Cohort Variations in the Membership of Voluntary Associations in Great Britain,1991

Abstract

This article uses longitudinal data from the British Household Panel Study to examine trends in the level of membership in a range of voluntary associations from 1991 to 2007. It has been suggested that the membership of voluntary associations is in decline in western societies and the analysis examines the trajectory of the number of memberships with age for four 10-year age cohorts born between 1935 and 1975. The results for men show lower levels of membership over time for the 1955–1964 and 1965–1974 cohorts in comparison to those of earlier born cohorts. For women, levels of membership were only notably lower for the 1965–1974 cohort. The differences in the probability of belonging to an organisation between cohorts were similar in magnitude to those between categories of social class and education. Cohort differences in the membership of voluntary associations are interpreted to reflect primarily the impact of changing social and economic conditions on individual’s capacity for involvement.

Keywords

cohort social capital social change volunteering

Introduction

This article examines the proposition that there has recently been a decline in the level of membership in voluntary associations in Great Britain. The debate concerning the decline in the membership of voluntary associations in western societies was largely initiated by Putnam (1995) who used data collected as part of the General Social Survey to argue that in America the average number of group memberships had fallen by more than 25 per cent over the period since 1970. Subsequent studies have examined the extent to which the decline in the membership of voluntary associations is a general process in western societies. The results of these studies have been mixed but, in a recent review of the literature, Stolle and Howard (2008) conclude that there is no conclusive evidence that participation levels are in general declining in European societies.

In the UK, the variation over time in the level of membership in voluntary associations has been examined by several studies (Hall, 1999; Li et al., 2003; Warde et al., 2003). Hall (1999) used data from a range of sources to conclude that there had been little change in the level of overall membership in voluntary associations over the period from 1950 to 1990. Warde et al. (2003) updated the analysis using data on the membership of organisations from the British Household Panel Survey (BHPS) from 1991 to 1998 and also concluded that there had been little change over time in the overall level of membership. Both studies suggested, however, that the distribution of memberships in voluntary associations was becoming more unequal and increasingly characteristic of individuals from middle-class backgrounds. The relationship between social and economic change and the level of membership in voluntary associations was more directly examined by Li et al. (2003) who compared the membership of voluntary associations reported in the Social Mobility Inquiry undertaken in 1972 with that in the BHPS collected in 1992 and 1999. Their results showed little change in the overall level of membership for individuals from middle-class backgrounds but a substantial decline in membership among the working class.

This article extends previous work using the BHPS to examine changes in the level of membership of voluntary associations between 1991 and 2007 for different birth cohorts. In a cross-sectional survey, differences in the level of membership in voluntary associations between individuals of different age cannot be distinguished from cohort differences. As successive waves of data are accumulated over time in a longitudinal survey, however, the ages at which individuals who were born in different cohorts are observed start to overlap. This allows an examination of whether there have been changes in the level of membership in voluntary associations between cohorts while also allowing control for individual differences in behaviour over time (Miyazaki and Raudenbush, 2000).

The distinction between age and cohort effects is important for conceptual reasons. For an individual, the probability of belonging to a voluntary association varies with age over the lifecourse due to changes in the contexts which influence participation, such as the workplace and children’s schools. Changes in the probability of membership between cohorts also reflects, however, the influence of shared social and economic circumstances which cohort members share and that differentiate one cohort from another (Ryder, 1965). Cohort differences in the propensity to join voluntary associations therefore provide a way of capturing the influence of changes in social and economic structure on individual’s behaviour. Analyses which do not take account of cohort differences assume that there is a common trajectory with age along which individual’s membership will develop across cohorts. In the light of the many social and economic changes that have taken place in the UK over the last 40 years, this assumption is worth further study.

The level of participation in voluntary associations is of interest not only to social scientists but increasingly to policy makers. One of the aims of the Big Society is that everyone should be a member of a community group. The government is also intending to introduce National Citizen Service for all young people between 16 and 25 years of age with the aim of developing the skills and motivation necessary to get involved in ongoing volunteering. Understanding how patterns of membership in voluntary associations have changed between successive cohorts is certainly crucial for understanding the extent to which these aims can be achieved.

The remainder of this article is organised as follows. The next section describes the rationale for the study. The data and methods used in the analysis are then discussed and the results of the analysis presented, before the article concludes with a discussion of the findings.

Study Rationale

The main objective of this study is to examine trends over time in the level of membership of voluntary associations for different birth cohorts. There are several reasons why the levels of membership in voluntary associations may be lower in more recent cohorts. Most importantly, the changes in patterns of income and employment that have taken place in the UK since the 1970s (Gallie, 1999) may have had a negative impact on the number of people joining voluntary associations in more recent cohorts. Employment provides the financial resources that are necessary to be able to take part in activities with other people and work is also a source of social contacts and skills that can be transferred to public activities (Verba et al., 1995). The changes in the UK labour market since the 1970s have resulted, however, in a growth in insecure and casual employment for many young people, including those with graduate level qualifications (Furlong and Cartmel, 2007). Such patterns provide few opportunities for the accumulation of human capital or the development of the social networks that are important to both career progression and involvement in voluntary groups.

Changes in the way that labour is organised in the workplace may also have influenced the level of organisational membership in different cohorts. In particular, there has been a marked decline in the level of union membership since 1980, which is largely attributable to an increase in the number of people in the labour force who have never belonged to a union (Bryson and Gomez, 2005). The membership of trade unions may influence levels of membership in voluntary associations both by shaping the organisational resources of union members and by providing personal experiences of the benefits of group membership. The decline in union membership may therefore be expected to have reduced the extent to which employment supports wider patterns of organisational membership.

Finally, as women have increasingly entered the labour market they have been faced with the challenge of combining motherhood and the demands of a job. Time-budget studies show that although women are increasingly working for pay there has been little adjustment to the division of domestic labour (Kan, 2008). In particular, among couples where both partners are working full time, the amount of time spent by men on domestic work has risen only marginally since the 1970s and women still do the bulk of the housework. The tradeoffs involved in meeting the demands of different roles have become more acute for many women and as a consequence it is likely that they have also experienced a reduction in the amount of time available for taking part in activities outside the household.

In addition to examining trends over time in the membership of voluntary associations, the analysis examines how controlling for individual and household characteristics influences the magnitude and statistical significance of cohort differences in the level of membership. Studies which have documented the characteristics that are associated with membership of voluntary associations have almost always found a significant association between membership of voluntary associations and a range of individual and household characteristics (Li et al., 2003, 2008). The economic changes that have taken place since the 1970s have been accompanied by changes in the distribution of social classes with a shift towards non-manual and away from manual jobs (Gallie, 1999). There has also been a significant rise over the period in the number of people achieving higher educational qualifications (Halsey, 1999). Changes over time in resources such as time, money and skills may not have taken place uniformly across cohorts and the present analysis examines whether controlling for individual and household characteristics influences the magnitude and statistical significance of cohort differences in the membership of voluntary associations.

Finally, the analysis also examines whether the relationship between individual and household characteristics and the membership of voluntary associations varies across cohorts. Resource-based explanations for why some people join voluntary associations, such as the civic voluntarism model of Verba et al. (1995), have often been considered to be inadequate on the grounds that they cannot explain why levels of membership should be in decline while the overall level of resources in society is rising (Pattie et al., 2004). This argument assumes, however, that the relationship between individual and household characteristics and membership is constant and does not vary across cohorts. To examine whether this assumption is reasonable, interactions between individual and household characteristics and cohort were included in the analysis. The interactions allow the effects of explanatory variables to vary across cohort and permit the statistical significance of cohort differences in the effects of the explanatory variables to be examined.

Data and Statistical Methods

Data

The data source used in this study is the British Household Panel Study (BHPS), an annual panel survey of a representative cross-section of British households (Taylor, 2009). The survey tracks all adults and their children who were in wave 1 households over time even if the household breaks up or they move home. The initial wave of interviews took place in 1991 and obtained full interviews from 9092 individuals in 5050 households, a response rate of 74 per cent of eligible households. The wave-on-wave response rate is comparable to that of similar surveys and has varied between 88 and 97 per cent once ineligible subjects are removed (Lynn, 2006).

The BHPS questionnaire includes questions asking respondents whether they belong to several different types of voluntary association. The questions concerning membership of voluntary associations have not been included at every interview and the analysis is based on the responses at every second interview between wave 1 and 17 inclusive. The types of organisations have been added to since the initial interview, but we only considered those types of organisation in which membership was asked at each interview. Voluntary associations may have professional staff but membership is not legally required and they are governed using voluntaristic mechanisms instead of through legislation or market incentives. The different types of organisation included in the analysis were: political party, environmental group, parents association, tenants group, religious group, voluntary service group, other community group, social group, sports club, women’s institute and women’s group. The overall number of different types of membership is used to provide a simple indicator of the overall level of participation.

From the information collected by the BHPS, social class, education and household type were chosen as explanatory variables for inclusion in the analysis. The social class measure used is the National Statistics Socio-economic Classification (NSSEC) (Rose and Pevalin, 2003). The NSSEC differentiates positions in the labour market according to their typical employment relations. In this study, the three-class version is used which distinguishes managerial and professional occupations, intermediate occupations and routine and manual occupations. Because of the importance of the family in transmitting capital in all its forms from one generation to the next, the measure of social class used was the higher of the social class positions of the respondent’s parents when the respondent was age 14. If neither of the respondent’s parents were in employment, the respondent was assigned to the manual category. Education was measured by assigning the respondent’s highest qualification to one of four categories: Higher, A-level, O-level or Other and None. The respondent’s household type was included in the analysis because of the association between participation in voluntary organisations and having children in the household, and the household type categories distinguished in the analysis were: couple with no children, couple with children and single (including lone-parent) or other.

In our analyses we include individuals in our dataset who do not have complete patterns of response over time. Individuals were, however, not included in the dataset beyond the first interview they had missed. To minimise the influence of dropout due to mortality on the results, the analysis was restricted to individuals who were less than 56 years of age at the initial interview. The analysis examines variation in the level of membership in voluntary associations for four cohorts defined using year of birth: 1935–1944, 1945–1954, 1955–1964 and 1965–1974. The sample consists of 3243 men and 3592 women who contributed 18,252 and 22,154 interviews respectively. For the purposes of analysis the response was recoded into three ordered categories corresponding to none, one and two or more memberships. The sample proportions in each category were 47.2, 34.8 and 17.9 per cent for men and 55.3, 29.2 and 15.3 per cent for women.¹

Statistical Methods

An ordered logistic model was used to study the variation in the number of types of membership for each individual as a function of time (Vermunt and Hagenaars, 2004). In the ordered logistic model, the logit of the cumulative probability of the response is modelled as a linear function of explanatory variables and the model used can be written as:

\log i t (P (Y_{i t} \leq s)) = κ_{s} - (π_{0 i} + π_{1 i} (a_{i t} - \bar{a} ..) + π_{2 i} {(a_{i t} - \bar{a} ..)}^{2})

(1)

where i and t are indicators for the individual and time, s = 1,2 is an indicator for response category, π_0i π_1i, and π_2i are model parameters, a_it is the individual’s age in years, ā.. is the median age in the sample and κ_s are unknown category thresholds. The dependence of the response on additional explanatory variables is captured by allowing the parameters π_0i and π_1i to depend on cohort membership and the set of explanatory variables X_it:

\begin{array}{l} π_{0 i} = \sum_{j = 2}^{4} β_{0 j} C o h o r t_{j i} + λ X_{i t} + u_{0 i} \\ π_{1 i} = β_{10} + \sum_{j = 2}^{4} β_{1 j} C o h o r t_{j i} + u_{1 i} \end{array}

(2)

In the model for π_0i, β₀₂ – β₀₄ are coefficients for the effect of cohort and λ is a vector of parameters associated with X_it. The model does not include a constant which is absorbed into the model thresholds, κ_s. In the model for π_1i, β₁₀ is the effect of age for the 1935–1944 cohort while the coefficients β₁₂ – β₁₄ are the effect of age for subsequent cohorts measured relative to that for the 1935–1944 cohort. A similar model was examined for the quadratic age term but was omitted from the final analysis. The coefficients for the explanatory variables can be interpreted as the effect of the associated variable on the log of the odds of the probability of being in a higher rather than in a lower response category with the exponentiated coefficients therefore giving the cumulative odds of the response categories.² Note that the model assumes that the effect of a coefficient is the same for all categories, s. If a coefficient increases the odds of a response in the first category by a factor of two, it increases the odds of a response in the second category or below by the same factor. The terms u_0i and u_1i are normally distributed random errors which have a joint bivariate normal distribution with standard deviations σ₀ and σ₁, respectively, and correlation ρ. The implications of the results can best be understood by plotting the population average predicted probability of the response against age for different values of the covariates. The population average predicted probability is calculated by integrating the response probability over the distribution of the random errors (Skrondal and Rabe-Hesketh, 2009). Models were estimated in Stata 11 using the Gllamm program and 20 point Gaussian quadrature.³

The analysis was undertaken in three steps. The first step (Model 1) included only age and cohort as explanatory variables and aimed to describe the overall time trends in membership for different cohorts. The second step then examined whether controlling for differences in individual and household characteristics influenced the coefficients from the previous model. Model 2 added social class and Model 3 additionally included education and household type as explanatory variables. Finally, interactions between each explanatory variable and cohort were added to the model to examine whether the effects of the explanatory variables varied across cohorts. Separate analyses were conducted for men and women. In order to adjust for the effects of dropout from the survey, two approaches were examined. The first approach included indicators of dropout due to non-response and whether the respondent was known to have died as explanatory variables. The second approach used the longitudinal weights available in the dataset to adjust for the effect of dropout. There were no significant differences in the results obtained using the different methods and, for reasons of transparency, we present the results from the analyses which included indicators of dropout as explanatory variables.

Results

For each cohort Table 1 shows the mean number of memberships for men at each age and, in the lower panel, at each wave. The number of memberships is lowest in the 1965–1974 cohort and highest in the 1935–1944 and 1945–1954 cohorts. The higher number of memberships in the 1935–1944 and 1945–1954 cohorts is expected given the range of ages over which the cohorts have been observed. A comparison of the number of memberships between different cohorts observed at the same age suggests, however, that in most cases membership levels are lower in the more recent cohort. Table 2 shows the corresponding results for women. The overall pattern of variation in the number of memberships appears similar to that for men. The number of memberships increases from the youngest to the oldest cohort while a comparison of the number of memberships between cohorts at the same age suggests that membership levels may be lower in the more recent cohorts.

Table 1.

Men: mean number of types of organisation and (number of observations), by cohort and age at time of interview.

	Cohort				All
	1935–1944	1945–1954	1955–1964	1965–1974
Age
15–19				0.54 (465)	0.54 (465)
20–24				0.75 (1111)	0.75 (1111)
25–29			0.99 (361)	0.80 (1260)	0.84 (1621)
30–34			1.03 (1259)	0.79 (977)	0.92 (2236)
35–39		1.16 (326)	1.10 (1537)	0.80 (515)	1.04 (2378)
40–44		1.24 (1101)	1.03 (1238)	0.76 (127)	1.11 (2466)
45–49	1.24 (312)	1.22 (1459)	1.02 (716)		1.17 (2487)
50–54	1.20 (883)	1.07 (1208)	1.09 (134)		1.12 (2225)
55–59	1.18 (1071)	0.97 (665)			1.10 (1736)
60–64	0.99 (831)	0.96 (158)			0.99 (989)
65–69	0.93 (455)				0.93 (455)
70–74	1.01 (83)				1.01 (83)
Wave
1	0.81 (626)	0.84 (821)	0.71 (918)	0.54 (878)	0.72 (3243)
3	0.94 (493)	1.01 (660)	0.82 (723)	0.59 (656)	0.83 (2532)
5	0.98 (437)	0.96 (589)	0.83 (632)	0.62 (553)	0.84 (2211)
7	0.88 (415)	0.92 (561)	0.78 (591)	0.58 (492)	0.79 (2059)
9	0.88 (389)	0.86 (525)	0.73 (550)	0.61 (453)	0.76 (1917)
11	0.82 (355)	0.77 (485)	0.71 (515)	0.58 (413)	0.72 (1768)
13	0.85 (327)	0.75 (457)	0.73 (478)	0.59 (368)	0.73 (1630)
15	0.85 (305)	0.75 (424)	0.73 (437)	0.57 (333)	0.72 (1499)
17	0.82 (288)	0.73 (395)	0.67 (401)	0.52 (309)	0.69 (1393)

Table 2.

Women: mean number of types of organisation and (number of observations), by cohort and age at time of interview.

	Cohort				All
	1935–1944	1945–1954	1955–1964	1965–1974
Age
15–19				0.41 (386)	0.41 (386)
20–24				0.59 (1132)	0.59 (1132)
25–29			0.65 (452)	0.64 (1565)	0.65 (2017)
30–34			0.77 (1513)	0.73 (1281)	0.76 (2794)
35–39		0.95 (383)	0.89 (1890)	0.81 (779)	0.87 (3052)
40–44		1.02 (1383)	0.89 (1579)	0.77 (193)	0.94 (3155)
45–49	0.89 (350)	1.00 (1786)	0.91 (862)		0.96 (2998)
50–54	1.00 (1029)	0.88 (1483)	0.75 (158)		0.92 (2670)
55–59	0.90 (1223)	0.82 (860)			0.87 (2083)
60–64	0.91 (1037)	0.84 (188)			0.90 (1225)
65–69	0.86 (535)				0.86 (535)
70–74	0.93 (107)				0.93 (107)
Wave
1	0.67 (701)	0.70 (971)	0.56 (1011)	0.37 (909)	0.57 (3592)
3	0.86 (559)	0.86 (817)	0.69 (862)	0.48 (716)	0.72 (2954)
5	0.89 (504)	0.89 (736)	0.75 (782)	0.53 (644)	0.76 (2666)
7	0.80 (482)	0.75 (708)	0.68 (739)	0.43 (611)	0.66 (2540)
9	0.81 (460)	0.74 (659)	0.67 (700)	0.50 (569)	0.67 (2388)
11	0.78 (433)	0.66 (605)	0.66 (655)	0.50 (524)	0.64 (2217)
13	0.78 (405)	0.62 (560)	0.64 (611)	0.52 (493)	0.63 (2069)
15	0.86 (382)	0.68 (530)	0.65 (566)	0.54 (449)	0.67 (1927)
17	0.78 (355)	0.61 (497)	0.59 (528)	0.52 (421)	0.62 (1801)

Table 3 shows the model results for men. Model 1 aimed to describe how the overall level of membership varied over time for the different cohorts. Likelihood ratio tests rejected the hypothesis of a linear cohort effect (χ² = 19.36, p-value < 0.001) and of a common age effect across cohorts (χ² = 23.21, p-value < 0.001). The model therefore includes separate dummy variables for cohort and an interaction term between age and cohort. The exponentiated coefficients for the category thresholds give the predicted cumulative odds of the response categories for a respondent in the base group for all covariates (i.e. at the median sample age of 42 years). The predicted cumulative odds of the response categories for the 1935–1944 cohort were therefore 0.443 (or exp(−0.813)) and 8.944 (or exp(2.191)) corresponding to cumulative probabilities of 0.307 and 0.899.⁴ The predicted proportions of respondents in the different categories of the response were therefore 0.307, 0.592 and 0.101. In comparison to the 1935–1944 cohort, the cohort coefficients show that the odds of being in a higher response category are significantly lower for members of the 1955–1964 and 1965–1974 cohorts. For example, for the 1965–1974 cohort the predicted proportions of respondents in the different categories at the median age in the sample were 0.676, 0.301 and 0.023.⁵

Table 3.

Men: coefficient estimates and standard errors of growth models of number of organisational memberships.

Fixed effects	Model 1	Model 2	Model 3
Cohort
1935–1944	_	_	_
1945–1954 (β₀₂)	0.167 (0.199)	0.129 (0.198)	–0.006 (0.198)
1955–1964 (β₀₃)	−0.493 (0.208)	−0.659 (0.207)	−0.873 (0.207)
1965–1974 (β₀₄)	−1.548 (0.207)	−1.700 (0.207)	−2.038 (0.209)
Age (β₁₀)	0.038 (0.018)	0.038 (0.018)	0.031 (0.018)
Age x cohort
Age x cohort 1945–1954 (β₁₂)	−0.073 (0.015)	−0.072 (0.015)	−0.070 (0.015)
Age x cohort 1955–1964 (β₁₃)	−0.099 (0.022)	−0.102 (0.022)	−0.102 (0.022)
Age x cohort 1965–1974 (β₁₄)	−0.123 (0.030)	−0.124 (0.030)	−0.126 (0.030)
Age-squared/100	−0.293 (0.047)	−0.300 (0.047)	−0.288 (0.047)
Social class
Professional / managerial		_	_
Intermediate		−0.407 (0.133)	−0.345 (0.131)
Manual		−0.911 (0.113)	−0.707 (0.114)
Education
Degree			_
A-level			−0.339 (0.097)
O-level / Other			−0.253 (0.089)
None			−0.794 (0.099)
Household type
Single, lone parent, other			_
Couple no children			−0.159 (0.078)
Couple children			−0.101 (0.070)
Dropout
Full response	_	_	_
Non-response	−0.459 (0.102)	−0.369 (0.101)	−0.325 (0.100)
Death	−0.771 (0.283)	−0.660 (0.283)	−0.587 (0.282)
Thresholds
κ₁	−0.813 (0.189)	−1.401 (0.205)	−1.803 (0.214)
κ₂	2.191 (0.189)	1.601 (0.204)	1.194 (0.213)
Random effects
σ₀	2.270 (0.054)	2.237 (0.054)	2.185 (0.053)
σ₁	0.115 (0.005)	0.114 (0.005)	0.113 (0.005)
ρ	0.333 (0.037)	0.340 (0.037)	0.338 (0.038)
Log-likelihood	−15047.49	15014.74	−14983.17

The interaction of age and cohort is best interpreted graphically and Figure 1 plots the predicted trajectories of belonging to one or more organisations for men who had complete responses. The figure illustrates the influence of the cohort effect on the probability of belonging to at least one organisation. The predicted proportion of respondents in the 1965–1974 cohort who belonged to at least one organisation is approximately 10 per cent lower than that in the 1955–1964 cohort observed at the same age. The corresponding gap between the 1955–1964 and 1945–1954 cohorts is of a similar order of magnitude while that between the 1955–1964 and 1965–1974 cohorts is somewhat less. The results therefore provide some significant evidence for a decline in the probability of organisational membership across cohorts.

Figure 1.

Men: predicted trajectories with age of the probability of belonging to one or more voluntary organisations for each cohort.

Model 1 also adjusts for selection effects resulting from dropout due to non-response and death. The coefficients for the indicators of dropout in Model 1 show that, in comparison to individuals with complete response patterns, the odds of a higher response category were lower among individuals who left the survey and among those who are known to have died. Controlling for the influence of dropout was necessary to reduce the dependence of the age and cohort estimates on the dropout process.

The main interest in Model 2 is the effect of adjusting for social class on cohort differences in the level of organisational membership. In comparison to the results in the previous model, the adjustment for social class resulted in a modest increase in the magnitude of the 1955–1964 and 1965–1974 cohort coefficients which also remain statistically significant. After adjustment for social class the odds of a higher response in the 1965–1974 cohort are 18.2 per cent (or exp(−1.700)) of those of the 1935–1944 cohort when calculated at the median age in the sample. The model coefficients for social class are also of interest and show that for intermediate and manual social class groups the odds of a higher response are 66.5 per cent (or exp(−0.407)) and 40.2 per cent, respectively, of those for the professional/managerial group.

Model 3 examines the effect of additionally controlling for education and household type on cohort differences in levels of organisational membership. The results show that the 1955–1964 and 1965–1974 cohort coefficients remain strongly statistically significant, with adjustment for the additional factors again resulting in a modest increase in the magnitude of the coefficients in comparison to the previous model. Cohort differences in the level of organisational membership cannot be explained by changes in the distribution of social class, education or household type between cohorts. The results also show significant associations between educational qualifications and household type and the number of organisational memberships. The magnitude of the coefficients for categories of education can be judged relatively large: for example, the odds of being in a higher response category for men with no qualifications are exp(−0.794) or 45.2 per cent of those for men with a degree. In comparison, the association between household type and the number of memberships was considerably weaker, but men in couple households without children had a lower odds of being in a higher category in comparison to those in single person households.

Finally, the results of likelihood ratio tests showed that the interaction of cohort with social class was statistically significant (χ² = 13.70, p-value = 0.03) while those with education (χ² = 8.82, p-value = 0.45) and household type (χ² = 5.29, p-value = 0.51) were not. The significant interaction between social class and cohort shows that the effect of social class varies across cohorts, and the variation in the predicted probability of belonging to at least one organisation with social class is plotted in Figure 2 for each cohort. The figure shows that while the social class gradient in the probability of belonging to at least one organisation is relatively flat for the 1935–1944 and 1945–1954 cohorts, it is significantly steeper for the 1965–1974 cohort. The relative differences between social classes in the probability of belonging to at least one organisation are therefore significantly greater in the more recent cohorts in comparison to those in earlier born cohorts.

Figure 2.

Men: predicted probability of belonging to at least one organisation by social class for each cohort, calculated with remaining variables set to zero.

Table 4 shows the corresponding results for women. In Model 1, the cohort coefficients show that the odds of a higher response are higher for the 1945–1954 cohort but lower for the 1965–1974 cohort in comparison to the omitted 1935–1944 cohort. The predicted proportions in each of the response categories for the 1945–1954 cohort were 0.399, 0.515 and 0.085 while those for the 1965–1974 cohort were 0.707, 0.268 and 0.025. The predicted age trajectories in the probability of belonging to one or more organisations are plotted in Figure 3 for women who responded at all interviews. The figure shows the expected relationship between age and organisational membership with the predicted probability of having at least one membership increasing from the late teens into middle age and then slowly declining. The figure also illustrates the influence of the cohort effect with the 1965–1974 cohort having a probability of belonging to an organisation which is notably lower than that of the remaining cohorts. Finally, the results from Model 1 show that dropout from the survey was selective. In particular, women who left the survey had a significantly lower odds of being in a higher response category in comparison to women with complete response patterns

Table 4.

Women: coefficient estimates and standard errors of growth models of number of organisational memberships.

Fixed effects	Model 1	Model 2	Model 3
Cohort
1935–1944	_	_	_
1945–1954 (β02)	0.689 (0.185)	0.532 (0.183)	0.325 (0.181)
1955–1964 (β03)	0.176 (0.192)	−0.050 (0.190)	−0.434 (0.190)
1965–1974 (β04)	−0.604 (0.191)	−0.869 (0.190)	−1.390 (0.191)
Age (β10)	0.078 (0.017)	0.075 (0.017)	0.071 (0.017)
Age x cohort
Age x cohort 1945–1954 (β12)	−0.106 (0.015)	−0.101 (0.015)	−0.102 (0.014)
Age x cohort 1955–1964 (β13)	−0.132 (0.021)	−0.129 (0.021)	−0.134 (0.021)
Age x cohort 1965–1974 (β14)	−0.150 (0.028)	−0.146 (0.028)	−0.151 (0.028)
Age-squared/100	−0.394 (0.044)	−0.392 (0.044)	−0.377 (0.044)
Social class
Professional / managerial		_	_
Intermediate		−0.685 (0.128)	−0.514 (0.124)
Manual		−1.398 (0.107)	−1.089 (0.107)
Education
Degree			_
A-level			−0.246 (0.094)
O-level / Other			−0.369 (0.080)
None			−1.075 (0.088)
Household type
Single, lone parent, other			_
Couple no children			−0.012 (0.068)
Couple children			0.165 (0.062)
Dropout
Full response	_	_	_
Non-response	−0.755 (0.100)	−0.687 (0.098)	−0.580 (0.095)
Death	−0.569 (0.307)	−0.492 (0.288)	−0.335 (0.278)
Thresholds
κ₁	0.281 (0.175)	−0.702 (0.190)	−1.093 (0.197)
κ₂	3.058 (0.176)	2.070 (0.190)	1.667 (0.197)
Random effects
σ₀	2.318 (0.053)	2.233 (0.052)	2.118 (0.051)
σ₁	0.124 (0.005)	0.120 (0.005)	0.118 (0.005)
ρ	0.354 (0.035)	0.337 (0.035)	0.340 (0.035)
Log-likelihood	−17104.14	−17021.04	−16956.62

Figure 3.

Women: predicted trajectories with age of the probability of belonging to one or more voluntary organisations for each cohort.

In Model 2, the introduction of social class attenuated the 1945–1954 cohort coefficient while the 1965–1974 cohort coefficient increased in magnitude by nearly 50 per cent. Calculated at the median age in the sample, the odds of a higher response for a respondent in the 1965–1974 cohort are now 41.9 per cent (or exp(−0.869)) of those of a respondent in the 1935–1944 cohort. In addition, the coefficients for the social class term also show that for respondents in intermediate and manual groups, the odds of a higher response are 50.4 per cent and 24.7 per cent, respectively, of those of respondents in the professional group. In Model 3, the adjustment for education and household type resulted in some further change in the cohort coefficients in comparison to the previous model. In particular, the 1965–1974 cohort coefficient has nearly doubled in magnitude in comparison to the previous model, while the 1945–1954 cohort coefficient has been notably attenuated and is now statistically insignificant. The increase in the magnitude of the 1965–1974 cohort coefficient is particularly striking and indicates the importance of controlling for individual characteristics for accurate estimation of cohort differences. The additional characteristic most strongly associated with the membership of organisations was education, with those women who had lower educational qualifications having an odds of being in a higher response category that is about 34.1 per cent of those for women with a degree. There was also, however, an association between household type and the level of memberships, with women living in couple households with children having a higher odds of being in a higher response category compared to women in single person households.

Finally, the results of likelihood ratio tests suggested that cohort differences in organisational membership varied by education (χ² = 24.84, p-value = 0.003), but not by social class (χ² = 3.01, p-value = 0.81) or household type (χ² = 6.80, p-value = 0.34). Figure 4 plots the predicted probability of belonging to at least one organisation by education for each cohort. While the figure illustrates the magnitude of cohort differences in the probability of belonging to an organisation, the interpretation of the interaction term is not straightforward due to the non-smooth variation in the probability of belonging to at least one organisation across education categories for the 1945–1954 cohort. In order to aid interpretation, a simpler model was examined which contrasted the effects of education in the combined 1935–1944 and 1945–1954 cohorts with those in the 1955–1964 and 1965–1974 cohorts. This suggests that the interaction between education and cohort is due principally to a decline in the probability of membership among women with O-level/Other qualifications relative to women with degree-level qualifications in the later born cohorts.

Figure 4.

Women: predicted probability of belonging to at least one organisation by education for each cohort, calculated with remaining variables set to zero.

Conclusions

This article has used longitudinal data to examine trajectories in organisational membership from 1991 to 2007 for four 10-year cohorts born between 1935 and 1975. The results suggest that men in the 1955–1964 and 1965–1974 cohorts and women in the 1965–1974 cohort have lower levels of membership across age in comparison to those shown by earlier born cohorts. The decline in the membership of voluntary associations in later cohorts could not be explained by variation in observed characteristics such as social class, education and household type. Indeed, adjustment for these characteristics tended to increase the magnitude of cohort differences found in statistical analyses.

The debate concerning the decline in the membership of voluntary associations in western societies has rarely discussed the interconnections between economic and social inequalities. Differences in the level of membership in voluntary associations between cohorts can be interpreted, however, to reflect the influence of changes in the macro-social and economic conditions experienced by different cohorts. The members of the 1935–1944 and 1945–1954 cohorts benefited from steady economic growth and the development of the welfare state throughout the 1950s and 1960s. These factors may have helped the members of these cohorts to accumulate the resources that support participation in voluntary associations. The experience of collective organisation through union membership and the peace movements of the 1960s may also have given these cohorts an awareness of the benefits of collective action. In contrast, the members of the 1965–1974 cohort experienced a labour market with unemployment of around 10 per cent throughout the 1980s and early 1990s. This period saw a decline in the membership of collective organisations of all types, including unions and political parties, and a retrenchment of the welfare state. These changing conditions may have influenced the level of organisational membership by reducing both the resources and opportunities for membership and individuals’ willingness to take part in collective activities of all kinds.

The reasons suggested here for the decline in organisational memberships across cohorts have a different emphasis to those provided by Putnam in America (1995). Putnam attributes the decline in the level of involvement in voluntary associations primarily to differences in the ‘civic mindedness’ of different cohorts while changing economic conditions are of lesser importance. It is unlikely to be possible to distinguish between these different interpretations empirically and it may be that there are different factors at work in the two countries. While the members of different cohorts may share similar attitudes and values, the preferred explanation in this work is that these are primarily an effect of changes in economic conditions. This explanation provides a link between changes in behaviour at the individual level and changing macro-social and economic conditions (Little, 1990), rather than relying on independent changes in a diffuse and hard to quantify characteristic such as ‘civic mindedness’.

In addition to changing economic and social conditions, the policies of successive governments may have influenced the extent to which men and women join voluntary associations. In particular, successive governments have sought to increase economic competitiveness by introducing programmes of privatisation, increased labour market flexibility and deregulation. Market-mimicking forms of governance have been introduced into many areas of the public sector, including health and education, despite an absence of evidence that they would increase the efficiency of resource allocation (Gibson and Asthana, 2000). Markets provide a weak basis for social order, however, and as numerous studies of the effects of contemporary working conditions on the attitudes of employees have shown, the removal of social protection and increase in competitive forces has undermined the habits of sociability and commitment that are needed for voluntary associations to prosper (Toynbee, 2003).

While the results presented here suggest that levels of membership in voluntary associations are declining, the overall number of voluntary associations in the UK has increased over the last several decades (NCVO, 2010). In America, Sampson et al. (2005) have argued that civic participation may now take the form of involvement in events organised by local community organisations rather than individual membership of organisations. The level of organisational membership may not therefore provide a complete account of the degree to which individuals have group interests. It has also been argued, however, that new organisations are less about mass mobilisation and more about managerial forms of civic organising (Jordan and Maloney, 2007). In the UK, the main beneficiaries of the social economy have been the state, the private sector and a growing group of social economy professionals (Amin et al., 1999), and some greater change in social relationships and economic institutions might perhaps have been expected if growth in the UK’s voluntary sector in recent years had reflected an increase in grassroots activities.

In terms of policy, the age trends in membership levels for different cohorts provide an understanding of how patterns of membership may develop for future cohorts as a result of ongoing patterns of social and economic change. These highlight the magnitude of the challenge for government policies that seek to increase the level of membership of voluntary associations. It is not readily apparent, however, what an appropriate policy response might be. It has been suggested that recent patterns of social and cultural change in the UK are mainly due to the loss of the defining role of the working class in shaping social and economic change (Savage, 2000). At present, however, a renewal of the labour movement in the UK appears unlikely. Although one aim of current policy is to increase the membership of community groups, community represents a rather weak basis for organising in modern societies (Sampson, 1999). The emphasis on localism in recent policy may have benefits in allowing local people to shape their own communities and services, but in a country with a population of around 60 million people, local movements are unlikely to influence the overall pattern of social change without the development of political institutions at an intermediate level to concentrate and organise their efforts (Olson, 1977). The introduction of National Citizen Service for young people seems to take the view that the decline in participation is confined to this group and is a result of a lack of individual ability. Families are, however, the main influence shaping the attitudes and experiences of children and youth, and policy might be more effective if it took a more inclusive approach to increasing the involvement of both adults and children.

A more complete understanding of changes in the level of membership in voluntary associations needs to examine how past experiences shape current behaviour and how individual behaviour accumulates over time in social structures. The present analysis treats social class as a time-invariant individual characteristic and further work might attempt to examine the process by which individuals accumulate capital of various kinds as they move across the lifecourse (Savage et al., 2005). It might also be possible to study the specific events associated with joining and leaving voluntary associations; however, measurement error can have severe consequences for the accuracy with which transitions are measured in longitudinal categorical data (Maccoby, 1956). These topics, however, lie outside the scope of the present study.

The analysis in this article has some weaknesses. The response is a count of the number of types of organisations that the respondent was involved in and not the overall number of organisational memberships. There may also be different trends for different groups of organisations (Rotolo and Wilson, 2004). The data used do not allow a complete picture of the types of organisations that respondents may take part in. Membership of professional organisations has not been asked at all interviews and involvement in this type of organisations was not considered here. In a society where markets play an increasing role, the membership of professional organisations may be expected to have increased. The lack of a specific category for cultural and arts organisations is also a weakness. The study also did not consider whether informal patterns of participation may have substituted for membership of organisations. In terms of the statistical analysis, the lack of any control for period effects might be considered a weakness; however, cross-sectional surveys have shown that the overall level of involvement in voluntary activities has remained largely unchanged in the UK since the early 1980s (Hilton et al., 2012).⁶ Despite these shortcomings the study provides information on trends over time in organisational membership for different cohorts and serves as a basis against which the effectiveness of developing government policies aimed at increasing membership can be gauged.

Footnotes

Funding

The Third Sector Research Centre is funded by the ESRC, the Office for Civil Society, and the Barrow Cadbury Trust (ESRC ref: RES-595-28-0001).

Notes

Andrew McCulloch is a social scientist whose main interests are in social stratification, education and health.

References

Amin

Cameron

Hudson

(1999) Welfare as work? The potential of the UK social economy. Environment and Planning A 31(11): 2033–51.

Bryson

Gomez

(2005) Why have workers stopped joining unions? The rise in never-membership in Britain. British Journal of Industrial Relations 43(1): 67–92.

Furlong

Cartmel

(2007) Young People and Social Change: New Perspectives. Maidenhead: McGraw-Hill/Open University Press.

Gallie

(1999) The labour force. In: Halsey

Webb

(eds) Twentieth-Century British Social Trends. Basingstoke: Macmillan, 281–323.

Gibson

Asthana

(2000) Local markets and the polarization of public-sector schools in England and Wales. Transactions of the Institute of British Geographers 25(3): 303–19.

Hall

(1999) Social capital in Britain. British Journal of Political Science 29: 417–61.

Halsey

(1999) Education. In: Halsey

Webb

(eds) Twentieth-Century British Social Trends. Basingstoke: Macmillan, 221–53.

Hilton

Crowson

Mouhot

McKay

(2012) A History of NGOs in Britain: Charities, Civil Society and the Voluntary Sector since 1945. Basingstoke: Mcmillan.

Jordan

Maloney

(2007) Democracy and Interest Groups: Enhancing Participation? Basingstoke: Palgrave Macmillan.

10.

Kan

(2008) Does gender trump money? Housework hours of husbands and wives in Britain. Work, Employment & Society 22(1): 45–66.

11.

Savage

Pickles

(2003) Social capital and social exclusion in England and Wales (1972–1999). British Journal of Sociology 54(4): 497–526.

12.

Savage

Warde

(2008) Social mobility and social capital in contemporary Britain. British Journal of Sociology 59(3): 391–411.

13.

Little

(1990) Varieties of Social Explanation. Boulder, CO: Westview Press.

14.

Lynn

(2006) Quality Profile: British Household Panel Survey. Colchester: University of Essex.

15.

Maccoby

(1956) Pitfalls in the analysis of panel data: A research note on some technical aspects of voting. American Journal of Sociology 61(4): 359–62.

16.

Miyazaki

Raudenbush

(2000) Tests for linkage of multiple cohorts in an accelerated longitudinal design. Psychological Methods 5(1): 44–63.

17.

NCVO (2010) The UK Civil Society Almanac. London: National Council for Voluntary Organisations.

18.

Olson

(1977) The Logic of Collective Action: Public Goods and the Theory of Groups. Cambridge, MA: Harvard University Press.

19.

Pattie

Seyd

Whiteley

(2004) Citizenship in Britain: Values, Participation and Democracy. Cambridge: Cambridge University Press.

20.

Putnam

(1995) Bowling alone: America’s declining social capital. Journal of Democracy 6(1): 65–78.

21.

Rabe-Hesketh

Skrondal

(2008) Multilevel and Longitudinal Modeling Using Stata. College Station, TX: Stata Press.

22.

Rose

Pevalin

(2003) A Researcher’s Guide to the National Statistics Socio-economic Classification. London: Sage.

23.

Rotolo

Wilson

(2004) What happened to the ‘long civic generation’? Explaining cohort differences in volunteerism. Social Forces 82(3): 1091–121.

24.

Ryder

(1965) The cohort as a concept in the study of social change. American Sociological Review 30(6): 843–61.

25.

Sampson

(1999) What ‘community’ supplies. In: Ferguson

Dickens

(eds) Urban Problems and Community Development. Washington, DC: Brookings Institution Press, 241–92.

26.

Sampson

McAdam

MacIndoe

Elizondo

(2005) Civil society reconsidered: The durable nature and community structure of collective civic action. American Journal of Sociology 111(3): 673–714.

27.

Savage

(2000) Class Analysis and Social Transformation. Buckingham: Open University.

28.

Savage

Warde

Devine

(2005) Capitals, assets, and resources: Some critical issues. British Journal of Sociology 56(1): 31–47.

29.

Skrondal

Rabe-Hesketh

(2009) Prediction in multilevel generalized linear models. Journal of the Royal Statistical Society: Series A 172(3): 659–87.

30.

Stolle

Howard

(2008) Civic engagement and civic attitudes in cross-national perspective: Introduction to the symposium. Political Studies 56(1): 1–11.

31.

Taylor

(2009) British Household Panel Survey User Manual: Volume A; Introduction, Technical Report and Appendices. Colchester: University of Essex.

32.

Toynbee

(2003) Hard Work: Life in Low-pay Britain. London: Bloomsbury.

33.

Verba

Schlozman

Brady

(1995) Voice and Equality: Civic Voluntarism in American Politics. Cambridge, MA: Harvard University Press.

34.

Vermunt

Hagenaars

(2004) Ordinal longitudinal data analysis. In: Hauspie

Cameron

Molinari

(eds) Methods in Human Growth Research. Cambridge: Cambridge University Press, 374–93.

35.

Warde

Tampubolon

Longhurst

Ray

Savage

Tomlinson

(2003) Trends in social capital: Membership of associations in Great Britain, 1991–98. British Journal of Political Science 33: 515–25.

Cohort Variations in the Membership of Voluntary Associations in Great Britain,1991–2007

Abstract

Keywords

Introduction

Study Rationale

Data and Statistical Methods

Data

Statistical Methods

Results

Conclusions

Footnotes

Funding

Notes

References