Role-relationships are the backbone and central nervous system of most organizations (Blau, Reference Blau1964; Katz & Kahn, Reference Katz and Kahn1978). Relationships at work, as such, both form a ‘contemporary embodiment of how most work gets accomplished’ (Ferris, Liden, Munyon, Summers, Basik, & Buckley, Reference Ferris, Liden, Munyon, Summers, Basik and Buckley2009: 1379) and how subordinates often find meaning in their jobs (Mao, Hsieh, & Chen, Reference Mao, Hsieh and Chen2012; Vandenberghe, Bentein, & Panaccio, Reference Vandenberghe, Bentein and Panaccio2017). As such, research on work-based role-relationships holds a popular place in both the academic (Dutton & Ragins, Reference Dutton and Ragins2007; Ferris et al., Reference Ferris, Liden, Munyon, Summers, Basik and Buckley2009; Hu & Judge, Reference Hu and Judge2017; Shen, Chou, & Schaubroeck, Reference Shen, Chou and Schaubroeck2019; Sias, Reference Sias2009; Sluss, Ployhart, Cobb, & Ashforth, Reference Sluss, Ployhart, Cobb and Ashforth2012; Topal, Reference Topal2015) and practitioner literatures (Grant, Reference Grant2013; Rath, Reference Rath2006; Sluss, Reference Sluss2020). Overall, role-relationships are viewed as a ‘nexus of transactions’ across which organizational members cooperate (and/or compete) in order to minimize costs and maximize benefits (cf. Williamson, Reference Williamson1985). Thus, it is no surprise that social exchange has become predominant when analyzing relationships across phenomena such as leadership, negotiation, psychological contracts, organizational commitment, and perceived support (Graen & Uhl-Bien, Reference Graen and Uhl-Bien1995; for review see Cropanzano, Anthony, Daniels, & Hall, Reference Cropanzano, Anthony, Daniels and Hall2017).
Nevertheless, repeated exchange-based transactions often go hand-in-hand with relational interactions – interactions wherein organizational members get to know each other (Allen, Eby, Chao, & Bauer, Reference Allen, Eby, Chao and Bauer2017). This personalization – getting to know each other – influences both the relationship's future as well as attitudes and behaviors toward the organization in which the relationship is embedded (Sluss et al., Reference Sluss, Ployhart, Cobb and Ashforth2012). As a result, role-relationships provide multiple opportunities for organizational members to fulfill not only task-based needs but psychosocial needs – thereby expanding and enriching work roles (e.g., Clark & Mills, Reference Clark and Mills1979; Gersick, Bartunek, & Dutton, Reference Gersick, Bartunek and Dutton2000; Parker, Reference Parker1998; Sluss & Ashforth, Reference Sluss and Ashforth2008; Walumbwa & Hartnell, Reference Walumbwa and Hartnell2011). One particularly poignant psychosocial need is that of identity (e.g., Ashforth & Mael, Reference Ashforth and Mael1989; Bednar, Galvin, Ashforth, & Hafermalz, Reference Bednar, Galvin, Ashforth and Hafermalz2020; Pratt, Reference Pratt, Whetten and Godfrey1998).
While not as prevalent as relational exchange, research has also explored identity-based components and processes within role-relationships via relational identification (RI) (Brewer & Gardner, Reference Brewer and Gardner1996; Brickson, Reference Brickson2000; Flynn, Reference Flynn2005; Leavitt & Sluss, Reference Leavitt and Sluss2015; Sluss et al., Reference Sluss, Ployhart, Cobb and Ashforth2012; Sluss & Ashforth, Reference Sluss and Ashforth2007; Sluss & Ashforth, Reference Sluss and Ashforth2008). RI is defined as the ‘partial definition of oneself in terms of [the] role relationship’ (Sluss & Ashforth, Reference Sluss and Ashforth2007: 15) – in short, including the relationship in ‘who we are.’ Indeed, RI significantly and positively influences important attitudes and behaviors such as: forgiveness and relationship resilience (Thompson & Korsgaard, Reference Thompson and Korsgaard2019); extra-role and in-role performance (e.g., Hu & Judge, Reference Hu and Judge2017; Shen, Chou, & Schaubroeck, Reference Shen, Chou and Schaubroeck2019); voice behavior (e.g., Zhu, He, Treviño, Chao, & Wang, Reference Zhu, He, Treviño, Chao and Wang2015); creative behavior (e.g., Qu, Janssen, & Shi, Reference Qu, Janssen and Shi2015; Yoshida, Sendjaya, Hirst, & Cooper, Reference Yoshida, Sendjaya, Hirst and Cooper2014); engagement (e.g., Liang, Chang, Ko, & Lin, Reference Liang, Chang, Ko and Lin2017); and organizational identification (e.g., Sluss et al., Reference Sluss, Ployhart, Cobb and Ashforth2012). As robust as these findings are, scholars have yet to weigh RI's influence alongside that of relational exchange quality (REQ) – the predominant explanation for how relationships influence important attitudes and behaviors.
We propose that investigating how RI complements REQ may provide a more complete view of relationships at work. Role-relationships, for example, do not exist as simplistic or unidimensional entities. Role-relationships encompass the breadth of dyadic experience – from tangible and intangible exchanges (due to organizationally-determined or other institutionally-determined role transactions) to the feelings of connectedness and belonging, often accompanying deeper, more personalized, interactions (Sluss & Ashforth, Reference Sluss and Ashforth2007; Szulc, Reference Szulc2020). We contend that role-relationships within organizations encompass both exchange-based and identity-based processes building upon Clark and Mills (Reference Clark and Mills1979) classic distinction that relationships are built upon two essential components: exchange and communal processes (see also Clark, Mills, & Powell, Reference Clark, Mills and Powell1986; Dutton & Ragins, Reference Dutton and Ragins2007). Our research focuses on how RI provides an important complementary glimpse (to that of REQ) into role-relationships at work. In short, we attempt to see how the communal (or identity) process can complement that of exchange (cf. Clark & Mills, Reference Clark and Mills1979).
In two temporally-lagged studies, we examine the dual influence of identification and exchange vis-à-vis the immediate supervisor–subordinate relationship. The supervisor–subordinate relationship is one of the most salient within organizational life (Lord, Brown, & Freiberg, Reference Lord, Brown and Freiberg1999; Masterson, Lewis, Goldman, & Taylor, Reference Masterson, Lewis, Goldman and Taylor2000), often having the potential to fulfill both the subordinate's task and psychosocial needs (e.g., Chen, Tsui, & Farh, Reference Chen, Tsui and Farh2002). To conceptualize and operationalize identification, we utilize RI with the supervisory relationship. RI – the ‘partial definition of oneself in terms of [the] role relationship’ (Sluss & Ashforth, Reference Sluss and Ashforth2007: 15) – focuses on the role-relationship – versus the relational other or the collective memberships involvedFootnote 1. To capture exchange within supervisory relationships, we focus on the relational exchange component of leader-member exchange quality (LMX) – what we call here, REQ (Bernerth, Armenakis, Feild, Giles, & Walker, Reference Bernerth, Armenakis, Feild, Giles and Walker2007; Graen & Uhl-Bien, Reference Graen and Uhl-Bien1995). Leader-member exchange, although built upon the logic of social exchange theory (Blau, Reference Blau1964), has grown over the years to capture a varied constellation of relationship quality outcomes such as liking, trust, and other components less focused on exchange (Bernerth et al., Reference Bernerth, Armenakis, Feild, Giles and Walker2007; Zhou & Schriesheim, Reference Zhou and Schriesheim2009). Therefore, we focus exclusively on the exchange-based component of the supervisory relationshipFootnote 2 to better capture how identification complements exchange. A significant body of research demonstrates that REQ is associated with important attitudes and behaviors (Gerstner & Day, Reference Gerstner and Day1997; Ilies, Nahrgang, & Morgeson, Reference Ilies, Nahrgang and Morgeson2007; Wayne, Shore, & Liden, Reference Wayne, Shore and Liden1997); meanwhile, a growing corpus of research suggests that RI also influences key outcomes (see cited research above). However, existing empirical research has not examined whether RI indeed makes a complementary contribution to key individual-level outcomes (after accounting for REQ).
Although important individual-level outcomes exist within organizational research, we focus on four outcomes across two major categories – relationally-focused and organizationally-focused. For relationally-focused outcomes, we examine the association of RI (after accounting for REQ) with interpersonal citizenship behaviors (ICBs; both person-focused and task-focused) wherein these prosocial helping behaviors are important facilitators of cooperation and workplace productivity (e.g., Podsakoff, Whiting, Podsakoff, & Blume, Reference Podsakoff, Whiting, Podsakoff and Blume2009; Settoon & Mossholder, Reference Settoon and Mossholder2002). For organizationally-focused outcomes, we examine the association of RI (again, after accounting for REQ) with job satisfaction and affective organizational commitment.
Investigating the dual influence of RI and exchange within role-relationships is important for at least three reasons. First, although having informed our understanding of role-relationships at work, social exchange also appears to have dominated scholarly research at the expense of other explanatory frameworks (cf. Cropanzano et al., Reference Cropanzano, Anthony, Daniels and Hall2017; Dutton & Ragins, Reference Dutton and Ragins2007). Our studies contribute to organizational research on relationships in that we explore RI and REQ side-by-side. Indeed, we do not negate the influence of exchange – yet we do suggest that identification also matters in a significant way. Second, our studies contribute to the burgeoning literature on positive role-relationships by supplying needed evidence on how both RI and REQ influence positive outcomes such as increased helping behaviors, satisfaction, and commitment at work (Dutton & Ragins, Reference Dutton and Ragins2007). We add to this line of research by exploring the positive effects of RI. Third, we contribute to research on how the subordinate's self-concept influences the leadership process (e.g., Epitropaki, Kark, Mainemelis, & Lord, Reference Epitropaki, Kark, Mainemelis and Lord2017; Lord & Brown, Reference Lord and Brown2004; see also seminal works for the relational self in social psychology; Aron & Aron, Reference Aron, Aron, Ickes and Duck2000; Brewer & Gardner, Reference Brewer and Gardner1996). Scholars have proposed that leadership effectiveness is, at least in part, determined by the follower's (in our case, the subordinate's) self-concept. Nevertheless, the bulk of empirical evidence focuses on the self as a member of a collective (e.g., work team, organization, occupation) but not as a partner in a relationship. Our study provides needed evidence that the relational self-concept matters and complements relational exchange in the leadership process.
Relational exchange quality and relational identification
REQ and RI represent relational components that are distinct (albeit interrelated) across theoretical underpinnings, conceptualizations, and nomological networks. Building upon social exchange theory (Blau, Reference Blau1964), scholars posit that leader–subordinate relationships vary in REQ, or the ‘perceived value of the tangible and intangible commodities exchanged’ within the relationship (Wayne, Shore, & Liden, Reference Wayne, Shore and Liden1997: 84). High-quality relationships are characterized by increased levels of support and effort (Graen & Uhl-Bien, Reference Graen and Uhl-Bien1995; Maslyn & Uhl-Bien, Reference Maslyn and Uhl-Bien2001). Subordinates who perceive higher REQ reciprocate the leaders' contributions through attitudes and behaviors that support the leader and the organization (Wayne, Shore, & Liden, Reference Wayne, Shore and Liden1997). As such, REQ is positively related to satisfaction, commitment, performance, intentions to stay, and citizenship behaviorsFootnote 3 (Gerstner & Day, Reference Gerstner and Day1997).
However, REQ is only part of the relational story at work. Recent research exploring RI offers a distinct but complementary view of supervisory relationships. Drawing from research on the relational self (e.g., Brewer & Gardner, Reference Brewer and Gardner1996), scholars suggest that role-relationships (such as that of the subordinate-supervisor) are integral to how individuals define themselves at work (Flynn, Reference Flynn2005; Sluss & Ashforth, Reference Sluss and Ashforth2007). RI is influenced by two major relational components: (1) the respective roles within the relationship and (2) the persons occupying these roles (Sluss & Ashforth, Reference Sluss and Ashforth2007). Specifically, the role-based component includes the goals, values, norms, tasks, and expectations attached to the respective roles (Ashforth, Reference Ashforth2001). At the same time, the role occupants influence the nature of the relationship via the personalized way they enact the role. These two components (and their perceived valence, whether positive or negative), in turn, influence the degree to which the focal individual identifies with the relationship. In short, role-relationships ‘that meet one's task and social-psychological needs are likely to be seen as attractive, and therefore foster identification’ (Sluss & Ashforth, Reference Sluss and Ashforth2008: 810). More generally, the development of RI is inextricably tied to the extent to which the relationship fulfills identity-relevant needs such as the need for belonging, self-enhancement, self-distinctiveness, and self-consistency (Cooper & Thatcher, Reference Cooper and Thatcher2010; Tajfel & Turner, Reference Tajfel, Turner, Worchel and Austin1986) – again, all tied to ones' working self-concept (Markus & Wurf, Reference Markus and Wurf1987).
Incorporating the relationship into one's self-definition should encourage positive attitudes and behaviors such as empathy, liking, cooperation, in-role performance, and extra-role performance (Cooper & Thatcher, Reference Cooper and Thatcher2010). RI also likely influences attitudes and behaviors that target the relational other's collective (e.g., team, organization, occupational group; Bullis & Bach, Reference Bullis and Bach1989; Carmeli, Atwater, & Levi, Reference Carmeli, Atwater and Levi2011; Marique, Stinglhamber, Desmette, & Goldoni (Reference Marique, Stinglhamber, Desmette and Goldoni2014); Sluss and Ashforth, Reference Sluss and Ashforth2008). RI (in extreme cases) may also take on negative qualities such as codependency (i.e., over-identification) and, at times, collusion with corrupt behavior (Aron & Aron, Reference Aron, Aron, Ickes and Duck2000; Ashforth & Sluss, Reference Ashforth, Sluss, Kyriakidou and Ozbilgin2006; Sluss & Ashforth, Reference Sluss and Ashforth2007).
RI and REQ coexist and are mutually accessed within work role-relationships (Gouldner, Reference Gouldner1960; cf. Clark & Mills, Reference Clark and Mills1979), but they differ in motivational processes. RI and REQ arise out of different underlying processes – self-consistency/enhancement of the role-relationship (Cooper & Thatcher, Reference Cooper and Thatcher2010) and reciprocity of the resources exchanged, respectively (Graen & Uhl-Bien, Reference Graen and Uhl-Bien1995; Maslyn & Uhl-Bien, Reference Maslyn and Uhl-Bien2001). As such, a role-relationship may be positively tied to one's working self-concept (cf. Ibarra, Reference Ibarra1999; Markus & Wurf, Reference Markus and Wurf1987) yet not provide a high-quality exchange of tangible and intangible resources. Extending the previous example, the novice engineer may experience higher RI (for the reasons listed above) yet report lower REQ if the senior engineer does not provide the expected and/or desired resources (such as not orienting the novice at the beginning of a new project or sharing expertise only occasionally; Graen and Uhl-Bien, Reference Graen and Uhl-Bien1995). On the other hand, the focal individual may receive abundant resources, engendering perceptions of higher REQ, yet not perceive the role-relationship as fulfilling needs of his/her working self-concept (i.e., lower RI). Although more than likely correlated in practice, RI and REQ remain independent relational components with distinct underlying processes.
Relationally-focused outcomes: interpersonal citizenship behaviors
Citizenship or helping behaviors are of particular interest for organizations and their members (Bergeron, Reference Bergeron2007; Grant, Dutton, & Rosso, Reference Grant, Dutton and Rosso2008). Today's work environment is flatter, less hierarchal, and therefore much more dynamic. Organizations and their members need to be flexible and adaptable in getting work done. Work is therefore facilitated when organizational members benefit from another member's knowledge, skills, and/or abilities. Indeed, helping behaviors seem to be an apt tool to maintain organizational flexibility whilst mitigating the chaos that, at times, accompanies such a dynamic working environment (Bednar et al., Reference Bednar, Galvin, Ashforth and Hafermalz2020; Ocampo, Acedillo, Bacunador, Balo, Lagdameo, & Tupa, Reference Ocampo, Acedillo, Bacunador, Balo, Lagdameo and Tupa2018). In this light, organizational scholars have investigated helping behaviors that benefit both the organization (i.e., organizational citizenship behaviors, Podsakoff et al., Reference Podsakoff, Whiting, Podsakoff and Blume2009) and organizational members (i.e., ICBs, Settoon & Mossholder, Reference Settoon and Mossholder2002). In our two studies, we focus on supervisor-directed ICBs – thereby focusing on behaviors that help a relational partner, yet indirectly help the organization.
Building upon previous research on helping behaviors (e.g., Van Dyne & LePine, Reference Van Dyne and LePine1998), Settoon and Mossholder (Reference Settoon and Mossholder2002) conceptualized (and operationalized) two distinct types of ICBs: (1) person-focused and (2) task-focused. They describe them in the following manner:
‘Person-focused ICB provides for self-esteem maintenance and deals with problems of a more personal nature…. often [having] an affilitative-promotive character (Van Dyne & LePine, Reference Van Dyne and LePine1998), [and] being grounded in friendship and social support. ….In contrast, task-focused ICB involves the resolution of work-related problems of a less personal nature and deals with organization-based issues…. [being] more instrumental, arising in the course of work-role performance and involving the exchange of job-related resources’ (Settoon & Mossholder, Reference Settoon and Mossholder2002: 256).
Although these behaviors (in their proposed content) revolve around expressive or instrumental concerns separately, Settoon and Mossholder found that both instrumental (i.e., network centrality) and expressive (i.e., empathetic concern; see also Grant and Mayer, Reference Grant and Mayer2009) variables predicted both types of ICBs. We also argue that both REQ and RI will have direct ties to supervisory-directed ICBs. Scholars, building upon exchange and reciprocity logic, have found consistent associations between REQ and citizenship behaviors (both interpersonal and organizational) – in short, the subordinate will ‘return the favor’ (of higher REQ) via increased helping behaviors (see meta-analysis; Ilies, Nahrgang, & Morgeson, Reference Ilies, Nahrgang and Morgeson2007).
In related fashion, RI (with the supervisory relationship) should also increase the likelihood that the subordinate will implement supervisory-directed ICBs. RI occurs when the role-relationship fulfills important task and psycho-social needs and, thus, becomes important for the working self-concept. The subordinate, as a result, extends his/her self-concept to include the supervisory relationship (Aron & Aron, Reference Aron, Aron, Ickes and Duck2000). Given the role-relationship is part of the subordinate's self-concept, the subordinate will be motivated to maintain a positive relationship, and ICBs will serve to do just that (Babalola, Reference Babalola2016; Marstand, Epitropaki, van Knippenberg, & Martin, Reference Marstand, Epitropaki, van Knippenberg and Martin2020; Thompson & Simkins, Reference Thompson and Simkins2017). As RI increases, the subordinate will engage in both person- and task-focused ICBs because these behaviors (1) are consistent with the relational schema (i.e., routinized relational scripts) (Baldwin, Reference Baldwin1995; Tsai, Dionne, Wang, Spain, Yammarino, & Cheng, Reference Tsai, Dionne, Wang, Spain, Yammarino and Cheng2017) and (2) help maintain a positive self-concept since the relationship is included in the self (Aron & Aron, Reference Aron, Aron, Ickes and Duck2000; Sluss & Ashforth, Reference Sluss and Ashforth2007). Hence:
Hypothesis 1: RI with the supervisory relationship will be positively associated with person-focused ICBs (after accounting for the association of REQ).
Hypothesis 2: RI with the supervisory relationship will be positively associated with task-focused ICBs (after accounting for the association of REQ).
Organizationally-focused outcomes: job satisfaction and affective commitment
The supervisory relationship, as mentioned, tends to be highly salient in one's work experience (Gerstner & Day, Reference Gerstner and Day1997; Hogg, Martin, Epitropaki, Mankad, Svensson, & Weeden, Reference Hogg, Martin, Epitropaki, Mankad, Svensson and Weeden2005; Lin et al., Reference Lin, Wang, Bamberger, Zhang, Wang, Guo and Zhang2016). Indeed, subordinates traditionally perceive the supervisor as an agent of the organization (Neves, Reference Neves2012). As such, how the subordinate perceives the supervisory relationship is likely to impact the subordinate's view and attitude toward the organization – in our case, both the subordinate's job satisfaction (i.e., emotional and cognitive appraisal regarding one's job experiences; Locke, Reference Locke and Dunnette1976) and affective organizational commitment (i.e., an emotional attachment and involvement with the organization; Meyer & Allen, Reference Meyer and Allen1997).
Drawing from prior research, REQ should be positively associated with job satisfaction. When REQ is high, subordinates gain a sense of satisfaction in that they obtain increased resources (both tangible and intangible) for completing job tasks (Graen, Novak, & Sommerkamp, Reference Graen, Novak and Sommerkamp1982). Indeed, there is strong empirical evidence that REQ with the supervisor is positively related with job satisfaction (Gerstner & Day, Reference Gerstner and Day1997). We also argue that RI is likely to facilitate job satisfaction. RI with the supervisory relationship helps fulfill one's needs for relatedness at work – which is associated with overall satisfaction and well-being (Baard, Deci, & Ryan, Reference Baard, Deci and Ryan2004; cf. Locke, Reference Locke and Dunnette1976). More specifically, positive affect due to RI will spill over to satisfaction toward the subordinate's job – given that the supervisory relationship and the job are generally intertwined via task and reward allocation. For example, Raabe and Beehr (Reference Raabe and Beehr2003) found that the subordinate's positive affect toward the supervisor was positively related with the subordinate's affective evaluation of the job (i.e., job satisfaction). Thus, we hypothesize:
Hypothesis 3: RI with the supervisory relationship will be positively associated with job satisfaction (after accounting for the association of REQ).
As with job satisfaction, REQ should be positively related with affective organizational commitment. In short, subordinates reciprocate resources in diverse ways. Although it is likely that resources provided by the supervisor will be reciprocated with attitudes and behaviors that are directed to the supervisor (Wilson, Sin, & Conlon, Reference Wilson, Sin and Conlon2010), it is also likely that the subordinate will reciprocate toward the organization – given the supervisor is an organizational agent. One such reciprocation is increased affective commitment to the organization. The increased resources, via high REQ with the supervisory relationship, provide a generalized sense of obligation or commitment to the organization (Stinglhamber & Vandenberghe, Reference Stinglhamber and Vandenberghe2003). Indeed, REQ has been found to be significantly related to affective organizational commitment (Epitropaki & Martin, Reference Epitropaki and Martin2005; Gerstner & Day, Reference Gerstner and Day1997; Hom, Tsui, Lee, Ping Ping, Wu, & Zhang, Reference Hom, Tsui, Lee, Ping Ping, Wu and Zhang2009). Furthermore, given that the supervisor is an organizational agent (Neves, Reference Neves2012), incorporating the supervisory relationship into one's self-definition (i.e., RI) should increase the subordinate's tendency to be concerned for and involved with the organization, thus increasing affective organizational commitment. For example, empirical research has found that including another into one's self-concept increases one's commitment to shared goals with the relational partner (Shah, Reference Shah2003). Indeed, supervisors (via institutionalized role expectations) traditionally regard increased subordinate attachment as a supervisory goal (e.g., Bass & Avolio, Reference Bass and Avolio1994; Stryker & Burke, Reference Stryker and Burke2000). In related fashion, Johnson and Chang (Reference Johnson and Chang2006), found that thinking of oneself in terms of the ‘we’ self-concept was related with affective commitment whereas the ‘I’ self-concept was not. We argue that RI increases the potential to think of oneself as ‘we’ with the organization (rather than ‘they’ or ‘I’; cf. Mael & Ashforth, Reference Mael and Ashforth1992; see also Sluss et al., Reference Sluss, Ployhart, Cobb and Ashforth2012) and therefore facilitate affective commitment with the organization (see also, Lux, Grover, & Teo, Reference Lux, Grover and Teo2019). Thus, we hypothesize:
Hypothesis 4: RI with the supervisory relationship will be positively associated with affective organizational commitment (after accounting for the association of REQ).
Overview of studies
Study 1 investigated the contribution of RI (after accounting for REQ) within a sample of (primarily) part-time employed undergraduate business students at a large university in the southeast United States (i.e., hypotheses 1 through 4). Next, in study 2, we generalized our findings from study 1 to a sample of primarily full-time professional master of business administration (i.e., PMBA) students from the same university. Our samples provided us access to individuals working in a wide range of organizations, occupations, and industries, and, thus allowed us to control for possible confounding organizational, occupational, or industry-specific effects (e.g., Dierdorff & Rubin, Reference Dierdorff and Rubin2007).
Study 1 method
Sample and procedure
We conducted a longitudinal study among working undergraduate business students at a large university in the southeast United States. Study 1 consisted of two temporally-lagged (i.e., one month) surveys. Study participants received extra credit in exchange for participation. Both surveys were administered by a member of the research team. The first survey measured RI, REQ, as well as demographic and control variables (see below). The second survey measured the relevant dependent variables (i.e., person-focused ICB, task-focused ICB, job satisfaction, and affective organizational commitment). In order to protect confidentiality, the respondents were instructed to create a code number in order to match responses across surveys.
Following Podsakoff, MacKenzie, Lee, and Podsakoff (Reference Podsakoff, MacKenzie, Lee and Podsakoff2003), we implemented various design elements to reduce the potential of common method bias – given our reliance on self-report measures. First, we separated, via a temporal lag of one month, our predictor and criterion variables, thereby reducing consistency motifs and implicit theories that may occur in a cross-sectional design. Second, we attempted to reduce social desirability bias through promoting participant anonymity. The participants received the surveys directly from (and returned them to) the research team. Neither the course instructor nor the employing organizations were privy to the data, thus promoting participant anonymity (i.e., the organization did not have to know they were participating in the study).
There were 298 respondents to the first survey and, approximately one month later, 311 respondents to the second survey. The number of respondents who completed both surveys was 231. However, there were 79 respondents that were not employed or provided incomplete data and, thus, were dropped from the study. As such, we obtained a final sample of 152 working respondents. Fifty-seven percent were female and 80% were Caucasian, with a mean age of 21.5 years (SD = 4.50). Seventy-six percent worked in part-time positions. Respondents had an average organizational tenure of 22 months (SD = 30.16), an average relationship tenure with their immediate supervisor of 37 months (SD = 62.64), and an average work experience of 37 months (SD = 43.90). They worked in a very diverse group of industries. The most dominant industries represented were leisure/hospitality (23.8%), food (19.7%), services (16.4%), and retail (14.5%). There were no mean differences across industries.
Attrition analysis
We conducted analyses to confirm that attrition did not create non-random effects. As prescribed by Goodman and Blum (Reference Goodman and Blum1996), we employed multiple logistic regression with the dependent variable being a dichotomous variable distinguishing between ‘stayers’ (respondents to both Time 1 and Time 2 surveys) and ‘leavers’ (respondents to Time 1 survey only). We tested for any systematic differences across ‘stayers’ and ‘leavers.’ We included the demographic control variables (age, gender, racioethnicity, organizational tenure, relationship tenure, and work experience) as well as the two main independent variables (i.e., RI and REQ). None of the variables significantly predicted ‘staying’ or ‘leaving’ the study. As such, we conclude that attrition bias was likely not present in the data (Goodman & Blum, Reference Goodman and Blum1996).
Measures
The first survey consisted of measures that assessed RI and REQ as well as control variables (i.e., age, racioethnicity, gender, relationship tenure with immediate supervisor [in months], tenure with the current organization [in months], and previous work experience [in months]). The second survey (one month later) measured person-focused ICB, task-focused ICB, job satisfaction, and affective organizational commitment.
Relational identification
We measured RI with the supervisory relationship via a three-item scale developed by Sluss (Reference Sluss2006). The response scale ranged from 1 (‘disagree strongly’) to 7 (‘agree strongly’). Sluss (Reference Sluss2006) obtained α reliabilities well above. 70 and significant factor loadings in both exploratory (N = 158) and confirmatory (N = 215) samples. The adapted scale includes the following items: ‘My relationship with my supervisor is important to how I see myself at work’; ‘If someone criticized my relationship with my supervisor, it would feel like a personal insult’; and ‘My relationship with my supervisor reflects the kind of person I am at work’ (α reliability = .89).
Relational exchange quality
Given our desire to measure global REQ, we followed Graen and Uhl-Bien's (Reference Graen and Uhl-Bien1995) recommendation to use a unidimensional scale. We adapted Graen, Novak, and Sommerkamp's (Reference Graen, Novak and Sommerkamp1982) LMX-7 scale, building upon subject matter expert (SME) findings in Bernerth et al., (Reference Bernerth, Armenakis, Feild, Giles and Walker2007). Bernerth et al. argued that current measures of LMX do not adequately address the relational exchange component of the supervisory relationship. As a result, Bernerth et al. relied on SMEs to rate which items of the LMX7 scale were most aligned with relational exchange. This analysis pointed to three items as most conceptually aligned with relational exchange: (a) ‘How well does your supervisor recognize your potential?’; (b) ‘Regardless of how much formal authority he/she has built into his/her position, what are the chances that your supervisor would use his/her power to help you solve problems in your work?’; and (c) ‘Regardless of the amount of formal authority your supervisor has, what are the chances that he/she would ‘bail you out,’ at his/her expense?’ We used these three items to operationalize REQFootnote 4. Respondents used a 5-point Likert-type scale (with each item having its own anchor descriptions) (αreliability = .75).
Interpersonal citizenship behaviors
The ICB scale consisted of six of the 15 items developed by Settoon and Mossholder (Reference Settoon and Mossholder2002) to measure ICBs. Consistent with prior research, the six items were the items that Settoon and Mossholder (Reference Settoon and Mossholder2002) found to have the highest factor loadings (see de Jong, van der Vegt, & Molleman, Reference de Jong, van der Vegt and Molleman2007; van der Vegt, Bunderson, & Oosterhof, Reference van der Vegt, Bunderson and Oosterhof2006). These items were set to a 5-point Likert-type scale ranging from 1 (‘seldom’) to 5 (‘always’). Of the six items, three were person-focused, including (as an illustration) ‘I listen to this supervisor when he/she has to get something off his/her chest’ and three were task-focused (as an illustration) ‘I help this supervisor with difficult assignments, even when assistance is not directly requested’ (α reliabilities [both scales] = .91).
As mentioned, we assessed ICBs directly from the subordinates (i.e., self-report). This warrants explanation. While there are many studies that address general citizenship behaviors (either those aimed at co-workers [generally] or the organization; e.g., Grant & Mayer, Reference Grant and Mayer2009; Mayer & Gavin, Reference Mayer and Gavin2005), there are fewer studies that examine citizenship behaviors directed at specific supervisors. While some of these studies rely on supervisor-report data (Choi, Reference Choi2008; Kamdar & Van Dyne, Reference Kamdar and Van Dyne2007; Liao & Rupp, Reference Liao and Rupp2005; Van Dyne, Kamdar, & Joireman, Reference Van Dyne, Kamdar and Joireman2008); these studies minimize social desirability by directly accessing the supervisors via the employing organization (independent of the subordinate). Our study design, on the other hand, required us to obtain data from employees in multiple organizations across multiple industries so as not to introduce confounding organizational or industry effects. Accessing the supervisor (in our study design) would mean doing so via the subordinate and thereby significantly lessening confidentiality. As such, other-report data would have the strong potential to elicit socially desirable responses from the supervisor (with regard to ICBs) and thereby threaten validity (cf. Shalley, Gilson, & Blum, Reference Shalley, Gilson and Blum2009; see also examples of self-report helping behaviors – de Cremer, Mayer, van Dijke, Schouten, & Bardes, Reference de Cremer, Mayer, van Dijke, Schouten and Bardes2009; Glomb & Welsh, Reference Glomb and Welsh2005).
Job satisfaction
We measured job satisfaction via a three-item global job satisfaction scale from Cammann, Fichman, Jenkins, and Klesh (Reference Cammann, Fichman, Jenkins and Klesh1983). The response scale ranged from 1 (‘disagree strongly’) to 7 (‘agree strongly’). A sample item is ‘All in all, I am satisfied with my job’ (α reliability = .85).
Affective organizational commitment
We measured affective organizational commitment via Meyer and Allen's (Reference Meyer and Allen1997) Organizational Commitment Scale (see also Meyer and Allen, Reference Meyer and Allen1990). A sample item from the six-item affective organizational commitment scale is ‘I do not feel ‘emotionally attached’ to this organization’ (reversed-coded) (response scale from 1 [‘disagree strongly’] to 7 [‘agree strongly’]). We obtained an α reliability of .89.
Control variables
Our control variables included age, racioethnicity, gender, supervisory relationship tenure (in months), organizational tenure (in months), previous work experience (in months), and part-time position (vs. full-time). The first three variables (i.e., age, racioethnicity, and gender) are fairly common demographic control variables having an influence on outcomes. However, we also suspected that the relationship tenure, organizational tenure, previous work experience, and part-time status might also covary with outcomes. In accordance with Becker (Reference Becker2005), we examined our model both with and without control variables. Because our parameter estimates were very similar with and without our control variables, we report our findings without control variables.
Discriminant analysis
Given our focus on discriminating between RI, REQ, ICBs, job satisfaction, and affective organizational commitment, we tested for discriminant validity via confirmatory factor analyses. We compared the hypothesized six-factor model (i.e., REQ, RI, person-focused ICB, task-focused ICB, job satisfaction, and affective commitment as separate factors) to: (1) a five-factor combined model (i.e., wherein REQ and RI combine into one factor with the other four variables remaining separate factors); (2) a four-factor combined model (i.e., wherein REQ and RI combine into one factor, satisfaction and commitment combine into another with the remaining two as separate factors); (3) a three-factor model (i.e., wherein REQ and RI combine into one factor, satisfaction and commitment combine into another, and task-focused and person-focused ICBs combine into a third factor); (4) a two-factor model (i.e., REQ and RI combine into one factor, the other four form an outcome factor); and finally (5) a one-factor omnibus model.
We consider model fit to be good if the comparative fit index (CFI) is >.90 (⩾.95 is excellent), the Tucker-Lewis coefficient (TLI) is >.90 (⩾.95 is excellent), and the root-mean-square error of approximation (RMSEA) is <.08 (⩽.06 is excellent) (Hu & Bentler, Reference Hu and Bentler1999; Jackson, Gillaspy, & Purc-Stephenson, Reference Jackson, Gillaspy and Purc-Stephenson2009; Kline, Reference Kline2005). The six-factor model demonstrated the best comparative fit as well as a good overall fit (χ2 = 265.00 [df = 172]; CFI = .95; TLI = .94; RMSEA = .06; p-value test for close fitFootnote 5 [RMSEA < .05] = .24). Note that the hypothesized model obtained a significant improvement in χ2 as compared to each of the alternative models (see Table 1). We also compared just RI and REQ (i.e., two-factor vs. a one-factor model) and found the RI and REQ to be factorially distinct. Thus, we treat the six variables as separate constructs in all analyses.
Table 1. Study 1 confirmatory factor analysis
N = 152. REQ, relational exchange quality; RI, relational identification; TICB, task-focused interpersonal citizenship behavior; PICB, person-focused interpersonal citizenship behavior; AC, affective organizational commitment; JS, job satisfaction.
a An insignificant p-value indicates good fit – that is, failure to reject the null hypothesis that the lower bound of the model's RMSEA confidence internal is below.05.
Analytical method
To test our hypotheses, we employed structural equation modeling (Bollen, Reference Bollen1989) using AMOS 17.0 (SPSS Inc, 2008). In accordance with prior research, we used several goodness-of-fit indices to assess the fit of the hypothesized model (Joreskog & Sorbom, Reference Joreskog and Sorbom1993; Kline, Reference Kline2005; Medsker, Williams, & Holahan, Reference Medsker, Williams and Holahan1994). The indices included the CFI, the TLI, and the RMSEA. Finally, we followed guidance from Kline (Reference Kline2005) in trimming our structural model in order to present the most appropriate final model. Trimming entails removing structural paths from the full model such that the final model (i.e., trimmed model) fits the data just as well as the full model (i.e., no significant χ2 difference) and is the most parsimonious model (Kline, Reference Kline2005; see example, Scott, Colquitt, & Zapata-Phelan, Reference Scott, Colquitt and Zapata-Phelan2007).
Study 1 results
Table 2 presents descriptive statistics, α reliabilities, and intercorrelations for study 1. To test our hypotheses, we employed full structural equation modeling (H1 through H4).
Table 2. Study 1 descriptive statistics, reliabilities, and correlations
N = 152. *Significant at p ⩽ .05. **Significant at p ⩽ .01. Reliability coefficients (α) are on the diagonal.
a Stated in months.
b Racioethnicity was coded as 0 = Caucasian, 1 = non-Caucasian.
c Gender was coded as 0 = female, 1 = male.
d Full-time status was coded as 0 = part-time, 1 = full-time.
For hypotheses 1 through 4, we first assessed the fit of the full model (RI and REQ with paths specified to all four outcomes), which demonstrated good overall fit (χ2 = 265.00 [df = 172]; CFI = .95; TLI = .94; RMSEA = .06; p-value test for close fit [RMSEA < .05] = .13). To ensure that we were using the most appropriate model, we trimmed the model until the data fit the final model as well as the full model (i.e., with no significant χ2 difference) while yielding the most parsimonious model. The trimmed model obtained overall good fit, not significantly different, and most parsimonious (χ2 = 266.17 [df = 175]; CFI = .95; TLI = .94; RMSEA = .06; p-value test for close fit [RMSEA < .05] = .16; Kline, Reference Kline2005). The final model entailed removing paths from REQ to affective commitment, person-focused ICBs, and task-focused ICBs. The outcomes of the trimmed final model, using standardized γs (i.e., standardized regression coefficients), were as follows: RI was positively related to person-focused ICB (H1; γ = .44, p < .01); task-focused ICB (H2; γ = .35, p < .01); and affective organizational commitment (H4; γ = .47, p < .01); but, only marginally related to job satisfaction (H3; γ = .25, p < .10). Therefore, hypotheses 1, 2, and 4 were supported (Figure 1). We should note that REQ was positively related to job satisfaction (γ = .47, p < .01).
Figure 1. Study 1 final model. Note. N = 152. †Significant at p⩽.10. *Significant at p⩽.05. **Significant at p⩽.01. χ2 = 266.17 (df = 175); CFI = .95; TLI = .94; RMSEA = .06; p-value test for close fit (RMSEA < .05) = .16. For ease of presentation, we only show the structural model. Note, per standard practice, RI and REQ were allowed to covary.
In sum, we found that RI was significantly associated with person-focused ICB, task-focused ICB, and affective organizational commitment – after accounting for REQ. However, our sample was mainly part-time employees (76%) in ‘non-professional’ jobs. As such, in study 2, we sought to generalize our findings to full-time employees within ‘professional’ occupations (e.g., engineering, accounting, financial analysis, and project management).
Study 2 method
Sample and procedure
Our sample consisted of 197 employed professionals that were enrolled in a PMBA program at one of two universities in the United States (one in the southeast and the other in the southwest). We followed the same temporal lag as in study 1. In the first survey, we measured RI, REQ, as well as demographic and control variables (see below). In the second survey, we measured the subordinate's person-focused ICB, task-focused ICB, job satisfaction, and affective organizational commitment. As we did in study 1, we followed suggestions provided by Podsakoff et al. (Reference Podsakoff, MacKenzie, Lee and Podsakoff2003) to reduce (albeit not erase) the potential of common method bias.
The surveys were administered via an encrypted and secure web survey. Out of a potential pool of 287 respondents, 243 respondents volunteered (in exchange for extra credit) and responded to the first survey. One month later, 237 of the 243 participants responded to the second survey. Forty respondents were not employed or provided incomplete data and therefore were dropped from the study. Thus, the final sample size was 197. Sixty-four percent were male, 84% were Caucasian, 90% were employed full-time, and mean age was 31.1 years (SD = 6.3). The sample had an average organizational tenure of 43 months (SD = 41.8); an average relationship tenure with their immediate supervisor of 41 months (SD = 48.5); and an average work experience of 91 months (SD = 72.7). Also, 54% reported that they occupied supervisory or management positions. The most dominant industries were manufacturing engineering (21%), professional services (14%), healthcare (12%), and financial services (12%) – with no mean differences across industries.
Attrition analysis
We conducted an attrition analysis (Goodman & Blum, Reference Goodman and Blum1996). Our analysis revealed that only racioethnicity was significantly related to staying or leaving the study. Nevertheless, we were able to conclude that non-random sampling bias was less likely because the mean difference (on racioethnicity) between ‘stayers’ and ‘leavers’ was non-significant.
Measures
We used the same measures (as we did in study 1) for REQ (5-point scale; α = .84), RI (5-point scale; α = .73), person-focused ICB (5-point scale; α = .90), task-focused ICB (5-point scale; α = .90), job satisfaction (7-point scale; α = .94), and affective organizational commitment (7-point scale; α = 89). Again, we implemented similar design features so as to reduce the effects of common method (Podsakoff et al., Reference Podsakoff, MacKenzie, Lee and Podsakoff2003). Anonymity was further enhanced via the encrypted-secure web survey and accessing the participant without the employer having knowledge of the study.
Control variables
We used the same variables as in study 1: age, gender, racioethnicity, work experience, employment status, as well as organizational and relationship tenure. As with study 1, because our parameter estimates were very similar with and without our control variables, we report our findings without control variables.
Discriminant analysis
Again, we tested the discriminant validity of our substantive variables. We compared our hypothesized six-factor model to the five alternative models. The six-factor model demonstrated the best fit and an acceptable overall fit (χ2 = 274.48 [df = 172]; CFI = .96; TLI = .96; RMSEA = .06; p-value test for close fit [RMSEA < .05] = .24; see Table 3). We also found RI and REQ to be factorially distinct (two-factor vs. one-factor models; see Table 3).
Table 3. Study 2 confirmatory factor analysis
N = 197. REQ, relational exchange quality; RI, relational identification; TICB, task-focused interpersonal citizenship behavior; PICB, person-focused interpersonal citizenship behavior; AC, affective organizational commitment; JS, job satisfaction.
a An insignificant p-value indicates good fit – that is, failure to reject the null hypothesis that the lower bound of the model's RMSEA confidence internal is below .05.
Study 2 results
Table 4 presents descriptive statistics, α reliabilities, and intercorrelations for study 2. To test our hypotheses, we employed full structural equation modeling (H1 through H4).
Table 4. Study 2 descriptive statistics, reliabilities, and correlations
N = 197. *Significant at p ⩽ .05. **Significant at p ⩽ .01. Reliability coefficients (α) are on the diagonal.
a Stated in months.
b Racioethnicity was coded as 0 = Caucasian, 1 = non-Caucasian.
c Gender was coded as 0 = female, 1 = male.
d Full-time status was coded as 0 = part-time, 1 = full-time.
For hypotheses 1 through 4, as in study 1, we first assessed the fit of the full model, which demonstrated good overall fit (χ2 = 274.48 [df = 172]; CFI = .96; TLI = .96; RMSEA = .06; p-value test for close fit [RMSEA < .05] = .24). Following Kline (Reference Kline2005), we again trimmed the model until the data fit the final model as well as the full model (i.e., with no significant χ2 difference) while yielding the most parsimonious model (χ2 = 279.43 [df = 174]; CFI = .96; TLI = .95; RMSEA = .06; p-value test for close fit [RMSEA < .05] = .14). The final model entailed removing paths from relational exchange to both person-focused ICBs and task-focused ICBs. The outcomes for the trimmed final model, using standardized γs, were as follows: RI explained significant incremental variance (after accounting for REQ) in person-focused ICB (H1; γ = .46, p < .01), task-focused ICB (H2; γ = .38, p < .01), job satisfaction (H3; γ = .26, p < .05), and affective organizational commitment (H4; γ = .25, p < .05). Thus, we found support for hypotheses 1 through 4 (see Figure 2). We should note that REQ was positively related to job satisfaction (γ = .34, p < .01) and affective commitment (γ = .26, p < .01).
Figure 2. Study 2 final model. Note. N = 197. *Significant at p ⩽ .05. **Significant at p ⩽ .01. χ2 = 279.43 (df = 174); CFI = .96; TLI = .95; RMSEA = .06; p-value test for close fit (RMSEA < .05) = .14. γs reflect the standardized regression coefficients. For ease of presentation, we only show the structural model. Note, per standard practice, RI and REQ were allowed to covary.
Discussion
Organizationally-bound role-relationships are indeed more complex than functional, reciprocity-based transactions. Although a deep and wide current of knowledge demonstrates that exchange makes an indelible mark on relational processes and organizational attitudes; the current research expands the purview of role-relationships to include not only exchange-based processes but also identity-based processes. Indeed, our studies underscore the notion that organizations or applied settings may be more tribal than we otherwise might think (e.g., Ashforth, Harrison, & Corley, Reference Ashforth, Harrison and Corley2008). We now elaborate on the theoretical implications of our findings as well as discuss managerial implications, study limitations, and future research directions.
Theoretical contributions and future research
The findings from our studies serve to extend prior research on both citizenship behaviors and organizational attitudes as well as to integrate identification and exchange within work role-relationships. We previously argued that research on role-relationships has been dominated by social exchange theory (i.e., SET). However, relational processes are not the only nomological networks that have experienced a somewhat myopic focus on exchange. Research exploring citizenship and organizational attitudes has also been dominated by explanations emanating from SET (with notable exceptions from research on affect; see Brief, Reference Brief1998). While certainly valid (see Cropanzano et al., Reference Cropanzano, Anthony, Daniels and Hall2017), social exchange is not the only mechanism by which these important outcomes may be explained (Dutton & Ragins, Reference Dutton and Ragins2007). In addition, these outcomes have traditionally been explained via either collective (i.e., organization, occupation, work group) or individual differences – with much less focus on relational differences. That said, when scholars have explored relational differences as predictors to these outcomes, they have again relied on SET. In addition, research on role-relationships at work can benefit from examining multiple types of relational differences (e.g., exchange, identification). Our studies show that RI provides a compelling and complementary lens by which scholars may explore role-relationships at work. Indeed, we are not able to ignore our findings in which RI was the sole predictor for person-focused ICB and task-focused ICB (in both studies) as well as for affective commitment (in study 1).
Our findings suggest that an important future direction for research is to work at integrating RI and REQ, thus further specifying a more holistic theory for relationships at work (cf., Ferris et al., Reference Ferris, Liden, Munyon, Summers, Basik and Buckley2009). How do we (in a more precise way) integrate both identification and exchange within research on role-relationships? Based upon our findings, we foresee two potential options: (1) an ‘independent effects’ approach; (2) a ‘met expectations/fit’ approach. We then speculate concerning an additional and third option: (3) a ‘substitution’ approach.
In these studies, we examined the independent effects approach by examining the direct relations between REQ and RI and outcomes. Our findings provide evidence that identification and exchange covary somewhat discriminantly (e.g., RI was the sole predictor for person-focused ICB, task-focused ICB, [in study 1 and 2], and affective commitment [in study 1]). These results raise interesting questions about the role of RI and REQ in relationships. Drawing from research that utilized more expansive measures of exchange, we assumed that REQ would have consistent positive relationships with both types of ICBs. However, using a more precise measure of REQ and analyzing REQ and RI simultaneously, our studies did not support these relationships. Perhaps because one's individual interests underlie the reciprocity in exchange, REQ may not induce subordinates to provide help beyond what is expected in the role-relationship. In contrast, when the role-relationship with the supervisor is perceived as part of one's self-definition, going beyond expectations to help the partner is perceived as benefiting the self (Brewer & Gardner, Reference Brewer and Gardner1996). Also surprising is that RI consistently related to affective commitment, while REQ only predicted affective commitment in study 2. We speculate that this difference may be due to the specific motivational processes underlying the predictors (RI and REQ) and the outcomes (ICBs, satisfaction, commitment). It may be that these relational components (RI and REQ) differentially related to outcomes based upon whether the outcomes are more closely rooted in exchange-based and short-term schemas or identity-based and longer-term schemas (Kivetz & Tyler, Reference Kivetz and Tyler2007; Settoon & Mossholder, Reference Settoon and Mossholder2002; Sluss & Ashforth, Reference Sluss and Ashforth2007). We also suggest that future research may make more precise predictions for these interrelationships and thus carry out more precise analyses – such as a relative weights analysis – to see, not only if RI explains additional variance, but to see how much variance can be attributed to each relational component. That said, our direct findings for study 2 do not bear this out consistently, which we explore in our next integrative approach.
Identification and exchange may become more predictive of outcomes depending upon the met expectations of and/or the ‘fit’ with the nested relationship. This may explain the difference in findings regarding affective commitment across non-professional (study 1) and professional (study 2) samples. The resources expected by subordinates and provided by supervisors in non-professional contexts are primarily tangible and directly related to task execution. In professional contexts, supervisors provide more intangible resources (i.e., trust, knowledge, developmental guidance), thus engaging subordinates' long-term schemas and their reciprocation through affective commitment to the organization. Indeed, a recent meta-analysis by Kossek, Pichler, Bodner, and Hammer (Reference Kossek, Pichler, Bodner and Hammer2011) found supervisor support in both work- and non-work-related contexts to correlate with similar positive outcomes.
We argue that RI and REQ may similarly depend on individual differences. For example, an individual that has a high relational self-construal (i.e., the general tendency to think of oneself in relational terms; Cross, Bacon, & Morris, Reference Cross, Bacon and Morris2000) may focus more on RI than REQ as a basis for positive organizational attitudes. For example, Clark and Mills (Reference Clark and Mills1979) found that those with a communal perspective did not respond well to relational others that adhered to strict reciprocity norms, and, in fact, tend to perceive strict reciprocity as a breach in behavioral expectations. In short, the form and function of the role-relationship should fit the relational expectations/needs. These relational expectations may come from the job context (as demonstrated in non-professional vs. professional jobs), individual differences (as speculated above) or possibly the organizational context (e.g. relational norms embedded in the organizational culture).
Third and most speculative, while we may observe independent effects for certain outcomes, it may be that these processes also substitute for each other. Although not tested, we speculate that RI and REQ may have substituting effects – especially for outcomes where both have somewhat equal influence. For example, it may be that if a subordinate experiences higher RI then the level of REQ (whether high or low) may not particularly matter or vice-versa for outcomes. Indeed, RI and REQ access different motivational processes (e.g., self-enhancement/self-consistency and reciprocity, respectively). Outcomes that are equally related to RI and REQ may just require one or the other to be satisfied. Future research should explore more deeply the possibility of substitution.
In addition to the integration possibilities revealed by these two studies, RI and exchange may also be integrated into that one may mediate the effect of the other. In our studies, we assumed that RI and REQ covary but are not causally related. Scholars have explored how both REQ (Bauer & Green, Reference Bauer and Green1996) and RI develop (Sluss & Ashforth, Reference Sluss and Ashforth2008). One intuitive approach is to conjecture that exchange forms first and, thus, REQ influences RI. That said ones' relational needs may determine which relational component (RI or REQ) take primacy in the nascent role-relationship (cf. Rusbult & van Lange, Reference Rusbult and van Lange2003). While the mediatory mechanism approach holds promise, RI and relational exchange may, in the end, be quite a reciprocal process – depending upon when and where one is in the relationship's formation and trajectory. In addition to individual differences, organizational culture or industry norms may also play a significant role in the mediatory order of RI and relational exchange. For example, certain organizations and/or industries might be more relational than others which promulgate relationships that could transcend (see Sluss & Ashforth, Reference Sluss and Ashforth2007) organizational membership. Future research should investigate how individual differences, contextual differences, or an interaction between the two variable sets determine the mediatory process within RI and REQ.
Practical implications
Our findings also provide insight for practicing managers. Not only do leaders need to focus on providing relevant and appropriate resources – as exchange theory would suggest – but also provide a context for personalized connection and identity negotiation. The manager plays an integral role in helping the subordinate define who they are in the organization and what that means for their attitude toward work. Our findings suggest that leaders should be aware of the differential needs of their subordinates. We (tentatively) recommend that leadership development include focusing not only task-related needs but psycho-social needs as well. A well-developed mentoring (formal vs. informal) program may be needed. In addition, managers need to go beyond just implementing degrees of task-focused and/or relationship-focused behaviors to forging relationships upon exchange-based and identity-based needs (Hill, Reference Hill2007).
Indeed, our findings demonstrate the increasing importance of selecting and training leaders that are able to effectively provide for both exchange and identification – thus, maintaining a level of flexibility. Our findings highlight the importance of self-monitoring as an important leadership capability. We suggest that self-monitoring will enable managers to be in tune with how their task and relationship-focused behaviors are influencing both exchange and identification processes within their relationships (cf. Zaccaro, Foti, & Kenny, Reference Zaccaro, Foti and Kenny1991).
Limitations and future research
Our two studies pose some limitations and, also as a result, directions for future research. First, all data were collected from one source. That said we temporally separated, via a one-month lag, the predictor and criterion variables. As such, we attempted to reduce the potential for consistency motifs, implicit theories, and illusionary correlations. Note that we, in our theory versus empirics, argue that RI and REQ influence the outcomes. As such, our temporal separation focuses more on reducing consistency motifs and the like rather than establishing the structural association – this would be done, for example, in a newcomer sample in which measures for baseline outcome variables could be controlled. We also promoted anonymity in that (1) the surveys were not presented via their employing organizations but via their educational institution and (2) the surveys were handled in a way that protected personal information. We also found significant evidence of discriminant validity while comparing multiple models. Many of the variables examined within this study encompass intrapersonal perceptions and attitudes that are quite inaccessible to others (cf. Podsakoff et al., Reference Podsakoff, MacKenzie, Lee and Podsakoff2003). In sum, while not eradicating concern for common method bias, our design features and analyses help in reducing the concern.
Second, we used self-report measures for ICBs (i.e., subordinates reported their own ICBs toward their supervisors). Recent research has pointed out that while there are indeed potential problems associated with individual bias in self-reports of citizenship behavior, supervisor reports of citizenship behavior are not completely insusceptible to bias either (Klotz, He, Yam, Bolino, Wei, & Houston, Reference Klotz, He, Yam, Bolino, Wei and Houston2018; O'Brien & Allen, Reference O'Brien and Allen2008; Vandenberg, Lance, & Taylor, Reference Vandenberg, Lance, Taylor and Turnipseed2005). O'Brien and Allen (Reference O'Brien and Allen2008) explain that problems with supervisor reports of citizenship behaviors may result from actor-observer bias or halo effect bias (see also Klotz et al., Reference Klotz, He, Yam, Bolino, Wei and Houston2018; Vandenberg, Lance, & Taylor, Reference Vandenberg, Lance, Taylor and Turnipseed2005). Nonetheless, future research would do well to examine these relationships by using a multi-source rating system rather than merely using ratings from solely the employee or the supervisor (Allen, Barnard, Rush, & Russell, Reference Allen, Barnard, Rush and Russell2000). Relatedly concerning measurement, we did not use the full Bernerth et al.'s (Reference Bernerth, Armenakis, Feild, Giles and Walker2007) eight-item scale. Our intent was to base our study on the most widely used measure of relational exchange (LMX-7). However, we do support using the full Bernerth et al.'s (Reference Bernerth, Armenakis, Feild, Giles and Walker2007) scale in future research.
Third, while focusing on the supervisory relationship, we ignored other potential salient role-relationships (e.g., coworker–coworker, consultant–client). We recommend that future research include exploring the impact of coworker relationships – especially within technical and non-managerial professionals. Expanding research to other relationships will also allow scholars to explore additional moderating conditions such as one's work orientation (Wrzesniewski, McCauley, Rozin, & Schwartz, Reference Wrzesniewski, McCauley, Rozin and Schwartz1997) or relational self-construal (Cross, Bacon, & Morris, Reference Cross, Bacon and Morris2000). In related fashion, we only focused on affective organizational commitment while leaving out two other forms of organizational commitment – normative and continuance. We suggest that future research should explore how RI and REQ integrate when predicting normative and continuance organizational commitment (Meyer & Allen, Reference Meyer and Allen1997).
Fourth, it may be that exchange is an antecedent of RI (or vice-versa). High-quality exchange may encourage RI yet RI may also provide a positive backdrop reducing transaction costs and, by default, creating more benefits (cf. Williamson, Reference Williamson1985). We recommend research examine the cross-lagged relationships between RI and REQ. Scholars should also take care to explore both nascent relationships and established relationships in transition. In sum, our studies show that it is important for theory and practice to better delineate the intrapersonal, interpersonal, and contextual antecedents for both REQ (Bauer & Green, Reference Bauer and Green1996; Gerstner & Day, Reference Gerstner and Day1997) and RI (Sluss & Ashforth, Reference Sluss and Ashforth2007, Reference Sluss and Ashforth2008). Similarly, we only implicitly hypothesized that RI influences outcomes via identity-based needs being fulfilled such as the need for belonging. However, we did not explicitly operationalize and test these mediatory associations. Likewise, REQ also has differing motivational processes such as reciprocity and exchange. We suggest that future research explore these mediatory processes between RI and its outcomes as well as REQ and its outcomes. Indeed, we may see (with regard to RI) more complex interrelationships between certain identity-based needs and outcomes. For example, the need for belonging may be particularly powerful in mediating the association between RI and a sense of collective membership whereas a need for self-expansion may be more integral for mediating the association between RI and helping the relational other.
Conclusion
Organizationally-bound role-relationships are indeed complex. Our studies show that role-relationships ‘do not live by’ either relational exchange or RI alone. Our studies provide a more expansive view of how individuals experience relational dynamics within their own particular organizational landscape as well as a view of how scholars may better understand these important underlying relational processes.
Acknowledgment
We acknowledge helpful comments from Jeremy Dawson, Jacqueline Coyle-Shapiro, Lynn Shore, Rolf van Dick, and Daan van Knippenberg on previous versions of this paper.
