Objective and Subjective Socioeconomic Status, Their Discrepancy, and Health: Evidence from East Asia

Abstract

Socioeconomic status (SES) is largely understood to be a fundamental determinant of health. Recently, subjective socioeconomic status (SSS) has emerged as a potentially important predictor of health above and beyond traditional (i.e., objective) SES indicators (OSS). The current study adds to this emerging body of research by examining the potentially important role of status discrepancies for health outcomes. We used nationally representative data from three East Asian countries (China, Japan, and South Korea) (2010 East Asian Social Survey) and a non-linear statistical technique (i.e., diagonal mobility model) to simultaneously model the independent contributions of OSS and SSS and their discrepancy for three health outcomes. Findings showed that SSS does, in fact, explain additional variation in health net of OSS in most cases, and status discrepancy is not associated with any of the three health outcomes. While status discrepancy was not found to be a driving factor in determining the predictive power of SSS net of OSS (at least in East Asia), the present study adds robustness to the accumulating evidence that challenges the social inequality hypothesis and provides a basis from which future research can build and contribute further to the understanding surrounding socioeconomic status and health outcomes.

1. Introduction

Socioeconomic status (SES) is widely recognized as a fundamental determinant of health, such that resources of knowledge, money, power, and prestige largely determine one’s ability to avoid risks for, and adopt protective strategies to prevent, morbidity and mortality (Link and Phelan 1995). Yet, the actual pathways that link health and SES are not well-established (Phelan et al. 2010; Schieman and Koltai 2017), and there is debate surrounding the utility of different SES components (see Andersson 2017). Specifically, a growing body of research suggests that traditional components (e.g., educational attainment, income, etc.) do not adequately capture underlying factors that determine one’s social position, which is argued to have socio-psychological effects on health independent from material circumstances (see Prag et al. 2016). Thus, two distinct SES components are at the center of this debate: (a) one that focuses on material [i.e., objective socioeconomic status (OSS)], and (b) another that highlights socio-psychological [i.e., subjective socioeconomic status (SSS)], factors (see Schnittker and McLeod 2005).

Studies focused on identifying the independent contributions of OSS and SSS to health are largely based on empirical evidence from select populations [e.g., pregnant women (Reitzel et al. 2007), and working men (Macleod et al. 2005)] and a small number of studies that examined representative samples of the general population (see Prag et al. 2016). This literature, despite a lack of generalizable evidence, has consistently shown that a wide range of health outcomes are associated with SSS after controlling for OSS (e.g., Cohen et al. 2008; Goldman et al. 2006; Ostrove et al. 2000; Wright and Steptoe 2005). Yet, it remains unknown whether SSS simply just explains additional variation in health, or if status discrepancies (i.e., OSS > or < SSS) are driving this relationship. Status discrepancy represents one’s perceived (dis)advantage when he or she compares him or herself to others with similar material resources (i.e., OSS), which is recognized as a major socio-psychological determinant of health (see Wilkinson and Pickett 2006; see also van de Brake et al. 2017). One key reason knowledge surrounding status discrepancies is scant is because researchers have generally relied on statistical techniques that are unable to simultaneously model the independent contributions from OSS and SSS and their discrepancy.

In this study, we examine the relative importance of OSS and SSS, and the role of their discrepancy, for three health outcomes (i.e., self-rated health, depression, and chronic conditions), with data from the 2010 wave of the East Asian Social Survey. The current study makes three important contributions: First, we use a methodological approach [i.e., the diagonal mobility model (DMM; Sobel 1981)] that is well-suited to deal with the linear dependent nature and for simultaneously modeling the independent contributions of OSS, SSS, and their discrepancy. Second, we use nationally representative data from China, Japan, and South Korea to build on an emerging body of generalizable evidence (see Prag et al. 2016). Finally, little is known about the associations between health and OSS/SSS in East Asia (Hanibuchi et al. 2012), and the current study throws light on this important, yet understudied, region of the world.

1.1. Background

Disentangling the contributions from OSS and SSS is an important step toward better identifying the pathways that link health and SES. While OSS is among the strongest known predictors of SSS, these two SES components are conceptually distinct (see Andersson 2017). Traditional socioeconomic indicators such as educational attainment, occupational status, and income typically reflect (individually or, though less frequently, collectively as an index) OSS, whereas SSS is a self-appraisal of one’s socioeconomic standing in his or her society (Adler et al. 2000). The former is useful for identifying one’s relative position in a given society in objective terms, and the latter is useful for identifying his or her perceived relative position—often measured by asking respondents to assess their social position with the use of a ladder analogy where the top rung represents people who are best off (e.g., those who have the most money, education, and respected jobs), and the bottom rung represents those who are worst off (see Singh-Manoux et al. 2003).

Both SES components generally reflect one’s socioeconomic standing within his or her society (Singh-Manoux et al. 2005), but there are two prominent arguments for why SSS tends to explain additional variation in health outcomes net of OSS. First, the health-SSS link has roots in a strand of research inspired by The Black Report (Black et al. 1982), which has shown that morbidity and mortality are more strongly tied to income inequality than absolute income (Wilkinson 1996, 1999). While related findings of the association between inequality and average population health are mixed and contentious (see Beckfield 2004), emerging research points to the limited nature of aggregate health measures and highlights the need to examine the social distribution of population health (Beckfield et al. 2013).

In fact, there is only modest evidence to support a link between national-level inequality and average population health, but a growing amount of evidence highlights the role of inequality in shaping health disparities (Truesdale and Jencks 2016). In general, population health disparities are understood to be linked to socioeconomic inequality through socio-psychological pathways related to stress processes (Brunner and Marmot 1999) and health-related behaviors (Marmot 2005) that are more closely related to one’s perceived social position than his or her material resources (see Wilkinson and Pickett 2006). The importance of hierarchical position for health outcomes has been substantiated in animal models (Sapolsky 2004), but this is still regarded as a controversial issue within the social sciences (see Lillard et al. 2015).

Indeed, there are few theories that explicitly address the socio-psychological mechanisms thought to drive the association between inequality and health (see Truesdale and Jencks 2016), and research has yet to fully address the potentially important role of status inconsistency (i.e., status discrepancy). According to inconsistency theory (Hughes 1945), status inconsistency (e.g., SSS > OSS) induces stress (see van de Brake et al. 2017). Previous research concerned with inconsistent subjective and objective socioeconomic status is limited, but it does suggest that those who assess their SSS to be greater than their OSS experience additional stress due this mismatch between expectations and actual circumstances (Macleod et al. 2005; Muntaner et al. 1998).

A second argument is more methodological in nature, and rests on the idea that SSS possesses a multidimensional quality that enables it to more fully reflect one’s socioeconomic resources compared to objective indicators (Singh-Manoux et al. 2003). For example, educational attainment typically reflects one’s accumulated years of education or highest degree, whereas such fine-grained factors as college selectivity are likely accounted for in self-appraisals of social position (Nielsen et al. 2015). In general, such self-appraisals may better capture an individual’s unique circumstances through a process of cognitive averaging that accounts for past, present, and expected objective socioeconomic resources, which suggests that SSS is both a socio-psychological and economic phenomenon (Singh-Manoux et al. 2005). Thus, this second argument is not fully disjoined from the first, but it is distinct because less emphasis is placed on the role of hierarchical rank.

Taken together, there is not a great deal of support for either argument. For example, a cross-national comparative study of twenty-nine countries found that health is associated with SSS net of OSS in all countries under study, but the strength of this association did not vary systematically in relation to national-level income inequality (Prag et al. 2016). Furthermore, cognitive averaging processes appear to differ by cultural context. For example, only present OSS explains variation in SSS among non-Hispanic whites in the U.S., while an average of past, present, and expected OSS best explains SSS in Japan (Andersson 2015). These recent studies, which do not closely align with prominent perspectives, point to the need for a more discriminative approach—one which focuses on countries that share a common culture.

1.2. East Asian Context

Cross-national studies are useful to establish the robustness of findings from single-context studies (Beckfield et al. 2013). Yet, cross-national studies that disentangle the contributions from OSS and SSS to health are rare (e.g., Adler et al. 2008; Prag et al. 2016). Moreover, the relative importance of OSS and SSS is largely unknown outside of a Western context (Euteneuer 2014). Therefore, the current study focuses on three East Asian countries (i.e., China, Japan, and South Korea). The present analyses are limited in terms of cross-national comparisons, but our focus on these three countries adds to the robustness of our findings and sheds light on an important, yet understudied, region of the world.

China, Japan, and South Korea arguably share a common culture (i.e., Confucianism). In contrast to most Western cultures, Confucianism emphasizes the importance of social harmony, collectivism, and a balanced approach to life (Inoguchi et al. 2005). Moreover, limited evidence does suggest that East Asians uniquely assess SSS from a temporally (e.g., past, present, and expected) balanced perspective (Andersson 2015). East Asia is also important to consider in light of its large population [e.g., China is the first, Japan is the eleventh, and South Korea is the twenty-eighth most populated country in the world (Central Intelligence Agency 2016)], and its economic prowess [e.g., China’s GDP is second, Japan’s is third, and South Korea’s is eleventh highest in the world (The World Bank 2015)].

In terms of national-level income inequality (see Wilkinson and Pickett 2006), China’s level of inequality is relatively high (Gini = 47), Japan’s is moderate (Gini = 38), and South Korea’s is among the lowest in the world (Gini = 30; Central Intelligence Agency 2013). In relation to institutional structures (Wilkinson and Pickett 2010), all three countries under study have social health insurance programs, but out-of-pocket health care expenditures differ substantially [i.e., 60% in China, 13% in Japan, and 50% in South Korea (O’Donnell et al. 2008)]. In regard to cross-national differences in economic background (see Deaton 2013), each country’s path toward development has been distinct [e.g., China is undergoing dramatic transformations, Japan is characterized by long-standing economic success, and South Korea has experienced rapid growth over the past four decades (Beeson 2014)].

2. Hypotheses

2.1. Relative Weight of OSS and SSS

Objective socioeconomic status is widely recognized as key for good health (Phelan, Link, and Tehranifar 2010). Whether subjective status explains variation in health over and above objective status in East Asia remains unknown. Thus, examining the relative weight of SSS versus OSS for health in three East Asian countries (China, Japan, and South Korea) is a major focus of this study. Yet, there is a lack of theory to develop informed hypotheses for whether SSS has predictive power over and above OSS for health in East Asia. Findings to these regards contribute to an emerging body of cross-national evidence (e.g., Prag et al. 2016), which ideally will lead to better identifying pathways that link health and SES.

2.2. Consistent Socioeconomic Statuses

An individual is recognized to have consistent socioeconomic statuses when her or his objective and subjective socioeconomic statuses are equivalent. Having less of both OSS and SSS is likely detrimental for one’s health (Singh-Manoux et al. 2005).

H1 Individuals with the highest consistent socioeconomic status (i.e., those who report both high objective and high subjective socioeconomic status) will be healthier than those with relatively lower consistent socioeconomic status. Conversely, individuals with the lowest consistent socioeconomic status (i.e., those who report both low objective and low subjective socioeconomic status) will be the least healthy consistent status group.

2.3. Inconsistent Socioeconomic Statuses

Status inconsistency (i.e., status discrepancy), according to inconsistency theory (Hughes 1945), induces stress (see van de Brake et al. 2017). Limited evidence shows that those who assess their SSS to be greater than their OSS experience additional stress due a mismatch between expectations and objective circumstances (Macleod et al. 2005; Muntaner et al. 1998). Relatedly, the dissociative thesis posits that social mobility is also linked to stress (see Sorokin 1927). However, an emerging body of social mobility research has found little evidence to support the dissociative thesis, as findings suggest that mobile individuals tend to adapt to new social positions through processes of acculturation (Houle and Martin 2011; Marshall and Firth 1999). Yet, processes of acculturation appear to be asymmetric, as downward mobility results in relatively higher levels of dissociation (Daenekindt 2017).

While the present study is not focused on social mobility, this closely related literature further points to the potentially negative consequences of having an SSS that is greater than one’s OSS. For example, OSS reflects one’s current (destination) status, and status inconsistencies likely emerge when one’s SSS is rooted in his or her origin status. In turn, SSS that exceeds OSS could be a sign of downward mobility. We are unable to test this extension of the dissociative thesis, but these speculations shed some light on potentially important underlying process that could link negative health outcomes with status inconsistencies (see Daenekindt 2017).

H2 Individuals whose SSS is greater than their OSS are more likely to have worse health compared to individuals with equivalent OSS (i.e., SSS > OSS = x will have worse health than SSS = OSS = x).

3. Data and Methods

Data used in this study came from the 2010 wave of the East Asian Social Survey (EASS). This biannual cross-sectional survey follows a similar format as the U.S.A.’s General Social Survey (GSS), and includes a common module in each of the Chinese, Japanese, South Korean, and Taiwanese samples. Each country’s sample is nationally representative of the non-institutionalized adult population, though, each country employed a slightly different sampling technique. The target ages of the samples were 18-years and older in China, South Korea, and Taiwan, and 20-years and older in Japan. These data include measures that tap both objective and subjective socioeconomic status, as well as a variety of health outcomes. Because some important health outcomes and demographic characteristics are not available in the Taiwanese sample, Taiwan was excluded from the current study. We further limited our samples to those age 30-years and older, as younger adults have not yet fully acquired an objective and/or subjective status. Accordingly, 3307 observations in the Chinese sample, 2260 observations in the Japanese sample, and 1281 observations in the South Korean sample are included in the current analyses. More detailed descriptions of these data are published elsewhere (e.g., Chen 2012), and further information can be obtained from the EASS website (http://eass.info).

3.1. Dependent Variables: Health Outcomes

Multiple health outcomes were examined to add to the robustness of our findings, as the EASS is limited to self-reported health measures and there is no agreed upon or “gold standard” health measure. The three health outcomes considered in the current study include self-rated health, depression, and chronic conditions. Self-rated health was measured by asking “In general, would you say your health is: excellent, very good, good, fair, or poor?” This measure was dichotomized to compare those who reported (0) negative health (i.e., fair or poor) to those who reported (1) positive health (i.e., good to excellent). Depression was measured by asking respondents to report how often they had felt depressed: all of the time, most of the time, some of the time, a little of the time, or none of the time. This measure was dichotomized to compare those who reported feeling depressed (0) none or a little of the time versus (1) some or more of the time. Chronic conditions included whether a responded reported having hypertension, diabetes, heart disease, respiratory problems, or other chronic conditions (0 = no conditions and 1 = one or more condition).

3.2. Independent Variables: Subjective and Objective Socioeconomic Status

We adopt multidimensional measures of SSS and OSS rather than relying on unidimensional measures such as income, education, and occupation (Geyer et al. 2006; Hoffmann et al. 2018; Torssander and Erikson 2009). Subjective socioeconomic status (SSS) was measured with the use of a ten-rung symbolic ladder, where the first and the tenth rung represent the lowest and the highest socioeconomic status, respectively. Participants were instructed: “Think of a ladder with 10 steps representing where people stand in your country. At step 10 are people who are the best off—those who have the most money, the most education, and the most respected jobs. At step 1 are the people who are worst off—those who have the least money, least education, and the least respected jobs or no job. Where would you place yourself on this ladder?” The ten-levels were divided into four categories (1–3, 4, 5, 6–10) that reflect relatively similar group sizes. Detailed validation of this instrument can be found elsewhere (Singh-Manoux et al. 2003; Adler et al. 2000). Collapsing categories could help account for possible non-linear effects. Additionally, with more than four categories of SES, some cells in the mobility table may not have enough individuals to generate good estimates.

Cross-country differences in educational systems and income levels for purchasing power complicate cross-national studies of objective socioeconomic status (OSS; Adler et al. 2008). Therefore, OSS was measured according to the International Socio-Economic Index (ISEI), which is advantageous over other objective measures (e.g., educational attainment, and income) because it acts as an intervening measure that optimizes the relationship between education and income (Ganzeboom et al. 1992). The ISEI scale is based on the respondent’s current or last job and relies on a set of weights assigned to occupations to maximize its role as an intervening variable between education and income. This OSS measure is understood to be more cross-nationally comparable than separate traditional measures (e.g., Ganzeboom and Treiman 1996; Ganzeboom et al. 1992; Treiman and Ganzeboom 2000), it has been widely used in East Asian studies (e.g. Wu and Treiman 2004; Lu and Treiman 2008; Wu 2011; Park 2013; Byun et al. 2012), and cross-national East Asian studies (e.g. Hanibuchi et al. 2012; Yu and Chiu 2014). OSS quartiles (i.e., lower, lower-middle, upper-middle, and upper) were computed for each country separately to match the SSS measure’s four-unit structure.

Inconsistent socioeconomic status (i.e., those who report a subjective socioeconomic status that is either higher or lower than their objective socioeconomic status) was taken into account by computing two dummy indicators: whether an individual’s SSS is (a) higher or (b) lower than his or her OSS. Given that OSS is based on four groups (i.e., quartiles) and some individuals were on the border between groupings, only the cases where an individual’s SSS is at least two steps higher or lower were used to indicate inconsistent socioeconomic status.

3.3. Covariates

Based on their relative importance in previous literature, the current analyses controlled for age, gender, marital status, medical insurance, and community type. Age was treated as a continuous measure, and a quadratic age term was included. Marital status accounts for whether respondents were coupled (i.e., married or cohabiting), uncoupled (i.e., widowed, separated, or divorced), or not coupled (i.e., never married and/or not currently cohabitation). Medical insurance is a dummy variable that indicates whether an individual had any type of medical insurance. Community type accounts for whether a respondent resides in a large city, the suburbs or outskirts of a city, a town or a small city, or a rural area.

3.4. Empirical Strategy

The analyses were designed to examine the independent contributions of, and the discrepancy between, OSS and SSS to health. Previous research that has simultaneously examined these two socioeconomic components has generally relied on linear regression techniques that are unable to separate the effects of status discrepancy from SSS. That is, the effect of SSS estimated in the linear regression framework is likely to be a mix between the net SSS effect and the effect of status discrepancy. Furthermore, given the linear dependency of OSS, SSS, and status discrepancy (i.e., SSS − OSS = discrepancy), it is impossible to simultaneously model OSS, SSS, and status discrepancy in a linear regression framework.

Thus, we apply a unique non-linear approach developed in the social mobility literature (Sobel 1981) to separate the distinct contributions of origin class, destination class, and social mobility—the diagonal mobility model (DMM). The DMM, also recognized as the diagonal reference model (DRM), has been widely used in sociology. This approach has been applied, more recently, to examine discrepancy between educational ideals and expectations on educational behavior (Vaisey 2010), the relationship between perceived social mobility and health (Jin and Tam 2015), and the association between social mobility and various consequences, such as happiness (Houle 2011; Zang and Dirk de Graaf 2016), mortality (Claussen et al. 2005; Billingsley 2012), and political values (Nieuwbeerta et al. 2000; Tolsma et al. 2009). One requirement of the DMM is that the two independent variables of interest (e.g., OSS and SSS) must have the same number of categories. Despite this requirement, the DMM has several advantages over generalized linear modeling techniques such as the traditional logistic regression in modeling our case, particularly for reducing the chances of obtaining false positive results (van der Waal et al. 2017).

First, the relative importance of OSS and SSS can be modeled with weighted parameters. The OSS and SSS weights sum to one, which allows for a meaningful comparison of their independent contributions. Second, the discrepancy between OSS and SSS can be modeled simultaneously (i.e., with both measures together). For example, in social mobility research, the DMM provides point estimates for the extent to which a mobile individual resembles the “core members” in the origin and destination social classes (Sobel 1981, 1985; Weakliem 1992). Analogously, the current model provides point estimates for the extent to which an individual with inconsistent socioeconomic status (i.e., SSS > or < OSS) resembles the “core members” with consistent socioeconomic status (i.e., SSS = OSS). For example, if an individual has OSS = 1 and SSS = 4, the model estimates will tell us how much this individual resembles his or her OSS “core members” (whose OSS = SSS = 1) or SSS “core members” (whose OSS and SSS = 4). Third, the DMM uses fewer degrees of freedom, which is more advantageous than many conventional models, particularly when the model is complex and has a large number of parameters (Vaisey 2010; Long 1997).

The baseline model for a dichotomous dependent variable is given by:

$$\text{prob}\left({\text{Y}}_{\text{ijk}}=1\right)=1/\left(1+{\text{e}}^{-\text{lin}}\right)$$ (1)

$$\text{lin}={\text{pm}}_{\text{i}}+(1-\text{p}){\text{m}}_{\text{j}}$$ (2)

$$\left(i=1,\dots ,\text{T};j=1,\dots ,\text{T};k=1,\dots ,{\text{N}}_{\text{ij}};\text{p}\in [0,1]\right)$$

Where Y_ijk is the health outcome of respondent k in the ijth cell; m_i stands for the expected mean health outcome of OSS “core members;” stands for the expected health outcome of SSS “core members;” T is the total number of social status groups; N_ij is the number of observations in the ijth cell of the table. For simplicity, we have omitted individual indices in all equations.

The DMM has two reference values: the diagonal mean of the OSS category m_i and that of the SSS category. Thus, p and (1 − p) are interpreted as OSS and SSS weights, respectively. When p > 0.5, the effect of OSS is larger than the effect of SSS. The opposite is true when p < 0.5. After adding status discrepancy variables and covariates, the full model is given by:

$$prob\left(Y_{ijk}=1\right)=1/\left(1+e^{-lin}\right)$$ (3)

$$lin=p{m}_i+(1-p)m_j+\sum _{w=1}^{\text{W}}{b}_w{M}_w+\sum _{z=1}^{\text{Z}}{b}_z{C}_z$$ (4)

$$\left(i=1,\dots ,T;j=1,\dots ,T;w=1,\dots ,W;k=1,\dots ,N_{ij};z=1,\dots ,Z;p\in [0,1]\right)$$

where b_w and b_z are interpreted as the parameters of the status discrepancy variables and covariates, respectively; W indicates total number of status discrepancy variables; Z indicates total number of covariates; M_w represent status discrepancy variables; C_z represent the covariates. It should also be noted that we deviate from traditional DMM studies by testing whether the parameters are different from zero. Here, we follow some recent DMM studies (e.g., Houle 2011; Zang and Dirk de Graaf 2016) to test whether OSS and SSS independently contribute to variation in health outcomes. In other words, if parameters are indistinguishable from zero OSS/SSS does not contribute to variation in health outcomes. All models are estimated using Stata package “diagref” (Lizardo 2007). Unfortunately, the diagref package is not presently available through Stata (presumably because the author stopped maintaining it). Therefore, we have provided the files needed to reproduce our analyses—the copy rights of these files belong to the original author (Lizardo 2007).

3.5. Missing Data

Only small percentages (e.g., generally less than 1%) for each covariate were missing in each sample due to non-response. Missing data were replaced using the multiple imputation by chained equations approach, which replaces missing values with predictions based on information observed in the sample and accounts for random uncertainty across each imputed dataset (Acock 2005; Rubin 1987). Theoretically and computationally, this approach can handle large percentages of missing values (preferable ≤ 40%; Acock 2005; Rubin 1987; Lee and Huber 2011). However, the effectiveness of this approach rests strongly on the assumption that missingness is unrelated to the outcome conditioning on the predictor—missing at random (MAR). The MAR assumption is reasonable if variables that are predictive of missing data in a covariate of interest are included in the imputation model (Sterne et al. 2009).

Approximately 40% of cases in each sample had missing values on OSS—largely due to the fact that housewives were not originally assigned a value on the ISEI scale. Thus, we imputed the data for each country separately. Besides all variables in our empirical model, we also include a wide range of variables in the imputation model: (a) Individual characteristics: education, self-rated physical health, self-rated mental health, religion, chronic conditions, BMI, various personality measures (e.g., calm, energy, trust, happiness), frequency of exercise; (b) Household characteristics: household income, number of household members, geographic location; (c) Spouse’s characteristics: age, education, and ISEI; (d) Parents’ characteristics: father’s and mother’s years of education. Because we have jointly missing covariates and outcomes, multiple imputations by chained equations can generate a complete dataset where the outcome variable is imputed as well and its imputed version is used to impute the independent variables. We imputed ten data sets using the ‘ice’ command (Royston 2005) in Stata 14.2, and then averaged results from the model across these ten imputed samples. A summary by country of all the variables included in the analysis before and after imputation, and percentages missing for each variable, are shown in Table 1.

Table 1.

Variable mean before and after imputation by country

	China			Japan			South Korea
	Percentage missing (%)	Mean before imputation	Mean after imputation	Percentage missing (%)	Mean before imputation	Mean after imputation	Percentage missing (%)	Mean before imputation	Mean after imputation
Health outcomes
Self-rated health	0.12	0.79	0.79	0.13	0.70	0.70	0.08	0.73	0.73
Depression	0.39	0.37	0.37	0.66	0.29	0.29	0.16	0.34	0.34
Chronic diseases	0.00	0.39	0.39	0.00	0.49	0.49	0.08	0.35	0.35
Objective social status	37.71			42.43			40.12
Lower status		0.50	0.39		0.33	0.28		0.30	0.31
Lower-middle status		0.06	0.14		0.27	0.25		0.30	0.23
Upper-middle status		0.24	0.24		0.21	0.24		0.16	0.23
Upper status		0.20	0.23		0.18	0.23		0.24	0.23
Subjective social status	0.45			1.19			1.01
Lower status		0.36	0.36		0.17	0.17		0.29	0.29
Lower-middle status		0.19	0.19		0.14	0.14		0.15	0.16
Upper-middle status		0.30	0.30		0.15	0.15		0.29	0.29
Upper status		0.14	0.14		0.54	0.54		0.26	0.26
Age	0.03	51.06	51.06	0.00	56.73	56.73		50.07	50.01
Marital status	0.54			0.04			0.31
Married or cohabiting		0.86	0.86		0.78	0.78		0.76	0.76
Widowed, separated or divorced		0.12	0.12		0.13	0.13		0.16	0.15
Never married		0.02	0.02		0.09	0.09		0.09	0.09
Community type	0.00			0.22			0.62
A big city		0.18	0.18		0.05	0.05		0.28	0.28
Suburbs or outskirts of a big city		0.07	0.07		0.16	0.16		0.27	0.27
A town or a small city		0.35	0.36		0.45	0.45		0.29	0.30
Country		0.39	0.39		0.35	0.35		0.15	0.15
Female	0.00	0.51	0.51	0.00	0.54	0.54	0.00	0.54	0.53
Have insurance	3.08	0.90	0.90	3.19	0.97	0.97	0.78	0.94	0.94
N		3307			2260			1281

3.6. Robustness Checks

There are theoretical and empirical debates about how to measure women’s subjective social status. Especially in East Asian countries, a large proportion of women may be housewives and thus “not employed,” or in a lower occupational status compared to their spouse for family reasons. For those not employed women, the multiple imputation should have effectively imputed their social status. But, for employed women who took a lower occupational status compared to their spouse for family reasons, the ISEI may have a large measurement error. Therefore, we restricted our sample to males and repeated all the analyses. The main results for the male sample are generally consistent with our overall findings (see “Appendix” Tables 7, 8 and 9). Additionally, we also conducted analyses with 3-category SSS and OSS measures, and results are largely comparable (see “Appendix” Tables 10, 11 and 12).

4. Results

4.1. Characteristics of the Analytic Samples

Mean values for the three health outcomes (i.e., self-rated health, depression, and chronic conditions) are respectively similar across each country (see Table 1). Approximately three-fourths of respondents in each sample have good or better self-rated health, one-third reported feeling depressed some or more of the time, and two in five have chronic health conditions. OSS is evenly distributed across the four categories (i.e., quartiles) in Japan and South Korea, and the lower status category is relatively larger in China. In regard to SSS, approximately half of the respondents in China and South Korea reported that they were either lower or lower-middle status, but this was only one-third in Japan. Mean age is approximately 50-years old in China and South Korea, and 57-years old in Japan. Eighty-six percent of respondents were married or cohabiting in China, 78% in Japan, and 76% in South Korea. Approximately 3 in 20 respondents were uncoupled in all three countries. The South Korean sample is more urban than the other two countries. Approximately one-half of each sample is female, and almost everyone has health insurance in all three countries.

4.1.1. Mean SSS by OSS

Table 2 shows the country-specific frequencies for SSS by OSS in a 4 × 4 format. The rows represent SSS and the columns reflect the corresponding OSS category. Thus, the difference on the diagonal indicates distributional differences in the correlation between OSS and SSS within each of three countries, respectively. The strongest pattern can be seen among the lower status groups. For example, in China, approximately one-half of those who reported lower SSS were also in the lower OSS group—this is relatively greater than the other proportions on the diagonal. Similar patterns are also found in Japan and South Korea, but with relatively smaller proportions at the lower level and generally more equal distributions across categories. Thus, SSS is more strongly correlated with OSS at the lower and upper level compared with other categories across all three countries, which can be clearly seen in the graphical presentation of these data in Fig. 1.

Table 2.

Average frequencies by objective and subjective social status, using ten imputed datasets

Subjective social class	Objective Social Class
	Lower class	Lower-middle class	Upper-middle class	Upper class
China (N = 3307)
Lower class	601	263	313	122
Lower-middle class	178	93	146	58
Upper-middle class	260	164	250	116
Upper class	166	121	282	174
Japan (N = 2260)
Lower class	154	81	88	312
Lower-middle class	118	90	95	259
Upper-middle class	72	77	80	306
Upper class	43	63	84	340
South Korea (N = 1281)
Lower class	163	67	97	70
Lower-middle class	98	45	82	69
Upper-middle class	71	44	94	89
Upper class	44	43	98	108

Fig. 1.

Mean subjective social status by objective social status. Note: x-axis represents Objective Social Status, y-axis represents Subjective Social Status. Proportions for each Subjective Social Status sum up to 1

4.2. Consistent OSS and SSS

4.2.1. Self-Rated Health

In all three countries, those with both lower OSS and SSS (i.e., consistent lower status) are less likely to report excellent/good self-rated health compared to those in other consistent status categories (see Table 3). The performance of other social statues in non-discrepant locations is context specific, and there appears to be no clear status gradient in self-rated health. Specifically, only those with both upper-level SSS and OSS in both China and Japan are more likely to report excellent/good self-rated health compared to those in other consistent status categories. Unlike the other two countries, consistent upper status is not associated with better self-rated health in South Korea.

Table 3.

Estimates from diagonal mobility models predicting self-rated health

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.237** (0.0940)	0.218 (0.201)	0.557* (0.280)
Subjective social status weight (r)	0.763*** (0.0940)	0.782*** (0.201)	0.443 (0.280)
Consistent social status (OSS = SSS)
Lower status	−0.830*** (0.114)	−0.425*** (0.143)	−0.465* (0.244)
Lower-middle status	0.124 (0.122)	−0.122 (0.133)	0.0751 (0.261)
Upper-middle status	0.186* (0.110)	0.177 (0.143)	0.280 (0.197)
Upper status	0.520*** (0.154)	0.371*** (0.0983)	0.110 (0.182)
Age	−0.138*** (0.0272)	−0.00888 (0.0259)	−0.0320 (0.0465)
Age square	0.000824*** (0.000239)	−0.000166 (0.000225)	−0.000194 (0.000413)
Female (1 = yes)	−0.464*** (0.0943)	−0.0303 (0.0975)	−0.440*** (0.153)
Have health insurance	−0.166 (0.168)	0.512* (0.287)	−0.0395 (0.336)
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	−0.225 (0.139)	0.0216 (0.148)	−0.268 (0.189)
Never married	−0.404 (0.353)	−0.489*** (0.176)	−0.338 (0.310)
Community type (Ref = A big city)
Suburbs or outskirts of a big city	0.344 (0.252)	0.0672 (0.260)	0.140 (0.201)
A town or a small city	−0.423*** (0.145)	−0.149 (0.237)	0.226 (0.193)
Country	−0.629*** (0.167)	−0.164 (0.240)	−0.334 (0.234)

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

4.2.2. Depression

Those with consistent lower status are more likely to report depression in China and South Korea, but not Japan, compared to those in other consistent status categories (see Table 4). In China, those with consistent upper-middle or upper status are less likely to report depression, while. In Japan, only those with upper status, and in South Korea only those with upper-middles status are less likely to report depression.

Table 4.

Estimates from diagonal mobility models predicting depression

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.223** (0.0880)	0.133 (0.487)	0.249 (0.213)
Subjective social status weight (r)	0.777*** (0.0880)	0.867* (0.487)	0.751*** (0.213)
Consistent social status (OSS = SSS)
Lower status	0.781*** (0.0945)	0.211 (0.124)	0.556*** (0.190)
Lower-middle status	0.0155 (0.0930)	0.0147 (0.147)	−0.0626 (0.178)
Upper-middle status	−0.197** (0.0849)	0.106 (0.129)	−0.377*** (0.139)
Upper status	−0.599*** (0.121)	−0.331** (0.149)	−0.117 (0.149)
Age	0.0549** (0.0216)	−0.00855 (0.0254)	0.00258 (0.0348)
Age square	−0.000435** (0.000197)	−1.25e-06 (0.000226)	−2.29e-05 (0.000314)
Female (1 = yes)	0.484*** (0.0765)	0.166 (0.0987)	0.665*** (0.132)
Have health insurance	−0.162 (0.127)	−0.564* (0.277)	−0.697** (0.278)
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	0.435*** (0.125)	0.0114 (0.152)	0.110 (0.194)
Never married	0.0289 (0.256)	0.257 (0.163)	0.276 (0.235)
Community type (Ref = A big city)
Suburbs or outskirts of a big city	−0.0612 (0.182)	−0.419* (0.240)	0.0799 (0.167)
A town or a small city	0.149 (0.113)	−0.325 (0.219)	−0.110 (0.168)
Country	0.214 (0.139)	−0.381 (0.224)	0.369* (0.209)

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

4.2.3. Chronic Conditions

Only in China are chronic conditions associated with consistent statuses. Specifically, those with consistent lower SSS and OSS in China are more likely to report having chronic health conditions. Additionally, Chinese respondents with consistent upper status are less likely to report chronic health conditions compared to those in other consistent status categories (see Table 5).

Table 5.

Estimates from diagonal mobility models predicting chronic diseases

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.170 (0.166)	0.211 (0.481)	0.936 (1.111)
Subjective social status weight (r)	0.830*** (0.166)	0.789 (0.481)	0.0643 (1.111)
Consistent social status (OSS = SSS)
Lower status	0.483*** (0.0992)	0.218 (0.161)	0.217 (0.215)
Lower-middle status	−0.0353 (0.0919)	−0.000525 (0.162)	0.146 (0.221)
Upper-middle status	−0.0566 (0.0827)	−0.0666 (0.131)	−0.234 (0.477)
Upper status	−0.391*** (0.125)	−0.151 (0.101)	−0.129 (0.434)
Age	0.167*** (0.0245)	0.123*** (0.0272)	0.128*** (0.0440)
Age square	−0.000972*** (0.000220)	−0.000608** (0.000238)	−0.000604 (0.000384)
Female (1 = yes)	0.392*** (0.0798)	−0.121 (0.0938)	0.319* (0.178)
Have health insurance	0.0980 (0.138)	0.655* (0.344)	0.235 (0.341)
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	0.133 (0.128)	−0.0268 (0.148)	0.217 (0.186)
Never married	−0.00882 (0.306)	0.438** (0.170)	0.273 (0.286)
Community type (Ref = A big city)
Suburbs or outskirts of a big city	−0.200 (0.178)	−0.289 (0.249)	0.0213 (0.189)
A town or a small city	−0.272** (0.112)	−0.206 (0.231)	0.108 (0.180)
Country	−0.0443 (0.142)	−0.381 (0.235)	0.190 (0.230)

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

4.3. Relative Weight of OSS and SSS

4.3.1. Self-Rated Health

SSS is a statistically significant predictor of self-rated health net of OSS and all covariates in China and Japan, but not in South Korea where the relative weight of SSS is not statistically different from zero (see Table 3). The relative weight of SSS for self-rated health in China and Japan is 0.76 and 0.78, respectively. Both OSS and SSS make independent contributions to explaining self-rated health in China. Only SSS, but not OSS, makes a contribution to explaining self-rated health in Japan, and visa-versa in South Korea

4.3.2. Depression

SSS explains a substantially greater amount of variations in depression compared to OSS across all three countries. The relative weight of SSS is 0.78 in China, 0.87 in Japan, and 0.75 in South Korea (see Table 4). OSS is not a statistically significant predictor of depression net off SSS and all other covariates in Japan and South Korea. However, in China, OSS does make a statistically significant contribution toward explaining variance in depression.

4.3.3. Chronic Conditions

SSS is a statistically significant predictor of chronic conditions net of OSS only in China where it carries 83% of the weight. Neither SSS nor OSS contribute toward explaining chronic diseases in Japan and South Korea net of one another and all other covariates (see Table 5).

4.4. Inconsistent OSS and SSS: Status Discrepancy

Table 6 shows the results from models that estimated the effects of having an inconsistent status on the three health outcomes, respectively. Results indicate that health for those with inconsistent statuses is not different from those with consistent statuses.

Table 6.

Estimates from diagonal mobility models on status discrepancy

	Self-rated health			Depression			Chronic diseases
	China	Japan	South Korea	China	Japan	South Korea	China	Japan	South Korea
Objective social status weight (p)	0.358** (0.137)	0.264 (0.566)	0.612 (0.441)	0.138 (0.183)	0.233 (0.642)	0.270 (0.301)	0.222 (0.222)	0.620 (0.637)	0.338 (0.518)
Subjective social status weight (r)	0.642*** (0.137)	0.736 (0.566)	0.388 (0.441)	0.862*** (0.183)	0.767 (0.642)	0.730** (0.301)	0.778*** (0.222)	0.380 (0.637)	0.662 (0.518)
SSS higher than OSS	0.134 (0.216)	0.114 (0.395)	−0.145 (0.273)	0.155 (0.226)	−0.0936 (0.275)	−0.0339 (0.244)	−0.174 (0.184)	−0.229 (0.218)	0.350 (0.274)
SSS lower than OSS	−0.232 (0.215)	0.162 (0.425)	−0.300 (0.335)	−0.0822 (0.214)	−0.0345 (0.301)	0.0241 (0.249)	−0.0611 (0.178)	−0.0214 (0.256)	−0.275 (0.322)

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

5. Discussion

A growing body of research has shown that health is associated with SSS independent of OSS (e.g., Cohen et al. 2008; Singh-Manoux et al. 2005). However, the role of status discrepancy has been neglected, and previous research was generally limited to select populations from Western countries (see Euteneuer 2014). Using an appropriate non-linear statistical technique (i.e., DMM; Sobel 1981) and nationally representative samples from three East Asian countries, we simultaneously modeled the independent contributions from SSS and OSS and their discrepancy in predicting three health outcomes. Findings showed that SSS does, indeed, explain additional variation in health net of OSS in most cases. However, status discrepancies did not explain any of the variation in all three health outcomes, net of the effects of OSS and SSS.

In terms of consistent statuses, having more or less of both SES components (i.e., OSS and SSS) was hypothesized to be protective or detrimental for health, respectively. On the one hand, people with lower OSS and SSS were systematically less healthy than those with higher consistent statuses—with the exception of depression in Japan and chronic conditions in Japan and China. On the other hand, health did not follow a strong gradient across the consistent status categories. Given that the East Asian socioeconomic gradient in health is underexplored, these findings point to a need for more research on underlying patterns across different health outcomes and socioeconomic measures in East Asia (see Hanibuchi et al. 2012).

In regard to status discrepancy, it was hypothesized, given previous findings (e.g., Macleod et al. 2005; Muntaner et al. 1998), that an individual whose SSS is greater than his or her OSS is more likely to have relatively worse health, but this was not the case. These findings suggest that either (a) status discrepancies are less important for health in East Asia compared to the West, or (b) previous status discrepancy findings are artifacts from linear modeling techniques, which points to the need for future cross-national examination between Eastern and Western nations that takes a non-linear approach (e.g., DMM; see Adler et al. 2000).

In terms of cross-national differences, the social inequality hypothesis posits that the link between health and SSS is exacerbated by socioeconomic inequality (see Wilkinson and Pickett 2010). Yet, current findings align with a previous cross-national test of this hypothesis (Prag et al. 2016), which suggests that it is insufficient to simply consider national-level socioeconomic inequality. That is, the relative weight of SSS versus OSS did not follow a pattern simply based on socioeconomic inequality. While the current study is limited in terms of comparing results across countries, it does appear that South Korea is particularly unique. Specifically, OSS seems to be especially important for chronic conditions in South Korea. We can only speculate, but these apparent cross-national differences may be partially due to the rapid economic development that has occurred in South Korea over the past four decades and its differential impact on health behaviors across socioeconomic groups (see Khang and Kim 2005). Cross-national differences within East Asia warrant future investigation.

The relative importance of SSS versus OSS differed by health outcome. In general, the influence of SSS on health is understood to be socio-psychological in nature (Singh-Manoux et al. 2003). Indeed, SSS consistently predicted depression in all three countries, and the predictive nature of SSS versus OSS for other health outcomes (i.e., self-rated health and chronic conditions) appeared to differ across countries. These findings are consistent with previous studies from Japan (Sakurai et al. 2010) and the UK (Demakakos et al. 2008), and among African Americans (Subramanyam et al. 2012), and they partially support the understanding of SSS to act through socio-psychological pathways. However, we cannot assert causality, and issues surrounding reverse causality must also be considered.

6. Limitations and Conclusions

There are several caveats when explaining our results. First, we were limited in our ability to make any causal statements due to potential issues of endogeneity and reverse causality (Singh-Manoux et al. 2005; Nobles et al. 2013). For example, it is possible that individuals could take into account their own health when evaluation their SSS. However, the goal of this study was not to examine the causal effect of SSS and OSS or their discrepancies on health. Given that self-rated health is based on a subjective assessment, and depression is socio-psychological in nature and could influence self-rated health assessments, it is somewhat concerning that SSS was not a strong predictor of chronic conditions (arguably a more objective health measure; see Garbarski 2010). Future studies could address these issues using longitudinal panel data, but such data are still limited in East Asia.

Second, the cross-sectional nature of our data makes it difficult to examine the interplay between SES and health over the life course. Since SSS and OSS are measured at different life stages for each individual, life-cycle bias in measuring SES is a concern (e.g. Haider and Solon 2006). For example, individuals with different occupations may have distinct OSS/SSS trajectories in terms of the intercept (starting point) and the slope (growth). In other words, if an individual has a steep OSS trajectory, but their SES is only measured in young adulthood, then the contribution of lifetime OSS to health is likely to be underestimated.

The aim of the current study was to examine the relative importance of OSS and SSS, and the role of their discrepancy, for health outcomes. Overall, our findings showed that SSS is generally an important predictor of health, and that status discrepancies don’t seem to matter. Previous literature was limited in terms of methodological approach, generalizability, and geographic scope. Our study is among the first to address the role of status discrepancy in shaping the association between health and OSS/SSS, and the present analytic approach (i.e., DMM) represents a promising strategy for future research. Few studies have drawn on nationally representative data to examine associations between health and OSS/SSS, and our use of large nationally representative data adds to the generalizability of consistent findings in this area of research. The actual pathways that link health and SES are not well-established, but one’s perceived social position appears to be an important factor in determining his or her health. The present study’s findings are specific to China, Japan, and South Korea, but they highlight a need to revisit the social inequality hypothesis through more discriminative cross-national research to better identify the underlying mechanisms responsible for the strong associations between health and subjective socioeconomic status.

Appendix

Tables 7, 8, 9, 10, 11 and 12.

Table 7.

Estimates from diagonal mobility models predicting male self-rated health The model for South Korea did not converge

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.259*	0.0952	-
	(0.133)	(0.420)	-
Subjective social status weight (r)	0.741***	0.905*	-
	(0.133)	(0.420)	-
Consistent social status (OSS = SSS)
Lower status	−1.005***	1.869	-
	(0.175)	(8.355)	-
Lower-middle status	−0.000819	−2.902	-
	(0.194)	(8.836)	-
Upper-middle status	0.266	3.326	-
	(0.215)	(9.836)	-
Upper status	0.740***	−2.292	-
	(0.242)	(8.850)	-
Age	−1.005***	−0.0175	-
	(0.175)	(0.0364)	-
Age square	−0.000819	−8.62e-05	-
	(0.194)	(0.000312)	-
Have health insurance	0.266	0.383	-
	(0.215)	(0.484)	-
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	0.740***	0.181	-
	(0.242)	(0.248)	-
Never married	−1.005***	−0.506**	-
	(0.175)	(0.220)	-
Community type (Ref = A big city)
Suburbs or outskirts of a big city	−0.000819	0.433	-
	(0.194)	(0.370)	-
A town or a small city	0.266	0.0172	-
	(0.215)	(0.339)	-
Country	0.740***	1.57e-05	-
	(0.242)	(0.344)	-
N	1595	1048	588

The model for South Korea did not converge

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

Table 8.

Estimates from diagonal mobility models predicting male depression

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.183 (0.149)	0.213 (0.338)	0.138 (0.256)
Subjective social status weight (r)	0.817*** (0.149)	0.787** (0.338)	0.862*** (0.256)
Social status
Lower status	0.741*** (0.135)	0.0105 (2.274)	0.717** (0.290)
Lower-middle status	−0.0810 (0.137)	0.459 (3.280)	−0.0736 (0.195)
Upper-middle status	−0.123 (0.122)	−0.278 (4.574)	−0.736** (0.305)
Upper status	−0.537** (0.206)	−0.191 (3.324)	0.0928 (0.290)
Age	0.0353 (0.0303)	−0.0371 (0.0377)	−0.0459 (0.0542)
Age square	−0.000216 (0.000275)	0.000245 (0.000331)	0.000391 (0.000492)
Have health insurance	−0.00600 (0.188)	−0.164 (0.502)	−0.938** (0.429)
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	0.548*** (0.194)	0.0984 (0.273)	0.945*** (0.315)
Never married	0.176 (0.305)	0.238 (0.222)	0.313 (0.303)
Community type (Ref = A big city)
Suburbs or outskirts of a big city	0.378 (0.263)	−0.235 (0.376)	0.0111 (0.274)
A town or a small city	0.312* (0.170)	−0.316 (0.351)	−0.0939 (0.268)
Country	0.369* (0.204)	−0.103 (0.353)	0.182 (0.330)
N	1595	1048	588

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

Table 9.

Estimates from diagonal mobility models predicting male chronic diseases The model for South Korea did not converge

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.0788 (0.228)	0.912 (12.77)	−
Subjective social status weight (r)	0.921*** (0.228)	0.0885 (12.77)	−
Social status
Lower status	0.456*** (0.128)	4.781 (22.55)	−
Lower-middle status	−0.113 (0.126)	−0.786 (15.58)	−
Upper-middle status	0.111 (0.109)	−3.171 (18.43)	−
Upper status	−0.454** (0.178)	−0.824 (15.57)	−
Age	0.131*** (0.0339)	0.147*** (0.0403)	−
Age square	−0.000668** (0.000302)	−0.000813** (0.000347)	−
Have health insurance	0.213 (0.203)	0.865 (0.509)	−
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	−0.00468 (0.206)	−0.177 (0.258)	−
Never married	0.0765 (0.353)	0.345 (0.226)	−
Community type (Ref=A big city)
Suburbs or outskirts of a big city	−0.219 (0.262)	−0.780* (0.411)	−
A town or a small city	−0.189 (0.163)	−0.627 (0.392)	−
Country	−0.0767 (0.186)	−0.633 (0.396)	−
N	1595	1048	588

The model for South Korea did not converge

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

Table 10.

Estimates from diagonal mobility models predicting male self-rated health

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.213** (0.0999)	0.130 (0.205)	0.446 (0.440)
Subjective social status weight (r)	0.787*** (0.0999)	0.870*** (0.205)	0.554 (0.440)
Consistent social status (OSS = SSS)
Lower status	−0.756*** (0.108)	−0.380*** (0.116)	−0.328* (0.166)
Middle status	0.225** (0.0933)	0.0279 (0.0845)	0.166 (0.212)
Upper status	0.531*** (0.128)	0.352*** (0.0912)	0.162 (0.201)
Age	−0.139*** (0.0272)	−0.00824 (0.0259)	−0.0320 (0.0465)
Age square	0.000827*** (0.000239)	−0.000178 (0.000224)	−0.000213 (0.000413)
Female	−0.466*** (0.0943)	−0.0329 (0.0972)	−0.452*** (0.156)
Have health insurance	−0.162 (0.168)	0.516* (0.289)	−0.0457 (0.335)
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	−0.223 (0.138)	0.0160 (0.148)	−0.264 (0.189)
Never married	−0.405 (0.353)	−0.496*** (0.175)	−0.335 (0.306)
Community type (Ref = A big city)
Suburbs or outskirts of a big city	0.336 (0.252)	0.0849 (0.259)	0.142 (0.202)
A town or a small city	−0.430*** (0.145)	−0.139 (0.237)	0.218 (0.192)
Country	−0.650*** (0.168)	−0.160 (0.240)	−0.370 (0.229)

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

Table 11.

Estimates from diagonal mobility models predicting male depression

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.210** (0.0865)	0.137 (0.531)	0.235 (0.184)
Subjective social status weight (r)	0.790*** (0.0865)	0.863 (0.531)	0.765*** (0.184)
Social status
Lower status	0.745*** (0.0886)	0.215* (0.110)	0.479*** (0.142)
Middle status	−0.138** (0.0682)	0.0971 (0.132)	−0.311*** (0.109)
Upper status	−0.607*** (0.103)	−0.312** (0.123)	−0.169 (0.132)
Age	0.0557** (0.0215)	−0.00869 (0.0254)	−0.00143 (0.0346)
Age square	−0.000440** (0.000197)	5.82e-07 (0.000225)	1.51e-05 (0.000314)
Female	0.487*** (0.0765)	0.168 (0.0982)	0.663*** (0.131)
Have health insurance	−0.168 (0.127)	−0.566* (0.276)	−0.702** (0.276)
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	0.434*** (0.125)	0.0113 (0.152)	0.112 (0.194)
Never married	0.0285 (0.255)	0.257 (0.163)	0.262 (0.234)
Community type (Ref = A big city)
Suburbs or outskirts of a big city	−0.0514 (0.181)	−0.417* (0.240)	0.0802 (0.167)
A town or a small city	0.158 (0.112)	−0.326 (0.220)	−0.109 (0.167)
Country	0.231 (0.138)	−0.380 (0.224)	0.376* (0.207)

^***p < 0.01,

^**p < 0.05,

^*p < 0.1

Table 12.

Estimates from diagonal mobility models predicting chronic diseases The model for South Korea did not converge

	China	Japan	South Korea
Objective and subjective social class weight
Objective social status weight (p)	0.155	−0.342	-
	(0.161)	(0.933)	-
Subjective social status weight (r)	0.845***	1.342	-
	(0.161)	(0.933)	-
Social status
Lower status	0.461***	0.134	-
	(0.0940)	(0.0995)	-
Middle status	−0.0626	−0.0251	-
	(0.0660)	(0.0607)	-
Upper status	−0.399***	−0.108	-
	(0.111)	(0.0892)	-
Age	0.167***	0.121***	-
	(0.0245)	(0.0271)	-
Age square	−0.000973***	−0.000587**	-
	(0.000220)	(0.000235)	-
Female	0.395***	−0.113	-
	(0.0797)	(0.0931)	-
Have health insurance	0.0960	0.644*	-
	(0.138)	(0.345)	-
Marital status (Ref = Married or cohabiting)
Widowed, separated or divorced	0.131	−0.0269	-
	(0.128)	(0.148)	-
Never married	−0.00835	0.439**	-
	(0.306)	(0.170)	-
Community type (Ref = A big city)
Suburbs or outskirts of a big city	−0.198	−0.299	-
	(0.178)	(0.249)	-
A town or a small city	−0.270**	−0.203	-
	(0.111)	(0.231)	-
Country	−0.0406	−0.365	-
	(0.142)	(0.235)	-

The model for South Korea did not converge

^***p < 0.01,

^**p < 0.05,

^*p < 0.1