Time Use Profiles, Chronic Role Overload, and Women’s Body Weight Trajectories from Middle to Later Life in the Philippines

Abstract

Although chronic life strain is often found to be associated with adverse health outcomes, empirical research is lacking on the health implications of persistent role overload that many women around the world are subject to, the so-called “double burden” of work and family responsibilities. Using data from Cebu Longitudinal Health and Nutrition Survey (1994–2012), we examine the linkage between time use profiles and BMI (body mass index) trajectories for Filipino women over an eighteen-year span. Out of the four classes of women with differential levels of a combination of work and family duties, the group with the heaviest double burden has the highest average BMI. In addition, those who have remained in this class for three or more waves of data not only have higher BMI on average but also have experienced the steepest rate of increase in BMI when they transition from midlife to old age.

With caregiving and domestic duties often considered the domain of women in most parts around the world, increasing female representation in the workforce means that more and more women are juggling between work and family responsibilities, often characterized by the term “double burden” or “second shift” (Bratberg, Dahl, and Risa 2002; Hochschild and Machung 2012). Researchers have explored the potential harm associated with work and family conflict issues (Floderus et al. 2009; Hewitt, Baxter, and Western 2006). However, we know surprisingly little of the long-term health effects of the combination of work and family duties for women, especially as they enter into and exit from midlife, a life stage characterized by complex demands from various life domains and potential early onset of chronic illness and diseases (Lachman 2004). While engagement in multiple roles could enhance psychological and emotional well-being, excessive demands from work, caregiving, and household chores at the same time could become a source for elevated stress, especially when such role overload is enduring and accumulates over the life course.

Existing research has suggested an association between demands or stress from paid or domestic work and changes in body weight (Block et al. 2009; Kivimäki et al. 2006). Being overweight or obese, as indicated by a high body mass index (BMI), is associated with higher risks for mortality and morbidity, such as heart disease, type 2 diabetes, and some forms of cancer (Brunner, Chandola, and Marmot 2007). The increasing prevalence of obesity has become a global epidemic (Atella and Kopinska 2014). While key behavior factors such as excessive calorie intake in food consumption and insufficient physical activity are well-identified causes for overweight, the understanding of psychosocial factors as underlying mechanisms remains unclear in the literature.

This study is set in Cebu, Philippines. The Philippines is in the middle of an epidemiologic transition, with a quarter of adults being overweight, and women are more affected than men (Pedro, Barba, and Benavides-de Leon 2008). Recent surveys identified obesity as an emerging public health issue in the Philippines (EIU 2017). Existing studies on obesity in the Philippines have identified socioeconomic status, urban residence, and hours of employment as significant predictors (Adair, Gultiano, and Suchindran 2011; Colchero and Bishai 2008). In this paper, we are interested in documenting an underexplored area of health research, the double demands from work and family over the life course, as a source of potential role overload for women. Using rich time use diary data from Cebu Longitudinal Health and Nutrition Survey (CLHNS), we developed time use profiles to capture the diversity in women’s work and family responsibilities. Previous research using these data have identified four classes of women with differential levels of a combination of work and family responsibilities (Chen et al, 2017). We expect those with a heavy double burden in work and family arenas to have a higher BMI than other groups. In addition, we maintain that excessive dual demands from both ends for an extended period could amount to chronic life strain. Such wear and tear on the body could have cumulative effects by not only elevating the level of BMI but also accelerating the rate of increase in BMI trajectory.

BACKGROUND

Cumulative Disadvantage and Chronic Overload for Women in Midlife

The view that health risks accumulate from birth and middle age into old age has been widely accepted in the field of health and aging. The cumulative advantage/disadvantage (CAD) theory posits that early adversity in life accumulates and translates into additional disadvantages and consequently leads to widening health disparities in later life (Dannefer 2003; O’Rand 2003). A large body of literature on health disparities over the life course has lent support to the cumulative disadvantage theory. Empirical evidence covers multiple dimensions of health, including self-reported health, composite measures such as physical functioning or health problems such as obesity, hypertension, heart diseases, and cognitive decline (Ferraro and Kelley-Moore 2003; Umberson et al. 2014). Validation of the theory has also been found in a variety of national settings, including the U.S., China, and European countries, although more work has been done on developed than developing countries (Chen, Yang, and Liu 2010; Leopold 2018).

Despite these major developments in the study of life course inequalities, more empirical research is urgently needed in the systematic examination of the cumulative disadvantage process of health. Existing health literature overwhelmingly centers on the impact of social stratification, most notably on SES or racial/ethnic inequality (Geronimus et al. 2006; Pavela and Latham 2016; Umberson et al. 2014), with relatively less attention to other chronic stressors or inequality processes at play. One of the underexplored research areas in the test of cumulative disadvantage theory in health research is a type of chronic strain that many women around the world are subject to on a daily basis, that is, an overload of work and family responsibilities. These duties can be particularly burdensome in midlife, as care demands for home, spouses, children, and parents intensify or even overlap. Persistent gender inequality in both the workplace and family arena is almost universally observed (Jacobs and Gerson 2005). Neither is there any shortage of literature on work and family conflict issues, as women still shoulder the lion share of the caregiving and housework, despite their increasing presence in the labor market (see review by Bianchi and Milkie 2010). However, many studies tend to be segmented, often exclusively considering one side of the burden, for example, job strain, or caregiver stress, with work status or family status as a control (Moen, Kelly, and Lam 2013; Perry-Jenkins et al. 2007). For those that explicitly address work-family conflicts, existing studies often focus on the subjective report of stress or strain (e.g., Roxburgh 2005). Only a few have examined the combination of work and family load and their health implications (Barnett and Hyde 2001; Bratberg et al. 2002; Strazdins et al. 2016), and none of these studies focus on the long-term health effects over the life course.

Theoretically, the role strain perspective proposes that the occupancy of multiple work and family roles can lead to role overload and competing demands from work and family duties could lead to job and family dissatisfaction, tension, and life stress (Davis et al. 2008). There are possible detrimental effects of multiple roles when the number of roles or the time spent on one role exceeds a certain limit (see review by Barnett and Hyde 2001). Some studies suggest that double burden of paid work and housework is associated with higher risks of having more physical health deficits and psychological distress for women, especially when they shoulder the main domestic responsibilities (Bratberg et al. 2002; Roxburgh 2002).

Other studies, however, find that the impact of double demands from full-time paid work and childcare on women’s physical and psychological health is rather inconclusive (Floderus et al. 2009; Hewitt et al. 2006; Krantz, Berntsson, and Lundberg 2005). As an alternative to role strain theory, role enhancement theory suggests that engaging in multiple work and family roles could enhance subjective well-being by promoting self-esteem and life satisfaction and is associated with fewer stress-related mental and physical health problems (Mirowsky and Ross 1986). The beneficial effect of multiple roles is particularly salient among women with higher education in terms of enhancing their autonomy (Ahrens and Ryff 2006). The benefits of multiple roles are also conditional on the number of roles, the time demands, and the quality of each role (Barnett and Hyde 2001).

Additionally, in the discussion on the health implications of double demands from work and family some researchers have pointed out the importance of taking a closer look at the time spent on each role. Recent studies on time use and health have enriched this line of investigation. Scholars have found that excessive time demands from work and family may indirectly lead to poor health through restraining physical activities and the time for preparing and eating healthy food (Strazdins et al. 2016; Venn and Strazdins 2017). Multitasking and feelings of rushing, as sources of stress that oftentimes occur due to time scarcity, can directly lead to poorer self-reported health and psychological well-being (Barnett and Hyde 2001; Offer and Schneider 2011). Time scarcity, as an inevitable result of the double burden of work and family responsibilities, offers explanations for cumulative disadvantage in health brought by chronic role overload over the life course.

Lastly, scholars on health and aging have increasingly moved away from the narrow scope of studying older people alone, and have taken the “long view of aging” by using longitudinal study designs and studying health trajectories over time (Ferraro 2006). Nonetheless, scholars have devoted more attention to the earlier and later periods of life than to the middle years (Lachman 2015). For example, studies have convincingly demonstrated the “long arm of childhood” by documenting the influence of an array of childhood conditions, including family SES and lifestyles on mortality and other health factors later on in life (Hayward and Gorman 2004; Pavela and Latham 2016). However, life course inequality and cumulative “insults” do not stop at childhood. Despite ongoing longitudinal studies such as the Midlife in the United States (MIDUS, beginning in 1995, Brim, Ryff, and Kessler 2004), empirical studies that focus on middle age and beyond have been limited, especially from the perspective of cumulative disadvantage theory. A recent review by Lachman (2015) has called for more attention to midlife trajectories, as it is a “pivotal period” in the life course of individuals, as they are “juggling multiple responsibilities and dealing with physical and cognitive signs of aging” (Lachman 2015: 330).

Body Mass Index (BMI) and Its Linkage to Time Scarcity and Role Overload

In our endeavor to study competing time demands and consequently potential chronic strain for women over the life course, we zero in on a particular type of health outcome, the body weight relative to body height, as measured by Body Mass Index (BMI) for Filipina women. The prevalence of overweight and obesity have been growing to an epidemic proportion in developing countries (Prentice 2006). An increasing body of literature has identified changing diets, rising adoption of sedentary work, genetics and early life nutrition as biological, behavioral, and environmental factors of the spreading obesity pandemic (Van Dyck et al. 2015; Prentice 2006).

Although we know that increasing intake of food high in saturated fats and sugars, coupled with reduced physical activities are the main culprits for weight gain, the mechanisms leading to these behavioral patterns are still not clear. Coming from a time availability perspective, competing demands in work and family responsibilities can lead to reduced time for exercise and preparing healthy food (Strazdins et al. 2016; Venn and Strazdins 2017).

At the same time, time scarcity as a result of the double burden from work and family could also translate into psychosocial stress and contribute to an unhealthy lifestyle including poor dietary behavior and sleep deprivation. A considerable amount of research on obesity has been directed toward the role of stress as a risk factor (Block et al. 2009; Shimanoe et al. 2015). While acute stress may reduce food intake in the short term, chronic stress is often associated with the release of cortisol, which has been reported to stimulate appetite and to influence eating more energy-dense foods that are high in sugar and fat (Torres and Nowson 2007). Some population-based studies have found job stress, or perceived stress in general, has no or weak association with obesity (Kouvonen et al. 2005; Shimanoe et al. 2015). Other longitudinal studies have found that job strain is associated with weight gain for men already overweight at baseline, but has no effect for women (Kivimäki et al. 2006). Others report chronic job strain coupled with low social support leads to obesity (Brunner et al. 2007). Block and colleagues (2009) have found that job strain and work stress turn out to be the most important predictors for weight gain, among multiple domains of psychosocial stress related to work, personal relationships, life constraints, and finances.

The Filipino Context

When it comes to the time bind between work and family responsibilities, Filipino women are not different from women in many other settings around the world. Parreñas (2007) has pointed out that there is a gender ideological clash in Filipino society, which simultaneously greets women’s labor force participation and emphasizes their domestic role. On the one hand, the work of women in the public sphere is considered valuable to the national economy and individual households. The female labor force participation rate in the Philippines remained above 46.6% since 1990 (International Labour Organization 2017). Many women have become a co-economic provider with their husbands to the household (Medina 2001).

On the other hand, domestic responsibilities are still the primacy of Filipino women’s gender role. Filipino women bear the major share of household chores, caregiving, and household economics (Alcantara 1994; Medina 2001). Although Filipino husbands increasingly share some household tasks, wives still do more than their husbands (Diefenbach 2002). Women’s participation in domestic tasks is associated with the family’s social and economic mobility; meanwhile, fulfilling a childrearing responsibility also promotes women’s status and decision-making power in the family (Alcantara 1994). As a result of this paradoxical but dual emphasis on women’s role both in the private and public sphere, we expect that many Filipino women are subject to a heavy double burden from work and family.

One of the distinctive characteristics of Filipino families is the reciprocal exchange among extended kin (Agree, Biddlecom, and Valente 2005; Miralao 1997). Women’s domestic responsibilities continue with the transition to grandparenthood. Close to 60 percent of Filipino women aged 60 and over co-reside with at least one of their grandchildren (Agree et al. 2005). Our prior work with the same data documented that as these women age, there was a substantial increase in the proportion them co-residing with grandchildren (from 20.6% in 2002 to 58.1% in 2012) (Chen et al. 2017). Behind the prevalence in migrant Filipina domestic workers around the world, female extended kin (e.g. grandmothers, aunts) often take over household tasks for migrant Filipinas and become a surrogate parent for left-behind children in transnational families (Parreñas 2000). Therefore, when Filipino women enter a later life stage, they may still need to divide their time for work, domestic labor, and caregiving for grandchildren. Our previous study shows that many of these Cebu women were subject to the double burden of work and family demands in a span of 18 years from mid- to later life (Chen et al. 2018).

During our study period (1994–2012), the total fertility rate in the Philippines was 4.1 in 1994 and slightly decreased to 3.1 in 2010–2015 (United Nations 2017). Only 11.5% of Filipino women were childless by the end of their reproductive span (Philippine Statistics Authority 2005). Our sample is a cohort of women with higher fertility than the general population because all of them entered the first survey wave while pregnant. Therefore, chronic role overload could be a more prominent issue for them compared to childless women in regard to their responsibilities as mothers and grandmothers over the life course. During the process of nutrition transition (Popkin and Gordon-Larsen 2004), with many developing countries experiencing shifts to a diet high in saturated fats and sugar, and patterns of lower levels of physical activity, increasing overweight and obesity has become a concern in the Philippines (Adair et al. 2011). Previous research shows that among the women participated in the same survey for our study, the percentage of women overweight and obese (combined) increased by 6-fold from 7% in 1983 to 43% in 2005 (Adair et al. 2011). Urbanization, household socioeconomic status, age, diet, reproductive history, and physical activity at work are predictors of weight gain (Adair et al. 2011). The effect of life stress such as work-family role overload is rather underexplored.

Research Goals

In this paper, we adopt the cumulative disadvantage theory, which provides central guidance to the study of health disparities over the life course in different societies. As Dannefer (2003: S327) stated, “(m)ore formally, cumulative advantage/disadvantage can be defined as the systemic tendency for interindividual divergence in a given characteristic (e.g., money, health, or status) with the passage of time.” Needless to say, money and status are fundamentally important characteristics in relevance to health. However, we believe interindividual divergence in other health-related characteristics deserves closer attention as well, such as chronic double burden of work and family responsibilities and the associated time scarcity and role overload.

First, we describe the diversity in women’s work and family responsibilities by using rich time use data from a longitudinal dataset. Based on the patterns of time use, we classify women into groups with different extent of “double burden” in work and family duties. The zero-sum nature of time inevitably means that work, caregiving, housework, and leisure time have competing demands on one’s time. A direct examination of actual time allocation for work and family thus can be an excellent gauge of the extent of overburden. We expect those with heavy demands from both work and family are likely to suffer from a disadvantage in the level of BMI. Second, while we expect diversity in women’s work and family responsibilities, that is, not all women have the same type of demands from the work and family arena, we posit that time use patterns also change throughout one’s lifetime. A previous study of our research team has documented that, as a woman moves from younger to older adulthood, together with changes in her family roles and household composition, the configuration of time devoted to work, caregiving and housework activities also shifts over time for many women while remaining stable for some others (Chen et al. 2018). Following the life course perspective (Elder, Johnson, and Crosnoe 2003), we expect the changes and continuities in women’s work and family responsibilities could affect their BMI level differently.

Third, we maintain that women’s time use profiles not only affect their BMI status (i.e., the average level) but also have consequences on their BMI trajectories (i.e., the rate of change). Cumulative disadvantage theory suggests that stressful experiences could accumulate over time, lead to stress proliferation, and therefore widen health inequality over the life course (Dannefer 2003). We expect women who are overburdened with work and family duties for a long period of time to face more cumulative stresses over the life course. The gap in BMI levels by age across women with different time use profiles (e.g., intensity and duration of work and family overburden) is expected to widen over time as a result.

DATA AND METHODS

Data

We use data from the Cebu Longitudinal Health and Nutrition Survey (CLHNS), an ongoing collaborative project of the Carolina Population Center at the University of North Carolina and the Office of Population Studies Foundation at the University of San Carlos in Cebu. The survey follows a cohort of mothers and an index child born in 1983–84 longitudinally. Using a single stage cluster sampling procedure, 17 urban and 16 rural barangays (local administrative units) were randomly selected from the 255 barangays in Metropolitan Cebu. Cebu is a province of the Philippines and Metropolitan Cebu is the second largest metropolitan area in this country. The 33 barangays, representing about 28,000 households, were surveyed to locate all pregnant women. Those who gave birth during the period of May 1983 through April 1984 were included in the sample (n = 3,327, response rate 90%) and followed up in 1991, 1994, 1998, 2002, 2005, 2007 and 2012. There is no comparable large sample of women in developing countries who are followed from the reproduction years to older adulthood, making CLHNS uniquely suitable to study changes in women’s life trajectories in the larger context of epidemiological transition (e.g., Dahly and Adair 2007; Schmeer 2010; Zimmer et al. 2017).

In 1994, time use diary data were collected for the first time, making it Wave 1 for the current analysis, and subsequent Waves 2 to 5 were based on data collected in 1998, 2002, 2005, and 2012. The 2007 data was excluded due to the lack of time use data. We start our analysis with a sample of women in 1994 with the mean age of 38. At the time of the last observation in 2012, they were between ages 45 and 65. After excluding observations with missing values on key variables (about 1% across waves), our analytical sample consisted of 2,262 persons in 1994; 1,986 in 1998; 2,098 in 2002; 2,010 in 2005; and 1,805 in 2012. Among them, 131 persons died by the survey of 2012, and 261 persons were lost to follow up across the surveys. On average, a person was observed more than four times in this longitudinal data, and this altogether yielded 10,161 person-year observations.

Measurement of Key Dependent and Independent Variables: BMI and Time Use Profiles

Our dependent variable is Body Mass Index (BMI), which is calculated as weight (in kilograms) divided by height (in meters) squared. Respondents’ weight and height were assessed by trained personnel at each wave of the CLHNS (1994, 1998, 2002, 2005, and 2012). A BMI between 18.5 and 24.9 is commonly considered to be within the normal range (WHO expert consultation 2004). The average BMI for women in our sample increased as they aged. It was 23.3 in 1994, 23.6 in 1998, 24.2 in 2002, 24.3 in 2005, and 24.8 in 2012.

The key independent variable, time use profiles, is based on time use data using 24-hr activity diaries (reported for a typical weekday, excluding weekends). Unfortunately, the data does not include any information on weekend time use. Time allocation can largely vary by days, especially between weekdays and weekends, and using a “typical” workday rather than the last 24 hours could induce recall error (National Research Council 2000). We acknowledge this as a limitation, as it could potentially underestimate leisure time, or underestimate double burden if a woman works on weekends. We first calculate the amount of time (in hours) spent on each daily activity, then combine these time use activities into five broad categories: working outside the home for pay, working at home for pay, household chores, caregiving, and personal time. After carefully examining the distribution of daily hours spent on each of these five categories, we collapse daily time use in each of the five categories into three levels of intensity: zero/low intensity, moderate intensity, and high intensity (cutoff points varied by each category, see Chen et al. 2018 for detailed criteria). Using these three-intensities across five time use categories as indicators, we conducted latent class analysis and developed six-group time use profiles: high intensity worker away from home, high intensity worker based at home, moderate intensity worker away from home, moderate intensity worker based at home, high intensity caregiver, and homemaker (see Chen et al. 2018 for detailed documentation of the time use profiles).

For this paper, we use a simplified version of the typology and classify women in four groups, i.e., high intensity worker, moderate intensity worker, high intensity caregiver, and homemaker. Description of each time use profile is shown in Table 1. The first two classes of women both shoulder a heavy double burden of work and household responsibilities. The high intensity worker is subject to a heavy workload (inside or outside the home for pay) (>8 hours/day), is most likely to engage in a moderate level of household chores (3–6 hours/day) and caregiving, and has a very low level of personal time. Moderate intensity worker’s workload is lighter (<=8 hours/day), but her household chores and caregiving duties tend to be heavier than the former, and the level of personal time is slightly higher (probability not shown in Table 1, see Chen et al. 2018 for a more detailed description). Neither high intensity caregiver nor homemaker engages in any income generating activities but does a heavy load of housework. What distinguishes these two groups is that the first carries a high load of caregiving duties and the latter enjoys the highest level of leisure time.

Table 1.

Description of Time Use Profiles, 1994-2012 (Pooled sample, 1994, 1998, 2002, 2005, and 2012)

	High intensity worker	Moderate intensity worker	High intensity caregiver	Homemaker
Work hours	> 8 hours	≤ 8 hours	0	0
Housework hours	3-6 hours	3-6 hours	> 6 hours	> 6 hours
Caregiving hours	≤ 1 hours	≤ 1 hours	> 1 hours	≤ 1 hours
Leisure hours	< 3 hours	3-6 hours	3-6 hours	> 6 hours
N	3.441	2.388	1.343	2.989

It is important to note that individuals’ time use patterns change over time, as the women in our sample age. As shown in Figure 1, membership in moderate intensity worker has declined sharply over time, from 32.7% in 1994 to 5.9% in 2012. The group of homemaker has more than doubled from 1994 to 2012. Interestingly, almost a third of the sample are classified as high intensity workers and the proportion remains relatively stable over the span of eighteen years.

Figure 1.

Distribution of Time Use Profiles, 1994–2012

In addition to documenting women’s membership in different time use profiles at each time point and its concurrent association with BMI, we are also interested in longitudinally examining how persistence in a certain time use profile may be associated with different BMI trajectories. To do so, we operationalize the duration of time use membership across waves of the survey. Table 2 demonstrates that a sizable proportion of the women are subject to heavy double demands over an extended period. Strikingly, about 26.8% of the sample are classified as the high intensity worker in three waves of the data or more. In contrast, only 18.6% of the sample are homemakers for three or more waves.

Table 2.

Percent in Duration of Time Use Profiles, 1994-2012 (N = 10,161) (Pooled sample, 1994, 1998, 2002, 2005, and 2012)

	None	One wave	Two waves	Three waves or more
High intensity worker	31.6	23.9	17.7	26.8
Moderate intensity worker	34.9	32.8	21.7	10.6
High intensity caregiver	56.8	28.3	11.1	3.7
Homemaker	28.7	30.8	22.0	18.6

Time Use Profiles and Its Association with Excessive Demands from Work and Family

We conceptualize women’s time use profiles as a source of chronic role overload. Membership in each four of the categories represents the combination of work, household chores, and caregiving duties, with the high intensity worker shouldering the heaviest double burden overall. Since we do not directly measure stress, we conduct additional analyses to examine whether our time use profiles indeed are associated with the manifestation of stress and strain in life. From its 2002 wave (but not available in earlier waves), CLHNS asked respondents to evaluate the levels of demandingness from work and household activities on the basis of seven characteristics: physical demandingness, dexterity, level of multi-tasking, mathematical skills, reading skills, the importance of teamwork, and stress. For each characteristic, the responses were recorded on a 0 to 3 scale, in which 0 indicates not demanding and 3 the most demanding. We summed responses of all these seven characteristics to form an additive scale that measures demandingness from work and home separately, ranging from 0 to 21, and a summary index that combines both. Table 3 shows averages of these scales by time use profiles. In addition, we also singled out two single-item characteristics, physical demandingness and stress, as both could be most relevant for BMI. As expected, high intensity workers and moderate intensity workers are exposed to higher demands from work than caregivers and homemakers. However, their mean score of demandingness from home are not significantly different from the other two groups, suggesting that they are in much higher risks of “double burden” from their work and family roles. High intensity workers have the highest mean score of demandingness overall and report the highest level of stress from work.

Table 3.

Mean Scale of Demandingness from Work and Family Responsibilities, by Time Use Profiles, 2002-2012

	High intensity worker	Moderate intensity worker	High intensity caregiver	Homemaker
Scale of Demandingiiess	10.65	9.99	5.14	4.44	*
from Work	(2.87)	(3.15)	(5.74)	(5.49)
Scale of Demandingiiess	10.10	10.21	10.95	10.67
from Home	(3.18)	(3.19)	(3.04)	(3.17)
Scale of Demandingiiess	20.75	20.20	16.09	15.11	*
from Both Work and Home	(5.45)	(5.68)	(7.10)	(6.58)
Physical Demandingiiess	1.73	1.66	0.81	0.70	*
from Work	(0.74)	(0.76)	(0.98)	(0.94)
Stress from Work	1.77	1.66	0.80	0.71	*
	(0.61)	(0.61)	(0.92)	(0.91)
Physical Demandingiiess	1.58	1.66	1.74	1.59
from Home	(0.66)	(0.66)	(0.67)	(0.64)
Stress from Home	1.53	1.59	1.66	1.59
	(0.57)	(0.58)	(0.59)	(0.57)
N	2,022	923	830	2,140

Notes: *p < 0.05 (signifies significant differences among different time use profiles on the basis of one-way ANOVA test)

Control Variables

We also control for potential confounders of the relationship between time use profiles and BMI in the analysis. They include two domains of covariates: demographic characteristics and socioeconomic status (SES) characteristics. For the former, we control for women’s marital status (1=married; 0=not married) and household composition (number of children aged 0–17, number of adults aged 18–59, and number of older adults aged 60 and older). We also control for women’s pregnancy status. As for SES, we include two measures at the individual and household level, years of education and a quartile index measuring household assets, with those in a higher quartile indicating a higher level of household SES. In terms of community-level SES, we use an urbanicity score that has been validated in earlier studies using CLHNS, which quantifies the concentration of amenities typically found in urban communities (Dahly and Adair 2007; Zimmer et al. 2017). It ranges from 7 to 61, and a higher value indicates a higher level of urbanization. Table 4 presents descriptive statistics of sample characteristics by survey year.

Table 4.

Descriptive Statistics of CHLNS Respondents by Survey Year, 1994-2012

	1994	1998	2002	2005	2012
BMI	23.3	23.6	24.2	24.3	24.8
	(3.9)	(4.1)	(4.3)	(4.3)	(4.7)
Age (in years)	37.9	41.9	45.1	47.9	55.1
	(6.1)	(6.1)	(6.1)	(6.1)	(5.9)
Education (in years)	7.3	7.4	7.3	7.3	7.3
Assets score (%)	(3.8)	(3.9)	(3.8)	(3.8)	(3.8)
1st Quartile (Lowest)	30.5	25.1	25.0	33.5	29.6
2nd Quartile	15.9	23.5	32.7	20.3	20.8
3rd Quartile	30.4	29.7	17.3	22.2	23.2
4th Quartile (Higliest)	23.2	21.7	25.0	24.0	26.4
Urbanicity	35.8	38.8	41.2	40.5	43.9
Household composition	(13.2)	(13.7)	(14.1)	(13.6)	(12.6)
Number of children (aged 0-17)	4.1	3.5	2.4	2.0	1.6
	(1.8)	(2.3)	(2.1)	(1.9)	(1.7)
Number of adults (aged 18-59)	2.7	3.1	4.0	4.4	4.0
	(1.4)	(1.4)	(1.5)	(1.5)	(1.9)
Number of older adults (aged 60+)	0.2	0.2	0.2	0.2	0.5
	(0.5)	(0.4)	(0.4)	(0.5)	(0.7)
Married (%)	93.5	91.1	88.5	86.1	75.4
Pregnant (%)	4.6	2.1	1.2	0.4	0.0
N of persons	2,262	1,986	2,098	2,010	1,805

Notes: Values for categorical variables are in percent (with “%” following the variable labels). The mean values, followed by standard deviations in parentheses, are presented for all other variables.

Methods

To investigate the effect of Filipino women’s time use profiles on their BMI from mid- to later life, we use growth curve models to account for intra-individual changes in time use profiles and then examine both intra-individual and inter-individual differences in BMI trajectories (Raudenbush and Bryk 2002). With time-variant covariates, this analytic approach allows individuals to serve as their own controls and thus we are able to control for both intra- and inter-individual confounds. By allowing random-effect variations across individuals, it also takes into account the unknown heterogeneity across individuals. The model is represented by equation 1.

$$BM{I}_{ti}={\beta }_{0i}+{\beta }_{1i}Age{C}_{ti}+{\beta }_{2i}Age{C_{ti}}^2+{\beta }_{3i}TU{P}_{ti}+{\beta }_{4i}Age{C}_{ti}\mathrm{x}TU{P}_{ti}+e_{ti}$$ (1)

Age is the analytical time metric and centered by mean of age (AgeC_ti) so that the intercept reflects the level of BMI at the average age of 45. We begin with a change trajectory model of BMI of individual i at time t (BMI_ti), as a function of age (AgeC_ti) and its quadratic term due to its better empirical fit and a theoretical expectation of a nonlinear pattern of BMI increase. We further add our key independent variable, measures of time use profiles (TUP), and other time-varying variables. Because we hypothesize that the effect of time use profiles is age-dependent, we also add the interaction terms between age and the time use profiles measures at level 1.

We then posit a level-2 sub-model for inter-individual differences in change, where the coefficients βs in the level-1 model are further modeled as dependent variables. Although technically it is possible to model all of the βs, we choose models based on our theoretical hypotheses. We begin with two unconditional models of the intercept model β_0i and linear rate of change β_1i at Level-2 but will also test a model on β_2i (quadratic rate of change) later. Other predictor variables are entered at level-1 for time-varying covariates (e.g., marital status) and at level-2 for time-constant covariates (e.g., education). We address the potential bias introduced by attrition and death by using a simple but effective strategy, namely, by entering dummy variables indicating the deceased and non-respondent identities (Raudenbush and Bryk 2002).

$${\beta }_{0i}={\gamma }_{00}+{\gamma }_{01}{X}_{i1}+{\gamma }_{02}{X}_{i2}+\dots +{\gamma }_{0k}{X}_{ik}+u_{0i}$$ (2)

$${\beta }_{1i}={\gamma }_{10}+{\gamma }_{11}{X}_{i1}+{\gamma }_{12}{X}_{i2}+\dots +{\gamma }_{1k}{X}_{ik}+u_{1i}$$ (3)

Our growth curve models consist of two major steps. First, we are interested in whether the four-category time use profiles predict the intercept (e.g., an average level of BMI) and the slope (rate of increase in BMI trajectory). Second, we model whether the duration of being engaged in a heavy double burden from work and family (e.g., being a high intensity worker in more than three waves of data) is particularly vulnerable to a higher average level of BMI as well as an accelerated rate of BMI increase.

RESULTS

Results from growth curve models are presented in Table 5 and 6. Table 5 presents the effect of time use profiles on Filipino women’s BMI trajectories. The main effect of time use profiles reflect its effect on average BMI at age 45 (the centered age value), and the interaction effects of time use profiles by age reflect the effects of time use profiles on the rate of change of BMI trajectories. From Model 1 to Model 3, we add time use profiles, their interactions with age, and other control variables gradually, and results show a statistically significant effect of time use profiles on BMI consistently across models, after controlling for sociodemographic and socioeconomic characteristics (Models 3). Women who are high intensity workers, on average, are highest in their BMI among the four time use profiles: at age 45, their BMI is 0.252 higher than those of moderate intensity workers; 0.334 higher than high intensity caregivers; and 0.235 higher than homemakers. Interaction effects of time use profiles by age, however, do not show any significant effect, which suggests that variation in BMI among time use profiles are stable and persistent across age.

Table 5.

Growth Curve Models Predicting the Effect of Time Use Profiles on BMI, 1994-2012

	Model 1	Model 2	Model 3
Time use profiles (ref =high intensity worker)
Moderate intensity worker	−0.370***	−0.261***	−0.252***
	(0.050)	(0.050)	(0.049)
High intensity caregiver	−0.385***	−0.350***	−0.334***
	(0.061)	(0.060)	(0.059)
Homemaker	−0.167***	−0.218***	−0.235***
	(0.051)	(0.051)	(0.050)
Age (centered on the mean)		0.103***	0.081***
		(0.005)	(0.006)
Age (centered), squared		−0.003***	−0.002***
		(0.000)	(0.000)
Moderate intensity worker ×' Age		−0.003	−0.003
		(0.006)	(0.006)
High intensity caregiver × Age		−0.004	−0.007
		(0.007)	(0.007)
Homemaker × Age		−0.007	−0.003
		(0.006)	(0.006)
Education (in years)			0.091***
			(0.020)
Assets score (ref.=lst quartile)
2nd Quartile			0.273***
			(0.052)
3rd Quartile			0.352***
			(0.060)
4th Quartile			0.673***
			(0.072)
Urbanicity			0.028***
			(0.003)
Household composition
Nmnber of children			−0.077***
			(0.013)
Number of adults			0.053***
			(0.013)
Nmnber of older adults			0.001
			(0.046)
Married			0.350***
			(0.080)
Pregnant			1.542***
			(0.122)
Attrition status
Died	−0.492	−0.596	−0.421
	(0.375)	(0.377)	(0.363)
Loss to follow-up	0.456	0.647*	0.346
	(0.267)	(0.270)	(0.261)
Random effects – variance components
Level 1: Witliin-person	1.858***	1.839***	1.766***
	(0.035)	(0.035)	(0.033)
Level 2: In intercept	16.583***	16.577***	15.255***
	(0.532)	(0.518)	(0.480)
Level 2: In linear growth rate	0.031***	0.021***	0.020***
	(0.001)	(0.001)	(0.001)
Constant	23.836***	24.389***	21.976***
	(0.095)	(0.099)	(0.226)
BIC (smaller is better)	46.080	45,567	45.183
N of persons	2,292	2,292	2,292
N of person-year observations	10,161	10,161	10,161

^***p < 0.001

^**p < 0.01

^*p < 0.05 (two-tailed tests)

Table 6.

Growth Curve Models Predicting the Effect of Duration of Being a High Intensity Worker on BMI, 1994-2012

	Model 1	Model 2	Model 3
Duration of being a high intensity worker (ref.=none)
One wave	0.538*	0.647**	0.603**
	(0.222)	(0.231)	(0.223)
Two waves	0.735**	0.838***	0.721**
	(0.249)	(0.260)	(0.250)
Three or more waves	1.137***	1.379***	1.097***
	(0.223)	(0.232)	(0.225)
Age (centered on the mean)		0.081***	0.061***
		(0.007)	(0.007)
Age (centered), squared		−0.003***	−0.002***
		(0.000)	(0.000)
High intensity worker in one wave × Age		0.016	0.016
		(0.011)	(0.010)
High intensity worker in two waves × Age		0.022	0.019
		(0.012)	(0.011)
High intensity worker in three or more waves × Age		0.043***	0.039***
		(0.010)	(0.010)
Education (in years)			0.088***
			(0.020)
Assets score (ref.=lst quartile)
2nd Quartile			0.269***
			(0.052)
3rd Quartile			0.342***
			(0.060)
4th Quartile			0.661***
			(0.073)
Urbanicity			0.028***
			(0.003)
Household composition
Number of children			−0.079***
			(0.013)
Number of adults			0.053***
			(0.013)
Number of older adults			0.003
			(0.046)
Married			0.341***
			(0.080)
Pregnant			1.542***
			(0.122)
Attrition status
Died	−0.281	−0.377	−0.253
	(0.377)	(0.379)	(0.365)
Loss to follow-up	0.636*	0.816***	0.480
	(0.269)	(0.272)	(0.263)
Random effects – variance components
Level 1: Witliin-person	1.864***	1.849***	1.777***
	(0.035)	(0.035)	(0.033)
Level 2: In intercept	16.443***	16.388***	15.156***
	(0.528)	(0.512)	(0.477)
Level 2: In linear growth rate	0.032***	0.021***	0.020***
	(0.001)	(0.001)	(0.001)
Constant	23.077***	23.542***	21.300***
	(0.154)	(0.160)	(0.253)
BIC (smaller is better)	46,124	45,560	45,194
N of persons	2,292	2,292	2,292
N of person-year observations	10,161	10,161	10,161

^***p < 0.001

^**p < 0.01

^*p < 0.05 (two-tailed tests)

To illustrate the effects visually, we display BMI trajectories in Figure 2 by using all the coefficients in Model 3 from Table 5 varying by our key predictor variable (time use profiles), while holding all the control variables at their means (for continuous variables) and modes (for dummy variables). Figure 2 shows a clear disadvantage for high intensity workers, with their BMI level being the highest across the life course compared with the other three groups. The differences among the BMI levels of the other three groups are not statistically significant (supplementary analysis with different reference groups not shown). After the age of 50, the BMI level of high intensity workers increases above 25, crossing the threshold for overweight (WHO expert consultation 2004), while the average BMI of the other three time use profiles remains under 25.

Figure 2.

Predicted BMI Trajectories by Time Use Profiles, 1994–2012

Table 5 also indicates that those with a higher level of education and household assets, and living in more urban areas are more likely to have a higher level of BMI. This is consistent with findings from other developing countries in the middle of an epidemiological transition and economic development, where higher household and individual SES may bring risks to adult overweight since higher SES is associated with increased consumption of high fat, high sugar diets, and reduced physical activity (Popkin and Gordon-Larsen 2004).

Table 6 further explores whether the duration of being a high intensity worker (one wave, two waves, three or more waves, compared with none) has any effect on BMI trajectories. Results from Model 3 show not only a clear main effect of longer duration but also a significant interaction effect with age (high intensity worker in 3 or more waves x Age). This suggests that the duration effect is significant at both the intercept and slope level. To begin with the main effect of duration, it shows that a longer duration of being a high intensity worker is associated with a higher BMI level at age 45 (the centered age value), with an increase of 0.603 for one wave, 0.721 for two waves, and 1.097 for three or more waves. Furthermore, the significant interaction effect of being a high intensity worker in three waves or more indicates a more rapid increase in BMI by age under this circumstance.

We again display the effect visually in Figure 3, which shows that being a high intensity worker for an extended period of time has a detrimental effect on BMI trajectories. Being a high intensity worker in three or more waves of the data not only translates into a higher BMI level, but the rate of increase in BMI is also considerably sharper compared with the other groups. To illustrate the magnitude of the divergence, we translate the BMI differences into weight differences. For someone with average height (1.51m) at age 30, the predicted weight difference between a woman who is a high intensity worker and who is not is 0.9kg. After having remained in this category for three waves or more, the gap between her and someone who has never been in this category diverge to as large as 2.7kg.

Figure 3.

Predicted BMI Trajectories by Duration of Being a High Intensity Worker, 1994–2012

DISCUSSION

The appeal of cumulative dis/advantage (CAD) theory to health research lies in its simple, basic premise that variation in strains and resources accumulates over the life course, leading to “interindividual divergence” in the health trajectory. Given that socioeconomic status (SES) is widely accepted as a fundamental cause of health inequality (Link and Phelan 1995), health studies on life course heavily focus on socioeconomic disadvantage as a source of numerous health problems and a catalyst of accelerated aging. Nonetheless, disadvantage and hardship can manifest in a wide range of ways, reflecting varying types of inequality in social systems. A recent overview of the growing body of work applying CAD theory insightfully points to its broad relevance as a general sociological construct (Dannefer 2018). We believe our study provides an important extension of the theory in the study of women’s health. Our empirical analyses have clearly identified chronic role overload over the life course as a source of disadvantage for women, with excessive demands from both work and family arena day in and day out taking a toll on their BMI trajectories as they transition from midlife into older adulthood.

Our time use profiles document that close to a third of the women in the sample fall into the “high intensity worker” category. These are women steadily engaged in a high level of work for pay (more than eight hours a day) and do three to six daily hours of housework and caregiving. The label itself is a misnomer because the extent to overwork spreads to both work and family duties. Further, over a quarter of the women were consistently categorized as “high intensity workers” (in >= 3 waves of data) during an eighteen-year span. In contrast to what is generally observed for older adults in developed countries, whom often experience a surge in leisure time (Gauthier and Smeeding 2003), it is striking that a sizable proportion of Filipino women in our sample are still subject to a heavy double burden as they transition from midlife to old age. The heavy workload, together with increasing caregiving responsibilities as grandmothers, could eliminate time for personal care.

We believe our use of time use profiles is an innovative way to document the multiple burdens women commonly face in a developing country, where boundaries of work and family are often fluid and standard survey questions on employment and household structure may fail to capture the full extent of these responsibilities. Although we do not directly measure subjective role strain, the zero-sum nature of time inevitably means that the time for work, caregiving, housework, and self-care (including physical exercise) crowd each other out. Our additional analyses on the association between time use profiles and perceptions of demandingness and stress confirm the validity of time use profiles as a gauge of chronic role overload. Those who are subject to heavy demands from work and family responsibilities thus face the consequence of time scarcity and an elevated level of stress and this, in turn, translates into an adverse BMI trajectory.

More importantly, our growth curve models on BMI trajectories provide direct evidence that time use profiles are significantly associated with BMI statuses as well as their rates of increase over the life course. Those who are subject to extreme demands of work and family, “high intensity workers”, have the highest level of BMI throughout the observed life course. In addition, those who remain “high intensity workers” for three or more waves of the survey not only have higher risks of overweight but also are subject to accelerated rates of increase of BMI over time.

It is important to note that the effects we observe are net of socioeconomic status (SES). Undoubtedly, the role of SES has been well documented in affecting BMI (Atella and Kopinska 2014; Colchero and Bishai 2008; Schmeer 2010). However, women with higher SES often report having less time and feeling more time pressure in both Asian and non-Asian settings (Cha and Suh 2017; Roxburgh 2002). Women who are high intensity workers (i.e., high time pressure) in our sample are more likely to come from higher SES households and report a higher level of education. Our study thus lends to the argument that time is also an important health resource thus shedding light on the link between exposure to chronic role overload and increasing health disparities in later life, extending cumulative dis/advantage theory in a new direction.

Although our analyses are exclusively on women and we are not able to compare men and women’s time use profiles, we believe our study helps to elucidate the health implications of gender roles and gendered life course. Although work and family conflicts are certainly not unique experiences to women alone, women around the world continue to shoulder more responsibilities for domestic and caregiving work than men, despite their increasing opportunities and participation in the labor market. The persistence of gendered household division of labor in many countries reflects deep-rooted gender stratification. Although a full study on the gendered process has to involve men, our study provides insight into the health implications of a heightened sense of burden and stress that results from work and family overload, which affects women more severely than men.

The Filipino women in our sample may be a particularly vulnerable group, caught between the high fertility norms in the 1980s and work roles that are often ascribed within a rapidly developing region (Alcantara 1994; Medina 2001). Even in developed countries such as the U.S., where gender ideology is more egalitarian although by no means equitable, women are still disproportionately more likely than men to do more housework and to be caretakers, and are increasingly becoming “sandwiched” between the care of children and elderly parents in midlife (Grundy and Henretta 2006; Keene and Prokos 2007). For example, U.S. studies have documented women to be considerably more likely to feel overburdened with work and family responsibilities and to have a heightened level of stress than men in dual-earner families (Offer and Schneider 2011). We speculate that our selective sample of Filipino women may not be unique in their experience and therefore our study could have enormous implications for studying women’s health in other settings.

We acknowledge that we only focus on the effect of age, or life course, in our longitudinal data analysis from 1994 to 2012. During this period, the Filipino society has experienced extensive social and economic growth. Rapid economic development and ongoing health transition means different cohorts of women could be impacted differently by changes in the larger social context. However, our study sample was drawn from women who gave birth in 1983–1984, with a mean age of 26 and a standard deviation of 6. The relative lack of heterogeneity in age does not lend itself to separate age and cohort effects. As the cumulative dis/advantage is a systemic tendency of the cohort as well as individual aging process, future studies using an accelerated longitudinal study design are warranted.

Another apparent limitation of using the time use profiles over the life course as a source of role strain is that we do not directly measure them. However, our results are remarkably consistent with existing studies on BMI that include perceived stress or strain in different settings. Although these studies vary by their operationalization of stress, ranging from reported job strain to an overall measure of psychosocial stress, they all have documented an association between stress level and weight change (Block et al. 2009; Kivimäki et al. 2006). In spite of the fact that these studies are based on panel data, none of them are able to track BMI trajectory over time such as how we have done in this paper, by following a cohort of women as they transition from midlife to old age. By contextualizing women’s body weight changes through the lens of the cumulative dis/advantage theory, we advance the literature by establishing linkages across various life domains during the process of aging. In future work, we will utilize dietary, health behavior and biomarker data such as allostatic load and thus map out the complex mechanisms through which time use patterns can be linked to BMI. Despite our limitations, we believe our study is a step forward in documenting the health implications of women’s work, family and social contributions over the life course, particularly during the critical period as they transition from midlife to old age. We believe our study should spark further research on the implications of role overload and time scarcity in diverse settings.