Loyal rebels? A test of the normative conflict model of constructive deviance
Summary
Constructive deviance is a voluntary behavior that violates organizational rules but is conducted with honorable intentions to benefit the organization or its stakeholders. Despite emerging interest in this behavior, the antecedents of constructive deviance remain unclear, with particular ambiguity concerning the relationship between organizational identity and constructive deviance. In this article, we address this ambiguity with the normative conflict model, which posits that organizational identity drives constructive deviance in the workplace only when people perceive normative conflict with organizational rules. In Studies 1a and 1b, we develop and validate a measure of normative conflict. In Study 2, we conduct a preliminary test of the model with employed students and find that identity is positively related to constructive deviance only when normative conflict is high. In Study 3, we replicate and extend the model to show that the moderating effect of normative conflict is mediated by experienced psychological discomfort and that organizational identity is positively related to constructive deviance among working adults only when discomfort is high. In total, our findings demonstrate the utility of the normative conflict model for explaining when constructive deviance is mostly likely to occur in the workplace.
1 INTRODUCTION
Employees sometimes challenge and break workplace rules not because they are disloyal but because they are passionate enough to dissent against practices that they see as stagnant, ineffective, or even dangerous to those around them. This behavior is referred to as constructive deviance, which involves voluntary actions that violate organizational norms, but are conducted with honorable intentions to benefit the organization and its stakeholders (e.g., Mainemelis, 2010; Morrison, 2006; Spreitzer & Sonenshein, 2003; Vardi & Weitz, 2004; Warren, 2003). Historically, organizational researchers focused only on understanding destructive deviance, which is conducted with the goal of harming the organization or benefitting oneself (e.g., Robinson & Bennett, 1995). However, researchers have more recently started to study constructive deviance because it has the potential to trigger beneficial change, increase productivity, and promote innovation in the workplace.
Despite the value of understanding and responsibly channeling constructive deviance in groups (Jetten & Hornsey, 2014), we know surprisingly little about its antecedents in the workplace, and a variety of contradictory findings have recently emerged in the literature (Vadera, Pratt, & Mishra, 2013). One of the most interesting ambiguities concerns the relationship between organizational identity and constructive deviance. As Vadera et al. (2013) observed, some studies document that identity is positively related to constructive deviance (Mellahi, Budhwar, & Li, 2010; Olkkonen & Lipponen, 2006). Other studies, however, have found a negative or null relationship between organizational identity and constructive deviance (Burris, Detert, & Chiaburu, 2008; Sims & Keenan, 1998; J. Zhou & George, 2001). Further, it is still theoretically unclear why and when highly identified employees would engage in constructive deviance instead of loyal conformity. If employees care strongly about the organization, why would they willingly break rules instead of channeling their effort into the best, rule-abiding performance they can provide?
To clarify the relationship between organizational identity and constructive deviance, we draw on a recent theory of group dissent called the normative conflict model (Packer, 2008; Packer, 2011). This model explains that the effects of organizational identity on deviance depend on the extent to which group members perceive normative conflict, which occurs when members perceive a discrepancy between the group's actual, practiced norms and some better, alternative standard for behavior. The model posits that identity predicts conformity when normative conflict is low but that identity predicts dissent instead when normative conflict is high. This interaction occurs because strongly identified group members who feel normative conflict exhibit dissent intended to help the group improve. Consequently, the model may offer insight into constructive deviance: We predict that organizational identity is only predictive of constructive deviance when people also feel high normative conflict.
Our research on the normative conflict model makes several contributions. First and foremost, we advance research on constructive deviance by resolving an emergent discrepancy in the literature concerning its relationship with social identity (Vadera et al., 2013). This is an important theoretical contribution because we know little about why constructive deviance occurs, and the contradictory findings concerning the relationship between organizational identity and constructive deviance points to a need to understand the boundary conditions that change the nature of this relationship (Busse, Kach, & Wagner, 2016).
Second, we introduce the normative conflict model to the organizational behavior literature, which is a promising framework for understanding identity-driven behavior in the workplace. Research on the normative conflict model to date is focused on informal or opinion-based social groups (e.g., Packer & Chasteen, 2010; Täuber & Sassenberg, 2012), but we show that the model has utility for predicting behavior in the workplace, where the stakes for engaging in deviant behavior are higher. Managers in the workplace heavily weigh deviant behavior when judging performance and making personnel decisions (Rotundo & Sackett, 2002), and employees may face sanctions or lose their jobs for deviating in ways that are suppressed by the organization (Berry, Ones, & Sackett, 2007). The steeper consequences of constructive deviance at work accordingly allow for a strong test of the theory in a formal setting where rules are more rigidly enforced and dissent is met with greater hostility (X. Zhou, 1993).
Last, we extend the normative conflict model in this research by elaborating on the psychological experience of normative conflict. One limitation of the normative conflict model is relatively little consideration of why normative conflict motivates action; implicit in the normative conflict model is the idea that normative conflict is a distressing experience (Packer, 2008). Consequently, we integrate the normative conflict model with cognitive dissonance theory (Festinger, 1957), which offers additional insight into why people experience psychological distress in the presence of normative conflict and subsequently act on it. We therefore make an important theoretical extension to the normative conflict model to better explain why normative conflict can trigger deviant behavior.
1.1 Constructive deviance in the workplace
When considering rule-breaking behavior, organizations mainly focus on destructive deviance, which is voluntary behavior that knowingly deviates from organizational norms in ways that threaten the organization or its stakeholders (Robinson & Bennett, 1995). Destructive deviance is threatening because it is committed with the intention to either harm others or benefit the self (Vardi & Weitz, 2004). For example, behaviors such as merchandise theft conducted in the interest of personal gain, or verbal aggression intended to hurt a coworker's feelings, would be categorized as instances of destructive deviance.
Constructive deviance similarly involves willful deviation from organizational norms, but the deviation is intended to benefit the organization or its stakeholders (Morrison, 2006; Spreitzer & Sonenshein, 2004; Warren, 2003). Because the standards for judging deviance can be highly subjective, Warren (2003) submitted that deviance is only constructive if employees reject organizational norms in favor of hypernorms (Donaldson & Dunfee, 1994), which are defined as “globally held beliefs and values” that encompass basic principles and satisfy fundamental human needs (Warren, 2003, p. 628). Because hypernorms typically stem from the values and practices of large, inclusive social groups and societies, they serve as a “higher authority” that reflect the needs and priorities of people both inside and outside of the organization. Relevant hypernorms that are used to judge behavior may vary across different nations, industries, organizations, and even roles. However, Warren (2003) suggested that the standards and ethical codes of relevant, large, and inclusive organizations can serve as a likely source of hypernorms. For example, the American Psychological Association's Ethical Code and Code of Conduct articulate hypernorms about protecting the privacy of clients. A psychologist could opt to constructively deviate to follow these hypernorms if she felt that her organization or workgroup's norms failed to adequately protect the confidential information divulged by clients. Thus, deviance that rejects local norms and practices, but complies with hypernorms, is constructive because it serves a greater good.
Consequently, constructive deviance is formally defined as behavior that (a) benefits the reference group, (b) deviates from reference group norms, and (c) conforms to broader hypernorms (Vadera et al., 2013). A specific example of the distinction between destructive and constructive deviance can be seen in a typical customer service context. If a regular customer of a restaurant presents an expired gift certificate, an employee can decide to follow procedure and refuse it, or break rules and accept it. The decision to break rules can occur due to different motivations. In this example, the rule-breaking behavior would be considered destructive deviance if the employee opts to accept the gift certificate because it benefits himself (e.g., through a larger tip) or because he hopes it will hurt the organization to lose money. On the other hand, the rule-breaking behavior would be considered constructive deviance if the employee opts to accept the gift certificate to help the customer and organization (e.g., this customer is important and would take his business elsewhere if refused, costing the organization more in the long run). In the latter case, the employee is still willingly deviating from an organizational rule, but his choice is made with the express aim of helping the company and complying with broader societal hypernorms concerning kindness that should be shown to others. As Vadera et al. (2013) observed, many organizational behaviors can be constructively deviant under certain circumstances, such as whistleblowing (Sims & Keenan, 1998) and expressing challenging voice (Burris et al., 2008), provided that they are conducted with the intention to help the organization or its stakeholders and comply with a broader hypernorm instead.
Constructive deviance can have important consequences in workgroups and the overall organization. Fundamentally, constructive deviance is valuable because it provides the group with alternative norms that can serve as the catalyst for change (Jetten & Hornsey, 2014; Prislin & Filson, 2009); challenging the existing norms that characterize the status quo can introduce much needed instability that unlocks the potential for positive, adaptive growth (Gioia, Schultz, & Corley, 2000). Consequently, constructive deviance may result in better performance and retention within groups (Vadera et al., 2013). Constructive deviance is also associated with desirable individual states, such as role breadth self-efficacy (Galperin, 2012). However, the consequences of constructive deviance are complex and likely dependent on organizational context. For example, constructive deviance that benefits one work group may cause complications for other, interdependent work groups (Vadera et al., 2013). Also, some instances of constructive deviance that are well intentioned can nevertheless lead to negative outcomes if rule breakers do not recognize the importance of violated norms (Dahling, Chau, Mayer, & Gregory, 2012). Although the outcomes of constructive deviance are consequently in need of further study (Jetten & Hornsey, 2014; Vadera et al., 2013), it is clear that constructive deviance can have powerful effects in organizational settings that warrant more research attention.
Several studies have identified antecedents of constructive deviance (e.g., Dahling et al., 2012; Galperin, 2012; Morrison, 2006). Vadera et al. (2013) recently published the first comprehensive review of this literature, in which they organized the predictors of constructive deviance into three conceptual categories: intrinsic motivation, felt obligation, and psychological empowerment. They also highlighted several emergent contradictions in the literature that require clarification, most notably the effect that organizational identity has on constructive deviance. As we explain in the following section, identity can theoretically exhibit a positive or negative relationship with constructive deviance. For example, organizational identity is positively related to behaviors that benefit the organization, which could include constructive deviance (Vadera, Aguilera, & Caza, 2009). On the other hand, organizational identity is also positively related to adherence to organizational norms and values (Pratt, 2000). Consistent with these contradictory arguments, the empirical literature on identity and constructive deviance is split: Some studies have found a positive relationship (e.g., Mellahi et al., 2010; Olkkonen & Lipponen, 2006; Taylor & Curtis, 2010), others have found a negative relationship (e.g., J. Zhou & George, 2001), and others have found no relationship at all (Burris et al., 2008; Sims & Keenan, 1998). What might predict, then, when organizational identity will lead an employee to adhere to norms versus break them? The normative conflict model introduces a key moderator that sheds light on this relationship.
1.2 The normative conflict model of group dissent
The normative conflict model (Packer, 2008) was developed to explain the interplay between group identity and the choice to either conform or dissent from group norms. The model builds on social identity theory (SIT; Ellemers & Haslam, 2012; Tajfel, 1974; Tajfel & Turner, 1979), which states that strongly identified group members usually conform to the group's salient norms and standards (e.g., Terry & Hogg, 1996). However, strong identifiers sometimes behave differently when they perceive that their group is threatened in some way that challenges its status or value (e.g., Crane & Platow, 2010; Dahling, Wiley, Fishman, & Loihle, 2016; Ellemers, Spears, & Doosje, 2002). The normative conflict model diverges from SIT by describing what happens when the group is specifically threatened from within by its own rules or norms that have become harmful, dangerous, or ineffective to the group (Packer, 2008). According to the model, a sense of normative conflict occurs when group members perceive a discrepancy between the group's actual norms and some better, alternative standard for behavior. For example, normative conflict might occur for a tenured professor if her university takes advantage of adjunct faculty who could instead be treated better, for a cook if her restaurant discards food that could alternatively be donated to a homeless shelter, and for a sales associate if store rules prevent her from serving customers as effectively as a competing business does.
People need to perceive a discrepancy between the actual norms practiced by the group and alternative norms to experience normative conflict, but where do these better, alternative norms come from? According to SIT, alternative norms arise from other aspects of personal or group identities that a person holds (Packer, 2008; van Veelen, Otten, Cadinu, & Hansen, 2016). For example, alternative norms can be derived from personal or moral values (e.g., Hornsey, Majkut, Terry, & McKimmie, 2003), leading to normative conflict as in the previous example of a cook who objects to wasting food when it could instead be used to help people in need. Alternative norms can also arise from the practices of other groups, much like benchmarking, as in the example of the sales associate who perceives that a competing business's norms offer a preferable standard of customer service quality (Packer, 2008). Similarly, Warren (2003) theorized that hypernorms are drawn from the values and practices of large, inclusive groups that reflect the values of many people. These groups might include humanitarian agencies, such as the World Health Organization, that seek to serve all people or smaller groups that specify the ethical practices of a particular profession, such as the Academy of Management. Thus, alternative norms can arise from a variety of personal and social sources, and the sources of the alternative norms that people perceive to be better than their actual norms will vary widely across contexts.
Regardless of the specific way in which normative conflict manifests, this discrepancy between the group's actual, practiced norms and alternative, better norms yields a sense of tension that people are motivated to resolve. To this end, the normative conflict model explains how the effect of group identification on behavior depends on normative conflict. When normative conflict is high, identity is positively related to “dissent,” which Packer (2008) defined as “nonconformist reactions motivated by a desire to change group norms and initiate improvement within the group” (p. 54). However, identity is positively related to conformity to group norms when normative conflict is low because the norms are judged as appropriate and acceptable. The model also specifies that identity is negatively related to behavioral or psychological disengagement when normative conflict is high. That is, people with low levels of group identity respond to normative conflict by simply disengaging from the group (Packer, 2008; Packer, 2011). An abundance of evidence from SIT supports this proposition, demonstrating that weak identifiers tend to respond to group threats by distancing themselves from the group, to the extent that they are able to do so (Ellemers et al., 2002). Thus, dissent should only occur for high identifiers, but only when they also feel high normative conflict.
To date, empirical research on the normative conflict model has tested the model in informal or opinion-based social groups (e.g., online chat groups; Packer & Chasteen, 2010; Täuber & Sassenberg, 2012). For example, Packer and Chasteen conducted a set of studies testing the model in informal groups of college students. One of their studies assessed college students' levels of identification with their university and asked them to think about possible negative consequences of a pro-alcohol norm at their school. In support of the model, students who were the most highly identified with the university were most willing to dissent and voice their concerns about binge drinking to the group. Similarly, Täuber and Sassenberg (2012) tested the normative conflict model among players in a university recreational soccer league. They hypothesized that players would experience normative conflict when their teams set unambitious performance goals that contrasted with alternative, more desirable goals that the group might set instead. Consistent with the normative conflict model, they found that team identity was positively related to dissent and the articulation of alternative goals among players whose teams initially set unambitious goals. More recently, Dupuis, Wohl, Packer, and Tabri (2016) tested the model with respect to political action, finding that strongly identifying Republicans were more willing to express dissent to other Republicans in favor of the Affordable Care Act if they also felt high normative conflict and group angst. These findings provide important support for the tenets of the normative conflict model. However, to date, the model has not been applied in business organizations to predict workplace behavior, where the stakes for dissent are higher.
1.3 Integrating cognitive dissonance theory
People experience normative conflict when they detect a discrepancy between existing organizational norms and some alternative, better norms. To extend the model, we believe that the reason that normative conflict strengthens the relationship between identity and constructive deviance is that normative conflict is a psychologically uncomfortable experience and that the urge to resolve this discomfort is what generates the motivation among high identifiers to engage in constructive deviance. Packer (2008) hinted at this idea, noting that people may be “bothered by” normative conflict in ways that motivate their actions (p. 65).
In this respect, we see a possible connection between the normative conflict model and cognitive dissonance theory, which clarifies that holding conflicting or inconsistent thoughts, such as abiding by actual norms that are less desirable than alternative norms, is psychologically uncomfortable. One way that people can reduce this tension is to change their behavior, such as by taking actions that reduce the discrepancy between those inconsistent thoughts (e.g., Festinger & Carlsmith, 1959). Empirical research strongly supports this proposition and demonstrates that feelings of tension and discomfort mediate the effects of cognitive dissonance on behavioral intentions (e.g., Elliot & Devine, 1994; Glasford, Pratto, & Dovidio, 2008). Similarly, we submit that the moderating effect of normative conflict on the relationship between organizational identity and constructive deviance can be explained by feelings of psychological discomfort. Discomfort should therefore mediate the moderating effects of normative conflict.
1.4 The present research
The normative conflict model provides a key to understanding the ambiguous relationship between organizational identity and constructive deviance observed in past research (Vadera et al., 2013): The relationship between organizational identity and constructive deviance should depend on perceptions of normative conflict with organizational rules. When normative conflict is high, identity should be positively related to constructive deviance because these employees feel a need to break rules for the perceived good of the organization. However, when normative conflict is low, identity should be negatively related to constructive deviance because strong identifiers are otherwise prone to conform to group norms (Ellemers & Haslam, 2012; Terry & Hogg, 1996). The moderating role of normative conflict should consequently explain why some studies have found a positive relationship between identity and constructive deviance (reflecting dissent), whereas others have not (reflecting conformity). Further, drawing on cognitive dissonance theory (Festinger, 1957), we expect that the reason normative conflict moderates the relationship between organizational identity and constructive deviance is that normative conflict generates dissonance-based psychological discomfort that people are motivated to resolve through behavioral change. In aggregate, these expectations yield a form of mediated moderation (Edwards & Lambert, 2007; Muller, Judd, & Yzerbyt, 2005) wherein the moderating effect of normative conflict on the relationship between organizational identity and constructive deviance occurs indirectly via psychological discomfort (cf. Grant & Berry, 2011). Figure 1 illustrates this model and our hypotheses:
Hypothesis 1.Normative conflict will moderate the relationship between organizational identification and constructive deviance. Specifically, this relationship will be significant and positive only for people with high (vs. low) perceptions of normative conflict.
Hypothesis 2a.Normative conflict will be positively related to psychological discomfort.
Hypothesis 2b.Psychological discomfort will moderate the relationship between organizational identification and constructive deviance. Specifically, this relationship will be significant and positive only for people with high (vs. low) psychological discomfort.
Hypothesis 2c.Psychological discomfort mediates the direct moderating effect of normative conflict on the relationship between organizational identification and constructive deviance.
We conducted four studies to explore these ideas. Studies 1a and 1b briefly describe the development and validation of a measure of normative conflict with organizational rules, the key moderating variable posited by the normative conflict model. Study 2 presents a preliminary test of Hypothesis , documenting the direct moderation of normative conflict on the relationship between organizational identity and constructive deviance among a sample of employed students. Study 3 tests the full model shown in Figure 1; we replicate Hypothesis and test Hypotheses 2a–2c with a predictive design and a large sample of experienced adult workers.
2 STUDY 1A: DEVELOPING THE MEASURE OF NORMATIVE CONFLICT WITH ORGANIZATIONAL RULES
Few studies have empirically tested the normative conflict model. In this research, normative conflict is operationalized in idiosyncratic ways (e.g., setting an unambitious soccer performance goal; Täuber & Sassenberg, 2012). Consequently, we needed to develop our own measure of normative conflict that could be generalized to different organizational contexts. Following best practices (Hinkin, 1998; Spector, 1992), we developed items and explored the factor structure of the measure with a preliminary sample (Study 1a) and then confirmed the factor structure and assessed validity evidence with a second sample (Study 1b).
2.1 Method
2.1.1 Sample and procedure
We initially wrote a pool of 10 items to assess normative conflict, drawing on explanations and examples provided by Packer (2008, 2011) in his description of the normative conflict model. We started with a fairly small item pool because perceived normative conflict with rules is a narrowly defined construct, and we believed that we could fully capture the construct domain space with a short measure (Spector, 1992). Three trained raters independently evaluated the item pool relative to the construct definition reported by Packer (2008); all 10 items were evaluated as highly aligned with the construct definition and were retained for subsequent analysis.
Through exploratory factor analysis (EFA; Thompson, 2004), our goal was to further reduce the item pool to yield a parsimonious scale that adequately represented the construct with high reliability and a simple factor structure (Spector, 1992). We collected data from 469 employed participants from the United States via Amazon Mechanical Turk (MTurk; http://www.mturk.com). MTurk matches “employers” with “workers” to complete tasks, such as research participation. Careful comparisons indicate that MTurk yields data that matches, and sometimes exceeds, the quality of data collected from in-person samples (e.g., Casler, Bickel, & Hackett, 2013; Paolacci & Chandler, 2014), provided that checks are put in place to screen respondents. Consistent with best practices, we tracked IP and MTurk identification numbers to prevent any repeat participants, screened participants on their self-rated comfort with the English language, and only accepted participants with a past-work approval rating of 95% or better (Peer et al., 2014). The sample was 52% female and 83.4% Caucasian, and participants reported a mean age of 33.63 years (SD = 10.90). Most participants worked in full-time or salaried jobs (69.7%).
2.2 Results
We first conducted a parallel analysis (PA) on the data to identify the maximum number of factors to extract (Velicer, Eaton, & Fava, 2000). PA generates random data sets with the same parameters as the actual data (in this case, N = 1,000), extracts the eigenvalues from the random data correlation matrix (in this case, with principal axis factoring), and averages those eigenvalues. The actual eigenvalues from the true data are compared to the average eigenvalues extracted from the parallel random data, and each factor is deemed interpretable only if its actual eigenvalue is larger than its average random eigenvalue. Results of this analysis indicated that no more than two factors should be extracted from the data.
We then conducted an EFA using principal axis factoring with a varimax rotation to examine the factor structure of the item pool (Thompson, 2004). The results of this analysis are shown in Table 1. All items loaded on a single factor, and thus the factor loadings reported in Table 1 are unrotated. This single factor explained 65.24% of the variability in the items. However, an analysis of the Scree plot (Velicer et al., 2000) indicated that a second factor was marginally interpretable (initial Eigenvalue = 0.99; 9.95% of variance explained). This factor seemed to represent a more extreme perception of normative conflict and was anchored by the two items with the weakest loadings in Table 1 (“This organization has rules or norms that are dangerous to follow” and “This organization has rules or norms that encourage immoral or unethical behavior”). We consequently eliminated these two items and concluded with a unidimensional, eight-item measure (α = .95) that we employed in our subsequent studies.
| Item | Factor loading |
|---|---|
| 1. I think this organization falls short of what it could be because of the rules and norms it enforces on employees. | .90 |
| 2. This organization could be so much better if it followed different rules or norms. | .87 |
| 3. I think this organization will never reach its true potential until it changes its practices. | .85 |
| 4. The standards of this organization encourage the wrong sort of behavior from employees. | .83 |
| 5. This organization has rules or norms that lead to wasteful or counterproductive behavior. | .81 |
| 6. This organization could be much more efficient if people could follow different rules or norms. | .79 |
| 7. The values of this organization are not accurately reflected in the rules and norms it sets. | .78 |
| 8. I think that the rules and norms of this organization are valid and reasonable. (R) | .75 |
| 9. This organization has rules or norms that encourage immoral or unethical behavior. | .64 |
| 10. This organization has rules or norms that are dangerous to follow. | .58 |
- Note. Items presented in italics were removed from the final measure. (R) = reverse-scored item.
3 STUDY 1B: VALIDATING THE MEASURE OF NORMATIVE CONFLICT WITH ORGANIZATIONAL RULES
Our goals in Study 1b were to confirm the factor structure of the normative conflict measure and to validate it relative to related psychological states. The specific psychological states that we focused on were identity fusion, disidentification, affective commitment, and normative commitment. Identity fusion refers to relational connectedness with group members that leads to a feeling of “oneness” with other members of the group (Gómez et al., 2011). In contrast, disidentification occurs when people expressly reject the qualities or values that they believe characterize the group (Kreiner & Ashforth, 2004). Affective commitment refers to an emotional attachment and involvement with a group, whereas normative commitment refers to a sense of felt obligation and responsibility to a group (Meyer & Allen, 1997). We chose these four variables for our validation design because, like normative conflict, they all involve an acceptance or rejection of salient group norms. We expected that normative conflict perceptions would be negatively correlated with states that involve acceptance of norms (identity fusion, affective commitment, and normative commitment), and positively correlated with states that involve rejection of norms (disidentification).
3.1 Method
3.1.1 Sample and procedure
We recruited 288 employed adults from the United States via MTurk. Participants from Study 1a were excluded from Study 1b with an unobtrusive screening process for Qualtrics developed by Peer, Paolacci, Chandler, and Mueller (2012) on the basis of individual MTurk identification numbers. For eligible participants, we followed the same best practices for data screening described in Study 1a. The sample was 54% female and 80.2% Caucasian, and participants reported a mean age of 33.90 years (SD = 11.68). Most participants worked in full-time or salaried jobs (63.9%). The most common industries reported were education (13.5%), retail sales (11.8%), and health care (11.1%). Measures of normative conflict and the correlate variables were presented in a fully randomized order.
3.1.2 Measures
We measured normative conflict using the eight items developed in Study 1a (α = .91). Responses were made on a 5-point scale where 1 = “Strongly Disagree” and 5 = “Strongly Agree.”
For identity fusion, we used the seven-item measure (α = .93) developed and validated by Gómez et al. (2011). A sample item reads, “I am one with my organization.” Responses were made on a seven-point scale where 1 = “Strongly Disagree” and 7 = “Strongly Agree.”
For disidentification, we used the six-item measure (α = .93) developed by Kreiner and Ashforth (2004). A sample item reads, “I am embarrassed to be part of this organization.” Responses were made on a 7-point scale where 1 = “Never” and 7 = “Daily.”
Affective and normative commitment were measured using the revised commitment scales developed by Meyer and Allen (1997). A sample item from the six-item (α = .91) affective commitment measure reads, “I would be very happy to spend the rest of my career with this organization.” A sample item from the six-item (α = .92) normative commitment measure reads, “This organization deserves my loyalty.” Responses to both measures were made on a seven-point scale where 1 = “Strongly Disagree” and 7 = “Strongly Agree.”
3.2 Results
Table 2 reports descriptive statistics and correlations for Study 1b variables. As expected, the normative conflict measure correlated negatively with identity fusion, affective commitment, and normative commitment, and positively with disidentification.
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 |
|---|---|---|---|---|---|---|---|---|---|---|
| 1. Gender | -- | -- | -- | |||||||
| 2. Age | 33.90 | 11.68 | .08 | -- | ||||||
| 3. Tenure (years) | 3.84 | 4.37 | .08 | .42*** | -- | |||||
| 4. Normative conflict | 3.89 | 0.98 | −.02 | .06 | .09 | (.91) | ||||
| 5. Identity fusion | 3.73 | 1.49 | −.06 | −.01 | .12* | −.43*** | (.93) | |||
| 6. Disidentification | 1.91 | 0.99 | −.08 | −.02 | .02 | .67*** | −.54*** | (.93) | ||
| 7. Affective commitment | 3.97 | 1.62 | .02 | .00 | .18** | −.45*** | .80*** | −.60*** | (.91) | |
| 8. Normative commitment | 3.71 | 1.59 | .06 | −.04 | .08 | −.46*** | .70*** | −.53*** | .80*** | (.92) |
- * p < .05.
- ** p < .01.
- *** p < .001
We conducted a series of confirmatory factor analyses (CFA; Brown, 2015) in MPlus version 7 (Muthén & Muthén, 1998–2012) to evaluate the normative conflict measure. All models were tested without allowing any residuals to covary or forming any item parcels. First, we examined the hypothesized, single-factor structure of the normative conflict measure in isolation. This model exhibited acceptable fit to the data, χ2(20) = 60.92, p < .001; CFI = 0.97; RMSEA = 0.08; SRMR = 0.03, with all eight items loading strongly on the single factor (all p < .001).
Second, given the strong correlation that we observed between normative conflict and disidentification (r = .67), we also used CFA to demonstrate the discriminant validity of the normative conflict measure by comparing the fit of a hypothesized two-factor model to an alternative one-factor model (Shaffer, DeGeest, & Li, 2015). Results indicate that the hypothesized two-factor model, χ2(76) = 245.37, p < .001; CFI = 0.95; RMSEA = 0.09; SRMR = 0.05, fit the data better than an alternative, one-factor measurement model in which the items from both measures acted as indicators of the same latent variable, Δχ2(1) = 553.3, p < .001.
Last, we tested the overall five-factor model. This model had modest-to-poor fit to the data, χ2(485) = 1,364.89, p < .001; CFI = 0.89; RMSEA = 0.08; SRMR = 0.06. However, it is likely that the cause of these subpar fit indices are the strong correlations between the measures of affective commitment, identity fusion, and normative commitment (see Table 2). The structure of the normative conflict measure replicated from Study 1a and fits the data well, and the normative conflict measure is empirically distinct from the only correlate with which it exhibited a strong relationship (i.e., disidentification). Consequently, we do not see the results of the five-factor measurement model as problematic to our research program given that our focus is on the validity evidence for the normative conflict measure.
In summary, we replicated the factor structure of the normative conflict measure in Study 1b. The measure correlated with related psychological states as expected and is discriminant from disidentification despite the strong relationship between these variables. Reliability exceeded .90 in both Studies 1a and 1b. Given this initial construct validity evidence, we proceeded to a preliminary test of the normative conflict model in Study 2.
4 STUDY 2: PRELIMINARY TEST OF THE NORMATIVE CONFLICT MODEL
In Study 2, we used the newly developed normative conflict measure to test the core proposition of the normative conflict model concerning constructive deviance. Following Packer (2008, 2011), we expect that the relationship between organizational identity and constructive deviance will depend on normative conflict perceptions. This relationship should be significant and positive only when normative conflict is high (Hypothesis ).
4.1 Method
4.1.1 Sample and procedure
The participants were 102 working students who were recruited from Psychology courses at a small college in the mid-Atlantic region of the United States. The mean age of participants was 20.25 years (SD = 1.99). The sample was 75.5% female and 78.4% Caucasian. Participants reported a mean organizational tenure of 16.76 months (SD = 17.50) and a mean of 18.38 hr worked per week (SD = 10.14). Participants completed a short, computer-based survey containing measures of normative conflict, organizational identity, and constructive deviance in a single session under supervised conditions. We collected self-report data because identity and normative conflict are perceptual measures that only the focal respondent can indicate, and constructive deviance involves a prosocial motive that observers may not clearly distinguish from destructive deviance (Dahling et al., 2012). All measures were presented in a fully randomized sequence to account for order effects.
4.1.2 Measures
Normative conflict was measured with the eight-item scale developed in Study 1a (α = .89). Responses were made on a 5-point scale where 1 = “strongly disagree” and 5 = “strongly agree.” Higher scores indicate stronger feelings of normative conflict.
Organizational identity was measured with a 10-item scale from Leach et al. (2008) that assesses in-group identity in terms of solidarity, centrality, and satisfaction with membership in a social group (α = .94). Sample items read, “I often think about the fact that I am a member of this organization” and “I am glad to be a member of this organization.” Responses were made on a 7-point scale where 1 = “strongly disagree” and 7 = “strongly agree.” Higher scores indicate stronger identification with the organization.
Few measures of constructive deviance exist in the literature. Although many organizational behaviors, such as issue selling, voice, and extra-role behavior, can be constructively deviant under certain circumstances, in many instances those behaviors are performed in nondeviant ways (e.g., if norms are not intentionally violated or if the motivation is not constructive; Vadera et al., 2013). Most available measures do not adequately distinguish between deviant and nondeviant instances of these behaviors, which severely limits the options for empirical studies of constructive deviance. In this study, we selected a five-item measure of organizationally directed constructive deviance (α = .89) developed by Galperin (2012). To our knowledge, this is one of only two generalizable measures in the literature that are explicitly intended to measure constructive deviance (Dahling et al., 2012). The items making up the scale were, “Sought to bend or break the rules in order to perform your job,” “Violated company procedures in order to solve a problem,” “Departed from organizational procedures to solve a customer's problem,” “Bent a rule to satisfy a customer's needs,” and “Departed from dysfunctional organizational policies or procedures to solve a problem.” Responses were made in terms of frequency on a 5-point Likert scale ranging from 1 = “never” to 5 = “weekly.” Higher scores indicate greater frequency of constructive deviance. Galperin (2012) developed and validated the measure over three studies relative to other organizational behaviors (e.g., exit, neglect, innovation) and individual differences (e.g., perceived sociopolitical support).
4.2 Results
4.2.1 Confirmatory factor analysis
We tested our measurement model using CFA in MPlus version 7 (Muthén & Muthén, 1998–2012). For the indicators of the organizational identity latent variable, we formed three parcels on the basis of the subscales of this measure, which is a recommended step given that these item sets should share secondary covariance that would otherwise distort the results of the CFA (Hall, Snell, & Foust, 1999). Following best practices for parceling (Williams & O'Boyle, 2008), we evaluated the reliability of each parcel. Specifically, the three parcels consisted of the means of the three solidarity items (α = .91), the three identity centrality items (α = .88), and the four identity satisfaction items (α = .92). Because the constructive deviance and normative conflict measures are not multidimensional, we used the individual items as the indicators for their respective latent variables.
The hypothesized, three-factor model exhibited acceptable fit to the data given the small sample size, χ2(101) = 172.75, p < .001; CFI = 0.93; RMSEA = 0.08; SRMR = 0.07, with all items loading strongly on their intended constructs (all p < .001). No error terms were allowed to covary. Because we observed that scale scores on the normative conflict and organizational identity measures exhibited a moderate correlation (r = −.41, p < .001), we tested an alternative, two-factor measurement model that placed the items from both of these measures as indicators of a single latent variable. Results indicated that this alternative model fit the data significantly worse than the hypothesized measurement model, Δχ2(2) = 128.15, p < .001. Consequently, we retained the three-factor model and moved forward to test Hypothesis .
4.2.2 Hypothesis test
Initial correlations indicated that organizational identity (M = 4.78, SD = 1.21) was unrelated to constructive deviance (M = 2.09, SD = 0.92; r = −.07, ns) and negatively related to normative conflict (M = 2.38, SD = 0.88; r = −.40, p < .001). Normative conflict was positively related to constructive deviance (r = .43, p < .001). We tested Hypothesis with moderated multiple regression conducted in PROCESS (Hayes, 2013). Consistent with the correlations, normative conflict had a significant relationship with constructive deviance (b = 0.51, SE = 0.10, t = 5.00, p < .001), but identity did not (b = 0.07, SE = 0.07, t = 0.99, ns). The interaction term was statistically significant (b = 0.18, SE = 0.07, t = 2.53, p < .05, ΔR2 = .05). We plotted the interaction in Figure 2 at high (+1 SD) and low (−1 SD) levels of organizational identity and normative conflict (Aiken & West, 1991). Figure 2 shows that the relationship between identity and constructive deviance is significant and positive only when normative conflict is high (t = 2.49, p < .05). In contrast, the relationship between identity and constructive deviance is not significant and trends in a negative direction when normative conflict is low (t = −0.87, p = .39). Hypothesis was therefore supported.
4.3 Discussion
The results of Study 2 fully supported Hypothesis . As shown in Figure 1, organizational identity was only associated with greater constructive deviance among employees who perceived high levels of normative conflict. Interestingly, identity was unrelated to constructive deviance when normative conflict was low, although it trended in a negative direction. Overall, the interaction shown in Figure 2 supports the normative conflict model concerning the conditions that promote dissent (high identity and high normative conflict) and indicate that this framework has utility for predicting deviance in the workplace, which is a novel application of the theory.
However, this preliminary test of the theory is limited in several key respects. First, the sample consisted of traditional-aged college students, who typically exhibit higher rates of organizational deviance than older adults, and we oversampled women, who typically exhibit lower rates of deviance than men (Berry et al., 2007). Study 2 also focused on part-time workers, who are generally less involved in their jobs than full-time workers (Thorsteinson, 2003). Consequently, these findings may not generalize to other contexts. Second, the design was cross-sectional, which increases the likelihood that common method variance (CMV) might have distorted our results (Podsakoff, MacKenzie, & Podsakoff, 2012). Third, our findings may be specific to the Galperin (2012) constructive deviance measure, which has limited validity evidence to date. These limitations point to a clear need to constructively replicate our findings with a different sample, design, and criterion measurement strategy in Study 3.
5 STUDY 3: REPLICATION AND EXTENSION OF THE NORMATIVE CONFLICT MODEL
In addition to addressing the limitations of Study 2, we wanted to provide a stronger test of the normative conflict model in Study 3 by testing the mediating state (psychological discomfort) that we expect will account for the moderating effect of normative conflict on the relationship between organizational identity and constructive deviance. Consequently, we evaluated our full model (Figure 1) using a second sample of working adults with a delay between the measurement of the predictor variables and the constructive deviance criterion.
5.1 Method
5.1.1 Sample and procedure
We recruited full-time employees from MTurk for Study 3, excluding any respondents who had participated in Studies 1a or 1b with the procedure developed by Peer et al. (2012). We used the same process of screening for data quality described in Study 1a. At Time 1, we collected 633 usable responses from full-time employees. At Time 2, which was 1 month later, we collected 539 usable matched responses for hypothesis testing (an 85% follow-through rate). The final sample was 54.5% female, 85.2% Caucasian, and reported a mean age of 35.83 years (SD = 11.17). With respect to their work, the majority of participants were individual contributors without supervisory responsibility (63.8%), and most of the participants had a long tenure with their employer (M = 5.21 years, SD = 5.46). The most commonly reported industries of employment in the sample were educational services (13.7%); professional, scientific, and technical services (11.1%); retail trade (10%); and health care and social assistance (9.8%).
5.1.2 Measures, Time 1
Normative conflict was measured with the eight-item scale developed in Study 1a (α = .94), and organizational identity was measured with the Leach et al. (2008) measure described in Study 2 (α = .97).
Psychological discomfort was measured with the three-item measure (α = .94) of dissonance-related discomfort developed by Glasford et al. (2008). Participants were asked to reflect on the way they typically feel when working in their organization; sample items include “uncomfortable with the way things are done” and “uneasy about things that are happening.” Responses were on a 7-point scale where 1 = “never” and 7 = “daily.” Higher scores indicate more frequent feelings of psychological discomfort.
5.1.3 Measure, Time 2
Constructive deviance was measured with the 13-item measure of prosocial rule breaking (PSRB; α = .95) developed by Dahling et al. (2012). PSRB is a form of constructive deviance identified by Morrison (2006) that is motivated by a desire to conform to broad hypernorms to avoid waste and show consideration to coworkers and customers or clients. We used the Dahling et al. measure in this study to demonstrate that our results are not specific to the Galperin (2012) measure. A sample item reads, “I violate organizational policies to save the company time and money.” Responses to both measures were made on a 5-point scale where 1 = “strongly disagree” to 5 = “strongly agree.” Dahling et al. (2012) developed and validated the PSRB measure over three studies and demonstrated its discriminant validity relative to other performance criteria (e.g., task performance and citizenship behaviors).
5.2 Results
5.2.1 Analysis strategy
Our analysis of the model shown in Figure 1 unfolded in several steps. First, we again conducted CFAs to evaluate the measurement model and discriminant validity of our measures and then analyzed correlations to examine the relationship between normative conflict and psychological discomfort (Hypothesis 2a). Second, we conducted moderated multiple regression to test the interactions of organizational identity with normative conflict, and with psychological discomfort, to predict constructive deviance (Hypotheses and 2b, respectively). Third, we tested the significance of the indirect moderating effect (Hypothesis 2c) using an adaptation of the moderated path analysis procedure recommended by Edwards and Lambert (2007) to evaluate mediated moderation. This approach employs bootstrapping to construct a confidence interval used to evaluate the statistical significance of the indirect effect.
5.2.2 Confirmatory factor analysis
We tested the hypothesized, four-factor measurement model using the same procedures described in Study 2. We formed item parcels to serve as indicators for the multidimensional organizational identity and PSRB measures. The three parcels for the identity measure again exhibited good reliability (solidarity, α = .95; centrality, α = .92; identity satisfaction, α = .97), as did the three parcels for the PSRB measure (efficiency, α = .92; coworker assistance, α = .92; customer assistance, α = .95). We used the individual items from the unidimensional normative conflict and psychological discomfort measures as the indicators for their respective latent variables (Hall et al., 1999; Williams & O'Boyle, 2008).
Results indicated that the hypothesized, four-factor measurement model exhibited acceptable fit to the data, χ2(113) = 410.86, p < .001; CFI = 0.96; RMSEA = 0.07; SRMR = 0.04, without allowing any error terms to covary. We compared the hypothesized model to two alternative models. First, we tested a three-factor alternative model on the basis of the strong observed correlation between normative conflict and psychological discomfort (r = .69, p < .001) that combined the items from these two measures on a single latent variable. Despite the strong correlation, alternative model #1 exhibited significantly worse fit to the data than the hypothesized measurement model, Δχ2(3) = 863.96, p < .001. Second, as in Study 1b, we tested a three-factor model that combined the items from the normative conflict and the organizational identity measures on a single latent variable. Alternative model #2 also showed significantly worse fit to the data than the hypothesized model, Δχ2(3) = 860.85, p < .001. Given support for the hypothesized model, we proceeded to test Hypotheses through 2c.
5.2.3 Tests of hypotheses
Table 3 reports descriptive statistics and correlations. Among the demographic variables, gender correlated with constructive deviance; women report somewhat less constructive deviance than men, which is a typical finding in past research on destructive deviance as well (e.g., Berry et al., 2007). We do not include gender as a statistical control in the results that follow because we did not theorize its effects in advance (Bernerth & Aguinis, 2016), but the pattern of hypothesis support is unchanged if gender is included in the analyses.
| Variable | M | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|---|
| 1. Gender | -- | -- | -- | ||||||
| 2. Age | 35.83 | 11.17 | .03 | -- | |||||
| 3. Tenure (years) | 5.21 | 5.46 | .03 | .45** | -- | ||||
| 4. Normative conflict | 2.70 | 1.06 | −.02 | −.01 | −.03 | (.94) | |||
| 5. Psychological discomfort | 3.18 | 1.70 | .03 | .02 | −.02 | .69** | (.94) | ||
| 6. Organizational identity | 4.60 | 1.46 | .03 | .01 | .15** | −.57** | −.55** | (.97) | |
| 7. Prosocial rule breaking | 2.14 | 0.84 | −.13** | −.01 | .07 | .33** | .26** | −.10* | (.95) |
- Note. Gender is coded such that 1 = male and 2 = female. Alpha coefficients are reported on the diagonal.
- * p < .05.
- ** p < .001.
We followed the process used by Grant and Berry (2011) to test our model. First, to test Hypothesis 2a, we regressed psychological discomfort on normative conflict while controlling for the effects of organizational identity and the interaction between identity and normative conflict. We mean centered prior to making all interaction terms (Aiken & West, 1991). The results of this analysis are shown in Table 4, where normative conflict has a strong, positive relationship with psychological discomfort. Hypothesis 2a was consequently supported.
| Predictors | Step 1 | Step 2 | ||||||
|---|---|---|---|---|---|---|---|---|
| b | SE | β | t | b | SE | β | t | |
| Organizational identity | −.27 | .04 | −.24 | −6.39*** | −.25 | .04 | −.22 | −5.94*** |
|
Normative Conflict |
.89 | .06 | .55 | 15.04*** | .89 | .06 | .56 | 15.34*** |
| Identity × normative conflict | −.13 | .03 | −.12 | −3.98*** | ||||
- *** p < .001
Next, we examined the predictors of constructive deviance to test Hypotheses and 2b. In Step 1, we regressed constructive deviance on organizational identity, normative conflict, and the interaction between identity and normative conflict, which tests Hypothesis . In Step 2, we added psychological discomfort and the interaction between identity and psychological discomfort to the regression equation, which tests Hypothesis 2b. The results of this regression analysis are shown in Table 5. In Step 1, normative conflict moderated the relationship between organizational identity and constructive deviance as expected, replicating our support for Hypothesis from Study 2. However, the moderating effect of normative conflict becomes nonsignificant in step 2, where psychological discomfort moderates the relationship between identity and constructive deviance instead (ΔR2 = .02). We plotted the interaction of identity and psychological discomfort on constructive deviance in Figure 3, which shows that identity is positively related to deviance when discomfort is high (t = 3.84, p < .001), but unrelated to deviance when discomfort is low (t = −.08, p = .94). This finding supports Hypothesis 2b.
| Predictors | Step 1 | Step 2 | ||||||
|---|---|---|---|---|---|---|---|---|
| b | SE | β | t | b | SE | β | t | |
| Organizational identity | .07 | .03 | .12 | 2.45* | .08 | .03 | .14 | 2.71** |
| Normative conflict | .32 | .04 | .41 | 8.36*** | .25 | .05 | .32 | 5.56*** |
| Identity × normative conflict | .07 | .02 | .14 | 3.37** | .02 | .03 | .04 | 0.64 |
| Psychological discomfort | .08 | .03 | .17 | 2.88** | ||||
| Identity × psychological discomfort | .05 | .02 | .16 | 2.43* | ||||
- Note. For Step 1, R2 = .14. For Step 2, ΔR2 = .02.
- * p < .05.
- ** p < .01.
- *** p < .001.
In sum, the piecemeal results suggest that psychological discomfort fully mediates the moderating effect of normative conflict: Using the logic for mediation introduced by Baron and Kenny (1986), normative conflict is positively associated with psychological discomfort, psychological discomfort moderates the relationship between identity and constructive deviance, and the direct-effect moderation of normative conflict becomes nonsignificant when the mediating moderator (discomfort) is added to the regression equation (see Figure 1). To conduct a formal test of mediated moderation, we adapted the path analytic framework recommended by Edwards and Lambert (2007), which uses bootstrapping to calculate a confidence interval around the indirect effect. Importantly, our model in Figure 1 differs from the models analyzed by Edwards and Lambert, who focused on decomposing a mediated effect that is moderated at the first and/or second stages. However, following Grant and Berry (2011), we modified the Edwards and Lambert (2007) reduced-form equations to calculate the significance of the indirect moderating effect of normative conflict on the relationship between organizational identity and constructive deviance via psychological discomfort. We then employed bootstrapping to construct a bias-corrected, 95% confidence interval by drawing 1,000 random samples with replacement from the full sample of 539 respondents. Our results indicated that the average bootstrap indirect effect was 0.04 and that this indirect effect was statistically significant because the confidence interval excluded zero (0.01 to 0.08). This is a weak indirect conditional effect because one component of it, the interaction of identity and psychological discomfort, had a small effect on constructive deviance, as is typical in field research (McClelland & Judd, 1993). However, in conjunction with the piecemeal results noted above, Hypothesis 2c was supported; the moderating effect of normative conflict on the relationship between organizational identity and constructive deviance is completely explained by the experience of psychological discomfort.
6 GENERAL DISCUSSION
People who strongly identify with groups are usually loyal conformists, but sometimes strong identifiers deviate in constructive ways for the good of the group. When does identity prompt constructive deviance rather than conformity to rules? We demonstrated that the relationship between organizational identity and constructive deviance depends on normative conflict and, more specifically, on the experience of psychological discomfort that is triggered by normative conflict. Studies 1a and 1b developed and validated a reliable measure of normative conflict with a stable factor structure. Study 2 provided a preliminary test of the normative conflict model to predict constructive deviance among employed students, demonstrating that identity only promotes constructive deviance when normative conflict is high. We then elaborated on these results in Study 3, which advanced the normative conflict model with a mediator, psychological discomfort, consistent with cognitive dissonance theory (Festinger, 1957). Study 3 replicated and extended Study 2 with a predictive design and a large sample of older, full-time workers.
To summarize our contributions in aggregate, our results first clarify a contradictory body of evidence concerning the relationship between group identity and constructive deviance, which is an important criterion that requires further study (Vadera et al., 2013). Organizational researchers typically take a dim view of deviant behavior, which overlooks the potential for constructive deviance to challenge norms and drive healthy change in business groups (Fuller, Marler, & Hester, 2006; Jetten & Hornsey, 2014). Understanding the conditions that shape constructive deviance is an important step toward responsibly managing this behavior. Second, our studies also introduced the normative conflict model to the organizational behavior literature to inform future research on organizational deviance and conformity. We provided the first evidence that the normative conflict model has utility for understanding behavior in the workplace, rather than in opinion-based social groups where the stakes for expressing dissent are oftentimes lower (e.g., Packer & Chasteen, 2010). Third, we expanded the normative conflict model by integrating it with cognitive dissonance theory (Festinger, 1957) to explicitly include psychological discomfort as a mediator, which is a core assumption of the model that has not previously been explained. This contribution clarifies why normative conflict can motivate strong identifiers to take drastic action to dissent against their organization.
6.1 Implications for theory
Our findings have important implications for theory concerning normative conflict and constructive deviance. First, with respect to the normative conflict model, we made an important contribution by articulating the role of psychological discomfort in this theory. Our integration of the normative conflict model with cognitive dissonance theory clarifies why normative conflict is a distressing experience; the results of Study 3 illustrate that the moderating effect of normative conflict is entirely explained by feelings of psychological discomfort. As cognitive dissonance theory explains, people are motivated to take action to resolve discrepancy-based discomfort (Elliot & Devine, 1994; Festinger & Carlsmith, 1959; Glasford et al., 2008). The theoretical arguments underpinning the normative conflict model hint at the importance of psychological discomfort to motivate strong identifiers to dissent (Dupuis et al., 2016; Packer, 2008), but ours is the first study to articulate how the normative conflict model is connected to cognitive dissonance theory and to provide evidence in support of this theoretical integration.
We also contributed to the development of the normative conflict model by illustrating its generalizability to formal business settings. Our studies provide a strong test of the model because existing empirical research on the model has only focused on informal, opinion-based social groups, such as discussion groups of university students (Packer & Chasteen, 2010). Although this type of context allows for a valid test of the model, informal social groups differ from business organizations in several key respects; for example, businesses have more regimented rules (X. Zhou, 1993), leaders who are generally intolerant of deviance (Rotundo & Sackett, 2002), and greater financial and personal stakes for deviance (e.g., job loss; Berry et al., 2007). Consequently, it was important to demonstrate in this research that normative conflict and social identity can still motivate dissent despite the prospect of steeper consequences. Support for the normative conflict model in organizational settings bolsters the generalizability of the theory and suggests that this model has potential for predicting other organizational behaviors as well, a point that we explore below as a future research direction.
Our work is also important to emerging theory concerning constructive deviance. Theory explaining the antecedents of constructive deviance is underdeveloped, and the empirical literature is presently characterized by a variety of contradictory findings (Vadera et al., 2013). Our results clarify the ambiguities surrounding the relationship between social identity and constructive deviance by pointing to the critical moderating role of normative conflict perceptions. As we noted previously, the normative conflict model offers valuable precision for helping organizational researchers (a) understand when constructive deviance is likely to occur and (b) better differentiate constructive and destructive deviance.
6.2 Implications for practice
Our findings demonstrate that feelings of normative conflict can trigger constructive deviance among highly identified employees, which underscores that managers need to talk with employees about their understanding of rules and policies in the workplace. Managers who want to guide and control constructive deviance can do so by carefully explaining the reasoning behind organizational rules and accepting employee input concerning alternative practices that might be preferable. We expect that employees are unlikely to dissent with extreme acts of rule breaking if they can instead resolve their dissonance-based discomfort by openly expressing voice and enacting change collaboratively with their supervisors and peers (Gutworth, Morton, & Dahling, 2013).
The results additionally indicate that managers should attempt to understand the intention behind rule-breaking behavior. Not all employee deviance is conducted with an intent to harm the organization, and constructive deviance may indicate that a current rule is outdated, inefficient, or harmful (Packer, 2011; Spreitzer & Sonenshein, 2003). By punishing all forms of deviance without consideration of the motivation, organizations may miss opportunities to adapt and grow. On the other hand, there may be employees who commit deviance with the intention to help the organization, but who lack the foresight to see how their acts may cause larger organizational problems. In this case, managers should be sure to set clear expectations for future behavior and suggest other ways that employees could advocate for changes that they see fit (Gutworth et al., 2013).
Last, our findings clearly demonstrate that dissenters are people who may otherwise be among the most loyal and supportive high identifiers in the organization. Managers should be thoughtful about how they react to deviance and take pains to understand the motivations behind deviant actions (Dahling et al., 2012). Managers who ignore the subtleties of deviance may wind up punishing, alienating, or firing employees who are making good-faith efforts to improve the organization's practices.
6.3 Future research directions
We applied the normative conflict model to better understand when employees may be motivated to engage in constructive deviance. However, the model may also provide insight into when an employee is likely to commit destructive behaviors. Although the model predicts that high normative conflict and high organizational identification are both necessary for predicting constructive deviance, it may be that the experience of high normative conflict under conditions of low organizational identification leads to disengagement that could manifest as destructive deviance, psychological withdrawal, or withdrawal behaviors such as absenteeism or quitting (Ellemers et al., 2002; Täuber & Sassenberg, 2012). Future research should test if the normative conflict model could be used to predict other workplace attitudes and behaviors, and especially if it can differentiate the conditions that promote destructive versus constructive deviance.
Although our focus in this study was on the conditions that enhance constructive deviance, future research should also focus on the ambiguities we found concerning conformity. In both Studies 2 and 3, identity had a null relationship with constructive deviance when normative conflict was low. In contrast, the normative conflict model generally posits that deviance should be lowered (or, conversely, that conformity should be increased) under these circumstances. We see several possible explanations that might account for this discrepancy. First, from a conceptual standpoint, loyal conformity is not merely the absence of dissent (Packer, 2008). We expect that the interaction of high identity and low normative conflict would yield a significant effect on a more direct measurement of organizational loyalty or rule conformity, which is an important issue to explore in future studies. A second explanation for this null finding is a simple floor effect; as shown in Figures 2 and 3, our samples reported overall low levels of constructive deviance, perhaps yielding an insufficient range for a negative relationship between identity and constructive deviance to emerge when normative conflict was low. A related, third explanation stems from the normative conflict model, which explains that high identifiers with low normative conflict may still engage in occasional instances of “strategic nonconformity” (Packer & Miners, 2014, p. 359). Strategic nonconformity involves dissenting from a norm that one actually endorses for the good of the group, such as by playing “devil's advocate” to build consensus and ensure that the group is making good decisions (Packer, 2011). Given the overall low levels of constructive deviance that we observed in our data, even very occasional strategic nonconformity might explain why we did not find a negative relationship between identity and constructive deviance when normative conflict was low.
Researchers may also gain additional insight into the normative conflict model by differentiating between self-focused versus group-focused normative conflict (Packer & Chasteen, 2010; Packer & Miners, 2014). Self-focused normative conflict occurs when problematic rules affect only the individual employee, whereas group-focused normative conflict occurs when problematic rules affect the group. Our measure was written to reference group-focused normative conflict, but it is likely that strong identifiers are not as willing to dissent when they perceive only self-focused normative conflict. By clarifying the focus of the normative conflict that people experience, future research may further explain when employees are likely to take different kinds of action to resolve their feelings of psychological discomfort.
Goal construal also has important effects within the normative conflict model that offer promising directions for future research. Packer and colleagues (Packer, 2014; Packer, Fujita, & Chasteen, 2014) have recently demonstrated that the relationship between group identity and dissent depends on participants' level of goal construal. When people consider goals at a high level of construal, they think in abstract, distal terms and are more open to the possibility of group change. In contrast, when people consider goals at a low level of construal, they think in concrete, immediate terms and prioritize group stability. Consequently, high identifiers are more likely to dissent when they adopt a high level of goal construal that makes the possibility of change and improvement salient (Packer et al., 2014). Future research could build on these findings in organizations by testing if constructive deviance is enhanced when high identifiers engage in activities such as strategic planning that encourage a high goal construal level.
Finally, future research should move beyond the study of antecedents to investigate the consequences of constructive deviance in organizations. Constructive deviance has the potential to yield a variety of desirable outcomes, such as enhanced efficiency (Dahling et al., 2012) and constructive organizational change (Fuller et al., 2006). However, constructive deviance may yield unintended, negative consequences as well; rules sometimes exist for valid reasons that employees do not understand, and breaking those rules may cause broader problems despite good intentions (Vadera et al., 2013; Vardi & Weitz, 2004). Although the normative conflict model sheds light on the motives and intended outcomes behind acts of workplace deviance, more research is needed to examine under what conditions these behaviors yield the desired results.
6.4 Limitations
Our research has some limitations that should be acknowledged. First, our samples are restricted in some respects. Racial and ethnic minorities were underrepresented in both Studies 2 and 3 that tested the normative conflict model. With respect to work experience, entry- and mid-level employees were overrepresented in both samples. It is therefore unclear whether and how individuals with greater power and supervisory scope engage in constructive deviance. Future research should consequently replicate our findings with more diverse samples.
Second, the variables in our studies were all self-reported. We made this choice because all of our antecedent variables (identity, normative conflict, and psychological discomfort) are perceptual variables that only the focal respondent could rate. Concerning the criterion variables, we chose self-ratings because managers, coworkers, and other raters may be unable to distinguish between the motive structures underlying constructive versus destructive deviance (Dahling et al., 2012). Fortunately, our focal hypotheses concerned interactions, which are not susceptible to CMV (Siemsen, Roth, & Oliveira, 2010), and our replication with a time-separated measure of constructive deviance in Study 3 also helps to allay concerns about CMV (Podsakoff et al., 2012). Nevertheless, our design would be stronger if an external observer could provide the deviance ratings.
Third, more validity data is needed on the new normative conflict measure and the constructive deviance measures utilized in our research; both the Dahling et al. (2012) and Galperin (2012) measures have limited validity evidence to date. Although we found consistent support for our hypotheses across multiple studies and with different measures of constructive deviance, more research is needed to collect evidence of the construct validity of these measures with additional samples and variables.
6.5 Conclusion
Constructive deviance may provide many benefits to organizations, such as improved organizational effectiveness and increased competitive advantage. Although research is increasingly focusing on this important criterion, conflicting findings exist concerning the relationship between organizational identity and constructive deviance. Our studies clarified this contradictory literature by applying the normative conflict model, designing a scale to measure normative conflict, and testing a mediated-moderation mechanism that explains how this experience motivates constructive deviance. We hope that our findings spark further applications of the normative conflict model to understand deviant behavior in organizations.
Biographies
Jason J. Dahling received a PhD in Industrial/Organizational Psychology from the University of Akron. His research focuses on self-regulation as it applies to performance management, emotion regulation, and career decision making. His previous work has been published in outlets such as Organizational Behavior and Human Decision Processes, Personnel Psychology, and Journal of Management.
Melissa B. Gutworth received a PhD in Industrial/Organizational Psychology from Penn State University. Her research examines creativity, deviance, leadership, and mechanisms behind the creativity–deviance relationship. Her research has been published in Journal of Business and Psychology, Journal of Creative Behavior, and Psychology of Aesthetics, Creativity, and the Arts.
