When trust matters—a multilevel analysis of the effect of generalized trust on political participation in 25 European democracies

ABSTRACT

Generalized social trust has been argued to have positive effects for both individuals and society, but despite solid theoretical arguments in favour of different contextual effects, the empirical evidence remains scant. We here examine whether and how the effect of generalized trust on the propensity to become politically active is moderated by the level of generalized trust in society. We inspect two different causal mechanisms: The rainmaker effect, which entails that the effect of generalized trust is weaker at the individual level when there is a high level of generalized trust in society; and the sunmaker effect, which entails that the effect of generalized trust is strengthened by a high level of generalized trust in society. We examine the links for three forms of political participation: voting, institutionalized participation, and non-institutionalized participation. The data come from the fourth round of the European Social Survey from 2008 [European Social Survey. (2008). Data file edition 4.0. Norwegian Social Science Data Services, Norway—Data Archive and distributor of ESS data], and we include 47,489 respondents in 25 democratic countries. The results from a series of multilevel logistic regressions suggest that the effect of generalized trust varies depending on the level of generalized trust in the surrounding community but the causal mechanisms differ for different kinds of political participation.

KEYWORDS:

• Generalized trust
• political participation
• social capital
• contextual effects
• multilevel analysis

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Maria Bäck http://orcid.org/0000-0003-2008-9391

Henrik Serup Christensen http://orcid.org/0000-0003-2916-0561

Notes

1. Israel has been excluded since it is geographically located outside the European context and Lithuania is excluded since the data set does not include appropriate weights.

2. Information on coding and descriptive statistics for all variables is available in the Appendix. VIF scores suggest that there is no need for concern about multicollinearity since all scores are below 2.5.

3. Self-reporting is not a perfect measure of actual voting (see Bäck, 2011, p. 78), but it is the best available for the present purposes. Belgium, Greece, and Cyprus have compulsory voting, but including dummy variables for these countries does not alter the results substantially.

4. An exploratory factor analysis for the six items included in institutionalized and non-institutionalized participation show that they load onto two separate dimensions in accordance with this distinction.

5. We tried to include the level of corruption since this may well be an important factor, but this caused problems with multicollinearity (VIF scores above 7 when only including this and the aggregate level of generalized trust at the country level). Nevertheless, the distinction between old and new democracy is closely related to the level of corruption (Pearson correlation coefficient 0.75), meaning this simplistic distinction at least partly controls for this aspect as well.

6. We also ran empty models for each form of participation to decompose the variance into group and individual level. The results suggested that the multilevel approach is warranted since the intra-class correlation was 0.12 for voting, 0.05 for institutionalized participation and 20% for non-institutionalized participation.

7. A simple bivariate regression between contextual generalized trust and non-institutionalized participation suggests that there is a significant linkage, but this evaporates when controlling for the other factors.

Additional information

Funding

This work was supported by the Academy of Finland [grant number 257055].