The Reputational Consequences of Generalized Trust

Abstract

The present research examines the reputational consequences of generalized trust. High-trust individuals are seen as moral and sociable, but not necessarily competent. When controlling for other traits, there is a negative relationship between trust and perceived competence (Studies 1 and 2). Compared with optimism, generalized trust has stronger effects on morality and sociability (Study 2). Furthermore, people judge those who do not discriminate between trustworthy and untrustworthy groups (unconditional trustors) more negatively than those who only trust groups that are, in fact, trustworthy (conditional trustors). Unconditional trust and unconditional distrust are both viewed negatively (Study 3), even after controlling for attitudinal similarity (Study 4). Critically, both generalized trust and discriminant ability (i.e., conditional trust) have independent reputational benefits (Study 5). These studies suggest that generalized trust plays an important role in how we perceive and judge others.

How do people judge those who are generally inclined to trust others? Is dispositional trust viewed as an indicator of benevolence, social sophistication, and intelligence, or are there negative interpersonal consequences for trusting too much? Dispositional trust is an important topic in psychology (Thielmann & Hilbig, 2015) and economics (Knack & Keefer, 1997; Zak & Knack, 2001). Most prior research has focused on trust as an outcome, asking when and why people engage in trust behavior (Evans & Krueger, 2016). Fewer studies have considered how people evaluate the trait of trust in others. This is surprising, as scholars have proposed that generalized trust has important consequences in dyadic relationships (Rotter, 1967) and societal functioning (Zak & Knack, 2001). We contribute to this literature by examining the information that generalized trust conveys about the motives and abilities of social agents.

The present research has three objectives: First, we ask how generalized trust influences perceptions of morality, sociability, and competence (Goodwin, Piazza, & Rozin, 2014; Landy, Piazza, & Goodwin, 2016). Second, we ask to what extent the trait of trust reveals distinct interpersonal information compared with the related construct of optimism (Scheier, Carver, & Bridges, 1994). Third, we ask how people judge conditional trustors, those who accurately discriminate between trustworthy and untrustworthy groups. Are conditional trustors judged more (or less) positively than those who show trust unconditionally? The present research emphasizes the important reputational consequences of generalized trust.

The Generalized Propensity to Trust

Rotter (1967) defined the general propensity to trust as “an expectancy held by an individual or group that that the word, promise, verbal or written statement of another person or group can be relied upon” (p. 651). This definition is motivated by social learning theory, which posits that generalized expectations are accumulated through interactions with various social agents (Rotter, 1960). Individuals learn through experience that others can be relied upon (or not), and apply this expectation when interacting with new social entities.

The correlates of generalized trust are heterogeneous: Five-factor models of personality define trust as related to agreeableness (Costa, McCrae, & Dye, 1991). Other studies also found that the trait of trust is positively correlated with extraversion and negatively correlated with neuroticism (Evans & Revelle, 2008), and there is also a link between dispositional trust and intelligence (Carl & Billari, 2014; Hooghe, Marien, & de Vroome, 2012; Sturgis, Read, & Allum, 2010). In addition, generalized trust shapes interpersonal behavior—it is positively correlated with social success (Rotter, 1967), and the willingness to help others (Peysakhovich, Nowak, & Rand, 2014; Yamagishi et al., 2013).

Perceptions of Morality, Sociability, and Competence

Previous studies found that generalized trust is associated with positive outcomes. However, to our knowledge, no study has systematically addressed its reputational consequences. When we learn that another person is generally inclined to trust others, what inferences do we draw about this person’s character? This question is important for understanding the dynamic role that the propensity to trust plays in close relationships (Simpson, 2007) and organizations (Kramer, 1999). We contend that perceptions of generalized trust shape personality inferences and expectations of future behavior.

In understanding the reputational consequences of generalized trust, we focus on three fundamental dimensions in person perception—morality, sociability, and competence (Brambilla, Sacchi, Rusconi, Cherubini, & Yzerbyt, 2012; Goodwin, 2015).¹ These three dimensions have far-reaching effects in both interpersonal and intergroup contexts (Brambilla & Leach, 2014). Moral individuals have positive intentions toward others, they are honest and good-natured; sociable individuals have interpersonal skills and are able to gain support from others, they are friendly and extraverted; and competent individuals are intelligent and efficient. In other words, morality refers to an individual’s intention to help (or harm) others, and sociability and competence refer to the individual’s various abilities to enact those intentions (Landy et al., 2016).

Perceptions of Trust Behavior

Although not much is known about the consequences of dispositional trust, recent work based on the economic investment game (Berg, Dickhaut, & McCabe, 1995) has asked how people perceive trust behavior. The investment game takes place in two stages and involves two players, the trustor and the trustee: In the first stage, the trustor has an endowment (US$10) that she can keep or send to the trustee. In the second stage of the game, any sent money is tripled by the experimenter and the trustee can then return some, all, or none of the tripled money to the trustor. The first player’s decision to send money is interpreted as a measure of trust because it involves accepting an uncertain outcome based on expectations of the other player’s behavior (Rousseau, Sitkin, Burt, & Camerer, 1998). The second player’s response is viewed as a measure of trustworthiness. It involves no vulnerability or uncertainty; instead, the trustee has a choice about whether to uphold the norm of reciprocity (Gouldner, 1960).

How do people perceive behavior in the investment game? To answer this question, Krueger, Massey, and DiDonato (2008) asked participants to judge the behavior of hypothetical players in both game roles. Those who trusted (e.g., invested money in the first stage of the game) were seen as more moral than those who kept the initial endowment, and trust behavior was judged to be as moral as trustworthy behavior (e.g., returning money in the second stage of the game). Those who trusted were also seen as more competent; however, this effect was relatively small compared with the effect of trust on morality, and the analyses did not control for a potential correlation between morality and competence.

Consistent with the idea that trust behavior is seen as relevant to morality, a program of research on principled trustfulness has argued that people feel a social obligation to trust in strangers (Dunning, Anderson, Schlösser, Ehlebracht, & Fetchenhauer, 2014; Schlösser, Fetchenhauer, & Dunning, 2016; Schlösser, Mensching, Dunning, & Fetchenhauer, 2015). This line of work suggests that people do not trust in others for instrumental reasons (i.e., the expected gain from the interaction), but rather they trust because they feel it is the right thing to do in the moment. Refusing to trust may be seen as personally insulting or even harmful (Evans & van Beest, 2017; Pillutla, Malhotra, & Murnighan, 2003).

Yet, not all findings are consistent with the idea that trust behavior is perceived as morally relevant: Bicchieri, Xiao, and Muldoon (2011) examined to what extent people believe that untrusting and untrustworthy behavior should be punished. People use punishment to signal moral disapproval (Jordan, Hoffman, Bloom, & Rand, 2016; Raihani & Bshary, 2015); hence, punishment decisions provide some insight into how observers perceive the morality of different behaviors. Bicchieri and colleagues (2011) asked participants to read a description of the investment game and then asked them whether they would punish (i.e., spend money to reduce the payoff of) players who did not send money in the first stage of the game or return money in the second stage. Participants indicated that they would punish untrustworthy, but not untrusting, behavior, indicating that trust behavior is not always perceived as a moral issue.

It is also unclear if people trust primarily because it is what they feel they should do. Consistent with the idea that trust is not exclusively moral, Thielmann and Hilbig (2017) found that ratings of what people want to do were a stronger predictor of trust behavior compared with feelings of obligation. These findings point to the idea that trust is also influenced by instrumental concerns (see also Evans & Krueger, 2016)—people trust, in part, because they hope to gain something from the interaction. The benefits may be direct (the wealth gained from a loan or investment) or indirect (a new relationship or a positive reputation). Given these conflicting findings, further research is needed to fully understand to what extent trust behavior is linked with perceptions of morality.

From Trust Behavior to Generalized Trust

This research focuses on the concept of trust as a generalized trait (Evans & Revelle, 2008; Rotter, 1967) as opposed to a specific behavior (Berg et al., 1995). Surprisingly, there is a relatively weak relationship between the trait of trust and trust behavior (Dunning et al., 2014; Glaeser, Laibson, Scheinkman, & Soutter, 2000; Johnson & Mislin, 2012). Trust behavior, as it is typically measured in the investment game, is associated with an array of dispositions, such as generalized prosociality (Kanagaretnam, Mestelman, Nainar, & Shehata, 2009) and individual risk-taking propensity (Evans & Krueger, 2011). Indeed, Dunning, Fetchenhauer, and Schlösser (2012) found that a majority of participants interpreted the investment game in terms of risk-taking and gambling (though they also believed the game was related to trust). Trust behavior is also influenced by situational factors, such as changes in economic payoffs (Evans & Krueger, 2011; Van de Calseyde, Keren, & Zeelenberg, 2017) and the social identities of the trustor and the trustee (Balliet, Wu, & De Dreu, 2014). Thus, previous studies that focused on how people perceive trust behavior may be confounded by inferences about related traits (e.g., risk-taking propensity) and the specific features of the situation (e.g., the potential payoffs associated with choosing trust). As such, behavioral and dispositional trust may have divergent reputational consequences.

Previous research has argued that generalized trust is important for both individuals and societies (Van Lange, 2015), but few studies have addressed the trait’s interpersonal consequences. Our work measures the effects of generalized trust on perceptions of morality, sociability, and competence. Note that prior work did not disentangle how the propensity to trust uniquely relates to these related dimensions; for example, people may think of trusting individuals as moral, or as friendly but ultimately self-interested. The present work has implications for our understanding of the downstream consequences of generalized trust, and our knowledge of how people form impressions of others.

Overview of Studies

We conducted five studies to investigate the effects of generalized trust on perceptions of morality, sociability, and competence: In Study 1, participants evaluated the personality and character traits of various acquaintances. Study 2 used experimental vignettes to compare perceptions of the traits of trust and optimism (Scheier et al., 1994). Studies 3 and 4 asked how people perceive conditional trust. We compared perceptions of those who trust unconditionally (i.e., people who believe that all groups are trustworthy) with those who are conditional and careful in their propensity to trust (i.e., people who accurately discriminate between trustworthy and untrustworthy groups). Finally, Study 5 tested the interactive effects of generalized trust and conditional trust on person perception.

We report how we determined our sample sizes, all data exclusions (if any), all manipulations, and all measures in our studies. No data collection occurred after data analysis. All of our raw data, syntax, and study materials are available on the Open Science Framework at osf.io/wpy3a.

Study 1

In our first study, we used an acquaintance rating task (Landy et al., 2016) to examine the effects of generalized trust on perceptions of morality, sociability, and competence. We consider three central hypotheses: First, we ask if generalized trust is associated with perceptions of morality. Work on principled trustfulness posits that people perceive trust in moral terms (Dunning et al., 2014), but other findings suggest that trust occurs for more self-interested reasons (Evans & Krueger, 2016; Thielmann & Hilbig, 2017). Second, we ask if generalized trust is associated with perceptions of sociability. Previous studies have found that high-trust individuals are more popular and well-liked (Rotter, 1967, 1980), which suggests that dispositional trust may be associated with better social functioning.² Third, we ask if there is a relationship between dispositional trust and perceived competence—Studies have found that there is an actual relationship between intelligence and generalized trust (Sturgis et al., 2010), but it is unclear if people perceive this relationship.

Method

Participants

We recruited 163 first-year students (138 women) from the psychology bachelor at Tilburg University. Our sample size was based on the number of participants we could recruit in 1 week. Participants received course credit as compensation; the experiment was conducted online using Qualtrics; and instructions were provided in English. Two participants quit the study during the task and were therefore excluded from our analyses, final N = 161.

Materials and procedure

Participants completed an acquaintance rating task (Landy et al., 2016). Following the procedure used by Landy and colleagues (2016), each participant rated the personality of four targets: a person they respected, a person they liked, a person they did not respect, and a person they did not like. We selected this task to generate targets with varying levels of perceived morality, sociability, and competence. Participants were instructed to think of four distinct individuals, and the order of the four targets was randomized for each participant. The rating task consisted of three pages per target: On the first page, participants were instructed to visualize the target and enter the target’s initials; on the second page, participants evaluated the dispositional trust of the target; on the third page, participants evaluated the morality, sociability, and competence of the target.

Generalized trust measure

We used three items from the General Trust Scale (Yamagishi & Yamagishi, 1994) to measure the extent to which each target was perceived as a trusting individual: (a) This person thinks that most people are basically honest. (b) This person thinks that most people are basically good-natured and kind. (c) This person generally trusts others. For each item, participants were asked to judge how accurate it was in describing the target (1 = extremely accurately; 5 = not accurately at all). The three items were reliable for each of the four targets: α_like = .89, α_respect = .89, α_{do not like} = .89, α_{do not respect} = .83.

Morality, sociability, and competence ratings

Eighteen items (from Landy et al., 2016) were used to measure perceptions of the targets’ morality (moral, principled, honest, trustworthy, fair, and responsible; αs > .82), sociability (sociable, warm, friendly, easygoing, extraverted, and playful; αs > .71), and competence (competent, capable, intelligent, effective, skillful, and talented; αs > .86). The items were presented in a randomized order, and for each trait participants were asked to rate to what extent the target possessed each trait (e.g., 1 = not at all moral, 9 = moral).

Additional measures

Participants also provided several ratings that were unrelated to this report. They answered several questions about the cognitive style of each target (e.g., whether the target was an intuitive or reflective thinker), and indicated the target’s gender.

Results and Discussion

Perceptions of morality, sociability, and competence

To account for the clustered nature of the data, we used the lme4 package in R (Bates, Maechler, Bolker, & Walker, 2014) and we used the lmertest package to estimate p values (Kuznetsova, Brockhoff, & Christensen, 2015).

Our first set of analyses focused on the effects of trust on perceptions of morality, sociability, and competence. We estimated simple multilevel models, with trust as predictor and perceptions of morality, sociability, and competence entered as dependent variables; random-intercepts were estimated for each participant and each target. Individuals who were seen as trusting of others were judged to be more moral, B = .28, confidence interval (CI) = [.18, .40], p < .001, and more sociable, B = .39, CI = [.30, .49], p < .001.³ There was, however, no significant zero-order relationship between trust and perceived competence, B = .01, CI = [–.10, .13], p = .82. See Figure 1 for an illustration of these results.⁴

Figure 1.

The effects of generalized trust on perceptions of morality, sociability, and competence (Study 1).

Note. High-trust individuals were seen as more moral and sociable, but not more competent.

We also estimated models to test if the effects of trust were different for different types of targets: There were large main effects of target type. Not surprisingly, the two positive targets (i.e., the liked and respected targets) were seen as more moral, sociable, and competent than the two negative targets (i.e., the disliked and disrespected targets). In addition, perceptions of morality were influenced by a significant trust by target-valence (positive vs. negative) interaction, B = –.23, p = .04. Trust was associated with perceived morality for both negative (i.e., disliked and disrespected) and positive (i.e., liked and respected) targets. However, the relationship was stronger for negative targets (B = .45, p = .01) than it was for positive targets (B = .14, p = .01). There were no other significant interactions involving target type.

The preceding analyses did not account for the potential correlations of morality, sociability, and competence. Therefore, we estimated models that included the corresponding dimensions as covariates. For example, we tested the relationship between trust and perceived morality while also controlling for sociability and competence. The results of these analyses are reported in Table 1. Trust was still associated with perceived morality (p < .001) and sociability (p < .001); interestingly, there was also a significant negative relationship between trust and perceived competence (p < .001).

Table 1.

Trust and Perceptions of Morality, Sociability, and Competence (Study 1).

Fixed effects	Morality			Sociability			Competence
	B	CI	p	B	CI	p	B	CI	p
Intercept	1.38	[–0.05, 2.63]	.13	3.16	[2.39, 4.04]	<.001	2.52	[2.15, 2.85]	<.001
Trust	0.26	[0.15, 0.35]	<.001	0.36	[0.27, 0.47]	<.001	−0.22	[–0.31, –0.11]	<.001
Morality	—	—	—	0.08	[0.00, 0.16]	.053	0.55	[0.50, 0.60]	<.001
Sociability	0.07	[–0.01, 0.14]	.058	—	—	—	0.19	[0.11, 0.26]	<.001
Competence	0.53	[0.47, 0.60]	<.001	0.22	[0.14, 0.30]	<.001	—	—	—
Random effects	SD	—	—	SD	—	—	SD	—	—
Participant	<0.01	—	—	.33	—	—	.36	—	—
Target	1.34	—	—	.69	—	—	<.01	—	—

Note. CI = confidential interval.

Summary

Generalized trust was positively associated with both morality and sociability. Individuals who were perceived to be trusting of others were seen as both more moral and more sociable, and both of these findings were robust when controlling for the correlations between morality, sociability, and competence. Interestingly, trusting individuals were not seen as more competent, and in fact generalized trust was negatively correlated with perceived competence when we controlled for morality and sociability as covariates. This pattern suggests that morality and sociability (which were positively correlated with both trust and competence) suppress the negative relationship between trust and competence (MacKinnon, Krull, & Lockwood, 2000).

Study 2

Why do people believe that high-trust individuals are moral and sociable? One possible explanation is that people have a general preference toward positivity in others. Prior work on kill-the-messenger effects suggests that actors who express positive or optimistic opinions are judged more positively than those who share negative or pessimistic views (Tesser & Rosen, 1975; Walther, Nagengast, & Trasselli, 2005). Hence, high-trust individuals may be judged positively because people generally prefer those who hold positive beliefs. To test this possibility, our second study asks whether people have different perceptions of trust (Rotter, 1967) and optimism (Scheier et al., 1994). If people prefer trustors because of a general bias in favor of positivity, then the traits of trust and optimism should have similar effects on person perception. However, if trust conveys unique social information (above and beyond preferences for positivity) then trust should have stronger effects on perceptions of morality and sociability (compared with optimism).

Method

We recruited 219 U.S. participants on Mturk. The average age was 33.24 (SD = 9.93) and there were 84 women. Participants were paid 60 cents and the study took less than 5 min to complete. Our planned sample size was based on 80% power to detect a small- to medium-sized effect (f = .20) using a 2 × 2 design, minimum N = 199. This effect size estimate was based on the smallest significant zero-order effect observed in Study 1 (B = .27). Power analyses were conducted using G*Power 3 (Faul, Erdfelder, Buchner, & Lang, 2009).

At the beginning of the study, participants read a short description of Mark, a hypothetical Mturk worker. We manipulated whether Mark was described in terms of his dispositional trust in others (or optimism), and whether he had a high (or low) score in the corresponding trait (see Table 2 for exact descriptions). The trust description was based on item content from Yamagishi and Yamagishi’s (1994) trust scale; the description of optimism was based on items from Scheier and colleagues’ (1994) Life Orientation Test, a measure of dispositional optimism versus pessimism.

Table 2.

Descriptions Used to Manipulate Trust and Optimism (Study 2).

	Low	High
Trust	Mark is a very distrustful person. He believes that most people are basically dishonest and self-interested. He thinks that most people will take advantage of others when the opportunity arises.	Mark is a very trustful person. He believes that most people are basically honest and kind. He thinks that most people will respond in kind when they are trusted.
Optimism	Mark is a very pessimistic person. Overall, he expects that more bad things will happen to him than good. In uncertain times, he expects the worst. He is very negative about the future.	Mark is a very optimistic person. Overall, he expects that more good things will happen to him than bad. In uncertain times, he expects the best. He is very positive about the future.

After reading this description, participants evaluated the target on the 18 traits used in Study 1 (α_morality = .93, α_sociability = .95, α_competence = .90). Participants were asked to evaluate the accuracy of one statement related to each adjective, for example, “Mark is a moral person.” Participants responded using a scale from 1 (not at all accurate) to 7 (very accurate).

To test the strength of each manipulation, we also included two items to measure the overall impression of Mark: (a) My overall impression of Mark is very positive and (b) My overall impression of Mark is very negative; r(217) = –.79, p < .001.

Results and Discussion

The consequences of generalized trust

Our first set of analyses attempted to replicate Study 1’s findings on trust and perceptions of morality, sociability, and competence (Table 3). First, we tested the zero-order effects of generalized trust. We estimated simple linear regressions using level of trust (high vs. low) to predict morality, sociability, and competence. Note that these analyses only included participants from the two trust conditions. High-trust individuals were seen as more moral (p < .001), more sociable (p < .001), and more competent (p < .001) than low-trust individuals.

Table 3.

Trust and Perceptions of Morality, Sociability, and Competence (Study 2).

	Morality						Sociability						Competence
	B	CI	p	B	CI	p	B	CI	p	B	CI	p	B	CI	p	B	CI	p
Intercept	4.74	[4.55, 4.91]	<.001	1.42	[0.56, 2.28]	.001	3.92	[3.76, 4.08]	<.001	1.37	[0.57, 2.18]	<.001	4.40	[4.24, 4.57]	<.001	1.63	[0.86, 2.40]	<.001
Trust	2.06	[1.69, 2.42]	<.001	0.72	[0.11, 1.32]	.02	2.96	[2.64, 3.28]	<.001	2.15	[1.74, 2.56]	<.001	0.73	[0.40, 1.06]	<.001	−0.79	[–1.34, –0.23]	.005
Morality				—	—	—				0.30	[0.13, 0.47]	<.001				0.38	[0.22, 0.54]	<.001
Sociability				0.34	[0.15, 0.53]	<.001				—	—	—				0.24	[0.06, 0.43]	.008
Competence				0.44	[0.25, 0.63]	<.001				0.25	[0.06, 0.44]	<.001				—	—

Note. CI = confidential interval.

We also conducted analyses controlling for the correlations between these three outcomes (following the procedure described in Study 1): high-trust individuals were still seen as more moral (p = .02) and more sociable (p < .001). Importantly, we also replicated the significant negative relationship between trust and perceived competence (p = .005).

These results were consistent with the findings of Study 1, with the exception that we observed a (comparatively small) zero-order relationship between trust and perceived competence. However, we again found that morality and sociability suppressed the negative relationship between trust and perceived competence. When we controlled for the correlations between morality, sociability, and competence, there was a negative relationship between trust and perceived competence.

Comparing generalized trust and optimism

Our next analyses asked to what extent people made different inferences from information about dispositional trust versus optimism. We estimated a series of regressions with morality, sociability, competence, and overall impression entered as dependent variables (see Table 4). The predictors were trait (optimism = –.5; trust = +.5), trait-score (low = –.5, high = +.5), and a trait by trait-score interaction term. The full analyses are reported in Table 4 and illustrated in Figure 2.⁵

Table 4.

The Effects of Optimism and Trust on Impression Formation (Study 2).

	Morality			Sociability			Competence			Overall impression
	B	CI	p	B	CI	p	B	CI	p	B	CI	p
Intercept	4.86	[4.68, 5.04]	<.001	4.15	[3.99, 4.31]	<.001	4.62	[4.45, 4.78]	<.001	4.67	[4.43, 4.92]	<.001
Trait (optimism vs. trust)	−0.13	[–0.38, 0.12]	.31	−0.23	[–0.45, –0.003]	.046	−0.21	[–0.44, 0.02]	.07	−0.26	[–0.60, 0.09]	.14
Score (low vs. high)	0.99	[0.65, 1.35]	<.001	2.53	[2.21, 2.85]	<.001	1.07	[0.75, 1.40]	<.001	2.63	[2.14, 3.12]	<.001
Trait × Score interaction	1.06	[0.57, 1.56]	<.001	0.43	[–0.02, 0.87]	.062	−0.35	[–0.81, 0.11]	.14	0.30	[–0.38, 0.99]	.38

Note. CI = confidential interval.

Figure 2.

Study 2 compared the effects of optimism and trust on perceptions of character.

Note. Trust had stronger effects on perceptions of morality and sociability compared with optimism. In our box plots, the solid lines show median values, the upper and lower edges of each box indicate the upper and lower quartiles, respectively, and individual dots signify outliers.

Morality

Perceptions of morality were influenced by a significant trait by trait-score interaction, p < .001. Given the significant interaction, we estimated the simple effects of trait-score (low vs. high) for trust and optimism. Optimism was associated with greater perceived morality, B = .99, CI = [.66, 1.34], p < .001; however, there was an even stronger positive relationship between trust and perceived morality, B = 2.06, CI = [1.70, 2.43], p < .001.

Sociability

Perceptions of sociability were also influenced by a marginally significant trait by trait-score interaction, p = .062. To understand this interaction, we estimated the simple effects of trait-score (low vs. high) for trust and optimism. Optimism was associated with greater sociability, B = 2.53, CI = [2.22, 2.85], p < .001; however, there was a slightly stronger positive relationship between trust and sociability, B = 2.95, CI = [2.64, 3.28], p < .001.

Competence

Trust and optimism had similar effects on perceptions of competence; there was no significant trait by trait-score interaction, p = .14. Instead, there was a significant main effect of trait-score, such that trusting and optimistic individuals were both perceived as more competent, p < .001 (but see our analyses from the previous section). There was also a marginal effect of trait (optimism vs. trust), such that those in the trust conditions were seen as somewhat less competent compared with participants in the optimism conditions, p = .072.

Overall impression

Finally, to test the strength of each manipulation, we tested the effects of generalized trust and optimism on overall evaluations. There was no significant trait by trait-score interaction, p = .38. Instead, there was a significant main effect of trait-score, such that trusting and optimistic people were judged more positively, p < .001. Finally, there was no significant effect of trait, p = .14. Importantly, the lack of a significant interaction term suggests that the two manipulations were comparable in strength.

Summary

As in our first study, we found further evidence that generalized trust was associated with perceptions of morality and sociability. Importantly, these two effects remained significant while controlling for the correlations between morality, sociability, and competence. We also replicated the finding that morality and sociability suppress a negative relationship between trust and perceived competence. Although there was a positive zero-order correlation between trust and competence, this relationship became significantly negative when we controlled for morality and sociability.

We also compared the perceptions of trust and optimism: Dispositional trust and generalized optimism were both associated with positive evaluations—trusting and optimistic individuals were both judged more positively. Consistent with previous studies, we observed that people positively evaluate those who express positive beliefs (Tesser & Rosen, 1975; Walther et al., 2005). Critically, trust and optimism had distinct effects on perceptions of morality and sociability. We observed two interactions such that trust had stronger positive effects on perceptions of morality and sociability (compared with the effects of optimism). Although the pattern of results was similar for perceptions of morality and sociability, the difference between trust and optimism was most pronounced for perceptions of morality. These results support the idea that trust is uniquely associated with morality (Dunning et al., 2014) and its reputational consequences go beyond mere positivity.

Study 3

Our third study focuses on the question of how people perceive conditional trustors, those who only express trust in groups that are generally known to be (or at least, believed to be) trustworthy. Studies 1 and 2 found that individuals who are high in generalized trust (those who trust strangers and “most people”) are seen as moral and sociable, but lacking competence; now, we ask whether there are limits to how far trust should be extended. Is it better to be blindly trusting of others, or do people also value the ability to discriminate between trustworthy and untrustworthy groups? We compared conditional trustors with two groups, those who show trust unconditionally and those who show distrust unconditionally.

Balance theory suggests that people evaluate actors based on whether they agree with how the actors evaluate other targets (Crandall, Silvia, N’Gbala, Tsang, & Dawson, 2007; Heider, 1958). Discriminative ability plays an important role in how we judge others: Observers react negatively when an actor evaluates positive targets negatively, or when an actor evaluates negative targets positively (Gawronski & Walther, 2008). This perspective suggests that conditional trustors should be judged more positively than those who are unconditionally trusting of others. Observers should be similarly critical of those who are unconditionally positive and those who are unconditionally negative toward others (Gawronski & Walther, 2008).

Method

Pretest

We conducted an initial study to identify trustworthy and untrustworthy groups to use as stimuli. American Mturk workers (N = 41) rated the trustworthiness of 12 professions from a recent Gallup survey on the perceived ethical standards of different American professions (“Honesty/Ethics in Professions,” 2014). We preselected six trustworthy professions (nurses, doctors, high school teachers, police officers, clergy, and funeral directors) and six untrustworthy professions (advertisers, stockbrokers, members of Congress, lobbyists, car salesmen, and telemarketers). Participants evaluated each profession on a scale from 0 (not trustworthy at all) to 100 (very trustworthy).

Reassuringly, the trustworthy professions were rated as much more trustworthy than the untrustworthy professions (trustworthy: M = 67.2, SD = 18.6; untrustworthy: M = 31.02, SD = 16.69), t(40) = 9.89, p < .001. Even the lowest rated trustworthy profession (funeral directors: M = 59.6, SD = 27.5) was still rated as much more trustworthy than the highest rated untrustworthy profession (stockbrokers: M = 40.0, SD = 22.2), t(40) = 4.7, p < .001.

Participants

We recruited 182 U.S. participants on Mturk. The average age was 32.3 (SD = 9.97) and there were 68 women. Participants were paid 30 cents and the study took less than 3 min to complete. Our planned sample size was based on 80% power to detect a medium-sized effect (f = .25) using a one-way ANOVA with three groups, minimum N = 159. We did not have a clear prediction about the expected effect size of our conditional trust manipulation. However, the effects of our experimental trust manipulation in Study 2 were relatively large; therefore, we considered a medium-sized effect to be a conservative estimate.

Materials and procedure

Participants were presented with information about another Mturk worker. They were told that we had recently conducted a study on the trustworthiness of different professions, and that they would learn about the responses of one person from this study, a worker named Mark.⁶ We then displayed a horizontal bar graph illustrating how much Mark trusted each of the 12 pretested professions. Mark’s ratings were displayed on a scale from 0 (not trustworthy at all) to 100 (very trustworthy).

In the conditional trust condition, Mark (the target) accurately discriminated between trustworthy and untrustworthy professions: Mark rated trustworthy professions positively (ratings from 60 to 76 out of 100) and rated untrustworthy professions negatively (ratings from 26 to 40). In the unconditional trust condition, Mark rated all professions positively (ratings from 60 to 76), and in the unconditional distrust condition, Mark rated all professions negatively (ratings from 24 to 40).

Then, participants evaluated Mark’s personality using three-item measures. They evaluated his morality (trustworthy, sincere, honest, α = .87), sociability (likeable, warm, and friendly, α = .90), and competence (capable, intelligent, reliable, α = .92). Participants rated the accuracy of each adjective on a scale from 1 (not at all accurate) to 7 (very accurate). We selected these shorter measures (as opposed to the six-item measures used in Studies 1 and 2) to minimize the duration of the survey.

Results and Discussion

We conducted a series of one-way ANOVAs to test the effect of experimental condition on perceptions of morality, sociability, and competence. Differences across condition are illustrated in Figure 3. In the event that an omnibus test was significant, we conducted pairwise comparisons of the three conditions using Tukey’s honestly significant difference (HSD) with a family-wise error rate of .05. We also ran the following analyses with covariates to control for the correlations between morality, sociability, and competence, but adding covariates did not change any of our results.

Figure 3.

Study 3 compared perceptions of conditional trust, unconditional trust, and unconditional distrust.

Morality

Perceptions of morality differed significantly across conditions, F(2, 180) = 21.3, p < .001, η² = .19. Perceptions of morality were higher in the conditional trust condition compared with the unconditional trust condition (difference = 1.02, p < .001) and the unconditional distrust condition (difference = 1.17, p < .001). Perceptions of morality were similar in the unconditional trust and unconditional distrust conditions, difference = .15, p = .73.

Sociability

Perceptions of sociability differed significantly across the three conditions, F(2, 180) = 8.76, p < .001, η² = .089. Perceptions of sociability were marginally higher in the conditional trust condition compared with the unconditional trust condition (difference = .42, p = .074) and significantly higher compared with the unconditional distrust condition (difference = .77, p < .001). Perceptions of sociability were similar in the unconditional trust and unconditional distrust conditions, difference = .34, p = .16.

Competence

To conclude, we examined the effects of condition on perceived competence: The pattern of means closely resembled the results for perceptions of morality: Competence differed across the three experimental conditions, F(2, 180) = 19.95, p < .001, η² = .18. Competence was higher in the conditional trust condition compared with the unconditional trust condition (difference = 1.17, p < .001) and the unconditional distrust condition (difference = 1.13, p < .001). Perceptions of competence were comparable in the unconditional trust and unconditional distrust conditions, difference = .04, p = .97.

Summary

Consistent with balance theory (Crandall et al., 2007; Heider, 1958), conditional trust was evaluated more positively than both unconditional distrust and unconditional trust. Conditional trustors, those who discriminated between trustworthy and untrustworthy professions, were seen as more moral and competent than unconditional trustors. Interestingly, people had similar (negative) impressions of unconditional trust and unconditional distrust. In this study, people did not differentiate between those who trusted too much and those who trusted too little; both groups were evaluated as less moral and less competent than those who expressed trust conditionally.

Study 4

In Study 3, participants rated conditional trust more positively than unconditional trust and unconditional distrust. An important question is whether observers’ perceptions of conditional trust can be explained by perceived similarity. In the previous study, the conditional trustor was a person who held average beliefs about the trustworthiness of different professions. This means that participants may have been more likely to hold beliefs about professions that were similar to those of the conditional trustor (compared with the other two conditions). To address this potential confound, we conducted a study where we measured and controlled for participants’ beliefs about the trustworthiness of different professions.

Method

Participants

We recruited 260 U.S. participants on Mturk. The average age was 34.3 (SD = 9.54) and there were 105 women. Participants were paid 85 cents and the study took less than 7 min to complete.

Our planned sample size was based on 80% power to detect a small- to medium-sized effect (f = .20) using a one-way ANOVA with three groups, minimum N = 246. We increased our sample size (compared with Study 3) to account for the possibility that controlling for similarity would reduce the strength of our conditional trust manipulation. Our smallest omnibus effect size in Study 3 was f = .30; we planned to have sufficient power to detect a comparable effect assuming that similarity explained up to one third of the effect of conditional trust.

Materials and procedure

The materials and procedure were identical to Study 3, except that we also asked participants to rate the perceived trustworthiness of the 12 professions used in the experimental manipulation. At the beginning of the survey, participants rated each profession on a scale from 0 (not trustworthy at all) to 100 (very trustworthy) using a slider bar. After providing their own trustworthiness ratings, participants then learned about Mark, who was presented as either a conditional trustor, an unconditional trustor, or an unconditional distrustor. As in Study 3, participants then rated Mark in terms of his perceived morality, sociability, and competence. We also clarified that the target (i.e., the worker named Mark) was hypothetical, meaning that the study was conducted without deception.

Results and Discussion

Conditional trust and similarity

First, we tested to what extent participants had similar (or dissimilar) beliefs to the targets presented in the three experimental conditions: To measure similarity, we calculated the absolute distance between each participant’s own trustworthiness ratings and the trustworthiness ratings of the target (i.e., Mark, the hypothetical Mturk worker). We then calculated an average absolute distance score for each participant (M = 25.79, SD = 9.52). A distance score of 0 would mean that the participant and the target had identical ratings; a score of 10 would mean that, on average, the participant’s trustworthiness ratings were 10 points away (on a 100-point scale) from the target’s ratings. Hence, a smaller distance score indicates greater similarity.

The average levels of distance were significantly different across the three experimental conditions, F(2, 257) = 49.6, p < .001, η² = .28. Given the significant omnibus test, we compared the three conditions using Tukey’s HSD with a family-wise error rate of .05: The absolute distance was lower (indicating greater similarity) in the conditional trust condition (M = 18.4, SD = 6.52), compared with the unconditional trust condition (M = 29.5, SD = 9.36), difference = −11.0, p < .001, and compared with the unconditional distrust condition (M = 28.9, SD = 8.14), difference = −10.55, p < .001. Levels of distance were comparable in the unconditional trust and unconditional distrust conditions, difference = –.50, p = .91. These results confirm that participants’ own beliefs were most similar to those of the conditional trustor.

Note that we also conducted analyses using a correlation-based measure of similarity. We estimated to what extent participants’ ratings of the different professions were correlated with the ratings of the target they were asked to evaluate, with a stronger correlation indicating greater similarity. This correlation-based approach produced similar results.

Similarity and perceptions of morality, sociability, and competence

Our primary analyses focused on how controlling for similarity would influence perceptions of conditional trustors. We conducted two sets of linear regressions: First, we tested the effects of experimental condition (conditional trust, unconditional trust, or unconditional distrust) on perceptions of character. In these models, the conditional trust condition was treated as a reference level, and dummy variables were created for the unconditional trust and unconditional distrust conditions. The results are reported in Table 5. Importantly, all of the results from Study 3 were replicated. Conditional trustors were seen as more moral, more sociable, and more competent than both unconditional trustors and unconditional distrustors.

Table 5.

Perceptions of Unconditional Trust and Distrust (Compared With Conditional Trust) While Controlling for Similarity (Study 4).

	Morality						Sociability						Competence
	B	CI	p	B	CI	p	B	CI	p	B	CI	p	B	CI	p	B	CI	p
Intercept	5.20	[4.97, 5.44]	<.001	5.43	[–0.08, 2.89]	<.001	5.04	[4.80, 5.28]	<.001	5.39	[5.00, 5.77]	<.001	5.11	[4.87, 5.35]	<.001	5.63	[5.25, 6.02]	<.001
Unconditional trust	−0.73	[–1.06, –0.40]	<.001	−0.59	[–0.97, –0.22]	.002	−0.40	[–0.74, –0.07]	.017	−0.20	[–0.58, 0.18]	.30	−1.00	[–1.34, –0.67]	<.001	−0.69	[–1.07, –0.32]	<.001
Unconditional distrust	−1.17	[–1.50, –0.85]	<.001	−1.04	[–1.41, –0.68]	<.001	−1.39	[–1.72, –1.06]	<.001	−1.19	[–1.56, –0.82]	<.001	−0.90	[–1.23, –0.56]	<.001	−0.60	[–0.97, –0.23]	.001
Distance				−0.012	[–0.028, 0.004]	.14				−0.018	[–0.035, –0.002]	.026				−0.028	[–0.04, –0.01]	<.001

Note. CI = confidential interval.

Next, we estimated models that included both experimental condition and similarity (i.e., the absolute difference in trustworthiness ratings) as predictors. When controlling for similarity, nearly all of the effects of experimental condition remained significant. There was one exception: There was no longer a significant difference between the perceived sociability of conditional trustors and unconditional trustors (p = .30). However, even after controlling for similarity in trustworthiness perceptions, conditional trustors were still seen as more moral and more competent than unconditional trustors.

Study 5

In Studies 3 and 4, we found that those who discriminated between trustworthy and untrustworthy groups (i.e., conditional trustors) were judged more positively than those who did not (i.e., unconditional trustors and unconditional trustors). These findings raise the question of whether discriminative ability is more important than generalized trust in person perception. Does discriminative ability supersede generalized trust in impression formation, or do both factors have independent effects in person perception? To examine this issue, in Study 5, we simultaneously manipulated conditional trust (i.e., the ability to discriminate between trustworthy and untrustworthy groups) and generalized trust (i.e., overall beliefs about whether groups are trustworthy).

Method

We recruited 220 U.S. participants on Mturk. The average age was 33.4 (SD = 9.97) and there were 91 women. Participants were paid 85 cents and the study took less than 7 min to complete. Our planned sample size was based on 80% power to detect a small- to medium-sized effect (f = .20) using a 2 × 2 ANOVA, minimum N = 199.

Materials and procedure

The materials and procedure were similar to Study 3, except that we simultaneously manipulated the target’s (i.e., Mark’s) levels of generalized trust and discriminative ability.

To manipulate generalized trust, we adjusted the target’s average rating across all 12 professions: In the low generalized trust conditions, Mark rated all professions as relatively untrustworthy (overall average = 31.25 out of 100); in the high generalized trust conditions, all professors were rated as trustworthy (average = 67.25).

To manipulate discriminative ability (i.e., conditional trust), we altered to what extent the target differentiated between trustworthy and untrustworthy professions. In the low discriminative ability conditions, the target rated the two types of professions as equally trustworthy. In the high discriminative ability conditions, the target rated trustworthy professions more positively than untrustworthy professions (average difference = 16 points out of 100).

Results and Discussion

We estimated a series of regressions with morality, sociability, and competence entered as dependent variables (see Table 6). The predictors were discriminant ability (low = –.5, high = +.5), generalized trust (low = –.5, high = +.5), and a discriminant ability by generalized trust interaction term. The full analyses are reported in Table 6.

Table 6.

The Effects of Discriminant Ability and Generalized Trust on Impression Formation (Study 5).

	Morality			Sociability			Competence
	B	CI	p	B	CI	p	B	CI	p
Intercept	4.86	[4.73, 4.99]	<.001	4.62	[4.45, 4.78]	<.001	4.77	[4.63, 4.91]	<.001
Discriminant ability	0.59	[0.33, 0.85]	<.001	0.44	[0.14, 0.74]	.004	0.50	[0.22, 0.77]	<.001
Generalized trust	0.51	[0.25, 0.77]	<.001	0.77	[0.47, 0.08]	<.001	0.50	[0.19, 0.75]	<.001
Discriminant × Generalized trust interaction	−0.24	[–0.77, 0.28]	.36	−0.067	[–0.67, 0.53]	.82	−0.72	[–1.28, –0.16]	.011

Note. CI = confidential interval.

Morality and sociability

There were significant main effects of discriminant ability and generalized trust on perceptions of morality and competence: Individuals who discriminated between trustworthy and untrustworthy groups were seen as more moral and more sociable, ps < .005, and individuals high in generalized trust were also seen as both more moral and more sociable, ps < .001.

Competence

Perceptions of competence were influenced by a significant discriminant ability by generalized trust interaction, p = .011. To understand this interaction, we estimated the simple effects of generalized trust within the low- and high-discriminant ability conditions. When the target had low discriminant ability (i.e., when Mark did not differentiate between trustworthy and untrustworthy groups), generalized trust was positively associated with competence, B = .82, CI = [.42, 1.23], p < .001. However, when the target displayed high discriminant ability, generalized trust had no significant effect on competence, B = .11, CI = [–.28, .50], p = .59. In other words, generalized trust only influenced perceptions of competence when participants judged targets that did not discriminate accurately between trustworthy and untrustworthy groups.

Summary

The present results suggest that both discriminant ability (i.e., conditional trust) and generalized trust influence person perception. When manipulated orthogonally, both factors had significant (positive) main effects on perceptions of morality and sociability. The findings for competence were more nuanced: Generalized trust was positively associated with competence, but only when the target was low in discriminant ability. For targets high in discriminant ability, there was no relationship between generalized trust and perceived competence. These findings fit with the broader pattern of results suggesting that the effects of generalized trust on perceived competence are fragile and inconsistent, at least compared with its effects on morality and sociability.

General Discussion

Five studies investigated the reputational consequences of generalized trust. High-trust individuals were perceived to be more moral and sociable, but not necessarily more competent (Studies 1 and 2). When controlling for other positive traits, there was a negative relationship between trust and perceived competence. Perceptions of trust also differed from perceptions of optimism—compared with optimism, generalized trust had stronger positive effects of perceived morality and sociability (Study 2).

Yet, people did not believe that trust should be applied indiscriminately. Rather, people believed that conditional trustors (those who accurately discriminated between trustworthy and untrustworthy groups) were judged more positively than those who trusted in others indiscriminately (Studies 3-4). Interestingly, both unconditional trust and unconditional distrust were judged negatively. That is not to say that discriminant ability fully supersedes trust in impression formation: When generalized trust and discriminant ability were manipulated orthogonally, both had positive reputational consequences (Study 5). Together, these results demonstrate that generalized trust has important reputational consequences.

Theoretical Implications

The question of how people perceive the trait of trust in others has implications for theories of trust and social perception. Our studies provide further support for the idea that trust in strangers is a moral issue (Dunning et al., 2014). Studies 1 and 2 found that high-trust individuals were seen as more moral, and this is consistent with the idea that people may trust in others because they feel it is what they should do. High-trust individuals were also seen as having better social skills and those skills can be used to pursue either selfish or selfless goals (Landy et al., 2016). We propose that trust is a moral, but not purely self-sacrificial, tendency. People may think of the propensity to trust as a socially rational belief that benefits both the trustor and the recipient of trust (Evans & Krueger, 2016).

We also found that the ability to accurately discriminate between trustworthy and untrustworthy groups is important. In Studies 3 to 5, individuals who expressed trust in untrustworthy social groups, such as lobbyists and members of Congress, were seen more negatively than those who accurately discriminated between trustworthy and untrustworthy groups (i.e., conditional trustors). Consistent with balance theory, those who believe that immoral groups are trustworthy may be seen as immoral themselves (Gawronski & Walther, 2008). Errors of unconditional trust and unconditional distrust were judged in similarly negative terms. Yet, our work goes beyond previous studies (e.g., Gawronski & Walther, 2008) in two important ways: First, we demonstrate that perceptions of conditional trustors cannot be explained purely by attitudinal similarity (Study 4). Second, we find that generalized trust and discriminative ability have independent effects on person perception (Study 5).

The finding that conditional trustors are viewed more positively than unconditional trustors also suggests an important limitation to models of social signaling, which suggest that people respond positively to those who trust and cooperate unconditionally (Hoffman, Yoeli, & Nowak, 2015). Previous work found people prefer to interact with partners who decide to act prosocially without calculating the costs and benefits of doing so (Jordan, Hoffman, Nowak, & Rand, 2016; Van de Calseyde, Keren, & Zeelenberg, 2014). Gestures of partial or incomplete trust can be seen as insulting to others (Falk & Kosfeld, 2006; Fehr & Rockenbach, 2003; Pillutla et al., 2003). However, people also believe that others should express some caution in deploying trust. It is important for future research to consider when people believe it is justified to withhold trust in others.

Are Perceptions of Generalized Trust Accurate?

An important question is whether participants’ beliefs about high-trust individuals reflect reality (Jussim, Crawford, & Rubinstein, 2015). Is generalized trust valid information to use when forming an impression of a stranger at zero-acquaintance? There is indeed evidence to suggest that high-trust individuals are more likely to be moral (Yamagishi et al., 2015) and sociable (Rotter, 1967). However, participants’ beliefs about trust and competence did not reflect the positive relationship between trust and general intelligence (Carl & Billari, 2014; Hooghe et al., 2012; Sturgis et al., 2010). When controlling for other positive traits, there was a negative relationship between trust and perceived competence, though we did observe a positive zero-order relationship between trust and competence in Study 2. This discrepancy may stem from the indirect mechanism through which intelligence fosters generalized trust; intelligent individuals are better at knowing the right times to trust in others, and through experience they have more social positive experiences; in turn, these positive experiences lead to higher generalized trust (Sturgis et al., 2010).

Limitations and Future Directions

In the present studies, we focused on perceptions of generalized trust, but we did not ask what leads an individual to be seen as generally trusting (or distrusting) of others. In real-world contexts, impressions of others’ traits may be based on inferences from observed behavior (Smith, Stewart, & Buttram, 1992). When considering the reputational consequences of trust behavior, it will be useful for future work to consider how perceptions change in different contexts. For example, the extent to which trust behavior is perceived as evidence of morality (vs. sociability) may depend on the stakes and the identity of the trustee. Trust behavior is more likely to be perceived in moral terms when it involves interactions with members of in-groups (Balliet et al., 2014) or when it involves the allocation of potential harms (Evans & van Beest, 2017; Van Beest, Van Dijk, De Dreu, & Wilke, 2005). On the contrary, people may see trust behavior as evidence of sociability when it conveys clear social benefits to the decision maker. There may also be specific cases where trust behavior is seen as a sign of competence; in particular, people may think of trust behavior as a sign of competence when it results in economic benefits for the trustor (Evans & Krueger, 2016) or when it is based on calculative reasoning (Wang, Zhong, & Murnighan, 2014).

In Studies 1 and 2, we found consistent evidence that generalized trust was associated with perceptions of morality and sociability. However, the relationship between generalized trust and perceived competence was more nuanced: There was a significant negative relationship between trust and competence, but only when we controlled for morality and sociability. We interpreted this pattern of results as evidence of statistical suppression—those who trust others were more likely to be seen as moral and sociable, and those who are moral and sociable were more likely to be seen as competent. Note that this pattern also explains why there were inconsistent zero-order correlations between trust and competence, as the zero-order correlation depends on the correlations between morality, sociability, and competence. Although the statistical evidence of suppression was consistent in Studies 1 and 2, the ideal test is an experimental study where generalized trust and the proposed suppressor variables are manipulated simultaneously.

We also observed circumstances where there were negative consequences for those who trust too much. In Studies 3 and 4, unconditional trustors were judged to be less moral and less competent than conditional trustors. Scholars have speculated about the potential downsides of trust (Hardin, 2002; Rotter, 1980), and recent empirical work suggests that trust can have negative consequences for both individuals and organizations. For example, trust and cooperation may serve as foundations for bribery (Jiang, Lindemans, & Bicchieri, 2015) and corrupt collaboration (Weisel & Shalvi, 2015). Further work is needed to better understand when trust leads to negative impressions and harmful organizational outcomes.

Conclusion

How do people judge those who give (or withhold) trust in others? We found that those who believe others are trustworthy are seen as moral and sociable, but not necessarily competent. Yet, the reputational benefits of dispositional trust are contingent on the assumption that trust is only given to deserving targets; those who trust indiscriminately are judged negatively. The present work suggests that trust (and how it is deployed) has important downstream consequences, and shapes how we form impressions of others’ intentions and abilities.