Abstract
Current theory and research suggest a positive linear relationship between perceived organizational support (POS) and employees’ taking charge, or change-suggesting, behaviors. Via a sample of 89 subordinate-supervisor dyads, we hypothesize and test an inverted U-shaped relationship between employees’ POS and their taking charge behaviors and the likelihood that this curvilinear pattern is accentuated when employees anticipate costs related to their taking charge. Results support both of these patterns. We conclude by discussing our findings’ implications regarding future needed sensitivity on the part of managers as well as management scholars regarding how to gauge optimal levels of POS.
Prior work has theoretically and/or empirically noted that employees’ proactive behaviors in general and their “taking charge” behaviors in particular (i.e., change-making or improvement-making actions) tend to be greater when employees are in an organizational environment that rewards them for taking charge (Zhou & George, 2001), have a supervisor who they perceive as open to taking charge (Ashford, Rothbard, Piderit, & Dutton, 1998; Detert & Burris, 2007; Morrison & Phelps, 1999; Saunders, Sheppard, Knight, & Roth, 1992), and have a positive relationship with their supervisor (Milliken, Morrison, & Hewlin, 2003; Van Dyne, Kamdar, & Joireman, 2008). All of these predictors seem to be indicators of some form of support, and as a result, it is not surprising that a more direct measure of perceived support—perceived organizational support (POS)—has also been positively linked to employees’ extent of proactive behaviors, including taking charge (Choi, 2007; Eisenberger, Armeli, Rexwinkel, Lynch, & Rhoades, 2001; Ohly, Sonnentag, & Pluntke, 2006). For this reason, we focus on POS as the predictor of employees’ proactive behaviors—specifically, taking charge. Consistent with social exchange theory (Blau, 1964), we note as past scholars have that this positive linear relationship between POS and taking charge is likely due to employees feeling a desire to reciprocate the support they have received. Yet the relationship that POS has with taking charge has been sometimes found to be negative, as we detail below, and as such fails to match a social exchange theory–based explanation (Lambert, 2000).
In this study, we propose that the negative relationship between POS and taking charge might be an indication that there is a dark side to employees perceiving greater than optimal levels of POS. By “dark side,” we mean that greater than optimal levels of POS may be associated, ironically, with undesirable employee behaviors such as less helpfulness and initiative taking in general, or less taking charge in particular. By analogy, children of overly nurturing parents have been found to be less confident and independent later in life (Clarke, Dawson, & Bredehoft, 2004); and similarly, students of overly nurturing teachers have been found to be weaker in pursuing their own interests or in solving problems on their own (Levine, 2002). These overnurturing effects can be explained by the threat-to-self-esteem model (Fisher, Nadler, & Whitcher-Alagna, 1982; Nadler & Fisher, 1986), which we explain in our literature review.
Since greater than optimal levels of support have been associated with less than optimal outcomes when the providers of support are parents or teachers, it is surprising that efforts to explain the inconsistent POS–taking charge relationship lack both theoretical complexity and the corresponding empirical testing. Guided by the likelihood that there is an optimal level of POS beyond which negative effects (e.g., less taking charge) associated with POS occur, we posit that the relationship POS has with taking charge takes the shape of an inverted U such that the positive effects of POS on taking charge occur before (not after) optimal levels of POS are reached. This curvilinear relationship may explain findings that are inconsistent with the positive linear point of view that employees with greater organizational support will take charge more frequently. We derive this inverted U-shaped hypothesis (a) from the sometimes-positive and sometimes-negative direct effect of POS on taking charge and other proactive behaviors, (b) from management scholars, such as Grant and Schwartz (2011) and Pierce and Aguinis (2013), suggesting the need to question when good things (such as supportiveness) may backfire when they are provided in excess, and (c) from our integration of social exchange theory and the threat-to-self-esteem model.
In addition to offering theory that supports an inverted U-shaped relationship between POS and taking charge, we theorize that this inverted U shape will be more likely to be observed when employees more strongly anticipate, or expect, that they may suffer negative consequences for making change suggestions, or taking charge—a moderator we refer to as anticipated costs. Our reasons for identifying anticipated costs associated with taking charge as a moderator of the inverted U-shaped curve between POS and taking charge described above are twofold. First, it has been theoretically and empirically documented that employees tend to pay more attention to contextual cues when they have experienced severe work-related outcomes (e.g., Brockner, Konovsky, Cooper-Schneider, Folger, Martin, & Bies, 1994; De Cremer, Brockner, Fishman, van Dijke, van Olffen, & Mayer, 2010; Shapiro, Buttner, & Barry, 1994) or when they are more highly concerned about potential harm coming to them (cf. Janis & Mann, 1977). Second and similarly, when employees are more uncertain about the stability of their work structures, they tend to be more preoccupied with how fairly treated they are (Lind & van den Bos, 2002; van den Bos, 2001; van den Bos & Lind, 2002). The uncertainty that is described in the latter works pertains to anticipated negative events. The latter reasons suggest that employees’ ability to pick up on environmental cues is greater when they are in a state of uncertainty about their well-being. Put differently, the reasoning shown here suggests that employees will be more aware of varying levels of POS (just as they are about how fairly they are being treated) when they anticipate potentially suffering negative consequences associated with taking charge.
Purpose and Contributions of This Study
The purpose of this article is to revisit the POS–taking charge relationship. By testing the relationships we do—namely, whether POS relates to taking charge in an inverted U shape and when this shape is most likely to be observed—our article makes three contributions. Our first contribution is that we offer an integrative theoretical framework based on social exchange theory and the threat-to-self-esteem model, in an effort to specify with more accuracy a curvilinear relationship between POS and taking charge. We see this theoretical extension as important because explanations based exclusively on social exchange cannot account for existing inconsistent empirical findings. Our second contribution is that we go beyond the linear perspective that dominates the POS literature. This contribution is important both from a theoretical perspective (increased model specification) and practically (organizations can optimize their allocation of resources if they are aware of an inflection point).
Our third contribution is that we posit and test one important boundary condition, in the form of employee anticipated costs of taking charge. Our specific moderator—anticipated costs—is important to account for especially in the context of taking charge behaviors since these behaviors (like “speaking up”) are seen as carrying risks for the employees who may engage in them. Our choice of moderator also responds to calls for research such as Baer and Oldham’s (2006) call to examine boundary conditions of nonlinear relationships, and extends work where such propositions were presented only in a theoretical form (cf. Morrison, 2011). The remainder of this article reviews literature guiding the hypotheses we test, describes the methodology by which we tested our hypotheses and our findings, and concludes with a discussion of implications for managers as well as scholars interested in conditions that may best encourage employees to take charge.
Literature Review and Hypotheses
As noted earlier, researchers have found POS or other types of support, such as supervisor support, to be positively associated with employees’ organizational-directed proactive behaviors (e.g., Choi, 2007). Yet not all studies find this positive linear relationship. Studies whose findings are inconsistent with a positive influence of support include Lambert (2000) and Ohly et al. (2006), who found negative relationships, whereby high levels of support were associated with fewer employees’ organizationally directed proactive behaviors. Furthermore, the studies of Moorman, Blakely, and Niehoff (1998), Baer and Oldham (2006), and Fuller, Hester, Barnett, Frey, Relyea, and Beu (2006) found a nonsignificant relationship between employees’ perceptions of support and the extent to which employees share ideas in the workplace about ways to improve individual performance, working conditions, quality, or procedures. This inconsistent set of findings may be due to an unexamined maximum threshold effect of POS that we posit exists. The speculations offered by scholars for the nonpositive findings involving POS and employees’ taking charge (or other organizationally directed proactive) behaviors are nonuniform and regard unmeasured variables. Yet one source of commonality across the latter studies is the linear standpoint underlying their theorizing and empirical testing—that is, the assumption of a positive upward-sloping line between support perceived by employees and their level of proactivity, such as taking charge.
POS—Taking Charge: An Inverted U Relationship Perspective
We integrate two theories to explain why an inverted U shape seems likely to characterize how employees’ POS relates to taking charge. These theories, which until now have been treated in isolation of each other, are social exchange (Emerson, 1976; Lavelle, Rupp, & Brockner, 2007) and the threat-to-self-esteem model (Fisher et al., 1982; Nadler & Fisher, 1986).
Social exchange theory
With regard to social exchange theory, this is based on two principles: first, that two or more associated parties (e.g., employees and organizations) are necessary to have a social exchange; and second, that each party feels an obligation to reciprocate positive rewards from the other party (Emerson, 1976; Lavelle et al., 2007). Social exchange theory has been used to connect POS and employee effectiveness, with the traditional theoretical explanation being that support generates reciprocation and increased positive affect (e.g., Eisenberger et al., 2001) that, in turn, increases worker effectiveness. Although social exchange explanations have been empirically supported when explaining how POS affects work-related effectiveness conceptualized as task performance and affiliative forms of citizenship (Organ, Podsakoff, & MacKenzie, 2006; Rhoades & Eisenberger, 2002), social exchange explanations have been less predictive of change-oriented behaviors such as taking charge. For example, an inability of social exchange explanations to predict proactive types of behaviors relating to change suggestions such as taking charge has been reported by Lambert (2000) when she failed to support a positive, instead finding a negative relationship between organizational support and employees’ making change-oriented suggestions. Such findings suggest that theory other than social exchange is needed to more accurately predict when POS will encourage employees to take charge, or offer improvement suggestions.
Threat-to-self-esteem model
A basic tenet of the threat-to-self-esteem model is that receiving social support at work will sometimes have a negative effect because it results in feelings of inferiority and incompetence (Fisher et al., 1982; Nadler & Fisher, 1986). Insights from the threat-to-self-esteem model, especially those associated with descriptions of this model’s dynamics (cf. Deelstra, Peeters, Zijlstra, Schaufeli, Stroebe, & van Doornen, 2003), may therefore help to explain why a weakening of taking charge on the part of employees is likely when they perceive an overabundance of support.
These reasons are all guided by Deelstra et al.’s (2003) explanation that an overabundance of support, for example in the form of helping employees with their tasks, can threaten recipients’ self-esteem; and if this indeed occurs, then the recipients will be less likely to respond in ways that reciprocate helpfulness, such as by behaving proactively (e.g., taking charge). Put more succinctly, Deelstra et al. explain that employees’ reaction to support depends on their perceptions of whether or not the support threatens their self-esteem. Deelstra et al. explain, furthermore, that too much support (e.g., too much task-related assistance) directed at employees can lead them to (a) feel incompetent and (b) feel restricted in their freedom and autonomy. Moreover, they note that these two dynamics threaten employees’ self-esteem. The reason why employees feel incompetent when they receive too much support is because an overabundance of support can inadvertently signal to the recipients that those offering the support do not trust them to be skilled enough to perform without help. Similarly, compared to children whose parents give them greater autonomy, children of overnurturing parents generally behave less independently, use their competencies to a lesser extent, and feel more self-doubt about their abilities (e.g., Clarke et al., 2004; Ogilvie, 2006). Deelstra et al. explain that an overabundance of support can also lead employees to perceive that their freedom of choice is limited when they receive support they did not ask for.
Empirical findings supportive of the latter dynamics are many. Specifically, these include Fisher et al.’s (1982) finding that support can sometimes lead recipients to feel incompetent, dependent, or embarrassed, Chow and Lowery’s (2010) finding that a lack of gratitude tends to be felt by employees who do not believe they are responsible for the support they are receiving (as may be the case if support is overly abundant), and several scholars’ finding that at high levels of received support at work there are higher levels of negative affect (Buunk, Doosje, Jans, & Hopstaken, 1993), physical symptoms (Hahn, 2000), and depersonalization (Iverson, Olekalns, & Erwin, 1998), and also more frequent and longer spells of absenteeism (Rael, Stansfeld, Shipley, Head, Feeney, & Marmot, 1995). Cumulatively, the latter observations support Miller’s (2005) contention that explanations other than social exchange are needed to more accurately predict or explain consequences of support. We find the ideal alternative model to be an integration of social exchange theory for explaining how employees likely respond to moderate levels of POS relative to lower levels of POS and the threat-to-self-esteem model (Fisher et al., 1982; Nadler & Fisher, 1986) for explaining how employees likely respond to supraoptimal levels of POS.
Integration of social exchange and threat-to-self esteem model
Extrapolating from the two theories reviewed above, it logically follows that when the POS employees perceive is moderate relative to low, they will likely respond positively to it, and thus behave proactively, an example of which is more taking charge to a greater extent. However, when the POS employees perceive exceeds a threshold and becomes “overabundant,” employees’ taking charge is likely to be weakened. Our theorizing this threshold effect of the “good thing” we call POS is consistent with recent propositions by Grant and Schwartz (2011) and by Pierce and Aguinis (2013), all of whom argue there is likely a maximum optimum level of “good things,” such as POS, beyond which unintended negative effects occur. Our theorizing a threshold effect associated with POS is consistent, also, with the threat-to-self-esteem model, an important premise of which is that receiving help is neither all good nor all bad; it is the relative degree of self-threat and self-support that ultimately determines the recipient’s reaction to received help. The model further predicts help that is perceived as self-supportive will elicit positive reactions, whereas help that is perceived as self-threatening (which we posit happens when the level of POS exceeds a moderate level) will elicit negative reactions.
Once we integrate the threat-to-self-esteem model with social exchange theory, it becomes clear why the POS–taking charge relationship is likely an ascending positive slope from lower to moderate levels of POS yet a descending slope once POS becomes supraoptimal (goes beyond moderate levels). Specifically, at lower levels of POS, employees likely feel no obligation to help their organization. The possibility that at higher levels of POS, employees may become inept was also suggested by Lambert (2000) as an explanation for why at high levels of POS she did not observe high levels of taking charge, but Lambert’s linear approach prevented her from directly testing the inverted U-shaped pattern described here. What might be distinct at moderate levels of POS is that this provides employees a balanced state in terms of the amount of care received from their organizations and the amount of support and organizationally directed contributions they are comfortable giving. Thus, we predict,
Hypothesis 1: There will be an inverted U-shaped relationship between employees’ POS and employees’ taking charge behaviors, such that taking charge is highest when POS is moderate and lower when POS is either lower or higher.
Contingency of Anticipated Costs on the Inverted U-Shaped Relationship
When might the inverted U relationship predicted by Hypothesis 1 be more likely to be observed? We propose that employees will scrutinize their work environments more carefully for cues that they are being supported when they are worried about suffering negative consequences if they take charge—hence when they anticipate costs associated with doing this. That employees evaluate the cost of their proactive behaviors results clearly from the findings of Milliken et al. (2003), whose 22.5% of respondents mentioned the possibility of retaliation or punishment for presenting ideas, or the more general stance in the voice literature mentioning fear as a motive for lack of involvement (Kish-Gephart, Detert, Treviño, & Edmondson, 2009; Van Dyne, Ang, & Botero, 2003). As a result of these risks, however, employees are likely to have concerns about potential (un)receptivity to their taking charge. The kinds of costs that employees may worry about if they make improvement suggestions include being replaced, penalized, or losing recognition associated with being a top performer or good citizen. Such issues have been found to be among those that employees express when they contemplate whether or not to speak up in general (Milliken et al., 2003).
We have two reasons for expecting employees who more strongly anticipate costs associated with taking charge to be more cognizant of varying levels of POS in their work environment—hence to behave in ways matching the inverted U hypothesis. First, theorizing and findings document the tendency for employees to pay more attention to contextual cues when they have experienced severe work-related outcomes (e.g., Brockner et al., 1994; De Cremer et al., 2010; Shapiro et al., 1994), or when they are more highly concerned about potential harm coming to them (cf. Janis & Mann, 1977). Second, theorizing and findings by scholars document the tendency for employees to be more preoccupied with how fairly treated they are when they are more uncertain about the stability of the work structures that exist (Lind & van den Bos, 2002; van den Bos, 2001; van den Bos & Lind, 2002). Such uncertainty thus seems related to anticipated negative events (e.g., instability). The latter reasons suggest that employees’ ability to pick up on environmental cues is greater when they are in a state of uncertainty about their well-being. Put differently, the reasoning shown here suggests that employees will be more aware of varying levels of POS (just as they are about how fairly they are being treated) when they anticipate potentially suffering negative consequences associated with taking charge. Thus, we predict,
Hypothesis 2: The tendency for employees’ taking charge to be greatest at moderate (rather than lower or higher) levels of POS, hence for taking charge to be related to POS in an inverted U-pattern (as predicted by Hypothesis 1), is stronger for employees who perceive higher costs related to their taking charge behaviors drawing negative reactions.
Method
Sample and Procedure
Our sample was drawn from an aluminum company on the Caribbean island of Jamaica. Employees participating in this study were from a variety of job categories, including accounting, finance, human resources, engineering, operations and maintenance, and administration. We invited 433 employees and their immediate supervisors from different departments to participate in this study. Informed consent by all study participants was obtained before the study began, and the data collection process ensured confidentiality of the responses. The employee questionnaire contained items measuring POS, employee anticipated costs of taking charge, and the control variables (age, gender, level of education, job category, organizational tenure). A total of 316 employees returned completed questionnaires, representing a 73.0% employee response rate. The supervisor questionnaire contained items measuring subordinates’ taking charge behaviors and requested that supervisors evaluate only the taking charge behaviors of employees working in their department. We received supervisors’ evaluations of the taking charge behaviors of 28.2% of those who participated in the employee data collection, resulting in 89 subordinate–supervisor matched pairs. Of the 89 employees in the final sample, 95% were male and 75% had college or higher degrees. The average age was 41 years (SD = 8.8), and average company tenure was 14.6 years.
Measures
To minimize common method variance (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003), data were collected from different sources (i.e., supervisors and subordinates). Subordinates completed measures of POS and anticipated costs, while supervisors rated subordinates’ taking charge behaviors.
Employees’ perceived organizational support
To assess POS, employees indicated (via a 7-point Likert-type scale ranging from 1 = strongly disagree to 7 = strongly agree) the strength with which they agreed with eight items from the Survey of Perceived Organizational Support (Eisenberger, Fasolo, & Davislamastro, 1990; Eisenberger, Huntington, Hutchison, & Sowa, 1986). The eight-item version of POS has been widely used in previous studies and demonstrated strong psychometric validity (Farh, Hackett, & Liang, 2007; Settoon, Bennett, & Liden, 1996). These eight items have previously been found to have high factor loadings on the POS construct and to be applicable to a wide array of organizations (Eisenberger, Cummings, Armeli, & Lynch, 1997). Sample items include “The company really cares about my well-being” and “The organization strongly considers my goals and values.” In our study, the estimated reliability of this scale was .88.
Employees’ anticipated taking-charge-related costs
Since no scales exist to measure employees’ cognitions about potential losses emerging from making improvement suggestions, we generated and validated new items, via a three-stage process (see the appendix for further details regarding the scale validation process). The results of this three-stage process led us to assess respondents’ anticipated reward loss, by asking them to indicate (via a 7-point scale where 1 = very unlikely and 7 = very likely) the extent to which various consequences associated with making improvement suggestions in their organization would likely result; these seven statements followed the stem “If I made suggestions for (task- and/or procedural-) improvements in my organization. . . .” The seven statements following this stem were (a) “I would be replaceable,” (b) “I would lose recognition associated with being a top performer,” (c) “I would lose recognition as a good citizen,” (d) “I would be overlooked as a candidate for a promotion,” (e) “I would miss out on bonuses associated with being a top performer,” (f) “I would receive less pay as a result of helping others,” and (g) “I would be penalized if helping others impedes my ability to complete my own work.” The estimated reliability of this scale was .90.
Employees’ taking-charge behavior
To assess employees’ taking charge behavior, supervisors indicated via a 7-point Likert-type scale (ranging from 1 = strongly disagree to 7 = strongly agree) the extent to which they agreed with descriptions of their employees’ taking charge behaviors, using six items from Morrison and Phelps (1999). An illustrative item is “This employee tries to adopt improved procedures for doing his or her job.” The estimated reliability of our six-item scale was .94.
Control variables
Because the participants were drawn from different job categories, which may influence employees’ taking charge behaviors, we controlled for their job type (coded as 1 = nonunionized and 0 = unionized). In addition, consistent with prior research on organizationally directed behavior antecedents, employees’ age, gender, education (coded as 1 = high school graduate and 0 = college graduate), and organizational tenure were included as controls (LePine, Erez, & Johnson, 2002; Morrison & Phelps, 1999).
Data Analysis
Following existing recommendations and recent empirical studies (Aiken & West, 1991; Cohen, Cohen, West, & Aiken, 2003; Le, Oh, Robbins, Ilies, Holland, & Westrick, 2011) we used hierarchical polynomial regression analyses to test the hypotheses. More specifically, we entered the predictors into the regression equation at five hierarchical steps, in this order: (a) the control variables, (b) POS, (c) quadratic term of POS-squared, (d) the moderator variable of anticipated costs and the interactions between the moderator and POS, and (e) the interactions between the moderator variable and the quadratic term. In addition to testing the significance of regression coefficients, we also interpreted the R2 change (ΔR2) associated with a particular step at which a term testing a certain hypothesis was entered (Cohen et al., 2003; Le et al., 2011; Zhou, Shin, Brass, Choi, & Zhang, 2009).
To increase interpretability (Aguinis, 2004; Dalal & Zickar, 2012), we standardized the predictor variables and then computed the higher order terms (Aiken & West, 1991; Liao, Liu, & Loi, 2010; Marquardt, 1980). Because high correlations among predictors may lead to biased interpretation of the regression results (Cohen et al., 2003), we further conducted collinearity diagnostic analyses to detect any potential problems. The variance inflation factors were below 3 for all predictors, ranging from 1.1 to 2.7, which fall below the normally accepted cutoff value of 10.0 (e.g., Kleinbaum, Kupper, & Muller, 1988).
The means, standard deviations, and correlations among variables are shown in Table 1. We performed a series of confirmatory factor analyses (CFAs) in LISREL (Jöreskog & Sörbom, 1993) to establish the discriminant validity of the measures. Given our relatively small sample size, we constructed item parcels in the CFAs to maintain a favorable indicator-to-sample-size ratio. Three indicators were randomly formed for each latent construct that has more than three items. Specifically, we compared the proposed three-factor model (anticipated costs, POS, and taking charge behavior) with one alternative model in which the correlation between each pair of factors is fixed to 1 by conducting chi-square difference tests to show that the model with the freely estimated correlations displays superior fit to each model with fixed correlations (Bagozzi, Yi, & Phillips, 1991). The baseline three-factor model fit the data well, χ2(24) = 25.47, root mean square error of approximation (RMSEA) = .03, nonnormed fit index (NFI) = 1.00, and comparative fit index (CFI) = 1.00. One alterative model in which the correlation between anticipated costs and POS were fixed to one exhibited significantly worse fit than the baseline model, Δχ2(1) = 49.08, p < .01, RMSEA = .15, NFI = .86, CFI = .90. Thus, we treated the three variables as independent constructs in all our subsequent analyses.
Means, Standard Deviations, Reliabilities, and Intercorrelations Among Study Variables
Note: N = 85-89, listwise deletion. Education: 1 = high school graduate and 0 = college graduate. Job category: 1 = nonunionized and 0 = unionized. Estimated reliabilities (Cronbach’s alphas) are shown on the diagonal. Correlations greater than |.22| are significant at p < .05. Correlations greater than |.28| are significant at p < .01.
Results
Relying on researchers’ recommendations on model and effect testing with higher order terms, we tested our hypotheses using the global model (Aiken & West, 1991: chap. 6; Friedrich, 1982). For completeness, we also report corresponding step-down hierarchical models (Aiken & West, 1991). Consistent with Hypothesis 1, we found evidence supporting an inverted U-shaped relationship between employees’ POS and taking charge behaviors. This evidence can be seen in Model 5 of Table 2 where the quadratic term is significantly (negatively) related to taking charge behavior (β = –.36, p < .05). For completeness, we also report the ΔR2 associated with the quadratic term, which was statistically significant (ΔR2 = .07, p < .05). We calculated the inflection point to determine at which point the effect of POS on taking charge declines. To calculate the inflection point, we used the following equation for the polynomial regression of POS: Y = B0 + B1X + B2 X2 + ε. We then computed the inflection point by the coefficients of the predictor X as shown by the following equation (Weisberg, 2005): Xinflection = – B1/2B2. Our inflection point results show that as POS increased, employees’ taking charge behavior also tended to increase; however, after a standardized inflection point of .06 was reached, increased levels of POS became associated with fewer taking charge behaviors. Additional support for Hypothesis 1 is shown via Figure 1, which we plotted using the steps recommended by Aiken and West (1991). Thus, Hypothesis 1 is supported.
Results of Hierarchical Regression Analyses Predicting Taking Charge
Note: N = 85-89. Education: 1 = high school graduate and 0 = college graduate. Job category: 1 = nonunionized and 0 = unionized.
p < .05, two-tailed test.
Curvilinear Relationship Between Perceived Organizational Support and Taking Charge
Consistent with Hypothesis 2, we found that the inverted U-shaped curvilinear relationship between employees’ POS and taking charge behaviors (described above) is significantly moderated by employees’ anticipated costs related to engaging in these behaviors. This evidence can be seen in two ways. First, this moderating effect is seen via significant interaction terms involving POS with anticipated costs in Model 5 of Table 2 (β = –.40, p < .05). The ΔR2 associated with the anticipated costs interaction-term was statistically significant (ΔR2 = .06, p < .05). We calculated the inflection point to determine at which point the effect of POS on taking charge declines at high and low levels of anticipated costs. To calculate the inflection point, we used the following equation for the polynomial regression of POS moderated by employee anticipated costs: Y = B0 + B1X + B2 X2 + B3Z + B4 ZX + B5ZX2 + ε. We reformulated the latter equation to compute the inflection points at the different values of the moderator, as follows: Y = B0 + B3Z + (B1 + B4Z)X + (B2 + B5Z)X2 + ε. We then computed the inflection point by the coefficients of the moderator Z as shown by the following equation (Weisberg, 2005): Xinflection = –(B1 + B4Z) / 2(B2 + B5Z). Our inflection point results show that when anticipated costs were high, as POS increased, employees’ taking charge behavior also tended to increase; however, after a standardized inflection point of –.30 was reached, increased levels of POS became associated with fewer taking charge behaviors. In contrast, when anticipated costs are low, the inflection point was 3.13, indicating that high levels of taking charge behaviors are associated with lower, moderate, and higher levels of POS.
Additional support for the moderating effect of employees’ anticipated costs on the inverted U-shaped relationship between employees’ POS and taking charge behaviors can be seen by the plot shown in Figure 2. More specifically, as seen there, the relationship between POS and taking charge behavior followed an inverted U-shaped function for employees with high, not low, levels of anticipated costs—as depicted by the solid versus dotted lines, respectively. Even though we obtained consistent results for our squared and interaction terms across various models presented in Table 2, we also advise researchers to exercise caution when interpreting models containing higher order terms (for a detailed treatment of this issue, see the discussion on sequential step-down procedures vs. the hierarchical step-up approach; Aiken & West, 1991: 105-113).
Curvilinear Interaction of Perceived Organizational Support and Anticipated Costs on Taking Charge
We probed this interaction further, using a procedure from previous research (e.g., Ng & Feldman, 2010) to provide a sense of whether employees’ taking charge behavior was significantly greater at moderate levels of POS compared to lower or higher levels of this for the employees with high levels of anticipated costs. To do so, we first estimated POS at lower, moderate, and higher levels of POS (i.e., at the 33rd, 50th, and 66th percentiles for each of these POS levels, respectively) and estimated low and high levels of taking-charge-related anticipated costs (i.e., at one standard deviation below and above the mean for low and high anticipated costs, respectively). Second, we ran pairwise comparisons of employees’ taking charge behavior at each of the three levels of POS, for employees with high versus low levels of anticipated costs. Consistent with Hypothesis 2, these pairwise comparisons showed that for the employees with high (but not low) levels of anticipated costs, there was a significantly higher level of reported taking charge behavior at moderate levels of POS compared to lower POS (d = .74, p < .05) and compared to high POS (d = .84, p < .05). In contrast, for the employees apparently unafraid about suffering negative consequences for sharing change suggestions (low anticipated costs), there was not a significantly higher level of reported taking charge behavior at moderate levels of POS compared to lower POS (d = .31, ns) and compared to high POS (d = .08, ns); all levels of POS had approximately an equal high degree of taking charge behavior. Thus, Hypothesis 2 was supported in the predicted direction, namely, the inverted U-shaped curve was observed for employees high, not low, in anticipated costs.
Discussion
At our article’s outset, we noted that higher levels of POS are presumed a more positive state of affairs by principles using a linear approach, as is typically the case in management research. Yet, more recently, researchers have entertained the possibility that even positive organizational attributes have “thresholds” beyond which their effects may become less positive (cf. Grant & Schwartz, 2011; Pierce & Aguinis, 2013). Collectively, our findings suggest two specific conclusions. We discuss these two conclusions’ theoretical and practical implications next, each in turn.
Conclusion 1: Moderate (Rather Than Lower or Higher) Levels of POS May Be Optimal for Encouraging Employees to Take Charge
Our first conclusion, guided by our theorizing and observing an inverted U-shaped relationship between employees’ POS and the extent to which they take charge, is that moderate levels of organizational support may be optimal for encouraging employees to take charge in their organization. A practical implication of this is that there is apparently a threshold of POS beyond which there is likely to be deleterious returns. Consistent with this pattern, too much support from organizations may “overload” employees since supraoptimal levels of “care” may be perceived as overwhelming, overly controlling, or otherwise excessive (cf. Ilies, Wilson, & Wagner, 2009; Kossek, Pichler, Bodner, & Hammer, 2011; Perlow, 1998). Indeed, recipients’ reactions to support are not always positive; at times, they may feel threatened or incompetent (Fisher et al., 1982; Lee, 2002), and even “overhelped” (Gilbert & Silvera, 1996: 678). Our current findings extend these lines of work by focusing on support from the organization and substantiating a specific pattern involving POS and taking charge—namely organizations may, rather counterintuitively, gain higher levels of taking charge from employees by providing less is more-oriented support programs. Such a view may be especially welcome during times, such as today, of a recessionary resource-constrained economy.
In addition to the latter practical implication, our first conclusion has important theoretical implications. First, our findings support the positive linear relationship previously described for the association that POS has with taking charge when we compare employees’ taking charge at moderate relative to lower levels of POS. As such, our findings are consistent with social exchange-oriented theories describing the tendency for employees who feel supported by the organization to reciprocate this with behaviors that benefit the organization (Lynch, Eisenberger, & Armeli, 1999; Rhoades & Eisenberger, 2002; Riggle, Edmondson, & Hansen, 2009). Second, our findings support a relationship between POS and taking charge that takes the form of an inverted U when we compare employees’ taking charge at moderate relative to supraoptimal levels of POS; as a result, a linear relationship does not capture the complexities of the POS–taking charge relationship. Instead, our finding the inverted U, as we predicted, supports the likelihood that there can be negative consequences for organizations providing their employees either too little or too much support. Providing evidence of this “POS threshold” enables us to help explain why POS may be positively related, negatively related, or unrelated to employees’ proactive behaviors (e.g., Lambert, 2000). In addition, providing evidence of this “POS threshold” enables us to reinforce theorizing and findings in other studies that regard curvilinear relationships between contextual variables, such as Harris, Kacmar, and Witt’s (2005) finding a U-shaped relationship between employees’ perception of the quality of their exchange relationship with their supervisors and their turnover intentions and van Ruysseveldt and van Dijke’s (2011) finding an inverted U-shaped relationship between employees’ levels of workload and their experience of opportunities for workplace learning. Collectively, our finding coupled with these others suggests that, indeed, the time may have come for management scholars to revisit many presumed linear relationships to question when too much of a good thing may not be good (see Grant & Schwartz, 2011, and/or Pierce & Aguinis, 2013, for an elaboration of this view). Such reflection seems especially critical for businesses whose employees are highly culturally diverse, given the tendency for employees’ preferences for various managerial practices, including the presumed desirability of having voice, or input into managers’ decision making, to be culturally guided (cf. Brockner et al., 2001).
Conclusion 2: Balancing (Moderate) Organizational Support May Matter More for Employees Who More Strongly Anticipate Costs for Taking Charge
Our second conclusion is that the need for organizations and managers to provide moderate levels of support to employees (as suggested above), thereby avoiding support that will be perceived as either too low or too high, may be more true for employees who fear that taking charge will cost them valuable rewards (e.g., result in their being replaced, not receiving deserved recognition or bonus pay, and/or losing opportunities for advancement). Although no other study has tested this as we have, this pattern is indirectly supported by findings by justice scholars documenting that how much fairness employees perceive is of greater importance to them when they have more rather than less uncertainty about the stability of work structures (Lind & van den Bos, 2002; van den Bos, 2001; van den Bos & Lind, 2002).
Practical implications of the moderated inverted U-shaped pattern observed in our study are as follows. First, there is need for managers to pay particular attention to ambiguities and insecurities in employees’ workplace since doing so can help managers more accurately predict when employees will be extra-attentive to how much support they are (or are not) receiving as well as how much fairness they are (or are not) receiving. Second, there may be many different types of support that employees perceive in their work environment. For example, organizational support (as captured by POS) may be one type of support. As another example, positive responsiveness to employees’ taking charge suggestions may be another type of support. Being mindful of these different types of support perceptions carries practical importance since our study found these two have interactive effects on employees’ level of taking charge. This reinforces researchers’ call for more research to be done on dynamics relating to listening in addition to voicing (Morrison & Milliken, 2000), or similarly, on dynamics relating to showing employees’ consideration when they speak up (cf. Barry & Shapiro, 2000; Victor, Trevino, & Shapiro, 1993). Theorizing and research regarding the importance of providing voicers’ consideration in order for voice to enhance perceptions of organizational justice (called the “voice-effect”; cf. Folger, 1977) may help to inform what the behaviors may be that reduce or eliminate voicers’ concerns about how recipients will respond to them. Since raising employees’ level of anticipated costs and emotions relating to this is contrary to views expressed in most social exchange theories, including POS, the practical implication we have been discussing here also has important theoretical implications—namely, the need to better understand, hence to study, the circumstances in which employees’ anticipation of taking-charge-related costs help desired organizational and employee outcomes.
Limitations and Needed Future Research
As with all studies, ours is not without some limitations. First, the correlational design of our study limits our ability to infer causality amongst our model’s variables. On the other hand, the pattern of our findings supports, in part, previous theorizing and findings from studies using different methodologies associated with employees’ tendency to be more helpful and to engage in more behaviors beneficial for the organization when they perceive higher levels of organizational support (e.g., Lynch et al., 1999; Rhoades & Eisenberger, 2002; Riggle et al., 2009) and supports, in part, previous theorizing and findings regarding an inverted U-shaped pattern among variables in the workplace and the conditions of this pattern’s pronouncement reported elsewhere (e.g., Lambert, 2000; van Ruysseveldt & van Dijke, 2011. As such, our theorizing supports prior research associated with inverted U-shaped relationships (cf. Grant & Schwartz, 2011). Moreover, the fact that employees’ POS was assessed by them and employees’ taking charge was assessed, instead, by their supervisors helps to minimize the possibility of reverse causality. For example, why would supervisors’ perception of employees’ degree of taking charge, especially when our data also show that some anticipate this leading to negative evaluations, cause the employees to perceive higher levels of organizational support?
Nevertheless, future research is needed to examine the hypothesized relationships in this study via methods that enable causal relationships amongst study variables to be more clearly observed. One way to achieve this may be for future studies to obtain several assessments over time of employees’ perceptions of organizational support to draw causal conclusions about the effects of implementing programs aimed at increasing employees’ perceptions of organizational support. Likewise, measuring employees’ anticipated taking charge costs at different points in time allows for a deeper understanding of the causes and effects of employees’ worries about potentially suffering possible unwanted outcomes as a result of their taking charge.
A third limitation of our study is that this is the first time we are using a scale that explicitly measures the degree to which employees anticipate that there are costs associated with taking charge. This may be viewed, however, as a strength too, given that direct measures of anticipated costs associated with speaking up have yet to be taken despite the fact that employees’ anticipated costs has long been part of theorizing regarding antecedents to organizational silence (Morrison & Milliken, 2000) or, similarly, to avoidance of speaking up (Tangirala & Ramanujam, 2008) or issue selling (Dutton, Ashford, O’Neill, & Lawrence, 2001). Moreover, we engaged in a process of validating the anticipated taking-charge-related costs scale used in our study, the results of which suggest that it can be distinguished from employees’ dispositions, including neuroticism and fear of invalidity. Future research is needed to test the extent to which the anticipated taking-charge-related costs scale introduced in this study helps explain POS-related dynamics, as it did here, and has the potential to be extended to other similar relationships.
Equally important is to clarify mechanisms and processes whereby POS, or POS in conjunction with anticipated taking-charge-related costs, lead to variations in taking charge. What are employees thinking when they anticipate suffering penalties if they take charge? Is this only whether or not to speak up or does it extend to other change-oriented and proactive behaviors (Bindl & Parker, 2010; Chiaburu, Lorinkova, & Van Dyne, in press)? Or, might this be broader depending on how much support employees perceive they have received to date and/or on employees’ anticipation (guided by many potential intrapersonal, interpersonal, and/or normative cues) that the specific change or improvement suggestion they wish to make will likely receive a supportive response? Better understanding the variables that lead to anticipation about taking-charge-related costs has practical as well as theoretical importance. If, for example, the uncertainty about whether to take charge is due to intrapersonal factors (e.g., lack of self-efficacy, confidence, or resilience), interventions for minimizing uncertainty will likely need to be oriented toward enhancing employees’ self-esteem. If, as another example, the uncertainty about whether to take charge is due to interpersonal factors (e.g., laissez-faire leadership), interventions for minimizing uncertainty will likely need to be oriented toward changing leader behavior, including how to help leaders better demonstrate to employees that they value the employees as full-fledged members of the work community. If, as still another example, the uncertainty about possible taking-charge-related costs is due to employees’ perceiving their workplace norms as resistant to innovation or change, then interventions for minimizing uncertainty will likely need to be oriented at the organization level to change its culture. Considering all of these possible sources of employee uncertainty as potentially important in determining how much employees will be likely to be aware of how supportively or fairly treated they are, hence how proactive they will likely behave, promises to offer a multilevel approach to uncertainty management theory (Lind & van den Bos, 2002; van den Bos, 2001; van den Bos & Lind, 2002), which, until now, has been directed only at the individual level. In summary, identifying and measuring one or more of these potential process mechanisms can further understanding about when (various levels of) POS will encourage versus discourage employees from taking charge.
Conclusion
Pierce and Aguinis (2013: 314) proposed the “too-much-of-a-good thing” effect as a meta-theoretical principle and noted that “the TMGT effect suggests that management researchers should hypothesize and test the possibility that relatively high levels of otherwise beneficial antecedents may lead to unexpected and undesired outcomes.” From another direction, in a recent review of the proactive behaviors literature, Bindl and Parker (2010: 590, italics added) argued that “research is needed to identify under which circumstances situational influences may promote or inhibit proactive behaviors at work.” In this study, we clarify inconsistencies by proposing a more complex yet parsimonious perspective to predict taking charge. We do so by integrating organizational support theory (Eisenberger et al., 1986) with new arguments on the utility of an inverted U-shaped model (Grant & Schwartz, 2011; Pierce & Aguinis, 2013) and with existing suggestions contrasting (positive and negative) effects of situational factors (Frese & Fay, 2001; Parker, Bindl, & Strauss, 2010). We thus advance existing knowledge related to the role of organizational support in differentially influencing taking charge across levels of support by theorizing and finding support for a more complex inverted U-shaped relationship for the influence of support. Furthermore, this study explicates the contingency effect of anticipated costs of taking charge and provides a theoretical rationale for why the inverted U-shaped relationship will likely be stronger when employees are highly concerned about suffering negative consequences if they do take charge.
This study benefits managers because it gives them insight regarding the appropriate amount of organizational support that employees need to maximize their taking charge behaviors. By providing ideal levels of support (which our theorizing and findings suggest is moderate levels rather than lower or higher levels of POS), managers will be more likely to have employees who take charge and thereby help the organization obtain positive outcomes, such as stronger innovativeness and competitiveness (Nonaka, 1991). Hopefully, our theorizing and findings will encourage more management studies to occur through the lens of inverted U-shaped principles so that managers as well as management scholars can more deeply understand when positive organizational attributes are truly positive when present, and when potentially too much of their presence may produce unintended and unwanted consequences.
Footnotes
Appendix
The item development procedures assessing the validity of the anticipated taking-charge-related costs construct (described above) consisted of the following three stages. In the first stage, we conducted a qualitative study aimed at establishing with more precision the anticipated taking-charge-related costs content in terms of specific individual behaviors. We administered an open-ended computer-based questionnaire to a sample of 74 undergraduate students (42 males, 32 females) enrolled at a large state university in the United States. Respondents, who on average reported 19 months of work experience, were first informed that employees at work engage in a variety of behaviors, including taking charge. To make sure that our respondents, who were part of a broader subject pool, understood the nature of these behaviors, we listed as behavioral examples all 10 taking charge items from Morrison and Phelps (1999; e.g., “adopt improved procedures for doing the job”). Participants were then asked to think of at least five issues they typically consider before engaging in the taking charge behaviors previously exemplified. Respondents were asked to write their concerns through open responses, in the space provided. In total, respondents provided 239 statements. We reduced the list to 42 statements by eliminating redundancies and statements that reflected their concerns about supervisors. The content of the remaining 42 open responses were coded by two coders (93% agreement using Anderson and Gerbing’s, 1991, percentage of substantive agreement formula). The coding process resulted in general statements reflecting seven categories of anticipated costs, leading to the generation of the specific items presented above (e.g., “I would be replaceable,” “I would lose recognition associated with being a top performer”).
In the second stage, we conducted a quantitative study aimed at establishing the discriminant validity of our scale and whether our construct is distinct from other theoretically related constructs (i.e., neuroticism and fear of invalidity). We provided a sample of 111 undergraduate students (61 male, 50 female) with 18 months of work experience a questionnaire containing our anticipated cost items, as well as these other constructs. To assess employees’ anticipated taking-charge-related costs, this sample’s participants were instructed to indicate (via a 7-point Likert-type scale ranging from 1 = very unlikely to 7 = very likely) the likelihood with which each of the seven consequences named in the anticipated costs scale (listed above) would affect their making improvement suggestions in their organization. An exploratory factor analysis (EFA) with varimax rotation (Fabrigar, Wegener, MacCallum, & Strahan, 1999) confirmed the scale was unidimensional, with 64.90% of variance explained by one factor and factor loadings between .67 and .88.
Furthermore, we also determined the extent to which employees’ anticipated taking-charge-related costs is separate from theoretically relevant constructs such as neuroticism and fear of invalidity by asking participants to provide responses to a number of items measuring the latter constructs. Specifically, we expected anticipated costs to be distinct from neuroticism (4 items; Donnellan, Oswald, Baird, & Lucas, 2006; “get upset easily”; α = .73) and from fear of invalidity (6 items; Thompson, Naccarato, Parker, & Moskowitz, 2001; “I wish I would not worry so much about making errors”; α = .81). As expected, anticipated costs were uncorrelated with neuroticism (r = .13, p > .05) and fear of invalidity (r = .09, p > .05). Based on the constitutive definition for anticipated costs, presented above, there are also no conceptual connections with uncertainty, suggesting a distinction between the two constructs. Furthermore, from an item content (face validity) standpoint, the items describing anticipated costs (e.g., “I would be overlooked as a candidate for promotion”) reflect no uncertainty. Rather, they are about the respondents being apprehensive—in various ways—concerning what can happen to them if they engage in a specific behavior. Finally, from a discriminant validity standpoint, we provide evidence that fear of invalidity, a construct that is likely to covary with uncertainty, is not correlated with our anticipated costs construct. Specifically, as Thompson and coauthors (2001: 29-30) noted, “individuals possessing higher levels of personal fear of invalidity (PFI) expressed greater ambivalence across a greater variety of social issues.” Ambivalence connotes being simultaneously attracted and repelled by an object or attitude, hence being conceptually associated with uncertainty. Yet our data did not support a correlation between anticipated costs and fear of invalidity, which suggests, by extension, an empirical distinction between anticipated costs and uncertainty.
In the third stage, we conducted a quantitative study in a sample of 135 Jamaican working adults (16 female, 119 male) to further establish the validity of our anticipated taking-charge-related costs scale. To assess anticipated costs, participants in this sample were given a questionnaire containing our anticipated taking-charge-related costs items and asked to think about the outcomes that are likely to occur if ideas for making changes to improve work outcomes are shared. They were then asked to indicate (via a 7-point Likert-type scale ranging from 1 = very unlikely to 7 = very likely) the likelihood that each of the seven consequences named in the anticipated taking-charge-related costs scale would affect their making improvement suggestions. An EFA with varimax rotation (Fabrigar et al., 1999) confirmed the scale as unidimensional, with 63.56% of variance explained by one factor and factor loadings between .71 and .84. The estimated reliability of our seven-item scale was .90. Based on the psychometric properties and on its distinctiveness, we used the seven-item anticipated taking-charge-related costs scale (listed above) in our field sample.
Acknowledgements
This article was accepted under the editorship of Deborah E. Rupp. The authors would like to acknowledge Elizabeth W. Morrison, In-Sue Oh, Paulette Johnson, and the two anonymous JOM reviewers for their guidance on this article.