Cohorts and Emerging Health Disparities

Abstract

Despite the recent and rapid worldwide rise in body mass index (BMI), little empirical research outside the developed world has systematically considered the role of cohorts in inaugurating emergent biomorphic disparities. This study integrates aspects of the life course perspective (attention to age- and cohort-level influences) with fundamental cause theory to investigate how BMI differences have unfurled in the Chinese population. We analyze growth-curve models of adults in the China Health and Nutrition Survey. The results indicate that more recent cohorts have higher levels of BMI and, among women, experience a steeper rate of age-related BMI growth. Moreover, biomorphic change has unfolded in complex ways related to social conditions across successive cohorts. The most pronounced changes are observed among women, who demonstrate an emerging disparity in BMI levels on the basis of education and urban versus rural residence.

Keywords

BMI China cohorts fundamental cause theory life course obesity rural/urban SES

Following the precedent of developed nations, China is expecting a continued rise in population weight. Current estimates suggest that more than 25 percent of its adults are considered overweight or obese (Cecchini et al. 2010), and though this overall prevalence statistic is lower than in most Western nations, the body mass index (BMI) growth rate in China is faster than in the United States, the United Kingdom, and Australia (Popkin 2008). Left unabated, the prevalence of overweight and obesity is expected to double before 2030 (Popkin 2008).

These general patterns of biomorphic change seem to be a common correlate of economic modernization. Developing nations in Latin America, Africa, and Asia have all experienced population-level increases in body mass (Cecchini et al. 2010). In each of these contexts, economic development has brought an increased availability of high-fat, energy-dense foods; the spread of sedentary occupations such as professionals and office workers; the adoption of motorized transportation; and the diffusion of passive leisure activities. Among developing nations, however, China stands out for the extent and pace of social change. The Chinese experience is one of unprecedented government involvement in the shaping of public health, leading to exceptional increases in life expectancy and suppression of infectious diseases during the second half of the twentieth century (Chen and Liu 2009). With its rapid, state-directed move toward industrialization and urbanization, China’s population of 1.3 billion people represents a crucial, though still understudied, context for examining the social determinants of individual health.

In the current study, we focus on trajectories of biomorphic change during China’s recent era of economic development (1989 to 2006). Specifically, we ask whether Chinese men and women born into different cohorts display differential growth rates of BMI and risk for overweight and obesity. Furthermore, we consider whether these biomorphic trajectories differ by fundamental social characteristics—household income, education, and urban versus rural residence—and whether the emerging social disparities tend to diverge or converge with age.

Several theoretical insights motivate this set of concerns. First, cohorts, long an important consideration in the field of demography, have come to represent an increasingly fertile issue in the study of individual health trajectories (Yang and Lee 2009). Cohort effects align with basic assumptions of the life course perspective, which emphasizes how people simultaneously exposed to historical experiences have potentially unique life pathways depending on the particular juncture of history with their age (i.e., the timing of events). The life course perspective also underscores the long view of people’s lives, whose trajectories through a given period are again shaped by their birth years. Second, fundamental cause theory (FCT) envisions social conditions as the underlying causes of health conditions in the population (Link and Phelan 1995). FCT has offered a cogent explanation for the persistence of socially patterned health heterogeneity over long stretches of historical time, but the theory has remained largely silent about cohorts, the other vital temporal dimension in which population health disparities unfold. We see a need for FCT to articulate a more explicit life course sensibility: a recognition that new birth cohorts may have distinctive affinities with evolving health risks and that they ultimately act as the transmitters of emerging health disparities. Below, we unpack this claim and then move on to a discussion of biomorphic change in the Chinese context and a presentation of the empirical results.

Background

FCT with a Life Course Sensibility

Historical time plays a central role in Link and Phelan’s (1995) FCT. One of the theory’s central claims is that health-enhancing technology and treatment options arise at specific historical junctures, and basic social conditions dictate who has access to those resources once they become available. The theory predicts that people with the greatest access to resources will be the first to take advantage of new knowledge and capitalize on medical innovation, leading to social inequality in prevalence of disease X. Over time, as these innovations become diffused more widely, disease X is no longer distributed so unequally in the population. But emergent knowledge and technology now arise to treat diseases Y and Z, thereby sustaining an overall pattern of social inequality in disease prevalence. Taken as a whole, the very ubiquity of population heath disparities across historical eras is seen as strong evidence to support FCT (Freese and Lutfey 2010).

Despite this clear illumination of historical time, two other temporal factors have not yet been given center stage in FCT’s explication. The life course perspective draws attention to these blind spots, cohort and age, as essential mechanisms for explaining social behavior (Alwin and McCammon 2003), which includes biomorphic change and the risk for overweight or obesity (Reither, Hauser, and Yang 2009). Principally, it may be that the rise and fall of emergent health disparities is not monolithic within a population during a specified period of historical time. Rather, these health issues could be manifested in unique ways among people born in different eras and progressing through different stages of the human life span. Without carefully examining age and cohort factors, most studies in the FCT tradition have implicitly assumed demographic uniformity in the unfurling of population health disparities. Assimilating age and cohort provides a more complete picture of how health conditions are socially distributed, particularly in cases in which early life development is consequential for lifelong disease risk (e.g., metabolic and endocrine processes and pathways; see Kuh and Ben-Shlomo 2004).

In the life course framework, aging effects refer to biological and psychological maturation, as well as passage through social roles and the experience of life events. Related to biology and biomorphic change, aging is associated with slowed resting metabolism and hormonal changes that lead to increased fat under conditions of stress (Holt and Balint 1993). Aging also signifies an opportunity to be exposed to risk factors (and variation in the duration of exposure) that influence biomorphic composition. Likewise, aging represents an opportunity for distinct pathways to play out over time and for a range of between-person advantages and disadvantages that may diverge or converge over time.¹

Cohort effects, on the other hand, represent a set of people experiencing the same historical events at the same age. Year of birth is important because the risks, opportunities, and constraints produced by society impinge on early life chances (Elder, Johnson, and Crosnoe 2003). Developmental trajectories are cumulative, meaning that the particular insults or advantages experienced early in life may influence susceptibility to health problems in middle or older age (Kuh and Ben-Shlomo 2004). Hence, membership in a particular birth cohort shapes individual aging via the set of historical contingencies it introduces for a set of people (Elder et al. 2003). Furthermore, people coming of age in a particular sociocultural milieu may adopt distinctive habits or perspectives that differentiate their birth cohort from those preceding or following it.

The presence of multiple cohorts also ensures that members of the population will have exposure to identical historical eras (e.g., market reform in China) at different stages of biological maturation (e.g., different stages of metabolic development). Earlier theorists such as Ryder (1965) maintained that social change is enabled by the constant influx of new birth cohorts that view the existing social conditions anew and whose minds are not yet fixed by their own life experiences. A similar argument can apply to the context of biomorphic change in modern China: The continued entrance of new bodies, still in formative and responsive metabolic phases, is an essential condition for triggering substantial population change in BMI and for shaping its nascent (or fluctuating) social gradients.

In this study, we observe one specific slice of historical time—the era of Chinese economic modernization from 1989 to 2006—but focus our attention on multiple cohorts of Chinese adults and their age-based biomorphic trajectories. Chinese men and women have likely experienced BMI change differently during this era depending on their years of birth. Our focus is on personal trajectories of BMI because biomorphic form is dynamic and modifiable over the course of adulthood. Indeed, our hope is to more closely align FCT with the life course perspective, which emphasizes a long view of people’s lives (Kuh and Ben-Shlomo 2004). We see the case of China during the era of modernization as an opportune context for this effort, as relatively recent cohorts (young adults) face exceptionally different conditions in the early stages of their life courses than did people born in the middle portion of the twentieth century. The particularities of modern Chinese history—its rapid ascension from undeveloped nation to industrialized giant—give us a clear view of how social disparities “unfold” cohort by cohort over the course of nearly two decades.

BMI in China: Overview and Hypotheses for Cohort and Age Effects

Cohorts likely play an important role in understanding how body mass increased in China during the era of economic modernization. Although mean levels of BMI have risen considerably since the early 1980s (Wu 2006), the changes have not likely been experienced equally by all birth cohorts of Chinese people. Specifically, recent cohorts have increased exposure to obesity-conducive lifestyles and are more likely to have broadly adopted Western eating and leisure habits. This leads us to hypothesize a positive association between recent cohort and BMI, net of age effects.

In light of past research on life course changes in obesity (Reither et al. 2009), we expect that age has a curvilinear association with BMI, net of cohort effects. BMI likely increases until middle age, but bone and muscle loss at older age should contribute to declining body mass and produce a curvilinear age-BMI relationship. Moreover, in keeping with our focus on birth cohorts, we anticipate nontrivial cohort variation in age-based BMI growth. In particular, we hypothesize more rapid age-related BMI growth for more recent cohorts. Correspondingly, for people at older ages in which BMI is expected to decrease, the deceases should be decelerated for each successive cohort.

Cohort differences by socioeconomic status (SES) and urban versus rural residence

Returning to the premise of FCT, we should expect that BMI levels are closely related to basic social conditions. BMI is a somewhat different case, however, from many of the health conditions that are often assessed in empirical tests of FCT. Although cancer or diabetes is an unequivocally unfortunate condition, BMI growth in the population could represent a good or a bad turn of events. That is, higher BMI could signify increasing risk for overweight and obesity, or it could be a positive health development if it shifts those in the lower part of the BMI distribution away from the risk for underweight. In underdeveloped and developing nations, poor people have little access to high-fat foods and motorized transportation. High BMI levels are found largely among the privileged. In developed nations, this pattern becomes reversed. With unhealthy foods plentiful, inexpensive, and mass marketed, wealthy and highly educated people are best positioned to negate these environmental inducements and to maintain recommended weight levels. In line with this contrast, as countries undergo the transition to modernization, the correlations between BMI and social class often reverse (McLaren 2007).

Without taking cohorts into account, however, the shifting nature of these social disparities is likely obscured. That is, birth cohorts that have already lived much of their lives before the advent of economic modernization may demonstrate a positive association between SES and BMI trajectories, whereas the pattern may be just the opposite among cohorts growing up in the period of economic modernization.² Related to the current study, people born in birth cohorts prior to the advent of development may be more likely to have higher BMIs if they are high in SES, net of age-related BMI change. People born in the 1960s and later, on the other hand, would be exposed to the changing social conditions associated with economic modernization during their childhood, teens, and 20s. During this period of modernization (after 1980), increases in BMI likely became increasingly concentrated among the lower SES segments of the population (Jones-Smith et al. 2011). Therefore, we hypothesize a negative interaction between SES and recency of birth cohort membership for levels of BMI, net of aging × SES effects.

Perhaps more so than in most Western nations, place of residence in China represents another crucial dividing line for population health disparities. Urban dwellers, for instance, have access to far superior medical care and exhibit better self-rated health than those in rural China (Chen, Yang, and Liu 2010). That being said, the pace of modernization is much faster in dense urban areas than in rural outposts, so city dwellers may be especially prone to higher levels of BMI and escalations in BMI growth as Chinese society becomes infiltrated with Western diets and lifestyles and professional and manufacturing jobs. Although receding somewhat, traditional foods and agricultural practices remain far more prevalent in rural China than in highly populated areas (Popkin 1999). For the purposes of this article, it is also important to note that, much like SES, rural versus urban residence may also have a changing role in BMI growth according to birth cohort. As Ryder (1965) noted, young people, especially those “in the big cities,” are in many ways the propellants of social change, because new conditions can form an “experiential chasm between cohorts” (pp. 849–50). Accordingly, we may expect more recent Chinese cohorts to have the most direct exposure to shifting social conditions and/or to be most amenable to adopting BMI-amplifying Western practices. Both conditions would imply becoming heavier. We therefore hypothesize that though urban residence is generally associated with higher BMI, there will be a positive interaction between urban residence and recent cohort membership, net of aging × urban residence.

Summary of Hypotheses and Additional Analytic Considerations

In summary of the preceding discussion, we evaluate four central hypotheses:

Hypothesis 1: There will be a positive association between cohort recency and BMI, net of aging effects.

Hypothesis 2: Age-related increases in BMI are intensified for more recent Chinese cohorts.

Hypothesis 3: There will be a negative interaction between recent birth cohort membership and SES and for predicting BMI levels, net of aging effects.

Hypothesis 4: There will be a positive interaction between urban residence and cohort recency for predicting BMI levels, net of aging effects.

In addition, we expand hypotheses 3 and 4 to examine how cohort differences change over time. A life course–attuned FCT would not only consider whether people born during different eras experience distinct health trajectories related to their position in society; it would also be interested in whether disparities jointly related to cohort and social position magnify or diminish as people age. Drawing from the studies that examine a cumulative advantage/disadvantage versus age-as-lever explanation for health trajectories over time (Yang and Lee 2009), we test a series of three-way interaction terms (cohort × social conditions × age) to investigate this issue. These interaction terms will help us understand whether the cohort-specific social disparities tend to grow or diminish over time.³

Given sex differences in metabolism and in norms of physical attractiveness (McLaren 2007), we also assess hypotheses 1 to 4 separately for men and for women (cf. Jones-Smith et al. 2011). Regarding the issue of physical attractiveness and body image, it may be the case that hypotheses 3 and 4 are observed only among women. That is, if weight gain has become an inauspicious issue more acutely among women than among men (McLaren 2007), then the emergent social gradients may be most pronounced among the former.

Finally, our analyses examine two biomorphic trajectories: BMI and overweight and obesity. These are clearly interrelated constructs, as overweight and obesity is simply a dichotomous category demarcating some threshold of the BMI continuum (e.g., ≥25 kg/m²). Overweight and obesity, however, has yet to besiege China (and other developing Asian nations) with the same intensity as in North America. As the BMI distribution continues to push upward, increasing numbers of Chinese adults are expected to fall within the range of overweight or obesity. With these considerations in mind, we anticipate that the hypothesized patterns may be more consistently observed for the linear BMI outcome than for the actual expression of overweight and obesity. China is growing heavier, but the early indicators of population health trends may be more apparent in the initial increase in BMI levels than in the prevalence or incidence of overweight and obesity.

Data and Methods

Sample

All analyses for this study are based on the China Health and Nutrition Survey (CHNS), the most comprehensive and trustworthy data source for documenting health change in China during the era of economic modernization. A detailed description about the aims and scope of the survey can be found at http://www.cpc.unc.edu/projects/china. In brief, study participants were selected from a multistage, random cluster of eight provinces. The provinces were stratified by income (low, middle, and high), and four counties from each province were randomly selected from a weighted sampling scheme. In addition, the research team included the provincial capital and a lower income city in the sampling frame.

As of this writing, CHNS respondents have been tracked through seven rounds of longitudinal data collection: 1989, 1991, 1993, 1997, 2000, 2004, and 2006. This, along with the wide range of age cohorts represented (adults born as early as 1898 and as recently as 1988), makes the CHNS a unique data source for studying cohort and age change. For the sake of consistency, we focus on adult BMI only and restrict the sample to adults aged 18 years and older at each wave (n = 18,380). The average age of the sample was 31.68 years for the first wave (1989) and 49.10 years for the final wave (2006).

Measures

BMI

Data for weight in kilograms and height in centimeters were available for each wave of the CHNS. These measurements were conducted by trained health workers using a portable stadiometer. Weight was measured in light clothing without shoes on a calibrated beam scale. BMI was computed as weight in kilograms divided by height in meters squared. In addition to this linear operationalization of BMI, we created a dummy variable to denote men and women who were considered overweight or obese (BMI ≥ 25 kg/m²).

Age and cohort

The ages of respondents in the baseline sample range from 18 to 79 years and are measured in single-year increments. To make parameters most interpretable, we centered age at the median age. We constructed nine 10-year birth cohorts, but with several exceptions. First, the initial cohort was broadened over a greater range of birth years (1898 to 1920) to ensure a sufficient number of subjects. Second, the 1951 to 1955 and 1956 to 1960 cohorts were separated because of the historical significance of the Great Leap Forward and the Three-Year Famine (see Chen et al. 2010). In the judgment of prior scholars, both of these events were significant enough to merit distinct cohort categorization. We code cohorts from 1 to 9 from the earliest (oldest) to the most recent cohort (youngest).

Social conditions

We focus on three key social indicators in China: educational attainment, household income, and urban residence. Education is divided into three levels: no education or primary school, junior high school, and senior high school or higher. Household income, inflated to 2006 Chinese yuan, was categorized into quartiles, with the lowest quartile as the reference category. Both education and household income are used to operationalize SES.

Urban residence is the other social condition hypothesized to intersect with cohort membership and explain variance in BMI trajectories. This variable was coded as a dummy variable (1 = urban residence, 0 = rural residence).

Controls

Covariates in our multivariate analyses include current smoking status, sex, marital status, working status, and province. Respondents who report current smoking are coded 1 and 0 otherwise. Consistent with another recent study using the CHNS data, we include the smoking covariate only in the analysis of men because of the low prevalence and the higher rate of missing data on this measure among women (Jones-Smith et al. 2011). Marital status is coded 1 for currently married and 0 otherwise. To control for potential regional variation in BMI, we control for the effects of province with dummy variables: coastal (Shandong and Jiangsu), northeast (Liaoning and Heilongjiang), inland (Henan, Hubei, and Hunan), and the southern mountain region (Guangxi and Guizhou). Table S1 (available in the online supplement) presents summary statistics of all variables used in the analyses for all waves combined.

Analytic Strategy

Using multiple birth cohorts in seven waves over a 17-year (1989 to 2006) period allows us to examine cohort analysis with more accumulated information on age and cohort than a single longitudinal cohort study (for an additional example, see Chen et al. 2010). We applied growth-curve models into the analysis of panel data to examine trajectories of BMI and overweight and obesity over Chinese adulthood. Equations shown below assume a linear growth model, though we also estimate binary growth-curve models, which apply a logistic link function and enable us to predict the risk for overweight and obesity over time. Growth-curve models account for within-person clustering effects and can handle an unbalanced survey design, that is, an inconsistent number of observations per respondents (Singer and Willett 2003). In the level 1 model, time-varying BMI scores were regressed on age terms (linear and quadratic) for each respondent. These regressions produced random intercept and slope estimates that describe subject-specific growth curves. At level 1, BMI scores at time t are nested within individuals (i):

Y_{t i} = π_{0 i} + π_{1 i} {Age}_{t i} + π_{2 i} {Age}_{t i}^{2} + e_{t i},

where Y_ti is the BMI for respondent i at time t, for i = 1, …, n and t = 1, … 7. Age_ti is the age of respondent i at time t. The intercept π_0i is the estimated BMI of respondent i at the median age and when all other covariates are controlled at their means; π_1i is the estimated growth rate per year of age for respondent i and is assumed to follow a normal distribution. Note that the level 1 specification incorporates X_pti time-variant covariates (e.g., working status, marital status)

The level 2 models for the intercept and for the linear rate of change (age) represent the average trajectory for each of the Chinese cohorts along with several interaction terms. Such terms include the interaction between cohort and social conditions (i.e., education, household income, and urban residence), the interaction between age and social conditions, and the three-way interaction among age, cohort, and social conditions. Also included in the model are time-invariant covariates (e.g., region). The equation takes the form:

π_{0 i} = β_{00} + β_{01} {Cohort}_{i} + β_{02} {SocialConditions}_{i} + β_{03} Cohort \times SocialConditions + r_{0 i},

π_{1 i} = β_{10} + β_{11} {Cohort}_{i} + β_{12} {SocialConditions}_{i} + β_{13} Cohort \times {SocialConditions}_{i} + r_{1 i},

where Cohort_i is coded 0 to 8 for cohorts 1 to 9, and social conditions include educational attainment, household income, and urban residence. In equation 2, β₀₀ is the estimated BMI at the median age in cohort 1 for the reference group (rural residence, primary or lower educated, and the first quartile family income) at time 1 (i.e., in 1989), β₀₁ is the average difference in BMI between cohorts in time 1, β₀₂ is the differences in BMI across people in varying social conditions in cohort 1, and β₀₃ is the average effect of the interaction between cohort and social conditions. In equation 3, β₁₀ is the estimated rate of change in BMI in cohort 1 at time 1, and β₁₁ and β₁₂ are coefficients for the effects of cohort and social condition on the rate of change with age, respectively, whereas β₁₃ represents interactive effect between cohort and social conditions on the rate of change (i.e., the three-way interaction among age, cohort, and social condition).

In a growth-curve model, level 1 analysis shows within-individual change in which the outcome is a function of time indicator. Because of the problem collinearity, the time indicator in the level 1 model should be either age or wave (Singer and Willett 2003). For the purpose of analysis, we choose age as a time indicator. Linear models used full maximum likelihood estimation in the mixed procedure in Stata 12.0 (StataCorp 2011). Binary growth-curve models used Mplus 6.12 (Muthén and Muthén 2011).

Results

Table S2 (available in the online supplement) presents the mean BMI scores for the study sample over all adulthood ages and across all birth cohorts represented in the CHNS data. As an overall pattern, mean BMI scores increase over adulthood, but at any given age, BMI scores are higher in the more recent cohort. To systematically examine the contribution of age and cohort effects in producing these trends, we estimated the growth-curve model described in detail above. Models for women are presented in Table 1 and models for men in Table 2.

Table 1.

Results from Growth-Curve Models of Age, Cohort, and Social Indicators on BMI for Chinese Women, Aged 18 Years and Older, CHNS 1989 to 2006 (n = 9,597)

	BMI		Overweight and Obese (BMI ≥ 25 kg/m²)
Variable	β	SE	β	SE
Fixed effects
Intercept β₀₀	19.710***	.286	7.324***	.656
Age β₁₀	.032*	.016	.134**	.054
Age squared β₂₀	−.001*	.0002	−.004***	.001
Cohort β₀₁	.734***	.050	.993***	.110
Age × cohort β₁₁	.014***	.003	.010	.009
Education β₀₂ (reference: primary school or lower)
Junior high school	.316	.387	1.916*	.740
Senior high school or higher	1.261**	.484	3.230***	.894
Household income β₀₂ (reference: first quartile)
Second quartile	.004	.305	.245	.574
Third quartile	.231	.323	.060	.587
Fourth quartile	.972**	.371	1.897**	.643
Urban residence β₀₂	2.382***	.336	3.391***	.580
Two-way cohort interactions β₀₃
Cohort × junior high school	−.085	.066	−.346**	.125
Cohort × senior high school or higher	−.318***	.082	−.687***	.151
Cohort × second quartile	.013	.058	−.036	.105
Cohort × third quartile	−.011	.059	.008	.106
Cohort × fourth quartile	−.119	.065	−.284*	.113
Cohort × urban residence	−.330***	.059	−.508***	.102
Social conditions on growth rate β₁₂
Age × junior high school	.053***	.016	−.015	.030
Age × senior high school or higher	.041*	.019	.000	.038
Age × second quartile	−.008	.009	−.017	.019
Age × third quartile	−.009	.010	.003	.020
Age × fourth quartile	−.037***	.011	−.024	.021
Age × urban residence	−.012	.011	.010	.022
Cohort × social conditions on growth rate β₁₃
Cohort × age × junior high school	−.008**	.002	−.001	.005
Cohort × age × senior high school or higher	−.009**	.003	−.008	.006
Cohort × age × second quartile	.002	.002	.002	.003
Cohort × age × third quartile	.004*	.002	.000	.003
Cohort × age × fourth quartile	.006***	.002	−.001	.003
Cohort × age × urban residence	.001	.002	−.007	.002
Controls
Married	.106	.065	.346**	.122
Currently working	−.205***	.046	−.346***	.076
Region (reference: coastal)
Northeast	.052	.098	−.086	.157
Inland	−.558***	.086	−.902***	.143
South	−1.432***	.092	−2.175***	.169
Random effects
Level 1: within person	5.256***
Level 2: intercept	6.503***		11.146***
Growth rate	.003***		.015***

Note: BMI = body mass index; CHNS = China Health and Nutrition Survey.

p < .05. **p < .01. ***p < .001 (two-tailed tests).

Table 2.

Results from Growth-Curve Models of Age, Cohort, and Social Indicators on BMI and Overweight for Chinese Men Aged 18 Years and Older, CHNS 1989 to 2006 (n = 8,783)

	BMI		Overweight and Obese (BMI ≥ 25 kg/m²)
Variable	β	SE	β	SE
Fixed effects
Intercept β₀₀	19.734***	.285	9.918***	.683
Age β₁₀	.127***	.015	.318***	.034
Age squared β₂₀	−.002***	.0002	−.006***	.001
Cohort β₀₁	.657***	.051	1.273***	.118
Age × cohort β₁₁	−.007**	.003	−.021***	.006
Education β₀₂ (reference: primary school or lower)
Junior high school	.277	.321	.495	.682
Senior high school or higher	.214	.386	.600	.772
Household income β₀₂ (reference: first quartile)
Second quartile	−.362	.294	−.536	.669
Third quartile	−.309	.311	.391	.679
Fourth quartile	−.293	.354	.907	.746
Urban residence β₀₂	.917**	.316	3.694**	.623
Two-way cohort interactions β₀₃
Cohort × junior high school	−.002	.057	−.020	.120
Cohort × senior high school or higher	.050	.068	.021	.134
Cohort × second quartile	.074	.056	.100	.122
Cohort × third quartile	.079	.057	−.047	.122
Cohort × fourth quartile	.112	.063	−.077	.130
Cohort × urban residence	−.034	.056	−.423***	.107
Social conditions on growth rate β₁₂
Age × junior high school	−.016	.012	−.020	.024
Age × senior high school or higher	−.029*	.014	−.051	.028
Age × second quartile	.011	.010	.032	.022
Age × third quartile	.005	.011	.013	.023
Age × fourth quartile	.005	.011	.000	.024
Age × urban residence	.013	.012	−.034	.023
Cohort × social conditions on growth rate β₁₃
Cohort × age × junior high school	.005*	.002	.007	.004
Cohort × age × senior high school or higher	.008**	.002	.011*	.005
Cohort × age × second quartile	.001	.002	−.004	.004
Cohort × age × third quartile	.002	.002	−.007	.004
Cohort × age × fourth quartile	.003	.002	−.004	.004
Cohort × age × urban residence	.002	.002	.003	.004
Controls
Married	.090	.061	.427**	.134
Currently working	−.191***	.051	−.215*	.098
Region (reference: coastal)
Northeast	−.099	.091	−.215	.161
Inland	−.801***	.081	−1.254***	.150
South	−1.734***	.086	−2.879***	.184
Current smoker	−.308***	.036	−.473***	.073
Random effects
Level 1: within person	4.273***
Level 2: intercept	5.045***	11.192***
Growth rate	.003***	.010***

Note: BMI = body mass index; CHNS = China Health and Nutrition Survey.

p < .05. **p < .01. ***p < .001 (two-tailed tests).

Beginning with Table 1 (women), the first set of columns show results from the linear growth-curve model, while binary growth-curve models are presented in the right half of the table. The cohort variable is positive and significant in the linear growth model, indicating that BMI levels have increased for each successive cohort by an average of .73 units, lending support to our first hypothesis. Moreover, the growth rate in BMI was greater for each successive cohort (β = .014, p < .001). This cross-cohort gradient in BMI growth trajectories supports the second hypothesis among women: age-related increases in BMI are intensified for more recent Chinese cohorts. A visual depiction of this age and cohort effect is presented in Figure 1a. Interestingly, the predicted values in Figure 1a indicate that although BMI levels are increasing, the average values do not reach the threshold for overweight (25 kg/m²), even for the highest observed data point (the 1956 to 1960 cohort at 48 years of age). Indeed, the binary growth-curve model shown in the rightmost columns reveals a positive and significant cohort effect for risk for overweight and obesity, but not a significantly increasing risk for that condition over time for more recent cohorts (.010, p > .05). Taken together, the linear and binary models suggest a clear biomorphic shift that presages, but has not yet resulted in, a spike in obesity prevalence among younger Chinese women.

Figure 1.

Predicted Trajectories of Body Mass Index (BMI) for Women

Regarding the third hypothesis (a negative interaction between recent birth cohort and SES), we see some support for the cohort-varying effect of education on the mean level of BMI, but only a marginally significant result for the effect of income. Evaluating the third hypothesis requires attention to several coefficients: main effects of SES variables (education and income), interactions between SES variables and birth cohort, interactions between SES variables and time (i.e., age), and three-way interactions among SES variables, birth cohort, and time (i.e., age). Considering that higher SES is associated with higher BMI levels among the earliest cohort (main effects of highest education and highest income quartile = 1.261 [p < .01] and .972 [p < .001], respectively), the negative interactions between higher SES variables and cohort imply that this positive association decreases across successive cohorts. However, this does not necessarily translate into reducing the SES gap within cohorts, because age-related SES effects can be reversed in more recent cohorts. For instance, the highest education × age coefficient (β = .041, p < .05) becomes reversed in more recent cohorts (three-way interaction coefficient = −.009, p < .01). Similarly, the significant interaction between highest income and age (β = −.037, p < .001), and the three-way interaction among age, cohort, and highest income (β = .006, p < .001) indicates that income BMI differentials narrow with age, and this trend of convergence strengthens in successive cohorts. In other words, the association between SES and BMI may actually change in direction (from a positive to a negative association) in more recent cohorts.

The above interpretation is well demonstrated in Figures 1b and 1c; the association between SES and BMI clearly shifts in successive cohorts. As shown in Figure 1b, highest education attainment is positively associated with BMI levels and growth in earlier cohorts (prior to 1955). An inverse relationship between highest education level and BMI emerges and then continues in cohorts 7 (1961 to 1970) and 8 (1981 onward). Similarly, Figure 1c indicates that the positive association of income with BMI, though not BMI growth, becomes marginal by the 1961 to 1970 cohort and reverses in direction for Chinese women born after 1971. This inverse relationship is most prominent for the most recent cohort (1981 to 1988).

Contrary to hypothesis 4, the main effect of urban residency was positively associated with BMI levels in the earliest cohort (β = 2.832, p < .001), but this positive effect decreases across successive cohorts as indicated by the negative urban × cohort coefficient (β = −.330, p < .001). The three-way interaction including age was nonsignificant. Together, these findings suggest that more recent female urbanite cohorts have higher mean BMI scores, though they are not getting heavier at a faster rate. Figure 1d shows the predicted values generated from the growth curve equation.

Overall, the results from the binary growth-curve model for predicting overweight and obesity (rightmost columns in Table 1) were largely similar to the linear growth-curve findings. Several distinctions, however, merit mention. The highest income × cohort term was significant and negative (β = −.284, p < .05), whereas this two-way interaction term was nonsignificant in the linear model. Also, two- and three-way interactions incorporating the age variable were not statistically significant in the binary growth-curve model. In other words, SES and cohort × SES interaction effects are apparent only in the intercept equation (mean levels of overweight probability), not in the growth-rate equation (trajectories of overweight probability). This pattern may arise if intraindividual change in overweight risk is less pronounced than intraindividual variability in BMI levels; on the other hand, the discrepancy in findings could reflect the fact that emerging biomorphic trends are likely more detectable when conceived in continuous form rather than identified by a relatively high binary cut point (i.e., overweight or obese).⁴

Turning to the results for men (Table 2), we also found a positive association between cohort recency and BMI intercept (β = .657, p < .001). Distinct from the female subsample, however, the interaction between age and cohort is negative (β = −.007, p < .01). This fails to support hypothesis 2 among men.⁵ Of note, this negative coefficient for the interaction between age and cohort was nonsignificant and positive until all covariates and the full series of interactions were included in the model (supplementary analyses are available upon request). Results from the fully adjusted model among men are presented in Figure 2a.

Figure 2.

Predicted Trajectories of Body Mass Index (BMI) for Men

We did not find any significant cohort variation in the association between SES and BMI among the male sample. However, three-way interaction terms including age, cohort, and education were positive and significant (e.g., β = .008, p < .01, for the highest education category), and a significant similar story emerged in the binary growth-curve model (β = .011, p < .05). These findings suggest cohort-specific educational differences in biomorphic growth trajectories and the risk for overweight and obesity over time. Paired with the negative coefficient for the age × highest education term (β = −.029, p < .05), the three-way interaction suggests that the educational differential tends to increasingly narrow with age over for successive cohorts. Figure 2b displays the predicted education-specific age trajectories of BMI for selected cohorts. The figure indicates that BMI levels were higher overall for the most educated men, but these education-based differentials have decreased in successive cohorts. For the most recent cohort, the gap has closed entirely, perhaps suggesting an imminent SES reversal in biomorphic trajectories like the one observed among women.

In contrast to education, there were no significant interaction effects observed for income among men (Figure 2c), though the general patterns are similar to the education findings in Figure 2b. This is consistent with the rise of an educational gap prior to that of an income gap observed among women.

Another noteworthy finding from the analysis of the male sample is that urban residence had a positive association with both mean BMI and risk for overweight and obesity, but its intercohort variation was significant only in the binary growth-curve model of overweight (β = −.423, p < .001). Contrary to hypothesis 4, the effect is negative; this interaction effect suggests that the positive association between urban residence and overweight or obese status among Chinese men has decreased across successive cohorts. Figure 2d presents the predicted BMI values from the linear growth models in Table 2. Although the cohort × urban residence term was nonsignificant in the linear model, the shrunken gap between rural and urban males in the most recent cohort is clearly apparent in Figure 2d. In all, the gender difference in cohort and urban versus rural patterns emphasizes the importance of studying male and female BMI trajectories separately. Perhaps future cohorts of rural men will surpass their urban counterparts in BMI (as has been the case among women), but future data will be needed to assess this possibility.

Discussion

This study considered four hypotheses related to BMI change in China. The time frame under consideration, 1989 to 2006, was one of economic growth and the spreading influence of Western market forces. Like other developing nations, China is undergoing a biomorphic transformation in which bodies increase in mass and overweight and obesity prevalence rises (Cecchini et al. 2010). Our analysis aimed to shed light on how cohort membership influences men’s and women’s BMI trajectories during the era of market reform and how cohort-based biomorphic change unfolds along a social gradient.

Our initial set of hypotheses was largely descriptive in nature. We first evaluated whether the Chinese situation parallels the Western context, in which subsequent generations are “getting heaver, younger” (Lee et al. 2010). Among women, belonging to more recent Chinese birth cohorts is associated with higher levels of BMI and a greater risk for overweight and obesity. Furthermore, the nature of age-related BMI growth is accelerated for each successive cohort. Specifically, BMI levels have increased at an average rate of .734 units for every successive generation, and the age-related BMI growth rate is .014 greater for a successive cohort than a prior cohort. Hypothesis 1 and 2 were therefore supported among women. For men, each successive cohort has an average BMI .657 units higher and a risk for overweight 3.5 times greater (exponentiated logit coefficient) than its predecessor. This supports hypothesis 1. On the other hand, we fail to find support for hypothesis 2 among men, because the rate of BMI growth and the risk for overweight does not increase with age for each successive cohort of Chinese men.

In general, these findings fall in line with cohort-based patterns observed in the United States (Lee et al. 2010; Reither et al. 2009). Upon evaluating the similarity of cohort trends across these westernized societies, Reither et al. (2009) noted the possibility of declining average life spans due to chronic morbidity and speculated that “the health and longevity of populations in other developed nations could also be in jeopardy” (emphasis added). Cohort-based research on developing nations, China in particular, is only beginning to emerge (Jones-Smith et al. 2011) but will be important for assessing when and to what extent Reither et al.’s forecast will apply in other global contexts. Future work is needed to assess population-level patterns in India, Pakistan, Brazil, and other developing nations.

Beyond describing intracohort variability in biomorphic change, the second set of hypotheses was more theoretically motivated and examined how social conditions intersect with BMI trajectories. Specifically, we expected a negative interaction between birth cohort recency and several indicators of SES (education and income) and a positive interaction between cohort recency and urban residency. The rationale for posing hypotheses 3 and 4 comes from several basic propositions: (1) Social conditions are at the root of how health outcomes are distributed in a population (Link and Phelan 1995), (2) social disparities in a given health issue emerge in some temporal context, and (3) a birth cohort encases individuals’ life trajectories in a channel that exposes unique groups of people to distinct social conditions at nearly simultaneous stages of development. Together, these propositions imply that cohort turnover is a temporal mechanism for a social health gradient emergence. In effect, we sought to link FCT with the sensibilities of a life course perspective.

Although uniting these two traditions proved useful in positioning our research questions, the findings that tested our hypotheses revealed considerable levels of complexity in how social disparities unfold over time and between cohorts. First, key differences surfaced for the effects of BMI between household income and education, our two indicators of SES. Second, these indicators of SES operated somewhat distinctly along gender lines.

Among women, education robustly modified the birth cohorts effects according to the patterns articulated in hypothesis 3. Using birth cohort as a linear term in our growth model, we found statistically significant negative interactions between cohort (recency) and each level of education relative to the reference group (no schooling or primary school education).⁶ This means that although each successive cohort is at increasing risk for obesity and has greater growth in BMI levels, the effect is partially offset among people with higher levels of formal education. Perhaps most interesting, however, is that the association between education level and BMI changes in direction, not only in magnitude, across successive cohorts of women. Linear growth-curve models indicate that in birth cohorts such as 1951 to 1955, the most highly educated women have considerably higher BMI levels than do the least educated Chinese women; by 1981 and onward—cohorts that demonstrate the steepest increases in BMI over time—the trend is reversed, and the least educated women have higher BMI levels.

Although the education effects among women were clearly in line with our expectations, the effects of income did not provide unequivocal support for hypothesis 3. In line with the hypothesis, the risk for overweight and obesity was indeed lower for women in the highest income quartile in more recent cohorts. At the same time, there was no evidence of less growth in the linear BMI term among the highest income women in the most recent cohorts. Figures 1b and 1c show that the education and income patterns are generally similar, but the social gradient is much more apparent in the case of educational differences.

Understanding the distinction between education and household income can be aided by a thoughtful recent review article by McLaren (2007). Synthesizing empirical associations between SES and obesity from a large body of international research, McLaren suggested that education is more closely aligned with conceptions of cultural capital than other dimensions of SES, thereby sensitizing people, particularly women, to socially favored conceptions of thinness and attractiveness. Whereas high household income entails availability of processed food and sedentary forms of entertainment and leisure, material resources may have less to do with factors that most directly work against harbingers of the “obesity epidemic” (Wang et al. 2006). These countervailing factors—cultural valuation of thinness or an internalized sense of achievement and agency over health—are more influenced by educational status (McLaren 2007; Mirowsky and Ross 2003). Given the intensifying ubiquity of unhealthy practices and technologies in a globalizing world, the most beneficial social resources may be those that help withstand the tides of BMI growth. For current and future Chinese cohorts of Chinese women, education may therefore represent the key axis of health inequality.

The men in our analyses differed somewhat in their expression of the SES gradients. Fewer of the interactions tested were statistically significant in the male subsample, but the emerging trend is toward a narrowing of the social gradient in BMI levels. Among women, there was some evidence of an SES reversal whereby the intercohort gap closed and then swapped direction of association. It is too early to tell whether Chinese male cohorts will repeat the progression of their female counterparts. Extrapolating Figure 2’s findings to future cohorts may imply that socioeconomic gradients among men are inevitable and only lagging behind the emerging disparities observed among women. On the other hand, if men are exposed to less pressure than women to maintain a low-BMI physique, then it makes sense that high-SES men would be less motivated to leverage their resources to offset fattening influences in the interest of remaining thin. Accordingly, it may be that low-SES Chinese men will not continue to increase in BMI relative to their higher SES counterparts; perhaps men across the social status distribution will equally experience the continued upward biomorphic shift. This possible gender distinction would add further complexity to the issue examined in the current article: how social conditions, as fundamental causes of health, unfold in a cohort-successive temporal sequence. Cultural sensitivities about body image are clearly linked to gender, and these cultural tastes and preferences may also be quite important for understanding social gradients in health (Freese and Lutfey 2010). Nevertheless, this set of connected issues has not yet been thoroughly fleshed out in FCT.

In contrast to the SES indicators, similar urban versus rural patterns were found among both men and women (though the effects were most robust among the latter). Interestingly, these findings contradicted our fourth hypothesis, which anticipated that urban residence would more strongly associated with BMI levels and risk for overweight and obesity in the most recent cohorts. In fact, the opposite pattern emerged for men and women. Urban living, though associated with higher overall BMI, was associated with reduced BMI levels (among women) and with lower risk for overweight and obesity in more recent cohorts. The growth rate in these outcomes, however, was nonsignificant. Hypothesis 4 was rooted in the assumption that urban dwellers are more likely to adopt Western practices, including higher fat and sugar diets, sedentary hobbies, and reliance on motorized transport. Yet we found that recent cohorts gain less weight in the city despite the environmental inducements of the urban setting. This finding is quite instructive in its similarity to the education patterns (among women in particular).

Taken together, the evidence suggests that urban residence in China, especially for women, represents a key basis of social status likewise related to effective control over life circumstances and the purposeful attainment of health-enabling choices (Mirowsky and Ross 2003). Other recent research points to the increasing urban-rural divide in overall health, noting that the cooperative medical schemes, once instrumental for providing health care to rural people, have been on the decline during the recent period of economic modernization (Chen et al. 2010). The current study builds on the observation that emergent Chinese health inequalities are becoming tightly linked to place of residence; though the rural environment has traditionally been amenable to lower BMI and risk for overweight and obesity, recent cohorts are becoming more apt to experience high levels of BMI if they dwell outside cities. Although there seems a certain paradox to this pattern, the trend is sensible if city residence affords higher status and resource access, enables people to exercise more control over their health, and produces heightened sensitivity to modernized images of thinness and attractiveness. All told, the emerging body-mass disparities in China seem more related to elements of social status most related to cultural capital, including education and urban residence, than to the ownership of material resources and wealth.

The biomorphic form of Chinese men and women entails a complex range of influences that are associated with basic sociological variables. We undertook this analysis to better understand how social disparities unfurl across birth cohorts, our ultimate aim being a more explicit integration of FCT with the life course perspective. Social disparities in health certainly emerge in historical time, coinciding with treatment innovations, better technologies, and increased knowledge about health (Freese and Lutfey 2010). But they also emerge through the temporal process of cohort turnover, as different groups of people are exposed to unique conditions at distinct stages of biological development. It is important to note that health disparities less tethered to metabolism may unfurl more uniformly across different age groups and cohorts. For example, Miech (2008) noted that cocaine use became strongly concentrated among low-SES Americans, of multiple cohorts, around 1990. Biomorphic trajectories, on the other hand, are critically influenced by the physiological malleability of early life (Lee et al. 2010), the subsequent pathways displaying considerable path dependency due to metabolic homeostasis. Accordingly, the cohort-staggered pattern of disparity emergence we observe in BMI may generalize more so to some conditions than to others.

High-fat, sweetened foods, motorized transport, and passive forms of leisure activity requiring minimal physical exertion have become fairly readily accessible and/or prevalent behavioral choices in China (cf. Popkin and Gordon-Larsen 2004). Yet the extent of their assimilation and the extent of their health effects are not uniform across all segments of the Chinese population. The same can be said of any other mechanism that potentially explains cohort variability in BMI. As an underlying determinant of population health patterns (Link and Phelan 1995), social conditions modify cohort biomorphic patterns by structuring basic constraints, risks, and opportunities. We identified SES and urban versus rural residence as two social conditions relevant for health and social change in China (Chen et al. 2010). Our results suggest that education, household income, and urban residence are indeed associated with BMI and with overweight and obesity but that these variables can differ in importance—and, as exemplified by Chinese women, can even take a complete turnabout in their direction of association. Future expression of biomorphic social disparities will likewise continue to unfurl in the temporal context of cohort succession.

In closing, several limitations of this study should be acknowledged. First, this study did not address explicit mechanisms of social patterns in BMI and its trajectories, that is, whether differential BMI level associated with social conditions result from unmeasured or omitted variables, such as the physical activities, dietary habits, total energy intake, and access to health-enhancing resources and information. The scope of the study and the lack of related data (particularly measures that are available in each wave of the CHNS) does not allow for explicitly and empirically specifying the underlying cause of the cohort variations and social patterns in BMI we observed in China.

Second, in a growth-trajectory model, such as the one specified in this article, age and time period are essentially the same variable. Similar to others using this approach, we did not explicitly estimate period effects. With longitudinal panel data, period and age are perfectly collinear terms (Chen et al. 2010; Yang and Lee 2009). Future research could be aided by using time-series cross-sectional data and incorporating period effects into the analysis. This strategy would more clearly untangle period, cohort, and age effects on BMI. As the data become available, such an analysis will become quite valuable given China’s increasing trend toward overweight and obesity (Popkin 2008).

Finally, this study focused only on adults. It is equally important to examine BMI or obesity among very recent cohorts that include children and adolescents.

Although ours is not the first study to focus on cohorts as a critical factor in understanding population BMI trends, little research has focused on developing nations—and one as exceptional and as influential as China (but see Jones-Smith et al. 2011). Understanding the nature of these cohort trends and how they intersect with social conditions represents not only an important way to think about fundamental causes but also a pressing public health issue. Higher BMI levels at earlier ages are a problem because they extend the duration of exposure to a health risk factor. Prior research demonstrates that greater duration of obesity, net of timing of onset or severity, increases the risk for diabetes (Brancati et al. 1999) and generates more hospital utilization later in life (Schafer and Ferraro 2007). This study shows that China, like the West, will confront a trend of lengthened durations of higher BMI levels for younger birth cohorts. Furthermore, the effects will be most pronounced among those with low education and dwelling in rural settings.

Footnotes

Acknowledgements

We are grateful to the JHSB editor and reviewers for their constructive critiques of an earlier version of this article. Data were made available through the Carolina Population Center at the University of North Carolina at Chapel Hill ().

Notes

Bios

Markus H. Schafer is an assistant professor of sociology at the University of Toronto. His past research on obesity has examined hospitalization, discrimination, risk accumulation, and psychological well-being. Currently, he is studying the relationship between health and social network structure among older adults.

Soyoung Kwon is a PhD candidate in the Department of Sociology at Purdue University. Her research interests include social stratification and inequality, health, and research methods. She is currently working on her dissertation, which examines changing patterns of health disparities during China’s economic reform.

References

Alwin

Duane F.

McCammon

Ryan J.

2003. “Generations, Cohorts, and Social Change.” Pp. 23–49 in Handbook of the Life Course, edited by Mortimer

J. T.

Shanahan

M. J.

New York: Springer.

Brancati

Frederick L.

Wang

Nae-Yuh

Mead

Lucy A.

Liang

Kung-Yee

Klag

Michael J.

1999. “Body Weight Patterns from 20 to 49 Years of Age and Subsequent Risk for Diabetes Mellitus: The Johns Hopkins Precursors Study.” Archives of Internal Medicine 159(9):957–63.

Cecchini

Michele

Sassi

Franco

Lauer

Jeremy A.

Lee

Yong Y.

Guajardo-Barron

Veronica

Chisholm

Daniel

. 2010. “Tackling of Unhealthy Diets, Physical Inactivity, and Obesity: Health Effects and Cost-Effectiveness.” The Lancet 376(9754):1775–84.

Chen

Feinian

Liu

Guangua

. 2009. “Population Aging in China.” Pp. 157–72 in International Handbook of Population Aging, edited by Uhlenberg

Dordrecht, The Netherlands: Springer.

Chen

Feinian

Yang

Liu

Guangua

. 2010. “Social Change and Socioeconomic Disparities in Health over the Life Course in China.” American Sociological Review 75(1):126–50.

Elder

Glen H.

Johnson

Monica K.

Crosnoe

Robert

. 2003. “The Emergence and Development of Life Course Theory.” Pp. 3–19 in Handbook of the Life Course, edited by Mortimer

J. T.

Shanahan

M. J.

New York: Springer.

Freese

Jeremy

Lutfey

Karen

. 2010. “Fundamental Causality: Challenges of an Animating Concept for Medical Sociology.” Pp. 677–81 in Handbook of the Sociology of Health, Illness, and for the 21st Century, Healing: A Blueprint, edited by Pescosolido

B. A.

Martin

J. K.

McLeod

Rogers

New York: Springer.

Holt

Peter R.

Balint

John A.

1993. “Effects of Aging on Intestinal Lipid Absorption.” American Journal of Physiology: Gastrointestinal and Liver Physiology 264(1):G1–G6.

Jones-Smith

J. C.

Gordon-Larsen

Siddiq

Popkin

B. M.

2011. “Emerging Disparities in Overweight by Educational Attainment in Chinese Adults (1989–2006).” International Journal of Obesity 36(4):866–75.

10.

Kuh

Diana

Ben-Shlomo

Yoav

. 2004. A Life Course Approach to Chronic Disease Epidemiology. Oxford, UK: Oxford University Press.

11.

Lee

J. M.

Pilli

Gebremariam

Keirns

C. C.

Davis

M. M.

Vijan

Freed

G. L.

Herman

W. H.

Gurney

J. G.

2010. “Getting Heavier, Younger: Trajectories of Obesity over the Life Course.” International Journal of Obesity 34(4):614–23.

12.

Link

Bruce G.

Phelan

. 1995. “Social Conditions as Fundamental Causes of Disease.” Journal of Health and Social Behavior 35(extra issue):80–94.

13.

McLaren

Lindsay

. 2007. “Socioeconomic Status and Obesity.” Epidemiologic Reviews 29(1):29–48.

14.

Miech

Richard

. 2008. “The Formation of a Socioeconomic Health Disparity: The Case of Cocaine Use during the 1980s and 1990s.” Journal of Health and Social Behavior 49(3):352–66.

15.

Mirowsky

John

Ross

Catherine E.

2003. Education, Social Status and Health. New York: Aldine de Gruyter.

16.

Muthén

Linda K.

Muthén

Bengt

. 1988–2011. MPlus User’s Guide. 6th ed. Los Angeles, CA: Muthén & Muthén.

17.

Popkin

Barry M.

1999. “Urbanization, Lifestyle Changes and the Nutrition Transition.” World Development 27(11):1905–16.

18.

Popkin

Barry M.

2008. “Will China’s Nutrition Transition Overwhelm Its Health Care System and Slow Economic Growth?” Health Affairs 27(4):1064–76.

19.

Popkin

B. M.

Gordon-Larsen

2004. “The Nutrition Transition: Worldwide Obesity Dynamics and Their Determinants.” International Journal of Obesity 28(Suppl. 3):s2–s9.

20.

Reither

Eric N.

Hauser

Robert M.

Yang

. 2009. “Do Birth Cohorts Matter? Age-Period-Cohort Analyses of the Obesity Epidemic in the United States.” Social Science & Medicine 69(10):1439–48.

21.

Ryder

Norman B.

1965. “The Cohort as a Concept in the Study of Social Change.” American Sociological Review 30(6):843–61.

22.

Schafer

Markus H.

Ferraro

Kenneth F.

2007. “Obesity and Hospitalization over the Adult Life Course: Does Duration of Exposure Increase Use?” Journal of Health and Social Behavior 48(4):434–49.

23.

Singer

Judith D.

Willett

John B.

2003. Applied Longitudinal Data Analysis: Modeling Change and Event Occurrence. New York: Oxford University Press.

24.

StataCorp. 2011. Stata Statistical Software: Release 12. College Station, TX: StataCorp.

25.

Wang

Shan

X. Y.

Wang

Q. J.

K. Y.

2006. “Is China Facing an Obesity Epidemic and the Consequences? The Trends in Obesity and Chronic Disease in China.” International Journal of Obesity 31(1):177–88.

26.

Yangfeng

. 2006. “Overweight and Obesity in China.” BMJ 333(7564):362–63.

27.

Yang

Lee

Linda C.

2009. “Sex and Race Disparities in Health: Cohort Variations in Life Course Patterns.” Social Forces 87(4):2093–2124.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.12 MB