Introduction
The older adult population in the United States of America (USA) is growing fast, with roughly 10,000 adults turning 65 every day since 2011 (Pew Research Center, 2010). These older adults live longer with better health than their preceding cohorts, resulting in a greater reservoir of older adults with talents and experiences. In the face of this ageing trend and related issues such as social security, long-term care and pensions, policy makers and scholars have been concerned with ways to promote health and well-being in those extended years. The productive ageing model was first developed as part of a collaborative effort to extend productive life for older adults and to counter stereotypes of older adults for being dependent on the productivity of younger generations (Butler, Reference Butler1989; Bass and Caro, Reference Bass, Caro and Bass1995). Robert Butler (Reference Butler1989), who first coined the term ‘productive ageing’, contended that a society should transform current social structures to create opportunities for older adults to stay active and healthy.
At the heart of the productive ageing model is the notion that all ageing experiences are embedded within various social contexts. Indeed, research focuses on factors that encourage or hinder later-life productive activities, showing that certain types of social resources or capital are more likely to influence productive activities than others (Musick and Wilson, Reference Musick and Wilson2007; McNamara and Gonzales, Reference McNamara and Gonzales2011). However, what is largely unexplored in the literature but has significant implications in the lives of older adults is predictors of multiple productive activities over time. Cross-sectional data lose information on whether resources promote long-term maintenance in productive activities. Since older adults participate in various activities to varying degrees over time, the question remains whether the resources that affect the initial participation level also influence the change over time. Studying a longer vector of activities also may have significant implications for interventions regarding the recruitment and retention of older adults in various organisations.
Employing the productive ageing model and the concept of capital as the main conceptual underpinnings, the current study investigates the effects of human, social and cultural capital on productive activities over time. Multiple forms of productive activity are considered prevalent among older adults: namely volunteering, employment and care-giving. The main research questions are twofold: (a) are human, social and cultural capital associated with the baseline-level participation in productive activities and (b) do these capitals predict the maintenance or initiation into these activities?
Productive activity: definition and prevalence
Though multiple scholars have offered conceptual and operational definitions of productive activity, there appears to be a lack of a unifying definition in the literature. A very narrow or a very broad definition can be equally problematic. On the one hand, limiting productive activity to strictly monetary criteria may lead to definitions that overlook several productive albeit rarely monetised activities. On the other hand, an extremely broad definition which includes all activities loses some of its utility as a concept for empirical research. Thus, adopting a widely used definition by Bass et al. (Reference Bass, Caro and Chen1993), productive activity is defined as any activity (either paid or unpaid) that makes a constructive contribution to community life. Housework and do-it-yourself activities are excluded because, contrary to the definition, these activities are typically accomplished in isolation and often benefit solely the actor.
Three specific activities are examined in this paper: (a) volunteering, (b) paid work and (c) care-giving. Volunteering and paid work are most commonly recognised as productive activities in multiple studies (Hinterlong, Reference Hinterlong2008; Morrow-Howell et al., Reference Morrow-Howell, Hinterlong and Sherraden2001; Musick and Wilson, Reference Musick and Wilson2003). The US Bureau of Labor Statistics (2013) estimates that about one in four Americans aged 65 and older is engaged in volunteer work. In addition, older volunteers devoted more time than their younger counterparts. For the same age group, around 17 per cent were employed in 2013, much higher than the European Union employment rate of 5 per cent (US Bureau of Labor Statistics, 2013). By 2020, older adults are expected to make up about 6 per cent of the total labour force compared to 3.6 per cent in 2006. Care-giving is also widely considered a significant form of productive activity; the age of care-givers is on the rise, and the contributions of older care-givers generate benefits for the nation and those served (AARP, 2012). Contrary to the popular depiction of older adults as mostly care recipients, AARP (2012) estimates that about 13 per cent of all care-givers are 65 years and older and are providing care to their spouse, parents, friends and relatives.
In sum, despite a description of older age as notably ‘roleless’ (Burgess, Reference Burgess1960), data show that the majority of older adults make both market and non-market contributions that generate substantial benefits for the nation and community. When each activity is examined separately, volunteering is more prevalent while care-giving and employment are less common types of activities in later life.
Antecedents of productive activity: the productive ageing model and capital
Why do some people engage in productive activities while others do not? Though Butler (Reference Butler1989) initially emphasised the role of social environments on productive ageing, Bass and Caro (Reference Bass, Caro and Bass1995) provided the conceptual model, focusing on how environmental, situational, individual and social policy factors are related to productivity in later life. Sherraden et al. (Reference Sherraden, Morrow-Howell, Hinterlong, Rozario, Morrow-Howell, Hinterlong and Sherraden2001) then extended this model to explicate that demographic factors (e.g. gender, race) influence individual capacity while public policies (e.g. programmes, regulations, taxation) influence institutional capacity for productive activity. The productive ageing model notes that individuals participate in productive activity based on an evaluation of their individual and environmental capabilities and options.
While this model provides the conceptual background on how older adults’ capacity leads to productive activity, the current paper situates activities within the concept of capital, demonstrating the role of human, social and cultural capital on a behavioural outcome that is productive activity. Capital refers to resources that have potentially tangible values and can be used to produce other forms of returns (Bourdieu, Reference Bourdieu and Richardson1986). The concept of capital sheds light on several critical questions regarding the productive ageing model. For example, according to the productive ageing model, health and social support are both considered components of individual capacity, but health is a resource largely embedded within an individual while social support emanates from one's relationships and networks. Studying capital elucidates whether individuals participate in productive activity mainly because of their attained status, social relationships, their values in helping others or a combination of all of these factors. Policy makers and planners can also consider ways in which intervention programmes can enhance different forms of capital to encourage older adults who desire to stay productive in later life, such as providing learning opportunities throughout the lifecourse, fostering new and existing social relationships, and creating inclusive organisational environments for individuals with fewer resources (Raymond et al., Reference Raymond, Grenier and Hanley2014).
Human capital can be broadly defined as resources embedded within the individuals, such as knowledge, education, health and functional ability which are valued in paid and unpaid labour markets (Becker et al., Reference Becker, Murphy and Tamura1990). Distinct from human capital, social capital exists when social relationships among individuals or organisations have the potential for benefiting the members (Coleman, Reference Coleman1988). Though there are varying definitions, scholars distinguish at least two distinct types of social capital: bonding capital includes intimate social ties including marital and family relationships, and bridging capital includes looser ties among friends and neighbours (Putnam, Reference Putnam2000; Zhang et al., Reference Zhang, Anderson and Zhan2011; Gonzales and Nowell, Reference Gonzales and Nowell2017). Structural traits for bonding and bridging ties offer different merits. Research has shown that strong normative obligations and long histories for bonding ties are ideal for providing support (Furstenberg, Reference Furstenberg2005; Keating and Dosman, Reference Keating and Dosman2009), while more heterogeneous bridging ties may operate as resources to learn about volunteer and employment opportunities (Musick and Wilson, Reference Musick and Wilson2007). Thus, the current project uses two distinctive sub-types of social capital with varying resources that may promote or hinder individuals who engage in productive activity.
Finally, cultural capital indicates shared symbolic meanings, values and ways of relating to others that have economic or social ‘profits’ (Portes, Reference Portes2000). In society, cultural values and behavioural expectations are exemplified by the role of religion in people's lives. Religious organisations often make conscious efforts to emphasise the connection between faith and one's active role in society, and this may profoundly influence parishioners’ orientation towards work (Davidson and Cadell, Reference Davidson and Cadell1994), care-giving (Stuckey, Reference Stuckey2001) and civic engagement (Musick and Wilson, Reference Musick and Wilson2007).
Empirical studies indicate that individuals equipped with human, social and cultural capital are generally more likely to partake in productive activities in multiple cultures (Kim et al., Reference Kim, Kang, Lee and Lee2007; Musick and Wilson, Reference Musick and Wilson2007; McNamara and Gonzales, Reference McNamara and Gonzales2011). Volunteers are relatively healthy, well-educated, more connected and religious than their non-participating counterparts (Bass and Caro, Reference Bass, Caro and Bass1995; Burr et al., Reference Burr, Caro and Moorhead2002; Choi, Reference Choi2003). Employment is highly correlated with human and social capital such as education, health, network size and marital status, which may indicate that there are more barriers to entry into the labour market for older adults, particularly when they lack human capital (Schmitz, Reference Schmitz2011). Though less attention has been paid to the role of cultural capital on employment, Weber's thesis that considers work as a ‘calling’ has been investigated in the literature (Davidson and Cadell, Reference Davidson and Cadell1994), showing that frequent religious service attenders were more inclined to view their work as a calling rather than a simple job to make ends meet.
Finally, older care-givers are more likely to be female, non-White, healthier, unemployed and have a larger social network compared to non-care-givers. Recognising the significance of health and physical functioning in the care-giving activity, the healthy care-giver hypothesis has been proposed, suggesting that healthier people might be selected into care-giving (Bertrand et al., Reference Bertrand, Saczynski, Mezzacappa, Hulse, Ensrud and Fredman2012; Fredman et al., Reference Fredman, Lyon, Cauley, Hochberg and Applebaum2015). Religious beliefs are also found to affect directly the motivation to care for frail family members or relatives (Stuckey, Reference Stuckey2001), a finding that signifies the role of cultural capital on the norms and ideas to provide help for those in need appropriately.
While considerable scholarly attention has been paid to social antecedents of productive activity, there is still inconsistency in the productive ageing literature on which form of capital predicts which productive activity. Such discrepancy may stem from the fact that most studies exclusively focus on one activity (e.g. volunteering) or a single time-point. The current project uses two-wave panel data in order to examine predictors of baseline participation as well as the changes of productive activity over time, taking into account the baseline participation level and non-random attrition effects. Further, given that older adults partake in multiple activities simultaneously (Morrow-Howell et al., Reference Morrow-Howell, Putnam, Lee, Greenfield, Inoue and Chen2014), the models adjust for other types of productive activities in order to isolate the effects of human, social and cultural capital effectively. Drawing on the conceptual and empirical research on productive activity discussed so far, the current study examines whether human, social and cultural capital predict multiple types of productive activity, namely volunteering, employment and care-giving.
Methods
The study uses two waves of data, hereafter Wave 1 (W1) and Wave 2 (W2), from the National Social Life, Health, and Aging Project (NSHAP), a representative, population-based sampling of older adults in the USA. NSHAP W1 was collected in 2005–2006 and comprised 3,005 respondents with a response rate of 75.5 per cent. The productive activity items except for employment were included in the leave-behind questionnaire (LBQ). The response rate for the LBQ was 84 per cent (N = 2,524) at W1. At W2, collected in 2010–2011, 89.4 per cent of the W1 respondents were re-interviewed and among those who completed the in-person interview, 89 per cent returned the LBQ. Statistical analysis for each productive activity has a slightly different sample size due to missing data.
Measures
Dependent variables: Productive activity
Three productive activities were examined: volunteering, employment and care-giving. A binary variable for each productive role is created to examine whether respondents participated or not at each wave. For volunteering, respondents were asked how often they had volunteered for religious, political, charitable, health-related or other organisations in the past 12 months. Responses ranged from never (0) to several times a week (6). For employment, respondents reported the number of hours they typically work during a week. Those who work 40 hours or more a week were defined as full-time workers, resulting in three categories: non-workers (0), part-time workers (1) and full-time workers (2). For care-giving, respondents were asked whether they are currently assisting an adult who needs help with day-to-day activities because of disability. If they answered yes, they were then queried how many days per week they typically spend caring for this person. Original responses ranged from zero to seven days with a bimodal distribution (reporting either providing no care or full-time care). Thus, three ordinal categories were created for non- (0), part-time (1) and full-time care-giver (2).
Independent variables
Human capital. Education, self-rated physical and mental health, and functional limitation were measured for human capital. For education, four categories of educational attainment (less than high school (0), high school (1), some college (2) and college or more (3)) were included. Self-rated physical and mental health was measured with the question ‘would you say your physical/mental health is excellent (4), very good (3), good (2), fair (1) or poor (0)?’ Functional limitations were indicated by reported difficulty with six items for activities of daily living (ADL). The ADL items included difficulty walking across a room, dressing, eating, bathing or showering, getting in and out of bed, and using the toilet. For each task, respondents were asked to indicate whether they had no difficulty (0), some difficulty (1), much difficulty (2) or were unable to do (3). The items were then summed for a range of 0–18, where higher scores indicate greater functional limitations. The supplementary analyses did not indicate a substantial difference between the original coding and alternate coding schemes (e.g. sum of dichotomous indicators).
Social capital. For bonding social capital, marital status, network size and social support were measured. For bridging capital, neighbourly socialisation was measured. Married was dichotomised with 1 indicating married or co-habiting with a partner. For close network size, respondents were asked to report the number of individuals with whom they can discuss matters that are important to them (the core discussion network). The answers ranged from none (0) to six or more (6). For perceived social support from family, respondents were asked ‘how often can you open up to members of your family if you need to talk about your worries?’ and ‘how often can you rely on them for help if you have a problem?’ The answer ranged from hardly ever or never (0) to often (2). The same items were repeated for spouse and friends, resulting in a total of six possible items of social support. Since some respondents were not asked the entire set of six questions based on their reports on number of friends (if reported having no friend) or marital status (if not married), the items were averaged by the number of valid answers to create an overall social support index ranging from 0 to 2, where higher numbers indicate greater social support. Neighbourly socialisation is measured by one item asking respondents ‘how often do you get together with any of your nearby neighbours just to chat or for a social visit?’, with answer categories from hardly ever (0) to daily or almost every day (4).
Cultural capital. Religious service attendance was measured for cultural capital that is conducive to productive activities. A measure for religious service attendance comes from an item that asked how often respondents attended religious services within the last 12 months. Reponses were rated on an ordinal-level scale ranging from never (0) to several times a week (6).
Additional covariates. In addition to central variables of interest, several covariates were included because of their association with productive activity and with multiple forms of capital in the previous literature (McNamara and Gonzales, Reference McNamara and Gonzales2011; Morrow-Howell, Reference Morrow-Howell2010). Age is coded in years, and sex is dichotomised with 1 indicating female. Race was divided into a series of binary variables (White, Black and other race), with non-Hispanic White serving as the reference group. Due to small sample sizes (6.42% of all analytic sample), other race groups were combined, consisting of Hispanic Americans, Native Americans, Asian or Pacific Islander, as well as those identifying themselves as ‘other race’. Depressive symptoms were included in the analyses based on the 11-item Center for Epidemiologic Studies Depression Scale (CES-D; Cronbach's α = 0.80). For health lifestyle variables as human and social capital resources, W1 tobacco use, physical activity and obesity were included. Participants who reported currently smoking cigarettes, pipes, cigars or chewing tobacco were classified as current tobacco users. Physical activity was measured with an item probing respondents’ engagement in physical activities such as walking, dancing or exercise (0 = never, 1 = once a month, 2 = one to three times a month, 3 = one or two times a week, 4 = three or more times per week). Obesity is defined by Body Mass Index (BMI) values equal to or higher than 30 kg/m2. Respondents’ weight and height were measured by the interviewer with a scale and measuring tape, not provided by the respondents. Finally, meeting attendance was measured by a single binary item asking if respondents attended any organised meeting in the past year (0 = no, 1 = yes).
Supplementary analyses considered additional variables including respondents’ income, wealth, contact with core network members (e.g. total volume of contact frequency, average contact per year, average contact per person), living arrangements (i.e. number of people living in the same household), stroke, hypertension and cancer. These variables were then omitted from the final analyses for more parsimonious models because they were not significant and did not alter the substantive conclusions in multivariate specifications.
Analyses
Logistic regression models were used to predict the participation in each activity role at W1 and W2. For volunteering and employment, a sub-sample of respondents who were undertaking a role at W1 was examined to predict the maintenance of activity during the five-year period. However, because a larger proportion of the NSHAP sample initiated care-giving across the two waves, respondents who were not providing care at W1 were included in the logistic regression model to examine what factors influence the initiation of care-giving across waves.
Separate models were estimated for the frequency of activity within each role. Frequency for volunteering ranges from zero to six and is normally distributed, and the analyses used ordinary least squares (OLS) regression to model the W1 frequency of activity and change in activity (by regressing W2 frequency of activity on the W1 level). Frequency for employment and care-giving ranges from zero to two and the ordinal logistic regression models were used to predict the W1 and W2 frequency of activities. For employment, analyses for W2 participation frequency included respondents who worked at W1. For care-giving frequency at W2, respondents who did not provide care at W1 were included in longitudinal analyses.
A series of supplementary analyses were conducted to test the robustness of the results. First, multinomial logistic regression estimated the changes in frequency of activity across waves with −1 indicating a decrease, 1 indicating an increase and 0 reflecting no change in the given activity between waves. Further, ordinal logistic regression was specified to predict the frequency of volunteering since categories were measured at the ordinal level. The substantive conclusions were unchanged, and OLS regression was used because of its greater ease in interpreting coefficients.
Although the current project does not suffer from an extensive amount of missing data, some baseline variables had missing values. Of particular concern were obesity and neighbourhood socialisation. To address this issue, a multiple imputation method was employed using Stata's ‘mi impute’ command. The imputation models included all independent variables and five data-sets were generated. In addition, in order to adjust for differential selectivity due to death, Heckman's (Reference Heckman1979) selection bias models were employed, since sample attrition primarily due to selective mortality may produce biased parameter estimates in longitudinal analyses. First, a probit model was estimated to distinguish participants at the follow-up interview from those who were deceased. Predictors of mortality in the probit model included age, female and tobacco use, along with variables that were not included in the substantive equation predicting productive activity (i.e. walking a block, underweight [BMI < 18.5], stroke, cancer, heart problems). The selection lambda based on the inverse Mills ratio was subsequently included in the models.
Results
Table 1 summarises the descriptive statistics for study variables. At baseline, a majority of respondents volunteered (63%) in the past year and one-third of the sample volunteered at least once a month. Employment and care-giving were less prevalent, with 30 per cent employees and 17 per cent care-givers in the sample. Though the vast majority of respondents did not provide care, among the 327 who did, 52 per cent were part-time and 48 per cent were full-time care-givers, meaning that they were assisting care recipients seven days a week. Among older employees, 48 per cent were part-time and 52 per cent were full-time workers.
Table 1. Descriptive statistics for variables
Notes: Sample size varies due to missing data. SD: standard deviation. Ref.: reference group. BMI: Body Mass Index. CES-D: Center for Epidemiologic Studies Depression Scale.
Source: National Social Life, Health, and Aging Project.
At W2, the prevalence of volunteering remained at 62 per cent, indicating considerable stability in the activity over time. In addition, the frequency of volunteer activity did not necessarily decrease over time; among W1 volunteers, 42 per cent showed a decrease, 24 per cent experienced no change and 34 per cent showed an increase in the frequency of volunteer activity. However, employment declined from 30 to 10 per cent in the entire analytic sample, with 44 per cent of the workers being employed full-time. Finally, though the proportion of care-givers remained stable (17% at W1, 16% at W2), many respondents moved in and out of the category. Among W1 care-givers, 71.9 per cent stopped providing care while 6.9 per cent of W1 non-care-givers started care-giving activity by W2. In general, descriptive analyses revealed that volunteering remained fairly stable over the five-year period, while employment declined over time among older adults. Care-giving showed a substantial amount of change over time.
Overall, the older adult sample in NSHAP was relatively well-equipped with human capital. More than two-thirds of the sample received at least high school education and did not report any functional limitation. The majority of respondents reported very good mental and physical health. In terms of social capital, 62 per cent of the sample were married. On average, respondents had about 3.5 friends in their core discussion network with whom they could discuss important matters. They also received an adequate amount of support from their spouse, friends and family members (mean = 1.43, standard deviation (SD) = 0.43) and socialised with neighbours at least several times a year (mean = 1.38, SD = 1.28). Respondents were fairly frequent religious service attenders; 46 per cent of the respondents had attended church at least every week in the past year.
Predictors of productive activities
Table 2 displays the results for volunteering at W1 and W2. The first equation is a logistic regression of W1 participation in volunteering regardless of their frequency in activity, and the second model predicts W2 maintenance in volunteering on a sub-sample of W1 volunteers. The cross-sectional analyses demonstrated that volunteering was influenced by all three forms of capital. Examining the participation of volunteering at W1 in the first equation, those with higher education (odds ratio (OR) = 1.60, p < 0.001), fewer functional limitations (OR = 0.89, p < 0.001), bigger network size (OR = 1.14, p < 0.001), greater social support (OR = 1.46, p < 0.01), a greater amount of neighbourly socialisation (OR = 1.11, p < 0.01) and church attendance (OR = 1.43, p < 0.001) were more likely to be active volunteers. However, analyses using two-wave panel data (second equation) showed that education (OR = 1.49, p < 0.01), neighbourly socialisation (OR = 1.22, p < 0.01) and religious service attendance (OR = 1.33, p < 0.001) remained as significant predictors of continued volunteer activity. Among W1 volunteers, education, neighbourly socialisation and religious service attendance were associated with 49, 22 and 33 per cent greater odds of continuing the same activity, respectively.
Table 2. Logistic regression and ordinary least squares (OLS) regression predicting volunteering, Waves 1 and 2
Notes: All models adjusted for age, gender, race, depressive symptoms, tobacco use, physical activity, obesity, meeting attendance and selection lambda. 1. Odds ratios (OR) are exponentiated coefficients based on logistic regression. 2. Unstandardised coefficients based on OLS regression. CI: confidence interval. SE: standard error.
Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.
The third and fourth equations show the results of OLS regression analyses for W1 and W2 volunteering frequency. Older adults with greater human, social and cultural capital not only participated in W1 volunteer activity, but also dedicated more time in doing so. However, over five years, education (b = 0.28, p < 0.001), network size (b = 0.09, p < 0.05), neighbourly socialisation (b = 0.13, p < 0.01) and religious service attendance (b = 0.30, p < 0.001) exerted the effects adjusting for non-random selection effects and W1 volunteering frequency. All in all, volunteers were generally well-equipped with human, social and cultural capital. Employment or care-giving were not significant predictors of volunteering. However, it is noteworthy that those who attended meetings more regularly were more likely to volunteer and did so more frequently at W2 (model not shown).
Table 3 shows the findings for employment. The first two equations show results from logistic regression estimating W1 and W2 employment status, and the third and fourth equations present findings from ordinal logistic regressions predicting W1 and W2 employment, while distinguishing part-time and full-time employment. The results indicated that employment was greatly influenced by human capital. At W1 (first equation), education (OR = 1.13, p < 0.05), physical health (OR = 1.36, p < 0.001) and functional limitation (OR = 0.78, p < 0.001) were associated with employment. Those who received a greater amount of social support were less likely to participate in paid work activities (OR = 0.75, p < 0.05). Other components for social capital and cultural capital did not predict the likelihood of paid work. Among covariates, those who are younger, female, Black and other race were less likely to be employed.
Table 3. Logistic regression and ordered logistic regression predicting employment, Waves 1 and 2
Notes: All models adjusted for age, gender, race, depressive symptoms, tobacco use, physical activity, obesity, meeting attendance and selection lambda. 1. Odds ratios (OR) presented based on binary logistic regression. 2. OR presented based on ordinal logistic regression. CI: confidence interval.
Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.
Among W1 workers, mental health was associated with 30 per cent higher likelihood of continued employment at the subsequent wave. The effects of education, physical health and functional limitation were attenuated in the analyses using data from both waves.
Finally, Table 4 presents the findings on care-giving. At W1 (first equation), married individuals were almost twice as likely as unmarried persons to provide care for adults with their day-to-day activities (OR = 1.81, p < 0.001). Among W1 non-care-givers (second equation), those with a larger network were more likely to initiate care-giving activity over the five-year period. Higher mental health was a negative predictor of care-giving initiation (OR = 0.77, p < 0.05). The results from the ordinal logistic regression replicate those from binary logistic regression. The findings provide little evidence for the healthy care-giver hypothesis or the role of cultural capital in predicting care-giving.
Table 4. Logistic regression and ordered logistic regression predicting care-giving, Waves 1 and 2
Notes: All models adjusted for age, gender, race, depressive symptoms, tobacco use, physical activity, obesity, meeting attendance and selection lambda. 1. Odds ratios (OR) presented based on binary logistic regression. 2. OR presented based on ordinal logistic regression. CI: confidence interval.
Significance levels: * p < 0.05, *** p < 0.001.
Discussion
The productive ageing model was developed to promote the value of age-integrated society, focusing on the predictors and consequences of productivity in later life. Building on this body of literature, the current study examined the role of human, social and cultural capital in shaping three types of productive activities over a five-year period.
The strong effects of education as a form of human capital on productive activity indicate that these activities are the enterprise of individuals with human resources. The current findings show that education plays a direct role in promoting productive activity, particularly volunteering and employment, consistent with other studies (Burr et al., Reference Burr, Caro and Moorhead2002; McNamara and Gonzales, Reference McNamara and Gonzales2011). Though older adults in the sample completed their education almost 30–40 years ago, education affected not only baseline participation in volunteering and employment, but also the long-term commitment to volunteer activity over the five-year period. Thus, education remains a powerful resource of multiple types of productive activities above and beyond schooling years. It does not necessarily indicate that older adults with fewer human resources do not participate in productive activities; rather, it points to the fact that education may be a barrier to volunteerism and employment, particularly for those with fewer resources (Raymond et al., Reference Raymond, Grenier and Hanley2014; Cramm and Nieboer, Reference Cramm and Nieboer2015). They still engage in informal helping and do-it-yourself activities that might not be captured within the productive ageing framework. Thus, policies aimed at reducing barriers to accessing volunteer and employment opportunities for those with limited resources may open up new avenues for both paid and unpaid contributions in later life (Loh and Kendig, Reference Loh and Kendig2013).
Mental health was a significant predictor of continued employment activity and initiation of care-giving. In line with the healthy worker hypothesis, mentally healthier older adults are more likely to consider continued work as a viable option (Schwingel et al., Reference Schwingel, Niti, Tang and Ng2009). It may also indicate that among already healthy and educated employees, mental health is a crucial resource for prolonged employment activity. Contrary to a healthy care-giver hypothesis (Fredman et al., Reference Fredman, Lyon, Cauley, Hochberg and Applebaum2015), mentally healthier individuals were less likely to initiate care-giving between waves. Given the age of the NSHAP sample, older adults with better mental health may be better at utilising professional and medical care resources such as arranging paid professional help compared to their less-healthy counterparts (Pinquart and Sorensen, Reference Pinquart and Sörensen2003). Further, care-givers who report lower baseline health are more likely to develop more severe depressive symptoms, exhibit worse health habits and report care-giver burden (Donelan et al., Reference Donelan, Hill, Hoffman, Scoles, Feldman, Levine and Gould2002; Pinquart and Sorensen, Reference Pinquart and Sörensen2007). Thus, from a policy perspective, existing policies such as the Family Caregiver Support Act should expand services for improving care-givers’ self-care capacity and psychological well-being.
Another notable finding is the lack of relationship between ADLs and productive activities over time. It is consistent with the literature demonstrating that functional ability is not an important determinant of volunteer and employee retention (Tang et al., Reference Tang, Morrow-Howell and Choi2010). However, in order to shed light on these counterintuitive findings, several additional analyses were conducted. First, W2 productive activities were re-estimated using the changes in ADL (i.e. increase, decrease, no change) along with other independent variables. The results showed that an increase in functional limitations had no significant effect on productive activities except for employment. Further, analyses were conducted to test the possibility of mediation by which functional limitations affect baseline activity and subsequently influence W2 activity. However, the Sobel–Goodman test did not reveal significant mediation effects. This can be due to lack of change in ADL over five years, since 68 per cent of the respondents did not experience any changes in ADL. More moderate functional limitation measures, such as Instrumental ADLs (IADLs), were only available at W2 and not used in the current project. However, supplementary analysis using only W2 data revealed that IADL was a marginally significant predictor of volunteering and a significant predictor of employment. Taken together, the results from this project, along with other studies (Li and Ferraro, Reference Li and Ferraro2006; Musick and Wilson, Reference Musick and Wilson2007), suggest that the effects of health selection for older adults may not be as substantial as one might expect.
The effects of social capital exhibit a less-consistent pattern than that of human capital. Those with greater social capital generally participate more in productive activities but the effects are not apparent over the five-year period. Not surprisingly, married individuals were more likely to be care-givers (possibly to their ill spouses). Network size is positively associated with the initiation of care-giving. Further, neighbourhood socialisation, a form of bridging capital, was associated with volunteering both cross-sectionally and over time. The finding is consistent with the benefits of bridging social capital, in that those with greater bridging capital are more likely to be recruited and stay as volunteers for a longer period of time (Putnam, Reference Putnam2000; Wilson, Reference Wilson2000; Musick and Wilson, Reference Musick and Wilson2007; Gonzales and Nowell, Reference Gonzales and Nowell2017). One practical example is Experience Corps in Baltimore city, who recruited older adults from diverse backgrounds and collaborated with various umbrella neighbourhood organisations, AARP, school districts, churches, retiree organisations and senior centres in order to increase the community effectiveness and volunteer retention (Martinez et al., Reference Martinez, Frick, Glass, Carlson, Tanner, Ricks and Fried2006). Thus, voluntary organisations may benefit from partnering with community organisations or recruit volunteering from the neighbourhood or city as a unit.
Perhaps the most consistent and notable effects are from cultural capital, indicated by the frequency of religious service attendance. Those who attend religious services more regularly are more likely to be volunteers than are less-frequent service-goers. This underscores the role of cultural capital in recruiting and retaining older individuals into volunteering. Though some studies consider religious service attendance as a form of social capital, the current project adopted a view that religious organisations are not simply a set of social networks, but rather a cultural context that promotes collective actions and helps their parishioners internalise the culture of benevolence (Lim and MacGregor, Reference Lim and MacGregor2012). This echoes Robert Wuthnow's (Reference Wuthnow1991: 284) argument that religious institutions teach values of giving back to the communities in their ‘churches, synagogues, fellowship halls, and meeting places … [and also] command valuable resources for mobilizing people, turning their good intentions into concrete actions, so that the needy are actually helped’.
Testing the hypotheses regarding the social antecedents of productive activity leads to a number of important conclusions. First, it should be noted that human capital is not the sole predictor of productive activity, particularly over time. Rather, older adults’ activities occur in the context of human, social and cultural capital. Though previous studies on the productive ageing model have rendered notable findings on how resources shape one specific activity, cross-sectionally or over time, they also have acknowledged that failing to take into account selective survival (Bertrand et al., Reference Bertrand, Saczynski, Mezzacappa, Hulse, Ensrud and Fredman2012), longer vectors of independent variables (McNamara and Gonzales, Reference McNamara and Gonzales2011) and changes in the activities over time (Choi, Reference Choi2003) may lead to over-estimation of the role of resources in predicting any productive activity of interest. The current study contributes to the literature on productive ageing that different patterns emerge in cross-sectional and two-wave panel data of productive activity, adjusting for baseline activity level and attrition.
It is also noteworthy that resources may hinder older adults’ productive activities. The underlying assumptions of the productive ageing model are widely accepted as a positive way of ageing; some critical gerontologists, such as Harry Moody (Reference Moody, Bass, Caro and Chen1993) and Carol Estes and Jane Mahakian (Reference Estes, Mahakian, Morrow-Howell, Hinterlong and Sherraden2001), are concerned that it may dichotomise ageing as productive or unproductive (i.e. dependent) processes and create yet another competitive system in later life. If activity patterns among older adults are mainly contingent on their capital (e.g. education or health), the productive ageing model may become an extension of market logic in later life, further marginalising disadvantaged groups. Thus, the ultimate goal of productive ageing should be to broaden the opportunities and eliminate barriers to activities (Raymond et al., Reference Raymond, Grenier and Hanley2014).
The findings further demonstrate how individual and social factors influence productive activity differently, calling attention to studying multiple types of productive activities. Some research pioneers this trend by examining multiple productive activities at the same time, using a person-centred approach using latent class analyses or a variable-centred approach employing exploratory factor analyses, in attempt to classify these types of productive activities and what factors predict these activities (Burr et al., Reference Burr, Mutchler and Caro2007; Morrow-Howell et al., Reference Morrow-Howell, Putnam, Lee, Greenfield, Inoue and Chen2014). Future research should focus on these profiles of activities.
Several aspects of the study are novel, but the conclusions should be interpreted with caution due to some study limitations. Studies on the trajectories of productive activity over the life course could benefit from adopting a longer view of social environments and behavioural outcomes that spans across middle and later life. The current study took a step forward by using two-wave data from a nationally representative sample of older adults, but did not have any information on duration, timing and previous experiences in productive activity. Investigating when, why and how long individuals partake in productive activities will shed light on how one can create opportunities for older adults to participate in these activities, promoting optimal ageing.
Further, though the current study focused on three types of important productive activity, a wider range of activities should be considered to capture a more comprehensive picture of older adults’ productive activity profiles. For example, civic participation is a type of productive activity (Burr et al., Reference Burr, Caro and Moorhead2002). Voting behaviour or participation in political activities are prevalent types of community engagement for older adults since they require a modest commitment of time and skills. Further, older adults undertake multiple roles and are active in these roles to varying degrees. Since some activities are complementary while others are not, investigating a wider array of activity patterns is as important as studying one activity in depth.
Though the current project adjusts for gender and race, there are observed differences in productive activity patterns in multiple gender and racial groups (Choi et al., Reference Choi, Tang and Copeland2016; Gonzales et al., Reference Gonzales, Matz-Costa and Morrow-Howell2015). For example, supplementary analyses revealed that a greater proportion of volunteers and care-givers are non-Hispanic White women. Future research projects should examine systematically gender- and race-specific effects of capital on productive ageing.
Finally, extensive measures of capital are needed. For cultural capital, religious service attendance is recognised as a valid proxy measure for the ‘culture of benevolence’ (Wuthnow, Reference Wuthnow1991) because religious organisations not only provide older adults with opportunities for contributing to the larger community but also promote values in helping others. Nonetheless, multiple or direct measures of cultural capital (e.g. values in ‘being productive’ or ‘helping others’) are preferred. For human capital, more subtle measures of functional limitation may be preferred, such as IADLs. Though ADLs indicate moderate to severe functional impairment including dressing or eating, future research projects should examine a wider range of nuanced measures of functional ability.
Despite limitations, the study adds to the literature by finding that human, social and cultural capital affect productive activities differently, and suggests that scholars cannot simply assume that capital will be related to productive activities in a similar way across activities or over time. It is evident that some forms of capital predict the changes of the activity while others do not. The findings also suggest that studies that examine one form of capital or a single activity of interest cross-sectionally may run a risk of over-estimating the effects of capital on any activity in question. Further, some of these factors, including social support, network size, neighbourhood socialisation and religious service attendance, are relatively amenable to policy changes or community interventions, which can help older individuals to stay active and productive.
Acknowledgements
The author appreciates the comments from Kenneth F. Ferraro, Elliot Friedman, Ann Howell, Sarah A. Mustillo, Patricia A. Thomas and two anonymous reviewers on an earlier version of this paper.
Ethical standards
No ethical approval was needed.
