Abstract
Trust is among the most critical factors in buyer–supplier relationships. In an effort to understand the origins of trust, power has become the focus of a burgeoning body of literature. However, research on the association between power and trust has been plagued by inconsistencies in terms of whose power and trust are being examined, which has led to confusion and hindered cumulative progress. We address this issue by disentangling the effect of the focal organization (actor effect) from the effect specific to its partner (partner effect) and accounting for both simultaneously. We further theorize and show that the partner's level of self-promotion communication about his or her own achievements and credentials moderates both actor and partner effects, thus adding knowledge about a contingency that accounts for a considerable degree of variation in the linkage between power and trust. Using multi-informant, dyadic survey data paired with archival information scraped from firms’ webpages, we find that (i) actors low (vs. high) in power tend to place more trust (ii) while simultaneously eliciting higher levels of trust from their partners; however, (iii) these effects differ markedly depending on the partner's level of self-promotion communication. Our study offers a novel, integrative perspective on power and trust, and we elaborate on its important implications for understanding buyer–supplier relationships.
Introduction
Trust is critical for achieving successful outcomes in buyer–supplier relationships. Prior research has shown that trust significantly enhances buyer–supplier relationships by supporting, for example, resilience (Kaufmann et al., 2018), satisfaction (Nyaga et al., 2013), and compliance with environmental standards (Villena et al., 2021). Accordingly, researchers have endeavored to identify relevant antecedents to trust and isolate critical conditions that facilitate or hinder its development in buyer–supplier relationships (e.g., Anderson et al., 2022; Brinkhoff et al., 2015). In particular, an extensive literature in operations and supply chain management (OSCM) addresses the important question of the extent to which power may shape trust in buyer–supplier relationships (Chen et al., 2016; Handley et al., 2019).
Trust broadly refers to one party's willingness to accept vulnerability to the actions of another party (Mayer et al., 1995; Özer and Zheng, 2017; Schilke et al., 2021), while power is defined as the degree to which one party in a relationship is perceived to have control over resources valued by the other party (Cook et al., 2006; Fiske and Berdahl, 2007). Power and trust are often regarded as two of the most important concepts in supply chain relationships (Ireland and Webb, 2007; Kaufmann et al., 2023), but there has been limited cumulative progress in understanding their relationship. In particular, the extant literature has been plagued by confusion that has led researchers to assess extant findings about the influence of power on trust as “inconsistent” (Zhao et al., 2008: 373) and “mixed” (Nyaga et al., 2013: 47). We argue that an important reason explaining this problem is the considerable ambiguity about whose power and whose trust are being theorized and investigated, which can be traced back to distinct intellectual traditions that have largely evolved in parallel in the OSCM literature. Prior research has either been ambiguous about their focal unit of analysis or has taken the exclusive vantage point of either the focal organization (e.g., buyer and client) or of the partner (e.g., supplier and manufacturer) in isolation. This observation is consistent with the “single-party blind spot” that Lumineau and Oliveira (2018) highlighted as a major limitation in research on interorganizational relationships. This limitation has led to misunderstandings of how each partner's power can influence its respective trust.
To address these limitations in prior research, we first take a step back to clarify the source of the problem. Extant research has typically built on the problematic assumption that power is a zero-sum game insofar as a focal organization's high power invariably means that the partner is low in power. This approach is misguided because it cannot account for the common situation in which both parties perceive both themselves and their partner as either high or low in power (Brito and Miguel, 2017; Reimann and Ketchen, 2017). Our article thus eschews this zero-sum fallacy (Pilditch et al., 2019; Roberts and Davidai, 2022) and instead approaches power as a firm-specific variable, such that one firm's high power does not necessarily imply that the power of that firm's counterpart is low, as both firms can have high (or low) levels of power simultaneously.
Based on this approach, we examine how distinct intellectual traditions have produced only a partial view of buyer–supplier dyads, with distinct foci. One research stream, drawing mainly on a rational choice perspective from economics (Hardin, 2002; Williamson, 1975), has typically examined the opportunism that can result from the partner's power. Building on social psychology (Blau, 1964; Homans, 1958), a second stream of research has broadly employed a behavioral perspective that tends to focus on the focal actor's power and how it influences its willingness to trust (Brattström and Faems, 2020; Schilke et al., 2015). The study of only one constituent part of a buyer–supplier dyad—either the focal entity or its partner—precludes theoretical clarity and the integration of insights from research following mainly different theoretical traditions.
The purpose of this paper is to develop a unifying approach by examining the extent to which the focal organization's trust in the partner is affected (i) by its partner's power and (ii) by its own power. We depart from prior research insofar as we theorize and capture both effects simultaneously—that is, the influence of the partner's power on the trust placed by the focal organization (which we call the partner effect) and the influence of the focal organization's power on its trust in the partner (which we call the actor effect). This bilateral approach allows us to reflect the truly relational nature of buyer–supplier dyads (Lumineau and Oliveira, 2018). To that end, this paper introduces the actor−partner interdependence model (APIM) to research on buyer–supplier relationships. The APIM is an approach to “dyadic relationships that integrates a conceptual view of interdependence […] with the appropriate statistical techniques for measuring and testing it” (Cook and Kenny, 2005: 101). In our study, the APIM specifically allows different predictions to be made regarding the effect of power on trust based on the rational choice and behavioral perspectives, thus offering the possibility of unifying seemingly diverging approaches. By drawing ideas from symbolic interactionism (Kaufmann and Denk, 2011; Omar et al., 2022) and cheap talk (e.g., Li et al., 2022; Özer et al., 2011) in the context of OSCM, we also develop an argument that the partner's level of self-promotion communication is a critical contingency that impacts these effects of power on trust.
To test our theorizing, we followed recent methodological recommendations by surveying both the buyer and the supplier within the interorganizational relationship; furthermore, we relied on multiple key informants within each of these organizations (Homburg et al., 2012). Specifically, we collected matching data from both the buyer and supplier sides of 250 dyads in China while also collecting data from two key informants in each organization (for a total of 2 × 2 × 250 = 1,000 data points in our dyadic dataset). In addition, we scraped the webpages of each organization to conduct a computerized, linguistic assessment of the company's level of self-promotion communication using Linguistic Inquiry and Word Count (LIWC) (Tausczik and Pennebaker, 2010). This combination of dyadic survey data and archival data addresses recent calls for the use of mixed-methods approaches in OSCM (Kaufmann et al., 2023).
We make two interrelated contributions to OSCM. First, we contribute by unifying extant findings about the influence of power on trust in buyer–supplier dyads. Our use of dyadic survey data from both sides of the dyadic relationships allows us to parse out the partner effect from the actor effect and demonstrate that the nature of the influence of power on trust depends on whose power within the dyad is being considered. This insight offers a long-overdue theoretical integration of different conceptual vantage points. Second, the actor and partner effects differ substantially based on the partner's level of self-promotion communication. This contingent analysis has important implications regarding our understanding of the conditions under which the effects suggested by the two traditions materialize, thus enhancing our knowledge of the scope of each perspective.
Conceptual Background and Hypotheses
Trust plays a central role in buyer–supplier relationships (Bachmann, 2001; Zhong et al., 2017). At the broadest level, trust can be defined as one party's willingness to accept vulnerability to the actions of another party (Mayer et al., 1995; Özer and Zheng, 2017; Schilke et al., 2021). The OSCM literature has investigated various themes related to trust in supply chain management, particularly the relationship between trust and trustworthiness (e.g., Choi et al., 2020; Özer et al., 2014, 2018), the development of trust (e.g., Cai et al., 2010; Gattiker et al., 2007; Huang et al., 2008), trust and cooperative behaviors (e.g., Brinkhoff et al., 2015; Cao and Lumineau, 2015; Johnston et al., 2004), and the influence of trust on knowledge transfer (e.g., Bolton et al., 2013; Ebrahim-Khanjari et al., 2012; Kraft et al., 2022; Özer et al., 2011; Wang et al., 2023). In much of this research, scholars have come to converge on the critical importance of goodwill trust (hereafter referred to as trust), which is defined as attributions concerning another party's interest in one's welfare and the other party's intention to act in one's interest, even when the opportunity to exploit the situation may arise (Ireland and Webb, 2007; Malhotra and Lumineau, 2011), in buyer–supplier relationships. For example, favorable attributions about another party's evenhandedness in social interactions and its tendency to keep promises are important markers of trust (Zaheer et al., 1998), which in turn can smooth adaptation to disruptive events (Kaufmann et al., 2018) and comply with environmental standards (Villena et al., 2021). 1
Researchers have devoted much attention to organizations’ power as a critical origin of trust (Handley and Benton, 2012; Ireland and Webb, 2007). Following prior research (e.g., Cook et al., 2006; Fiske and Berdahl, 2007), we define power as the degree to which one party in a relationship is perceived to have control over resources valued by the other party. As such, power is inherently relational and perceptual in nature (Bourdieu, 1977). As an essential facet of buyer–supplier relationships, power has been approached in terms of its nonmediated and mediated bases (Handley and Benton, 2012; Zhao et al., 2008). The former suggests that power may not necessarily be exercised by the source and instead hinges on others’ perception of the source (e.g., attribution about the partner's expertise), while the latter refers to the use of pressure and withholding of rewards to bring about a specific action by the partner (Benton and Maloni, 2005; Reimann and Ketchen, 2017). In developing our theory, we use power and trust as short forms, but we return to the nonmediated and mediated distinction to specify the effect of power on trust in our findings.
The Influence of Power on Trust
Power and trust are the bedrock of buyer–supplier relationships. The relationship between power and trust essentially relates to the extent to which organizations can transform perceived vulnerability into positive relational dynamics (Ireland and Webb, 2007). However, a key limitation in prior research has to do with conceptual ambiguity and/or one-sidedness regarding the unit of analysis in conceptualizing power. Some studies have been silent regarding the party on whose power they focused (e.g., Ireland and Webb, 2007; Jain et al., 2014), while others have adopted divergent foci on either the focal organization or the partner organization that together form the buyer–supplier dyad (Table 1 summarizes prior research on the effect of power on trust in buyer–supplier relationships). This practice is problematic on two counts: first, it produces considerable confusion about whose power and whose trust is being studied; second, it leads to the erroneous assumption of power as a zero-sum game. To our knowledge, we are the first to make this observation, which we argue can help bring together different theory streams and therefore warrants further examination. On the one hand, the rational choice perspective has generally focused on the impact of the power of the partner on the focal organization's trust (e.g., Handley et al., 2019). Here, the central issue relates to the extent to which the focal organization is more likely to place trust in a high-power partner or a low-power partner. On the other hand, a behavioral perspective largely prioritizes the analysis of the impact of the focal organization's power on its willingness to trust its partner (e.g., Brattström and Faems, 2020; Schilke et al., 2015). The core issue in the behavioral perspective is the extent to which a low-power or high-power actor is more trusting of others. Rather than representing a strict conceptual split, the rational-choice and behavioral perspectives capture broadly different theoretical foundations about our understanding of power and trust dynamics.
Summary of the empirical literature on the effect of power on trust in buyer−supplier relationships.
a
Summary of the empirical literature on the effect of power on trust in buyer−supplier relationships. a
SEM = structural equation modeling; OLS = ordinary least squares; APIM = actor−partner interdependence model.
We conducted a systematic review of articles about power and trust in buyer–supplier relationships by searching the Web of Knowledge (1990–2022). We included articles even if the exact measures were not entirely related to power and trust. By doing so, we ensured the inclusion of all relevant articles about power and trust and therefore can provide a comprehensive review of the power-trust effect in buyer–supplier relationships.
Offering a conciliatory integration, we suggest that research streams following the rational choice and behavioral perspectives, respectively, are in fact not in direct opposition to each other. Instead, each perspective adopts different—but ultimately incomplete—lenses to understand and analyze the dyadic relationship. Researchers have theorized about the effect of the partner's power (i.e., partner level—emphasized largely by the rational choice perspective) or the focal organization's power (i.e., actor level—usually emphasized by the behavioral perspective) on trust in these dyads. For example, Maloni and Benton (2000) report how a partner organization's use of power influences the focal organization's trust (partner-level effect), while Perrons (2009) examines how Intel's platform leadership strategy is built on both its limited use of power and its high trust in its suppliers (actor effect). However, as we will discuss, trust in buyer–supplier dyads is affected by both (i) the partner's power and (ii) one's own power, and these two effects are at play simultaneously. Accounting for both partner and actor effects is thus critical for obtaining a more holistic understanding of the fundamental power–trust linkage in buyer–supplier dyads. Accordingly, this study's integrative research question is as follows: to what extent is the focal organization's trust in the partner affected by (i) its partner's power and (ii) its own power?
To answer the foregoing question, we bring together research drawing primarily on the rational choice or behavioral perspectives (see Schilke and Cook, 2015 and Zajac and Olsen, 1993 for a discussion of these perspectives in interorganizational scholarship) to pursue two main goals. Figure 1 summarizes our approach vis-à-vis prior research. First, we study the effect of power on trust when approached from the vantage points of the partner and of the actor, the two basic entities forming the buyer–supplier dyad. Second, we examine the strength of the partner effect and the actor effect as a function of a key contingency—namely, the partner's self-promotion communication. As we elaborate in greater detail below, the extent to which a partner engages in self-promotion to purposefully develop a favorable image of itself can significantly alter the effect of power on attributions about that partner's trustworthiness. Table 2 summarizes the main arguments that underlie the hypotheses proposed in this study.
Approaches to studying the effect of power on trust in prior research. Notes: The matching of the rational-choice perspective with the partner effect on the one hand and of the behavioral perspective with the actor effect on the other is motivated by our reading of the extant literature and is aimed at easing the communication of our synthesis of the literature.
Summary of the arguments in support of our hypotheses.
Research on power largely builds on economics and follows a rational choice perspective that emphasizes the threat of being taken advantage of in buyer–supplier dyads (Hardin, 2002; Williamson, 1975). Indeed, “large parts of the literature on power in the supply chain have worked under the assumption that the more powerful party will tend to exploit its power for its own benefit” (Reimann and Ketchen, 2017: 4).
In research mainly following a rational choice perspective, the possibility of exerting pressure and granting privileges creates opportunities for partners to behave in a selfish manner in their interactions (Cheng et al., 2021). In the context of global chip shortages (J.P. Morgan Research, 2022), for instance, a carmaker (focal organization) may anticipate the possibility of a powerful chip supplier (partner) changing the delivery items that would disrupt their production, and this carmaker will, in turn, place low trust in this supplier. High-power partners may also enjoy greater bargaining leverage, providing them with the opportunity to dictate terms that are favorable for themselves (Brown et al., 1995; Crook and Combs, 2007). From a normative viewpoint, the focal organization should thus be watchful regarding the intentions of high-power partners in supply chains (Table 2).
Conversely, low-power partners may plausibly be expected to be trustworthy, as they tend to place high value on a relationship that promises them access to resources and opportunities they would not be able to obtain otherwise. Low-power partners often face challenges and extra costs in finding alternative firms with which to work. Therefore, the relations that low-power partners have are all the more valuable (Brito and Miguel, 2017). Low-power partners may be highly motivated to maintain their existing relationships and thus behave in an accommodating manner. Anticipating these considerations, focal organizations view relatively powerless partners as trustworthy and thus place high trust in them.
What is rarely made explicit, however, is that the rational choice perspective of trust, sometimes also referred to as the “encapsulated interest account” (Farrell, 2004; Farrell and Knight, 2003; Hardin, 1999, 2002), primarily offers arguments about the partner effect—that is, the influence of the partner's perceived power on the focal organization's trust. The emphasis on opportunistic behavior as a function of power explains the focus on the partner and the extent to which it can be trusted, depending on whether it has high or low power. 2 The central assumption is that trustors will put themselves in the position of the trustee to predict how the trustee will behave (Handley et al., 2019). For example, a manufacturer of medical equipment (buyer) may make efforts to understand the position of a supplier of high-quality cotton gauze (supplier) to anticipate the interests and likely actions of this supplier, such as the likelihood of delayed deliveries and increased prices. Such considerations are highly pertinent in light of the current shortage of high-quality cotton, where “lower global supplies [are] sending cotton prices higher”. If the cotton gauze supplier is high in power, then this supplier can be expected to have greater liberty to act opportunistically, in turn leading the medical equipment manufacturer to display low trust in its partner. Conversely, if the gauze supplier is relatively low in power, then this supplier is expected to place high value on the relationship and thus behave cooperatively, in turn leading the buyer to display high trust in the partner. Based on the above reasoning, we propose the following hypothesis regarding the partner effect:
Hypothesis 1 (H1). The lower the partner's power is, the higher the focal organization's trust in its partner.
Behavioral Perspective
At first glance, however, the aforementioned prediction appears to be at odds with evidence gathered in several field studies on buyer–seller relationships (Brattström and Faems, 2020) and experimental laboratory studies on the influence of power on trust (Schilke et al., 2015), which highlight that low-power parties place high trust in the counterpart. To appreciate such apparent discrepancies, it is important to identify relevant differences in foci and assumptions.
First, scholars following mainly a behavioral perspective question whether parties factor in the perspective of the partner to the extent assumed by research that follows a rational choice perspective (Connelly et al., 2012; Schilke et al., 2015). The criticism is that parties might, in fact, be more concerned about their own situation than about considering their partners’ interests (Table 2). This viewpoint resonates with reports by professional associations (e.g., Association for Supply Chain Management) and consulting firms consistently indicating that operations managers are often primarily concerned with their own goals. The PwC Digital Trends in Supply Chain Survey (PwC, 2022) reports that one of managers’ core concerns is securing materials for their own firm and developing their own local sourcing strategies. Reflecting such self-focus, the behavioral account addresses the focal organization's own power rather than its perceptions of the partner's power. In other words, the emphasis is on the actor effect in studying the effect of power on trust in the buyer–supplier dyad.
Second, the key explanatory mechanism proposed to produce an effect of power on trust differs across camps. While research following a rational choice account tends to emphasize how power can explain opportunistic behavior, the behavioral account instead stresses how power can shape motivated cognition. Motivated cognition denotes that managers may strive to arrive at conclusions they want to arrive at as part of mitigating cognitive dissonance (Chae et al., 2017; Graebner, 2009). As such, the decision to trust may be based more on one's own motivations and desires than on calculations of the partner's likely deliberations. Especially regarding perceptions of low power, feelings of vulnerability and helplessness can engender ambiguity if not distress (Molm et al., 2000; Pulles and Loohuis, 2020). To elude these undesirable states, managers of low-power focal organizations may simply hope that their partners will turn out to be reliable and trustworthy (Kaufmann et al., 2018; Mooijman et al., 2015). To minimize cognitive dissonance, low-power focal actors may display high levels of trust, while high-power focal organizations have little reason to engage in motivated reasoning (Kramer, 1996) and are thus relatively less trusting. It is noteworthy that the cognitive dissonance underlying motivated reasoning is the result of the focal firm's own precarious power position rather than the partner's strength. Accordingly, our approach is in line with prior research conducted from the behavioral perspective, which has consistently emphasized the focal actor's own situation in explaining motivated reasoning. Consistent with this line of argument, we propose the following hypothesis about the actor effect 3 :
Hypothesis 2 (H2). The lower the focal organization's power is, the higher that focal organization's trust in its partner.
As we elaborated above, prior research tends to adopt only a partial view of the buyer–supplier dyad by accounting for either the partner effect (rational choice perspective, H1) or the actor effect (behavioral perspective, H2). However, both effects are at play in the buyer–supplier dyad. Without accounting for the partner effect when estimating the actor effect, theoretical predictions about the effect of power and trust in buyer–supplier dyads are necessarily incomplete, and empirical estimates are possibly biased.
We acknowledge that the suggested matching of the rational-choice perspective with the partner effect on the one hand and of the behavioral perspective with the actor effect on the other is primarily motivated by our reading of the extant literature—in principle, however, it may be possible to adopt a rational-choice stance on the actor effect and a behavioral view on the partner effect, 4 even though we are not aware of any prior research using such approaches. Specifically, one could argue that from a rational-actor perspective, a high-power actor may anticipate that its resources render it a valuable collaborator, thereby reducing the likelihood of being treated opportunistically and positioning it to exhibit greater trust than a lower-power actor. Intriguingly, this reasoning suggests a reversal of the hypothesis presented in H2. Similarly, a behavioral perspective on the partner effect may suggest that a high-power partner can lead a focal firm to realize it has more to lose than with a low-power partner, hence making it more prone to exhibit motivated reasoning and place more trust. As a result, the effect proposed in H1 would be reversed. While conceptually plausible, we have found no precedence for such “cross” theorizing in the literature, which is why we have refrained from formally stating competing hypotheses, but we will revisit this issue in interpreting our findings in Section 5.
Self-Promotion Communication as a Critical Contingency in Buyer–Supplier Relationships
Our argument about the effect of power and trust has thus far considered the power of the parties engaged in a buyer–supplier dyad (i.e., a focal organization and a partner) while ignoring how these parties represent themselves and how such representation may shape their social interaction and thus the effect of power on trust. However, as Dill (1962) insists, “it is not the supplier or customer himself that counts [i.e., the parties], but the information that he makes accessible” (cited in Heide and John, 1990: 32). Indeed, a focal firm is likely to try to anticipate the potential behavior of partners, in large part based on information provided to the firm by these partners. It is well known that the way firms present themselves is a crucial component in creating opportunities for forming new ties, securing long-lasting business relationships, and building reputation in the market (Oliveira et al., 2022). In particular, the extent to which the partner engages in impression management directly impacts social interactions with the focal organization, thus influencing how the focal organization perceives the partner, makes attributions about the partner's intentions, and subsequently places trust in this partner.
To theorize the implications of partners’ impression management, we mainly build on symbolic interactionism (Fine, 1993; Goffman, 1959), which focuses on how entities purposefully present information in specific ways in hopes of creating a favorable image of themselves when interacting with others. The literature defines impression management as the process by which actors attempt to control how others perceive them (Leary and Kowalski, 1990). 5 In particular, self-promotion communication has been singled out as a key form of impression management aimed at publicly presenting organizations in a positive light and favorably influencing the perceptions of other parties (Bansal and Clelland, 2004; Diestre et al., 2023). High self-promoters tend to share more information about their achievements and virtues than low self-promoters, who tend to be discreet and modest by not boasting about their achievements and even concealing information (Reimann et al., 2022; Sampath et al., 2022). The notion of self-promotion communication is further related to OSCM research on cheap talk (e.g., Li et al., 2022; Özer et al., 2011), which zooms in on the communication of empty language (while the concept of self-promotion communication is agnostic with regard to the truth value of information). The concept of self-promotion is generally more likely to reflect a firm's capability than mere empty language. When buyers disseminate false information about awards and certificates, industry peers can verify these claims, and such behavior can increase the cost of doing business (Li et al., 2022).
As supply chain practitioners understand, how a firm presents itself to external partners—and in particular, how it attempts to promote its business virtues and accomplishments—differs considerably among firms. Two suppliers might produce largely identical products in terms of features and price, but they can pursue very different degrees of self-promotion communication. An organization might go to great lengths to boast about its awards and ties with highly reputable firms (Beer et al., 2022; Busse, 2016), while another might deliberately conceal such information and restrain itself from attracting attention from stakeholders (Carlos and Lewis, 2018; Gehman and Grimes, 2017). Building on social interactionism as applied to the supply chain context (Beer et al., 2022), we argue that the consequences of power may differ markedly as a function of the partner's level of self-promotion communication. Therefore, our second research question is as follows: to what extent does the partner's self-promotion communication moderate the relationship between power and trust in buyer−supplier dyads?
In what follows, we argue that the partner and actor effects proposed in H1 and H2, respectively, differ as a function of the partner's level of self-promotion communication.
Moderation of the Partner Effect
As discussed above, two important, interrelated preconditions for the proposed partner effect of power on trust are that the focal organization (i) is watchful regarding the partner's intentions and (ii) engages in the effort to take its partner's perspective. If these conditions hold, the greater the partner's power is, the less the focal organization will trust the partner (H1). Considering the important role of self-promotion communication in how organizations present themselves to stakeholders (Bansal and Clelland, 2004; Beer et al., 2022), we believe that these conditions tend to be less true when the partner engages in high levels of self-promotion communication. As a result, the negative partner effect weakens when the partner engages in high (vs. low) levels of self-promotion communication.
A well-understood aspect of impression management is that strong self-promoters direct much of their efforts to share information about themselves. Consistent with symbolic interactionism (Beer et al., 2022; Heide and John, 1990), a partner's self-promotion communication inevitably involves sharing information about its goals and intentions such that the focal organization can actively use this information as input to weigh how much trust to place in the partner, given the level of the partner's power. A focal organization working with a high self-promotor has more access to information about this partner's achievement and goals, such that the focal organization's perceived threat of being taken advantage of is effectively reduced. Extensive self-promotion by the partner also tends to engender a superior reputation and market visibility for this partner that, if tainted, can have detrimental consequences for the partner (Schilke and Cook, 2015). Because high self-promoters are highly committed to being perceived in a positive light, partners who are high self-promoters typically take extra steps to engage with and accommodate the focal organization as part of reputation management (Turnley and Bolino, 2001). Research about self-representation by companies finds that the language of self-promotion typically entails actively caring for and showing an understanding about others’ well-being (Busse, 2016). For example, the Dutch firm ASML actively promotes on its webpage that “we work with our suppliers closely to ensure we all adhere to our high standards in quality, logistics, technology, cost, and sustainability”; such communication emphasizes joint work to overcome the challenges that suppliers might face. In sum, the focal organization's perceived threat of being taken advantage of or not being understood by a high-power partner tends to be lower when that partner engages in high levels of self-promotion communication, thereby weakening the negative effect of the partner's power on the focal organization's trust (Table 2).
In contrast, low self-promoters make fewer efforts to create and share information about themselves. Because of the partner's low self-promotion communication levels, the focal organization has limited information about the partner's goals and intentions, and thus, fears of being taken advantage of may be aggravated. The focal organization may deal with this ambiguity by searching for information about the partner from sources that might be noisy, such as industry rumors (Currall and Judge, 1995). Partners who have low self-promoter potential can be expected to be less concerned about their reputation. By virtue of their low visibility in the industry, such partners are likely to go unnoticed by industry peers. While focal organizations can be aware that self-promoters might exaggerate their claims about caring for their supply chain counterparts, the lack of information on low self-promoters can exacerbate the focal firm's fears of possibly being exploited. Because the focal organization lacks information, it may become extra cautious when dealing with low self-promoters who are powerful, thereby strengthening the negative effect of the partner's power on the focal organization's trust. The aforementioned arguments lead us to propose the following hypothesis.
Hypothesis 3a (H3a). The relationship between the partner's power and the focal organization's trust in its partner (i.e., the partner effect proposed in H1) is weaker when the partner engages in high (vs. low) levels of self-promotion communication.
Moderation of the Actor Effect
Continuing to build on symbolic interactionism (Beer et al., 2022; Heide and John, 1990), we posit that the extent to which the partner engages in self-promotion communication also has important implications for the actor effect of power on trust, albeit in a different way. In developing H2, we suggested that the focal organization's power will be inversely related to that focal organization's trust in its partner in the buyer–supplier dyad. Central to this argument is the fact that managers in low-power focal organizations typically experience higher ambiguity and distress than those in high-power focal organizations, and these managers respond by placing high levels of trust in the hope of reducing cognitive dissonance (Mooijman et al., 2015; Pulles and Loohuis, 2020). Here, we extend this line of argument by proposing that the magnitude of the negative actor effect will be higher when the focal organization faces a partner that engages in low levels (compared to high levels) of self-promotion communication.
When interacting with partners low in self-promotion communication, the focal organization has little information about the partner. As a result, managers from low-power focal organizations have a greater motive to engage in strategies designed to cope with this ambiguity. As the degree of ambiguity increases when interacting with partners low in self-promotion communication, actors have a greater propensity to engage in motivated cognition stemming from a low-power position, thus strengthening the actor effect of power on trust in buyer–supplier dyads (Table 2).
When interacting with partners high in self-promotion communication, however, we argue that the confidence displayed by such partners may offer (perceived) assurances and convey implicit promises about the conduct of the partner. For example, as a buyer (partner) boasts about long-term relationships with high-profile suppliers and publicly advertises their awards, a supplier (the focal organization) may feel reassured about working with this buyer and feel less anxious, even when working with high-power buyers. If ambiguity is attenuated when interacting with partners high in self-promotion communication, then we expect that actors are less likely to engage in the motivated cognition that underlies the actor effect of power on trust in buyer–supplier dyads. Accordingly, we hypothesize the following:
Hypothesis 3b (H3b). The relationship between the focal organization's power and trust in its partner (i.e., the actor effect proposed in H2) is stronger when the partner engages in low (vs. high) levels of self-promotion communication.
Methods
Research Design and Sample
Testing our hypotheses required data from both sides of the buyer–supplier relationship to capture each party's perceptions of the relationship (Lumineau and Oliveira, 2018). We therefore employed a dyadic survey design while recruiting two informants from each side of the buyer–supplier relationship (Kumar et al., 1993), resulting in four informants for each dyad. Specifically, we developed four customized questionnaires with distinct target respondents: the buyer's middle manager, the buyer's top manager, the supplier's middle manager, and the supplier's top manager. As different informants from the buyer and supplier are used to collect data for the dependent and independent variables, our multi-informant approach reduces the threat of common method bias (Podsakoff et al., 2003).
Our data collection took place in China. Prior research recommends that survey data collection in China be carried out by first contacting the respondents via phone to verify that they qualify for participation in the study and to assure them that their responses will be kept confidential and that the results will be reported only at the aggregate level (Poppo and Zhou, 2014; Villena and Craighead, 2017). This confidentiality assurance aims to encourage participants to provide honest and accurate responses, without fear of repercussions or exposure. In addition, multiple incentives, including organizing a free seminar and providing a formal report, are offered to improve response rates and enhance data reliability (e.g., Flynn et al., 2018; Forza, 2002). Collecting dyadic data required the willingness of matching organizations to participate in our study, making data collection more challenging. Particularly in the context of emerging economies, working with local institutions is considered critical for obtaining organizations’ buy-in and collecting valid information (Hoskisson et al., 2000). As such, we appointed All China Marketing Research (ACMR), a company with considerable expertise in data collection in China. ACMR has a strong track record of data collection for scientific research (e.g., Bai et al., 2016), and it owns a proprietary database called the ACMR Enterprise Information System (ACMR EIS). This database integrates several other archival databases developed by government institutions 6 to provide relevant background information on Chinese companies (e.g., industry and age) and contact details of their executives.
Prior to commencing data collection, we took several steps to ensure the clarity of the questionnaire and the relevance of the survey items. First, we extensively reviewed the literature to identify valid survey items to measure our constructs. Second, we pretested the survey with 27 managers (12 purchasing managers and 15 supply managers) to assess the clarity of the wording and face validity of the measures. We then used the feedback we received to refine the questionnaire. Third, we discussed the survey items with experienced recruiters at ACMR and finalized the questionnaire based on their suggestions. Finally, we developed an English version of the questionnaires and used back-translation procedures to ensure conceptual equivalence.
A random sample of 1,776 manufacturing firms 7 was drawn from the EIS database. Manufacturing is a major contributor to the Chinese economy, representing 24.87% of China's GDP in 2024 (World Bank National Accounts data, and OECD National Accounts), and it is a well-known field for salient power dynamics among buyers and suppliers (Bensaou, 1999; Wilhelm and Villena, 2021). Because specialists tend to provide more reliable survey responses than generalists (Homburg et al., 2012), we aimed to recruit dedicated procurement managers who could serve as qualified key informants about the supplier relationship. Since dedicated procurement managers are relatively rare among very small firms, we excluded 166 buyers with fewer than 100 employees. ACMR recruiters called each of the remaining 1,610 buyers to inquire about their willingness to participate. Among them, 584 were unreachable (in part due to incorrect phone numbers), 446 answered the call but refused to participate, and 109 agreed to participate but turned out not to meet at least one of our three inclusion criteria (i.e., operating in the manufacturing industry, employing at least 100 employees, and being a middle or top manager familiar with the firm's supply chain unit). Eventually, 471 buyers signaled their firm's willingness to participate and put us in touch with managers who qualified to complete this study. ACMR recruiters then mailed the questionnaire to the middle and top managers of each of these 471 buyers. In total, 261 buyers completed and returned questionnaires from both the middle and top managers. Based on the number of distributed questionnaires, the response rate was 55.41%, which compares favorably to the response rates of other dyadic surveys about buyer–supplier relationships (e.g., Poppo and Zhou, 2014: 35.40% and Villena and Craighead, 2017: 25.00%). We compared the responding and nonresponding buyers using a multivariate analysis of variance. The results did not point to substantial differences in terms of key buyer characteristics (i.e., industry, ownership, number of employees, and annual sales revenues) (Wilks's Λ = 0.98; F = 1.38; p = 0.23). Therefore, nonresponse bias did not appear to be a major concern in our study. After the data collection, one of the authors called 50 randomly selected top managers from the buyers to confirm whether the survey had been conducted, and found no evidence of cheating.
To gather information for the buyer-side questionnaire and using a referral procedure that aimed to increase suppliers’ willingness to cooperate, ACMR recruiters contacted the middle and top managers of the fourth largest supplier of each buyer. 8 Eventually, 250 suppliers returned two completed supplier-side questionnaires (one from a middle manager and one from a top manager). 9
In total, this study's dataset contains information from 1,000 managers representing 500 firms paired in 250 buyer–supplier dyads. We aimed to ensure the suitability of the key informants by asking them about their tenure in the organization and in their current job. On average, managers had over five years of experience in the employing organizations (buyers’ top managers: Mean = 10.13; SD = 6.43; buyers’ middle managers: Mean = 7.37; SD = 5.47; suppliers’ top managers: Mean = 9.57; SD = 5.60; suppliers’ middle managers: Mean = 6.53; SD = 3.97) and in their current job (buyers’ top managers: Mean = 6.34; SD = 4.36; buyers’ middle managers: Mean = 4.60; SD = 3.20; suppliers’ top managers: Mean = 7.00; SD = 4.52; suppliers’ middle managers: Mean = 4.82; SD = 3.15).
In addition to conducting a multi-informant, dyadic survey, we collected secondary data from the companies’ websites to assess their level of self-promotion communication. We collected data on the 500 organizations included in our dataset, and we used both the English and (simplified) Chinese versions of the text that amounted to over 3,500 pages of additional data. A multidata source approach was desirable to reduce the dependence on a single data source, and it provided us with the actual information that companies use to convey their public image. The combination of different data sources not only offers “a better understanding than one data type alone” (Harrison et al., 2020: 474) but can also “produce insights that exceed the sum of the individual […] components” (Molina-Azorin et al., 2017: 183).
Measures
Trust
For our dependent variable, we assessed the extent to which the focal organization placed goodwill trust in its partner. Goodwill trust refers to the focal organization's belief that its partner attributions has an interest in one's welfare and will act in one's interest even when the opportunity to exploit the situation may arise (Ireland and Webb, 2007; Malhotra and Lumineau, 2011). To measure goodwill trust, we used scales from prior research on buyer–supplier relationships (Seppänen et al., 2007). For example, we asked participants to respond to the following statement: “This buyer/supplier has always been evenhanded in its negotiations with us” (Zaheer et al., 1998). Table 3 presents the questions included in our measures. Trust was measured using three items; all items were measured on a seven-point Likert scale (1 = strongly disagree; 7 = strongly agree). We asked buyers’ top managers to answer questions regarding their suppliers, and suppliers’ top managers answered questions regarding their buyers.
Constructs and measures.
Constructs and measures.
Notes: In the dyadic survey, we adjusted the items used to measure power and trust based on whether the respondent was a buyer or a supplier. AVE = average variance extracted.
In measuring our focal independent variable, we built on prior research asking managers to rate their partner's power, that is, the extent to which an actor is perceived to have control over resources valued by another party (Cook et al., 2006; Fiske and Berdahl, 2007). More specifically, power includes both a mediated basis, relating to extrinsic motivation and overt forms of pressure, and a nonmediated basis, relating to intrinsic motivation and settled forms of pressure (Skowronski et al., 2022: 4). Accordingly, the items include, for instance, “We usually get good advice from this customer” (nonmediated power) and “We feel that by going along with this customer, we are favored on some other occasions” (mediated power) (Table 3). A high score indicates that the partner is high in power (i.e., has control over resources valued by the other party). We used a total of six items to capture power; all items were measured on a 7-point Likert scale (1 = strongly disagree; 7 = strongly agree) 10 (in the findings section, we further explore the more nuanced effect of power on trust according to mediated vs. nonmediated bases of power).
Self-Promotion Communication
Our moderator concerns the partner's level of self-promotion communication, defined as its attempts at positively influencing perceptions about the organization (Bansal and Clelland, 2004; Diestre et al., 2023). To measure our moderator, we followed prior research leveraging the fact that managers actively use information on their company websites (e.g., policies, reports, and press releases) to influence the perceptions of how their firm is perceived by external audiences (Bansal and Clelland, 2004). Because company websites offer relevant insight into the company's impression management, we manually scraped the content of every buyer's and supplier's website in our sample.
We examined the level of self-promotion communication using the Linguistic Inquiry and Word Count (LIWC) approach. LIWC is “a transparent text analysis program that counts words in psychologically meaningful categories” (Tausczik and Pennebaker, 2010: 24). Importantly, LIWC classifies words into specific categories, controlling for the length of the text, based on dictionaries that are internally reliable and externally valid according to multiple tests and experimental research (Pennebaker et al., 2015). In line with our research objectives, we used a widely tested measure labeled “clout,” which captures the extent to which a text displays properties related to self-promotion (Pennebaker et al., 2015). For example, Piezunka and Dahlander (2019) used this LIWC measure to capture the extent to which new ventures boast about their crowdsourcing accomplishments. In our study and data, a firm boasting about its “excellent, experienced, hard-working R&D team” and the fact that it “is greatly appreciated by a variety of different markets throughout the world” would return a high score on this measure. 11 Table A in the E-Companion contains further examples from the data we collected. Consistent with our understanding of self-promotion, the data could include aspects of cheap talk (empty claims) and boasting based on facts (e.g., awards and certifications). Self-promotion communication is operationalized as a continuous variable ranging from 0 to 100.
Controls
We included multiple controls at the respondent, firm, and buyer–supplier dyad levels, particularly with regard to factors that have been discussed within research largely based on the rational choice and behavioral perspectives. First, we accounted for the tenure of the respondent (tenure) to control for differences in experience and managerial routines that are accumulated over time, which could influence not only answers concerning power and trust but also the respondent's perspective-taking ability (which is central in the behavioral perspective). We measured the number of years for which the person had worked at the company. Second, we controlled for the following firm attributes that have typically been linked to the rational choice perspective on power and trust: age, industry, and R&D intensity. The longer an organization has operated, the greater the knowledge acquired about the production process and its reputation in the market (Saparito et al., 2004), which might affect the relationship between trust and power. Therefore, we controlled for age (number of years since founding). The different industries in which firms operate can differ in terms of power dynamics and norms about trust. While we focused on firms working in the manufacturing and electronics industry, we also controlled for whether the firm operated in a low-tech (= 0) versus high-tech industry (= 1) (industry). High-technology industries tend to be volatile, which can influence power and trust as well as practices related to self-promotion communication. Consistent with the national context of our study, we followed the National Standards of the People's Republic of China (GB/T 4754-2017), which specifies six high-tech manufacturing industries: medical and pharmaceutical manufacturing; aviation, spacecraft and equipment manufacturing; electronic and telecommunication equipment; computer and office equipment manufacturing; medical equipment and instrument manufacturing; and information chemical manufacturing. We also controlled for the firm's level of R&D investment (R&D) because some firms might operate in low-tech industries while making major R&D investments; however, they can also operate in high-tech industries while making minor investments in R&D. Organizations that make large investments in R&D might be more prone to self-promotion communication as a way to create a public image of centers of expertise relative to their competitors. We measured R&D as the ratio of R&D expenditures to total sales. Finally, we accounted for the possibility that firms can systematically differ in terms of the extent to which they proactively share substantive information (Özer et al., 2014; Zhou and Benton, 2007) by controlling for the volume of information pertaining to a firm's business on its corporate website. To measure this variable, we added the number of words used on the firm's webpage as a control variable.
Finally, we accounted for the development of relational norms in the dyad by controlling for the frequency of interaction between buyers and suppliers (Akkermans et al., 2019; Cao and Lumineau, 2015). Frequent interactions may be related to commitment between parties (Lin and Germain, 1999), which can affect both the partner and actor effects included in our conceptual model. For example, parties who interact frequently may display high levels of trust toward a counterparty based on commitment rather than power (an antecedent of trust we theorize about in this article). We measured interaction frequency on a 7-point scale (once per day = 1; three times a week = 2; once a week = 3; twice a month = 4; once a month = 5; once per 6 months = 6; less than once every 6 months = 7).
Measurement Properties
Table 3 shows the coefficient alphas (α), composite average variances extracted (AVE), and descriptive statistics for each of the latent constructs (namely, power and trust). Overall, these statistics suggest that we adopted reliable and valid measures. We performed a confirmatory factor analysis (CFA) including all study constructs using structural equation modeling and following the maximum likelihood (ML) procedure (Hair et al., 2006). We found a satisfactory goodness of fit when estimating only the direct effects to test H1 and H2 (χ2 = 282.231; df = 132; χ2/df = 2.138; CFI = 0.851; TLI = 0.828; SRMR = 0.080) and when estimating the moderation effects to test H3a and H3b (χ2 = 434.033; df = 198; χ2/df = 2.192; CFI = 0.941; TLI = 0.931; SRMR = 0.079) (Hu and Bentler, 1999). The square root of each factor's AVE is larger than the absolute value of the correlation of that factor's measure with all measures of the other factors in the model, demonstrating satisfactory discriminant validity (Fornell and Larcker, 1981).
Robustness Checks
Single-Respondent Bias
The use of a single respondent could have introduced a potential bias that is particularly problematic in survey research on buyer–supplier relationships between organizations (Lumineau and Oliveira, 2018). We investigated this problem by collecting data from two respondents from each of the two organizations in the dyad. While most constructs were captured from either the top or the middle manager, we decided to collect information on trust from both of them, which allowed us to triangulate their responses (Homburg et al., 2012). We found a high correlation between the responses of middle managers and those of top managers regarding trust in the partner organization (rbuyers = 0.560; p < 0.001; rsuppliers = 0.568; p < 0.001), which further demonstrates the reliability of our measurement.
Common Method Bias
We combined survey data and secondary data, thus limiting concerns about common method bias. Nonetheless, the data for several variables were collected via the dyadic survey, which could still raise concerns about potential common method bias (Podsakoff et al., 2003). We aimed to assuage such concerns by asking the top managers of the buyers and suppliers in our sample questions about trust (dependent variable) while collecting information about power (independent variable) from their middle managers. We also carried out Harman's one-factor test of this study's scale items using an unrotated exploratory factor analysis (McFarlin and Sweeney, 1992). The findings of this analysis showed that no single factor accounted for more than 50% of the total variance in the variables (buyers = 43.590%; suppliers = 42.311%), thus providing additional evidence that common method bias was unlikely to be a serious problem. Following Krishnan et al. (2006), we also used the survey respondent's job tenure as a marker variable when employing the partial correlation procedure recommended by Lindell and Whitney (2001). This analysis showed that all zero-order correlations remained essentially unaffected by this adjustment. Overall, common method bias does not appear to be a significant concern in this study.
Endogeneity Bias
We followed a theory-driven approach to endogeneity, first, by building on theory to evaluate potential sources of endogeneity (Wooldridge, 2010) and, second, by assessing the implications of endogeneity in light of the study's specific aims (Shmueli, 2010). Our approach followed advice to offer “an explicit analysis of how sensitive inferences are to endogeneity in the specific research situation” (Ketokivi and Mcintosh, 2017: 3), thereby directly addressing concerns that “ambiguity regarding the cause(s) of endogeneity impairs readers’ ability to assess if the approach used in the study alleviated endogeneity or exasperated the problem” (Hill et al., 2021: 116).
From a theoretical viewpoint, power tends to be to a large extent structurally determined (Cook and Emerson, 1978); thus, potential endogeneity should be less of a concern than with choice variables that organizations flexibly self-select into (Shaver, 1998). Based on an extensive set of endogeneity checks, a recent study concluded that “mediated and non-mediated power can be treated as exogenous rather than endogenous” (Handley et al., 2019: 1705); more recently, Skowronski et al. (2022: 31) reached the similar conclusion that “coercive power and expert power (…) can be treated as exogenous.” These findings converge toward the idea that power is largely exogenous. In addition, with respect to the goals of our study, we are primarily concerned with predicting an actor effect and a partner effect about the influence of power on trust in buyer–supplier relationships. Because of our focus on predicting, received wisdom suggests that endogeneity—which affects the accuracy of the estimates—is not a key concern (Ketokivi and McIntosh, 2017; Shmueli, 2010).
Nonetheless, we endeavored to further explore the possibility of endogeneity bias and its potential impact on our results by addressing the relevance of specific sources of endogeneity in our study: (1) omitted variables, (2) simultaneity, (3) measurement error, and (4) selection (Wooldridge, 2010). First, for omitted variables, we included several control variables across various levels of analysis. Specifically, interaction frequency (dyad) was included as a control. If omitted, the effect of interaction frequency was added as part of the residual, which would cause the residual to correlate with the independent variable “power.” We also followed the approach suggested by Frank et al. (2013) by assessing the extent of bias that would be necessary to invalidate our inferences. When applying this technique to the actor and partner effects, we found that to challenge the validity of our hypotheses-related conclusions, it would be necessary for a nonincluded variable to have an impact of at least 0.202. Comparing this magnitude to our included controls suggests that such a significant bias is unlikely. Likewise, recent empirical research reporting similar estimates has been interpreted as robust (e.g., Busenbark et al., 2017; Oliver et al., 2018). Second, true to this study's aims, we favored a multi-informant, dyadic survey over a longitudinal analysis (as a key way to address simultaneity) because we are interested in minimizing biases in the data while capturing the relational nature of power and trust in buyer–supplier relationships. Notwithstanding this limitation, prior research that has systematically addressed simultaneity as a source of endogeneity has concluded that power is exogenous (Handley et al., 2019; Skowronski et al., 2022). Third, a potential source of endogeneity could be measurement error. We have therefore strived to preempt this source of endogeneity through research design, that is, the use of multiple informants per dyad and different respondents answering questions relating to power and trust, the usage of clearly defined constructs and focused questions (e.g., goodwill trust), and the combination of a survey (primary data) and web scraping (secondary data). Relatedly, we also addressed measurement error throughout our data analysis; that is, we used SEM—reportedly less prone to endogeneity stemming from measurement error (Hill et al., 2021)—to test our hypotheses. In addition, we used a computerized analysis (LIWC)—shown to yield convergent and discriminant factor loadings (Pennebaker and King, 1999)—to measure the moderation variable. Fourth, we also considered selection. We aimed to minimize selection bias as a source of endogeneity by drawing a random sample of firms and employing a company that was well-versed in data collection for academic research purposes in China (Bai et al., 2016).
For these reasons, we concluded that we are able to establish plausible exogeneity and, as a result, relaxing the exclusion restriction may not materially jeopardize inference (Conley et al., 2012). Our conclusion is set against the following background: (i) power is reportedly structurally determined, thus alleviating concerns about endogeneity, as supported by prior research, and (ii) this study focuses on predicting, such that endogeneity concerns play a much less salient role.
Analytical Strategy
Actor−Partner Interdependence Model (APIM)
True to the nature of our research questions, we employed the APIM for our empirical tests (Kenny et al., 2006; Kenny and Judd, 1986; Kraemer and Jacklin, 1979). The APIM is particularly suitable for analyzing dependence between observations in dyadic research designs (Krasikova and LeBreton, 2012: 742), especially when using data on perceptual constructs, such as trust (Yakovleva et al., 2010; Yao et al., 2017).
Our APIM uses data collected from the buyer and supplier and explicitly accounts for the nestedness of the data within the buyer–supplier dyad (Kashy and Kenny, 2000; Kenny, 2018; Kenny and Cook, 1999). The independent variables (i.e., the focal organization's power and the partner's power) are modeled as correlated with one another (Cook and Kenny, 2005; Kenny and Ledermann, 2010); APIM also accounts for the correlation of the residuals for the focal organization's and the partner's trust. Distinctively, the APIM enables the partitioning of “the influence of a person's own causal variable on his or her own outcome variable, which is called the actor effect, and on the outcome variable of the partner, which is called the partner effect” (Kenny and Ledermann, 2010: 359). In our study of power and trust in buyer–supplier relationships, the partner effect refers to the effect of the partner's power on the focal organization's trust (H1), while the actor effect refers to the effect of the focal organization's power on that focal organization's trust in the partner (H2). The handling of nonindependence in dyadic data and the estimation of actor and partner effects make the APIM a powerful analytical technique. The APIM has been used to study a wide range of topics in interpersonal dyads, such as work engagement (trustworthiness and cooperation) and emotions in negotiations (for an overview on the use of the APIM, see Krasikova and LeBreton, 2012), but we believe our study is the first to bring this technique to study important questions in an interorganizational setting.
Statistical Modeling
The APIM can be estimated using either structural equation modeling or multilevel modeling. We opted for structural equation modeling as our default methodology, given that it tends to produce a better model fit (Hong and Kim, 2019). We followed the modeling approach developed by Stas et al. (2018), who implemented an APIM with moderation following Garcia et al. (2015). We further augmented their approach by including control variables—not included in previous implementations of the APIM—using lavaan, which is an open-source software for latent variable modeling (Rosseel, 2012), also used by Stas et al. (2018). Our implementation directly follows Garcia et al. (2015) and Olsen and Kenny (2006) 12 (the full lavaan code used for the data analysis is available upon request).
Computerized Linguistic Analysis
The LIWC enabled us to study the properties of the content of the firms’ websites that are indicative of self-promotion communication (see Table A in the E-Companion). Compared to a manual content analysis, a computerized approach was desirable because it is well documented and displays convergent and discriminant factor loadings (Pennebaker and King, 1999). Complementing survey information with text data has become increasingly common in the field of organization theory (e.g., Piezunka and Dahlander, 2019; Sytch and Kim, 2021) but has yet to find its way into OSCM (for notable exceptions, see Oliveira et al., 2022; Reimann and Ketchen, 2017).
Results
Table 4 reports the means, standard deviations, and correlations for each of the variables in our model. The only moderately large correlation is between buyer's and supplier's trust (r = 0.419; p < 0.001). The relatively moderate and low correlations among variables, particularly the main variables tested in our conceptual model, alleviate potential concerns about multicollinearity in the models.
Descriptive statistics and correlations.
Descriptive statistics and correlations.
Categorical variables: Industry: low-tech industry (= 0) and high-tech industry (= 1). Volume of information refers to the word count (excluding functional information, such as contact and location).
Table 5 shows the estimates for the actor and partner effects while controlling for several theoretically relevant variables. Regarding H1, we found empirical support for the partner effect of power on trust (b = −0.109; p = 0.008). Thus, H1 (research mainly following a rational choice perspective) is empirically supported in our full sample. In turn, H2 was tested by estimating the actor effect. The unstandardized coefficient for the actor effect is −0.267 (p < 0.001). As such, we also found support for H2 (research that tends to follow a behavioral perspective). We included controls to address potential confounding factors at the manager, organization, and dyad levels. None of the control variables were found to be significant in our model.
APIM results (H1 and H2).
APIM results (H1 and H2).
Notes: Since our theorizing does not differentiate between buyers and suppliers, we modeled the data as indistinguishable dyads (for a detailed discussion of data structures in the APIM, see Kenny et al., 2006). Therefore, the estimates do not differ between buyers and suppliers. APIM = actor−partner interdependence model.
To test H3a and H3b, we estimated the moderation effects of self-promotion communication on both the partner and actor effects of power on trust in buyer–supplier dyads, the results of which are shown in Table 6. Regarding H3a, we found empirical support for the moderating role of self-promotion communication in the partner effect of power on trust (b = 0.005; p = 0.026). The coefficient for the effect of power on trust (H1) becomes nonsignificant, and further analyses revealed that the partner effect does not run afoul of H1. 13 Turning to H3b, we estimated the moderation effects of self-promotion communication on the actor effect of power on trust in buyer–supplier dyads. We found no support for this moderation effect (b = −0.002; p = 0.440). That is, the negative actor effect of power on trust is not significantly different across varying levels of self-promotion communication. Like in the previous analysis (shown in Table 5), no control variable was found to be significant. In line with our theorizing about self-promotion communication, we anticipate that the findings about the moderation effect hold when there is a degree of costliness and verifiability of the claims (i.e., not mere cheap talk).
APIM results: Self-promotion communication (H3a and H3b).
APIM results: Self-promotion communication (H3a and H3b).
Notes: We model the data as indistinguishable dyads (for a detailed discussion of data structures in the APIM, see Kenny et al., 2006). APIM = actor−partner interdependence model.
Power Bases
Cognizant that power has different bases (Maloni and Benton, 2000; Skowronski et al., 2022), our post hoc analyses further explored the effects of nonmediated and mediated power separately. We followed Skowronski et al. (2022: 6) by “examining one form of non-mediated power and one form of mediated power.” We also examined expert power (related to nonmediated power) and reward power (related to mediated power) because these are two of the most important forms of power in buyer–supplier dyads (Skowronski et al., 2022).
Regarding mediated power, we found that the partner effect is −0.207 (SE = 0.042; z value = −4.942; p < 0.001). This result is consistent with those reported above; H1 is again supported. In contrast, a nonsignificant partner effect was found for nonmediated power (Table D in the E-Companion), with a point estimate of 0.027 (SE =0.024; z value = 1.127; p = 0.260). This suggests that the negative partner effect of power on trust is primarily driven by mediated power. For H2, the actor effect based on mediated power is −0.332 (SE =0.042; z value = −7.922; p < 0.001), and the actor effect for nonmediated power is −0.054 (SE =0.024; z value = −2.257; p = 0.024). In line with our main results, we therefore also found support for H2 using either nonmediated or mediated power as the independent variable (Table D in the E-Companion).
Next, we estimated the moderation effect of the partner's self-promotion communication on the relationships (i) between mediated power and trust and (ii) between mediated power and trust in buyer–supplier dyads. The results are largely consistent with those of the main analysis, but they also elucidated a few nuances. Table E in the E-Companion shows the estimates of the moderation effect of self-promotion communication on the partner effect (H3a) and actor effect (H3b) using mediated and nonmediated power, respectively. H3a (moderation of the partner effect) continues to be supported when using mediated power (b = 0.006; SE =0.002; z value = −2.448; p = 0.014), and we found a positive—although not significant—effect when using nonmediated power (b = 0.001; SE =0.001; z value = 0.654; p = 0.513). For H3b, the partner's self-promotion communication negatively impacts the negative effect between power and trust when using mediated power (b = −0.001; SE =0.002; z value = −0.167; p = 0.867) or nonmediated power (b = 0.001; SE =0.001; z value = 0.638; p = 0.524) (see Table E in the E-Ccompanion). Although not statistically significant, these findings are in line with the hypothesized moderation of the actor effect in buyer–supplier dyads. For the control variables, R&D intensity is significant when examining the mediated basis but not for nonmediated power in buyer–supplier dyads.
Combinations of Power
It is plausible that the dyad's specific combination of power can influence the effect of power on trust. By combination of power, we mean a distinct setup of a low versus high power actor and a low versus high power partner in a given dyad. Accordingly, there are four possible combinations of power that can exhibit different levels of trust, as shown in Figure 2. A dyad may consist of a high-power actor and a low-power partner (combination I: mean = 5.85; SD = 0.799), and vice-versa (combination IV: mean = 5.65; SD = 0.691). Alternatively, the dyad may consist of a low-power actor and partner (combination II: mean = 5.94; SD = 0.604) or of a high-power actor and partner (combination III: mean = 5.39; SD = 0.776). Similar power between actor and partner (combinations II and III) can occur because power is not zero-sum; just because one party in a relationship has high power, the other party does not necessarily have to be low in power (Roberts and Davidai, 2022). While not hypothesized ex ante, the difference in trust between low- and high-power actors appears larger, specifically when the partner has high levels of power.
Trust level according to combinations of power. Notes: I, II, III, and IV refer to the specific combinations of power in buyer–supplier dyads.
Using an ANOVA, we found that the mean trust is significantly different among the four combinations of power, F(3, 240)=8.633; p = 0.001. This result is corroborated by a Kruskal-Wallis ANOVA test (test statistic = 23.743 [df = 3]; p = 0.001). Next, we examined pairwise differences between combinations using Tukey HSD (Honestly Significant Difference) tests. Our findings show significantly different levels of trust for the following two (out of six possible) pairwise comparisons: combination I versus combination III (mean difference = 0.462; p = 0.003) and combination II versus combination III (mean difference = 0.555; p = 0.001). However, whether the dyad includes a high-power actor matched with a low-power partner or a low-power actor matched with a high-power partner does not yield statistically significant different levels of trust (mean difference = 0.197; p = 0.568). We acknowledge that non-significant findings might be a byproduct of dichotomizing the continuous variable power and unequal sample sizes across the different power configurations (Aguinis et al., 2017).
The relevance of these findings is twofold. First, the mean level of trust is the lowest when both the actor and the partner are high-power, thus underscoring the generalized negative association between power and trust examined under H1 and H2. Second, the interesting finding that mean trust levels vary across some combinations calls for additional studies, as we discuss in greater detail in the section about future research.
A rich body of literature in organizational science and management about competition uncertainty and demand uncertainty (Carson et al., 2006; Robertson and Gatignon, 1998) promises to enhance our understanding of power–trust dynamics. We examined whether uncertainty was a boundary condition for our findings; specifically, we used survey data on managers’ perceived competition and demand uncertainty. 14
We found that competition uncertainty positively moderates the partner effect only (b = 0.083; SE = 0.039; z value = 2.153; p < 0.001; Table F in the E-Companion), while demand uncertainty moderates the actor effect only (b = −0.102; SE = 0.031; z value = −3.279; p < 0.001; Table F in the E-Companion). The interaction between competition uncertainty and the partner effect suggests that in contexts characterized by high levels of competition uncertainty, the negative impact of partner power on trust is weakened. This finding is in line with the rational choice perspective, as high levels of competition uncertainty may lead firms to rely more on partners and their resources, thereby reducing the negative effect of partner power on trust. Furthermore, the interaction between demand uncertainty and the actor effect indicates that in situations characterized by high demand uncertainty, the negative effect of focal firm power on trust is strengthened. This finding is in line with research adopting a behavioral perspective, as high levels of demand uncertainty may lead firms to prioritize their own interests and engage in self-protective behaviors, thereby potentially exacerbating the negative effect of focal firm power on trust.
Discussion and Conclusions
Trust is central in buyer–supplier relationships, in which parties have to make decisions that make them vulnerable to their supply chain partners. Thus, researchers have long been intrigued by the precursors of trust, especially power. A considerable number of studies have addressed the relationship between power and trust in buyer–supplier dyads (Brinkhoff et al., 2015; Handley et al., 2019). However, prior research hailing from economics and social psychology has presented apparently partial and inconclusive evidence—due to ambiguity about whose perspective is being examined—that our article endeavors to illuminate and contextualize with the goal of developing a unifying perspective.
Adopting an integrative approach that explicitly incorporates distinct disciplinary traditions and both parties of the buyer–supplier dyad, we test how the partner's power influences the focal organization's trust (i.e., the partner effect) as well as how the focal organization's power shapes its trust (i.e., the actor effect). This first analytical layer accounts for the different entities (i.e., focal and partner organizations) in the buyer–supplier dyad. Both perspectives assume the criticality of information to enable focal firms to make decisions about placing trust in their partners. Drawing on ideas from symbolic interactionism (Beer et al., 2022; Carlos and Lewis, 2018), we extend our study of dyads by analyzing the partner's self-promotion communication. We examine the extent to which the strength of actor and partner effects varies across low versus high self-promoters.
Our main findings reveal an effect of the partner's power on the focal organization's trust (partner effect). We also find that a low-power focal organization tends to display higher levels of trust (actor effect). Our contingency analysis indicates that the influence of the partner's power on the focal organization's trust (partner effect) is weaker when the partner engages in high (vs. low) levels of self-promotion communication. Consequently, whether the predictions regarding the power−trust effect based on the rational choice perspective (partner effect) or the behavioral perspective (actor effect) are more informative varies according to the partner's levels of self-promotion communication.
Theoretical Contributions to OSCM
To address equivocal findings in prior research, we theorize and test distinct partner and actor effects (who the entities are) on the influence of power on trust, and we then examine the partner's level of self-promotion communication (how the entities present themselves) as a critical contingency to both effects. This study provides several implications for the OSCM literature.
First, and on the broadest level, we hope to set an example for embracing the relational complexity inherent to buyer–supplier dyads and overcoming criticism related to the single-party blind spot, which has limited much of the prior interorganizational research (Lumineau and Oliveira, 2018). One-sided theoretical accounts are necessarily ill-equipped to grasp a subject that is multiparty by definition, and empirical tests relying on data from only one organization can give substantially biased results when applied to the study of dyadic relationships (Campbell and Kashy, 2002; Ledermann et al., 2011). Our bilateral APIM approach allows us to theorize about both partner organizations’ unique viewpoints and embrace partner-specific contingencies while avoiding empirical misspecifications (Krasikova and LeBreton, 2012). We see much potential for future interorganizational research to follow this route and shed new light on relational phenomena that a unilateral approach may misrepresent.
Second, we contribute to the continuing quest to understand the sources of trust, which is one of the most active areas of inquiry in contemporary research on interorganizational relationships (Anderson et al., 2022; Schilke and Lumineau, 2025). Recent investigations have singled out power as a critical antecedent to trust (Graebner, 2009; Ireland and Webb, 2007), but extant insights have resulted in confusing predictions, a theoretical puzzle that our study addresses. Research that tends to adopt a rational choice perspective focuses on partners’ potential actions to predict that organizations will place little trust in their high-power partners. With a distinct focus on the focal organization, research mainly following a behavioral perspective suggests that low-power parties strive to ameliorate their inferior structural position by placing trust in their partners. The corollary of these two perspectives is a set of distinct predictions about the influence of power on trust in buyer–supplier relationships. To explore this theoretical conundrum empirically, we re-examined the influence of power on trust while leveraging data from both parties in buyer–supplier dyads, which allowed us to represent both theoretical accounts simultaneously. We adopted a unifying approach to the various aspects of the dyad as opposed to pitting two perspectives against each other. We tested the effect of the partner's power on the focal organization's trust while simultaneously accounting for the effect of the focal organization's power on the focal organization's trust. In so doing, we extend prior research by capturing the relational origins of trust and thereby uniting predictions from research guided by the rational choice and behavioral perspectives within a single model. Hence, this study reinforces the importance of making the relational nature of trust the central starting point for scholarly inquiry.
Importantly, our findings provide concurrent support for both perspectives rather than positioning them as mutually exclusive and contradictory. What is more, the empirical support for the first two hypotheses gives further credence to our theoretical mapping of the rational-choice lens to the partner effect and the behavioral lens to the partner effect. As noted in our theoretical development, despite the absence of precedence, one could theoretically swap this matching—applying a behavioral lens to the partner effect and a rational-choice perspective to the actor effect—which would generate predictions directly opposing our hypotheses. However, the fact that our proposed relationships received empirical support challenges the plausibility of such cross-mapping and instead suggests that the theoretical logic underlying our hypotheses captures an essential aspect of how power and trust intersect in interorganizational relationships. Nonetheless, future research should further unpack the precise mechanisms at play to deepen our understanding of these dynamics, a point we expand upon below.
Third, we sought to examine the relational complexity in buyer–supplier dyads not only in terms of who the entities are but also how these entities present themselves as a critical contingency of the power-trust effect. We identify the highly contingent nature of the partner's level of self-promotion communication by drawing on symbolic interactionism (Goffman, 1959) to differentiate high vs. low self-promoters. Specifically, we found that the negative partner effect is weaker when interacting with a partner that engages in high (vs. low) levels of self-promotion communication, which may be explained by the fact that a focal organization's ability to collect more information on its partner alleviates some of the uncertainty regarding a high-powered partner's conduct. This finding suggests that the power-trust association derived from the rational choice perspective is more applicable to interactions with low (rather than high) self-promoters. However, the pattern of low-power focal organizations placing high trust in their partners (the actor effect) does not vary significantly between low and high self-promoters, thus suggesting that the power-trust association derived from the behavioral perspective is applicable independent of the partner's level of self-promotion. Theoretically, the predictions resulting from the rational choice and behavioral perspectives are not fundamentally contradictory. Instead, we highlight their different foci and assumptions and their ultimate complementarity to understand the power-trust effect in buyer–supplier dyads. As such, the crucial question for researchers is not so much which of the two theoretical perspectives is “right” but rather when.
Fourth, our findings speak to recent debates surrounding trust asymmetries in buyer–supplier dyads (Brito and Miguel, 2017; Graebner et al., 2020). In discussions of why one party to an exchange may place higher levels of trust than its counterpart, prior research tends to assume that such situations would be only transitional in nature and that, ultimately, partner organizations’ trust levels would match (Jeffries and Reed, 2000). Our research suggests that this does not have to be the case whenever persistent power differences exist. Low-powered organizations tend to have high levels of trust, while trust tends to be lower among organizations that are high in power. Therefore, our study points to the importance of future investigations accounting for the challenge of theorizing trust as a bilateral and possibly enduring asymmetric phenomenon that cannot be fully understood without accounting for each party's unique viewpoint in buyer–supplier relationships.
Finally, our post hoc analysis indicates that the effect of actor power on trust may depend on the level of partner power, and vice versa. Although not hypothesized, we found suggestive evidence of an interaction effect such that the difference in trust between low- and high-power actors appears larger when the partner has high (vs. low) levels of power. These findings offer a novel—and more nuanced—insight about the dyadic dynamics that affect the relationship between power and trust in buyer–supplier relationships.
From a methodological viewpoint, we demonstrate the importance of employing techniques that are able to capture the dyadic nature of buyer–supplier relationships. We used a multi-informant, matched dyadic design to study buyers and suppliers. Our study is clearly not the first to use dyadic data in interorganizational scholarship (e.g., Villena and Craighead, 2017), but it moves research forward by conceptualizing and testing fundamental aspects of the relational complexity of these exchanges. Taking advantage of developments in the statistical modeling of dyadic data, we adopted the APIM, which allowed us to simultaneously model partner and actor effects (Kenny, 2018). To our knowledge, our study is the first to apply APIM to the study of interorganizational relationships. Without using proper dyadic techniques, researchers risk developing only a partial understanding of buyer–supplier relationships, no matter what conceptual perspective they draw on.
Limitations and Future Research
As in any other research study, we had to make conceptual and empirical choices, which resulted in certain limitations that open up directions for future research. First, we acknowledge that our direct mapping of rational-actor theory to the partner effect and the behavioral perspective to the actor effect may be somewhat reductive, and that both rational calculations and motivated reasoning may jointly contribute to explaining both effects. Instead of a rigid rational-versus-behavioral divide, the partner and actor perspectives reflect two broad theoretical streams within existing work on power–trust dynamics. As such, we need future conceptual and empirical research to disentangle the precise underlying mechanisms and provide a more nuanced understanding of how power-trust dynamics unfold in interorganizational relationships. Beyond juxtaposing encapsulated-interest and motivated-cognition mechanisms, such research may also endeavor to identify and test mechanisms not elaborated here. For instance, it may be possible that expected reciprocity may serve as yet another mechanism that aligns with a behavioral perspective, such that a low-power actor trusts the partner more because they expect high levels of trust from their counterpart (e.g., Korsgaard, 2018). While the APIM model does account for such interdependencies by controlling for the correlation between the outcomes, future research is required to further develop the theoretical argument and design appropriate (ideally, longitudinal) studies to test it.
Second, we prioritized a multi-informant and matched-sample design and collected dyadic data from a large and diverse group of respondents, enhancing the external validity of the findings. However, an important direction for future studies involves complementing our methodological approach with experimental research designs, which are uniquely suitable for ruling out potential confounds and identifying relevant underlying mechanisms (Levine et al., 2023). We see great value in both behavioral experiments (e.g., Weber and Bauman, 2019) and vignette experiments (e.g., Eckerd et al., 2022), which can be conducted either in the lab or online with the goals of manipulating each partner's power and observing the causal effects on levels of trust. Such experiments could also endeavor to capture the theoretical mechanisms suggested by the rational choice and behavioral perspectives (namely, anticipated opportunism and motivated cognition), thereby making important contributions to disambiguating these perspectives more conclusively. While ambitious in dyadic research, scholars should aim for large samples and study designs that boost statistical power. In addition, we encourage longitudinal data collection efforts to examine more systematically, for example, how a focal organization's trust can be a response to the partner's trust (or vice versa). Longitudinal data would facilitate a deeper analysis of the dynamics and processual mechanisms that characterize the relationship between power and trust.
Third, we encourage researchers to further examine our exploratory finding that the impact of actor (partner) power on trust may depend on the level of partner (actor) power (using robust research designs. For example, experimental research can be used to create a scenario where an actor (e.g., a carmaker) and a partner (e.g., a manufacturer) negotiate a contract for the supply of electric vehicle batteries. Both perceive differing levels of power, which can be manipulated by providing each participant with a role description tailored to manipulate (perceived) power. Using a 2 × 2 experimental design, participants are randomly allocated to each condition. Future research that examines whether the interaction between the impact of actor power and partner power on levels of trust can unpack the complex relational dynamics governing buyer–supplier relationships.
Fourth, we conducted our empirical study in China, which, like any other country, is characterized by a certain set of cultural factors—such as high collectivism and power distance—that may differ from those of other parts of the world. On the one hand, we have no reason to believe that any of these factors act as strong scope conditions that eliminate the effects we examined. Although our study focuses on China, the findings can still provide valuable insights into buyer–supplier relationships more generally, to the very least in other countries with similar levels of power distance as China, such as Mexico, Indonesia, and Nigeria (Hofstede, 2001). On the other hand, we acknowledge that China's distinct cultural and political institutions harbor certain peculiarities that may limit the extent to which our exact findings can be replicated elsewhere (Özer et al., 2014; Zhou and Poppo, 2010). Therefore, we specifically encourage future research to further extend our research with a multicountry approach. By considering cultural nuances and conducting comparative research across cultures, future research can gain a more comprehensive understanding of the intricacies of culture-specific power dynamics in buyer-supplier relationships.
Fifth, it would be worthwhile to explore how our findings can be applied and generalized to other types of dyadic exchanges (e.g., horizontal alliances). Relatedly, future research should also be extended to study the complex role of power in supply chain relationships involving nonprofit organizations. Such research would facilitate the development of research on operations management that addresses the basic success factors of operating socially and environmentally sustainable supply chains (Sunar and Swaminathan, 2022).
Finally, our study focused on the partner's self-promotion communication as a critical contingency. Particularly, we tested the moderating role of self-promotion when there is some degree of costliness and verifiability of the claims (i.e., not mere cheap talk), but future research might also focus on cheap talk. We also see exciting opportunities for exploring additional contingencies to the effect of power on trust in buyer–supplier relationships. For instance, task interdependence (e.g., pooled, sequential, and reciprocal) and corporate culture (e.g., transparency and hierarchy) may plausibly have relevant implications for the effect of power on trust in buyer–supplier relationships.
Supplemental Material
sj-pdf-1-pao-10.1177_10591478251371270 - Supplemental material for The Influence of Power on Trust in Buyer–Supplier Relationships: An Actor−Partner Interdependence Approach
Supplemental material, sj-pdf-1-pao-10.1177_10591478251371270 for The Influence of Power on Trust in Buyer–Supplier Relationships: An Actor−Partner Interdependence Approach by Nuno Oliveira, Oliver Schilke, Fabrice Lumineau and Baofeng Huo in Production and Operations Management
Footnotes
Acknowledgements
We gratefully acknowledge the comments on earlier versions of this article provided by Anna Brattström, Zhi Cao, and Sarah Doyle. Maksim Sitnikov provided valuable research assistance. We also acknowledge comments made by participants at the Academy of Management Annual Conference (2024) and the Strategic Management Conference (2024), as well as attendees of research seminars at EM Lyon (STORM), Groningen University, Hong Kong University, and Hong Kong Baptist University.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by funding from the National Natural Science Foundation of China (#72432009, #72091210/72091214, and #71525005) to the last author. Research support was also provided by a National Science Foundation CAREER Award (#1943688) granted to the second author. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation.
Notes
How to cite this article
Oliveira N, Schilke O, Lumineau F and Huo B (2026) The Influence of Power on Trust in Buyer–Supplier Relationships: An Actor–Partner Interdependence Approach. Production and Operations Management 35(4): 1285–1312.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
