Life-Course Transitions, Socioeconomic Status and Health Behaviours

It has been widely reported that major life transitions, such as going to college, finding a stable job, or becoming unemployed, are associated with poor health outcomes, such as ….  (1-3). Moreover, there is solid evidence that life transitions are also associated with health behaviours: some studies have reported poorer dietary intake during college (4), or higher rates of smoking (5, 6) and heavy drinking (7, 8) among those facing unemployment. In fact, health behaviours have been reported to account for about one-quarter of socioeconomic disparities in health (9).
Traditionally, two alternative explanations are offered for the association between specific life events and health (behaviours) that differ in the directionality of the causality: the social causation hypothesis and the health selection hypothesis (10, 11). 
The social causation hypothesis argues that life events, e.g. job loss, lead to unhealthy behaviours through different mechanisms: first, high stress levels associated with unemployment may lead to unhealthy behaviours like smoking or alcohol use as coping mechanisms (3). Alternatively, a decrease in material resources may lead to unhealthy behaviours like poor diet due to a higher availability and affordability of industrially-processed products (9). 
Alternatively, the selection hypothesis claims that pre-existing health behaviours make individuals more or less prone to experiencing certain events, like finding a partner (12) or becoming unemployed (13). In the specific case of young adults, there is evidence that health selection mainly operates through mental (e.g.. ..) and behavioural disorders (e.g. ..) rather than physical conditions (e.g. ..)  (14). For instance, substance use may lead to difficulties to cope with duties of daily life (15), and thus to being less eligible for a job and more prone to unemployment (16, 17).   
In fact, there is evidence that both social causation and health selection processes are involved in the social gradient of health behaviours (10, 18, 19, 20). Furthermore, there is evidence that both operate simultaneously, in a bidirectional association (13, 21, 22). 
Traditionally, studies have mostly focused on testing either causation or selection (3, 16, 17, 19, 20, 23). The few studies that test both simultaneously, tended to consider the general association between life events -e.g. unemployment- and health behaviours –e.g. alcohol use- (13, 21), without specifically looking at potential subgroup effects. 
However, the relative contribution of social causation or health selection to inequalities in health -and health behaviours- may be modified by external factors, i.e. moderators. For instance, a meta-analysis investigating the association between unemployment and health was moderated by occupational status and social support (i.e. stronger effect for manual workers and weaker effect among those with strong social networks) (19).  Similarly, studies on health selection (i.e. poor health) into unemployment have shown that gender moderates the association, with selection being stronger among women than among men (24).
To the best of our knowledge, no studies have focused on testing both causation and selection, and at the same time addressing the role of moderators. Thus, important questions like ‘for whom is one or the other effect stronger?’ and ‘in which socioeconomic contexts is health selection stronger than causation, and vice versa?’ remain unresolved.
In order to tackle these questions, we simultaneously analyse social causation and health selection processes in the association between life transitions and health behaviours, and which factors moderate the strength of these associations.
Life transitions include events like entry into a partner relationship, parenthood, and completing an education. A special focus is placed on unemployment, since it has been reported to “represent the greatest threat to worker’s well-being” (25). 
Health behaviours include smoking, alcohol and illicit drug use, dietary patterns and physical activity. Finally, key moderators are SES –i.e. educational level, income, and occupation-, (perceived) family and social support, and gender. Age, as well as stressful life events will be taken into account as covariates. 
Life course perspective: the case of unemployment and health behaviours
Studies that analyse the causal relationship between life transitions and health behaviours have increasingly adopted a life-course perspective (26, 27). The rationale behind it is that individuals navigate through different events and transitions during their life-course and progressively acquire new educational, personal and professional achievements. Moreover, since a bidirectional causal association is assumed, health behaviours in one stage of life may be both cause or consequence of their previous and current life events, as well as a determinant of future ones (11). 
From this point of view, health behaviours can be understood as diverging responses of different groups of people to certain conditions or exposures imposed by the socioeconomic context (28). These exposures have traditionally been analysed in the life-course literature from two perspectives: the “accumulation of risk”  and the “critical period” approach. The first approach (accumulation of risk) assumes that being exposed to one disadvantage increases the likelihood of subsequent ones (11). In this model, causation and selection may interact, for instance because unemployment in one stage may cause poor health in the next, which in turn may cause a “downward spiralling effect” (19). The second approach (critical period) focuses on specific exposures, such as in early life, that may have  “irreversible” effects (26, 29). For instance, diet and physical activity have been reported to be strongly associated with inherited childhood conditions (28, 30), such as parental SES, whereas smoking has been shown to be more strongly associated with exposures during adolescence and young adulthood, mainly to education and occupation (28, 31). 
The role of moderators
Examining the role of moderators in the causal association between life transitions and health behaviours is one of the key goals of this study. For instance, SES may act as a moderator, as a lack of material and symbolic resources may make lower SES individuals more vulnerable to stressful life transitions (e.g. unemployment) or other life events (e.g. death of a spouse), thus responding with heavier alcohol use than their higher SES counterparts (11). Likewise, other variables may moderate the association between stressful events and health behaviours, either by buffering it (e.g. strong family network or high educational level) or increasing it (e.g. low income).  Drawing on this perspective, we will focus on several potential moderators: 
Socioeconomic Status: There is evidence that both causation (19) and selection (18) are stronger among lower SES than among higher SES individuals. However, SES can be measured in different ways, such as by level of education, occupational class, income or housing tenure status. Educational level is considered to be an accurate proxy for SES (32). It is strongly associated with income and wealth and it can also explain social inequalities in health behaviours per se, e.g. by means of “health literacy” (9). Occupational class has the advantage that  (33), whereas level of education largely remains constant in adulthood, occupational class can change throughout life. Therefore,  it is argued that the scope for health selection in older adults may be larger when occupational class is considered (11). Recently, some studies suggested that housing tenure is especially relevant after the financial crisis and its big impact on mortgage debt and housing evictions (34, 36). Still, how each dimension of SES moderates the association between life transitions and health behaviours remains unclear.
Family support has been suggested to modify the association between life events and alcohol abuse. Potential explanations are that having a solid family network provides both emotional and instrumental support that can lower the burden of unemployment in terms of mental health and thus result in less alcohol-related problems to cope with it (34). Nevertheless, research on its moderating effect is scarce and findings are inconclusive (37). 
Social support: Perceived social support has been found to be a significant moderator of the association between unemployment and depression, with the effect of job-seeking stress on depression decreasing as levels of perceived social support increase (38). However, it has also been suggested that family and social networks can be “coercive and (act as) sources of strain” (39). Hence, the moderating effect of social support in the association between unemployment and health behaviours remains unclear.  
Gender: Finally, there is evidence pointing at diverging causal effects by gender. However, findings are mixed, with some pointing at unemployment allegedly having greater impact on health behaviours among men (40) and others pointing at stronger causation effects in women (41). The moderating role that gender may play, for instance, in the health selection from health behaviours to unemployment in women remains unclear. 
In sum, it remains unknown to which degree causation and selection account for the association between life transitions and health behaviours. It is also unclear to what extent the aforementioned variables modify this association. 
For that reason, this study aims to analyse for which population subgroups social causation or health selection play a bigger role in the association between life transitions and health behaviours (see Figure 1). 
As it has been pointed out, evidence on the role of moderators is scarce. One of the main reasons for this scarcity is that, in order to be able to infer causality and establish moderation effects, it is essential to use large samples of longitudinal data (11). The Lifelines panel data allow us to unravel this conundrum between social causation and health selection processes and address the role of moderators in that association.