Particular and general social trust

For some, particular trust is based on knowledge of and close contact with others gained from close and constant contact with others. Hence, Hardin states, ‘for me to trust you, I have to know a fair amount about you’ (Hardin, Reference Hardin2000: 34). Similarly, Rose (Reference Rose1994: 29) finds that east Europeans ‘know who they can trust and trust who they know’, which is generally a fairly small circle of family, friends, and colleagues (see also Luhmann, Reference Luhmann1979: 43; Yamagishi and Yamagishi, Reference Yamagishi and Yamagishi1994; Offe, Reference Offe1999: 56). In contrast, Uslaner (2000–Reference Uslaner2001: 573, Reference Uslaner2008a: 102) refers to knowledge-based trust as strategic trust, and distinguishes it from particular trust, which is ‘faith only in your own kind’. Strangers, he says, are presumed to be untrustworthy, and he gives the example of religious fundamentalists who see non-believers as heathens.

Empirically, there is probably a good deal of overlap between ‘knowledge based’ particular trust and ‘own-kind’ particular trust, but theoretically they are distinct. In daily life, most people probably come into closest contact with people like themselves, and they probably develop their most trusting relations with people they come into close and constant contact with, but we cannot assume either of these claims. Do people usually, invariably, or inevitably trust others because they are of the same family, church, ethnic group, neighbourhood, profession, class, status, gender, or generation? Do we trust the person in the next office because he or she is of the same class, status, education, age, ethnic group, and nationality as us? These are subjects for empirical research rather than matters of definition.

In this article, we try to avoid empirical speculation by defining particular trust in a neutral way. Following the standard Oxford English Dictionary definition, the word ‘particular’ is used here as an adjective describing the noun ‘trust’ where social trust is associated with specific people or groups of people, whether known or in-group others. General trust is not limited in this way. It extends in a more abstract manner to people as a whole in an unselective and unspecific manner. General trust is inclusive. General trust is the belief that most people can be trusted, even if you do not know them personally, and even if they are not like you socially (Uslaner, Reference Uslaner2001–Reference Uslaner2002: 573). Accordingly, Misztal (Reference Misztal1996: 72) suggests that social trust ranges along a continuum from the personal (particular) to the abstract (general), while others see this as a continuum of ‘thick’ and ‘thin’ trust, the latter being a basis of the weak, low-density ties of large-scale society (Granovetter, Reference Granovetter1973, Reference Granovetter1983).

Since it is often assumed that general trust is most important for modern western society, most comparative research ignores particular trust and what little research is published on the subject is impressionistic or else based on plausible assumptions or limited case studies. To our knowledge, there was no systematic comparative study of particular social trust until the World Values survey of 2005–07, the basis of this work, which breaks new research ground and produces new and novel conclusions. However, it seems reasonable to hypothesize with Uslaner (Reference Uslaner1999: 123) that everyone must trust someone, and consequently that particular trust is fairly widespread in all societies. Much more is known about general social trust, which is a relatively rare commodity in some countries because it involves risks, unknowns, and short-cuts (Luhmann, Reference Luhmann1988; Kollock, Reference Kollock1994: 319; Misztal, Reference Misztal1996: 18). In only eight of the 93 nations covered in the World Values study of 2000 does more than half the population say that ‘most people can be trusted’, and five of these are in the small group of affluent nations in northwestern Europe. In another 19 nations, mostly Organisation for Economic Co-operation and Development countries, between a third and a half express attitudes of general trust, but in much of central and east Europe, Africa, Latin America, and Asia, a minority of 25% is willing to express general trust (Newton, Reference Newton2007: 346). Nevertheless, general trust is vital to modern, large-scale, mobile, and heterogeneous society in which the weak ties of daily life require risk-taking with strangers and casual acquaintances.

The relationship between particular and general social trust

Recent studies have begun to uncover the relationship between particular and general social trust, but with very different results. Some have found that they are distinct and sometimes incompatible sets of attitudes (Yamagishi and Yamagishi, Reference Yamagishi and Yamagishi1994: 144; Yamagishi et al., Reference Yamagishi, Jin and Miller1998; Newton, Reference Newton1999: 180; Stolle, Reference Stolle2001, Reference Stolle2002; Uslaner, Reference Uslaner2002: 54), while others find that the two can coexist, or that specific trust can promote general trust (Whiteley, Reference Whiteley1999; Yosano and Hayashi, Reference Yosano and Hayashi2005; Glanville and Paxton, Reference Glanville and Paxton2007; Freitag and Traunmüller, Reference Freitag and Traunmüller2009). A third possibility is suggested in an experimental study finding cultural differences between Americans, who tend to trust strangers if they are of the same in-group as themselves, and Japanese, who place greater trust in strangers if they are members of the same personal social network, irrespective of in-group/out-group differences (Yuki et al., Reference Yuki, Maddux, Brewer and Takemura2005). The inconsistent results of research add to the puzzle of trust; are particular and general social trust mutually exclusive or mutually supportive, or does this depend on culture and context?

Social and political trust

Another major part of the trust puzzle is the association between different forms of social and political trust. For some years, empirical research failed to deliver clear support for the social capital theory that social and political trust are linked at the individual level. At best, the evidence was weak and patchy, and at worst it showed no clear associations between social and political trust (Wright, Reference Wright1976: 104–110; Craig, Reference Craig1993: 27; Orren, Reference Orren1997; Hall, Reference Hall1999; Kaase, Reference Kaase1999; Newton, Reference Newton1999: 180–185, Reference Newton2001, Reference Newton2006a: 84–85; Newton and Norris, Reference Newton and Norris2000: 62–66; Uslaner, 2000–Reference Uslaner2001: 586, Reference Uslaner2002, Reference Uslaner2008a: 111; Delhey and Newton, Reference Delhey and Newton2005). At the same time, there is better evidence that cross-national levels of aggregate social and political trust are positively associated (Inglehart, Reference Inglehart1997, Reference Inglehart1999; Rothstein, Reference Rothstein1998; Putnam et al., Reference Putnam, Pharr and Dalton2000: 26; Newton and Norris, Reference Newton and Norris2000: 52–73; Newton, Reference Newton2001; Paxton, Reference Paxton2002; Torpe, Reference Torpe2003; Freitag, Reference Freitag2003a; Van der Meer, Reference Van der Meer2003; Delhey and Newton, Reference Delhey and Newton2005; Kumlin and Rothstein, Reference Kumlin and Rothstein2005; Kumlin, Reference Kumlin2007; Oskarson, Reference Oskarson2007; Svallfors, Reference Svallfors2007; Gabriel and Walter-Rogg, Reference Gabriel and Walter-Rogg2008; Listhaug and Ringdal, Reference Listhaug and Ringdal2008; Rothstein and Stolle, Reference Rothstein and Stolle2008a, Reference Rothstein and Stolleb; Freitag and Bühlmann, Reference Freitag and Bühlmann2009).

More recent studies, however, have found stronger associations between social and political trust at the individual level. Country studies in the United States, Switzerland, Germany, Sweden and Japan find a close tie between general social and political trust (Freitag Reference Freitag2003a, Reference Freitagb; Glanville and Paxton, Reference Glanville and Paxton2007; Bäck and Kestilä, Reference Bäck and Kestilä2008) and cross-national studies at the individual levels have produced similar evidence (Jagodzinski and Manabe, Reference Jagodzinski and Manabe2004: 85–97; Zmerli and Newton, Reference Zmerli and Newton2008; Freitag and Bühlmann, Reference Freitag and Bühlmann2009). The conclusion that social and political trust are, indeed, associated at both individual and cross-national aggregate levels revives a range of research questions concerning the political importance of social capital.

Although bits of the picture have been filled in, the puzzle of trust remains, not least the triangular set of relations between particular and general social trust and political trust. A reading of the large and growing social science literature suggests that there are three main models of trust. One claims that different types of social and political trust tend to fit together in a mutually reinforcing pattern, another argues the opposite, and the third suggests a more complicated pattern in which some but not all forms of trust fit together. We will refer to these as the compatible model, the incompatible model and the conditional model.

Model 1: the compatible model

The simplest model presents all three types of trust as a single syndrome; those who are trusting in one realm of life are usually trusting in the others. Two main schools of thought contribute to this view. The first is the macro approach to social capital arguing that social and political structures and institutions are major influences on levels of trust. Societies with dense networks of social relations and voluntary associations, and with embedded institutions that enforce or encourage trustworthy behaviour (police, courts, civil service and welfare institutions), will develop high levels of social and political trust. This, in turn, will reinforce the institutions and norms of civil society and so create a virtuous spiral (Putnam, Reference Putnam1993, Reference Putnam2000; Rothstein, Reference Rothstein1998; Tyler, Reference Tyler1998; Weingast, Reference Weingast1998; Rahn et al., Reference Rahn, Brehm and Carlson1999; Knack, Reference Knack2000; Levi and Stoker, Reference Levi and Stoker2000; Newton and Norris, Reference Newton and Norris2000; Paxton, Reference Paxton2002; Delhey and Newton, Reference Delhey and Newton2005; Inglehart and Welzel, Reference Inglehart and Welzel2005; Uslaner, Reference Uslaner2008b).

Social psychology offers a second theory for expecting that most forms of trust are closely associated in the same individual. Glanville and Paxton (Reference Glanville and Paxton2007) label it ‘the psychological propensity model’. It argues that trust is a core personality characteristic, learned mainly in early life and intimately linked with other personality characteristics, especially a sense of control over life, a belief in interpersonal cooperation and a sunny and optimistic disposition. In contrast, the distrusting are a misanthropic breed with a dismal view of fate, human nature, and the possibility of cooperation (Erikson, Reference Erikson1950; Rosenberg, Reference Rosenberg1956, Reference Rosenberg1957; Allport, Reference Allport1961; Cattell, Reference Cattell1965; Uslaner, Reference Uslaner1999: 138, Reference Uslaner2002: 79–86; 2000–Reference Uslaner2001: 571). If the social psychology school is right, trust and distrust of different kinds are unlikely to mix in the same individual.

The two theories are supported by some empirical evidence. Uslaner (Reference Uslaner2002: 32–33), for example, finds a small and positive correlation between particular and general social trust, and Herring et al. (Reference Herring, Jankowski and Brown1999) find little evidence in their study of African–Americans to suggest that strong in-group identity is associated with strong out-group dislike. Whiteley (Reference Whiteley1999) finds that measures of particular and general trust form a single cumulative scale. Bahry et al. (Reference Bahry2005: 525, 529) find positive correlations between in-group and out-group trust and conclude that ‘Faith in one’s own and faith in major out-groups are not mutually exclusive, but complementary’. Glanville and Paxton (Reference Glanville and Paxton2007: 240) find no evidence that trust in any one domain (particular trust) hinders the development of general trust.

Other research finds a mutually reinforcing relationship between democracy, social trust, and political trust (Putnam, Reference Putnam1993, Reference Putnam2000: 136–137; Knack and Keefer, Reference Knack and Keefer1997; Inglehart, Reference Inglehart1999; Booth and Richard, Reference Booth and Richard2001; Newton, Reference Newton2001; Paxton, Reference Paxton2002; Delhey and Newton, Reference Delhey and Newton2005). Zmerli and Newton (Reference Zmerli and Newton2008) and Zmerli et al. (Reference Zmerli, Newton and Montero2007) find a strong and significant positive correlation between individual general social trust and political trust in two separate studies covering democracies in Europe and the United States. Conversely, the political systems of the communist bloc generated both low political and general social trust (Sztompka, Reference Sztompka2000; Mishler, and Rose Reference Mishler and Rose2001).

The empirical implications of the compatibility model are that (hypothesis 1) the three types of trust are positively associated. However, (hypothesis 2) while social psychologists argue that this is a feature of individual personality characteristics, (hypothesis 3) social capital theorists argue that the association may also be influenced by top–down social climates of trust and by social and political institutions that reinforce trustworthiness.

Model 2: the incompatible model

The second model is almost, but not quite, the reverse of the compatible model. It is based on reasons for believing that particular social trust may be incompatible with both general social trust and political trust. In his influential study of ‘Montegrano’ in southern Italy, Banfield (Reference Banfield1958) argues that the local culture of amoral familism entailed trust in the family and automatic distrust of all others, including politicians who are presumed to be only self-interested. More recent work also argues that interpersonal trust (particular trust) is not an instance of a more general impersonal phenomenon, and nor can it simply be transferred to others or to other contexts (Cohen, Reference Cohen1999: 221). Hooghe (Reference Hooghe2008: 578) and Hardin (Reference Hardin2002) say it should not be assumed that we will trust strangers simply because we trust the people we know. There is little theory or evidence to tell us how particular trust can be extended to a general form, which leaves a hole in the theory claiming that the particular can lead to the general.

Some research goes an important step further by claiming that particular and general trust are inimical. A recent study of Russia points out that particular trust is generally assumed to be a zero-sum entity (Bahry et al., Reference Bahry2005: 525). Uslaner (Reference Uslaner1999: 124–125) writes, ‘the more dependent we are on our close associates and kin, the more we think of the world in terms of “we” and “they”. We won’t trust “most people” ’ (Uslaner, Reference Uslaner1999: 124). Warren (Reference Warren1999: 318) remarks that particular trust in members of the same family, clan, or group is often combined with general distrust of strangers. There is evidence that particular and general trust load on discrete factors (Yamagishi and Yamagishi, Reference Yamagishi and Yamagishi1994; Uslaner, Reference Uslaner2002: 54), and that general trust is weak among members of associations with strong in-group trust (Stolle, Reference Stolle1998: 503–504).

Other empirical work finds little evidence of associations between social and political trust at either individual or aggregate levels (Kaase, Reference Kaase1999: Newton, Reference Newton2006a: 84–86, Reference Newton2006b). This may be because the two are different things, social trust being influenced largely by social variables, including class, education, and membership of voluntary associations, and political trust being associated mainly with political variables such as political interest, party identification, and use of the political media (Abramson, Reference Abramson1983; Lawrence, Reference Lawrence1997; Mishler and Rose, Reference Mishler and Rose2001; Anderson and LoTempio, Reference Anderson and LoTempio2002).

The empirical implications of the incompatibility model are that there will be (hypothesis 4) a negative correlation between particular and general social trust, and (hypothesis 5) a negative or random association between social trust and political trust.

Model 3: the conditional model

The third model of trust suggests that there is no necessary incompatibility between particular and general social trust, and that the two may be positively associated in some cases, but not necessarily in most or all cases. There are three reasons for advancing this possibility, one logical and two empirical.

The logical argument is simple: it is difficult to conceive of an individual who is high on general trust but low on particular trust, since someone who trusts in general must logically trust in particular. The reverse is not true, however: to trust particular and selected others does not necessarily entail trusting people in general. At the same time, since everyone must trust someone, total distrust is probably as rare as total trust. Even the inhabitants of Montegrano had, perforce, to trust their own family, while distrusting all others.

The second argument for the conditional model is drawn from social psychology research. Much of it assumes, at least implicitly, that in-group identity is necessarily associated with out-group hostility (cf. Brewer, Reference Brewer1999: 430), but recent work shows that in-group attachment is independent of attitudes towards out-groups. Distrust of out-groups depends partly on competition for resources, how much the in-group feels threatened, and on the trade-off between the benefits of closure of in-group boundaries versus the opportunities of opening up to outside groups (Brewer, Reference Brewer1979, Reference Brewer1999, Reference Brewer2007; Yamagishi et al., Reference Yamagishi, Jin and Miller1998; Hewstone et al., Reference Hewstone, Rubin and Willis2002: 575–604; Voci, Reference Voci2006).

The third argument supporting the conditional model of trust concerns the wider cultural and institutional context in which in-groups and out-groups exist. Lines of demarcation can change turning out-groups into in-groups, or vice versa. Bates et al. (Reference Bates, Figueiredo and Weingast1998) gave an example of how political change caused Serbs and Croats to turn from being peaceful social groups with a high rate of inter-marriage into hostile in- and out-groups. In times of war and crisis, national populations tend to forget their internal differences and concentrate on common and external problems. Liberal and egalitarian cultures may make it easier in some countries to combine particular and general social trust compared with other countries, and countries with established, corruption-free, and power-sharing democratic governments, low levels of inequality, Protestant traditions, a strong rule of law, and universal social services seem to make it easier for individuals to combine all three forms of trust (Freitag, Reference Freitag2003a; Rothstein and Stolle, Reference Rothstein and Stolle2003; Delhey and Newton, Reference Delhey and Newton2005; Rothstein and Uslaner, Reference Rothstein and Uslaner2005; Kumlin and Rothstein, Reference Kumlin and Rothstein2005; Neller, Reference Neller2008; Freitag and Bühlmann, Reference Freitag and Bühlmann2009), compared with non-democratic countries where particular social trust seems to accompany a lack of political and general social trust.

Evidence of conditionality can be found in the very different social contexts of Russia and Germany. Along with other studies of Russia, Bahry et al. (Reference Bahry2005: 530) find that some individuals express high particular with low general social trust, but they also find others who combine high particular and general social trust. Freitag and Traunmüller (Reference Freitag and Traunmüller2009) also find evidence of an overlap between general and particular trust in Germany. Trusting one’s own group, or particular other ethnic groups, does not necessarily preclude a high level of general social trust, but is not necessarily associated with it either. It may depend upon circumstances.

The conditional model suggests (hypothesis 6) that while all those expressing general trust must necessarily express specific trust, the reverse is not necessarily true and that (hypothesis 7) not all of those who trust in particular will necessarily trust in general, although (hypothesis 8) some will combine both forms of social trust. Similarly, (hypothesis 9) the more open, democratic, and egalitarian a society the more its citizens will combine all three forms of trust.

A note of caution

Often assumed to be much less important than general social trust for modern society, particular trust has slipped into the background of research. The recent Handbook of Social Capital (Castiglione et al., Reference Castiglione, van Deth and Wolleb2008) mentions it once, in passing, in its 720 pages. There are a few single-country surveys that include questions about both general and particular social trust, alongside questions about political trust, but until the most recent World Values survey we know of no cross-national study that includes batteries of questions about all three kinds of trust. Therefore, the three models outlined above are built on weak foundations of formal logic, brave assumptions, and some contradictory evidence, some of it circumstantial or based on single case studies. All three models have a prima facie plausibility, but Models 1 and 2 are contradictory and both cannot be right.

Social trust

The World Values survey asks a set of six questions about social trust.

The first three deal with forms of particular trust involving known others with whom respondents have close ties (family and those they know personally) and people in their neighbourhood. These three measures cover particular social trust in the sense that they ask about particular, specific, and selected others. The last three questions cover general trust in unspecified and possibly different others who are not known personally. In addition, World Values asks: ‘Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?’. This is the ‘standard’ general social trust question used in many surveys. Respondents are offered a ‘Yes/No’ option for this question, as against the four rating scale of the other six questions, which makes comparison difficult, but it provides a useful benchmark against which to judge the general nature of responses to the last three questions, and vice versa.

Principal component analysis of the pooled social trust data for 22 countries yields a two component solution, one weighted primarily on the measures of particular trust in known others and the other on the four measures of general trust (Table 1). This confirms the validity of the three general trust questions, and vice versa (see also Uslaner, Reference Uslaner2002: 54). Although the six or seven trust measures cluster in two groups, these are not completely distinct, because trust in the neighbourhood and in people known personally show positive loadings on both components, though not the heaviest ones. This suggests that there are two types of social trust corresponding to the particular–general distinction, but that they overlap to some degree.Footnote 4 Similar results have appeared in national studies of the United States, Germany, and Japan (Yosano and Hayashi, Reference Yosano and Hayashi2005; Glanville and Paxton, Reference Glanville and Paxton2007; Freitag and Traunmüller, Reference Freitag and Traunmüller2009), but Table 1 provides cross-national confirmation of these studies. It is consistent with empirical propositions 6 and 7 of the conditional model.

Table 1 Principal component analyses of measures of social trust, promax rotation, structure matrix, with and without the standard general trust item

The newly released World Values survey integrated data set presents equilibrated weights for the 2005 survey in which N = 1000 or 1500. All tables in this paper are based on equilibrated data in which N = 1000.

Political trust

The World Values survey asks the same questions about a set of six political and governmental institutions as follows:

The organisations named are parliament, the government, political parties, the justice system, the civil service, and the police. The World Values survey also asks about a further set of private institutions, but since this study focuses on interpersonal social trust and political trust, they are not included in this study.

In common with previous research (Listhaug and Wiberg, Reference Listhaug and Wiberg1995; Zmerli and Newton, Reference Zmerli and Newton2008), Table 2 shows that principal component analysis of the six World Values questions tapping political confidence reveals a single component. As with social trust, the principal component analysis shows a common cross-national pattern of responses to political confidence questions.

Table 2 Principal component analysis of six measures of confidence in political institutions

Countries weighted by equilibrated weight (N = 1000).

At this stage, we could conduct two kinds of analysis, one focusing on country variations and explanations of them, the other trying to build high-level cross-national generalizations about trust, irrespective of national differences. Notwithstanding the existence of national variations that are both interesting and important, this article chooses the second approach. The fact that different types of trust cluster in a clearly patterned way when individual data for 22 nations are pooled shows that it is also meaningful to take individuals as the unit of analysis, leaving national comparisons for later work. Correlation and multi-level analysis (see below) support this approach. In addition, analysing pooled individual data across 22 nations helps to set a broadly based framework of research within which particular country variations may be more meaningfully examined at a later stage.

The second choice involves using trust measures based either on factor loadings or indices. We choose to use indices here.Footnote 5 The first index measures particular trust in family, neighbours, and people known personally, and the second measures general trust in people met for the first time and those of other nationalities and religions. Both indices are based on the 4-point rating scale of the trust questions, ranging from 1 (low trust) to 4 (high trust) and have 13 parameter values. The standard general social trust question is dropped from the analysis for three reasons. First, as the authors have shown elsewhere (Zmerli and Newton, Reference Zmerli and Newton2008), interviewees are offered a ‘Yes/No’ response option that does not discriminate as effectively as a 4-item rating scale, and second, a single-item measure of social trust is not as sensitive as a 3-item scale. Third, as the figures in Table 1 show, the 3-item scale in the World Values survey is a better measure of general social trust than the standard question. The political trust index is built in the same way using the six political confidence questionnaire items. It ranges from 1 (low trust) to 4 (high trust) and has 37 parameter values.

The study proceeds by way of correlation analysis to show the strength, direction, and consistency of the associations between the three types of trust in the 22 countries, before moving to multi-level analysis to test the strength of their associations with each other and with other individual and contextual variables.

Cross-national patterns

Table 3 shows the distribution of national trust scores ranked in descending order of general trust. As hypothesized, particular trust is higher than general trust in every country and usually by a substantial margin. The average general trust score for all countries weighted equally is 2.3 on a 4-point scale. For particular trust it is 3.3, which is 43% higher. However, the hypothesis that general social trust is more common than political trust is not supported by the figures. Across the 22 countries, political and general social trust have the same mean of 2.3 and in 13 countries political trust is equal to, or higher than, general social trust. High levels of political compared with general trust are found in a mixture of countries, including South Africa, Switzerland, Finland, India, Cyprus, and Germany, which adds to the puzzle.

Table 3 Means of social and political trust by country, ranked by general social trust

^aEntries are based on the weighted means of the three trust indices. Average scores and Ns are based on World values equilibrated weight (N = 1000). Countries in order of descending mean of general social trust.

Correlations between the three types of trust show that they are positively and significantly associated (Table 4). All correlations except two are statistically significant at 0.01, and those between particular and general trust are most generally the highest, as one might expect.Footnote 6 The correlations are generally within the same, rather small range in each column showing, once again, strong cross-national consistency. One surprise, however, is that in 10 of the 22 cases, the association between political and particular social trust is stronger than that between political and general social trust. Therefore, as social capital theory would have it, there is a close and positive correspondence between social and political trust at the individual level in all countries, but in some countries particular trust is a more important correlate than general trust. These results are not consistent with the incompatible model of trust, and seem to fit the conditional model better than does the compatible model. Nevertheless, the strong association between particular social and political trust is unexpected.Footnote 7 It also seems from Table 3 that all three forms of trust are generally higher in wealthier and established democracies, a finding that is consistent with previous cross-national studies (Inglehart, Reference Inglehart1997, Reference Inglehart1999; Delhey and Newton, Reference Delhey and Newton2005; Newton, Reference Newton2007: 347).

Table 4 Correlations between indices of three types of trust. Pearson’s r, by country

^aResults are based on equilibrated weights (N = 1000); ***P < 0.001, **P < 0.01, *P < 0.05.

The claim that everyone must trust someone is confirmed by the fact that of the 28,031 individuals giving valid answers, only 31 (0.1%) claimed to have no social trust of any of the six kinds. Similarly, no more than 0.6% professed the strange combination of general but not particular trust, a finding that is consistent with hypothesis 6. It may be that 0.1% and 0.6% figures are noise created by interviewer error, miscoding, or failure on the part of those interviewed to understand or hear the questions.Footnote 8 Particular trust is the normal position for nine out of ten people in democratic countries, where 92% express high particular trust (score 2.6 or more on the particular trust index). It is virtually impossible to develop a sense of general trust in the absence of particular trust, but particular trust is not automatically associated with general trust. No less than 99% of those with high general trust have high particular trust, whereas only 45% of those with high particular trust also have high general trust. This combination of figures is consistent with the conditional model, but not with the compatible model or the incompatible model. Nor are the figures consistent with the social psychology school claiming that there is a personality type for whom trust is a core characteristic. It seems that people choose where to place their trust, distinguishing between those close to them (family, neighbours, and those they know well) and those who are not (members of other nations, religions, and those they have just met). Trust is not a general personality characteristic, but something involving choices and distinctions, a suggestion that fits the conditional model.

There is a similar asymmetry between social and political trust, in which the great majority of those with political trust are socially trusting, but only one third to 40% of those who are socially trusting are also politically trusting. This also suggests that social trust, either particular or general, is a necessary but not a sufficient cause of political trust. This also suggests that social trust conditions political trust, although it also indicates an unexpectedly important role for particular trust.

In the next stage of the analysis, we use multi-level modelling in order to put the associations between different forms of trust to a more rigorous test and to estimate the effects of individual and contextual variables on them.

Multi-level analysis

A reading of the already voluminous literature on trust suggests a fairly short and consistent list of individual variables associated with it, namely membership of voluntary associations, education, church attendance, age, gender. For political trust, we can add political identity and political interest.Footnote 9 The list of contextual variables covers corruption, ethnic heterogeneity, social equality, and various measures of democratic development, durability, and performance.

Contextual variables

Homogeneity and ethnic fractionalization

Social heterogeneity is often argued to be a cause of low trust (Helliwell, Reference Helliwell1996; Knack and Keefer, Reference Knack and Keefer1997; Hero, Reference Hero1998, Reference Hero2003; Alesina and La Ferrara, Reference Alesina and La Ferrara2000; Glaeser et al., Reference Glaeser, Laibson, Scheinkman and Soutter2000; Zak and Knack, Reference Zak and Knack2001; Costa and Kahn, Reference Costa and Kahn2003; Eisenberg, Reference Eisenberg2006). The indicator used here is the Alesina and La Ferrara index of ethnic fractionalization (Reference Alesina and La Ferrara2002).

Corruption

Corruption in public life is also argued to be a cause of low trust (Svensson, Reference Svensson1998; Alesina et al., Reference Alesina, Baqir and Easterly1999; Annett, Reference Annett1999; Easterly, Reference Easterly2000; Uslaner, Reference Uslaner2005).

Democratic quality and durability

The better established and developed a democratic system, the more likely it is to sustain comparatively high levels of social and political trust (Dunn, Reference Dunn1993; Putnam, Reference Putnam1993: 111–115; Anderson and Guillory, Reference Anderson and Guillory1997; Brehm and Rahn, Reference Brehm and Rahn1997; Inglehart, Reference Inglehart1999; Newton and Norris, Reference Newton and Norris2000: 70; Booth and Richard, Reference Booth and Richard2001: 55; Paxton, Reference Paxton2002). This study uses the duration of democracy and the Economist quality of democracy index as its measures.

Government effectiveness and performance

The more effectively and impartially a political system performs for its citizens, the more likely it is to generate high levels of political trust and to sustain high levels of social trust (Knack and Keefer, Reference Knack and Keefer1997: 1275–1276; Tyler, Reference Tyler1998; Levi and Stoker, Reference Levi and Stoker2000; Uslaner, Reference Uslaner2002: 223–229; Delhey and Newton, Reference Delhey and Newton2005; Rothstein and Stolle, Reference Rothstein and Stolle2008a, Reference Rothstein and Stolleb). Three World Bank indicators are used, measuring government effectiveness, rule of law, and regulatory quality.

Equality and inequality

Previous research shows social and economic equality to have a positive association with social and political trust (Putnam, Reference Putnam2000: 354–361; Warren, Reference Warren2004: 143; Delhey and Newton, Reference Delhey and Newton2005; Uslaner, Reference Uslaner2008a: 113). The indicator used here is the Gini index.Footnote 10

In addition, we intend to scrutinize whether and to what extent the interrelationships between our three types of trust are affected by country-specific variations of the contextual variables. For this purpose, we create cross-level interactions and include them in the multi-level models.

Tables 5, 6 and 7 present multi-level analyses of each type of trust at the individual level. Since the macro variables create problems of colinearity, each table presents them separately. Each set of tables includes both macro and micro measures of the other two types of trust as independent variables in order to test for their aggregate (‘rainmaker’ effects) and individual level effects on other types of trust.Footnote 11 Each set of tables includes both fixed effects based on the cross-national consistencies in the pooled individual-level data for 22 countries and cross-level interactions.

Table 5 Macro- and micro-level determinants of particular social trust, multi-level analysis, and random intercept models

Table 5 (Continued)

***P < 0.001, **P < 0.01, *P < 0.05; for the coding of the variables see Appendix. Predictors are centred on their grand means. Countries weighted by equilibrated weight (N = 1000); explained variances calculated according to Bryk and Raudenbush (Reference Bryk and Raudenbush1992). The intra-class correlation coefficient based on the empty model indicates that 83.1% of the variance are explained by the first level and 16.9% by the second level.

Table 6 Macro- and micro-level determinants of general social trust, multi-level analysis, and random intercept models

Table 6 (Continued)

***P < 0.001, **P < 0.01, *P < 0.05; for the coding of the variables see Appendix. Predictors are centred on their grand means. Countries weighted by equilibrated weight (N = 1000); explained variances calculated according to Bryk and Raudenbush (Reference Bryk and Raudenbush1992). The intra-class correlation coefficient based on the empty model indicates that 74.7% of the variance are explained by the first level and 25.3% by the second level.

Table 7 Macro- and micro-level determinants of political trust, multi-level analysis, random intercept models

Table 7 (Continued)

***P < 0.001, **P < 0.01, *P < 0.05; for the coding of the variables see Appendix. Predictors are centred on their grand means. Countries weighted by equilibrated weight (N = 1000); explained variances calculated according to Bryk and Raudenbush (Reference Bryk and Raudenbush1992). The intra-class correlation coefficient based on the empty model indicates that 81.2% of the variance are explained by the first level and 18.8% by the second level.

Together, the multi-level tables present a rich collection of statistics, from which we may draw six general conclusions. First, the numerous significant cross-level interaction effects show significant country variations, but the fixed effects also show strong patterns in the pooled individual-level data for the 22 countries. There are, in other words, generally strong consistencies across countries as well as variations between them. Second, the trust levels of a country as a whole have a significant impact on individual levels. Consistent with the rainmaker hypothesis (hypothesis 3), aggregate social trust of the particular and general variants have large and significant impacts on individual levels of social and political trust. This evidence is also supported by the significant cross-level interactions, which reveal that aggregate trust levels impact on the strength and direction of micro level determinants of trust. Third, and following from this point, all three forms of trust form a fairly tight triangle of interdependent relationships, although particular and general social trust are more closely associated with each other than with political trust, again consistent with the conditional model. Fourth, particular trust is not incompatible with either general social trust or political trust, being positively and significantly associated with them at both the aggregate and individual levels, at least in the democratic nations covered in this study.Footnote 12 Particular social trust seems to play an important part in creating and sustaining high levels of general and political trust. Fifth, contextual political variables do not make much of a contribution to particular social trust, and they are more important for general and (not surprisingly) for political trust. The overarching, society-wide impact of democratic government, rule of law, government effectiveness, and regulatory quality would seem to have an impact on general trust in unknown others, but less so on particular trust in close and known others. This suggests a slight modification of hypothesis 3 to read that aggregate and institutional macro variables have a stronger impact on general social and political trust than on particular forms of trust.