Network Ties,Upward Status Heterophily,and Unanticipated Health Consequences

Abstract

Using cross-national data containing information on the status rank of network alters, this study investigates the potential negative effects of “upward status heterophily,” ties to and perceived interaction with higher status others. According to our main finding, upward status heterophily is associated with poor physical health and lower subjective well-being. We also find that this focal relationship varies across individual and contextual moderators. For subjective well-being only, it is weaker among people who are better educated, have larger nonkin network, and possess greater self-efficacy. Moreover, there is a significant cross-level interaction: For both health outcomes, the relationship is more pronounced in subnational regions that are economically more unequal. Our findings shed light on the mechanisms of the “dark side of social capital” by operationalizing perceived status differential as a proxy for upward social comparison and showing its deleterious consequences in the East Asian context.

Keywords

physical health relative deprivation social comparison subjective well-being upward status heterophily

It is a sociological axiom that similarity drives human relations (McPherson, Smith-Lovin, and Cook 2001). By demarcating the boundaries of communication and interaction, homophily also plays a powerful role in reinforcing inequality (DiMaggio and Garip 2012). While it is the case that individuals typically gravitate toward homophilous others, people everywhere compete for limited resources (i.e., social capital) through forging heterophilous ties to “higher status” others, that is, people in higher socioeconomic positions.¹ Given that society is largely defined and organized by homophily, those who fare better do so partly because of access to social capital inherent in horizontal as well as vertical interpersonal connections (DiMaggio and Garip 2012; Lin 2001). On the one hand, for lower status actors, higher status counterparts are sought after since they can provide instrumental (information, financial support) and other resources (e.g., sponsorship and endorsement). Actors located higher on the status hierarchy, on the other hand, may be incentivized to interact with those below them because doing so can affirm their relative privilege, power, and prestige (Lin 2001; Lin and Erickson 2012). If favors are channeled downward, then deference may be said to flow upward. Relatedly, there are institutional constraints pressuring individuals to retain those who are socially superior, including “difficult alters” (Offer and Fisher 2018).²

Regardless of how they emerge, heterophilous interactions can have profound implications for health inequality. Indeed, under the general rubric of “social determinants of health” (Kawachi and Berkman 2003), a wealth of evidence points to purportedly salubrious effects of social capital (Ehsan et al. 2019; Xue, Reed, and Menclova 2020). Across multiple contexts, studies emphasize the functional roles of social trust, interaction, and participation in improving health outcomes (Thoits 2011). By providing emotional/material support and through positive behavioral influences, social ties can and may operate as stress buffers throughout the life course (Umberson, Crosnoe, and Reczek 2010). While fully recognizing this fact, the present study adds to the literature by probing the “other” side of social embeddeness or mechanisms underlying the “dark side of social capital” (Villalonga-Olives and Kawachi 2017). To that end, we introduce the concept “upward status heterophily,” perceived frequent interaction with people (nonkin others) who are socioeconomically higher.

Our research has a twofold purpose. First, using large cross-national data, we examine the associations between upward status heterophily and (physical and mental) health outcomes. Second, we explore whether these associations vary across individual (respondent-level) and contextual (country-level) moderators. When gauging the impact of social capital, prior research has mostly operationalized variables such as network size or degree of social engagement. Heuristically useful, these measures do not capture perceptions of status distinction among network alters. They refer to the quantity of connections—how many people one knows, for example—but do not reveal the qualitative nature of those connections. That is, network size in and of itself cannot gauge the relative social standings of one’s (ego’s) contacts (alters). As a result, studies based on such variables fall short of demonstrating the health implications of ties to others of unequal (i.e., higher) socioeconomic status. Our study contributes to the scholarship by analyzing this critical yet underexamined factor: perceived status differentials between ego and alters.

Conventional wisdom has it that network connections with actors located higher on the status hierarchy can serve as a health-promoting conduit. Undoubtedly, this idea is fundamental to the research on social capital and health (Kawachi and Berkman 2003). Social relationships are deemed important generally. But more important is knowing people in “desirable” (i.e., socioeconomically upward) positions—because they are more capable of channeling the kinds of resources that really matter for personal health and well-being (Moore et al. 2009; Song 2011). Contrary to this commonly accepted view, we hypothesize that upward status heterophily may lead to harmful, albeit unanticipated, consequences. The reason, we contend, is relative deprivation: Knowing and interacting with higher status alters can make the focal actor (ego) feel inferior emotionally and psychologically (Smith et al. 2012). Upward social comparison and envy, in other words, can produce deleterious results (Lee and Kawachi 2017).

Based on representative samples drawn from East Asian countries, we show that upward status heterophily is negatively associated with both subjective well-being and self-rated health. To our knowledge, no prior study has systematically investigated the direct and interactive processes underlying health implications of perceived status differentials using multilevel, comparative data.³ For the most part, existing research “focuses on the supportive function of personal relationships” (Offer and Fisher 2018:112). In this article, we show that social capital constitutes a double-edged phenomenon by shifting our attention to the potential “downside of social relationships” (Umberson and Montez 2010). To that end, we emphasize the mechanisms by which social capital can have deleterious consequences by operationalizing perceived frequent interaction with higher status contacts as the main predictor.

Background

Social Capital and Health Outcomes

The basic premise of social capital theory is that ties to others—in particular, those in upper socioeconomic positions—can confer information and resource benefits for attaining one’s goals (Lin and Erickson 2012; Verhaeghe and Li 2015). That is, the status position of network alters is a positive correlate of the quality of social capital available through them (Lin 2001; Song and Pettis 2020). People with greater access to social capital are shown to be healthier, happier, and more satisfied in life (e.g., Hoogerbrugge and Burger 2018; Huang et al. 2019; Maass et al. 2016; Moore et al. 2018; Story and Glanville 2019). There are “behavioral,” “psychosocial,” and “physiological” pathways through which interpersonal relationships can influence health outcomes (Umberson and Montez 2010). Social ties are also essential for the emergence of “emotional sustenance” and “active coping assistance” (Thoits 2011). In sum, a theoretical common denominator cuts across myriad empirical findings: Social capital provides a stress-buffering function that enhances physical and mental health.

This prevailing narrative, however, tends to overlook and not problematize an essential fact of social life: Interpersonal ties and the benefits of social capital are not costless. Clearly, no relationship can be maintained without some degree of reciprocal obligation. That is, one cannot unilaterally take advantage of relationally embedded resources without contributing to them. An unreciprocated tie is bound to disintegrate over time. A critical aspect of human interaction is reciprocity because what you received is often expected to be repaid, which can lead to relational stress (Huang et al. 2019; Jou and Fukuda 2002; Offer 2012). As such, interaction with high status others can certainly be beneficial. However, it comes with a cost by way of having to reciprocate the received favor in the future (Offer and Fisher 2018). Aside from obligatory reciprocity, upward status heterophily can result in something even more burdensome or costly: relative deprivation.

In this study, we expound and demonstrate that people can experience deleterious consequences because of their (unequal) network connections. Our proposed mechanism is social comparison, a universal constant (Baldwin and Mussweiler 2018), which, under unfavorable conditions, can produce feelings of envy and relative deprivation (Alderson and Katz-Gerro 2016; Olivos, Olivos-Jara, and Browne 2021; Pham-Kanter 2009; Smith and Huo 2014; Song 2015). And this outcome is more likely to the extent that the object of comparison is someone higher on the socioeconomic hierarchy. Our study thus highlights that while they “are the central source of emotional support for most people, social relationships can be extremely stressful” (Umberson and Montez 2010:S57).

Although previous research is mostly limited to health-promoting dimensions of social connectedness (Offer and Fisher 2018:112; see also Rook 1984), a growing number of studies has emerged underscoring its harmful aspects. At the individual level of analysis, for example, it is found that “relational turbulence” negatively influences psychological and behavioral health (Weigel and Shrout 2020), “unsolicited support” from others can induce distress (Song and Chen 2014), and “relational constraints and exclusion” lower subjective well-being (Huang et al. 2019). Moreover, access to higher status others is linked with poor physical health (Pham-Kanter 2009), excessive demands from others raise loneliness among older adults (Kim and Jung 2022), and higher accessed status positively relates to depression especially in more collectivistic societies (Song and Pettis 2020). Others have conceptualized and operationalized the focal concept at an aggregate or ecological level, thereby showing, for example, that neighborhood bridging social capital leads to higher depression (Murayama et al. 2015), low-trust individuals face reduced health in high-trust societies (Campos-Matos, Subramanian, and Kawachi 2016), and stronger social cohesion predicts more depressive symptoms for newer members of the community (Takagi et al. 2013).

Also related to the main purpose of this study, a literature exists on the specific link between social comparison/relative deprivation and health outcomes (e.g., Alderson and Katz-Gerro 2016; Firebaugh and Schroeder 2009; Pham-Kanter 2009; Song and Petis 2020; Yang, Hu, and Schieman 2019). The basic argument is that some people are less healthy and/or less happy not only because of their limited financial means but because of the perceived gap between themselves and those who are better off in material terms. According to a systematic review, there are several processes by which the dark side of social capital operates, including restrictions on individual freedom, negative behavioral contagion, exclusion of outsiders, and excessive claims on individual members (Villalonga-Olives and Kawachi 2017; see also Portes 1998). A more recent review highlights role strain, asymmetric social relations, burdensome relational investment, and demanding social obligations, among others (Song et al. 2021).

Although the articles analyzed in these two review articles highlight the downside of social ties, none explicitly conceptualizes or operationalizes the ego-alter status distinction in the context of network relations. Nor do they systematically discuss the dual mechanisms of social comparison and relative deprivation with respect to (perceived) status differentials. To the extent that this issue is addressed at all, status differentials are often measured in terms of occupational prestige scores based on the position generator (see Song et al. 2021) or indirectly gauged from positions of social status. Moreover, prior research mostly relies on conventional social capital measures—such as interpersonal trust, network size, and organizational membership—in making inference about its reputedly harmful impact based on their negative statistical associations with health outcomes (see Villalonga-Olives and Kawachi 2017).

In short, despite a burgeoning scholarship on the potential health-related costs of social capital, empirical evidence remains limited when it comes to the unanticipated consequences of perceived interaction with socioeconomically superior others. The current study fills this gap by explicitly investigating the role of upward status heterophily in the unequal distribution of health and well-being across East Asia. In formulating our hypotheses, we draw on the social cost theoretical framework laid out by Song et al. (2021), which prioritizes the “social cost model.” As they point out, earlier studies are largely based on the “social resource model” emphasizing health benefits of relational connectedness. By contrast, our research shifts the analysis toward costs and disadvantages stemming from the perceived status difference between ego and alters. That is, we examine whether and to what extent upward status heterophily—as a component of the structural composition of ego’s networks (Song et al. 2021)—can compromise mental and physical health by way of reinforcing hierarchical social comparison and thereby exacerbating feelings of relative deprivation.

Research Hypotheses

For the most part, prior studies on the link between social comparison and health outcomes measure the “perceived gap” by using subjective socioeconomic status (SES), or relative income, as the main explanatory variable (for systematic review, see Tan et al. 2020). As such, they view low status in and of itself as a health-damaging stressor (Yang et al. 2019). By contrast, we use a behavioral measure based on the perceived frequency of interaction with high status others, a prerequisite for the interplay between unfavorable social comparison, envy, and relative deprivation (Gartrell 2002; Lee and Kawachi 2017). Put differently, we conceptualize upward status heterophily as a proxy for the mechanism of “asymmetric social comparison” (Olivos et al. 2021) with potentially deleterious implications. Based on the foregoing discussion, we present our first set of hypotheses as follows:

Hypothesis 1: Upward status heterophily is negatively related to subjective well-being (SWB).

Hypothesis 2: Upward status heterophily is negatively related to self-rated health (SRH).

If our hypotheses are valid—if relative deprivation via social comparison/relative deprivation does undergird the proposed relationships—then their strengths should increase (decrease) under higher (lower) measures of “vulnerability.” That is, the focal relationships ought to vary partly as a function of how more or less vulnerable the individual subjects are in our data. More specifically, the negative associations between upward status heterophily and outcomes should be greater for those who are more vulnerable or susceptible to relative deprivation. To test this possibility, we introduce three individual-level (self-efficacy, educational attainment, and size of nonkin contacts) and one contextual-level (regional inequality) moderators.

First, self-efficacy is shown to be related negatively to depression but positively to life satisfaction (Tak et al. 2017). According to social cognitive theory, self-efficacy refers to the general belief in one’s capacity to be in control of and to perform the tasks necessary toward a desired end (Bandura 2012). It is thus intricately related to a person’s overall confidence in competently dealing with the social environment, including various life stressors (Sahu and Rath 2003). In other words, an individual with a relatively higher sense of self-efficacy would be less vulnerable to feelings of relative deprivation. By contrast, another with lower perceived self-efficacy—one who lacks control, competence, and confidence—would fall more susceptible to the negative consequences of upward social comparison. As previously mentioned, this can create feelings of inferiority coupled with negative self-evaluation and discontent. Self-efficacy can thus serve as a buffer by providing protection against its adverse effects on health outcomes. This reasoning, then, leads to our third hypothesis:

Hypothesis 3: The negative relationship between upward status heterophily and SWB/SRH is stronger for persons with lower levels of perceived self-efficacy.

Second, along with income and occupation, education constitutes one of the main indicators of objective SES. Health benefits of education are well known: Adults who are better educated live healthier and longer lives (for review, see Zajacova and Lawrence 2018). In contemporary society, education is a power determinant of stratification and inequality (Fischer and Hout 2006) since educational institutions function as “sorting machines” that place people into hierarchical slots (Domina, Penner, and Penner 2017). Given this role, we treat education as an important source of protective security that can guard against, or lessen, the feelings of deprivation stemming from unhealthy social comparison. All things equal, compared to a better educated counterpart, a less educated person would face higher odds of feeling the burden of reciprocity (in case a favor was distributed) and, at the same time, the stress of negative self-perception in the presence of socially higher others or those who are more “successful” (see Fiske 2011:57–63). Based on this argument, we thus anticipate the following:

Hypothesis 4: The negative relationship between upward status heterophily and SWB/SRH is stronger for persons with lower levels of educational attainment.

As the last individual-level moderator, we use a variable measuring the total number of daily nonkin contacts. If ties to higher status people can produce mental stress and emotional anxiety, then knowing more people who are not of higher status ought to help diffuse some of the noxious effects. Imagine Person A who interacts regularly with 20 nonkin contacts, whereas Person B has 40 such daily contacts. Let us assume that for both, the proportion of “higher status alters” is constant at .8. That is, Person A has four nonkin contacts who are not socially higher; for Person B, that number is eight. In this hypothetical case, Person B has twice the number of nonkin contacts who are on par with or below his or her own status than does Person A. In this scenario, the level of status heterophily would also be greater for Person B. We conjecture, though, that the effect of social interaction with higher status others does not rise monotonically. Rather, there may be a curvilinear (i.e., an inverted J-curve) relationship. That is, we anticipate a “law of diminishing marginal returns” such that after a certain threshold, adding an additional high-status alter would not significantly increase the degree of relative deprivation felt by the focal actor (ego). The lack of data, unfortunately, prevents us from directly testing this.

The main point here is that larger social networks (containing homophilous alters), all things equal, should help dilute the adverse consequences of upward status heterophily. Even if one should, willingly or not, socialize with many who are socioeconomically higher, that individual may find counterbalancing protection and comfort among social equals. Another possibility is that having many nonkin ties (larger network) is a proxy for popularity or sociability, both of which are known to have salubrious effects.⁴ Conversely, one who has a relatively small nonkin network would be less able to counterbalance the emotional and mental costs associated with ties to socially superior others. In short, we expect that “status homophily” (interaction among similar others) can serve to protect against “status heterophily” (interaction with dissimilar [i.e., unequal] others). This discussion introduces our fifth hypothesis:

Hypothesis 5: The negative relationship between upward status heterophily and SWB/SRH is stronger for persons with smaller numbers of total nonkin contacts.

Lastly, we consider the possibility that the proposed link between upward status heterophily and health measures may shift across regional-level inequality. According to research, health prospects are generally worse in more unequal places (Rambotti 2015). More specifically, the poor, on average, experience worse outcomes, and this relationship grows in societies with greater national inequality (Wilkinson and Pickett 2006). Across more than 300 peer-reviewed studies, evidence points to such a “causal connection” (Pickett and Wilkinson 2015), which exists cross-nationally net of poverty level (Rambotti 2015). Growth in inequality can change people’s own income, hence directly impacting health prospects. Indirectly, by altering others’ income, it can also shift the health outcomes of those whose own income remains unchanged—the latter shaping the slope of the income–health relationship (for review, see Truesdale and Jencks 2016). Importantly, for our purposes, “larger income differences increase social distances, accentuating social class or status differences,” leading to devastating consequences in terms of physical and mental health (Pickett and Wilkinson 2015:323).

Our main hypothesis is that interaction with higher status others can be harmful. If so, this proposed relationship ought to become amplified in economically more unequal contexts, where social distances are farther and status differences more pronounced among individual members (Melita, Willis, and Rodríguez-Bailón 2021; Truesdale and Jencks 2016). Extending the logic, in places with relatively low economic inequality (e.g., where social distances are closer), it can be conjectured that a higher proportion of individuals of unequal status may interact with one another. But as inequality grows, social division and in-group solidarity rise with it, thereby making this less likely—that is, making people of different social status unlikely to socially mingle (Kawachi and Kennedy 1997; Lin et al. 2017; Rothstein and Uslaner 2005). When that happens, we conjecture that access to higher status individuals would become more limited and competitive, shifting the power balance in favor of those who occupy higher socioeconomic positions. In the end, this situation would create greater relational burden and stress on the part of lower status (focal) actors, which leads us to our final hypothesis:

Hypothesis 6: The negative relationship between status heterophily and SWB/SRH is stronger in regions that are more economically unequal.

Data and methods

Data Source

Testing our hypotheses required data on the relative social status of network partners, namely, information on interaction with “higher status others.” To that end, we utilized the East Asian Social Survey 2012 (EASS 2012) designed to provide comparative data across the East Asian region consisting of nationally representative samples on China, Japan, South Korea, and Taiwan. To the authors’ knowledge, no other cross-national survey contains the information necessary to test our hypotheses.⁵ Modeled after the General Social Survey, EASS is conducted biennially centered on a special theme or module. The 2012 version contains the “Module on Network Social Capital,” the only one of its kind ever conducted across the four countries, offering a wealth of survey information on networks and related sources among the East Asian population (for details, see Bian and Ikeda 2018).

We obtained the micro datafile for EASS 2012 from its online data archive (http://www.eassda.org). Multistage area stratified probability-proport ional-to-size method was used for data collection. Face-to-face interviews were conducted in the local language. Response rates were 71% for Chinese General Social Survey (N = 5,819), 59% for Japanese General Social Survey (N = 2,335), 56% for Korean General Social Survey (N = 1,396), and 52% for Taiwanese Social Change Survey (N = 2,134). Importantly, EASS 2012 provides geocodes across 65 subnational regions, in which individual respondents are nested. To address cases with missing values, we used multiple imputation by chained equation to create 10 multiply imputed data sets (Royston and White 2011).⁶ The two dependent variables (health and well-being) and the main predictor (upward status heterophily)—along with age, gender, and marital status—were not imputed. Poststratification weights (that match on age, gender, and employment status of samples to relevant country characteristics) were used to adjust for probabilities of selection, sampling error, and potential nonresponse bias. The analytic sample had 11,112 residents living in 65 regional clusters across four countries.

Measures

There were two dependent variables: self-rated health and subjective well-being. The former was based on answers coded on a 5-point scale (e.g., 1 = very unhealthy, 3 = neither unhealthy nor healthy, 5 = very healthy) to a survey item about personal health status. The latter, also coded on a 5-point scale (e.g., 1 = very unhappy, 5 = very happy), was based on a question about general happiness (i.e., “a person’s cognitive and affective evaluations of his or her life”; Diener, Lucas, and Oishi 2002:63). Our main predictor, upward status heterophily, was operationalized using the following question in the EASS 2012 Module on Network Social Capital:

Among non-kin contacts with whom you frequently socialize, which of the following best describes the social characteristics of these people? From your viewpoint, [Circle One]: (1). You have contact more with people who are socially higher than those who are lower; (2). The majority of people with whom you have contact are socially equivalent to you; (3). You have contact more with people who are socially lower than those who are higher.

Original answers were dichotomized so that the Choice 1 is coded 1 and all other responses are coded 0. Consistent with sociological expectations, most social interactions in East Asia are driven by social equivalence or homophily: About 85% of the sample chose Choice 2. About 1 in 10 are frequently engaged in heterophilous interactions with higher status others.⁷

For a conservative and unbiased estimate of the associations between upward status heterophily and the two outcomes (SRH and SWB), our models adjusted for a variety of demographic, socioeconomic, and social capital variables at individual and regional levels—including the three moderators (education, self-efficacy, and nonkin network size). At the individual level, we adjusted for the following demographic controls: age (in years), female, marital status (coded 1 if married), employment status (coded 1 if currently working), religiosity (coded 1 if religiously affiliated; reference = religious none), and urban residence (coded 1; reference = suburban and rural). Also included was a measure for self-efficacy to tap whether survey participants believe they have the “power to make important decisions to change the course of your life” (e.g., 4 = mostly able to change). The list of socioeconomic controls included education (years of formal schooling), subjective social standing (coded on a 10-point, ladder-type scale, where, for example, 1 = lowest and 10 = highest position in society), household income (converted to 2012 USD and log-transformed), and relative income rank (measured by the household income rank within the subnational region of residency).

Social capital was gauged using standard measures found in the literature, such as kin network (“On an ordinary day, with how many family members or relatives, excluding those who live with you, do you have contact through telephone, mails, internet, or face-to-face?”) and nonkin network (“On an ordinary day, with how many people other than family members or relatives do you have contact through telephone, mails, internet, or face-to-face?”). Both were coded on an ordinal scale (1 = 0, 2 = 1–2, 3 = 3–4, 4 = 5–9, 5 = 10–19, 6 = 20–49, 7 = 50–99, 8 = 100 or more). Social (particular) trust was based on a question about how much respondents trust their relatives, friends, and neighbors (4 = a great deal, 3 = to some extent, 2 = not very much, 1 = hardly at all). Original answers were averaged to create an index (Cronbach’s alpha = .69).

We also operationalized organizational involvement using a series of items about the level of participation in online and offline activities via formal affiliation (political associations, residential/neighborhood associations, social service clubs, religious groups, etc.). Due to the highly (right-tailed) skewed distribution of data, we recoded the original answers to range from 0 to a maximum of 3 memberships. Sociability was another variable that may have confounded the focal relationships under consideration. It measured the frequency of eating out or having casual dinners with nonkin others (e.g., 1 = never, 3 = sometimes, 5 = very often), a proxy for outward-oriented, outgoing, and socially active nature that would have covaried with the main predictor and the outcome measures. Finally, we took advantage of the position-generator item in the data set to tap people’s access to network diversity (Van der Gaag, Snijders, and Flap 2008). The question asked whether “you have friends, relatives, or acquaintances that fit the following occupational descriptions.” The list consisted of professor, lawyer, nurse, computer programmer, middle school teacher, personnel manager, farmer, hairdresser, receptionist, and policeman. We weighted each job category using the updated International Socio-Economic Index of Occupational Status scores (Ganzeboom and Trieman 1996) and then log-transformed the summed scores to measure network prestige.⁸

At the regional level, our main predictor was Gini inequality index of the household income. In the EASS, household income was measured differently by country. For China, it was measured as a continuous variable without a top-coding. For Japan, South Korea, and Taiwan, it was measured as an interval variable with an open-ended top bracket.⁹ We thus converted the interval measure to a continuous variable by taking the midpoint of each income bracket and approximated the top-coded values with the assumption of a log-normal distribution. We limited our computation to positive income. Two other regional-level covariates controlled in our models were the mean years of education (average education) and the mean household income in 2012 USD (average income).

Analytic Strategy

EASS 2012 has a hierarchically nested data structure (individual clustered in subnational regions) allowing us to check for cross-level moderation as stated in one of our hypotheses. To do so and to address the problem of data dependence, we estimated a series of two-level mixed-effects models. As robustness checks, we also analyzed the data using ordinary least squares (OLS) regression and ordered logit models with region fixed effects and clustered standard errors. In addition, we used the inverse probability of treatment weighting (IPTW) method by first calculating the propensity scores for status heterophily and then balancing the data on “pretreatment” baseline characteristics by IPTW (Austin and Stuart 2015). In the following, we juxtapose results from these alternative strategies and those from our main analysis (mixed-effects modeling) for a stringent test of our hypotheses.

Multilevel analysis was conducted using Stata (mi estimate: mixed option; StataCorp 2019) and HLM version 8 (Raudenbush, Bryk, and Congdon 2019), yielding comparable statistical results. Findings tabulated and discussed in the following are from the Stata output. We estimated the following two-level models:

\begin{array}{l} Y_{i j} = β_{0 j} + β_{1 j} U p w a r d s t a t u s h e t e r o p h i l y_{i j} \\ + \sum^{} δ_{q} X_{q i j} + ε_{i j} \end{array}

(1)

\begin{array}{l} β_{0 j} = γ_{00} + \sum^{} γ_{0 s} W_{s j} + u_{0 j} \\ β_{1 j} = γ_{10} + \sum^{} γ_{1 s} W_{s j} + u_{1 j}, \end{array}

(2)

where $Y_{i j}$ is the predicted value of the physical/mental health outcome; $β_{0 j}$ is the random intercept varying across subnational regions; $β_{1 j}$ is the random slope of the main independent variable, upward status heterophily; and $X_{q i j}$ is the value of covariate q associated with respondent i in regional cluster j. The error term $ε_{i j}$ is the individual-level random effect assumed to be independently and normally distributed with constant variance σ². Equation 2 shows our regional-level models, where $W_{s j}$ is a vector of regional-level covariates and $u_{0 j}$ and $u_{1 j}$ are the regional random effects. We assumed that the covariance between two regional-level random effects was zero net of covariates. To make sure that influential cases in cross-sectional multilevel modeling did not affect our results (Van der Meer, Te Grotenhuis, and Plezer 2010), we applied DFBETAS diagnostic tests and controlled for outlying regions with dummy variables.

Results

Table 1 contains the descriptive statistics. Concerning the outcome variables, the mean for subjective well-being measured by the 5-point Likert scale is 3.81; for self-rated health, it is 3.51. Across four East Asian countries, the average scores for SWB (≈3.74–3.87) and SRH (≈3.37–3.57) are quite similar. With respect to our main predictor (upward status heterophily), around 11% of the respondents interact mostly with higher status others. The proportions of heterophilous interactions are relatively higher in Japan (14.3%) and Taiwan (14.1%) compared to China (9.2%) and Korea (11.7%).

Table 1.

Descriptive Statistics.

	Mean (SD)	Minimum	Maximum
Subjective well-being	3.812 (.872)	1 (low)	5 (high)
Self-rated health	3.506 (1.026)	1 (low)	5 (high)
Upward status heterophily	.113 (.317)	0	1
Demographic variables
Age	47.2 (16.7)	18	95
Female	.498 (.500)	0	1
Married	.730 (.444)	0	1
Employed	.654 (.475)	0	1
Religious	.345 (.475)	0	1
Socioeconomic variables
Years of schooling	10.41 (4.60)	0	23
Subjective social standing	4.550 (1.726)	1 (lowest)	10 (highest)
Relative HH income rank within region	6.145 (3.109)	1 (bottom 10%)	10 (top 10%)
Log HH income (2012 USD)	9.152 (3.232)	0	13.81
Social capital variables
Kin network	2.250 (1.103)	1 (low)	8 (high)
Nonkin network	3.433 (1.697)	1 (low)	8 (high)
Particular trust	3.153 (.488)	1 (low)	4 (high)
Number of organizations	.750 (1.058)	0 (low)	3 (high)
Self-efficacy	2.670 (.881)	1 (low)	4 (high)
Sociability	2.301 (1.040)	1 (low)	5 (high)
Network prestige	4.346 (1.685)	0	6.342
Regional variables
Gini inequality index	.426 (.064)	.291	.630
Average income (USD)	29,873 (25,956)	3,763	79,481
Average education	2.759 (.819)	1.00	5.25
Total sample size (N regions)	11,112 (65)
Sample size by country (n regions)
China	5,601 (29)
Japan	2,267 (6)
Korea	1,396 (13)
Taiwan	1,848 (17)

Note: The data source is the East Asian Social Survey 2012. Sample weights are applied. Missing values are imputed by chained equations (10 times). The dependent variable and the main independent variable, heterophilic interaction, are not imputed. Age, gender, and marital status are not imputed either. HH = household.

Upward Status Heterophily and Health Outcomes

To examine how social interaction with higher status (nonkin) others is associated with SWB and SRH, we first estimate linear mixed-effects models, as shown in Table 2. The intercepts of SWB and SRH and slopes for upward status heterophily vary substantially across the 65 subnational regions, justifying the use of multilevel modeling. The introduction of the individual-level covariates accounts for 5% to 20% of the random components. According to the bivariate results in Models 1 and 3, respectively, the main predictor is negatively associated with SWB and SRH. That is, compared to those whose daily interactions are mostly with people of “similar” or “lower status” others, people who are in regular contact with socially superior others are significantly less happy and less healthy. The average raw SWB score for status homophily is 3.83, while that for (upward) status heterophily is 3.66. Adjusting for an extensive set of individual- and group-level covariates as shown in Models 2 and 4—including sociodemographic, SES, social capital, and regional covariates—attenuates the magnitude of the association between upward status heterophily and SWB (from –.143 in Model 1 to –.081 in Model 2, a reduction of 43% in effect size). The significantly negative relation, however, remains unchanged. As for the association between upward status heterophily and SRH, it becomes slightly enhanced from –.084 (in Model 3) to –.102 (in Model 4), indicating the presence of suppressor effect. For both outcomes, the central finding is that net of individual and regional confounders and country fixed effects, network interaction with alters located relatively higher on the status hierarchy does not enhance, as predicted in the mainstream literature on social capital, but diminishes one’s health and well-being in support of our two main hypotheses (Hypothesis 1 and Hypothesis 2).

Table 2.

Mixed Linear Model Estimates of Upward Status Heterophily for Subjective Well-Being and Self-Rated Health.

	Subjective Well-Being		Self-Rated Health
	Model 1	Model 2	Model 3	Model 4
	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)
Upward status heterophily	–.143*** (.038)	–.081* (.035)	–.084* (.041)	–.102* (.041)
Demographic covariates
Age		–.023*** (.004)		–.031*** (.005)
Age-squared		.0002*** (.000)		.0002*** (.0000)
Female		.020 (.017)		–.081** (.026)
Currently married		.186*** (.030)		.106*** (.029)
Currently employed		–.068 (.075)		.035 (.029)
Religious		.011 (.023)		–.040 (.026)
Living in urban area		–.002 (.023)		.030 (.028)
Socioeconomic status covariates
Years of schooling		.006 (.003)		.019** (.005)
Subjective social standing		.106*** (.008)		.070*** (.009)
Relative income rank in region		.027*** (.004)		.015** (.006)
Log household income		–.007 (.007)		.000 (.008)
Occupation (ISCO one-digit)		Controlled		Controlled
Social capital covariates
Kin network		.028*** (.008)		.053*** (.011)
Nonkin network		.009 (.006)		–.008 (.006)
Particular trust		.234*** (.024)		.128*** (.026)
Number of organizations		.041*** (.009)		.020 (.012)
Self-efficacy		.072*** (.013)		.065*** (.014)
Sociability		.037*** (.010)		.058*** (.014)
Network prestige		.003 (.006)		–.006 (.007)
Regional covariates
Gini		–.191 (.330)		.106 (.355)
Mean education		.008 (.048)		.037 (.355)
Mean income		.000 (.000)		–.000 (.000)
Country fixed effects	Controlled	Controlled	Controlled	Controlled
Constant	3.887*** (.035)	2.550*** (.271)	3.586*** (.049)	3.123*** (.283)
Random components
SD (heteophilic interaction)	.199 (.038)	.188 (.034)	.180 (.046)	.160 (.044)
SD (constant)	.188 (.027)	.113 (.023)	.188 (.025)	.142 (.021)
SD (residual)	.853 (.013)	.800 (.012)	1.000 (.018)	.921 (.012)
N regions (second level)	65	65	65	65

Note: The data source is the East Asian Social Survey 2012. Sample size is 11,112 for all models. Missing values are imputed by chained equations (10 times). The dependent variable and the main independent variable, heterophilic interaction, are not imputed. Age, gender, and marital status are not imputed either. Sampling weights are applied. The influential regions—one region for the models of subjective well-being and two regions for the models of self-rated health—are controlled by dummy variables. Stata’s mixed models are applied.

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

Estimating Within- and Cross-Level Interaction Effects

Next, in Table 3 (with SWB) and Table 4 (with SRH), we explore whether the association between upward status heterophily and the outcome varies across individual and regional conditions. To that end, three individual-level moderators (years of schooling, self-efficacy, and the size of nonkin network) and one region-level moderator (index of household income inequality) are introduced in the analysis. Our general argument, as expressed in Hypotheses 3 to 5, is that the negative associations between heterophilous interaction and SWB/SRH would be stronger (weaker) among individuals who are in a more (less) “vulnerable” position (i.e., who are more [less] susceptible to relative deprivation). According to Table 3, estimated coefficients of the three interaction terms between status heterophily and within-region moderators (education, self-efficacy, nonkin network) are all positive, while the main effects of status heterophily are strongly negative. Figures 1a to 1c illustrate how the link between upward status heterophily and SWB shifts across the susceptibility levels of relative deprivation while holding constant other covariates. Specifically, the negative impact of network ties to higher status others is weaker or even positive for those who are better educated (Figure 1a based on Model 1), possess a higher sense of perceived self-efficacy (Figure 1b based on Model 2), and have a larger nonkin network (Figure 1c based on Model 3).

Table 3.

Mixed Model Estimates of Various Moderators on the Effects of Upward Status Heterophily on Subjective Well-Being.

	Model 1	Model 2	Model 3	Model 4
	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)
Control variables
Demographic covariates	〇	〇	〇	〇
Socioeconomic status covariates	〇	〇	〇	〇
Social capital covariates	〇	〇	〇	〇
Regional covariates	〇	〇	〇	〇
Upward status heterophily	–.331*** (.075)	–.319** (.094)	–.200** (.069)	.628** (.205)
Moderator
Years of schooling	.004 (.003)
Self-efficacy		.060*** (.014)
Nonkin network			.004 (.006)
Regional Gini inequality				–.058 (.328)
Interaction between heterophily and moderator
Heterophily × Years schooling	.024*** (.006)
Heterophily × Self-efficacy		.087** (.030)
Heterophily × Nonkin network			.033* (.016)
Heterophily × Gini				–1.658** (.483)
Random components
SD (heterophilic interaction)	.163 (.041)	.143 (.034)	.186 (.035)	.143 (.034)
SD (second-level constant)	.132 (.023)	.140 (.021)	.131 (.023)	.140 (.021)
SD (residual)	.800 (.012)	.801 (.012)	.800 (.012)	.801 (.012)
N regions (second level)	65	65	65	65

Note: The data source is the East Asian Social Survey 2012. Sample size is 11,112 for all models. Missing values are imputed by chained equations (10 times). The dependent variable and the main independent variable, heterophilic interaction, are not imputed. Age, gender, and marital status are not imputed either. Sampling weights are applied. Demographic, socioeconomic status, and social capital covariates are the same as Table 2. Country fixed effects are controlled for. Regional covariates include Gini, regional mean levels of education, and regional mean household income. The influential region—one region for the models of subjective well-being—is controlled by a dummy variable. Stata’s mixed models are applied.

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

Table 4.

Mixed Model Estimates of Various Moderators on the Effects of Upward Status Heterophily on Self-Rated Health.

	Model 1	Model 2	Model 3	Model 4
	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)
Control variables
Demographic covariates	〇	〇	〇	〇
Socioeconomic status covariates	〇	〇	〇	〇
Social capital covariates	〇	〇	〇	〇
Regional covariates	〇	〇	〇	〇
Upward status heterophily	–.244* (.095)	–.178 (.113)	–.210** (.080)	.500** (.192)
Moderator
Years of schooling	.017** (.006)
Self-efficacy		.061*** (.015)
Nonkin network			–.012 (.007)
Regional Gini inequality				.233 (.424)
Interaction between heterophily and moderator
Heterophily × Years schooling	.014 (.007)
Heterophily × Self-efficacy		.028 (.033)
Heterophily × Nonkin network			.029 (.018)
Heterophily × Gini				–1.408** (.424)
Random components
SD (heterophilic interaction)	.139 (.046)	.160 (.044)	.145 (.045)	.138 (.047)
SD (second-level constant)	.142 (.021)	.143 (.021)	.142 (.021)	.142 (.021)
SD (residual)	.921 (.012)	.921 (.012)	.921 (.012)	.921 (.012)
N regions (second level)	65	65	65	65

Note: The data source is the East Asian Social Survey 2012. Sample size is 11,112 for all models. Missing values are imputed by chained equations (10 times). The dependent variable and the main independent variable, heterophilic interaction, are not imputed. Age, gender, and marital status are not imputed either. Sampling weights are applied. Demographic, socioeconomic status, and social capital covariates are the same as Table 2. Country fixed effects are controlled for. Regional covariates include Gini, regional mean levels of education, and regional mean household income. The influential regions—two regions for the models of self-rated health—are controlled by dummy variables. Stata’s mixed models are applied.

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

Figure 1.

Varying Associations between Upward Status Heterophily and Subjective Well-Being. (a) Years of Schooling and Subjective Well-Being. (b) Self-Efficacy and Subjective Well-Being. (c) Nonkin Network and Subjective Well-Being. (d) Gini Inequality and Subjective Well-Being.

In Model 4, we evaluate whether the heterophily–SWB relationship fluctuates across the regional income inequality. The main effect of upward status heterophily is significantly positive, but its interaction with regional inequality is significantly negative. That is, the magnitude of the focal relationship increases in a region where income inequality is higher. In regions where it is sufficiently low, however, frequent contact with higher status others is positively associated with well-being. This is graphically displayed in Figure 1d. Substantively, among those whose daily interactions are mostly with similar or lower status others, an increase in the Gini score is not associated with the change in SWB (i.e., regional inequality is not necessarily related to lower well-being without upwardly heterophilous interaction). The estimated threshold for the Gini index (i.e., where the solid red line crosses the horizontal zero line in Figure 1d) is .393 (= –.628 / [.058 – 1.658]), after which the effect of upward status heterophily becomes increasingly more negative. The average regional Gini index in the sample is .426, with the range between .291 and .630. Out of 65 subnational regions in our study, the index scores in 21 regions are below the threshold. Based on these results, we conclude that Hypothesis 3, Hypothesis 4, Hypothesis 5, and Hypothesis 6 specifically in relation to SWB receive empirical support.

Next, as summarized in Table 4, we proceed with investigating how and the extent to which the association between upward status heterophily and SRH varies partly as a function of the same set of individual- and region-level moderators. Contrary to the earlier case with SWB, individual-level moderators do not significantly moderate the association under investigation. Parameter estimates for the interaction terms in Models 1 to 3 all fall below the conventional level of statistical significance, while the estimate for upward status heterophily remains significantly negative, except in Model 2. Turning to the moderating role of regional income inequality concerning SRH in Model 4, the result is akin to what we reported earlier with respect to SWB. Figure 2 (based on Model 4 in Table 4) provides a visual description of this contingent relationship. The association is strongly negative for those who interact mostly with higher status others. Even for individuals whose main contacts are of similar or lower status, regional inequality is negatively associated with self-rated health, although the association is not statistically significant. Regional household income inequality worsens the negative impact of social interaction with higher status others. The main effect of upward status heterophily in Model 4 is significantly positive. However, the crossover point on the Gini index where its influence changes from positive to negative is .426 (= –.500 / [.233 – 1.408]). Out of the total 65 subnational regions, more than a half of them (35) exhibit the Gini index higher than this threshold. All in all, Hypotheis 6 is further borne out empirically with respect to SRH.

Figure 2.

Association between Upward Status Heterophily and Self-Rated Health Moderated by Regional Income Inequality.

Robustness Checks

Our main analyses are based on two-level linear mixed models. In this section, we analyze the data using alternative models to check whether our results are sensitive to methodological strategies. Also, based on the subset of employed respondents only, we calculated the difference between ego’s own occupational prestige score and the average score of alters’ respective occupational prestige. The correlation between upward status heterophily and the average score is .46, indicating that the subjective measure of status difference and the objective measure based on occupational prestige scores are related. As expected, the measured gap is larger for respondents with upward status heterophily (6.6 in terms of Treiman Prestige Score) than that for the reference group (3.5). However, we found that this alternative measure based on the prestige score difference was not significantly related to the outcome (models not shown). A plausible reason may be that the position generator item asks survey participants whether they have friends, family, or acquaintances in one of the listed job categories. As such, it is entirely possible to select, for example, five categories based on socially distant ties (acquaintances) or socially intimate ties (close friends), making it problematic to accurately gauge the basis of upward social comparison.

As a further robustness check, we reassess the associations between the main predictor and outcome measures using the OLS models with region fixed effects and clustered standard errors. Although mixed models incorporate random components, when it comes to cross-level interactions, OLS regression with clustered standard errors may be a preferred strategy over multilevel analysis (Primo, Jacobsmeier, and Milyo 2007). As shown in Table 5, the earlier results are replicated, confirming our main thesis. The dependent variables of this study, SWB and SRH, are coded on a 5-point Likert scale. To see whether our conclusions are altered when the violation of the assumptions of the linear models is factored in, we also estimate ordered logit models with region fixed effects and clustered standard errors. Once again, all our hypotheses are supported.

Table 5.

Ordinary Least Squares and Ordered Logit Estimates of Upward Status Heterophily for Subjective Well-Being and Self-Rated Health.

	Ordinary Least Squares	Ordered Logit	IPW
	With Region Fixed Effects and Clustered SE	With Region Fixed Effects and Clustered SE	Balanced on Individual and Regional Covariates Including Region Dummies
	Coefficient (SE)	Coefficient (SE)	Coefficient (SE)
I. Subjective well-being
Model 1
Upward status heterophily	–.353*** (.069)	–.739*** (.156)	–.335*** (.069)
Years of schooling	.004 (.003)	.003 (.008)	.019*** (.004)
Heterophily × Years school	.025*** (.006)	.058*** (.013)	.025*** (.006)
Model 2
Upward status heterophily	–.341*** (.091)	–.682** (.200)	–.313** (.105)
Self-efficacy	.058*** (.014)	.153*** (.035)	.109*** (.016)
Heterophily × Self-efficacy	.095** (.030)	.206** (.067)	.083* (.035)
Model 3
Upward status heterophily	–.210** (.065)	–.384** (.140)	–.175* (.066)
Nonkin network	.003 (.006)	.012 (.015)	.024** (.007)
Heterophily × Nonkin network	.034* (.014)	.069* (.031)	.025 (.016)
Model 4
Upward status heterophily	.588** (.187)	1.361** (.478)	.393 (.215)
Gini	.331 (74.1)	–1.994 (.787)	–.219 (.397)
Heterophily × Gini	–1.585** (.444)	–3.510** (1.118)	–1.131* (.489)
II. Self-rated health
Model 4
Upward status heterophily	.478* (.210)	.889* (.395)	.431 (.260)
Gini	–.029 (.429)	.172 (.845)	–.632 (.606)
Heterophily × Gini	–1.352** (.464)	–2.490** (.869)	–1.281* (.576)
N regions (clusters)	65	65	65

Note: The data source is the East Asian Social Survey 2012. Sample size is 11,112 for all models. Missing values are imputed by chained equations (10 times). The dependent variable and the main independent variable, heterophilic interaction, are not imputed. Age, gender, and marital status are not imputed either. Sampling weights are applied. Demographic, socioeconomic status, and social capital covariates are the same with Table 2. Regional covariates include Gini, regional mean levels of education, and regional mean household income. IPW = Inverse Propensity Weighting.

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

Given the cross-sectional nature of our data, we cannot draw causal inference from these results. As a further check of robustness, we assess whether the reported findings are driven mainly by nonrandom selection into the “treatment group” of upwardly heterophilous interaction by applying the IPTW technique (Austin and Stuart 2015; Granger, Sergeant, and Lunt 2019). Using the balanced data produces results that are relatively weaker in terms of the statistical significance compared to those from the mixed-effects models. Nevertheless, findings from the IPTW approach are consistent with those from our main analyses. In addition to the ones already discussed, we estimated another interaction between upward social status and occupational category (managerial and professional coded 1; 0 otherwise). For both health outcomes, the estimate is significantly positive: The harmful effect of upward social comparison is weaker among employed individuals with higher occupational status. In other words, managerial and professional occupations provide a buffering mechanism. These results (Appendix 2 in the online version of the article) provide further evidence in support of our argument.

Lastly, we estimate country-disaggregated models to visualize where the results vary across countries (see Appendix 3 in the online version of the article). Coefficients for the individual-level moderators are mostly consistent across countries, although the significance levels are diminished due to the relatively smaller sample sizes for each country in comparison with the pooled data. As for the region-level moderator, Gini inequality index, results for China, Korea, and Taiwan are consistent with what was previously reported. Results for Japan, on the other hand, deviate from the general pattern. It should be noted that the number of subnational regions in Japanese data is only six, far below what would be necessary to reliably estimate cross-level interactions. Part of Japan’s peculiarity has to do with the country’s strong education credentialism. Those who attended school but did not acquire a matching degree show a substantially large negative impact on health assessment. While these cases account for only 100 respondents, their influence on the estimates is substantial.

Conclusion

According to conventional wisdom, people connected to those who are better positioned on the socioeconomic hierarchy ought to gain more. Conversely, exclusion from such social ties is thought to result in, among others, poor physical and mental health (Song 2015; Verhaeghe et al. 2012). Against this backdrop, we explored a new topic: the downside of social capital or liability of connectedness to socioeconomically superior others. Using population-based data on four East Asian countries, we tested a set of hypotheses concerning our central claim that while interpersonal relationships can serve as a channel for resources, information, and support, they can also be a source of envy, burden, and stress. Surely, higher status others can enrich and improve one’s life chances. To the extent that the reference group is socioeconomically more superior or higher, however, unequal social comparison can produce feelings of relative deprivation (Smith and Huo 2014). Does upward status heterophily (i.e., measured by perceived interaction with more socioeconomically successful others) downgrade personal health and well-being? Using various methods and model specifications, we addressed this critical yet underexamined question and found that indeed it has deleterious consequences in the East Asian context.

Our study also revealed that the strength of the relationship between upward status heterophily and the outcome (subjective well-being, in particular) is modified by individual-level moderators, namely, education, self-efficacy, and network size. Specifically, we reasoned that individuals more prone to relative deprivation ought to suffer more from the negative impact of heterophilous interaction. And this is exactly what we found: People who are less educated, lack a strong sense of self-efficacy, and have limited nonfamilial contacts faced a greater cost of upward social comparison in terms of poor mental health. That the harmful consequence of relative deprivation is worse, for both health and well-being, in more unequal regions gives further credence to our thesis. Of course, ties to socioeconomically higher others do not, by definition, diminish personal happiness and/or health. Nevertheless, that our main findings confirmed the (within- as well as between-region) statistical moderation supports the proposition that social connections and interactions are not always “functional.” Rather, as comparative reference group theory would suggest (Song 2020), who the network alters are in relation to the ego matters. To the extent that they are located higher on the status hierarchy, thereby imposing unfavorable/unequal social comparison, they can compromise ego’s overall well-being and health.

Since our analyses use data exclusively from a particular region of the world, a question inevitably arises: Can the findings of this study be replicated using data on other (e.g., Western) countries? Recent research using East Asian data stresses the need for more comparative research (e.g., Sung and Son 2020). According to one study, the strength and direction of associations between social comparison and health outcomes may partly depend on the cultural setting (Song 2015). That is, there is a greater tendency for upward, as opposed to downward, social comparison in collectivistic rather than individualistic societies (Sasaki, Ko, and Kim 2014). If so, are the members of individualistic societies such as those in North America and Europe less susceptible to the negative impact of status heterophily vis-à-vis those in Confucian cultures? Answering this question is beyond the scope of this article. With novel findings across four East Asian countries, where social networks tend to be more hierarchical or unequal (Bian and Ikeda 2018), our study provides a rationale for broader comparative research in this area.

In analyzing EASS 2012, we encountered some data limitations. First and foremost, because of its cross-sectional design, despite adjusting for confounders, drawing conclusive causal inference is problematic due to the possibility of reverse causation. More specifically, we cannot rule out the scenario that individuals of worse health conditions intentionally seek out socioeconomically higher others. Theoretically, selection process driven by self-devaluation is possible, although empirically, it is unlikely. Forging and maintaining ties to high status members for individuals who are socioeconomically lower requires strenuous efforts. One must invest time and energy in getting to know them, which the physically and mentally weak, by definition, would find both costly and demanding. Second, EASS 2012 does not contain network information using a name generator. It would have been ideal to ask respondents to name a list of close confidants and give details on their respective social rank along with socioeconomic and demographic characteristics. The absence of such information may be a source of omitted variable bias.¹⁰ Third, also missing is contextual information with respect to workplace characteristics and community environment. Organizational size and industry types, for example, may be important determinants of network formation and interaction. Factors associated with one’s residential context may also shape the outcome.

Fourth, we lack data on the reasons for or motivations behind maintaining ties with them. There may be relevant contextual (e.g., normative or institutional) factors as well (Offer and Fisher 2018).¹¹ Data limitations, unfortunately, prevented us from probing them. Fifth, our main predictor, a self-assessed measure, may not accurately capture the precise level of social interaction with high status others. Since survey responses concerning the frequency of interaction with higher status others are based on subjective perception, they may be biased or inaccurate. Future research can benefit by directly gauging the degree to which participants have experienced relative deprivation from upward social comparison. Access to sociocentric network data would make it possible to objectively assess the number or proportion of contacts in “higher” versus “lower” status positions. Equipped with this information, the analyst can evaluate how objective measures of high status network contacts independently and interactively influence subjective assessments of happiness and health. Lastly, more refined measures of inequality at the regional level aside from what we used, the Gini coefficient, would also improve the accuracy of results.

Despite data shortcomings, aside from empirical findings, our study offers theoretical contributions. Previous research linking relative deprivation and health outcomes almost exclusively emphasize subjective social status as the key predictor (e.g., Alderson and Katz-Gerro 2016; Präg, Mills, and Wittek 2016). The limitation with this approach is that researchers must assume that people compare themselves invariably (exclusively) with those who rank higher on the status hierarchy, which may or may not be the case. By contrast, without making such an assumption, we measured and examined the impact of perceived frequency of interaction with higher status others, net of subjective social status and other controls. In addressing the issue of unequal distribution of health and well-being among individuals, the social capital argument is often invoked in the literature to explain why, all else equal, some are physically and/or psychologically better off than others. Diverging from this prevailing perspective, we demonstrated that network ties can be a source of envy and distress: Regularly socializing with nonkin contacts perceived to have higher socioeconomic characteristics is found to undermine physical and emotional well-being among East Asian adults.

An oft-cited systematic review concerning the dark side of social capital has identified two main mechanisms: negative health–behavioral contagion and interaction between social cohesion and individual-level mistrust (Villalonga-Olives and Kawachi 2017:126). Based on findings from the present study, we propose a third that has largely escaped systematic attention: dual processes of social comparison and relative deprivation. In fact, this article’s primary contribution lies in furthering our understanding of the mechanism by which the downside of social connectedness is related to health outcomes. The colloquial expression “It’s not what you know but who you know” connotes that relationships contain instrumental value. While this has some general validity across societies, our findings indicate that interpersonal networks are by no means always valuable but can be, at times, costly. Networks operate in such ways as to produce both salubrious and deleterious effects. Additional cross-national research would help us in better understanding how and the conditions under which relational ties promote and compromise the health and well-being of individuals embedded in them.

Supplemental Material

sj-docx-1-hsb-10.1177_00221465231155892 – Supplemental material for Network Ties, Upward Status Heterophily, and Unanticipated Health Consequences

Supplemental material, sj-docx-1-hsb-10.1177_00221465231155892 for Network Ties, Upward Status Heterophily, and Unanticipated Health Consequences by ChangHwan Kim and Harris Hyun-soo Kim in Journal of Health and Social Behavior

Footnotes

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: For the current study, ChangHwan Kim received support from the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2020S1A3A2A03096777).

ORCID iDs

ChangHwan Kim

Harris Hyun-soo Kim

Supplemental Material

Appendices 1 through 3 are available in the online version of the article.

Notes

Author biographies

ChangHwan Kim is professor of sociology at the University of Kansas. His areas of interests include labor-market inequality, stratification, education, and Asian Americans studies. His recent work has appeared in, among other outlets, Social Forces, Social Science Research, and Research in Social Stratification and Mobility.

Harris Hyun-soo Kim is professor of sociology at Ewha Womans University in Seoul, Korea. His primary interests include social capital, sociology of immigration, health inequality, and political participation and civic engagement. Some of his recent research has been published in Social Science & Medicine, Research on Aging, Youth & Society, and Journal of Applied Gerontology, among others.

References

Alderson

Arthur S.

Katz-Gerro

Tally

. 2016. “Compared to Whom? Inequality, Social Comparison and Happiness.” Social Forces 95(1):25–53.

Austin

Peter C.

Stuart

Elizabeth C.

2015. “Moving towards Best Practice when Using Inverse Probability of Treatment Weighting (IPTW) Using the Propensity Score to Estimate Causal Treatment Effects in Observational Studies.” Statistics in Medicine 34:3661–79.

Baldwin

Matthew

Mussweiler

Thomas

. 2018. “The Culture of Social Comparison.” Proceedings of the National Academy of Sciences 115(43):E9067–74.

Bandura

Albert

. 2012. “On the Functional Properties of Perceived Self-Efficacy Revisited.” Journal of Management 38(1):9–44.

Bian

Yanjie

Ikeda

Ken’ichi

. 2018. “East Asian Social Networks.” Pp. 679–836 in Encyclopedia of Social Network Analysis and Mining, edited by Alhajj

Rokne

New York, NY: Springer.

Campos-Matos

Ines

Subramanian

S. V.

Kawachi

Ichiro

. 2016. “The ‘Dark Side’ of Social Capital: Trust and Self-Rated Health in European Countries.” European Journal of Public Health 26(1):90–95.

Davies

James B.

Shorrocks

Anthony F.

1989. “Optimal Grouping of Income and Wealth Data.” Journal of Econometrics 42:97–108.

Diener

Lucas

Richard E.

Oishi

Shigehiro

. 2002. “Subjective Well-Being: The Science of Happiness and Life Satisfaction.” Pp. 63–73 in Handbook of Positive Psychology, edited by Lopez

S. J.

Snyder

C. R.

New York, NY: Oxford University Press.

DiMaggio

Paul

Garip

Filiz

. 2012. “Network Effects and Social Inequality.” Annual Review of Sociology 38:93–118.

10.

Domina

Thurston

Penner

Andrew

Penner

Emily

. 2017. “Categorical Inequality: Schools as Sorting Machines.” Annual Review of Sociology 43:311–30.

11.

Ehsan

Annahita

Klaas

Hannah Sophie

Bastianen

Alexander

Spini

Dario

. 2019. “Social Capital and Health: A Systematic Review of Systematic Reviews.” SSM – Population Health 8:100425. doi:10.1016/j.ssmph.2019.100425.

12.

Firebaugh

Glenn

Schroeder

Matthew B.

2009. “Does Your Neighbor’s Income Affect Your Happiness?” American Journal of Sociology 115(3):805–31.

13.

Fischer

Claude S.

Hout

Michael

. 2006. Century of Difference: How America Changed in the Last One Hundred Years. New York, NY: Russell Sage Foundation.

14.

Fiske

Susan

. 2011. Envy up, Scorn down: How Status Divides US. New York, NY: Russell Sage Foundation.

15.

Ganzeboom

Harry B. G.

Trieman

Donald J.

1996. “Internationally Comparable Measures of Occupational Status for the 1988 International Standard Classification of Occupations.” Social Science Research 25(3):201–39.

16.

Gartrell

C. David

. 2002. “The Embeddedness of Social Comparison.” Pp. 164–84 in Relative Deprivation: Specification, Development, and Integration, edited by Walker

Smith

H. J.

Cambridge, UK: Cambridge University Press.

17.

Granger

Emily

Sergeant

Jamie C.

Lunt

Mark

. 2019. “Avoiding Pitfalls When Combining Multiple Imputation and Propensity Scores.” Statistics in Medicine 38(5):1–13.

18.

Hoogerbrugge

Marloes M.

Burger

Martijn J.

2018. “Neighborhood-Based Social Capital and Life Satisfaction: The Case of Rotterdam, the Netherlands.” Urban Geography 39(10):1484–509.

19.

Huang

Xianbi

Western

Mark

Bian

Yanjie

Yaojun

Côté

Rochelle

Huang

Yangtao

. 2019. “Social Networks and Subjective Well-Being in Australia: New Evidence from a National Survey.” Sociology 53(2):401–21.

20.

Jou

Yuh Huey

Fukuda

Hiromi

. 2002. “Stress, Health, and Reciprocity and Sufficiency of Social Support: The Case of University Students in Japan.” The Journal of Social Psychology 142(3):353–70.

21.

Kawachi

Ichiro

Berkman

Lisa F.

2003. Neighborhoods and Health. New York, NY: Oxford University Press.

22.

Kawachi

Ichiro

Kennedy

Bruce P.

1997. “Health and Social Cohesion: Why Care about Income Inequality?” BMJ 314(7086):1037–40.

23.

Kim

H. Harris

Jung

Jong Hyun

. 2022. “Relational Burden, Loneliness, and Depression among American Older Adults: An Inquiry into the ‘Dark Side of Social Capital.’” Aging & Mental Health. doi:10.1080/13607863.2022.2045564.

24.

Lee

Min-Ah

Kawachi

Ichiro

. 2017. “The Company You Keep: Is Socialising with Higher Status People Bad for Mental Health?” Sociology of Health & Illness 39(7):1206–26.

25.

Lin

Nan

. 2001. Social Capital: A Theory of Social Structure and Action. Cambridge, UK: Cambridge University Press.

26.

Lin

Nan

Erickson

Bonnie

, eds. 2012. Social Capital: An International Research Program. New York, NY: Oxford University Press.

27.

Lin

Yanwei

Zhang

Qui

Chen

Wen

Ling

. 2017. “The Social Income Inequality, Social Integration, and Health Status of Internal Migrants in China.” International Journal of Equity in Health 16:139. doi:10.1186/s12939-017-0640-9.

28.

Maass

Ruca

Kloeckner

Christian A.

Lindstrøm

Bengt

Lillefjell

Monica

. 2016. “The Impact of Neighborhood Social Capital on Life Satisfaction and Self-Rated Health: A Possible Pathway for Health Promotion?” Health & Place 42:120–28.

29.

McPherson

Miller

Smith-Lovin

Lynn

Cook

James M.

2001. “Birds of a Feather: Homophily in Social Networks.” Annual Review of Sociology 27:415–44.

30.

Melita

Davide

Willis

Guillermo B.

Rodríguez-Bailón

Rosa

. 2021. “Economic Inequality Increases Status Anxiety through Perceived Contextual Competitiveness.” Frontiers in Psychology 12:637365. doi:10.3389/fpsyg.2021.637365.

31.

Moore

Spencer

Child

Stephanie

Yun-Hsuan

Mandelbaum

Jennifer

. 2018. “Network Approaches to the Study of Social Capital and Health.” Pp. 221–31 in Elgar Companion to Social Capital and Health, edited by Folland

Nauenberg

Cheltenham, UK: Edward Elgar Publishing.

32.

Moore

Spencer

Daniel

Mark

Gauvin

Lise

Dubé

Laurette

. 2009. “Not All Social Capital Is Good Capital.” Health & Place 15:1071–77.

33.

Murayama

Hiroshi

Nofuji

Matsuo

Eri

Nishi

Mariko

Taniguchi

Fujiwara

Yoshinori

Shinkai

Shoji

. 2015. “Are Neighborhood Bonding and Bridging Social Capital Protective against Depressive Mood in Old Age? A Multilevel Analysis in Japan.” Social Science & Medicine 124:171–79. doi:10.1016/j.socscimed.2014.11.042.

34.

Offer

Shira

. 2012. “The Burden of Reciprocity: Processes of Exclusion and Withdrawal from Personal Networks among Low Income Families.” Current Sociology 60(6):788–805.

35.

Offer

Shira

Fisher

Claude S.

2018. “Difficult People: Who Is Perceived to Be Demanding in Personal Networks and Why Are They There?” American Sociological Review 83(1):111–42.

36.

Olivos

Francisco

Olivos-Jara

Pablo

Browne

Magdalena

. 2021. “Asymmetric Social Comparison and Life Satisfaction in Social Networks.” Journal of Happiness Studies 22:363–84.

37.

Pham-Kanter

Genevieve

. 2009. “Social Comparisons and Health: Can Having Richer Friends and Neighbors Make You Sick?” Social Science & Medicine 69(3):335–44.

38.

Pickett

Kate E.

Wilkinson

Richard G.

2015. “Income Inequality and Health: A Causal Review.” Social Science & Medicine 128:316–26.

39.

Portes

Alejandro

. 1998. “Social Capital: Its Origins and Applications in Modern Sociology.” Annual Review of Sociology 24:1–24.

40.

Präg

Patrick

Mills

Melinda C.

Wittek

Rafael

. 2016. “Subjective Socioeconomic Status and Health in Cross-National Comparison.” Social Science & Medicine 149:84–92.

41.

Primo

David M.

Jacobsmeier

Matthew L.

Milyo

Jeffrey

. 2007. “Estimating the Impact of State Policies and Institutions with Mixed-Level Data.” States Politics & Policy Quarterly 7(4):446–59.

42.

Rambotti

Simone

. 2015. “Recalibrating the Spirit Level: An Analysis of the Interaction of Income Inequality and Poverty and Its Effect on Health.” Social Science & Medicine 139:123–31.

43.

Raudenbush

Stephen W.

Bryk

Anthony S.

Congdon

Richard T.

2019. HLM 8 for Windows [Computer software]. Skokie, IL: Scientific Software International, Inc.

44.

Rook

Karen

. 1984. “The Negative Side of Social Interaction: Impact on Psychological Well-Being.” Journal of Personality and Social Psychology 46(5):1097–108.

45.

Rothstein

Uslaner

Eric M.

2005. “All for All: Equality, Corruption, and Social Trust.” World Politics 58(1):41–72.

46.

Royston

Patrick

White

Ian R.

2011. “Mutiple Imputation by Chained Equations (MICE): Implementation in Stata.” Journal of Statistical Software 45(4):1–20.

47.

Sahu

F. M.

Rath

Sangeeta

. 2003. “Self-Efficacy and Well-Being in Working and Nonworking Women: The Moderating Role of Involvement.” Psychology and Developing Societies 15(2):187–200.

48.

Sasaki

Joni Y.

Deborah M.

Kim

Heejung S.

2014. “Culture and Self-Worth: Implications for Social Comparison Processes and Coping with Threats to Self-Worth.” Pp. 230–52 in Communal Functions of Social Comparison, edited by Križan

Gibbons

F. X.

Cambridge, UK: Cambridge University Press.

49.

Smith

Heather J.

Huo

Yuen J.

2014. “Relative Deprivation: How Subjective Experiences of Inequality Influence Social Behavior and Health.” Policy Insights from the Behavioral and Brain Sciences 1(1):231–38.

50.

Smith

Heather J.

Pettigrew

Thomas F.

Pippin

Gina M.

Bialosiewicz

Silvana

. 2012. “Relative Deprivation: A Theoretical and Meta-analytic Review.” Personality and Social Psychology Review 16(3):203–32.

51.

Song

Lijun

. 2011. “Social Capital and Psychological Distress.” Journal of Health and Social Behavior 52(4):478–92.

52.

Song

Lijun

. 2015. “Does Knowing People in Authority Protect or Hurt? Authoritative Contacts and Depression in Urban China.” American Behavioral Scientist 59(9):1173–88.

53.

Song

Lijun

. 2020. “Social Capital, Social Cost, and Relational Culture in Three Societies.” Social Psychology Quarterly 83(4):443–62.

54.

Song

Lijun

Chen

Wenhong

. 2014. “Does Receiving Unsolicited Support Help or Hurt? Receipt of Unsolicited Job Leads and Depression.” Journal of Health and Social Behavior 55(2):144–60.

55.

Song

Lijun

Pettis

Philip J.

2020. “Does Whom You Know in the Status Hierarchy Prevent or Trigger Health Limitation? Institutional Embeddedness of Social Capital and Social Cost Theories in Three Societies.” Social Science & Medicine 257:111959. doi:10.1016/j.socscimed.2018.09.035.

56.

Song

Lijun

Pettis

Philip J.

Chen

Yvonne

Goodson-Miller

Marva

. 2021. “Social Cost and Health: The Downside of Social Capital and Social Networks.” Journal of Health and Social Behavior 62(3):371–87. doi:10.1177/00221465211029353.

57.

StataCorp. 2019. Stata Statistical Software: Release 16 [Computer software]. College Station, TX: StataCorp LLC.

58.

Story

William T.

Glanville

Jennifer L.

2019. “Comparing the Association between Social Capital and Self-Rated Health in Poor and Affluent Nations.” SSM – Population Health 9:100508. doi:10.1016/j.ssmph.2019.100508.

59.

Sung

Pildoo

Son

Joonmo

. 2020. “In-Group Trust and Self-Rated Health in East Asia Using Quadrinational Survey Data.” International Journal of Comparative Sociology 61(1):27–58.

60.

Tak

Yuli R.

Brunwasser

Steven M.

Lichtwarck-Aschoff

Anna

Engels

Rutger C. M. E.

2017. “The Protective Associations between Self-Efficacy and Depressive Symptoms from Early to Middle Adolescence: A Cross-Lagged Model.” Journal of Youth Adolescence 46:744–56.

61.

Takagi

Daisuke

Kondo

Katsunori

Kondo

Naoki

Cable

Noriko

Ikeda

Kenichi

Kawachi

Ichiro

. 2013. “Social Disorganization/Social Fragmentation and Risk of Depression among Older People in Japan: Multilevel Investigation of Indices of Social Distance.” Social Science & Medicine 83:81–89. doi:10.1016/j.socscimed.2013.01.001.

62.

Tan

Jacinth J. X.

Kraus

Michael

Carpenter

Nichelle C.

Adler

Nancy E.

2020. “The Association between Obejctive and Subjective Socioecononic Standing and Subjective Well-Being: A Meta-analytic Review.” Psychological Bulletin 146(11):970–1020.

63.

Thoits

Peggy A.

2011. “Mechanisms Linking Social Ties and Support to Physical and Mental Health.” Journal of Health and Social Behavior 52(2):145–61.

64.

Truesdale

Beth C.

Jencks

Christopher

. 2016. “The Health Effects of Income Inequality: Averages and Disparities.” Annual Review of Public Health 37:413–30.

65.

Umberson

Debra

Crosnoe

Robert

Reczek

Corinne

. 2010. “Social Relationships and Health Behavior across Life Course.” Annual Review of Sociology 36:139–57.

66.

Umberson

Debra

Montez

Jennifer K.

2010. “Social Relationships and Health: A Flashpoint for Health Policy.” Journal of Health and Social Behavior 51(Suppl.):S54–66.

67.

Van der Gaag

Martin

Snijders

Tom A. B.

Flap

Henk

. 2008. “Position Generator Measures and Their Relationship to Other Social Capital Measures.” Pp. 27–48 in Social Capital: An International Research Program, edited by Lin

Erickson

New York, NY: Oxford University Press.

68.

Van der Meer

Tom

Te Grotenhuis

Manfred

Plezer

Ben

. 2010. “Influential Cases in Multilevel Modeling: A Methodological Comment.” American Sociological Review 75(1):173–78.

69.

Verhaeghe

Pieter-Paul

Yaojun

. 2015. “Socioeconomic and Ethnic Inequalities in Social Capital from the Family among Labour Market Entrants.” European Sociological Review 29(4):683–94.

70.

Verhaeghe

Pieter-Paul

Pattyn

Elise

Bracke

Piet

Verhaeghe

Mieke

Van De Putte

Bart

. 2012. “The Association between Network Social Capital and Self-Rated Health: Pouring Old Wine in New Bottles?” Health & Place 18:358–65.

71.

Villalonga-Olives

Ester

Kawachi

Ichiro

. 2017. “The Dark Side of Social Capital: A Systematic Review of the Negative Health Effects of Social Capital.” Social Science & Medicine 194:105–27.

72.

Weigel

Daniel J.

Shrout

M. Rosie

. 2020. “Relational Turbulence and Health-Compromising Behaviors: Integrating Relational Turbulence and Stress Theories.” Personal Relationships 27(4):785–800.

73.

Wilkinson

Richard G.

Pickett

Kate E.

2006. “Income Inequality and Population Health: A Review and Explanation of the Evidence.” Social Science & Medicine 62(7):1768–84.

74.

Xue

Xindong

Reed

W. Robert

Menclova

Andrea

. 2020. “Social Capital and Health: A Meta-analysis.” Journal of Health Economics 72:102317. doi:10.1016/j.jhealeco.2020.102317.

75.

Yang

Xiaozhao Yousef

Anning

Schieman

Scott

. 2019. “Relative Deprivation in Context: How Contextual Status Homogeneity Shapes the Relationship between Disadvantaged Social Status and Health.” Social Science Research 81:157–69.

76.

Zajacova

Anna

Lawrence

Elizabeth M.

2018. “The Relationship between Education and Health: Reducing Disparities through a Contextual Approach.” Annual Review of Public Health 39:273–89.

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

0.03 MB