Cluster Analysis Results
We employed cluster analysis on the sequences of self-reported caregivers and identified seven main clusters using the weekday sample and six clusters using the weekend sample. These clusters or categories of sequences represent different types of daily activity patterns among Japanese family caregivers. Tables 1 and 2 summarize the averages for main activities for different caregivers’ clusters on weekdays and weekends, respectively. It is worth noting that non-caregivers, presented in Tables 1 and 2 for comparison, spend substantially less time on adult care—2 min versus almost 50 min a day on average for family caregivers.
Based on the mean estimates for certain activities, we can identify which activities are prevalent in each cluster. For instance, Group 5 on weekdays and Group 6 on weekends show the highest average for the time spent on adult care (574 and 315 min, respectively). Therefore, we named them the ‘Intensive Carers’.
The average amount of paid work for three groups out of seven on weekdays and two groups out of six on weekends is substantially higher than in other groups, particularly for Group 6 on weekdays. We can call all these groups collectively ‘Workers’. Among them, Group 6 caregivers report more than 11 h of paid work on an average weekday. Due to the longest working hours, they can be referred to as ‘Overworkers’, while Group 4 on weekdays and Group 2 on weekends can be referred to as ‘Full-time Workers’ because the hours for this category are close to those of full-time employment (8 h a day). Considering that both groups of ‘Full-time Workers’ report considerable time spent on housework, these groups may also include the second shifters, discussed in the theoretical part. Group 2 on weekdays and Group 1 on weekends can be referred to as ‘Part-time Workers’ because their time spent on work is closer to part-time employment hours. We also note that ‘Full-time Workers’ spend more time on total work (work + housework + adult care + childcare) than ‘Part-time Workers’.
Group 1 on weekdays and Groups 4 and 5 on weekends do the highest amount of housework and very little paid work (5 to 6 min a day). We refer to them as ‘Houseworkers’ category. On weekends, Group 4 reports a longer time, on average, spent on childcare than Group 5. Thus, Group 1 on weekdays (also a cluster with higher childcare time) and Group 4 on weekends may also include the sandwich caregivers discussed in the literature review.
Groups 3 on weekdays and weekends report the longest hours of leisure time. Therefore, we refer to them as ‘Leisurely Carers’. The other group that stands alone (similar to ‘Overworkers’) is Group 7 on weekdays. They report long sleeping hours and personal care time (including medical care). After confirming their diaries’ details, we can report that these diaries belong to people who needed medical attention during the diary day. We decided to refer to these as the ‘Emergency Diaries’.
In order to visualize the daily patterns of caregivers, Figs. 1 and 2 summarize the tempograms (sequence density plots) for different groups of clusters on weekdays and weekends. In all tempograms, each step within sequences represents a 15-min interval, totaling 96 such intervals in a 24-h day. Each sequence starts at 00:00 am and ends at 11:59 pm. The x-axis in all sequence density plots (tempograms) represents the hours and minutes of the diary day from 12 am to 11:59 pm.
The leftmost panels in Fig. 1 represent ‘Full-time Workers’. The tempograms show that they work during the day and do some housework with adult and childcare in the mornings and evenings (or during what can be called their ‘second shift’). The panels in the middle of Fig. 1 represent the daily patterns of ‘Part-time Workers’. Their tempograms show that generally speaking, some proportions of ‘Part-time Workers’ work the morning shifts and some work afternoon shifts. The hours of housework and care work for ‘Part-time Workers’ stretches over the entire day. Finally, the rightmost panels show the daily patterns of ‘Intensive Carers’. These tempograms reveal that most of the day for this type of caregivers is taken up by adult care.
In Fig. 2, the top leftmost and rightmost tempograms show the unique categories of ‘Overworkers’ (leftmost) and those who needed medical attention ‘Emergency Diaries’ (rightmost). For ‘Overworkers’, the entire day, including evenings, is devoted to working activities. In ‘Emergency Diaries’, most of the day is spent on sleep and personal medical services. We also note that there is a considerable amount of travel and some adult care reported throughout the day in these diaries.
In the middle, the tempograms represent the daily patterns of 'Houseworkers' on weekdays and weekends. They summarize that most of the day for this category is spent on doing housework, as well as some adult care and childcare. As was discussed above, Group 4 (weekends) reports more childcare instances than Group 5 (weekends). Thus, for both Group 1 on weekdays and Group 4 (‘Houseworkers 1’) on weekends, childcare responsibilities stretch out throughout the day, whereas for Group 5 (‘Houseworkers 2’), most instances of childcare take place in the afternoons. However, the latter group spends a longer time on adult care than the ‘Houseworkers 1’ group (see Tables 1 and 2). Additionally, in Fig. 2, Groups 3 on weekdays and weekends spend most of the day on leisure activities. The only difference between the two ‘Leisurely Carers’ is that this category of caregivers spends a longer time on travel on weekends than on weekdays, suggesting travel related to leisure activities.
To show the proportional sizes of the different categories of caregivers, Fig. 3 summarizes the proportional (to the samples) sizes of the groups identified by the cluster analysis. It shows that the largest group among caregivers is the ‘Houseworkers’ category. Most family caregivers fall into this category. On weekdays, the next largest category is ‘Full-time Workers’. On weekends, both ‘Full-time Workers’ and ‘Leisurely Carers’ represent a considerable percentage of caregivers. ‘Part-time Workers’, ‘Overworkers’, ‘Intensive Carers’, and those who needed medical attention (‘Emergency Diaries’) represent comparatively smaller proportions of all family caregivers.
Multinomial Regression of the Caregiver Cluster Membership
The primary dependent variable used for the multinomial regression analysis is the categories of at-home eldercare-providers identified from the cluster analysis of sequences by the optimal matching technique described in the section above.
The multinomial logistic regression models include a number of factors relating to the demographic profiles of caregiver categories. These models are run separately for clusters of weekday and weekend diaries. We use the ‘Houseworkers’ category on weekdays and ‘Part-time Workers’ on weekends as the referent. In the model outputs, we report average marginal effects to avoid confusion with reference groups and facilitate results’ interpretation.
We control for employment status, household income, education, the number of children under 15 in the household, the number of women in the household, respondents’ gender, homeownership, and marital status.
Tables 3 and 4 summarize the results of the average marginal effects from the multinomial logistic regression models for weekdays (Table 3) and weekends (Table 4). The dependent variables for both models are family caregivers’ categories identified in the cluster analysis, seven for the weekday diaries and six for the weekend diaries.
Multinomial models indicate that employed caregivers are more likely than non-employed to belong to one of the three groups of workers (‘Overworkers’, ‘Full-time Workers’, and ‘Part-time Workers’). The reverse is true among all other types of caregivers.
Employed caregivers are significantly more likely to be in the 'Overworkers' group than non-employed. Caregivers in the second quartile of household income are statistically more likely to be in this group than those in the lowest quartile. Caregivers over 65 years of age are statistically significantly linked to a lower likelihood of belonging to this group than those under 65. Caregivers with more children are significantly more likely to be in this group than those with fewer children, and women have a significantly lower likelihood to belong to this group than men.
Both on weekdays and weekends, women are statistically significantly less likely to belong to ‘Full-time Workers’ than men. This is an interesting finding that men in Japan are more likely to belong to the group, including the proverbial ‘second shifters’ than women when it comes to eldercare provision to their parents or parents-in-law. It is also explained by the labor force structure in Japan, where lower proportions of women land full-time jobs, and men are compelled to continue working even if they have to take care of their elderly parents. The higher quartiles of household income are significantly linked to a higher likelihood of belonging to the ‘Full-time Workers’ group among weekday diaries, but the same is not true in weekend diaries. This finding indicates that regular employment in Japan is more normative and brings a steadier and higher income for families. However, the education level higher than high school is surprisingly statistically less likely to indicate this group’s belonging than to have education up to middle school.
As to ‘Part-time Workers’, being married is significantly less likely to indicate the belonging to this group in weekday diaries. Thus, the associations show that this pattern is more common among those who are not married than among those who are. In weekend diaries, caregivers with lower education levels are significantly more likely to display time-use patterns of ‘Part-time Workers’.
Caregivers in the lowest quartile of household income both on weekdays (Table 3) and weekends (Table 4) have a statistically more significant likelihood to belong to the ‘Houseworkers’ group, in particular those with higher childcare responsibilities (sandwich caregivers). Additionally, women caregivers are statistically more likely to belong to ‘Houseworkers’ than men caregivers. Married caregivers are also statistically more likely to belong to ‘Houseworkers’ than non-married caregivers. Thus, the results suggest that married Japanese women-caregivers are more likely to have daily patterns of ‘Houseworkers’. Moreover, considering that this type of family caregivers is the largest, it represents a larger fraction of caregivers’ daily routines (see Fig. 3).
In weekday diaries, caregivers of 65 and above are significantly more likely to belong to ‘Leisurely Caregivers’. The same is not true for weekend diaries. Thus, it appears that the retired elderly are more likely to represent the daily patterns for the weekday ‘Leisurely Caregivers’ than those still in the workforce. In weekend diaries, women and married caregivers are less likely to be ‘Leisurely Caregivers’ than men and non-married people. Furthermore, caregivers living in households with more co-residing adult women are more likely to belong to this group, perhaps because this type of household structure allows large amounts of leisure activities to be undertaken by the caregivers due to social gender roles expectations.
No particular demographic stands out significantly with the likelihood to belong to the ‘Intensive Caregivers’, which might indicate that the need for intensive adult care for the elderly may arise regardless of the demographic factors tested in our models. The only significant factor is that non-employed caregivers are more likely to be ‘Intensive Carers’ than employed. This indicates that as soon as the caregiving burden increases, the employment is rarely maintained, even in Japan, with its progressive eldercare policies (Tables 5 and 6).
Non-employed caregivers with higher levels of education, men, and homeowners are statistically more likely to be among those who needed medical attention on the diary day (‘Emergency Diaries’) than employed, lower educated, women, and those who live in rented housing.