Introduction
Although ageing without a partner has recently received considerable attention for its consequences for an individual's health and wellbeing (Djundeva et al., Reference Djundeva, Dykstra and Fokkema2019; Smith and Victor, Reference Smith and Victor2019; Quashie et al., Reference Quashie, Arpino, Antczak and Mair2021), the channels through which the absence of kin might affect an individual's outcomes are still under-researched. The loss of a spouse in older ages, as a result of death, separation or divorce, might have detrimental effects on an individual's health by, for example, affecting daily-life habits, including eating behaviours (Vesnaver et al., Reference Vesnaver, Keller, Sutherland, Maitland and Locher2016). Indeed, the quantity and type of food consumed appears to be strongly influenced by social context (Higgs, Reference Higgs2015) and – although based on small samples and mainly qualitative research – a body of literature examining also young adults has presented convincing evidence that widowhood is associated with altered diets and eating behaviours (Heuberger and Wong, Reference Heuberger and Wong2014; Vesnaver et al., Reference Vesnaver, Keller, Sutherland, Maitland and Locher2016).
The benefits that accrue from being in a couple in older ages have been mainly accounted for in relation to two theoretical concepts: the spousal resources/support model and social control theory. However, little is known about the relationship between stability/changes in marital status and modifications in protein consumption (frequency) in older ages, a behaviour deemed to be a significant risk factor for the development of frailty among older people (Haider et al., Reference Haider, Grabovac, Drgac, Mogg, Oberndorfer and Dorner2020). To the best of our knowledge, few quantitative studies using longitudinal data have been conducted to date to address this relationship. In general, previous studies have focused on widowhood to the exclusion of other partnership statuses (such as divorce/separation or never having been in a couple). Thus, our objective here is to contribute to the existing literature by addressing the following research questions:
RQ1a: Do respondents living stably without a partnerFootnote 1 (i.e. never-married, divorced and widowed) consume proteins less frequently than those living in a couple (in-couple)?
RQ1b: Do individuals whose status shifts from being in-couple to living without a partner (non in-couple) report a greater reduction in the frequency of protein consumption than that reported by both in-couple and long-standing single respondents?
RQ2: Does the subjective evaluation of economic resources (or changes in that evaluation) moderate the association between the status shift from in-couple to non in-couple and the frequency of protein consumption?
Our study contributes to the extant literature by using – as far as we can determine, for the first time in relation to this research question – a large representative longitudinal sample of Europeans aged 50+. More specifically, we use data from Waves 4–8 (W4–W8) of the Survey of Health, Ageing and Retirement in Europe (SHARE) and estimate longitudinal models that are better able to account for possible selection and reverse causality issues than has been the case in related studies published previously.
Background
The importance of protein consumption at older ages
Low protein consumption appears to be a particularly risky behaviour among older adults. Indeed, late-life protein inadequacies can contribute to frailty (Machón et al., Reference Machón, Mateo-Abad, Vrotsou, Zupiria, Güell, Rico and Vergara2018) by increasing the loss of physical strength (Hanach et al., Reference Hanach, McCullough and Avery2019), impacting disability trajectories (Mendonça et al., Reference Mendonça, Granic, Hill, Siervo, Mathers, Kingston and Jagger2019) and elevating the risk of chronic diseases (McLean et al., Reference McLean, Mangano, Hannan, Kiel and Sahni2016). Various studies have recently reported that frail individuals present an insufficient consumption of protein (e.g. Machón et al., Reference Machón, Mateo-Abad, Vrotsou, Zupiria, Güell, Rico and Vergara2018), while those who consume appropriate amounts present a significantly lower risk for presenting exhaustion, weight loss, slowness, weakness and low activity (Haider et al., Reference Haider, Grabovac, Drgac, Mogg, Oberndorfer and Dorner2020). This suggests that increasing protein consumption at older ages should improve muscle health and prevent the development of sarcopenia (Wolfe, Reference Wolfe2012), while it might also help maintain energy balance and, thus, reduce the risk of under-nutrition.
Partnership status and eating behaviours
Numerous studies report definite advantages of living in-couple for several life-related dimensions including social participation (Bolano and Arpino, Reference Bolano and Arpino2020), and better physical health (Grundy and Tomassini, Reference Grundy and Tomassini2010), psychological health (Stokes and Moorman, Reference Stokes and Moorman2018) and wellbeing (Solé-Auró and Cortina, Reference Solé-Auró and Cortina2019), although being in a domestic partnership might be associated with a selection-effect given that individuals in better health present a greater probability of getting married (Braithwaite and Holt-Lunstad, Reference Braithwaite and Holt-Lunstad2017). Marital relationships tend to become even more salient with ageing (Thomas et al., Reference Thomas, Liu and Umberson2017) since other social relationships are often lost as a result of geographic relocation or death (Liu et al., Reference Liu, Waite and Shen2016). Married individuals tend to be less engaged in risky behaviours and are more likely to seek recommended medical treatment (Van Jaarsveld et al., Reference Van Jaarsveld, Miles, Edwards and Wardle2006). In the specific case of eating behaviours, living in a couple seems to be particularly relevant in establishing correct conduct since food behaviour is the result of a complex negotiation about what and how much we eat (Vesnaver et al., Reference Vesnaver, Keller, Sutherland, Maitland and Locher2015). Moreover, it has been found that individuals in a couple adhere to a healthy diet (Meltzer et al., Reference Meltzer, Novak, McNulty, Butler and Karney2013).
As discussed above, two main theories have been formulated in accounting for the health-related benefits of being in-couple. According to the first of these, the spousal resource/support model, being in-couple is associated with a greater availability of social, emotional, instrumental and economic support (Umberson et al., Reference Umberson, Thomeer and Williams2013) and healthier eating behaviours (Choi et al., Reference Choi, Ailshire and Crimmins2020). Thus, a partner can be a source of social engagement by motivating individuals to cook and eat healthily (Choi et al., Reference Choi, Ailshire and Crimmins2020), while a partner can also provide instrumental and economic support, including help with shopping and cooking (Sidenvall et al., Reference Sidenvall, Nydahl and Fjellstrom2001; Salehi et al., Reference Salehi, Eftekhar, Mohammad, Tavafian, Jazayery and Montazeri2010; Conklin et al., Reference Conklin, Forouhi, Surtees, Khaw, Wareham and Monsivais2014). Consistent with this theory, research has shown that being widowed is associated with a lower frequency and variety of food consumed (Conklin et al., Reference Conklin, Forouhi, Surtees, Khaw, Wareham and Monsivais2014). The second theory, that of social control, argues that intimate family relationships can better provide control of healthy behaviours by indirectly affecting the internalisation of norms for such behaviours, and by directly providing informal sanctions for deviating from healthy behaviours (Umberson et al., Reference Umberson, Pudrovska and Reczek2010b). Thus, having a partner may lead to a significantly better regulation of eating behaviours, as a result of spousal control (Umberson et al., Reference Umberson, Crosnoe and Reczek2010a). This means that losing a partner might lead to a loss of social control, since the partner constitutes an important source of supervision and care ensuring the adoption of heathy behaviours (Umberson et al., Reference Umberson, Pudrovska and Reczek2010b, Reference Umberson, Thomeer, Reczek and Donnelly2016).
Finally, empirical research has also shown that depression and a lack of motivation to eat might explain a deterioration in the eating habits of older individuals not in a partnership. Indeed, standard mealtimes may increase feelings of loneliness, given the painful memories of this activity previously shared with a former partner, and result in a change in appetite due to a negative emotional status and depression (Johnson, Reference Johnson2002). The widowhood effect in terms of increased depressive symptoms is widely reported in the literature (Peña-Longobardo et al., Reference Peña-Longobardo, Rodríguez-Sánchez and Oliva-Moreno2021; Yu et al., Reference Yu, Kahana, Kahana and Han2021), with no difference by gender and region (Schmitz, Reference Schmitz2021). Indeed, drawing on longitudinal data from different European regions, Schmitz (Reference Schmitz2021) found no empirical support for the hypothesised stronger effects of widowhood on the mental health of women living in Southern and Western Europe.
Gender differences
While women tend to receive more financial benefits from living in-couple, men tend to receive greater support for, and more regulation of, their health behaviours that, directly or indirectly, improve their health (Revenson et al., Reference Revenson, Griva, Luszczynska, Morrison, Panagopoulou, Vilchinsky and Hagedoorn2016). However, studies conducted to date do not consistently provide evidence as to whether being in-couple has a differential impact on the variety and quantity of food consumed by men and women. However, men and women appear to have very different habits as regards protein consumption, with men typically reporting a higher daily protein intake (see e.g. Schütz and Franzese, Reference Schütz and Franzese2018). In general, though, a reduction in protein consumption is likely with ageing, with recent evidence showing that more than a third of older men and 41 per cent of older women do not consume enough protein (Baum et al., Reference Baum, KIM and Wolfe2016; Wolfe, Reference Wolfe2012).
Yet, older women are more likely to be economically vulnerable and to report greater financial hardships than men (Lyon and Colquhoun, Reference Lyon and Colquhoun1999; Denton et al., Reference Denton, Prus and Walters2004), affecting their ability to buy food (Sidenvall et al., Reference Sidenvall, Nydahl and Fjellstrom2001), especially as regards meat products, an outcome that has been shown to be influenced by income availability (see e.g. Maguire and Monsivais, Reference Maguire and Monsivais2015). Additionally, older men living without a partner may be at a higher risk than women living without a partner of having an inadequate diet (Davis et al., Reference Davis, Murphy, Neuhaus and Lein1990) as the former are more likely to be dependent on others for food-related decision-making and largely unprepared for food cooking, particularly those belonging to older cohorts (Moss et al., Reference Moss, Moss, Kilbride and Rubinstein2007; Vesnaver et al., Reference Vesnaver, Keller, Sutherland, Maitland and Locher2015). In addition, studies have shown that men are generally less motivated to cook when living alone as they might be less skilled than women (Bennett et al., Reference Bennett, Hughes and Smith2003).
Finally, although previous studies on fruit and vegetable consumption have shown that men fare worse than women as regards the negative associations between non-partnered individuals and healthy eating behaviours (Conklin et al., Reference Conklin, Forouhi, Surtees, Khaw, Wareham and Monsivais2014; Choi et al., Reference Choi, Ailshire and Crimmins2020), we expect a greater influence of changes in their partnership status on the frequency of protein consumption of women.
Methods
Study population
The data were drawn from SHARE, a multi-disciplinary longitudinal survey, representative of the non-institutionalised population aged 50 and above (Börsch-Supan et al., Reference Börsch-Supan, Hank and Jürges2005).
The analytic longitudinal sample is composed of data from W4 (2011), W5 (2013), W6 (2015), W7 (2017) and W8 (2019) waves of SHARE,Footnote 2 including 19 European countries that participated in at least two waves.Footnote 3 We use information from W4–W7 to measure independent variables (baseline wave) and information from W5–W8 (follow-up wave) as our outcome variables.Footnote 4 We excluded the first two waves because they did not collect information about eating behaviours and the third wave because it comprises largely retrospective information about the respondents. In the model devised here, we control for the difference between the baseline and follow-up waves in months.
We restricted the analytic sample to women and men aged 50 and over.Footnote 5
Measures
Dependent variable
The frequency of protein consumption by European older adults was measured using three questions as to how often (i.e. every day; 3–6 times a week; twice a week; once a week; less than once a week) respondents usually eat: (a) meat, fish or chicken; (b) eggs and legumes; and (c) dairy products. Although alternative operationalisations were considered (see the Robustness Checks section), these measures were equalised to their corresponding weekly frequency (i.e. 7, 5, 2, 1 and 0), standardised and used to compute a principal component analysis to obtain a continuous index. More specifically, we considered the first component extracted, which was the only one with an eigenvalue greater than 1.
Explanatory variable
The main independent variable is a measure identifying changes or stability in partnership status. We used information on marital status collected at the baseline (W4 or W5 or W6 or W7) and at the follow-up (W5 or W6 or W7 or W8) to assess any changes over time. We coded the partnership variable as follows: remaining in-couple (married or co-habitant at the baseline and at the follow-up); remaining never married; from divorced to divorced (this includes separated); from widowed to widowed; from in-couple to divorced (married or co-habitant at the baseline and divorced/separated at the follow-up); from in-couple to widowed (married or co-habitant at the baseline and widowed at the follow-up). Respondents moving from never married/divorced/separated or widowed to in-couple were excluded as they represented about 0.44 per cent (women) and 0.53 per cent (men) of the initial sample.
Control variables
We controlled for a set of socio-demographic variables including a measure of the respondents' employment status (being in paid work; retired; other) – where ‘other’ includes being unemployed, permanently sick or disabled, homemaker or other; a measure of the highest educational qualification attained re-coded into three categories according to the International Standard Classification of Education (where low educational level is defined as below secondary education; high educational level refers to a university education or above); and a measure assessing whether respondents live in a rural or urban area (Alcañiz et al., Reference Alcañiz, Riera-Prunera and Solé-Auró2020). We also included age and age squared as continuous variables to account for the non-linear relationship between age and changes in eating behaviours and continuous variables for the number of children and grandchildren. As a robustness check (see Table S1 in the online supplementary material), we incorporated a further six measures of health and health behaviours that the literature suggests are likely to be associated with the outcome (Gregório et al., Reference Gregório, Rodrigues, Eusébio, Sousa, Dias, André, Grønning, Coelho, Mendes, Graça, Espnes, Branco and Canhão2017). Specifically, we included in the model: smoking cigarettes (being a current smoker versus being a former/never-smoker); physical activity (being vigorously active versus not being vigorously active); self-reports of doctor-diagnosed long-standing illness such as heart disease, hypertension, stroke or cancer (recoded as having no conditions; one condition; two or more conditions), a self-report measure of perceived general health (good or excellent versus fair, poor or very poor); Euro-D scale for depression (counting the number of self-reported depressive symptoms out of a maximum of 12); and a measure of obesity recoded into four categories based on the Body Mass Index (below 18.5 – underweight; 18.5–24.9 – normal; 25–29.9 – overweight; 30 and above – obese). The time distance between waves was controlled in months.
Moderator variable
A perceived measure of being financially vulnerable was used to address the economic status of the respondents. Specifically, they were asked whether the household is able to make ends meet (fairly easily or easily versus with some difficulty or great difficulty). As we sought to determine whether the respondent's ability to make ends meet had changed over time, the variable was recoded as follows: from easily to easily; from with difficulty to easily; from with difficulty to with difficulty; from easily to with difficulty.
Statistical analyses
Descriptive analyses were conducted to explore the distribution of the analytic sample's main characteristics. We used conditional change multiple regression models to examine associations between stability/changes in partnership status and changes in the frequency of protein consumption at the follow-up, controlling for protein consumption at the baseline as well as for the baseline covariates described above. Indeed, in this way (see also Di Gessa et al., Reference Di Gessa, Bordone and Arpino2020), the regression coefficients indicate how the explanatory variable is associated with changes in the frequency with which respondents consume proteins at the follow-up, given that the respondents' eating behaviour at the baseline is controlled for (Twisk, Reference Twisk2007).
We carried out two separate analyses to investigate our research questions specifically as outlined above. The first (Model 1) examines the relationship between experiencing a change in partnership status and a change in the frequency of protein consumption to test whether compared with in-couple respondents: (a) non in-couple respondents and those who did not change their partnership status report a decrease in the frequency with which they consume proteins (RQ1a; Model 1); and (b) those moving from being in-couple to being non in-couple report an even greater decrease in the amount of protein regularly consumed (RQ1b; Model 1). The second analysis repeats Model 1 but focuses specifically on the consumption of meat, chicken and fish, given that they are more likely to be affected by economic resources (RQ2; Model 2). Model 2 introduces an interaction term capturing whether or not respondents report experiencing a change in the difficulty of making ends meet in different scenarios.
The analyses were conducted separately for men and women to identify any gender differences.
Results
Descriptive statistics
Table 1 shows the summary statistics for the main variables. A slight difference is observed in the types of protein regularly consumed by gender, with male consumption being more frequent, especially of meat, chicken and fish. In the case of the main independent variable, our summary statistics show that the prevalence of respondents living in a stable couple across waves was about 79 per cent for men and 60 per cent for women. About 5–6 per cent of both men and women reported never having been married, whereas there was a slightly higher prevalence of women reporting themselves as being long-standing divorced. A substantial difference is observed in the prevalence of long-standing widowhood, with women being almost four times more likely to record this status than men. As for changes in partnership status, our descriptive statistics show a small percentage of both men and women getting divorced between waves and a slightly higher percentage of women (about 3.5%) reporting becoming widows (versus about 1.7% of men).
Summary statistics for the outcome and explanatory variables by gender
Notes: N = 134,313. Missing moderator values were imputed using the multiple imputation technique. Dependent variables were estimated at the baseline. Independent and moderator variables were constructed as changing variables between the baseline and follow-up.
Source: Survey of Health, Ageing and Retirement in Europe (SHARE) 2011 (Wave 4), 2013 (Wave 5), 2015 (Wave 6), 2017 (Wave 7) and 2019 (Wave 8).
To address RQ1, Table 2 shows the results from the linear regression models investigating the association between stability/changes in partnership status and changes in the frequency of proteins regularly consumed controlling for socio-economic and demographic characteristics.
Ordinary least squares regression coefficients for the conditional change model of eating proteins at follow-up
Notes: N = 134,313. SE: robust standard errors. Ref.: reference category. The model controls for: age, age2, number of children, number of grandchildren, rural versus urban area of residence, country dummy variables, education, working status and financial vulnerability. Authors' own calculations. Results stratified by gender. Controls are measured at the baseline (t − 1). Standardised outcome variable.
Source: Survey of Health, Ageing and Retirement in Europe (SHARE) 2011 (Wave 4), 2013 (Wave 5), 2015 (Wave 6), 2017 (Wave 7) and 2019 (Wave 8).
Significance levels: ** p < 0.05, *** p < 0.01.
For men and women alike, we detected a more marked reduction in the frequency of protein consumption in individuals living without a partner and whose partnership status did not change, on the one hand, than in individuals in-couple and whose partnership status did not change, on the other. Indeed, males and females whose partnership status did not change from never married, divorced and widowed at the outset present a significant reduction in the frequency of their protein consumption. This outcome being slightly higher in magnitude for the long-standing divorced. More specifically, we observe that, on average, remaining divorced is associated with a reduction in regular protein consumption in both women and men, while the reduction in those who never married was relatively higher among women. Among those who remained widowed, this reduction in frequency of protein consumption was similar, with a relatively stronger effect among women (β = −0.082; p < 0.01) than on men (β = −0.065; p < 0.01). Turning to examine the impact of changes in partnership status, only women who got divorced between the observed time-points were more likely to reduce the frequency of their protein consumption. Finally, the transition to widowhood was associated with a decrease in the frequency of protein consumption among both women and men (p < 0.01). These results are very stable when health-related covariates are included (see Table S1 in the online supplementary material), indicating that the effect is robust to the inclusion of potential mediators.
To address RQ2, Table 3 reports the conditional change in eating meat-related protein for women and men with the inclusion of an interaction term between changes in partnership status and changes in subjective financial conditions.
Ordinary least squares regression coefficients for the conditional change model of eating meat-related proteins at follow-up with an interaction term (financial vulnerability)
Notes: N = 134,313. SE: robust standard errors. Ref.: reference category. The model controls for: age, age2, number of children, number of grandchildren, rural versus urban area of residence, country dummy variables, education, working status and financial vulnerability. Authors' own calculations. Results stratified by gender. Controls are measured at the baseline (t − 1). Standardised outcome variable.
Source: Survey of Health, Ageing and Retirement in Europe (SHARE) 2011 (Wave 4), 2013 (Wave 5), 2015 (Wave 6), 2017 (Wave 7) and 2019 (Wave 8).
Significance levels: * p < 0.1, ** p < 0.05, *** p < 0.01.
Some of the interactions are statistically significant but none is associated with a change in partnership status (p > 0.05). This means that even if the experiencing of a worsening in own economic situation was negatively associated with the frequency of protein consumption, it did not significantly moderate the relationship between moving to a non in-couple status and the outcome.
Robustness checks
A number of checks were conducted to test the robustness of our main findings in Model 1. In addition to incorporating a set of covariates related to health conditions at the baseline (see Table S1 in the online supplementary material), we repeated the analyses using binary measures of each type of protein. The results were broadly similar to our main findings (see Tables S1 and S2 in the online supplementary material). We then replicated the models by excluding the outcome at the baseline so as to check for the possible underestimation of the main explanatory variable effects due to differences in the initial frequency of protein consumption. These findings (available upon request) point to an even stronger effect for the long-standing never married, divorced and widowed, suggesting a partial wash-out effect as the conditions of status stability had an impact on the outcome at baseline.
Discussion
Malnutrition is a significant public health problem in European countries, affecting older individuals above all (Leij-Halfwerk et al., Reference Leij-Halfwerk, Verwijs, van Houdt, Borkent, Guaitoli, Pelgrim, Heymans, Power, Visser, Corish and de van der Schueren2019). However, the channels via which an individual's health and wellbeing are affected by ageing, especially with or without a partner, are still under-researched. Moreover, undergoing a change in partnership status appears to have significant consequences for an individual's daily-life habits, including a reduction in the frequency with which older people consume the proteins that are fundamental to protect them from such risks as sarcopenia and disability.
Our findings have shown that individuals ageing without a partner report a significantly greater reduction in the frequency with which they consume proteins compared to those living in-couple, regardless of gender. This result, which provides a response to RQ1a, suggests that both older men and women living alone run a higher risk of having a diet poor in protein.
Changes in partnership status in older ages significantly reduced the frequency of protein consumption only in the case of individuals who had been widowed (RQ1b). Spousal loss is one of the most traumatic events an individual can experience and, since mealtime may serve as a reminder of shared activities with the spouse, widows are likely to report poorer appetite, meal skipping, and poorer diet quality and variety (Johnson, Reference Johnson2002). Here, we found that women who were long-standing widows consistently presented a reduction in their frequency of protein consumption. In line with the main prevailing theories, one of the advantages of being in-couple is the supposed greater availability of economic resources. Our results provide some support for this idea, as is suggested by the negative association between increasing subjective economic vulnerability and overall protein consumption among never married women and among long-standing divorced men. However, no significant effects have been found in relation to the role of financial vulnerability among those who underwent a change in their partnership status (RQ2). The effect of widowhood on the frequency of protein consumption among women could be due to changes in food-related roles and habits. Some qualitative studies have shown, mostly in relation to women, that cooking shared meals in a marriage may be a marital obligation and that this commitment to sharing food also means having to negotiate eating patterns (Vesnaver et al., Reference Vesnaver, Keller, Sutherland, Maitland and Locher2016). Thus, reducing the frequency of protein consumption may also be a result of the newly acquired opportunity to eat according to one's own personal preferences (Quandt et al., Reference Quandt, McDonald, Arcury, Bell and Vitolins2000). While we control for depressive symptoms, a lack of interest in eating or the absence of motivation might contribute to explaining the deterioration in the eating behaviour of widowed women. In the case of men, eating patterns are likely to be directly related to the loss of supervision or care leading to the adoption of unhealthy behaviours (Umberson et al., Reference Umberson, Pudrovska and Reczek2010b, Reference Umberson, Thomeer, Reczek and Donnelly2016).
Strengths and limitations
This study seeks to contribute to the literature by providing quantitative evidence based on a large representative longitudinal sample of European individuals aged 50+. Indeed, extant evidence mainly derives from qualitative research or quantitative studies conducted with small samples. More specifically, our study contributes by, first, extending the group of individuals ageing without a partner to the divorced and never married using longitudinal European data; and, second, by testing a possible mechanism that might contribute to accounting for the reduction in protein consumption.
Various limitations should be recognised. First, the measurement of the frequency of protein consumption relies on a self-report measure and does not provide a precise amount of grams consumed per day. Second, although a robustness check has been undertaken, the high prevalence of missing data for the variable capturing the date of divorce/widowhood (15%) meant we were unable to make additional differentiations between the long-standing divorced and widowed. Indeed, the SHARE data do not allow us to differentiate clearly the effects of widowhood and/or divorce from the effects of the circumstances surrounding partner loss, such as the marital conflicts or health problems that may precede the loss of a partner. Further studies need to address these specific aspects. Third, in this study we have only been able to consider average frequency of protein consumption but not whether or how frequently the respondent eats with other persons (such as family members, friends or care-giver).
This study should serve to motivate further research that can help shed light on the specific role of partnership status and changes in that status on eating behaviours, in general, and on protein intake, in particular, among older people. Indeed, low-frequency protein consumption is a key modifiable risk factor for multiple chronic diseases and, as such, many national and international policies recognise the importance of ensuring individuals obtain an adequate intake of protein so as to reduce inequalities in the frailty of older individuals. In short, polices need to allocate the resources required to meet the changing social and contextual needs of older people.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0144686X22001040.
Data
This paper uses data from SHARE Waves 4, 5, 6, 7 and 8 (DOIs: 10.6103/SHARE.w4.800, 10.6103/SHARE.w5.800, 10.6103/SHARE.w6.800, 10.6103/SHARE.w7.800, 10.6103/SHARE.w8.800), see Börsch-Supan et al. (Reference Börsch-Supan, Brandt, Hunkler, Kneip, Korbmacher, Malter, Schaan, Stuck and Zuber2013) for methodological details. The SHARE data collection has been funded by the European Commission through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-2006-028812), FP7 (SHARE-PREP: GA No. 211909, SHARE-LEAP: GA No. 227822, SHARE M4: GA No. 261982), Horizon 2020 (SHARE-DEV3: GA No. 676536, SERISS: GA No. 654221), and by DG Employment, Social Affairs and Inclusion. Additional funding from the German Ministry of Education and Research, the Max Planck Society for the Advancement of Science, the US National Institute on Aging (U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815, R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04-064, HHSN271201300071C) and from various national funding sources is gratefully acknowledged (see www.share-project.org).
Conflict of interest
The authors declare no conflicts of interest.
Ethical standards
Ethical approval is not required.
