Trust Propensity Across Cultures: The Role of Collectivism

Conceptualization of Trust Propensity

In a business environment increasingly reliant on various forms of interfirm cooperation, the role of trust as a foundation for effective economic exchange and development has been well established (e.g., Knack and Keefer 1997). The literature distinguishes between at least three perspectives on trust. The first focuses on the relationship, viewing trust as a willingness to rely on an exchange partner in whom one has confidence (Moorman, Deshpandé, and Zaltman 1993) or as “existing when one party has confidence in an exchange partner’s reliability and integrity” (Morgan and Hunt 1994, p. 23). The second perspective focuses on the characteristics of the trustee. This concept has generally been called “trustworthiness” (Mayer, Davis, and Schoorman 1995; Tullberg 2008) and, in short, suggests that some people are more likely to be considered trustworthy.

The third perspective, which is the focus of this study, centers on a person’s predisposition to trust others: that is, trust propensity. We note that others have used different labels for this concept, such as dispositional trust (Kramer 1999), generalized trust (Nee, Holm, and Opper 2018), social trust (Bergh and Öhrvall 2018), or interpersonal trust (Rotter 1967). Trust propensity is an important concept because the characteristics of the trustor play a stronger role than those of the trustee in developing trust, especially in the early stages of a relationship (Jones and Shah 2016).

Rotter (1967) conducted some of the foundational research into what he called “interpersonal trust,” or the general expectation that the word, promise, verbal or written statement of another could be relied on. Following this seminal research, subsequent studies established that some people have a greater tendency than others to be trusting (Colquitt, Scott, and LePine 2007). This perspective views trust propensity as a general disposition that a person would carry from one specific situation to another.

Trust Propensity Across Cultures

Empirical research on the relationship between trust propensity and individualism–collectivism has been inconclusive. Some have found evidence of a positive relationship between collectivism and trust propensity (Zeffane 2017). Others have found evidence of a negative relationship (Huff and Kelley 2003), and others still have found no relationship between trust and collectivism (Torres and Bligh 2012).

Some of this empirical ambiguity may be due to the frequent use of what we consider a psychometrically suboptimal way to operationalize trust propensity. The WVS includes a single item, which researchers have used as a proxy for trust propensity: “Generally speaking, would you say that most people can be trusted or that you can’t be too careful in dealing with people?” The dichotomous response categories were “Most people can be trusted” and “Can’t be too careful.” Drawing on this question, researchers have calculated the percentage of people within a country responding that most people can be trusted (Johnson and Cullen 2002). Most psychometric guidelines would advise against drawing firm conclusions of data using a single item with a dichotomous response category. Given the shortcomings of this measure, it is perhaps not surprising that the results are inconsistent across the different waves of the WVS. Indeed, over the seven waves of data collected by the WVS Association from 1984 to 2020, the correlation between the WVS trust item and Hofstede’s collectivism is sometimes positive and sometimes negative and ranges from .14 to −.56. Furthermore, depending on which wave is being analyzed, a variety of countries that are quite different from each other (e.g., China, Sweden, Saudi Arabia) have had the highest level of agreement with the WVS trust question. Thus, empirical analysis based on WVS data is psychometrically suboptimal, and empirical findings would be more credible if combined with other studies using alternative measures of trust propensity.

In addition to ambiguous empirical results, a review of the literature reveals conflicting perspectives on how individualism–collectivism is expected to relate to trust propensity. Given the lack of empirical consensus and theoretical counterarguments, we review both perspectives.

Collectivism and Trust Propensity

The individualism–collectivism value reflects the relationship between individuals and society as being either individual-focused or group-focused. “Collectivism” refers to a society in which people from birth onward are integrated into strong, cohesive in-groups that protect them throughout their lives in exchange for unquestioning loyalty. In contrast, “individualism” refers to a society in which the ties between individuals are loose. Everyone is expected to look after themselves and their immediate family (Hofstede, Hofstede, and Minkov 2010). Individualist values stress personal responsibility and achievement, and individuals are self-oriented rather than group-oriented. In collectivist societies, individuals are integrated into cohesive in-groups, and group goals and norms outweigh personal goals and attributes in guiding behavior (Triandis 1995).

Collectivist societies have been described as trust-based societies where reciprocity norms and mutual interdependence govern relationships (Hofstede 1980). For example, Japan has been used as an illustration of a country that emphasizes trust and cooperation (Hagen and Choe 1998), epitomized by the Japanese concept of Wa, which emphasizes sharing, cooperation, warmth, and fellowship (Rohlen 1974). It has been suggested that through an institutional environment that fosters goodwill, trust, and cooperation, Japanese firms are able to lower transaction costs and generate higher relational rents compared with Western counterparts (Dyer and Singh 1998). Similarly, when conflicts arise in societies with strong collectivist values, conflicts should preferably be worked out through mutual discussion or mediation. The use of legal documents and enforcements is evidence of mistrust, which tends to be a strong contrast with more litigious individualist societies (Sullivan and Peterson 1982). In effect, trust serves as a foundation for social and business relationships. In contrast, individualist societies tend to be dominated by arm’s-length relationships (Samaha, Beck, and Palmatier 2014). Members of individualist societies generally desire independence from any sort of group affiliation, and relationships are formed primarily with a self-serving purpose (Steensma et al. 2000). Combined, these arguments suggest a positive relationship between collectivism and trust propensity.

However, several theorists have advanced a competing perspective. For example, Huff and Kelley (2003) suggest that aspects of collectivist culture inhibit trust and that collectivists are relatively ineffective with strangers. Yamagishi (2011, p. 1) opens his book on trust with the claim that “the collectivist society produces security but destroys trust" [emphasis added]. Rather than trust, collectivist societies depend on a system of mutual monitoring and social sanctioning to enforce compliance with social norms. Yamagishi suggests that this collective system of mutual sanctioning guarantees mutual cooperation in tight-knit groups, but it leads to insecurity in a broader social environment where such a system does not exist. An example of a public policy initiative that demonstrates this perspective is the recent development of a social credit system in China. It is designed to punish untrustworthy behavior (e.g., spreading false information, breaking traffic rules) and reward trustworthy behavior (e.g., helping neighbors) (Kostka 2019). Such a system goes beyond the informal institution of social norms by introducing formal guidelines and punishment and could be viewed as an indicator of low social trust and the necessity for formal social regulation to provide assurances. This competing explanation for cross-national differences in trust propensity has been referred to as “cultural tightness,” which suggests that strong social norms and sanctions compel others to act in predictably reliable ways and that violating trust would be punished with social sanctions (Gelfand et al. 2011). The social norms and sanctions argument has also been advanced as a theoretical explanation linking collectivism and trust propensity (Hagen and Choe 1998). However, at the same time, cultural tightness has been argued as distinct from collectivism (Triandis 1989) and has been found to be only moderately correlated with collectivism (Carpenter 2000). For example, Gelfand, Nishii, and Raver (2006) identify Brazil as being collectivist but culturally loose and Germany as individualist but culturally tight.

We propose that the focus on harmony and collective good embodied by the value of collectivism suggests that collectivism has a positive relationship with trust propensity. The compliance and predictably reliable behavior found in culturally tight societies is less reflective of trust and more reflective of coercive compliance as a remedy to low trust. In summary, trust seems to be more instrumental to the functioning of society in collectivist cultures, suggesting that trust propensity is higher in collectivist societies.

H₁: There is a positive relationship between collectivism and trust propensity.

Social Projection of Individual-Level Values

Collectivism is not only a country-level variable. There is an individual-level equivalent to individualism–collectivism often referred to as independent versus interdependent self-construal (Markus and Kitayama 1991), which has been investigated as such in the international marketing literature (e.g., Hui, Ho, and Wan 2011). Importantly, individuals from an individualist heritage (e.g., North American, Western European) are more likely to have an independent self-construal; whereas individuals with a background originating from a collectivist heritage (e.g., East Asian, Latin American) have a tendency to have an interdependent self-construal (Fischer and Poortinga 2012; Peterson and Barreto 2018; White, Argo, and Sengupta 2012).

Accepting that collectivism is a value that can be held at the individual level, social projection then is a theoretical lens that could offer valuable insights into how collectivism influences trust propensity and account for trust variation within cultures. For example, in collectivist cultures social projection theory would account for people with low trust propensity by virtue of their individual-level self-construal and whether it was independent (individualist) or interdependent (collectivist). Social projection theory has a long history in social psychology, and its key thesis is the tendency to assume that others are similar to oneself; in other words, people use their own characteristics as the basis for judging others’ characteristics (Overbeck and Droutman 2013). Katz and Allport (1931) introduced the term “social projection” after finding that students who admitted to cheating on an exam were more likely to expect that others were also cheating. Assigning similar attributes or attitudes toward others (i.e., “most other people cheat”) as one holds for oneself is a way to reaffirm and justify one’s own actions (i.e., “it’s okay for me to cheat”). Social projection serves as a fundamental pillar of a functioning society and rests on the assumption that “humans continually perceive others and predict what these others think, feel, and, most importantly, what they will do” (Krueger 1998, p. 163).

Empirical evidence has shown that social projection tends to be strong regardless of whether people predict attitudes, behaviors, or personality traits. For example, people who support a particular political candidate are more likely to believe that most other people also support the same candidate (Van Boven, Judd, and Sherman 2012). People’s own behavioral preferences for competition or cooperation predict whether they think others prefer competition or cooperation (Dawes, McTavish, and Shaklee 1977). Social projection extends to values in general and to individualism–collectivism in particular. For example, Overbeck and Droutman (2013) found that, with respect to collectivism values, participants showed social projection onto others’ beliefs. In other words, the participant’s collectivism value, which we refer to as one’s own collectivism, resulted in a belief that others also held collectivism values, which we refer to as the perceived collectivism of others. In summary, social projection appears to be a well-established phenomenon and emerging evidence suggests that it extends to cultural values.

Applying social projection theory to the context of cultural values and trust propensity, we suggest that people who hold collectivism values will tend to socially project those same values onto others, creating a perceived collectivism consensus. Whether the other person actually shares the collectivism value is irrelevant. The important factor is the perception and belief that the other person shares the collectivism value. While one’s own degree of collectivism may indicate a greater propensity to trust others, it is logical that the perception of whether others share that collectivism value plays a role. Indeed, trust propensity is the belief by the trustor that the trustee will act benevolently (e.g., Lewicki and Wiethoff 2000). If the trustor believes the trustee has a shared concern for the group, then the trustor would feel more confident of the trustee’s intentions (i.e., that the trustee would act in an honest and fair manner). Thus, social projection helps us understand the mechanism that connects collectivism with trust propensity.

H₂: The positive relationship between one’s own collectivism and trust propensity is mediated by the perceived collectivism of others.

Sample

Study 1 offers a test of H₁ with a large sample of respondents from multiple countries. The sample was obtained in collaboration with the X-Culture project (Taras 2019). X-Culture is a global team project with participants from universities around the world. As part of the project, participants completed a questionnaire, which included the individual-level trust propensity construct. We gathered trust propensity data from 6,429 participants from 44 countries. We only included data from countries for which at least 20 responses had been collected, a quantity consistent with other international marketing research estimating multilevel models with respondents nested within countries (e.g., Steenkamp and Geyskens 2006). This reduced our sample to 6,326 responses from 36 countries. ¹ Table 2 reports the sample characteristics with average age of 24 years (SD = 5.24) and nearly even gender split (51.4% female). ² This matched-sample approach, in which participants share key characteristics (e.g., young, college educated) while differing primarily on nationality, helps control for nuisance variables and is similar to sampling techniques used by Schwartz (teachers and students) and Hofstede (IBM employees).

OPEN IN VIEWER

Table 2.

Sample Characteristics.

Characteristic	Study 1		Study 2 (U.S.)		Study 3 (U.S.)		Study 3 (China)		Study 4 (U.S.)
	Frequency	%	Frequency	%	Frequency	%	Frequency	%	Frequency	%
Total n	6,326		185		106		100		211
Countries	36		1		1		1		1
Gender
Male	2,832	48.6	99	53.5	37	34.9	68	68.0	102	51.7
Female	2,996	51.4	86	46.5	69	65.1	32	32.0	109	48.3
Age (years)
18–29	5,497	86.9	183	98.9	75	70.8	39	39.0	74	35.1
30–39	629	9.9	2	.2	21	19.8	43	43.0	53	25.1
40–49	155	2.5	0	.0	8	7.5	11	11.0	30	14.2
50–59	41	.6	0	.0	2	1.9	6	6.0	34	16.1
60 +	4	.1	0	.0	0	.0	1	1.0	20	9.5
Education
Some high school	N.A.		0	.0	1	.9	0	.0	1	.5
High school diploma	N.A.		26	14.1	11	10.4	8	8.0	15	7.1
Some college	N.A.		158	85.4	20	18.9	29	29.0	34	16.1
Bachelor degree	N.A.		1	.5	45	42.5	54	54.0	110	52.1
Graduate degree	N.A.		0	.0	29	27.4	9	9.0	51	24.2
Annual Household Income
<$24,000	N.A.		17	9.2	15	14.2	11	11.0	15	7.2
$24,000–$44,999	N.A.		4	2.2	42	39.6	34	34.0	13	6.3
$45,000–$74,999	N.A.		17	9.2	25	23.6	31	31.0	50	24.0
$75,000–$120,000	N.A.		55	29.7	16	15.1	17	17.0	58	27.9
>$120,000	N.A.		92	49.7	8	7.6	7	7.0	72	34.1

Notes: Income band cutpoints for Chinese sample in RMB are 25,000, 75,000, 125,000, and 175,000.

Measures

We measured trust propensity with a four-item construct adapted from Mayer and Davis (1999). The items include the following: (1) “Most people can be counted on to do what they promise to do,” (2) “Most people can be trusted,” (3) “Most people are honest about their skills and abilities,” and (4) “People are generally good and are trying their best.” The measure’s reliability was strong (α  = .88), and we used an average trust propensity index in the mixed model.

To ensure robustness of results, we merged the individual-level data with cultural values data from three separate values frameworks: Hofstede, Hofstede, and Minkov (2010), Schwartz (1994), and Minkov (2018). For the analysis based on Hofstede, data were not available for four countries in our sample, which reduced the sample size to 6,112 from 32 countries. We complement the analysis using the cultural framework based on Schwartz. At the national, cultural level, Schwartz developed a circular model of seven cultural values. However, he further condensed the values into two dimensions: conservatism and mastery. The conservatism dimension measures the extent to which individuals focus on the group versus the self and is conceptually similar to what has generally been referred to as individualism–collectivism (Shao, Kwok, and Guedhami 2010). Unavailability of Schwartz data for a few countries led to a final sample size of 5,889 from 28 countries for the Schwartz-based analysis. Finally, Minkov has provided a recently updated cultural values framework based on new data gathered from 53,000 respondents across 56 countries. Minkov attempted to replicate Hofstede’s framework, but his analysis only found support for the existence of individualism–collectivism and long-term orientation as coherent cultural dimensions. He suggests that power distance is a subdimension of collectivism and that masculinity and uncertainty avoidance lack coherence. Thus, we conduct a third analysis based on Minkov’s data with a sample size of 5,830 from 28 countries.

Control Variables

For each cultural values framework, we control for the other dimensions suggested by the framework. ³ We also control for two individual-level sociodemographic variables (age and gender). Finally, we control for two country-level economic differences that may affect trust propensity: societal inequality, measured by the Gini index, and gross domestic product (GDP) per capita (Bergh and Bjørnskov 2014).

Model Estimation

The sample data is multilevel with respondents nested within countries. Applying ordinary least squares regression to multilevel data is inappropriate, because the resulting estimates will be biased and the estimated standard errors of the effects will be too small (Aitkin, Anderson, and Hinde 1981). Hierarchical linear modeling (HLM), or mixed modeling, has been developed to deal with multilevel data. It enables the simultaneous estimation of relationships of variables at two (or more) levels, using iterative maximum likelihood estimation. Following best-practice guidelines, we centered the respondent-level predictors (Level 1) within countries, and grand-mean-centered the country-level predictors (Level 2) (Raudenbush and Bryk 2002).

Results

Table 3 reports descriptive characteristics and correlations between variables, and results of the HLM model are presented in Table 4. We report unstandardized coefficients because standardized coefficients are problematic due to variance being partitioned across different levels. We first estimated the null model using the full sample of 6,326 observations. The intercepts for trust propensity vary significantly across countries (Wald Z = 3.70, p < .001), and the intraclass correlation suggests that 10.48% of the variance in trust propensity is due to between-country differences. Next, we estimated a series of cross-level main effects models to observe differences in trust propensity using three cultural value frameworks. We began with Hofstede, Hofstede, and Minkov’s (2010) framework, which revealed a positive relationship between collectivism and trust propensity (b = .008, p < .001). The results are consistent using Schwartz’s (1994) conservatism index (i.e., collectivism), which has a positive effect on trust propensity (b = .389, p = .02), and the findings are consistent when using scores provided by Minkov et al. (2017) (b = .002, p = .02). Combined, these findings suggest consistent support for H₁.

OPEN IN VIEWER

Table 3.

Study 1 Correlation Matrix and Descriptive Statistics.

	1	2	3	4	5	6	7	8	9	10	11	12	13
1. Trust propensity	1
2. Conservatism (S)	.15**	1
3. Mastery (S)	.04**	.48**	1
4. Collectivism (H)	.20**	.50**	.02	1
5. Masculinity (H)	.03*	.25**	.34**	−.11**	1
6. Uncertainty avoid (H)	.04**	−.24**	−.70**	.43**	−.18**	1
7. Long-term orient (H)	.02	−.47**	−.28**	.01	−.35**	.18**	1
8. Collectivism (M)	.16**	.84**	.29**	.76**	.23**	.10**	−.29**	1
9. Flexibility (M)	−.06**	−.44**	.07**	−.57**	−.17**	−.37**	−.63**	−.66**	1
10. GDP per capita	−.14**	−.49**	−.04**	−.89**	.12**	−.38**	.08**	−.79**	.54**	1
11. Gini	.01	.54**	.59**	.16**	.41**	−.32**	.75**	.46**	−.57**	−.09**	1
12. Age	−.05**	.07**	.11**	−.06**	−.02	−.14**	.05**	−.01	.05**	.00	.05**	1
13. Gender (male = 1)	−.05**	−.06**	−.01	−.09**	−.06**	−.06**	−.04**	−.10**	.09**	.06**	−.05**	.04**	1
Mean	3.72	3.66	4.02	38.63	56.25	60.64	38.48	4.49	−12.30	36,901	43.94	24.08	.51
SD	.70	.27	.15	30.03	11.48	18.61	21.86	70.60	83.98	19,484	6.74	5.24	.48

*p < .05.

**p < .01.

Notes: H = Hofstede framework, S = Schwartz framework, M = Minkov framework. Collectivism variable estimates are the reverse-coded results for the Hofstede and Minkov individualism indices for consistency with the other studies.

OPEN IN VIEWER

Table 4.

Study 1 HLM Results.

	Analysis Based on Hofstede				Analysis Based on Schwartz				Analysis Based on Minkov
Individual N	6,112				5,889				5,830
Number of Countries	32				28				28
Predictor	Estimate	SE	t	p	Estimate	SE	T	p	Estimate	SE	t	p
Intercept	4.184	(.433)	9.65	.00	1.111	(1.165)	.95	.35	3.861	(.301)	12.82	.00
Collectivism (H)	.008	(.002)	4.79	.00
Masculinity (H)	.004	(.002)	2.09	.05
Uncertainty avoid (H)	−.002	(.002)	−1.34	.19
Long-term orient (H)	.000	(.001)	.19	.84
Conservatism (S)					.389	(.157)	2.48	.02
Mastery (S)					.364	(.261)	1.40	.17
Collectivism (M)									.002	(.000)	2.54	.02
Flexibility (M)									.001	(.000)	2.43	.02
GDP per capita	.000	(.000)	1.35	.19	.000	(.000)	.15	.88	.000	(.000)	−.29	.77
Gini	−.004	(.007)	−.49	.63	−.004	(.007)	−.55	.58	.000	(.006)	.02	.98
Gender (male = 1)	−.035	(.018)	−1.94	.05	−.028	(.018)	−1.50	.13	−.029	(.018)	−1.59	.11
Age	−.003	(.002)	−1.73	.08	−.003	(.002)	−1.80	.07	−.003	(.001)	−1.68	.09

Notes: H = Hofstede framework, S = Schwartz framework, M = Minkov framework. Collectivism variable estimates are the reverse-coded results for the Hofstede and Minkov individualism indices for consistency with the other studies.

With respect to control variables, the Hofstede values of uncertainty avoidance and long-term orientation and Schwartz’s mastery values are nonsignificant. Hofstede’s masculinity value (b = .004, p = .05), and Minkov’s flexibility dimension (b = .001, p = .02) are significant. Women have marginally higher trust propensity, when analyzed with the available data used in the Hofstede analysis, but not with the slightly smaller data sets used in the Schwartz and Minkov analyses. None of the other covariates, age, GDP per capita, or Gini index are significant. In summary, across all three values frameworks, we find consistent evidence that collectivism is positively related to trust propensity.

The presence of strong cultural norms and sanctions referred to as cultural tightness (Gelfand, Nishii, and Raver 2006) has been advanced as an alternative to collectivism as an explanation for cross-national differences in trust propensity (Yamagishi 2011). Gelfand et al. (2011) measured cultural tightness with items such as “In this country, there are very clear expectations for how people should act in most situations,” and “In this country, if someone acts in an inappropriate way, others will strongly disapprove.” See Web Appendix A for the full list of items.

To test cultural tightness as an alternative explanation, we reestimated the multilevel model including the most recent country-level cultural tightness scores reported in Eriksson et al. (2021), which included scores for 57 countries. Matching that list of countries to those in our data produced smaller data sets to analyze in terms of number of countries (Hofstede: 24, Schwartz: 23, Minkov: 23) compared with the results in Table 4. When including cultural tightness in the multilevel model of these smaller samples, the effects of collectivism on trust propensity remain significant across all three cultural frameworks (Hofstede: b = .010, t = 4.65, p < .001; Schwartz: b = .459, t = 2.34, p = .03; Minkov: b = .002, t = 2.22, p = .04). In contrast, cultural tightness is nonsignificant when estimated alongside the Hofstede framework (tightness: b = −.125, t = −1.06, p = .30), the Schwartz framework (tightness: b = −.091, t = −.60, p = .55) and the Minkov framework (tightness: b = −.259, t = −1.31, p = .21). We also estimated a model that included tightness without any other values scores. This resulted in 24 countries with scores, and the model produced results consistent with the previous model. Cultural tightness was not a significant predictor of trust propensity (b = .092, t = .70, p = .49).

Discussion

The finding that collectivism is positively related to trust propensity provides some clarity to a research question that has long plagued the field. Reliance on the WVS, with its single trust question and dichotomous response options, inadequately addresses the important question of how cultural values influence trust propensity. Importantly, we also tested a competing theoretical explanation, cultural tightness, and the findings based on the analysis of a large sample of individual respondents nested within countries on a multiple-item trust propensity measure make a meaningful contribution to this literature by highlighting the role of collectivism on trust propensity.

Although these findings have high external validity by employing country-level data on a large sample of participants and using multiple cultural values frameworks, the study is subject to certain limitations. Notably, the study is correlational in nature, leaving the possibility for the reverse relationship (i.e., a society consisting of individuals with high trust propensity fosters the development of the collectivism value). Further, the theoretical mechanism for how collectivism influences trust propensity is still unclear. We attempt to systematically address these limitations in the subsequent studies. Although we believe that the multi-item latent construct employed to measure trust propensity is superior to the WVS single item, it only measures people’s attitudes. Therefore, in Study 2, we add a behavioral outcome measure of trust propensity using the trust game. In Study 3, we prime collectivism and individualism values to establish causal evidence and test the social projection hypothesis, and finally in Study 4 we examine a managerially relevant outcome variable, trust in the firm, using advertising as the priming mechanism.

Participants and Procedure

We recruited 185 business student participants (53% male) and conducted the experiment in the behavioral lab of a U.S. university. In our version of the game, there is only one live person playing (the trustor), because we are concerned only with the initial trust propensity of the trustor. Participants were informed that the amount returned by the trustee would be determined by a virtual player whose responses are based on those of live players in prior experiments, specifically Berg, Dickhaut, and McCabe (1995) and Cox (2004); thus, the amounts returned reflected the decisions of actual participants from other studies. Participants were also informed that although they would be using play money during the experiment, 20% of them would be randomly selected to be paid in U.S. dollars in the amount as determined by the probability table. Random payment procedures have been used as cost-saving devices in other trust game studies (e.g., Cox 2009), and research suggests that random payment does not systematically affect behavior (Bolle 1990). Participants were asked whether they really believed that there was a chance that they could win some money. The mean response was significantly above the midpoint of the five-point scale (M = 3.91, t = 12.17, p < .001), which indicates that the game was perceived to have real behavioral consequences.

Participants first completed a computer questionnaire that explained the trust game process and offered two training exercises. Afterward, participants were taken to a breakout room, where a lab assistant repeated the instructions of the game. Participants were given $10 (play money) and asked to insert any amount of money into an envelope to send to the trustee. Immediately following, we asked participants to complete a questionnaire that included the collectivism items. Collectivism was measured with a four-item scale based on Yoo, Donthu, and Lenartowicz (2011); items are listed in Web Appendix A. The scale demonstrated good reliability (α = .86), and we averaged all items to create a collectivism index.

Results

The mean amount of money participants inserted into the envelope to be entrusted was $6.94. The most frequent amounts entrusted were $5 and $10, which is consistent with other trust game studies (Berg, Dickhaut, and McCabe 1995; Cox 2004). To estimate the effect of collectivism on trust propensity, we conducted regression analysis using the amount of money participants inserted into the envelope as the dependent variable and the participant’s collectivism index as antecedent. The results indicate a significant effect of collectivism on the amount of money entrusted (b = .34, t = 2.71, p < .01) and remain consistent when controlling for age, gender, and income.

Discussion

Study 2 results offer evidence of a relationship between collectivism at the individual level and an alternative measure of trust propensity. Instead of a self-report of whether others can be trusted, a rigorous behavioral measure was used, implying that collectivism influences individual-level behavior and underscoring the relevance of understanding the role of collectivism. Combined, Studies 1 and 2 offer correlational evidence of the effect of collectivism on trust propensity. To provide robust evidence and causally isolate the effect of a specific value, it is also advisable to attempt to isolate the effect of values and offer more direct evidence of causation. To accomplish this goal, in Studies 3 and 4, we make the value temporarily accessible through cultural priming (Oyserman and Lee 2008), which allows us to test for causal evidence.

Participants and Procedure

We conducted two separate experiments in two distinct cultural environments as a check on robustness and generalizability of the manipulation. The first experiment was conducted in the United States, where we recruited 100 U.S. participants from Amazon Mechanical Turk (MTurk) consumer panel. The second experiment was conducted in China, where we recruited 106 Chinese participants using WeChat, a multipurpose mobile application used for messaging, social media, and mobile payments. A limited number of messaging groups were invited to participate in the survey and be entered into a lottery for a nominal financial award (150 RMB). The use of crowdsourced samples, such as MTurk and WeChat, have been deemed suitable for experimental research, especially when multiple sources are used (Hulland and Miller 2018). Table 2 provides complete sample characteristics for both experiments.

We employed the same one-way between-subjects analysis of variance (ANOVA) design (individualism vs. collectivism) in both experiments. Thus, to examine the effect of collectivism on trust propensity in an experimental setting, we prime individualism–collectivism at the individual level, as has been done in other research (Oyserman and Lee 2008). Participants were randomly assigned to one condition and completed the priming task (see Web Appendix B), which was an adapted version of the similarities and differences with family and friends task (Trafimow, Triandis, and Goto 1991). Participants in the collectivism condition described three things they had in common with their family and friends and then described an experience when they sacrificed something for the good of the group. Conversely, participants in the individualism condition described three things that made them unique compared with their family and friends and then described an experience when they accomplished a goal independently.

We measured trust propensity (dependent variable), perceived collectivism of others (mediator), and collectivism of self (manipulation check) with four-item scales. Trust propensity (U.S.: α = .90; CN: α = .83) was measured with the same scale as in Study 1, based on Mayer and Davis (1999). We created a new scale to measure the perceived collectivism of others by modifying the scale used to measure one’s own collectivism (U.S.: α = .93; CN: α = .91). One’s own collectivism was measured with the same scale based on Yoo, Donthu and Lenartowicz (2011) as in Study 2. Items demonstrated good reliability (U.S.: α = .88; CN: α = .90) and were averaged to create a collectivism index. All items and experimental procedures were pretested in a pilot study ⁴ with U.S. participants, and all items are listed in Web Appendix A.

Results

First, to check whether the priming task was effective, we compared the two groups on collectivism using ANOVA. As expected, participants assigned to the collectivism condition scored higher on collectivism than those assigned to the individualism condition in both the U.S. sample (M_collectivism = 4.88, M_{individualism} = 3.95; F(1, 98) = 10.76, p = .001) and the Chinese sample (M_collectivism = 5.23, M_{individualism} = 4.36; F(1, 104) = 10.24, p = .002). With respect to trust, participants in the collectivism condition scored higher on trust propensity than participants in the individualism condition for both the U.S. sample (M_collectivism = 3.75, M_{individualism} = 3.23; F(1, 98) = 7.60, p < .01) and the Chinese sample (M_collectivism = 3.68, M_{individualism} = 3.40; F(1, 104) = 4.28, p = .04). These results offer causal evidence and support for H₁. Participants’ perceptions of the collectivism of others were also greater in the collectivism condition than in the individualism condition for both the U.S. sample (M_collectivism = 4.73, M_{individualism} = 4.01; F(1, 98) = 5.35, p = .02) and the Chinese sample (M_collectivism = 4.89, M_{individualism} = 4.32; F(1, 104) = 5.01, p = .03).

To test whether social projection mediates the relationship between collectivism and trust propensity, we used Model 4 of the Hayes (2017) PROCESS macro to detect any significant indirect effects. The experimental condition (individualism–collectivism prime) was modeled as the focal antecedent, and the respondent’s estimate of others’ collectivism as the mediator. We also included age and gender as covariates. Results reported in Table 5 indicate a significant effect of the condition (individualism = 0, collectivism = 1) on the mediator, perceived collectivism of others, in both the U.S. sample (b = .72, t = 2.26, p = .03) and the Chinese sample (b = .46, t = 1.83, p = .07). When trust propensity is regressed onto both the condition and the perceived collectivism of others, the condition is marginally significant in the U.S. sample (b = .34, t = 1.88, p = .06) and nonsignificant in the Chinese sample (b = .10, t = .86, p = .39). More importantly, the respondent’s perceived collectivism of others is significant in both the U.S. sample (b = .25, t = 4.41, p < .01) and the Chinese sample (b = .28, t = 6.21, p < .01). A 90% bootstrap confidence interval based on 5,000 bootstrap samples for the indirect effects (U.S.: b = .18; CN: b = .13) is entirely above zero for the United States (.040, .360) and China (.003, .271), indicating a significant indirect effect of collectivism on trust propensity.

OPEN IN VIEWER

Table 5.

Study 3: Indirect Effect of One’s Own Collectivism on Trust Propensity.

	Model 1: Perceived Collectivism of Others			Model 2: Trust Propensity
	b	t	p	b	t	p
U.S. Sample (n = 100)
Condition (COL = 1)	.72	2.26	.03	.34	1.88	.06
Perceived COL of others				.25	4.41	<.01
Control Variables
Age	−.00	−.00	.99	−.00	−.46	.65
Gender (male = 1)	.02	.06	.96	.05	.27	.79
R2 = .23
Indirect effect	b = .18 (CI: [.040, .360])
Chinese Sample (n = 106)
Condition (COL = 1)	.46	1.83	.07	.10	.86	.39
Perceived COL of others				.28	6.21	<.01
Control Variables
Age	.05	3.02	<.01	.00	.47	.64
Gender (male = 1)	.04	.15	.88	−.31	−2.66	.01
R2 = .36
Indirect effect	b = .13 (CI: [.003, .271])
Combined Sample (U.S. and China) (n = 206)
Country (CN = 1)	.43	1.86	.06	−.07	−.61	.54
Perceived COL of others				.28	7.89	<.01
Control Variables
Age	.03	2.11	.04	−.00	−.46	.65
Gender (male = 1)	.12	.54	.59	−.12	−1.06	.29
R2 = .24
Indirect effect	b = .12 (CI: [.014, .237])

Notes: We estimated the indirect effect using 5,000 bootstrap samples to create a 90% confidence interval (CI). b = unstandardized regression coefficient. COL = collectivism. CN = China.

As a robustness test and to offer a more direct cross-cultural comparison, we repeated the mediation analysis using country as the focal antecedent instead of the experimental condition. Country (United States/China) is not a perfect proxy for individualism–collectivism, especially because half the sample in each country was primed with individualism and the other half collectivism. However, the United States and China are near cultural polar opposites with respect to individualism–collectivism and regression analysis using country (U.S. = 0, CN = 1) as the focal antecedent produced supportive results. The effect of country on the perceived collectivism of others was nearly significant (b = .43, t = 1.86, p = .06). When regressing trust propensity on both country and perceived collectivism of others, the effect of country weakened (b = −.07, t = −.61, p = .54), whereas the effect of perceived collectivism of others (b = .28, t = 7.89, p < .01) was strong. Further, the bootstrap confidence interval is entirely above zero (.014, .237), supporting the indirect effect.

Discussion

The experimental results from Study 3 offer causal evidence that is consistent with the findings from Studies 1 and 2. Combined, these studies offer strong support for the positive relationship between collectivism and trust propensity. Further, the results support the social projection hypothesis, which suggests that the effect of one’s own collectivism value indirectly influences trust propensity through one’s perception of the collectivism of others. In other words, people project their own level of collectivism onto others; then, believing that others share a benevolent concern for others, people are more likely to trust others.

Participants and Procedure

We recruited U.S. consumers for the experiment by employing graduate students trained in recruiting techniques to collect responses and direct participants to an online survey. The sample consisted of 211 participants (51.7% male; M_age = 38.6 years). Table 2 provides complete sample characteristics. To begin the experiment, participants were randomly assigned to view one of two different sets of advertisements (one emphasizing individualism values and the other collectivism values) to prime the participant’s level of collectivism. Each set included a print and video advertisement for a fictitious insurance company called CAPCO (see Web Appendix C). Participants viewed the video advertisement, followed by the print advertisement. We then measured the participant’s trust in the firm (α = .91), trust propensity (α = .82), the participant’s own level of collectivism (α = .68), and perceived collectivism of the firm as a manipulation check (α = .94). All measurement items are reported in Web Appendix A.

Results

First, we conducted an ANOVA to test whether participants who viewed the collectivism advertisements perceived the company to be positioned as more collectivist than participants who viewed the advertisements emphasizing individualist values. The results indicate that the advertisements were perceived as intended with respect to the perceived collectivism of the firm (M_{collectivism ads} = 4.74, M_{individualism ads} = 2.25; F(1, 209) = 500.43, p < .001).

To assess whether the collectivism- versus individualism-positioned advertisements had a significant effect on trust in the firm, we conducted a serial mediation analysis in SPSS using Model 6 of the PROCESS macro, which executes a series of regression models (Hayes 2017) (results are reported in Table 6). The first model results indicate that the condition (individualism ads = 0, collectivism ads = 1) had a significant effect on participants’ level of collectivism (b = .31, t = 2.91, p < .01). The second model revealed that participants’ collectivism significantly predicted their trust propensity (b = .17, t = 3.18, p < .01). More importantly, the final model, which assessed the effect on trust in the firm, implied the presence of serial mediation with a significant effect of trust propensity (b = .40, t = 4.99, p < .01) but nonsignificant effects of the condition (b = .14, t = 1.49, p = .14) and the participant’s collectivism (b = .02, t = .32, p = .75). We tested the significance of the serially mediated indirect effect of the condition (condition → collectivism → trust propensity → trust in the firm) using 5,000 bootstrap samples, which estimated an effect of b = .02 with a 95% confidence interval that did not include zero (.004, .047). Thus, the viewing of advertisements emphasizing collectivism values had a significant indirect effect on trust in the firm.

OPEN IN VIEWER

Table 6.

Study 4: Effect of Collectivism on Trust in the Firm.

	Model 1 Collectivism			Model 2 Trust Propensity			Model 3 Trust in the Firm
	b	t	p	b	t	P	b	t	p
Condition (collectivism ad = 1)	.31	2.91	<.01	.18	2.25	.03	.14	1.49	.14
Collectivism				.17	3.18	<.01	.02	.32	.75
Trust propensity							.40	4.99	<.01
Control Variables
Age	−.00	−.59	.56	.00	.26	.80	−.00	−1.11	.27
Gender (male = 1)	.20	1.90	.06	−.24	−2.90	<.01	−.20	−2.10	.04
R2 = .17
Indirect effect: condition → collectivism → trust propensity → trust in the firm	b = .02 (CI: [.004, .047])

Notes: We estimated the indirect effect using 5,000 bootstrap samples to create a 95% confidence interval (CI). b = unstandardized regression coefficient. Condition = individualism versus collectivism ad condition.

Discussion

Study 4 offers evidence of a causal relationship between collectivism and trust propensity, building on the correlational evidence from Studies 1 and 2 and adding to the causal evidence from Study 3. Further, an indirect effect on trust in the firm was observed. Notably, the level of participants’ collectivism was the result of a priming induced by positioning in advertisements, which could be used as practical guidance for managers to enhance consumer trust in their firm.

Theoretical Implications

Prior research on the drivers of trust propensity often centered on personality traits (e.g., Alarcon et al. 2018), and efforts to examine cultural value influences on trust propensity are characterized by competing perspectives and conflicting results. The theoretical perspective we adopt is that collectivism’s focus on harmony and collaboration with others is associated with trust propensity through the mechanism of social projection. Others have suggested that individualism can have a positive effect on trust development (Doney, Cannon, and Mullen 1998), and still others suggest that the strength of social norms and sanctioning in societies, a concept Gelfand, Nishii, and Raver (2006) terms “cultural tightness,” is a predictor of trust propensity.

The results observed in the multilevel model with data from 6,326 participants from 36 countries suggest significant cross-cultural differences with respect to trust propensity influenced by the collectivism value. The data offer evidence of the relationship; however, we also recognize that significant within-country variance on values is common. So, although there appear to be cross-cultural differences, just as with cultural values, it is only on the average that trust propensity varies from country to country, and it cannot be assumed that any particular individual from a given country has more or less trust propensity. Although cultural tightness is a conceptually appealing predictor of trust, our multilevel model based on a large sample of individuals nested within countries suggests that cultural tightness does not predict trust propensity. Instead, the evidence suggests a positive relationship between collectivism and trust propensity, helping clarify the role of collectivism with respect to competing perspectives.

The evidence for the relationship between collectivism and trust propensity appears particularly strong for several reasons. First, not only does this investigation establish correlational evidence of the relationship, but, by experimentally priming the collectivism value, it offers causal evidence of the relationship. Second, trust propensity was measured several ways. Instead of relying on the psychometrically weak WVS single-item dichotomous measure, we assessed trust propensity using a multi-item measure, which may explain some inconsistencies in previous research; in the trust game in Study 2, we used a behavioral measure of trust propensity.

In our view, perhaps the most impactful theoretical implication revolves around independent versus interdependent self-construal, the individual-level manifestation of individualism versus collectivism. We successfully activated the collectivism value experimentally even when using advertising messages as the stimuli. The consistent findings at the societal and individual levels across these studies offer additional support for the assertion by Oyserman and Lee (2008) that some cultural values can be investigated at the individual level using experiments, as has been done with individualism–collectivism (White, Argo, and Sengupta 2012), power distance (Gao, Winterich, and Zhang 2016), and long-term orientation (Bearden, Money, and Nevins 2006). In so doing, we established causal evidence of the relationship between collectivism and trust propensity as well as identified the potential for marketers to do the same to induce trust propensity.

Social projection offers a new lens through which to view the collectivism–trust propensity relationship. Social projection clarifies how collectivism influences propensity to trust others. A person’s own collectivism value, through the process of social projection, creates a perceived collectivism consensus between the self and others. Assuming a shared concern for the welfare of others, the trustor develops a greater trust propensity. It is important to note that a perceived collectivism consensus is merely a perception and does not necessarily reflect reality; it can sometimes be a false consensus (Ross, Greene, and House 1977). In other words, the perception of others to some degree merely reflects one’s own preferences. Although the prevalence of collectivism itself is predictive of trust propensity, Study 3 offers evidence that it is the perceived collectivism of others that drives trust. Indeed, in the U.S. sample the direct effect of collectivism is rather weak, and it is not close to significance in the Chinese sample when the mediator perceived collectivism of others is included in the model.

We focus on a trust-inducing value; however, other values and attitudes that potentially facilitate or hinder trust development could also be socially projected. It is not simply the perceived consensus that facilitates trust; rather, it is the perceived consensus in a trust-inducing value that develops trust propensity. Indeed, in some of the earliest research on social projection, students who cheated believed that the majority of other students cheated as well (Katz and Allport 1931). Thus, people who have a disposition toward dishonesty would be likely to believe that others are dishonest as well, therefore hindering trust development.

Managerial Implications

Implications for managers are highlighted primarily in the results of Study 4, which tested the theoretical model using a priming process that more closely resembles reality and measured a downstream consequence of trust propensity (i.e., trust in the firm). Trust in the firm has become increasingly important. According to a 2019 Gallup survey, the view of corporations as being corrupt is a majority view worldwide (Khoury and Crabtree 2019). Thus, understanding one of the antecedents to trust in the firm is a step in the right direction for firms to address the problem.

The priming task from Study 4 highlights a potentially useful tool for firms. Priming values in research studies has typically been accomplished by asking study participants to complete a cognitive task such as arranging words in an order that produces logical sentences or writing about an experience identifying similarities and differences with family and friends, as employed in Study 3. Study 4 demonstrated that the collectivism value could also be activated by asking study participants to view advertisements. Because this procedure more closely resembles how firms communicate with buyers, it implies that trust in the firm may be influenced indirectly through promotional messaging that stresses collectivism values.