The impact of work and family responsibilities on healthy sleep habits

Abstract

Healthy sleep habits include sufficient sleep, regular bedtimes and established sleep routines. The responsibilities of paid and unpaid work that arise during the daytime are assumed to haunt us at night as well, eventually affecting these sleep habits. A long-term comparison of sleep duration from 1966 and 1999 time-diary data shows that sleep duration has not declined to the large extent that is generally assumed. Moreover, analyses of timing of sleep and sleep routines using time-diary data from 1999 and 2004 also do not show much evidence of this assumed decline. On the contrary, increasing work and family responsibilities positively affect regular bedtimes and sleep routines.

Keywords

Sleep habits sleep routine sleep time time use work-family responsibilities

1. From first to fourth shift

We live with an ongoing struggle today between the responsibilities that come from the various roles we play within different social contexts. Time pressure increases as we combine paid work and household and childcare responsibilities and the count of our daily shifts based on these responsibilities has climbed to four. During the day, the first shift of paid work seamlessly passes into the second shift of domestic work and childcare while thwarted by the third shift of ongoing emotional involvement with family responsibilities (Hochschild, 1990, 1997). And at night it is not over as we are struck by a fourth shift that disrupts our sleep because the sentient activity of the third shift continues while we are asleep (Venn et al., 2008). Sleep has become embedded in our social realities, reflecting and being structured by the social roles we play (Burgard and Ailshire, 2009; Chatzitheochari and Arber, 2009; Hislop and Arber, 2003b).

It seems that, although sleep is naturally prompted (‘chief nourisher in life’s feast’ – Macbeth), it is increasingly becoming a social phenomenon that needs to be studied as such, especially since sleep is susceptible to changes in other social domains such as paid work and leisure. In contemporary society, ‘being busy’ has become the new standard, both in paid work (Gershuny, 2005) and in leisure (Glorieux et al., 2010; Sullivan, 2007), to symbolize a full and valued life (Darier, 1998). We want it all: having a successful career, enjoying qualitative and daring leisure time, and, last but not least, being a beloved mother or father. However, when all this is squeezed into a day that will never get any longer than 24 hours, something has to make way. And so – or at least it is generally assumed – in our quest for more time, it is sleep that will be sacrificed to get more things done (Basner et al., 2007; Biddle and Hamermesh, 1990). In the past, we might have been a little sceptical about this idea since the night really serves as a frontier between diurnal and nocturnal activity and, as is the case with borders, the culture of time-use on the other side of this frontier is different from what we are used to (Melbin, 1978).

However, the assumption that we will sacrifice sleep for daily activities is based on the idea that sleep has been used as a social status marker for a long time. Taylor notes that ‘The way in which sleep is commonsensically conceived will depend on an individual’s social location and their economic function’ (1993: 468). He puts forward that those who see sleep as a leisurely activity have attained a particular socio-economic status compared to a working class that are much more likely to perceive sleep as respite from work. Also, during the Victorian period, sleep was regarded as indulgence. Sleeping beyond 7–8 hours was frowned upon as a sign of idleness. Nowadays, this social normalization of sleep continues: being able to work long hours (and run a household and raise children) without much sleep is integrally part of a successful career from which a higher social position in society is derived (cf. Gershuny, 2005).

With sleep being normatively valued and strongly related to paid and domestic work there is no doubt that there is inequality in sleep time. The work of Hislop and Arber showed that mid-life women are seriously disadvantaged in their sleep, since these women ‘do sleep in the way they do their activities’ (2003b: 709, italics in original). Their sleep is deeply embedded in their social realities, structured by a multiplicity of their social roles and responsibilities (Hislop and Arber, 2003a, 2003b, 2003c). We find this same ‘fourth shift’ for couples with young children in the study of Venn and colleagues (Venn et al., 2008). Mothers were more likely than their partners to find their sleep disrupted because they ‘respond to their children’s physical needs at night… [which] resulted in them undertaking sentient activity at night’ (2008: 96). The fourth shift of fathers will disrupt their sleep as a result of anxiety about their older, teenage children. Maume, Sebastian and Bardo (2009) come to the same conclusions: they find gender inequality in waking role obligations and prove – for a non-population based sample – that greater caregiving responsibilities result in more sleep disruptions for women and ‘pro-feminist’ men compared to men in general. Women may take on more night shift caregiving even if they have the same roles and responsibilities as men (Burgard, 2011).

Although the studies mentioned above mostly related domestic chores to a gendered sleep inequality, the work of Chatzitheochari and Arber (2009) related the number of hours asleep to paid work and found the opposite gendered conclusion. Working men are much more likely to report sleeping fewer than 6.5 hours per day compared to their female colleagues. Burgard and Ailshire (2009) relate work stressors and home stressors to lower sleep quality and conclude that negative experiences at home are not more strongly related to sleep quality than negative experiences at work. In other words, they seem to prove that first and second shift responsibilities contribute to lower sleep quality.

In summary, it seems that the perception of sleep as a social and normative concept that is subjected to work and family responsibilities translates itself into two major concerns about sleep. Firstly, the quantity of sleep is assumed to decline as a result of the increasing time-consuming activities of the first and second shift, that is, working and running a household. Secondly, and related to the first concern, the quality of sleep declines because responsibilities from the first two shifts not only burden our daily activities (cf. the third shift) but also disrupt our sleep time (cf. the fourth shift) and eventually will affect our health (Patel et al., 2006; Steptoe et al., 2006).

When studying the impact of work and family responsibilities on sleep time in terms of the four shifts identified herein, we need to make a conceptual distinction between the content of these four shifts. The first two shifts can be fairly easily quantified in terms of the time spent on paid work activities (first shift) and household and childcare activities (second shift) (as is partially done in the work of Chatzitheochari and Arber (2009)). The last two shifts capture a perception, a feeling, about constantly worrying or thinking about whether one is capable of meeting the responsibilities that are quantified in the first and second shift (as is done in the work of Burgard and Ailshire, 2009). The problem here is that reliable, quantitative information on time spent on paid work, household work and sleep time, on the one hand, and qualitative information about responsibilities and sleep on the other hand, are hardly ever found together in large-scale social surveys (Chatzitheochari and Arber, 2009).

In our study we will use quantitative information from time-diary data to study the impact of work and family responsibilities on sleep time, since time-diary data are assumed to give the best and most objective proxies of daily behaviour (Robinson and Godbey, 1997). Time-diary data are only sparsely available in sleep research (e.g. Basner et al., 2007; Burgard, 2011; Chatzitheochari and Arber, 2009) but are not a priori designed to study sleep behaviour and thus might still underestimate, for example, sleep interruptions or the difference between actually sleeping and lying awake in bed (Burgard, 2011). Nevertheless, time-diary data do have some important advantages over other survey data. Since time-diaries always have to add up to 24 hours per day and are recorded ‘on-the-moment’, they are much less prone to over- and underestimations. Moreover, time-diary data not only provide information on the duration of activities, but also on the timing or the moment of the day when activities are undertaken and the tempo or the recurrence of activities during the day or week (Zerubavel, 1982). We will argue that this additional information coming from time-diary data will provide good approximations of healthy sleep habits: sleeping sufficiently (i.e. duration of sleep), going to bed at fixed times (i.e. timing of sleep) and covering the same period of sleep every night (i.e. tempo of sleep).

In our study, we will analyse the impact of work and family responsibilities on these three elements of healthy sleep habits. Firstly, in section 2 we evaluate long-term changes in sleep duration to see whether those who are assumed to suffer from first up to fourth shift responsibilities sleep less than their ‘peers’ did decades ago. Secondly, in section 3 we will analyse the regular timing of sleep and sleep routines and investigate for both measures the effects of both socio-demographic characteristics and household composition and the time spent on paid work and domestic work and childcare. We will summarize and conclude in section 4. Each section will commence by introducing the data and method used to perform the required analyses. Technical reports on the data used are available on request.

2. Long-term changes in sleep duration

As already mentioned, most large-scale social surveys do not contain much information about sleeping habits, but time-diary surveys have been shown to be a useful source to link sleep to other non-sleep activities (e.g. in the USA: Basner et al., 2007; Burgard, 2011) or social determinants (e.g. in the UK: Chatzitheochari and Arber, 2009). Time-use surveys are assumed to give the best representation of daily behaviour because respondents continuously report their actions by filling in time-use diaries (including the location and presence of others). Time-use surveys using time-use diaries provide objective measures of the amount of time people spend at different activities; they are considered more reliable than conventional surveys (Robinson and Godbey, 1997; Robinson et al., 2011). These diaries are mostly combined with household and individual questionnaires, which provide an insight into socio-demographic characteristics, household composition, and, in some more extensive questionnaires, norms, values and attitudes towards different time-use activities.

The first of the abovementioned analyses is a long-term comparison of sleep duration to see whether changes in sleep time have occurred over the years and, if so, whether there has been an increase or reduction in sleep time.

2.1. Time-diary data from 1966 and 1999

To make a long-term comparison of sleep time, we use the Belgian data from the Multinational Comparative Time-Budget Research Project of 1966 (mtus66) held in 12 countries (Szalai, 1972) and a derivative of the Belgian time-use survey of 1999 (btus99) (indicated with an asterisk). Although the data from 1999 are somewhat outdated, we will still use them because they are the only data that has been made comparable to the data of 1966 due to different, extensive, weighting and selection procedures. In the mtus66, for a randomly assigned day, primary and secondary activities and location and interaction partners were recorded by 2077 19- to 65-year-old Belgians in a continuous self-completion diary (Javeau, 1970).

The National Institute for Statistics conducted the Belgian Time-Use Survey of 1999 according to Eurostat guidelines. In this survey, respondents of 12 years and older completed a self-completion diary with fixed 10-minute time slots for one randomly designated weekday and one randomly designated weekend day. The btus99 contains information on 8382 Belgians from 4275 households. To make long-term comparison possible, the btus99* is derived from the btus99 in such a way that it is as equally representative as the mtus66 data and contains information on 3279 respondents from an equal number of households.

We selected the active population (i.e. students are excluded) who were between the ages of 19 to 64 and not living with their parent(s).

Table 1 shows the results of the long-term comparison of sleep time for different social categories. The figures are in hours per weekday, per weekend day and per week on an aggregated level and have been tested for significant differences between both years of the survey. All adjusted figures are controlled for all other categories in the table.

Table 1.

Long-term comparison of time spent sleeping by socio-demographic characteristics (1966–1999) in hours (mca-anova).

	Weekday						Weekend day						Synthetic week
	mtus66 (n = 1425)		btus99* (n = 3205)				mtus66 (n = 1425)		btus99* (n = 3205)
	%	Adj.	%	Adj.	Diff.^a	Sig.	%	Adj.	%	Adj.	Diff.	Sig.	Diff.	Sig.
General controlled mean
100.0	8.2	100.0	7.9	−0.27	***	100.0	8.7	100.0	8.9	0.21	*	−0.85	**
Gender
Male	50.2	8.1	47.7	7.7	−0.38	***	48.6	8.8	47.9	8.9	0.12	n.s.	−1.41	**
Female	49.8	8.3	52.3	8.1	−0.18	*	51.4	8.7	52.1	9.0	0.26	°	−0.45	n.s.
100.0	100.0	100.0	100.0
Age
19–24 years	11.7	8.5	7.5	8.4	−0.02	n.s.	8.3	8.8	7.4	9.1	0.34	n.s.	0.85	n.s.
25–39 years	8.0	8.1	13.3	7.9	−0.20	n.s.	7.1	8.9	13.2	9.1	0.24	n.s.	0.01	n.s.
40–49 years	23.9	8.1	32.0	7.8	−0.23	*	24.0	8.6	31.8	8.9	0.36	*	−0.59	n.s.
50–55 years	26.9	8.1	20.4	7.6	−0.45	***	28.9	8.7	20.6	9.0	0.28	n.s.	−1.76	**
55 + years	29.5	8.4	26.8	8.1	−0.34	***	31.7	8.9	27.0	8.7	−0.16	n.s.	−1.70	**
100.0	100.0	100.0	100.0
Highest level of education
Low	38.9	8.2	26.3	8.2	−0.02	n.s.	45.6	8.7	26.3	8.9	0.24	n.s.	0.45	n.s.
Medium	49.7	8.3	29.1	7.9	−0.37	***	39.6	8.8	29.1	9.0	0.25	n.s.	−1.27	**
High	11.4	8.2	44.6	7.8	−0.48	***	14.8	8.8	44.6	8.9	0.10	n.s.	−2.20	**
100.0	100.0	100.0	100.0
Number of children
No children	43.5	8.3	64.6	8.0	−0.36	***	48.1	8.8	64.7	8.8	0.03	n.s.	−1.57	***
1 child	23.5	8.0	15.7	7.8	−0.31	*	24.9	8.9	15.7	9.3	0.35	°	−0.62	n.s.
2–3 children	25.7	8.0	17.8	7.8	−0.20	°	21.0	8.5	17.7	9.1	0.59	**	−0.02	n.s.
4 or more children	7.3	8.4	1.9	8.3	−0.11	n.s.	6.0	8.6	1.9	8.6	0.06	n.s.	−1.08	n.s.
100.0	100.0	100.0	100.0
Hours of work per week
Unemployed	38.0	8.5	38.3	8.3	−0.25	**	38.5	8.8	38.2	8.9	0.01	n.s.	−0.93	°
Part‐time (1–25 hrs)	3.1	8.1	10.2	8.1	−0.02	n.s.	3.0	9.2	10.2	9.0	−0.23	n.s.	0.25	n.s.
Full-time (26–38 hrs)	6.2	8.2	29.3	7.7	−0.43	*	7.5	8.7	29.3	9.0	0.30	n.s.	−1.33	n.s.
Full-time (38–50 hrs)	40.0	7.9	17.3	7.6	−0.38	***	38.8	8.8	17.4	9.1	0.25	n.s.	−1.26	*
More than 50 hrs	12.6	8.0	4.8	7.5	−0.47	*	12.2	7.7	4.9	9.0	1.23	***	−1.87	n.s.
100.0	100.0	100.0	100.0
Weekend day
Saturday	53.7	8.2	50.0	8.4	0.20	n.s.
Sunday	46.3	9.3	50.0	9.5	0.25	°
100.0	100.0

Note ^aDifferences between mtus66 and btus99* are significant at (***) p ≤ .001; (**) p ≤ .010; (*) p ≤ .050; (°) p ≤.100; n.s. not significant.

2.2. Long-term comparison

From Table 1 we can see that over the last 3.5 decades we lost a significant 0.27 hours (16 minutes) of sleep per weekday, which represents an average decline of 25 seconds of sleep per year. At first sight, we might conclude that this decrease in sleep duration between 1966 and 1999 is relatively low, especially when we take into consideration that on the weekends we gain 0.21 hours of sleep per weekend day. With a few exceptions, this seems to be the general pattern, even if we take into account different social categories – we lose sleep during the week but regain part of that lost sleep during the weekend.

Women sleep ‘considerably’ less than they did 33 years earlier but the difference is still only 10 minutes per weekday and they gain 15 minutes of sleep per weekend day. Men, on the other hand, have lost 23 minutes of sleep per weekday and have regained only 7 minutes on weekend days (although the difference is not significant). Young adults have not significantly given up sleep, but, surprisingly, the people aged 50 to 64 years old have and they even do not sleep significantly longer during the weekend. This is an interesting finding since people tended to resign from the labour market earlier in 1999 than they did 33 years earlier and because their waking time is less tied to working hours, they have time to sleep longer.

No significant differences in time spent sleeping are found for the lower educated people. However, people with higher levels of education in 1999 sleep significantly less per weekday compared to 1966. The highest educated people in particular have the largest decline in sleep time (almost half an hour) and none of them significantly makes up for that loss during the weekend. This could be partly due to the great difference in the proportion of people who are highly educated in 1999 (44.6%) compared to 1966 (11.4%).

Even though the presence of children is less common in 1999 than it was 33 years earlier, people with no children sleep significantly less in 1999 than they did in 1966. The statistics show the presence of children made similar demands on people’s time in 1966 and 1999; there are hardly any significant changes in the time spent sleeping over the 33 years, although people with children sleep significantly longer during the weekends.

From Table 1 we derive another interesting finding. In 1966, people either did not work (38%) or they worked at least 38 hours per week (52.6%). Part-time working solutions were relatively rare at that time. In 1999, we find many more people being employed part-time (10.2%) or having maximum working weeks of 38 hours (29.3%). Working more than 38 hours per week is very uncommon nowadays and working more than 50 hours is quite exceptional compared to 1966. However, despite these changes in working weeks, the unemployed and the full-time employed sleep less than they did 33 years earlier, but these differences vary only between 15 and 30 minutes. Moreover, those with the longest workweeks catch up with this loss of sleep during the weekend.

We can conclude that, yes, we do sleep less than in 1966, but we should view these results with some caution. Firstly, they are estimates of weekdays and one major difference between 1966 and 1999 is the length of the working week. Four decades ago, the 6-day working week was the standard and so people tended to go to bed much earlier on Friday nights because they had to work on Saturday. Nowadays, with the 5-day working week being the standard, the weekend starts on Friday night, meaning that people go to bed at later hours because they can sleep late on Saturday. As a consequence of this, Monday to Friday sleep durations recorded are less in 1999 than they were in 1966.

Secondly, it is often hypothesized that we make up for this diminished sleep during the weekends. As we can see in the second part of Table 1, we might conclude that this is true for the general (controlled) mean. Furthermore, with some exceptions, we find that sleep during the weekend has actually increased between 1966 and 1999. Indeed, those people that are assumed to have lost some of their sleep time to paid and unpaid work responsibilities (i.e. women and those who have children or are employed full-time) did so during the working week, but partially compensated for this loss during the weekend.

However, since we cannot compare the difference in sleep on an average weekday with the difference in sleep on an average weekend day, we computed a synthetic week by multiplying sleep recordings on weekdays by five and on weekend days by two. Next, we redid the analysis. The last two columns of Table 1 show the differences and their significances between 1966 and 1999, but now in hours per week. This shows that we lost 0.85 hours (51 minutes) of sleep over the whole week between 1966 and 1999. Surprisingly, those social categories that were believed to have lost sleep over that time actually did not. Women do not sleep significantly less in 1999 than they did in 1966, nor do people with children, people in the age group of 19 to 49 years or people that are employed for a minimum of one and a maximum of 38 hours per week. On the other hand, men, higher educated people, people of 50 years and older and those employed for 38 to 50 hours a week sleep between 1.3 and 2.2 fewer hours per week.

Thus, we might conclude that it is the first shift of paid work responsibilities, especially among men and higher educated people, that is to blame for a decline in sleep duration over the past decades. This supports the findings of Chatzitheochari and Arber (2009) and concurs with other research on the impact on (social) time-use of the inflexibility of paid work as a contracted responsibility that cannot be postponed or neglected (for example, research on the time partners spent together (Glorieux et al., 2010)). So some of us have given up some sleep time over the 33-year period, but not those who we expected to have lost sleep and certainly not in the vast amounts we presumed. Nor do we see any evidence of diminished sleep duration serving as a contemporary social status marker.

3. Healthy sleep habits

The idea of a growing threat of paid and domestic work, combined with less sleep, as a social status marker that restrains the time spent sleeping has been partially rejected in the previous section. Nonetheless, we still face the impact of continuous emotional involvement resulting from these responsibilities. The impact of the fourth shift, in particular, will not be seen explicitly in the average duration of sleep but rather in the quality of sleep.

Quality of sleep is a controversial concept because it is highly subjective and difficult to measure, let alone to relate it to the influence of the fourth shift (which is both subjective and probably partly subconscious). Someone might, for example, think they have quality of sleep if they reach a certain duration of uninterrupted sleep, whereas another person might think they have quality of sleep if they – despite the presumed omnipresent fourth shift – are able to go from superficial sleep to the state of slow-wave sleep, which is the sleep phase at which mental and physical recuperation takes place. Since this qualitative information is not available in our data we turn to the idea that healthy sleep habits relate to a good quality of sleep. In medical research, we find that regular timing of sleep (i.e. going to sleep at a fixed time every day), on the one hand, and sleeping behaviour routines, on the other, are considered to be important parameters for the generic concept of healthy sleep habits (see e.g. Monk et al., 1991, 1994) that enables people to recover from and cope with shifting work schedules (Horrocks and Pounder, 2006). A discussion of how healthy sleep habits relate to the quality of sleep lies beyond the scope of this paper. Our focus is on revealing the effects of work and family responsibilities on the regular timing and routines of sleep that make good sleep habits.

3.1. Time-use data from week diaries

To study the rigidity of timing and the routine of sleep we need time-use information of multiple, consecutive days. For this purpose, the Flemish time-use surveys of 1999 and 2004, conducted by the Research Group TOR of the Vrije Universiteit Brussel, provide useful information (TOR99 and TOR04). (Flanders is the Dutch speaking part of Belgium; it has over 6 million inhabitants, accounts for 60% of the total population of Belgium, and is representative of modern North-European countries.) The tor99 surveys 1474 respondents between 18 and 75 years of age and the tor04 covers 1780 respondents in the same age range. In both surveys, respondents completed their time-use dairies over seven consecutive days and also filled in a questionnaire. The respondents recorded in the diaries their moment-to-moment time use in sequences by noting the exact starting and ending time, the type of activity and the context of the activities (e.g. motivation, transportation and social contact). The individual questionnaire inquired about socio-demographic characteristics and items that measure, for example, eating habits, political preferences or didactic values. We pooled both samples (tor9904), as statistical tests do not indicate substantial changes in time use between the two datasets spanning five years. Our analyses are restricted to the active population (i.e. students are excluded) between 18 and 64 years of age who are not living with their parent(s). The selected and pooled sample consists of 2264 individuals.

3.2. First measure healthy sleep habits: regular timing

Equal or regular timing of sleep, that is, going to sleep at more or less the same time every day, is positively related to healthy sleep habits (cf. Monk et al., 1991, 1994). To create an indicator of regular timing of sleep we measured the variation in the timing of sleep by computing the differences between the times respondents went to sleep on the five nights before the five workdays. These nights include Sunday, since it is the night before the working week starts, and exclude Friday, because it is the night before the weekend starts.

We have transformed the moment-to-moment time-use recordings of the tor9904 data to a 10-minute interval registration by keeping the activity performed on the 10th minute of each interval. We start our day at 10 a.m. instead of at midnight for an obvious reason. In each 24 hours, we need to cover only one night. If our days run from midnight to midnight and a respondent goes to sleep at 1 a.m. on the first night and at 11 p.m. on the second night, both his moments of going to sleep will take place on the same day.

In order to include respondents who sleep during the daytime (e.g. because they work night shifts) and in order to exclude midday naps, we selected the beginning of the longest registered sleeping period for each day. Apart from restricting our selection to those in the age range of 18 to 64 years, non-students and those not living with parents, we also only selected respondents for whom all five periods of sleep are at least three hours long. This reduces the sample to 2199 respondents.

We then compared all moments of going to sleep with each other, which gives us 10 comparisons (Sunday–Monday, Sunday–Tuesday, Sunday–Wednesday, Sunday–Thursday, Monday–Tuesday, Monday–Wednesday, et cetera until Wednesday–Thursday). The differences between the starting times of sleeping are calculated in absolute numbers and then divided by 10 (i.e. the number of comparisons). This gives us the average variation in the timing of sleep.

Table 2 gives an illustration of this calculation. Someone’s bedtimes for all five days (Sunday to Thursday) are given in the main diagonal of the table. The cells to the right of the diagonal show the absolute differences resulting from the comparisons of the bedtimes for all days and the sum of these figures divided by 10 gives the average timing of sleep for this person, which is 43 minutes. Since we transformed our time-use data into 10-minute intervals, we measure the average variation in the timing of sleep in numbers of 10-minute intervals. Consequently, the average timing of sleep for the example in Table 2 is 4.3 10-minute intervals.¹ (Note that the variation of the timing of sleep is not the same as the interval within which someone’s moments of sleep fall; it is the difference between the earliest and latest bedtime, which, in the example in Table 2, is 90 minutes.)

Table 2.

Illustration of the measure of variation of timing of sleep.

Sunday night	Monday night	Tuesday night	Wednesday night	Thursday night
Sunday night	9 pm	10 min.	30 min.	70 min.	20 min.
Monday night	9:10 pm	20 min.	60 min.	30 min.
Tuesday night	9:30 pm	40 min.	50 min.
Wednesday night	10:10 pm	90 min.
Thursday night	8:40 pm
Average variation in timing of sleep	43 min. (4.3 10-minute time-intervals)
Interval of timing of sleep	90 min. (8:40 pm – 10:10 pm)

With our measure constructed, we are now able to analyse the impact of changes in individual and family characteristics, as well as changes in the time spent on paid work, domestic work and child care, on the variation of going to sleep. In order to do so, we use a multiple regression technique (see Table 3). Model 1 investigates two individual characteristics that are related to the unequal effects of first to fourth shift responsibilities (i.e. gender and age). Model 2 adds family characteristics that relate to the emergence of a third and fourth shift (i.e. family situation and presence of children) and model 3 adds the time spent on first and second shift activities.

Table 3.

Multiple regression analysis of average ‘timing of sleep’ in 10-minute intervals per working week (Sunday night to Thursday night).

	TOR9904 (n = 2199)
	Model 1			Model 2			Model 3
	B	Beta	Sig.	B	Beta	Sig.	B	Beta	Sig.
Constant	5.832	.252	***	7.879	.448	***	8.413	.469	***
Individual characteristics
Gender (ref. is men)
Women	−1.208	.250	***	−1.217	.250	***	−1.510	.279	***
Age (ref. is 25–39 years)
18–24 years	1.934	.746	**	1.461	.752	n.s.	1.478	.751	*
40–54 years	−.618	.286	*	−.837	.330	*	−.862	.335	**
55–64 years	−1.948	.349	***	−2.432	.400	***	−2.655	.429	***
Family characteristics
Family situation (ref. is living alone)
With partner without children	−1.786	.446	***	−1.777	.445	***
Single parent family	−.157	.952	n.s.	−.149	.950	n.s.
Two parent family	−2.410	.798	**	−2.341	.797	**
Age of youngest child (ref. is no children)
Younger than 7 years	−.322	.798	n.s.	−.278	.825	n.s.
Between 7 and 25 years	.245	.744	n.s.	.215	.743	n.s.
First and second shift activities
Paid work (ref. is 4–8 hrs a day)
None or fewer than 4 hrs a day^a	−.082	.330	n.s.
More than 8 hrs a day	−1.446	.345	***
Domestic work (ref. is >4 hrs a day)
More than 4 hrs a day	.014	.347	n.s.
Childcare (ref. is >30 min a day)
More than 30 min a day	−.145	365	n.s.
Adjusted R Square	.027	.045	.051

Note ^aThis category includes the unemployed and early retired. (***) p≤.001; (**) p≤.010; (*) p≤.050; n.s. not significant

Results

From the constant value in model 1 in Table 3 we derive that men in the age range of 25 to 39 years have an average variation in their timing of sleep of almost 1 hour, or six 10-minute intervals. Women have a significantly less varied timing of sleep than men and this difference remains significant after controlling for family characteristics in model 2 and the time spent on paid and unpaid work in model 3. Also, the effects of age remain persistent over all three models. Younger people have more variation and older people have significantly less variation in their timing of sleep. When we look at the standardized coefficients (i.e. the betas), we see that the age effects are among the largest effects in the models; growing older means having a more regular timing of sleep.

Based on the effects of family situations in model 2, we might conclude that the more people there are present in a household or family, the more activities have to be synchronized or negotiated, and the less variation occurs in the timing of sleep. Going from living alone without children to living with a partner (still without children) decreases the variation in the timing of sleep by 1.8 10-minute intervals and this almost doubles when going from living alone to living with a partner with children. Together with age, these effects are the largest in this regression model. The absence of an effect for single parent families is probably due to the small number of this type of family (only 5% of the sample). The age of the children, if present, does not influence the timing of sleep.

Finally, in model 3 we added the time spent on paid and unpaid work, but again we find almost no indication of the impact of paid and domestic work on the regular timing of sleep. The only significant effect we find is for people working more than 8 hours a day as compared to people working 4 to 8 hours a day. However, the effect runs in the opposite direction; after being controlled for all other variables in the model, working long hours a day still decreases the variation in the timing of sleep only by 15 minutes.

We use regular timing of sleep as an indicator of healthy sleep habits and we almost did not find any instances of individual characteristics, changes in family situation and work and household responsibilities negatively affecting the timing of sleep; in fact, having to synchronize daily behaviour with a partner, with children and with paid working hours led to a more regular timing of sleep, even when controlled for age and gender.

Besides going to sleep at regular moments during the week, we also argued that a nightly sleep routine, that is, sleeping during the same moments of time every night, is also part of healthy sleep habits. This routine of sleep is our second indicator: apart from going to sleep at the same time, we question whether people cover the same period of the night with sleep on every day.

3.3. Second measure of sleep quality: routine

Again we use the tor9904 because this allows us to analyse a single person’s time-use activities for multiple days, in this case, for all five weekdays (Monday to Friday); it allows us to analyse the time spent sleeping on each weekday and mutually compare these days to derive the percentage of sleep that occurs every weekday at the same time. We will call this measure ‘sleep routine’. For example, suppose someone sleeps every weekday from 11 p.m. until 6 a.m. (indicated with an ‘ × ’ in Table 4), obviously, he or she will have a sleep routine of 100 per cent (indicated by the grey shading in Table 4). Now suppose someone sleeps on Monday and Tuesday from 11 p.m. until 6 a.m. and on Wednesday, Thursday and Friday from 12 p.m. until 6 a.m., with Wednesday and Friday night’s sleep interrupted for one hour. This person’s total sleep duration is 35 hours, of which 25 hours are routine sleep. This gives a sleep routine of 71.4 per cent for this person.

Table 4.

Examples of sleep routine.

	Person 1					Person 2
	Mon	Tue	Wed	Thu	Fri	Mon	Tue	Wed	Thu	Fri
11 pm	x	x	x	x	x	x	x
12 pm	x	x	x	x	x	x	x	x	x	x
1 am	x	x	x	x	x	x	x	x	x	x
2 am	x	x	x	x	x	x	x	x	x
3 am	x	x	x	x	x	x	x	x	x	x
4 am	x	x	x	x	x	x	x	x	x
5 am	x	x	x	x	x	x	x	x	x	x
6 am	x	x	x	x	x	x	x	x	x	x
Total sleep duration			40 hrs	Total sleep duration			35 hrs
Total sleep routine duration			40 hrs	Total sleep routine duration			25 hrs
Percentage sleep routine			100.0	Percentage sleep routine			71.4

The example given in Table 4 is just a hypothetical and rough estimation of sleep routine because the time intervals span one hour. More detailed and reliable estimates can be derived when these time intervals are narrowed. For our proposed measure of routine of sleep we again transformed the recorded moment-to-moment time use of all respondents in the tor9904 data into a time-use registration of 10-minute intervals. For each day, we kept the activity performed on the 10th minute of each interval and we started the first interval at midnight. In this way, we can compare every 10-minute interval over all weekdays. Basically, this means that we can divide all five weekdays into the 144 10-minute intervals that make up the 24 hours available each day. To measure sleep routine, we then compared each 10-minute interval on day one with the homologous 10-minute interval on the other days of the week. The comparison is based on a ‘count’ of equal time-use activities on the same 10-minute intervals on every day, and, for the purpose of our study, we only selected sleep as the time-use activity of interest.

Results

Table 5 shows the distribution of routine of sleep for different social categories and for the time spent on paid work and on domestic work and childcare. The figures are shown as a percentage of routine of sleep for all five weekdays, on an aggregated level. All adjusted figures are controlled for all other categories in the table.

Table 5.

MCA-ANOVA of ‘routine of sleep’ in percentage per working week (Monday to Friday).

Covariates	TOR9904 (n = 2199)
Covariates	B (Unadj.)	Sig.^a	B (Adj.)	Sig.^b
Paid work [hrs/weekday]	.098	n.s.	.629	°°°
Domestic work [hrs/weekday]	.660	***	1.369	°°°
Childcare [hrs/weekday]	−.130	n.s.	−.293	n.s.
Factors	%	Unadj.	Sig.^a	Adj.	Sig.^b
Gender	***	°°°
Male	49.7	75.7	75.8
Female	50.3	80.2	80.2
100.0
Age	***	°°°
18–24 years	3.5	70.5	72.3
25–39 years	34.8	76.7	76.7
40–54 years	41.8	77.9	77.5
55–64 years	19.9	81.7	82.4
100.0
Highest level of education	*	°
Low	38.2	78.9	78.3
Medium	34.3	76.1	76.7
High	27.5	79.0	79.3
100.0
Occupational status	***	°°°
Part-time employed	15.9	80.8	79.6
Full-time employed	55.8	76.1	74.2
Unemployed	28.3	80.1	84.6
100.0
Family situation	*	°°
Living alone, no children	10.9	76.5	76.2
Living with partner, no children	26.6	77.3	76.4
Single parent family	5.1	75.7	75.4
Two parent family	57.3	78.8	79.3
100.0
R Squared^c	.068

Note ^aUnadjusted effects are significant for (***) p ≤ .001; (**) p ≤ .010; (*) p ≤ .050; n.s. not significant; ^bAdjusted effects are significant for (°°°) p ≤ .001; (°°) p ≤ .010; (°) p ≤ .050; n.s. not significant; ^cAdjusted for factors and covariates.

On average, about 78 per cent (not in Table 5) of all the time we spend sleeping is done to a routine. We recall that routine of sleep is the percentage of the time spent sleeping that takes place at the same period of time every weekday, so this 78 per cent means that of all the time we spend sleeping, over three quarters of this sleep time occurs at the exact same period of every night from Monday to Friday.

Now let us firstly examine the second half of Table 5 containing the socio-demographic factors that might influence the percentage of sleep routine. When looking at the sleep routine of men and women, we immediately identify a striking finding: women have a much higher sleeping habit than men. Even after controlling for all other variables in the table, women have a routine of sleep of around 80 per cent, whereas for men it is only 76 per cent.

When we focus on age, we find a (logical) increase in sleep routine when people get older. The young adolescents – probably still without many domestic or childcare responsibilities – have the lowest level of sleep routine (note here that students are excluded in this analysis). For the level of education, the middle group has the lowest level of sleep routine, whereas the highest educated group has the largest percentage of sleep routine when controlled for all other variables. Full-time employed people have a sleep routine below average and the unadjusted figures suggest that part-time employed and unemployed people have equal amounts of sleep routine, but this does not hold once controlled for the other variables. Unemployed people have the highest overall sleep routine of almost 85 per cent, whereas the part-time employed have almost 80 per cent. Finally, adjusting for the effects of the family situation shows that single parent families have the second-lowest sleep routine (75.4%).

Secondly, we return to the effects of the covariates in the first half of Table 5 and here we find some striking results: once adjusted, paid work and domestic work increase the percentage of sleep routine. Spending one hour per weekday more on domestic work will increase sleep routine by 1.4 per cent. Moreover, the (expected) negative influence of childcare on sleep routine turns out to be insignificant. When searching for an explanation for these, at first sight, contradictory findings, we might argue that, although full-time employed and single parent families have the lowest percentage of routine sleep, both paid work and domestic work contribute at some level to a routine daily behaviour. Both tasks and responsibilities cannot be postponed: these people have to go to work every workday and do certain household chores (such as cooking dinner or, at a certain point, clean the home, wash the dishes and do laundry) (e.g. Baxter, 2000; Hochschild, 1990; Noonan, 2001).

As we found with the sleep duration and the timing of sleep segments, again, we do not find sustainable evidence that work and family responsibilities affect routine of sleep. On the contrary, the least sleep routine is found in the youngest age group that is assumed to be still relatively free from these responsibilities. Single parents have the second-lowest routine of sleep, but they still cover more than three-quarters of the same moments at night with sleep. In fact, based on the results in Table 5, we might conclude that the effect runs in the opposite direction. Indeed, full-time employment and parenthood have arrears with regard to routine sleep due to paid and domestic work obligations; however, it is in fact these responsibilities that bring back the routine to their daily behaviour. As we have seen, an increase in paid work and domestic work increases the percentage of routine sleep. Departing from the idea that routine sleep and regular sleep timing is good for mental and physical recuperation, we suggested that measures of both sleep routine and regular timing of sleep can be used as indicators of healthy sleep habits. When using the most objective social survey data available (in the form of time-diary data), we find little evidence that suggests an impact of a first and second shift on the rigidity of timing and routine of people’s rest at night.

4. Concluding remarks

Paid work (first shift), domestic work and childcare (second shift), emotional involvement with responsibilities derived from these two shifts during the day (third shift) and during the night (fourth shift) are assumed to have our daily lives in a strong grip. The growing notion that sleep needs to be studied not only as a pure biological activity but also as a social phenomenon has led to an increase in research relating these four shifts to the quantity and quality of sleep. However, since most population surveys are not based on sleep research, information on sleeping habits often is not present in enough detail. A partial solution came from the use of time-diary data. The advantage of time-diary data over other social surveys is threefold: they provide an objective measurement of people’s daily time use, they are large-scale surveys and they combine time-use registration with an individual and a household questionnaire. This makes it possible to analyse sleeping habits to a certain objective degree while taking into account different individual characteristics, different family or household settings and time spent on different waking activities. However, these data are not purely aimed at sleep research and they lack sufficient qualitative information on sleep. Nevertheless, time-diary data provide additional information compared to standard surveys in a way that enables them to capture the moment or timing and the recurrence of sleep activity.

It is this strength of time-diary data that we have exploited in this study to analyse the impact of work and family responsibilities on sleep. We measure the first and second shift by the amount of time spent on paid work, domestic work and childcare. Since the third and fourth shifts are the diurnal and nocturnal emotional involvements in the first and second shifts and no such qualitative indicators were present in our data, we replaced these shifts in our analyses with socio-demographic characteristics as found in the existing literature that predict inequality in the effects of these shifts on sleep (gender, presence of children, family situation and occupational status). We hypothesized that the first and second shifts and certain categories (women, presence of young children, single parent families and dual-earner families) will negatively impact healthy sleep habits; these habits are sleeping long enough (i.e. sleep duration), regular bedtimes (i.e. timing of sleep) and routine sleep (i.e. covering the same moments of the night with sleep). This occurs partially because these shifts occupy us more and more and since there are only 24 hours a day available to us, some other activities have to make way; the other reason is that leading a busy life without needing much sleep becomes more and more a sign of living a valued life.

We used different sorts of time-diary data in our research to, firstly, study the quantity of sleep by analysing the long-term evolution in sleep duration. To account for the influences of the responsibilities of the four shifts occupying our minds during the day and night, we controlled for social and family background (e.g. gender, age, education, occupational status, family composition) in the analyses and the time spent on paid and unpaid work. Surprisingly, we hardly found any major decline in sleep duration and those that we did find were among working men and not among those social categories (i.e. women, mothers and single parents) that are assumed to experience the greatest impact of work and family responsibilities on their sleep time. In other words, sleeping long enough as a key element of healthy sleep habits seems not to be violated by work and family responsibilities.

Secondly, we analysed the regular timing of sleep (i.e. going to sleep at more or less the same time every weekday) and the ‘routine of sleep’ (i.e. the percentage of sleep enjoyed at the same time every weekday) as two other key elements of healthy sleep habits. Both indicators are constructed based on comparisons of 10-minute time intervals of sleep time over all five weekdays, making them reliable proxies. We investigated the impact of different individual and social and family background characteristics as well as the impact of time spent on paid work, domestic work and childcare. Surprisingly, once again we barely found any results that, as often assumed, could be assigned to responsibilities arising from paid and unpaid work negatively affecting the quality of sleep (Chatzitheochari and Arber, 2009; Hislop and Arber, 2003a, 2003b, 2003c; Venn et al., 2008). Even more astonishing was the finding that paid work, domestic work and individual and family characteristics that suggest the increased impact of third and fourth shift responsibilities (e.g. women, full-time employment and two-parent families) have a positive impact on both routine of sleep and regular timing of sleep. Apparently, the responsibilities that arise from these shifts lead individuals and households to structure their daily behaviour in a routine way; a routine that continues during the night. To all appearances, people indeed do sleep in the way they do their activities, but, from our point of view, this does not necessarily need to be interpreted in a negative way.

We know from the existing literature that work and family responsibilities have a negative impact on the quality of sleep because they haunt us at night as a fourth shift of continuous emotional involvement in first and second shift tasks. We also know that these negative impacts are the greatest for women, mothers, dual-earner families, et cetera. What we have demonstrated in our study is that sleep duration, regular timing of sleep and routine of sleep as objective indicators of healthy sleep habits are the most strongly present among these disadvantaged groups. Our sleeping habits have barely changed. We sleep almost as long as we did over three decades ago and our work and life cycle changes have hardly affected routine sleep behaviour and regularities in bedtimes. Therefore, we might conclude that the problem of first to fourth shift responsibilities is even bigger than we thought: those who suffer the most from it have been proven to score among the highest on sleep duration, regular timing of sleep and routine of sleep. Thus, although it is to a certain degree difficult to level a busy and restless mind with actual behaviour (cf. Venn et al., 2008), a new hypothesis could be that having healthy sleep habits is no guarantee that we will be able to better cope with work and family responsibilities.

Footnotes

Note

References

Basner

Fomberstein

Razavi

Anks

William

Rosa

Dinges

(2007) American time use survey: Sleep time and its relationship to waking activities. SLEEP 30(9): 1085–1095.

Baxter

(2000) The joys and justice of housework. Sociology 34(4): 609–631.

Biddle

Hamermesh

(1990) Sleep and the allocation of time. Journal of Political Economy 98(5): 922–943.

Burgard

(2011) The needs of others: Gender and sleep interruptions for caregivers. Social Forces 89(4): 1189–11215.

Burgard

Ailshire

(2009) Putting work to bed: Stressful experiences on the job and sleep quality. Journal of Health and Social Behavior 50(4): 476–492.

Chatzitheochari

Arber

(2009) Lack of sleep, work and the long hours culture: Evidence from the UK Time Use Survey. Work, Employment and Society 23(1): 30–48.

Darier

(1998) Time to be lazy: Work, the environment and modern subjectivities. Time & Society 7(2): 193–208.

Gershuny

(2005) Busyness as the badge of honor for the new superordinate working class. Social Research 72(2): 287–314.

Glorieux

Laurijssen

Minnen

Van Tienoven

(2010) In search of the harried leisure class in contemporary society: Time-use surveys and patterns of leisure time consumption. Journal of Consumer Policy 33(2): 163–181.

10.

Glorieux

Minnen

Van Tienoven

(2010) Spouse ‘together time’: Quality time within the household. Social Indicators Research 101(2): 281–287.

11.

Hislop

Arber

(2003a) Sleep as a social act: A window on gender roles and relationships. In: Arber

Davidson

Ginn

(eds) Gender & Ageing: Changing Roles and Relationships, Maidenhead: McGraw Hill/Open University Press, pp. 186–205..

12.

Hislop

Arber

(2003b) Sleepers wake! The gendered nature of sleep disruption among mid-life women. Sociology 37(4): 695–711.

13.

Hislop

Arber

(2003c) Understanding women’s sleep management: Beyond medicalization-healthicization? Sociology of Health & Illness 26(6): 815–837.

14.

Hochschild

(1990) The Second Shift: Working Parents and the Revolution at Home, New York: Avon Books.

15.

Hochschild

(1997) The Time Bind: When Work Becomes Home and Home Becomes Work, New York: Metropolitan Books.

16.

Horrocks

Pounder

(2006) Working the night shift: Prepraration, survial and recovery. A guide for junior doctors. Clinical Medicine 6(1): 61–67.

17.

Javeau

(1970) Les vingt-quatre heures des Belges, Bruxelles: Editions de l’Institute de Sociologie.

18.

Maume

Sebastian

Bardo

(2009) Gender differences in sleep disruption among retail food workers. American Sociological Review 74(6): 989–1007.

19.

Melbin

(1978) Night as frontier. American Sociological Review 43(1): 3–22.

20.

Monk

Kupfer

Frank

Ritenour

(1991) The social rhythm metric (SRM): Measuring daily social rhythms over 12 weeks. Psychiatry Research 36(2): 195–207.

21.

Monk

Petrie

Hayes

Kupfer

(1994) Regularity of daily life in relation to personality, age, gender, sleep quality and circadian rhythms. Journal of Sleep Research 3(4): 196–205.

22.

Noonan

(2001) The impact of domestic work on men’s and women’s wages. Journal of Marriage and Family 63(4): 1134–1145.

23.

Patel

Malhotra

Gottlieb

White

(2006) Correlates of long sleep duration. SLEEP 29(7): 881–889.

24.

Robinson

Godbey

(1997) Time for Life: The Surprising Ways Amercians Use their Time, Pennsylvania: Pennsylvania State University Press.

25.

Robinson J, Martin S, Glorieux I and Minnen J (2011) The overestimated workweek revisited. Monthly Labour Review June: 43–53.

26.

Steptoe

Peacey

Wardle

(2006) Sleep duration and health in young adults. Archives of Internal Medicine 166(16): 1689–1692.

27.

Sullivan

(2007) Cultural voraciousness – a new measure of the pace of leisure in a context of ‘harriedness’. electronic International Journal of Time Use Research 4(1): 30–46.

28.

Szalai

(1972) The Use of Time: Daily Activities of Urban and Suburban Populations in Twelve Countries, The Hague: Mouton.

29.

Taylor

(1993) Unconsciousness and society: The sociology of sleep. International Journal of Politics, Culture and Society 6(3): 463–471.

30.

Venn

Arber

Meadows

Hislop

(2008) The fourth shift: Exploring the gendered nature of sleep disruption among couples with children. The British Journal of Sociology 59(1): 79–98.

31.

Zerubavel

(1982) Temporal regularity. Hidden rhythms, Berkeley: University of California Press, pp. 1–30.