Explaining the Impact of Formal Institutions on Social Trust: A Psychological Approach

Abstract

By drawing on psychological models of action choice, this study distinguishes between four key factors that determine trust building: (a) knowledge to trust, (b) others-regarding, (c) cognition and (d) contexts. These four factors are combined into a single analytical framework that is used for establishing channels through which the institutional context impacts social trust formation. Our theoretical and empirical evidence suggests that context is the strongest determinant of trust, with its overall effect being, however, modified by the degree to which the individual’s knowledge of trusting, cognition and others-regarding is developed. The Programme for the International Assessment of Adult Competencies (PIAAC) data from the year 2012 are utilized for testing our propositions.

JEL: D02, D03, Z10, Z13

Keywords

Social trust trust formation formal institutions action choice multi-level analysis

Introduction

Trust is traditionally considered as a cultural attribute that is influenced by an individual’s internal values and formed during early socialization processes (Fukuyama, 2000). Recent studies have dissociated trust building from an individual’s personal characteristics and externalized it to contextual properties (Nooteboom, 2007; Rothstein & Stolle, 2001), with formal institutions often viewed as one such contextual factor (Farrell, 2005; Farrell & Knight, 2003).

Analyses of institutional contexts’ impact on trust are drawn upon a twofold definition of institutions. First, institutions are viewed as a set of public organizations that individuals interact with over the course of their lives. The process of trust formation is affected by such institutions, as citizens evaluate the quality of their performance (Edlund, 2006; Mishler & Rose, 2001) or that of elected officials (Thomas, 1998). Social trust is interpreted as a product of institutional trust (Letki, 2006; Murphy, 2004; Rothstein & Eek, 2009; Rothstein & Stolle, 2001, 2002, 2003; Tyler, 2006): A positive experience with public institutions or their officials may result in the belief that these institutions perform their task in a fair and unbiased manner, which motivates individuals to exhibit more trust towards them and to extrapolate this trust to other people (Kumlin, 2004; Kumlin & Rothstein, 2007; Rothstein & Eek, 2009; Rothstein & Stolle, 2001, 2010). In eliciting trust, public authorities need not be objectively trustworthy, but they must be perceived as such by citizens (Levi, 1998; Scholz & Lubell, 1998).

Second, institutions are viewed as a set of rules that define legal boundaries within which individuals are allowed—and expected—to act. Efficient formal institutions are conducive to establishing trust, since they enforce third-party agreements (Herreros & Criado, 2008). They enable individuals to pursue redress and restitution when wronged, thereby reducing the risk involved in trusting someone (Rothstein & Stolle, 2001; Tillmar & Lindkvist, 2007) and serving as a safety net (Farrell, 2005). If sanctions and penalties are imposed when a contract is breached, formal institutions may also increase the cost of betrayal (Bohnet & Baytelman, 2007) and overcome the information deficit problem by indicating how others are likely to act (Farrell & Knight, 2003). Formal institutions’ impact is especially strong when formal rules are duly enforced (Oskarsson, Öberg & Svensson, 2009) and perceived by individuals as being fairly applied to various population groups (Oskarsson et al., 2009).

Although both strands to explaining the context–trust nexus found empirical evidence for a positive relationship between the quality of formal institutions and trust levels, they still exhibit a common deficiency: They lack a clear formalization of the mechanisms through which institutional contexts influence trust. Most empirical studies either solely examine the associations that exist between trust levels and institutional scores while controlling for the socio-demographic characteristics of respondents (Herreros & Criado, 2008). Or, they offer mathematical models derived from the rational choice perspective (Zak & Knack, 2001), which do not account for non-cognitive (cultural) forces beyond rational thinking that underlie an individual’s decision-making regarding whether or not to trust.

This research’s main objective is to introduce a new comprehensive model of trust formation in institutional contexts by drawing on various theories of psychology. Since trusting others is a mental operation governed by brain structures, we argue that action choice models can explain how trust emerges and describe what role formal institutions play in this process. Applying a psychological approach to explain trust formation enables addressing cognitive and non-cognitive, as well as conscious and subconscious, mechanisms of an individual’s decision-making simultaneously. This allows us to integrate cultural and contextual factors of trust into a single analytical framework.

Literature Overview

Psychology’s point of departure is the assumption that behaviour is a function of the organism and its environment (Stern, 2000; Triandis, 1977). This integrated approach suggests that behaviour is an interactive product of personal sphere variables and contextual determinants (Stern, 2000). The former represents a variety of factors that are internal to the individual, such as specific personal beliefs, norms, values, attitudes or general predispositions to act in certain ways as well as personal capabilities (see Cees & Bruel, 2007 for a detailed overview). The latter may include numerous influences that are external to the individual, such as institutional and legal factors, public policy, interpersonal influences or broader dimensions of the social context (Cees & Bruel, 2007). The structural dynamics between the influence of internal and contextual factors is a focal point in the psychological models of behaviour.

The majority of models interpret behaviour as a two-stage process (DeScioli & Kurzban, 2009; Strack & Deutsch, 2004). First, one needs to form values, norms or action knowledge reflecting an individual’s understanding of the world, a so-called moral identity (Bandura, 1991, 2001; Higgins, 1996; Narvaez, Lapsley, Hagele & Lasky, 2006; Shao, Aquino & Freeman, 2008). Second, the individual must act in accord with this knowledge or values, whereas the context predefines whether or not sufficient motivation and opportunity for behavioural compliance are present (Broom, 2006; DeScioli & Kurzban, 2009).

More specifically, the brain stores action knowledge in the form of moral identity defined as an organized mental representation (scheme) of how an individual with certain norms, values is likely to think, feel and act (Shao et al., 2008). The individual is believed to possess multiple and sometimes competing mental representations that are formed through both intergenerational transmissions and further influenced by later experiences with others. These mental representations might not exist chaotically, as they are organized according to one’s internal understanding of the world (Taatgen, Huss, Dickison & Anderson, 2008). Each of the mental representations is not stored in the individual’s memory as mere facts but is augmented with (a) preconditions under which they can be carried out and (b) a representation of their expected outcomes (Taatgen et al., 2008).

Regardless of the actual volume of one’s stock of mental representations, only one of them can be activated for processing social information and hence be mapped onto an action at any given moment (Higgins, 1996; Markus & Kunda, 1986). Which one is ultimately activated mainly depends on the environment within which the individual is placed (Bargh, Bond, Lombardi & Tota, 1986; Shao et al., 2008). Environmental or contextual properties are often viewed as the key determinant in choosing the mental representation that will be mapped onto an action.

The individual’s interaction with the context unfolds through the perceptual system and results in collecting perceptual input (Prinz, 1997). The brain uses the perceptual input to align the observed contextual properties with the preconditions and outcomes of mental representations. When preconditions and expected outcomes match the perceptual input, the corresponding mental representation is mapped onto an action (Taatgen et al., 2008).

Psychology has identified two major problems that may arise during the match process. First, all of the information about the contextual properties is not available or observable. In this case, the individual’s internal understanding of the world governs the selection of a mental representation leading to the action (Taatgen, 2005, 2007), and hence, culture defines the level of trust displayed by the individual. This suggests that the culture–behaviour link is strongest when the context is weak or contextual factors are non-existent. Conversely, there is virtually no link between culture and behaviour when the context is strong.

Second, the mental representation containing the necessary contextual characteristics as a precondition might not be present in one’s declarative memory. Experiments show that, in this case, participants simply discover the relevant knowledge of action by taking a random action and observing whether it can bring them closer to the goal. Once the correct action is selected, a new record (mental representation) is created with the original perceptual state as a precondition and the resulting perceptual state as a post-condition (Taatgen et al., 2008).

Personal experience with the context hence shapes the individual’s behaviour by activating or triggering the formation of a mental representation that contains the knowledge required for optimal action in the given context. Psychology further suggests that the context’s influence may go beyond the individual’s own behaviour and can be extended to their expectations regarding other people’s behaviour. One’s personal experience with the new context may prompt the individual to expect others to act similarly as a result of a similar experience in the given context (Jones, 1996; Lewis & Weigert, 1985). An individual who is positively affected by the context may hence make references from their own behaviour to others’ behaviour, re-assessing their perception of others regarding their intentions and motives (Jones, 1996; Lewis & Weigert, 1985).

Analytical Framework

We define trust as an individual’s expression of confidence in the intentions and motives of others (Deutsch, 1958; Mellinger, 1956). As such, trust arises through interactions with other people, whereas the decision of whether or not to trust someone is based on assessing others’ trustworthiness (Coleman, 1982; Hardin, 2001). We further expand this conventional understanding of trust formation with psychological determinants of action choice while viewing the selection of the level of trust to display as an action. We therefore suggest that trust is determined by four key factors:

T r u s t = f (K n o w l e d g e, O t h e r s - r e g a r d i n g, C o g n i t i o n, C o n t e x t),

(1)

where Knowledge refers to an individual’s set of mental representations, with each containing the knowledge of how much trust to exhibit towards others in every particular situation. In line with the psychological models of action choice, we assume that these mental representations are structured according to the individual’s internal understanding of the world and that only one of them is activated in any given moment. This results in displaying the level of trust that is embedded within the activated mental representation. We further assume that any knowledge of trust is formed through both early socialization processes (Rotenberg, 2007) and an individual’s past course of actions or past experience with other people (Hardin, 2001). The knowledge facet hence refers to both dispositional and experiential sources of trust emergence that the literature on trust offers (see Dinesen & Bekkers, 2015 for a detailed overview).

Others-regarding refers to an individual’s perception of other people’s motives and intentions. We believe that it is influenced by direct interactions with other individuals and may include evaluating others’ trustworthiness in deciding which level of trust to display as the traditional approach to interpreting trust as a function of the trustee’s trustworthiness suggests (Hardin, 2001). However, we deviate from this conventional understanding of how regarding others enters the trust equation. We focus on the individual’s ability to learn from acting in the institutional context and to generalize the impact of formal institutions from one’s own behaviour to the behaviour of others, which shapes the individual’s expectations about other people’s actions.

Cognition refers to the individual’s general cognitive abilities to perceive, process and record information, such as declarative memory, procedural memory and processing speed. This component is expected to predefine the quality and the speed of mental operations involved in collecting the perceptual input about the contexts’ properties, the choice of mental representation and its mapping onto an action. The cognition facet relates to the intelligence view on trust that defines social trust as a product of the individual’s cognitive abilities. More intelligent individuals are more able to detect signs of untrustworthiness and hence engage in more interactions with trustworthy others, as a result of which they upgrade their trust levels (see Dinesen & Bekkers, 2015 for a detailed overview).

Context refers to the properties of the institutional context that the individual perceives and utilizes to match the preconditions and outcomes of mental representations in choosing the mental representation that will be mapped onto an action. Good context is expected to be characterized by properties that facilitate activities or social relations relevant to the individual’s decision-making regarding trust. Context is a multi-faceted concept and we limit the scope of our analysis to the political environment in which the individual must act. In line with the institutional approach to trust, the political context is defined as the recognized structure of rules and principles within which laws are created, enforced and applied (Rothstein & Eek, 2009; Rothstein & Stolle, 2007). The political context is considered to be good when it mediates conflicts efficiently and provides representation for the populous (Rothstein & Eek, 2009; Rothstein & Stolle, 2007).

We model trust formation, as defined by the four factors, as follows (see Figure 1). An individual possesses (a) a cognitive system of certain qualities and (b) a set of mental representations that store information regarding how much they can trust others in various institutional contexts. The knowledge stored in these mental representations is assumed to stem from the individual’s cultures, experiences and overall goals. In accordance with the psychological models, we expect that these mental representations are augmented by (a) the precondition that specifies which level of trust should be displayed in a certain institutional context and (b) the expected outcome that this particular level of trust may lead to in this institutional context. In choosing the level of trust, the individual evaluates contextual properties and forms expectations about other people’s motives or intentions in this context. Both pieces of information are linked and matched to the preconditions and outcomes of mental representations. The one that corresponds to the match criteria is selected for mapping onto an action and the level of trust stored in this mental representation will be displayed. If no match is found, the individual randomly selects the level of social trust. It will be stored as a new mental representation if the new level of trust is sufficient to bring the individual closer to their goals. If the chosen trust level is insufficient for the given context, random selection is repeated until the match between one’s goals and the outcomes of exhibiting this certain level of trust is found. Personal experience with the context is further extrapolated by the individual to other people acting in the same context, reshaping this individual’s expectations regarding other people’s behaviour in the given contextual properties. The individual uses this information to determine which level of trust to exhibit towards others in later interactions, provided that the contextual properties and one’s personal experience with them remain unchanged.

Figure 1.

Source: Authors’ own computations.

We further argue that the context’s ultimate impact on the individual’s level of trust is related to the individual’s knowledge, others-regarding and cognition.

Proposition 1: We expect that the institutional context’s positive impact on trust is greater for individuals who already possess mental representations with the knowledge of exhibiting sufficiently high levels of trust. Good institutional contexts may encourage the up-levelling of trust by activating a mental representation corresponding to high trust levels. In order to become active, the brain should previously have recorded such mental representation. If this is not the case, the individual is less likely to display the sufficient level of trust in good institutional contexts. When lacking a match, the individual must randomly select a trust level and analyse this action’s consequences. Since this process usually involves errors, it may take time before the right level of trust is determined and stored as a new record in the brain. Additionally, the brain chooses a new level of trust, even when randomly or not, through drawing analogies with the existing knowledge of trust. This suggests that any positive, even substantial, developments in the institutional context are unlikely to lead to a sudden leap from distrust to complete trust for the individual with no, or little, knowledge to trust. Rather, trust levels will gradually increase, as the individual chooses to display more and more trust until the right level of trust is attained.

Proposition 2: We expect that the institutional context’s positive impact on trust will be greater for individuals who can better extrapolate their own positive experiences from the context to others. To trust others, one must account for others’ motives and intentions. One’s expectations regarding others’ motives and intentions can change when the new institutional context provides positive experiences. When an improved institutional environment triggers a positive change in one’s own behaviour, the individual may assume that this context can also encourage other individuals to act more honestly, ameliorating this individual’s expectation regarding others’ motives and intentions, which leads to higher trust levels towards them. The overall change in trust levels through this extrapolation mechanism may hence depend on the extent to which the individual is able or willing to make references from his or her own experience with the new context to other people’s behaviour. The individual who has a more developed extrapolation mechanism is expected to display more trust towards others when the institutional context improves.

Proposition 3: We expect that the context’s positive impact on social trust is stronger for individuals who possess a better cognitive system. The action choice process includes numerous mental operations, such as collecting perceptual input, coupling between perceived contextual properties and preconditions or outcomes of mental representations, and mapping the selected mental representation onto an action. As such, an individual’s general cognitive abilities may determine the speed and quality with which the brain performs these operations. Because the matching and learning processes are usually faster and more efficient for individuals with better cognition, we expect that such individuals are more responsive to positive changes in contextual properties, and hence the contextual improvements’ positive impact on trust is greater for them. We also acknowledge that this impact is strong in the periods immediately following a change in the institutional context and tends to subside when an individual attains the right level of trust in the new or changed institutional environment.

We use the aforementioned observations to postulate our hypotheses:

Hypothesis 1: The institutional context’s positive impact on trust is greater for individuals who possess better knowledge to trust.

Hypothesis 2: The institutional context’s positive impact on trust is greater for individuals who have a more developed extrapolation mechanism.

Hypothesis 3: The institutional context’s positive impact on trust is greater for individuals who have a better cognitive system.

Data and Methods Description

To test our hypotheses, we utilize the public use data from the Programme for the International Assessment of Adult Competencies (PIAAC) conducted by the OECD in 2012. The survey evaluates adults’ proficiency in key information-processing skills and gathers information on how adults use their skills at home, at work and in the wider community.¹ It offers measures for the key cognitive and workplace skills needed for individuals to participate in society, achieve their goals and develop their potential (OECD, 2016). Since social trust is closely linked to the individual’s stock of knowledge and skills or often regarded as a skill in itself that brings a return and is worth investing in (Keeley, 2007; OECD, 2001), the PIAAC database contains a question about respondents’ trust levels. The question’s formulation is unique, because it limits social trust to faith in other people without relating it to the individual’s caution levels as is often the case in socio-economic surveys (Ben-Ner & Halldorsson, 2010; Miller & Mitamura, 2003).²

Our sample includes Austria, Belgium (Flanders), the Czech Republic, Denmark, Estonia, Finland, France, Germany, Ireland, Italy, Japan, Korea, the Netherlands, Norway, Poland, the Russian Federation, the Slovak Republic, Spain, Sweden, the UK (England and Northern Ireland) and the USA. Australia and Canada are excluded from the analysis, since data for many variables were suppressed for these countries. The sample is restricted to respondents aged between 16 and 65 years. We use the following set of variables to empirically test our hypotheses (see Table 1 for descriptive statistics).

Table 1.

Descriptive Statistics for the Key Variables Used in the Analysis

	N	Minimum	Maximum	Mean	Std. Deviation
Social trust	76,558	1.000	5.000	2.362	1.161
Knowledge	76,658	1.000	5.000	1.642	1.019
Others-regarding	76,055	6.000	30.000	21.950	4.359
Cognition	76,711	37.870	426.120	278.129	42.914
Context	76,289	1.000	5.000	2.767	1.259
Respondent’s education level
Highly educated	73,509	0.000	1.000	0.396	0.489
Middle educated	73,509	0.000	1.000	0.461	0.498
Yearly income percentile	70,921	1.000	6.000	3.335	1.532
Age	76,722	1.000	5.000	3.108	1.266
Health status	76,651	1.000	5.000	2.480	0.985
Living with a spouse or partner	66,705	0.000	1.000	0.741	0.438
Children in the household	76,659	0.000	1.000	0.649	0.477
Born in the country	76,689	0.000	1.000	0.909	0.287
Lifelong learning	75,372	0.000	4.000	0.903	0.967
Unrealized learning	75,388	0.000	1.000	0.269	0.443
Learning in the job	70,006	3.000	15.000	9.847	3.111
Job tasks
Teaching, presentations, planning for others	76,570	3.000	15.000	6.802	3.514
Cooperating or sharing information with others	70,243	2.000	10.000	7.633	2.213
Selling, advising or influencing others	76,373	4.000	20.000	11.464	4.952
Planning or organizing activities for oneself	76,502	2.000	10.000	8.079	2.704
Managerial responsibilities	66,329	0.000	1.000	0.289	0.454
Economic sector (Private)	76,550	0.000	1.000	0.724	0.447
Economic sector (Public)	76,550	0.000	1.000	0.249	0.433
Economic sector (Non-profit)	76,550	0.000	1.000	0.027	0.161
Number of firms worked in	76,604	0.000	7.000	1.810	1.267
Flexibility with working hours	76,649	1.000	5.000	2.658	1.407
Job satisfaction	76,657	1.000	5.000	1.993	0.843
Working hours per week	73,163	1.000	125.000	38.319	13.441
Skill mismatch	76,032	0.000	1.000	0.832	0.374

Source: Authors’ own computations.

Dependent Variable

Social trust is approximated by asking the respondents to specify whether they agree with the statement that they can only trust a few people completely. The response scale varies from 1 ‘strongly agree’ to 5 ‘strongly disagree’.

Independent Variables

In operationalizing knowledge, we adopted the conventional assumption that involvement in voluntary associations in the past approximates the individual’s current understanding of trust and hence their current stock of knowledge to trust. Volunteering is expected to offer regular and close contact with others, and such contacts help to develop reciprocity, cooperation, empathy for others, an understanding of the common interest and common good and, as a result, the knowledge of trust (Brehm & Rahn, 1997; Paxton, 2002; Putnam, 2000). The knowledge variable is operationalized through the question asking how often the respondent did voluntary work in the last 12 months, with the response values ranging between 1 ‘never’ and 5 ‘every day’.

Others-regarding is expected to capture whether the respondent tends to link or extrapolate their own knowledge/experiences to others and is measured through a synthetic construct obtained by summing up the responses to six questions about the extent to which the respondent agrees with the following statements: (a) When I learn or read about new ideas, I try to relate them to real-life situations to which they might apply, (b) I like learning new things, (c) When I come across something new, I try to relate it to what I already know, (d) I like to get to the bottom of difficult things, (e) I like to figure out how different ideas fit together and (f) If I do not understand something, I look for additional information to make it clearer. The final construct has a scale ranging between 6 and 30, with higher values corresponding to a greater subjective propensity to extrapolate information.

Cognition is measured through the individual’s cognitive abilities and is constructed by averaging out the cognitive test results in literacy and numeracy, as in Blau and Kahn (2005). Each of the two constructs represents a mean of 10 possible values that the PIAAC survey provides.

Context is operationalized through the perceived legitimacy of governmental policies. The legitimacy variable is expected to reflect the quality of the country’s political institutions and is measured through the extent to which the respondents agree with the following statement: ‘People like me do not have any say about what the government does’. The response scale ranges from 1 ‘strongly agree’ to 5 ‘strongly disagree’.

Control Variables

We control for the conventional determinants of trust: respondents’ education level, income, health condition, living with a spouse or partner, the presence of children in the household, immigration status and age. The education level is measured by two dummies. The first takes the value of 1 if the respondent has obtained a higher level education, and the second takes the value of 1 if the respondent has obtained a middle-level education. A low education is used as a reference category. The income variable is operationalized through the PIAAC derived variable that specifies the decimal to which the respondent’s yearly income belongs. The health condition is measured through a question asking respondents to self-rate their health status from 1 ‘excellent’ to 5 ‘poor’. The living with a spouse or partner variable is a dummy that takes the value of 1 if the respondents declare living with a spouse or partner. The presence of children in the household is measured by a dummy variable that takes the value of 1 if the respondents declare that the household has at least one child. Immigration status is captured by a dummy variable that takes the value of 1 if the individual was born in the country where the survey was conducted. The respondents’ age is measured in 10-year bands.

Additionally, we include learning-related and work-related controls in the trust regression. Participation in lifelong learning is measured by questions where respondents specify whether they did any of the following learning activities within the last 12 months: (a) courses or private lessons, (b) seminars or workshops, (c) courses conducted through open or distance education or (d) any organized on-the-job training sessions or training by supervisors or co-workers. The final construct’s response scale ranges from 0 ‘no participation in lifelong learning’ to 4 ‘active participation in lifelong learning’. Unrealized learning is measured by a dummy variable that takes the value of 1 if the individual specifies that, in the last 12 months, there were learning activities the individual wanted to participate in but did not. Learning on the job is a synthetic variable constructed by summing up responses to three questions asking how often the current job involves: (a) learning new work-related things from co-workers or supervisors, (b) learning-by-doing from tasks one performs and (c) keeping up to date with new products or services. Each item has a response scale ranging from 1 ‘never’ to 5 ‘every day’ so that the final construct has values between 3 ‘no learning on the job’ and 15 ‘active learning on the job’.

We also control for job tasks and emotions-related variables. Based on a factor analysis, we combined job-task questions into four constructs, as presented in the social trust base model. The respondents’ involvement in managerial responsibilities is captured by a dummy variable that takes the value of 1 if the respondent currently manages or supervises other employees. The level of workplace stress due to the workload is operationalized through the number of weekly work hours. Skill mismatch is measured through a dummy that takes the value of 1 if the respondents specify that they can cope with more demanding duties than those they are currently required to perform.

Finally, we control for the respondent’s job sector, job stability, freedom at workplace and job satisfaction. The job sector is measured through two dummies specifying whether the individual’s job belongs to the public or non-profit sector, with the private sector representing a reference category. Job stability is measured by the number of different firms or organizations the respondent has changed in the last five years. The freedom at workplace is derived by asking whether the individual can choose or change their working hours. The response scale varies from 1 ‘not at all’ to 5 ‘to a very high extent’. Job satisfaction is operationalized through the question in which the respondents need to self-rate their satisfaction with job by choosing between the value of 1 ‘extremely satisfied’ and 5 ‘extremely dissatisfied’.

Methods Used in the Analysis

The main method of analysis is multi-level modelling, which accounts for our data’s hierarchical structure and prevents the unmodelled country information from being pooled into the single individual error term (Kreft & de Leeuw, 1998; Luke, 2004). The base model takes the following form:

\begin{matrix} T_{i j} = γ_{00} + γ_{10} K n o w l e d g e_{i j} + γ_{20} O t h e r s - r e g a r d i n g_{i j} + \\ γ_{30} C o g n i t i o n_{i j} + γ_{40} C o n t e x t_{i j} + γ_{50} X_{i j} + m_{o j} + ε_{i j} \end{matrix}

(2)

Here, T_ij stands for the individual level of social trust. Knowledge_ij, Others-regarding_ij, Cognition_ij and Context_ij are the measures for the knowledge, others-regarding, cognition and contextual factors, respectively. X_ij is a set of control variables, m_oj is the variance at the country level and ε_ij is the variance at the individual level. We further augment the base model by including interaction terms between the context and the three other factors as follows:

\begin{matrix} T_{i j} = γ_{00} + γ_{10} K n o w l e d g e_{i j} + γ_{20} O t h e r s - r e g a r d i n g_{i j} + \\ γ_{30} C o g n i t i o n_{i j} + γ_{40} C o n t e x t_{i j} + γ_{50} K n o w l e d g e * C o n t e x t_{i j} + \\ γ_{60} O t h e r s - r e g a r d i n g * C o n t e x t_{i j} + γ_{70} C o g n i t i o n * C o n t e x t_{i j} + \\ γ_{80} X_{i j} + m_{o j} + ε_{i j} \end{matrix}

(3)

where T_ij is trust scores, Knowledge_ij, Others-regarding_ij, Cognition_ij and Context_ij are the measures for the four trust factors as described previously. Knowledge *Context_ij, Others-regarding*Context_ij and Cognition*Context_ij are the interaction terms between the contextual factor and the three other trust factors that are (sequentially) included in the base model. Again, X_ij is the set of individual-level control variables, m_oj is the country-level variance, whereas ε_ij is the individual-level variance.

The STATA gllamm command is utilized to calculate the model’s parameters. Since social trust is an ordinal variable, the ologit link is specified together with the binominal family sub-options. Additionally, we include the GLLAMM adapt option, which causes adaptive quadrature to be used instead of ordinary quadrature.

Analysis and Results

The results of the social trust base model are consistent with previous findings concerning the impact of conventional trust determinants and work-related factors on trust levels (see Table 2). Augmenting the base model with the four factors substantially improves the model’s fit. Overall, our results indicate that individuals with more knowledge of trust tend to have higher trust scores (see Table 3). Similarly, individuals with a better cognitive system display more trust. Surprisingly, a more developed extrapolation mechanism is associated with lower trust levels. Finally, individuals who assign higher scores to formal institutions have higher trust scores.

The formal institutions’ role in trust building increases after accounting for the interactions between the institutional context and one’s knowledge, others-regarding and cognition (see Table 4). In line with Hypothesis 1, the institutional context’s positive impact on social trust is significantly greater for individuals who possess better knowledge of trusting others. The contextual effects on trust are also stronger for individuals who have a more developed extrapolation mechanism and can better generalize new knowledge to new domains or to others’ behaviours, which is consistent with Hypothesis 2. This positive interaction effect has been found regardless the fact that the others-regarding variable is negatively associated with trust levels. We do not detect any interaction between the context and the cognition variables and hence more intelligent or more able individuals do not show a stronger impact of good formal institutions on trust levels. We do not receive any support for Hypothesis 3. Nonetheless, we reveal a strong interaction between the knowledge and cognition variables, suggesting that the formation of knowledge to trust occurs faster for more able individuals.

These impacts do not change after applying the selected strategies of robustness check, such as (a) controlling for sample size variations (see Appendix 1 for results) or (b) controlling for endogeneity in the relationship between the four components and social trust (see Appendix 2 for results).

Overall, the empirical analysis supports our four factor model of trust creation. Our results confirm that the institutional context is not only a strong determinant of trust building but also stands in a close relationship to the three other factors in the trust formation process. To analyse how imbalances in the four factors’ maturity levels may impact trust emergence, we group countries according to their values on these four factors. By applying a hierarchical cluster analysis to the aggregated data set (see Figure 2), we distinguish between four country groups (see Table 5). The first encompasses Finland, Japan, the Netherlands, Norway and Sweden, where all the four components take high values, resulting in high trust levels. The second group includes France, Ireland, Korea, Poland, the UK and the USA, where the knowledge component is slightly better developed than the cognition and context components, resulting in average levels of trust. The third group consists of Austria, Belgium, the Czech Republic, Denmark, Estonia, Germany, the Russian Federation and the Slovak Republic, with lower scores on volunteering but relatively good values on the cognitive and contextual components, with trust scores being close to those of the second group. The fourth group includes Italy and Spain, where all the four components are underdeveloped compared to other countries, resulting in the sample’s lowest trust levels. One should note that the others-regarding component does not show a substantial variation across the selected countries. In Figure 3, we provide a four-dimensional visualization of this country grouping.

Table 2.

Social Trust Base Model

Variables	(1)	(2)	(3)	(4)
Respondent’s education level
Highly educated		0.732***	0.646***	0.598***
		(0.026)	(0.030)	(0.031)
Middle educated		0.212***	0.195***	0.179***
		(0.024)	(0.028)	(0.028)
Yearly income percentile		0.050***	0.030***	0.048***
		(0.006)	(0.007)	(0.008)
Health status		–0.122***	–0.119***	–0.107***
		(0.009)	(0.009)	(0.009)
Living with a spouse or partner		0.073***	0.072***	0.068***
Living with a spouse or partner		(0.020)	(0.022)	(0.022)
Children in the household		–0.107***	–0.099***	–0.114***
		(0.022)	(0.023)	(0.023)
Born in the country		0.229***	0.217***	0.182***
		(0.028)	(0.031)	(0.030)
Age		0.075***	0.084***	0.067***
		(0.008)	(0.009)	(0.009)
Lifelong learning			0.101***	0.087***
			(0.009)	(0.009)
Unrealized learning			–0.031	–0.021
			(0.019)	(0.019)
Learning on the job			0.009***	0.005
			(0.003)	(0.003)
Job tasks
Teaching, presentations, planning for others			0.025***	0.021***
Teaching, presentations, planning for others			(0.003)	(0.003)
Cooperating or sharing information with others			0.001	0.005
Cooperating or sharing information with others			(0.004)	(0.004)
Selling, advising or influencing others			–0.004**	–0.003
Selling, advising or influencing others			(0.002)	(0.002)
Planning or organizing activities for oneself			0.008**	0.002
Planning or organizing activities for oneself			(0.004)	(0.004)
Managerial responsibilities			–0.078***	–0.079***
			(0.022)	(0.022)
Working hours per week				–0.004***
				(0.001)
Skill mismatch				–0.086***
				(0.023)
Economic sector (Public)				0.134***
				(0.020)
Economic sector (Non-profit)				0.260***
				(0.051)
Number of firms worked in				–0.014**
				(0.007)
Flexibility with working hours				0.052***
				(0.007)
Job satisfaction				–0.097***
				(0.011)
Between-class variance	0.272	0.336	0.279	0.333
	(0.084)	(0.048)	(0.084)	(0.065)
Log likelihood	–102,739.970	–77,694.231	–66,789.354	–66,530.996
Number of level 2 units	21	20	20	20
Number of level 1 units	75,718	58,277	50,245	50,149

Source: Authors’ own computations.

Notes: Standard errors in parentheses.

*** p < 0.01, ** p < 0.05 and * p < 0.1.

To estimate the four groups’ potential to increase social trust levels by improving their context, we conduct a prediction exercise in which we assign the context variable’s highest value of 5 to the individuals’ contextual component while keeping the three other components unchanged. The STATA gllapred option (see Rabe-Hesketh & Skrondal, 2008 for more details) is used to make the predictions. The procedure requires first conducting a multi-level analysis of actual trust scores and then calculating predictions. In doing so, we calculate the marginal cumulative probabilities and conditional cumulative probabilities with random intercept equal to zero. We further calculate a mean value of such predictions for each country and report them in comparison to the actual value of trust (see Table 6). Our analysis reveals that even if all people in the selected countries have a similarly positive perception of a government’s effectiveness, the trust scores still remain heterogeneous across the four groups. Group 1, with the best scores on the four factors, has the highest predicted trust levels. Group 4, with the lowest scores on the four factors, has the lowest predicted trust scores. Interestingly, our results suggest that Group 3, with a poor volunteering tradition but high values on the cognitive component, will improve trust scores to a slightly greater extent than Group 2, with a more developed tradition of volunteering but relatively worse investments in formal and non-formal education. Both groups however still remain behind Group 1 in their predicted trust levels. This finding’s key implication is that an individual’s response to a positive change in the context depends not only on the actual size of this change, but also on the quality of the individual’s stock of knowledge to trust, cognition and the degree to which this individual applies the extrapolation mechanism in analysing their own and others’ behaviour.

Table 3.

Social Trust Augmented Model, with the Four Psychological Factors

Variables	(1)	(2)	(3)	(4)	(5)
Knowledge	0.083***				0.064***
	(0.009)				(0.009)
Others-regarding		–0.008***			–0.021***
		(0.002)			(0.002)
Cognition			0.004***		0.003***
			(0.000)		(0.000)
Context				0.377***	0.372***
				(0.008)	(0.008)
Constant
Cut 1	–0.712***	–0.918***	0.153	–0.078	0.474***
	(0.120)	(0.142)	(0.104)	(0.093)	(0.105)
Cut 2	1.375***	1.167***	2.245***	2.085***	2.646***
	(0.120)	(0.142)	(0.104)	(0.094)	(0.106)
Cut 3	1.986***	1.777***	2.857***	2.712***	3.276***
	(0.121)	(0.142)	(0.105)	(0.094)	(0.106)
Cut 4	3.907***	3.696***	4.780***	4.672***	5.238***
	(0.122)	(0.143)	(0.107)	(0.096)	(0.108)
Control variables	Yes	Yes	Yes	Yes	Yes
Between-class variance	0.244	0.254	0.262	0.347	0.262
	(0.065)	(0.081)	(0.046)	(0.037)	(0.025)
Log likelihood	–65,961.077	–65,999.889	–65,876.738	–64,743.043	–64,606.693
Number of level 2 units	20	20	20	20	20
Number of level 1 units	49,766	49,766	49,766	49,766	49,766

Source: Authors’ own computations.

Notes: Standard errors in parentheses. The list of controls includes the full set of variables from the social trust base model.

*** p < 0.01, ** p < 0.05 and * p < 0.1.

Table 4.

Interactions between the Key Trust Factors

Variables	(1)	(2)	(3)	(4)
Knowledge	0.011	0.064***	0.064***	–0.239***
Knowledge	(0.022)	(0.009)	(0.009)	(0.062)
Others-regarding	–0.021***	–0.035***	–0.021***	–0.021***
	(0.002)	(0.005)	(0.002)	(0.002)
Cognition	0.003***	0.003***	0.003***	0.001***
Cognition	(0.000)	(0.000)	(0.000)	(0.000)
Context	0.343***	0.255***	0.322***	0.372***
Context	(0.013)	(0.036)	(0.048)	(0.008)
Context * Knowledge	0.018***
Context * Knowledge	(0.006)
Context * Others-regarding		0.005***
Context * Others-regarding		(0.002)
Context * Cognition			0.001
Context * Cognition			(0.000)
Knowledge * Cognition				0.001***
Knowledge * Cognition				(0.000)
Control variables	Yes	Yes	Yes	Yes
Between-class variance	0.262	0.244	0.222	0.258
Between-class variance	(0.025)	(0.025)	(0.025)	(0.025)
Log likelihood	–64,603.251	–64,600.768	–64,605.357	–64,593.684
Number of level 2 units	20	20	20	20
Number of level 1 units	49,766	49,766	49,766	49,766

Source: Authors’ own computations.

Notes: Standard errors in parentheses. The list of controls includes the full set of variables from the social trust base model.

*** p < 0.01, ** p < 0.05 and * p < 0.1.

Conclusion and Discussion

Overall, our study supports the four factor conceptual framework of trust formation. Both theoretical and empirical elaborations indicate that formal institutions can considerably modify the dynamics and patterns of trust building processes. Our results are in line with the previous findings, demonstrating that ‘the public’ is an important determinant of ‘the private’ (Kumlin & Rothstein, 2007), but the extent to which ‘the public’ ultimately intervenes into ‘the private’ depends on the actual properties of this ‘private’. The context’s overall impact on trust is a function of a particular individual’s actual properties, such as current stocks of knowledge to trust, ones’ cognition, and the ability to learn and extrapolate one’s own knowledge and experiences to other individuals. These properties set certain constraints in the relationship between the formal institutional context and social trust levels.

Figure 2.

Source: Authors’ own computations.

Table 5.

The Four Factors’ Mean Values, by Country Group

	Country	Knowledge	Others-regarding	Cognition	Context
Group 1	Finland Japan The Netherlands Norway Sweden	1.8 (Strong)	21.8 (Average)	291.3 (Strong)	3.1 (Strong)
Group 2	France Ireland Korea Poland The UK The USA	1.6 (Relatively Strong)	21.9 (Average)	270.5 (Relatively Weak)	2.7 (Relatively Weak)
Group 3	Austria Belgium The Czech Republic Denmark Estonia Germany The Russian Federation The Slovak Republic	1.5 (Relatively Weak)	21.9 (Average)	280.7 (Relatively Strong)	2.8 (Relatively Strong)
Group 4	Italia Spain	1.4 (Weak)	22.9 (Relatively Strong)	258.4 (Relatively Weak)	2.3 (Weak)

Source: Authors’ own computations.

Figure 3.

Source: Authors’ own computations.

Our psychological approach to the analysis of the context–trust nexus may have substantial theoretical implications. The four facet model introduced by this study provides the foundation for uniting the contextual strand with the majority of existing theories on trust formation into a single analytical framework. The knowledge component incorporates the dispositional theories that view trust as a deeply rooted disposition or belief formed by either genetic transmission or early life socialization (Dinesen & Bekkers, 2015). It can be further broadened to the individual’s later experiences by integrating the experiential perspective on trust into the conceptual framework. The approach to explaining trust as a function of trustworthiness can be linked to the others-regarding component. The cognition facet relates to the intelligence perspective on social trust (Yamagishi, 2001) and education-related predictors of trusting others (Helliwell & Putnam, 2007; Knack & Keefer, 1997). Finally, the contextual component can be expanded to a wide range of environmental factors stemming from family, neighbourhood, workplace (Grund & Harbring, 2009) or the overall institutional framework (Herreros & Criado, 2008). This integrated approach may provide a more complete view on how contextual properties affect trust formation processes by expanding the context’s direct influences to interactions with dispositional and numerous experiential predictors in trust-related behaviour over the individual’s life course.

Table 6.

Predicted Levels of Social Trust, by Country Group

		Actual trust scores	Predicted social trust scores
			Linear	Conditional cumulative probabilities				Marginal cumulative probabilities
			Linear	above 1	above 2	above 3	above 4	above 1	above 2	above 3	above 4
Group 1	Finland	2.785	2.742	0.902	0.523	0.373	0.080	0.893	0.522	0.379	0.088
	Japan	2.329	2.745	0.902	0.524	0.373	0.079	0.894	0.523	0.380	0.088
	The Netherlands	2.733	2.762	0.903	0.527	0.378	0.081	0.894	0.526	0.385	0.089
	Norway	2.846	2.804	0.906	0.538	0.389	0.085	0.897	0.536	0.394	0.094
	Sweden	2.839	2.802	0.906	0.537	0.387	0.085	0.898	0.536	0.393	0.093
	Group average	2.706	2.771	0.904	0.530	0.380	0.082	0.895	0.529	0.386	0.090
Group 2	France	1.939	2.589	0.887	0.486	0.339	0.069	0.878	0.487	0.347	0.077
	Ireland	2.206	2.672	0.894	0.506	0.358	0.075	0.886	0.506	0.365	0.083
	Korea	2.239	2.539	0.882	0.474	0.328	0.067	0.872	0.475	0.337	0.074
	Poland	2.306	2.486	0.878	0.460	0.316	0.062	0.861	0.449	0.314	0.066
	The UK	2.259	2.670	0.894	0.506	0.358	0.075	0.885	0.505	0.365	0.083
	The USA	2.341	2.642	0.890	0.499	0.353	0.075	0.881	0.499	0.360	0.082
	Group average	2.215	2.600	0.888	0.489	0.342	0.071	0.877	0.487	0.348	0.078
Group 3	Belgium	2.315	2.703	0.898	0.514	0.365	0.078	0.889	0.513	0.372	0.085
	The Czech Republic	1.956	2.538	0.883	0.474	0.328	0.066	0.873	0.475	0.336	0.073
	Denmark	3.181	2.811	0.906	0.539	0.390	0.086	0.898	0.537	0.396	0.094
	Estonia	2.008	2.590	0.887	0.486	0.339	0.069	0.878	0.487	0.347	0.077
	Germany	2.263	2.669	0.894	0.506	0.358	0.075	0.885	0.505	0.365	0.083
	The Russian Federation	2.332	2.486	0.878	0.461	0.316	0.062	0.868	0.463	0.325	0.069
	The Slovak Republic	2.009	2.496	0.879	0.463	0.318	0.063	0.869	0.465	0.327	0.069
	Group average	2.295	2.613	0.889	0.492	0.345	0.071	0.880	0.492	0.353	0.079
Group 4	Italy	1.948	2.434	0.872	0.449	0.306	0.060	0.861	0.451	0.315	0.067
	Spain	2.323	2.459	0.872	0.455	0.313	0.062	0.862	0.457	0.321	0.069
	Group average	2.136	2.447	0.872	0.452	0.310	0.061	0.862	0.454	0.318	0.068

Source: Authors’ own computations.

Note: The model used for calculating the predictions is as follows: Trust = 0.064Knowledge – 0.021Others_regarding + 0.003Cognition + 0.372Context + 0.393Respondent_highly_educated + 0.095Respondent_middle_educated + 0.022Income + 0.079Age – 0.092Health_status + 0.061Living_with_spouse – 0.088Children + 0.051Born_in_the_country + 0.061Lifelong_learning – 0.045Unrealized_learning + 0.008Leanring_in_the_job + 0.017Job_tasks_1 + 0.009Job_tasks_2 – 0.006Job_tasks_3 – 0.001Job_tasks_4 – 0.077Managerial_responsibilities + 0.115Economic_sector_PUBLIC + 0.187Economic_sector_NON-Profit – 0.022Number_of_firms_worked + 0.035Flexibility – 0.082Job_satisfaction – 0.003Working_hours – 0.080Skill_mismatch. Predictions are calculated by assuming that all respondents assign the institutional context the highest value of ‘5’.

Further research is needed to eliminate two major drawbacks of our study. First, an alternative set of operationalizations should be found for each of the four factors to further validate the complex structure of the trust formation processes introduced by our model. Particular attention should be given to the choice of trust knowledge measures. We utilize one’s active participation in voluntary associations in the past as a measure for the individual’s current stock of knowledge to trust. However, trust knowledge can be formed through mechanisms different from volunteering and hence the operationalization item used in our analysis might not fully approximate the individual’s overall stock of knowledge relevant to trusting others. Additionally, there are studies that failed to provide evidence of a positive impact that one’s participation in voluntary associations has on social trust (Delhey & Newton, 2003; Letki, 2006; Stolle, 2003), thereby suggesting that volunteering may be a poor measure of one’s current knowledge to trust. Second, it is necessary to check the dynamic nature of the relationship between the four factors and social trust by testing the proposed analytical framework with longitudinal data.

Footnotes

Acknowledgements

We thank Stephanie Vogel for proofreading the manuscript.

Appendix

Appendix 2.

Robustness Check: Controlling for Endogeneity in Trust Regressions

Variables	(1)	(2)	(3)	(4)
Knowledge	1.306***	0.086***	0.101***	0.044***
	(0.059)	(0.006)	(0.005)	(0.006)
Others-regarding	–0.027***	–0.043***	–0.006***	–0.016***
Others-regarding	(0.002)	(0.007)	(0.001)	(0.002)
Cognition	0.001	0.003***	0.002***	0.001***
	(0.000)	(0.000)	(0.000)	(0.000)
Context	0.178***	0.247***	0.256***	0.758***
	(0.007)	(0.005)	(0.004)	(0.019)
Constant	0.324***	–0.047	0.937***	0.418***
	(0.057)	(0.124)	(0.083)	(0.043)
Observations	55,323	40,173	63,919	55,537

Notes: Robust standard errors in parentheses. Column (1) contains the results of an instrumental variable (IV) regression, in which knowledge to trust is instrumented with an individual’s working hours per week and countries’ mean values for respondents’ altruism levels. The altruism variable is considered to be the foundation of trust and is operationalized through the question asking the extent to which respondents agree that it is important to help other people and care about others’ well-being. The European Social Survey (ESS) data from the year 2012 are used as a source for the altruism variable. Column (2) reports the results of an IV regression, in which we instrument the others-regarding component with the following set of PIAAC variables: the respondent’s work experience length in years, the type of the respondent’s employment contract, the intensity of computer use at work, the need for further training, company size, number of people living in the respondent’s household and the respondent’s job industry code. Column (3) reports the results for the IV regression, in which the cognition component is instrumented with the mother’s immigration background and the number of people living in the respondent’s household. In addition, we employ a conventional psychological approach to instrument the respondent’s genetics through the mother’s level of education (Cunha & Heckman, 2008). Column (4) presents the results of an IV regression in which we instrument the contextual component with the Freedom House civil liberties index (as in Fidrmuc, 2003) and countries’ morale culture calculated as the average of responses to two ESS questions asking whether respondents consider it wrong to (a) make an exaggerated or false insurance claim and (b) buy something that they think might be stolen.

*** p < 0.01, ** p < 0.05 and * p < 0.1.

Notes

References

Bandura

(1991). Social cognitive theory of moral thought and action. In Kurtines

W. M.

& Gewirtz

J. L.

(Eds.), Handbook of moral behaviour and development (pp. 45–103). Hillsdale, NJ: Erlbaum.

Bandura

(2001). Social cognitive theory: An agentic perspective. Annual Review of Psychology, 52(1), 1–26.

Bargh

J. A.

Bond

R. N.

Lombardi

W. J.

& Tota

M. E.

(1986). The additive nature of chronic and temporary sources of construct accessibility. Journal of Personality and Social Psychology, 50(5), 869–878.

Ben-Ner

& Halldorsson

(2010). Trusting and trustworthiness: What are they, how to measure them and what affects them. Journal of Economic Psychology, 31(1), 64–79.

Blau

F. D.

& Kahn

L. M.

(2005). Do cognitive test scores explain higher U.S. wage inequality? The Review of Economics and Statistics, 87(1), 184–193.

Bohnet

& Baytelman

(2007). Institutions and trust: Implications for preferences, beliefs and behaviour. Rationality and Society, 19(1), 99–135.

Brehm

& Rahn

(1997). Individual level evidence for the causes and consequences of social capital. American Journal of Political Science, 41(3), 999–1023.

Broom

(2006). The evolution of morality. Applied Animal Behavioural Sciences, 100(1–2), 20–28.

Cees

& Bruel

(2007). Nothing is as practical as a good theory: Analysis of theories and a tool for developing interventions to influence energy-related behaviour. Retrieved from http://www.cres.gr/behave/pdf/paper_final_draft_CE1309.pdf

10.

Coleman

J. S.

(1982). Systems of trust: A rough theoretical framework. Angewandte Sozialforschung, 10(3), 277–299.

11.

Cunha

& Heckman

J. J.

(2008). Technology of cognitive and non-cognitive skill formation. The Journal of Human Resources, 43(4), 738–782.

12.

Delhey

& Newton

(2003). Who trusts? The origins of social trust in seven societies. European Societies, 5(2), 93–137.

13.

DeScioli

& Kurzban

(2009). Mysteries of morality. Cognition, 112(2), 281–299.

14.

Deutsch

(1958). Trust and suspicion. Journal of Conflict Resolution, 2(4), 187–197.

15.

Dinesen

P. T.

& Bekkers

(2015). The foundation of individuals’ generalized social trust: A review. In Van Lange

P. A. M.

Rockenbach

& Yamagishi

(Eds.), Trust in social dilemmas (pp. 77–99). Oxford: Oxford University Press.

16.

Edlund

(2006). Trust in the capability of the welfare state and general welfare state support: Sweden 1997–2002. Acta Sociologica, 49(4), 395–417.

17.

Farrell

(2005). Trust and political economy: Institutions and the sources of interfirm cooperation. Comparative Political Studies, 38(5), 459–483.

18.

Farrell

& Knight

(2003). Trust, institutions and institutional evolution: Industrial districts and the social capital hypothesis. Politics and Society, 31(4), 537–556.

19.

Fidrmuc

(2003). Economic reform, democracy and growth during post-communist transition. European Journal of Political Economy, 19(3), 583–604.

20.

Fukuyama

(2000). Social capital and civil society (IMF Working Paper No. 00/74). Washington, DC: IMF.

21.

Grund

& Harbring

(2009). Trust and control at the workplace: Evidence from representative samples of employees in Europe (IZA Discussion Paper No. 4297). Bonn: IZA.

22.

Hanushek

E. A.

Woessmann

& Zhang

(2011). General education, vocational education and labour-market outcomes over the life-cycle (CESifo Working Paper No. 3614). Munich: IFO.

23.

Hardin

(2001). Conceptions and explanations of trust. In Cook

K. S.

(Ed.), Trust in society (pp. 2–39). New York, NY: Russell Sage Foundation.

24.

Helliwell

J. F.

& Putnam

R. D.

(2007). Education and social capital. Eastern Economic Journal, 33(1), 1–19.

25.

Herreros

& Criado

(2008). The state and the development of social trust. International Political Science Review, 29(1), 53–71.

26.

Higgins

T. E.

(1996). The ‘self-digest’: Self-knowledge serving self-regulatory functions. Journal of Personality and Social Psychology, 71(6), 1062–1083.

27.

Jones

(1996). Trust as an affective attitude. Ethics, 107(1), 4–25.

28.

Keeley

(2007). Human capital: How what you know shapes your life. Paris: OECD.

29.

Knack

& Keefer

(1997). Does social capital have an economic payoff? A cross-country investigation. Quarterly Journal of Economics, 112(4), 1251–1288.

30.

Kreft

& de Leeuw

(1998). Introducing multilevel modelling. Thousand Oaks, CA: SAGE Publications.

31.

Kuckulenz

& Zwick

(2003). The impact of training on earnings: Differences between participant groups and training forms (ZEW Discussion Paper No. 03–57). Manheim: Centre for European Economic Research.

32.

Kumlin

(2004). The personal and the political: How personal welfare state experiences affect political trust and ideology. New York, NY: Palgrave Macmillan.

33.

Kumlin

& Rothstein

(2007). Making and breaking social capital: The impact of welfare state institutions. Comparative Political Studies, 38(4), 339–365.

34.

Letki

(2006). Investigating the roots of civic morality: Trust, social capital, and institutional performance. Political Behaviour, 28(4), 305–325.

35.

Levi

(1998). A state of trust. In Levi

& Braithwaite

(Eds.), Trust and governance (pp. 77–101). New York, NY: Russell Sage Foundation.

36.

Lewis

D. J.

& Weigert

(1985). Trust as a social reality. Social Forces, 63(4), 967–985.

37.

Luke

D. A.

(2004). Multilevel modeling. Thousand Oaks, CA: SAGE Publications.

38.

Markus

& Kunda

(1986). Stability and malleability of the self-concept. Journal of Personality and Social Psychology, 51(4), 858–866.

39.

Mellinger

G. D.

(1956). Interpersonal trust as a factor in communication. Journal of Abnormal Social Psychology, 52(3), 304–309.

40.

Miller

A. S.

& Mitamura

(2003). Are surveys on trust trustworthy? Social Psychology Quarterly, 66(1), 62–70.

41.

Mishler

& Rose

(2001). What are the origins of political trust? Testing institutional and cultural theories in post-communist societies. Comparative Political Studies, 34(1), 30–62.

42.

Murphy

(2004). The role of trust in nurturing compliance: A study of accused tax avoiders. Law and Human Behaviour, 28(2), 187–209.

43.

Narvaez

Lapsley

D. K.

Hagele

& Lasky

(2006). Moral chronicity and social information processing: Tests of a social-cognitive approach to the moral personality. Journal of Research in Personality, 40(6), 966–985.

44.

Nooteboom

(2007). Social capital, institutions and trust. Review of Social Economy, 65(1), 29–53.

45.

OECD (2001). Well-being of nations: The role of human and social capital. Center for Educational Research and Innovation. Paris: Author.

46.

OECD (2016). The survey of adult skills: Reader’s companion. OECD Skills Studies (2nd ed.). Paris: Author.

47.

Oskarsson

Öberg

& Svensson

(2009). Making capitalism work: Fair institutions and trust. Economic and Industrial Democracy, 30(2), 294–320.

48.

Paxton

(2002). Social capital and democracy: An interdependent relationship. American Sociological Review, 67(2), 254–277.

49.

Prinz

(1997). Perception and action planning. European Journal of Cognitive Psychology, 9(2), 129–154.

50.

Putnam

(2000). Bowling alone. New York, NY: Simon & Schuster.

51.

Rabe-Hesketh

& Skrondal

(2008). Multilevel and longitudinal modeling using Stata (2nd ed.). College Station, TX: Stata Press.

52.

Rotenberg

(2007). The socialization of trust: Parents’ and children’s interpersonal trust. International Journal of Behavioural Development, 18(4), 713–726.

53.

Rothstein

& Eek

(2009). Political corruption and social trust: An experimental approach. Rationality and Society, 21(1), 81–112.

54.

Rothstein

& Stolle

(2001, November). Social capital and street-level bureaucracy: An institutional theory of generalized trust. Paper prepared for ‘Trust in Government Conference’ at the Centre for the Study of Democratic Politics, Princeton University, NJ.

55.

Rothstein

& Stolle

(2002, August 29–September 2). How political institutions create and destroy social capital: An institutional theory of generalized trust. Paper prepared for the 98th annual meeting of the American Political Science Association, Boston, MA.

56.

Rothstein

& Stolle

(2003). Introduction: Social capital in Scandinavia. Scandinavian Political Studies, 26(1), 1–26.

57.

Rothstein

& Stolle

(2007). The quality of government and social capital: A theory of political institutions and generalized trust (QoG Working Paper Series 2007:2). Department of Political Science, Göteborg University, Göteborg.

58.

Rothstein

& Stolle

(2010). The state and social capital: An institutional theory of generalized trust. In Kesselman

(Ed.), Readings in comparative politics: Political challenges and changing agendas (pp. 123–128). Boston, MA: Wadsworth Cengage Learning.

59.

Scholz

J. T.

& Lubell

(1998). Trust and taxpaying: Testing the heuristic approach to collective action. American Journal of Political Science, 42(2), 398–417.

60.

Shao

Aquino

& Freeman

(2008). Beyond moral reasoning: A review of moral identity research and its implications for business ethics. Business Ethics Quarterly, 18(4), 513–540.

61.

Stern

P. C.

(2000). Toward a coherent theory of environmentally significant behaviour. Journal of Social Issues, 56(3), 407–424.

62.

Stolle

(2003). The sources of social capital. In Hooghe

& Stolle

(Eds.), Generating social capital: Civil society and institutions in a comparative perspective (pp. 19–42). New York, NY: Palgrave Macmillan.

63.

Strack

& Deutsch

(2004). Reflective and impulsive determinants of social behaviour. Personality and Social Psychology Review, 8(3), 220–247.

64.

Taatgen

N. A.

(2005). Modelling parallelization and flexibility improvements in skill acquisition: From dual tasks to complex dynamic skills. Cognitive Science, 29(3), 421–455.

65.

Taatgen

N. A.

(2007). The minimal control principle. In Gray

(Ed.), Integrated models of cognitive systems (pp. 368–379). Oxford: Oxford University Press.

66.

Taatgen

N. A.

Huss

Dickison

& Anderson

J. R.

(2008). The acquisition of robust and flexible cognitive skills. Journal of Experimental Psychology: General, 137(3), 548–565.

67.

Thomas

C. W.

(1998). Maintaining and restoring public trust in government agencies and their employees. Administration and Society, 30(2), 166–193.

68.

Tillmar

& Lindkvist

(2007). Cooperation against all odds: Finding reasons for trust where formal institutions fail. International Sociology, 22(3), 343–366.

69.

Triandis

(1977). Interpersonal behaviour. Monterey, CA: Brooks/Cole.

70.

Tyler

T. R.

(2006). Psychological perspectives on legitimacy and legitimation. Annual Review of Psychology, 57(2), 375–400.

71.

Yamagishi

(2001). Trust as a form of social intelligence. In Cook

(Ed.), Trust in society (pp. 121–147). London: SAGE Publications.

72.

Zak

P. T.

& Knack

(2001). Trust and growth. The Economic Journal, 111(470), 295–321.