far relatively small – but over time the time diary evidence, available from the start of the process of diffusion of informatics, allowed us to track larger changes over longer time scales. (Anderson et al [eds.] 2007).
Methodological Research:
Light/heavy diary comparison
The aim of this project was to investigate the effects of non-response by comparing 'light' diaries (relatively low respondent burden diaries, with short observation windows and precoded activity categories) with 'heavy' (own words diaries with high respondent burden) diary studies. This research initially was jointly undertaken with the ONS, which had an interest in comparing the pre-coded diary instrument attached to the Omnibus survey of 2001 and the ONS time use survey 'own words' diary conducted in 2000 (as the UK contribution to the HETUS) in order to design the diary component of the 2005 Omnibus survey. The main publication was a joint one between Gershuny and members of the ONS staff (Lader, Short and Gershuny 2006), in which conclusions and recommendations resulting from the comparison were outlined. In brief, it was found that the light Omnibus diary provides comparable results to the full scale diary survey for the main sub-categories of the population, and therefore could provide a broad indicator of major changes in time use between full time use surveys. For certain categories of the population (notably the elderly), the interviewer-administered light Omnibus diary proved to be more representative than the self-completed TUS. Although weighting within the main subcategories of the population made little difference to the within-category results, the light Omnibus diary could be weighted for non-response to make the whole sample representative. This procedure was implemented in the 2005 ONS Omnibus diary.
Using questionnaires with diary estimates
The overall aim of this previously untested methodology was to investigate the possibility of arriving at reliable long-term time use accounts at an aggregate level through combining regression-based estimates of time use for population subgroups from diary data with large-scale questionnaire based survey data. One major methodological outcome of this research contradicted a view (fashionable among some journal reviewers at the time) that time diary data on infrequent activities (with large proportions of respondents not participating in activities during the diary observation period) should be modelled using Tobit rather than OLS regression. Gershuny and Egerton concluded that this view was incorrect. They used surveys with long diaries (7 and 21 days) which therefore show relatively few zero time-use values, which have been split into single days with more zero values as a result—even though they sample the same population of individuals and days—to show empirically that Tobit-based analyses lead to inappropriate estimations of the volume and distribution of time spent in activities across populations.
Calibration of BHPS-type stylized questions
Panel data are desirable to investigate how daily activity patterns have effects on individuals over the lifecourse. However, diary panel studies were very rare during this project. Because of concerns about sample attrition rates, many panel studies with a special interest in respondents’ time use deployed questionnaire-type instruments rather than time use diaries in order to arrive at an estimate of the time spent in particular activities. But although they are widely used, the general view is that stylised estimate questions are subject to systematic biases. Gershuny and Robinson used the 7-day work schedule instrument attached to the 2001 UK national time use study, which also collected two days of diary material, to demonstrate that much of the error relates to irregularity in daily and weekly work patterns, such that respondents to the standard LFS “how many hours did you work last week?” questions, simply did not know the answers: for example, men tended to over-estimate their work time, whilst women consistently under-estimated their paid work. Gershuny and Robinson used the panel element of the HoL study (see below) to demonstrate that, while a random 'regression to the mean' effect is evident, around four fifths of the variation between the diary and questionnaire estimates is systematic.
The Home on Line study (HoL) probably was the only panel diary study really suitable for the investigation of such effects. HoL was a national sample, a household time use 3-wave panel study which contains more than 2500 seven-day diaries. It used both a precoded diary instrument and stylized questionnaire estimates of the same time use categories drawn from the BHPS—which enables reliability testing and data fusing. In respect of reliability testing HoL provided the opportunity to compare stylised with diary estimates from the same individuals. In addition, it included exactly the same time-use related stylised questionnaire items as the British Household Panel Study (BHPS), permitting (on the assumption that the diary estimates are the more reliable) calibration of the BHPS questions. Kan and Pudney in Sociological Methodology compared stylised questions and diary reports for the HoL study, and stylised questions from the BHPS (Kan and Pudney 2008). They found that stylised estimates are subject to both random and systematic biases, and provide methods for correcting for the biases introduced by random measurement errors in both types of estimate. Kan and Gershuny (2006) described the results of the data fusion exercise, providing a calibrated index of time-use patterns based on BHPS questionnaire items the major substantive contribution of which is to the analysis of the gender division of labour (see Gender and Time Allocation project), and the resulting dataset was deposited in the UK Data Archive as a BHPS add-on. Both CTUR and ISER prioritised this joint work.
Multinational Time Use Study (MTUS) was central to the activities of the time use group and was our longest running project. It was a continually growing collection of national studies, harmonized for the purposes of comparative research, and currently contained around 42 studies from 15 countries at the end of this project.
The major objective was the extension of MTUS Version 5.0 to include more countries and the most recent datasets, and the initial pilot development of MTUS Version 6 dataset, the first post-collection harmonised dataset to include sequential and episode data.
The older versions of the MTUS (World 5.0 and earlier) covered only the working-age population (20-59) in 25 studies from 20 countries. Version 5.52 doubled the numbers of surveys included in the study, covered the full adult age range and added a number of further background variables. The version at the end of this project (World 5.53) added a further 14 background variables which provided greater compatibility with the HETUS surveys (incorporated into the MTUS) including couple identifiers (see sub-project 11 above); age of youngest child; individual occupation; public or private sector employment; whether the household owned a computer; how many cars the household owned; whether the individual was responsible for caring for a dependent adult or child with disabilities and self-reported health. Data files also were released for the first time in STATA format. A total of 47 studies were included in versions 5.52 and 5.53 combined, representing over 20 countries and more than 1/2 million diary days. Versions 5.8 and 6.0 were under development, and were being tested for a few key countries. Version 5.8 expanded the activity code list to an updated 69 activity code list (the original MTUS only covered 41 activity categories), while Version 6.included activity sequence data (for those countries for which this is available) covering main and secondary activities, location, mode of transport, use of ICTS, and co-presence. Future releases of updated MTUS data were planned.
Continuing efforts were made to add more studies to the MTUS from countries which had contributed data thus far to the MTUS, in particular from Asia. Researchers and the official statistical office in Japan invited Kimberly Fisher to give a presentation on the MTUS with the hope of attracting future data contributions from Asia.