Introduction
Political scientists agree that informed political participation is the foundation of democracy. It empowers the citizen and ensures checks on and balance of the authority and power of the state and other entities. Turning out to vote is the most common and easily measured form of citizen political participation in a democracy. In fact, all sides of the ideological spectrum agree that voting is the symbolic essence of the democratic government and responsible citizenship. It is an indication of a nation's political health (Mishler, Reference Mishler1979; Miller, Reference Miller1980). Voter turnout is so fundamental that political scientists have been alarmed by lack of political attitude consistency among voters, the non-voting of a large segment of the population, and not having enough knowledge or a very good reason to vote among those who do (e.g., Johnston, Reference Johnston2001; Pattie and Johnston, Reference Pattie and Johnston2003). It is not surprising, then, that voter turnout is considered “a major theoretical puzzle” (Aldrich, Reference Aldrich1993).
Scholars and politicians have also been alarmed by the growing citizen disenchantment with government and politics in many Western democracies (Pharr et al., Reference Pharr, Putnam and Dalton2000). Some have referred to such discontent as a “crisis” (Norris, Reference Norris and Norris1999: 257) or “sickness” (Johnston, Reference Johnston2001: 4) of electoral democracies, as evidenced by the reluctance of voters to play the game of turning out to vote. In Canada, turnout steadily declined from 75 per cent in the 1984 election to 61 per cent in the 2000 election. The 2004 election showed an increase in voter turnout (64.9%). The decline in turnout is not specific to Canada. A similar trend is also apparent in other countries, such as the US and the UK (Howe and Northrup, Reference Howe and Northrup2000; Blais et al., Reference Blais, Gidengil, Nadeau and Nevitte2002; Pattie and Johnston, Reference Pattie and Johnston2003).
Despite the importance of turnout for the health of democracies, we do not have a clear picture as to why citizens turn out (or do not) to vote. We are less able to describe and/or explain variations in voting among various groups. Moreover, past research often focussed on socio-demographics and the psychological underpinnings of voting. Less attention was paid to social contexts and social relationships. Models of voting turnout that ignore social context and social relations, and only emphasize individual characteristics, have low predictive power (Aldrich and Simon, Reference Aldrich, Simon and Long1986: 277–278), fail to specify the voting process (Duncan and Stenbeck, Reference Duncan, Stenbeck and Cogg1988), and result in biased estimates (Straits, Reference Straits1990) and premature conclusions.
Thus, this paper has three objectives: a) highlight the importance of social relations or social capital; b) take into account sources of social capital, socio-demographics and socio-economic forces; and c) empirically investigate voting turnout in federal, provincial and municipal elections in Canada.
Roles of Social Capital and Civic Engagement
Previous research on voter turnout highlighted the importance of the socio-economic makeup of the voters, generational differences in voting, legal restrictions on voting, and political contexts of elections (Miller, Reference Miller1992; Pattie and Johnston, Reference Pattie and Johnston1998; Lyons and Alexander, Reference Lyons and Alexander2000; Blais et al., Reference Blais, Gidengil, Nevitte and Nadeau2004). Less attention was given to the features of social life that help produce social connections, benefitting both individuals and communities that share similar objectives. Social features such as norms of reciprocity, mutual obligations, trust and social networks help to build tolerance of alternative views and commitment to democratic values. Such features tend to foster attitudes, values and orientations that encourage political participation. As well, they tend to provide opportunities and organize an individual's interests in the political process, and to enhance citizen involvement in politics through political information, friendly pressures and influence.
The importance of social capital for democracy was highlighted by Putnam (Reference Putnam1993), whose work showed how the effectiveness of Italian Northern and Southern regional governments hinged on patterns of social networks and informal associations. Southern Italy was said to lack a tradition of horizontal associations and norms of reciprocity and trust, while Northern Italy, with its stronger civic community, including association membership, represented a higher democratic institutional performance. In his more recent work, Putnam laments the decline of league bowling in the United States and sees such a decline as part of a general civic disengagement. Others have taken his lead and evaluated the importance of social capital in the domain of politics (Jackman and Miller, Reference Jackman and Miller1998; Hemingway, Reference Hemingway1999; Sullivan and Transue, Reference Sullivan and Transue1999; Dolfsma and Dannreuther, Reference Dolfsma and Dannreuther2003).
Although named differently, the importance of social capital for politics was recognized long ago. Almond and Verba (Reference Almond and Verba1963) emphasized the importance of social capital when they argued that the development of effective and stable government depends significantly upon the orientation that people have to the political process. For them, the culture of trust and tolerance, and norms of social responsibility, ensure political cooperation and action. Subsequent research also showed that membership in voluntary participation stimulates political participation (Wolfinger and Rosenstone, Reference Wolfinger and Rosenstone1980; Verba et al., Reference Verba, Schlozman and Brady1995). Nonvoters are revealed to be less likely to live in the same community for a long time and/or belong to organizations or associations. They are socially isolated and are less likely to believe in their own abilities to exert influence (Eagles and Erfle, Reference Eagles and Erfle1989; Straits, Reference Straits1990; Sabucedo and Cramer, Reference Sabucedo and Cramer1991; Ragsdale and Rusk, Reference Ragsdale and Rusk1993; Pattie and Johnston, Reference Pattie and Johnston1998). Timpone's (Reference Timpone1998) study of the 1992 American National Election Studies showed that social connectedness, measured by knowing neighbours, volunteering, church-related charity and attempting to solve community problems, is related to turnout after controlling for demographic variables and political attitudes. Likewise, in Canada, Saxena (Reference Saxena2003) established that those who voted in 2000 are more likely than non-voters to hold membership in a variety of organizations. However, Milner's (Reference Milner2001) multi-nation study revealed little evidence that people in high-trust democratic societies, aggregated at the national level, are more likely to turn out to vote than those in low-trust societies. The replication of results for municipal elections showed a similar pattern. The relationship between membership in voluntary groups and more active forms of political participation at the individual level was also found to be insignificant (Verba et al., Reference Verba, Schlozman and Brady1995).
Putnam (Reference Putnam2000: 218) also argued that newspaper readership is a “mark of substantial civic engagement.” He pointed out that newspaper readership, organizational membership and associational involvement are hallmarks of successful Italian regions (Reference Putnam1995a: 66). Similarly, Milner (Reference Milner2001) pointed to the importance of civic literacy, as measured by newspaper readership, as the link between social capital and voting. Milner's (Reference Milner2001) multi-nation study revealed significant relationships between both illiteracy rates and daily newspaper readership with a tendency to vote in municipal elections. This is an important consideration for political scientists, since the democratic process is argued to be effective if voters are knowledgeable and informed. However, further analysis by Pattie and Johnston (Reference Pattie and Johnston2003) showed that such a relationship was insignificant when appropriate controls were included in the model. It may be suggested that the lack of civic literacy effect in Pattie and Johnston's model is due to their measure, “frequency of newspaper readership.” People read newspapers for a variety of reasons—some may not pay attention to current affairs, which is the underlying theme of civic literacy. In general, it is understood that social capital lubricates and facilitates the political process. Trust and reciprocity develop the individual's sense of fairness and efficacy. Voluntarism and cooperation enhance an individual's sense of belonging and civic responsibility. Similarly, participation in associations and clubs creates fertile ground for political socialization and participation. Thus, social capital tends to stimulate turnout by providing channels for political information, resulting in social pressure and persuasive arguments, shared responsibility, cooperation and friendly competition.Footnote 1 Nevertheless, there are empirical disagreements on whether or not social capital, or some aspects of it, aid with electoral participation.
Moreover, it is unclear as to whether social capital affects turnout independent of socio-economic and demographic forces. Some view social capital as a “precondition” to socio-economic inequalities (Putnam Reference Putnam1995a: 66). Granovetter (Reference Granovetter1995) showed that access to information channels accounts for a significant portion of the socio-economic gap between black and white people in the US. Others see social capital as unequally distributed among various groups. Tilly (Reference Tilly1998) argued that the position occupied by an individual in the social structure results in differential access to social capital. Uslaner (Reference Uslaner2002) showed that economic inequality is directly and significantly related to level of trust in the United States. Wuthnow (Reference Wuthnow1997) pointed to a higher decline in associational involvement among the marginalized section of the population.
Similarly, it is unclear whether social capital complements or competes with socio-economic status as the explanation of electoral participation. Education, occupation and income have been standard predictors of turnout (Heath et al., Reference Heath, Jowell and Curtice1991; Heckelman, Reference Heckelman1997; Saxena, Reference Saxena2003; but see Pattie and Johnston, Reference Pattie and Johnston1998). Social status is shown to structure political opportunities and help develop political attitudes that motivate participation. Individuals of higher socio-economic status are more likely to have a stake in the political system, to be gatekeepers and opinion leaders, and to influence the system. Traditionally, education was used as an indicator of socio-economic status. However, it is also very important in political decision making. Education levels affect literacy and political knowledge, both of which are highly related to voting (Verba et al., Reference Verba, Schlozman and Brady1995; Pattie and Johnston, Reference Pattie and Johnston2001; but see Gray and Caul, Reference Gray and Caul2000). Higher education increases one's capacity to understand complex political issues (Straits, Reference Straits1990), offers access to political awareness and information, contributes to an appreciation of the relevance of political decisions and encourages civic responsibility. The general wisdom is that those with higher education, better jobs and more income have a higher propensity to vote than their counterparts (see Pattie and Johnston, Reference Pattie and Johnston1998, Reference Pattie and Johnston2001, Reference Pattie and Johnston2003; Howe and Northrup, Reference Howe and Northrup2000).
In sum, social capital describes the extent to which actors participate in the life of the community and may be correlated with or predict other types of activity, such as voting. However, past empirical evidence does not seem to agree. The relative importance of social capital compared to other standard predictors of voting turnout is also not established. Therefore, it is worth investigating how electoral participation is related to individuals' stock of social capital. Are those active in civic society also active in the electoral process? Does social capital affect turnout independent of socio-economic status? Are social capital effects consistent across three levels of Canadian government?
Data
The data source is the National Survey of Giving, Volunteering and Participation (NSGVP; 2001), a public-use microdata file administered by Statistics Canada. The NSGVP is a stratified and multistage survey with unequal probabilities of selection. Therefore, survey weights are used so that estimates are unbiased (Statistics Canada, 2001: 120). Moreover, since only Canadian citizens can vote in an election and the public-use file does not distinguish between citizens and others, those born outside Canada are excluded from the analysis. Finally, the public-use file combined the 15-to-24 age group into one category. Since those under 18 years of age are unable to vote, the 15-to-24 age group is also excluded from the analysis. This age restriction is important because it assures that respondents had the opportunity to complete their schooling. The final sample for this analysis includes 9,314 individuals who were 25 years of age and older at the time of the survey. Although only those who can vote are included in the analysis, the age effect may be underestimated, since the younger age group, the generations X and post-X, are shown to have a lower tendency to vote (for Canada see Blais et al., Reference Blais, Gidengil, Nevitte and Nadeau2004).
Measurement
Past research in Canada has had a tendency to focus on only one level of government (see Kushner et al., Reference Kushner, Siegel and Stanwick1997). This study focuses on three levels of government (municipal, provincial and federal). Miller argues that there are significant differences between local and national elections: for example, local elections are less likely to be distorted by “emotion-laden television campaigns” (2001: 88). The NSGVP asked respondents the following question: Did you vote in the last federal, provincial and municipal elections? Each individual's response is coded as yes = 1 and no = 0. While there is a tendency for self-report surveys to overestimate voting when compared to actual turnout figures (see Miller, 2001: 86; Blais et al., Reference Blais, Gidengil, Nevitte and Nadeau2004), there is little evidence that such overestimation is differentially distributed among various groups.
As I alluded to above, social capital is a concept that can include many things, resulting in conceptual confusion and absence of a consensus on how to measure it. In Bowling Alone, Putnam (Reference Putnam2000) conceived social capital as a property of the community and social groups. However, he saw it originate in individual subjectivity (e.g., values). According to Putnam, social capital is both a private and public good. As a private good, the individual's connections and ties may result in benefits for that individual or others close to him/her. As a public good, a poorly connected individual may benefit from a well-connected community. Thus, some of the benefits from investment in social capital may go to bystanders (Putnam, Reference Putnam2000: 20). Similarly, Coleman (Reference Coleman1988) viewed social capital as being rooted in the individual. Not only is there disagreement on the unit of analysis but also on the concept itself (see Bourdieu, Reference Bourdieu and Richardson1983; Coleman, Reference Coleman1988, Reference Coleman1990; Portes, Reference Portes1988; Wall et al., Reference Wall, Ferrazzi and Schryer1988; Putnam, Reference Putnam1993, Reference Putnam1995a, Reference Putnam2000; Fukuyama, Reference Fukuyama1995). Most see social capital results from memberships in dense networks, neighbourhoods, workplaces, organizations, associations and kin groups. These memberships are fertile grounds for development of obligations, expectations, trust, information channels, norms, sanctions and authority relations that constrain and/or facilitate specific behaviours. Among these, social networks, ties, organizational membership, voluntarism and giving are seen as the core of the concept of social capital (see Putnam, Reference Putnam2000: 19, 117, 134; Coleman, Reference Coleman1988: S98).
In this paper, I use social networks, religious participation, giving, voluntarism and frequency of giving as measures of social capital. Social networks are measured by asking: “How often do you socialize … with parents or other relatives, with friends who live outside the neighbourhood, with friends participating in sports or recreation activities, or watching family members participate in sports or recreation activities?” For each of these questions, respondents could choose from the following options: every week = 3, a few times a month = 2, a few times a year = 1, and not at all = 0. These variables are summed with a minimum of 0 and maximum of 12. Principal Component Analysis was used to confirm unidimensionality of multi-item measures such as this. Religious service attendance is measured by the question: “In the past 12 months, other than on special occasions (such as weddings, funerals or baptisms), how often have you attended religious services or meetings?” (The choices are least once a week = 5, at least once a month = 4, at least 3 or 4 times = 3, only once or twice = 2, and not at all = 1.Footnote 2) The data file provides a variable that distinguishes volunteers and givers from volunteers but not givers, from givers but not volunteers and from those who have done neither (reference category). Another variable measures the frequency of giving/donating, starting at 1 and truncated at 11.
Civic engagement is distinct but related to social capital and is measured by attentiveness to the news. Putnam (Reference Putnam1995a: 6) states that each hour reading a newspaper is associated with more social trust and more group membership. This variable is indicative of voters' level of political and/or social sophistication. It provides the necessary social and political knowledge to make an informed decision and thus is the link between social capital and voting (Milner, Reference Milner, Dowding, Hughes and Margetts2001b). The focus is on how often citizens pay attention to current affairs and not so much what type of current affairs they pay attention to. Certain types of news, such as political corruption, may, in fact, increase the sense of distrust and reduce voting. Respondents were asked the frequency with which they follow news and current affairs (international, national, regional or local). Choices included rarely or never = 1, several times each month = 2, several times each week = 3 and daily = 4.
Social capital is not distributed equally among all groups and individuals. Individuals need some level of residential stability in order to build on their stock of social capital. Putnam (Reference Putnam2000) considers community rootedness as the foundation of social capital. Community rootedness is measured by length of time living in a community: two years and less = 1, 3 to 5 years = 2, 6 to ten years = 3, more than ten years = 4. Similarly, marriage is understood as a site of “thick trust” (Putnam, Reference Putnam2000) and is measured by marital status: married = 1 and others = 0.
Moreover, middle- and upper-class individuals, or those belonging to the dominant groups, are more likely to develop diverse networks and friends and are better able to benefit from such relationships. Likewise, the information channels available to these groups differ. As an example, research has pointed to the importance of education in inculcating civic culture (Helliwel and Putnam, Reference Helliwel and Putnam1999). Educated people are more likely to join groups and to trust others, partly because they are better off economically but also because of the skills and resources that were imparted to them at home and at school (Putnam, Reference Putnam1995a: 7). More educated individuals get involved in their communities, vote, pay taxes give to charities and volunteer their time (Uslaner, Reference Uslaner2002: 18–19).
Education is dummy-coded into: less than high school = (reference category), graduated from high school, some post-secondary, post-secondary diploma and university degree. The data file is limited in its classification of occupations. Occupation is coded into white-collar, sales, blue-collar (reference category) and not in the paid labour force. Household income includes five categories, from less than $20,000 (reference category) to $100,000 and more.
In addition to social capital, its sources and socio-economic status, “post-materialist values” (Inglehart, Reference Inglehart1997) and socio-demographic variables also play important roles in building social capital and/or turning out to vote. That is, such variables either have a direct relationship to political participation or the relationship is mediated through social capital (Sullivan and Transue, Reference Sullivan and Transue1999; Veenstra, Reference Veenstra2002). Inglehart (Reference Inglehart1997) emphasized the importance of self-expression and quality of life for development and maintenance of democracy. Such values ensure a subjective sense of well-being and allow for cognitive mobilization, essential to informed political participation. However, it is unclear whether or not these post-materialist values have an independent effect on turnout when social capital and other forces are taken into account. Here, I use sense of control and life satisfaction. Sense of control is measured by the following question: “How much control do you have in making decisions that affect your everyday activities?” The choices are no control = 1, control over few or some decisions = 2, control over most decisions = 3, control over all decisions = 4. Life satisfaction is measured by asking respondents: “How satisfied are you with your life in general?” The choices are very satisfied = 4, somewhat satisfied = 3, somewhat dissatisfied = 2 and very dissatisfied = 1.
In Canada, it has been said that regionalism is one of the preeminent facts of Canadian politics (Mishler, Reference Mishler1979). Provinces differ in economic development, wealth, power and other resources, which in turn affect the stock of social capital and political participation. Canadian regions are divided into East (reference category), Quebec, Ontario, West and BC.
Evidence also suggests that elderly, pre-boom and boom generations, Protestants and Catholics are more likely to vote than their counterparts. Similarly, women are somewhat less likely to vote than men, though the difference may have diminished (Mishler, Reference Mishler1979; Howe and Northrup, Reference Howe and Northrup2000; Blais et al., Reference Blais, Gidengil, Nadeau and Nevitte2002, Reference Blais, Gidengil, Nevitte and Nadeau2004; Pammett and LeDuc, Reference Pammett and LeDuc2003; Saxena, Reference Saxena2003). Although life cycle, period and particularly generation are all important for voting (see Miller, Reference Miller1992; Blais et al., Reference Blais, Gidengil, Nevitte and Nadeau2004), this study is based on cross-sectional data and thus is able to control only for age. Age has five categories ranging from 25 to 34 years of age to 65 years and over. Religious affiliation is shown to be related to voluntarism and voting (Archer, Reference Archer1987). Lipset (Reference Lipset1990) pointed to the importance of Protestantism in providing for the needs of the community. Mishler (Reference Mishler1979) showed that Protestants are more likely to vote and Catholics are more involved in political campaigns. Religious affiliations include Catholic, Protestant, other religions and no religion (reference category). Research also points to the importance of ethnicity for voting (see Chui et al., Reference Chui, Curtis and Lambert1991; Lapp, Reference Lapp1999). Unfortunately, public-use microdata precluded measurement of ethnicity.
Analysis without controls
Table 1 shows the percentage of respondents voting in municipal, provincial and federal elections in Canada for each category of variables. Note that some of the measures are recoded into more manageable categories for the purpose of this table only. Consistent with previous research, it is apparent that respondents are continually less likely to vote in municipal than provincial or federal elections (see Kushner et al., Reference Kushner, Siegel and Stanwick1997). Among the key sources of social capital, community rootedness is an important predictor of voting. It has the potential to increase electoral participation in federal and provincial elections by about 18 and 20 per cent, respectively, and by 32 per cent in municipal elections. Marriage also increases electoral participation by 6 to 12 per cent. More importantly, social capital measures show a clear linear relationship with voting (e.g., religious participation, voluntarism and giving, frequency of giving). Those who participate in religious services at least once a week are about 13, 14 and 20 per cent more likely to vote in federal, provincial and municipal elections, respectively, than those who do not participate. The difference in voting among those who frequently give to the community and those who do not at all is from 26 to 30 per cent for the three elections. Similarly, the difference between those who give and volunteer and those who do not give at all and/or do not volunteer is around 25 per cent in these elections. The social network effect is, however, weak. Civic engagement is also a strong predictor of voting. Those who pay attention to current affairs on a daily basis vote between 24 to 30 per cent more than those who rarely pay attention to current affairs.
Table 1 Frequency Distribution of Voting Turnout for All Variables

Two measures of post-materialist values score in the expected direction. Sense of control increases the likelihood of voting by 9 to 18 percentage points, depending on the level of government. Similarly, life satisfaction increases the electoral vote by 9 to 16 per cent.
Among socio-economic measures, household income tends to be more related to voting than education and occupation. Those in higher-income households are about 10 per cent more likely to vote than those in the marginalized income category.Footnote 3 Individuals in white-collar occupations and those not working are about 10 per cent more likely to participate than those in sales and blue-collar occupations. The effect of education is somewhat weaker than that of other social inequality measures; the main dividing line seems to be between those with university degrees and higher and those with less education.
Among the socio-demographic measures, age is an important predictor of voting. Age increases turnout by about 16 to 18 per cent for provincial and federal elections and by 29 per cent in municipal elections. This evidence is consistent with Pammett and LeDuc's (Reference Pammett and LeDuc2003) study of the federal election in Canada, which indicated a large generational or age gap in turnout in the 2000 election. Catholics and Protestants are 10 to 18 per cent more likely to vote than others, depending on the level of government. There is little gender difference in voting. Finally, as one moves from East to West, the tendency to vote decreases.
In sum, most social capital measures, civic engagement, community rootedness and age are strongly associated with turnout. Other variables are either not related to turnout or the relationships are weaker than those of social capital measures.
Analysis with control
In order to evaluate the relationship between the various measures and voter turnout, I first created an index of turnout by summing up the response to each of three elections. This index of voting frequency has four categories, ranging from zero (not voting in any election) to three (voting in all three elections). Regression analysis is used in order to analyze this variable (see Table 2). Second, I evaluated participation in each of the three levels of government individually, using logistic regression analysis (see Table 3).
Table 2 Unstandardized and Standardized Regression Coefficients of Voting Turnout and Predictors

a: p < .001, b: p < .01, c: p < .05
Table 3 Logistic Regression of Voting Turnout by Predictors

a: p < .001, b: p < .01, c: p < .05
Table 2 shows standardized and unstandardized regression coefficients for the relationship between voter turnout and predictors. Variables are entered into the equation as a group in several models, allowing us to observe what happens to a set of variables as we include more variables in the equation. Tests for model 5 revealed that the assumptions of the model were not violated; in particular, the collinearity across predictors was always at acceptable levels, as measured by Variance Inflation Factors (VIF), which ranged from 1.1 to 3.5. Similarly, the tolerance level was never below .29. According to Cohen et al. (Reference Cohen, Cohen, West and Aiken2003), these figures are all acceptable and therefore multicollinearity is not a problem for this study.
In model 1, socio-demographic measures, particularly age, are shown to be related to turnout, accounting for 6.2 per cent of the explained variance. Inclusion of socio-economic measures (model 2) increases the explained variance to 10.1 per cent. Although the addition of new variables to each model changes and generally decreases the coefficients in the previous models, basically the same socio-economic and demographic variables are statistically related to turnout. Given that all of these variables, with the exception of age, are categorical, they can be directly compared, thus pointing to the strength of household income and religious denomination. For example, those with household incomes of $60,000 to $100,000 are more likely to vote, by a factor of .494, than those with less than $20,000. This amounts to movement just over one-eighth of the length of the voting index, or about a 13 per cent change in voting. Similarly, the households in the category of $100,000 or more move about one-ninth of the length of the voting index, or about 11 per cent, compared to the lowest income group. However, the age effect is the strongest among both socio-economic and socio-demographic forces. Age has five categories. Thus, moving from the youngest to the oldest age cohort increases voting by about one-third of the index length. The age effect would have been even larger if I was able to include the younger age group in the analysis.
Next, I introduced key sources of social capital. Both community rootedness and marriage are significantly related to voting. Although community rootedness shows a strong effect similar to age in model 3, its effect is lower than that of age. It has four categories, while age has five. These two measures, as expected, help reduce the effects of socio-economic and demographic variables, particularly income and age. This suggests that older individuals, as well as those in higher-income households, are more likely to vote, partly because they are more likely to have a stable community residency and partly because they are married. Community rootedness and marital status together increase the explained variance by another 2.6 per cent over the 10.1 per cent explained by both socio-economic and demographic variables.
Model 4 includes social capital measures, all of which show independent and statistically significant relationships with voting. They increase the explained variance to 16.3 per cent and to 17.7 per cent when the civic engagement and post-materialist measures are included in the model (compare models 4 and 5).Footnote 4 The largest change in explained variance emerges when measures of norms of generalized reciprocity (volunteerism and giving) are included in the model. Inclusion of social capital measures also reduces the importance of socio-economic and demographic variables for turnout. In fact, the wealthy higher-income household ($100,000 and up) is now not statistically different from the marginalized lower-income groups ($20,000 and less). Similarly, Protestants and those with high school diplomas and/or post-secondary education are no longer statistically different from the non-religious and lowest education groups, respectively. The age effect also declines somewhat. These examples are clear indications that social capital mediates the relationship between socio-economic and demographic variables and voter turnout. Stated differently, part of the reason for the higher levels of political activity among those who belong to higher socio-economic status groups (particularly those in higher education and income) or follow the dominant religions (Protestant and Catholic) is that they have a higher stock of social capital (see models 3 to 4 and 5). Also note that socio-economic measures are sensitive to attentiveness to current affairs. This evidence is supportive of Downs' (Reference Downs1957) rational voter model in that the cost of information (monetary or time) tends to disenfranchise citizens. But more importantly, introduction of this variable results in an insignificant relationship between education and voting, pointing to the importance of literacy for electoral information. The difference between those who rarely pay attention to current affairs and those who do so on a daily basis is about .620, or about a 20 per cent change in the index of voting. Finally, the introduction of this variable slightly reduces the importance of volunteering and giving on turnout. This suggests that those who give and volunteer are also more attentive to current affairs. They are civic-minded individuals.
These models demonstrate the importance of volunteering and giving in explaining voter turnout. Other important variables are community rootedness, age and attentiveness to news. We should remember that community rootedness has four categories, age has five categories and attentiveness to news has four categories. Thus, they have substantial potential effects on voting. For example, the difference between those with less than two years of residence in a community and those with more than ten years is .417, or more than a 10 per cent change in the index of voting.
An analysis of standardized coefficients for the final model reveals the importance of volunteering and giving (beta = .192) followed by age (beta = .170), giving but not volunteering (beta = .165), community rootedness (beta = .138), and attentiveness to news (beta = .114). The next highest standardized coefficient in this model is for white-collar occupations, at .070. These coefficients, therefore, point to the importance of social capital and civic engagement, as well as the contexts that help develop the stock of social capital, as a means to motivate citizens to vote.
Figure 1 substantiates the importance of social capital for voter turnout. Overall, social capital measures have a stronger effect than socio-demographic or socio-economic measures alone. Thus, the proportion of variance explained by social capital measures is 8.1 per cent. It is substantially more than the socio-demographics at 6.2 per cent and more so than the socio-economic measures at 3.1 per cent. Similarly, the effects of community rootedness at 4.7 per cent, and both sources of social capital at 5.8 per cent, are stronger than those of education, occupation and income combined.Footnote 5

Figure 1. Proportion of Variance
Table 3 shows the log odds of voter turnout for each level of government. These estimates (b coefficients) show us the differences in the log odds on voting for those in a category compared to the reference category for each predictor (Aldrich and Nelson, Reference Aldrich and Nelson1984). Since the log odds may have little intuitive meaning, the exponentiated coefficients are also presented. The exponential of a coefficient is the factor by which the unlogged odds on vote are multiplied for one unit of change in the predictor variable (e.g., gender). For interval and ratio variables, one needs to exponentiate the coefficient and then take it to the power of the desired category.
The table shows that the effects of social capital measures on turnout are more consistent than those of socio-economic statuses and demographics. Religious participation, volunteering and giving or giving without volunteering and frequency of giving are all consistent correlates of voting turnout at all levels of government. As an example, those who volunteer and give have 1.814, 2.561 and 2.815 times higher odds of voting in municipal, provincial and federal elections, respectively, than those who do not volunteer or give. Each unit increase in number of giving/donating increases the odds of voting in municipal, provincial and federal elections exponentially by 1.063, 1.069 and 1.082, respectively. Note that this variable has 12 categories. Similarly, community rootedness as the social context in which social capital can be developed has a consistent effect on turnout. Each unit increase in community rootedness increases the odds of turnout exponentially by 1.422 at the municipal level, by 1.252 at the provincial level and by 1.254 at the federal level. The effect of marital status is specific to provincial and particularly federal turnout.
Among socio-economic predictors, only white-collar occupations and a household income of $60,000 to $100,000 consistently influence turnout at all three levels. Among socio-demographic predictors, age shows a consistently strong effect across the three different levels of government. Each additional increase in age category exponentially increases the odds of voting by about 1.206 to 1.56 times. Regions to the west of Quebec turn out to vote significantly less than those to the east of Ontario.
Discussion and Conclusion
In the mid-nineteenth century, de Tocqueville (Reference Tocqueville1945) discussed three important forces that help maintain American democracy: laws; situation and context; and manners and customs of people. In the Canadian context, I have provided support for his emphasis on the latter two forces. “Carriers of creed,” who are the engines of the democratic system, are more likely to volunteer, to give, to have social networks and to participate in religious services. Such community and cooperative norms, values and behaviours entail a sense of social responsibility and political participation that is rarely matched by other forces.
In addition, the civic responsibility of being abreast of current affairs is also fundamentally important for an informed and rational voting process. This is consistent with Milner's argument on the importance of civic responsibility. Even though Milner (Reference Milner2001a) showed that newspaper readership was the most important aspect of civic responsibility, we have argued that newspaper readership may not necessarily include knowledge of current affairs. It is the attentiveness to current affairs in newspaper readership or other information media that seems to be important to turnout. The civic literacy argument points to the importance of citizens' knowledge of the political environment. Such individuals are well informed about the government and political issues and follow political news consistently. However, to read the newspaper does not mean that one reads news or pays attention to politics. One may read only the sports or the entertainment section of the newspaper, not paying attention to current affairs. The measure used here, on the other hand, is not about newspaper readership, but specifically about attention to current affairs, which could come from newspapers, magazines, TV or other sources. Therefore, the analysis confirms the relationship between civic literacy and electoral participation (but see Pattie and Johnston, Reference Pattie and Johnston2003). The effect of news attentiveness is also testimony to the rational and informed participation of voters. The health of democratic societies is based not just on voters' participation but, more importantly, is predicated on substantial information. Attentiveness to current affairs ensures such information.
To some extent, social capital and civic engagement are nurtured in specific social contexts. Community rootedness shows a strong relationship with electoral turnout at all levels of government and particularly in the municipal election. Putnam (Reference Putnam1993, Reference Putnam2000) uses the analogy of frequent repotting of plants to refer to residential instability. Such repotting tends to disrupt the root system, which delays an individual's attempt to put down new roots. Those living in the same community for a greater length of time are more likely to own their own home, be involved in the affairs of their community in general, and the immediate community in particular, and to establish networks and relationships. Such involvement increases the required information and social pressure for political involvement. As well, those who are residentially stable are more likely to be included in the voting registry. Pammett and LeDuc (Reference Pammett and LeDuc2003) revealed that a large percentage of non-voters reported that their names were not on the list, this percentage being higher among the younger age group than the older group. This may be understandable since those in the younger age group are more geographically mobile.
Similarly, married individuals are significantly more likely to vote than their counterparts. The marriage effect is consistent with Putnam's view as the site of “thick trust” and with Coleman's (Reference Coleman1990: 590–97) view that the family is a key institution through which social capital is transmitted. For example, family members ensure strong moral values and transmit social capital by articulating guidelines and communicating clear rules, thus prescribing actions that they view acceptable or unacceptable. Spouses establish trusting and supporting relationships that act as conduits through which individuals effectively conform to specific political values and behaviours. They pressure each other to vote, possibly with the reminder: “Don't forget to vote, otherwise you can't complain.”
Although a voter's decision to turn out is related to other status variables, their effects are not as consistent as those of social capital measures across three levels of government. Education's effect is specific to provincial elections and only for those with some post-secondary education. A consistent occupational effect is specific only to white-collar occupations. The consistent effect of income is also specific to those households in the $60,000 to 100,000 income bracket. The income effect for municipal election is, in fact, curvilinear. Except for age and BC, there emerged no consistent effect across the three elections for the socio-demographic variables, either. As an example, Ontario and the western provinces are less likely to vote in federal and provincial elections and more in municipal when compared to the eastern provinces. Quebec residents are significantly more active in provincial and municipal level elections, but do not significantly differ from the eastern provinces in federal electoral participation. Residents of BC, however, tend to vote consistently less than those of other provinces at all levels of government. This finding seems to be a peculiar effect of this survey. In the thirty-seventh general election (2000), BC had about the same voter turnout as did the eastern provinces and slightly higher than that in Ontario. It seems that BC residents are less likely than residents of other regions to over-report voting in the survey.
In general, the most consistent results can be placed within the rational choice theory. Social capital theory is predicated on the rational choice theory, which sees individuals as self-interested and as maximizers, whose social participation can, under certain circumstances, be an investment. Yet, in this theory, individuals are not seen as completely atomized, as in the rational choice theory. Nevertheless, the basic premise of rational choice, that actors are selfish and competitive, is consistent with social capital theory. Individuals learn from experience that cooperation has tangible economic, social and political benefits. Social capital theory, therefore, sees the actor as an active participant rather than a passive recipient of the political or other processes (Walters, Reference Walters2000). As well, sense of social responsibility, inherent in social capital theory, seems to counteract the self-interested elements of rational choice theory. That is, rational individuals are also motivated by their sense of civic duty, independent of short-run losses such as the cost of voting. Similarly, the effects of education, occupation, income and attentiveness to current affairs on voting are related to rational choice perspective. Those in upper statuses are more likely to have an interest in the system and invest more in the political process. Attentive individuals are also better able to calculate the expected utility of their action.
Canada, unlike Australia, Belgium and Greece, does not have a system of compulsory voting. In fact, Canadian public opinion polls show that the Canadian public is not supportive of such a system, either. Therefore, in order to deal with declining turnout, various other policy suggestions are made. These include having a fixed date for elections rather than allowing the government in power to hold an election whenever it wishes (within five years of office), increasing the power of parliamentary committees, ensuring regular referenda on major policy issues, or making individual votes more important by giving the parties with more votes more seats. Although these are important considerations, this paper suggests that instilling norms of social and civic responsibilities, cooperativeness and social obligations will have the expected outcome of an increase in turnout. Voluntarism and giving reinforce participatory norms and encourage generalized reciprocity, trust and cooperation, which are essential in solving community problems. Civic engagement, or attentiveness to current affairs, also increases knowledge and political interest. In other words, norms of civic responsibility, cooperativeness, social obligations and involvement, as well as attentiveness to socially relevant issues, will go a long way in improving the health of a “sick” (Johnston, Reference Johnston2001: 4) Canadian democracy. All of these factors are related to increased turnout, which is essential to the political vitality of democracy. In particular, a strong government advertising campaign that encourages citizens to volunteer, to give to charities and to pay attention to the social world around them could increase citizen turnout.
However, it is not clear whether or not social capital can be instilled in individuals late in the life course. Putnam (Reference Putnam1993) suggests that social capital is time-invariant. He claims that national differences in value orientations are stable and thus social capital measured in 1990 is an effective proxy for social capital measured in 1920. Similarly, Johnston and Soroka (Reference Johnston, Soroka, Dekker and Uslaner2001) view social capital as cultural values that are embedded in a group's psyche. The implication is that communities and individuals have more or less of civic culture and/or social capital, which present themselves in personal attitudes and behaviours and that remain constant across generations, and even oceans (Putnam, Reference Putnam1993; Fukuyama, Reference Fukuyama1995; Verba et al., Reference Verba, Schlozman and Brady1995).
Hanks and Ecklands (Reference Hanks and Ecklands1978: 481, 488) demonstrate that participation in extracurricular school activities has a relatively strong direct effect on participation in adult secondary association. This relationship is also shown to be significant independent of social class background, academic aptitude or group performance. In other words, adolescent extracurricular activities seem to be a training ground for adult organizational and associational involvements. Such experiences and personal histories catalogue significant events that condition citizens' current attitudes, thereby influencing both motivations and finally behaviour (also see Janoski and Wilson, Reference Janoski and Wilson1995). Therefore, it seems that social capital is and can be shaped during the school years, and by the time one exits the educational institutions, he or she is already a “carrier of the creed” who will become an active member of the community. Therefore, the development of a sense of civic duty, voluntary and community work and attentiveness to current affairs should develop early in life and perhaps should be the focus of mandatory courses offered in high schools. Such efforts will nurture a new generation of aware and active citizens who may be able to reverse the tide of declining turnout. As an example, the requirement in some provinces that high-school students perform certain voluntary community work may have the effect of instilling norms of civic and generalized reciprocity, with a consequent increase in turnout among the younger generation. Such long-term strategies seem to offer the only means of reversing the tide of declining turnout that has been building up for decades and across generations.
Acknowledgments
Financial support from the Social Sciences and Humanities Research Council of Canada is greatly appreciated (grant # 410-2003-1594). I thank two anonymous JOURNAL reviewers for providing thoughtful comments on an earlier version of this paper.