Volunteering and the long civic generation

Nonprofit and Voluntary Sector Quarterly http://nvs.sagepub.com/ Volunteering and the Long Civic Generation Kristin A. Goss Nonprofit and Voluntary Sector Quarterly 1999 28: 378 DOI: 10.1177/0899764099284002 The online version of this article can be found at: http://nvs.sagepub.com/content/28/4/378 Published by: http://www.sagepublications.com On behalf of: Association for Research on Nonprofit Organizations and Voluntary Action Additional services and information for Nonprofit and Voluntary Sector Quarterly can be found at: Email Alerts: http://nvs.sagepub.com/cgi/alerts Subscriptions: http://nvs.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Citations: http://nvs.sagepub.com/content/28/4/378.refs.html Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic GossGeneration ARTICLES Volunteering and the Long Civic Generation Kristin A. Goss Harvard University By virtually every conceivable measure, civic participation is on the decline in America. Volunteering is one important exception. An analysis of a newly available archive of national surveys finds that the frequency with which Americans volunteer has increased 20% since the mid-1970s. However, nearly all of that increase is concentrated among older Americans, who are volunteering twice as frequent in the late 1990s as their sameaged predecessors did in the 1970s. Meanwhile, volunteering has actually decreased among middle-age adults, who once were the voluntary sector’s most reliable source of donated labor. The reasons for increased volunteering among seniors remain elusive. Tests of various hypotheses, from improved health and financial conditions to increased spare time, do not explain the explosive increase. Nonetheless, it is clear that a powerful and mysterious force is pushing seniors toward greater volunteer involvement, and nonprofit groups should tap into this particularly civic age group before the Indian summer of volunteering reaches its end. Americans are less civically and socially engaged today than at any time in recent memory. Over the past generation, they have become less likely to attend public meetings, sign petitions, vote, join voluntary associations, give generously to charity, trust one another, or even hold dinner parties and play cards with friends (Putnam, in press). Note: An earlier version of this article was presented at the annual ARNOVA conference in November 1998 in Seattle, Washington. The project from which this research was drawn was supported by grants from the Carnegie Corporation of New York, Lilly Endowment, John D. and Catherine T. MacArthur Foundation, C. S. Mott Foundation, Pew Charitable Trusts, Rockefeller Brothers Fund, Rockefeller Foundation, and Surdna Foundation. The material is based on work supported under a National Science Foundation Graduate Fellowship. The author would like to thank Chris Callahan, Jim Crimmins, Marty Horn, and Doug Hughes of DDB Needham; Sid Groeneman of Market Facts; Kate Fitzpatrick, Louise Kennedy, and Tom Sander of the Saguaro Seminar: Civic Engagement in America; Dalene Stangl of Duke University; Steve Yonish, of the University of Wisconsin and Harvard Business School, for his heroic efforts on the data set; and Steven Rathgeb Smith and the anonymous reviewers for their helpful comments. A particular note of thanks goes to Harvard Professor Robert Putnam, who first noticed the key puzzles explored in this study, discovered and named the long civic generation, and provided the funding and intellectual leadership that made this research project possible. Nonprofit and Voluntary Sector Quarterly, vol. 28, no. 4, December 1999 © 1999 Sage Publications, Inc. 378-415 378 Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 379 Amid pervasive signs of declining social capital, there is one broad and important exception. Americans today volunteer substantially more often than they did a generation ago. Although the share of adults who volunteer in any given year has remained steady, Americans volunteer, on average, 20% more times per year now than they did in the late 1970s, according to a newly available archive of annual national surveys. To my knowledge, these surveys represent the longest consistent data series on volunteer patterns ever amassed, and enable researchers for the first time to draw reliable conclusions about volunteering in recent decades. This article describes some of the more intriguing developments captured in the 1 data, the findings of which have never before been published. The rise in volunteering that the surveys document is perhaps the most intriguing puzzle. In many ways, this increase should not have happened. For example, we would not have predicted a rise in volunteering during a time when married women—the stalwart volunteers of yesteryear—have been flocking into full-time, paid jobs outside the home. We would not have predicted a rise in volunteering during a period when other measures of charitable spirit—particularly, contributions as a share of income—have been eroding (Kaplan, 1998). We would not have expected a rise in volunteering during a time when the traditional recruiting grounds for volunteers—churches and clubs—have been losing members (Putnam, in press). What is more, several demographic characteristics traditionally correlated with volunteer effort—notably, being married and having children—have become less com2 mon during this time. In addition, even demographic correlates that have become more common—such as having a college degree and a good income— have not become pervasive enough to account for such a large increase in volunteerism. So, what explains the large increase in volunteer effort in America over the past generation? Cohort analysis helps to answer the question. Seniors—those who are at least 60 years old—are responsible for virtually all of the increase in volunteer effort. Indeed, today’s seniors volunteer nearly twice as often as seniors did in 1975. Not only are seniors today more committed to volunteering than were their same-age predecessors in previous decades, but today’s seniors are also more committed than are today’s middle-age and young adults. These findings are striking, and, as demonstrated in the next section, go against the received wisdom about the life cycle of volunteering. Indeed, the findings provide additional evidence that there is a long civic generation of exceptionally participatory individuals (Putnam, 1996, in press). These civic torchbearers, born roughly between 1910 and 1930, not only volunteer more often than younger adults do, but they also attend church more frequently, go to more club meetings, and even hold more dinner parties. Although seniors’ engagement in these activities is going down along with everyone else’s, their commitment to volunteering continues to rise. The most vexing question is why. Despite an exhaustive search for an explanation of older people’s increased commitment to volunteering, a Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 380 Goss definitive answer remains elusive. I have been able to rule out, with varying degrees of certainty, several possible explanations of seniors’ extraordinary new efforts. These include improved health and economic circumstances, more free time, increased psychic and actual engagement in public life, the desire to fill a void left by family disintegration, and even disillusionment with government social programs. The only explanations that I have are partial. First, advances in education among older Americans are surely fueling some of the increase, although it is not even close to all of it. The second explanation is based on gender. Much of the increase in seniors’ volunteering is due to older women. For reasons that I have been unable to isolate, females 60 and older have become the vanguard of voluntary action. The growth in their volunteer output is not a byproduct of any identifiable development in their lives, but it is very real and quite profound. This article is organized as follows. First, findings from previous studies of demographic and social determinants of volunteering are summarized. Knowing which factors boost or suppress the donation of one’s time, I constructed hypotheses about the trends that one would expect to see in volunteering, given the changes over time in the relevant demographic, economic, and social conditions. Second, the data set and the methods of analysis employed in this article are described. Third, the findings with respect to the long civic generation and, particularly, the women therein are presented. Finally, the article concludes with some thoughts about this study’s implications, both for the nonprofit world and for academic research. THE VOLUNTEERING LITERATURE Studies of volunteering have reached a general consensus on three points. First, people from dominant-status groups are far more likely to donate their time than are people from low-status groups (for a review, see Smith, 1994). Second, middle-age people are more likely to volunteer than are young and older people. Third, people who are involved in community-based organizations, such as churches and civic clubs, are much more likely to volunteer than are socially isolated individuals. DOMINANT-STATUS GROUPS Perhaps the strongest and most robust predictor of volunteerism is educational attainment. A 1996 survey by the Gallup Organization for Independent Sector, for example, found that college graduates were twice as likely as people with a high school education or less to have volunteered in the past year—71% compared to 36% (Hodgkinson & Weitzman, 1996). Education remains a substantively and statistically significant determinant of volunteering, even after other factors correlated with education (income, race, etc.) are controlled. Having a good income also has been found to be a consistent pre- Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 381 dictor of volunteering, even when its correlates are controlled (Auslander & Litwin, 1988; Freeman, 1996; Menchik & Weisbrod, 1987; Romero, 1987; Sasser, 1983). Two explanations have been offered for the income effect. Wealthy people contribute more to charity, which in turn draws them into donating time (Berger, 1991). In addition, wealthy people, on average, may have more verbal and writing skills and be more confident, lowering the psychological barriers to volunteer involvement (Wilson & Musick, 1997). Finally, several studies have found that holding a paid job is associated with the likelihood of volunteering (Hodgkinson & Weitzman, 1996; Segal, 1993; but see Dempsey, 1988). However, studies also find that employed people who work long hours tend to donate fewer hours to charity than do employed people with part-time or otherwise reasonable work schedules (Clotfelter, 1985; Segal, 1993; Statham & Rhoton, 1986). There are a number of possible explanations for the seemingly paradoxical finding that employment increases the likelihood of volunteering but decreases the hours donated. For one, people who are employed join professional associations that require volunteer labor. In addition, people who work are exposed to volunteer recruitment efforts by coworkers. Yet, in both cases, their time is more constrained than is that of someone who is not expected to be on the job 8 hours a day. Although social-class variables are strongly linked to volunteering, three other demographic characteristics—gender, race, and marital status—have a less straightforward effect. In most surveys, women volunteered more than men (although men belonged to more voluntary associations than did women); Whites volunteered more than Blacks; and married people volunteered more than single, divorced, or widowed people. However, most scholars who have looked for independent effects have found them only rarely. In short, most evidence suggests that gender, race, and marital status do not predict voluntary participation once other variables—notably, education, income, and employment status—are statistically controlled. The cumulative evidence here suggests two hypotheses for this study’s longitudinal analysis. First, we should expect volunteering to have increased among all age groups, because people of all ages (young adults, middle-age adults, etc.) are both better educated and wealthier than were their counterparts in the 1970s. Second, we should expect that the increase in volunteer effort that is theoretically possible in a wealthier, better educated population has been mitigated by the increase in the work hours of women and middleage professional men over the past 25 years (Robinson & Godbey, 1997), and by the decline in club membership and church attendance (Putnam & Yonish, 1998). Third, we expect that the percentage of working-age people who volunteer has increased, propelled by the influx of women into paid employment (a known correlate of volunteer likelihood) and by the improved socioeconomic status of the public as a whole. In short, because of the offsetting effects of socioeconomic status and increased work hours among key segments of the population, we expect no more than a modest increase, if any, in volunteer effort. On the other hand, because of improved socioeconomic status and the Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 382 Goss fact that employment provides new avenues for volunteering, we expect a jump in the simple fraction of people who donate time each year. VOLUNTEERING AND THE LIFE CYCLE A wealth of research concludes that volunteerism is highly sensitive to an individual’s place in the life cycle. For example, numerous studies have found that the likelihood and breadth of volunteering increase dramatically as young adults marry and have children (although the onset of volunteering may not occur until the children have reached school age) (Berger, 1991; Clotfelter, 1985; Freeman, 1996; Knoke & Thomson, 1977; Segal, 1993; Wilson & Musick, 1997; Wilson & Musick, 1998a). The most recent Gallup-Independent Sector survey (1996) nicely illustrates the typical bell-shaped pattern. Volunteer participation rates, by age cohort, are shown in Figure 1. The heightened volunteer involvement of parents with children at home is attributable to any number of factors. These include the demands that youth organizations and schools place on parents to serve as troop leaders, coaches, board members, and so forth; the fact that parents of young children may be setting down roots in the neighborhood and feel a greater stake in its future; and the fact that parents may step up church involvement as part of the childrearing process. Whatever happens to propel young parents into the volunteer labor force, there is some early evidence that their interest wanes as the children leave home and retirement dawns (Menchik & Weisbrod, 1987). Because a rising fraction of young adults is getting married later or not at all, and because they are having fewer children, we expect the life-cycle effect to decline in importance over time. CLUBS AND CHURCHES: THE ROUTES TO VOLUNTEERISM As suggested above, volunteering has traditionally been associated with membership in local organizations. Civic groups, fraternal lodges, booster clubs, recreational associations, and churches all rely on volunteers to organize their activities and serve on governing boards. Likewise, these organizations have traditionally been fertile ground for outside individuals and groups trying to recruit volunteers. The Gallup-Independent Sector surveys provide a wealth of evidence for the centrality of associations and churches to volunteerism. For example, in 1995, nearly half (46%) of all volunteers got involved “through participation in an organization.” This is far more than the 25% who had sought out the activity on their own (Hodgkinson & Weitzman, 1996). Among those who learned about their volunteer activities through participation in an organization, the survey found that more than half (60%) said that the organization was a religious institution (Hodgkinson & Weitzman, 1996). Moreover, among people who had been asked to volunteer, about one third (31%) said that they were asked by “someone in a religious institution”—the second most popular Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 383 Figure 1. Volunteering and the Life Cycle Source: Adapted from data in Hodgkinson & Weitzman, 1996. answer next to “a friend” (Hodgkinson & Weitzman, 1996). Based on the findings above, it is not surprising that members of religious and nonreligious organizations are three to four times more likely to volunteer than are people who did not belong to any organization. Among members of nonreligious organizations, 73% donated time in 1995; among religious-group members, 55%; and among nonmembers, 19% (Hodgkinson & Weitzman, 1996). Rates of volunteering surpassed 80% among members of service clubs, alumni organizations, civic associations, professional organizations, charitable organizations, and religious groups (Hodgkinson & Weitzman, 1996). The correlation between organizational membership and volunteering holds even when other factors (for example, education, income, gender, work status, and age) are controlled. Associational (including church) involvement is a strong and statistically significant predictor of the likelihood of being asked to volunteer (Berger, 1991), of the range of volunteer activities performed (Wilson & Musick, 1998b), of commitment to volunteering over time (Wilson & Musick, 1998a), of hours donated (Jackson, Bachmeier, Wood, & Craft, 1995), and of volunteer tasks performed (Amato, 1990). Several analyses have suggested that organizational participation in one’s youth—belonging to a scouting group or a school club, for example—is a strong predictor of volunteering in one’s adulthood, all else being equal (Hodgkinson & Weitzman, 1996; Wilson & Janoski, 1995). Because the surveys from which my data are based did not ask respondents whether (or how often) they had been asked to volunteer, or about their youthful activities, I have no direct way of testing the effects of these factors on observed trends in volunteering. However, I do know from the data that club membership and church attendance are down among all age groups. Based on those findings, one would expect volunteering—both the frequency and the overall rate—to have declined as well. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 384 Goss In the next section, I test a range of hypotheses that stem from these findings. In particular, these tests illuminate whether trends in the known correlates of volunteering—having a high socioeconomic status; being in the middle stages of the life cycle; and belonging to civic, social, and religious organizations—explain changes in voluntary effort over time. DATA AND METHODS The findings in this article are based mainly on a proprietary archive of national household survey data going back nearly a quarter of a century. The Life Style surveys were commissioned by the advertising agency DDB Needham and have been conducted each year since 1975 by the commercial polling company Market Facts. Because the surveys are intended for use by commercial clients, many of the questions center on consumer tastes—in automobiles or laundry detergent, for example. However, the surveys also contain a wealth of information about Americans’ recreational, social, and political activities, as well as about their friendships, family life, and attitudes toward society at large. These data, coupled with information about respondents’ demographic characteristics and labor-market experience, make it possible to put together a picture of American volunteering over nearly a quarter of a century. One is able to describe how the typical volunteer has changed over time, how different life circumstances affect voluntary output, and how the social context of volunteerism has evolved since the mid-1970s. The Life Style surveys have several noteworthy advantages over the other volunteering surveys. For one, unlike the periodic surveys conducted by the federal government, the DDB Needham volunteerism questions have been asked in precisely the same way in each year. Neither the definition of volunteer activity nor the memory prompts have changed. In addition, as far as researchers can ascertain, the procedures by which data have been collected have not been altered (Putnam & Yonish, 1998). Second, unlike the biennial surveys conducted by the Gallup Organization for Independent Sector, the DDB Needham volunteering data were collected every year. The DDB Needham data were collected over a far longer period—24 years and counting, compared to just 8 years for Gallup-Independent Sector surveys. Although the Life Style data are a rich and promising new source of information, they have certain limitations. For example, the sample slightly under represents racial minorities, single people, and people without a high school degree. Because these groups tend to volunteer less, there is a chance that their underrepresentation is causing an exaggeration in the level of volunteering observed in any given year. However, so long as these groups are underrepresented by about the same amount in each survey year, this sampling flaw should not affect the variable of greatest interest here—change in voluntary output over time. Putnam and Yonish (1998) find no evidence that the less educated or racial minorities or singles have become proportionately scarcer Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 385 in the samples over time. The surveys have tended to overrepresent homeowners, but Putnam and Yonish (1998) see no increase in their overrepresentation in more recent survey years. A second limitation of the data is that, until 1985, the surveys reached married people almost exclusively. For purposes of this study, a midcourse change in the sampling frame could, in theory, pose a significant problem, particularly when the two samples differ along a dimension (marital status) that some studies have found to covary with volunteer activity. Being conscious of this problem, I have examined separately volunteering by married people and by single people during the period from 1985 to 1998. I found virtually no difference between the two groups in the level of volunteer output in any given year. Therefore, the married and single trend lines are virtually indistinguishable. However, when married and single volunteering is broken down further by age group, I do find modest differences. Younger married adults volunteer more often than do younger single adults. However, for older adults, the situation is reversed. Single seniors volunteer more often than do married seniors. This is an issue that would theoretically affect how one is to interpret the findings, as will become clear. However, even after correcting for the change in the sampling frame, the basic results are unaltered. The third limitation—and potentially the most serious—is that the DDB Needham surveys are not based on random samples of Americans. Rather, they are based on a form of quota sampling called mail panel. Unlike a panel survey, in which the same respondents are requestioned regularly over many years, a mail panel consists of respondents who are asked a battery of questions over the course of 1 year, with a new panel being selected the following year. Respondents are selected as follows. Market Facts obtains lists of names, addresses, and, occasionally, demographic characteristics of a large number of people. A sample of those people is invited to agree to respond periodically to mail and phone surveys about products, services, and, occasionally, public 3 issues. The rate of favorable response varies by demographic group, but it is largest among middle-class, middle-age, middle Americans. Demographically representative samples are drawn from among those who agreed to be contacted. Those people receive the DDB Needham Life Style surveys (as well as hundreds of others). The response rate is 70% to 80%, which is higher than that typically observed for random samples. Because there are so many opportunities for sampling bias, which would undermine the credibility of any results drawn from the surveys, extensive validity tests have been run on the DDB Needham data (Putnam & Yonish, 1998). These tests involved comparing the DDB Needham results of certain public-opinion questions to the results found by two highly reputable random-sample surveys—the General Social Survey and Roper Social and Political Trends. These tests found high (and sometimes near perfect) agreement between DDB Needham surveys and the General Social Survey and Roper Social and Political Trends in three crucial regards. These include the distribution of responses on comparable questions, the trends in the variables over time, and the demographic correlates of Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 386 Goss the answers given (for a fuller description of the results, see Putnam & Yonish, 1998). I am therefore quite confident in the DDB Needham data and in the validity of the results presented here. A final limitation surrounds the volunteering question itself. The DDB Needham survey was intended to gather a wide array of information about Americans’ attitudes and behaviors, and in exchange for that breadth, there is little depth on any one aspect of American life. Most notably, for present purposes, the surveys ask only about the number of times that respondents volunteered, and not how many hours that they spent doing so. In principle, there should be a correlation between times and hours. However, the data do not allow one to measure that correlation, to assess trends in hours volunteered, or to describe differences in volunteer hours across demographic groups. Likewise, the surveys do not ask respondents about the types of groups for which they volunteer. The bulk of the statistical analysis in this article is descriptive. First, it seeks to understand what has been happening with volunteering over time. Are people becoming more likely, or less likely, to volunteer? Is volunteer output up or down? Which population groups are responsible for the changes, if any? Besides charting these trends, I use ordinary-least-squares regression analysis to understand why volunteering has become more frequent since the 1970s. Although the DDB Needham data, in principle, permit the investigation of many aspects of volunteer activity, I focus on the volunteering over time questions because this data set has a unique leverage on finding the answers. FINDINGS Over the past quarter century, and contrary to expectation, the share of the American population that volunteers at least once per year has remained remarkably flat. Except for a slight uptick in very recent years, the volunteer participation rate has hovered around 50%, as Figure 2 shows. In addition, the number of times that the typical American volunteers has likewise held steady. Today, the median number of volunteer activities is two per year, the same as in the 1980s and the 1970s. At first glance, one might think that Americans are no more or no less committed to helping others today than they were a generation ago. However, these statistics hide a startling development. The average number of volunteer episodes is up dramatically, a full 20% since the mid-1970s. During those years, Americans volunteered about six times per year. By the mid-1990s, that number was closer to seven or eight, as Figure 3 shows. Together these findings—same volunteer participation rate and median number of episodes, and far greater mean number of episodes—suggest that some subset of Americans is intensifying its volunteer effort. In addition, these people are a big enough and dedicated enough group to be driving a noticeable increase in overall volunteer effort. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 387 Figure 2. Volunteer Participation Rate Source: DDB Needham Life Style Surveys. So, who is that subgroup? Given other evidence of a strong generational component to civic activity, I examined whether any particular birth cohort may be responsible for the increase in volunteer activities. To facilitate the presentation, following Putnam (in press), I divided respondents into the following four birth cohorts: • • • • Civics: people born before 1930 Silents: the Silent Generation, born between 1930 and 1945 Boomers: the Baby Boom Generation, born between 1946 and 1960 Busters: the Baby Bust Generation, born after 1960 As Figure 4 shows, three of the four birth cohorts (Civics, Boomers, and Busters) have increased their volunteering over the past quarter century. The Silents, on the other hand, have posted a modest decline. In some respects, these findings are not surprising. Over the 24 years that the survey has been done, the Boomers (age 15 to 29 in the survey’s first year) have gotten married and had children—two factors expected to increase their voluntary activity. The post-Boomers (who were 20 or younger in the first year that they were surveyed) have likewise come of age. They have finished their schooling, gotten jobs, and perhaps gotten married, bought a house, and had children. Meanwhile, the Silents (age 30 to 45 in the survey’s first year) have seen their children grow up and leave home, freeing this generation of the volunteer responsibilities that accompany parenthood. What is less clear is why the oldest generation—the Civics—have seen such a dramatic increase Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 388 Goss Figure 3. Frequency of Volunteering Source: DDB Needham Life Style Surveys. Figure 4. Frequency of Volunteering by Birth Cohort Source: DDB Needham Life Style Surveys. in volunteer intensity. These people, who were older than 45 in the survey’s first year, have gotten more involved in volunteer activity with nearly every passing year. In 1998, the survey’s most recent year, these people—now in their 70s and early 80s—were volunteering an average of 12 times a Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 389 year—twice as often as they volunteered in 1975 and far more than the younger generations volunteer today.4 AGE COHORT ANALYSIS To get around the problems posed by generational analysis, I examined volunteer patterns by age cohort over time. This method implicitly controls for life-cycle effects by comparing same-age people in different years. Here, the role of the long civic generation in boosting volunteerism becomes clear. As Figure 5 shows, since 1975, there has been a dramatic increase in the frequency of volunteering by seniors 60 years and older. They have nearly doubled their volunteer output, from 6 activities per year to 11. On the other hand, volunteer output by middle-age people (30 to 59) has hovered around seven times per year for the past two decades, and output by young adults (under 30) has hovered around four times per year. In other words, seniors have become twice as altruistic as they used to be, whereas everyone else has stayed the same. What is more, seniors are responsible for nearly all of the increase in voluntary activity observed over the past quarter century. Another simpler measure of volunteering is the fraction of people who volunteered in the past year. Using this measure, one finds an increase among seniors, although it is not as pronounced. As Figure 6 shows, in the mid-1970s, roughly 48% of people 60 and older volunteered at least once a year, but that figure had risen to about 55% by the mid-1990s. Interestingly, and contrary to the findings for volunteer effort above, young adults have also become more likely to volunteer at least once a year. In 1975, just 37% reported volunteering in the past 12 months, but that share had risen to the 47% to 48% range in 5 the 1990s. Meanwhile, volunteer participation by middle-age people has remained flat, or even dropped by a few percentage points, since the 1970s. To summarize the findings so far, Americans are slightly more likely to volunteer now than they were in the mid-1970s. However, Americans volunteer significantly more often now than they did then. The number of times that Americans volunteer per year has jumped by 20% since the 1970s, from 6.3 6 times on average to 7.6 times. Yet, when volunteer effort is broken down by age cohort, one sees that only one group—senior citizens—has actually increased its volunteer output. It is seniors, rather than middle-age parents with children, who have become the new torchbearers for voluntary activity. These findings call into question the long-held assumption that there is a predictable life cycle to volunteering. The life-cycle theory holds that volunteer participation follows a bell-shaped curve, with rates rising through adulthood, peaking in middle age, then declining as people move into retirement and old age. As Figure 7 shows, the DDB Needham data do reveal a bellshaped curve for simple volunteer participation (the fraction of people who 7 reported volunteering in the past 12 months). Here, participation rises from 48% among young adults to 60% to 61% among middle-age adults (35 to 64), then falls to 57% for seniors 65 and older. However, the bell-shaped pattern Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 390 Goss Figure 5. Frequency of Volunteering by Age Cohort Source: DDB Needham Life Style Surveys. Figure 6. Volunteer Participation Rate by Age Cohort Source: DDB Needham Life Style Surveys. disappears when one examines volunteer effort (mean times volunteered over the past 12 months). As shown in Figure 8, volunteer effort rises in three steps and shows a pronounced peak among seniors. Young adults volunteer the least often (an Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 391 Figure 7. Life Cycle of Volunteer Participation Source: DDB Needham Life Style Surveys. average of five times per year), middle-age adults volunteer more (eight to nine times a year), and seniors volunteer the most (12 times a year). EXPLORING THE INCREASE IN VOLUNTEERING AMONG SENIORS The analysis so far begs two separate, although possibly related, questions. • Why do seniors in the 1990s volunteer more often than did seniors in the 1970s? • Why has seniors’ volunteering increased dramatically while volunteer- ing by other age groups has posted only a slight increase or no increase at all? This analysis has shown that seniors—or some subset thereof—are volunteering increasingly often. To identify which factors might be driving the increase, I looked to the existing literature on the correlates of volunteering. Prior studies have found that an array of demographic, social, and attitudinal variables influence volunteer output. To assess which, if any, of these variables might be behind the upward trend in seniors’ volunteering, I adopted a two-pronged approach. First, I constructed a multivariate ordinary least squares regression model that included variables that have been found in prior studies to affect volunteering. The regression analysis included only Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 392 Goss Figure 8. Life Cycle of Volunteer Frequency Source: DDB Needham Life Style Surveys. those respondents age 60 and older. Because the sample size was so large (15,500 seniors), I had the freedom to include a large number of explanatory variables (17) in the regression model. The model included the following variables, which were introduced statistically in four blocks. Social capital variables. These include the frequency of giving and attending dinner parties, frequency of attending club meetings, frequency of attending church, degree of agreement that “most people are honest,” and reliance on television as one’s primary form of entertainment. Also included was a contextual-effects variable that represents the mean volunteer effort among people in the respondent’s county. The contextual variable is the best proxy possible in the data for the likelihood that a respondent might have been recruited to volunteer. Demographic and economic variables. These include education, gender, employment status, marital status, self-reported physical condition, selfreported financial worries, and the degree of feeling “hassled.”8 Psychic variables. These include the degree to which they agree with “I wish for the good old days” and “I like to be considered a leader.” The nostalgia variable was included on the theory that older people might volunteer to right the wrongs of modern life. The leadership variable was included because it best captured respondents’ psychic engagement with the larger world. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 393 Generational variable. This is a dummy variable for whether the respondent was born between 1910 and 1930, and therefore a member of the long civic generation (Putnam, 1996).9 In addition to the blocks above, I included a fifth block—a dummy variable for year of study.10 This variable represents the secular increase in seniors’ volunteering, net of the effects of other variables that might be behind the increase. To the extent that the model captures the positive effects of these other variables, the year variable should become substantively small and statistically insignificant. If that is achieved, one can rest assured that the analysis has identified the underlying changes in seniors’ demographic, social, and attitudinal characteristics that have driven the rise in volunteering. RESULTS The results of the multivariate regression analysis were unexpected. In brief, although nearly all of the variables were statistically significant—and some were substantively important—none of the factors traditionally associated with increased volunteering at the individual level appears to explain why volunteering among seniors has risen over time. The full results are presented in Table 1. As Table 1 shows, 12 of the 17 variables were highly significant (p < .01). In addition, four of them were significant at the p < .05 level. Only one (belief in others’ honesty) was not significant at conventional levels. Most of the variables had the expected sign. In line with prior research, the frequency of volunteering rises with more education, status as a female, belief in the honesty of others, greater self-regard as a leader, and greater social engagement (dinner parties, club meetings, and church services). These last two variables—club and church attendance—have, by far, the strongest impact on volunteer frequency, with standardized beta scores that are roughly three to five times as large as those for most other variables (such as education and television viewing). The contextual variable was also substantively large. Surprisingly, volunteering also rises with a greater degree of feeling hassled, which perhaps signifies a reverse causality—people feel pressed precisely because they are engaged in a lot of activities, such as volunteering and others not captured in my model. On the other hand, and not surprisingly, 11 volunteering declines with a rise in one’s time commitment to a job, one’s financial worries, one’s reliance on television for entertainment, and one’s marital status. In addition, the analysis finds that less nostalgic people—those who do not spend their time wishing for the good old days—are more frequent volunteers, as are those who value their role as leaders. Finally, and consistent with Putnam’s (in press) thesis that a long civic generation has helped to stem a general decline in social capital, I find that—even after controlling for myriad demographic, work, and social factors—being a member of the 1910 to 1930 birth cohort is a significant predictor of the frequency of Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 394 Goss Table 1. The Determinants of Volunteering: All Seniors Age 60 and Older (dependent variable: times volunteered over past 12 months) Model 1 2 3 4 R2 (adjusted) .008 (.008) .158 (.157) .172 (.170) .174 (.172) Independent Variables Year of survey Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Respondent’s education Respondent’s gender Respondent’s employment status (4-point scale) Married (1 = yes) Agree they are in good physical condition (1 to 6) Extent of financial worries (factor score) Extent of feeling hassled (factor score) Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) B Standardized b t Significance .212 .334 .089 .140 7.395 12.417 .000 .000 .147 .079 6.845 .000 .271 .248 21.341 .000 .109 .160 14.078 .000 .333 .027 2.445 .014 –.625 –.071 –6.313 .000 .977 .288 .109 .121 9.821 10.066 .000 .000 .122 .066 5.657 .000 .255 .234 20.094 .000 .102 .150 13.148 .000 .236 .019 1.728 .084 –.454 –.052 –4.495 .000 .963 .794 .732 .107 .068 .023 9.746 5.741 1.898 .000 .000 .058 1.233 –1.175 .078 –.033 6.357 –2.653 .000 .008 .301 .029 2.570 .010 –.664 –.040 –3.381 .001 .813 .279 .037 .117 3.141 9.752 .002 .000 .115 .062 5.343 .000 .251 .229 19.677 .000 .102 .150 13.118 .000 .208 .017 1.524 .128 –.435 –.049 –4.297 .000 Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation Table 1 Model 5 R2 (adjusted) .175 (.173) 395 Continued Independent Variables Average annual times people in respondent’s county volunteered Respondent’s education Respondent’s gender Respondent’s employment status (4-point scale) Married (1 = yes) Agree they are in good physical condition (1 to 6) Extent of financial worries (factor score) Extent of feeling hassled (factor score) Agree they wish for the good old days (1 to 6) Agree they like to be considered a leader (1 to 6) Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Respondent’s education Respondent’s gender Respondent’s employment status (4-point scale) Married (1 = yes) Agree they are in good physical condition (1 to 6) Extent of financial worries (factor score) Extent of feeling hassled (factor score) Agree they wish for the good old days (1 to 6) Agree they like to be considered a leader (1 to 6) Respondent born 1910 to 1930 B Standardized b t Significance .964 .677 .855 .107 .058 .027 9.764 4.820 2.190 .000 .000 .029 1.265 –1.216 .080 –.034 6.520 –2.745 .000 .006 .265 .026 2.259 .024 –.554 –.034 –2.803 .005 .748 .034 2.858 .004 –.277 –.030 –2.627 .009 .481 .274 .047 .115 4.072 9.525 .000 .000 .115 .062 5.315 .000 .251 .230 19.724 .000 .101 .150 13.117 .000 .212 .017 1.559 .119 –.438 –.050 –4.323 .000 .964 .680 .828 .107 .058 .026 9.769 4.844 2.121 .000 .000 .034 1.259 –1.184 .080 –.033 6.493 –2.672 .000 .008 .261 .026 2.230 .026 –.563 –.034 –2.846 .004 .727 .033 2.774 .006 –.267 –.029 –2.531 .011 .481 .855 .047 .023 4.072 2.040 .000 .041 Source: DDB Needham Life Style surveys (1975 to 1998). Note: The method used was ordinary least squares pairwise regression. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 396 Goss volunteering. All else being equal, being a member of the long civic generation boosts one’s volunteering by about one time per year. Most of these results are not particularly surprising. What is puzzling, however, is the persistent effect of the year variable, which captures the upward secular trend in volunteering. The year variable can be thought of as the effect of factors not explicitly controlled in the model—the mysterious wind that has blown through and lifted seniors’ volunteering. Even after all the factors known to affect volunteering are controlled, the year coefficient is still strongly positive. What this means, in essence, is that the increase in seniors’ volunteering is not attributable to conventional explanations—an increase in education or an improvement in health and financial well-being, for example. The regression method used here—entering blocks of related variables sequentially—helps to illustrate the forces that have been pushing for, and against, greater volunteering by seniors. The first block was simply the secular-trend year variable. Its coefficient was .212, meaning that, on average, the simple passage of time increased seniors’ volunteer output by about one annual episode every 5 years. After adding the next block of variables, those capturing seniors’ social capital, the effect of the secular trend jumped to .334. What this means is that the secular trend is even more bizarre. Whatever mystery winds are pushing volunteering upward, they were strong enough to overcome countervailing forces that were exerting a downward pressure on volunteering. The next three blocks of variables helped to explain some of the secular trend. With the addition of each of these blocks—demographic, attitudinal, and generational variables—the coefficient on year became progressively smaller. Some of the observed increase in seniors’ volunteering is explained by factors in the model, most especially educational advances, increased leadership values, and membership in the long civic generation. Still, there is much that is unexplained. Even after all the variables were entered, the coefficient on year remained strongly positive (.274). The simple, secular trend has a substantial effect. With a standardized beta coefficient of .115, the year variable is stronger than virtually any other factor, including education, television viewing, leadership values, employment status, financial worries, and marital status. The unmodeled mystery winds have been strong enough to push volunteering up when, by all conventional accounts, it 12 should have gone down. To illustrate this, I isolated some of the variables that I had hypothesized to be behind the rise in volunteering among older Americans and describe their trends over time. Seniors’ health has improved. With increases in national wealth, greater attention to healthy living and preventive care, and advances in the treatment of disease, Americans live longer now than ever before. It is clear from the multivariate regression analysis that good health is a positive (but not significant) predictor of volunteering frequency among seniors, even after controlling for Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 397 myriad demographic and social factors. One might expect, then, that the improved health of seniors would be a major force behind the observed increase in volunteering. Interestingly, there is no support for that proposition. First, the data suggest that seniors’ self-reported health status has not improved over the past quarter century. In fact, it has declined modestly. The fraction of seniors 60 and older who considered themselves to be in good physical condition has dropped steadily over the years, from 72% in 1975 to 58% in 1998. There are several reasons why seniors’ self-reported health might have declined during a period when medical treatment has improved. First, seniors who might have died of their infirmities 25 years ago may be living with them today. Second, given the advances in medical testing and diagnosis, seniors might simply be more aware of their health conditions than they were in decades past. Third, the recent DDB Needham samples might simply include more frail elderly people than did samples in the 1970s. With the data, I have no way of testing the first two explanations. The third—that a skew toward very old people is causing reduced self-reported health—does not appear to be the case. The trend line for self-reported health of seniors 60 to 80 years old is nearly identical to the trend line for the entire senior population, with the mean health score declining from 4.2 to about 3.7 during the span of the surveys. Even more intriguing is the finding that both healthy and unhealthy seniors have increased their volunteer output relative to their counterparts in the 1970s, and the two groups have done so at the same rate. Although healthy people do volunteer more often than do unhealthy people, both groups have 13 doubled their annual volunteer output over the past quarter century. Improved health could not have driven the increase in seniors’ volunteering because it apparently has not improved. However, if a greater fraction of seniors had become healthier, we would have seen an even greater increase in volunteer output than we have actually observed. Seniors have become more civically and socially engaged. The most perplexing finding concerns social capital. For various reasons, including increases in education and relative income, it is expected that seniors today will be more civically and socially engaged than were their same-age predecessors in the 1970s and 1980s. If this were the case, there might be an answer to the puzzle. Social capital emerged in the model (and in other studies) as the strongest predictor of volunteer frequency. If social capital indeed had been on the rise, this increase could very well account for the rise in volunteering. Unfortunately, both for this study’s detective work and for the nation, the data show that seniors’ engagement with public life has been declining since the 1970s. They are attending church services and club meetings less frequently and working on fewer community projects. They are becoming less willing to trust in the honesty of others. This disengagement is true for all seniors and for the subset of seniors who report being in good physical condition. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 398 Goss It is important to note that older people’s participation, at every point in time, is still higher than that of younger age cohorts, and so, without seniors, the rate of participation among the public at large would have sunk even lower than it has. Nonetheless, the decline in seniors’ civic engagement makes the increase in their volunteering doubly perplexing. First, it is perplexing because church going, club membership, and working on community projects almost, by definition, entail volunteering, and in regression models, participation in such activities is the strongest predictor of volunteer output. Thus, if these activities are declining, volunteering should be declining as well. Second, the rise in volunteering is perplexing because, in so many other ways, seniors have become less likely to take the initiative to associate, either formally or informally, with other people. Given this broad, if gradual, disengagement, it is paradoxical that the most costly form of engagement—helping others through volunteer work—should be on the rise. Seniors have more time on their hands. Recent studies have suggested that older Americans have more time on their hands now than ever before, largely because more of them are taking early retirement (Robinson & Godbey, 1997). However, according to the DDB Needham data, older people are participating in fewer planned leisure-time activities. They are joining fewer clubs than they used to, going to church less frequently, and socializing with friends less often. Therefore, I expected that seniors would be volunteering more often to fill the newly available time. To test this hypothesis, three variables were examined. These included seniors’ job commitments, their self-reported degree of feeling hassled, and their self-reported spare time. Contrary to expectation, older people actually appear to feel more burdened now than they did in the 1970s. In the DDB Needham samples, the fraction of seniors working full-time (or selfemployed) has not declined since the mid-1970s, and the fraction working part-time has actually risen slightly and steadily since the early 1980s. Parttime work among seniors is more common now than it was in 1975. At the same time, although the fraction of seniors reporting that they have “a lot of spare time” is higher now than in 1975, there has been no movement in the figure since the early 1980s. In other words, during an 18-year period (1980 to 1998) when volunteer output was steadily rising, spare time was not. Finally, although the pattern is irregular, seniors’ mean score on the hassle index is higher now than in the 1970s, and it has not moved over the past decade. One might speculate that these time constraints are connected to volunteering. Seniors who might otherwise have had spare time, or who might have felt more relaxed, do not because they are volunteering all the time. There are two reasons to reject this possibility. First, seniors certainly do volunteer more than others do—but an older person’s average volunteer commitment is just once a month, not every day. In addition, recall that seniors are no less likely to be working full-time today than they were before, and that they are actually more likely to be working part-time. Therefore, seniors have not simply Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 399 substituted volunteering for paid work. The data are not conducive to the image of a nation of seniors freed of the job burdens faced by earlier generations and wondering what to do with all that time on their hands. Seniors have become more financially secure. In theory, it is possible that an improvement in seniors’ economic circumstances might have driven the rise in increased volunteerism, as people who are better off may have the time and the psychic freedom to give back to the community. Again, there is little evidence for this explanation. First, although it is true that seniors’ income has increased relative to what it was in the past, income does not show up in the multivariate regression models as a statistically significant predictor of volunteer output. What is a significant (although substantively small) predictor of seniors’ volunteering in some models is the extent of their financial anxiety.14 In general, as financial worries rise, volunteering frequency declines. This finding could help account for the rise in seniors’ volunteering only if they have become less financially worried over time. However, the fact is that their worries have risen relative to other people’s. Hence, improved financial status represents another explanation that cannot be supported. Seniors use volunteering to compensate for social isolation. One might imagine that a decline in club and church work has left seniors feeling more lonely and isolated. Perhaps, then, they are finding new, nontraditional routes to volunteer work in an effort to overcome a feeling of being cut off from the larger world. Three variables allow one to examine the social isolation of seniors. These include whether they are married, whether they spend a lot of time with 15 friends, and whether the family eats dinner together frequently. If seniors are using volunteer work as a next-best substitute for family and informal socializing, one would expect to see a negative correlation between volunteering and private socializing. Contrary to this prediction, supplementary regression analysis (not shown) finds that both the friends and family-dinner variables are significantly and positively correlated with volunteering. Seniors who spend more time with friends and family members volunteer more often. I therefore find no support for the substitution effect. In addition, among seniors, both of these activities have been declining noticeably since the 1970s, which should have pushed volunteering down. Thus, far from explaining the rise in volunteering, the trend away from informal socializing only makes the increase that much more perplexing. Marital status, on the other hand, appears to have some impact on seniors’ volunteering—but only among women. Being married, all else being equal, depresses senior women’s volunteer frequency and may do the same for men (although the effect is not statistically significant). This finding is consistent with the hypotheses that seniors use volunteer work as a means of overcoming the loneliness that often accompanies the absence of a spouse. Hence, if a greater proportion of seniors consists of unmarried women, this shift in the Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 400 Goss population composition could be driving the aggregate rise in seniors’ volunteering. Although the fraction of senior women who are unmarried has risen slightly (about 5%) since the mid-1980s, it is not enough to account for the overall increase in volunteering, which is present among both married and unmarried senior women. Hence, the social isolation explanation is, at best, a partial explanation for the rise in volunteering among one group of seniors. Seniors’ changing attitudes toward the welfare state. One could imagine that seniors’ increased volunteering was prompted by a shift in their attitudes toward the relative role of private organizations versus governmental organizations in meeting social needs. Although the DDB Needham surveys do not contain longitudinal data on this question, there is such a question in the 1996 DDB Needham survey. That year, respondents were asked about the extent to which they agreed that “private charities are more effective than government.” Older people were significantly more likely to agree with that statement than were middle-age and young adults. Older people were far more likely than others to say that charities are definitely more effective (24%); this is compared to 16% to 17% for other adults. In theory, if such beliefs predict greater volunteering—and if seniors have increasingly embraced these beliefs over time—it might help to explain the rise in their volunteer frequency. Unfortunately, multivariate regression analysis found that one’s belief in the relative effectiveness of private charity was not a statistically significant predictor of volunteer output, either among all seniors or among senior women. Thus, there is little reason to have faith in this explanation. Nonprofit groups’ are asking more of senior volunteers. Perhaps seniors’ increased volunteering is the result of an increased demand for their time. In other words, it is important to consider whether seniors are being targeted more often by nonprofit groups looking for volunteers. This is anecdotally plausible, because the traditional pool of volunteers—middle-age, nonworking women with children—is getting smaller. Several earlier studies, using data sets with far more detailed information on volunteering than the DDB Needham surveys contain, have found that one of the strongest predictors of volunteer likelihood is simply being asked to give time (Berger, 1991; Freeman, 1996; Hodgkinson & Weitzman, 1996). The recruitment effect is strong and positive, even after controlling for demographic, socioeconomic, and contextual factors (Berger, 1991; Freeman, 1996). In principle, the inclusion of club-meeting and church attendance should have picked up some of the effect of being asked, as clubs and churches are fertile recruiting grounds for volunteers. However, there is some evidence that nonorganizational recruitment is on the rise, at least over the past decade or so. According to the biennial Gallup-Independent Sector surveys, the fraction of all adults who got into their volunteer activities because they were asked by someone grew from 40% in 1987 to 46% in 1995. Yet, the fractions who said that someone was a friend, a family member, a coworker, or a person in a Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 401 religious institution have all declined. Superficially, at least, it appears that, although volunteer recruitment is on the rise, it is increasingly less likely to take place through traditional channels. Although the data do not allow a direct test of the recruitment-of-seniors theory, an indirect test is possible. If seniors are increasingly entering their voluntary activity through routes other than clubs, one would expect to see a decline over time in the correlation between the frequency of club-meeting 16 attendance and volunteering among older people. The intuitive explanation for the declining correlation is simple. For example, some people would continue the existing pattern of one club meeting for every two volunteer episodes, whereas others would maintain their club schedules but see fluctuations in volunteer episodes based on the ebb and flow of recruitment by other organizations. Unfortunately, the correlation between club meetings and volunteer episodes has held steady, both for older men and for older women, since the late 1970s. In fact, the correlation for women is actually up a little bit since that time, which is contrary to what would be expected if these women were being recruited through channels other than clubs. Another voluntary-group–related explanation for seniors’ unusually large increase in volunteer activity is that the organizations (or clubs) with which these people are involved have been asking more of their time. One could imagine that seniors are attending fewer group meetings, but the groups with which they are affiliated are asking a different commitment from them. One commentator has suggested that the number of organizations catering to aging-related concerns has increased dramatically—Alzheimer’s groups, hospices, cancer-prevention groups, and cultural organizations, and so forth—and that these groups actively involve seniors in volunteer work (S. R. Smith, personal communication, April 26, 1999). There is no truly satisfactory way, given the limitations of my data, to test that hypothesis. However, the data do show that, among seniors who reported attending at least one club meeting, the ratio of club meetings to volunteer episodes has increased fourto fivefold since 1975. In other words, today’s club members are far more involved in volunteering than their club-member predecessors were. There was also an increase in the meeting-to-volunteering ratio for middle-age and younger people, but the increase was less striking. Given the finding that the correlation between club meetings and volunteer episodes among seniors has held steady, the increasing meeting-to-volunteering ratio suggests a wholesale shift from meetings to volunteering as the central activity of organizations with which seniors are affiliated. This possibility is intuitively appealing and is worthy of future research. WHAT GIVES? The cumulative evidence suggests several possible reasons why seniors have been intensifying their voluntary activities. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 402 Goss First, we know that education is a strong predictor of volunteer output and that older people today are more educated than were their counterparts in the 1970s. In 1975, the average senior in the DDB Needham sample had a high school degree but no more. By the late 1990s, the average senior has attended some college. More significant is the impact of those at the very top of the educational pyramid. People with a graduate degree have increased their volunteering at a faster rate than have people with less education, and graduatedegree holders have become more numerous (15% of the senior sample in 1998, up from less than 3% in 1975). However, even with an increase in these people’s numbers and volunteer output, further regression analysis (not shown) found that the highly educated are responsible for no more than a tiny 17 fraction (if any) of the secular increase in volunteering. The second possible explanation revolves around the seeming paradox that club-going seniors have intensified their volunteering more than club-goers have in other age groups. Although data limitations preclude a definitive answer, there is strong circumstantial evidence that seniors’ organizational activity increasingly involves volunteering. This may be because seniors are gravitating to newer organizations (that respondents consider clubs) in which the central activity is volunteering, or because traditional clubs have evolved to emphasize volunteering more than meetings. Either way, the finding has a potentially broad significance. It means that, as Baby Boomers are shifting their organizational involvement from face-to-face activities to monetary contributions, often to advocacy groups in state capitals and Washington D.C. (Skocpol, 1999), seniors are finding new avenues through which to engage in traditional, face-to-face ways. The third partial explanation for the observed rise in volunteering among seniors revolves around gender. Throughout its 24 years, the DDB Needham sample of senior citizens has gone from being dominated by males to being dominated by females. While females constituted 41% to 48% of the sample in the 1970s, their numbers had risen to 55% to 57% in the 1990s. Because women have consistently volunteered more frequently than have men, the growing representation of females in the sample could well be driving the overall rise in volunteering. That said, the data also suggest that volunteering frequency has been rising steadily and consistently among both men and women. Thus, there appears to be a real secular increase in volunteering that is not simply the product of a change in sample composition. What is less clear, however, is whether the explanation for the increase in female seniors’ volunteering is the same as the explanation for the increase among their male counterparts. To investigate this question, I ran separate regression analyses on the female and male subsamples, with a surprising and paradoxical result. Simply put, the underlying forces pushing volunteering upward are far stronger and more difficult to understand for senior women than for senior men (see Table 2 and Table 3). In the full model of senior men, the coefficient on year is .210, whereas the comparable number in the women’s model is .344. I interpret this to mean that, whatever mysterious forces are Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 403 Table 2. The Determinants of Volunteering: Male Seniors Age 60 and Older (dependent variable: times volunteered over past 12 months) Model 1 2 3 4 R2 (adjusted) .007 (.006) .143 (.141) .152 (.149) .154 (.150) Independent Variables Year of survey Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Respondent’s education Respondent’s employment status (4-point scale) Married (1 = yes) Agree they are in good physical condition (1 to 6) Extent of financial worries (factor score) Extent of feeling hassled (factor score) Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) B Standardized b t Significance .168 .255 .082 .124 4.541 7.303 .000 .000 .137 .080 4.648 .000 .240 .233 13.424 .000 .09896 .165 9.681 .000 .311 .029 1.745 .081 –.471 –.060 –3.532 .000 .875 .223 .111 .108 6.644 6.108 .000 .000 .116 .068 3.894 .000 .230 .224 12.878 .000 .09708 .162 9.465 .000 .211 .020 1.174 .240 –.345 –.044 –2.531 .011 .865 .515 .110 .054 6.588 2.973 .000 .003 1.023 –.899 .070 –.021 3.816 –1.209 .000 .227 .228 .025 1.442 .149 –.572 –.039 –2.178 .029 .725 .217 .040 .105 2.154 5.915 .031 .000 .113 .066 3.766 .000 .227 .221 12.671 .000 .09668 .161 9.423 .000 .189 .018 1.050 .294 –.338 –.043 –2.472 .013 (continued) Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 404 Goss Table 2 Model 5 R2 (adjusted) .155 (.150) Continued Independent Variables B Standardized Average annual times people in respondent’s county volunteered .863 Respondent’s education .442 Respondent’s employment status (4-point scale) 1.042 Married (1 = yes) –.842 Agree they are in good physical condition (1 to 6) .202 Extent of financial worries (factor score) –.504 Extent of feeling hassled (factor score) .669 Agree they wish for the good old days (1 to 6) –.162 Agree they like to be considered a leader (1 to 6) .347 Year of survey .210 Gave/attended dinner party (times/year) .112 Attended club meetings (times/year) .228 Attended church services (times/year) .09671 Agree most people are honest (1 to 6) .191 Agree TV is their primary entertainment (1 to 6) –.339 Average annual times people in respondent’s county volunteered .865 Respondent’s education .444 Respondent’s employment status (4-point scale) 1.046 Married (1 = yes) –.834 Agree they are in good physical condition (1 to 6) .204 Extent of financial worries (factor score) –.508 Extent of feeling hassled (factor score) .645 Agree they wish for the good old days (1 to 6) –.149 Agree they like to be considered a leader (1 to 6) .343 Respondent born 1910 to 1930 .738 b t Significance .110 .046 6.576 2.510 .000 .012 .071 –.020 3.877 –1.172 .000 .241 .022 1.278 .201 –.034 –1.911 .056 .037 1.964 .050 –.019 –1.131 .258 .038 .102 2.190 5.693 .029 .000 .066 3.754 .000 .222 12.714 .000 .161 9.428 .000 .018 1.064 .288 –.043 –2.477 .013 .110 .047 6.587 2.527 .000 .012 .072 –.020 3.894 –1.121 .000 .262 .022 1.285 .199 –.034 –1.926 .054 .035 1.892 .059 –.018 –1.040 .298 .038 .023 2.166 1.353 .030 .176 Source: DDB Needham Life Style surveys (1975 to 1998). Note: The method used was ordinary least squares pairwise regression. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 405 pushing volunteering upward, older women are feeling the effects half again as much as are senior men. After controlling for all known correlates of volunteering, there remained a secular rise in senior women’s volunteer output equivalent to eight extra volunteer acts in 1997, relative to the 1974 benchmark. To put it another way, a senior woman in the 1998 survey volunteered eight more times over the prior year than did her demographic and social twin in the 1975 study. This is an enormous effect, given that senior women today perform only 12 acts per year on average. Having isolated senior women’s volunteering as the key variable to be explained, I did further regression analysis on this subsample to test, in particular, whether older women are volunteering more to overcome social isolation. Unfortunately, I again found no compelling evidence for such a substitution effect. After accounting for the decline in having family dinners and spending time with friends, the volunteering trend is still positive and just as strong.18 Likewise, accounting for the positive impact of unmarried status on older women’s volunteering and for the slight rise in the number of older unmarried women in the sample did not meaningfully affect the unexplained 19 portion of the increase in volunteering. Over all, one can draw three intriguing conclusions. • The increase in senior women’s volunteering is particularly remarkable because, by virtually all indications, it should have been declining over the past 25 years. • There is a mystery force (or forces) driving this increase that is not among the standard retinue of sociodemographic factors known to influence volunteering. • Whatever this mystery force is, it is strong and appears not to have inspired men to the same extent that it inspired women. It is unlikely that the DDB Needham data set, for all its richness, allows for the identification of this mystery factor(s). I have modeled every attitudinal, social, and demographic correlate of volunteering available in the data, and inclusion of those variables did not make the underlying secular increase go away statistically. For example, one might sensibly hypothesize that today’s senior women are volunteering more than their predecessors were because of some lingering effect from World War II. During the war, when today’s seniors were in their late teens and 20s, they were thrown into extraordinary service on the domestic front. They rallied around charitable activities to aid the war effort, and many got jobs to support their families. Perhaps a taste for civic involvement was developed during those formative years, and has remained strong as the women have aged. Plausible as this explanation sounds, the war 20 effect should have been picked up by the generational variable —and indeed, it might well have been. However, after accounting for the war, the mysterious secular rise in older women’s volunteering remains. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 406 Goss Table 3. The Determinants of Volunteering: Female Seniors Age 60 and Older (dependent variable: times volunteered over past 12 months) Model 1 2 3 4 R2 (adjusted) .007 (.007) .156 (.154) .172 (.169) .176 (.172) Independent Variables Year of survey Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Respondent’s education Respondent’s employment status (4-point scale) Married (1 = yes) Agree they are in good physical condition (1 to 6) Extent of financial worries (factor score) Extent of feeling hassled (factor score) Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) B Standardized b t Significance .223 .396 .084 .150 5.224 9.804 .000 .000 .152 .078 5.006 .000 .288 .254 16.149 .000 .111 .149 9.706 .000 .352 .026 1.742 .082 –.698 –.074 –4.851 .000 1.058 .358 .108 .135 7.252 8.225 .000 .000 .128 .066 4.206 .000 .273 .241 15.322 .000 .106 .142 9.250 .000 .274 .020 1.360 .174 –.548 –.058 –3.744 .000 1.037 1.202 .106 .087 7.160 5.533 .000 .000 1.489 –1.335 .085 –.038 5.363 –2.311 .000 .021 .352 .032 2.077 .038 –.748 –.042 –2.603 .009 .945 .349 .038 .132 2.428 8.014 .015 .000 .119 .061 3.898 .000 .267 .235 14.951 .000 .106 .142 9.259 .000 .244 .018 1.209 .227 –.517 –.055 –3.519 .000 (continued) Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation Table 3 Model 5 R2 (adjusted) .176 (.173) 407 Continued Independent Variables Average annual times people in respondent’s county volunteered Respondent’s education Respondent’s employment status (4-point scale) Married (1 = yes) Agree they are in good physical condition (1 to 6) Extent of financial worries (factor score) Extent of feeling hassled (factor score) Agree they wish for the good old days (1 to 6) Agree they like to be considered a leader (1 to 6) Year of survey Gave/attended dinner party (times/year) Attended club meetings (times/year) Attended church services (times/year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Respondent’s education Respondent’s employment status (4-point scale) Married (1 = yes) Agree they are in good physical condition (1 to 6) Extent of financial worries (factor score) Extent of feeling hassled (factor score) Agree they wish for the good old days (1 to 6) Agree they like to be considered a leader (1 to 6) Respondent born 1910 to 1930 B Standardized b t Significance 1.041 1.039 .107 .075 7.200 4.713 .000 .000 1.537 –1.412 .088 –.040 5.540 –2.445 .000 .015 .307 .028 1.811 .070 –.597 –.034 –2.063 .039 .881 .036 2.236 .025 –.387 –.039 –2.535 .011 .597 .344 .053 .130 3.469 7.901 .001 .000 .118 .061 3.866 .000 .268 .236 14.979 .000 .106 .142 9.253 .000 .252 .019 1.250 .000 –.522 –.055 –3.557 .211 1.040 1.045 .106 .076 7.194 4.744 .000 .000 1.521 –1.371 .087 –.039 5.484 –2.373 .000 .018 .298 .027 1.753 .080 –.613 –.035 –2.118 .034 .864 .035 2.191 .029 –.380 –.038 –2.492 .013 .602 1.134 .054 .027 3.499 1.802 .000 .072 Source: DDB Needham Life Style surveys (1975 to 1998). Note: The method used was ordinary least squares pairwise regression. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 408 Goss Even if one cannot directly model the mystery factor, a comparative analysis of volunteering by age cohort might offer clues to its identity. If the mystery wind has hit all age groups equally, one might search for a broad social explanation. By contrast, if the force seemed only to affect older people, one might narrow the search to the particular context and circumstances in which seniors find themselves, and the ways that those things might have changed over time. This inquiry brings us to the second big question revealed in the data. Why has seniors’ volunteering risen so much faster than that of other age groups? As Figure 5 shows, the frequency of volunteering by people older than 60 has risen steadily since the mid-1970s. By contrast, volunteering by young adults has been essentially flat (at least since 1980), and volunteering by middle-age people has actually declined over the past quarter century. This begs a question: Are the underlying processes driving these three age-related trends the same or different? That is, are the static trend in young adult volunteering and the downturn in middle-age volunteering attributable to changes in the same factors that propelled seniors’ volunteering upward? In particular, I was curious whether the mystery force, as captured by the year coefficient, would be as strongly positive for other age groups as it was for seniors. To answer these questions, I expanded the multivariate analysis to two 21 other age groups: adults younger than 30 and middle-age adults 30 to 59. The key results are presented in Table 4. To summarize, once all the demographic, social capital, and attitudinal variables have been added, there remains a positive and significant secular trend effect on volunteering. However, the size of the effect varies greatly by age group. For adults younger than 30, the year coefficient is tiny (.08), meaning that the trend in volunteer output is well explained by conventional variables (e.g., employment status, gender, and social participation). Among middle-age adults, the secular trend effect is larger (.168), but not nearly as large as it is for older adults (.278). As revealed in the coefficients on year, the unmodeled mystery winds are responsible, ceteris paribus, for an increase of two volunteer acts for young adults and four volunteer acts for middle-age folks, compared to seven additional acts for 22 older folks, over the surveys’ 24 years. Table 5 shows the effect of the mystery winds by comparing the predicted and actual volunteer output for young, middle-age, and older adults. Based on the foregoing analysis, I reached four conclusions. • Some new wind has been blowing over the past 25 years that has lifted volunteering among those who volunteer the most. • Whatever this wind is, young adults do not feel it, middle-age people feel it as a breeze, and older people feel it as a gale force. • Even though it did exert pressure on middle-age people, this breeze was not strong enough to counteract the countervailing forces that were working simultaneously to drive down volunteering in this group. Several key factors known to boost volunteering became less common Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 409 Table 4. Determinants of Volunteering by Young, Middle-Age, and Senior Adults (dependent variable: times volunteered over past 12 months) B Coefficient Model 1 2 3 4 a Independent Variables Young Adults Middle-Age Seniors 60 Under 30 Adults 30 to 59 and Over Year of survey Year of survey Respondent’s employment status (4-point scale) Agree they are in good physical condition (1 to 6) Extent of feeling hassled (factor score) Extent of financial worries (factor score) Respondent’s education Married (1 = yes) Respondent’s gender Children at home (1 = yes) Year of survey Respondent’s employment status (4-point scale) Agree they are in good physical condition (1 to 6) Extent of feeling hassled (factor score) Extent of financial worries (factor score) Respondent’s education Married (1 = yes) Respondent’s gender Children at home (1 = yes) Gave/attended dinner party (times last year) Attended club meetings (times last year) Attended church services (times last year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Year of survey Respondent’s employment status (4-point scale) Agree they are in good physical condition (1 to 6) Extent of feeling hassled (factor score) Extent of financial worries (factor score) Respondent’s education Married (1 = yes) Respondent’s gender Children at home (1 = yes) Gave/attended dinner party (times last year) Attended club meetings (times last year) Attended church services (times last year) Agree most people are honest (1 to 6) Agree TV is their primary entertainment (1 to 6) Average annual times people in respondent’s county volunteered Agree they wish for the good old days (1 to 6) Agree they like to be considered a leader (1 to 6) .04159 .02579 .567** .06395 .909** –.417* .618** –.229 –.250 –.389 .08803** .441** .04669 .691** –.01712 –.109 –.468 –.754* –.445 –.09102** .276** .07599** .06604 –.367** –.009996 .02896 1.382** .235** .884** –.635** 1.504** 1.052** 1.564** 1.468** .174** 1.141** .05093 .465** –.09454 .704** .224 .166 1.256** .09869** .326** .09071** .181* –.403** .212** .171** 1.458** .634** 1.286** –1.277** 1.445** –.933* 2.309** –1.329** .287** 1.223** .299* .850** –.636** .797** –1.168** .727 –.822 .121** .255** .102** .231 –.455** .601** .08400** .454** –.01610 .625* .02835 –.185 –.493 –.640* –.454 .08639** .272** .07609** .05069 –.353** .759** .168** 1.158** .02490 .282 .001949 .543** .207 .411 1.199** .08824** .320** .09148** .159* –.381** .959** .278** 1.255** .262* .786** –.525** .678** –1.208** .852* .857 .114** .250** .102** .203 –.436** .602** –.116 .275* .763** –.158* .655** Source: DDB Needham Life Style surveys (1975 to 1998). Note: The method used was ordinary least squares pairwise regression. a. R2 for Model 4 are .141 (young adults), .187 (middle-age adults), and .175 (seniors). *p < .05. **p < .01. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 .960** –.276** .486** 410 Goss Table 5. Effect of the Mystery Wind Lifting Volunteers Number of Volunteer Acts Population Group All seniors Women Men Middle-age Young adults Extra Output Left Unexplaineda 6.6 8.3 5.0 4.0 2.0 Actual Output (1998) Output Expectedb (1998) 10 12 9 7 5 3 to 4 3 to 4 4 3 3 a. Extra volunteer acts reported in 1998, relative to 1975, after social, demographic, and attitudinal factors are controlled. b. Output given social and demographic changes. among middle-age people, and the key factors that suppress volunteering became more common. For example, club participation (a positive correlate) declined more dramatically among middle-age adults than among any other age group. In addition, having children at home (a positive correlate) became substantially less common. The fraction of adults with kids at home fell by more than 10% between 1985 and 1998. At the same time, two negative correlates of volunteering—reliance on television for entertainment and full-time work—increased among middle-age people more than among any other age group.23 These four developments conspired to suppress volunteering, with the effect that what would otherwise have been an increase was, in fact, a slight decline in voluntary output. Hence, the trend line shows that middle-age people’s volunteer frequency fell from eight acts (over the past 12 months) in the 1975 survey to seven acts in the 1998 survey. • The gale-force wind felt by seniors was strong enough to counteract the depressing effects of declining social participation, increased work, and greater financial worries in this age cohort. Hence, we see that seniors volunteer nearly twice as often today as they did in the mid-1970s. Another way to show the differential effects is to compare the proportional change in volunteering that we actually observed with the change we would have seen if the mystery wind had not blown through. The analysis suggests that, without the mystery forces, seniors’ volunteering would have fallen by 33% (relative to 1975); instead, it rose by 80%. The net effect of the mystery wind is +113%. Meanwhile, the effect on middle-age and younger adults is far less pronounced. Without the mystery push, middle-age people’s volunteering would have fallen by 64%; instead, it fell by 12%, for a net effect of +52%. In addition, among young adults, volunteering would have risen by 10%; instead, it rose by 77%, for a net effect of +67%. Thus, the net effect of the mystery Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 411 force on seniors is approximately twice that of the effect on middle-age and young adults. CONCLUSION Analysis of the DDB Needham data has identified and solved some important puzzles relating to trends in American volunteering. However, in the process, these findings have created other puzzles that are, in some ways, far more perplexing. This study has documented, for the first time, a substantial increase in volunteering over the past generation. The study has identified one segment of the population—senior citizens, especially women—as the vanguard of this 24 movement toward greater participation in civic life. The analysis has shown how truly paradoxical this movement is at a time when virtually every other form of public and social involvement, from club meetings to dinner parties, is down quite dramatically, even among seniors. What this study has not been able to state definitively, however, is why seniors are so much more volunteer-spirited now than they were in the 1970s and 1980s. I have discounted seven of the most plausible explanations, including improved health, better economic circumstances, more free time, increased psychic and actual engagement in public life, the desire to fill a void left by family disintegration, and disillusionment with government social programs. Increased educational attainment among older people is almost certainly part of the story, but a substantial secular increase in volunteer output remains unexplained. Some commentators have suggested that, over the past couple of decades, there has been a profound cultural shift in seniors’ perception of retirement—from a time of relaxation to a time of activity and adventure. This conception of retirement was most recently trumpeted by former President (and active volunteer) Jimmy Carter in his 1998 book The Virtues of Aging. Such a cultural shift, if it has occurred, would be hard to detect in the survey data used in this project, but it is intuitively plausible. It is fully consistent with the idea that the increase in seniors’ volunteering may be attributable to an increase in volunteer-led organizations focused on aging-related concerns. If these observers are correct, the increase in seniors’ volunteering may well be attributable to an interaction between increased individual motivation and increased institutional focus on volunteer involvement. Regardless, we do know that important clues to the puzzle rest with older women. Even without definitive answers to the “why” question, this research has important implications for voluntary agencies. First, it is clear that senior citizens are both an eager and reliable source of volunteer labor. To the extent that voluntary agencies are not specifically recruiting this group, they would do well to begin such targeted appeals. Second, because volunteering among middle-age people is actually lower now than it was in the 1970s, it appears Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 412 Goss that, as the century comes to a close, America is enjoying not a springtime of volunteerism, but an Indian summer.25 Voluntary agencies must take this reality to heart and try to find new ways to energize Baby Boomers as they pass into retirement. Many nonprofit groups are already doing so, and at least one organization has been formed to create model volunteer programs for aging Boomers (Hall, 1997).26 One can anticipate that the urgency of such efforts will be felt increasingly as the civic generation dies and volunteer agencies find themselves scrambling for assistance from new sources. Finally, this research has posed an interesting mystery for academic researchers to try to solve. Data sets richer in volunteer-related information may allow scholars to identify the gale-force wind that has had such a profound effect on the public spiritedness of so many older Americans. Notes 1. The surveys were conducted by DDB Needham, a private marketing company based in Chicago. The data have been extensively tested for both internal and external validity and reliability (Putnam & Yonish, 1998). 2. In this article, volunteer effort will refer to the frequency of volunteering in a given year. 3. The respondents are generally not paid, although they might occasionally receive a small sum if extra effort (e.g., shopping) is required to complete the survey. 4. In 1998, the Silent Generation volunteered eight times over the past 12 months, the Boomers seven times, and the post-Boomers five times. 5. The rise in young-adult volunteering shows up in a variety of independent surveys and is, therefore, almost certainly not a statistical artifact. It is also surprising, given that those who are now in their 20s—Generation X—have gained a reputation as a generation of cynics and slackers. Although the DDB Needham data do not allow for a test of the hypothesis, I suspect that the recent upturn in young-adult volunteering is at least partly a legacy of community service learning programs that were proliferating as Generation X was going through school. Champions of such mandatory volunteering have long argued that these programs would inculcate a taste for volunteering that would last into students’ adult years. These data certainly do not disconfirm that claim. 6. Figures are averaged over the 1975 to 1979 and 1994 to 1998 periods to smooth out random year-to-year fluctuations in the survey responses. 7. I use 1996 data alone to illustrate the pattern. However, the bell shape also appears when data from all years are included. 8 The hassle index (created by Robert Putnam) was a composite of answers to four questions, each measured on a 6-point agree-disagree scale. These questions include the extent to which the respondent agrees that his or her family (a) is too heavily in debt, (b) can never seem to get ahead, (c) has more to spend on extras, and (d) has enough income to meet important desires. For consistency’s sake, the latter two variables were rescaled so that higher agreement translated into a lower score on the 6-point scale. 9. I originally included four other variables that were thought to affect volunteering: seniors’ income (measured as a percentile score), race, residence in a rural area or small town, and the belief that “religion is important in my life.” None of these was significant, and all were dropped from the final model presented. 10. Year = 1975 to 1998. 11. The employment variable was measured on a 4-point scale, with 1 being full-time or selfemployed, 2 being employed part-time, 3 being retired, and 4 being unemployed. Hence, a Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 413 positive sign on the regression coefficient signifies that, as one moves up the scale toward less work, one’s volunteering rises. The same results are obtained when individual employment categories are dichotomized—that is, when I included separate control variables for retired (yes/no), employed full-time (yes/no), and so forth. 12. I was concerned that the midcourse change in the surveys’ sampling frame—adding unmarried people beginning in 1985—may have been skewing my results. Because older married people volunteer less often than do their unmarried counterparts, and because only married people were included in the pre-1985 samples, it seemed probable that the volunteering trend line started at an artificially low point, and hence that the observed growth in seniors’ volunteering was exaggerated. To get around this problem, I ran separate multivariate regression analyses on married seniors and unmarried seniors. The coefficient on the survey year variable (representing the secular trend) was larger, as expected, for unmarried seniors than for the married subsample (.385 vs. .263). However, both of these coefficients remained large relative to other coefficients in the model—meaning the upward secular trend was still pronounced—and both coefficients were highly significant. In short, the change of sampling frame does not explain the basic puzzle of the strong upward trend in seniors’ volunteering. Having performed this test on separate samples of married and unmarried seniors, I decided simply to include a married control variable in subsequent models. In those cases, as expected, the coefficient on year (.274) was in the same range and was highly significant. 13. Healthy seniors in 1975 volunteered an average of 6 times in the past year, whereas healthy seniors in 1998 volunteered 12 times, for an overall increase of 100%. By contrast, unhealthy seniors in 1975 volunteered 4 times in the past year, whereas unhealthy seniors in 1998 volunteered 8 times per year, also a 100% increase. 14. Financial worries is significant in models that omit hassled. These two variables are significantly (p < .01) correlated and appear to pick up each other’s effects when both are included. 15. Unfortunately, the survey questionnaire does not define what a family dinner might mean, particularly to widowed or elderly people. Presumably, then, depending on how broadly one defines family dinner, this question might be subject to response error among certain groups of respondents. However, unless frequent volunteers systematically overstate the frequency of family dinners (and nonvolunteers do not), the conclusions drawn in this section will be valid. 16. In principle, one could observe the same correlation between club and church attendance and volunteering even if the former is declining and the latter is increasing over time. However, to the extent that the alternate-routes-to-volunteering dynamic adds a degree of randomness to individual volunteering, the nexus between club and church attendance and volunteering is broken and the correlation should decline. 17. I reran the model, this time inserting a term representing the interaction between graduate degree and year of study. This term will pick up whatever fraction of the overall secular increase in volunteering is due to highly educated people. In this model, the coefficient on the year variable was reduced by only a fraction, from .274 to .273. In addition, the coefficient on the interaction term was not statistically significant. Hence, I conclude that this interaction effect is not an important causal mechanism. 18. The coefficient on year is .344 in the basic model, compared to .343 in the model that includes visiting friends and .343 in the model that includes family dinners. Visiting friends has a positive and statistically significant effect on senior women’s volunteering. By contrast, having frequent family dinners is negatively associated with volunteering, but the size of the effect is miniscule and not even close to statistically significant. 19. Using the data from 1985 (when both married and unmarried people were surveyed), I ran two versions of the full model using the subsample of senior women. In the first version, I included all the demographic, attitudinal, and social capital variables, with the exception of marital status. In the second, I ran the full model including marital status, which was coded 0 (unmarried) or 1 (married). The inclusion of the marital-status variable reduced the coefficient on year of study from .476 to .460, a modest 3% decline that left the coefficient substantively large and highly statistically significant (p < .0001). Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 414 Goss 20. Women who were 18 to 30 years old during America’s involvement in World War II (1941 to 1945) would have been born between 1911 and 1927. The civic-generation variable controls for births between 1910 and 1930—roughly the same span of time. 21. In the models run on seniors only, I did not include a variable for children in the home. Because that has been found to be a significant predictor of volunteering for nonseniors, I included it in this analysis. For consistency’s sake, I also reran the regression on seniors with the children-in-the-home variable included. The results were basically identical to those from the previous model. The children variable is dichotomous, and it was coded 0 (do not have children at home) or 1 (have children at home). Because of problems with other questions about children, researchers (Putnam & Yonish, 1998) have found this variable to be the one best suited for longitudinal analysis. In the interest of comparability, I omitted the civic-generation variable from all three models. This omission did not alter the seniors findings to any appreciable degree. 22. The coefficient on year in the senior model implies a seven-act increase relative to 1975 because .278 × 24 survey years ≈ 7. Likewise, the coefficient on year in the middle-age model implies a four-act increase relative to 1975 because .168 × 24 survey years ≈ 4. 23. Interestingly, financial worries, physical health, and marital status turned out to be nonsignificant (and substantively negligible) factors in middle-age adults’ volunteer frequency. 24. Chambré (1993) reviews a dozen studies of volunteering among older people and finds evidence that “a larger proportion of elders are currently engaged in volunteer work than 25 years ago” (p. 227). She attributes the change to increases in education and wealth, as well as to an expansion of private and government-sponsored volunteer programs. 25. I thank Robert Putnam for this metaphor. 26. The organization, Civic Ventures, was started with $1.3 million in foundation grants. Its headquarters is in Berkeley, California (Hall, 1997, p. 28). References Amato, P. R. (1990). Personality and social network involvement as predictors of helping behavior in everyday life. Social Psychology Quarterly, 53, 31-43. Auslander, G. K., & Litwin, H. (1988). Sociability and patterns of participation: Implications for social service policy. Journal of Voluntary Action Research, 17, 25-37. Berger, G. (1991). Factors explaining volunteering for organizations in general, and social welfare organizations in particular. Unpublished doctoral dissertation, Brandeis University, Waltham, MA. Carter, J. (1998). The Virtues of Aging. New York: Ballantine. Chambré, S. M. (1993). Volunteerism by elders: Past trends and future prospects. The Gerontologist, 33, 221-228. Clotfelter, C. T. (1985). Federal tax policy and charitable giving. Chicago: University of Chicago Press. Dempsey, N. P. (1988). Women and volunteerism: An examination of the changing patterns, types and motivations. Unpublished doctoral dissertation, Bryn Mawr College, PA. Freeman, R. B. (1996, January). Working for nothing: The supply of volunteer labor (NBER Working Paper 5435). Cambridge, MA: National Bureau of Economic Research. Hall, H. (1997, November 27). A new force in volunteerism. The Chronicle of Philanthropy, p. 28ff. Hodgkinson, V. A., & Weitzman, M. S. (1996). Giving and volunteering in the United States: Findings from a national survey, 1996 edition. Washington, DC: Independent Sector. Jackson, E. F., Bachmeier, M. D., Wood, J. R., & Craft, E. A. (1995). Volunteering and charitable giving: Do religious and associational ties promote helping behavior? Nonprofit and Voluntary Sector Quarterly, 24, 59-78. Kaplan, A. E. (1998). Giving USA 1998. New York: AAFRC Trust for Philantropy. Knoke, D., & Thomson, R. (1977). Voluntary association membership trends and the family life cycle. Social Forces, 56, 48-65. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011 Volunteering and the Long Civic Generation 415 Menchik, P. L., & Weisbrod, B. A. (1987, March). Volunteer labor supply. Journal of Public Economics, 32, 159-183. Putnam, R. D. (1996, Winter). The strange disappearance of civic America. The American Prospect, 24, 34-48. Putnam, R. D. (in press). Bowling alone: Civic disengagement in America and what to do about it. New York: Simon & Schuster. Putnam, R. D., & Yonish, S. (1998). New evidence on trends in American social capital and civic engagement: Are we really “bowling alone”? Unpublished manuscript. Robinson, J. P., & Godbey, G. (1997). Time for life: The surprising ways Americans use their time. University Park: The Pennsylvania State University Press. Romero, C. J. (1987). Retirement and older Americans’ participation in volunteer activities. In S. H. Sandell (Ed.), The problem isn’t age: Work and older Americans (pp. 218-227). New York: Praeger. Sasser, S. L. (1983). The relationship between selected demographic variables and participation in voluntary activities by mature women. Unpublished doctoral dissertation, Texas Woman’s University. Segal, L. M. (1993). Four essays on the supply of volunteer labor and econometrics. Unpublished doctoral dissertation, Northwestern University, Evanston, IL. Skocpol, T. (1999). Advocates without members: The recent transformation of American civic life. In T. Skocpol & M. P. Fiorina (Eds.), Civic engagement in American democracy. Washington, DC: Brookings Institution. Smith, D. H. (1994). Determinants of voluntary association participation and volunteering: A literature review. Nonprofit and Voluntary Sector Quarterly, 23, 243-263. Statham, A., & Rhoton, P. (1986, February). Mature and young women’s volunteer work, 1974-1981 (Center for Human Resource Research working paper). Columbus: Ohio State University. Wilson, J., & Janoski, T. (1995). The contribution of religion to volunteer work. Sociology of Religion, 56, 137-152. Wilson, J., & Musick, M. (1997). Who cares? Toward an integrated theory of volunteer work. American Sociological Review, 62, 694-713. Wilson, J., & Musick, M. (1998a). Attachment to volunteering. Unpublished manuscript. Wilson, J., & Musick, M (1998b). The contribution of social resources to volunteering. Unpublished manuscript. Kristin A. Goss is a graduate student in the Department of Government, Harvard University, and a recipient of a National Science Foundation research fellowship. For 6 years, she was a reporter for The Chronicle of Philanthropy. She has a master of public policy degree from Duke University and a B.A. from Harvard College. Downloaded from nvs.sagepub.com at DUKE UNIV LIBRARY on June 11, 2011