Citizenship and Civic Engagement: Attitudes and Behaviour in Britain



Analysis of patterns of participation revealed three dimensions of civic activism in modern Britain: individualistic activism; contact activism; and collective activism. Three alternative theories of participation were examined to account for these dimensions: general incentives; social capital; and civic voluntarism. None proved sufficient in itself to account for civic activism in modern Britain: each provided only part of the explanation.

Citizenship is a central political concept (for example Almond and Verba, 1963; van Gunsteren, 1998; Norris, 1999; Pharr and Putnam, 2000). It is also increasingly an object of policy concern. Since the early 1990s, electoral turnout has fallen in several major states (Pattie and Johnston, 2001; Johnston, 2001; Gray and Caul, 2000; but see Burden, 2000; McDonald and Popkin, 2001). At only 59 percent, turnout at the 2001 British general election was the lowest since 1918. For some, there is a looming crisis of democratic accountability, as the public abandons conventional electoral politics (but see Heath and Taylor, 1999; Pattie and Johnston, 2001). At the same time, the social capital literature has suggested a correlation between active community life and good government (for example Putnam, 1993; Whiteley, 2000). Even here, however, there are signs of crisis: associational life may be in decline (Putnam, 1995, 2000; but see Norris, 1995; Hall, 1999). Citizenship is important. But existing studies are generally undertaken from just one theoretical perspective (for example Verba and Nie, 1972; Verba et al., 1995; Parry et al., 1992).

Theories of Civic Engagement

We examined three alternative theories of civic behaviour: rational choice, social capital and civic voluntarism.

Rational Choice

In the rational choice model, economically rational actors maximise their benefits from any activity while minimising their costs (Downs, 1957). Civic engagement occurs where the costs of involvement are low and the benefits of successful action are high. But if the costs outweigh the likely advantages, no rational actor will engage.

However, the logic of the rational choice model can lead to paradoxical conclusions, such as Downs's voting paradox. Most elections are decided by more than one or two votes. Rational actors discount potential benefits by their personal efficacy in achieving them. The chances that any one voter's participation will be crucial to the outcome are minuscule. But the less likely it is that one rational actor's participation will be crucial to the outcome, the less likely that individual is to participate at all. But in a truly rational electorate, all should abstain. Paradoxically, the logic of collective action, far from encouraging civic engagement, would seem to result in paralysis.

Why should rational actors participate at all? Riker and Ordeshook (1968) added non-collective benefits, such as personal satisfaction derived from participating in civic life. However, this does not fit easily within the rational choice framework (Mueller, 1987, 1989; Brennan and Lomasky, 1993; Aldrich, 1993; Overbye, 1995). More consistent with the rational choice model, Olson (1965) differentiated between collective and selective benefits. Collective benefits are available to all, whether or not they participate in ensuring the good is provided, whereas selective benefits are restricted to those who participate. For Olson, participation only occurs if some of the benefits of participation could be restricted to those who play an active part and denied to those who free-ride. Hence, cause groups will try to offer privatised incentives for members. Actors are more likely to participate if they receive more benefits, whether collective (accruing to all, irrespective of whether they participate) or selective (available only to those who participate), and they become less likely to participate as the costs increase.

Whiteley and Seyd have recently outlined a ‘general incentives’ rational action model that includes a wider range of incentives as part of the decision-making criteria, combining both rational choice and socio-psychological models of participation (Seyd and Whiteley, 1992; Whiteley et al., 1994; Whiteley and Seyd, 1996, 1998). In this model, participation is a function not only of costs and benefits, but also of people's expressive attachments: those who feel strongly attached to a group should be more likely to act on its behalf than those who do not share that attachment. A sense of duty (referred to here as ‘system benefits’) can encourage civic engagement. Selective benefits are broken down into three elements: process benefits (those people receive as a result of participation in the political process); outcome benefits (privatised advantages accruing as a result of action, such as personally achieving relatively high office); and group benefits (privatised advantages accruing to groups people are concerned about). And people are likely to be influenced by the social norms of those they live among: the more that friends and family think participation is a waste of time, the less likely people should be to participate. The ‘general incentives’ model can be outlined in the following terms:1


where A is activism, p is efficacy, B is collective benefits, C is cost, S is selective benefits, SY is system benefits, SN is social norms for participation and EX is expressive motives for participation.

Social Capital

The social capital model claims that joining and taking part in local organisations helps foster a sense of trust in others (Putnam, 1993). Compared with societies where most citizens do not routinely undertake voluntary activities, societies in which many engage in social and voluntary activity outside of the home are more likely to be trusting, well-governed, affluent and successful (Putnam, 1993; Knack and Keefer, 1997; Hall, 1999; Whiteley, 2000; but see Claibourn and Martin, 2000). Therefore, levels of political activism in a society should be positively correlated to levels of both trust and voluntary activity. In addition, we included measures of how well established people are in their local neighbourhoods: how long they have lived at their current address and whether they have family members living nearby. Those who are settled in an area should also have more opportunity to build social capital than those who are recent arrivals.

Social capital must be nurtured to flourish, however, and can wither if neglected. Television-watching has been claimed to be detrimental to community life (Putnam, 1995, 2000). Time spent in front of the television is time not spent socialising with one's fellow citizens, threatening the store of social capital. This claim has proved controversial, however. Some argue that what matters is not how much, but what sort of, television is watched (Norris, 1995): active citizens may use television to keep abreast of the news, whereas inactive citizens may be interested primarily in entertainment. Be that as it may, we can evaluate an enhanced version of Putnam's social capital thesis:


where A is activism, T is trust (in other people and in institutions), M is membership in groups, N is networks of civic engagement, TV is hours spent watching television, Y is years the respondent has lived at current address and F is whether the respondent has family living nearby.

Civic Voluntarism

The civic voluntarism model is based on a socio-economic model of participation. Resources are important in participation: the better educated, more affluent and more middle class people are, the more likely they are to participate (Verba et al., 1995; Brady et al., 1995; Parry et al., 1992). However, civic voluntarism extends the socio-economic model in several important respects. First, Verba et al. (1995) conceptualise resources widely, including not only economic and educational resources, but also resources of time. Some people are so busy they have little free time to engage in political activism.

Second, as in the rational choice model, citizens’ sense of efficacy is important. The more people feel their opinions and actions are likely to have an influence on the outcome of decisions, the more likely they are to engage in political action.

Thirdly, the civic voluntarism model stresses the importance of general involvement in the political system. Party identification and political engagement should encourage civic engagement.

Finally, Verba et al. (1995) stress the importance of mobilisation. Even when people are resource-rich, have plenty of free time and have a strong sense of efficacy, they may still fail to participate if they are unaware of the importance of their involvement or if no one has tried to elicit their co-operation. Being asked to participate by other people is an important catalyst for individual participation. For instance, local campaigning by political parties can play an important role in mobilising the vote (Bochel and Denver, 1971, 1972; Denver and Hands, 1997; Pattie et al., 1995; Seyd and Whiteley, 1992; Whiteley et al., 1994).

We therefore operationalise the civic voluntarism model as follows:


where A is activism, p is efficacy, R is resources (time, income, etc.), M is mobilisation (‘Have you been asked to get involved?’), I is party identification and E is political engagement.

Testing the Models

We evaluated the three models using the Citizen Audit, a large multi-wave survey of citizenship in the United Kingdom at the start of the twenty-first century. Just over 3,400 individuals in 100 local authorities across Great Britain were interviewed face to face in autumn 2000. At the same time, a much larger sample of randomly selected individuals in the same local authorities were sent a shorter mail survey: just under 10,0000 replied. Finally, in the autumn of 2001, we re- interviewed as many of the original face-to-face survey's respondents as possible (around 1,200 took part). For the following discussion, however, we employ the first wave of the face-to-face survey, since this provides us with the greatest range of potential explanatory variables.

Defining Civic Engagement

We based our analysis of civic engagement on a battery of questions that asked respondents whether, in the previous 12 months, they had undertaken any of a series of different forms of action aimed at influencing rules, laws or policies. The list was deliberately wide ranging (later analyses do not include participating in strike action: exploratory analysis not reported here suggests that this form of action is different in kind to the others covered by the survey).

Not surprisingly, some forms of action were more common than others (Table 1). Relatively low-cost actions were the most widely reported. By far the most common action, for instance, was donating money to an organisation (the vast majority almost certainly covering charitable donations). Voting in local elections was the next most commonly reported action, followed by signing a petition. In no other case did more than a third of respondents say they had undertaken a particular action in the previous 12 months, although substantial minorities said they had undertaken several. And most of these involved more than minimal effort on the part of participants. ‘Ethical consumerism’ was quite important, for instance, with 28 percent saying they had either boycotted or bought particular products for ethical reasons. And 27 percent said they had been active fund-raisers. But in general, forms of action that involved working collectively were less widely indulged in than those that could be carried out by individuals. Only 5 percent said they had attended a political meeting in the previous 12 months, only 4 percent said they had participated in a public demonstration or had formed a group of like-minded citizens, and only 1 percent said they had participated in an illegal protest.

Table 1.  : Levels of Civic Activism
Form of actionPercentage that said they had undertaken the action in the previous 12 months
  1. Source: Citizen Audit Wave 1 face-to-face survey.

Donated money to an organisation56
Voted in local government election45
Signed a petition38
Boycotted a product28
Raised funds for an organisation27
Bought a product for ethical reasons26
Contacted a public official23
Wore a campaign badge/sticker20
Contacted a solicitor18
Contacted a politician12
Contacted an organisation10
Contacted the media 8
Attended a political meeting 5
Participated in a public demonstration 4
Formed a group of like-minded people 4
Participated in a strike 2
Participated in an illegal protest 1
Undertook at least one of the above78
Average number of acts 3

Overall, a large majority of respondents (78 percent) said they had undertaken at least one of the above actions in the previous 12 months, and the average had undertaken three. This measure of civic engagement is not distorted unduly by the relatively high levels of donation to voluntary organisations: even if the latter is excluded, 74 percent said they had undertaken at least one of the other actions and the average had undertaken 2.7. Therefore, the audit suggests a relatively wide range of civic engagement, although the extent of people's involvement, and the time and energy they devoted to their actions, is likely to have varied considerably.

These figures may overestimate civic engagement among the general public: the less active are also less likely to respond to questionnaires (Swaddle and Heath, 1989). We tried to minimise the problem by weighting our survey to the 1991 British Census (details of the 2001 census were not yet available). But there is also a risk of systematic errors in recording civic engagement if respondents give answers consonant with perceptions of socially normative actions: respondents might under-report no activity and over-report engagement. Analyses comparing self-reported and actual electoral turnout in voter surveys suggest that this is an issue, but not a major one (Swaddle and Heath, 1989). That said, even if the estimates of levels of participation are not exact, the numbers of respondents that claimed to have undertaken each action are intuitively sensible, from frequent low-cost actions such as signing petitions, to infrequent high-cost actions such as participating in an illegal protest. Therefore, even if the absolute numbers are not exactly right, we are confident we have the correct relative values.

The above discussion treats all forms of civic engagement as elements of a seamless whole. This is unlikely in practice, however. People who undertake some forms of action may be unlikely to undertake others. For instance, those who write to their MPs when a problem arises are expressing, at some level, faith in how the conventional political system operates and might therefore be unlikely to participate in illegal protests. To analyse whether there are different ‘types’ of civic engagement, we conducted a principal components analysis (responses were coded 1 if respondents said they had undertaken the action, and 0 if they did not). The results reveal three distinct types of civic activism, accounting collectively for 42 percent of the variation in the original data (Table 2).

Table 2.  : Types of Civic Activism: Principal Components Analysis
Action undertaken in the previous 12 monthsType of civic activism
  1. Note: Only the strongest component loadings for each variable are reported.

Donated money to an organisation0.69  
Boycotted a product0.68  
Bought a product for ethical reasons0.67  
Signed a petition0.61  
Raised funds for an organisation0.54  
Voted in a local government election0.52  
Wore a campaign badge/sticker0.52  
Contacted a public official 0.66 
Contacted a politician 0.65 
Contacted the media 0.60 
Contacted an organisation 0.57 
Contacted a solicitor 0.49 
Participated in a public demonstration  0.73
Attended a political meeting  0.66
Participated in an illegal protest  0.66
Formed a group of like-minded people  0.51
Percentage of variance accounted for2697

The first dimension, ‘individualistic activism’ (accounting for 26 percent of the variation in the original data), loads most strongly on ethical consumption, donations, petition-signing, fund-raising, voting in local elections and wearing a campaign badge. People who undertake one of these actions are also likely to undertake others. On the whole, these are relatively individual acts. The dimension has an average score of 0 (positive scores indicate people relatively likely to engage in this type of activism, whereas negative scores indicate those who are relatively unlikely to do so).

The second, ‘contact activism’ (accounting for 9 percent of the variation), picks up on actions that focus on contacting those in authority. People who contacted a public official are also likely to contact politicians, to write to the local media, to contact a solicitor or to contact an organisation. Again, a positive score indicates people particularly likely to undertake this type of activism.

The third, ‘collective activism’ (accounting for 7 percent of the variation), loads most strongly with participation in a public demonstration, but it is also strongly correlated with attending political meetings and participating in illegal protests and (more weakly) with the propensity to form a group of like-minded people. Again, positive scores indicate the more active.

These are quite different types of civic activism, potentially involving different groups of people at different times. People who engage in individualistic activism, for instance, are no more or less likely to engage in collective activities or to con-tact the authorities than those who are not ‘individualistic activists’. Our results are similar to those from an earlier study of political participation in Britain (Parry et al., 1992, pp. 50–54), which uncovered six dimensions of political participation in the mid-1980s: voting, party campaigning, collective action, contacting, direct action and political violence. We have no measure for political violence or party campaigning, but their ‘collective’, ‘contact’ and ‘direct action’ dimensions are clearly echoed in our survey. The underlying structure of political participation remains clearly recognisable after 15 years.

Accounting for Civic Engagement

We measured each of these types of civic activism using the scores for the appropriate component as dependent variables in regression models. As the regressions were designed to test the various theories of civic engagement outlined above, we derived the independent variables from those models. We consider each model in turn, before analysing them together. In each case, as well as including the ‘signature’ variables associated with the relevant theory, we also included controls for respondents’ age, gender, social class, educational attainments and ethnicity. Analyses (not reported here) show that the more middle class, better educated and older people are, the more civically active they are.

‘General Incentives’ Rational Choice

As outlined above, there are nine key variables in the rational choice model: political efficacy, collective benefits, selective process benefits, selective outcome benefits, selective group benefits, costs, system benefits, social norms and expressive incentives. We begin by examining each in turn.

Political efficacy measured the extent to which people feel the political system is responsive to public wishes in general and, more particularly, to the views and opinions of the individual voter. We focus here on the extent to which people felt they were able to make a difference in their daily lives. We asked respondents whether, in the previous 12 months, they had been able to influence either their employment conditions or any medical treatment they or their relatives had received. Although 61 percent said they had tried to have some influence on medical treatment, only 3 percent felt they had succeeded (Table 3). Fewer (32 percent) said they had tried to influence their work conditions, but more (15 percent) felt they were successful when they tried. We created an ‘efficacy’ score by summing responses to both questions: the scale has a minimum score of 0 (those who did not try to influence either medical treatment or work conditions) and a maximum of 4 (those who were successful in influencing both). The average was 1.1 on the scale, suggesting some attempt by people to influence their lives, but with little success. However, the rational choice model rests on the perceived costs and benefits of action. As outlined above, three groups of benefits can be identified: collective, selective and system.

Table 3.  : Defining the ‘General Incentives’ Rational Choice Model
‘Collective benefits’ scale: principal components analysisComponent loadingsAverage score
Satisfied with the way democracy works in Britain0.723.03
Satisfied with own working conditions0.642.26
Satisfied with medical treatment0.482.71
Percentage of variance accounted for39 
‘Efficacy’ scalePercentage that did not try to influencePercentage that had no influencePercentage that had some influence
Able to influence medical treatment3958 3
Able to influence work conditions691715
Average summed score 1.09 
‘Costs of involvement’ scalePercentage that agreed
Politics would take time away from one's family60
Politics can be tiring after a hard day's work69
People like me are too busy to vote18
Average summed score10.0
‘Selective benefits’ scalesPercentage that agreed
Process incentives
 Participating in politics is not much fun56
 Participating in politics is a good way to meet people26
 Average summed score5.2
Outcome incentives
 Person like me could do a good job as a councillor19
 Politics would be more effective if people like me in parliament27
 Average summed score5.6
Group incentives
 Politics is a good way to get benefits for oneself and one's family23
 Politics is a good way to get benefits for groups that I care about36
 Average summed score6.0
‘System benefits’ scalePercentage that agreed
Every citizen should get involved in politics if democracy is to work40
If person dissatisfied with government policy, duty to do something54
Average summed score6.8
‘Social norms’ scalePercentage that thought their friends would agree
People who get involved in politics are often a bit odd32
Members of Parliament are respected figures in the community30
Only way to change anything in society is to get involved52
Average summed score9.4
‘Expressive incentives’ scalePercentage that agreed
  1. Note: Details of the codings for the variables are available from the lead author.

Very proud to be a British citizen30

Collective benefits were measured by a principal components analysis of responses to three questions on satisfaction with the way democracy works in Britain, own working conditions and medical treatment in the previous 12 months. Each of the three variables reflects outcomes that are affected by collective action, hence their relationship to perceived collective benefits. Each variable was measured on a five-point scale, ranging from 1 (satisfaction) to 5 (dissatisfaction). The average score was near the middle of the scale (i.e. a score of 3 – neither satisfied nor dissatisfied) in terms of satisfaction with British democracy, but there was relative satisfaction with working conditions and medical treatment. Responses to all three questions are closely correlated, and a principal components analysis shows that one common dimension underlies all three (the component itself accounts for 39 percent of the variation in the original variables). We used the standardised component scores as our measure of respondents’ evaluations of collective benefits. Given the way the original variables were coded, positive component scores suggest people who are relatively dissatisfied and hence do not think there are effective collective benefits to be had, whereas negative scores suggest people who are relatively satisfied and who do think there are collective benefits.

Three different sorts of selective benefits were assessed: process, outcome and group incentives. Each was measured using Likert scales, ranging from 1 (responses suggesting few selective benefits) to 5 (responses suggesting considerable selective benefits). The scores were then summed to create an incentives scale for each sort of selective benefits, ranging from 2 (few incentives) to 10 (considerable incentives). Process incentives reflect the personal pleasures people can gain from political involvement. We asked respondents whether they agreed with the following statements: ‘Participating in politics is not much fun’ (56 percent agreed) and ‘Participating in politics is a good way to meet interesting people’ (26 percent agreed). Outcome incentives reflect people's perceptions of privatised benefits, such as the desirability of becoming political representatives. Those with strong outcome incentives are likely to value situations where politicians are identifiably people ‘like themselves’. We asked respondents whether they agreed with the following statements: ‘A person like me could do a good job as a local councillor’ (19 percent agreed) and ‘Politics would be a lot more effective if people like me were elected to parliament’ (27 percent agreed). Group incentives are related to the results of participation. We asked respondents whether they agreed with the following statements: ‘Being active in politics is a good way to get benefits for oneself and one's family’ (23 percent agreed) and ‘Being active in politics is a good way to get benefits for groups that I care about’ (36 percent agreed).

System benefits were measured by summing responses to two questions: we asked respondents whether they agreed with the following statements: ‘Every citizen should get involved in politics if democracy is to work properly’ (40 percent agreed) and ‘If a person is dissatisfied with the policies of the government, he/she has a duty to do something about it’ (54 percent agreed). The ‘system benefits’ scale ranges from 2 (few benefits) to 10 (considerable benefits).

Costs of involvement are important in rational choice. We asked respondents whether they agreed with the following statements: ‘Being involved in politics would take time away from one's family’ (60 percent agreed); ‘Getting involved in politics can be tiring after a hard day's work’ (69 percent agreed); and ‘People like me are so busy these days that they don’t have the time to vote’ (18 percent agreed). Responses to each were coded on a five-point scale (with a high score suggesting perceptions of significant costs). Again, we summed the scores to create, in this case, a ‘costs of involvement’ scale that ranged from 3 to 15 (with 15 indicating perceptions of considerable costs).

A ‘social norms’ scale was created by summing responses to three questions: we asked respondents whether they felt people whose opinions were important to them would agree with the following statements: ‘People who get involved in politics are often a bit odd’ (32 percent thought their friends would agree); ‘Members of Parliament are respected figures in the community’ (30 percent thought their friends would agree); and ‘The only way to change anything in our society is to get involved’ (52 percent thought their friends would agree). In each case, responses were coded so that high scores reflected perceptions of social norms that would encourage participation. The ‘social norms’ scale ranges from 3 (weak perceptions of social norms for participation) to 15 (strong perceptions of social norms favouring participation).

Finally, expressive incentives reflect people's general feelings of attachment to key groups or ideas. We asked respondents a question about attachment to the British polity: ‘Thinking about what it means to be a citizen of Britain today, would you say that you were very proud, somewhat proud, not very proud, or not at all proud to be a British citizen?’ (30 percent felt very proud). The variable was coded from 0 (not at all proud) to 3 (very proud).

Regression analysis was used to examine how well the ‘general incentives’ rational choice model accounts for civic engagement (Table 4). Interestingly, perceptions of collective benefits seem to discourage participation. The relevant coefficients are positive and significant in all but collective activism: people are more likely to free-ride if they think the system delivers. But, as Olson (1965) suggests, selective benefits encourage participation. The more selective benefits people perceive as emerging from participation, the more activities they are likely to engage in. Selective process incentives are significant and correctly signed in all but collective activism, and outcome incentives encourage higher levels of both total and contact activism. Only group incentives are different to expectations: they are insignificant in most types and are negatively related to contact activism. In other words, the more people think participation might benefit groups they feel close to, the less likely, other things being equal, they are to contact officials.

Table 4.  : Accounting for Civic Activism: The ‘General Incentives’ Rational Choice Model – OLS Regressions
VariableType of civic activism
  1. Note: *Significant at P=0.05; **significant at P=0.01. OLS, ordinary least squares. The collective benefits scale is a standardised component score derived from this principal component analysis in Table 3.

Collective benefits0.30**0.06**0.08**0.03
Selective process benefits0.13**0.03*0.04*0.02
Selective outcome benefits0.18**0.020.08**0.02
Selective group benefits−0.06−0.00−0.04**−0.00
System benefits0.28**0.08**0.03*0.04**
Social norms0.01−*
Expressive motives0.000.02−0.02−0.04
Gender (comparison=male)
Ethnicity (comparison=white)
 Other ethnic groups−0.64**−0.28**0.04−0.08
Years in education0.28**0.09**0.010.07**
Social class0.20**0.06**0.05**−0.01

A sense of efficacy also encourages most types of activism (except collective). Similarly, the more people feel that citizens in general should participate for the good of society (as measured by system incentives), the more likely they are to be active. And social norms are an incentive for collective action: the more people perceive that others close to them value political participation, the more likely they are to engage in collective activism. However, expressive motives of action are not related to any of the types of activism.

Perceived costs of involvement, though, are negatively related to both total and individualistic activism (as theory suggests they should be). The more troublesome people think civic engagement is likely to be, the less likely they are to engage. However, this is not true for all types of activism: perceptions of cost are not related to either contact or collective activism.

Therefore, the ‘general incentives’ model, at first glance, holds up well. Most variables perform as expected for most types of activism. Part of the process underlying the decision whether or not to engage in civic activism rests on a calculus of costs and benefits.

Social Capital

The social capital model estimates the effects of four sets of independent variables: trust (divided into trust in others and trust in institutions; Newton, 1999); membership of voluntary organisations and of informal networks beyond the family; television-watching habits; and embeddedness in a local community. The first two sets of measures supposedly contribute positively to social capital; the third is possibly inimical to it (Putnam, 1993, 2000). We hypothesised that the fourth group also contributes positively to social capital.

Our measure of trust in others is based on three questions: we asked respondents to indicate, on an 11-point scale that ranged from 0 (the least trusting response) to 10 (the most trusting), whether people with whom they had contact tried to be helpful or mostly looked out for themselves; whether they would try to take advantage of them if they had the chance or would try to be fair; or whether they could be trusted. This is somewhat different to conventional measures of trust, which tend to ask not about people that respondents are in contact with, but about people in general. However, the conventional formulation is too general to mean much: trust is only meaningful where some form of reciprocal action is expected (Hardin, 1999; Offe, 1999). We prefer our more specific formulation, therefore, as it asks people to think about those who they encounter rather than about large social abstracts.

The results reveal that people are relatively, but not overwhelmingly, trusting of their fellow citizens. The average scores on all three questions were around 6.5, just over the theoretical mid-point of the scales and closer to the most trusting than to the most cynical possible responses (Table 5). In a principal components analysis, all three measures contribute to the same underlying component, which accounts for 74 percent of the original variation. The standardised component scores provide us with a ‘trust in others’ scale, with positive scores indicating trusting people and negative scores indicating the less trusting.

Table 5.  : Defining the Social Capital Model
‘Trust in others’ scale: principal components analysisComponent loadingsAverage score
How helpful are people0.876.1
How fair are people0.866.3
Level of trust in people0.846.6
Percentage of variance accounted for74 
‘Trust in institutions’ and ‘trust in non-political institutions’ scalesComponent loadingsAverage score
Trust in political institutionsTrust in state non-political institutions
Trust in government0.90 3.7
Trust in House of Commons0.89 3.9
Trust in politicians0.77 3.3
Trust in local government0.57 4.6
Trust in the police 0.816.3
Trust in the courts 0.815.5
Trust in the civil service 0.775.5
Percentage of variance accounted for5815 
‘Group membership’ scale
Average number of associations respondent a member of0.84
‘Informal networks’ scalePercentage that were in a network
Belong to informal network19
‘Television-watching’ scalePercentage that watched 4+ hours daily
Hours of television watched on average weekday34
Hours of television watched on average day at weekend43
‘Years at address’ scalePercentage that had lived at current address for 5+ years
Years lived at current address55
‘Family near’ scalePercentage that had family nearby
Family living nearby57

We also asked a series of questions about trust in a variety of institutions: the government, the House of Commons, politicians, local government, the police, the courts and the civil service. These, too, were on 11-point scales, ranging from 0 (the institution was not trusted at all) to 10 (the institution was trusted completely). The results are striking. There is a clear distinction in the public mind between political and non-political state institutions. People are not very trusting of political institutions (for example the average ‘trust in politicians’ score was only 3.3), but they are much more likely to trust non-political state institutions such as the police, the courts and the civil service (all of which averaged 5.5 or above). This extends to the underlying structure of opinion revealed by principal components analysis. Two components emerge from the analysis. The first (accounting for 58 percent of the original variance) is strongly related to trust in political institutions such as government, the House of Commons and politicians. The second component (accounting for a further 15 percent of the variance) loads most heavily on trust in non-political state institutions: the police, the courts and the civil service. The political and non-political wings of the state are clearly seen as quite distinct. Our measure of institutional trust concentrates on the component scores for the first dimension (trust in political institutions): the more positive the score, the higher the trust.

Social activity is central to Putnam's analysis of social capital. We measured this using a battery of questions that asked respondents which of a wide range of clubs, societies and organisations they were involved with. The list includes cause groups like Greenpeace, hobby groups, leisure organisations (for example sports clubs), citizen groups (for example parent–teachers’ associations), charities and groups that sell benefits to members (for example the main motoring organisations). The voluntary activism index counts the number of types of organisations people are members of (this is a relatively conservative measure, as it ignores voluntary action that does not necessarily involve membership – helping out with some activity, for instance). Just over 40 percent of all respondents said they were members of at least one type of group, and the average was a member of 0.8 types of groups. In addition, citizens can also be active in a variety of informal social networks of friends and neighbours. We asked respondents whether they belonged to an informal network of friends or acquaintances that met on a regular basis: 19 percent said they did.

The next independent variable is television-watching. We asked respondents to assess how many hours they watched television on an average weekday and on an average weekend. Our television-watching index is the average of the scores on two 13-point scales (one for watching on weekdays, the other at weekends), each ranging from 1 (no television-watching), 2 (an average of up to an hour per day), 3 (an average of 1–2 hours), and so on, up to 13 (11 or more hours). Only 1 percent said they watched no television at all. The median respondent, though, watched television for around 3–4 hours per day.

Finally, we included two measures of social embeddedness in the local community. The first takes into account how long people have lived at their current address, the implication being that the longer they have lived there, the more embedded they are in the local community: 55 percent said they had lived at their current address for five or more years. Our second measure is whether people had close family members living nearby: 57 percent said they did.

Some aspects of the social capital model fare well in accounting for civic engagement, but others do not (Table 6). Participation in voluntary organisations and in informal networks are both highly significant correlates of civic activism. The more groups people are members of and the more active they are in informal networks, the more civic actions they are likely to undertake and the more likely they are to engage in all three types of activism. As Putnam suggests, the socially active are most likely to be politically active.

Table 6.  : Accounting for Civic Activism: The Social Capital Model – OLS Regressions
VariableType of civic activism
  1. Note: *Significant at P=0.05; **significant at P=0.01. OLS, ordinary least squares.

Trust in others−0.100.01−0.05*−0.05*
Trust in political institutions0.000.02−0.030.01
Group membership0.51**0.16**0.10**−0.01
Informal networks0.93**0.19**0.20**0.19**
Hours of television watched daily−0.05−0.03**0.01−0.00
Years lived at current address0.000.04*−0.04−0.03
Family nearby0.200.030.11*−0.05
Gender (comparison=male)
Ethnicity (comparison=white)
 Other ethnic groups−0.76**−0.31**−0.01−0.05
Years in education0.22**0.06**0.010.06**
Social class0.15**0.04**0.05**−0.01

However, other aspects of the social capital model are less convincing. Trust hardly features in our models. Trust in others is significant, and negatively signed, in just two types of civic activism: contact and collective. The more people trust in others, the less likely they are to contact officials or to participate in collective action. But trust in political institutions is insignificant in all types. Insofar as it matters at all, therefore, trust operates through the opinions of fellow citizens and is a negative influence on activism: trusting people are, if anything, less active than their less trusting fellows – the opposite to the relationship posited by Putnam. The generally weak results for trust are something of a surprise, since this is a key variable in the social capital model. One possibility, of course, is that trust ‘fails’ in our models because trust and group membership are closely correlated. However, that is not the case: the correlation is significant, but weak. People who are likely to join many groups are, on the whole, slightly more trusting than those who do not join groups, but group membership accounts for only trivial amounts of the variation in either trust measure. And when we repeated our social capital models, excluding both group membership and activity in informal networks, trust still failed to reach significance (results not reported here). One possibility is that trust is a double-edged sword. People who trust a lot may become active because they feel their actions are likely to be successful, but they may remain inactive because they feel the ‘powers that be’ can be trusted to get on with the job. Equally, those who do not trust may include some activists (who want to ensure that the rascals in office are brought to book) as well as some disengaged cynics (who feel the rascals cannot be controlled by the public). If this is true, then trust might have opposite and self-cancelling effects on activism for different people.

Nor does television-watching have quite the effects suggested by Putnam. In line with his ‘bowling alone’ thesis, the more people watch television, the less likely they are to engage in individualistic activism (the relevant coefficient is negative and significant). However, contrary to expectations, television-watching is not associated with contact or collective activism or with the total number of activities engaged in. Television is not the culprit Putnam claims. As Norris (1995) points out, it is not just a question of how much television is watched, but also of what is watched. Blaming television is almost technological determinism. But, as with any technology, it is what you do with it that counts.

Nor is embeddedness in the local community consistently related to activism. The longer people have lived at their current address, the more likely they are to participate in some form of individualistic civic action. But period of residence has no effect on any other type of activism. Similarly, having family members in the local community is related only to contact activism. Other things being equal, people with family living locally are more likely to contact officials than are those with families not living locally.

The most consistent result from the social capital models, therefore, suggests that voluntary activity, broadly defined, encourages civic activism. To encourage civic participation, therefore, encourage people to be active in their communities (even if that means joining a local gym). Building trust and changing viewing habits, or even creating more stable, family-centred communities, will not work.

Civic Voluntarism

A number of the factors underlying the civic voluntarism model are already included among the control variables (such as class and education). But we also need some additional measures. Like the rational choice model, civic voluntarism requires a measure of political efficacy. However, the exact specification of the efficacy measure is different here. Whereas efficacy in rational choice rests on people's abilities to have a real effect (and hence our measure focused on actual achievements), efficacy in civic voluntarism has more to do with generalised perceptions of the chances of having an impact (or ‘subjective confidence’; Parry et al., 1992, p. 173; Almond and Verba, 1963, p. 137). We looked at whether respondents agreed with the following four statements: ‘Sometimes politics and government seem so complicated that a person like me cannot really understand what is going on’; ‘People like me have no say in what the government does’; ‘It really matters which party is in power, because it will affect our lives’; and ‘People like me can have a real influence on politics if they are prepared to get involved’. In each case, responses were coded from 1 (perceptions of limited efficacy) to 5 (a strong sense of efficacy). A principal components analysis reveals two separate structures of opinion (Table 7). The first, loading most heavily on measures for ‘politics and government are so complicated’ and ‘have no say in what government does’, reflects feelings of personal efficacy and accounts for 41 percent of the original variance. The second (accounting for another 26 percent of the original variance) loads most strongly on measures for ‘it really matters which party is in power’ and ‘can have a real influence on politics if involved’; in other words, this second component reflects a more general sense of what we might term ‘system’ efficacy – does the system deliver? We concentrated on respondents’ sense of personal efficacy, using their scores on the first component as our measure: those with positive scores have a relatively strong sense of personal efficacy, whereas those with negative scores do not.

Table 7.  : Defining the Civic Voluntarism Model
‘Efficacy’ scale: principal components analysisComponent loadings
Personal efficacySystem efficacyAverage score
Politics and government are so complicated0.82 2.6
Have no say in what government does0.72 2.5
It really matters which party is in power 0.853.4
Can have a real influence on politics if involved 0.713.1
Percentage of variance accounted for4126 
‘Free-time’ scalePercentage that had 4+ hours free time daily
Hours in other activities on average weekday25
‘Household income’ scalePercentage that earned £20,000+
Pre-tax annual household income29
‘Mobilisation’ scalePercentage that had been asked to participate
Mobilisation by another22
‘Engagement’ scalePercentage that were fairly/very interested
How interested in local politics32
How interested in regional politics30
How interested in national politics44
How interested in European politics23
How interested in international politics29
Average summed score4.9
‘Partisanship’ scaleNoneNot very strongFairly strongVery strong
Strength of party identification42232510

The resources available to people are important in the civic voluntarism model. Two forms of resources were included. Civic voluntarism stresses the importance of free time. We asked respondents how much time they had left on an average weekday, once they had completed their other commitments to paid employment, study, domestic work, leisure activities and sleep. The variable was coded on a 13-point scale, ranging from 1 (no free time), 2 (up to an hour of free time), and so on, in hourly increments, up to 13 (11 or more hours of free time). Only 25 percent said they had more than 4 hours of free time, and 17 percent said they had no free time at all. We also include economic resources, as measured by respondents’ total household annual incomes. This is measured using an ordinal scale that rises in £10,000 steps, from incomes under £10,000 per annum (coded 1) to incomes of over £80,000 per annum (coded 9). The average was between £10,000 and £20,000 per annum.

A fourth important element of the civic voluntarism model is mobilisation – being invited to participate. We measured this by combining responses to two questions. First, we asked respondents whether, in the previous 12 months, they had been personally asked to undertake any of the forms of civic action discussed above: 23 percent said they had been approached. We then asked who had approached them: a family member, a close friend, a work colleague, a stranger or someone else? Around 5 percent of all those who had been asked to participate said they had been approached by a family member; 19 percent by a close friend; 12 percent by a work colleague; and 51 percent by a stranger (the remainder were unable or unwilling to say who had approached them). By combining these two questions, we derived a four-point scale, ranging from 0 (not asked at all) to 4 (asked by a family member or close friend). We expect that the closer the relationship with the person making the request, the more likely people would be to comply with it.

The civic voluntarism model suggests that people are more likely to be active if they are interested in the issue. We measured this psychological engagement with a battery of questions that asked respondents how interested they were in local, regional, national, European and international politics. Responses to each were coded on a four-point scale, ranging from 0 (no interest) to 3 (a great deal of interest). Interest in all aspects of politics was quite low: only in the case of national politics did more than a third of respondents indicate some interest – and then only 44 percent. To produce the overall ‘engagement’ scale, we summed responses to each of the ‘interest in politics’ questions to create a scale that ranged from 0 (not interested in any aspect of politics) to a theoretical maximum of 15 (interested in all aspects of politics).

Some elements of the civic voluntarism model emerge well from the regression analyses (Table 8). Mobilisation is a very consistent influence on civic engagement. People who are asked to undertake a particular action are more likely to be civically active than those who are not, particularly if they are asked by a close acquaintance. Therefore, people respond most readily to requests from friends and family. But failing that, they are also responsive to requests from more distant acquaintances – better to be asked than not to be asked at all. Similarly, political engagement is positively related to all types of civic activism: the more interested people are in politics generally, the more likely they are to become active.

Table 8.  : Accounting for Civic Activism: The Civic Voluntarism Model – OLS Regressions
VariableType of civic activism
  1. Note: *Significant at P=0.05. **significant at P=0.01. OLS, ordinary least squares.

Personal efficacy0.070.030.01−0.02
Free time−0.03−0.01−0.010.00
Household income0.09**0.08**−0.04**−0.03*
Strength of party identification0.16**
Political engagement0.19**0.05**0.03**0.02**
Gender (comparison=male)
Ethnicity (comparison=white)
 Other ethnic groups−0.56**−0.25**0.04−0.07
Years in education0.11**0.04**−0.010.05**
Social class0.11**0.02**0.05**−0.02

Resources were more ambiguous. Class and education behave broadly as the literature suggests: the longer people have been in formal education and the more middle class and the older they are, the more likely they are to undertake most forms of civic action. Similarly, household income is a significant factor in all types of civic activism. But it is related to each in different ways. The more affluent people's households are, the more likely they are to undertake some form of individualistic action and the more activities they will engage in overall. But people from affluent households are less likely than those from poorer ones to engage in contact and collective activism. And whereas income-based resources are related, albeit in different ways, to different types of activism, time-based resources are not. Free time is not related to activism on any of the scales. Busy people are no less or more likely to engage in civic activism than are those with time on their hands.

Personal efficacy is not significant in the civic voluntarism model, in contrast with the somewhat different rational choice measure of efficacy. Partisanship, though, was related to total activism. The more strongly attached to a party people feel, the more civic activities in total they are likely to engage in. But it was not related to particular types of activism.

Comparing the Models

So far, we have looked at each of the models in isolation. The actual decision to participate is likely to be influenced by a wide combination of factors, however, and it is unlikely that any one theory will capture all the various influences on civic engagement. Therefore, we compared the models using encompassing tests (Granger, 1990; Hendry, 1995; Whiteley and Seyd, 1996, 2002). The basic principle was to discover whether one theory can account for another. We tested this by using the data in Tables 4, 6 and 7 to generate predicted values of activism for each of our models. These predicted values give us measures of that part of activism that can be accounted for by each theory. If the theory is encompassed by another theory, then the coefficient for the predicted values will be insignificant when the predicted values for the original theory are added to the second theory. If, however, the latter does not encompass the former theory, then the predicted values of activism for the former theory will still be significantly related to activism even when added to a model examining the latter.

The results of the encompassing tests are reported in Table 9 (we concentrate on the coefficients for the predicted activism values). The predicted value coefficients are all positive and significant in every case. This indicates that none of the models is encompassed by any of the others. Rather than being rival theories, each throws some extra light on a different aspect of the same phenomenon. They all have a part to play. Understanding civic engagement, therefore, requires all three theories.

Table 9.  : Comparing the Civic Activism Models: Encompassing Tests
General incentivesSocial capitalCivic voluntarism
  1. Note: *Significant at P=0.05; **significant at P=0.01.

Total activism models
 Predicted general incentives 0.51**0.54**
 Predicted social capital0.59** 0.62**
 Predicted civic voluntarism0.82**0.04 
Individualistic activism models
 Predicted general incentives 0.51**0.51**
 Predicted social capital0.61** 0.64**
 Predicted civic voluntarism0.79**0.77** 
Contact activism models
 Predicted general incentives 0.71**0.74**
 Predicted social capital0.73** 0.75**
 Predicted civic voluntarism0.79**0.79** 
Collective activism models
 Predicted general incentives 0.64**0.68**
 Predicted social capital0.74** 0.73**
 Predicted civic voluntarism0.92**0.95** 

We can summarise the analysis, therefore, by constructing a parsimonious reduced-form model for each type of civic activism, based on variables from all three theories (Table 10). Only those independent variables that were significant when all relevant variables were entered into a multiple regression were included in the final models. As the encompassing tests suggest should be the case, variables from each of the perspectives are significant in each type of activism. From the ‘general incentives’ perspective, perceptions of collective benefits and efficacy feature in most types (except collective): as before, the greater people's perceptions that collective benefits might accrue, the less likely they are to act; but the greater their perceptions of efficacy, the more likely they are to become civically active. Selective benefits fared less well, but did play some part: outcome benefits were positively related to total activism, and group benefits were positively related to total activism but negatively related to contact activism. Perceived levels of system benefits, though, were positively related to individualistic and collective activism. The main surprise for the ‘general incentives’ approach in these final models, however, was how little costs matter. They are negatively related only to individualistic activism. Once other factors are taken into account, it seems people pay attention to collective benefits and to judgements of how effective action is likely to be. And they pay some, though less consistent, attention to selective benefits. But they do not, by and large, worry about the costs of action.

Table 10.  : The Final Models – OLS Regressions
VariableType of civic activism
  1. Note: *Significant at P=0.05; **significant at P=0.01. OLS, ordinary least squares.

Collective benefits0.25**0.06**0.06** 
Selective outcome benefits0.06*   
Selective group benefits0.17** −0.03* 
System benefits 0.05** 0.04**
Group membership0.38**0.11**0.09*−0.05**
Informal networks0.39** 0.090.08
Hours of television watched daily −0.02**  
Years lived at current address  −0.05* 
Family nearby  0.10* 
Household income 0.06**−0.05** 
Strength of party identification0.16**   
Political engagement0.17**0.05**0.02**0.01*
Age −0.01**0.01* 
Gender (comparison=male)
Ethnicity (comparison=white)
 Other ethnic groups−0.40*−0.22  
Years in education0.11**  0.06**
Social class0.02 0.04** 

Some elements of the social capital model emerge well from this exercise. Notably, voluntary activity, as measured by group membership, has the predicted effect, being positively related to all measures of civic engagement. And being part of an informal network is associated with greater levels of total civic activism. But television-watching still discourages individualistic activism, although it is not associated with any other measure of civic activism. Television viewers are not, pace Putnam, particularly disengaged from civic life. Embeddedness in the local community, though, is related only to contact activism, and in inconsistent ways: the longer people have lived at a particular address, the less likely they are to contact officials; but those with families in the neighbourhood are more likely to contact officials than are those with no family nearby. But despite the emphasis placed upon it in much of the social capital literature, trust does not feature in any of these models.

Two elements of the civic voluntarism model stand out in all four models. Mobilisation remains very important. Being asked to participate makes it more likely people will participate. And political engagement matters too: the more interested people are in politics, the more likely they are to engage in all types of civic activism. Other aspects of civic voluntarism are less consistent in their effects. Household income is positively related to levels of individualistic activism but negatively related to contact activism (perhaps because poorer citizens need to contact state bureaucracies more frequently than their more affluent fellows). Surprisingly, given how often they feature in most discussions of political activism, class and education were significant in less than half of the models: the more educated people are, the more total and collective activism they engaged in. But education had no net impact on either individualistic or contact activism, once other factors were taken into account. And class is related only to contact activism: the more middle class people are, the more likely they are to contact officials.


Civic engagement is a diverse phenomenon. There are three distinct types of civic activism: individualistic-based activities, contacts with those in authority and collective action. Citizens who engage in activities encompassed by one of these types need not be engaged in those linked to other types. But it is also noteworthy that most citizens do seem to engage in some sort of civic activism, even though most are relatively low-cost actions.

In addition, civic engagement is multi-causal. No single explanatory framework fully accounts for it. All those we looked at here play some part in influencing levels of activism. Access to resources, positive evaluations of the benefits of involvement, involvement in associational life and informal networks (though not trust; Newton, 1999) and mobilisation all seem to be important correlates of most types of civic activism.

This implies that encouraging civic engagement among Britain's citizens at the start of the twenty-first century will require several things. People need to be persuaded of the existence of benefits emerging from involvement, and they need to be able to see that their actions are likely to have an effect, at least some of the time (as indicated by the ‘general incentives’ efficacy variable). However, perceived costs of involvement generally do not seem to be a concern: efforts to reduce those costs are unlikely to have much effect. As Putnam suggests, it seems that encouraging people to join clubs, societies and voluntary organisations helps engender other forms of civic action, but this does not work through trust. However, these are all difficult to legislate for. More amenable to political intervention by governments, though still difficult to enact, are some of the key findings to emerge from the civic voluntarism perspective. Crucially, raising people's interest in politics generally makes them more active citizens. And there is no substitute for mobilisation. Left to their own devices, people are not particularly likely to become active citizens. But when they are asked to participate (especially by those close to them), they are very likely to do so. Analysts of party campaigning have rediscovered the importance of the constituency campaign in recent years (for example Pattie et al., 1995; Whiteley and Seyd, 1998). Asking voters for their support pays electoral dividends. It seems that the same is true of other forms of civic action. Exhortations to virtue are unlikely to succeed. Invitations to partipate in (specific) activities may well be more successful. If you don’t ask, you don’t get.

About the Authors

Charles Pattie, Department of Geography, University of Sheffield, Elmfield, Winter Street Sheffield S10 2TN, UK; email:

Patrick Seyd, Department of Politics, University of Sheffield, Elmfield, Northumberland Road, Sheffield S10 2TU, UK; email:

Paul Whiteley, Department of Government, University of Essex, Wivenhoe Park, Colchester C04 3SQ, UK; email:


  • 1

    The Citizen Audit survey was funded by a grant from the Economic and Social Research Council (ESRC award no L215252025) of its Democracy and Participation programme. We are grateful to the ESRC for its support.

  • 1

    Theoretically, we could include a multiplicative interaction between efficacy and collective benefits in the model, on the grounds that people who feel efficacious and who perceive large collective benefits should be particularly likely to become involved. The specification would therefore be


    However, experimentation with alternative model specifications suggests the additive form reported here performs best.