Copyright 1995 by New Prospect, Inc. Readers may redistribute this article to other individuals for noncommercial use, provided that the text, all html codes, and this notice remain intact and unaltered in any way. This article may not be resold, reprinted, or redistributed for compensation of any kind without prior written permission from the author. If you have any questions about permissions, please contact The Electronic Policy Network (email@example.com), P.O. 383080, Cambridge, MA 02238, or by phone at (617) 547-2950 (voice) or (617) 547-3896 (fax).
Preferred Citation: Robert D. Putnam, "The Strange Disappearance of
The American Prospect no. 24 (Winter 1996
See correction to this article.
by Robert D. Putnam
A more extended version of this article, complete with references, appears in the Winter 1995 issue of PS, a publication of the American Political Science Association. This work, originally delivered as the inaugural Ithiel de Sola Pool Lecture, builds on Putnam's earlier articles, "Bowling Alone: America's Declining Social Capital," Journal of Democracy (January 1995) and "The Prosperous Community," TAP (Spring 1993).
For American Prospect subscriptions and bulk reprints, call 1-800-872-0162.
For the last year or so, I have been wrestling with a difficult mystery. It is a classic brainteaser, with a corpus delicti, a crime scene strewn with clues, and many potential suspects. As in all good detective stories, however, some plausible miscreants turn out to have impeccable alibis, and some important clues hint at portentous developments that occurred before the curtain rose.
The mystery concerns the strange disappearance of social capital and civic engagement in America. By "social capital," I mean features of social life--networks, norms, and trust--that enable participants to act together more effectively to pursue shared objectives. (Whether or not their shared goals are praiseworthy is, of course, entirely another matter.) I use the term "civic engagement" to refer to people's connections with the life of their communities, not only with politics.
Although I am not yet sure that I have solved the mystery, I have assembled evidence that clarifies what happened. An important clue, as we shall see, involves differences among generations. Americans who came of age during the Depression and World War II have been far more deeply engaged in the life of their communities than the generations that have followed them. The passing of this "long civic generation" appears to be an important proximate cause of the decline of our civic life. This discovery does not in itself crack the case, but when combined with other data it points strongly to one suspect against whom I shall presently bring an indictment.
Evidence for the decline of social capital and civic engagement comes from a number of independent sources. Surveys of average Americans in 1965, 1975, and 1985, in which they recorded every single activity during a day--so-called "time-budget" studies--indicate that since 1965 time spent on informal socializing and visiting is down (perhaps by one-quarter) and time devoted to clubs and organizations is down even more sharply (by roughly half). Membership records of such diverse organizations as the PTA, the Elks club, the League of Women Voters, the Red Cross, labor unions, and even bowling leagues show that participation in many conventional voluntary associations has declined by roughly 25 percent to 50 percent over the last two to three decades. Surveys show sharp declines in many measures of collective political participation, including attending a rally or speech (off 36 percent between 1973 and 1993), attending a meeting on town or school affairs (off 39 percent), or working for a political party (off 56 percent).
Some of the most reliable evidence about trends comes from the General Social Survey (GSS), conducted nearly every year for more than two decades. The GSS demonstrates, at all levels of education and among both men and women, a drop of roughly one-quarter in group membership since 1974 and a drop of roughly one-third in social trust since 1972. (Trust in political authorities, indeed in many social institutions, has also declined sharply over the last three decades, but that is conceptually a distinct trend.) Slumping membership has afflicted all sorts of groups, from sports clubs and professional associations to literary discussion groups and labor unions. Only nationality groups, hobby and garden clubs, and the catch-all category of "other" seem to have resisted the ebbing tide. Gallup polls report that church attendance fell by roughly 15 percent during the 1960s and has remained at that lower level ever since, while data from the National Opinion Research Center suggest that the decline continued during the 1970s and 1980s and by now amounts to roughly 30 percent. A more complete audit of American social capital would need to account for apparent countertrends. Some observers believe, for example, that support groups and neighborhood watch groups are proliferating, and few deny that the last several decades have witnessed explosive growth in interest groups represented in Washington. The growth of such "mailing list" organizations as the American Association of Retired People and the Sierra Club, although highly significant in political (and commercial) terms, is not really a counterexample to the supposed decline in social connectedness, however, since these are not really associations in which members meet one another. Their members' ties are to common symbols and ideologies, but not to each other. Similarly, although most secondary associations are not-for-profit, most prominent nonprofits (from Harvard University to the Ford Foundation to the Metropolitan Opera) are bureaucracies, not secondary associations, so the growth of the "third sector" is not tantamount to a growth in social connectedness. With due regard to various kinds of counterevidence, I believe that the weight of available evidence confirms that Americans today are significantly less engaged with their communities than was true a generation ago.
Of course, American civil society is not moribund. Many good people across the land work hard every day to keep their communities vital. Indeed, evidence suggests that America still outranks many other countries in the degree of our community involvement and social trust. But if we examine our lives, not our aspirations, and if we compare ourselves not with other countries but with our parents, the best available evidence suggests that we are less connected with one another.
Reversing this trend depends, at least in part, on understanding the causes of the strange malady afflicting American civic life. This is the mystery I seek to unravel here: Why, beginning in the 1960s and accelerating in the 1970s and 1980s, did the fabric of American community life begin to fray? Why are more Americans bowling alone?
Many possible answers have been suggested for this puzzle:
The classic questions posed by a detective are means, motive, and opportunity. A solution, even a partial one, to our mystery must pass analogous tests.
Is the proposed explanatory factor correlated with trust and civic engagement? If not, that factor probably does not belong in the lineup. For example, if working women turn out to be more engaged in community life than housewives, it would be harder to attribute the downturn in community organizations to the rise of two-career families.
Is the correlation spurious? If parents, for example, were more likely than childless people to be joiners, that might be an important clue. However, if the correlation between parental status and civic engagement turned out to be entirely spurious, due to the effects of (say) age, we would have to remove the declining birth rate from our list of suspects.
Is the proposed explanatory factor changing in the relevant way? Suppose, for instance, that people who often move have shallower community roots. That could be an important part of the answer to our mystery only if residential mobility itself had risen during this period.
Is the proposed explanatory factor vulnerable to the claim that it might be the result of civic disengagement, not the cause? For example, even if newspaper readership were closely correlated with civic engagement across individuals and across time, we would need to weigh the degree to which reduced newspaper circulation is the result (not the cause) of disengagement.
Against those benchmarks, let us weigh the evidence. But first we must acknowledge a trend that only complicates our task.
Education is by far the strongest correlate that I have discovered of civic engagement in all its forms, including social trust and membership in many different types of groups. In fact, the effects of education become greater and greater as we move up the educational ladder. The four years of education between 14 and 18 total years have ten times more impact on trust and membership than the first four years of formal education. This curvilinear pattern applies to both men and women, and to all races and generations.
Sorting out just why education has such a massive effect on social connectedness would require a book in itself. Education is in part a proxy for social class and economic differences, but when income, social status, and education are used together to predict trust and group membership, education continues to be the primary influence. So, well-educated people are much more likely to be joiners and trusters, partly because they are better off economically, but mostly because of the skills, resources, and inclinations that were imparted to them at home and in school.
The expansion of high schools and colleges earlier this century has had an enormous impact on the educational composition of the adult population during just the last two decades. Since 1972 the proportion of adults with fewer than 12 years of education has been cut in half, falling from 40 percent to 18 percent, while the proportion with more than 12 years has nearly doubled, rising from 28 percent to 50 percent, as the generation of Americans educated around the turn of this century (most of whom did not finish high school) died off and were replaced by the baby boomers and their successors (most of whom attended college).
So here we have two facts--education boosts civic engagement sharply, and educational levels have risen massively--that only deepen our central mystery. By itself, the rise in educational levels should have increased social capital during the last 20 years by 15-20 percent, even assuming that the effects of education were merely linear. (Taking account of the curvilinear effect in figure 1, "Education and Civic Life," the rise in trusting and joining should have been even greater, as Americans moved up the accelerating curve.) By contrast, however, the actual GSS figures show a net decline since the early 1970s of roughly the same magnitude (trust by about 20-25 percent, memberships by about 15-20 percent). The relative declines in social capital are similar within each educational category--roughly 25 percent in group memberships and roughly 30 percent in social trust since the early 1970s, and probably even more since the early 1960s.
While this first investigative foray leaves us more mystified than before, we may nevertheless draw two useful conclusions. First, we need to take account of educational differences in our exploration of other factors to be sure that we do not confuse their effects with the consequences of education. And, second, the mysterious disengagement of the last quarter century seems to have afflicted all educational strata in our society, whether they have had graduate education or did not finish high school.
Many studies have found that residential stability and such related phenomena as homeownership are associated with greater civic engagement. At an earlier stage in this investigation I observed that "mobility, like frequent re-potting of plants, tends to disrupt root systems, and it takes time for an uprooted individual to put down new roots." I must now report, however, that further inquiry fully exonerates residential mobility from any responsibility for our fading civic engagement.
Data from the U.S. Bureau of the Census 1995 (and earlier years) show that rates of residential mobility have been remarkably constant over the last half century. In fact, to the extent that there has been any change at all, both long-distance and short-distance mobility have declined over the last five decades. During the 1950s, 20 percent of Americans changed residence each year and 6.9 percent annually moved across county borders; during the 1990s, the comparable figures are 17 percent and 6.6 percent. Americans, in short, are today slightly more rooted residentially than a generation ago. The verdict on mobility is unequivocal: This theory is simply wrong.
But if moving itself has not eroded our social capital, what about the possibility that we have moved to places, especially suburbs, that are less congenial to social connectedness? In fact, social connectedness does differ by community type, but the differences turn out to be modest and in directions that are inconsistent with the theory.
Controlling for such characteristics as education, age, income, work status, and race, citizens of the nation's 12 largest metropolitan areas (particularly their central cities, but also their suburbs) are roughly 10 percent less trusting and report 10-20 percent fewer group memberships than residents of other cities and towns (and their suburbs). Meanwhile, residents of very small towns and rural areas are (in accord with some hoary stereotypes) slightly more trusting and civicly engaged than other Americans. Unsurprisingly, the prominence of different types of groups does vary significantly by location: Major cities have more political and nationality clubs; smaller cities more fraternal, service, hobby, veterans', and church groups: and rural areas more agricultural organizations. But overall rates of associational membership are not very different.
Moreover, this pattern cannot account for our central puzzle. In the first place, there is virtually no correlation between gains in population and losses in social capital, either across states or across localities of different sizes. Even taking into account the educational and social backgrounds of those who have moved there, the suburbs have faintly higher levels of trust and civic engagement than their respective central cities, which should have produced growth, not decay, in social capital over the last generation. The central point, however, is that the downtrends in trusting and joining are virtually identical everywhere--in cities, big and small, in suburbs, in small towns, and in the countryside.
Of course, Evanston is not Levittown is not Sun City. The evidence available does not allow us to determine whether different types of suburban living have different effects on civic connections and social trust. However, these data do rule out the thesis that suburbanization per se has caused the erosion of America's social capital. Both where we live and how long we've lived there matter for social capital, but neither explains why it is eroding everywhere.
Americans certainly feel busier now than a generation ago: The proportion of us who report feeling "always rushed" jumped by half between the mid-1960s and the mid-1990s. Probably the most obvious suspect behind our tendency to drop out of community affairs is pervasive busy-ness. And lurking nearby in the shadows are economic pressures so much discussed nowadays, from job insecurity to declining real wages.
Yet, however culpable busy-ness and economic insecurity may appear at first glance, it is hard to find incriminating evidence. In the first place, time-budget studies do not confirm the thesis that Americans are, on average, working longer than a generation ago. On the contrary, a new study by John Robinson and Geoffrey Godbey of the University of Maryland reports a five hour per week gain in free time for the average American between 1965 and 1985, due partly to reduced time spent on housework and partly to earlier retirement. Their claim that Americans have more leisure time now than several decades ago is, to be sure, contested by other observers, notably Juliet Schor, who in her 1991 book The Overworked American reports evidence that work hours are lengthening, especially for women.
But whatever the resolution of that controversy, other data call into question whether longer hours at work lead to lessened involvement in civic life or reduced social trust. Results from the GSS show that employed people belong to somewhat more groups than those outside the paid labor force. Even more striking is the fact that among workers, longer hours are linked to more civic engagement. The patterns among men and women on this score are not identical: Women who work part-time appear to be somewhat more civicly engaged and socially trusting than either those who work full-time or those who do not work outside the home at all--an intriguing anomaly, though not relevant to our basic puzzle, since female part-time workers constitute a relatively small fraction of the American population, and the fraction is growing, up from about 8 percent to about 10 percent between the early 1970s and early 1990s.
But what do workaholics do less? Robinson reports that, unsurprisingly, people who spend more time at work do feel more rushed, and these harried souls do spend less time eating, sleeping, reading books, engaging in hobbies, and just doing nothing. Compared to the rest of the population, they also spend a lot less time watching television, almost 30 percent less. However, they do not spend less time on organizational activity. In short, those who work longer forego Nightline, but not the Kiwanis club; ER, but not the Red Cross.
So hard work does not prevent civic engagement. Moreover, the nationwide falloff in joining and trusting is perfectly mirrored among full-time workers, among part-time workers, and among those outside the paid labor force. So if people are dropping out of community life, long hours do not seem to be the reason.
If time pressure is not the culprit, how about financial pressures? It is true that people with lower incomes and those who feel financially strapped are somewhat less engaged in community life and somewhat less trusting than those who are better off, even holding education constant. On the other hand, the downtrends in social trust and civic engagement are visible among people of all incomes, with no sign whatever that they are concentrated among those who have borne the brunt of the economic distress of the last two decades. Quite the contrary, the declines in engagement and trust are actually somewhat greater among the more affluent segments of the American public than among the poor and middle-income wage-earners. Moreover, personal financial satisfaction is wholly uncorrelated with civic engagement and social trust. In short, neither objective nor subjective economic well-being has inoculated Americans against the virus of civic disengagement; if anything, affluence has slightly exacerbated the problem. Poverty and economic inequality are dreadful, growing problems for America, but they are not the villains of this piece.
Most of our mothers were housewives, and most of them invested heavily in social capital formation--a jargony way of referring to untold unpaid hours in church suppers, PTA meetings, neighborhood coffee klatches, and visits to friends and relatives. The movement of women out of the home and into the paid labor force is probably the most portentous social change of the last half century. However welcome and overdue the feminist revolution may be, it is hard to believe that it has had no impact on social connectedness. Could this be the primary reason for the decline of social capital over the last generation?
Some patterns in the survey evidence seem to support this claim. All things considered, women belong to somewhat fewer voluntary associations than men do. On the other hand, time-budget studies suggest that women spend more time on those groups and more time in informal social connecting than men. Although the absolute declines in joining and trusting are approximately equivalent among men and women, the relative declines are somewhat greater among women. Controlling for education, memberships among men have declined at a rate of about 10-15 percent a decade, compared to about 20-25 percent a decade for women. The time-budget data, too, strongly suggest that the decline in organizational involvement in recent years is concentrated among women. These sorts of facts, coupled with the obvious transformation in the professional role of women over this same period, led me in previous work to suppose that the emergence of two-career families might be the most important single factor in the erosion of social capital.
As we saw earlier, however, work status itself seems to have little net impact on group membership or on trust. Housewives belong to different types of groups than do working women (more PTAs, for example, and fewer professional associations), but in the aggregate working women are actually members of slightly more voluntary associations (though housewives, according to Robinson and Godbey, spend more time on them). Moreover, the overall declines in civic engagement are somewhat greater among housewives than among employed women. Comparison of time-budget data between 1965 and 1985 seems to show that employed women as a group are actually spending more time on organizations than before, while housewives are spending less. This same study suggests that the major decline in informal socializing since 1965 has also been concentrated among housewives. The central fact, of course, is that the overall trends are down for all categories of women (and for men, too, even bachelors), but the figures suggest that women who work full-time actually may have been more resistant to the slump than those who do not.
Thus, although women appear to have borne a disproportionate share of the decline in civic engagement over the last two decades, it is not easy to find any micro-level data that tie that fact directly to their entry into the labor force. Of course, women who have chosen to enter the workforce doubtless differ in many respects from women who have chosen to stay home. Perhaps one reason that community involvement appears to be rising among working women and declining among housewives is that precisely the sort of women who, in an earlier era, were most involved with their communities have been disproportionately likely to enter the workforce, thus lowering the average level of civic engagement among the remaining homemakers and raising the average among women in the workplace.
No doubt the movement of women into the workplace over the last generation has changed the types of organizations to which they belong. Contrary to my own earlier speculations, however, I can find little evidence to support the hypothesis that this movement has played a major role in the net reduction of social connectedness and civic engagement. On the other hand, I have no clear alternative explanation for the fact that the relative declines are greater among women, both those who work outside the home and those who don't, than among men. Since this evidence is at best circumstantial, perhaps the best interim judgment here is the famous Scots verdict: not proven.
Another widely discussed social trend that more or less coincides with the downturn in civic engagement is the breakdown of the traditional family unit--mom, dad, and the kids. Since the family itself is, by some accounts, a key form of social capital, perhaps its eclipse is part of the explanation for the reduction in joining and trusting in the wider community. What does the evidence show?
First of all, evidence of the loosening of family bonds is unequivocal. In addition to the century-long increase in divorce rates (which accelerated from the mid-1960s to the mid-1970s and then leveled off), and the more recent increase in single-parent families, the incidence of one-person households has more than doubled since 1950, in part because of the rising number of widows living alone. The net effect of all these changes, as reflected in the General Social Survey, is that the proportion of all American adults currently unmarried climbed from 28 percent in 1974 to 48 percent in 1994.
Second, married men and women do rank somewhat higher on both our measures of social capital. That is, controlling for education, age, race, and so on, single people--both men and women, divorced, separated, and never married--are significantly less trusting and less engaged civicly than married people. (Multivariate analysis hints that one major reason why divorce lowers connectedness is that it lowers family income, which in turn reduces civic engagement.) Roughly speaking, married men and women are about a third more trusting and belong to about 15-25 percent more groups than comparable single men and women. (Widows and widowers are more like married people than single people in this comparison.)
In short, successful marriage, especially if the family includes children, is statistically associated with greater social trust and civic engagement. Thus, some part of the decline in both trust and membership is tied to the decline in marriage. To be sure, the direction of causality behind this correlation may be complicated, since it is conceivable that loners and paranoids are harder to live with. If so, divorce may in some degree be the consequence, not the cause, of lower social capital. Probably the most reasonable summary of these arrays of data, however, is that the decline in successful marriage is a significant, though modest part of the reason for declining trust and lower group membership. On the other hand, changes in family structure cannot be a major part of our story, since the overall declines in joining and trusting are substantial even among the happily married. My own verdict (based in part on additional evidence to be introduced later) is that the disintegration of marriage is probably an accessory to the crime, but not the major villain of the piece.
Circumstantial evidence, particularly the timing of the downturn in social connectedness, has suggested to some observers that an important cause--perhaps even the cause--is big government and the growth of the welfare state. By "crowding out" private initiative, it is argued, state intervention has subverted civil society.
Some government policies have almost certainly had the effect of destroying social capital. For example, the so-called "slum clearance" policies of the 1950s and 1960s replaced physical capital, but destroyed social capital, by disrupting existing community ties. It is also conceivable that certain social expenditures and tax policies may have created disincentives for civic-minded philanthropy. On the other hand, it is much harder to see which government policies might be responsible for the decline in bowling leagues and literary clubs. Some community institutions sponsored, organized, or subsidized by government, such as National Service, agricultural extension programs, and Head Start, may enhance trust and social capital. Which effect prevails needs to be resolved with evidence, not ideology.
One empirical approach to this issue is to examine differences in civic engagement and public policy across different political jurisdictions to see whether enlarged government leads to shriveled social capital. Among the U.S. states, however, differences in social capital appear essentially uncorrelated with various measures of welfare spending or government size. Citizens in free-spending states are no less trusting or engaged than citizens in frugal ones.
Cross-national comparison can also shed light on this question. Among nineteen member countries of the Organization of Economic Cooperation and Development (OECD) for which data on social trust and group membership are available from the 1990-1991 World Values Survey, these indicators of social capital are, if anything, positively correlated with the size of the state. This simple bivariate analysis, of course, cannot tell us whether social connectedness encourages welfare spending, whether the welfare state fosters civic engagement, or whether both are the result of some other unmeasured factor(s). Even this simple finding, however, is not easily reconciled with the notion that big government undermines social capital.
Some observers have noted that the decline in social connectedness began just after the successes of the civil rights revolution of the 1960s. That coincidence has suggested the possibility of a kind of sociological "white flight," as legal desegregation of civic life led whites to withdraw from community associations.
The erosion of social capital, however, has affected all races. In fact, during the 1980s the downturns in both joining and trusting were even greater among African Americans (and other racial minorities) than among the white majority. This fact is inconsistent with the thesis that "white flight" is a significant cause of civic disengagement, since black Americans have been dropping out of religious and civic organizations at least as rapidly as white Americans. Even more important, the pace of disengagement among whites has been uncorrelated with racial intolerance or support for segregation. Avowedly racist or segregationist whites have been no quicker to drop out of community organizations during this period than more tolerant whites.
This evidence is far from conclusive, of course, but it does shift the burden of proof onto those who believe that racism is a primary explanation for growing civic disengagement over the last quarter century. This evidence also suggests that reversing the civil rights gains of the last thirty years would do nothing to reverse the social capital losses.
Our efforts thus far to identify the major sources of civic disengagement have been singularly unfruitful. In all our statistical analyses, however, one factor, second only to education, stands out as a predictor of all forms of civic engagement and trust. That factor is age. Older people belong to more organizations than young people, and they are less misanthropic. Older Americans also vote more often and read newspapers more frequently, two other forms of civic engagement closely correlated with joining and trusting.
"Civic Engagement by Age" shows the basic pattern. Civic involvement appears to rise more or less steadily from early adulthood toward a plateau in middle age, from which it declines only late in life. This humpback pattern seems naturally to represent the arc of life's engagements. That, at least, was how I first interpreted the data. But that would be a fundamental misreading of the most important clue in our whole whodunit.
Evidence from the General Social Survey enables us to follow individual cohorts as they age. If the rising lines in the figure indeed represent deepening civic engagement with age, we should be able to track this same deepening engagement as we follow, for example, the first of the baby boomers, born in 1947, as they aged from 25 in 1972 (the first year of the GSS) to 47 in 1994 (the latest year available). Startlingly, however, such an analysis, repeated for successive birth cohorts, produces virtually no evidence of such life cycle changes in civic engagement. In fact, as various generations moved through the period between 1972 and 1994, their levels of trust and membership more often fell than rose, reflecting a more or less simultaneous decline in civic engagement among young and old alike, particularly during the second half of the 1980s. But that downtrend obviously cannot explain why, throughout the period, older Americans were always more trusting and engaged. In fact, the only reliable life cycle effect visible in these data is a withdrawal from civic engagement very late in life, as we move through our eighties.
The central paradox posed by these patterns is this: Older people are consistently more engaged and trusting than younger people, yet we do not become more engaged and trusting as we age. What's going on here?
Time and age are notoriously ambiguous in their effects on social behavior. Social scientists have learned to distinguish three contrasting phenomena:
Life cycle effects represent differences attributable to stage of life. In this case individuals change as they age, but since the effects of aging are, in the aggregate, neatly balanced by the "demographic metabolism" of births and deaths, life cycle effects produce no aggregate change. Everyone's close-focus eyesight worsens as we age, but the aggregate demand for reading glasses changes little.
Period effects affect all people who live through a given era, regardless of their age. Period effects can produce both individual and aggregate change, often quickly and enduringly, without any age-related differences. The sharp drop in trust in government between 1965 and 1975, for example, was almost entirely this sort of period effect, as Americans of all ages changed their minds about their leaders' trustworthiness. Similarly, as just noted, a modest portion of the decline in social capital during the 1980s appears to be a period effect.
Generational effects affect all people born at the same time. Like life cycle effects (and unlike typical period effects), generational effects show up as disparities among age groups at a single point in time, but like period effects (and unlike life cycle effects) generational effects produce real social change, as successive generations, enduringly "imprinted" with divergent outlooks, enter and leave the population. In pure generational effects, no individual ever changes, but society does.
Returning to our conundrum, how could older people today be more engaged and trusting, if they did not become more engaged and trusting as they aged? The key to this paradox, as David Butler and Donald Stokes observed in another context, is to ask, not how old people are, but when they were young. The figure "Social Capital and Civic Engagement by Generation," addresses this reformulated question, displaying various measures of civic engagement according to the respondents' year of birth.
In effect, the figure on the bottom of page 43 lines up Americans from left to right according to their date of birth, beginning with those born in the last third of the nineteenth century and continuing across to the generation of their great-grandchildren, born in the last third of the twentieth century. As we begin moving along this queue from left to right--from those raised around the turn of the century to those raised during the Roaring Twenties, and so on--we find relatively high and unevenly rising levels of civic engagement and social trust. Then rather abruptly, however, we encounter signs of reduced community involvement, starting with men and women born in the early 1930s. Remarkably, this downward trend in joining, trusting, voting, and newspaper reading continues almost uninterruptedly for nearly 40 years. The trajectories for the various different indicators of civic engagement are strikingly parallel: Each shows a high, sometimes rising plateau for people born and raised during the first third of the century; each shows a turning point in the cohorts around 1930; and each then shows a more or less constant decline down to the cohorts born during the 1960s.
By any standard, these intergenerational differences are extraordinary. Compare, for example, the generation born in the early 1920s with the generation of their grandchildren born in the late 1960s. Controlling for educational disparities, members of the generation born in the 1920s belong to almost twice as many civic associations as those born in the late 1960s (roughly 1.9 memberships per capita, compared to roughly 1.1 memberships per capita). The grandparents are more than twice as likely to trust other people (50-60 percent compared with 25 percent for the grandchildren). They vote at nearly double the rate of the most recent cohorts (roughly 75 percent compared with 40-45 percent), and they read newspapers almost three times as often (70-80 percent read a paper daily compared with 25-30 percent). And bear in mind that we have found no evidence that the youngest generation will come to match their grandparents' higher levels of civic engagement as they grow older.
Thus, read not as life cycle effects, but rather as generational effects, the age-related patterns in our data suggest a radically different interpretation of our basic puzzle. Deciphered with this key, the figure on page 43 depicts a long "civic" generation, born roughly between 1910 and 1940, a broad group of people substantially more engaged in community affairs and substantially more trusting than those younger than they. (Members of the 1910-1940 generation also seem more civic than their elders, at least to judge by the outlooks of relatively few men and women born in the late nineteenth century who appeared in our samples.) The culminating point of this civic generation is the cohort born in 1925-1930, who attended grade school during the Great Depression, spent World War II in high school (or on the battlefield), first voted in 1948 or 1952, set up housekeeping in the 1950s, and watched their first television when they were in their late twenties. Since national surveying began, this cohort has been exceptionally civic: voting more, joining more, reading newspapers more, trusting more. As the distinguished sociologist Charles Tilly (born in 1928) said in commenting on an early version of this essay, "We are the last suckers."
To help in interpreting the historical contexts within which these successive generations of Americans matured, the figure also indicates the decade within which each cohort came of age. Thus, we can see that each generation that reached adulthood since the 1940s has been less engaged in community affairs than its immediate predecessor.
Further confirmation of this generational interpretation comes from a comparison of the two parallel lines that chart responses to an identical question about social trust, posed first in the National Election Studies (mainly between 1964 and 1976) and then in the General Social Survey between 1972 and 1994. If the greater trust expressed by Americans born earlier in the century represented a life cycle effect, then the graph from the GSS surveys (conducted when these cohorts were, on average, 10 years older) should have been some distance above the NES line. In fact, the GSS line lies about 5-10 percent below the NES line. That downward shift almost surely represents a period effect that depressed social trust among all cohorts during the 1980s. That downward period effect, however, is substantially more modest than the large generational differences already noted.
In short, the most parsimonious interpretation of the age-related differences in civic engagement is that they represent a powerful reduction in civic engagement among Americans who came of age in the decades after World War II, as well as some modest additional disengagement that affected all cohorts during the 1980s. These patterns hint that being raised after World War II was a quite different experience from being raised before that watershed. It is as though the postwar generations were exposed to some mysterious X-ray that permanently and increasingly rendered them less likely to connect with the community. Whatever that force might have been, it--rather than anything that happened during the 1970s and 1980s--accounts for most of the civic disengagement that lies at the core of our mystery.
But if this reinterpretation of our puzzle is correct, why did it take so long for the effects of that mysterious X-ray to become manifest? If the underlying causes of civic disengagement can be traced to the 1940s and 1950s, why did the effects become conspicuous in PTA meetings and Masonic lodges, in the volunteer lists of the Red Cross and the Boy Scouts, and in polling stations and church pews and bowling alleys across the land only during the 1960s, 1970s, and 1980s?
The visible effects of this generational disengagement were delayed by two important factors. First, the postwar boom in college enrollments raised levels of civic engagement, offsetting the generational trends. As Warren E. Miller and J. Merrill Shanks observe in their as yet unpublished book, The American Voter Reconsidered, the postwar expansion of educational opportunities "forestalled a cataclysmic drop" in voting turnout, and it had a similar delaying effect on civic disengagement more generally.
Second, the full effects of generational developments generally appear several decades after their onset, because it takes that long for a given generation to become numerically dominant in the adult population. Only after the mid-1960s did significant numbers of the "post-civic generation" reach adulthood, supplanting older, more civic cohorts. The figure "The Rise and Decline of a Civic Generation" illustrates this generational accounting. The long civic generation born between 1910 and 1940 reached its zenith in 1960, when it comprised 62 percent of those who chose between John Kennedy and Richard Nixon. By the time that Bill Clinton was elected president in 1992, that cohort's share in the electorate had been cut precisely in half. Conversely, over the last two decades (from 1974 to 1994) boomers and X-ers (that is, Americans born after 1946) have grown as a fraction of the adult population from 24 percent to 60 percent.
In short, the very decades that have seen a national deterioration in social capital are the same decades during which the numerical dominance of a trusting and civic generation has been replaced by the dominion of "post-civic" cohorts. Moreover, although the long civic generation has enjoyed unprecedented life expectancy, allowing its members to contribute more than their share to American social capital in recent decades, they are now passing from the scene. Even the youngest members of that generation will reach retirement age within the next few years. Thus, a generational analysis leads almost inevitably to the conclusion that the national slump in trust and engagement is likely to continue, regardless of whether the more modest "period effect" depression of the 1980s continues.
To say that civic disengagement in contemporary America is in large measure generational merely reformulates our central puzzle. We now know that much of the cause of our lonely bowling probably dates to the 1940s and 1950s, rather than to the 1960s and 1970s. What could have been the mysterious anticivic "X-ray" that affected Americans who came of age after World War II and whose effects progressively deepened at least into the 1970s?
Our new formulation of the puzzle opens the possibility that the zeitgeist of national unity, patriotism, and shared sacrifice that culminated in 1945 might have reinforced civic-mindedness. On the other hand, it is hard to assign any consistent role to the Cold War and the Bomb, since the anticivic trend appears to have deepened steadily from the 1940s to the 1970s, in no obvious harmony with the rhythms of world affairs. Nor is it easy to construct an interpretation of the data on generational differences in which the cultural vicissitudes of the sixties could play a significant role. Neither can economic adversity or affluence easily be tied to the generational decline in civic engagement, since the slump seems to have affected in equal measure those who came of age in the placid fifties, the booming sixties, and the busted seventies.
I have discovered only one prominent suspect against whom circumstantial evidence can be mounted, and in this case, it turns out, some directly incriminating evidence has also turned up. This is not the occasion to lay out the full case for the prosecution, nor to review rebuttal evidence for the defense, but I want to present evidence that justifies indictment.
The culprit is television.
First, the timing fits. The long civic generation was the last cohort of Americans to grow up without television, for television flashed into American society like lightning in the 1950s. In 1950 barely 10 percent of American homes had television sets, but by 1959, 90 percent did, probably the fastest diffusion of a major technological innovation ever recorded. The reverberations from this lightning bolt continued for decades, as viewing hours grew by 17-20 percent during the 1960s and by an additional 7-8 percent during the 1970s. In the early years, TV watching was concentrated among the less educated sectors of the population, but during the 1970s the viewing time of the more educated sectors of the population began to converge upward. Television viewing increases with age, particularly upon retirement, but each generation since the introduction of television has begun its life cycle at a higher starting point. By 1995 viewing per TV household was more than 50 percent higher than it had been in the 1950s.
Most studies estimate that the average American now watches roughly four hours per day (excluding periods in which television is merely playing in the background). Even a more conservative estimate of three hours means that television absorbs 40 percent of the average American's free time, an increase of about one-third since 1965. Moreover, multiple sets have proliferated: By the late 1980s three-quarters of all U.S. homes had more than one set, and these numbers too are rising steadily, allowing ever more private viewing. Robinson and Godbey are surely right to conclude that "television is the 800-pound gorilla of leisure time." This massive change in the way Americans spend their days and nights occurred precisely during the years of generational civic disengagement.
Evidence of a link between the arrival of television and the erosion of social connections is, however, not merely circumstantial. The links between civic engagement and television viewing can be instructively compared with the links between civic engagement and newspaper reading. The basic contrast is straightforward: Newspaper reading is associated with high social capital, TV viewing with low social capital.
Controlling for education, income, age, race, place of residence, work status, and gender, TV viewing is strongly and negatively related to social trust and group membership, whereas the same correlations with newspaper reading are positive. Within every educational category, heavy readers are avid joiners, whereas heavy viewers are more likely to be loners. In fact, more detailed analysis suggests that heavy TV watching is one important reason why less educated people are less engaged in the life of their communities. Controlling for differential TV exposure significantly reduces the correlation between education and engagement.
Viewing and reading are themselves uncorrelated--some people do lots of both, some do little of either--but "pure readers" (that is, people who watch less TV than average and read more newspapers than average) belong to 76 percent more civic organizations than "pure viewers" (controlling for education, as always). Precisely the same pattern applies to other indicators of civic engagement, including social trust and voting turnout. "Pure readers," for example, are 55 percent more trusting than "pure viewers."
In other words, each hour spent viewing television is associated with less social trust and less group membership, while each hour reading a newspaper is associated with more. An increase in television viewing of the magnitude that the U.S. has experienced in the last four decades might directly account for as much as one-quarter to one- half of the total drop in social capital, even without taking into account, for example, the indirect effects of television viewing on newspaper readership or the cumulative effects of lifetime viewing hours. Newspaper circulation (per household) has dropped by more than half since its peak in 1947. To be sure, it is not clear which way the tie between newspaper reading and civic involvement works, since disengagement might itself dampen one's interest in community news. But the two trends are clearly linked.
Time displacement. Even though there are only 24 hours in everyone's day, most forms of social and media participation are positively correlated. People who listen to lots of classical music are more likely, not less likely, than others to attend Cubs games. Television is the principal exception to this generalization--the only leisure activity that seems to inhibit participation outside the home. TV watching comes at the expense of nearly every social activity outside the home, especially social gatherings and informal conversations. TV viewers are homebodies.
Most studies that report a negative correlation between television watching and community involvement (see figure "The TV Connection") are ambiguous with respect to causality, because they merely compare different individuals at a single time. However, one important quasi-experimental study of the introduction of television in three Canadian towns found the same pattern at the aggregate level across time. A major effect of television's arrival was the reduction in participation in social, recreational, and community activities among people of all ages. In short, television privatizes our leisure time.
Effects on the outlooks of viewers. An impressive body of literature suggests that heavy watchers of TV are unusually skeptical about the benevolence of other people--overestimating crime rates, for example. This body of literature has generated much debate about the underlying causal patterns, with skeptics suggesting that misanthropy may foster couch-potato behavior rather than the reverse. While awaiting better experimental evidence, however, a reasonable interim judgment is that heavy television watching may well increase pessimism about human nature. Perhaps too, as social critics have long argued, both the medium and the message have more basic effects on our ways of interacting with the world and with one another. Television may induce passivity, as Neil Postman has claimed.
Effects on children. TV consumes an extraordinary part of children's lives, about 40 hours per week on average. Viewing is especially high among pre-adolescents, but it remains high among younger adolescents: Time-budget studies suggest that among youngsters aged 9 to 14 television consumes as much time as all other discretionary activities combined, including playing, hobbies, clubs, outdoor activities, informal visiting, and just hanging out. The effects of television on childhood socialization have, of course, been hotly debated for more than three decades. The most reasonable conclusion from a welter of sometimes conflicting results appears to be that heavy television watching probably increases aggressiveness (although perhaps not actual violence), that it probably reduces school achievement, and that it is statistically associated with "psychosocial malfunctioning," although how much of this effect is self-selection and how much causal remains much debated. The evidence is, as I have said, not yet enough to convict, but the defense has a lot of explaining to do.
More than two decades ago, just as the first signs of disengagement were beginning to appear in American politics, the political scientist Ithiel de Sola Pool observed that the central issue would be--it was then too soon to judge, as he rightly noted--whether the development represented a temporary change in the weather or a more enduring change in the climate. It now appears that much of the change whose initial signs he spotted did in fact reflect a climatic shift.
Moreover, just as the erosion of the ozone layer was detected only many years after the proliferation of the chlorofluorocarbons that caused it, so too the erosion of America's social capital became visible only several decades after the underlying process had begun. Like Minerva's owl that flies at dusk, we come to appreciate how important the long civic generation has been to American community life just as its members are retiring. Unless America experiences a dramatic upward boost in civic engagement (a favorable "period effect") in the next few years, Americans in 2010 will join, trust, and vote even less than we do today.
In an astonishingly prescient book, Technologies without Borders, published in 1991 after his death, Pool concluded that the electronic revolution in communications technology was the first major technological advance in centuries that would have a profoundly decentralizing and fragmenting effect on society and culture. He hoped that the result might be "community without contiguity." As a classic liberal, he welcomed the benefits of technological change for individual freedom, and in part, I share that enthusiasm. Those of us who bemoan the decline of community in contemporary America need to be sensitive to the liberating gains achieved during the same decades. We need to avoid an uncritical nostalgia for the fifties. On the other hand, some of the same freedom-friendly technologies whose rise Pool predicted may indeed be undermining our connections with one another and with our communities. Pool defended what he called "soft technological determinism" because he recognized that social values cans condition the effects of technology. This perspective invites us not merely to consider how technology is privatizing our lives--if, as it seems to me, it is--but to ask whether we like the result, and if not, what we might do about it. Those are questions we should, of course, be asking together, not alone.
Figure 1."Education and Civic Life" shows a strong correlation between social trust and group membership on the one hand and and years of education on the other.
Figure 2."Civic Engagements by Age (education controlled)" shows civic engagement to increase with age.
Figure 3."Social Capital and Civic Engagement by Generation (education controlled)" shows an overall decline in social capital and civic engagement in the age cohorts that turned 18 after the 1940s.
Figure 4. "The Rise and Decline of a Civic Generation" shows that the civically-engaged generation born between 1911 and 1940 (and its sub-cohort born between 1921 and 1935) was at its greatest numerical concentration in 1960.
Figure 5. "The TV Connection" shows that group membership tends to decline as television viewing increases among those having twelve or more years of education.