Health benefits of volunteering in the Wisconsin Longitudinal Study * Jane Allyn Piliavin and Erica Siegl University of Wisconsin-Madison

Word count: 8345 Tables: 8

Running head: Health Benefits of Volunteering in the WLS

* Jane Allyn Piliavin Department of Sociology 8128 Social Science Building 1180 Observatory Drive University of Wisconsin-Madison Madison WI 53706-5983

Erica Siegl Department of Sociology 8128 Social Science Building 1180 Observatory Drive University of Wisconsin-Madison Madison WI 53706-5983

The research reported in this paper was supported by NIA Grant #1 R03 AG21526-01. The authors would like to thank Shelley Correll, Robert Hauser, and Hal Winsborough for statistical help, and Jeremy Freese, Cheryl Bowdre, and all of the other members of the CDHA working group for extremely valuable feedback during the preparation of this analysis and manuscript.

Health benefits of volunteering in the Wisconsin Longitudinal Study

Abstract

Positive effects of volunteering on psychological well-being and self-reported health are investigated, using all four waves of the Wisconsin Longitudinal Study (WLS). Confirming previous research, volunteering was positively related to both outcome variables. Volunteering consistently over time significantly improved well-being and self-reported health. Volunteering for one or two organizations had a similar positive effect, but greater diversity of participation added no further benefits. The relationship of volunteering to psychological well-being was moderated by level of social integration, such that those who were less well integrated benefitted the most. Mattering appears to mediate the volunteering - well-being link. Controls for proxy measures of well-being in 1957, for other forms of social participation, and for the predictors of volunteering are employed in analyses of 1992 well-being. Volunteering effects on psychological well-being in 2004 are found controlling for 1992 well-being, providing a strong argument for a causal effect.

Health benefits of volunteering in the Wisconsin Longitudinal Study

When praised for their altruistic actions, blood donors and volunteers commonly respond, Oh, no, what I do is really selfish. I get so much more out of it than I give. Is this simply American individualist rhetoric? Or, is there evidence that doing work that serves others has moodenhancing, social-integrating, health-promoting, or even death-delaying power? Based on a review of the literature, Piliavin claims that the basic answer is ... essentially yes. One does well by doing good. (2003, p. 227). Numerous studies reveal protective effects of volunteering on mental and physical health. The strongest evidence for the benefits of time spent volunteering comes from longitudinal research. Young and Glasgow (1998) found that self-reported health status increased as instrumental social participation increased for both men and women, using a longitudinal sample of 629 non-metropolitan elderly. Moen, Dempster-McClain, and Williams (1989), following a sample of women who were between the ages of 25 and 50 when first interviewed in 1956, found that participation in clubs and volunteer activities had a significant protective effect on mortality in 1986. The analysis controlled for many other relevant factors, including the number of other roles and health in 1956, and the article makes clear that the activities were indeed largely community-oriented (PTA, scouting, book drives, etc.). In a second more complex analysis, based on interviews done in 1986 with the 313 surviving women, Moen, et al. (1992) find effects on three measures of health: self-appraised health, time to serious illness, and functional ability. Oman, Thoresen, and McMahon (1999) also examined volunteering and mortality in a 1990-91 prospective study of 2,025 community-dwelling elderly aged 55 and older in Marin County, California. Mortality was assessed through November, 1995. Controlling for health habits,

physical functioning, religious attendance, social support, and many other factors, high volunteers (two organizations) had 44% lower mortality than non-volunteers. A number of excellent studies employ the Americans Changing Lives data set (House, 1995). Using the first two waves of those data, Thoits and Hewitt (2001) discover that both the number of volunteer hours at time one (1986), and the change in volunteer hours to time two (1989) show significant effects on six measures of well-being: happiness, life satisfaction, selfesteem, mastery, depression, and physical health at time two. There is also a significant effect of well-being at time one on volunteering at time two, but it is small. Controls for demographic factors as well as for other forms of community participation (and change in participation), such as church attendance and participation in other organizations, do not eliminate these effects. The most highly significant effects are on life satisfaction. The present research seeks to extend this literature by testing the volunteering-well-being relationship on a longitudinal data set never before used for this purpose: the Wisconsin Longitudinal Study (WLS). Although previous research has been done with longitudinal samples, only Moen, Dempster-McClain, and Williams (1989) followed their respondents for a long period (30 years) and they had a very small sample, all of them women. The WLS began in 1957 with a 1/3 sample of all of that years graduates from Wisconsins high schools. They were re-interviewed in 1975, 1992, and 2004, and social participation information is available from all three waves. In 1992 and 2004, information was obtained about psychological wellbeing and self-reported health. Thus both an analysis of the impact of volunteering on wellbeing and an analysis of the mediating role of psychological resources on perceived health can be carried out. 4

We are focusing on these two measures because (1) there is good evidence for the importance of self-perceived health for the prediction of harder health measures of morbidity and mortality, above and beyond medical indicators (Benyamini, Idler, Levenhthal, and Leventhal, 2000), (2) psychological well-being is associated with physical health outcomes and as an aspect of mental health is important in its own right (Keyes, 2005; Keyes & Haidt, 2003), and (3) at the age of 64 in the most recent wave, the WLS sample is too young to have experienced much mortality (about 10%) and their health is on the whole quite good. In short, this study contributes by pursuing the questions: why is volunteering different from other forms of social engagement, for whom is the effect of volunteering most beneficial, how much participation is optimal, and through what mechanisms does it operate. Why should volunteering be special? In previous research, there has been little discussion of why volunteer work should be particularly beneficial as compared to activities such as doing amateur theater, playing on a sports team, participating in a garden club, or carrying out non-institutionalized informal helping or family caregiving. A number of theories (e.g., role accumulation or activity theory) as well as Durkheims social integration hypothesis would propose that any form of social participation should have the effect of decreasing alienation. Thus social participation in a variety of ways was seen as equally beneficial. Others (e.g., Friedland et al., 2001) have even found positive health effects of simply keeping active in a variety of ways in older age. So, why not also include community participation of all sorts? The answer stems from both theoretical and empirical evidence. In early arguments over the costs and benefits of multiple roles, proponents of the scarcity approach (Coser, 1974;Goode, 1960; Sarbin and Allen, 1968) which posited that the 5

demands of different roles will conflict and that more roles would lead to greater strain largely lost out to proponents of the role accumulation approach. Adherents to this view (Marks, 1977; Sieber, 1974; Thoits, 1986) assumed that social roles provide status, role-related privileges, and ego-gratification, and that identities associated with these roles give individuals meaning and purpose and therefore adding roles will enhance psychological well-being. Both increasing the number of roles (Miller, Moen, Dempster-McClain, 1991; Thoits, 1986) and holding particular roles such as that of spouse (Baruch and Barnett, 1986; Gove and Geerken, 1977; Menaghan, 1989) enhance psychological well-being. There is also evidence of physical health benefits (Adelmann, 1994a; Moen, Dempster-McClain, and Williams, Jr., 1992; Verbrugge, 1983; 1987; Waldron and Jacobs, 1989) contingent on performing spouse, employee, and parent roles. Thoits (1992; 1995), however, suggests that voluntary roles such as friend or group member may be more responsible for the positive effects of multiple roles than are obligatory roles such as parent or spouse. Informal helping of neighbors and friends also my carry some sense of obligation. It would be difficult to stop driving ones elderly neighbor to the doctor once one has begun. Caretaking roles with regard to family are also obligatory; thus we are looking only at volunteering, as a non-obligatory role identity. The distinction between eudaimonic and hedonic well-being is useful for understanding the impact of different kinds of activities (Ryan & Deci, 2001). The constructs of life satisfaction and happiness are measures of hedonic well-being: feeling good about ones situation in life. Eudaimonic well-being, on the other hand, is defined in terms of meaning and self-realization. This involves not only feeling good, but also feeling good about oneself. We propose that this is the power of volunteering and similar other-oriented activities (e.g. political 6

activism). When engaging in social activities or hobbies, one experiences hedonic well-being. But in other-oriented activity such as volunteering, one can enjoy the activity itself but also feel a sense of satisfaction that one is serving society. It is our contention that it is this focus outside oneself that provides the greatest benefit to mental health, perhaps in part through enhancing self-esteem and the sense of mattering feeling that one makes a difference in the world (Elliott, Kao,& Grant, 2004; Rosenberg & McCullough, 1981). Thus, although all freely chosen activity should increase psychological well-being and have the potential to increase physical health, volunteering should give one that extra boost. Who will benefit most from volunteering? Thus far, the positive effects of volunteering on physical and mental health appear to be particularly strong among the elderly, and other adults with limited social networks. Van Willigen (2000) presents an elegant study demonstrating the benefits of volunteering for the well-being of the elderly, and comparing it with its impact on younger adults, using the first two waves of the Americans Changing Lives data (1986, 1989). For both older and younger adults, volunteering predicts greater life satisfaction and better perceived health, regardless of what measure of volunteering is used. However, the relationships are significantly stronger in the elderly sample for two of the three measures. Oman et al. (1999) also found -- and report on studies by others -- that the impact of volunteering on mortality increases with increasing age; that is, those more at risk are helped the more. Musick, Herzog, and House (1999) tracked respondents aged 65 and older at the first wave of the Americans Changing Lives (House, 1995) data set, using the National Death Index, from the year of the survey (1986) through March, 1994. The protective effect of volunteering on mortality was found only among those with low informal social interaction (measured by how 7

often they talk on the telephone with friends, neighbors, or relatives in the typical week and how often they get together with them). How much volunteering is optimal? Amounts of volunteering can be considered in at least different dimensions: the number of different organizations (diversity), the amount of effort or time spent per week, year, or month (intensity), and how regularly one volunteers across the lifespan (consistency). Friedland, Fritsch, Smyth, et al. (2001) make the useful distinction between diversity and intensity of involvement in activities in general. They find that the diversity of engagement (i.e. number of different activities) at midlife across three categories: passive, intellectual, and physical, is protective against Alzheimers disease at age 70. Intensity of intellectual activities at midlife also distinguished between the control-group members and the Alzheimers patients. Controlling for health, race, age, income, physical activity, and initial health and impairment, Musick, Herzog, and House (1999) found that moderate volunteering (< 40 hours per year or for only one organization) had a protective effect against mortality, but volunteering for additional hours did not contribute further. Luoh and Herzog find that volunteering leads to better health and lower mortality but that . . .the quantity of volunteer and paid work beyond 100 annual hours is not related to health outcomes. . . (2002: 490). The third dimension of volunteering, consistency, can be defined as the extent to which one has engaged in the activity regularly over a period of time. Using three waves of the ACL data set, Musick and Wilson (2002) find positive effects of volunteering on depression, mainly in the over 65 group, with a dose-response effect over time. Volunteering in only one wave has no effect, in two it has some, and in three has a large positive effect. This is consistent with van Willigens (2000) findings, mentioned previously, but with different dependant variables. 8

Mediating mechanisms. Thoits and Hewitt note that an important question is, . . . how the positive effects of volunteer work on well-being are generated. What are the mechanisms through which happiness, life satisfaction, self-esteem, a sense of control, good health and lower depression result from volunteer work (2001:128)? Musick, Herzog, and House (1999) suggest that one mechanism by which their finding that volunteering led to better physical and mental health among those with low social interaction could have been through preventing alienation and anomie, since the index measures social integration. Musick and Wilson, (2002) also find very small mediating effects of psychological and social resources on the relationship between volunteering and depression. So, what might be the actual process by which health and well-being are generated by volunteer participation? Among psychologists, current research indicates that good feelings alter peoples bodily systems (Fredrickson & Losada, 2005, p. 678). Davidson et al. (2003) have found that positive affect increases immune function. Prospective longitudinal studies have demonstrated that frequent positive affect predicts, among other outcomes, psychological growth (Fredrickson et al, 2003), lower levels of cortisol (Steptoe, Wardle, & Marmot, 2005), resistance to rhinoviruses (Cohen, et al, 2003), and even how long people live (cites from Fredrickson). Elliott, Colangelo, and Gelles (2005) have demonstrated that among adolescents, those with higher scores on a measure of mattering have significantly lower levels of suicide ideation. Mattering appears to be related to higher self-esteem and in turn to lower depression, which directly affects thoughts of suicide. The proposed sequence, then, is that volunteering leads to both hedonic well-being and, through the mechanism of mattering, eudaimonic well-being feeling better about oneself. The 9

effects of volunteering on well-being should be strongest among those who have fewest other sources of well-being, such as employment, social ties, and social support. Although both of these positive emotions probably strengthen the immune system, eudaimonic well-being should contribute more. The final result is decreased morbidity and mortality. Since we have no way to measure the actual physiological links postulated, and because effects on morbidity and mortality have largely been shown in those older than 60 or 65, this paper will concentrate more on psychological well-being than on physical health. Hypotheses: 1. Volunteering will be positively related to psychological well-being and self-reported health, and will be more strongly related than more self-oriented social participation. 2. The dose-response curve will be such that volunteering for more organizations, and more continuous involvement in those organizations, will lead to more positive effects. 3. The relationship of volunteering to psychological well-being will be moderated by level of social integration: those who are less well integrated will benefit the most. 4. 5. The impact of volunteering on well-being will be mediated by the sense of mattering. The relationship of volunteering to self-reported health will be mediated by psychological well-being. That is, the effect of volunteering on self-reported health will be indirect. Methods Sample. The WLS began with a 1/3 random sample (N = 10,317) of women and men who graduated from Wisconsin high schools in 1957. The next two waves of survey data were collected from the graduates or their parents in 1964 and 1975; these early waves concentrated

on status attainment processes. In 1992, telephone and mail surveys of 8500 WLS graduates were conducted. These surveys updated measurements of marital status, child-rearing, education, labor force participation, jobs and occupations, social participation, and future aspirations and plans. It expanded the content of earlier follow-ups to include psychological well-being, mental and physical health, and many other variables. A major strength of the WLS lies in its exceptional sample retention. A major weakness is that the sample reflects the Wisconsin of the late 1950's, in that very few minority group individuals are included, and, because the sample members are all high school graduates, the level of education, occupational status, and income are above average. Measures . Information will be used from the 1957, 1975, 1992, and 2004 waves of data. Independent variables. The key independent variable volunteering is taken from measures of social participation asked in 1975, 1992 and 2004. Respondents were asked how much involvement they had in each of a list of possible kinds of groups such as labor unions, PTAs, or sport teams. In the 1975 wave involvement was measured as none, some, and very much (coded 0, 1, 2); the later waves used a five-point scale, from not involved to involved a great deal. There are a number of ways to conceptualize and measure volunteering. In this study it is defined as taking actions, within an institutional framework, that potentially provide some service to one or more other people or to the community at large. Thus we selected five types of organization deemed to be clearly other-oriented in nature: PTAs, youth groups, neighborhood organizations, community centers, and charity or welfare groups and considered participation in any of these to be volunteering. A number of different measures using this information were constructed. Our basic measures, Volunteering 1975 , 1992, and 2004 are 11

the sums of the answers regarding these five kinds of organizations at each time period.1 Initial scores ranged from zero to 10 in 1975 and from zero to 25 in the other two waves, and were highly positively skewed. Thus we use the logs10 of the sums in our analyses. Although the data on social participation do not include a record of the number of hours contributed -- the dose-response curve can be explored in relationship to the effect of diversity the number of organizations and consistency over time. With regard to moderating effects, it will be possible to test whether the positive effects of volunteering are the strongest where they are most needed: among those who are least well integrated into the society through other social ties. Unfortunately, because the sample members are all roughly the same age around 64 when last interviewed in 2004 the moderating effects of age cannot be investigated. One distinction we made earlier is between intensity and diversity (Friedland, et al. , 2001) of activities. Our basic measure conceptually combines these. We have constructed a separate measure of diversity as the count of the number of different organizations (out of five) in which respondents took part. We have been unable to devise a measure of intensity from these data, lacking a question on hours contributed. Attempts to use respondents degree of involvement to construct such a measure proved either conceptually inadequate or statistically problematic.2 However, we have calculated a measure of consistency of volunteering across the first two waves: This measure is zero if the respondent volunteered in neither 1975 or 1992, one if s/he volunteered in one of the two waves, and two of s/he volunteered at both time periods. Dependent variables. A problem for any causal analysis, unless repeated measures of the dependent variables are available, is attempting to demonstrate that the effects are not simply the result of a 12

continuing state that existed prior to the purported effect. The earlier waves of the WLS did not include mental or physical health measures. However, a few items from the 1957 high school surveys can be employed as proxies for positive and negative psychological states at that time.3 When the analysis is of well-being in 2004, well-being in 1992 can be used as a control. Many other variables, including IQ, education, income, marital status, and other factors known both to be related to volunteering and to contribute to mental and physical health are employed as controls. Because of our contention that other-oriented activity is particularly efficacious, a measure of participation in more self-oriented activities (referred to as social participation) is included as a stringent control, along with other measures of social contact: participation in church-related groups, church attendance, extent of visiting with friends, perceived social support, and work and marital status. 1. Psychological well-being. The WLS includes the six scales of psychological well-being developed by Ryff and her colleagues (Ryff, 1989), each consisting of seven items measured on six-point Likert scales, from agree strongly to disagree strongly with no midpoint. The measure used here in the 1992 and 2004 waves combines the four scales of environmental mastery, personal growth, purpose in life, and self-acceptance. 4 2. Self-reported health. This single item asks, How would you rate your health at the present time? The alternatives are very poor, poor, fair, good, excellent coded 1 to 5. Because the measure is highly skewed towards the negative end, the square root transformation is used. 5 Moderating and mediating variables. 1. Index of social integration. One hypothesis driving this research was that the positive impact of volunteering on psychological well-being will be greater among individuals who are less 13

socially integrated. An index of Integration was thus created from the sum of four variables as measured in 1992: current marital status (0, 1), work status (0, 1), rural-urban residence (1, 2, 3), visits with friends (1, 2, 3)6, and social support (0, 1, 2). The social support measure is the sum of yes-no responses to the two questions, Is there a person in your family (a friend outside your family) with whom you can really share your very private feelings and concerns? The index has a range of 2-10. A score of 2" reflects an unmarried, non-working individual living in a

rural area who has no kin or friends they can count on for support and few social visits.7 Not all
of these measures were available in 1975; thus only perceived social support, marital status, and work status are included in 1975. The term for the interaction between volunteering and social integration is explained later. 2. Mattering. The concept of mattering was introduced in 1981 by Morris Rosenberg. It is defined as the perception that, to some degree, we are a significant part of the world around us that people notice us, care that we exist, and value us. We hypothesize that the sense of mattering will mediate the volunteering - psychological well-being link. Elliott, et al. (2004) devised a 26item measure of three dimensions of mattering: awareness, importance, and reliance. The measure in the WLS questionnaire is the sum of four awareness items (e.g., People are usually aware of my presence.) and two reliance items (e.g., People tend to rely on me for support.). The items are measured on a five-point agree strongly to disagree strongly scale. Additional variables. 1. Controls for social participation. The most important control variable, Social participation, is a measure of a more self-oriented kind of organizational participation that is clearly not other oriented. The measure consists of the log10 of the sum of participation in veterans 14

organizations, fraternal organizations, organizations of the same nationality, sports teams, and country clubs in 1975, 1992, and 2004. Other controls for social participation in 1975 include participation in church-related groups, church attendance, and visits with friends.8 The number of children aged 6-12 in 1975 is included, because participation in many types of organizations is related to having children. 2. Controls for past psychological well-being. The initial 1957 questionnaire was focused almost entirely on students plans and aspirations for their lives after high school, and thus has no direct measures of psychological well-being or health. However, four variables have been developed that appear to relate to positive or negative psychological states or to perceptions of parental and teacher encouragement and support that might contribute to mental health. One item, with very badly worded double-barreled alternatives, was as follows: High school studies: ___ have been interesting; I want to learn more; ___ have been uninteresting; I would rather work than study; ___ have had no effect on my decision. Due to the order of the alternatives, and the leading words being interesting and uninteresting, we argue that these are the parts of the question on which students would have focused. Based on this reading, the item yielded two dummy variables: Interest and Disinterest, leaving the third alternative as the excluded group in regression analyses. Csikszentmihaly and his colleagues have produced a sizable literature on intrinsic interest, flow, and boredom (Csikszentmihaly, 1975; Hunter & Csikszentmihaly, 2003). In the most recent publication, using the experience sampling method (ESM), the authors find two groups of adolescents: those who experience chronic interest in everyday life experiences and 15

another who experience widespread boredom. The two groups differed significantly on measures of both level and stability of self-esteem, locus of control, optimism, and pessimism. The literature on boredom proneness has similar results. McLeod and Vodanovich (1991) found their measure of boredom-proneness to relate to self-actualization, while Sommers and Vodanovich (2000) found that those high on the scale had higher scores on anxiety, depression, and obsessive-compulsiveness. Based on this literature, we expect both of the dummy variables to function as proxies for psychological well-being (one positive, the other negative). Another item from the 1957 questionnaire asks, How sure are you that you will be doing what you plan? This is a dichotomous variable with the alternatives certain and uncertain; the dummy variable is named Certain. The sense of control is central to concepts of psychological well-being; one of the six scales we use from the Ryff inventory is a measure of environmental mastery. Hunter and Csikszentmihalyi (2003) include locus of control as one measure of wellbeing. House includes lack of self-efficacy/control and negative affect/hopelessness/ pessimism in his broad range of psychosocial risk factors for health (2002:125). A large literature on self-efficacy (Bandura,1977) indicates that perceptions of the ability to control ones outcomes have positive mental and physical health outcomes. A meta-analysis of 56 studies of self-efficacy and health-related outcomes (exercise, smoking-cessation, etc.) done from 1977 to 1989 (Holden, 1991) found strong evidence for a relationship, with an overall effect size of .276. Thus one would expect a measure of self-efficacy a concept closely related to certainty or confidence to relate to health. Thus the dummy variable Certain is also being used as a proxy for psychological well-being. Finally, two items concerning the encouragement of significant others parents and teachers for college studies were summed to make an index 16

of perceived Support in high school. A supportive environment for an adolescent should be predictive of later psychological well-being. 3. Controls for other sources of well-being. On the assumption that intelligence, education, occupational status, and income contribute to both mental and physical well-being, the variables of IQ measured in high school, the Duncan SEI of the head of household in 1957, respondents achieved educational level, the 1970 occupational education score,9 and combined family earnings in 1974 were included. *** Insert Tables 1a and 1b about here *** 4. Measures of health behaviors. These include the respondents body mass index, whether s/he is a smoker, number of packs a day smoked, and extent of exercise, which was the sum of two items: How often do you participate in light physical activity such as walking, . . .(vigorous physical exercise or sports such as aerobics, . . .), each with four possible answers, from three or more times a week to less than once per month. These are all measured in 1992. Percentages for categorical measures and means and standard deviations for continuous variables are presented in Tables 1a and 1b. Results The overall conceptual path model from which we are working was presented in Figure 1. We assume that family background factors and demographic and personal factors predict volunteer participation. Volunteering, social integration in 1992, and the interaction of integration and volunteering, in addition to the background factors, predict psychological wellbeing in 1992. All of the factors just mentioned, mediated through the sense of mattering, then 17

predict psychological well-being in 2004. Health habits such as smoking, weight control, and exercise, along with all of the foregoing variables, are expected to predict self-reported health. The prediction of volunteering. The literature tells us that volunteers come from the wealthier, more educated segment of society, that they are more likely to be female than male, to be married and have children, and to be active in their religion. Evidence for a reciprocal causal role of psychological well-being and volunteering (Thoits & Hewitt, 2001) also suggests including our proxy measures of well-being in the equation. Table 2 presents the regression of the log of Volunteering 1975 and Volunteering 1992 on the relevant variables. The strongest predictors of volunteering in 1975 are being female, church attendance, marital status, and number of children, quite consistent with the research in this area. Also significant are several measures of social class: occupation of the head of the respondents family of origin, and respondents family earnings and occupation education score. Finally, three of the early proxy measures of psychological well-being: high school support, interest, and disinterest (negatively) are significant or borderline significant, supporting Thoits contention that well-being predicts volunteering in addition to the reverse. Slightly over 15% of the variance in volunteering in1975 is predicted by the set of 14 independent variables. The analysis of 1992 volunteering shows similarities but some interesting differences. Marital status is no longer an important factor, and the size of the coefficient for being female has been more than halved. Education, not significant in 1975, is highly significant now, whereas SES of family of origin, significant in 1975, no longer has an effect. The number of children is no longer a positively contributing factor; and in fact is significantly negative; for most of these respondents, the children will be either in high school or gone.10 The amount of 18

variance explained is lower, even after the addition of past volunteering, work for church groups, and the index of more self-oriented activity, all of which are highly important predictors. *** Insert Table 2 about here *** Relationship of volunteering to well-being. Hypothesis 1 proposes that there will be a significant relationship between volunteering and well-being, and that this relationship will be greater than the relationship of engaging in more self-oriented activities (social participation) and well-being. As a test of this, partial correlations were calculated between the measure of volunteering and well-being, controlling for sex11 and social participation. Partials for self-oriented activities and well-being, controlling for sex and volunteering were also calculated. In every case tested (concurrent measures at the last two waves, and lagged measures of 1975 and 1992 social participation with well-being at the following wave) the correlation for volunteering was higher than the correlatlion for selforiented activities (See Table 3.). Based on this very simple analysis, there does appear to be an extra boost to well-being for doing unto others. *** Insert Table 3 about here *** Table 4 shows the first-order relationship of self-reported health and psychological wellbeing in 1992 to two measures of volunteering: diversity: the number of organizations out of five categories with which the respondent was involved in 1975, and consistency: the extent to which volunteering was done consistently over time. Significant relationships are found for both measures of volunteering with both psychological well-being and self-reported health. The strongest relationships are with the measure of consistency of volunteering over time. The actual

19

measure of self-reported health is shown here so that the reader can relate the means to the measure. ** Insert Table 4 about here** The dose-response curve. From the analysis presented in Table 4, it is not possible to tell how much volunteering (that is, for how many organizations) is optimal or whether consistent volunteering over time is better than volunteering only intermittently. To explore this question, the dependent variables were regressed on dummy variables described below. The omitted category is no volunteering in all analyses. Table 5a presents analyses of the 1992 measures of health and well-being. In the top panel of Table 5a, these measures are regressed on three dummy variables for diversity: participation in 1, 2, or 3 or more organizations. The second analysis, of consistency, employs two dummy variables, for participation in one wave (1975 or 1992) or in two waves. Two interesting findings come out of this table. First, for both dependent variables, the effect of participation in one organization in 1975 has a relatively small (although highly significant) effect, only about half the size of the other two dummies. Second, the effects of two organizations and of three or more are about the same, indicating that there is not a linear increase in well-being as diversity increases. We are unable to perform an analysis of the effects of intensity of volunteer involvement, because we lack information on the number of hours of volunteering that were performed. The results for the two-wave consistency analysis are different. For both dependent variables, the effect of volunteering in one wave is significant. Second, the effects of volunteering over two waves appear to be additive. For psychological well-being, the effect of the two waves dummy is about double that of the wave one dummy; for self-reported health, it is 20

larger by about 50%. There is no evidence for either diversity or consistency that one can get too much of a good thing. That is, there is never a decrease in the impact of volunteering as the amount increases. ** Insert Tables 5a and 5b about here** Table 5b presents analyses of the health and well-being measures from the 2004 wave. In this case we are able to control for the corresponding health measure at the 1992 wave. Before controlling for the measure in the previous wave, the results look very similar to those for the 1992 analysis. The exception is for the impact of 1992 diversity on self-reported health. In this case, there is an increase and then a decrease in the positive impact of volunteering. When the corresponding health measure from the previous wave is entered into the analysis all of the effects are greatly decreased. However, they remain significant in most cases. Again, the impact of consistency is more robust than that of diversity by our measures. In the regression models analyzing the 1992 data, we have measured volunteering with the use of one variable that combines diversity, consistency, and intensity. We do so because the way the diversity and consistency measures are calculated means they have a part-whole correlation and thus should not be used in the same analysis. The measure we use is simply the sum of the two indexes of volunteering in 1975 and 1992.12 Analyses of the 1992 measure of psychological well-being use this combined measure and an interaction term computed by multiplying the volunteering measure by the index of integration in 1992. Analyses of psychological well-being in 2004 retain these measures, and add both the 1992 measure of wellbeing and the index of volunteering in 2004. This allows for a stringent test of the causal impact of 2004 volunteering, by holding constant both prior volunteering and prior well-being. 21

The overall model to be tested on the 1992 data is a path model, in which volunteering (with its predictors13), current social integration, and the interaction of volunteering and integration predict psychological well-being in 1992. Table 6 presents the regression of psychological well-being on these variables. The first step of the regression included all the variables used to predict volunteering, plus the combined measure of volunteering in waves one ***Insert Table 6 about here*** and two. The highly significant effects on well-being in 1992 of all four 1957 measures intended as proxies for well-being supports their interpretation as such. On the second step, integration, and the interaction of volunteering and integration are added. On this step, the effects of gender and of marital status and visits with friends in 1975 drop somewhat. These latter decreases are largely due to the integration variable, since it includes more contemporary measures of marital status and visiting. The addition of these variables, however, increases the predictive ability of 1975 volunteering and the overall R2. It is worth noting that the effect of the four proxy measures of past well-being are not decreased much by the addition of the volunteering measures. This analysis provides support for Hypothesis 1: Volunteering even controlling for its predictors and other sources of well-being contributes significantly to psychological well-being, and Hypothesis 3: The effect is greatest for those who are least integrated in society. It should be mentioned that there is no effect of the social participation index, supporting our contention that other-oriented activities provide more of a boost to psychological well-being than more self-focused activities. As noted earlier, it is difficult to make a strong case that the relationship between volunteering in 1975 and 1992 and psychological well-being in 1992 is causal because of the 22

absence of a well-being measure in the 1975 wave. Although the 1957 proxy measures behave as they should, they in no sense can be seen as an equivalent measure to the Ryff scales used in 1992. In order to make this argument, we turn to an analysis of the 2004 data. Table 7 presents the regression of psychological well-being in 2004 on the same variables used in the previous analysis. In addition, however, the model includes volunteering as reported in 2004 and psychological well-being in 1992. Our expectation is that inclusion of the most recent volunteering will add to the prediction of well-being, and that inclusion of the measure of wellbeing in 1992 will decrease or eliminate the impact of earlier volunteering (which, we argue, was a contributing factor to that state) and 1992 social integration, but will not alter the impact of volunteering carried out between the two waves volunteering reported in 2004, providing support for the causal role of volunteering in psychological well-being. ** Insert Table 7 about here** Our expectations are borne out in the analysis. Before the addition of 1992 well-being and 2004 volunteering, the results are essentially the same as those for the 1992 wave. The combined measure of 1975-1992 volunteering, integration, and their interaction all have significant effects on well-being in 2004. Three of the four proxy measures of well-being from 1957 continue to show a relationship to psychological well-being, nearly 50 years later. When 2004 volunteering is added, there are very few changes in the other coefficients, and it has a highly significant effect. As expected, when 1992 well-being enters the equation, the impact of earlier volunteering becomes borderline and the interaction term also drops in size and significance, but there is a smaller decrease in the size of the coefficient for recent volunteering. It is our contention, of course, that earlier volunteering was causally related to 1992 well-being, 23

and that what we have shown is an indirect effect of earlier volunteering on 2004 volunteering through its effect on 1992 well-being. This cannot be unequivocally demonstrated due to the absence of a 1975 measure of well-being. We would like to point out, however, that all of the three proxy measures of well-being from 1957 that were significant initially decrease in size, and only one of them remains statistically significant when 1992 well-being is entered. This is what one would expect if the proxy measures are, indeed, less effective measures of well-being. In any case, the finding that recent volunteering reported in 2004 affects current well-being, controlling for prior well-being, is strong evidence for a causal effect. The final question we asked regarding psychological well-being is whether the effect of volunteering is mediated by mattering. That is, does volunteering lead to increased psychological well-being because volunteering leads individuals to feel that they matter in the world? This is tested in the final column of Table 7. When mattering is entered into the equation, the impact of recent volunteering is greatly reduced. This is, we believe, evidence for the mediating role of mattering. Volunteering increases psychological well-being because it leads people to feel that they have an important role in society, that their existence is important.14 We would also like to point out that the coefficient for the integration term was greatly decreased when 1992 wellbeing entered the equation, and was further decreased when mattering was entered. A question for further research may be to what extent connections between people of all kinds may have their impact on well-being through the medium of an increased sense of mattering. Finally, we arrive at the last step in the path, self-reported health. Regressions on perceived health in 1992 are presented in Table 8.15 Prediction overall is quite good: R2 is .162. ** Insert Table 8 about here** 24

The bulk of predictive power, however, comes from the measures of health habits: smoking, obesity, and exercise. Our predictions regarding the impact of volunteering, and the mediating role of psychological well-being, are not strongly supported. The combined measure of 19751992 volunteering is significant at traditionally accepted levels only until the health behavior measures are entered. The interaction term remains significant, if small. When well-being is entered into the equation, the interaction term drops to borderline significance and the impact of volunteering disappears entirely. It is hard to argue that this provides strong support for hypothesis 5, which stated that the impact of volunteering on perceived health is indirect, working through psychological well-being. Others have indeed found that the impact of volunteering on physical health and mortality does not appear until the later years. Thus we also analyzed the 2004 data, in which respondents are on the average 64 years old. The findings were about the same. When entered, both 1975-1992 volunteering and 2004 volunteering had borderline relationships with selfreported health. These again disappeared, however, when measures of health habits were entered. Conclusions We have once again confirmed the hypothesis that volunteering defined in this research as doing other-oriented community participation is positively related to psychological wellbeing and, to some degree, to self-reported health. There is evidence that the first relationship is causal, based on controls for psychological well-being measured when respondents were seniors in high school. Further analyses, using well-being measured with the same scales in 1992 and in 2004, confirmed this more conclusively. In addition, controls for more self-focused social 25

participation have no impact on well-being when volunteering is controlled, indicating that the effect is specific to altruistic forms of social engagement. Our second hypothesis, that volunteering for more organizations, and more intense and continuous involvement in those organizations, will lead to more positive effects, was also borne out. There was evidence that working for two organizations provided more benefit than working for one or none; diversity of participation beyond this point conveyed no further benefit. This replicates the research of Luoh and Herzog (2002). The relationship of continuity of involvement, on the other hand, appears to be linear, at least over the range we tested. Participation in both 1975 and 1992 leads to greater benefit measured in 1992 than participation in neither, and participation in three waves led to greater benefit than none, one, or two in analyses of well-being measured in 2004. This is consistent with the research of Musick and Wilson (2002) using the House (1995) data set. The third hypothesis, that the relationship of volunteering to psychological well-being will be moderated by level of social integration, with those who are less well integrated benefitting the most, was also confirmed. This result is consistent with the findings of Musick, Herzog, and House (1999). Finally, we predicted that the relationship of volunteering to psychological well-being would be mediated by mattering. This prediction was also borne out. Although the direct effect of volunteering on well-being was not eliminated, it was greatly reduced when mattering was introduced. This study helps solidify the volunteer-well-being link, but raises questions about differences between altruistic actions and other forms of social integration. Both contribute positively to psychological well-being; do they both operate via mattering? Or, is there a 26

different path for social integration? Do other factors mediate psychological well-being via an increased sense of mattering? This study restricted itself to a definition of formal volunteering, but it may be useful to further investigate the effects of different types of other-oriented community activities such as political participation and social protest. Analyses of self-reported health, controlling for other variables, indicate that the initial finding that self-reported health was slightly influenced by volunteering was largely illusory. Once the measures of health behaviors were introduced in multiple regression, the effect became borderline. Since the integration measure had a highly significant effect, it is clear that the general finding that connection to others as indexed by this measure is good for ones health is supported. Although we were unable to further substantiate the link between volunteering and self-reported physical health, its strong relationship to psychological well-being both directly and via mattering, and the known relationship between psychological and physical well-being provide good reasons to pursue this line of questioning. Others have found a relationship between volunteering and measures of physical health. Why not here? First, it is possible that more controls were employed in this analysis than in previous studies. Second, the WLS is not a random sample. The respondents in this sample are well above average in a number of respects. They have all graduated from high school. If we can trust their self-reports of height and weight, they are nowhere close to the national average for obesity. Very few of them even in 2004 when they are 64 years old report fair or poor health. Thus the power of the measure may simply not be enough to detect a very small effect, when controls are in place.

27

A striking finding in all of these analyses is the continuing effect of high school interest, disinterest, and certainty of plans, variables intended to be used only as an early (1957) proxy for psychological well-being. It is not clear what to make of this. Perhaps interest and disinterest reflect a biological arousal or temperamental factor. But we saw that interest predicted volunteering in 1975, and we also see it influencing psychological well-being and perceived physical health in 1992. The fact that the impact of these variables on psychological well-being in 2004 are almost completely eliminated when the 1992 measure is introduced does strongly suggest that whatever else the variables measure, using them as a proxy for psychological wellbeing is not unreasonable. The Wisconsin Longitudinal Survey is not a random sample, therefore it is possible that these findings are unique to this population. Considering the similarity of our findings to research conducted using a national, random sample (House, 1995), involving different age cohorts, this is unlikely. However, replication of the pathway we propose with a random or national sample would be a welcome addition, in particular our finding that the effects of volunteering are mediated by a sense of mattering. . As a new generation of Americans approach older adulthood, the search is on for ways to help them maintain the quality of their mental and physical health. This study demonstrates a direct, protective effect of volunteering on psychological well-being, and cautions against the assumption that simply encouraging people to adopt an active lifestyle is enough to promote positive mental health. There is no evidence that the mental or physical health of older adults benefit from participation in self-oriented activities beyond their possible contribution to social

28

integration. Therefore, while others may claim that, doing something is better than doing nothing, our research lends stronger support to the idea of doing well by doing good.

29

References Adelmann, P.K. 1994. Multiple roles and physical health among older adults. Research on Aging, 16: 142-166. Baruch, G. K. & Barnett, R.C. 1986. Role quality, multiple role involvement, and psychological well-being in midlife women. Journal of Personality and Social Psychology, 51: 578-585. Benyamini, Yael, Ellen L. Idler, Howard Leventhal, and Elaine A.Leventhal. 2000. Positive affect and function as influences on self-assessments of health: Expanding our view beyond illness and disability. Journal of Gerontology: Psychological Sciences. Vol. 55B: P107-P116. Cohen, Sheldon, William J. Doyle, Ronald B.Turner, Cuneyt M. Alper, & David P. Skoner. 2003. Emotional style and susceptibility to the common cold. Psychosomatic Medicine. 65: 652-657. Coser, L. (1974). Greedy institutions. New York: Free Press. Csikszentmihaly, Mihaly. 1975. Beyond Boredom and Anxiety. San Francisco: Jossey-Bass. Davidson , Richard J., Jon Kabat-Zinn, Jessica Schumacher, Melissa Rosenkranz, Daniel Muller, Saki F. Santorelli, Ferris Urbanowski, Anne Harrington, Katherine Bonus, and John F. Sheridan, 2003. Alterations in brain and immune function produced by mindfulness meditation. Psychosomatic Medicine. 65: 564-570. Elliott, Gregory C., Melissa F. Colangelo, & Richard J. Gelles. 2005. Mattering and suicide ideation: Establishing and elaborating a relationship. 68: 223-238. Aocial Psychology Quarterly.

30

Ellliott, Gregory C., Suzanne Kao, & Ann-Marie Grant. 2004. Mattering: Empirical validation of a social-psychological concept. Self and Identity, 3: 339-354. Fredrickson, Barbara L., & Marcial F. Losada. 2005. Positive affect and the complex dynamics of human flourishing. American Psychologist. 60: 678-676. Fredrickson, Barbara L., Michele M. Tugade, Christian E. Waugh, C.E., & G. Larkin. 2003. What good are positive emotions in crises? A prospective study of resilience and emotions following the terrorist attacks on the United States on September 11th, 2001. Journal of Personality and Social Psychology, 84: 365-376. Friedland, Robert P., Thomas Fritsch, Kathleen A. Smyth, Elisabeth Koss, Alan J. Lerner, Chien H. Chen, Grace J. Petot, & Sara M. Debanne. 2001. Patients with Alzheimers disease have reduced activities in midlife compared with healthy control-group members. Proceedings of the National Academy of Sciences 98(6): 3440-45. Goode, W. J. 1960. A theory of role strain. American Sociological Review, 25: 483-496. Gove, W. R. & Geerken, M. R. 1977. The effect of children and employment on the mental health of married men and women. Social Forces, 56: 66-76. Harris, Mary B. 1977. Effects of altruism on mood. Journal of Social Psychology. 91:3741. Holden, Gary. 1991. The relationship of self-efficacy appraisals to subsequent health related outcomes: A meta-analysis. Social Work in Health Care. 16:53-93. House, James.S. 1995. Americans changing lives: Waves I and II, 1986 and 1989. Ann Arbor, MI: Interuniversity Consortium for Political and Social Research.

31

House, James S. 2002. Understanding social factors and inequalities in health: 20th century progress and 21st century prospects. Journal of Health and Social Behavior. 43:125-143. Hunter, Jeremy P. & Mihaly Csikszentmihaly. 2003. The positive psychology of interested adolescents. Journal of Youth and Adolescence. 32:27-35. Keyes, Corey L. M. 2005. Mental illness and/or mental health? Investigating axioms of the complete state model of health. Journal of Consultling and Clinical Psychology. 73 539-548. Luoh, Ming-Ching; & A. Regula Herzog. 2002. Individual consequences of volunteer and paid work in old age: Health and mortality Journal of Health and Social Behavior, 43:490-509. Marks, S. R. 1977. Multiple roles and role strain: Some notes on human energy, time and commitment. American Sociological Review, 42: 921-936. McLeod, Carol R., & Stephen J. Vodanovich 1991. The relationship between self-actualization and boredom proneness. pp. 137-149 in Handbook of self-actualization. [Special Issue.] Journal of Social Behavior and Personality 6:137-146. Menaghan, E. G. 1989. Role changes and psychological well-being: Variations in effects by gender and role repertoire. Social Forces, 67: 693-714. Moen, Phyllis, Donna Dempster-McClain, & Robin. M. Williams, Jr. 1992. Successful aging: A life-course perspective on womens multiple roles and health. American Journal of Sociology, 97:1612-1638. Moen, Phyllis, Donna Dempster-McClain, & Robin. M. Williams, Jr. 1989. Social integration and longevity. American Sociological Review, 54:635-647.

32

Musick , Marc W., A. Regula Herzog , & James S. House. 1999. Volunteering and mortality among older adults: Findings from a national sample. The Journals of Gerontology : Psychological sciences and social sciences, 54B, S173-S180 . Musick, Marc W., & John Wilson . 2003. Volunteering and depression: The role of psychological and social resources in different age groups. Social Science and Medicine, 56:259-269. Oman, Doug, Carl E.Thoresen, & Kay McMahon. 1999. Volunteerism and mortality among the community-dwelling elderly. Journal of Health Psychology, 4:301-316. Piliavin, Jane Allyn. 2003. Doing well by doing good: Benefits for the benefactor. Pp. 227247 Keyes, Corey Lee M. & Jon Haidt (Eds.). Flourishing: The Positive Personality and the Life Well Lived. Washington, D.C.: American Psychological Association. Rosenberg, Morris & B. Claire McCullogh. 1981. Mattering: Inferred significance and mental health among adolescents. Research in Community and Mental Health. 2:163-182. Ryan, Richard M. & Edward L. Deci. 2001. On happiness and human potentials: A review of research on hedonic and eudaimonic well-being. Annual Review of Psychology. 52: 131166. Ryff, Carol D. 1989. The parental experience in midlife. Chicago: University of Chicago Press. Ryff, Carol D. & Corey L.M. Keyes, 1995. The structure of psychological well-being revisited. Journal of Personality and Social Psychology. 69:719-727. Sarbin, T. R., & Allen, V.L. 1968. Role theory. In G. Lindzey & E. Aronson (Eds.). Handbook of social psychology (2nd Ed.) , vol. 1, (pp. 487-567). Cambridge, MA:Addison-Wesley.

33

Sieber, S. D. 1974. Toward a theory of role accumulation. American Sociological Review, 39: 567-578. Sommers, Jennifer & Stephen J. Vodanovich 2000. Boredom proneness: Its relationship to psychological- and physical-health symptoms. Journal of Clinical Psychology. 56:149155. Springer, Kristen. 2001. Dimensions of psychological well-being at midlife: Results from the Wisconsin Longitudinal Study. Unpublished seminar paper, University of Wisconsin. Springer, Kristen, & Hauser, Robert. (Unpublished) The measurement and meaning of psychological well-being among aging adults. Steptoe, Andrew, Jane Wardle & Michael Marmot. 2005. Positive affect and health-related neuroendocrine, cardiovascular, and inflammatory responses. Proceedings of the National Academy of Sciences, USA. 102: 6508-6512. Thoits, P. A. 1995. Identity-relevant events and psychological symptoms: A cautionary tale. Journal of Health and Social Behavior, 36: 72-82. Thoits, P. A. 1992. Multiple identities: Examining gender and marital status differences in distress. Social Psychology Quarterly, 55: 236-2 56. Thoits, P. A. 1986. Social support and psychological well-being: Theoretical possibilities. In I.G. Sarason & B.R. Sarason (Eds.). Social support: theory, research, and application (pp. 51-72). Dordrecht, The Netherlands: Martinus Nijhoff Publilshers. Thoits, Peggy A., & Lyndi N. Hewitt. 2001. Volunteer work and well-being. Journal of Health and Social Behavior, 42:115-131. Van Willigen, Marieke. 2000. Differential benefits of volunteering across the life course. Journal of Gerontology: Social Sciences. 55B: S1-S11. 34

Van Willigen, Marieke. 1998. Doing good, feeling better: The effect of voluntary association membership on individual well-being. Paper presented at the annual meeting of the Society for the Study of Social Problems. Verbrugge, L. M. 1987. Role responsibilities, role burdens, and physical health. In F.J. Crosby (Ed.) Spouse, parent, worker: on gender and multiple roles (pp. 154-166). New Haven, MA: Yale Universtiy Press. Waldron, I. & Jacobs, J.A. 1989. Effects of multiple roles on womens health evidence from a national longitudinal study. Women and Health, 15: 3-19. Young, Frank W., & Nina Glasgow. 1998. Voluntary social participation and health. Research on Aging, 20:339-362.

35

Table 1a. Major independent and control variables. Means and standard deviations of continuous variables and percentages of categorical variables. Independent Variables Categorical (percentages): Female HS interest HS disinterest HS certainty of plans Married 75 Working 75 Continuous (means): HS support Visits w. friends in last 4 weeks, 1975 Family annual earnings, 1974/100 1970 Occ Ed score Duncan SEI of head of household in 1957 IQ in High School years of education number of kids 6-12, 1975 frequency of church attendance involvement in church-connected groups index of social participation, 1975 index of social participation, 1992 volunteer index, 1975 volunteer index, 1992 volunteer index, 2004 Index of integration, 1992 Central Tendency 53.6 47.2 9.1 74.7 89.3 76.2 3.07 5.32 166.21 387.91 348.79 102.09 13.69 1.26 4.81 .54 .86 1.44 1.28 1.79 1.50 7.27 Standard Deviation - - .95 4.87 112.46 286.78 233.76 14.62 2.30 1.08 1.58 .82 1.19 2.03 1.60 2.43 2.06 1.28 Valid sample size 6875 8170 8170 7785 6536 6546 8004 6520 6537 6596 6719 6875 6874 6875 6533 6529 6529 6593 6526 6601 6138 6127

36

37

Table 1b. Dependent variables. Means and standard deviations.

Independent Variables Psychological well-being, 1992 Psychological well-being, 2004 Self-reported health, 1992 Self-reported health, 2004

Mean 129.25 102.62 4.15 4.02

Standard Deviation 17.97 13.66 .679 .674

Valid sample size 6840 6151 6862 6255

38

Table 2. Regression of log (Base 10) of volunteering in 1975 and 1992 on predictor variables Cells: Unstandardized coefficient (standard error) ; standardized coefficient Volunteering 1975 Independent Variables Female Duncan SEI, head of household in 1957 HS support for education HS interest in school HS disinterest in school Certainty of 1958 plans IQ Education in years Number of children aged 6-12 in 1975 Married in 19751 Working in 19751 Church attendance, 1975 Family earnings, 1974 1970 occupation education score Involvement with church groups, 1975 Social activity score, 1975 .061*** Volunteering, 1975 R2 (Adjusted) ; df .156*** 14/5234 .131*** .107*** (.015) (.016) .056 .123 17/5029 B Std. error (.034) (.000) (.019) (.037) (.054) (.035) (.001) (.009) (.014) (.051) (.041) (.010) Beta .142 .038 .036 .032 -.027 .009 .012 .031 .226 .077 -.017 .145 .059 .095 B Volunteering 1992 Std. error Beta .062 -.006 .040 .007 -.034 .014 -.010 .123 -.053 -.015 .009 .081 .030 .069 .045

.329 *** .00019** .044* .076* -.100@ .023 .00094 .0150 .245 *** .280 *** -.049 .106 ***

.154*** (.039) -.00003 .053* .017 -.138* .040 -.00071 (.000) (.022) (.042) (.061) (.040) (.0008)

.065*** (.011) -.061*** (.017) -.058 .027 (.057) (.046)

.063*** (.012) .00024*** (.000) .00032*** (.000) .068 *** (.023)

.00062*** (.000) .00038*** (.000)

Significance: @ = p< .10; * = p < .05; ** = p <.01; *** = p < .001 1 Using work and marital status in 1992 for those analyses makes no difference.

39

Table 3. Partial correlations of volunteering with psychological well-being and health, holding constant gender and social participation and partial correlations of social participation with psychological well-being and health, holding constant gender and volunteering. Concurrent and lagged correlations are shown. Independent Variable Volunteering, 1975 Social participation, 1975 Volunteering, 1992 Social participation, 1992 Volunteering, 2004 Social participation, 2004 Well-being, 1992 .096*** .034** .140*** .055*** .146*** .035* .143*** .039** Well-being, 2004 Self-reported health, 1992 .045*** .010 .067*** .024* .058*** .014 .081*** .026@ Self-reported health, 2003

Significance: @ = p< .10; * = p < .05; ** = p <.01; *** = p < .001

40

Table 4: Means of two measures of well-being as a function of two measures of volunteering. Measure of well-being 1992 Diversity of Volunteering in Previous Wave None 1 Organization 2+ Organizations Consistency of Volunteering Over time None 1 Wave 2 Waves 3 Waves 4.07*** 4.17 4.21 125.77*** 129.00 132.71 3.92*** 4.02 4.05 4.11 99.70*** 101.57 104.05 106.34 Self-reported health (5-pt) 4.12*** 4.17 4.21 Psychological Well-being 127.49*** 129.67 132.37 Measure of well-being 2004 Self-reported health (5-pt) 3.99*** 4.07 4.07 Psychological Well-being 101.27*** 103.76 105.36

Indicators of significance: *** = p < .001

41

Table 5a. Regressions of psychological well-being and self-reported health in 1992 (square root X 10) on dummy volunteering variables. Cell entries are unstandardized regression coefficients (standard errors). Dummy Variables for Volunteer Diversity One organization 1975 Two organizations 1975 Three or more organizations 1975 Adj. R2 ; df Dummy Variables for Volunteer Consistency One wave Two waves Adj. R2 ; df Psychological Well-being 2.181*** 4.722*** 5.269*** .010 (.505) (.661) (.960) 3/6494 Self-Reported Health .146** (.049) .220*** (.064) .257** (.093) .002 3/6858

Psychological Well-being 3.499*** (.551) 7.129*** (.585) .023 2/6495

Self-Reported Health .257*** (.054) .346*** (.057) .006 2/6516

Indicators of significance: ** = p < .01;

*** = p < .001

42

Table 5b. Regression of psychological well-being and self-reported health in 2004 (square root X 10) as a function of diversity of volunteering in 1992 and in 2004, and consistency of volunteering, controlling for the level of the dependent variable in 1992. Cell entries are unstandardized regression coefficients (standard errors). a Psychological Well-being Dummy Variables One organization 1992 Two organizations 1992 Three or more organizations 1992 Health measure, 1992 R2 ; df One organization 2004 Two organizations 2004 Three or more organizations 2004 Health measure, 1992 R2 ; df One wave Two waves Three waves Health measure, 1992 R2 ; df .029 3/5054 Step one 2.489*** (.441) 3.865*** (.533) 4.420*** (.641) Step two .687 @ (.356) 1.304 * (.431) 1.446 ** (.517) .458 *** (.008) .016 3/5344 .366 4/5343 .003 3/5425 Self-reported Health Step one .184** (.058) .232*** (.071) .164@ (.085) Step two .091@ .105@ .012 (.051) (.062) (.074)

5.234*** (.127) .241 4/5424

3.104*** (.429) 4.538*** (.545) 4.712*** (.717)

1.477*** (.345) 2.299*** (.439) 2.851*** (.576) .455*** (.008) 4/5280

.183*** (.057) .216** (.027) .411*** (.036)

.099* (.050) .093 (.063) .277*** (.084) 5.225 (.128) 4/5338

.020

3/5281

.371

.004

3/5339

.242

1.888*** (.565) 4.376*** (.560) 6.659*** (.592)

.838@ (.457) 1.565*** (.456) 3.058*** (.484) .448*** (.009) .366 4/5053

.258*** (.075) .334*** (.074) .478*** (.079)

.163* (.066) .172** (.065) .260*** (.069) 1.971*** (.050)

.007

3/5109

.239

4/5108

Indicators of significance: @ p < .10 ;

* p < .05 ; ** p < .01; *** p < .001

43

Table 6: Regression of psychological well-being on control variables, combined measure of volunteering, and the interaction of integration and volunteering in 1992. Cell entries are unstandardized coefficients (standardized errors) a Dep. Var.: Psychological Well-being 1992 Variables Controls: b Female HS interest HS disinterest HS certainty of plans HS support Education Married 1975 Working 1975 Visits w. friends, 1975 Family earnings, 1974/100 Social participation index Combined measure of volunteering 1975-1992 (VOL) Integration Integration X VOL R2; df
a c

Step 1 1.915*** 1.876** -2.325** 2.947*** 1.113** .378 ** 2.299** 2.097*** .216*** .945*** .224 1.514*** (.582) (.611) (.886) (.578) (.317) (.133) (.847) (.656) (.051) (.232) (.208) (.215)

Step 2 1.264* 1.750** -2.397** 2.606*** .946** .339** 1.431@ 1.698** .107* .755** .078 3.575*** 3.446*** -.347* (.571) (.597) (.866) (.566) (.310) (.130) (.830) (.642) (.052) (.228) (204) (.451) (.289) (.149) 19/5082

.067***

17/5084

.109***

Indicators of significance: *** = p < .001; ** = p < .01; * = p < .05; @ = p <.10. Controls shown were significant when added at p<.05; other controls(n.s.): Duncan SEI of head of household in 1957, IQ in H.S., number of kids 6-12, involvement in church-connected groups, frequency of church attendance, and the index of social participation, all measured in1975. c Significance shown is for significance of change in R2.
b

44

Table 7: Regression of psychological well-being in 2004 on control variables, combined measure of volunteering, integration, their interaction, volunteering in 2004, and mattering. Cell entries are unstandardized coefficients (standardized errors) a Dep. Var.: Psychological Well-being 2004 Variables Controls: b Female HS interest HS certainty of plans HS support IQ in high school Years of Education Church attendance 75 Visits w. friends, 1975 Family earnings, 1974/100 Combined measure of volunteering 19751992 (VOL) Integration Integration X VOL Volunteering in 2004 Well-being, 1992 Mattering R2; df
c

Step 1 1.651*** 2.031*** 1.737*** .634* -.038* .724*** .294* .206@ .802*** (.493) (.520) (.491) (.269) (.016) (.110) (.149) (.054) (.201)

Step 2 1.489** 1.997*** 1.743*** .630* -.035* .692*** .284@ .094* (.493) (.519) (.489) (.269) (.016) (.110) (.149) (.045)

Step 3 .718@ 1.022* .691@ .219 -.032* .579*** .253* .061@ .472** (.406) (.428) (.404) (.221) (.013) (.091) (.122) (.037) (.165)

Step 4 .883* (.385) .710@ (.403) .588 (.380) .023 (.209) -.028* (.013) .532*** (.085) .236* (.116) .027 (.035) .302@ 1.127 .616** -.192@ .206* .370*** 1.342*** (.156) (.751) (.201) (.100) (.087) (.010) (.060)

.773*** (.201) 3.169*** (.966) 2.314*** (.256) -.395** (.129) .521*** (.106)

3.283*** (.969) 2.342*** (.257) -.372** (.130)

1.496@ (.797) .852*** (.214) -.226* (.107) .378*** (.087) .434*** (.010)

.109***

19/3817

.114***

20/3816

.399***

21/3815

.469*** 23/3813

Indicators of significance: *** = p < .001; ** = p < .01; * = p < .05; @ = p <.10. Controls shown were significant when added at p<.05; other controls(n.s.): High school disinterest, Duncan SEI of head of household in 1957, work status, married, number of kids 6-12, involvement in church-connected groups, and the index of social participation (all measured in1975), volunteer identity 2004. c Significance shown is for significance of change in R2.
b

45

Table 8: Regression of self-reported health in 1992 on control variables, combined measure of volunteering, integration, their interaction, and psychological well-being. Cell entries are unstandardized coefficients (standardized errors) a Dep. Var.:self-reported health in 1992 Variables Controls: b Female HS interest HS disinterest HS certainty of plans Duncan SEI of head of household in 1957 Years of Education Church attendance 75 Family earnings, 1974/100 Combined measure of volunteering 19751992 (VOL) Integration Integration X VOL Body mass Smoke now? Packs a day Exercise Well-being, 1992 R2; df c
a

Step 2 .012* .014* -.026** .011@ (.006) (.006) (.009) (.006)

Step 3 .003 .012* -.021* .013* .000023* .004** .002 .006** .020@ .019*** -.019* -.008*** -.026*** -.006** .018*** (.006) (.006) (.009) (.006) (.000) (.001) (.002) (.002) (.011) (.003) (.001) (.001) (.007) (.002) (.001)

Step 4 .001 (.006) .008 (.006) -.016@ (.008) .007 (.003) .000026* (.000) .003** (.001) .002 (.002) .004@ (.002)

.000024* (.000) .007*** (.001) .003@ (.002) .007** (.002)

.024* (.011) .022*** (.003) -.003* (.002)

.013 (.011) .012*** (.003) -.002@ (.001) -.007*** -.030*** -.006* .015*** .002*** (.001) (.007) (.002) (.001) (.000)

.049***

19/4806

.124***

23/4802

.166*** 24/4801

Indicators of significance: *** = p < .001; ** = p < .01; * = p < .05; @ = p <.10. Controls shown were significant when added at p<.05; other controls(n.s.):, HS support, IQ in H.S., Marital status, work status, number of kids 6-12, involvement in church-connected groups, the index of social participation, Visits w. friends (all measured in1975). c Significance shown is for significance of change in R2.
b

46

Notes 1. For the 1975 measures, scores range from 0-10 before logging; for 1992, they range from 020. Distributions were highly skewed. 2. The measures of intensity that were devised were too highly correlated with the measure of diversity. 3. Reviewers of this manuscript pointed out that they are not adequate controls. Nevertheless, I retain them because they do indeed provide some indication of psychological well-being in youth. 4. Springer(2002); Springer & Hauser (in preparation) have found, using confirmatory factor analysis, that the six Ryff scales do not form distinct factors in the WLS sample. Rather, these four scales essentially load on one factor. The alpha reliability of the measure used here is .909. 5. The item is so skewed that analyses employing the commonly used dichotomy of excellent + good vs. all other responses has very little power. Although the number of illnesses (of 17) and symptoms (of 27) reported are available in the WLS, volunteering appears not to be related to these measures. Similarly, relatively few people have died among those who responded in 1992; thus mortality cannot be used as a dependent variable. 6. This measure is a trichotomy based on a continuous measure of the number of visits with friends in the past four weeks. A score of 1" indicates 0-2 visits; 2" = 3-5, 3 = 6 or more. 7. All five variables were positively related to the measures of mental and physical health; analyses were also conducted using the five variables separately. Reviewers questioned the inclusion of the rural-urban variable, but in Wisconsin, where the majority of respondents still live, the lowest level on this measure a town size of up to 5000 really reflects considerable isolation. Over 1500 in the sample live in such communities. Results using an index that omits the rural-urban variable show similar, although weaker, findings. 47

8. These are measured on ordinal scales; thus the means provided in Table 1a cannot be interpreted in times per week or month. 9. This score is the percentage of persons in the 1970 Census in an occupation/ industry/classof-worker category who completed one year of college or more. A housewife or unemployed person was coded for a previous job, if any. Individuals who had never been in the labor force were coded as missing data. 10. There is a resultant decrease in work with youth groups and the PTA, although the overall level of volunteering remains the same as in 1975 or increases somewhat. 11. Women score higher on the volunteering index; men score higher on the self-oriented index, in part because it includes sports participation. 12. Since the mean and variance of volunteering in 1992 are greater than those in 1975, due to the difference in scoring the separate items, this has the effect of weighting 1992 volunteering somewhat more heavily. When the two scores are standardized and then summed, the resulting variable behaves essentially the same. If instead of using the sum, the two variables are both entered into the equation, the impact of 1975 volunteering is not significant while 1992 is. Thus weighting it somewhat more seems suitable. The sum of the two measures seems to more appropriately represent the known increase in well-being due to consistency demonstrated in earlier analyses. 13. A question asked by a reviewer how housewives and the unemployed were coded on the variable Occupational Education 1970" led us to remove that predictor from the model. These respondents had been coded as missing data, and thus were not included in the analyses. This allowed for about 300 mainly female respondents to be retained in the remaining analyses, which do not look very different but are presumably more representative of the total sample. 48

14. A reviewer asked why we never tested for the mediating effect of perceived physical health on psychological well-being. That is, why was it not considered as a possible path by which volunteering might affect well-being? Although we have no theoretical reason why this might be the case, we did add perceived health to the above equations. While it does add to the overall prediction of psychological well-being a reasonable expectation the coefficients for volunteering and the interaction of volunteering with integration do not decrease when it is added. 15. Logistic regressions were also run on a dichotomous measure, but the sample of individuals with poor health was simply too small.

49