Advancing the social identity theory of leadership: A meta-analytic review of leader group prototypicality

Leader group prototypicality as the average versus ideal group member

The concept of prototypicality has its origins in categorization theory where it is used primarily to refer to nonhuman categories (e.g., animals, objects; Rosch, 1978). In the social identity approach, the concept of group prototypicality has been applied to human groups (Turner et al., 1987). This introduces an important deviation from its use for nonhuman groups. For most practical purposes, prototypes of nonhuman groups can be understood to capture the typical or average group member—they are most similar to other members and descriptive of the group as it is (perceived to be). For human groups, in contrast, prototypes also capture an ideal or aspirational element that includes attributes such as future-oriented goals and ambitions. To a certain extent, goals and ambitions also capture what it is that guides group members’ attitudes and behavior. Importantly, however, goals and ambitions also speak to what we strive for and in that sense reflect the ideal-type of the group—positive deviations from the typical or average group member. For human groups, group prototypes thus should not be understood in purely descriptive terms—who we are—but also in aspirational terms—who we would ideally be (as with collective possible selves; Stam et al., 2014; van Knippenberg et al., 2004). These ideas also resonate with ideas put forward about implicit leadership theories (e.g., Epitropaki et al., 2013; Junker & van Dick, 2014) that distinguish between average and ideal prototypes of leaders in general. Even though research on implicit leadership theories does not concern leader group prototypicality, and thus does not concern the distinction between average and ideal-type group prototypical leaders, it provides more general support for the idea that we can form categories (e.g., whether these pertain to leaders or social groups more generally) based on more average or ideal representatives of that category.

Accordingly, considerations of leader group prototypicality have emphasized this normative and aspirational quality (Steffens et al., 2014; Steffens et al., 2013; van Knippenberg, 2011). Importantly, however, others have understood leader group prototypicality as referring to the typical or average group member rather than to such an ideal-type (e.g., Halevy et al., 2011), in part because of its origins in categorization theory where for nonhuman groups this distinction is less relevant. Nevertheless, this raises the question of whether an understanding of leader group prototypicality that includes the aspirational, ideal-type aspect is more strongly related to leadership outcomes than an understanding that only reflects a more descriptive, average member notion of group prototypicality. As noted above, we predict that this is likely to be the case, because goals and aspirations capture not necessarily how things are but beliefs about how things could be (and, more specifically, who we want to be) that provides the basis for leaders’ capacity to mobilize followers’ energies and hence their effectiveness. These beliefs are therefore important in guiding group members’ attitudes and behaviors.

Although there has been no direct test of this reasoning within the leader group prototypicality literature, our current analysis aligns closely with work on deviance and adherence to in-group norms (e.g., Crane & Platow, 2010; Packer, 2008). A critical component of this latter work lies in the distinction between descriptive and injunctive norms (e.g., Cialdini et al., 1991). Much like our distinction between descriptive and aspirational group prototypes, descriptive norms reflect what group members are doing, while injunctive norms are more aspirational, describing what group members approve of doing. Across three studies with both preexisting and laboratory-created groups, Crane and Platow (2010) observed that participants who self-categorized as group members were more likely to align their behaviors with injunctive than descriptive norms. Indeed, these group members were willing to deviate from descriptive group norms to align themselves with the injunctive, aspirational group norms.

Accordingly, we hypothesize that these aspirational elements will be more closely aligned with leaders’ ability to motivate group members to pursue group objectives:

H1: Leader group prototypicality will be more strongly associated with leadership outcomes for conceptualizations of group prototypicality that capture the prototype as an ideal rather than an average group member.

Leader group prototypicality as implicitly versus explicitly capturing intergroup context

The very notion of a social group implies an intergroup context where membership in a group derives its meaning in part from the fact that other people do not belong to the group (Turner et al., 1987). The conceptualization of group prototypicality captures this in so far as the group prototype reflects not only what group members have in common but also what distinguishes their group from other groups (Postmes et al., 2005; Turner et al., 1987). From this perspective, an operationalization of leader group prototypicality would capture not only perceptions that a leader is representative of attributes that are shared by in-group members but also that the leader is representative of those things that distinguish the in-group from other out-groups.

A review of the literature suggests that studies vary in this respect, with some operationalizing prototypicality explicitly with reference to the intergroup comparative context (i.e., how the group differs from other groups; e.g., Cicero et al., 2008, Study 2; Koivisto et al., 2013, Study 2; Pierro et al., 2009: Study 2; Platow & van Knippenberg, 2001), but others operationalizing prototypicality in ways that are noncomparative in terms of the intergroup context (e.g., De Cremer et al., 2010, Study 1; Giessner et al., 2009, Study 1; Leicht et al., 2014). This is an issue that is closely connected to the concept of leader group prototypicality. Accordingly, addressing whether such conceptual differences are consequential is not just a methodological matter, but also speaks to theoretical issues related to the psychology of prototypicality perception (i.e., whether or not this is enhanced by intergroup comparison).

One reading of these differences in operationalization is that they are largely inconsequential because there always is an implicit intergroup comparison when defining a group prototype. That is, what group members have in common is, at least in part, what distinguishes them from members of other groups (i.e., so that these are not separate elements but necessarily interlinked). Another reading could be that noncomparative operationalizations may include characteristics that are not seen to be related to what makes the group distinctive and therefore are less impactful in driving social influence. This reading would see explicit distinctiveness as a key element of group identity (cf. Brewer, 1991; Turner, 1985) such that conceptualizations which are comparative in intergroup terms more accurately capture leader group prototypicality. This latter reading leads us to hypothesize that:

H2: Leader group prototypicality will be more strongly associated with leadership outcomes when conceptualizations of group prototypicality capture the prototype in intergroup comparative terms rather than noncomparative terms.

Leader group prototypicality and prototype strength (indexed by group longevity)

The group prototype is not an objective given but reflects group members’ mental representation of the group (Turner et al., 1987). Such mental representations differ in the extent to which they are well-developed and clear in perceivers’ minds (Oakes et al., 1994; Turner et al., 1994). We conceptualize such variations in the extent to which group members’ mental representations of group identity are well-developed as variations in prototype strength. For some groups, members will have a better, more developed understanding of what the group identity is—a stronger group prototype—than is the case for others. Self-categorization theory would suggest that group longevity is one fundamental determinant of these differences in prototype strength. In line with this point, group longevity has been observed to be one of the fundamental characteristics that differentiates between social and organizational groups (Hollenbeck et al., 2012). At one extreme, there are short-lived groups that exist for such a limited period of time that they do not really have a history or significant time horizon into the future (e.g., student teams working on a course assignment or a team put together for one short-term project). At the other extreme, there are social groups that have existed for decades or even for hundreds of years and that have a long horizon into the future, as is the case for many organizations and a range of societal groups. Such differences in group longevity are a  basis for differences in prototype strength, because it takes time for a stable and well-developed prototype to emerge (Reicher & Hopkins, 2001; Sani et al., 2007; Shamir et al., 1993).

Accordingly, even though there are, to the best of our knowledge (based on the search reported below), no studies of leader group prototypicality that have directly assessed prototype strength, for reasons outlined above, we can use differences in the longevity of the groups that have been studied as a proxy for prototype strength—while also acknowledging that proxies of this form allow only for an initial indirect test of the influence of the construct of interest. In this context, it is apparent that studies of leader group prototypicality have relied on groups of varying longevity. For example, some laboratory experiments have relied on ad hoc groups that have had no prior interaction (e.g., De Cremer et al., 2010: Study 1; Giessner & van Knippenberg, 2008: Studies 1, 2, and 4) and short-lived groups in the field (e.g., Fielding & Hogg, 1997), whereas others have studied leadership in preexisting and ongoing groups in the field (e.g., Hirst et al., 2009; Lipponen et al., 2005; Sluss et al., 2012) and in the laboratory (e.g., Platow & van Knippenberg, 2001; Platow et al., 2006). Across these studies, we hypothesize that:

H3: Leader group prototypicality will be more strongly associated with leadership outcomes in groups that have stronger prototypes, as indexed by group longevity.

Leader group prototypicality of members in formal versus nonformal leader positions

The study of leadership is not limited to people who have formal leadership roles (i.e., those with senior positions in formal organizational hierarchies) but also includes people who have nonformal leadership roles (e.g., Fransen et al., 2014; Hollander & Julian, 1969; Shamir & Howell, 1999; Wheelan & Johnston, 1996). Research informed by the social identity theory of leadership has included studies of the impact of leader group prototypicality on the effectiveness of both leaders in formal and nonformal leadership positions (e.g., Fielding & Hogg, 1997; van Knippenberg et al., 2000) and does not make formal distinctions concerning the impact of leader group prototypicality for these two types of leaders. Nevertheless, we would argue that a formal (vs. a nonformal) leader role should have a bearing on the impact of leader group prototypicality as a consequence of the expectations and obligations that people associate with formal leader roles. Formal leader roles put individuals in positions where they are expected by subordinates to fulfill leadership functions associated with the group (e.g., to initiate and organize group activities, to coordinate members), but these same expectations are likely to be less strong for people in nonformal roles. Moreover, as a function of its structural (hierarchical) legitimacy, it is also possible that a formal leadership position will be more directly associated with subordinates’ expectations that the leader should be representative of the group.

Such role expectations should affect the salience of formal and nonformal leaders’ group prototypicality. Formal leaders are more likely to stand in the spotlight and act on behalf of the group in interactions with other groups, which should make evaluations of their group prototypicality particularly salient for followers and mean that leader group prototypicality has a stronger bearing on followers’ responses. This is not to say that group prototypicality would not influence the leadership effectiveness of nonformal leaders; theory and evidence suggests it does (e.g., Fielding & Hogg, 1997); rather the issue is that leader group prototypicality should exert a stronger influence on the effectiveness of formal leaders. On this basis we hypothesize that:

H4: Leader group prototypicality will be more strongly associated with leadership outcomes for members in formal than in nonformal leadership positions.

Follower group identification and leader group-serving behavior as widely examined moderators of leader group prototypicality effects

Early accounts of the social identity theory of leadership suggested two key moderating influences of leader group prototypicality effects: follower social identification with the group and leader group-oriented behavior (van Knippenberg & Hogg, 2003). First, because leader group prototypicality revolves around shared group membership, it has been argued that the influence of group prototypicality should be more pronounced to the extent that followers identify with the group (Hogg, 2001). By way of example, members of a team should care little about whether or not the leader embodies what the team is about if psychologically they fail to see themselves as a member of the group. But they are likely to be more sensitive to the leader’s representativeness of the team when that team is an important part of their identity.

Second, leader group prototypicality and leader group-serving behavior have been proposed to interact such that the one substitutes for the other (van Knippenberg & Hogg, 2003)—a prediction that has been subjected to empirical tests (Giessner & van Knippenberg, 2008; Giessner et al., 2009; Platow & van Knippenberg, 2001; Platow et al., 2006; van Knippenberg & van Knippenberg, 2005). An important element in group members’ willingness to follow a leader and be open to the leader’s influence is the expectation that the leader acts with the group’s best interest in mind—and is “doing it for us” (Haslam & Platow, 2001). This expectation is positively impacted by leader group prototypicality; group norms, values, and interests are an integral part of the group prototype, and being perceived to be group prototypical is intertwined with the perception that the leader has these norms, values, and interests at heart. Group-serving behavior (e.g., self-sacrifice on behalf of the group; van Knippenberg & van Knippenberg, 2005) also feeds into this expectation that past behavior will predict future behavior. Leaders are not expected to continuously prove their group-serving motivations, however. Once trust in leader group-servingness is established, further signals to that effect are less consequential. In effect, this means that leader group prototypicality and leader group-serving behavior substitute for each other’s influence in that either one is sufficient to instill trust in leader group-serving motivations that feeds into leadership effectiveness (Giessner & van Knippenberg, 2008).

Although these moderation effects have been summarized in narrative reviews (e.g., Hogg et al., 2012), they have not been subjected to a meta-analytic examination. This is important because while some studies have found support for the moderating role of follower group identification (e.g., van Dijke & De Cremer, 2008) and leader group-serving behavior (e.g., van Knippenberg & van Knippenberg, 2005) others have found that follower group identification (e.g., Hogg et al., 1998: Study 1) and leader group-oriented behavior (Platow et al., 2006: Study 1) do not moderate the effect of leader group prototypicality. To provide a more comprehensive picture of the state of the science, we therefore seek to establish meta-analytically whether these are robust effects that hold across the available body of evidence. Thus, in line with early theorizing, we hypothesize:

H5. Leader group prototypicality will be more strongly associated with leadership outcomes to the extent that followers identify strongly with the group.

Overview of present research

In the present research, we advance the social identity theory of leadership by subjecting the four hypotheses outlined in the previous sections to a meta-analytic test. In addition, the available evidence also allows us to rigorously explore the impact of two moderators examined in primary research. Furthermore, we sought to examine whether leaders who are group prototypical are effective not only for subjective evaluations of perceived leadership (as shown in previous work; Barreto & Hogg, 2017) but also for behavioral outcomes. In the study of leadership, behavioral outcomes (e.g., performance, creativity, citizenship) are of greater interest because they are not only more closely aligned with the understanding of leadership effectiveness in terms of the pursuit of group goals (Kaiser et al., 2008) but also influenced to a lesser extent by perceptual biases and common method variance (van Knippenberg, 2012). In this regard, while there are no direct tests from primary leader group prototypicality research, it is plausible to expect that subjective evaluations of leadership will be more strongly impacted by perceptions of leader group prototypicality than behavioral outcomes. Importantly, though, the effect of leader group prototypicality should also be apparent for behavioral outcomes.

Finally, we will take advantage of the unique benefits afforded by meta-analyses to assess the presence of various indicators of publication bias (e.g., due to small study effects, Kepes et al., 2012) to obtain an indication of the cumulative reliability of the present literature (Cumming, 2014; Cumming & Calin-Jageman, 2017; Schmidt & Oh, 2016).

Study search procedure

A flow diagram displaying an overview of the literature search procedure is presented in Figure 1. Three trained independent coders with a degree in psychology conducted a literature search for relevant studies in Web of Science, PsycINFO, and Google Scholar (so that each data base was searched by two different coders) using the following keywords: leader group prototypicality OR leader prototypicality OR prototypical leader OR leader embodiment OR leader representativeness OR supervisor organizational embodiment OR leader organizational embodiment OR social identity and leadership. To be included, all articles were required to have been conducted and reported by January 2020. Overall, this search identified a total of 6,008 articles. In the first phase, 356 duplicate articles were removed, and the remaining were screened for potential relevance on the basis of their title and abstract, and in cases in which the title and abstract did not allow determination of the relevance of the article, the full text was inspected. The phase identified 278 unique articles for further full-text review. In the second phase, we conducted a full-text search of these remaining articles and excluded irrelevant articles based on our inclusion and exclusion criteria (as detailed below), resulting in a total sample of 59 unique articles. Furthermore, to ensure the completeness of this procedure, we also searched all articles that cited influential reviews by Hogg (2001), Haslam et al. (2011), Hogg et al. (2012), and van Knippenberg (2011). At the time of review, 2,234 articles cited Hogg (2001), 1,198 articles cited Haslam et al. (2011), 273 cited Hogg et al. (2012), and 257 articles cited van Knippenberg (2011). Reviewing these citations resulted in 11 additional articles.

OPEN IN VIEWER

Figure 1.

Flow diagram displaying process of identification and selection of relevant studies.

Finally, to address publication bias and obtain a more accurate estimation of the effect size based on all existing data, we sought to obtain and include unpublished data (including from dissertations, conference presentations, book chapters, and unpublished manuscripts). With this in mind, we sent out calls for data via leading associations in social and organizational psychology and management including the Society for Personality and Social Psychology, the Academy of Management’s Organizational Behavior Division, the European Association of Social Psychology, the European Association of Work and Organizational Psychology, the German Psychological Society, and the Society for Australasian Social Psychology. In addition, we contacted the authors of articles reporting primary data that our literature search identified to seek additional unpublished data/reports they may have. These calls allowed us to obtain an additional 11 reports. Altogether, this search strategy resulted in a final sample of 128 independent samples (251 effect sizes, N = 32,834). The sample included data from published journal articles (k = 90), dissertations/theses (k = 25), unpublished manuscripts (k = 7), unpublished data (k = 3), and conference presentations (k = 3). All references to studies reporting primary data that have contributed to the analysis are marked with and an overview of the included samples is presented in the Supplemental material (see Table S1).

Inclusion criteria

Data were included on the basis of the following inclusion and exclusion criteria. We included data concerning quantitative assessments of the relationship between leader group prototypicality and at least one leadership outcome. We included data based on studies employing experimental, cross-sectional, and longitudinal methods. Data based on studies employing qualitative methods as well as theory and review articles that did not report primary data were excluded. Data were included if they were gathered by means of an experimental manipulation or a quantitative measure of a leader’s prototypicality of a collective group (i.e., manipulating or assessing the extent to which the leader is prototypical of a group, team, or organization). Data were excluded if they related to (a) leaders’ prototypicality of other entities including prototypicality of leaders in general (rather than the specific group that they were leading; e.g., van Quaquebeke et al., 2014), (b) in-group versus out-group leaders (e.g., Duck & Fielding, 1999), or (c) a group’s prototypicality of a superordinate group (e.g., Thomas et al., 2017).

Data were included if they related to the explicit assessment of at least one leadership outcome (e.g., perceived leader effectiveness, trust in the leader, follower performance; see overview of variables and corresponding number of effect sizes in Table 1). Finally, to be included, studies had to be reported in English and provide necessary statistical information for analysis (this led to eight papers being excluded: one in Arabic, one in Korean, two in Chinese, one in Lithuanian, and three in Spanish). In cases in which data were missing or coding was hindered by ambiguous reports, we contacted the author(s) of the paper via email to request missing data (we sent out two additional reminders at an interval of one week if the author(s) had not responded within two weeks). We were unable to contact four authors due to the absence of valid contact details from the report and no available details online. Of the 16 authors who we contacted, 12 authors responded with missing data, three authors responded by informing us that they were not able to provide missing data, while one author did not respond. In cases in which we were not able to obtain missing data, we computed effect sizes from statistics provided in the paper where possible (in two cases there was insufficient information reported and so the report was excluded).

OPEN IN VIEWER

Table 1.

Results of categorical moderation analyses of the relationship between leader group prototypicality and leadership outcomes.

Moderator	r	95% CI	t	Effect sizes (k)	F(df1, df2)	σ level 2 2	R2	σ level 3 2	R 2
Effect size moderators (Level 2)
Leadership outcome					F(16,234) = 5.51***	.018***	42.74	.062***	0.00
Evaluative vs. behavioral index					F(1,249) = 37.69***	.025***	21.70	.055***	3.58
Behavioral indices	.19	.11, .27	4.68	44 (12)
Follower extra-role performance	.17	.06, .27	3.18	20
Follower in-role performance	.35	.20, .49	4.26	19
Counterproductive work behaviors	.12	-.05, .29	1.36	6
Evaluative indices	.41	.37, .45	17.62	207 (116)
Leader effectiveness	.36	.26, .45	12.62	49
Leader endorsement	.43	.36, .50	10.06	38
Perceived leader fairness	.37	.19, .53	6.90	24
Trust in the leader	.53	.45, .60	11.50	19
Job satisfaction	.34	.24, .44	6.35	15
Leader charisma	.49	.38, .58	8.05	10
Perceived leader influence	.45	.38, .51	3.96	9
Identification with leader	.21	.12, .31	4.47	8
Well-being	.23	.09, .35	3.21	8
Team cohesion	.39	.26, .51	5.45	7
Openness to organization change	.22	.03, .40	2.26	6
Commitment to organization	.61	.48, .72	7.34	5
Follower perceived organization support	.35	.18, .50	3.93	5
Turnover intentions	.26	.04, .46	2.34	3
Between-study moderators (Level 3)
Prototypicality conceptualization					F(1,249) = 14.99***	.032***	0.69	.050***	13.47
Average	.35	.30, .39	14.03	203 (110)
Ideal	.55	.46, .63	10.14	48 (18)
Intergroup context					F(1,77) = 2.10	.006	0.00	.037***	3.10
Implicit context	.19	.12, .27	5.02	52 (30)
Explicit context	.29	.18, .38	5.46	27 (16)
Prototype strength					F(1,249) = 9.50**	.031***	1.26	.053***	7.51
Short-lived group	.29	.22, .37	7.34	86 (47)
Enduring group	.43	.38, .48	15.08	165 (81)
Leader role					F(1,249) = 7.81**	.032***	0.00	.053***	8.23
Formal	.41	.37, .46	16.03	209 (101)
Nonformal	.26	.16, .36	4.82	42 (27)
Leader group prototypicality scale					F(4,167) = 2.41	.039***	0.00	.045***	9.97
Platow & van Knippenberg (2001)	.51	.43, .58	11.08	49 (29)
van Knippenberg & van Knippenberg (2005)	.45	.34, .54	7.70	36 (19)
Steffens et al. (2014)	.50	.39, .60	7.48	33 (12)
Eisenberger et al. (2010)	.27	.12, .41	3.43	25 (10)
Residual others	.44	.32, .56	6.24	29 (12)
Study methodology					F(1,249) = 28.30***	.031***	2.53	.045***	22.45
Experimental	.23	.16, .30	6.00	79 (46)
Correlational/longitudinal	.46	.41, .50	17.35	172 (82)

Note. r = weighted sample size corrected correlation coefficient between leader group prototypicality and leader outcome; 95% CI = lower and upper bounds of 95% confidence interval; T = t value for coefficient when compared to zero; k = number of effect sizes with the number in parentheses representing the number of studies; σ² = the variance not accounted for by the moderating variable at each level, respectively; F(df₁, df₂) = the test of heterogeneity between the levels of a moderating variable.

*p < .05; **p < .01; ***p < .001.

Study coding procedure

For each study, we coded several variables as described in more detail below including author names, year of publication, country in which the sample resided, sample size, publication status, leader outcome, average vs. ideal member conceptualization, prototype strength: short-lived vs. enduring group, implicit vs. explicit intergroup context, formal vs. nonformal leader position, leader outcome index, leader group prototypicality scale, effect size, and study methodology. In the case of two moderators (the two most often examined moderators)—followers’ social identification with the group and leaders’ group-serving behavior—several studies reported explicit tests of the moderation of the leader group prototypicality–effectiveness relationship by respondents’ social identification and by leaders’ group-serving behavior. For these two factors, we thus coded the effect size of the interactions between leader group prototypicality and followers’ social identification and between leader group prototypicality and leaders’ group-serving behavior to meta-analyze the effect size of the moderation test.

Two independent coders, who both had a degree in psychology, received training concerning the variables of interest before coding all variables independently. Agreement between coders was high (95.96%, Cohen’s κ = 0.919; Cohen, 1960). Discrepancies between coders were resolved in discussion with the first author.

Leader group prototypicality scale

We coded effect sizes based on the inventory used in cross-sectional and longitudinal studies to assess leader group prototypicality. Common scales to assess leader group prototypicality are presented in Table 2. The most commonly used scales were the leader group prototypicality scale developed by Platow and van Knippenberg (2001; 49 effect sizes), the leader prototypicality scale developed by van Knippenberg and van Knippenberg (2005; 36 effect sizes, which extends Platow and van Knippenberg’s scale), the leader identity prototypicality subscale of Steffens et al.’s (2014) Identity Leadership Inventory (33 effect sizes), and Eisenberger et al.’s (2010) supervisor organizational embodiment scale (25 effect sizes). Because van Knippenberg and van Knippenberg’s (2005) scale is based on Platow and van Knippenberg’s (2001) scale, some studies referenced both scales and in these cases we coded the effect size as being based on Platow and van Knippenberg (2001). The remaining 29 effect sizes based on residual other scales created specifically for the study in question were coded as ‘other’.

OPEN IN VIEWER

Table 2.

Common scales assessing leader group prototypicality.

Scale name (Reference)	k	Scale items
Leader Relative In-group Prototypicality Scale(Platow & van Knippenberg, 2001)	49	1. [This leader] represents what is characteristic about [in-group].2. [This leader] is representative of [in-group] members.3. [This leader] is a good example of the kind of people who are members of [in-group].4. [This leader] stands for what members of [in-group] have in common.5. [This leader] is not representative of the kind of people who are members of [in-group]. (reversed)6. [This leader] is very similar to most people in [in-group].
Leader Group Prototypicality Scale(van Knippenberg & van Knippenberg, 2005)	36	1. [This leader] is a good example of the kind of people that are members of [in-group].2. [This leader] represents what is characteristic about [in-group].3. [This leader] has a lot in common with members of [in-group].4. [This leader] is very similar to the members of [in-group].5. [This leader] represents what [in-group] stands for.6. [This leader] is an embodiment of our [in-group] group norms.
Identity Leadership Inventory–Leader Identity Prototypicality Subscale(Steffens et al., 2014)	33	1. [This leader] embodies what it means to be a member of [in-group].2. [This leader] is representative of members of [in-group].3. [This leader] is model member of [in-group].a 4. [This leader] exemplifies what it means to be a member of [in-group].a
Supervisor Organizational Embodiment Scale(Eisenberger et al., 2010)	25	1. When [this leader] encourages me, I believe that [in-group] is encouraging me.2. When [this leader] is pleased with my work, I feel that [in-group] is pleased.3. When [this leader] compliments me, it is the same as [in-group] complimenting me.4. When [this leader] pays attention to my efforts, I believe that [in-group] is paying attention to my efforts.5. [This leader] is characteristic of [in-group].6. [This leader] and [in-group] have a lot in common.7. When I am evaluated by [this leader], it is the same as being evaluated by [in-group].8. [This leader] is representative of [in-group].9. [This leader] is typical of [in-group].
Sample of other residual (less common) scales assessing leader group prototypicality
Leader Group Prototypicality Scale created for purpose of study(Janson et al., 2008): Study 2)	1	1. In terms of interests and values, [this leader] does not have a lot in common with the team. (reversed)2. [This leader] is a great fit for the team.3. [This leader] exemplifies what is best about the team.a
Leader Group Prototypicality Scale created for purpose of study(León et al., 2009)	1	1. [This leader] represents the most characteristic traits of the group.2. [This leader] would be elected by the group members to represent the group before others.3. [This leader] is well-liked by the group members.4. [This leader] personifies the most positive values of the group.a
Leader Group Prototypicality Scale created for purpose of study(Sluss, 2006)	1	1. [This leader] is an exemplary member of the organization.a 2. [This leader] acts according to the organizational goals and values.a 3. [This leader] represents what the organization stands for.

Note. k = number of independent samples that used a given scale.

^a Indicates items assessing ideal-type notion of leader group prototypicality as coded in present analysis. The sixth item from van Knippenberg and van Knippenberg (2005) was used only in Study 4 and has not been used commonly in subsequent research. When researchers referenced both Platow and van Knippenberg (2001) and van Knippenberg and van Knippenberg (2005) as the scale they used, the effect size was coded as Platow and van Knippenberg (2001).

Analytic procedure

The majority of primary research reported correlations, and so we used mean-weighted r as effect size of choice. Given that our data included a substantial number of studies reporting mean differences, we followed suggestions from Borenstein and colleagues (2009) to combine effect sizes from different study designs by converting to a common effect size. To perform the meta-analysis we followed procedures developed by Borenstein et al. (Borenstein et al., 2009; Hedges & Olkin, 1985; Hedges & Vevea, 1998). All analyses were conducted using R open-source software (version 3.6.1; R Core Team, 2016) within the RStudio environment (version 1.1.463; RStudio Team, 2016) using the packages Metafor (Viechtbauer, 2010), plyr (Wickham, 2011), and QuantPsyc (Fletcher, 2012), following guidelines provided by Viechtbauer (2016) and Assink and Wibbelink (2016).

A large portion (53.90%) of studies in our data set reported multiple effect sizes, thereby violating assumptions of independence of effect sizes required for traditional meta-analytic approaches. Dependency among effect sizes has commonly been resolved by choosing one effect size from each study or by averaging across effect sizes (e.g., Borenstein et al., 2009; Lipsey & Wilson, 2001). However, this results in overestimation of the standard errors leading to an increase in Type-2 error (i.e., failure to reject a false null hypothesis; for a review and demonstration, see Moeyaert et al., 2017), and a loss of information about potential meaningful effect size variation as a function of the outcome (Van den Noortgate et al., 2014). To account for nonindependence of effect sizes, we therefore used a three-level meta-analytic model that explicitly models the effects of independent samples and of different effect size indices (for reviews, see Moeyaert et al., 2017; Van den Noortgate et al., 2014; Van den Noortgate & Onghena, 2003). This is the most appropriate and accurate estimation technique for the present data structure in which there are multiple outcomes and where different outcomes can be meaningfully clustered (around the dependent variables described above). The three-level model has the advantage that it splits the total variance into the sampling variance for each effect size (level 1), the within-study variance (level 2), and the between-study variance (level 3) and combines the regression equation at each level into a three-level meta-analytic regression equation (for a review, see Moeyaert et al., 2017)—something which is not possible with alternative estimation techniques such as traditional aggregated (univariate) models or robust variance estimation (RVE; Hedges et al., 2010). All parameters of the summary effect size and the moderators were estimated using the Restricted Maximum-Likelihood Estimator (Viechtbauer, 2005). In industrial and organizational psychology (not so much in other areas), researchers have sometimes used the Hunter and Schmidt method which was designed to correct for measurement in psychometric testing and aims to estimate corrected correlation coefficients (which tend to be larger than the raw uncorrected correlations coefficients) after taking into consideration measurement error (counteracting attenuation due to measurement bias). This method is useful for meta-analyses in industrial and organizational psychology that are often based exclusively on correlational (or longitudinal) studies. This method is not used in other fields (e.g., cognitive or social psychology) that tend to use experimental studies because there is no measurement error in a psychometric sense in experiments where the independent variable is manipulated (even though in experiments as in other methods, there can be various other potential sources of error). A large proportion of the present studies are experimental studies and so for the sake of comparability of effect sizes, the present analyses refrains from conducting additional analyses using this approach.

To assess moderators, we conducted a multilevel mixed-effects model with variance pooled across the groups (in the Metafor package this is akin to subgroup analysis with pooled variance as in Borenstein et al., 2009; Viechtbauer 2010; Viechtbauer 2016). As in the summary model, parameters were estimated with the Restricted Maximum-Likelihood Estimator (Viechtbauer, 2005) and one-sided log-likelihood-ratio tests were used to examine whether the variance (level 2 and level 3) not accounted for by the moderator was significantly different from zero.

We followed recommendations by Kepes et al. (2012) by inspecting publication bias using a number of techniques as reported below. Finally, we calculated a summary effect size for the moderating effect of (a) followers’ social identification and (b) leaders’ group-serving behavior using a three-level random-effects model as in the summary effect model.

Summary effect analyses

A three-level random-effect model across the studies (251 effect sizes, k = 128, N = 32,834) indicated that the mean-weighted relationship was positive, r = .38, 95% CIs [.34, .43], t(250) = 16.00. A forest plot displaying the distribution of the effect sizes (and confidence intervals) across studies can be found in the Online Supplemental Figure S1). The overall heterogeneity in all effect sizes was larger than expected due to sampling error, Q(250) = 6351.39, p < .001, and a substantial amount of the total variation was due to true differences both within-studies (Level 2: I² = 34.26%) and between-studies (Level 3: I² = 61.85%), rather than to sampling error (Level 1: 3.89%). There was significant within-study variance, σ level 2 2 = .032, χ²(1) = 2155.33, p < .001, and between-study variance, σ level 3 2 = .057, χ²(1) = 59.73, p < .001, indicating the likely presence of moderating variables—a possibility we explore further below. The 80% prediction interval (Raudenbush, 2009) indicates a probability of 80% that the true effect size of a new study is likely to fall somewhere within the range of .02 to .66.

Publication bias analyses

We used several techniques to estimate publication bias. For the funnel plot asymmetry test and trim-and-fill method, the multilevel data were aggregated (averaged across effect sizes) as multilevel methods of these tests have not been developed. An integrative summary of the publication bias results are presented in Supplemental Table S2.

Hypothesis tests

Intercorrelations between variables are presented in Table 3. Results from moderation analyses are presented in Table 1.

OPEN IN VIEWER

Table 3.

Intercorrelations between variables.

Variable	1	2	3	4	5	6	7	8
1. Prototypicality conceptualization	—
2. Intergroup context	n/aa	—
3. Prototype strength	.12	.21	—
4. Target leader	.22	.14	.53	—
5. Outcome index	−.01	−.19	.22	.18	—
6. Study methodology	.33	n/ab	.76	.55	.20	—
7. Publication status	.29	.39	−.02	.11	−.08	.03	—
8. Sample size	.28	.03	.18	.11	.12	.15	−.08	—

Note. All correlations based on 251 effect sizes/k = 128 except those pertaining to intergroup context which is based on 79 effect sizes/k = 46; n/a = correlation cannot be calculated because the correlation pertains to two categorical variables where one variable of the pair falls within one level of the respective other. Prototypicality conceptualization: 0 = average, 1 = ideal; Intergroup context: 0 = implicit, 1 = explicit; Prototype strength: 0 = short-lived group, 1 = enduring group; Target leader: 0 = non-formal, 1 = formal; Outcome index: 0 = subjective indices, 1 = behavioral indices; Study methodology: 0 = experimental, 1 = correlational/longitudinal. N for correlation is the lower n of the pair. For n = 251, all r ≥ |.15| significant at .05 level (two-tailed); for n = 79 (pertaining to intergroup context), all r ≥ |.39| significant at .05 level (two-tailed).

^a All studies with information about intergroup context fall within average operationalizations.

^b All studies with information about intergroup context are subsumed within experimental studies.

Moderation by method factors

Theoretical implications

Leader group prototypicality is a core concept in the social identity theory of leadership, and our analysis advances theory by specifying four key insights related to the nature and impact of this variable. First, we established that the relationship between leader group prototypicality and outcomes is stronger when prototypicality is conceptualized in aspirational terms of the ideal-type for the group rather than simply in descriptive terms with reference to a group average. This is consistent with the notion that it is the aspirational element of group prototypicality—who we would be ideally—that renders group prototypicality a basis for leaders’ social influence. This finding is important because it allows us to link leader group prototypicality more explicitly to what is arguably at the core of leadership—motivating and mobilizing group members to pursue objectives that are central to the group (e.g., the team, unit, or organization). Teams and organizations may pursue many objectives, and some are more central to the team or organization than others. Mission, vision, and strategy arguably are most fundamental to organizational goal pursuit (Stam et al., 2014), and these are objectives closely tied to the organization’s aspirations and to the organization’s prototype (van Knippenberg, 2020). Arguably, then, here in particular the aspirational element of leader group prototypicality would be important, because this gives leaders credibility in advocating the pursuit of these group-prototypical objectives.

Second, leader group prototypicality is more strongly related to leadership outcomes with stronger (vs. weaker) prototypes, as indexed by group longevity. Whereas no proxy variable can be equated with the concept, group longevity can be seen as foundational to prototype strength (Oakes et al., 1994). With the caveat, then, that our analysis does not demonstrate this directly, we would suggest that leader group prototypicality has stronger effects in more enduring groups because of more firmly established (stronger) mental representations of group prototypes in such groups (as compared with shorter-lived groups). This finding is important in understanding differences between groups in terms of how strong a basis they provide for social identity-based influence. That said, primary research is needed to provide more direct analyses of prototype strength to further clarify its contribution to leader effectiveness. Such follow-up research would also be valuable to develop two implications of the notion that leader group prototypicality has more pronounced effects for stronger group prototypes. First, it seems plausible that there can be variations in prototype strength between members of the same group. Some group members may hold stronger group prototypes (i.e., better developed mental representations) than others as a function of, for example, their own group tenure (i.e., their status as long-standing members vs. newcomers). Groups may also vary over time in prototype strength not only in terms of group tenure but also in terms of changes to the group that may at least temporarily render the group identity less well-defined (e.g., in times of organizational change; Venus et al., 2019).

The notion of prototype strength as a moderating influence should apply to a broader range of differences within and between groups than just differences captured by group longevity and is a research avenue worth exploring in its own right. In this regard, the notion of prototype strength is also relevant to Reicher and Hopkins’ (2001) observation that leaders are “entrepreneurs of identity” who play an active role in shaping people’s understanding of group identity (see also Haslam et al., 2011; Reicher et al., 2005; Steffens et al., 2014). Combining the notion of prototype strength with these insights in leader agency in group members’ social identity construction, we also propose that it is partly by building the strength of the group prototype (i.e., members’ mental representation of group identity) that group prototypical leaders are able to secure a basis for their identity-based influence.

Third, our understanding of leader group prototypicality is also enriched by the finding that effects are more pronounced for leaders in formal than in nonformal leadership positions. In particular, this shows that there is a distinct leader role component to leader group prototypicality effects and that these effects do not simply revolve around the group prototypicality of any group member who happens to be fulfilling a leadership role. This suggests that there is further scope for theory to clarify and examine further what it is that makes prototypicality powerful for the discharge of formal leader roles. One answer may be that this is where leadership matters most and hence where followers are most sensitive to leadership prototypicality. Support for this hypothesis also begs the question of whether perceived authority (or legitimacy), responsibility for the group, or the task of representing the group (e.g., for elected officials) are active ingredients here and whether the importance of prototypicality for formal leaders varies as a function of these factors. Again, these are important issues to address in primary research in the interests of further developing theory.

We found no support for the prediction that leader group prototypicality has stronger effects when it was explicitly operationalized in terms of intergroup comparison. Clearly in the articles we reviewed, some primary studies highlighted a salient intergroup context, while others did not. Given this variability in primary findings, we recommend future analyses seek to deconstruct this variable further. In particular, it may be fruitful to understand more fully the nature of the intergroup context. For example, in conflictual intergroup contexts, the salience of an out-group may enhance a group prototypicality effect but not have the same effect in more pallid intergroup contexts where the out-group seems more irrelevant to how group members define themselves. Alternatively, it may be noted that even when a measure or experimental induction does not explicitly invoke the intergroup context, such a context may be implied. Indeed, there are strong grounds for thinking that the perception of being a group inherently implies the existence of other groups (Turner et al., 1987). Accordingly, responses to measures or experimental inductions may be infused by the perceived intergroup context even when such context is not made explicit.

Together, then, the above findings provide robust evidence for core propositions of the social identity theory of leadership. Specifically, they show that the main effect of leader group prototypicality does not only hold for subjective leadership evaluations (Barreto & Hogg, 2017) but also for behavioral outcomes that are seen as fundamental toleadership effectiveness (Burns, 1978; House et al., 2001; Kaiser et al., 2008). In this regard, results show that leader group prototypicality exerts no small influence on leadership outcomes. Regarding the magnitude of the summary effect, it is worth noting that the overall effect size of r = .38 that we obtain from our analysis is substantially larger than the average effect size of r = .29 found in a review of 69 leadership meta-analyses summarizing more than 2000 studies on a range of leadership variables and outcomes (Paterson et al., 2016). Furthermore, results also consolidate support for two moderators that have been key propositions in the theory: follower group identification and leader group-serving behavior (e.g., Hogg et al., 2012). Finally, a sign of the robustness of the evidence is that the research that has examined the social identity theory of leadership is characterized by a large proportion of experimental studies that can shed light on causality—something which is not the case for the general evidence base in leadership research (Avolio et al., 2009; Podsakoff & Podsakoff, 2019).

Beyond this, the meta-analytic approach we have taken also allows us to assess the influence of various methodological factors and publication bias. This is important because a cumulative and replicable body of evidence is essential for any field of science and is a necessary foundation for scientific advance (Schmidt & Oh, 2016). In the present case, it is noteworthy that we addressed issues of publication bias by gathering unpublished data and subjecting the literature to analysis using a range of advanced techniques. These provided little indication that research evidence was adversely affected by publication bias.

In sum, then, the present evidence underscores the value of studying leadership as a process that is shaped by shared social identity and leader group prototypicality. At the same time, it is notable that a comprehensive review of leadership research since 2000 by Dinh and colleagues (2014) showed that the social identity analysis of leadership informed only a small minority (4%) of studies in this area. In view of the evidence that leader group prototypicality has stronger than average effects on leadership outcomes (cf. Paterson et al., 2016) and the fact that the core propositions of the theory are well-supported, it would seem that there is a case for leadership research to embrace the social identity theory of leadership and explore ways to integrate the theory with other well-supported perspectives on leadership (Haslam et al., 2011; van Knippenberg & Hogg, 2003).

Implications for practice

To some extent, we need to be careful not to be too bold in identifying implications for practice, given that the tradition in leadership selection and development—the obvious foci for practical implications of leadership research—is more focused on leader traits and behaviors than on follower responses to those leaders. However, as we noted in the previous section, the effect size for leader group prototypicality suggests that prototypicality is far from a trivial determinant of leader outcomes. Indeed, this effect appears to be stronger than that associated with leader personality or behavioral aspects of leadership identified in other meta-analyses (e.g., Judge et al., 2002). Moreover, the solid support for the social identity theory of leadership would suggest that organizations might seek to build leadership effectiveness by selecting and developing group prototypical leaders. However, the social identity analysis of leadership also cautions those members of organizations doing the selecting and doing the developing. This is because the analysis recognizes that the normative content of “who we are” is always actively debated among group members. Empirically, we know this from the in-group self-stereotype literature (Haslam et al., 1998), the schism literature (Sani, 2008), the in-group projection literature (Wenzel et al., 2008), and the normative conflict literature (Crane & Platow, 2010). What this means is that those who appoint (or anoint) themselves to select and develop may have views about what is and is not prototypical of the salient group that is at odds with the views of prospective followers, which requires negotiation at best and risks failure at worst.

In part, then, this selection would entail selecting on the basis of a match with organizational identity and broader associated norms, values, and ambitions in ways that are recognized as procedurally legitimate. In part it would also involve recognizing that leaders need to be entrepreneurs of identity who actively shape the perception of their own and others’ group prototypicality and the group prototypicality of the objectives they advocate (in ways suggested by Reicher & Hopkins, 2001; see also Haslam et al., 2011; Steffens et al. 2014; van Dick et al., 2018). As noted above, this would also include shaping an understanding of group identity to build stronger prototypes. At the same time, the actions that leaders might take to that effect—the focus of leadership development efforts—are as yet poorly understood (for an exception, see Haslam et al., 2017). The accumulated evidence suggests, however, that it may be worth investing in further developing this understanding for practical application.

Limitations and directions for future research

Meta-analyses have clear strengths in systematically synthesizing a body of evidence. At the same time, there are limitations to meta-analysis that we should be mindful of and that also apply to the current analysis. An obvious limitation is that meta-analyses are limited by the available evidence. This means that some relationships can only be studied by inference (e.g., as seen in our attempt to assess the impact of prototype strength by examining the effects of group longevity, which we could encourage future work to follow up with explicit measures and manipulations) and also that some relationships for which theory and evidence exist cannot be studied due to the limited number of effect sizes available for meta-analytic integration (e.g., as with research identifying motives that render followers more sensitive to leader group prototypicality; Pierro et al., 2005). As a result, meta-analysis has to be seen as complementing primary research and conceptual reviews, and a full evaluation of the social identity theory of leadership necessarily takes all three into account.

Meta-analytic tests of new theory also do not render future tests in primary research superfluous. Hypothesis tests based on coding of cross-study differences do not constitute causal evidence and may capture unmeasured variations between studies that covary with the coded variables. There is currently no study directly measuring and testing the effects of the four novel moderators of the relationship between leader group prototypicality and leader outcomes. The present evidence for these effects is strong in the sense that it is based on the quantitative integration of a large number of studies, but it is not based on direct, within-study assessment of these variables. Accordingly, future research to replicate and extend the current findings through within-study assessments of these effects would clearly be valuable.

Below, we outline a number of key questions and limitations that future work should overcome to further our understanding in this field. First, we would note that many conceptualizations and operationalizations of prototypicality captured a mixture of average versus ideal group prototypes. In this regard, there would be value in extending the analysis of the present relationships not only by distinguishing more clearly between these (and providing clear matching conceptualizations and operationalizations) but also by examining whether these have differential impact on different types of outcomes and given particular tasks. Furthermore, it will also be interesting to examine a range of additional questions including how follower responses to different group prototypes are influenced by (a) the size of the discrepancy between current and ideal collective self as well as (b) variations in identity definitions (e.g., the degree to which a group is defined by ideal vs. average elements). Second, speaking to the idea of prototype strength, our examination of this idea was limited by the nature of the data. This would be important to interrogate further both conceptually and empirically and to extend by examining a range of additional questions such as how follower responses to leader group prototypicality are influenced by the degree to which perceptions are shared among group members.

It would also be interesting in future research efforts to expand upon the present analysis by exploring whether, when, and how intergroup contexts affect the impact of leader group prototypicality. For example, future work might examine how the impact of leader group prototypicality is affected by additional factors that influence the salience of intergroup contexts (e.g., specific events or contexts) as well as the nature of the relationship between the in-group and other relevant out-groups (e.g., conflict in values and goals). Regardless of the specific questions that seem most pressing in future work, at a meta-research level, there would also be particular value in experimental work that can disentangle these effects while also establishing causality.