Followership styles scrutinized: temporal consistency and relationships with job attitudes and self-efficacy

Introduction

The tradition of a classic leader-centric view (i.e., the focus on the leader’s role and characteristics) is still dominant in the field of leadership research and practice (Avolio, Walumbwa & Weber, 2009; Banks et al., 2018; Dinh et al., 2014). However, the specific roles and contributions of followers in the leadership process (i.e., followership) have gained more attention in recent years (see, for instance, Khan, Busari & Abdullah, 2019; Uhl-Bien et al., 2014). In their integrative literature review and “Formal Theory of Followership” (FTF), Uhl-Bien et al. (2014) highlighted several ways to study how followers construe and enact their follower role, and how this may affect leaders and followership outcomes (i.e., the role-based approach to followership; see Uhl-Bien et al., 2014). In addition, they introduced the constructionist perspective on followership, which centers on how individuals mutually interact to (co-)construct leadership and followership identities. Thus, according to the FTF (Uhl-Bien et al., 2014), followership either describes the role enactment from a predefined follower rank or position, or reflects an individual’s actual act of following in a social interaction process. While the co-construction approach to followership is a rather novel approach within the field of followership research, the role-based approach has been developed over several decades (see, for instance, Chaleff, 1995; Hurwitz & Hurwitz, 2015; Kelley, 1988; Kelley, 1992).

Although Kelley (1988) provided one of the first (role-based) theories on the positive impact that followers can have in the leadership process, some of his assumptions are still awaiting empirical testing. Specifically, Kelley (1992) conceptualized different followership styles as rather stable behavior patterns based on the interaction of the followers’ active engagement (AE) in the leadership process and their independent, critical thinking (ICT) toward their leader. The notion of stable followership behaviors was also echoed by subsequent approaches to followership styles, such as Kelley (2008), Khan, Busari & Abdullah (2019), or Uhl-Bien et al. (2014). However, prior studies on Kelley‘s proposed followership behaviors¹ (Blanchard et al., 2009; Gatti et al., 2014; Gatti, Ghislieri & Cortese, 2017; Ribbat, Krumm & Hüffmeier, 2021) used cross-sectional designs and thus could not test his assumption of stable followership patterns. Indeed, there are various prominent approaches that consider behaviors at work as dynamic and variable (see, for instance, Beal et al., 2005; Weiss & Cropanzano, 1996). Therefore, followership behavior could also depend on situational factors such as followers’ mood or the current environment in which they work (Benson, Hardy & Eys, 2015; Weiss & Cropanzano, 1996).

Furthermore, followership approaches stress the role of followers by considering followership as an important independent variable in leadership research instead of considering it as the dependent variable (Uhl-Bien et al., 2014; Uhl-Bien & Carsten, 2018). Thus, according to followership theory (Kelley, 1988; Kelley, 1992; Uhl-Bien et al., 2014), followership behavior should be a major predictor for followership outcomes (e.g., individual follower outcomes such as job satisfaction or organizational commitment). On the individual follower level, prior research in fact demonstrated that AE and ICT correlate with job attitudes such as job satisfaction and organizational commitment (Blanchard et al., 2009; Gatti et al., 2014; Ribbat, Krumm & Hüffmeier, 2021). In contrast to followership approaches (Uhl-Bien et al., 2014; Uhl-Bien & Carsten, 2018), the literature on job attitudes and job performance (e.g., Riketta, 2008), however, suggests that job attitudes precede follower behavior. Since prior research (Blanchard et al., 2009; Gatti et al., 2014; Ribbat, Krumm & Hüffmeier, 2021) cannot provide insights into the direction of relationships due to their cross-sectional designs, the direction of the relationship between followership and job attitudes is still unclear.

In the current research, we conducted two studies to test the temporal consistency of Kelley’s (1992) followership behaviors in a longitudinal design, along with their relations to critical job attitudes (i.e., job satisfaction and organizational commitment) and self-efficacy for the first time. Specifically, we test both the construct stability of followership behaviors and the cross-lagged relationships between followership, job attitudes, and self-efficacy in two different samples. In Study 2, we additionally test two potential new moderator variables that were not part of Study 1 (i.e., leader humility and perceived organizational support (POS)).

Thus, our studies extend current research in several ways. First, we explore whether followership behavior can be characterized as rather stable or trait-like behavior patterns as proposed by Kelley (1992). This is important because Kelley’s (1992) conception of styles suggests that followership should be conceptualized as rather trait-like behavioral tendencies that are related to rather stable personal characteristics and/or general circumstances. A state-like nature of followership behavior, however, would necessarily shift the focus of future research from general and typical factors to more specific, situational, and contingent factors as followership would then rather be spontaneous, dynamic, or variable. Second, we examine whether followership behaviors are antecedents of job attitudes (i.e., job satisfaction and organizational commitment) as conceptualized by Kelley (1992) and/or vice versa in multistate models (Geiser, 2020; Prenoveau, 2016). These relationships have been investigated only in cross-sectional studies and, therefore, prior research could not yet provide a rigorous test of the direction of these relationships (see Byun, Lee & Kang, 2018; Ribbat, Krumm & Hüffmeier, 2021). Third, we explore the link of AE and ICT with self-efficacy, an important variable in the organizational context that has, however, not been studied thus far. According to Kelley (1992), p. 143), active and critical followers are goal-oriented, success-oriented, and effective, which culminates in a “can do aura”. Therefore, active and critical followership should be related to the followers’ perception of self-efficacy. Finally, we consider leader humility and POS as moderators. Hence, we also explore two potential new conditions under which the relationships of followership with job attitudes might be fostered. In this way, we contribute to a better understanding of the followership construct (as conceptualized by Kelley, 1992) and potential followership outcomes.

Followership

Followership behavior is defined as “behaviors of individuals acting in relation to a leader(s)” (Carsten et al., 2010, p. 545). In our study, we refer to the followership concept by Kelley (1992) that describes followership styles based on the interaction of the followers’ active engagement (AE) in the leadership process and their independent, critical thinking (ICT) towards their leader. According to Kelley (1992), the best followers are those who participate actively in the leadership process and take initiative. At the same time, they independently think for themselves and provide constructive criticism for their leader and group. By contrast, the worst followers do not independently think for themselves, simply take directions, and do not challenge their leader and group. Moreover, they are passive, lazy, and require constant supervision. Kelley (1992) proposes that the different combinations of AE and ICT result in five styles of followership behavior, which he describes as “passive” (i.e., low in both dimensions), “conformist” (i.e., high in AE, but low in ICT), “alienated” (i.e., low in AE, but high in ICT), “exemplary” (i.e., high in both dimensions), and “pragmatist” (i.e., with medium levels in both dimensions). The exemplary followership style is considered as most effective and most valuable to the organization (Kelley, 1988; Kelley, 1992).

In our current study, we will examine whether followership behavior can be characterized as rather stable behavior patterns as proposed by Kelley (1992). To do so, we apply a latent state-trait approach (Geiser, 2020; Steyer et al., 2015). This approach can provide answers to the question of whether measurement instruments assess more trait-like or more state-like attributes. Specifically, a latent state-trait approach quantifies to which degree observed and/or underlying latent state (τ) variables reflect trait effects that indicate consistency (i.e., coefficient Con_τ; see Geiser, 2020) or situation effects/person by situation interaction effects that indicate occasion specificity (i.e., coefficient Osp_τ; see Geiser, 2020). Measures can be considered as trait-like, when more than 50% of their true state variance is due to trait effects (Geiser, 2020; Steyer, Schmitt & Eid, 1999). In accordance with Kelley’s (1992) conceptualization of followership styles as rather stable (i.e., trait-like) behavior patterns we predict:

Hypothesis 1: (a) AE and (b) ICT will have a higher proportion of state variance at each time point that is due to trait effects (i.e., Con [ τ_t1,t2] >.50) than state residual variance (i.e., Osp [ τ_t1,t2] <.50).²

Followership and job attitudes

Again, in accordance with Kelley’s conceptualization, we posit that followership behavior will be related to important job attitudes (i.e., job satisfaction and organizational commitment). Our Hypotheses are visualized in Figs. 1 and 2. Job satisfaction is defined as “a pleasurable or positive emotional state resulting from the appraisal of one’s job or job experiences” (Locke, 1976, p. 1304). Organizational commitment “(a) characterizes the employee’s relationship with the organization, and (b) has implications for the decision to continue or discontinue membership in the organization” (Meyer & Allen, 1991, p. 67).

Empirical evidence suggests a positive relationship of AE with job satisfaction (e.g., Blanchard et al., 2009; Ribbat, Krumm & Hüffmeier, 2021). With regard to ICT, several studies found no significant relationship of ICT with job satisfaction (Gatti et al., 2014; Gatti, Ghislieri & Cortese, 2017; Ribbat, Krumm & Hüffmeier, 2021). Blanchard et al. (2009) even reported that ICT was negatively associated with job satisfaction. Their results further suggest that the interaction of AE and ICT increases intrinsic job satisfaction (i.e., satisfaction with job aspects that are task-related such as responsibility) but decreases extrinsic job satisfaction (i.e., satisfaction with job aspects that are unrelated to the task such as the supervisor; Blanchard et al., 2009). Specifically, followers with high AE and high ICT had the highest levels of intrinsic job satisfaction. The more followers engaged in ICT, the lower was their extrinsic job satisfaction, while this effect was weaker when followers were also actively engaged (Blanchard et al., 2009). The findings of Favara (2009) and Saraih et al. (2018) also showed a positive relation of an exemplary followership style (i.e., high scores both on AE and ICT) with job satisfaction. Furthermore, two studies (Blanchard et al., 2009; Ribbat, Krumm & Hüffmeier, 2021) showed that AE was positively associated with organizational commitment, whereas ICT was negatively related to organizational commitment.

To derive our first set of hypotheses we apply Cognitive Dissonance Theory (Festinger, 1957; Harmon-Jones & Mills, 2019). According to Cognitive Dissonance Theory, employees tend to harmonize conflicting cognitions (e.g., attitudes and behaviors). We follow the effort justification tenet of Cognitive Dissonance Theory (Aronson & Mills, 1959; Harmon-Jones, Armstrong & Olson, 2018), which refers to the strategy of adding consonant cognitions to dissolve experienced dissonance when undertaking demanding or exhausting activities and achieving outcomes that may not be satisfying. Thus, we expect followers to value their job and their organization more to justify the considerable effort they spend on engaging actively in the leadership process (see also Blanchard et al., 2009; Ribbat, Krumm & Hüffmeier, 2021). Hence, their earlier active engagement (at t1) is expected to increase their later job attitudes (at t2). This reasoning is in line with the dominant view that AE antecedes followers’ job attitudes (Blanchard et al., 2009; Gatti et al., 2014; Ribbat, Krumm & Hüffmeier, 2021). Thus, we predict:

Hypothesis 2a: AE at t1 will be positively related to job satisfaction at t2.

Hypothesis 2b: AE at t1 will be positively related to organizational commitment at t2.

However, we expect that ICT at t1 exerts a negative influence on attitudinal variables at t2, because it increases the followers’ awareness of the problems and negative aspects of their job. Hence, rather than valuing their job and organization more due to the effort that has also to be expended on independent, critical thinking, followers with high levels of ICT should become less satisfied and committed when faced with the problems and negative aspects of their job (see also Blanchard et al., 2009; Ribbat, Krumm & Hüffmeier, 2021). Therefore, we hypothesize:

Hypothesis 3a: ICT at t1 will be negatively related to job satisfaction at t2.

Hypothesis 3b: ICT at t1 will be negatively related to organizational commitment at t2.

Besides effort justification, another strategy to avoid cognitive dissonance according to Cognitive Dissonance Theory (Festinger, 1957; Harmon-Jones & Mills, 2019) would be to bring behaviors in line with one’s attitudes. That is, attitudes can guide and facilitate behavior. This notion suggests that job attitudes can cause later job behaviors (Hinojosa et al., 2017; Riketta, 2008). Therefore, we expect satisfied and committed followers to actively engage in the leadership process. We state:

Hypothesis 4a: Job satisfaction at t1 will be positively related to AE at t2.

Hypothesis 4b: Organizational commitment at t1 will be positively related to AE at t2.

However, we do not expect satisfied and committed followers to question the leader and the organization frequently. Since satisfaction and commitment are pleasurable and positive emotional states (Locke, 1976; Meyer & Allen, 1991), there should be no incentive for the followers to create conflicting cognitions and risk psychological discomfort (Festinger, 1957; Harmon-Jones & Mills, 2019). Thus, satisfied and committed followers might avoid becoming aware of negative aspects of their job through ICT in order to prevent cognitive dissonance. Hence, we predict that job attitudes at t1 will be negatively related to ICT at t2.

Hypothesis 5a: Job satisfaction at t1 will be negatively related to ICT at t2.

Hypothesis 5b: Organizational commitment at t1 will be negatively related to ICT at t2.

Additionally, we argue that the effort justification mechanism (cf. Hypothesis 2) is stronger than the “attitudes as guidelines” mechanism (cf. Hypotheses 4 and 5), because followership styles, which are conceptualized as relatively stable or trait-like behavior patterns (Kelley, 1992; see also our Hypothesis 1), should be more difficult to change than job attitudes. Therefore, we expect stronger relations of AE and ICT at t1 with the attitudinal variables at t2 due to effort justification than the opposite relations between attitudes (at t1) and AE and ICT (at t2). Thus, we hypothesize:

Hypothesis 6: The relationships of AE at t1 with (a) job satisfaction at t2 and (b) organizational commitment at t2 is stronger as compared to the relationships of (a) job satisfaction at t1 and (b) organizational commitment at t1 with AE at t2.

Hypothesis 7: The relationships of ICT at t1 with (a) job satisfaction at t2 and (b) organizational commitment at t2 is stronger as compared to the relationships of (a) job satisfaction at t1 and (b) organizational commitment at t1 with ICT at t2.

Since Kelley’s (1992) typology of followership styles refers to the interaction of both followership dimensions (i.e., AE and ICT), we also investigate this interaction. Continuing the argumentation above, the less followers engage actively in the leadership process, the more should the negative influence of ICT on job attitudes prevail over effort justification. However, independent, critical followers might also use active engagement to change undesired working conditions and, thus, they should become more satisfied and committed. Therefore, following the logic of both Cognitive Dissonance Theory (Festinger, 1957) and Kelley’s (1992) conceptualization, we predict:

Hypothesis 8: ICT at t1 will be less negatively related to (a) job satisfaction at t2 and (b) organizational commitment at t2 when AE at t1 is high as compared to when AE at t1 is low.

Followership and Self-Efficacy

Self-efficacy is defined as “people’s judgements of their capabilities to organize and execute courses of action required to attain designated types of performances” (Bandura, 1986, p. 391). According to Kelley (1992), active and critical followers are goal-oriented, success-oriented, and effective. He describes effective followers as those who determine early and accurately the criteria for success in the leadership environment, who track a record of successes in tasks that are important to the leader, and who “seize smaller victories that accumulate into a ‘can do’ aura” (Kelley, 1992, p.143). Therefore, active and critical followership should raise the followers’ awareness for their own capabilities to attain personal and leadership goals. Consequently, active and critical followership should be related to the followers’ perception of self-efficacy. Prior meta-analyses (Sadri & Robertson, 1993; Stajkovic & Luthans, 1998) demonstrated that self-efficacy is positively related to work-related behaviors (e.g., changes in career tracks or intentions to show assertiveness) and job performance. However, the relation of this important individual-level construct with followership behavior has not been tested so far. Therefore, we will investigate this relationship in our study.

According to Social Cognitive Theory (Bandura, 1997), self-efficacy emerges from four sources: (1) physiological and affective states, (2) verbal persuasion, (3) vicarious experience, and (4) mastery experience. Mastery experience is considered to have the strongest impact on the development of self-efficacy, because it provides authentic evidence for one’s capability to succeed, and, thus, builds a robust belief in one’s personal efficacy (Bandura, 1997). When followers simply take directions, there is no evidence for the followers’ own capability to succeed, because success fully depends on the leader’s decisions. Active and critical followership, however, allow for evaluating the followers’ own contributions to leadership success (e.g., followers’ high individual or collective job performance) and, thus, raises the awareness for mastery experiences. Therefore, we expect that the followers’ self-efficacy substantially results from ascribing (at least parts of) the leadership success to own efforts when carrying out an active and independent follower role. Hence, we expect that both AE at t1 and ICT at t1 are positively associated with self-efficacy at t2, which is also in line with the dominant view of followership research (Kelley, 1992; Uhl-Bien et al., 2014) that followership should be a major predictor for followership outcomes:

Hypothesis 9a: AE at t1 will be positively related to self-efficacy at t2.

Hypothesis 9b: ICT at t1 will be positively related to self-efficacy at t2.

Paralleling our above hypotheses on job attitudes, the relationship of ICT (at t1) with self-efficacy (at t2) might depend on the level of AE. While it may be obvious to ascribe leadership success to own efforts (i.e., to allow for mastery experience; see Bandura, 1997) when active engagement is high, the contribution of ICT to leadership success might not always be apparent. This should, for instance, be the case when a follower expresses concerns about the leader’s direction and consequently does not (strongly) contribute to goal attainment. According to Kelley (1992), followers may become frustrated and “alienated”, when their concerns are ignored repeatedly, which keeps them away from their goal achievements and day-to-day job satisfaction (see Kelley, 1992). When active engagement is high, however, ICT could be considered as an additional contribution to leadership success and therefore raise even more awareness for mastery experiences than the active engagement alone. Thus, we predict:

Hypothesis 10: ICT at t1 will be more positively related to self-efficacy at t2 when AE at t1 is high than when AE at t1 is low.

Furthermore, according to Social Cognitive Theory (Bandura, 1997), self-efficacy is also considered to be a predictor for one’s choices of activities, effort expenditure, persistence, thought patterns, and emotional reactions when confronted with obstacles (Lent, Brown & Hackett, 1994). Followers with high self-efficacy should therefore be more likely to actively engage in the leadership process and rather be willing to act independently and critically. Thus, we predict:

Hypothesis 11a: Self-efficacy at t1 will be positively associated with AE at t2.

Hypothesis 11b: Self-efficacy at t1 will be positively associated with ICT at t2.

Study 1

Results

Table 1 shows the descriptive statistics, the intercorrelations of the latent variables and the reliabilities of the measures. For ICT at t2, ω_s = .69 was slightly below the commonly used minimum value (Dunn, Baguley & Brunsden, 2014; Kline, 1998). However, we addressed this problem by specifying latent state-trait models with indicator-specific residuals (Eid, Schneider & Schwenkmezger, 1999; Geiser, 2020; see also above).

Followership and job attitudes

The first cross-lagged model explored the relationship of AE with job satisfaction. Both autoregressive paths were significant (for AE: β = .75, p < .001; for job satisfaction: β = .71, p < .001). The relationship of AE (at t1) with job satisfaction (at t2) was not significant, β =  − .08, p = .12, thereby not supporting Hypothesis 2. However, job satisfaction (at t1) was positively related to AE (at t2), β = .13, p = .02, thereby supporting Hypothesis 4a. The model explained 60% of the variance in AE and 49% of the variance in job satisfaction and the model fit was good: χ²(df) = 524.84(371), p < .001, χ2/df = 1.42, RMSEA = .05, CFI = .95, SRMR = .06.

The second cross-lagged model included AE and organizational commitment. Both autoregressive paths were significant (for AE: β = .74, p < .001; for organizational commitment: β = .87, p < .001). The relationship of AE (at t1) with organizational commitment (at t2) was not significant, β = .02, p = .39, which does not support Hypothesis 2. However, organizational commitment (at t1) was positively related to AE (at t2), β = .11, p = .049, thereby supporting Hypothesis 4b.⁵ The model explained 59% of the variance in AE and 75% of the variance in organizational commitment and the model fit was good: χ²(df) = 325.76(236), p < .001, χ2/df = 1.38, RMSEA = .05, CFI = .96, SRMR = .05.

The relationship of ICT with job satisfaction was explored in the third cross-lagged model. While we had predicted this relationship to be negative, our results rather pointed in a positive direction. Thus, we applied a two-tailed test for this model. We, therefore, could neither find a significant relation of job satisfaction (at t1) to ICT (at t2), β = .16, p = .08, nor between ICT (at t1) and job satisfaction (at t2), β = .09, p = .21. These results do not support Hypotheses 3a and 5a. Both autoregressive paths were significant (for ICT: β = .74, p < .001; for job satisfaction: β = .71, p < .001). The model explained 53% of the variance in ICT and 50% of the variance in job satisfaction and the model fit was again good: χ²(df) = 201.37(146), p = .002, χ2/df = 1.38, RMSEA = .05, CFI = .97, SRMR = .06.

The fourth cross-lagged model explored the relationship of ICT with organizational commitment. Again, we applied a two-tailed test for the model, because our results pointed in another direction than we had expected. Both autoregressive paths were significant (for ICT: β = .72, p < .001; for organizational commitment: β = .86, p < .001). No significant relationship of ICT (at t1) with organizational commitment (at t2) could be found, β = .06, p = .41. Organizational commitment (at t1) was also not related to ICT (at t2), β = .17, p = .07. These results do not support Hypotheses 3b and 5b. The model explained 55% of the variance in ICT and 75% of the variance in organizational commitment and the model fit was good: χ²(df) = 80.71(65), p = .09, χ2/df = 1.24, RMSEA = .04, CFI = .98, SRMR = .06.

In Hypotheses 6 and 7, we proposed the relationships of AE (at t1) and ICT (at t1) with job attitudes (at 2) to be stronger than those of job attitudes (at t1) with AE (at t2) and ICT (at t2). Both hypotheses were not supported by our results, because we found no significant relationship of AE or ICT (at t1) with any job attitude (at t2). The relationship of AE (at t1) with job satisfaction (at t2) was not significant, β =  − .08, p = .12. The relationship of AE (at t1) with organizational commitment (at t2) was not significant, β = .02, p = .39. The relationship between ICT (at t1) and job satisfaction (at t2) was also not significant, β = .09, p = .21. Finally, no significant relationship of ICT (at t1) with organizational commitment (at t2) could be found, β = .06, p = .41.

Furthermore, in Hypothesis 8 we posited ICT (at t1) to be less negatively related to job satisfaction and organizational commitment (at t2) when AE (at t1) was high, as compared to when AE (at t1) was low. In our analysis, the interaction of AE (at t1) and ICT (at t1) was not a significant predictor for job satisfaction (at t2), β =  − .10, p = .14. However, the interaction (at t1) predicted organizational commitment (at t2), β = .14, p = .04, while controlling for the autoregressive effect of organizational commitment (at t1), β = .86, p < .001. The interaction effect is plotted in Fig. 3. Overall, the model explained 78% of the variance. Since the interaction of AE and ICT (at t1) was associated with a higher score in organizational commitment, Hypothesis 8b was supported.

Brief Discussion

Overall, we found evidence for some, but not for all of our hypotheses. The results of Study 1 support Kelley’s (1992) assumption that followership behaviors are rather stable behavior patterns. Since we found no evidence for the relationships of AE or ICT (at t1) and later job attitudes or self-efficacy (at t2), our results of Study 1 challenge the original idea of followership theory, that followership behaviors are significant predictors for important organizational variables (see, for instance, Kelley, 1992; Uhl-Bien et al., 2014). However, the interaction of AE and ICT (at t1) was positively related to organizational commitment (at t2), which corresponds to Kelley’s (1992) assumption that exemplary followership leads to desired organizational variables. One reason for the absence of the expected effects could be the length of the time lag between the two measurement waves (nine to 12 months). That is, some effects might have simply vanished over time and, thus, might not have been possible to detect. This would be the case, for instance, if the respondents’ leader changed between the two measurement waves, if the proposed effects unfolded too quickly, or if the effects did not last long enough to be detected (for a related analysis of time-lags in leadership research, see Fischer, Dietz & Antonakis, 2017).

Another reason could be that we did not include conditions under which the relationships of followership with job attitudes might be fostered in our first study. Kelley (1988) argued that active and critical followers are the most effective and, thus, active and critical behaviors should be positively related to followership outcomes. However, he also admitted that an exemplary followership style might not always be the preferred style of a certain leader or organization. Therefore, exemplary followership might not always be acknowledged or rewarded. The missing acknowledgement of the followers’ efforts of engaging actively and critically in the leadership process could explain why followership behavior does not necessarily increase the follower’s job satisfaction.

Since our results have considerable implications both scientifically and practically, we conducted a second study to once again test our hypotheses and check whether the results of Study 1 are robust with a bigger sample size and, thus, more statistical power. In addition, we include potential moderator variables to test whether leader-related or organizational variables that may facilitate or even reward exemplary followership can foster the relationship of followership with job attitudes.

Study 2

Compared to Study 1, we made the following changes in Study 2: In addition to the variables that were considered in Study 1, we also tested two moderator variables that were not part of Study 2 (i.e., leader humility and perceived organizational support, see below). Thus, Study 2 went beyond a mere replication of Study 1 and examined two potential new conditions under which the relationships of followership with job attitudes may be fostered. Finally, we used a shorter time-lag than in Study 1 and assessed whether the followers’ direct leader (i.e., the person they referred to when responding to the followership questionnaire) had changed between t1 and t2 to rule out potential shortcomings of our initial study design.

Materials and Methods

We again conducted a two-wave survey. In contrast to the time lag between the two measurement waves in Study 1 (i.e., nine to 12 months), we realized a time lag of four months for Study 2. With a shorter time lag, we intended to rule out possible shortcomings of Study 1: Some effects might have simply disappeared over time and thus could not be detected. This would be the case, for instance, if the respondents’ leader changed between the two surveys, if the proposed effects unfolded too quickly, or if the effects did not last long enough to be detected (for a related analysis of time-lags in leadership research, see Fischer, Dietz & Antonakis, 2017). Since there were no other longitudinal studies that studied followership behaviors when we planned our study, we drew from the meta-analysis on relationships between performance and job attitudes by Riketta (2008) to obtain information about suitable time lags. On that basis, we considered four months to be a reasonable time lag to detect effects on attitudinal variables in the context of work. This was also in line with the general recommendation of pertinent literature to use, in case of doubt, a rather shorter time-lag (e.g., Dwyer, 1983; Griep et al., 2021; Voelkle et al., 2012). This study was preregistered at the Open Science Framework (https://osf.io/9q48p/?view_only=b02c98f57f7841129cf6daf42405b410).

Results

Table 2 shows the descriptive statistics, the intercorrelations of the latent variables and the reliabilities of the measures of Study 2.

Discussion

The results of both of our studies support Kelley’s (1992) assumption that followership behaviors are rather stable behavior patterns. We found significant relations of job attitudes with active, engaged followership behavior in the cross-lagged models, above and beyond the autoregressive effects in both studies. We also found significant relations of AE (at t1) with job satisfaction (at t2) and of ICT (at t1) with self-efficacy (at t2) in Study 2. The interaction of AE and ICT (at t1), however, predicted organizational commitment (at t2) only in Study 1.

Although the regression coefficients (.08 ≤ β ≥ .14) can be described as rather small (Cohen, 1988), the identified effects indicate important findings for three reasons. First, cross-lagged effects are generally hard to find. That is, autoregressive effects are often strong and therefore explain much of the variance by themselves (Adachi & Willoughby, 2015; Geiser, 2020). Hence, there is often little variance left to be explained by the cross-lagged effects. Second, since our results demonstrate that followership styles are rather stable behavior patterns, predicting change in followership behavior is even more noteworthy. Third, the reported effect sizes are indeed in a typical range as compared to similar studies with cross-lagged models (see, for instance, Riketta, 2008; Sonnentag, Binnewies & Mojza, 2010). However, across both studies, we found mixed results regarding our hypotheses. We will elaborate on these findings below.

Theoretical Implications

Our studies contribute to a better understanding of followership (Kelley, 1992) and its outcomes (Uhl-Bien et al., 2014). To go beyond prior research, we tested Kelley’s (1992) followership behaviors longitudinally, and explored their relationships to critical job attitudes (i.e., job satisfaction and organizational commitment) and self-efficacy as potential predictors and consequences. Our findings support Kelley’s (1992) assumption that followership styles (i.e., AE and ICT) are rather stable behavior patterns. This is an important finding, because a more state-like nature of followership behavior would necessarily shift the focus of future research from general and stable factors to more specific, situational, and contingent factors as followership would then rather be spontaneous, dynamic, or variable. For instance, the common perspective on leadership styles (see Anderson & Sun, 2017; Bass & Avolio, 1995) as consistent behavior patterns of leaders has been challenged most recently. An increasing number of authors argue for considering leader behaviors as more dynamic (see, for instance, Kelemen, Matthews & Breevaart, 2020; McClean et al., 2019). In contrast, our findings suggest that an opposite perspective is adequate for Kelley’s (1992) followership styles. However, several studies could show how leader and follower identities (that did not necessarily correspond to their formal ranks as “leaders” and “subordinates”) were (co-)constructed in different leadership situations (e.g., Blom & Alvesson, 2014; Larsson & Nielsen, 2021; Van De Mieroop, 2020). Adopting the co-construction approach to followership (see Uhl-Bien et al., 2014), Larsson & Nielsen (2021), for instance, found that leader and follower roles remained abstract in the workplace interactions that they analyzed. Participants rather focused on negotiated, task-oriented, expert or non-expert identities (Larsson & Nielsen, 2021). Hence, a more dynamic or situational perspective on followership can be helpful in those contexts, in which leaders and followers are not determined by their position. Our findings, however, suggest that subordinates tend to enact their follower role rather consistently within a continuum between active and passive, independent and uncritical (see Kelley, 1992), when they interact with their superior leader.

It is noteworthy and plausible, however, that ICT is less stable or trait-like than AE, a finding that we consistently observed across our studies. Speaking up in front of the leader can be a risky behavior for followers, particularly when it is done in challenging rather than supportive ways (Burris, 2012). Thus, followers probably consider their concerns carefully before they actually express them to the leader (Bashur & Oc, 2015; Detert et al., 2013). Consequently, they might hold back their contrary view, when they feel that their concern is less important, but they rather have the courage to speak up in situations where they perceive urgency. Therefore, it is plausible that potentially risky critical followership is more dependent on the evaluation of the urgency and appropriateness of a specific situation than the willingness to support the leader through active engagement. Still, even if followership behaviors cannot be seen as fully invariable personality traits, both AE and ICT are more trait-like than state-like according to our findings.

With regard to the relationships of AE and ICT with critical job attitudes (i.e., job satisfaction and organizational commitment) and self-efficacy, we obtained mixed results. In Study 1, we did not find significant relationships of AE or ICT (at t1) with job attitudes (at t2) that we had predicted. Furthermore, we did not detect any significant relation of followership with self-efficacy. In contrast, job satisfaction and organizational commitment (at t1) were positively related to AE (at t2). In Study 2, we obtained similar results with three exceptions. First, in Study 2, we did not find a relationship of job satisfaction (at t1) with AE (at t2), but AE (at t1) predicted later job satisfaction (at t2). Second, ICT (at t1) was negatively related to self-efficacy (at t2) in Study 2, although we had predicted a positive relation. This relationship was not significant in Study 1. Third, our results of Study 1 suggest that high levels of both AE and ICT (i.e., an exemplary followership style, see Kelley, 1992) lead to higher organizational commitment, thereby supporting Kelley’s (1992) assumption that high levels of both AE and ICT imply desirable organizational behaviors. However, we could not detect this interaction effect in Study 2. One reason for the relatively low bivariate associations across time and for the missing interaction effects in our studies could be that our analysis was restricted by low variance in followership styles. Most participants in both studies adopted either the pragmatist or the exemplary followership style (see SM2 for more details). Furthermore, the high mean values of both followership dimensions in both studies (see Tables 1 and 2) might indicate a certain risk of social desirability of the questions, which has also been discussed in previous studies (see, for instance, Gatti et al., 2014; Ribbat, Krumm & Hüffmeier, 2021).

While some of our results indicate that followership behaviors can predict later job attitudes or self-efficacy, most of our results either point in the opposite direction or indicate no significant relationships at all. Those results challenge the original idea of followership theory (Kelley, 1992; Uhl-Bien et al., 2014) that followership behaviors are significant predictors for organizational variables. However, our studies do provide at least a few hints for the potential of followership behaviors as predictors for job attitudes and self-efficacy. In sum, our findings raise important questions for future followership research.

First, the mixed findings across our two studies suggest that future research needs to elaborate on time lags in longitudinal research. An increasing number of authors calls for the appropriate inclusion of time aspects both in research designs and in theory, which has been neglected in leadership and organizational research for a long time (see Fischer, Dietz & Antonakis, 2017; Griep & Zacher, 2021; Griep et al., 2021; Shamir, 2011). In Study 1, we used data from a preexisting, more comprehensive data set that had a time lag of nine to 12 months. Since there are some comparable studies that used shorter time lags to detect interrelations of job attitudes and behaviors (for an overview, see, for instance, Riketta, 2008), we used a shorter time lag (i.e., four months) in Study 2. We indeed found two relationships of followership behaviors (at t1) with later job satisfaction and self-efficacy (at t2) that we did not detect in Study 1. These findings suggest that followership behaviors might affect job attitudes or self-efficacy in the short-term rather than in the long-term. Hence, our findings correspond to the analysis of leadership research by Fischer, Dietz & Antonakis (2017) in view of two points: Fischer, Dietz & Antonakis (2017) concluded that effects at the team- or individual level unfold rather quickly, while they do not last very long. Furthermore, they stated that effects on behaviors typically take longer to unfold and persist longer than do effects on cognitions or emotions (Fischer, Dietz & Antonakis, 2017). Still, we need more information on the role that time plays both in followership and leadership research to develop theory further and to better understand the nature of the studied effects (Castillo & Trinh, 2018; Griep et al., 2021). We, therefore, highly recommend future research to consider different time lags in order to learn in which time frames the effects of followership behaviors occur and when they potentially decline. In our two studies, we could demonstrate that followership behaviors are relatively stable, thereby supporting Kelley’s (1992) conceptions of rather consistent followership styles. Hence, our findings can guide future followership research, since “the temporal stability of variables and the stability of effects on these variables are criteria for deciding on repeated-measures designs” (Fischer, Dietz & Antonakis, 2017, p. 1740).

Second, we encourage future research to explore potential mechanisms that might appear to be the missing links for the relationships we could not detect in our studies. Benson, Hardy & Eys (2015), for instance, suggest that the situational context affects how leaders see followership behaviors. That is, attempts to influence a leader’s decisions in front of others might not be appreciated by the leader. However, while this might be obvious for independent and critical behavior, we would still expect the leader to acknowledge the follower’s support through active engagement. Hoption (2016), for instance, demonstrated that the followers’ provision of help to leaders corresponds to greater leader relationship satisfaction. Thus, it is somewhat surprising that we could find a positive relation of AE (at t1) with job satisfaction (at t2) only in Study 2, and no positive relation of AE (at t1) with organizational commitment (at t2) in both of our studies. Future research should, therefore, explore potential moderator and mediator variables that might uncover the most detrimental and beneficial conditions for positive followership outcomes.

We suggest that the leaders’ preferences and reactions to certain followership styles should be a good starting point to detect the missing links between active and critical followership and followership outcomes. We firstly tested leader humility, which was not confirmed as a moderator in our study despite our prediction. However, Shen & Abe (2022), for instance, found an indirect effect of followership behaviors on job performance through perceived supervisor support. Moreover, we expected POS to strengthen the employees’ beliefs that the organization recognizes and rewards increased performance (Rhoades & Eisenberger, 2002) and thereby foster the relationships between active and critical followership with job attitudes. The corresponding hypotheses, however, were not supported by our data. We could neither find a three-way interaction of followership behavior, leader humility and POS. One reason for this could be that leader humility, and POS were highly correlated (r = .60; see Table 2). Future research could further explore which organizational environments affect followership behaviors and its potential outcomes. Blair & Bligh (2018), for instance, argue that values and norms focused on hierarchy and control limit active follower beliefs in shared responsibility for leadership. The findings of Carsten et al. (2010) echo this argument by suggesting that followers’ ability to take initiative is diminished by strong bureaucracy.

Third, we explored the relationship of followership with self-efficacy for the first time. In our hypotheses, we postulated that the followers’ self-efficacy substantially results from ascribing (at least parts of) the leadership success to own efforts when carrying out an active and independent follower role as they experience mastery in this way. According to Social Cognitive Theory (Bandura, 1997), mastery experience is considered to have the strongest impact on the development of self-efficacy. We could only detect a negative relation of ICT (at t1) with later self-efficacy (at t2) in Study 2, although we had predicted a positive relation. One reason could be again that ICT was not appreciated by the leaders. The followers’ efforts to contribute independently then might had been not successful and, thus, they were experienced as a personal failure rather than personal mastery. In addition, we did not detect any significant relationship of self-efficacy (at t1) with later followership behaviors (at t2), which might result from strong autoregressive paths or inconvenient time-lags. Future research should further investigate whether followership can predict self-efficacy or vice versa with different time-lags. Thereby, research should also clarify whether or under which circumstances exemplary followership implies mastery experience and to which extent Social Cognitive Theory (Bandura, 1997) is applicable to followership research. Moreover, future studies could additionally build on other sources for self-efficacy. If other sources than mastery experience played a more important role than we had expected based on Social Cognitive Theory (Bandura, 1997), different effects could have cancelled each other out. This would be the case, for instance, if simply taking directions (i.e., uncritical followership) led to vicarious experience for some followers, while high levels of independent thinking led to mastery experience for others.

Practical Implications

Our findings provide interesting information for followers, leaders, and organizations. We could demonstrate that followership behaviors are relatively stable behavior patterns, even though it is possible that they change under certain circumstances. Thus, Kelley’s (1992) “Identify Your Followership Style Questionnaire” indeed allows for assessing one‘s own general behavioral tendencies and preferences regarding the interaction with the leader. Hence, our findings support Kelley’s (1992) conceptual idea of followership styles in this respect. Since followership, here, is enacted from a formal follower rank or position (see also Uhl-Bien et al., 2014), followership styles should be particularly relevant in the contexts of hierarchical organizations. It may be more difficult to “Identify Your Followership Style”, when the reference leader is not apparent in non-hierarchical organizational structures. This would be the case, for instance, in self-organized or agile teams, in which leader and follower roles shift more fluently (see, for instance, Srivastava & Jain, 2017; Zhu et al., 2018).

Our finding regarding the consistency of Kelley’s (1992) followership behaviors is valuable for followers, because knowing one’s own style allows for reflecting on how the style might fit or not with the leader and/or organization. This is important, because many studies have demonstrated that person-organization-fit, for example, increases satisfaction and reduces the employee’s intention to leave the organization (Verquer, Beehr & Wagner, 2003; Jin, McDonald & Park, 2018). Furthermore, person–supervisor fit is associated with greater leader satisfaction and with a better relationship between leader and follower (Kristof-Brown, Zimmerman & Johnson, 2005; Marstand, Martin & Epitropaki, 2016). For leaders, the assessment of the followers’ individual styles may help to understand why followers tend to behave in certain ways. This, in turn, can facilitate an adequate handling of leader-follower-interactions.

For organizations, recognizing followership styles as rather stable behavioral patterns can help to inspire development programs for followership styles that are needed or preferred within the organization. An increasing number of authors calls for such development programs (see, for instance, Bufalino, 2018; Hoption, 2014; Logan & Ganster, 2007). While complementing effective leadership training with followership development programs (Bufalino, 2018; Hoption, 2014) might be a useful approach to foster desired followership behaviors, our findings suggest that it might not be sufficient to simply call for active and/or critical followership to obtain positive organizational outcomes. According to our study, satisfied and committed followers are more likely to participate actively in the leadership process. Thus, our findings at least indirectly point to satisfactory working conditions and appropriate organizational goals as beneficial settings for active followership.

Following Andersson’s (2018) argumentation that followership is an important social resource for organizational resilience, organizations might care about developing desired followership to successfully meet future challenges that could emerge from a pandemic or from other far-reaching developments like, for instance, technological change, disruptive innovation, or climate change. Consequently, followers, leaders, and organizations should reflect on what styles of followership they want to have or show, and which styles they need. Kelley’s (1992) operationalization of followership behaviors (i.e., AE and ICT) can be a useful tool to learn about the way followers actually carry out their follower role. Still, there is more research needed to better understand the impact that different followership styles can have both on the individual and on the organizational level.

Limitations

There are several limitations to this study. First, we used only self-report data, which implies certain risks of common-method bias (Podsakoff et al., 2003). We addressed related problems by using a time-lagged design. However, future research could complement the evaluations of the followers with their leader’s perceptions or with independent observations to further decrease such risks (for related recommendations, see also Gatti et al., 2014; Ribbat, Krumm & Hüffmeier, 2021).

Second, an increasing number of authors (see, for instance, Antonakis et al., 2010; Bastardoz et al., 2023; Wulff et al., 2023) pointed to the threat of making false causal claims due to endogeneity problems. Since our analysis relied on self-assessment assessed with survey measures both for behaviors and attitudes, our analyses were at risk to suffer from such endogeneity problems. However, the inclusion of multiple measurement points and the cross-lagged design should have reduced the risks for simultaneity bias and reverse causality in our studies. Still, causal inferences should be made with due caution. Future studies, therefore, could apply a variety of methods to further test the causal relationships between followership behaviors and followership outcomes (e.g., experimental designs or multi-method designs with independent observations).

Third, we used data from different samples that had been collected within different time-lags. We used data from a preexisting more comprehensive data set that had a time lag of nine to 12 months in our first study. Since we assumed that a positive impact of followership may have vanished by this time and since there are some comparable studies that used shorter time lags to detect interrelations of job attitudes and behaviors (for an overview, see, for instance, Riketta, 2008), we used a shorter time-lag (i.e., four months) in our second study. While our findings suggest that followership behaviors affect job attitudes and self-efficacy rather short-term than long-term, change in relatively stable followership behaviors might unfold rather slowly. However, several relationships that we had hypothesized were neither significant in Study 1 nor in Study 2. While our studies still can be an orientation for designing time-lags for followership research, we recommend future research to consider different time lags in order to learn in which time frames the effects of followership behaviors occur and when they potentially decline. Furthermore, future studies could include more than two time points for a more comprehensive understanding of the longitudinal relationships.

Finally, the sample of our first study consisted of employees that were employed within 11 organization from the service sector. The sample of our second study was a heterogeneous convenience sample from an online respondent pool. Hence, we cannot claim generalizability of our findings, even if several findings were consistent across both of our studies. Future research should, however, further investigate followership behavior in various samples and different sectors to study the generalizability of the impact of active and critical followership.

Conclusion

Our study contributes to a better understanding of Kelley’s (1992) followership behaviors and its relations to important job-related variables (i.e., job attitudes and self-efficacy). Active engagement and independent critical thinking were both found to be stable and rather consistent followership behavior patterns across two different samples and within two different time-lags. Furthermore, we could detect positive relationships between followership behaviors, job satisfaction and organizational commitment, and a negative relationship between independent, critical thinking and self-efficacy above and beyond the autoregressive effects. However, across the two studies, we obtained mixed results for several relationships that we had predicted. While some of our hypotheses were only supported in one of the two studies, other relationships that we had predicted were neither significant in Study 1 nor Study 2. Hence, more research is necessary to explore potential mechanisms (including time) that link followership with relevant outcomes. Still, our studies open up promising avenues for future research and provide starting points for its conceptual designs (i.e., for instance, with regard to the treatment of followership behaviors as rather consistent behavior patterns or the definition of appropriate time-lags for longitudinal followership research).

Supplemental Information