It is widely recognized that differences in socioeconomic status (SES)—defined based on education, income, and/or occupation—are associated with inequalities in the occurrence of chronic disease. Theories to explain the social gradient in health predominantly focus on absolute SES differences. Such explanations, which are well-established, pose that a lower SES implies lower access to resources and therefore worse living conditions and increased chronic stress resulting from these conditions, which contribute—directly and indirectly—to risk factors, e.g., overweight, high blood pressure, and insulin resistance, for chronic disease [1, 2]. In addition, there is increased attention for explanations that look to relative SES differences. These propose that a lower SES may cause poor health not only because it implies lower access to resources but also because it might induce feelings of inferiority resulting from social comparisons. The main difference between the absolute and relative explanations is that the former assume that SES differences would also affect people if they were completely isolated from each other, whereas the latter rely on social comparisons [3]. The impact of social comparisons on individual wellbeing has long been studied in the social sciences, e.g., in Merton’s (1968) Reference Group Theory [4]. The hypothesized importance of explanations that rely on relative differences for the SES–health relationship is supported by empirical evidence from countries where most people can be presumed to have sufficient access to resources but where a clear social gradient in health is still observed [5].

While there are various mechanisms that could be relevant when it comes to relative explanations for the social gradient in health (as we will lay out in the discussion of our results), the most well-known mechanism through which relative SES differences are postulated to affect health is the emotional stress response resulting from social-evaluative threat [5–8], as brought to the fore in Wilkinson and Pickett’s distinguished Psychosocial Theory. It is believed that social inequality “intensifies social hierarchies” [9] and as a consequence “drives people into social comparisons” [6, 7], causing such stress responses. This type of stress is said to result from the feeling of not counting much in the eyes of others [6], “subjectively experienced inferiority,” or “the sense of inferiority” [8] for those at the losing side of the social comparison, i.e., status anxiety (SA). However, empirical evidence for SA as a mechanism generating health inequalities is limited [7, 8]. As posed in Wilkinson and Pickett’s Psychosocial Theory, societies that are more unequal are worse off on a range of different outcomes, e.g., crime, political involvement, life satisfaction, and population health [6, 8, 10, 11]. Population-level associations between inequality and health have repeatedly been observed [12, 13]. However, in the context of this theory, SA as a mechanism connecting inequality to health has been inferred rather than demonstrated. The ecological fallacy is therefore a prominent point of critique on the Psychosocial Theory [14, 15].

Like the Psychosocial Theory, studies that have tested SA as a mechanism have generally been of an ecological nature: they examine how SA and health are related by comparing countries that vary in inequality. This was the case for the majority of the studies that we identified that, first, incorporated a measurement of SA and, second, investigated either the relationship between inequality and SA [7–9, 16–21] or SA as a mediator of the relationship between inequality and health [22–25]. SA was found to in part explain the relationships between inequality and happiness, life satisfaction [24], and mental wellbeing [22] but not between inequality and depression [23] and an index covering life expectancy, infant mortality, obesity, teenage birth rate, homicides, and imprisonment [25]. Moreover, the previous studies that have incorporated a measurement of SA were largely focused on mental health outcomes, i.e., happiness, life satisfaction [24], mental wellbeing [22], and depression [23]. Only one of the studies explored the mediating role of SA for physical health outcomes, i.e., life expectancy, infant mortality, obesity, and teenage birth rate, as part of an index covering six outcomes [25], finding no mediation through SA. Because inequality is not the only aspect that countries vary in and because the observed population-level associations may not apply to individuals, ecological studies are insufficient to establish SA as a mechanism generating health inequalities, as they limit causal inference. To facilitate causal inference, individual-level analyses that assess how SA depends on SES and affects health are warranted.

We therefore conduct individual-level analyses to assess SA, operationalized as feelings of inferiority, as a mediator of the relationship between SES and type 2 diabetes (T2D) using data from HELIUS, an Amsterdam-based multi-ethnic cohort study [26]. To shed further light on the relative and absolute explanations for the SES–health relationship, we additionally assess financial difficulties—as an indicator of access to resources—as a mediator of the same relationship. We opt for T2D as the health outcome because this common chronic disease has a clear social patterning [27] and the impact of chronic stress on T2D risk has been found to be especially pronounced [28]. This is relevant as SA is hypothesized to affect health through the emotional stress response. This study is informed by the following hypotheses:

H1

Socioeconomic status will be negatively associated with type 2 diabetes and status anxiety, showing a social gradient.

H2

Status anxiety will be positively associated with type 2 diabetes while controlling for socioeconomic status.

H3

Both status anxiety and financial difficulties will partly mediate the relationship between socioeconomic status and type 2 diabetes.

We conducted individual-level analyses to assess SA—operationalized as feelings of inferiority—as a mediator in the relationship between SES and T2D. To shed further light on the relative and absolute explanations for the SES–health relationship, we additionally assessed financial difficulties as a mediator of the same relationship. We found that the prevalence of SA was higher among participants with a lower SES. SES was negatively associated with both T2D and SA. In turn, SA was associated with higher odds of having T2D. As hypothesized, both SA and financial difficulties played small but consistent mediating roles in the SES–T2D relationship. The mediated proportion through SA (2%–3%) was smaller than through financial difficulties (9%–11%).

These results should be seen in light of a number of limitations. First, all analyses were cross-sectional, where SES, SA, financial difficulties, and T2D were measured simultaneously. This prohibits establishing causality. Reverse causality is possible, e.g., those living with T2D may feel inferior due to their disease or its consequences (e.g., unemployment), and low SES may result from disease. Second, although a natural effect model is considered a non-parametric structural equation model, it still relies on the assumption that there are “no unmeasured confounders for the exposure-outcome, exposure-mediator, or mediator-outcome relations” [31]. This is particularly difficult to ensure for the mediator-outcome relations, as there are factors, e.g., underlying mental health problems, that may affect both SA and T2D risk. Mental health problems may also be possible consequences of SES, making them “exposure dependent confounders,” which are complicated to adjust for. Third, SA was measured by a single item, out of 12, of the Neuroticism subscale of the Neo Five Factor Inventory. While it may not be optimal to use a single item to operationalize SA, it is not unprecedented [8, 22]. Fourth, building on that, this measurement was not designed to capture SA. It is possible that participants perceived the statement “I often feel I’m inferior to other people” as normative, which could cause them to refrain from agreeing with the statement. Nevertheless, we did find that SA was prevalent in the entire population. In addition, while the statement does include a social subordination element, i.e., “I often feel I’m inferior to other people,” it does not specify that this is caused by relative SES differences (e.g., “More educated people look down on people like me”). The frequently used EQLS measurement does particularize a SES element—i.e., in item (1) “Some people look down on me because of my job situation or income.” The observed social gradient in SA as it was operationalized in this study does build confidence in the use of the measurement in the context of relative SES differences. Furthermore, it could conversely be argued that leaving the causes of the feelings of inferiority undefined functions as an advantage. Specifically, as considered by Delhey et al. (2017) in their discussion of the EQLS measurement, if the SES elements are particularized, “the phrasing may favor job and income as potential reasons for status anxiety” [7], above other indicators, e.g., education and employment status. Another strength of the statement used is that it reflects possible feelings attached to social position—in contrast to, e.g., analyses that use subjective social status (SSS) as an indicator for how people perceive themselves in relation to others or analyses that only rely on item (1) from the EQLS [8, 22]. That is, a low SSS does need to be accompanied by feelings of inferiority. To address possible measurement errors and account for any potential noisiness of the SA measurement, the existence of which we cannot ascertain in this study, we opted for dichotomizing SA—as well as financial difficulties—in order to function as a more robust indicator. To ascertain that this decision did not distort the results, we repeated the NEM analyses employing SA and financial difficulties as un-dichotomized, continuous mediators and found that this yielded similar relative results, i.e., a mediated proportion of 3%–5% through SA and of 11%–12% through financial difficulties (see Supplementary Table S1 in the Additional material for the results with the un-dichotomized mediators).

Much like previous ecological studies, these individual-level analyses imply that relative SES differences play a role in health inequalities. The notion that resource-based explanations are only part of the story is supported by our finding that, in this study, about 10% of the SES–T2D relationship could be ascribed to financial difficulties. While the proportion of the SES–T2D relationship mediated by SA was smaller than the proportion mediated by financial difficulties, the consistent association observed across different SES indicators indicates a robustness of this relationship. Furthermore, the proportion of the SES–T2D relationship mediated by SA may seem small but is of the same order of magnitude as the proportion attributable to other, widely accepted mediators, e.g., smoking—for which the reported proportion mediated ranges from 1% to 13% [36–38].

This study substantiates the theory that social comparisons are connected to individual wellbeing: feeling inferior to other people, resulting from social comparisons, was associated with T2D. Although these individual-level analyses support SA as a mechanism generating health inequalities, complementary analyses are required to further our understanding of the relationship between inequality, SA, and individual wellbeing. Arguably, a variety of analyses should be used in combination to appreciate how different constructs of inequality affect individual wellbeing and to eventually understand what we could do to address this.

First, while our analyses provide insight into how people perceive inequality, multilevel analyses using, e.g., the Gini coefficient could be used to ascertain whether more inequality at the societal level leads to more SA among individuals [2, 39].

Second, while our analyses link SES to SA and T2D, research has shown that, when it comes to health, results in relation to ‘objective’ SES do not always correspond to those in relation to SSS [1]. Analyses that incorporate SSS can provide additional information about the relationship between social position and SA. In addition, the concept of SSS relates to debates about the importance of the terminology used to refer to different types of education and occupation in that it may carry a value judgement. For example, in the Netherlands, it has been argued that instead of “low” and “high” education, the terms “practically trained” and “theoretically trained” should be used [40], consistent with the idea that they differ but should be equally valued by society.

Third, while the majority of identified studies operationalized SA as feelings of inferiority [7, 8, 17, 20, 22, 24, 25], other aspects of SA could be put at the forefront, which may lead to additional insights. Aspects identified as important may include status seeking [9, 18, 19, 41], desire for wealth and status [21], financial satisfaction and perceived relative income [42], economic worries [43], and perceived importance of income comparisons [23]. That other aspects of SA may lead to additional insights is reflected in the diversity in findings from various studies. For instance, Paskov et al. (2017) found that status seeking was more pronounced among those with a high social position and less prevalent among the unemployed [9], while Delhey, Schneickert, and Steckermeier (2017) and Layte and Whelan (2014) found that feelings of inferiority occurred more often among those with a low income [7, 8] and those excluded from the labor market [7]. Moreover, SA has also been defined as “a worry that we are currently occupying too modest a rung [on the social ladder] or are about to fall to a lower one” [44], implying the importance of the fear of losing one’s social position as an aspect of SA. Day and Fiske (2016) have introduced, as recently published by Melita et al. (2020), a designated Status Anxiety Scale consisting of five items covering different aspects of SA, e.g., “I worry that my social status will not change” and “I sometimes worry that I might become lower in social standing” [45, 46]. This has been employed in recent studies showing that SA uniquely explained job satisfaction [47], that the relationship between perceived economic inequality and SA is mediated by perceived competitiveness [16], and, through experiments, that the relationship between perceived economic inequality and consumption is mediated by status seeking [48].

Fourth, while the Psychosocial Theory has focused attention on SA, SA is arguably not the only possible cause of the social gradient in health that depends on the way society is organized. For example, the belief that we live in a meritocracy, a society in which people acquire their status based on their talents and efforts, could affect how people feel about their own social position. Other relative explanations for the social gradient in health should also be considered. For example, the Fundamental Cause Theory of health disparities includes power and prestige as assets that are primarily held by those with a high SES [49]. Here, power and prestige are presumably only favorable if they exceed the power and prestige of others. Apart from SA, growing relative SES differences could also breed feelings of unfairness. Such feelings may become increasingly influential in light of the “disproportionate share of global wealth growth” that has been attained by multimillionaires over the last 3 decades and especially during the COVID-19 pandemic [50]. In this regard, Bosma et al. (2012) found, while an effect of SES on health via perceived unfairness could not be detected, that perceived unfairness was more commonly experienced by people with a lower SES and that perceived unfairness at baseline, in turn, was related to worse mental as well as physical functioning 7 years later [51].