Loneliness and health of older adults: The role of cultural heritage and relationship quality
Abstract
EconStor is a publication server for scholarly economic literature, provided as a non-commercial public service by the ZBW.
Full text
Casabianca, Elizabeth J.; Kovacic, Matija Working Paper Loneliness and health of older adults: The role of cultural heritage and relationship quality JRC Working Papers in Economics and Finance, No. 2022/5 Provided in Cooperation with: Joint Research Centre (JRC), European Commission Suggested Citation: Casabianca, Elizabeth J.; Kovacic, Matija (2022) : Loneliness and health of older adults: The role of cultural heritage and relationship quality, JRC Working Papers in Economics and Finance, No. 2022/5, European Commission, Ispra This Version is available at: https://hdl.handle.net/10419/268931 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/
Loneliness and health of older adults The role of cultural heritage and relationship quality Casabianca, E. Kovacic, M. EUR XXXXX XX 2022 JRC Working Papers in Economics and Finance, 2022/5
This publication is a Technical report by the Joint Research Centre (JRC), the European Commission’s science and knowledge service. It aims to provide evidence-based scientific support to the European policymaking process. The scientific output expressed does not imply a policy position of the European Commission. Neither the European Commission nor any person acting on behalf of the Commission is responsible for the use that might be made of this publication. For information on the methodology and quality underlying the data used in this publication for which the source is neither Eurostat nor other Commission services, users should contact the referenced source. The designations employed and the presentation of material on the maps do not imply the expression of any opinion whatsoever on the part of the European Union concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. Contact information Name: Matija Kovacic Address: European Commission, Joint Research Centre (JRC), Via Enrico Fermi, 2749, 21027 Ispra (VA), Italy Email: matija.kovac[email protected]uropa.eu EU Science Hub https://ec.europa.eu/jrc JRC129380 Ispra: European Commission, 2022 © European Union, 2022 The reuse policy of the European Commission is implemented by the Commission Decision 2011/833/EU of 12 December 2011 on the reuse of Commission documents (OJ L 330, 14.12.2011, p. 39). Except otherwise noted, the reuse of this document is authorised under the Creative Commons Attribution 4.0 International (CC BY 4.0) licence (https://creativecommons.org/licenses/by/4.0/). This means that reuse is allowed provided appropriate credit is given and any changes are indicated. For any use or reproduction of photos or other material that is not owned by the EU, permission must be sought directly from the copyright holders. All content © European Union, 2022 How to cite this report: Casabianca, E. and Kovacic, M., Loneliness and health of older adults: the role of cultural heritage and relationship quality, European Commission, Ispra, 2022, JRC129380.
Executive summary We estimate the direct causal effect of loneliness on a variety of health outcomes using a sample of secondgeneration immigrants among older adults drawn from the Survey of Health, Ageing and Retirement in Europe. In an effort to account for the endogeneity of self-declared loneliness, we explore the link between loneliness and a specific trait of maternal cultural background strongly associated with quality of relations and use the latter as an instrument for loneliness. We thus also assess the importance of cultural heritage in shaping individuals' perceptions of loneliness. Additionally, we investigate one pathway by which some specific ancestral factors may influence the formation of cultural traits in the modern era. Our results suggest that loneliness has a significant impact on health, both mental and physical. Notably, our identification strategy allows us to uncover a larger effect of loneliness on health than that found in an OLS setting. These findings are robust to a battery of sensitivity checks.
Loneliness and health of older adults The role of cultural heritage and relationship quality Elizabeth Casabianca 1 and Matija Kovacic * 2 1 European Commission, Joint Research Centre (JRC), Ispra, Italy. ORCID: https://orcid.org/0000-0002-0451-5701 2 European Commission, Joint Research Centre (JRC), Ispra, Italy. ORCID: https://orcid.org/0000-0002-3267-5518 This version: 04/04/2022 Abstract We estimate the direct causal eect of loneliness on a variety of health outcomes using a sample of second-generation immigrants among older adults drawn from the Survey of Health, Ageing and Retirement in Europe. In an eort to account for the endogeneity of self-declared loneliness, we explore the link between loneliness and a specic trait of maternal cultural background strongly associated with quality of relations and use the latter as an instrument for loneliness. We thus also assess the importance of cultural heritage in shaping individuals' perceptions of loneliness. Additionally, we investigate one pathway by which some specic ancestral factors may inuence the formation of cultural traits in the modern era. Our results suggest that loneliness has a signicant impact on health, both mental and physical. Notably, our identication strategy allows us to uncover a larger eect of loneliness on health than that found in an OLS setting. These ndings are robust to a battery of sensitivity checks. Keywords: Loneliness, relationship quality, culture, mental health, physical health. JEL Classication: I12, I14, J14, D91, Z13. * Corresponding author : E-Mail: matija.kov[email protected]; [email protected]; Joint Research Centre (JRC), Via E. Fermi 2749 - I-21027 - Ispra (Va) - ITALY. 1
Declarations Conicts of interest/Competing interests : None of the authors have actual or potential conict of interest. The scientic output expressed does not imply a policy position of the European Commission. Neither the European Commission nor any person acting on behalf of the Commission is responsible for the use which might be made of this publication. Availability of data and material : Available upon request/acceptance for publication. Code availability : Available upon request/acceptance for publication. 2
1 Introduction Loneliness is generally understood as the negative experience (or feeling) arising when an individual perceives a signicant mismatch between actual and desired (or ideal) social interaction (Perlman and Peplau, 1981; Peplau et al., 1982). In other words, individuals feel lonely when current circumstances do not fulll their ideal targets (Erber and Gilmour, 2013). According to this denition, one might feel lonely despite enjoying a large social network and a high quantity of social connections (van Baarsen et al., 2001; Albert, 2021). Loneliness is widely recognized as being a public health issue. A meta-analytic review of nearly 150 studies by Holt-Lunstad et al. (2015) nds that the risk of premature mortality linked to loneliness is stronger than the risk associated with obesity and physical inactivity. Extensive literature, prevalently psychological, also documents consistent associations between loneliness and mental and physical health. Studies reveal that lonelier individuals are at higher risk of depression, suicidal ideation and suicide attempts, cardiovascular disease and cognitive decline (Cacioppo et al., 2014a; Cacioppo and Cacioppo, 2018; Cacioppo and Hawkley, 2009; Cacioppo et al., 2014b; Leigh-Hunt et al., 2017; Steptoe et al., 2013; Stickley et al., 2013; Stickley and Koyanagi, 2016; Valtorta et al., 2016, among many others). Besides being a public health issue, loneliness is an economic problem, too. Lonely people are more likely to use healthcare services (e.g. doctor visits, hospital admissions). Thus, loneliness may be associated with higher healthcare expenditure (Kung et al., 2021). The cost of loneliness to employers can be substantial and linked to increased absence, loss of productivity and increased voluntary turnover resulting from low job satisfaction (Michaelson et al., 2021). A rst attempt at estimating the eects of loneliness on economic growth, nds that regions with a higher share of lonely people have a more limited capacity to generate additional wealth (Burlina and Rodríguez-Pose, 2021). Despite this large body of evidence, studies so far are mainly based on descriptive analysis or simple multivariate regression models. Hence, they are only able to reveal correlations and can say little about causation . Courtin and Knapp (2017) call for more research on causal pathways that better link the evidence of the risk factors for loneliness [...] and their evidence on their impact on health. We are responding to this call and with this paper, we ll existing gaps in the related literature. We estimate the causal impact of loneliness on a variety of health-related measures using individual3
level data drawn from the Survey of Health, Ageing and Retirement in Europe (SHARE, Börsch-Supan, 2008). SHARE is a multidisciplinary longitudinal survey on ageing which focuses on individuals aged 50+ and their spouses. It contains rich information on individuals' mental and physical health statuses and allows us to link information on the respondents' current situation to retrospective childhood data, including parents' country of origin. It also includes questions that allow us to build metrics that correspond to widely recognized measures of loneliness: the three-item version of the Revised UCLA Loneliness scale (Russell et al., 1978) - an indirect measure of loneliness - and the single-item loneliness scale, which asks about loneliness directly. This set of features makes SHARE particularly suitable for our purposes. Moreover, the sample is representative of the older population, who is especially vulnerable to loneliness and its healthrelated implications (Vozikaki et al., 2018). While loneliness can aect people of all ages, older adults are more exposed to the risk of feeling lonely because they are more likely to experience life transitions and disruptive life events, such as retirement and bereavement, that increases loneliness. The absence of supportive relationships, low-quality or unfullling relationships is another factor associated to increased loneliness among older people (Victor et al., 2005). To isolate the impact of loneliness on health, our identication strategy consists of three primary elements. First, we explore how certain cultural dimensions aect loneliness. A number of studies show the importance of culture, dened as the set of social values and norms shared by people in a place or time, in determining individual's experiences of loneliness. This literature mainly focuses on the distinction between individualistic and collectivistic cultures (Hofstede et al., 2010) based on the quantity and extension of social networks within a society, which are more limited in the former compared to the latter. Hence, people in individualistic societies with weaker social connections should feel more lonely compared to those in collectivistic societies. Yet, when it comes to cross-cultural comparisons, this association does not always hold and the evidence is still rather mixed (for example, Lykes and Kemmelmeier, 2014; Van Tilburg et al., 2004). We argue that one reason for these discrepancies is conceptual in nature. Loneliness is not just about the quantity or frequency of social interactions. Instead, it is the negative feeling arising from the perceived discrepancy between actual and ideal quality of social relationships. The key role of quality of social contacts in predicting 4
loneliness has been already uncovered to a degree (Pinquart and Sörensen, 2003; Hawkley et al., 2008; Beller and Wagner, 2018). When evaluating actual social relations, individuals are more concerned with the quality of their social ties independent of the perceived desired size of their social networks. Based on this evidence, we link loneliness to indulgent vis -àvis restraint cultures (Hofstede et al., 2010), a specic cultural trait strongly associated with quality of relations, extraversion, happiness and high importance of having friends and positive feelings. While individuals in indulgent societies place more value in enjoying life without social restrictions, in restraint societies the prevalent belief is that enjoyment is somewhat wrong. Our focus on this dimension of culture represents one of the contributions of our work to the related literature. Second, we limit our main analytical sample to native individuals with at least one foreign-born parent (i.e. second-generation immigrants) to whom we assign the cultural trait belonging to their parents' country of origin. This epidemiological approach (Galor et al., 2020; Bernhofer et al., 2021, among other) allows us to separate the eect of culture from other country-specic factors, otherwise captured by a single country eect. In other words, we exploit the exogenous variation in parental cultural background while keeping the other country-specic factors invariant. Our model is based on two key assumptions: parents' foreign cultural background is transmitted to their children and it is long lasting, meaning that it aects individuals' beliefs, emotions and choices throughout life. Although this assumption might be strong, existing evidence shows that cultural values are transferred from parents to children, and rarely change in life (Beugelsdijk and Welzel, 2018). Third, we use the degree of restraint of maternal country of origin as an instrument for individual self-declared loneliness. More specically, we nd that maternal cultural background shows a statistically signicant eect on self-reported loneliness. This is in line with extant evidence on the inter-generational transmission of attitudes and behaviour (among the most recent contributions see Cipriani et al., 2013; Sgroi et al., 2020). In an extension of our analysis, we investigate one possible pathway by which some specic ancestral factors may have inuence on the formation of cultural traits associated with higher levels of restraint in the modern era. The combination on these three elements allow us to estimate a direct causal eect of loneliness on 5
Galor et al., 2020, Bernhofer et al., 2021), and focuses on native individuals with one or both foreign-born parents ( i.e., second-generation immigrants). For the cultural hypothesis to be consistent, immigrants who are identical in all aspects except for their cultural backgrounds should experience dierent levels of loneliness. The epidemiological approach relies on three main assumptions: i) cultural values and beliefs are vertically transmitted from parents to children, ii) cultural values systematically vary across individuals having dierent cultural backgrounds; and iii) despite the heterogeneity in their cultural backgrounds, individuals living in the same country (or region) face identical economic and institutional arrangements. Our main analytical sample, therefore, consists of native individuals with one or both foreign-born parents. This identication strategy allows us to exploit the exogenous variation in parental cultural backgrounds while keeping the other country-specic factors invariant. 5 3.2 Hypotheses Our framework consists of a set of hypotheses that we test empirically using data described in Section 4. The rst hypothesis tests the assumption that a greater general tendency to evaluate actual social relationships negatively as a result of the stricter social norms and prohibitions that characterize restraint societies increases the risk of loneliness: Hypothesis 1 Relationship quality and loneliness Individuals with cultural backgrounds characterized by negative cognitive evaluations of one's life and feelings are, on average, more likely to feel lonely, ceteris paribus . The relationship between negative evaluations of relationship quality and loneliness should hold independently of the ideal variety of social connections. We do not rule out the possibility of loneliness occurring at all levels of social embeddedness (van Staden and Coetzee, 2010; Heu et al., 2021a), but argue that individual satisfaction with social situations is more important than the size of social networks or the degree of physical isolation. Formally, two individuals, i and j , with the same ideal targets in terms of social embeddedness, 5 Another key assumption of our framework is that cultural heritage is long-lasting, meaning that it aects individual's beliefs, emotions and choices throughout their life. 12
E(xi) = E(xj) = E(x) , may have dierent evaluations of actual relationships, x < E(x) . Suppose that xi> xj . Since loneliness is conceptualized as an unpleasant feeling arising from the perceived discrepancy between existing and optimal quality, i.e., L=|x−E(x)| , the individual with a higher evaluation of actual social embeddedness will feel less lonely, i.e., L(xi)< L(xj) . This yields our second hypothesis: Hypothesis 2 Relationship quality and quantity, and loneliness Individuals with cultural backgrounds characterized by negative cognitive evaluations of their lives and feelings are, on average, more likely to feel lonely, regardless of the extent of social networks, frequency of social interactions, and degree of integration into social groups, ceteris paribus . Finally, to analyze the relationship between culture, loneliness and health, we empirically validate the following hypothesis: Hypothesis 3 Loneliness increases the likelihood of mental and physical health problems Increasing loneliness negatively aects mental and physical health-related outcomes and functional decline, ceteris paribus . 3.3 Empirical model According to our hypotheses, the perception of relationship quality may have a direct inuence on loneliness, which in turn aects individuals' mental and physical health. This chain mechanism can be analyzed by means of a two-stage estimation model. In the rst stage, we estimate the relationship between parental cultural background and loneliness (Hypotheses 1 and 2). We choose restraint as our preferred cultural dimension as it is strongly associated with perceived quality of relations. In the second stage we quantify the causal eect of loneliness on health (Hypothesis 3). The empirical problem, therefore, consists in estimating the following causal relationship: Healthi,p,c =α+βLi+ψXi+ρF Ei+ηi, (1) where Healthi,p,c is an indicator measuring mental or physical health of individual i with parental ancestry 13
p , born and currently residing in country c , Li denotes a measure of the individual i 's loneliness, Xi is a full set of individual level characteristics, and F Ei are the country of current residence and wave xed eects. In the rst stage we empirically validate Hypothesis 1 by estimating the eects of parental cultural background and other covariates on loneliness: Li=α+πi1RESi+πi2Xi+πi3F Ei+ζi (2) where RESi denotes the index of restraint of the parental country of origin. In order to verify Hypothesis 2, we separately estimate Equation 2 by adding the index of individualism associated to individual i0s parental country of origin, INDi : Li=α+πi1RESi+πi2INDi+πi3Xi+πi4F Ei+ζi (3) Since we expect that more restraint cultural backgrounds increase the risk of loneliness, we anticipate coecient πi1 to be positive. By plugging the rst stage tted values from Equation (2) in the second stage equation we obtain the reduced form model for health-related outcomes: Healthi,p,c =α+β c Li+ψXi+γFEi+errori. (4) Since loneliness is expected to aect the incidence of mental and physical health problems, the empirical validation of Equation (4) should yield a positive coecient of L . Given the particularities of our empirical strategy, in all model specications robust standard errors are clustered both at the country of residence and the parental country of origin level. For the two-stage empirical model (2) and (4) to work, the cultural marker we employ must satisfy two basic requirements: a) it must be correlated with the endogenous variable (instrument relevance), and b) it must be uncorrelated with the error term (independence). Moreover, the exclusion restriction requires that the level of restraint of the parental country of origin should not have any direct impact on individuals' health other than through its direct impact on loneliness. Despite the fact that the exclusion restriction cannot be directly tested and the literature lacks evidence of direct cultural eects on health outcomes, we provide some reassuring statistical evidence showing that RESi has no direct impact on health, and that it 14
does not indirectly inuence health through other factors, such as unhealthy behaviours (physical inactivity, smoking, drinking, and unhealthy dietary habits). In addition, we run an over-identication test, which provides further proof that the exclusion restriction is satised. 4 Data The individual-level data employed in this study are drawn from the Survey of Health, Ageing and Retirement in Europe (SHARE, Börsch-Supan, 2008). SHARE is a multidisciplinary longitudinal survey on ageing which focuses on individuals aged 50+ and their spouses. The survey contains both the regular and retrospective waves (SHARELIFE). The regular rounds collect information on the individuals' current situation, such as health, working situation, social network/relations, accommodation, economic situation/assets, behavioural risks, and expectations. In addition, two survey rounds add retrospective information on multiple dimensions of the respondents' past (health, health care, accommodation, working career, household situation and performance at school during childhood, number of children, childbearing for women, emotional experiences in early life, relationship with parents, adverse childhood experiences, etc.). Our sample includes individuals who were interviewed in four consecutive waves, namely 5, 6, 7, and 8 (release 8.0.0), for which information on loneliness is available. 6 What makes SHARE data particularly suited for the purposes of our analysis is the availability of a rich set of information on individuals' physical and mental health conditions, as well as drug consumption. In addition, the retrospective component of the SHARE data allows to link the information on the respondents' current situation to the retrospective childhood/adulthood data. Below we describe the variables used in the analysis. Table C.1 (in Appendix C) reports summary statistics, while Table C.3 reports the list of countries included in the analysis. Health outcomes 6 We consider a pool of individuals in which respondents interviewed in more than one wave are treated as separate observations since loneliness and some other individual-specic characteristics may change from one wave to another. When considering the pooled sample of individuals, however, the number of observations increases. Thus, we perform a robustness check by only considering refreshers, i.e., if the same individual appears in more than one wave, we only include in our sample the rst recorded observation. Section 6 includes results using the sample of refreshers. 15
We consider six health indicators: EURO-D depression scale (Prince et al., 1999), which ranges from 1 (absence of depressive symptoms) to 12 (severe depressive symptoms); 7 number of limitations with activities of daily living (ADL); body mass index (BMI); number of chronic diseases, and self-assessed health (ranging from excellent to poor). In addition, we consider a set of binary variables referring to the consumption of drugs (medicines) for six health problems, namely anxiety, sleeping problems, cholesterol, diabetes, pain and high blood pressure. As for the ADL measure, the respondents are given a list of ten everyday activities and asked to declare whether they have any diculty doing each of these activities excluding any diculties that they expect to last less than three months. Similarly, for chronic diseases, the respondents are given a list with 21 dierent items and asked how many of them they have been diagnosed or for how many they are currently being treated for or bothered by. Together with the overall prevalence (intensity of occurrence) of chronic diseases, we also estimate separately the eect of loneliness on ve dierent physical health-related factors, namely diabetes, high blood pressure, stomach or duodenal ulcer, and peptic ulcer, high blood cholesterol and stroke. The onset of these factors is captured by a set of binary variables. Loneliness To assess loneliness, a short three-item version (Hughes et al., 2004) of the Revised UCLA Loneliness scale (henceforth, R-UCLA) (Russell et al., 1978) was used. The exact wording of the items are: How often do you feel isolated from others? , How often do you feel you lack companionship? , How often do you feel left out? . In each case, the available responses are: 1. Often, 2, Some of the time, 3. Hardly ever or never. A sum score was computed, therefore the scale ranges from 3 (not lonely) to 9 (very lonely). It has been shown that this tool has favorable psychometric characteristics (Hughes et al., 2004). A multi-item measure that does not mention loneliness directly can be particularly useful when dealing with particular population groups, such as older people. This is because older people are often reluctant to admit to loneliness (Qualter et al., 2021). Also, there is variation in how people understand the term "loneliness" and a multi-item 7 The EURO-D depression scale consists of 12 elements connected to psychological health: depression, pessimism, willingness to die, guilt complexes, sleeping diculties, lack of interests, irritability, lack of appetite, fatigue, lack of concentration, inability to take pleasure from normal activities and a tendency to cry. Each item is of equal weighting and is reported with a 0 if the symptom is absent and a 1 when it is present. 16
measure that does not mention loneliness directly can help to alleviate these concerns. In addition to a multi-item measure of loneliness, we also consider a direct explicit measure on how often the respondent feels lonely. The question is as follows: How often do you feel lonely? . The available response categories are: 1. Often; 2. Some of the time, and 3. Hardly ever or never . We dene the individual as "lonely" if s/he answers "1. Often" or "2. Some of the time". Other controls As for the other individual-level characteristics, we consider a rich set of demographic and socio-economic information. Among demographics, we include age, gender, marital status, number of children, and whether a respondent lives alone in the household. Marital status is dichotomized into a binary variable, assigning value 1 if the respondent is legally married, or in a legally registered civil union, and 0 corresponds to separated, divorced or never married. Socio-economic variables include the highest educational attainment and occupational status. In addition, we control for the highest level of parental educational attainment (whether either one or both parents hold a tertiary degree). Finally, to account for the household nancial situation, we include the household net worth (assets) and total expenditure (in deciles). 8 Retrospective data allows us to consider a set of childhood circumstances. We control for a specic set of early-life conditions called "Adverse Childhood Circumstances". According to the adult attachment theory proposed by Hazan and Shaver (1987), early experiences in attachment relationships are likely to inuence adult attachment styles and general perception of social relations. The authors suggest that individuals with secure attachments early in life tend to be more positive about themselves and their relationships than their peers with insecure early-life attachments. In order to control for these condition we rely on the retrospective SHARELIFE component of the survey which asks respondents to report information on exposure to child neglect and childhood physical abuse, either from mother, father or third parties. More precisely, we consider the following item capturing the quality of the child-parent relationship: How would you rate the relationship with your mother/your father (or the woman/man that raised you)? 1. Excellent 2. Very good 3. Good 8 We also considered total household income as an alternative to household expenditure and net assets. The results do not change signicantly. 17
4. Fair 5. Poor . The relationship with mother/father in childhood is rated as problematic/negative, if the respondent answers "4. Fair" or "5. Poor". As a sensitivity check, we consider an additional set of childhood circumstances including physical abuse during childhood, absence of a parent, and the respondents' health status when they were 15 years old. 9 In addition, we check whether the inclusion of the frequency of contact with children resizes the eect of loneliness. Finally, we also control for genetic and linguistic distances between country of residence and parental country of origin. As shown by Becker et al. (2020), these measures signicantly correlate with dierences in preferences such as risk aversion, altruism, positive and negative reciprocity, patience and trust, with the eects being particularly pronounced for prosocial traits. Linguistic distances measure the degree to which two countries' languages dier from each other (Fearon, 2003). Genetic distances, on the other hand, quantify the expected genetic distance between two randomly drawn individuals, one from each country, according to the contemporary composition of the population (Spolaore and Wacziarg, 2009, 2018). We use the composite measure of ancestral or temporal distance that is computed as the unweighted average of the standardized values (z-scores) of linguistic and genetic distances. 5 Results Cultural heritage, relationship quality and loneliness Table 1 reports the coecients from a linear model in which the individual's self-declared loneliness is regressed on their parental cultural backgrounds (approximated with the indices of restraint and/or individualism) and the full set of covariates (see Equation 2 and 3). Together with the standard denition of second-generation immigrants, i.e. , natives with either one or both foreign-born parents (columns 1 and 2), we also consider two alternative denitions, namely, native individuals with a foreign-born mother and a native or foreign-born father (columns 3 and 4), and those with a foreign-born father and a native or foreign-born mother. (columns 5 and 6). 10 The results do not change signicantly even when we restrict our 9 Physical harm and lack of understanding are addressed by the following questions: How often did your mother/your father push, grab, shove, throw something at you, slap or hit you? 1. Often 2. Sometimes 3. Rarely 4. Never , and How much did your mother/your father (or the woman/man that raised you) understand your problems and worries? 1. A lot 2. Some 3. A little 4. Not at all . 10 Even though second-generation immigrants (approximately 10% of the sample) were born and raised in the same economic 18
sample to individuals with foreign-born mothers and native fathers, and/or with foreign-born fathers and native mothers. We do not report the results for the subset of immigrants with both foreign-born parents only because the sample coverage was not sucient to produce reliable estimates. The results strongly support Hypotheses 1 and 2. Indeed, the coecients of parental restraint conrm that cultural backgrounds with a stronger tendency to evaluate negatively the quality of actual social relationships positively aect the risk of loneliness. This is true independently of the variety of connections or the extent of social networks as approximated by the index of individualism (column [2], [4] and [6]). Interestingly, only the mother's cultural heritage shows a signicant impact, indicating that parents' cultural origins have unequal eects on shaping children's attitudes in the process of socialization and perception of the quality of their social relationships (column 3 and 4 vis -ávis column 5 and 6). This is in line with the existing empirical evidence on inter-generational transmission of attitudes and behavior (Fernández et al., 2004, Cipriani et al., 2013, Dohmen et al., 2012, Farré and Vella, 2013, Dohmen et al., 2011, Sgroi et al., 2020, among others). In light of this empirical evidence, we believe that the degree of restraint associated with maternal country of origin represents a suitable instrument for loneliness in our model of health. and institutional environment as native individuals, they may still be of feel "marginalised" compared to their peers because of their parental foreign origin and/or because they belong to ethnic enclaves (minorities), which may aect the risk of loneliness (Madsen et al., 2016). Figure B.2 (in Appendix B) shows the incidence of loneliness (measured on the R-UCLA scale) among natives and second-generation migrants. The dierence in means of loneliness between second-generation immigrants and the rest of the population is not statistically dierent from 0, which alleviates potential concerns related to the representativity of our sample. The t-test statistic is -1.0470 with a corresponding two-tailed p-value 0.2951 > 0.05. Furthermore, according to the Kolmogorov-Smirnov test, the null hypothesis of equal distribution of loneliness between second-generation immigrants and the rest of the sample cannot be rejected ( p= 0.315 ). 19
Table 1: The eect of parental restraint on loneliness. SG immigrants (either one or both parents born abroad; foreign-born mothers; foreign-born fathers). Variable Either Either Foreign Foreign Foreign Foreign or both or both Mother Mother Father Father Restraint_Mother 0.005*** 0.004*** 0.006*** 0.005*** (0.001) (0.001) (0.002) (0.002) Restraint_Father -0.000 -0.001 0.001 0.001 (0.002) (0.002) (0.001) (0.001) Individualism_Mother -0.001 -0.001 (0.002) (0.002) Individualism_Father 0.001 -0.001*** (0.002) (0.000) Age 0.002 0.000 0.002 0.000 0.000 -0.001 (0.003) (0.003) (0.004) (0.004) (0.004) (0.004) Female 0.045 0.026 -0.010 -0.026 0.096** 0.090* (0.040) (0.050) (0.060) (0.074) (0.047) (0.050) Low Education 0.146** 0.105* 0.199*** 0.175** 0.143*** 0.101** (0.058) (0.055) (0.071) (0.069) (0.053) (0.050) High Education -0.079 -0.076 -0.027 -0.024 -0.115 -0.108 (0.057) (0.060) (0.058) (0.055) (0.093) (0.097) Retired -0.067 -0.070 -0.143* -0.127 -0.145 -0.134 (0.084) (0.068) (0.083) (0.096) (0.129) (0.142) Unemployed 0.071 0.046 0.066 0.048 -0.021 -0.053 (0.117) (0.144) (0.111) (0.152) (0.134) (0.178) Disabled 0.578*** 0.574*** 0.436*** 0.483*** 0.491*** 0.493*** (0.129) (0.138) (0.038) (0.033) (0.133) (0.157) Employed -0.277*** -0.304*** -0.409*** -0.416*** -0.368*** -0.388*** (0.104) (0.082) (0.124) (0.126) (0.123) (0.126) Married -0.200*** -0.188** -0.270*** -0.313*** -0.193*** -0.154* (0.071) (0.089) (0.095) (0.107) (0.066) (0.082) Divorced -0.114 -0.098 -0.131 -0.184 -0.172** -0.148 (0.105) (0.142) (0.130) (0.157) (0.083) (0.117) Widowed -0.113 -0.102 -0.100 -0.110 -0.189 -0.154 (0.146) (0.163) (0.203) (0.221) (0.158) (0.177) Number of children -0.035** -0.041** -0.044*** -0.053*** -0.054** -0.067*** (0.016) (0.016) (0.010) (0.013) (0.025) (0.024) Live alone 0.407*** 0.413*** 0.343*** 0.311*** 0.497*** 0.520*** (0.107) (0.086) (0.113) (0.103) (0.154) (0.121) Education parents (tertiary) -0.085 -0.076 -0.084 -0.073 -0.122 -0.092 (0.102) (0.099) (0.120) (0.119) (0.112) (0.110) Relationship parents (adverse) 0.120** 0.120** 0.125* 0.122* 0.107** 0.111** (0.048) (0.057) (0.066) (0.073) (0.048) (0.051) Absent parent 0.106* 0.060 0.137* 0.118 0.091 0.037 (0.056) (0.060) (0.074) (0.083) (0.073) (0.077) HH net wealth -0.058*** -0.059*** -0.058*** -0.063*** -0.054*** -0.057*** (0.008) (0.009) (0.009) (0.009) (0.010) (0.010) HH expenditure 0.007 0.003 -0.001 -0.003 0.002 -0.004 (0.014) (0.012) (0.015) (0.014) (0.018) (0.017) Additional controls: Country of residence Yes Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes Yes N. Observations 4395 3956 3315 3022 2944 2604 Notes : The table reports the coecients of linear estimations (OLS). Robust standard errors are clustered at the country of residence and parental country of origin level. "Foreign Mother" indicates individuals with foreign-born mothers (and native or foreign-born fathers). "Foreign Father" indicates individuals with foreign-born fathers (and native or foreign-born mothers). Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . Loneliness and health Table 2 reports the coecients from a two-stage model where self-reported loneliness has been instrumented with the maternal country of origin's degree of restraint. 11 According to the Stock and Yogo (2005) 11 Our baseline specication considers second-generation immigrants dened as native individuals with one or both foreign20
rule of thumb, the F-statistic (F-test=24.04) conrms the strength of our instrument (shown at the bottom of column 1). Since it is not possible to directly test the exogeneity of the instrument, we cannot exclude a-priori that maternal culture may directly inuence health or that its eect may impact health through some other factors, such as unhealthy behaviours. Table C.4 (in Appendix C) shows the eects of the maternal restraint on a wide range of health outcomes and four indicators of unhealthy behaviour for lonely and non-lonely individuals (as measured by R-UCLA scale). 12 Estimates suggest that there is no direct association between the restraint indicator and health, and unhealthy behaviour. The only exception is EURO-D (and the probability of practicing vigorous sports), where the eect of maternal restraint is statistically signicant, but only among lonely individuals. While the lack of a direct link between maternal restraint and health does not imply that the exclusion restriction is satised, this nding may be viewed as reassuring, suggesting that the eect of maternal cultural background inuences health only indirectly through loneliness. Moreover, when instrumenting loneliness with both the maternal and paternal restraint, the Sargan test statistic (0.91) conrms that the over-identication restrictions are valid. 13 In line with the results from Table 1, the rst stage coecients in Table 2 (column 1) suggest that loneliness is particularly pronounced for individuals living alone and among those with disabilities. Similarly, adverse early life conditions such as the absence of a parent or a low-quality parent-child relationship correlate positively with loneliness. Meanwhile, wealthier, married, and/or employed individuals with more kids feel less lonely. 14 These ndings are in line with previous research (Beutel et al., 2017; ; Soest et al., 2018; Hajek and König, 2020). The results from the second-stage (columns 2-6) suggest that loneliness directly increases the probability of depression (Eurod), functional decline (Adl), and high body mass index (Bmi). More specically, a gradual increase in loneliness causes a 1.81 point increase in the intensity of depressive symptoms as measured by the born parents. The maternal restraint for individuals with foreign-born fathers and native mothers is identical to their country of birth restraint score. The results, however, are robust to the exclusion of these individuals, i.e., when we focus only on second-generation immigrants with foreign-born mothers (Table C.7, in Appendix C). 12 Individuals scoring 0 on the R-UCLA scale are considered as "non-lonely"; individuals scoring R-UCLA > 0 are dened as "lonely"). 13 This additional evidence is available upon request. 14 The interpretation of the association between loneliness and emotional experiences such as the parent-child relationship requires caution since it may be subject to recall bias and "colouring". However, by assessing the internal and external consistency of the measures of childhood socio-economic status and health, Havari and Mazzonna (2015) found that overall respondents seem to remember fairly well their childhood conditions. 21
in Table 6, where we regress loneliness on the predicted restraint from Table 5 (column 5). In line with the results from Table 1, only the mother's cultural background shows a signicant impact on loneliness. Finally, the ndings in Table 7 show that culture, even when approximated by historical agro-climatic conditions, is a good instrument for loneliness. The eects of loneliness on mental health, functional decline, and BMI, are very similar to those reported in Table 2 for the baseline specication. Table 6: The eect of parental restraint on loneliness. SG immigrants (either one or both parents born abroad; foreign-born mother; foreign-born father). Variable Either Either Either Foreign Foreign or both or both or both Mother Father Restraint_Mother (tted) 0.008*** 0.009*** 0.009** (0.003) (0.003) (0.003) Restraint_Father (tted) 0.002 -0.000 0.002 (0.003) (0.004) (0.003) Full set of v. Table 2 Yes Yes Yes Yes Yes Additional controls: Country of residence Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes N. Observations 4402 4202 4030 3151 2758 Notes : The table reports the coecients of linear estimations (OLS). Robust standard errors are clustered at the country of residence and parental country of origin level. "Foreign Mother" indicates individuals with foreign-born mothers (and native or foreign-born fathers). "Foreign Father" indicates individuals with foreign-born fathers (and native or foreign-born mothers). Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . Table 7: The eect of loneliness on health. Baseline specication. SG immigrants (either one or both parents born abroad). Variable 1ST 2ND 2ND 2ND 2ND 2ND Loneliness Eurod Adl Bmi Chronic SAH Fitted values 0.008*** (0.003) Loneliness (R-UCLA) 1.893*** 0.412* 4.957* 0.285 0.474 (0.705) (0.218) (2.892) (0.565) (0.407) Additional controls: Country of residence Yes Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes Yes 1st stage F 9.40 1st stage p-value 0.0074 N. Observations 4402 4402 4402 4402 4402 4402 Notes : The table reports the coecients of two-stage linear estimations. Abbreviations: 1ST - rst stage estimates; 2ND - second stage estimates. Robust standard errors are clustered at the country of residence and parental country of origin level. Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . Additional sensitivity checks Tables 8 to 11 report a battery of additional robustness checks for our baseline specication of Table 2. 28
The evidence in Table 8 conrms the strength of parental restraint even when the individuals are directly asked whether and how often they feel lonely. In line with the previous ndings, the eect of loneliness is signicant for mental health, functional decline, and body mass index. Being lonely translates into a 9.2 unit increase in depression and 2.2 more limitations in daily activities, and increases the body mass index by 23.8 points. Table 8: The eect of loneliness on health. SG immigrants (either one or both parents born abroad), direct question on loneliness. Variable 1ST 2ND 2ND 2ND 2ND 2ND Loneliness Eurod Adl Bmi Chronic SAH Restraint_mother 0.001*** (0.000) Lonely (direct question) 9.157*** 2.198*** 23.785* 0.541 1.333 (2.234) (0.504) (13.352) (2.740) (2.029) N. Observations 4699 4699 4699 4699 4699 4699 Full set of v. Table 2 Yes Yes Yes Yes Yes Yes Country of residence Yes Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes Yes 1st stage F 59.44 1st stage p-value 0.0056 Notes : The table reports the coecients of two-stage linear estimations. Abbreviations: 1ST - rst stage estimates; 2ND - second stage estimates. Robust standard errors are clustered at the country of residence and parental country of origin level. Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . Similarly, the inclusion of an additional adverse childhood condition (harm from parents) and health conditions in early life does not alter the eect of loneliness (Table 9). 29
Table 9: The eect of loneliness on health. SG immigrants (either one or both parents born abroad), additional adverse childhood conditions and health status during childhood Variable 1ST 2ND 2ND 2ND 2ND 2ND Loneliness Eurod Adl Bmi Chronic SAH Restraint_Mother 0.005*** (0.001) Loneliness (R-UCLA) 1.765*** 0.487*** 5.330* -0.146 0.224 (0.663) (0.139) (2.994) (0.480) (0.422) Relationship parents (adverse) 0.058 -0.011 -0.088*** -0.431 -0.131** -0.036 (0.050) (0.046) (0.013) (0.329) (0.059) (0.036) Harm from parents 0.098* -0.068 -0.023 -0.441 0.131 -0.039 (0.056) (0.114) (0.021) (0.451) (0.081) (0.064) Harm from third parties 0.252*** -0.214 -0.058 -1.264 0.153 -0.034 (0.053) (0.266) (0.059) (0.960) (0.269) (0.167) Understanding parents (lack of) 0.161*** 0.111 -0.013 -0.887* 0.172** 0.017 (0.050) (0.161) (0.031) (0.537) (0.084) (0.065) Health when 15yo 0.040 0.073 0.012 -0.275 0.147*** 0.112*** (0.028) (0.046) (0.018) (0.270) (0.028) (0.028) N. Observations 4040 4040 4040 4040 4040 4040 Full set of v. Table 2 Yes Yes Yes Yes Yes Yes Country of residence Yes Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes Yes 1st stage F 20.70 1st stage p-value 0.0005 Notes : The method of estimation is a two-stage linear estimation. Abbreviations: 1ST - rst stage estimates; 2ND - second stage estimates. Robust standard errors are clustered at the country of residence and parental country of origin level. Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . The likelihood of loneliness may be inuenced as well by the frequency of parent-ospring interactions (contacts). In Table 10 we control for three dierent frequency levels (with contact on a daily basis as the reference category): frequent contact (several times a week), fair contact (once a week or every two weeks) and rare contact (once a month, less than once a month, never). Loneliness is positively associated with rare contact with children, and the eects are similar to those shown in Tables 2 and 9. 30
Table 10: The eect of loneliness on health. SG immigrants (either one or both parents born abroad). Sub-sample with kids: frequency of the contact with children Variable 1ST 2ND 2ND 2ND 2ND 2ND Loneliness Eurod Adl Bmi Chronic SAH Restraint_Mother 0.007*** (0.001) Loneliness (R-UCLA) 1.748*** 0.369*** 4.648*** 0.265 0.181 (0.455) (0.095) (1.155) (0.448) (0.299) Frequent contact 0.029 -0.007 -0.022 0.038 0.079* -0.027 (0.052) (0.093) (0.023) (0.238) (0.047) (0.046) Fair contact 0.130** -0.140 -0.046 -0.671* -0.003 0.039 (0.057) (0.229) (.) (0.405) (0.053) (0.081) Rare contact 0.362*** -0.283 -0.082 -1.995** 0.073 0.072 (0.111) (0.331) (0.067) (0.819) (0.139) (0.132) N. Observations 2806 2806 2806 2806 2806 2806 Full set of v. Table 2 Yes Yes Yes Yes Yes Yes Country of residence Yes Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes Yes 1st stage F 33.33 1st stage p-value 0.0000 Notes : The method of estimation is a two-stage linear estimation. Reference category: every-day contact. Abbreviations: 1ST - rst stage estimates; 2ND - second stage estimates. Robust standard errors are clustered at the country of residence and parental country of origin level. Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . As expected, genetic and linguistic distances reduce the eect of loneliness, although the eect is not signicant (Table 11, column 1). This is not surprising evidence since these measures strongly correlate with cross-cultural dierences in pro-social traits. The eect of loneliness on mental health remains highly signicant. Loneliness still positively relates to functional decline and higher BMI, although the relationship is statistically weaker compared to our baseline results. 31
Table 11: The eect of loneliness on health. SG immigrants (either one or both parents born abroad), temporal distances from Becker et al. (2020) Variable 1ST 2ND 2ND 2ND 2ND 2ND Loneliness Eurod Adl Bmi Chronic SAH Restraint_Mother 0.005*** (0.001) Loneliness (R-UCLA) 1.342*** 0.294* 6.382* -0.180 0.209 (0.468) (0.176) (3.757) (0.745) (0.554) Ancestral distance (z-score) -0.034 0.016 0.001 0.350 -0.056 -0.036 (0.029) (0.060) (0.011) (0.325) (0.061) (0.039) N. Observations 3609 3609 3609 3609 3609 3609 Full set of v. Table 2 Yes Yes Yes Yes Yes Yes Country of residence Yes Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes Yes 1st stage F 11.44 1st stage p-value 0.0038 Notes : The method of estimation is a two-stage linear estimation. Abbreviations: 1ST - rst stage estimates; 2ND - second stage estimates. Robust standard errors are clustered at the country of residence and parental country of origin level. Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . Finally, we show that our baseline results are robust to alternative denitions of our sample. Table C.6 reports results on the sample of refreshers, i.e. , we only consider the rst observation of respondents who appear more than once in the main sample. Our ndings also hold if we exclude from our sample rst-generation immigrants and consider respondents with mothers born abroad only (Table C.8 and C.7, respectively). 7 Conclusions In this paper, we estimate the direct causal eect of loneliness on a variety of health outcomes using a sample of second-generation immigrants among older adults in Europe. In an eort to account for the endogeneity of self-declared loneliness, we use a specic trait of maternal cultural background strongly associated with perceived quality of relationships as an instrument for loneliness. We therefore also assess the importance of cultural heritage is shaping individuals' perceptions of loneliness. We contribute to the related literature in a number of ways. First, we estimate a direct causal eect on loneliness on multiple health outcomes. Second, we link loneliness to the degree of restraint of a society, a specic cultural trait strongly associated with quality of relations, extraversion, happiness and high importance of having friends and positive feelings. In an extension of our analysis, we also investigate one 32
possible pathway by which some specic ancestral factors may have inuence on the formation of cultural traits associated with higher levels of restraint in the modern era. Third, in order to separate the eect of culture from other country-specic factors, we focus on a sample of second-generation immigrants to whom we assign the cultural trait belonging to their mothers' country of origin. We nd that loneliness has a signicant impact on individual's health status. More specically, loneliness increases the probability of depression, suicidal ideation, and functional decline. Among physical healthrelated factors and medications, loneliness increases the likelihood of diabetes, hypertension and stroke. Our baseline results are robust to a number of sensitivity checks. We also uncover a strong relationship between the quality of relations and loneliness. The risk of loneliness signicantly increases among individuals with more restraint cultural backgrounds compared to their peers with more indulgent cultural heritages. This result holds independently of the variety and frequency of social interactions. Based on this evidence, we use maternal cultural background as an instrument for individuals' self-declared loneliness in our model of health, which proves to be a strong and valid instrument. This is in line with extant evidence on the inter-generational transmission of attitudes and behaviour from mothers to their children. Our results are in line with the related literature, which however only show correlations and no causation between loneliness and health. More importantly, our identication strategy allows us to uncover a larger eect of loneliness on health than that found in an OLS setting, which conrms the presence of a strong reverse causality between loneliness and health. A few implications derive from our work. Loneliness poses a serious threat to health with repercussion both from a social and economic point of view. Cultural heritage and perceived quality of social relations play a prominent role in shaping individuals' experiences of loneliness. In this context, the need to implement eective policies and interventions targeted at reducing loneliness becomes imperative. However, one-size- ts-all solutions may not be the most appropriate. When designing loneliness interventions, policy makers will also have to account for the diverse ways in which individuals experience loneliness across societies with dierent cultural backgrounds. 33
References I. Albert. Perceived loneliness and the role of cultural and intergenerational belonging: the case of portuguese rst-generation immigrants in luxembourg. 18(3):299310, 2021. C. A. Anderson. Attributional style, depression, and loneliness: A cross-cultural comparison of american and chinese students. Personality and Social Psychology Bulletin, 25(4):482499, 1999. doi: 10.1177/ 0146167299025004007. URL https://doi.org/10.1177/0146167299025004007 . A. Becker, B. Enke, and A. Falk. Ancient Origins of the Global Variation in Economic Preferences. AEA Papers and Proceedings, 110:319323, may 2020. ISSN 2574-0768. doi: 10.1257/pandp.20201071. URL https://pubs.aeaweb.org/doi/10.1257/pandp.20201071 . A. K. Bekhet and J. A. Zauszniewski. Mental health of elders in retirement communities: Is loneliness a key factor? Archives of Psychiatric Nursing, 26(3):214224, 2012. ISSN 0883-9417. doi: https://doi.org/10.1016/j.apnu.2011.09.007. URL https://www.sciencedirect.com/science/article/ pii/S0883941711001294 . J. Beller and A. Wagner. Disentangling loneliness: Dierential eects of subjective loneliness, network quality, network size, and living alone on physical, mental, and cognitive health. Journal of Aging and Health, 30(4): 521539, 2018. doi: 10.1177/0898264316685843. URL https://doi.org/10.1177/0898264316685843 . PMID: 28553795. J. Bernhofer, F. Costantini, and M. Kovacic. Risk attitudes, investment behavior and linguistic variation. Journal of Human Resources, 2021. doi: 10.3368/jhr.59.2.0119-9999R2. URL http://jhr.uwpress.org/ content/early/2021/05/04/jhr.59.2.0119-9999R2.full.pdf+html . S. Beugelsdijk and C. Welzel. Dimensions and dynamics of national culture: Synthesizing hofstede with inglehart. Journal of Cross-Cultural Psychology, 49(10):14691505, 2018. doi: 10.1177/0022022118798505. URL https://doi.org/10.1177/0022022118798505 . PMID: 30369633. 34
M. Beutel, E. Klein, E. Brähler, I. Reiner, C. Jünger, M. Michal, J. Wiltink, P. Wild, T. Munzel, K. Lackner, and A. N. Tibubos. Loneliness in the general population: Prevalence, determinants and relations to mental health. BMC Psychiatry, 17, 03 2017. doi: 10.1186/s12888-017-1262-x. A. Börsch-Supan. First results from the Survey of Health, Ageing and Retirement in Europe (2004-2007): Starting the longitudinal dimension. Univerza v Ljubljani, Ekonomska fakulteta, 2008. A. Buchman, P. Boyle, R. Wilson, B. James, S. Leurgans, S. Arnold, and D. Bennett. Loneliness and the rate of motor decline in old age: the rush memory and aging project, a community-based cohort study. BMC geriatrics, 10:77, 10 2010. doi: 10.1186/1471-2318-10-77. C. Burlina and A. Rodríguez-Pose. Alone and lonely. the economic cost of solitude for regions in europe. CEPR Discussion Paper 16674, 2021. URL https://cepr.org/active/publications/discussion_ papers/dp.php?dpno=16674 . J. T. Cacioppo and S. Cacioppo. Chapter three - loneliness in the modern age: An evolutionary theory of loneliness (etl). volume 58 of Advances in Experimental Social Psychology, pages 127197. Academic Press, 2018. doi: https://doi.org/10.1016/bs.aesp.2018.03.003. URL https://www.sciencedirect.com/ science/article/pii/S0065260118300145 . J. T. Cacioppo and L. C. Hawkley. Perceived social isolation and cognition. Trends in cognitive sciences, 13 (10):447454, 2009. J. T. Cacioppo, S. Cacioppo, and D. I. Boomsma. Evolutionary mechanisms for loneliness. Cognition & emotion, 28(1):321, 2014a. ISSN 0269-9931. doi: 10.1080/02699931.2013.837379. URL https:// europepmc.org/articles/PMC3855545 . S. Cacioppo, J. P. Capitanio, and J. T. Cacioppo. Toward a neurology of loneliness. Psychological bulletin, 140(6):1464, 2014b. M. Cipriani, P. Giuliano, and O. Jeanne. Like mother like son? experimental evidence on the transmission of values from parents to children. Journal of Economic Behavior and Organization, 90:100111, 2013. 35
ISSN 0167-2681. doi: https://doi.org/10.1016/j.jebo.2013.03.002. URL https://www.sciencedirect. com/science/article/pii/S0167268113000358 . E. Courtin and M. Knapp. Social isolation, loneliness and health in old age: a scoping review. Health & social care in the community, 25(3):799812, 2017. T. Dohmen, D. Human, J. Schupp, A. Falk, U. Sunde, and G. G. Wagner. Individual risk attitudes: Measurement, determinants, and behavioral consequences. Journal of the European Economic Association, 9(3):522550, 2011. ISSN 15424766, 15424774. URL http://www.jstor.org/stable/25836078 . T. Dohmen, A. Falk, D. Human, and U. Sunde. The intergenerational transmission of risk and trust attitudes. The Review of Economic Studies, 79(2):645677, 2012. ISSN 00346527, 1467937X. URL http://www.jstor.org/stable/23261346 . J. Domènech-Abella, J. Mundó, J. M. Haro, and M. Rubio-Valera. Anxiety, depression, loneliness and social network in the elderly: Longitudinal associations from the irish longitudinal study on ageing (tilda). Journal of Aective Disorders, 246, 12 2018. doi: 10.1016/j.jad.2018.12.043. P. A. Dykstra. Older adult loneliness: myths and realities. European journal of ageing, 6(2):91100, June 2009. ISSN 1613-9372. doi: 10.1007/s10433-009-0110-3. URL https://europepmc.org/articles/ PMC2693783 . R. Erber and R. Gilmour. Theoretical Frameworks for Personal Relationships. Taylor & Francis, 2013. ISBN 9781134754465. URL https://books.google.it/books?id=hfpBz22oMVMC . A. Falk, A. Becker, T. Dohmen, B. Enke, D. Human, and U. Sunde. Global Evidence on Economic Preferences*. The Quarterly Journal of Economics, 133(4):16451692, 05 2018. ISSN 0033-5533. doi: 10.1093/qje/qjy013. URL https://doi.org/10.1093/qje/qjy013 . L. Farré and F. Vella. The intergenerational transmission of gender role attitudes and its implications for female labour force participation. Economica, 80(318):219247, 2013. ISSN 00130427, 14680335. URL http://www.jstor.org/stable/24029846 . 36
J. Fearon. Ethnic and cultural diversity by country*. Journal of Economic Growth, 8:195222, 2003. doi: https://doi.org/10.1023/A:1024419522867. URL https://link.springer.com/article/10.1023/ A:1024419522867 . R. Fernández. Chapter 11 - does culture matter? volume 1 of Handbook of Social Economics, pages 481510. North-Holland, 2011. doi: https://doi.org/10.1016/B978-0-444-53187-2.00011-5. URL https: //www.sciencedirect.com/science/article/pii/B9780444531872000115 . R. Fernández, A. Fogli, and C. Olivetti. Mothers and Sons: Preference Formation and Female Labor Force Dynamics*. The Quarterly Journal of Economics, 119(4):12491299, 11 2004. ISSN 0033-5533. doi: 10.1162/0033553042476224. URL https://doi.org/10.1162/0033553042476224 . T. Fokkema, J. D. J. Gierveld, and P. A. Dykstra. Cross-national dierences in older adult loneliness. The Journal of Psychology, 146(1-2):201228, 2012. doi: 10.1080/00223980.2011.631612. URL https: //doi.org/10.1080/00223980.2011.631612 . PMID: 22303621. O. Galor and O. Özak. The agricultural origins of time preference. American Economic Review, 106(10): 30643103, October 2016. doi: 10.1257/aer.20150020. URL https://www.aeaweb.org/articles?id=10. 1257/aer.20150020 . O. Galor, Özak, and A. Sarid. Linguistic Traits and Human Capital Formation. AEA Papers and Proceedings, 110:309313, May 2020. doi: 10.1257/pandp.20201069. URL https://ideas.repec.org/a/aea/apandp/ v110y2020p309-13.html . P. Giuliano. Living arrangements in western europe: Does cultural origin matter? Journal of the European Economic Association, 5:927952, 02 2007. doi: 10.2139/ssrn.894038. A. Hajek and H.-H. König. Which factors contribute to loneliness among older europeans? ndings from the survey of health, ageing and retirement in europe: Determinants of loneliness. Archives of Gerontology and Geriatrics, 89:104080, 2020. ISSN 0167-4943. doi: https://doi.org/10.1016/j.archger.2020.104080. URL https://www.sciencedirect.com/science/article/pii/S0167494320300741 . 37
A Hofstede et al. (1991) and Hofstede et al.'s (2010) dimensions of culture In addition to the two cultural dimensions described in Section 2, (Hofstede et al., 1991) and Hofstede et al.'s (2010) six-dimensional model of national culture includes the following four cultural dimensions: 1. Uncertainty Avoidance . This dimension refers to a society's tolerance for ambiguity. It is conceptually dierent from risk avoidance. Cultures oriented to uncertainty avoidance are more prone to support stricter rules, laws, and norms aimed at reducing the ambiguity and oering "one-size-tsall" solutions. On the other side, cultures accepting ambiguity prefer fewer rules and more freedom in expressing dierent opinions. According to Hofstede et al. (2010), this dimension is conceptually associated to indulgence. 18 2. Long Term versus Short Term Orientation . This dimension denes the extent to which a society looks toward the future rather than resorting to the past. Short-term oriented societies look to the past experiences to deal with the current challenges, and maintain a rather static and xed mindset. Long-term oriented cultures, on the other side, are more exible, susceptible to change, and ready to cope with uncertain future challenges. Moreover, long term oriented societies value relationships while short term oriented societies focus more on tradition. 3. Power Distance . Power distance measures how people in a society relate to each other on a hierarchical scale. High power distance cultures assign a higher weight to a person or authority, while low power distant societies emphasize the equal treatment of everyone. 4. Masculinity versus Femininity . Masculinity is about the distinction of gender roles. In masculine societies gender roles are clearly distinct. Men are supposed to be assertive, tough, and focused on material success; women are supposed to be more modest, tender, and concerned with the quality of life. Femininity stands for a society in which social gender roles overlap: both men and women are supposed to be modest, tender, and concerned with the quality of life. 19 18 However, the authors did not nd objective ways of measuring and theorizing this association. 19 Hofstede et al. (2010), page 140. 44
Table C.9-C.11 report results using the rst three dimensions of maternal cultural background as an instrument for loneliness. They show that these alternative dimensions of culture do not represent a valid instrument. B Additional gures Figure B.1: Degree of dissatisfaction (%) with personal relationships versus restraint and individualism (a) Belgium Denmark France Germany Hungary Italy Latvia Netherlands Sweden Switzerland Austria Czech Republic Estonia Finland Lithuania Luxembourg Poland Portugal Slovakia Slovenia Spain 3 8 13 18 20 40 60 80 100 Restraint Dissatisfaction (high Ind.) Fitted values Dissatisfaction (interm. and low Ind.) Fitted values (b) Austria Belgium Czech Republic Denmark Estonia Finland France Germany Hungary Italy Latvia Lithuania Luxembourg Netherlands Poland Portugal Slovakia Slovenia Spain Sweden Switzerland 3 8 13 18 20 40 60 80 100 Individualism Dissatisfaction Fitted values Note : The country-level data on satisfaction comes from the 2018 World Gallup survey, available at https://stats.oecd.org/index.aspx?lang=en (Accessed: December 7, 2021). Satisfaction with personal relationships refers to the mean score of survey respondents who rate their satisfaction with their personal relationships on an 11-point scale, from 0 (not at all satised) to 10 (completely satised). The variable refers to the respondent's opinions or feelings Source : 2018 Gallup Survey and Hofstede et al. (2010). 45
Figure B.2: Loneliness across natives and second-generation immigrants, R-UCLA loneliness scale. 0 20 40 60 % 2 4 6 8 10 Loneliness (short version of R-UCLA Loneliness Scale) Full sample (natives) SG immigrants Source : SHARE data. 46
C Additional tables Table C.1: Summary statistics Variable Mean Std. Dev. Min. Max. N Loneliness (R-UCLA) 3.883 1.355 3 9 4579 Feels lonely 0.234 0.423 0 1 4578 Restraint_Mother 63.637 16.905 0 96 4579 Restraint_Father 63.211 17.774 22 96 4395 Individualism_Mother 56.619 15.787 12 91 4286 Individualism_Father 56.969 15.667 20 91 4068 Age 64.303 8.547 50 94 4579 Female 0.584 0.493 0 1 4579 Depression scale EURO-D 2.305 2.126 0 12 4579 Suicidality (part of EURO-D) 0.066 0.248 0 1 4573 Adl 0.165 0.614 0 6 4579 Body mass index (bmi) 27.408 5.078 15.571 73.462 4579 Number of chronic diseases 1.694 1.557 0 11 4579 Self-perceived health (SAH) 3.083 1.078 1 5 4579 High blood cholesterol 0.24 0.427 0 1 4575 High blood pressure or hypertension 0.389 0.487 0 1 4575 Diabetes or high blood sugar 0.133 0.34 0 1 4575 Stomach or duodenal ulcer, peptic ulcer 0.045 0.207 0 1 4575 Chronic lung disease 0.057 0.232 0 1 4575 Stroke 0.029 0.167 0 1 4575 Drugs for: sleep problems 0.076 0.265 0 1 4576 Drugs for: high blood cholesterol 0.226 0.418 0 1 4576 Drugs for: anxiety or depression 0.075 0.263 0 1 4576 Drugs for: joint pain 0.136 0.343 0 1 4576 Drugs for: diabetes 0.119 0.323 0 1 4576 Drugs for: high blood pressure 0.402 0.49 0 1 4576 Low Education 0.237 0.425 0 1 4579 Medium Education 0.461 0.499 0 1 4579 High Education 0.302 0.459 0 1 4579 Retired 0.497 0.5 0 1 4579 Unemployed 0.026 0.16 0 1 4579 Disabled 0.045 0.207 0 1 4579 Employed 0.368 0.482 0 1 4579 Married 0.679 0.467 0 1 4579 Divorced 0.131 0.338 0 1 4579 Widowed 0.112 0.315 0 1 4579 Number of children 2.236 1.363 0 14 4579 Lives alone 0.211 0.408 0 1 4579 Parental education (tertiary, either one or both) 0.161 0.368 0 1 4579 Relationship parents (adverse) 0.549 0.498 0 1 4579 Absent parent 0.116 0.32 0 1 4579 Harm from parents 0.278 0.448 0 1 4527 Harm third parties 0.083 0.276 0 1 4579 Understanding parents (lack of) 0.436 0.496 0 1 4552 Source : SHARE data, (Hofstede et al., 2010) and (Becker et al., 2020). 47
Variable Mean Std. Dev. Min. Max. N Every-day contact 0.279 0.449 0 1 3195 Frequent contact 0.271 0.445 0 1 3195 Fair contact 0.279 0.449 0 1 3195 Rare contact 0.17 0.376 0 1 3195 Ancestral distance (z-score) -0.769 0.816 -3.35 0.683 3609 Health when 15yo 2.314 1.155 1 6 4112 HH net assets (deciles) 6.609 2.58 1 10 4579 HH expenditure (deciles) 4.917 2.935 1 9 4579 48
Table C.2: List of countries included in (Hofstede et al., 2010) and the corresponding index of individualism and restraint Country Individ. Restr. Country Individ. Restr. Albania . 85 Italy 76 70 Algeria . 68 Jamaica 39 . Andorra . 35 Japan 46 58 Argentina 46 38 Jordan . 57 Australia 90 29 Korea, Republic of 18 71 Austria 55 37 Kosovo . 85 Azerbaijan . 78 Kyrgyzstan . 61 Bangladesh 20 80 Latvia 70 87 Belarus . 85 Lithuania 60 84 Belgium 75 43 Luxembourg 60 44 Benin 78 . North Macedonia . 65 Bosnia and Herzegovina . 56 Malaysia 26 43 Brazil 38 41 Mali . 57 Bulgaria 30 84 Malta 59 34 Burkina Faso . 82 Mexico 30 3 Canada 80 32 Moldova, Republic of . 81 Central African Republic 73 . Morocco 46 75 Chile 23 32 Netherlands 80 32 China 20 76 New Zealand 79 25 Colombia 13 17 Nigeria . 16 Costa Rica 15 . Norway 69 45 Croatia 33 67 Pakistan 14 100 Cyprus . 30 Panama 11 . Czech Republic 58 71 Peru 16 54 Denmark 74 30 Philippines 32 58 Dominican Republic . 46 Poland 60 71 Ecuador 8 . Portugal 27 67 Egypt . 96 Puerto Rico . 10 El Salvador 19 11 Romania 30 80 Estonia 60 84 Russian Federation 39 80 Finland 63 43 Rwanda . 63 France 71 52 Saudi Arabia . 48 Georgia . 68 Serbia 25 72 Germany 67 60 Singapore 20 54 Ghana . 28 Slovakia 52 72 Greece 35 50 Slovenia 27 52 Guatemala 6 . South Africa 65 37 Hong Kong 25 83 Spain 51 56 Hungary 80 69 Suriname 47 . Iceland . 33 Sweden 71 22 India 48 74 Switzerland 68 34 Indonesia 14 62 Taiwan 17 51 Iran 41 60 Tanzania, United Republic . 62 Iraq . 83 Thailand 20 55 Ireland 70 35 Trinidad and Tobago 16 20 Israel 54 . Turkey 37 51 49
Country Individ. Restr. Country Individ. Restr. Uganda . 48 Venezuela 12 0 Ukraine . 86 Viet Nam 20 65 United Kingdom 89 31 Zambia . 58 United States of America 91 32 Zimbabwe . 72 Uruguay 36 47 Source : (Hofstede et al., 2010). Table C.3: List of countries included in the analysis SG immigrants Country of interview (27) Austria, Germany, Sweden, Netherlands, Spain, Italy, France, Denmark, Greece, Switzerland, Belgium, Israel, Czech Republic, Poland, Luxembourg, Portugal, Slovenia, Estonia, Croatia, Lithuania, Bulgaria, Cyprus, Finland, Latvia, Malta, Romania, Slovakia. SG immigrants Country of origin Mother (62) Albania, Algeria, Argentina, Australia, Austria, Azerbaijan, Belarus, Belgium, Bosnia and Herzegovina, Brazil, Bulgaria, Canada, Chile, China, Colombia, Croatia, Cyprus, Czech Republic, Denmark, Dominican Republic, Egypt, Estonia, Finland, France, Georgia, Germany, Greece, Hungary, India, Indonesia, Iran, Iraq, Ireland, Italy, Kosovo, Latvia, Lithuania, Luxembourg, Macedonia, Malta, Moldova, Republic of Morocco, Netherlands, Norway, Poland, Portugal, Romania, Russian Federation, Serbia, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland, Turkey, Ukraine, United Kingdom, United States of America, Uruguay, Venezuela, Viet Nam. 50
Table C.4: The eect of maternal restraint on health and unhealthy behaviour. SG immigrants (either one or both parents born abroad). L=0 L=1 L=0 L=1 L=0 L=1 L=0 L=1 L=0 L=1 Variable Eurod Eurod ADL ADL BMI BMI Chronic Chronic SAH SAH Restraint_Mother 0.002 0.015*** 0.002** 0.002 0.021 0.022 0.001 -0.002 0.000 0.001 (0.003) (0.005) (0.001) (0.002) (0.011) (0.014) (0.002) (0.003) (0.002) (0.002) N. Observations 2677 1902 2677 1902 2677 1902 2677 1902 2677 1902 Variable Cholesterol Cholesterol Diabetes Diabetes Pressure Pressure Ulcer Ulcer Stroke Stroke Restraint_Mother -0.001 -0.001 -0.000 0.000 0.001 -0.000 -0.000 0.000 0.001 0.000 (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.000) (0.000) (0.000) (0.000) N. Observations 2674 1901 2674 1901 2674 1901 2674 1901 2674 1901 Variable No vig. FA No vig. FA No FA No FA Smoking Smoking Fruit Fruit Drinking Drinking Restraint_Mother 0.000 0.002** 0.001 0.001 -0.002** -0.000 -0.000 -0.002 -0.001 -0.000 (0.001) (0.001) (0.000) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) (0.001) N. Observations 2676 1900 2674 1899 2677 1902 2677 1902 1324 850 Full set of v. Table 2 Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Country (of residence) FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Year (round) FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Notes : The method of estimation is OLS. L= 0 indicates the subset of individuals with R−UCLA = 0 (not lonely), while L= 1 indicates individuals with R−CLA > 0 (lonely). Robust standard errors are clustered at the country of residence and parental country of origin level. Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . 51
Table C.5: The eect of loneliness on single components of the EURO-D scale. SG immigrants (either one or both parents born abroad). Variable 2ND 2ND 2ND 2ND 2ND 2ND 2ND 2ND 2ND 2ND 2ND Depr. Pess. Suicid Sleep Interest Irrit. Appet. Fatigue Concen. Enjoy Tear Lonely (dich. UCLA) 0.094 0.043 0.050** 0.080 0.093*** -0.144 -0.018 0.047 0.075*** 0.089*** 0.103*** (0.251) (0.030) (0.022) (0.122) (0.030) (0.117) (0.038) (0.100) (0.021) (0.016) (0.028) Age 0.000 0.001** 0.000* 0.001 0.000 0.000 -0.000 0.002** -0.000 0.000* -0.000 (0.001) (0.000) (0.000) (0.001) (0.000) (0.001) (0.000) (0.001) (0.000) (0.000) (0.000) Female 0.061 -0.008 0.008** 0.058** 0.005 0.014 0.023** 0.043*** -0.000 -0.004 0.069*** (0.042) (0.007) (0.003) (0.027) (0.005) (0.015) (0.009) (0.015) (0.004) (0.004) (0.015) Low Education 0.025 0.012* 0.009* 0.018 0.009*** 0.042** 0.018** 0.041*** 0.021*** 0.014*** 0.024** (0.026) (0.007) (0.005) (0.015) (0.003) (0.018) (0.009) (0.015) (0.006) (0.004) (0.010) High Education -0.011 -0.011* -0.008** -0.022 -0.006** -0.017 -0.003 -0.016 -0.008** -0.002 -0.005 (0.014) (0.006) (0.003) (0.015) (0.003) (0.011) (0.005) (0.012) (0.003) (0.002) (0.005) Retired 0.010 -0.009 -0.002 -0.012 0.008 -0.021 -0.003 0.011 0.005 -0.006 0.005 (0.017) (0.010) (0.006) (0.017) (0.007) (0.024) (0.011) (0.016) (0.006) (0.005) (0.009) Unemployed 0.032 0.032* 0.008 -0.021 0.010 -0.024 0.020 0.006 -0.024** -0.003 0.037** (0.035) (0.017) (0.010) (0.028) (0.008) (0.038) (0.016) (0.028) (0.011) (0.008) (0.017) Disabled 0.134 0.030** 0.022*** 0.090* 0.033*** 0.114*** 0.035 0.135*** 0.042*** 0.013** 0.063*** (0.096) (0.014) (0.007) (0.047) (0.009) (0.032) (0.022) (0.046) (0.010) (0.006) (0.018) Employed -0.024 -0.025** -0.014** -0.034 0.000 -0.069*** -0.033** -0.025 -0.019** -0.017*** -0.009 (0.035) (0.012) (0.006) (0.023) (0.006) (0.027) (0.016) (0.019) (0.008) (0.005) (0.008) Married -0.030 -0.007 -0.010 -0.025 -0.003 -0.049 -0.011 -0.032 -0.010 -0.004 -0.010 (0.033) (0.010) (0.006) (0.023) (0.006) (0.039) (0.015) (0.022) (0.009) (0.005) (0.009) Divorced -0.015 0.003 0.003 -0.008 0.001 -0.029 -0.005 -0.027 -0.001 -0.001 -0.006 (0.025) (0.010) (0.005) (0.018) (0.006) (0.038) (0.011) (0.022) (0.008) (0.004) (0.009) Widowed -0.021 0.005 -0.004 -0.017 0.002 -0.030 0.002 -0.017 -0.007 -0.000 0.003 (0.029) (0.009) (0.006) (0.027) (0.006) (0.031) (0.016) (0.024) (0.010) (0.005) (0.011) Number of children -0.006 -0.005** 0.001 -0.002 -0.000 -0.008 -0.000 -0.004 -0.001 -0.002* -0.000 (0.007) (0.002) (0.001) (0.003) (0.001) (0.006) (0.002) (0.004) (0.001) (0.001) (0.002) Live alone 0.058 0.008 0.005 0.032 0.005 0.043 0.018 0.029 0.003 0.007 0.030** (0.058) (0.011) (0.005) (0.024) (0.004) (0.045) (0.013) (0.027) (0.005) (0.005) (0.012) Education parents (tertiary) -0.007 0.002 -0.004 -0.010 0.002 -0.027 -0.004 -0.010 -0.009 -0.004 -0.010 (0.019) (0.006) (0.005) (0.015) (0.006) (0.025) (0.008) (0.017) (0.006) (0.004) (0.007) Relationship parents (adverse) 0.043 0.004 0.004 0.023 0.005** 0.056*** -0.000 0.038*** 0.011*** 0.003 0.011* (0.037) (0.006) (0.003) (0.019) (0.002) (0.015) (0.005) (0.014) (0.003) (0.002) (0.006) Absent parent 0.030 0.016** 0.005 0.011 0.002 0.038** 0.005 0.036** 0.008* 0.005 0.015** (0.023) (0.007) (0.004) (0.011) (0.003) (0.017) (0.007) (0.014) (0.004) (0.005) (0.008) HH net wealth -0.005 -0.005*** -0.002** -0.004* -0.001*** -0.009** -0.004** -0.011*** -0.002* -0.002*** -0.002** (0.005) (0.002) (0.001) (0.003) (0.001) (0.004) (0.002) (0.003) (0.001) (0.001) (0.001) HH expenditure 0.002 0.000 0.001 0.002 0.000 0.002 0.001 0.001 0.001** 0.000 0.000 (0.002) (0.001) (0.000) (0.001) (0.000) (0.002) (0.001) (0.001) (0.001) (0.000) (0.001) Additional controls: Country of residence Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes 1st stage F 10.10 1st stage p-value 0.0058 N. Observations 4576 4567 4573 4579 4575 4571 4579 4576 4570 4574 4572 Notes : The table reports the average marginal eects of bivariate probit estimations. Lonely (dich. UCLA) is the dichotomized version of the R-UCLA scale. Robust standard errors are clustered at the country of residence and parental country of origin level. Abbreviations: 1ST - rst stage estimates; 2ND - second stage estimates. Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . 52
Table C.6: The eect of loneliness on health. Baseline specication. SG immigrants (either one or both parents born abroad). Only refreshers. Variable 1ST 2ND 2ND 2ND 2ND 2ND Loneliness Eurod Adl Bmi Chronic SAH Restraint_Mother 0.006*** (0.001) Loneliness (R-UCLA) 1.866*** 0.469*** 3.238* -0.045 0.558 (0.215) (0.108) (1.849) (0.537) (0.404) Age 0.001 0.001 0.004 -0.042*** 0.038*** 0.017*** (0.003) (0.005) (0.003) (0.013) (0.006) (0.003) Female 0.033 0.627*** -0.006 -0.521 0.021 -0.046 (0.055) (0.076) (0.029) (0.432) (0.099) (0.035) Low Education 0.224*** -0.112 -0.042 -0.077 0.212** 0.087 (0.082) (0.100) (0.047) (0.665) (0.105) (0.088) High Education -0.143*** 0.051 0.044 0.247 -0.107 -0.130 (0.049) (0.164) (0.039) (0.386) (0.130) (0.088) Retired 0.046 0.143 -0.037 0.407 0.097 0.038 (0.109) (0.221) (0.047) (0.446) (0.135) (0.076) Unemployed 0.113 0.305 -0.079 -0.965 -0.129 0.126* (0.126) (0.207) (0.076) (0.835) (0.185) (0.076) Disabled 0.764*** 0.293 0.159 -1.716 1.536*** 0.610* (0.120) (0.231) (0.105) (1.418) (0.478) (0.328) Employed -0.189 0.274 0.020 0.300 -0.217** -0.005 (0.120) (0.248) (0.080) (0.774) (0.103) (0.126) Married -0.333*** 0.552** 0.116 0.285 -0.112 0.082 (0.101) (0.220) (0.089) (1.175) (0.182) (0.170) Divorced -0.217** 0.642*** 0.083 -0.052 0.063 0.105 (0.104) (0.214) (0.096) (0.963) (0.113) (0.122) Widowed -0.177 0.386 0.110 0.494 0.225** 0.136 (0.160) (0.278) (0.090) (1.063) (0.106) (0.131) Number of children -0.047*** 0.064 0.027* 0.293* -0.047* 0.006 (0.017) (0.052) (0.016) (0.153) (0.025) (0.015) Live alone 0.349*** -0.621*** -0.186** -1.776** -0.259 -0.327* (0.118) (0.142) (0.088) (0.750) (0.184) (0.183) Education parents (tertiary) -0.049 0.125 -0.011 -1.012*** -0.047 -0.059 (0.084) (0.116) (0.046) (0.269) (0.049) (0.061) Relationship parents (adverse) 0.089* -0.013 -0.057** -0.399 -0.058 -0.034 (0.050) (0.080) (0.023) (0.322) (0.071) (0.073) Absent parent 0.177* -0.101 -0.124** 0.173 0.121 -0.141* (0.094) (0.199) (0.057) (0.536) (0.128) (0.083) HH net wealth -0.038*** 0.017 0.004 0.059 -0.045 -0.034** (0.012) (0.021) (0.008) (0.082) (0.027) (0.017) HH expenditure 0.013 -0.022 0.000 -0.072 0.021* -0.007 (0.011) (0.014) (0.004) (0.049) (0.011) (0.005) Additional controls: Country of residence Yes Yes Yes Yes Yes Yes Year (SHARE wave) Yes Yes Yes Yes Yes Yes 1st stage F 27.80 1st stage p-value 0.0001 N. Observations 2655 2655 2655 2655 2655 2655 Notes : The table reports the coecients of two-stage linear estimations. Robust standard errors are clustered at the country of residence and parental country of origin level. Abbreviations: 1ST - rst stage estimates; 2ND - second stage estimates; SAH - self-assessed health. Signicance levels: * p < 0.1 , ** p < 0.05 , *** p < 0.01 . 53