GENDER CONVERGENCE IN THE AMERICAN HERITAGE TIME USE STUDY (AHTUS)

Abstract

We present evidence from a new comprehensive database of harmonized national time-diary data that standardizes information on almost 40 years of daily life in America. The advantages of the diary method over other ways of calculating how time is spent are reviewed, along with its ability to generate more reliable and accurate measures of productive activity than respondent estimates or other alternatives. We then discuss the various procedures used to develop these harmonized data, both to standardize reporting detail and to match with Census Bureau population characteristics. We then use these data to document historical shifts in Americans’ use of time, particularly focusing on gendered change in paid and unpaid work. We explore these data to find new and more complex evidence of continuing gender convergence, not just in aggregated totals of hours worked, but also in (1) the distributions of activity through the day and the week, (2) the sorts of activities that marital partners do together, as well as (3) the processes of construction of the diary accounts themselves.

1. INTRODUCTION

This article introduces the American Heritage Time Use Study (AHTUS), developed by the Centre for Time Use Research (CTUR) at the Institute for Social and Economic Research at the University of Essex, UK. The AHTUS merges the new American Time Use Survey, collected on a continuous basis beginning in 2003 by the US Bureau of Labor Statistics (BLS) with four previous national time-use studies collected by two academic survey centers, in a harmonized format appropriate for a wide range of economic and sociological analyses.¹ We use it, in this article, to identify several dimensions of gender convergence in the US across the latter half of the 20th century. We focus on changes in the distribution of paid and unpaid work tasks over the period 1965–2003, looking not simply at total time devoted to these activities, but at changes in the underlying “time-profiles” of activities through the day, and how husbands and wives arrange their time together.

One major advantage of time diary-based evidence is that time spent in different sorts of activities (paid work, unpaid work, personal care and so forth) can be added together to sum to exactly the 1440 minutes of the day. Previous studies (for instance Robinson and Godbey, 1999; Bianchi, 2000; Aguiar and Hurst, 2006) have focused on overall averages of time devoted to each activity, ignoring the rich contextual information available in these historical diary accounts, concerning who else was present during an activity, and when or where the activity occurred. The analysis in this article moves beyond previous assessments of patterns of historical change, both in terms of the quality of data employed and the complexity of time questions examined.

This article begins with a brief review of the previous national U.S. time-use surveys on which the AHTUS archive is based. Our substantive analyses first concentrate on the overall changes in paid and unpaid work. They show an overall reduction in the total productive work time (the sum of paid and unpaid work), a virtual diachronic constancy of time devoted to sleep and personal care, and a relative increase in time for activities outside production (including leisure and consumption). The general direction of this activity change suggests that women and men lead more similar lives now than 40 years ago, although women’s time patterns have been altered more considerably than men’s. Moreover, these data illustrate the persistence of gender difference in the time when partners are together – when there is unpaid labor to be done during leisure time, women still remain more likely to be the people who carry out the chores.

2. MEASURING TIME USE IN THE USA

The Bureau of Labor Statistics (BLS) initiated the first continuous time-use data collection (the American Time Use Survey – ATUS), interviewing a sub-sample of more than 20,000 respondents aged 15+ from the last wave of the Current Population Survey (CPS) across all months from the start of 2003. This provided the impetus behind the Glaser Foundation’s funding of the systematic harmonization of previous US time-diary use materials for analysis alongside the new ATUS data. The archive draws on the experience of the University of Essex time use group in harmonizing earlier European and other diary studies.

Much of what has been previously surmised about time use in the USA came from time estimate questions – questionnaire items that asked respondents to estimate how much time they spend on some activity over a particular time period, such as per year or in a “usual” week. Among the activities on which national historical data used such estimates include time spent working, doing voluntary work, attending religious services, traveling, socializing, and watching television, as in the General Social Survey (GSS) of the University of Chicago.

Probably the most widely used time-estimate questions have been for market work hours (also collected for BLS in the CPS), in which survey respondents are asked to report how many hours they worked last week and provide an estimate of the usual hours per week they worked in the preceding year. This CPS question has been considered the “gold standard” for assessing change in the work patterns of men and women. One great advantage of CPS estimate questions is that they usually take a respondent less than 10 seconds to answer, in contrast to the full time diary, which can take from 10 to 40 minutes to complete (depending on the degree of detail requested). This makes it far more cost-effective to ask estimate questions in the CPS, which surveys all workers in about 50,000 households every month across the full year. In contrast, time-diary studies have in the past been based on samples of 1000–10,000 respondents, with fresh diary samples typically collected only every 10 years. Another advantage of the paid work estimate question is that it has been administered since the late 1940s, with parallel data going back much earlier, whereas the first US national diary study was only first conducted in 1965 and had notably lower response rates than the estimate surveys. The size of the CPS sample also makes it possible to examine quite detailed breakdowns of work hour estimates by gender, marital status, presence and ages of children, and other household and personal characteristics.

Estimate questions have drawbacks, however. Recalling details about time spent in an activity involves complicated calculations. Asking someone “How many hours did you work last week?” assumes that each respondent interprets “work” the same way, searches memory for all episodes of work, and is able to correctly identify and sum all the episode lengths across the day or across days in the last week. Obtaining accurate responses regarding time use is particularly difficult in the survey context, in which respondents are expected to provide on-the-spot answers in a few seconds. What seems at first to be a simple estimate task turns out to involve several steps that are quite difficult to perform, even for a respondent with regular and clear work hours and a repetitive daily routine. One consequence is that, when asked to provide daily and weekly estimates of several activities, survey respondents give estimates that add up to considerably more than the 168 hours of time each week (Hawes et al., 1975; Verbrugge and Gruber-Baldine, 1993).

The appeal of the time-diary approach is that respondents are not asked to make complex yet vague calculations, but to simply recall their activities sequentially for a speciffic period, usually the previous day. In that way, it becomes possible to reduce the respondents’ recall period and reporting task, first to cover all daily activities, and second to ensure that the resulting account preserves the “zero-sum” property of time – that the activities total exactly 24 hours in a day. That means that if one activity increases across time, it must be compensated-for by a decrease in other ways of spending time.

Diaries² are well-established tools for social scientists. By the time Sorokin and Berger (1939) introduced “time budget” analysis to American sociologists, time diaries has been deployed in a variety of contexts,³ from studies of farm households (Vanek, 1974; Walker and Woods, 1976), unemployed industrial workers in 1930s’ Austria (Jahoda et al., 1972), low paid workers’ families in London and Liverpool (Pember-Reeves, 1913; Jones, 1934) as well as measures of national productivity in the USSR (Strumilin, referenced in Zuzanek, 1980, pp. 10–14). An extensive literature confirms the reliability and validity of diary data (Juster and Stafford, 1985; Robinson and Godbey, 1999; Michelson, 2005).

Research comparing diary methods with questionnaire estimates of paid work vary from articles favoring diary-based estimates (Niemi, 1983; Robinson and Bostrom, 1994; Robinson and Gershuny, 1994) – in terms of work estimates being higher than work time reported in diaries – to those finding only slight advantages to the diary approach (Bonke, 2005). However, there is less research comparing diaries with other methods of estimating hours of unpaid work. Studies which have collected both diary and questionnaire estimates at the same time demonstrate that diaries produce more defensible estimates. However, the nature of systematic differences between diary and questionnaire estimates may be inter-related with other factors, with Danish data suggesting that women’s diary and questionnaire estimates of unpaid work being more divergent than men’s (Bonke, 2005), the reverse pattern emerging in British data (Kan, 2006), and no gender-related divergence, but varying degrees of disparity in the estimates of different age groups in Norway (Kitterød and Lyngstad, 2005). In the U.S, both men and women overestimate their housework by about 50% (Marini and Shelton, 1993; Press and Townsley, 1998; Robinson and Presser, 2000). Some contributors to this literature suggest that aggregate estimates generated by the far less expensive direct questions and the more expensive diaries are sufficiently similar that for research questions where the simple aggregate of time in work is central, either diaries or estimates may suffice (e.g. Bonke, 2005). However, diaries hold the clear advantage where the dynamics or context of work is central to the research (such as examining work episodes in relation to time spent with household members as in Figure 6, the spread of work across the day (as in Figure 7), how work is associated with travel or time at home, and the integration of work life with leisure time) (Gershuny, 2000; Fisher and Layte, 2004; Michelson, 2005).

Fig. 6.

Change in activities through the day, weekdays, 1965 and 2003. Source: AHTUS data released 2006.

Fig. 7.

Married or cohabiting: co-presence with partner, 1965 and 2003. Source: 2003 vs. 1965 AHTUS data.

That is not to say the diary method is without flaws. Respondents can still embellish accounts, but in order to portray themselves as hard workers or light television viewers, respondents must fabricate the activities that precede and follow the one they want to exaggerate. As diaries cover only a short period, respondents probably realize that they may work less or watch television more than usual on any given day, so the incentive to distort is reduced (though certain activities, particularly criminal or violent activities seldom explicitly appear in time-diary accounts). Diary keepers who simply cannot remember what they did at a particular time may substitute a habitual activity for what actually took place. Even so, the diary still presents us with a far richer and more persuasive source of individual and family activity patterns than any present alternative (Robinson and Godbey, 1999; Gershuny, 2000; Michelson, 2005).

2.1. Time-diary Methodology in the USA

Most time-diary studies in the United States retain elements of the first national-scale USA diary study, one carried out as part of the most extensive comparative time-diary studies of its time – the 1965 Multinational Time Budget Study (Szalai, 1972). That survey collected a single-day 24-hour time diary registering main activity, simultaneous activity, location, and who else was present, as reported in the respondent’s own words and were coded using a harmonized coding frame. Diaries were collected from one person aged 19–65 in urban households in which at least one member was employed and from each of 12 different countries. Until the 2003 ATUS, which collected half of diaries on weekend days and half on weekdays, studies were designed to collect diaries from an even distribution of days of the week. Since the 1975–76 study, the national USA studies have spread the collected data across all seasons of the year. Except for 1965, survey weights balance the distribution of AHTUS diaries by season and day of the week. A schematic comparison of methodological differences across the surveys is provided by Table I, with more detail outlined in the Appendix.

TABLE I.

Methodological details of surveys currently in the AHTUS database

	1965–66	1975–1976	1985	1992–1994	2003
Survey organization	Survey Research Center, University of Michigan,	Survey Research Center, University of Michigan	Survey Research Center, University of Maryland	Survey Research Center, University of Maryland	United States Bureau of Labor Statistics and United States Census Bureau
Funder	National Science Foundation (NSF)	NSF, US Depart- ment of Health, Education, & Wel- fare	NSF; ATT	Environmental Pro- tection Agency	USA Department of Labor
Sample	Jackson 759 diaries; National 1262 dia- ries	4584 diaries (main respondent only)	2636 diaries	7514 diaries	19,663 diaries
Age range	19–65 (some up to 69)	18 +	12+ (18+ in AH- TUS)	0+ (18+ in AH- TUS)	15+ (18+ in AHTUS)
Months	November–Decem- ber 1965; January– February; March– May 1966	October–November 1975; February– March; May–July; September–October 1976	January–December 1985	September 1992– October 1994	January– December 2003
Response rate	82% Jackson; 74% national sample	72% first wave; 45% did all 4 waves	55% overall, 51% for mail back sample	63%	58%
Mode of collection	Self-completion with guidance from inter- viewer	Self-report waves 1&4; phone waves 2– 3	3 samples: CATI; self-completion, per- sonal interview (only self-completed in AHTUS)	CATI	CATI
Diary type	Mostly designated day	Previous day	Designated day	Previous day	Previous day
Notes on sample	Sample of urban households having at least one employed member; separate national and Jackson MI samples	Sample of all house- holds; longitudinal–4 waves; reduced dia- ries for spouses (if married)	Sample of telephone households; all eligi- ble household mem- bers asked to keep diaries	Sample of telephone households	Sample of all households; former respondents in CPS Wave 8; Par- ents with dependent children over-sampled

2.2. Producing the AHTUS

The harmonization exercise was designed to retain the highest level of accurate detail possible that could be found across the various surveys.⁴ Though the activity information was not always collected in the same way in each survey (as is evident in Table I), the activity variables are consistent enough across studies to be used for cross-time analysis (Egerton et al., 2005) – with the exception of the 2003 ATUS secondary childcare estimates, which are starkly out of line with secondary care recorded in the earlier surveys (Bianchi et al., 2005; Fisher, 2005). This harmonization effort revealed the signifficance of using all the variables in the diary when classifying activities, rather than simply relying in the entry in the main activity column (Fisher, 2006). We did not over-write information recorded in the original surveys, though we have corrected a number of data-entry errors in the time diary episode files of some of the older datasets.

We conducted filtering and consistency checks for all variables, and for the same variables across waves for respondents in longitudinal 1975–76 survey. We have included a number of flag variables that mark cases of inconsistent information. Though error levels for demographic variables were low in most surveys, we encountered a number of problems from corrupted episode data files from the 1985 survey (that may be corrected in the future by matching them with the apparently uncorrupted simple time-in-activity files).

We made a number of adjustments to the diary files, imputing activity codes in short gaps for cases where we could make logical inferences of the unrecorded activity by using information respondents recorded in the location and social partner columns, as well as from constraints imposed by the combinations of activities before and after these short gaps. We also used information in these other fields in the diary to disaggregate some original “primary” activity codes,⁵ in order to construct the activity codes for the harmonized file. These imputed time activities have separate codes, and thus are easily distinguished from the originally coded activities.

We amended original survey weights to match CPS age and sex profiles and to balance the distribution of diaries by day of the week and season after excluding poor quality diaries,⁶ and also to compensate for attrition in the 1975–76 survey. These weights ensure an appropriate cell distribution of days-of-the-week for each sex and age group in each of the component surveys after the exclusion of the low quality diaries. The resulting AHTUS variable distributions did not differ markedly from population statistics, with the important exception that larger proportions of the AHTUS respondents were well educated than found in the population (Egerton et al., 2005) – a typical problem in U.S. surveys, particularly those conducted by telephone. More detail appears in Appendix, with full documentation and the data available at http://www.timeuse.org/ahtus/.

3. CHANGES IN THE DAILY ACTIVITIES OF WOMEN AND MEN IN THE USA

Our tabulation of change in the allocation of productive time is for US adults aged 19–64, an age range used both because the 1965–66 samples only collected diaries from working age people and because trends in the changes in total work are most plainly evident amongst the population for the ages that perform the majority of paid work. We focus on the evolution of work, in the broadest sense – following the conceptualizations of Walker and Woods (1976) and Hawrylyshn (1977) – of activities that someone might commission a “third party” to carry out for pay without losing the main sorts of direct utility derived from that activity.⁷

3.1. Trends in Paid and Unpaid Work

The AHTUS dataset produces striking pictures of change in total work time, and of continuing convergence between the different sorts of work undertaken by men and women in the USA, which are, even after the extensive recoding and standardization described above, not dissimilar to those reported in Robinson and Godbey (1999). The first sequence of diachronic graphs (Figures 1, 4, and 5) traces trends in work for men and women aged 19–64 separately, showing trend averages and their 95% confidence intervals. Figure 1 first shows time in paid work, education, training, and other activities associated with work or education (commuting, applying for jobs or courses) from 1965 to 2003. Figure 4 then displays time devoted all unpaid work (including yard work and shopping but excluding – for reasons explained below – child care). Figure 5 then sums Figures 1 and 2, and additionally includes time in child care (as a “primary activity”), to show patterns of total work time.

Fig. 1.

Paid work: 95% confidence. Source: AHTUS data released 2006.

Fig. 4.

All domestic work: 95% confidence. Source: AHTUS data released 2006.

Fig. 5.

Total work including childcare: 95% confidence. Source: AHTUS data released 2006.

Fig. 2.

Mean minutes of paid work and commuting – excluding breaks at work (weighted). Source: AHTUS data release 2006.

Figure 1 shows the statistically signifficant decline in men’s time devoted to paid work along with other activities in the workplace and activities which enhance or facilitate work prospects (job applications, training) over the first three decades. The second two decades show a small, but probably not statistically signifficant increase, which still leaves the men’s 2003 mean time in activities associated with paid work and education substantially (and signifficantly) below the 1965 estimate. The women of this age group show a partially opposing trend, with an initial rise in paid work and related time, but with the major statistically signifficant increase concentrated during the 1985–1995 decade, as signifficantly more women entered the labor force.

Others who looked at change in the four broadest categories of time use (paid work and related, unpaid work, sleep and personal care, non-committed time) noted this trend before the release of the 2003 ATUS (for instance Robinson and Bostrom, 1994; Robinson and Godbey, 1999). The finding has raised some controversy from other researchers with a particular interest in hours in paid work, who argue that life in the United States has grown more harried with passing decades, but who base their conclusions on the BLS work estimate data rather than diary data. Schor (1993) argued that work estimate data between 1979 and 1991 showed that hours actually increased signifficantly, that peculiarities of the smaller diary samples offer a better explanation than a actual behavior change for this drop; and that the general expansion of women’s participation in the labor force mean that hours worked by the population of the USA as a whole have risen. Jacobs and Gerson (2004) argue that estimate hours worked have held roughly constant over the last four decades, but that the proportion of people working has increased and the total number of holidays taken by Americans has remained constant or even fallen, meaning that the total time committed by the population to paid work has increased (29–31). However, Figure 1 confirms that men’s work time has been roughly constant – but over the last two (not four) decades, while women’s has obviously increased as more of them have entered the labor force. In that way, the findings presented here do not necessarily conflict with the conclusions of Jacobs and Gerson.

This article considers the broad domain of activity associated with paid work, education and training for work among people of working age – a somewhat broader category that the speciffic matter of hours of paid work of interest to Schor and Jacobs and Gerson. Hidden within this summary pattern lies a range of divergent social changes, including the early retirement in some industries, the longer work-days now undertaken by the better educated, both of which are outside the scope of the present article. While the figures here do support the suggestion that the 1965–66 sample is somewhat idiosyncratic (reffecting the special sample selection criteria of the Szalai study), nevertheless Figure 1 indicates trends from 1975–76 onwards which are generally consistent with those from the 1965–66 data.

Figure 2 examines paid work time more closely by age and indicates that women’s hours of paid work (plus commuting) only increased from 1975– 76, then leveled from the mid 1990s. Figure 3 lends more support for the work intensiffication hypothesis. The relatively level proportion of time dedicated to paid employment tasks in Figure 2 contrasts sharply with the time taking a break at work (as opposed to doing something else, like making private phone calls or going to the bank during break time). Figure 3, which shows the average minutes on break while at paid work (only among workers reporting some paid work activity in their diaries) has dropped steadily from 20 minutes in 1965–66 to roughly only half a minute in 2003. Correspondingly, the proportion of workers taking a work break dropped from over half in 1965–66, to under half in 1975–76, around a third in 1985, less than 7% in 1992–94, and less than 2% in 2003. While some of this decline may be a function of fewer activities reported in later surveys (so that short episode activities like breaks would be less likely to be reported) or that work breaks are less formal and scheduled, this is the clearest possible diary evidence of a more harried work life.

Fig. 3.

Mean minutes of breaks at work (weighted). Source: AHTUS data release 2006.

Figures 1 to 4 also show how the increasing proportion of women in the workplace has increased the total daily participation rate of paid work in the United States. While people are at the workplace, women’s hours of paid work – and thus total hours of paid work for the whole population – have slightly increased, as “down time” at work in (i.e. breaks not involving other activities), has faded from the conscious daily life of people in the USA. Taking the whole range of activities associated with paid work and training for or applying for work (i.e. including this “down time”), the overall time devoted by men has declined and by women has increased. That is the reverse of the pattern for unpaid work examined in Figure 4.

Figure 4 charts the trends for all unpaid domestic work (including cooking, cleaning, yard work and shopping, but excluding childcare) over the period. Men’s unpaid work rose signifficantly during the first half of the period, and showed a (non-signifficant) decline – perhaps best interpreted for the moment as “no change” – during the second half of the period. Men of all ages performed an average of two episodes of unpaid work per day, and between 85% and 90% of men participated in unpaid work on any given day.

In contrast, nearly 100% of women engaged in some form of domestic work each day. In contrast with men, as Figure 4 suggests, women’s total time in unpaid work declined signifficantly across the period essentially offsetting their increased time in paid work. Though this drop primarily results from reduction in women’s time food preparation and cooking (Hamermesh, 2005), the average number of episodes of unpaid work for women of all ages dropped from 5 in 1965–66, to 4 in the 1970s and 1980s, to 3 in the 1990s and in 2003.

The AHTUS data do exhibit a surprising increase in time main activity childcare – an apparent doubling for women and a tripling for men (not shown here). Bianchi et al. (2005) compare main activity care time in the ATUS with two more recent USA time-use studies not presently included in the AHTUS (that is 1998–1999 and 1999–2001) that show a similar jump in childcare time at the end of the century. Perhaps increased concern with children’s safety prompted by events such as Columbine and September 11 may partially account for this rise,⁸ though we cannot at present exclude the possibility that these results are also influenced by the 2003 ATUS instrument itself.

Irrespective of these concerns, childcare clearly falls within the “third person” criterion of work, so, in the estimates of the trends in the totals of work given in Figure 5, we have added in childcare to the paid and unpaid work estimates set out in the previous two figures. We see what can only be described as a striking reaffirmation of the almost 15% fall in total work time from 1965 to 1985. That decline, as well as the amount of total productive activity, is remarkably similar for both men and women – especially so given the different composition in terms of paid and unpaid ratios for men and women. The small increase in the total of work time since the mid-1980s is plainly not statistically signifficant. But this aggregate view of the trend in the two more recent decades again masks the substantial changes in these distributions (particularly of paid work participation rates and work times, among both men and women): those with higher levels of educational or career attainment work more, and those with lower levels of attainment work less.

3.2. Daily Sequences in 1965 and 2003

So far we have considered the simple averages of time in “primary” activities. But diary evidence presented in this way masks much of its unique information about the times of day that the various activities are undertaken by different sorts of people. Figure 6 presents just the very simplest sorts of comparisons, of primary activities through the course of weekdays during 1965 and 2003, for the same population aged 19–64.⁹ Reading from left to right through each of these four charts, one can see that the bottom band, representing sleep and personal care, diminishes rapidly from 7 AM (at which time around 40% of men and 50% of women are still asleep). By 10 AM much the largest part of the society’s time is devoted to the two central bands of activity of paid and unpaid work. Thereafter the top band, representing leisure, rather continuously increases through the day, reaching its maximum at around 9 PM, after which an increasing proportion of the population goes to bed – an activity which approaches 90% of the population by midnight.

Comparing first the pairs of charts vertically, we find a considerable difference between men’s and women’s activities in the 1960s in terms of the relative thickness of the bands representing paid and unpaid work. For women as a group, we can see that paid work represented a little less than half of all work activity throughout the majority of the day, whereas for men it represents something more like 90%–95% until mid-evening, and unpaid work – at no point more than about one-fifth of all work. Considering this difference in types of work activity, it is perhaps surprising that the overall width of the two central bands is quite similar, so that men’s and women’s daily leisure profiles in 1965 resemble each other quite closely.

The changes in men’s activities through the weekday from 1965 to 2003 reffect exactly what we have already seen in the aggregate time-trends in Figures 1 and 6. Men have substantially reduced their paid work – reffected in Figure 6 as a reduced width band throughout the day. Their increase in unpaid work is also distributed throughout the day, although in this case apparently concentrated more in the evening. Still, the changes remain relatively minor: men in 2003 behaved in their weekdays pretty much as men in 1965.

But the change is far more dramatic for women. By 2003, their unpaid work in the middle of the day had diminished to occupy less than one quarter of the total of work time. Women in 2003 still do more unpaid work – more than twice as much unpaid work as men at any point in the day. Nonetheless, as much as the men of 2003 still look like the men of 1965, the profile of a woman’s day in 2003 resembles the men’s 2003 daily profile more than it does the women’s day-profile in 1965.

3.3. Multiple Activities from Event Data: Spouse Co-presence and Housework

Figure 7 compares the profile of co-presence with marital or cohabiting partners throughout the day for 1965 and 2003, separately for husbands and for wives. We see that the male and female accounts correspond reasonably well with each other, with some small divergences in reported levels of copresence (in both years, and for the moment unexplained) emerging only around midnight. There is a clear historical change in the profile during the afternoon and early evening, which is explored further by combining activity and co-presence data in Table II.

TABLE II.

Minutes per day of co-presence, ages 18–64, married or cohabiting: 1965–2003

	Means						Standard errors
	1965			2003			1965		2003
	Total	With partner	CoPres Ratio	Total	With partner	CoPres Ratio	All	With partner	All	With partner
All co-presence
Men	1440	249	0.17	1439	271	0.19		6.4		3.4
Women	1440	245	0.17	1439	266	0.18		6.3		3.3
Unpaid work
Men	65	20	0.31	88	33	0.38	3.2	1.4	1.8	1
Women	293	47	0.16	176	45	0.25	5.4	2.2	2.1	1.2
M/F ratio	0.22	0.42	1.9	0.5	0.75	1.49
Paid work
Men	489	55	0.11	394	35	0.09	8.5	1.8	3.9	0.6
Women	200	47	0.23	257	32	0.12	7.3	1.4	3.5	0.6
M/F ratio	2.44	1.19	0.49	1.54	1.09	0.71
Consumption
Men	363	169	0.47	435	201	0.46	6.6	5.3	3.1	2.8
Women	394	148	0.38	451	188	0.42	5.9	4.7	2.8	2.6
M/F ratio	0.92	1.14	1.24	0.96	1.07	1.11

The first panel of Table II (labeled All Co-presence) simply calculates time spent with the spouse or partner through the whole day. The men’s and women’s totals differ only insignifficantly – by less than 5 minutes per day. There does seem to have been a small and but statistically signifficant increase in partner co-presence of about 20 minutes per day over the almost 40-year period. The “partner co-presence ratio”, (defined as the proportion of available time spent in the company of the partner) only rises from .17 in 1965 to .18 or .19 in 2003.

By calculating the product of the co-presence and the primary activity fields, the three lower panels in Table II relate the three broad categories of unpaid work, paid work and consumption time to the partner’s co-presence. These three scores very nearly sum to the total of partner co-presence (the fourth general category, personal care and sleep, is omitted because very little co-presence is reported). In these lower panels we see more evidence both of historical change and some considerable gender differences. This latter phenomenon is quite remarkable, and potentially revealing of the process through which the diary accounts are constructed.

While the surveys in the AHTUS dataset do not allow us to directly test the co-presence reports of spouses/partners,¹⁰ we can compare the accounts given by women in couples of their time with husbands with the accounts given by men of their time with their wives. Consider the 1965 entries for partner co-presence in the unpaid work panel. The sample of “wife” diarists report spending 47 minutes engaged in domestic work while with their male partners, whereas the sample of “husband” diarists report spending only 20 minutes engaged in domestic work with their partners. Most of this 27-minute difference could be accounted for by the 21 extra minutes that men account for as leisure with their spouse present. This implies that when couples were together in 1965, the husband often took leisure while the wife spent her time doing housework (as for example where he sits in front of the television, perhaps talking with her while she sets the table and finishes food preparation).

More important for our present purpose is the historical convergence in the two columns of male/female co-presence ratios. This difference between the partners, while still evident in 2003, is much reduced in scale: the processes of personal representation within the social construction of time use appear to become more similar, just as men’s and women’s overall patterns of time use come to resemble each other more closely.

4. CONCLUSIONS

We have documented the reduction in American men’s paid work over the period 1965–2003, which was accompanied by a smaller increase in their unpaid work, a change that was concentrated within the first three decades of the five-decade period. We saw an even more substantial decrease in women’s unpaid work, again concentrated mostly in the first three decades of the period, more or less offset by their increases in paid work. Women’s and men’s aggregate allocation of time – and also the sequential organization of their days – look more similar over time. But differences still remain; earlier convergence processes appear to have stalled. And this process was disproportionately that of women’s activities coming to look like men’s, rather than the reverse.

The new harmonized cross-time dataset provides, in addition to standardized demographic variables, a new activity-coding scheme – intended to constitute benchmark categories for future comparisons of US time allocation studies. Together with the other diary fields, harmonized for the first time within the AHTUS, these provide a much more fiexible and comprehensive approach to time-use research, one in which analysts can exploit all the features of the diary instrument, rather than the almost exclusive focus on averages of primary activities that characterizes previous time-diary research. The provision of the harmonized, diachronically comparative, episode-file data, including primary and secondary activities, location and co-presence, now allows analysts to reconceptualize the daily activity focus, combining evidence from different diary fields to make use of the extensive sequence and time-of-day information recorded in the original diary instruments – factors entirely lost in the traditional time-diary analyses to date.

This new data archive also provides a major opportunity for international comparisons with the new Harmonized European Time Use Study (HETUS) coordinated by Eurostat, with harmonized diary studies from more than a dozen EU states collected around 2000, and with the Multinational Time Use Study (MTUS) which allows international comparisons with more than 20 countries, stretching back to the 1960s. Our group is presently working to merge the MTUS and the HETUS with the AHTUS to further extend the range of national coverage, with the goal of providing a source of micro-level information on time use similar to the information on money resources provided by the widely used Luxembourg Income Study.

Since the ATUS is now established as a continuous survey; and the HETUS exercise is scheduled for a new round of data collection around 2010, the international comparative time-use research framework, as envisioned in the pioneering UNESCO-funded Szalai study more than a half-century ago, now moves much closer to reality.

APPENDIX – BRIEF INTRODUCTION TO THE AHTUS

Summary Description

This appendix covers additional information on the AHTUS. AHTUS data and documentation may be downloaded from http://www.timeuse.org/ahtus/.¹¹

The open-ended diary reports from each of the original USA surveys were coded using a standard activity coding scheme, largely based on the code list developed for the 1965 Szalai (1972) project, consisting of about 100 (or which 85 are available in surviving datasets) general (“2-digit”) activity codes, and sometimes broken down into a more detailed “3-digit” classiffication with approximately 250 activity categories. The designers of the BLS survey devised a new classiffication scheme, influenced by the Eurostat (2004) 167 category activity classiffication from the HETUS and the Australian Bureau of Statistics code frame (215 activity codes), but which also reffected the priorities of various US government agencies, such as time spent completing security procedures. The ATUS code includes 564 categories, which have been reduced in the AHTUS to 92 categories which appear in the majority of the surveys (detailed below).

In additional to making the harmonization programs available to researchers, the dataset includes three harmonized data files for each original survey:

• a respondent-level file with harmonized information about individuals and households

• a diary-level file coded into 92 main activity categories

• an episode-level file in which each row contains each activity recorded by each diarist

The episode level file contains the full breakdown of context information (to the extent recorded) for each episode – the main activity, any simultaneous secondary activity, its location (see below), mode of transport (see below), and who else was present.

The AHTUS’ provision of this episode-level data is unique among harmonized comparative time-use archives. The diary-level file with its aggregated totals of time devoted to primary activities is made available for the simplest sorts of summary statistical calculations, but we expect that a growing proportion of analysts will start with the episode file, using relevant context information to construct a summary file appropriate to the analyst’s needs. The episode file also allows analysis of patterns of activity and timing of activities through the day.

Surveys Currently Included in the AHTUS

1965–1966 Time-use survey

The oldest dataset included in the AHTUS is the 1965 survey collected by the Survey Research Center at the University of Michigan. This study has two relatively small samples, one which followed the Szalai survey methodology (to sample a typical industrial mid-sized urban location), and a second national sample of all urban areas (with 2021 diaries collected across both samples). Both surveys sampled households where at least one member was employed in an industry other than agriculture, then selected one adult aged 19–65 to keep a single-day diary of activities. Respondents in this 1965 survey completed “tomorrow” diaries, that is respondents were visited by an interviewer who explained and left the diary to be filled out for the following day; the interviewer then returned on the day after that “diary day” to check over, correct and collect the completed diary (Robinson, 1977). Sayer et al. (2004) compared the 1965 sample characteristics with parallel characteristics from the March 1965 CPS, and concluded that its sample closely approximates U.S. population characteristics. An analysis of the full national sample of 1975 diaries indicated that the activities reported by that full sample matched those who would have met the 1965 criteria (Juster and Stafford, 1985).

1975–1976 American’s use of time: time use in economic and social accounts survey

In 1975, the Survey Research Center, University of Michigan, personally interviewed 1519 adult respondents aged 18 and over, who reported diaries for the previous day in the Fall of that year (Robinson, 1976); in addition, diaries were obtained from 887 spouses of these designated respondents, which increased the sample size to 2406 respondents. These respondents became part of a panel, who were subsequently re-interviewed in the Winter, Spring, and Summer months of 1976.¹² High levels of attrition in the later panel waves and problems in using the original file (which is not at all user-friendly, and contains some hitherto unidentiffed major errors¹³) explain why virtually all previous analyses (including Aguiar and Hurst, 2006) have simply ignored the subsequent waves, and analyzed only the first wave (including spouses) – adopting appropriate weights to compensate for the over-representation of couple households. In the AHTUS files, we have adopted precisely the opposite approach, using all four waves of data (with additional sample weights to compensate for non-response). As the spouse diaries include less information than main respondent diaries (spouses were asked to record main activity and location only, while main respondent diaries include main and secondary activity, location, and presence of others), we produced a separate supplementary file that included both the main respondent and the spouse diaries for all four waves with distribution and attrition weights. In this paper we use only the main respondent file.

1985 American’s use of time survey

In 1985, the Survey Research Center at the University of Maryland conducted a national study in which single-day diaries were collected from more than 5300 respondents aged 12 and over. This study employed the same basic open-ended diary approach as the 1965 and 1975 national studies. An important innovation in the 1985 study was the explicit attempt to spread the collection of diary days across the entire calendar year, from January through December 1985.

The 1985 study included experimentation with mode of data collection. The majority of diaries in the 1985 study were collected by a mail-back method from a sample of Americans who were first contacted and completed a “yesterday” diary by telephone, using the random-digit-dial (RDD) method of selecting telephone numbers. If the respondent agreed, diaries were then mailed out for each member of the participating household, aged 12 or over, to complete for a particular day for the subsequent week. Respondents completed and then mailed back their time diaries for coding and analysis.

Some 3340 diaries from 997 households were returned using this mail-out procedure during the 12 months of 1985. The other 1985 data included parallel diary data from 808 additional respondents interviewed in a separate personal-interview sample in the summer and fall of 1985, in addition to the 1210 “yesterday” diaries obtained by telephone as part of the initial contact. Unfortunately, the episode level data are no longer available for the personal-interview and telephone-interview samples. The AHTUS episode file consequently covers only the mail-back sample, and early testing of this file suggests that some degree of error remains in the data (Gershuny, 2005). We use only aggregated data from the 1985 mail-back sample in this article.

1992–1994 National human activity pattern survey (NHAPS)

The University of Maryland’s Survey Research Center conducted national RDD telephone interviews between September 1992 and October 1994, collecting 9386 diaries about the previous day from respondents of all ages (parents were asked to complete diaries for young children when a young child was selected as the diary keeper in the household). Only those respondents aged 18 and above are included in the main AHTUS files, with diaries from younger people in separate supplementary files (not used in the foregoing analysis). This study did not include pivotal questions about marital status and income. A 1995 survey followed a similar methodology (for people aged 18 and older) but asked the income and marital status questions. This is not currently included in the AHTUS, but may be added in the future.

2003 ATUS

The BLS began collecting time diaries from one person per household in a sub-sample of households that completed the eighth and final wave of the CPS. The survey collects diaries throughout the year. This sub-sample over-samples households with young children and only included people aged 15 and older. All diaries are collected over the telephone (with people in households without a phone sent a voucher to call and complete the diary from a pay phone) about the previous day’s activities. Half of diaries were collected on weekdays and the other half on weekend days. The large sample size permits breakdown of time by more detailed population groups than is possible in the smaller and older datasets. While the ATUS is a continuous and on-going study, only the 2003 data are included at this time.

Harmonized Activity Categories in the AHTUS

		1965–66	1975–76	1985	1992–94	2003
Harmonized activity categories in the AHTUS
−8	Item missing	×	×	×	×	×
1	General or other personal care	×	×	×	×	×
2	Imputed personal or household care	×	×	×	×	×
3	Sleep	×	×	×	×	×
4	Imputed sleep	×	×	×	×	×
5	Naps and rest	×	×	×	NO	×
6	Wash, dress, personal care	×	×	×	×	×
7	Personal medical care	×	×	×	×	×
8	Meals at work	×	×	×	NO	×
9	Other meals & snacks	×	×	×	×	×
10	Main paid work (not at home)	×	×	×	×	×
11	Paid work at home	×	×	×	×	×
12	Second job, other paid work	×	×	×	×	×
13	Work breaks	×	×	×	×	×
14	Other time at workplace	×	×	×	NO	×
15	Time looking for work	NO	×	×	×	×
16	Regular schooling, education	×	×	×	×	×
17	Homework	×	×	×	×	×
18	Short course or training	×	×	×	×	×
19	Occasional or other education/training	×	×	×	×	×
20	Food preparation, cooking	×	×	×	×	×
21	Set table, wash/put away dishes	×	×	×	×	×
22	Cleaning	×	×	×	×	×
23	Laundry, ironing, clothing repair	×	×	×	×	×
24	Home repairs, maintain vehicle	×	×	×	×	×
25	Other domestic work	×	×	×	×	×
26	Purchase routine goods	×	×	×	×	×
27	Purchase consumer durables	×	×	×	×	×
28	Purchase personal services	×	×	×	×	×
29	Purchase medical services	×	×	×	×	×
30	Purchase repair, laundry services	×	×	×	×	×
31	Financial/government services	×	×	×	×	×
32	Purchase other services	×	×	×	×	×
33	Care of infants	×	×	×	×	×
34	General care of older children	×	×	×	×	×
35	Medical care of children	×	×	×	×	×
36	Play with children	×	×	×	×	×
37	Supervise child or help with	×	×	×	×	×
38	Homework	×	×	×	×	×
39	Read to, talk with child	×	×	×	×	×
40	Adult care	×	×	×	×	×
41	General voluntary acts	×	×	×	×	×
42	Political and civic activity	×	×	×	×	×
43	Union and professional activities	NO	×	×	×	NO
44	Volunteer child/family organization	NO	×	×	×	NO
45	Volunteer fraternal organization	NO	×	×	×	NO
46	Other formal volunteering	×	×	×	×	NO
48	Acts for religious organization	×	×	×	×	NO
49	Worship and religious acts	×	×	×	×	×
50	General out-of-home leisure	×	×	×	NO	×
51	Attend sporting event	×	×	×	×	×
52	Go to cinema	×	×	×	×	×
53	Theater, concert, opera	×	×	×	×	×
54	Museums, exhibitions	×	×	×	×	×
55	Attend other public event	×	×	×	×	NO
56	Restaurant, cafe bar	×	×	×	×	×
57	Parties or receptions	×	×	×	×	×
58	Imputed time away from home	×	×	×	×	×
60	Sports & exercise	×	×	×	×	×
62	Walking	×	×	×	×	×
63	Cycling	NO	×	×	×	×
64	Outdoor recreation	NO	×	×	×	×
65	Physical activity, sports with child	×	×	×	×	×
66	Hunting, fishing, boating, hiking	×	×	×	NO	×
67	Gardening	×	×	×	×	×
68	Pet care, walk dogs	×	×	×	×	×
70	General indoor leisure	×	×	×	×	×
71	Imputed in-home social	×	×	×	×	×
72	Receive or visit friends	×	×	×	×	×
73	Other in-home social, games	×	×	×	×	×
74	Play musical instrument, sing, act	×	×	×	×	NO
75	Artistic activity	×	×	×	×	×
76	Crafts	×	×	×	×	×
77	Hobbies	×	×	×	×	×
78	Relax, think, do nothing	×	×	×	×	×
81	Read books	×	×	×	×	×
82	Read periodicals	×	×	×	×	NO
83	Read newspapers	×	×	×	×	NO
84	Listen to music (CD etc.)	×	×	×	×	×
85	Listen to radio	×	×	×	×	×
86	Watch television, video	×	×	×	×	×
87	Writing by hand	×	×	×	×	×
88	Conversation, phone, texting	×	×	×	×	×
89	Use computer	NO	NO	×	×	×
90	Imputed travel	×	×	×	×	×
91	Personal or adult care travel	×	×	×	×	×
92	Travel as part of paid work	NO	NO	NO	×	×
93	Travel to/from work + other work travel	×	×	×	×	×
94	Travel related to education	×	×	×	×	×
95	Travel related to consumption	×	×	×	×	×
96	Travel related to child care	×	×	×	×	×
97	Travel for volunteering or worship	×	×	×	×	×
98	Other travel	×	×	×	×	×
Location variables and category codes in the AHTUS
INOUT – outside, inside or in vehicle
−8	Location unknown	×	×	×	×	×
1	Outside	×	×	×	×	×
2	Inside	×	×	×	×	×
3	In a vehicle	×	×	×	×	×
ELOC – location, includes implied from activity codes as well as diary columns
−8	Location unknown	×	×	×	no	×
1	Own home	×	×	×	×	×
2	Other home	×	×	×	×	×
3	Workplace	×	×	×	×	×
4	School	×	×	×	×	×
5	Services or shops	×	×	×	×	×
6	Restaurant, café, bar	×	×	×	×	×
7	Place of worship	×	×	×	×	×
8	Traveling	×	×	×	×	×
9	Other	×	×	×	×	×
MTRAV – mode of travel
−8	Not answered	Not present	Not present	×	×	No
−7	Not traveling			×	×	×
1	Car, truck, motorcycle			×	×	×
2	Public, mass transport			×	×	×
3	Walk (including child carried)			×	×	×
4	Cycle			Limited	×	×
5	Other or unspecified mode			×	×	×

Sample Distribution by Selected Classificatory Characteristic AHTUS

Weighted distribution (frequency and column %) of age by survey

		1960s	1970s	1980s	1990s	2003
18–24	Count	337	850	403	856	2205
	%	16.95	19.45	15.79	12.39	12.49
25–34	Count	418	936	605	1472	3288
	%	21.03	21.41	23.7	21.3	18.63
35–44	Count	470	579	475	1514	3685
	%	23.64	13.25	18.61	21.91	20.88
45–54	Count	437	659	339	1135	3409
	%	21.98	15.08	13.28	16.42	19.32
55–64	Count	298	589	331	767	2331
	%	14.99	13.48	12.97	11.10	13.21
65plus	Count	28	758	400	1167	2731
	%	1.41	17.34	15.67	16.89	15.47
All	Count	1988	4371	2553	6911	17,649
	%	100	100	100	100	100

Weighted distribution (frequency and column %) of sex by survey

		1960s	1970s	1980s	1990s	2003
Men	Count	942	1991	1179	3074	8407	15,593
	%	47.38	45.55	46.16	44.47	47.63	46.58
Women	Count	1046	2380	1375	3839	9242	17,882
	%	52.62	54.45	53.84	55.53	52.37	53.42
All	Count	1988	4371	2554	6913	17,649	33,475
		100	100	100	100	100	100

Weighted distribution (frequency and column %) of education level by survey

		1960s	1970s	1980s	1990s	2003
0–8th grade	Count	257	618	163	200	726
	%	13.01	14.21	6.44	2.91	4.11
9–11th grade	Count	403	629	241	513	1464
	%	20.39	14.47	9.53	7.46	8.29
High school graduate	Count	775	1677	1094	2371	5101
	%	39.22	38.57	43.24	34.5	28.9
Some college	Count	289	687	455	1731	3549
	%	14.63	15.80	17.98	25.19	20.11
College graduate	Count	206	391	393	1182	4921
	%	10.43	8.99	15.53	17.2	27.88
Post college	Count	46	346	184	876	1889
	%	2.33	7.96	7.27	12.75	10.7
All	Count	1976	4348	2530	6873	17,650
	%	100	100	100	100	100

Weighted distribution (frequency and column %) of economic activity by survey

		1960s	1970s	1980s	1990s	2003
Employed full-time	Count	1368	2370	1349	3887	9822
	%	69.41	54.91	53.94	56.51	55.65
Employed part-time	Count	54	269	234	738	2335
	%	2.74	6.23	9.36	10.73	13.23
Not employed	Count	549	1677	918	2253	5492
	%	27.85	38.86	36.71	32.76	31.12
	Count	1971	4316	2501	6878	17,649
	%	100	100	100	100	100

Weighted distribution (frequency and column %) of marital status by survey

		1960s	1970s	1980s	2003
Married	Count	1594	2796	1636	10,181
	%	80.18	64.07	64.43	57.68
Separated, divorced	Count	109	418	200	2261
	%	5.48	9.58	7.88	12.81
Widowed	Count	83	483	183	1269
	%	4.18	11.07	7.21	7.19
Never married	Count	202	667	520	3939
	%	10.16	15.28	20.48	22.32
All	Count	1988	4364	2539	17,650