Abstract
This study examines the simultaneous effect of leader-member exchange (LMX) and coworker exchange (CWX) on employees’ helping behavior. It proposes that the congruence and incongruence of LMX and CWX have both a linear and curvilinear relationship with helping behavior. Hypotheses were tested on a final sample of 881 participants at hospitals in Taiwan. After controlling for relevant variables, results of a polynomial regression and response surface methodology indicate a positive linear relationship between the simultaneous effect of the congruence of LMX and CWX and employees’ helping behavior, but do not confirm a curvilinear relationship, whereas the simultaneous effect of the incongruence of LMX and CWX has a curvilinear relationship with employees’ helping behavior but not a linear relationship. The theoretical and practical implications, strengths and limitations, and directions for future research are discussed.
Keywords
Introduction
Organizations need employees to fulfill their formal role requests, as described in their employment contract, and to invest extra efforts in helping their coworkers achieve their goals and survive in the fast-changing business environment. Many studies have indicated that a high exchange relationship between leaders and members (leader-member exchange; LMX; Dansereau et al., 1975; Graen & Uhl-Bien, 1995) positively influences employees’ helping behavior (Dulebohn et al., 2012; Gerstner & Day, 1997; Ilies et al., 2007; Martin et al., 2016; Young et al., 2021). Research has also shown that when members develop high exchange relationships with each other (coworker exchange; CWX; Sherony & Green, 2002), their helping behavior increases (Baker & Omilion-Hodges, 2013; Tang et al., 2021).
Based on the relational perspective, significant others might influence employees’ behaviors (S. Chen et al., 2006) such as helping behavior. Furthermore, Katz and Kahn (1978) pointed out that in the workplace, leaders and other members also affect employees’ behaviors, as employees work with both leaders and other members (Cole et al., 2002; Graen, 1976; Lai et al., 2019). While studies have provided abundant findings regarding the relationship between high exchange relationships and employees’ behaviors, they often only consider the effect of one type of interpersonal relationships (e.g., either LMX or CWX) on employees’ behaviors (e.g., Eby & Allen, 2012). To date, few studies have examined LMX and CWX and the effects of both simultaneously (e.g., Lai et al., 2019). Further, although some researchers have conducted moderation analyses to understand the effect of these two exchange relationships (e.g., LMX and team-member exchange [TMX]; L. C. Wang & Hollenbeck, 2019) on related outcomes, but these only provide part of the fuller picture. An incomplete focus on these workplace interpersonal relationships might inhibit our understanding of their effect on employees’ behavior. Research should consider both LMX and CWX simultaneously (e.g., Lai et al., 2019) and examine their influence on employees’ behaviors through a more comprehensive method (e.g., polynomial regression; Edwards, 1994; Edwards & Parry, 1993).
In addition to LMX and CWX, Seers (1989) highlights two other types of interpersonal relationships in a group, namely lackeys and rebels. Lackeys have a high exchange relationship with a leader, but not with coworkers. In contrast, rebels are employees who develop high exchange relationships with coworkers but not with their leaders. Research has revealed that employees do not necessarily develop high exchange relationships with both their leaders and coworkers (Cole et al., 2002; Cropanzano & Mitchell, 2005), but LMX and TMX have a compensatory effect (Lai et al., 2019). As such, this study proposes that to increase employees’ helping behavior, LMX and CWX might compensate each other.
Furthermore, although researchers have pointed out that a curvilinear relationship might exist between the variables of interest (e.g., Farh et al., 2010; Lam et al., 2014; Le et al., 2011), previous studies often classified employees’ exchange relationships as high (positive) or low (negative; Eby & Allen, 2012) and examined their linear relationship with relevant outcomes (e.g., Bala et al., 2021; Bauer et al., 2006; Erdogan & Enders, 2007; Golden & Veiga, 2008; Harris et al., 2009; Hui et al., 1999; Kacmar et al., 2003; Lai et al., 2021; Liu et al., 2021; Ng et al., 2014; Schriesheim et al., 2000; Settoon et al., 1996). These studies ignore the possible curvilinear relationship (e.g., a moderate level of LMX and CWX) between LMX, CWX, and relevant variables. Adopting a polynomial regression (Edwards, 1994; Edwards & Parry, 1993), for instance, Lai and colleagues found that employees with a moderate level of LMX and TMX have the lowest turnover intention (2019), indicating that investigating a curvilinear relationship might help researchers comprehensively understand interpersonal relationships in the workplace. Therefore, this study aims to use a polynomial regression to examine the relationship between moderate levels of exchange relationships and behavioral outcomes. Specifically, this study proposes that the simultaneous effect of the (in-)congruence of LMX and CWX has a curvilinear relationship with employees’ helping behavior.
This study advances the theories of workplace interpersonal relationships and helping behavior in several ways. First, unlike previous research, which often considered only one side of interpersonal relationships in the workplace, this study responds to the call of prior studies (G. Chen & Sharma, 2012; Lai et al., 2019; Seers, 1989) to consider LMX and CWX simultaneously and examine their simultaneous effect on employees’ helping behavior. Second, in contrast with previous research, which often used moderation models to examine LMX and CWX, this study conducts a polynomial regression to analyze the simultaneous effect of LMX and CWX. The advantage of this method is that it enables analyzing both linear and curvilinear relationships among variables. Moreover, combined with the response surface methodology (Box & Draper, 1987; Myers, 1971), the polynomial regression model can produce a three-dimensional rather than two-dimensional surface (e.g., moderation model), the results of which are easier to interpret (Edwards, 2002). Third, based on the results of Lai et al.’s (2019) study, it is important to consider moderate levels of exchange relationships and their simultaneous effect on employees’ behavior. Thus, in addition to focusing on the congruence and incongruence of LMX and CWX, this study also examines how moderate levels of exchange relationships influence the employee’s helping behavior.
The subsequent sections of this paper elucidate the theory and hypotheses of the study, followed by a discussion of the study methods. Then, the results are provided and discussed. Finally, the practical implications of the study and directions for future research are highlighted.
Theory and Hypotheses
Linear Relationship Between LMX, CWX, and Helping Behavior
LMX and CWX
According to Graen and Scandura (1987), high LMX employees find it easier to access more information and valued resources and will be afforded greater latitude than low LMX employees. In addition, high LMX employees possess higher capabilities than low LMX employees; thus, they might be classified as a core cadre (Dansereau et al., 1975). When there is a high level of LMX, it can foster feelings of trust, respect, and mutual obligation between the leader and employees (Graen & Uhl-Bien, 1995). Although high LMX employees may receive various valued resources from a leader, the leader might also require the employees to invest extra efforts (e.g., help the other employees; Ilies et al., 2007) that are not described in their contracts as reciprocation (Lai et al., 2019). In contrast, employees who possess low LMX are classified as hired hands (Dansereau et al., 1975) and are usually only required to fulfill their contractual job responsibilities. Sherony and Green (2002) contend that CWX shares many similar characteristics with LMX, such that high CWX employees may receive more assistance and help from their coworkers than low CWX employees. To maintain high CWX, employees also need to offer their assistance to other coworkers when asked (Lai et al., 2019).
Seers (1989) proposed that when employees navigate the role expectations of their leader and coworkers, four types of exchange relationships may emerge: Employees might develop low LMX and low CWX (LL), high LMX and high CWX (HH), high LMX and low CWX (HL), or low LMX and high CWX (LH) with their leader and coworkers. Employees might develop different exchange relationships with their leader and coworkers (Cropanzano & Mitchell, 2005). For instance, Cole et al. (2002) pointed out that one employee might have a positive (i.e., high) and negative (i.e., low) exchange relationship simultaneously when an organization’s or team’s goals are misaligned. Hence, in the workplace, an employee may develop HL or LH exchange relationships instead of developing congruent HH or LL relationships.
Effect of Congruence of LMX and CWX on Helping Behavior
Van Dyne and LePine (1998) defined helping behavior as “promotive behavior that emphasizes small acts of consideration. Helping is cooperative behavior that is noncontroversial. It is directly and obviously affiliative; it builds and preserves relationship; and it emphasizes interpersonal harmony” (p. 109). In addition, unlike job performance, helping behavior is a discretionary behavior not specified in role prescriptions. As these extra efforts are not required, whether employees perform them or not, no rewards or punishment follow (Van Dyne & LePine, 1998). However, helping behavior could be considered a means of reciprocity (Blau, 1964; Gouldner, 1960), which could be used to develop and maintain good interpersonal relationships with the others such as leaders and coworkers (e.g., Baker & Omilion-Hodges, 2013; Dulebohn et al., 2012; Gerstner & Day, 1997; Ilies et al., 2007; Martin et al., 2016; Tang et al., 2021; Young et al., 2021). As such, HH employees might perform more helping behavior to maintain their high exchange relationships with both their leaders and coworkers. Conversely, LL employees are less likely to perform helping behavior because they do not consider themselves obligated to do so and do not believe they have enough resources or capability to do so. Thus, we propose that:
Hypothesis 1: There is a positive relationship between the congruence of LMX and CWX and helping behavior, such that employees with high LMX and high CWX would demonstrate higher levels of helping behavior, whereas employees with low LMX and low CWX exhibit lower levels of helping behavior.
Effect of Incongruence of LMX and CWX on Helping Behavior
As previously mentioned, Seers (1989) noted that in addition to HH and LL employees, two other types of exchange relationships exist in the workplace, namely lackeys (i.e., HL employees) and rebels (LH employees). HL employees develop a high exchange relationship with their leader but fail to build such exchange relationships with coworkers, so they primarily receive resources from their leader. Thus, unless requested by their leader, HL employees may not feel motivated to invest extra efforts in helping their coworkers, especially since helping behavior is discretionary rather than prescribed. HL employees prioritize maintaining a high exchange relationship with their leader rather than their coworkers, so they would likely only exhibit helping behavior as a means to maintain their core cadre position in their leader’s mind.
In contrast, LH employees develop high exchange relationships with their coworkers but fail to do so with their leader. Based on the characteristics of CWX, offering assistance and helping when needed is a crucial and important prerequisite to developing high exchange relationships among employees. Thus, LH employees would likely engage in helping behavior even in the absence of a formal request to do so. In other words, they are willing and ready to help their coworkers. Thus, LH employees would likely exhibit more helping behavior than HL employees, who assist others primarily for the sake of impressing their leader. Further, based on social comparisons theory, when employees discover their exchange relationship with the leader is inferior to that of their coworkers, they tend to develop a high exchange relationship with each other by exhibiting increased affiliative behaviors (Tang et al., 2021). Thus, LH employees might exhibit more helping behavior than HL employees due to their comparatively weaker relationship with the leader. We therefore propose that:
Hypothesis 2: There is a negative relationship between the incongruence of LMX and CWX and helping behavior, such that employees with low LMX and high CWX would demonstrate higher levels of helping behavior, whereas employees with high LMX and low CWX exhibit lower levels of helping behavior.
Curvilinear Relationship Between LMX, CWX, and Helping Behavior
Effect of Congruence of LMX and CWX on Helping Behavior
Furthermore, although previous research found that both high LMX and high CWX are associated with helping behavior (Baker & Omilion-Hodges, 2013; Dulebohn et al., 2012; Gerstner & Day, 1997; Ilies et al., 2007; Martin et al., 2016; Tang et al., 2021; Young et al., 2021), since researchers may have only considered one side of the exchange relationship, this may only be partially true. Further, many studies, such as the high-low categorization used by Seers (1989), overlook moderate exchange relationships between leaders and coworkers. This study fills the research gap and aims to reveal the full picture of the effect different levels of exchange relationships have on helping behavior.
To develop a high level of exchange relationship with their leader and coworkers (i.e., HH employees), employees need to understand and accomplish each assignment to meet their leader’s expectations, while also satisfying their coworkers’ expectations and navigating harmonious interactions with them; otherwise, they are LL employees who have a low level of exchange relationship with both their leader and coworkers. While HH and LL employees may represent the extreme ends of the helping behavior spectrum, there are employees who fall somewhere in between, where only a moderate level of exchange relationship with the leader and coworkers is developed (Lai et al., 2019). Compared to HH employees, these employees might receive less resources or help from their leader or coworkers. However, compared to LL employees, they receive more resources. Because the resources are neither abundant nor scarce, these employees may face challenges in determining the appropriate timing to offer help to others and in deciding how to distribute their extra resources to accomplish their tasks. Once employees acquire the knowledge of interacting appropriately with their leader and coworkers, as well as distributing their resources efficiently, they can engage in helping behaviors more effortlessly.
As such, the relationship between the simultaneous effect of LMX and TMX and employees’ behavior may not be linear. While direct evidence is lacking, previous studies have demonstrated a curvilinear relationship between the simultaneous effect of LMX and TMX on employees’ turnover intention (Lai et al., 2019). Specifically, employees who possess a moderate level of LMX and TMX have the lowest turnover intention. Thus, we propose that a U-shaped relationship may exist between the congruence of LMX and CWX and helping behavior. In other words, an inflection point may exist within this relationship. Therefore, we propose that:
Hypothesis 3: The relationship between the congruence of LMX and CWX and helping behavior is U-shaped, such that employees with moderate levels of both LMX and CWX would exhibit lower levels of helping behavior compared to employees with high LMX and high CWX and employees with low LMX and low CWX.
Effect of Incongruence of LMX and CWX on Helping Behavior
Employees with high capability may develop high exchange relationships with both their leaders and coworkers. Other employees may choose to develop their relationships with either their leader or coworkers (i.e., HL or LH). Still some other employees whose capabilities are better than those of LL employees but inferior to those of HH employees may only develop moderate exchange relationships with both their leader and coworkers. Compared to HL employees, these employees would only receive limited resources and support from their leader. Thus, they might exhibit minimum reciprocal helping behavior and try to reserve resources to accomplish their tasks. Furthermore, compared with LH employees, moderate level employees may not feel obligated to help because they do not have a strong bond with their coworkers. Taken together, employees who develop moderate exchange relationships with their leader and coworkers may exhibit less helping behavior than both HL or LH employees, suggesting a U-shaped relationship between the simultaneous effect of LMX and CWX on helping behavior. Thus, we propose that:
Hypothesis 4: The relationship between the incongruence of LMX and CWX and helping behavior is U-shaped, such that employees with moderate levels of both LMX and CWX would exhibit lower levels of helping behavior compared to employees with low LMX and high CWX and employees with high LMX and low CWX.
Methods
Sample and Procedure
Data were collected from work groups in six hospitals in Taiwan. To deal with common method variance (CMV; Podsakoff et al., 2012), this study adopted a time-lagged research design, collecting data from two time points that are 6 months apart. The group leader (i.e., the ward’s head nurse) was contacted to help with data collection. After explaining the goal of this study, in the first wave, participants (i.e., ward nurses) completed the questionnaires, responding to questions regarding their demographic information (age, gender, education level, and tenure), LMX, CWX, and helping behavior. In the second wave, information regarding helping behavior was collected from participants. All questionnaires were returned in a sealed envelope in both waves.
The final sample size was 881 participants. The average age of participants is 34.32 years, nearly all are female (96.6%), and most have a senior high vocational school diploma or higher (98.2%). The average tenure in the healthcare profession was 11.60 years (SD = 8.95), with an average of 9.02 years as a hospital staff member in the organization (SD = 8.15), and an average of 7.11 years working in the ward (SD = 6.60).
Measures
In the study, all measures were rated on a 5-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree). In addition, all measures used in this study were translated and back-translated from English into Chinese (Brislin, 1980).
Leader-Member exchange
Participants completed a 7-item measure adapted from Scandura and Graen (1984) to evaluate their exchange relationship with their leader. An example item is: “My head nurse understands my problems and needs well” (α = .91).
Coworker exchange
Participants completed a 7-item measure adjusted from Scandura and Graen’s (1984) 7-item leader-member exchange measure (1984) to evaluate their exchange relationships with their coworkers. An example item is: “I usually know how satisfied my ward coworker(s) is (are) with what I do.” (α = .89)
Helping behavior
Participants completed a 4-item measure adopted from Van Dyne and LePine (1998) to evaluate their helping behavior. An example item is: “Volunteers to do things for this ward.” (α = .86 in Time 1 and .91 in Time 2).
Control Variables
In order to partial out the potential effect of pervious helping behavior (Time 1) on later helping behavior (Time 2), we followed the procedures of previous studies (Glomb et al., 2011; Tepper et al., 2004) by collecting data on employees’ helping behavior in the first wave and treating it as a control variable (α = .86 in Time 1). In addition, perceived organizational support (POS; Eisenberger et al., 1986), which describes the exchange relationship between employees and the organization, is another type of exchange relationships in the workplace (Cole et al., 2002) and might influence employees’ helping behavior (Choi, 2009). Thus, we also treated it as a control variable to partial out its effect on employees’ helping behavior. We adapted Eisenberger et al.’s (1990) 6-item measure to evaluate POS and collected related data in the first-wave. An example item is: “My hospital strongly considers our goals and values” (α = .94).
Analysis
Most studies use the difference scores between two measures to operationalize the congruence (e.g., fit; Shin et al., 2017) or incongruence (relative LMX; Henderson et al., 2008) between the two constructs. However, previous methods used to calculate difference scores, including the algebraic, absolute, or squared difference between two measures, have limitations such as confounding effects, reduced reliability, and ambiguous interpretation (Edwards, 2002). On the other hand, a polynomial regression is more advantageous. First, it retains the independent effect of each construct and their joint effect on the outcome variable (Edwards, 2002; Humberg et al., 2019; Schönbrodt, 2016a, 2016b). Second, polynomial regression can test both the linear and curvilinear relationships between two constructs and the outcome variable (Edwards, 2002). In addition, when polynomial regression is combined with response surface methodology, a three-dimensional surface is produced (Box & Draper, 1987; Myers, 1971), helping researchers interpret the results (Edwards, 2002). Thus, this study follows prior research (e.g., Lai et al., 2019) and conducts polynomial listwise regressions combined with response surface methodology as an alternative technique to test our hypotheses (Edwards, 1994, 2002; Edwards & Parry, 1993). To address issues pertaining to multicollinearity, LMX and CWX were scale-centered before testing the hypotheses to facilitate the interpretation of results (Edwards & Parry, 1993). We also followed the methods in prior research (Humberg et al., 2019; Schönbrodt, 2016a, 2016b) to test the curvature of the effect of the congruence and incongruence of LMX and CWX on helping behavior, and to plot the figure.
Results
Table 1 shows the means, standard deviations, and correlations between the variables. The alpha coefficients of the variables are presented on the diagonal.
Descriptive Statistics and Intercorrelations Matrix of the Study Variables.
Note. Cronbach’s alphas appear across the diagonal in parentheses. n = 881; Tenure 1 = professional tenure; Tenure 2 = hospital tenure; Tenure 3 = ward tenure; LMX = leader-member exchange; CWX = coworker exchange; POS = perceive organizational support.
p < .05. **p < .01. *** p < .001.
Confirmatory Factor Analysis
To ensure the discriminant validity of measures, a confirmatory factor analysis (CFA) of four models was conducted before testing the hypotheses (see Table 2). Because Time 1 helping behavior had relative high correlations with Time 2 helping behavior (r = .43; p < .001), it was included in the CFA. Similarly, because POS also had relative high correlations with LMX and CWX (for LMX, r = .43; p < .001; for CWX, r = .35; p < .001), it was also included in the CFA. The null Model is a one-factor model where all items are loaded onto one variable (χ2 = 11,208.35; df = 350; RMSEA = 0.19; SRMR = 0.15; NNFI = 0.37; CFI = 0.42). Alternative Model 1 is a two-factor model (χ2 = 9,075.15; df = 349; RMSEA = 0.17; SRMR = 0.13; NNFI = 0.49; CFI = 0.53) in which Time 1 helping behavior, LMX, CWX, and POS were combined into one factor because these data were collected at the same time. Alternative Model 2 is a three-factor model (χ2 = 7,808.11; df = 347; RMSEA = 0.16; SRMR = 0.12; NNFI = 0.56; CFI = 0.60) in which the three types of workplace relationships (LMX, CWX, and POS) were combined into one factor. The baseline Model 1 is a 5-factor model (χ2 = 2,056.75; df = 340; RMSEA = 0.08; SRMR = 0.05; NNFI = 0.90; CFI = 0.91).
Confirmatory Factor Analysis.
In the two factors model, leader-member exchange, coworker exchange, perceived organizational support, and Time 1 helping behavior were combined into one factor.
In the three factor model, leader-member exchange, coworker exchange, and perceived organizational support were combined into one factor.
p < .001.
The fit indices of the null model were compared with those of alternative Model 1, the fit indices of alternative Model 1 were compared with those of alternative Model 2, and the fit indices of alternative Model 2 were compared with the baseline Model. The Chi-square difference test and fit indices showed that alternative Model 1 is better than the null Model (Δχ2 = 2,133.20; df = 1; p < .001), alternative Model 2 is better than alternative Model 1 (Δχ2 = 1,267.04; df = 2; p < .001), and the baseline Model is better than alternative Model 2 (Δχ2 = 5,751.36; df = 7; p < .001).
Further, the fit indices in the five-factor baseline Model are all acceptable (Hooper et al., 2008; H. Wang et al., 2005). However, some of our items may have been influenced by the wording, leading to potential measurement issues (Podsakoff et al., 2012; Wu, 2008). To address this, we conducted a baseline Model 2 analysis, correlating these items based on the modification indices from baseline Model 1 (see Table 2). Results showed that the fit indices of baseline Model 2 were improved (χ2 = 1,339.81; df = 335; RMSEA = 0.06; SRMR = 0.05; NNFI = 0.94; CFI = 0.95), and the Chi-square difference test also showed that baseline Model 2 is better than baseline Model 1 (Δχ2 = 7,16.94; df = 5; p < .001). Overall, these results support the discriminant validity of the measures.
Hypotheses Testing
To test our hypotheses, we used the R package (Schönbrodt, 2016a, 2016b) and Humberg et al. (2019) checklist. We expected a convex surface of the theoretical model; thus; the line of congruence (x = y) runs parallel to the second principal axis, and the line of incongruence (x = −y) runs parallel to the first principal axis (Edwards, 1993). The intercept (p20) and slope (p21) of the second principal axis, and the intercept (p10) and slope (p11) of the first principal axes were tested. If results of the intercept and slope for both axes are non-significant, it suggests the presence of the congruence and incongruence effects (Humberg et al., 2019). Results showed that the intercept (p20 = 0.06, p = .81; p10 = 8.76, p = .54) is non-significant but the slope is significant (p21 = 0.79, p < .01; p11 = −1.27, p < .01), indicating that the intercept of the second and first principal axis does not differ from the line of congruence, but the slope of both axes does. Despite the contradiction of our results and some of the hypotheses (i.e., Hypothesis 1 and Hypothesis 2) with the checklist of Humberg et al. (2019), we proceeded with hypothesis testing based on our own arguments and prior research (e.g., Lai et al., 2019). We believe that the contrasting findings hold valuable implications, which we will discuss in the subsequent section on theoretical implications.
To test Hypothesis 1, that the congruence of LMX and CWX has a positive relationship with helping behavior, we ran hierarchical regression analyses (see Table 3). Control variables were entered in Model 1, and LMX and CWX were added in Model 2. In Model 3, three interaction terms were entered (LMX2, CWX2, and LMX × CWX). The results showed that the quadratic terms of both LMX (β = .06, p = .19) and CWX (β = .09, p = .92) were non-significant, but the interaction term (i.e., LMX × CWX) was significant (β = −.17, p < .01). We also tested the linear effect of LMX and CWX on helping behavior. Results showed that the slope on the congruence of LMX and CWX (i.e., LMX = CWX) on helping behavior was significant (a1 = 0.16, p < .001; see Table 4). We plotted the simultaneous effect of LMX and CWX on helping behavior (Figure 1), which shows that when LMX and CWX are congruent (i.e., LMX = CWX), there is a positive relationship between LMX, CWX, and helping behavior. In addition, we also calculated each point to plot (see Table 5). The point of HH employees is 2.13 and that of LL employees is 1.54, indicating the congruence of LMX and CWX has a positive relationship with helping behavior. As such, Hypothesis 1 is supported. Hypothesis 2 proposes that the incongruence of LMX and CWX (LMX = −CWX) has a negative relationship with helping behavior. Results showed that the slope coefficient was non-significant (a2 = 0.03, p = .61.; see Table 4). Thus, Hypothesis 2 was not supported.
Hierarchical Regression Analysis of the Simultaneous Effect of LMX and CWX on Outcomes.
Note. Tenure 1 = professional tenure; Tenure 2 = hospital tenure; Tenure 3 = ward tenure; POS = perceive organizational support; LMX = leader-member exchange; CWX = coworker exchange.
p < .05. **p < .01. ***p < .001.
Testing Slope and Curve Effect of LMX and TMX on Outcomes.
p < .01.
Response surface analysis for testing the simultaneous effect of LMX and CWX on helping behavior.
Results of Points to Plot.
Hypothesis 3 proposes that the congruence of LMX and CWX has a U-shaped relationship with helping behavior. Results showed that the curve effect of LMX and CWX on helping behavior was non-significant (curvature coefficient a2 = −0.02, p = .65.; see Table 4). Thus, Hypothesis 3 was not supported. Hypothesis 4 proposes that the incongruence of LMX and CWX (LMX = −CWX) has a U-shaped relationship with helping behavior. Results showed that the curvature coefficient was significant (curvature coefficient a4 = 0.24, p < .001; see Table 4). Figure 1 shows that the incongruence of LMX and CWX (LMX = −CWX) has a U-shaped relationship with helping behavior. Employees with moderate levels of both exchange relationships engage in the lowest helping behavior (1.89; see Table 5). Therefore, Hypothesis 4 was supported.
Discussion
This study proposes a simultaneous effect of the (in-)congruence of LMX and CWX on employees’ helping behavior. Specifically, this study performed a polynomial regression combined with response surface methodology to test both the linear and curvilinear relationships between LMX, CWX and helping behavior. Results indicated that the congruence of LMX and CWX has a linear positive relationship but not a U-shaped relationship with helping behavior. That is, employees with high LMX and high CWX demonstrated higher levels of helping behavior, whereas employees with low LMX and low CWX exhibited lower levels of helping behavior. Further, employees with moderate levels of both LMX and CWX did not exhibit lower levels of helping behavior compared to employees with high LMX and high CWX or employees with low LMX and low CWX. One the other hand, the incongruence of LMX and CWX does not have a linear negative relationship with helping behavior, but rather a U-shaped relationship. Specifically, compared with those who only develop high exchange relationships with either their leader or coworkers (i.e., HL or LH employees), those who develop moderate exchange relationships with both their leader and coworkers exhibit the lowest helping behavior.
Theoretical Implications
The main contribution of this study is that we addressed the call of previous researchers (G. Chen & Sharma, 2012; Cole et al., 2002; Lai et al., 2019; Seers, 1989) to examine the simultaneous effect of two exchange relationships (LMX and CWX) on employees’ helping behavior. In the workplace, although the frequency of interactions might vary, employees typically collaborate with both their leader and coworkers simultaneously rather than exclusively with one or the other. Thus, they develop exchange relationships with both parties through the role-making process. However, most studies still only consider the effect of one type of exchange relationship on various outcomes, overlooking the simultaneous effect of these two exchange relationships. Furthermore, to better understand the simultaneous effect of LMX and CWX on helping behavior, this study conducted a polynomial regression and used response surface methodology to examine both the linear and curvilinear relationships between these variables.
This study found that, after controlling the potential effect of previous helping behavior (Time 1) and POS, the congruence effect of LMX and CWX demonstrated a positive linear relationship with helping behavior. This finding is consistent with that of previous research (e.g., Baker & Omilion-Hodges, 2013; Dulebohn et al., 2012; Gerstner & Day, 1997; Ilies et al., 2007; Martin et al., 2016; Tang et al., 2021; Young et al., 2021), although these prior studies only examined one type of interpersonal relationship in the workplace (e.g., LMX or CWX). This study integrated the two stems of research and provided an integrated theoretical model to understand how the two types of workplace relationships influence employees’ helping behaviors simultaneously. It was found that, when employees develop high exchange relationships with both the leader and coworkers, they exhibit more helping behavior than those who have low exchange relationships with their leader and coworkers. For these HH employees, exhibiting helping behavior could be considered a means to reciprocate the benefits derived from their leader and coworkers and to maintain these high exchange relationships. In contrast, LL employees tend to limit their behavior to the obligations outlined in their employment contract, resulting in less frequent displays of helping behavior. We also proposed that an inflection point exists in this relationship, such that the relationship between the congruence of LMX and CWX and helping behavior is not linear but curved. However, this contention was not supported by our results. Our findings help advances the theories of workplace interpersonal relationships and helping behavior by offering a more comprehensive view on workplace interpersonal relationships and their effect on employees’ behavior.
The second contribution of this study is that we followed Lai et al.’s (2019) suggestion by examining the effect of moderate LMX and CWX on helping behavior. Results showed that, compared with HL and LH employees, those with moderate exchange relationships with both the leader and coworkers exhibit the lowest helping behavior. Most prior research only considered the effect of high or low exchange relationships on workplace outcomes, overlooking those with moderate exchange relationships with the leader and coworkers, who often have an awkward position in the team. From the leader and coworkers’ viewpoint, employees with a moderate level of LMX and CWX have capabilities better than those of LL employees but worse than those of HH employees, so they might receive some but limited resources from their leader and coworkers. As such, they might preserve their resources and energies for accomplishing tasks, or they might not feel obligated to offer help. Contrastingly, to maintain their exchange relationships and reciprocate what they receive from the leader or coworkers (e.g., resources and help), HL and LH employees must exhibit more helping behavior in return. Therefore, a U-shaped relationship exists between the incongruence of LMX and CWX and helping behavior.
Furthermore, previous studies tended to focus on the effect of the exchange relationship with the leader (LMX) but ignore the effects of coworkers. Although our results do not support a linear negative relationship between the incongruence of LMX and CWX and help behavior, our findings indicate that employees who develop high exchange relationships with either their leader or coworkers exhibit a similar level of helping behavior. Our findings could be explained by leadership substitutes theory (Kerr & Jermier, 1978), wherein exchange relationships with coworker in the workplace can serve as substitutes for the exchange relationship with the leader. Therefore, those who fail to develop high exchange relationships with the leader might not be appropriately classified as hired hands when they develop high exchange relationships with coworkers. This is consistent with Cole et al. (2002) and Lai et al.’s (2019) viewpoints that workplace exchange relationships are primarily compensatory. While this explanation seems reasonable, caution is needed when interpreting our findings. Leaders and coworkers have different levels of authority and different exchange resources (Cole et al., 2002; Cropanzano & Mitchell, 2005). For instance, both leaders and coworkers might offer emotional support for employees, but only a leader can offer a promotion. Thus, this substitute effect may only exist for certain types of resources.
Researchers have pointed out that employees spend more interacting time with coworkers than their leaders (Halbesleben, 2012; Seers, 1989), and coworkers have a greater influence on employees’ behavior than leaders (Chiaburu & Harrison, 2008). Similarly, we found that CWX has a greater effect than LMX on helping behavior under congruence conditions, despite the discrepancy in the slope of the first and second principal axes with the checklist by Humberg et al. (2019). These results are also consistent with the nature of organizational citizenship behavior directed toward individuals (OCBI; Williams & Anderson, 1991; e.g., helping behavior), such that helping behavior is discretionary and not specified in role prescriptions. Even though high LMX levels increase helping behavior (Ilies et al., 2007), it does not necessarily imply that high LMX employees are obligated to exhibit such behavior, especially in the absence of any consequences (Van Dyne & LePine, 1998). Moreover, at the minimum level, employees have a formal relationship (e.g., an employment contract) that establishes their connection with the leader and provides access to the associated resources. However, there is no such relationship between employees; thus, employees need to invest their efforts (e.g., offer help) to develop and maintain positive relationships with their coworkers. Thus, based on our results and previous research (e.g., Lai et al., 2019), future research should consider the influence of coworkers when examining employees’ behavior. Furthermore, it is important to examine different patterns of influence on employees’ behavior. For instance, we might expect that a leader has greater influence on employees’ organizational citizenship behavior directed toward the organization (OCBO; Williams & Anderson, 1991) than coworkers.
Our results also revealed interesting findings beyond the scope of this study. One might assume HH employees would exhibit the highest level of helping behavior among all five types, but our results showed that HL and LH employees exhibited the highest level of helping behavior (see Table 5 and Figure 1). One possible explanation is that the norms of reciprocity surrounding helping behavior may vary across different type of employees (Rious & Penner, 2001). Incongruence employees (i.e., HL and LH employees) and congruence employees (i.e., HH employees) might have different norms of reciprocity dictating their helping behavior. Low exchange relationships are characterized by equal exchanges, whereas high exchange relationships often involve unequal exchanges (Uhl-Bien & Maslyn, 2003), where one party may receive more benefits than the other (Liden et al., 1997; Sparrowe & Liden, 1997). For HH employees, because they trust, respect, and are committed to their leader and coworkers, they may engage in exchanges that are characterized by unequal value (Uhl-Bien & Maslyn, 2003). Consequently, they may be less motivated to engage in helping behavior compared to HL and LH employees. In contrast, the norms of reciprocity for HL and LH employees might be more complicated. Unlike HH employees who can develop high exchange relationships with both their leader and coworkers, HL and LH employees are only able to develop such high exchange relationships with one part (e.g., leader or coworkers). Consequently, they cannot fulfill their needs through their low exchange relationships and are compelled to maintain their high exchange relationships (Cole et al., 2002), such as by exhibiting more helping behavior than HH employees. For instance, prior research found that employees would seek their leader’s support when they could not meet their needs via their relationships with other coworkers (Major et al., 1995).
In sum, future research should consider both exchange relationships (LMX and CWX) simultaneously to understand the full picture of workplace relationships and their effect on various outcomes. In addition, to advance workplace exchange relationship theory, future research should consider both the linear and curvilinear relationships among these variables.
Practical Implications
This study provides several practical implications for leaders and practitioners. First, to increase employees’ helping behavior, leaders and practitioners should develop high exchange relationships with all—or at least most—employees (Graen & Uhl-Bien, 1995). Results of this study suggest that leaders and practitioners should not only focus on LMX, but also on CWX. According to our findings, three types of exchange relationships (i.e., HH, HL, and LH) can increase employees’ helping behavior. However, while employees with low LMX can make valuable contributions to the organization by developing high exchange relationships with their coworkers, leaders and practitioners often overlook or disregard these employees. Based on our findings, leaders and practitioners should encourage low LMX employees to develop high exchange relationships with their coworkers to foster helping behavior. Providing training programs for interpersonal relationship development might help leaders and practitioners reach this goal (Cole et al., 2002).
Second, our results showed that employees with moderate exchange relationships with both their leader and coworkers also deserve attention (Lai et al., 2019). Compared with the other four types of employees (HH, HL, LH, and LL), these employees hold an ambiguous position in the team or organization, which could potentially reduce their inclination to engage in helping behavior. Because they do not have enough capabilities to develop high exchange relationships with their leader and coworkers, leaders and practitioners should help them in cultivating such relationships with either their leader or coworkers (i.e., to become HL or LH employees), which can potentially enhance their helping behavior. In addition, leaders and practitioners should avoid categorizing employees with high CWX and low LMX as rebels (Seers, 1989). Rather, they should also recognize these employees as valuable contributors, similar to their HH and HL counterparts, because they offer help when requested and thus contribute to the achievement of organizational goals.
Future Research and Limitations
This study provides some insightful avenues for future research on workplace relationships. First, future research might examine other outcome variables, such as voice. Given that voice is an extra-role behavior that may damage one’s relationships with others (Van Dyne et al., 1995), research should examine the effect of the (in-)congruence of LMX and CWX on voice. In addition, researchers should consider both the linear and curvilinear relationships between these variables. Second, it might be meaningful to examine possible underlying mechanisms among these variables. For instance, work engagement, which refers to “the simultaneous employment and expression of a person’s ‘preferred self’ in task behaviors that promote connections to work and to others, personal presence (physical, cognitive, and emotional), and active, full performances” (Kahn, 1990, p. 700), may mediate the effect of the (in-)congruence of LMX and CWX on helping behavior. Work engagement might mediate the linear relationship between the congruence of LMX and CWX and helping behavior, as well as the U-shaped relationship between the incongruence of LMX and CWX and helping behavior, such that employees with HH, HL, or LH relationships with their leader and coworkers might be inclined to help others because they are more engaged in their work than moderate-level employees. In sum, we encourage future research to examine other possible outcomes and possible underlying mechanisms.
This study has some strengths. First, we adopted a time-lagged research design and collected data from different time points to mitigate the potential influence of CMV. Second, consistent with prior research, this study controlled for the helping behavior at Time 1 to account for its potential effect on subsequent helping behavior. Third, based on a workplace exchange relationship perspective, we controlled for POS to mitigate its impact on employees’ helping behavior. Alongside the strengths, the study also has some limitations. The first limitation pertains to methodological issues. This study collected data from the same source (employees) at Time 1 and 2, and while controlling for Time 1 helping behavior to address potential concerns of CMV (Podsakoff et al., 2012), this approach may have attenuated our results. Future research should consider collecting data from multiple sources, such as leaders or coworkers, to address this concern. Moreover, in future research, it is important to be mindful of the potential wording effect that may exist in the measures and take steps to minimize its effect on the data (Podsakoff et al., 2012; Wu, 2008). Second, nearly all our participants are female and work at the same organization (a hospital). In addition, the interactions between employees and their leaders and coworkers might be influenced by other factors (e.g., organizational structure and culture; Cole et al., 2002). Thus, explaining and generalizing our results should be done with caution to other occupations, organizations, and industries. Future research is encouraged to replicate our theoretical model in different settings.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Ministry of Science and Technology in Taiwan (grant number: MOST-107-2410-H-020-011-SS3).
