Copyright Sociological Research Online, 2002


Yaojun Li, Mike Savage, Gindo Tampubolon, Alan Warde and Mark Tomlinson (2002) 'Dynamics Of Social Capital: Trends And Turnover In Associational Membership In England And Wales, 1972-1999'
Sociological Research Online, vol. 7, no. 3, <>

To cite articles published in Sociological Research Online, please reference the above information and include paragraph numbers if necessary

Received: 15/10/2002      Accepted: 15/10/2002      Published:


Following the important recent work of Robert Putnam, there is considerable current debate about whether the volume of 'social capital' in western societies is in decline and if so what might be the implications for political democracy. Evaluations of the arguments are difficult both because the concept of social capital is a contested one and because measuring social capital is difficult. This paper focuses on membership of voluntary associations in England and Wales as a key measure of social capital and analyses trends in associational membership and their social determinants using the Oxford Mobility Study and British Household Panel Survey. We show that focusing on seven associations there is a broad pattern of stability in membership with the striking and remarkable exception of falling male membership of trade unions and working-men?s clubs. We see this as testimony to a class polarisation in membership in which working class men have been increasingly marginalised from associational memberships. Our conclusion argues that if the membership of voluntary associations is to be used as an index of social capital, there is an increasing social skewing of membership and an intensifying service class hegemony over social capital which poses major concerns for its potential to sustain democratic politics.



It is striking that in a period frequently heralded to have witnessed far reaching processes of globalisation (see the overview in Ray, this volume), there has been a striking resurgence of interest from social scientists in the significance of small-scale 'communal' practices (see the discussion in Crow, this volume). The recent work of Robert Putnam (1993, 1995, 2000), in particular, has focused attention on the way that small-scale solidarities, defined as 'social capital', have crucial implications for social well being, but may be undergoing decline. He claims that social capital is crucial for ensuring the vitality of democratic institutions, but the argument has been extended so that it is also seen as important for generating good health (Wilkinson, 1996), trust and economic prosperity (Fukuyama 1995), and so forth. Although Putnam does not make the link, his argument that social capital has been declining in the US since the 1950s can be held to be consistent with theories of globalisation which imply the erosion of local attachment. Indeed, it could be argued that measuring trends in social capital can thereby become an important way of documenting globalising trends, especially valuable given the imprecision of this concept (see also the remarks of Urry 2002). Consideration of the argument, however, is difficult because the concept of social capital is a contested one and there is no agreement about how it can best be measured. This paper analyses the dynamics of associational membership in England and Wales since the 1970s using the Oxford Mobility Study (OMS) of 1972 and the British Household Panel Study (BHPS). We build on the analysis of the BHPS by Warde et al (2003) to show that there has been general stability of membership for most associations since the 1970s, but that a dramatic decline in male membership in trade unions and working men's clubs has had significant effects on the distribution of social capital. We develop the significance of this finding by showing increasing social polarisation of between members and non-members of associations over time. A final section considers the importance of turnover of membership. In conclusion, we offer an account of the progress of social capital in England and Wales in the 1990s and an empirically-based critique of the concept of social capital.

1. The Debate on Social Capital

The analysis of the nature of political participation has recently been enlivened by the extensive debate about the work of Putnam on the decline of social capital in America provides the background to this paper. Social capital is a contested concept (compare, for example, Bourdieu 1984; Bourdieu and Wacquant 1992; Coleman 1988, 1990; and Putnam 1995, 2000; and see the discussion in Portes 1998), but in all cases it imputes links between forms of trust, social networks and attachments. For Putnam, 'social capital refers to connections among individuals – social networks and the norms of reciprocity and trustworthiness that arise from them' (Putnam 2000: 19). It is a private and a public good. Benefits are greater from connections in well-connected communities; and generalised reciprocity is more efficient than a distrustful society. 'Trustworthiness lubricates social life' (Putnam 2000: 21). He admits that social capital comes in many forms, and it 'can be directed towards malevolent, antisocial purposes' (22). However, Putnam mostly considers it as overwhelmingly positive, and as an emergent collective property which increases social trust.

Putnam recognises that there are different types of social capital and distinguishes between bonding and bridging capital. Implicitly drawing on Granovetter's (1973) distinction between strong and weak ties, he claims that 'of all the dimensions along which forms of social capital vary, perhaps the most important is the distinction between bridging (or inclusive) and bonding (or exclusive). … Bonding social capital is good for undergirding specific reciprocity and mobilizing solidarity. … Bridging networks, by contrast, are better for linkage to external assets and for information diffusion' (22). He goes on to say that 'bonding and bridging are not "either/or" categories into which social networks can be neatly divided, but "more or less"' (23). This distinction is, however, difficult to operationalise, as indeed is the concept in any of its definitions.

Social capital has been measured in many ways (see eg World Bank 1997). Putnam, in Bowling Alone, relies on a neo-Tocquevillian framework which attaches particular importance to involvement in voluntary associations and charitable activity, though he is also interested in engagement in social movements, workplace interaction, informal social connections, and in the presence of altruistic values and trust. One of the merits of Putnam's work is that he is able to thereby suggest that voluntary association membership, hitherto regarded as largely of little social importance, is in fact of pressing social importance. His analysis of membership trends in civic, religious, voluntary and political groups indicates that the overall level of social capital in America is diminishing. This is of concern to him because he thinks that this entails decline of social and political trust, and that low levels of trust have detrimental consequences. It is possible to detect in Putnam's account an implicit nostalgic veneration of small town American life: he celebrates their voluntary associations and civic associations, but is suspicious of television, car driving and virtual networks linked to the internet (see also Urry 2002). He thus attributes the decline of social capital in minor part to the effect of the spread of dual earner households and longer commuting time, more substantially to increased television viewing, and in even greater part to a generational shift which involves the replacement of a highly engaged inter-war generation by younger cohorts, born since the second world war, who are much less involved in social and civic life. With few counter-trends apparent, Putnam is pessimistic about the survival of a trusting civic culture. His argument intimates that the same process will be occurring in other western societies where similar social trends can be observed.

Hall's (1999) account of trends in social capital in Britain argues that the declining social capital found in the USA does not apply. Many of the key indicators used by Putnam (2000) show no obvious parallel tendencies in Britain. Thus electoral turnout has not fallen; overall voluntary association membership has fluctuated and increased somewhat between 1959-1990; the number of, level of financial support for, and voluntary work for, charities has probably increased; time spent on informal sociability outside the home was at least as high in 1984 as in 1961. Disentangling life-cycle from cohort effects is difficult, but there was no evidence that the immediate post-war ('baby boom') generation was any less involved than its predecessor, though there was some evidence of change in values among persons born after 1960. This appears partly as decline in social trust – the proportion of people who would say 'in general you can trust other people', though even that was probably a period, rather than a generational, effect.

Hall goes on to explain the maintenance of levels of social capital in Britain in terms of three factors: increased levels of education, changes in the class structure and extensive state support for voluntary associations. Among the conclusions he draws is that the levels of social capital are being 'most strongly sustained by the middle class' (439). Hall explains changes in social trust not in terms of the impact of urbanization or Thatcherism, but in terms of three types of explanation: material insecurity, changing values, especially the opportunism of a more individualistic society brought to light in the attitudes of the under-30s, and a change in the character of associational life. Two changes in associational involvement are worth considering. First, consistent with a globalisation perspective, 'associations that involve people in the kind of face-to-face interaction thought to build social capital may have been replaced by others that involve little such interaction' (449). Putnam considered it important and regrettable that in the USA charitable and political commitments were increasingly being marked by 'members' contributing money which professional officers then used to support good causes and political lobbying. Second, 'associations dedicated to advancing some common or public interest may have diminished in size, while associations dedicated primarily to the private needs of their members could have grown' (449). The limited evidence indicated some diminution of membership of public interest associations, like political parties, churches and trade unions. However, Hall was only able to conclude that 'beneath the apparent stability of organizational involvement in Britain, changes in the character of that involvement may be taking place that would not only explain some of the decline in social trust but might also indicate some erosion in the quality of civic engagement' (450). Hall thinks that while political efficacy and political trust may have tended to decline, that has not reduced political engagement as such. Against this generally optimistic picture, Hall notes that there are sharp class differences between middle and working classes, the latter, along with the young, tending to be 'disconnected' and perhaps increasingly marginalised.

Hall's account of trends is open to further examination. His measurement of trends in associational membership relies predominantly on one off cross sectional surveys which do not necessarily ask similar questions, and whilst he is clear about the social skewing of membership to the middle classes, he relies on dated measures of class to establish this. Like Putnam, he does not examine the determinants of membership using multivariate analysis. In addition, his account does not cover more recent trends into the 1990s. To rectify this point, Warde et al (2003)[1] use BHPS data between 1991 and 1997 to show that there is general stability of membership over this six year period, though with a slight trend towards increase in the early 1990s and decrease in the mid 1990s. In this paper we are able to take this analysis a step further by considering trends since 1972 through a further comparison with the Oxford Mobility Survey (1972). We are cautious in extrapolating trends at the aggregate level from these surveys because of the different way that questions about associational membership were posed. However we are able to gain comparable information on a number of key organisations to allow us to demonstrate that there are important variations between types of associations. We are thereby able to show more fully how an apparent decline in membership at the aggregate level is almost entirely driven by falling male membership of two types of associations and that in general levels of social capital are stable, though increasingly skewed towards the middle classes.

Section 2 describes the OMS and the BHPS data sets that we use for this paper. Both are national sample surveys and the latter is a panel study which has been collecting information on the same individuals and households annually since 1991. Section 3 examines the number of types of association and organisation of which respondents to the OMS and the BHPS were members and describes change over time. We examine the socio-demographic characteristics of people who join associations and thus address inequalities in the distribution of associational social capital. Having distinguished those who are members of at least one core organization (joiners) from the rest, a series of logistic regression models are run for each data set to seek to identify any changes in the socio-demographic features of participants. In Section 4 we explore in a preliminary fashion movement in and out of different types of organization. The panel data permit analysis of careers, separating out, so to speak, the stocks and flows of social capital. Estimates of the volume of social capital circulating in any one year can fruitfully be complemented by an appreciation of the levels of volatility in memberships. We inquire into the volatility of membership in associations and political participation. We then examine what factors encourage respondents, year on year, to join, leave or remain a member of an association.

2. Data and Method

We use two data sets for this research: the Oxford Mobility Survey (OMS) of 1972 and the British Household Panel Survey (BHPS). The OMS has a sample of 10,309 men in England and Wales in 1972 and is the largest survey carried out during the 1970s (see Goldthorpe 1987). The BHPS began in 1991 as the premier British panel study, and samples around 5,000 households and 10,000 individuals each year. In the OMS and the BHPS a range of questions are asked on memberships in different civic organisations, ranging from trade unions, sports/hobby clubs, workmen's or social clubs, professional organisations, church or religious groups, tenants' or residents' groups, parent-teacher associations, political party, voluntary service groups, environmental groups, other community or civic groups, women's institutes, scouts or guides organizations and an unspecified 'other'.[2] Both data sets contain information on, among other things, respondent's class position, their mobility trajectories, educational qualifications, friendship ties and associational involvement.

To compare the two surveys, we had to restrict our analysis in various ways. Firstly, because the OMS only sampled men, we present BHPS data for men and women separately, so that the findings for men can be compared directly with those of the OMS. This allows data for women to be used for the 1990s, and for OMS data on men to be compared with BHPS data on men. Secondly, we made the BHPS comparable with the OMS by restricting the BHPS analysis to those of working age living in England and Wales. Thirdly, we had to handle questions on associational membership in a way that allowed comparisons to be made. There are significant differences in the way that questions on voluntary association membership were asked. Whereas the OMS asks about all associations that individuals belonged to, the BHPS only asks if respondents belonged to a particular type of organisation, which will therefore tend to understate the full number of associations that people belong to, since people may belong to more than one association of a similar type. We can alleviate this problem by focusing not on absolute numbers of memberships but on types of membership for both the OMS and BHPS, but it should be recognised that this will tend to underestimate total memberships. Finally, as the types of associations thus defined vary both between the OMS and the BHPS, and between different waves of the BHPS, we had, for some purposes, to restrict our analysis to those associations which are asked in all relevant surveys to ensure consistency. This will again have the effect of undercounting the scope of memberships.

Our analysis uses descriptive statistics for disclosing the observed patterns and trends in absolute terms, and binomial logit regression models for measuring participation across time and between gender groups. By employing the cross-sectional weights in the data, we present, in Section 3, patterns and trends from the OMS and the two waves in the BHPS as if they were all from cross-sectional surveys.[3] In section 4 we focus on the longitudinal elements of the BHPS. To implement this, we use weighting variables in the respective BHPS waves that represent the cross-sectional population weights. We use Stata 7 which allows for analytical weights in descriptive analysis and probability weights in the modelling procedures.

3: An Analysis of Associational Involvement in (1972- 1999)

3.1: Changing profiles and social bases of associational memberships

Table 1: Participation in voluntary associations in England and Wales (1972-1999)
Membership in organizations (%)
Trade unions39.526.422.315.716.3
Sports/Hobby Clubs25.025.626.111.114.0
Working men's or Social Clubs27.621.117.9 7.6 7.0
Professional associations11.2-13.8- 7.7
Church or religious groups 9.5 8.1 7.012.411.1
Tenants'/Residents' groups 3.5 6.8 7.0 8.2 9.0
Parent-Teacher Associations 5.4 3.5 2.4 7.2 6.5
Political party 7.4 3.5 3.1 2.6 2.0
Voluntary services groups- 3.1 2.5 5.3 4.5
Environmental group- 4.5 3.8 4.2 3.2
Other community/civic groups- 2.9 1.7 3.3 1.9
Women's institutes/groups--- 4.3 3.6
Scouts/Guides organizations-- 1.1- 1.7
Pensioners' group-- 0.7- 0.4
Other15.112.2 7.9 9.1 7.2
Mean number of organizations
All listed organisations1.441.181.170.910.96
Seven common organisations1.180.950.860.650.66
Five common organisations0.510.470.450.410.43
% non-participants22.636.340.646.948.6
Note: 1 For respondents aged 20-64 and resident in England and Wales at the time of interview.
2 The seven organisations are trade unions, sports/hobby clubs, working men's or social clubs, church or religious groups, tenants/residents groups, parent- teacher associations and political party. The five organisations are sports/hobby clubs, church/religious groups, tenants/residents groups, parent- teacher associations and political party.
3 Cross-sectional weights are used for each of the data sets.
Source: The Oxford Mobility Survey (OMS) of 1972, and the British Household Panel Survey (BHPS) of 1992 and 1999.

Table 1 reports membership in different civic organizations, and indicates that there are elements of both decline in membership (as Putnam argues for the US) and stability (as Hall argues for England and Wales). The mean number of types of listed organisations that men belong to fell from 1.44 for men in 1972 to 1.18 in 1992 and 1.17 in 1999. These figures for 1972 and 1992 are actually very similar to those offered by Hall for male memberships of associations in 1973 (1.46) and 1990 (1.13). This suggests sharp decline in the 20 years after 1972 and then stability. However, these data underestimate change in the 1990s, an underestimation resulting from the introduction of new categories of organization into the BHPS questionnaire in the mid-1990s.[4] Another way of approaching this, where later change is more apparent, is to look at the proportion of individual non- participants. This indicates a striking rise in the proportion of men who did not join any association between 1972 and 1992, from 23% to 36%, with a further rise to 41% in 1999.

However, caution needs to be exercised in interpreting these findings and we should not take too seriously a claim that there has been a precipitate decline in social capital at a general level. There is considerable diversity in how many types of organisations are listed both between the OMS and the BHPS, and within the BHPS. We can compensate for this in part by taking just the seven types of organisations which were asked about in both surveys. This indicates a decline, with the mean number of types that men belonged to falling from 1.18 in 1972, to 0.95 in 1992 and 0.85 in 1999. However, female membership of similar associations did not fall during the 1990s, and indeed Hall (1999) indicates that female membership has been rising since the 1950s. Moreover, Warde et al (2003) show that membership of those associations which were inquired about in the BHPS between 1992 and 1997 remained broadly constant,[5] showing a slight rise in membership between 1991 and 1995, and then a slight fall between 1995 and 1997. A further decline between 1997 and 1999, evident in Table 1, suggests overall an aggregate reduction of individual membership during the 1990s.

A further vital consideration is relevant here. Insofar as there was a decline in associational membership for men since 1972, this is almost entirely caused by declining membership of two types of associations. In 1972 trade unions were by some distance the most popular association for men, yet by 1999 they had been overtaken by sports and hobby clubs. Male membership of trade unions fell by 21 percentage points from 1972 to 1999, a fall of 54 percent. There was also a remarkable decline in working men's club membership and (27.6 to 16.7 per cent, a fall of 39.5 percent). However, if we restrict our analysis to the other five organisations, omitting trade unions and working men's clubs, namely, sports/hobby clubs, church or religious groups, tenants/residents groups, parent-teacher associations and political party, there was actually stability in the 1990s in the mean number of these organisations, for both men and women. Indeed men's membership of these five organizations remains more or less stable from 1972 to 1999. There were some categories where membership rose between 1972 and 1999, such as tenants' groups (+3.4%), and professional organizations (+2.6%) and, as Warde et al (2003) indicate, during the 1990s several other types of associations increased in numbers.

Comparing women's membership profiles with those of men, we see that they are less likely to be in associations than men, especially in trade unions, sports clubs, working men's clubs and professional associations. This is related to the concentration of women's employment in the service sectors and their under-representation in professional or managerial occupations (see Li 2002). But women are more likely than men to join religious groups, tenants' groups, parent-teacher associations or voluntary or civic organizations.

In general, then, these data suggest sharply declining participation in some categories of association, especially trades unions and social clubs, particularly among men, between 1972 and the end of the 1990s. Some other categories of association included in our data set showed increases, sometime proportionately substantial, but because starting from a comparatively low base are insufficient themselves to prevent absolute decline in membership. However, decline in membership is a male phenomenon, and does not apply to women. In addition, new types of association emerged and increased their affiliations in the period between 1972 and the late 1990s – environmental organizations, women's groups, new social movements among others – the effect of which is not recorded because of our techniques for assuring direct comparisons between the samples of the OMS and the BHPS. What our data reveal for certain, however, is the decline of participation in associations which had previously attracted working men in very substantial numbers (ie trades unions and social clubs). This is an issue which we explore below, as it suggests that rather than there being an aggregate decline of social capital in the way that Putnam argues, there may actually only be a decline in the memberships of a relatively small section of the population, and that stability, or even a slight increase in membership is found elsewhere.

This point is an important issue to reflect on for a further reason. Putnam sees positive civic and social qualities flowing from membership of associations, but associations vary in the extent to which they may provide this. To be a member does not necessarily entail active engagement, which is almost always necessary if membership is to be translated into acquisition of social capital. If there is no contact between the members, then there is no mechanism for the breeding of mutual trust and tolerance of difference which are some of the presumed sources of the positive effects of associational membership for the growth of social capital. Table 2 shows that levels of activism among members of associations varies from type to type. Unsurprisingly, very few people claim to be active in an association without being members. However, it can readily be seen that the majority of trade union members are inactive. Hence the decline of membership of this category of organization may not be very significant in reducing stocks of social capital - though perhaps unions inspired somewhat greater member activism in the highly contested years of the 1970s than they did in the 1990s. Possibly of greater relevance is the decline of membership of social clubs, for they do have a higher proportion of active members. Since the social club used to be a central neighbourhood institution, usually cooperatively organized and providing recreational services for members within working class communities, its decline may be of particular social and political significance. This offers further support to the argument that the decline of social capital evident in Table 1 may be more apparent than real.

Table 2: Respondents claiming to be a member and active, and to be active but not a member, various types of association, Wave 7 (percentages)
Members who are active (%)Persons active but not member (%)
Political party470.1
Trade union240.1
Environmental group450.3
Parent/school association792.3
Tenants/ residents440.6
Religious group861.9
Voluntary service870.7
Other civic group870.3
Social club801.6
Sports club884.2
Women's Institute940.1
Women's/feminist group830.1
Other group651.9

3.2: Trends in the social composition of membership

One way of interpreting the trends from Table 1 and 2 is that rather than seeing a decline in social capital across the board, we are seeing a decline in social capital for particular social groups, namely working class men. Warde et al (2003) do indeed establish that there was increased service class dominance of associational membership between 1991 and 1997. It is of course well known that the middle class is over-represented in associational memberships (see Parry et al 1992; Goldthorpe 1987; Li et al 2002a, b), but we are able to assess whether this skew is becoming more marked. We explore this by assessing the likelihood of respondents in different class positions belonging to associations in 1972, 1992 and 1999. We adopt a version of the prevalent class schema used in British research (Goldthorpe 1987; Erikson and Goldthorpe 1992), to wit, a six-way schema for the respondent's current, first-job and parental, class that distinguishes:

I&II: The service class: professionals, administrators and managers;
III: Routine-non-manual employees;
IV: Petty bourgeoisie: self-employed workers with or without employees;
V: Foremen and manual supervisors; lower-grade technicians;
VI: Skilled manual working class;
VII: Semi/unskilled manual workers, including agricultural labourers.

In addition, we are also able to assess whether a respondent's mobility trajectory affects the propensity to join associations, and thus to see, for example, how easy it is for the upwardly mobile to join association in equivalent ways to their stable service class peers. To do this, we measure their class at three points in time: their current class, their class at the entry to the labour market after leaving full-time education (first-job class), and their parental class. The respondent's current class is based on his or her current job (or last main job if the information on current occupation is not available). Parental class refers to father's job when the respondent was aged around 14. Where father's class is not available, mother's class is used, which is the practice in the OMS and is here applied to the BHPS.[6] First-job class in the BHPS is obtained from work-histories at Wave 3 and the subsequent updates.

We are further able to extend our analysis by examining whether having friends from a higher social class affects membership. Here we are only able to use a two-way class schema for the respondent's friend's class: service and non-service class. This is because, whilst the full-version schema for each of the three friends' class is available in the OMS, only the SOC (standard occupational classification) codes for the 'best friend' are available in the BHPS (Waves 2 and 8)[7]. A careful prior analysis shows that it is appropriate to use the class position of the first-mentioned 'most-frequent spare-time associate' in the OMS with that of the 'best friend' in the BHPS for the analysis.[8] In order to control for the effects of educational qualifications on membership we use a three-way variable: degree or professional qualifications; A/O Levels; vocational or no qualifications.[9] The effect of these procedures is to allow us to present a more sophisticated analysis of the way that particular social classes may be the 'bearers' of social capital than was attempted by Hall (1999).

Table 3 shows the impact of class, mobility trajectory, friendship homogamy and education on the propensity to join an association. We divide the sample between joiners versus non-joiners, using binomial logit models. Suppose we use ? to denote the probability of having memberships in voluntary organizations. The ratio of ?/(1-?) constitutes the odds of participation rather than non-participation. The logit modelling uses the log of the odds: log[?/(1-?)]. For brevity (see Agresti and Finlay 1997, ch. 15 for an excellent exposition of logit models), the model can be written as:

log[?/(1-?)] = logit(?) = ? + ?X
where ?X is a vector for b1x1 + … + bnxn with b1 … bn being the coefficients for the explanatory variables: class, education, mobility trajectories and friendship homogamy.

Table 3: Logistic regression models for associational involvement (joiners versus non-joiners)
Service -.128 .462 .622* 1.219*** 1.767***
Routine non-manual .048 -.086 -.328 .395** .627***
Petty bourgeoisie -.428*** -.425* -.580*.233 .980***
Foremen and technicians .164 .068 -.317 .084 .462
Skilled workers .397***-.115-.082.278.378

Degree/professional .935*** .691*** .489** .681*** .743***
A/O Levels .168 .311*.104.367**.218

Intergenerational mobility
Up into service .319 .070-.365-.271-.155
Down from service-.044- .105-.027.377* .387*

Work-life mobility
Up into service-.054- .391-.211-.357 - .719*
Down from service.157.527.114.028.343

Friendship homogamy
Having SV friends .184 .317*-.071 .375**.309*
Constant 1.321*** .385** .526** -.573*** -.836***
Pseudo R2.
1 * p<0.05, ** p<0.01, and *** p<0.001.
2 Weighted data are used with robust standard errors in the models.

Table 3 shows the patterns and trends of associational involvement in Britain (1972-1999). Our reference group is the least advantaged: the inter- and intra- generationally stable semi/unskilled manual working class with only vocational or no educational qualifications and with no service-class friends. We compare the coefficients (in terms of log odds) across the different categories of the independent variables whilst the effects of all other variables in the model are simultaneously controlled for. The interpretation of the models is straightforward. For instance, if the service class respondents were as likely as the reference group to be joiners, the odds ratio would be 1 and the log of this would be 0. Thus a positive coefficient would mean more, and a negative coefficient less, propensity between a given group as compared with the reference group in terms of engagement rather than withdrawal, controlling for all other factors in the model. Another thing to note in this regard is that, for ease of presentation and greater visibility of mobility effects, we have created inter/intra-generational mobility trajectory variables from a combination of parental, first-job and current class variables.[10] Mobility is, following Goldthorpe (1987: 43; see also Clifford and Heath 1993, and De Graaf et al 1995 for a similar arrangement of mobility trajectories), defined as moving into or out of the service class. [11]

The data in Table 3 indicate a significant degree of social closure around associational membership, with current class position being the best correlate of associational membership in 1999 for both men and women. In 1972, only men with higher educational qualifications (88% of them were in the service class) and those in skilled manual working-class positions were significantly more likely to participate, and the petty bourgeoisie were significantly less likely to participate when all other factors were simultaneously controlled for. However, over time, the association between education and participation weakens and that between the service class and participation strengthens. The effect for the service class rose from a non-significant -.128 in 1972 to a significant .600 in 1999, and the effects for having degree/professional qualifications fell from .935 in 1972 to .480 in 1999. The skilled manual workers were no longer different from their semi/unskilled peers in the 1990s. None of the mobility or friendship variables prove very important although one could see that intergenerational upward mobility effects were on the decrease (from .319 in 1972 to -.169 in 1999) and work-life downward mobility effects were on the increase (from .157 to .363). This suggests whereas the upwardly mobile were still showing signs of trying to integrate themselves into the associational patterns of their stable service-class peers in the earlier period, this tendency tends to fall over the years. The reason why those with work-life downward mobility trajectories were increasingly more likely to participate is difficult to explain. It may be that those men with a good start in life (having the first job in service-class positions) but with a downward trajectory were more likely to use civic engagement as a channel for possible counter-mobility. Since we do not have data on the timing of their associational engagement, we cannot fully explore the causal mechanisms involved.

The class effects for women were stronger than for men, but the importance of education seems similar for both sexes, although the coefficient for women's higher education seems higher than that for men in 1999. Women with downward intergenerational mobility experience were more likely to be participants than men, suggesting that origin effects had a stronger impact on women in terms of associational involvement. As women were generally less likely than men to be in the service class, having a service-class friend also proved more important for women's civic engagement.

To further explore the association between class and civic engagement, we use more data from the BHPS waves 1-7 (the membership questions were asked in Waves 1-5 and Wave 7; see also, Warde et al 2003). Figure 1 shows the mean number of memberships, by class, for each year. The relative position between non-service classes remains almost constant, but the rate of increase of memberships by the service class is constantly increasing with the results that the gap between this class and all the rest is widening over the years. The data confirm Hall's (1999) argument that the service class has gained greater social capital relative to all the other classes in British society[12].

Figure 1

A comparison of our results with the findings of Parry et al (1992) suggests some new features. Parry et al found disproportionate involvement by the service class, but engagement among the manual working class in proportion to their presence in the population was largely sustained by membership of trade unions and the Labour Party. In the mid-1980s it was the intermediate classes (especially the petty bourgeoisie) who were under-represented among activists. The position was slightly different among the 1991 BHPS sample, with intermediates other than the self-employed more active than skilled or unskilled manual workers. Since associational membership not only increases political participation, and hence political influence (Savage and Li 2002), but also delivers benefits of sociability including instrumental networks, we have further evidence of polarisation in British society where the most advantaged class is consolidating its privileges relative to the rest.

4: Turnover and Fluctuation

In their conclusion, Parry et al (1992:422-4) speculate about the likely future of voluntary group membership and its effect on political participation. They note, as one of their strongest conclusions, the importance of membership for activism: 'affiliation to groups has been shown to be one of the best predictors of a propensity to participation' (419). They argue, in support of Barnes et al (1979), that 'since actual political behaviour is contingent on political events, it is more important to study the readiness of people to be mobilised' (423). Warde et al (2003) have shown a considerable turnover of membership of political parties, showing that social capital fluctuates. We consider this further here by exploring the dynamics of membership of sports clubs, which are important since they are now the most popular form of membership, and one that has increased in popularity since the 1970s (see Table 1 above).

We looked at movement in and out of sport clubs over the five years covered by waves 1 to 5. A five year period allows us to explore churning over a mid-range period: exploring it over 7 or 10 years would probably show even more complexity.

Table 4: Patterns of joining and leaving a sports club, waves 1-5
Year and membershipFrequencyValid PercentCumulative Percent
XXOOX 22.390.7
X: member; O: non-member.

Table 4 shows that around two thirds of the sample never belonged to a sports club during a five year period. An 'X' indicates that a respondent was a member in a particular year, a 'O' that s/he was not. Thus the pattern 'OOOOO' indicates that the respondent was never a member, 'XXOOO' that s/he was a member in the first two waves but had departed never to return by wave 3, while 'XOXOX' shows someone whose membership lapsed before wave 2 and wave 4 but who rejoined in the intervening years. Only 6 per cent belonged to a sports club consistently over the five years (XXXXX). Over one quarter of the sample were members of a sport club at some point but not constantly. Most of this category do not just make one transition: from member to non member or vice versa, but move in and out of membership, sometimes even on an annual basis. 6.3 per cent of the sample were non-members at wave 1, joined at some point and then remained consistent members. A further 4.9 per cent were members at wave 1, then stopped being members, and then remained as non-members. However, 14.9 per cent moved in and out of membership more than once in the five-year period.

We conducted the analysis of turnover for all the types of organization mentioned in the questionnaire and patterns can be observed from Table 5. The proportion of people who were members in at least one of the five years outweighed those present in any given year by between 1.5 and 3 times. Thus, for example, 28.1 per cent of people held trade union membership during the period, but even in the best year showed an aggregate membership of only 17.7 per cent (a ratio of 1.5:1). For environmental groups the ratio was about 2:1, for women's groups about 3:1.

Table 5: Turnover of memberships of organizations, Waves 1-5 (1991- 1995), percentages
Member throughoutMember in at least one yearMaximum members in any year
Political party1.6 6.8 3.5
Trade Union9.328.117.7
Environmental group1.4 8.5 4.3
Parents association1.510.3 4.7
Tenants association2.021.710.3
Religious group7.519.912.7
Voluntary service1.012.1 5.1
Other civic0.1 9.8 3.5
Social club4.426.413.5
Sports club6.032.118.5
WI1.2 3.5 1.9
Women's group0.2 3.4 1.2

Therefore, a very substantially larger proportion of the population were likely to join at some point during the five years than were likely to remain a member throughout. The ratio for parties was approximately 4:1 (ie 6.8:1.6 per cent), that for unions 3:1, for environmental groups 6:1, for women's groups 17:1. A similar range can be seen among the civic and religious groups with tenants and residents associations exhibiting a ratio of 11:1, religious groups 2.5:1. Membership of religious groups is proportionately the most stable of all, but even it exhibits considerable rates of turnover with only 7.5 per cent claiming membership throughout compared with a maximum membership of 12.7 per cent, while 20 per cent had opted in at some point between 1991-1995.

This information could be interpreted in radically opposed ways. It might be seen as evidence of low levels of attachment to associations and an almost flighty attitude towards activism. On the other hand it might signify that people reflect carefully on their attachments at regular intervals in order better to direct their energies towards currently important and pressing activities. When we look at the careers of individuals over time we get a significantly different impression of the volume of social capital in circulation. Another instance concerns the likelihood of being a member of any association. While in any one year at least two out of five persons claim to have no memberships, very few (15.6 per cent) never held any during the 1990s. This suggests, first, that there are comparatively few people who have never experienced the proposed benefits of associational membership. Though many are not active at a given time, they are probably neither hostile to membership nor unavailable for mobilisation through organizations should an appropriate occasion arise. Second, probably the weak ties developed during a bout of membership will not immediately erode, such that individuals will be better connected than might be imagined from considering only the total volume available to them at any one time. The differential levels of volatility can be observed in the trellis graph Figure 2. This figure shows, class by class, the movement in the score for the number of associations of which a person was a member, over the seven waves. It is clear that there are class differences in volumes of social capital. Here it is also clear that there is a great deal of movement from year to year. The thickness of any line indicates the number of individuals retaining, increasing or decreasing the number of their affiliations from year to year. Thus, for instance, a rather greater number of service class respondents moved from zero to one membership between 1991 and 1992 than remained at zero. About the same number moved from one to zero over the same year. It appears, by comparing the thickness of all horizontal lines with all those moving diagonally, that change is at least as common as stasis.

Figure 2


Interpreting patterns of change in associational membership remains difficult and it is difficult to adjudicate between the projections of Putnam and Hall. Using the best available comparable survey data indicates some decline in membership among men of working age since the beginning of the 1970s, a decline which has continued among this group, though more slowly, during the 1990s. Nearly all the decline was due to the falling numbers of male members of trade unions and working men's clubs. Female involvement increased during the 1990s. We therefore do not think that Putnam's diagnosis of a general decline in social capital applies in the British case. The trend among men of working age is, however, of considerable significance because they mark a particular crisis for one distinctive form of social capital linked to the culture of working class men. By contrast, there is a concentration among the middle class in the generation and possession of social capital, especially through civic associations, supporting the claims of Hall (1999), Li et al (2002) and Warde et al (2003) about the distribution of social capital.

At the conclusion to his paper Hall (p 42) notes that 'the British case reminds us to be attentive not only to the aggregate levels of social capital but to its distribution, that some may be organised "in" and others "out" by the same set of developments…. It may be that the forms of civic involvement available to the less privileged are more fragile in the face of secular trends'. We have developed this argument here, and argue that the growing predominance of the service class in associational membership is likely to exacerbate it. An unequal class distribution of membership should be expected to carry over into differential levels of political participation. There is clearly an increasing divide between the service class and all other classes. This division is stronger among women than men, and its growth is being accelerated as participation of women belonging to the service class increases along with that of those others women whose closest friend is in the service class.

The evidence from the BHPS panel data suggest a somewhat more optimistic view of the issue of social capital. The large numbers of people who move in and out of organizations from year to year makes Britain appear a more participatory and active society than might otherwise be imagined. 84 per cent of adults had membership of at least one association in the first half of the 1990s. Associational and political experience touches many more people over a five year period than cross-sectional data would indicate. This suggests both that a great many people, and the community as a whole, obtain the benefits of membership at frequent intervals, and that a large proportion of the population might be considered available for mobilisation even if they are not currently active. Of course, such volatility might be seen as a sign of limited commitment; it certainly implies transient commitment, and as we have shown, membership is, in addition, not simply translated into regular and active involvement. The consequence is a substantial and steady rate of movement in and out of membership which affects most associations, including political parties and social movements. However, on balance, we see no reason why this churning process should be seen as very problematic and submit that the evidence of such 'careers' is a positive aspect of social capital formation.

This consideration returns us to questions about the nature of social capital and of how exactly associational membership is connected to social trust and political participation. Most studies show that there is a positive relationship between associational membership and political participation. There is however comparatively little evidence about how this process works. How exactly does membership of a non-political association translate into greater political involvement? Specifically, does it matter for the generation of trust what type of organization people join? Does membership of a sports club have the same series of effects as membership of a parent-teacher association, or a religious group? The standard technique, which we have also used, for estimating societal levels of social capital presumes that one organization is as efficacious as another. While recognising that some organisations are positively harmful to general social trust, Putnam asserts that they are mostly positive. However, it seems likely that differences are not simply along an axis of good to bad for the promotion of social capital, but that there are several different dimensions of variance. In the analysis of the BHPS we found that there was a tendency for individuals' memberships to cluster, in the 1990s, in such a way that labour organizations attracted one group of people, social and sports clubs another group, and all the other associations - political, community, religious, etc – a further group. The existence of this last group implies that there are networks of people who are particularly likely to participate in what might best be called 'civic' associations. The concentration of overlapping memberships implies that the social networks arising from participation in civic associations will be socially exclusive and the ties arising will offer advantages to particular sectors of the population.

Our, somewhat speculative, interpretation of these findings is that in the 1970s it was possible to detect two rather different patterns of associational membership among men. Working class men were well represented in trades unions and workingmen's social clubs, while service class men (and service class women too probably, though we have no data to prove the point) were disproportionately active in civic associations. Overall, associational membership was divided along class lines but in such a way that working class men had their own, distinctive sources of social capital, separate and independent from that provided through civic associations to the middle classes. Thus, in the 1970s, the associational field was fractured, with labour and civic associations having very different constitutencies. This is less the case in the 1990s, the field of associational membership having become more homogeneous and even more dominated by the service class. This produces a qualitatively different type of division between the service class, which supplies the core and active members of most associations, and other classes, who are less frequently represented than before.

Our findings are consistent with the view that there has been a shift from a world of class allegiances to one of disaffection by a significant part of the population from the political and civic process. What has happened is, essentially, a restructuring of the political field as the independent working class associations shrink, those which sustained class solidarity. In their place has emerged a more homogeneous set of associations (characterised by organizational isomorphism and hence social exclusion) which themselves tend effectively to exclude the working class (both men and women), who previously were better represented than the intermediate classes, thus contributing to the appearance of disaffection among all subordinate classes with associations. To the extent that associational membership is, indeed, directly a cause of political participation, this would help explain (to the extent that it is the case) the withdrawal from voting and other forms of political participation of the working class. It is almost as if it was the existence of institutionalised class struggle that provided the basis for extensive political participation by the working class in the past and that as institutions for the conduct of class conflict have been dismantled or circumvented, levels of participation have fallen. This presents dominant groups with a different type of legitimation problem – that arising from the fact that civic and political institutions cannot be said to be directly representing the people, for the people are apparently becoming apathetic and less ready to express support for, or trust in, political processes.

We are unable to specify in detail at present who are the most likely to move in and out of associations. However, it would seem plausible in the light of Figure 2 to suggest that the service class besides holding the most memberships are also the most mobile. By contrast the manual working class groups seem to be more constant in their memberships, as indicated by the much more solid horizontal lines on the graph. To the extent that transient membership will enhance social contact, and thus most probably weak ties, and because, as Granovetter (1973) and Burt (1992) among others have shown, weak ties provide instrumental benefits in social competition, the current situation provides ever more advantages to people in the service class. It might also be suggested that the predominance of the service class is associated less with their positions and degrees of integration within local communities and more with their possession of generalised skills and aptitudes relevant to organizational activity and management which allows them effectively to colonise most associations. This leads us to doubt whether associational membership is likely, in this conjuncture, to lead to the type of generalised trust that Putnam postulates. Rather, as Bourdieu might have anticipated, it is more likely to lead to greater social division as members become increasingly like to possess high levels of economic and cultural capital which only serve to exclude less privileged people. This might explain the findings of both Paxton (1999) and Hall (1999) that stability in levels of associational membership can go hand in hand with the decline of social and political trust.

Disclaimer for use of BHPS

The data used in this publication were made available through the ESRC Data Archive. The data were originally collected by the ESRC Research Centre on Micro-social Change at the University of Essex (now incorporated within the Institute for Social and Economic Research). Neither the original collectors of the data nor the Archive bear any responsibility for the analyses or interpretations presented here.


1 Though written earlier, Warde et al (2003) is still in press at the time that this current paper is to be published.

2 For full details of the questions used in the study, see the schedules held by the ESRC Data Archive at the University of Essex; The wording of the relevant questions here is as in OMS, Q38; BHPS, Wave 2, V13, and Wave 9, V19.

3 We have carefully consulted the BHPS User Manuals. Particular guidance is obtained from Section V, Volume A, for weighting purposes. We would like to thank Dr Brendan Halpin for his kind help in this and in deriving the first-job class.

4 The introduction of a specific question on professional association membership almost certainly made a major difference. Some of the memberships described as other in 1992 were surely of professional associations, but their number was almost certainly under-recorded in 1992. The decline in mean male membership probably decreased by approximately 0.06 points.

5 It should be noted that Warde et al's (2003) findings relate to all adults over 16 years of age.

6 We are aware of the debates on individual, conventional and dominance approaches to the family class allocation (Heath and Britten 1984; Goldthorpe and Payne 1986; Goldthorpe 1987; Erikson 1984; Sorensen 1994) but the positions we adopt are necessitated by the need for strict compatibility between the OMS and the BHPS.

7 It is to be noted that, in the OMS, such a person is termed 'most frequent spare-time associate' rather than 'friend'. To be compatible with the terminology in the BHPS and with the general practice in previous research using the OMS (Heath 1981; Mitchell and Critchley 1985), he/she is called 'friend' in the present paper. In addition, for Wave 2, we use associational involvement and all the socio-cultural variables including friendship homogamy in the wave. For Wave 9, we use associational involvement and socio-cultural variables in the wave and, since the friendship data are in Wave 8, we have matched the data on best friend's class onto Wave 9.

8 Respondents in different class positions varied in the scope of friendship ties but not in the class distribution of their friends. In the OMS, respondents were asked to name up to three 'most frequent spare-time associates'. While 29% of the service-class men reported having 3 friends, the figures for the intermediate and the working classes were 23% and 19% respectively. However, with regard to different friends' class positions, there were scarcely noticeable differences. The proportions of friends in service-class positions by respondents in service, intermediate and working classes were 57%, 24% and 11% for the first-mentioned friend; 58%, 24% and 11% for the second-mentioned friend; and 58%, 26% and 15% for the third-mentioned friend. Given this, we feel well justified in using the class position of the first-mentioned friend in the OMS in conjunction with the class of the 'best' friend in the BHPS.

9 We are grateful to Professor Anthony Heath for suggestions over the algorithms used in constructing the educational variable in the OMS.

10 We are interested in estimating the main and the 'mobility' effects of class. For this purpose, we have designed the class variables as follows (taking intergenerational mobility as an example):

Respondent's class
Parent class

A: Intergenerationally stable service class
B: Routine non-manual
C: Petty bourgeoisie
D: Foremen and technicians
E: Skilled manual workers
F: Semi/unskilled manual workers
G: Intergenerationally long-range downwardly mobile
H: Intergenerationally long-range upwardly mobile

The same is done for the first- and current- job categorization. The advantage of this is that it clearly highlights the mobility effects. Strictly speaking, we ought to refer to classes as the inter-/intra-generationally stable service class, the inter-/intra-generationally upward or downward mobile, etc. However, for the ease of expression, we simply refer to them as the service class, the routine non- manual etc.

11 We could use 'parent class' and 'first- job class' as main-effects variables and add interaction terms between current and parent class, and between current and first-job class. The results would be similar but the presentation and interpretation much less straightforward.

12 In this instance a class position is allocated to all respondents. However, when the exercise was repeated to include only those currently in employment, the pattern was more or less the same.


This research has been supported by two ESRC research grants: (1) Savage, Longhurst, Warde and Tomlinson, 'Social Capital and Social Networks: the careers of political activists' (Ref L215045) (funded under the auspices of the programme 'Democracy and Participation) and (2) Li, Savage and Pickles, 'Social capital: developing a measure and assessing its value in social research' (Ref R000223671).


AGRESTI, A. and Barbara F. (1997) Statistical Methods for the Social Sciences, Third Edition, New Jersey, Prentice Hall.

BARNES S, KAASE M et al (1979) Political Action: mass participation in five western democracies, Beverley Hills, Sage.

BOURDIEU, P. (1984) Distinction: a social critique of the judgment of taste, London, Routledge & Kegan Paul.

BOURDIEU, P. and WACQUANT, L. (1992) An Invitation to Reflexive Sociology, Cambridge, Polity.

CLIFFORD, P. and HEATH, A. F. (1993) The Political Consequences of Social Mobility, Journal of the Royal Statistical Society, Series A (Statistics in Society), 156(1): 51-61.

COLEMAN, J. (1988) Social capital in the creation of human capital, American Journal of Sociology, 94: S95-S120.

COLEMAN, J. (1990) Foundations of Social Theory, Chicago, Chicago UP.

DE GRAAF, N. D., NIEUWBEERTA, P. and HEATH, A. (1995) Class Mobility and Political Preferences: Individual and Contextual Effects, American Journal of Sociology, 100(4): 997-1027.

DiMAGGIO, P. & POWELL W (eds) (1991) The new institutionalism in organizational analysis, Chicago, University of Chicago Press.

DOWSE, R. and HUGHES, J. (1977) 'Sporadic interventionists', Political Studies, 25: 84- 92.

ERIKSON, R. (1984) Social Class of Men, Women and Families, Sociology 18:500-14.

ERIKSON, R. and GOLDTHORPE, J.H. (1992) The Constant Flux, Oxford, Clarendon Press.

FUKUYAMA, F. (1998), Trust, the social virtues and the creation of prosperity, London, Hamish Hamilton.

GOLDTHORPE, J.H. and PAYNE, C. (1986) On the Class Mobility of Women: Results from Different Approaches to the Analysis of Recent British Data, Sociology 20: 531-555.

GOLDTHORPE, J. H. (1987) Social Mobility and Class Structure in Modern Britain, 2nd edition, Oxford, Clarendon Press.

GRANOVETTER, M. (1973) 'The strength of weak ties', American Journal of Sociology, 78: 1360-80.

HALL, P. (1999) Social capital in Britain', British Journal of Political Science, 29: 417-61.

HALPIN, B. (1998) Unified BHPS Work-life Histories: Combining Multiple Sources into a User- Friendly Format, Bulletin de Méthodologie Sociologique No. 60.

HEATH, A.F. (1981) Social Mobility, London, Fontana.

HEATH, A.F. and BRITTEN, N. (1984) Women's jobs do make a difference: A Reply to Goldthorpe, Sociology 18: 475-90.

LI, Y. (2002) Falling off the ladder? Professional and managerial career trajectories and unemployment experiences, European Sociological Review, 18(3): 1- 18.

LI, Y., SAVAGE, M. and PICKLES, A. (2002a) 'Social Mobility, Social Capital and Civic Participation in England and Wales (1972-1998)', presented to BSA, Leicester, 26 March, 2002.

LI, Y., SAVAGE, M. and PICKLES, A. (2002b) Social inequalities and associational involvement in England and Wales (1972-1999), mimeo, University of Manchester.

MITCHELL, C.J. and CRITCHLEY, F. (1985) Configurational Similarity in Three Class Contexts in British Society, Sociology 19: 72-92.

PARRY, G., MOYSER, G,. and DAY, N. (1992) Political Participation and Democracy in Britain, Cambridge, Cambridge University Press.

PAXTON, P. (1999) 'Is social capital declining in the United States: a multiple indicator assessment', American Journal of Sociology, 105: 88-127.

PORTES, A. (1998) Social capital: its origins and applications in modern sociology, Annual Review of Sociology, 24:1-24.

PUTNAM, R. D. (1993) Making Democracy Work: civic traditions in modern Italy Princeton NJ, Princeton University Press.

PUTNAM, R. D. (1995) 'Tuning in and tuning out: the strange disappearance of social capital in America', Political Science & Politics, December 664- 683.

PUTNAM, R .D. (2000) Bowling Alone, New York, Simon and Schuster.

SORENSEN, A. (1994) Women, Family and Class, Annual Review of Sociology 20: 27-47.

SAVAGE, M. and LI, Y. (2002) 'Social Class, Associational Involvement and Political Preferences in England (1972-1999)', mimeo, University of Manchester.

URRY, J. (2002) 'Mobility and Proximity', Sociology, 36(2): 255- 274.

WARDE, A., TAMPUBOLON, G., LONGHURST, B., RAY K., SAVAGE, M., and TOMLINSON, M., (2003) 'Trends in social capital: membership of associations in Great Britain 1991-1998', Research Note in British Journal of Political Science.

WASSERMAN, S. & Faust, K. (1994) Social Network Analysis: methods and applications Cambridge, Cambridge University Press.

WELLMAN, B. & BERKOWITZ, S. D. (eds) (1988) Social Structures: a network approach Greenwich Conneticut, Jai Press Inc.

WILKINSON, R. (1996) Unhealthy Societies: the Afflictions of Inequality, London: Routledge.

WORLD Bank (1997) Expanding the Measure of Wealth: indicators of environmentally sustainable development, Washington, DC, chapter 6 'Social capital: the missing link?'

Copyright Sociological Research Online, 2002