Literature Review

The association between socioeconomic status and health is a well-known fact, and even in the countries with a relatively low level of income inequality and poverty, such as the Nordic countries, these inequalities in health persist (Marmot, Reference Marmot2005; Mackenbach et al., Reference Mackenbach, Stirbu, Roskam, Schaap, Menvielle, Leinsalu, Kunst, van Oyen, Demarest, Rychtarikova, Dzurova, Andersen, Ekholm, Judge, Tekkel, Prättäla, Martikainen, Desplanques, Jusot, Helmert, Kovacs, Marton, Layte, Costa, Vannoni, Villerusa, Kalediene, Klumbiene, Geurts, Dahl, Strand, Wojtyniak, Santana, Madarasova Geckova, Artnik, Borrell, Esnaola, Regidor, Burström, Fritzell, Lundberg, Bopp and Glickman2008; Hu et al., Reference Hu, van Lenthe, Borsboom, Looman, Bopp, Burströ, Dzúrová, Ekholm, Klumbiene, Lahelma, Leinsalu, Regidor, Santana, de Gelder and Mackenbach2016; Lahelma et al., Reference Lahelma, Pietiläinen, Pentala-Nikulainen, Helakorpi and Rahkonen2019). Mental health problems are no exception to this pattern (Murali and Oyebode, Reference Murali and Oyebode2004; Hudson, Reference Hudson2005; Aneshensel, Reference Aneshensel2009; Reiss, Reference Reiss2013; Allen et al., Reference Allen, Balfour, Bell and Marmot2014).

While the association between one’s socioeconomic status and health is well established in the literature, the bi-directional and complex causality behind this association leaves many questions open for examination. The association has been mainly explained with the social causation and selection hypotheses. The first asserts that poor socioeconomic status increases the risk of health problems through, for example, reduced access to health care, occupational hazards, poorer diets, and stress related to being poor or unemployed. According to the second hypothesis health problems prevent people from attaining higher positions and they can lead to work incapacity and decrease in earnings (e.g. Goldman, Reference Goldman2001). It could also be that there is a third factor affecting both health and socioeconomic status; for example, genetic endowment, risky behaviour or childhood living conditions might predict both worse health and lower socioeconomic status in adulthood (Case et al., Reference Case, Lubotsky and Paxson2002; Hoffmann et al., Reference Hoffmann, Kröger and Geyer2019).

Both social causation and selection in explaining inequalities in health are supported by evidence, and the choice of indicators for health and socioeconomic status, societal context, and population group under focus can affect the results (Kröger et al., Reference Kröger, Pakpahan and Hoffmann2015; Vaalavuo, Reference Vaalavuo2016a, Reference Vaalavuo2016b; Hoffmann et al., Reference Hoffmann, Kröger and Pakpahan2018; Kane et al., Reference Kane, Harris, Morgan and Guilkey2018). Age can be an important element in the conundrum and the relative importance of the two explanations has been shown to vary with age: according to Hoffmann et al. (Reference Hoffmann, Kröger and Pakpahan2018, Reference Hoffmann, Kröger and Geyer2019) both selection and causation are important mechanisms during childhood and young adulthood, but selection loses its importance with age, making causation a more relevant mechanism for older people. It seems logical that health problems early on in life are important for future socioeconomic status attainment as this is the period when individuals make decisions about their education, while at older ages one has already achieved a certain socioeconomic position in occupational and educational terms.

Life course studies have shown that both health and socioeconomic status depend on their prior status (Hoffmann et al., Reference Hoffmann, Kröger and Pakpahan2018). Moreover, the two are likely to be intertwined so that health and human capital accumulation reflect strong interdependence (Kane et al., Reference Kane, Harris, Morgan and Guilkey2018). From this perspective early life course – childhood, adolescence, and transition to adulthood – is essential in understanding social inequalities in health later in life. Haas (Reference Haas2006) and Haas et al. (Reference Haas, Glymour and Berkman2011) have shown that those with poor childhood health have substantially lower earnings over their working career. The effect seems to be mediated by lower educational attainment and poorer adult health.

The type of illness may matter too for how social causation and selection work. For example, social causation explains much of the association between depression and socioeconomic position, while in the case of schizophrenia social selection seems to be more important (Dohrenwend et al., Reference Dohrenwend, Leav, Shrout, Scwartz, Naveh, Link, Skodol and Stueve1992). Similarly, Miech et al. (Reference Miech, Caspi, Moffitt, Entner Wright and Silva1999) found that anxiety and depression during adolescence did not affect educational transitions, while those with conduct disorder were less likely to earn a school certificate in New Zealand. These findings suggest that different mental health disorders are related to social status in different ways.

One in six people in the European Union has a mental health problem costing the economy EUR 240 billion, or 1.6 per cent of the GDP, due to loss in productivity and lower employment according to the OECD estimates (2018). Mental health disorders can create a barrier to education and employment (Pacheco et al., Reference Pacheco, Page and Webber2014) and among adolescents and young adults they are associated with lower education, higher unemployment and substantially lower earnings later in life (Hakulinen et al., Reference Hakulinen, Elovainio, Arffman, Lumme, Pirkola, Manderbacka, Keskimäki and Böckerman2019). Mental disorders in young adulthood also increase the risk of sickness absence and receiving disability pension (Helgesson et al., Reference Helgesson, Tinghög, Niederkrotenhaler, Saboonchi and Mittendorfer-Rutz2017) and overall the risk of not succeeding in the transition from school to employment and of becoming excluded from the labour market (Olesen et al., Reference Olesen, Butterworth, Leach, Kelaher and Pirkis2013; Baggio et al., Reference Baggio, Iglesias, Deline, Studer, Henchoz, Mohler-Kuo and Gmel2015; Power et al., Reference Power, Clarke, Kelleher, Coughlan, Lynch, Connor, Fitzpatrick, Harley and Cannon2015; Rodwell et al., Reference Rodwell, Romaniuk, Nilsen, Carlin, Lee and Patton2017).

The understanding of what constitutes a disability has changed over time from a medical view to a more holistic social model (Mitra, Reference Mitra2006). According to the latter, disability is not a characteristic of the individual, but a social construction that is affected by personal attributes (e.g. impairment, age, gender, ethnic background), the individual’s resources, and the physical, social, economic and political environment (Mitra, Reference Mitra2006; Grover and Piggott, Reference Grover and Piggott2015). The way in which the society responds to mental illness is essential in the promotion of social inclusion of people with mental health issues (Grover and Piggott, Reference Grover and Piggott2015). Mental disorders can cause significant work incapacity but they are also heavily stigmatized compared to most somatic diseases. Research has shown that discrimination, negative stereotypes and prejudices against individuals with mental disorders are common, and discrimination is associated with poor employment outcomes (Mehta et al., Reference Mehta, Kassam, Leese, Butler and Thornicroft2009; Sharac et al., Reference Sharac, Mccrone, Clement and Thornicroft2010; Baldwin and Marcus, Reference Baldwin, Marcus, Schultz and Rogers2011).

Many factors associated with impaired mental health, such as unemployment and low education, are also among the biggest risk factors for receiving social assistance and predicting longer social assistance spells (Bäckman and Bergmark, Reference Bäckman and Bergmark2011; Ilmakunnas and Moisio, Reference Ilmakunnas and Moisio2018). Overall, several studies across rich Western democracies have found that social assistance recipients have worse physical and mental health compared to the rest of the population (Overbye and Saebo, Reference Overbye and Saebo1998; Taylor and Baruch, Reference Taylor and Baruch2004; Vozoris and Tarasuk, Reference Vozoris and Tarasuk2004; Kaplan et al., Reference Kaplan, Siefert, Ranjit, Raghunathan, Young, Tran, Danziger, Hudson, Lynch and Tolman2005; Morris et al., Reference Morris, Rosenbluth, Scott, Livingstone, Lix, McNutt and Watsonj2006; Ayala and Rodriguez, Reference Ayala and Rodriguez2007; Baigi et al., Reference Baigi, Lindgren, Starrin and Bergh2008; Ohram Naper, Reference Ohram Naper2009; Hammer and Hyggen, Reference Hammer and Hyggen2010; Løyland et al., Reference Løyland, Miaskowski, Dahl, Paul and Rustoen2011; Königs, Reference Königs2013; Vaalavuo, Reference Vaalavuo2016a; Shahidi et al., Reference Shahidi, Ramraj, Sod-Erdene and Siddiqi2019). It has also been shown that mental health disorders are the most significant disease group among social assistance recipients in Finland compared to the overall population, especially among young adults (Vaalavuo, Reference Vaalavuo2016a). While a lot of the research on the topic is descriptive and based on cross-sectional data, there is also some evidence that worse health precedes the receipt of social assistance or other forms of income support (e.g. Morris et al., Reference Morris, Rosenbluth, Scott, Livingstone, Lix, McNutt and Watsonj2006; Kiely and Butterworth, Reference Kiely and Butterworth2014; Vaalavuo, Reference Vaalavuo2016a). Nevertheless, a more dynamic understanding of the association with health and social assistance seems to be somewhat lacking.

We will test three hypotheses based on the earlier literature: first, health problems at the start of a social assistance spell are associated with longer social assistance duration as suggested by the literature on health selection; second, mental health problems are more strongly associated with spell length than somatic health problems; and third, younger individuals are more strongly affected by health problems than older individuals who are more likely to have established themselves in the labour market. Furthermore, we will test whether the results hold for various different diagnoses of mental health disorders.

Institutional Context

Finland is a welfare state characterized by universal and relatively generous welfare benefits together with a comprehensive social and health care sector. Individuals are entitled to basic income benefits based on residence, while earnings-related benefits also exist and they depend on prior earnings. Social assistance is a last-resort means-tested benefit that is designed to cover basic necessities in life. As the level of basic benefits has decreased relative to wages and rising housing expenditures, the use of social assistance has increased. This is especially typical for young adults who are not necessarily entitled to other benefits.

In 2019 the basic monthly amount of social assistance was 497 euros for those living alone. Supplementary social assistance can be granted to cover housing costs and other necessary expenses that are needed to secure a decent livelihood. Medical expenses can also be covered through social assistance, and recipients get their prescribed medicines automatically reimbursed. The amount can also be lowered if the recipient refuses to register as an unemployed jobseeker or take part in other activation measures such as training. However, even in these situations, a necessary level of subsistence should not be compromised. Overall, around 4 per cent of the recipients are sanctioned annually (Jokela et al. Reference Jokela, Kivipelto and Ylikännö2019).

8.5 per cent of the population received social assistance in 2018. In the last decade the number of recipients has been growing steadily. As in most parts of Europe, the original idea of social assistance was to cover for short-term needs and economic hardship, but since the 1990s it is more frequently received on a longer-term basis. In 2018 almost a third of all the recipients received social assistance for more than ten months (Tanhua and Kiuru Reference Tanhua and Kiuru2019). Since young adults usually do not have sufficient working experience to receive unemployment benefits and the transition to the labour market is somewhat uncertain, they are overrepresented among social assistance recipients.

The high share of young people receiving social assistance and the increasing share of long-term beneficiaries has led to many demands to reform the system and increase work incentives or availability of appropriate occupational rehabilitation services. The new government program (Government program, 2019) aims at decreasing social assistance dependency among young adults, which makes the results of this study very topical in Finland. However, the underlying mechanisms of social exclusion apply to other countries as well, and academics and policy-makers interested in health inequalities and welfare policies will find the results interesting.

Data, Analysis Sample, and Methods

Independent variables

The main explanatory variable in our analyses is health status, measured as the use of public special health care services during a calendar year and main diagnoses received based on the ICD-10 classification. Both in- and out-patients visits are taken into account. The health variable has three categories: 1) no diagnosis, 2) any psychiatric diagnosis, and 3) only a somatic diagnosis (pregnancy-related diagnoses are excluded). The second category includes also individuals with both psychiatric and somatic diagnoses, while the third category excludes individuals with any psychiatric diagnoses. Our main focus is on psychiatric diagnoses, but we have included a category for somatic diagnoses as well to see whether mental health issues, in particular, are associated with social assistance duration. We also examine how different psychiatric diagnoses are associated with the exit from social assistance.

Since special health care does not account for all health care visits, we have tried to measure mental health problems also more broadly by combining information on psychiatric diagnoses and/or use of psychiatric drugs (identified based on ATCs that internationally classify medicines). We have created a dummy variable on mental health problems based on this information (0=no mental health problems, 1=mental health problems). This variable is used mainly to investigate the robustness of the results based on our first health variable. In Finland, milder mental health disorders are usually treated in primary health care, and only people with more severe disorders end up in special health care. It can therefore be expected that the broader operationalization of the health variable will capture more people with milder mental disorders. It should be noted that some psychiatric drugs can also be used as painkillers (Ananth, Reference Ananth2012). Furthermore, a single purchase of sleeping pills does not necessarily indicate having a mental health disorder, while having a psychiatric diagnosis from special health care is likely to reflect more severe mental disorders. Both health variables are measured at the starting year of the first social assistance spell.

Table 1 presents the number of observations with a psychiatric diagnosis or purchases of psycho-pharmaceuticals in our analysis sample. Altogether almost 12 percent had some psychiatric diagnosis and 18 percent had bought psychiatric drugs during the first year of social assistance. Multiple diagnoses are possible, and they are common. For example, among those with a mood disorder more than a half also had another psychiatric diagnosis. There are some differences between age groups in the prevalence of specific diagnoses. Most importantly, schizophrenia is more common among those aged 30-34 (3.6 % compared to 1.9 % among those aged 18-24), while behavioural disorders with the onset in childhood and adolescence are more common in the youngest age group (1.4 % compared to 0.4 % among those aged 30-34).

Table 1.

Mental health problems at the start of the first social assistance spell based on special health care visits and purchases of psycho-pharmaceuticals

The use of health care services does not directly measure the health status of the person but it can be considered a good proxy for it. For example, Miilunpalo et al. (Reference Miilunpalo, Vuori, Oja, Pasanen and Urponen1997) have shown that poor self-assessed health is correlated with the use of health care services, while others have shown that self-assessed health is a good predictor of mortality and morbidity (Idler and Benyamini, Reference Idler and Benyamini1997; Benjamins et al., Reference Benjamins, Hummer, Eberstein and Nam2004). All residents in Finland are entitled to public special health care services after referral from primary health care, while user fees apply. However, we are not able to account for unmet needs for health care that are usually more common among low-income individuals. Unfortunately, we do not have information on the use of occupational or private health care services, but they would mostly impact on the measured use among employed and higher income groups. Despite these challenges, using health indicators based on register data and objective measures can be more accurate than relying on self-reported health that can suffer from justification bias (e.g. Bago d’Uva et al., Reference Bago d’Uva, O’Donnell and van Doorslaer2008).

Control variables

Our control variables are presented in Table 2. Age is measured at the start of the first social assistance spell. The level of education, labour market status, and family type are all used as time-varying variables and are measured on an annual basis. Labour market status measures the main activity of a person in the last week of the year. The “main activity” variable from Statistics Finland has originally been coded in line with the recommendations of the International Labour Organization. The ILO hierarchy of activities places labour market participation at the top so it prioritizes employment and unemployment over other statuses. In other words, if a full time student is working during the weekends they are coded as employed instead of student. Even one hour of work is enough to override the student status. (Asplund and Vanhala, Reference Asplund and Vanhala2013.) This is not the most accurate measure for our purposes when studying young adults, of whom many are primarily students. To account for this, the main activity variable has been recoded: if a person received educational allowance during the year, they have a student status, regardless if they were also working. This is important also because students are obliged to use their student loans before being eligible for social assistance.

Table 2.

Descriptive statistics based on personal characteristics at the beginning of the first social assistance spell

Family type has been created with a family identification number available in the data. According to this identification, a family consists of two adults who are married, live together or are in a registered relationship with or without children and single parents living with their children. People who are not included in this definition of a family and thus do not have a family identification number are coded as living alone. Children above the age of 18 living with their parents are thus coded as living alone. However, it is noteworthy that they are eligible for receiving social assistance regardless of their parents’ income, although the amount is lower than for those living independently.