Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

Contemporary Educational Psychology 58 (2019) 186–203

Contents lists available at ScienceDirect

Contemporary Educational Psychology


journal homepage: www.elsevier.com/locate/cedpsych

Teacher self-efficacy profiles: Determinants, outcomes, and generalizability T


across teaching level

Harsha N. Perera , Celeste Calkins, Rachel Part
Department of Educational Psychology and Higher Education, College of Education, University of Nevada, Las Vegas, 4505 Maryland Pkwy, Box 453003, Las Vegas, NV
89154-3003, United States

A R T I C LE I N FO A B S T R A C T

Keywords: There have been considerable advances in understanding teacher self-efficacy in the past two decades. However,
Teacher self-efficacy profiles the core theoretical postulate that teachers may simultaneously possess a variety of teaching-related self-efficacy
Social cognitive theory beliefs at varying levels has not been systematically examined. Adopting a person-centered, multidimensional
Sources of self-efficacy perspective on teacher self-efficacy, the current study aims to identify distinct profiles of teachers’ self-efficacy
Job satisfaction
beliefs in lower and upper secondary teachers and examine the replicability of the profiles across these groups. In
Profile invariance
Multiple-group latent profile analyses
addition, we draw on social cognitive perspectives to posit several predictors of teacher self-efficacy profile
membership, including professional development provisions and needs, mentoring experiences, gender, and
teaching experience, and outcomes of profile membership, including job satisfaction, perceived classroom cli-
mate, and teacher collaboration. Results revealed six teacher self-efficacy profiles, which were found to replicate
entirely across lower and upper secondary teachers. Job satisfaction, classroom climate, and teacher colla-
boration were found to differ as a function of the profiles, and these relations were found to be moderated by
teaching level, suggesting context-differentiated processes in the outcomes of teacher self-efficacy profile
membership. Finally, profile membership was shown to be predicted by professional development needs and
provisions, mentoring experiences, gender, and years of teaching experience. Implications of the results for
teacher self-efficacy theory and research are discussed.

1. Introduction Philipp, & Kunter, 2013), teacher engagement (Durksen, Klassen, &
Daniels, 2017; Granziera & Perera, 2019), job satisfaction (Klassen &
There has been an increased scholarly interest in teacher self-effi- Chiu, 2010), well-being (Zee & Koomen, 2016), and occupational
cacy in recent years (Zee & Koomen, 2016). This has been attributed to commitment (Klassen & Chiu, 2011). Teacher self-efficacy has also been
an increased policy focus on promoting teacher effectiveness and sa- found to be implicated in student-level outcomes, including academic
tisfaction and reducing rates of teacher attrition (Klassen & Tze, 2014). self-efficacy beliefs (Ross, Hogaboam-Gray, & Hannay, 2001), school
Given the considerable proportion of investment in education devoted engagement (Reyes, Brackett, Rivers, White, & Salovey, 2012), and
to teacher compensation, ineffective teachers impose an undue eco- achievement (Caprara, Barbaranelli, Steca, & Malone, 2006). Although
nomic burden on school districts as productivity fails to match invest- this suggests the importance of teacher self-efficacy at several levels of
ment. Furthermore, ineffective teachers become dissatisfied with classroom ecology (Zee & Koomen, 2016), this evidence is limited by
teaching, leading to increased absenteeism, illness, and ultimately at- the assumption that individual teachers possess uniformly high or low
trition (Billingsley & Cross, 1992; Darr & Johns, 2008). Given these levels of their capability beliefs across specific teaching domains. Here,
concerns, scholars and policy-makers have turned their attention to teaching domain refers to core clusters of teaching tasks and responsi-
teacher self-efficacy as a determinant of teacher effectiveness (Klassen bilities in the classroom environment that constitute teachers’ daily
& Tze, 2014) that may explain why some teachers disengage from, activities (Zee, Koomen, Jellesma, Geerlings, & de Jong, 2016) as dis-
become dissatisfied with, and ultimately leave the profession tinguished from academic domains (e.g., math, English). Yet, from a
(Granziera & Perera, 2019). social cognitive theory standpoint (Bandura (1997)), teachers may not
There is accumulating evidence that teacher self-efficacy is a key be equally efficacious across all domains (e.g., classroom management
determinant of instructional quality and student support (Holzberger, vs. instructional strategies) (Goddard, Hoy, & Woolfolk-Hoy, 2000).


Corresponding author.
E-mail address: Harsha.Perera@unlv.edu (H.N. Perera).

https://doi.org/10.1016/j.cedpsych.2019.02.006

Available online 20 February 2019


0361-476X/ © 2019 Elsevier Inc. All rights reserved.
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

Indeed, teachers’ domain-specific capability beliefs are not isolated capability to develop relations with students and promote their moti-
constructs; instead, they co-exist within individuals at different levels. vation and engagement in learning (Tschannen-Moran & Woolfolk Hoy,
As such, differing configurations of domain-specific teacher self-efficacy 2001).
beliefs may yield differential outcomes for teachers. However, little is Although it is assumed in this multidimensional model that teachers
known about how these distinct teacher self-efficacy beliefs combine may be differentially efficacious in each of these domains of teaching,
within teachers, how such combinations can be inferred from data, this theoretical complexity has not been reflected in research. Hitherto,
whether these configurations can be predicted by relevant socio-de- no research has been conducted on how domain-specific teacher self-
mographic variables and learning experiences, and the role of these efficacy beliefs co-exist at different levels within teachers. This dearth
capability belief combinations in outcomes. of research is attributable, at least in part, to evidence for considerable
Accordingly, drawing on social cognitive perspectives, we in- generality in teacher self-efficacy data (Duffin, French, & Patrick, 2012;
vestigated latent profiles of teacher self-efficacy representing distinct Perera, Wiens, McIlveen, Calkins, & McLennan, 2019). For instance,
configurations of teachers’ self-efficacy beliefs. Plausible predictors of studies have found support for a unidimensional structure of teacher-
the probability of profile membership were also examined, including self-efficacy data obtained from the widely-used Teacher Sense of Self-
years of teaching experience, gender, involvement in mentoring and Efficacy Scale (TSES), which is itself predicated on the three-dimen-
professional development activities, and perceptions of the need for sional perspective (Duffin et al., 2012; Tschannen-Moran & Woolfolk
professional development. In addition, job satisfaction, classroom dis- Hoy, 2001). Even where a three-factor structure has been supported
ciplinary climate, and teacher collaboration were examined as out- (Klassen and Chiu, 2010; Klassen et al., 2009; McLennan, McIlveen, &
comes of teacher self-efficacy profile membership. Notably, as the dis- Perera, 2017), factor correlations tend to be so high as to undercut the
tinct configurations of teachers’ self-efficacy beliefs, and their relations dominant multidimensionality perspective. This evidence for con-
with these predictors and outcomes, may differ as a function of teaching siderable generality has resulted in empirical practices of combining
level (Klassen & Chiu, 2011), we also examined the generalizability of domain-specific teacher self-efficacy dimensions into single composite
teacher self-efficacy profiles and their relations with the predictors and scores or modeling higher-order teacher self-efficacy factors (Granziera
outcomes across lower and upper secondary school teachers. & Perera, 2019; McLennan et al., 2017). The assumption underlying
these approaches is that individual teachers possess uniformly high or
1.1. Teacher self-efficacy: theoretical overview low self-efficacy beliefs across domains, which is inconsistent with the
theoretical position that teachers may be differentially efficacious in
Teacher self-efficacy theory (Tschannen-Moran, Woolfolk Hoy, & specific domains (Goddard et al., 2000).
Hoy, 1998) is predicated on Rotter’s (1966) attribution-based locus of One reason for inflated associations may be the improper dis-
control theory and, to a greater extent, social cognitive theory (Bandura aggregation of general and specific variance in teacher self-efficacy data
(1997)). From this integrative perspective, teacher self-efficacy is in previous work (Duffin et al., 2012; Klassen et al., 2009). Teachers’
viewed as self-referent judgments of capability to organize and execute ratings of their self-efficacy beliefs may simultaneously reflect a general
actions required to successfully perform teaching tasks and positively overarching teaching confidence construct together with self-referent
impact student learning. Notably, in line with this perspective, teacher judgments of capability with respect to specific domains (e.g., class-
self-efficacy beliefs reflect judgments of personal teaching capabilities, room management, instructional strategies) (Morin and Marsh (2015).
based on an appraisal of strengths and deficits, and judgments about the This position aligns with the theoretical view that teacher self-efficacy
requirements of domain-specific teaching tasks, which include apprai- beliefs encompass self-perceptions of personal teaching capabilities in
sals of external constraints and resources (Tschannen-Moran et al., addition to judgments about the requirements of domain-specific
1998). The convergence of judgments of personal teaching capabilities teaching tasks (Tschannen-Moran et al., 1998). At least some compo-
with task-specific requirements in the conceptualization of teacher self- nent of these self-perceptions may reflect individuals’ global judge-
efficacy suggests that teachers may not necessarily be efficacious across ments of their capability to perform their professional role as a teacher
all teaching domains. A central postulate of this theoretical perspective generally based on an assessment of personal teaching capabilities (i.e.,
is that teachers may be expected to feel more or less efficacious in “I can perform my role as a teacher”) quite apart from capability
specific domains (Goddard et al., 2000; Tschannen-Moran et al., 1998; judgments that are tied to specific tasks or domains (e.g., “I can im-
Zee & Koomen, 2016). For instance, a secondary school science teacher plement a classroom reward system”), which are informed by both
may view herself as a capable teacher; she may feel highly efficacious assessments of personal teaching capabilities and task-specific re-
for developing and implementing instructional routines and engaging quirements. This differentiation of global levels of confidence for per-
her students in this learning process; yet, she may feel highly in- forming a professional role from task-specific self-efficacy beliefs may
efficacious in managing classroom behaviors. This teacher’s self-effi- be particularly salient in teaching where there tends to be a greater
cacy reflects information drawn from judgments of personal teaching decoupling of expected professional role and enacted domain-specific
capabilities and perceptions of the requirements of specific domains tasks (Aiken et al., 2001).
and her strengths and weaknesses with respect to those domains. In factor models, the lack of statistical control for generality induces
Emerging from this theoretical view is the three-dimensional model upwardly biased estimates of relations as the generality underlying all
of teacher self-efficacy (Tschannen-Moran & Woolfolk Hoy, 2001). This items is absorbed by (inflated) correlations among domain-specific di-
model is designed to accommodate the centrality of domain-specificity mensions of teacher self-efficacy beliefs (Morin, Arens, & Marsh, 2016).
to teacher self-efficacy judgments while avoiding the efficiency pitfalls This failure to disentangle generality from specificity may obscure
of microscopic operationalization. In this regard, this multidimensional distinct intra-individual patterns of domain-specific teacher self-effi-
model is intended to reflect an intermediate level of domain-specificity cacy beliefs. Comparably, Morin and Marsh (2015) noted that failure to
on a continuum ranging from teachers’ general perceptions of con- account for teachers’ global levels of effectiveness makes it difficult to
fidence to perform their professional role to highly task-specific cap- isolate specific areas of perceived strengths and weaknesses. Consistent
ability beliefs (Zee et al., 2016). From this multidimensionality per- with the position for generality and specificity, Perera et al. (2019)
spective, teacher self-efficacy reflects self-referent judgments of found support for a bifactor structure of teacher self-efficacy data,
capability in three teaching domains. Self-efficacy for classroom man- among competing unidimensional, correlated-factors, and higher-order
agement refers to teachers’ perceived capability to establish and main- models, drawn from the short-form of the TSES (TSES-SF) in Australian
tain order in the classroom. Self-efficacy for instructional strategies re- and US teachers. The bifactor structure was characterized by well-de-
flects the perceived capability to use alternate methods in teaching and fined and orthogonal general teacher confidence and specific self-effi-
assessment. Self-efficacy for student engagement refers to the perceived cacy for classroom management, instructional strategies, and student

187
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

engagement factors. Notably, the orthogonality among the general instance, one profile of teachers may be characterized by uniformly
confidence and domain specific self-efficacy beliefs is consistent with high levels of the self-efficacy dimensions whereas another profile may
the theoretical view that teachers may be differentially efficacious in contain teachers with uniformly low levels of these dimensions. Profiles
specific teaching domains. ordered in terms of quantitative level differences can be adequately
In this study, we extend the variable-centered work of Perera et al. represented using variable-centered techniques (Marsh et al., 2009;
(2019) by adopting a person-centered perspective in investigating Muthén, 2001). Contrariwise, profiles that are qualitatively different
profiles of teachers’ multidimensional self-efficacy involving distinct have distinct patterns of teacher self-efficacy dimensions on which they
configurations of their general teaching confidence and domain-specific are high, medium, or low (i.e., shape-differentiation). For example, one
self-efficacy beliefs. We do so by drawing on the bifactor representation profile may be characterized by higher levels of instructional strategy
of teacher self-efficacy supported in Perera et al. (2019) that teases self-efficacy but lower levels of self-efficacy for classroom management
apart generality and specificity. In person-centered representations, and student engagement. These differences in profile shape cannot be
using preliminary bifactor measurement models may be necessary to straightforwardly accommodated by variable-centered methods. A
disentangle distinct patterns of self-efficacy dimensions on which in- person-centered approach would allow for the first direct empirical
dividuals may be high, medium, or low (i.e., shape-differentiated pro- examination of the postulate that teachers may simultaneously possess
files) from level-differentiated profiles reflecting the tendency for in- multiple capability beliefs at varying levels (Goddard et al., 2000).
dividuals to be uniformly high, medium, or low across all teacher self- Accordingly, drawing on a person-centered perspective, we aimed to
efficacy dimensions (Morin and Marsh, 2015). This is because gen- investigate whether there are quantitatively and qualitatively distinct
erality in data makes the identification of shape-differentiated profiles profiles of teacher self-efficacy beliefs.
more difficult as strong level effects, induced by generality, generate
strong level differences between profiles (Morin and Marsh, 2015). 1.3. Outcomes of profile membership

1.2. Teacher self-efficacy: towards a person-centered representation Distinct configurations of teacher self-efficacy beliefs may be dif-
ferentially implicated in important teacher outcomes. From a social
The theoretical position that teachers may not be equally efficacious cognitive perspective, teachers experience favorable outcomes, such as
across all teaching domains suggests that they may emerge from dis- satisfaction, greater collaboration, and heightened quality of classroom
tinct unobserved subpopulations presenting unique configurations of processes (Zee & Koomen, 2016), when they possess a strong sense of
self-efficacy dimensions. However, this presumed heterogeneity in self-efficacy to perform required tasks (Klassen and Chiu, 2010). Tea-
teacher self-efficacy beliefs has not been investigated in prior work; chers’ self-efficacy beliefs influence the decisions they make about
rather, extant research has been limited to examining teacher self-ef- choosing, investing attention and effort in, and persisting with goal-
ficacy from a variable-centered perspective. Although this variable- directed activity (Granziera & Perera, 2019). However, no work has
centered research has been central to establishing the nomological net been conducted to investigate the way in which distinct intraindividual
of teacher self-efficacy (Zee & Koomen, 2016), and demonstrating the patterns of teachers’ self-efficacy beliefs are associated with important
mechanisms through which teacher self-efficacy is related to mean- outcomes. Yet, as teachers are expected to be differentially efficacious
ingful outcomes (Granziera & Perera, 2019), this approach assumes that across domains (Goddard et al., 2000), investigating whether different
all individuals in a sample belong to the sample population and, by configurations of these beliefs differentially relate to outcomes would
implication, share the same set of parameter estimates and neglects the represent an important advancement to both teacher self-efficacy
possibility of unobserved subpopulations. theory and prior work by providing a fuller picture of the way in which
A holistic, multidimensional person-centered approach may yield self-beliefs interact to shape teachers’ perceptions of their work en-
important insights into teacher self-efficacy theory, namely how tea- vironment. Furthermore, examining such profile-outcome relations
chers’ general capability and domain-specific self-efficacy beliefs provides a critical test of profile validity (Perera & McIlveen, 2018).
meaningfully interact to reflect an integrated system of their confidence Per social cognitive perspectives (Lent & Brown, 2006; Zee &
beliefs. Indeed, from a theoretical standpoint, teachers possess differ- Koomen, 2016), we consider job satisfaction, teacher collaboration, and
ential levels of these self-efficacy beliefs, and it is likely that a combi- (positive) classroom disciplinary climate as outcomes of profile mem-
nation of these self-beliefs, rather than domain-specific beliefs uniquely bership. Job Satisfaction is a positive cognitive-affective state resulting
considered, governs how they navigate teaching and learning en- from a favorable evaluation of work-related experiences (Lent & Brown,
vironments. For instance, a teacher’s high self-efficacy for instructional 2006). Research shows that teachers with higher global self-efficacy
strategies, considered in isolation, reflects his or her confidence in using experience greater work satisfaction (Caprara et al., 2006; Skaalvik &
a variety of assessment strategies, providing alternative explanations of Skaalvik, 2014; Granziera & Perera, 2019). Beyond global self-efficacy,
content, and adjusting lessons to suit different abilities (Tschannen- significant relations of teacher self-efficacies for classroom management
Moran & Woolfolk Hoy, 2001). Possessing a high level of self-efficacy and instructional strategies with job satisfaction have been found
for instructional strategies may, prima facie, suggest that this teacher is (Klassen and Chiu, 2010). Teacher collaboration refers to cooperative
more likely to use learner-centered and constructivist approaches in practices and activities in which teachers are engaged towards
their teaching (Dunn and Rakes (2011)). However, when combined achieving common educational goals (Ning, Lee, & Lee, 2015). Re-
with low levels of classroom management and student engagement self- search shows that efficacious teachers are better able to collaborate
efficacy, the potentially beneficial effects of high self-efficacy for in- with colleagues in achieving shared educational goals (Caprara et al.,
structional strategies on instructional practices may be inhibited by 2003; Goddard & Kim, 2018). For instance, Goddard and Kim (2018)
doubts about abilities to control disruptive behavior and engage stu- found that elementary teachers’ self-efficacy was positively associated
dents in the learning process. Such a teacher, notwithstanding higher with their collaboration on instructional policy, formal collaboration,
self-efficacy for instructional strategies, may revert to traditional and informal collaboration. Finally, positive classroom disciplinary
teaching formats that allow for stricter control of the learning en- climate refers to perceptions of orderliness and predictability in the
vironment. implementation of classroom activities (Ning, Van Damme, Van Den
A person-centered approach allows for the identification of distinct Noortgate, Yang, & Gielen, 2015). For efficacious teachers, perceptions
subpopulations of teachers that differ both quantitatively and qualita- of positive classroom climate may be attributed to (a) being better able
tively on the teacher self-efficacy dimensions. Profiles that differ to cope with problem behavior (Almog and Shechtman, 2007), (b)
quantitatively vary with respect to absolute levels of the teacher self- using more constructive conflict management strategies (Morris-
efficacy dimensions (Marsh, Lüdtke, Trautwein, & Morin, 2009). For Rothschild & Brassard, 2006), and (c) adopting more learner-centered

188
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

teaching approaches (Nie, Tan, Liau, Lau, & Chua, 2013). Indeed, ef- theory (Lent et al., 2002). Gender-role socialization experiences, in-
ficacious teachers have been shown to perceive less student mis- cluding the divergent communal and agentic social roles of women and
behavior (Tsouloupas, Carson, Matthews, Grawitch, & Barber, 2010). men, respectively (Perera, 2016), may inform differences in how tea-
Although this literature is suggestive of relations of teacher self-efficacy chers perceive their gender roles in school settings. These differences
with satisfaction, collaboration, and climate, it is limited by in- may have implications for teachers’ beliefs about their ability to per-
vestigating the unique and additive relations of teacher’ self-efficacy form their professional roles generally as well as in specific domains
beliefs with these outcomes. We extended this work by investigating (Drudy, 2008; Ross, Cousins, & Gadalla, 1996; Skaalvik & Skaalvik,
how a range of multidimensional teacher self-efficacy beliefs combine 2007). Like gender, initial work on the associations between years of
in their relations with these outcomes. experience and teacher self-efficacy yielded inconsistent results
(Wolters & Daugherty, 2007; Woolfolk Hoy & Spero, 2005). Klassen and
1.4. Predictors of profile membership Chiu (2010) clarified these inconsistent results by finding a curvilinear,
inverted U-shaped association, such that self-efficacy dimensions ex-
A challenge for teacher self-efficacy theory development is under- perienced increases up to mid-career followed by declines into later
standing how a teacher’s configuration of multidimensional self-effi- career stages. These results have been explained in terms of the teacher
cacy beliefs may emerge from their learning experiences and socio- career development model of Huberman (1989). However, this work
demographic characteristics. According to teacher self-efficacy theory, does not account for the possible existence of distinct subpopulations of
teachers’ self-efficacy beliefs are developed as they interpret informa- teachers who may show differential capability belief configurations as a
tion from mastery experiences, vicarious experiences, social persua- function of years of experience. Indeed, Bandura (1997) suggested that
sions, and physiological and emotional arousal (Bandura, 1997; the expected decline in self-efficacy beliefs in later career stages may
Tschannen-Moran et al., 1998). From this perspective, teachers’ pro- not be universally experienced, raising the possibility of distinct sub-
fessional development participation and mentoring experiences may be populations characterized by differential self-efficacy configurations
viewed as learning experiences that serve as sources of mastery ex- that may be tied to different career stages. In this study, we examined
periences, vicarious learning, and social persuasions that shape and the extent to which teachers’ gender and years of experience predict the
reshape their role-specific self-efficacy beliefs. Consistent with this probability of membership into the teacher self-efficacy profiles.
view, accumulating evidence demonstrates positive effects of profes-
sional development participation on teachers’ self-efficacy beliefs 1.5. Teaching level and teacher self-efficacy
(Carney, Brendefur, Thiede, Hughes, & Sutton, 2016; Henson, 2001;
Ross & Bruce, 2007; Watson, 2006; Tschannen-Moran & McMaster, Teaching level may also be implicated in profiles of teacher self-
2009). Furthermore, research also shows benefits of mentoring ex- efficacy and their relations with the theorized predictors and outcomes.
periences for mentee (LoCasale-Crouch, Davis, Wiens, & Pianta, 2012; Teachers’ work contexts, including teaching level (Klassen and Chiu
Richter et al., 2013) and mentor (Holloway, 2001) teachers. In this (2011)), are posited to be implicated in their self-efficacy beliefs per
study, we investigated whether membership into the teacher self-effi- models of teacher motivation and engagement (Bandura, 1997; Lent
cacy profiles differs as a function of (a) serving as a mentor or mentee et al., 2002; Perera, Vosicka, Granziera, & McIlveen, 2018). The small
and (b) participation in professional development activities in the fol- amount of available evidence shows that teachers of elementary school-
lowing five critical domains of teacher practice: subject field knowl- aged students report higher self-efficacy than middle-and-high-school
edge; pedagogical competencies; assessment practices, classroom teachers (Fives and Buehl, 2009; Klassen and Chiu, 2010; Wolters &
management practices; and individualized learning. Daugherty, 2007). Within specific education levels, Klassen and Chiu
From a social cognitive standpoint, teachers’ perceptions of pro- (2010) found that teachers of kindergarten students reported higher
fessional development needs may also inform their configurations of self-efficacy for classroom management and student engagement than
multidimensional teaching self-efficacy. Teachers form perceptions of those from other elementary grades. However, within secondary
their needs for specific professional development as they encounter schools, a small amount of evidence shows that global levels of teacher
novel teaching tasks and evaluate the requirements of these tasks with self-efficacy are higher as secondary grade level increases (Raudenbush
respect to their capabilities in the self-efficacy development process et al., 1992). Although this variable-centered evidence is suggestive of
(Tschannen-Moran et al., 1998). Higher levels of perceived professional differences in teacher self-efficacy beliefs across teaching level, the way
development needs, particularly when those needs remain unmet, may in which these differences are reflected in distinct profiles of teacher
be viewed as self-efficacy-relevant barriers that undermine teachers’ self-efficacy remains unclear. One possibility is the emergence of pro-
domain-specific self-efficacy beliefs (Lent, Brown, & Hackett, 2002). files that are teaching-level-specific. A second possibility is that the
Consistently, prior research shows that cognitive appraisals of en- structure of the profiles may be invariant across teaching levels; how-
vironmental resources and barriers inform self-efficacy beliefs, in- ever, the probability of profile membership may vary across teaching
cluding in teachers (Tschannen-Moran & Woolfolk Hoy, 2007). How- levels. We investigated the extent to which the teacher self-efficacy
ever, this work has been limited to investigating the unique and profiles replicate across lower and upper secondary teachers. We also
additive relations of perceived barriers with teacher self-efficacy be- examined the invariance of relations of teacher self-efficacy profile
liefs. In this study, we investigated the relations of teachers’ perceptions membership with the theorized predictors and outcomes across these
of professional development needs in the five domains of teacher teaching levels.
practice noted above—subject field knowledge, pedagogical compe-
tencies, assessment practices, classroom management practices, and 1.6. The present study
individualized learning—with their configurations of multidimensional
self-efficacy beliefs. Drawing on social cognitive perspectives, this study aimed to ex-
Finally, socio-demographic factors are posited to be implicated in plore suspected heterogeneity in teacher self-efficacy data by ex-
teacher self-efficacy beliefs per social cognitive perspectives (Lent et al., amining profiles of teachers’ hierarchical and multidimensional self-
2002; Klassen and Chiu, 2010). Two such factors are gender and years efficacy beliefs. In doing so, we addressed a gap in the teacher self-
of teaching experience (Bandura, 1997; Klassen and Chiu, 2010; efficacy literature concerning the largely untested proposition that
Wolters & Daugherty, 2007). Although research on gender differences teachers can simultaneously hold multiple self-efficacy beliefs at dif-
in teacher self-efficacy has yielded inconsistent results (Brennan & ferent levels. As teaching level has been theorized to be an important
Robison, 1995; Klassen and Chiu, 2010; Raudenbush, Rowan, & contextual factor in social cognitive models of teacher motivational
Cheong, 1992), gender effects should be expected per social cognitive beliefs (Klassen and Chiu, 2010), we also investigated the degree to

189
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

Fig. 1. Depiction of the models tested in the present study. The hexagon depicts the latent categorical variable representing C latent profiles estimated from the
teacher self-efficacy indicators. In this model, the paths linking the predictors to the latent categorical variables implies that the predictors predict the likelihood of
profile membership. Paths from the Professional Development Participation and Professional Development Needs predictors are stylistic and designed to suggest that
all five domains of professional development examined are specified to predict the probability of profile membership. The paths from the latent categorical variable
to the outcomes indicate that the likelihood of profile membership predicts the outcomes. The paths from teaching level to the latent categorical variable as well as to
the paths linking the predictors and outcomes with the latent categorical variables indicate that (a) the profiles may differ as a function of teaching level and (b) the
predictor and outcome relations may be moderated by teaching level, respectively. Note: SE = Self-efficacy.

which the profiles generalize across lower and upper secondary tea- 2. Method
chers. Furthermore, because teachers are expected to be more or less
efficacious in different domains, it may be that it is a combination of 2.1. Participants and procedure
these beliefs that has implications for teacher outcomes and can be
predicted by meaningful socio-demographic factors and learning ex- We relied on Australian data from the TALIS 2013 study (Rutkowski
periences. Accordingly, we (a) examined teacher job satisfaction, per- et al., 2013). TALIS 2013 is an international study examining teachers’
ceived classroom disciplinary climate, and collaborative activity as teaching practices, learning environments, and working conditions
outcomes of the profiles, (b) investigated professional development conducted under the auspices of the Organisation for Economic Co-
participation and needs, mentoring activities, gender, and years of ex- operation and Development (OECD). The TALIS 2013 comprises data
perience as predictors of profile membership, and (c) considered the from 33 countries and sub-national entities, including data from over
degree to which these profile-outcome and predictor-profile relations 10,000 schools and approximately 170,000 teachers (OECD, 2014).
generalize across teaching level. Fig. 1 depicts the models to be tested in Country-specific datasets constitute data obtained from nationally re-
the present study. presentative samples (Scherer, Jansen, Nilsen, Areepattamannil, &
As there has been, to the authors’ knowledge, no prior person-cen- Marsh, 2016). All participating countries surveyed lower secondary
tered research on teacher self-efficacy beliefs, we adopt an exploratory teachers (i.e., ISCED Level 2). In addition, the survey of ISCED Level 1
approach guided by the following research questions: (i.e., Primary) and Level 3 (i.e., Upper secondary) teachers was optional
for countries. The TALIS 2013 involved the collection of data in six
Research Question 1 (RQ #1): Are there quantitatively and qua- areas as follows: (a) self-efficacy and job satisfaction; (b) learning en-
litatively distinct profiles of teacher self-efficacy beliefs? vironment; (c) appraisal and feedback; (d) teaching practices and
Research Question 2 (RQ #2): Do the retained profiles generalize classroom environment; (e) development and support; and (f) school
across lower and upper secondary teachers? leadership. Data on key socio-demographics were also collected (e.g.,
Research Question 3a (RQ #3a): Do teacher job satisfaction, gender, age, years of experience). Following the survey development
classroom disciplinary climate, and teacher collaboration differ as a and field trial, Australian teachers responded to the TALIS items in the
function of profile membership? fourth quarter of 2012. The full battery comprised 50 items that re-
Research Question 3b (RQ #3b): Do the relations of profile quired approximately 45–60 min to complete. Participation was vo-
membership with the teacher outcomes generalize across lower and luntary, and participants completed the survey online. Completion and
upper secondary teachers? return of the survey was indicative of implied consent.
Research Question 4a (RQ #4a): Can profile membership be The Australian sample from TALIS 2013 used in this study com-
predicted by professional development participation, perceived prises 37351 teachers working in 247 schools. The mean age of
professional development needs, mentoring activities, gender, and
years of experience?
Research Question 4b (RQ #4b): Do the predictor-profile relations 1
Please note that the present sample of 3735 teachers includes 1915 lower-
generalize across lower and upper secondary teachers? secondary teachers constituting the Australian sample in Perera et al. (2019)

190
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

participants was 43.78 (SD = 11.48), and 58.2% (n = 2174) of the activities in these domains using a dichotomous response scale
sample was female. Participants reported diverse teaching experience, (0 = Yes, 1 = No).
ranging from less than a year to 49 years (M = 17.05, SD = 11.16). Perceived professional development needs. Teachers’ perceived
ISCED2 teachers constituted 51.3% (n = 1915) of the sample whereas need for professional development was assessed using five items cor-
the remaining teachers were ISCED 3 teachers.2 Participation rates for responding to the professional development domains noted above. For
Australian teachers in participating schools at ISCED 2 and ISCED 3 each domain, teachers rated the extent to which they perceive a current
were 86.8% and 83.9%, respectively (OECD, 2014). The Australian need for professional development using a four-point response scale,
sample comprised schools from all Australian states and territories and ranging from 1 (No need at present) to 4 (High level of need).
included schools from the Government, Catholic, and Independent Mentoring activities. Teachers’ participation in mentoring activ-
sectors as well as metropolitan, rural, and remote regions. ities was indexed by two items from the TALIS 2013. First, teachers
indicated, on a dichotomous response scale (0 = Yes, 1 = No), whether
they currently have an assigned mentor to support them (i.e., Mentee).
2.2. Measures For the second item, teachers indicated, on a dichotomous scale
(0 = Yes, 1 = No), whether they serve as an assigned mentor for one or
Teacher self-efficacy. Teacher self-efficacy was measured using a more teachers (i.e., Mentor).
self-report scale based on the TSES-SF (Tschannen-Moran & Woolfolk Socio-demographics/Contextual variables. In the TALIS 2013,
Hoy, 2001). The instrument is predicated on the multidimensional teachers reported their gender on a binary response scale (0 = Female,
teacher self-efficacy perspective proposed by Tschannen-Moran and 1 = Male). Years of teaching experience was indexed as the number of
Woolkfolk Hoy, and is designed to measure self-efficacy for classroom years’ teachers reported working in the profession cumulatively.
management, instructional strategies, and student engagement. The Finally, teaching level was indexed as a dichotomous variable
measure comprises 12 items, rated on a four-point Likert-type scale (0 = lower secondary, 1 = upper secondary).
(1 = Not at all to 4 = A lot), with four items measuring each purported
factor. Scores generated from the measure have been shown to be re-
2.3. Statistical analyses
liable and validity evidence has also been obtained (OECD, 2014). Scale
items and reliability estimates in this sample are shown in Appendix A.
Analyses were conducted in four phases. First, preliminary multiple-
Job satisfaction. Teachers’ job satisfaction was measured using the
group bifactor exploratory structural equation modeling analyses
Satisfaction with Current Work Environment scale administered in the
(BESEM) of the TSES-SF data across the lower and upper secondary
TALIS 2013 (OECD, 2014). The scale comprises four items, rated on a
teacher samples were conducted to obtain factor scores on the self-ef-
four-point Likert-type scale (1 = Strongly Disagree to 4 = Strongly
ficacy dimensions from the most invariant measurement model to serve
Agree), designed to measure teachers’ cognitive-affective evaluations of
as LPA indicators. Factor score mixture indicators from multi-group
satisfaction with work. Reliability and validity evidence of the scale
BESEM models should be preferred for four reasons (Morin et al., 2017;
scores has been obtained (OECD, 2014). Scale items and reliability
Morin and Marsh, 2015). First, factor score indicators give greater
estimates in the present sample are shown in Appendix A.
weight to more reliable items and, in this regard, provide partial control
Classroom disciplinary climate. Teachers’ perceptions of the dis-
for measurement errors. Second, factor scores can be generated based
ciplinary climate of their classrooms was appraised using the TALIS
on standardized information, thereby facilitating profile labeling and
2013 Classroom Disciplinary Climate Scale (OECD, 2014). This scale
interpretability. Third, factor-scores based on BESEM models accom-
comprises four items, rated on a four-point Likert-type scale
modate construct-relevant multidimensionality due to (a) general and
(1 = Strongly Disagree to 4 = Strongly Agree), and is designed to measure
specific constructs underlying item data and (b) item fallibility (Perera
teachers’ perceptions of the extent of orderliness and composure in the
et al., 2019; Scherer et al., 2016). Notably, Morin et al. (2017) note
classroom. Evidence of reliability and validity of the scale data has been
that, where there is a general construct underlying data, as is the case
obtained (OECD, 2014). Scale items and reliability estimates for this
with the TSES-SF data, LPA based on factors scores obtained from bi-
sample are shown in Appendix A.
factor models results in an appropriate disaggregation of shape and
Teacher collaboration. Teachers’ collaborative activity was mea-
level effects. Finally, TSES-SF factor scores obtained from multiple-
sured using the TALIS 2013 Exchange and Coordination for Teaching
group models of invariance ensure comparability of the self-efficacy
Scale. This scale contains four item, rated on a six-point Likert-type
scores across the independent samples (Morin, Arens et al., 2016;
scale (1 = Never to 6 = Once a week or more), intended to measure the
Perera & McIlveen, 2018).
extent of teachers’ professional cooperative practices and activities in
For the BESEM measurement model, each TSES-SF item was speci-
the service of achieving common educational goals. Scores from the
fied to have a primary loading on one of the three self-efficacy di-
instrument have been shown to be reliable and valid (OECD, 2014).
mensions as well as a primary loading on the general teacher self-effi-
Scale items and reliability estimates in this sample are shown in Ap-
cacy factor. In addition, the solution was rotated using the target
pendix A.
(orthogonal) rotational criterion with cross-loadings “targeted” to be
Professional development participation. Participation in profes-
approximately zero but not restricted to zero (Asparouhov & Muthén,
sional development over the previous 12 months up to the TALIS as-
2009; Perera, McIlveen, Burton, & Corser, 2015). The target rotational
sessment was operationalized using five items corresponding to parti-
criterion is particularly useful as there is an expectation for construct-
cipation in the following five professional development domains:
relevant multidimensionality due, in part, to item fallibility. Condi-
subject field knowledge; pedagogical competencies; assessment prac-
tional on the acceptable fit of the BESEM model in each sample, the
tices; classroom management; individualized learning. Teachers in-
invariance of the TSES-SF data was examined across lower and upper
dicated whether they participated in professional development
secondary teachers. These multiple-group invariance tests were con-
ducted in line with Millsap and Yun-Tein’s (2004) taxonomy of in-
(footnote continued) variance tests for models based on ordered categorical data adapted for
and available as part of the public-use TALIS database. ESEM (Perera et al., 2015). This involves the sequential testing of
2
ISCED Level 2 denotes lower secondary years (As at 2013, this corresponds configural invariance and the invariance of item factor loadings,
to years 7–10 in the Australian Capital Territory, New South Wales, Tasmania, thresholds,3 uniquenesses, factor variances and covariances, and factor
and Victoria; years 8–10 in the Northern Territory, Queensland, South
Australia, and Western Australia). ISCED Level 3 denotes upper secondary years
3
(i.e., years 11–12 in Australian schools). When ordinal data are examined, both item thresholds and intercepts are

191
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

means (Perera et al., 2015).4 Phrase 3 of the analyses involved tests of profile solution general-
Analyses of the preliminary measurement structures were con- izability across teaching levels. These tests were conducted in line with
ducted using Mplus 7.4 (Muthén & Muthén, 1998–2015). Solutions the taxonomy of LPA similarity tests proposed by Morin, Meyer,
were estimated using weighted least squares with diagonal weight Creusier, and Biétry (2016), comprising sequential and comparative
matrices for the estimation of parameters and full weight matrix for the tests of configural similarity, structural similarity, dispersion similarity,
standard errors and mean-and-variance-adjusted chi-square tests sta- and distributional similarity. Configural similarity can be inferred from
tistic. In Mplus, this estimation routine is operationalized as the the single-sample LPA tests to the degree that the sample-specific
Weighted Least Squares Mean-and-Variance adjusted (WLSMV) esti- analytic solutions converge in the number of profiles identified. How-
mator. As the TALIS uses a cluster sampling design, with teachers ever, configural similarity also requires the simultaneous estimation of
nested within schools, we used the complex design-based correction of the retained k-profile model in both groups, which serves as a baseline
standard errors in Mplus to account for these dependencies model against which the more restrictive profile similarity models are
(Asparouhov, 2005). Also, given different selection probabilities in the compared. From configural similarity, the model of structural similarity
sampling of schools and teachers, all analyses were based on the TALIS can be tested via constraining the within-profile indicator means to
final teacher weights (i.e., TCHWGT) that incorporate sub-weights ac- equality across the groups. Conditional on support for structural simi-
counting for different probabilities of being selected as a school and larity, dispersion similarity can be examined by imposing equality
being selected as a teacher within a country (OECD, 2014). Fit assess- constraints on the indicator variances across groups, thereby providing
ment was inclusive and involved an evaluation of fit indices, parameter tests of the equality of the within-profile variability of the indicators
estimates, and alternative models. As the χ2 can be oversensitive to across samples. Finally, the model of distributional similarity requires
minor model misspecifications given even moderate-sized samples and additive equality restrictions imposed on the class probabilities across
contains a restrictive hypothesis test (i.e., exact fit), three approximate group, testing whether the relative sizes of the profiles are equal across
fit indices were considered: RMSEA, less than 0.050 and 0.080 for close groups (Morin, Meyer, et al., 2016).
and reasonable fit; Comparative fit index (CFI) and Tucker-Lewis Index Finally, we tested the postulated outcomes and predictors of profile
(TLI), > 0.900 and 0.950 for acceptable and excellent fit, respectively. membership across teaching level. For the outcome tests, per Morin,
For nested model comparisons, because the adjusted χ2 difference (MD Meyer et al.’s (2016) taxonomy of profile similarity tests, two models
Δχ2) test appropriate for the WLSMV estimator also tends to be sensi- were examined. An initial model was estimated in which the outcomes
tive to even trivial differences, changes in the CFI (ΔCFI) and RMSEA were incorporated into the most-invariant model retained in Phase 3 of
(ΔRMSEA) were primarily used. A decrease in the CFI and increase in the analysis. In this model, mean level differences across pairs of pro-
the RMSEA of less than 0.010 and 0.015, respectively, are indicative of files within the teaching level groups were tested via the multivariate
support for a more restrictive model (Chen, 2007; Cheung and delta method operationalized using the MODEL CONSTRAINT com-
Rensvold, 2002). mand in Mplus (Raykov & Marcoulides, 2004). This model was com-
Phase 2 of the analyses involved LPA with factor scores from the pared against a more parsimonious model of explanatory similarity in
most invariant BESEM model serving as mixture indicators. LPA is a which the within-profile means on the outcomes were constrained to
statistical model for modeling unobserved population heterogeneity. equality across groups. Support for the model of explanatory similarity
LPA models assume that samples drawn from a heterogeneous popu- indicates that mean levels on the outcomes in specific profiles are equal
lation generate data that are a mixture of k profile-specific distributions across the teaching levels. An important assumption of LPA is that the
where k is the number of profiles (k > 1). LPA models capture het- unconditional LPA solutions remain unchanged by the inclusion of
erogeneity by grouping individuals into latent profiles based on simi- covariates (Marsh et al., 2009). To ensure the profiles remained un-
larities in their response variable data. The LPA models were initially changed by the inclusion of distal outcomes, these models were esti-
tested separately in each sample to determine if the same number of mated using start values from the most-invariant model retained in
profiles could be identified. Models including one to eight profiles were Phase 3 of the analysis. Furthermore, the observed distal outcomes
estimated. Across the models, means of the profile indicators were incorporated herein were generated as factor scores from the most-in-
freely estimated, but indicator variances were equality constrained variant measurement models of the outcomes across teaching level (see
across the profiles consistent with the homogeneity assumption of the Appendix B).
classical LPA model (Lubke and Neale (2006).5 For the tests of the predictors of profile membership, per the profile
similarity taxonomy (Morin, Meyer et al., 2016), we tested two models.
First, a categorical latent variable multinomial logistic regression model
(footnote continued) was tested in which the predictors were incorporated into the most-
not simultaneously identified. Item thresholds refer to points on the unobserved invariant latent profile solution retained in Phase 3 of the analyses. In
response variate underlying the observed ordinal item at which observed
this model, the relations of the predictors with profile membership were
scores, on average, change from one response category to another. For every k
freely estimated. Second, a model of predictive similarity was estimated
response categories constituting a response scale, there are k-1 thresholds di-
viding the latent response variate distribution. Contrariwise, item intercepts
in which the multinomial logistic regression coefficients were con-
refer to the intercept of the link between the latent factor and latent response strained to equality across teaching levels, and compared to the less
variate. The default in Mplus is the modeling of item thresholds as reported restrictive model in which the predictor-profile relations were freely
herein. estimated. These models were also estimated using start values from the
4
An important consideration in these tests of invariance across groups is the most invariant model retained in Phase 3 of the analysis.
test of factor variance-covariance invariance in bifactor models. In a typical The LPA analyses were performed using robust maximum likelihood
Bifactor CFA model, the invariance of the factor covariances cannot be ex- (MLR) estimation in Mplus 7.4 (Muthén and Muthén, 1998–2015). The
amined given the orthogonality constraints on covariances typically imposed single-group models were estimated using 5000 random sets of start
(Perera & Ganguly, 2016). However, for the BESEM with orthogonal bifactor values with 500 iterations each and the 200 best solutions retained for
target rotation, orthogonality is a function of rotation, and, as such, invariance
final stage optimization. These values were increased to 10,000, 500,
constraints may be imposed on the unrotated solution, thereby allowing the
invariance of the factor covariances in the BSEM to be tested. However, as
Arens and Morin (2017) note, the value of this test is questionable in light of the
orthogonality of the final (rotated) solution. Nevertheless, the current oper- (footnote continued)
ationalization of ESEM models requires invariance constraints to be simulate- indicator variances; however, these models tended to converge on inadmissible
naouly imposed on factor variances and covariances. solutions or failed to converge altogether, suggesting over-parameterization of
5
We also estimated alternative models specifying the free estimation of the models (Bauer & Curran, 2003).

192
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

and 500, respectively, for the multiple-group models (Morin, Meyer 3.2. Teacher self-efficacy profiles
et al., 2016). In all models, (a) the nesting of teachers within schools
was accounted for using the complex design-based correction of stan- Fit indices for the LPA solutions estimated separately in the lower
dard errors in Mplus and (b) TCHWGT was incorporated to account for and upper secondary samples are shown in Table 3. For both samples,
different selection probabilities in the sampling of schools and teachers. the AIC, BIC, SaBIC, and CAIC continued to decrease with the addition
An inclusive approach to single-group model selection was used, in- of profiles. Plots of the information criteria as a function of the k-profile
volving an evaluation of the theoretical consistency of the solutions and solutions showed that decreases in the BIC, CAIC, and SaBIC perma-
information criteria (Henson, Reise, & Kim, 2007; Nylund, Asparouhov, nently flattened at about six profiles for both groups (see Appendix D).
& Muthén, 2007). Specifically, the Bayesian Information Criterion The six-profile solutions and adjacent five-and-seven profile solutions
(BIC), the sample-adjusted BIC (SaBIC), and the consistent Akaike In- were statistically admissible. Notably, in the six profile solutions, pro-
formation Criterion (CAIC) were used with lower values on the criteria files were well-defined with non-trivial class sizes and clear shape dif-
indicative of a better-fitting model (Henson et al., 2007). We also report ferences whereas estimating a seventh profile yielded solutions char-
the Akaike Information Criteria for information purposes. Although acterized by profiles with small class proportions or an additional
simulation research has found that the BIC, SaBIC, and CAIC are useful profile that could not easily be discriminated from the existing profiles.
in deciding on the optimal number of profiles (Nylund et al., 2007; Classification accuracy of the six-profile solutions was reasonable for
Tofighi & Enders, 2008), in large samples, these criteria continue to both samples as indexed by entropy (see Table 3). Average posterior
improve with the addition of profiles without reaching a minimum.6 In probabilities of membership in the assigned profile ranged from 0.765
this case, “elbow plots” of the information criteria can be useful in to 0.894 (M = 0.835), with generally low cross probabilities
detecting the point at which the information criteria plateau (Morin, (0.000–0.103; M = 0.033), in lower secondary teachers. For upper
Meyer et al., 2016). In addition to these statistical indices, we report the secondary teachers, these average posterior probabilities ranged from
entropy values for the models tested, with higher values indicative of 0.794 to 0.901 (M = 0.856) with systematically lower cross prob-
greater classification precision. Finally, for comparative tests of the abilities (0.000–0.079; M = 0.026). The six-profile solution was re-
multiple-group models of LPA similarity, the BIC, SaBIC, and CAIC were tained in each sample for further analysis.
used (Lubke and Neale, 2008; Morin, Meyer et al., 2016), with lower
values indicating better fit. Profile similarity is inferred if at least two 3.3. Latent profile similarity
information criteria suggest support for a more parsimonious model
(Morin, Meyer et al., 2016). Fit indices for the profile similarity tests across teaching level are
shown in Table 3. First, a multiple-group six-profile model was esti-
mated to serve as the configural similarity model against which the
3. Results more restrictive similarity models were compared. Relative to this
baseline model, the more restrictive structural similarity model resulted
3.1. Preliminary measurement model in lower BIC, SaBIC, and CAIC values. Support was also found for the
model of dispersion similarity, with lower values on the BIC, SaBIC, and
Tests of the BESEM model resulted in a good fit in both the lower CAIC relative to the structural similarity solution. Finally, the dis-
secondary, χ2 (24) = 120.872, p < .001, CFI = 0.995, TLI = 0.987, tributional similarity model had lower values on the information cri-
RMSEA = 0.046 (90% CI = 0.038, 0.054), and upper secondary, χ2 teria relative to the dispersion similarity model. Together, the profile
(24) = 139.859, p < .001, CFI = 0.994, TLI = 0.984, RMSEA = 0.052 similarity tests support the equality of (a) the number of latent profiles,
(90% CI = 0.043, 0.060) samples. For both groups, the general factor (b) within-profile mean levels of the mixture indicators, (c) within-
was well-defined with uniformly moderate to strong standardized profile interindividual variability, and (d) the relative sizes of the latent
loadings (lower secondary: |λ| = 0.607–0.784, M = 0.693; upper sec- profiles, respectively. Classification accuracy of the retained distribu-
ondary: |λ| = 0.514–0.802, M = 0.672). Target loadings on the specific tional similarity model is reasonable with an entropy value of 0.839 and
classroom management (SE-CM) (lower secondary: |λ| = 0.279–0.634, average posterior probabilities of membership in the assigned profile,
M = 0.488; upper secondary: |λ| = 0.362–0.692, M = 0.561) and stu- ranging from 0.767 to 0.898 (M = 0.840) and low cross probabilities
dent engagement (SE-SE) (lower secondary: |λ| = 0.238–0.663, (0.000–0.101, M = 0.015).
M = 0.494; upper secondary: |λ| = 0.220–0.671, M = 0.474) factors The profiles from the retained distributional similarity solution are
were uniformly significant and largely moderate to strong in magni- shown in Fig. 2, and Table 4 shows mean levels of the profile indicators
tude. The specific self-efficacy for instructional strategies (SE-IS) factor as a function of profiles and provides a brief description of each profile.
was also relatively well-defined, with three of the four target loadings Profile 1 (9.3%) was characterized by well-below average scores on
not-trivial and statistically significant in both samples (lower sec- both general teacher confidence and self-efficacy for student engage-
ondary: |λ| = 0.020–0.515, M = 0.324; upper secondary: ment and slightly below average scores on self-efficacy for classroom
|λ| = 0.015–0.600, M = 0.308). The complete loadings matrices for the management and instructional strategies. This profile was labeled
sample-specific solutions are provided in Appendix C. “Highly-Inefficacious”. The second profile (18.9%) was characterized
Next, the multiple-group BESEM models were examined. The con- by moderately above average levels of general teacher confidence and
figural invariance model resulted in an excellent fit to the data (see specific self-efficacy for instructional strategies, slightly lower but still
Table 1). Support was also found for the invariance of factor loadings, above average classroom management self-efficacy, and well-below
thresholds, uniquenesses, factor variance-covariance matrix, and factor average student engagement self-efficacy. This profile was labeled
means. Taken together, these results support complete factorial in- “Moderate Globally-and-Instructionally Confident”. Profile 3 (26.2%)
variance. Factor scores from this final solution were saved and used as was characterized by very high levels of general teacher confidence,
the LPA indicators. Table 2 shows the means, standard deviations, and moderately high levels of specific self-efficacy for student engagement,
correlations for the teacher self-efficacy factor scores as well as the and slightly above average levels of specific self-efficacy for classroom
predictors and outcomes for each subsample. management and instructional strategies. This profile was labeled
“Highly-Efficacious”. Profile 4 (26.6%) was characterized by well-
6
Note: The Bootstrap Likelihood Ratio Test (BLRT), which has been shown to below average general teacher confidence, approximately average
be a consistent indicator of profiles across a range of mixture models, is not specific classroom-management self-efficacy, slightly-below average
available in Mplus when the design-based correction of standard errors for specific instructional strategies self-efficacy, and slightly above-average
clustered sampling is used. specific student engagement self-efficacy. This profile was labeled

193
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

Table 1
Fit statistics for the multiple-group models of invariance across the teaching levels.
Model χ2 df CFI TLI RMSEA 90% CI MD χ2 ΔCFI ΔRMSEA

***
MGM1 (Configural IN) 260.693 48 0.995 0.986 0.049 0.043, 0.055
MGM2 (IN FL) 205.939*** 80 0.997 0.995 0.029 0.024, 0.034 36.627 (32) +0.002 -0.020
MGM3 (IN FL + Th) 207.594*** 100 0.997 0.997 0.024 0.019, 0.029 26.388 (20) 0.000 -0.005
MGM4 (IN FL + Th + Uniq) 233.886*** 112 0.997 0.996 0.024 0.020, 0.028 36.730 (12)*** -0.001 0.000
MGM5 (IN FL + Th + Uniq + FVCV) 194.566*** 122 0.998 0.998 0.018 0.013, 0.022 14.886 (10) +0.001 -0.006
MGM6 (IN FL + Th + Uniq + FVCV + FM) 208.323 126 0.998 0.998 0.019 0.014, 0.023 11.984 (4)* 0.000 +0.001

Note. df = degrees of freedom; MD χ2 = change in χ2 relative to a more complex model computed using the Mplus DIFFTEST function; ΔCFI = change in com-
parative fit index; ΔRMSEA = change in root mean square error of approximation; MGM = multiple-group model; IN = invariance; FL = factor loadings;
TH = thresholds; Uniq = Uniquenesses; FVCV = factor variances and covariances; FM = factor means.
* p < .05, ** p < .01.
*** p < .001.

Table 2
Means/proportions, standards deviations, and correlations for the self-efficacy indicators, predictors, and outcomes.
Variable Ma SD 1 2 3 4 5 6 7 8 9

1. GTC −0.082/0.026 0.888/0.864 – 0.133 0.167 0.184 0.098 −0.184 0.071 −0.249 −0.044
2. SE-CM −0.017/-0.030 0.716/0.777 0.157 – −0.069 −0.142 −0.030 0.025 0.064 −0.048 0.020
3. SE-IS 0.001/-0.014 0.629/0.632 0.172 −0.109 – −0.159 −0.045 −0.024 0.039 −0.069 −0.010
4. SE-SE −0.002/-0.016 0.706/0.701 0.165 −0.136 −0.170 – −0.003 0.024 −0.042 −0.084 −0.024
5. Years’ Experience 16.745/17.471 11.151/11.220 0.172 0.057 0.026 0.043 – 0.090 0.336 −0.199 0.065
6. Gender 0.586/0.585 – −0.162 −0.008 −0.075 −0.031 0.083 – 0.010 −0.051 0.047
7. Mentee 0.164/0.140 – 0.095 0.046 −0.050 −0.017 0.268 0.002 – 0.127 0.193
8. Mentor 0.278/0.305 – −0.222 −0.089 −0.103 −0.073 −0.233 0.017 0.087 – 0.096
9. PD-Subject Field 0.778/0.804 – −0.084 0.057 −0.046 −0.066 0.034 0.114 −0.059 0.061 –
10. PD-Pedagogical Competencies 0.656/0.659 – −0.076 0.085 −0.152 −0.016 0.057 0.166 0.102 0.094 0.524
11. PD-Assessment Practices 0.584/0.655 – −0.149 0.038 −0.047 −0.005 −0.003 0.061 0.013 0.157 0.298
12. PD-Classroom management 0.349/0.347 – −0.034 0.021 −0.030 −0.041 −0.008 −0.025 0.074 0.111 0.084
13. PD-Individualized Learning 0.523/0.467 – −0.105 0.093 −0.101 −0.012 −0.019 0.078 0.052 0.094 0.220
14. PDN-Subject Field 1.854/1.857 0.772/0.744 −0.171 −0.061 −0.026 0.012 −0.199 0.080 −0.063 0.081 −0.074
15. PDN-Pedagogical Competencies 1.962/1.914 0.770/0.744 −0.277 −0.066 −0.051 −0.006 −0.195 0.090 −0.129 0.100 −0.072
16. PDN-Assessment Practices 2.087/2.048 0.795/0.777 −0.171 −0.048 −0.027 −0.012 −0.162 0.069 −0.130 0.056 −0.009
17. PDN-Classroom Management 1.879/1.778 0.804/0.773 −0.345 −0.256 −0.006 −0.025 −0.233 0.060 −0.081 0.147 −0.085
18. PDN-Individualized Learning 2.293/2.195 0.826/ −0.154 −0.054 −0.052 −0.057 −0.117 0.023 −0.132 0.011 −0.090
19. Job Satisfaction −0.092/0.029 0.883/0.830 0.193 0.129 −0.003 0.072 0.030 −0.004 −0.127 −0.073 −0.097
20. Classroom Climate −0.019/0.0444 0.888/0.838 0.318 0.157 −0.053 0.070 0.149 −0.093 0.048 −0.160 −0.106
21. Teacher Collaboration −0.023/-0.005 0.870/0.864 0.253 0.020 0.114 0.039 −0.010 −0.109 −0.053 −0.210 −0.085

Variable 10 11 12 13 14 15 16 17 18 19 20 21

1. GTC −0.049 −0.118 −0.020 −0.103 −0.161 −0.193 −0.155 −0.261 −0.163 0.188 0.252 0.244
2. SE-CM 0.014 −0.013 −0.032 0.041 0.011 −0.057 −0.019 −0.195 0.004 0.074 0.062 −0.018
3. SE-IS −0.017 −0.050 −0.047 −0.099 −0.035 −0.034 −0.043 0.044 −0.035 0.020 −0.064 0.058
4. SE-SE −0.031 0.000 −0.063 −0.052 0.000 0.005 0.006 −0.070 −0.055 0.125 0.082 0.106
5. Years’ Experience 0.012 −0.041 −0.012 0.027 −0.137 −0.176 −0.168 −0.217 −0.099 0.020 0.179 −0.079
6. Gender −0.046 0.070 −0.004 0.009 0.062 0.043 0.038 0.001 0.029 −0.071 −0.081 −0.180
7. Mentee 0.098 0.196 0.176 0.191 −0.072 −0.107 −0.118 −0.138 0.008 −0.063 0.108 −0.120
8. Mentor 0.120 0.139 0.036 0.142 0.055 0.096 0.047 0.150 0.069 −0.040 −0.053 −0.143
9. PD-Subject Field 0.509 0.292 0.067 0.253 −0.162 −0.123 −0.065 −0.066 −0.046 −0.039 0.001 −0.091
10. PD-Pedagogical Competencies – 0.368 0.202 0.431 −0.107 −0.180 −0.141 −0.158 −0.150 −0.050 0.041 −0.149
11. PD-Assessment Practices 0.419 – 0.320 0.497 0.001 −0.008 −0.071 −0.042 −0.060 −0.071 0.024 −0.148
12. PD-Classroom management 0.202 0.305 – 0.402 −0.050 −0.075 −0.078 −0.161 −0.055 −0.033 0.119 −0.168
13. PD-Individualized Learning 0.395 0.483 0.331 – −0.041 −0.098 −0.050 −0.086 −0.112 −0.017 0.021 −0.129
14. PDN-Subject Field −0.029 0.047 −0.085 −0.004 – 0.683 0.495 0.406 0.333 −0.040 −0.142 −0.023
15. PDN-Pedagogical Competencies −0.092 0.028 −0.059 −0.031 0.679 – 0.591 0.531 0.501 −0.067 −0.147 −0.009
16. PDN-Assessment Practices −0.033 −0.004 −0.091 −0.008 0.490 0.600 – 0.477 0.508 −0.050 −0.091 0.019
17. PDN-Classroom Management −0.092 0.015 −0.176 −0.022 0.384 0.504 0.474 – 0.443 −0.169 −0.292 −0.009
18. PDN-Individualized Learning −0.088 0.005 −0.083 −0.025 0.294 0.464 0.505 0.473 – −0.067 −0.074 0.023
19. Job Satisfaction −0.023 −0.082 0.026 −0.066 −0.006 −0.014 −0.027 −0.151 −0.087 – 0.192 0.192
20. Classroom Climate 0.021 −0.050 0.089 −0.025 −0.130 −0.170 −0.172 −0.334 −0.142 0.299 – 0.052
21. Teacher Collaboration −0.199 −0.198 −0.098 −0.181 −0.009 −0.032 −0.035 −0.050 −0.044 0.198 0.089 –

Note. N = 3735 (nISCED2 = 1915; nISCED3 = 1820). Estimates before the forward slash and correlations below the diagonal are for the ISCED 2 sample. aFor di-
chotomous items, this value is the proportion of observations in the first category. GTC = general teacher confidence; SE-CM = self-efficacy for classroom man-
agement; SE-IS = self-efficacy for instructional strategies; SE-SE = self-efficacy for student engagement. PD = professional development; PDN = professional de-
velopment needs. Correlations above 0.045 are significant at p < .05 for ISCED 2 teachers, and correlations above 0.046 are significant at p < .05 for ISCED 3
teachers.

“Globally-Unconfident”. Profile 5 (11.6%) was characterized by very efficacy for classroom management and instructional strategies and
high levels of specific student engagement self-efficacy and moderately- thus was labeled “Student-Engagement-Efficacious”. Finally, Profile 6
below average levels of general teaching confidence and specific self- (7.4%) was characterized by very low levels of specific student

194
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

Table 3
Fit indices and classification accuracy for the LPA models.
Lower Secondary (n = 1915) k LL #fp AIC BIC SaBIC CAIC Entropy

One profile 1 −8448.724 8 16913.447 16957.907 16932.491 16965.908 –


Two Profiles 2 −8382.356 13 16790.712 16862.959 16821.658 16875.959 0.503
Three Profiles 3 −8288.608 18 16613.215 16713.250 16656.063 16731.251 0.553
Four Profiles 4 −8200.908 23 16447.815 16575.637 16502.566 16598.638 0.677
Five Profiles 5 −8149.155 28 16354.310 16509.919 16420.963 16537.919 0.700
Six Profiles 6 −8092.750 33 16251.501 16434.897 16330.056 16467.897 0.768
Seven Profiles 7 −8039.522 38 16155.043 16366.227 16245.501 16404.228 0.772
Eight Profiles 8 −7997.283 43 16080.566 16319.538 16182.926 16362.537 0.752
Upper Secondary (n = 1820)
One profile 1 −8123.360 8 16262.719 16306.772 16281.357 16314.773 –
Two Profiles 2 −8015.506 13 16057.013 16128.598 16087.298 16141.598 0.921
Three Profiles 3 −7910.419 18 15856.837 15955.956 15898.771 15973.957 0.735
Four Profiles 4 −7832.044 23 15710.087 15836.739 15763.669 15859.739 0.636
Five Profiles 5 −7785.946 28 15627.892 15782.077 15693.122 15810.077 0.712
Six Profiles 6 −7701.577 33 15469.154 15650.871 15546.032 15683.872 0.794
Seven Profiles 7 −7631.159 38 15338.318 15547.568 15426.844 15585.568 0.812
Eight Profiles 8 −7588.047 43 15262.093 15498.877 15362.268 15541.877 0.818
Tests of Profile Similarity
Configural – −18382.024 67 36898.048 37315.157 37102.263 37382.157 0.842
Structural – −18439.771 43 36965.542 37233.238 37096.605 37276.239 0.839
Dispersion – −18444.412 39 36966.823 37209.618 37085.695 37248.619 0.839
Distributional – −18449.237 34 36966.475 37178.142 37070.106 37212.141 0.839
Explanatory Similarity
Profile-outcome relations freely estimated – −31845.649 73 63837.299 64291.760 64059.802 64364.759 0.843
Profile-outcome relations invariant – −31982.967 55 64075.933 64418.336 64243.572 64473.337 0.842
Predictive Similarity
Predictor-profile relations freely estimated – −16354.100 174 33056.200 34122.265 33569.387 34313.4375 0.847
Predictor-profile relations invariant – −16405.258 104 33018.516 33655.705 33325.248 33769.968 0.844

Note. k = number of profiles; #fp = number of free parameters; LL = mode log-likelihood; AIC = Akaike information criteria; BIC = Bayesian information criteria;
SaBIC = sample-size adjusted BIC; CAIC = Consistent Akaike information criteria.

Fig. 2. Graphical depiction of the final six-profile solution.

engagement self-efficacy, moderately low levels of general teacher profile.


confidence, approximately average specific classroom management
self-efficacy, and slightly above-average specific instructional strategies
self-efficacy. “Student-Engagement-Inefficacious” was used to label this

195
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

3.4. Explanatory similarity

Moderately below average GTC; approximately-average CMSE; slightly above-average ISSE; well-below average SESE
Well-below average GTC; approximately average CMSE; slightly below-average ISSE; slightly above-average SESE
Moderately above-average GTC and ISSE; slightly lower, but still above-average, CMSE; well-below-average SESE
The teacher outcomes were added to the retained distributional si-
milarity model to examine the equality of profile-outcome relations
across teaching levels. As shown in Table 3, the model of explanatory

Note. GTC = General teacher confidence; CMSE = Specific classroom management self-efficacy; ISSE = Specific instructional strategies self-efficacy; SESE = specific student engagement self-efficacy.
similarity resulted in higher values on the BIC, SaBIC, and CAIC relative
Well-above average GTC; slightly above-average CMSE and ISSE; moderately above-average SESE

to a model in which the within-profile levels of the outcomes were


freely estimated across the teacher samples, providing evidence against
explanatory similarity. Thus, the more complex model was retained for
interpretation. Mean levels of each outcome across the profiles within
the lower and upper secondary teacher samples, and tests of between-
profile differences, are reported in Table 5. For lower secondary tea-
Moderately below-average GTC, CMSE, and ISSE; Well-above average SESE
Well below-average GTC and SESE; slightly below average CMSE and ISSE

chers, job satisfaction was lowest in Profile 1 (Highly-Inefficacious) and


significantly lower than in all other profiles, excepting Profile 6 (Stu-
dent-Engagement-Inefficacious). In contrast, Profile 3 (Highly-Effica-
cious) had the highest job satisfaction, which was significantly higher
than in all profiles other than Profile 2 (Moderate Globally-and-In-
structionally-Confident). Levels of job satisfaction in Profile 2 (Mod-
erate Globally-and-Instructionally-Confident) significantly exceeded
levels in Profile 4 (Globally-Unconfident) and Profile 6 (Student-En-
gagement-Inefficacious) but not Profile 5 (Student-Engagement-Effica-
cious). Profiles 4, 5, and 6 did not significantly differ on job satisfaction.
A comparable pattern of results was observed for (positive) classroom
climate and teacher collaboration. Both outcomes were lowest in Profile
1 (Highly-Inefficacious) and significantly lower in this profile than in all
other profiles. The highest levels of classroom climate and teacher
collaboration were observed in Profile 3 (Highly-Efficacious), followed
by Profile 2 (Moderate Globally-and-Instructionally-Confident), and
Profile Description

though these two subgroups did not differ from each other on the
outcomes, they significantly differed from levels observed in the re-
maining profiles. Profiles 4, 5, and 6 did not significantly differ on
climate and collaboration.
For upper secondary teachers, levels of job satisfaction were sig-
nificantly lower in Profile 1 (Highly-Inefficacious) than in all other
−0.602
−0.552

−1.402

profiles. No significant differences were observed between Profile 4


0.403
0.109
1.054
SESE

(Globally-Unconfident) and Profile 6 (Student-Engagement-


Inefficacious) on job satisfaction, though levels of the outcomes in these
profiles were significantly lower than in Profiles 2, 3, and 5, which
Mean levels on the profile indicators as a function of latent profiles and profile description.

−0.096

−0.175
−0.299

themselves were not significantly different. Classroom climate for upper


0.213
0.113

0.184
ISSE

secondary teachers was lowest in Profile 1 (Highly-Inefficacious) and


significantly lower than in all other profiles. There were no significant
differences between Profile 4 (Globally-Unconfident) and Profile 6
−0.148

−0.042
−0.494
CMSE

(Student-Engagement-Inefficacious) on classroom climate; however,


0.158
0.103

0.024

these profiles had significantly lower levels of the outcome than tea-
chers in Profiles 2 and 3. In addition, levels of classroom climate were
significantly lower in Profile 4 (Globally-Unconfident) than Profile 5
−1.236

−0.541
−0.323
−0.339
0.213
0.961

(Student-Engagement-Efficacious) and in Profile 5 (Student-


GTC

Engagement-Efficacious) relative to Profile 3 (Highly-Efficacious). No


significant differences on climate were observed between Profiles 2 and
3. Finally, for teacher collaboration, the highest levels were observed in
Moderate Globally and Instructionally Confident

Profile 3 (Highly-Efficacious), which significantly exceeded levels in all


profiles, excepting Profile 2 (Moderate Globally-and-Instructionally-
Confident). Profile 2 (Moderate Globally-and-Instructionally-Confident)
and Profile 5 (Student-Engagement-Efficacious) had higher collabora-
Student-Engagement-Inefficacious
Student-Engagement-Efficacious

tion than Profiles 1, 4, and 6, and did not significantly differ from one
another on the outcome. No significant differences on collaboration
were observed among Profiles 1, 4, and 6.
Globally-Unconfident
Highly Inefficacious

Highly-Efficacious

3.5. Predictive similarity

Predictors were added to the final distributional similarity model to


examine the equality of predictor-profile relations across lower and
upper secondary teachers. As shown in Table 3, the model of predictive
1:
2:
3:
4:
5:
6:

similarity resulted in lower values on the BIC, SaBIC, and CAIC, relative
Table 4

Profile

Profile
Profile
Profile
Profile
Profile
Profile

to the model in which the predictor-profile relations were freely esti-


mated across teaching level, thereby supporting the equality of

196
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

Table 5
Associations of profile membership with job satisfaction, classroom disciplinary climate, and teacher collaboration.
Outcomes Profile 1 (a) Profile 2 (b) Profile 3 (c) Profile 4 (d) Profile 5 (e) Profile 6 (f)

Satisfaction
ISCED 2 −0.629bcde 0.127adf 0.129adef −0.155abc −0.141ac −0.340bc
ISCED 3 −0.540bcdef 0.166adf 0.223adf −0.070abce 0.158adf −0.213abce

Climate
ISCED 2 −0.734bcdef 0.284adef 0.339adef −0.130abc −0.206abc −0.279abc
ISCED 3 −0.182bcdef 0.628adf 0.713adef 0.241abce 0.534acd 0.348abc

Collaboration
ISCED 2 −0.532bcdef 0.138adef 0.271adef −0.101abc −0.196abc −0.143abc
ISCED 3 −0.393bce 0.127adf 0.281adef −0.238bce 0.054acdf −0.192bce

Note: ISCED 2 = lower secondary; ISCED 3 = upper secondary. Subscripts denote profiles that differ significantly at p < .05.

predictor-profile relations. Table 6 shows the results from the catego- and-Instructionally-Confident) and Profile 3 (Highly-Efficacious) re-
rical latent variable multinomial logistic regressions for the relations of lative to Profile 4 (Globally-Unconfident), Profile 5 (Student-
the predictors with the probability of profile membership. Teachers Engagement-Efficacious), and Profile 6 (Student-Engagement-
with greater years of experience were significantly more likely to be in Inefficacious). Finally, teachers who perceived a greater need for pro-
Profile 1 (Highly-Inefficacious) and Profile 3 (Highly-Efficacious) re- fessional development in the areas of individualizing student learning
lative to Profile 6 (Student-Engagement-Inefficacious). No other sig- were significantly less likely to be in Profile 3 (Highly-Efficacious) than
nificant relations of years of experience were obtained. For gender, Profile 6 (Student-Engagement-Inefficacious). These teachers had a
males had a significantly greater likelihood of membership into Profile greater likelihood of membership in Profile 2 (Moderate Globally-and-
1 (Highly-Inefficacious) relative to Profile 2 (Moderate Globally-and- Instructionally-Confident) relative to Profile 1 (Highly-Inefficacious),
Instructionally-Confident), Profile 3 (Highly-Efficacious), Profile 5 Profile 3 (Highly-Efficacious), Profile 4 (Globally-Unconfident), and
(Student-Engagement-Efficacious), and Profile 6 (Student-Engagement- Profile 5 (Student-Engagement-Efficacious).
Inefficacious). Females were significantly more likely to be in Profile 2
(Moderate Globally-and-Instructionally-Confident) and Profile 3 4. Discussion
(Highly-Efficacious) than Profile 4 (Globally-Unconfident) and Profile 5
(Student-Engagement-Efficacious). Drawing on social cognitive perspectives, we investigated profiles of
Some significant relations of mentoring and professional develop- teacher self-efficacy. Additionally, we investigated theoretically plau-
ment involvement with profile membership were observed. Teachers sible outcomes and predictors of profile membership and examined the
who had an assigned mentor (i.e., mentees) had a significantly greater extent to which the profiles and their relations generalized across
likelihood of membership into Profile 4 (Globally-Unconfident) relative teaching levels. Although no prior work has examined unobserved
to Profile 2 (Moderate Globally-and-Instructionally-Confident) and heterogeneity with respect to teacher self-efficacy, per RQ#1, we
Profile 3 (Highly-Efficacious). No other relations of being a mentee with identified six profiles of teacher self-efficacy, which, notably, align with
profile membership were observed. Teachers who served as a mentor teacher self-efficacy theory. A central tenet of the multidimensional
were less likely to be in Profile 1 (Highly-Inefficacious) than in all other perspective of teacher self-efficacy is that teachers may not be equally
profiles. Contrariwise, these mentor teachers had a greater likelihood of efficacious across all domains (Tschannen-Moran et al., 1998); instead,
membership in Profile 3 (Highly-Efficacious) than in Profile 4 they may simultaneously possess multiple capability beliefs at varying
(Globally-Unconfident), Profile 5 (Student-Engagement-Efficacious), levels as they navigate their professional roles. To the authors’ knowl-
and Profile 6 (Student-Engagement-Inefficacious), and were more likely edge, this is the first study to obtain evidence showing that teachers
to be in Profile 2 (Moderate Globally-and-Instructionally-Confident) simultaneously possess multiple teaching self-efficacy beliefs with
than in Profile 4 (Globally-Unconfident). Teachers who participated in varying configurations, manifested as the six teacher self-efficacy pro-
professional development in individualized learning in the prior files. As teachers’ work spans multiple domains that consistently
12 months had a significantly greater likelihood of membership in overlap, the distinct configurations may serve as an integrated system
Profile 3 (Highly-Efficacious) and Profile 5 (Student-Engagement- of confidence beliefs that guides their interactions with the learning
Efficacious) relative to Profile 4 (Globally-Unconfident). environment. The simultaneous operation of these teacher self-efficacy
Several significant associations of perceived professional develop- beliefs suggests that each self-efficacy belief may serve as a context for
ment needs with profile membership were found. Teachers who per- other self-efficacy beliefs (Litalien, Morin, & McInerney, 2017). For
ceived a greater need for professional development in subject field instance, take the Student-Engagement-Inefficacious profile, char-
knowledge had a significantly greater likelihood of membership in acterized by, inter alia, above-average specific self-efficacy for instruc-
Profile 2 (Moderate Globally-and-Instructionally-Confident) relative to tional strategies. Although teachers in this profile may be somewhat
Profile 6 (Student-Engagement-Inefficacious). Those who perceived confident about their capability to implement instructional strategies,
that they required more professional development in pedagogical any advantages in the classroom conferred by this belief will likely be
competencies were significantly less likely to be in Profile 2 (Moderate diminished when combined with doubts about capabilities to engage
Globally-and-Instructionally-Confident) and Profile 3 (Highly- students in the learning process as well as low general confidence for
Efficacious) relative to Profile 4 (Globally-Unconfident) and Profile 5 performing their professional role.
(Student-Engagement-Efficacious). For teachers who perceived a Notably, the profiles comprise both general teacher confidence and
greater need for professional development in assessment practices, domain-specific teacher self-efficacy beliefs. This generality and speci-
there was a significantly reduced likelihood of membership in Profile 1 ficity is consistent with the view that teacher self-efficacy reflects an
(Highly-Inefficacious) relative to Profile 3 (Highly-Efficacious). integration of perceptions of personal teaching capability with judg-
Teachers who perceived a need for professional development in class- ments about the requirements of domain-specific roles and tasks
room management had a greater likelihood of membership in Profile 1 (Goddard et al., 2000; Tschannen-Moran et al., 1998; Zee & Koomen,
(Highly-Inefficacious) relative to all other profiles. These teachers were 2016). The generality and specificity also align with the view that
also approximately half as likely to be in Profile 2 (Moderate Globally- teacher self-efficacy may reside on a continuum incorporating both

197
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

Table 6
Results from the multinomial logistic regression models of the effects of the predictors on latent profile membership.
1 Vs 6 2 Vs 6 3 Vs 6 4 Vs 6 5 Vs 6

Predictors Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR

* **
Years’ Experience 0.029 (0.013) 1.029 0.018 (0.011) 1.018 0.026 (0.009) 1.026 0.017 (0.010) 1.017 0.010 (0.010) 1.010
Gender 0.729 (0.295)* 2.073 −0.235 (0.224) 0.790 −0.296 (0.209) 0.744 0.401 (0.225) 1.493 0.180 (0.222) 1.197
Mentee −0.385 (0.435) 0.680 0.204 (0.361) 1.227 0.117 (0.345) 1.124 −0.432 (0.315) 0.649 0.028 (0.363) 1.028
Mentor 1.050 (0.364)** 2.858 −0.315 (0.235) 0.730 −0.549 (0.200)** 0.577 0.211 (0.213) 1.235 −0.041 (0.295) 0.959
PD-Subject Field −0.109 (0.346) 0.897 −0.375 (0.274) 0.688 −0.313 (0.295) 0.731 −0.256 (0.248) 0.774 −0.462 (0.328) 0.630
PD-Pedagogical Competencies 0.188 (0.291) 1.207 0.061 (0.288) 1.063 −0.224 (0.243) 0.799 0.017 (0.237) 1.017 0.179 (0.215) 1.195
PD-Assessment Practices 0.033 (0.304) 1.034 −0.186 (0.233) 0.831 −0.239 (0.219) 0.787 0.002 (0.237) 1.002 0.141 (0.219) 1.151
PD-Classroom management −0.086 (0.339) 0.917 −0.024 (0.230) 0.976 −0.350 (0.203) 0.704 −0.080 (0.223) 0.923 −0.175 (0.266) 0.839
PD-Individualized Learning 0.207 (0.287) 1.229 −0.036 (0.215) 0.965 −0.129 (0.173) 0.879 0.269 (0.186) 1.308 −0.225 (0.206) 0.798
PDN-Subject Field 0.288 (0.276) 1.334 0.453 (0.198)* 1.573 0.296 (0.212) 1.345 0.321 (0.190) 1.379 0.275 (0.259) 1.316
PDN-Pedagogical Competencies 0.104 (0.316) 1.110 −0.316 (0.208) 0.729 −0.247 (0.184) 0.782 0.178 (0.204) 1.194 0.152 (0.187) 1.164
PDN-Assessment Practices −0.193 (0.200) 0.825 0.011 (0.161) 1.011 0.170 (0.163) 1.186 0.033 (0.146) 1.034 −0.106 (0.251) 0.900
PDN-Classroom Management 0.929 (0.244)*** 2.533 −0.649 (0.181)*** 0.523 −0.654 (0.163)*** 0.520 0.036 (0.166) 1.037 −0.046 (0.224) 0.955
PDN-Individualized Learning −0.383 (0.240) 0.682 −0.015 (0.178) 0.985 −0.335 (0.168)* 0.716 −0.241 (0.197) 0.786 −0.303 (0.196) 0.739

1 Vs 5 2 Vs 5 3 Vs 5 4 Vs 5 1 Vs 4

Predictors Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR

Years Experience 0.018 (0.011) 1.018 0.008 (0.011) 1.008 0.016 (0.010) 1.016 0.006 (0.010) 1.006 0.012 (0.011) 1.012
Gender 0.549 (0.240)* 1.732 −0.416 (0.166)* 0.660 −0.476 (0.187)* 0.621 0.221 (0.201) 1.247 0.328 (0.225) 1.389
Mentee −0.413 (0.366) 0.661 0.176 (0.257) 1.193 0.089 (0.238) 1.093 −0.460 (0.270) 0.632 0.046 (0.292) 1.047
Mentor 1.092 (0.325)** 2.979 −0.274 (0.226) 0.761 −0.508 (0.227)* 0.602 0.252 (0.234) 1.287 0.839 (0.303)** 2.315
PD-Subject Field 0.353 (0.277) 1.423 0.087 (0.244) 1.091 0.149 (0.230) 1.160 0.206 (0.227) 1.229 0.147 (0.265) 1.158
PD-Pedagogical Competencies 0.010 (0.245) 1.010 −0.117 (0.242) 0.889 −0.403 (0.211) 0.669 −0.162 (0.194) 0.851 0.172 (0.275) 1.187
PD-Assessment Practices −0.107 (0.230) 0.898 −0.327 (0.227) 0.721 −0.380 (0.221) 0.684 −0.138 (0.239) 0.871 0.031 (0.289) 1.031
PD-Classroom management 0.089 (0.343) 1.093 0.151 (0.214) 1.163 −0.175 (0.212) 0.839 0.095 (0.197) 1.100 −0.006 (0.286) 0.994
PD-Individualized Learning 0.432 (0.255) 1.540 0.190 (0.177) 1.209 0.096 (0.170) 1.101 0.494 (0.185)** 1.639 −0.062 (0.228) 0.940
PDN-Subject Field 0.013 (0.202) 1.013 0.178 (0.172) 1.195 0.021 (0.175) 1.022 0.046 (0.168) 1.047 −0.033 (0.192) 0.968
PDN-Pedagogical Competencies −0.048 (0.307) 0.953 −0.468 (0.176)** 0.626 −0.399 (0.151)** 0.671 0.026 (0.205) 1.026 −0.074 (0.235) 0.929
PDN-Assessment Practices −0.087 (0.219) 0.917 0.117 (0.195) 1.124 0.276 (0.196) 1.318 0.139 (0.230) 1.149 −0.226 (0.156) 0.798
PDN-Classroom Management 0.976 (0.178)*** 2.653 −0.602 (0.182)** 0.548 −0.608 (0.160)*** 0.545 0.082 (0.159) 1.086 0.893 (0.169)*** 2.443
PDN-Individualized Learning −0.081 (0.161) 0.922 0.288 (0.116)* 1.333 −0.032 (0.120) 0.969 0.061 (0.117) 1.063 −0.142 (0.155) 0.868

2 Vs 4 3 Vs 4 1 Vs 3 2 Vs 3 1 Vs 2

Predictors Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR Coef. (SE) OR

Years Experience 0.001 (0.008) 1.001 0.009 (0.008) 1.009 0.003 (0.011) 1.003 −0.008 (0.007) 0.992 0.011 (0.011) 1.011
Gender −0.636 (0.189)** 0.529 −0.697 (0.153)*** 0.498 1.025 (0.228)*** 2.787 0.061 (0.148) 1.062 0.965 (0.227)*** 2.624
Mentee 0.636 (0.234)** 1.889 0.549 (0.214)* 1.731 −0.503 (0.313) 0.605 0.087 (0.201) 1.091 −0.590 (0.337) 0.554
Mentor −0.526 (0.188)** 0.591 −0.760 (0.135)*** 0.468 1.599 (0.289)*** 4.950 0.234 (0.171) 1.264 1.365 (0.310)*** 3.916
PD-Subject Field −0.119 (0.200) 0.888 −0.057 (0.206) 0.944 0.204 (0.265) 1.226 −0.061 (0.183) 0.941 0.265 (0.282) 1.304
PD-Pedagogical Competencies 0.045 (0.178) 1.046 −0.241 (0.154) 0.786 0.412 (0.267) 1.511 0.285 (0.154) 1.330 0.127 (0.276) 1.135
PD-Assessment Practices −0.188 (0.161) 0.829 −0.242 (0.153) 0.785 0.272 (0.268) 1.313 0.053 (0.164) 1.055 0.219 (0.308) 1.245
PD-Classroom management 0.056 (0.187) 1.058 −0.270 (0.155) 0.763 0.264 (0.316) 1.302 0.326 (0.175) 1.386 −0.062 (0.319) 0.940
PD-Individualized Learning −0.304 (0.169) 0.738 −0.398 (0.156)* 0.672 0.336 (0.230) 1.399 0.094 (0.157) 1.098 0.242 (0.255) 1.274
PDN-Subject Field 0.132 (0.127) 1.141 −0.025 (0.131) 0.975 −0.008 (0.195) 0.992 0.157 (0.125) 1.170 −0.165 (0.209) 0.848
PDN-Pedagogical Competencies −0.493 (0.192)* 0.611 −0.424 (0.164)* 0.654 0.350 (0.265) 1.420 −0.069 (0.160) 0.933 0.420 (0.306) 1.521
PDN-Assessment Practices −0.022 (0.139) 0.979 0.137 (0.114) 1.147 −0.363 (0.151)* 0.696 −0.159 (0.111) 0.853 −0.204 (0.165) 0.816
PDN-Classroom Management −0.685 (0.153)*** 0.504 −0.690 (0.25)*** 0.501 1.584 (0.181)*** 4.872 0.006 (0.144) 1.006 1.578 (0.201)*** 4.846
PDN-Individualized Learning 0.226 (0.113)* 1.254 −0.093 (0.125) 0.911 −0.049 (0.175) 0.952 0.320 (0.088)*** 1.377 −0.368 (0.168)* 0.692

Note. PD = professional development; PDN = professional development needs; Coeff = logit coefficient; SE = standard error of the coefficient; OR = Odds Ratio.
Gender (0 = female, 1 = male); Mentee (0 = yes, 1 = no); Mentor (0 = yes, 1 = no); Professional development (PD) participation (0 = yes, 1 = no).
* p < .05,
** p < .01,
*** p < .001.

confidence beliefs pertaining to a broad domain (i.e., teaching) and and Marsh, 2015). Such differentiation of teacher self-efficacy beliefs is
increasingly task-specific capability judgments (e.g., developing a critical insofar as it accords with the theoretical view that teachers can
classroom management system) (Lent and Brown, 2006). Here the be more or less efficacious in different domains (Tschannen-Moran
general teacher confidence dimension may reflect a gestalt of capability et al., 1998; Zee & Koomen, 2016).
self-beliefs spanning their professional roles as distinct from the do- Per RQ#2, the six-profile solution replicated across lower and upper
main-specific capability beliefs that emerge from both judgments of secondary teachers. Specifically, support was found for the configural,
capability and the evaluation of task-or-role-specific requirements structural, dispersion, and distributional similarity of the profiles. These
(Perera et al., 2019). The failure to account for teachers’ general con- findings suggest, respectively, that across lower and upper secondary
fidence beliefs may obscure the multidimensional teacher self-efficacy teachers (a) the same number of profiles were identified, (b) the six
profiles insofar as unmodeled generality typically leads to less struc- profiles were characterized by similar mean levels of the general and
tural differentiation on the specific teacher self-efficacy beliefs (Morin specific teacher self-efficacy dimensions, (c) there is equivalent within-

198
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

profile variability of the teacher self-efficacy indicators, and (d) the satisfaction was observed. This moderating effect implies the greater
relative sizes of the profiles are equal. The present research provides the dependence of job satisfaction on general teacher confidence and spe-
first evidence that multidimensional teacher self-efficacy belief com- cific self-efficacy for classroom management and instructional strate-
binations are equivalent across distinct teaching contexts. The profile gies in teachers of lower secondary classes.
similarity results obtained support the cross-sample consistency of the Although the most suboptimal profile for both lower and upper
teacher self-efficacy profiles and underpin profile validity. secondary teachers is the Highly-Inefficacious subgroup (Profile 1) in
line with prior variable-centered work (Caprara et al., 2006; Skaalvik
4.1. Outcomes of profile membership and Skaalvik, 2010, 2014), the extent to which other profiles with a
configuration of mostly low teacher self-efficacy beliefs were deleter-
Theoretically-informative differences in job satisfaction, classroom ious differed somewhat across the teaching levels. At the lower sec-
climate, and collaboration as a function of profile membership were ondary level, teachers in the Globally-Unconfident (Profile 4), Student-
found in investigating RQ#3a. Notably, in investigating the invariance Engagement-Efficacious (Profile 5), and Student-Engagement-In-
of these profile-outcome relations per RQ#3b, some observed relations efficacious (Profile 6) subgroups reported comparably low levels of
differed across teaching level, indicating context differentiation in the satisfaction, climate, and collaboration. Even though each of these
outcomes of profile membership. The two profiles characterized by profiles is characterized by one teacher self-efficacy dimension ex-
above-average levels of general teacher confidence and specific self- ceeding average levels, higher-levels of one belief may not be sufficient
efficacy for classroom management and instructional strategies—the in lower secondary teachers to buffer the deleterious effects of posses-
Highly-Efficacious (Profile 3) and Moderate Globally-and- sing a configuration of mostly low or average teacher self-efficacy be-
Instructionally-Confident (Profile 2) profiles—generally had the most liefs. Contrariwise, at the upper secondary-level, whereas Globally-un-
positive outcomes across both lower and upper secondary teachers. confident and Student-Engagement-Inefficacious teachers did not
Lower and upper secondary teachers in these profiles reported greater significantly differ on the outcomes, Student-Engagement-Efficacious
job satisfaction, positive classroom climate, and collaboration than teachers reported greater satisfaction and collaboration than those in
those in the Highly-Inefficacious (Profile 1), Globally-Unconfident the other two profiles, and better classroom climate than Globally-un-
(Profile 4), and Student-Engagement-Inefficacious (Profile 6) profiles. confident teachers. The Student-Engagement-Efficacious profile is
These results suggest that possessing a configuration of higher levels of characterized by below-average levels of general teacher confidence
most of the teacher self-efficacy beliefs yields better outcomes than and the lowest levels of specific self-efficacy for classroom management
possessing a configuration of capability beliefs that are uniformly or and instructional strategies, but well-above-average levels of self-effi-
largely below-average. These results converge with previous variable- cacy for student engagement. In the upper secondary years, as classes
centered research showing that higher levels of teacher self-efficacy are become more specialized and high-stakes and oftentimes smaller, sa-
associated with greater job satisfaction (Caprara et al., 2006; Klassen tisfaction and positive climate may hinge more heavily on the devel-
and Chiu, 2010), perceived positive classroom climate (Tsouloupas opment of affective teacher-student relationships that provide crucial
et al., 2010), and collaboration (Caprara et al., 2003). emotional supports, which is theorized to be linked to self-efficacy for
Across the teaching levels, Highly-Efficacious (Profile 3) teachers student engagement in particular (Ryan, Kuusinen, & Bedoya-Skoog,
did not significantly differ from Moderate Globally-and-Instructionally- 2015).
Confident (Profile 2) teachers on the outcomes. This is somewhat sur-
prising as Highly Efficacious teachers possess uniformly above-average 4.2. Predictors of profile membership
beliefs, including general teacher confidence at approximately one
standard deviation above the mean, whereas teachers in Profile 2 re- We examined predictors of profile membership in RQ#4a and the
ported comparatively lower levels of general teacher confidence and generalizability of these relations in RQ#4b. There were some sig-
below-average levels of specific self-efficacy for student engagement. nificant associations between mentoring involvement and profile
For Moderate Globally-and Instructionally-Confident teachers, slightly membership, which were found to generalize across the lower and
higher levels of specific self-efficacy for classroom management and upper secondary teachers. Teachers with an assigned mentor had a
implementation of instructional strategies may buffer the potentially greater likelihood of membership in the Globally-Unconfident profile
deleterious impact of comparatively lower levels of general teacher (Profile 4) than in the Moderate-Globally-and-Instructionally-Confident
confidence and specific self-efficacy for student engagement on the (Profile 2) and Highly-Efficacious (Profile 3) profiles. These results may
teacher outcomes. This is consistent with the view that teachers’ dif- seem somewhat inconsistent with not only the theoretical perspective
ferential self-efficacy beliefs simultaneously operate as an integrative that mentoring serves as a learning experience that fosters the devel-
system, such that one or another belief may serve as a context for the opment of self-efficacy but also prior work that has found positive as-
operation of the remaining beliefs (Litalien et al., 2017; Perera et al., sociations of mentoring with mentee teachers’ self-efficacy beliefs
2019). (LoCasale-Crouch et al., 2012; Richter et al., 2013). One possibility is
Although lower secondary teachers in the Moderate Globally-and- that, after accounting for years of teaching experience, having an as-
Instructionally-Confident (Profile2) profile reported better classroom signed mentor may be indicative of teacher underperformance. Indeed,
climate than Student-Engagement-Efficacious (Profile 5) teachers, for mentoring may be a part of remediation activities initiated by admin-
upper secondary teachers, no significant differences were found. istrators for underperforming teaching staff (Goldstein, 2007). These
Efficacious teachers are better able to cope with problem behavior, use teachers may have existing doubts about their teaching capabilities, and
more constructive behavior management strategies, and implement the assignment of a mentor may lead to further doubts about cap-
more learner-centered teaching approaches (Almog and Shechtman, abilities. Teachers who served as a mentor were more likely to be in the
2007; Morris-Rothschild & Brassard, 2006; Nie et al., 2013), all which Highly-Efficacious profile than in the Globally-Unconfident, Student-
may be more important for maintaining a positive climate in lower Engagement-Efficacious (Profile 5), and Student-Engagement-In-
secondary classrooms typically characterized by greater teacher control efficacious (Profile 6) profiles, and were more likely to be in the
and discipline and fewer positive teacher-student relationships (Eccles Moderate-Globally-and-Instructionally-Confident profile than the
et al., 1993). This interpretation is supported by observed teaching level Globally-Unconfident profile. These teachers were also less likely to be
differences in profile-membership-job satisfaction relations. Specifi- in the Highly-Inefficacious (Profile 1) profile than in all other profiles.
cally, whereas Highly-Efficacious (Profile 3) lower secondary teachers Taken together, these findings are consistent with the results of prior
reported greater job satisfaction than Student-Engagement-Efficacious variable-centered research showing positive associations of mentoring
teachers, for upper secondary teachers no significant difference on with mentor teachers’ self-efficacy beliefs (Hobson, Ashby, Malderez, &

199
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

Tomlinson, 2009; Yost, 2002). Notably, we extend this existing vari- general teacher confidence and specific self-efficacy for classroom
able-centered work by showing that serving as a mentor is predictive of management and instructional strategies (Tschannen-Moran &
teachers’ possession of varying configurations of teaching self-efficacy Woolfolk Hoy, 2007).
beliefs. For mentor teachers, the experience of mentoring may serve as Teachers’ perceived need for individualized learning professional
a crucial learning experience through self-and-critical-reflection development was also implicated in profile membership. Teachers who
(Hobson et al., 2009), which may enhance their confidence beliefs perceived a greater need for this professional development had a sig-
(Bandura, 1997). nificantly greater likelihood of membership in the Student-
Relations of teachers’ professional development participation in Engagement-Inefficacious (Profile 6) profile than the Highly-Efficacious
most domains with profile membership were non-significant. One (Profile 3) subgroup, and were also more likely to be in the Moderate-
reason for the absence of significant relations is that mere exposure to Globally-and-Instructionally-Confident (Profile 2) profile than the
professional development may not reflect the mastery experiences Highly-Efficacious (Profile 3), Globally-Unconfident (Profile 4), and
posited by Bandura (1997) to influence self-efficacy beliefs. Instead, as Student-Engagement-Efficacious (Profile 5) subgroups. Notably,
research has shown, teachers may need to use the new knowledge Student-Engagement-Inefficacious and Moderate-Globally-and-
learned during professional development in their classrooms to allow Instructionally-Confident teachers reported among the lowest levels of
for sufficient experiences of success, constituting mastery experiences self-efficacy for student engagement. These results suggest that other-
(Ross & Bruce, 2007; Tschannen-Moran & McMaster, 2009; Ross, 1994). wise efficacious teachers, who perceive that they require professional
Future research would do well to examine teachers’ fidelity of im- development to better individualize student learning, are more likely to
plementation and perceptions of mastery in implementing professional doubt their specific capabilities to engage students in the learning
domain knowledge as potential moderators of the professional devel- process (Dixon et al., 2014). More surprisingly, teachers who perceived
opment-profile membership relation (Ross & Bruce, 2007). Only the a greater need for individualized learning professional development
relation of individualized learning professional development with pro- were more likely to be in the Moderate-Globally-and-Instructionally-
file membership was significant. Results showed that teachers who Confident profile (Profile 2) than the Highly-Inefficacious (Profile 1)
participated in this professional development domain were about one profile. Importantly, both profiles are characterized by self-efficacy for
and half times as likely to be in Profile 3 (Highly-Efficacious) and student engagement well-below the sample average. However, for
Profile 5 (Student-Engagement-Efficacious) than in Profile 4 (Globally- Moderate-Globally-and-Instructionally-Confident teachers, confidence
Unconfident), and these relations were found to be invariant across in their capability to perform their professional role in general and
lower and upper secondary teachers. Notably, Profile 3 and Profile 5 are manage classrooms and implement instructional routines specifically
those with the highest levels of specific self-efficacy for student en- may preserve their investment in the profession (Klassen and Chiu,
gagement. The findings suggest that teachers who were involved in 2011), such that they seek to develop in areas of perceived weakness.
professional development designed to strengthen adaptation of curri- Contrariwise, Highly-Inefficacious teachers may be disenfranchised and
culum materials and activities to students’ skill levels and interests are uncommitted (Rots, Aelterman, Vlerick, & Vermeulen, 2007; Ware &
more likely to possess configurations of self-efficacy beliefs in which Kitsantas, 2007), and proceed in their professional roles without con-
beliefs about engaging students in the learning process are strong. This cern for professional development (Perera, Granziera, & McIlveen,
is consistent with not only prior work that has shown that greater dif- 2018).
ferentiated instruction professional development is positively asso- Teaching experience was also found to be linked with profile
ciated with teachers’ self-efficacy beliefs (Dixon, Yssel, McConnell, & membership, with the relations generalizable over teaching level. More
Hardin, 2014) but also the theorized domain-specificity of teachers’ experienced teachers had a greater likelihood of being in the Highly-
self-efficacy beliefs (Bandura, 1997) insofar as individualization of Inefficacious (Profile 1) and Highly-Efficacious (Profile 3) profiles than
learning is conceptually most relevant to the domain of student en- in Profile 6 (Student-Engagement-Inefficacious). Although these results
gagement (Brophy, 1986; Connor et al., 2009; Davalos and Griffin, may seem paradoxical, they are in line with theoretical considerations.
1999). As cultivating student engagement has only recently been a The results indicate that increasing years of experience may be asso-
focus in mainstream teacher professional learning (Tschannen-Moran & ciated with two distinct subpopulations of teachers: (a) those teachers
Woolfolk Hoy, 2007), even exposure to individualized learning pro- in the Highly-Inefficacious profile who experience uniformly low self-
fessional development activities, which may allow for vicarious efficacy beliefs in line with moving into a period of career disengage-
learning experiences, may be important to teachers’ development of ment in later career stages as they withdraw from professional activities
confidence to engage students (Tschannen-Moran & Hoy, 2007). and prepare for retirement (Huberman, 1989; Klassen and Chiu, 2010);
Significant differences in profile membership as a function of tea- and (b) those teachers in the Highly-Efficacious profile who maintain
chers’ perceived professional development needs were observed, and their occupational identity and remain confident in their ability to
these differences were found to generalize across teaching levels. perform their roles due, at least in part, to accumulated mastery ex-
Teachers who perceived a greater need for classroom management periences and job characteristics (e.g., ongoing professional challenges)
professional development were significantly more likely to be in the (Bandura, 1997). This multi-sub-population perspective is consistent
Highly-Inefficacious profile (Profile 1) than all other profiles. These with Bandura’s (1997) view that the ostensible decline in self-efficacy
teachers were also less likely to be in the Moderate-Globally-and- beliefs in later career stages may not be universal. These findings also
Instructionally-Confident (Profile 2) and Highly-Efficacious (Profile 3) align with Huberman’s (1989) teacher career development model,
profiles than the remaining profiles. Furthermore, teachers in the which proposes that some teachers at later career stages may experi-
Moderate-Globally-and-Instructionally-Confident and Highly- ence reinvigoration.
Efficacious subgroups perceived less need for pedagogical competence There were several associations of gender with profile membership,
professional development than those in the Globally-Unconfident which were found to generalize across lower and upper secondary
(Profile 4) and Student-Engagement-Efficacious (Profile 5) profiles. teachers. Female teachers were significantly less likely to be in the
These results align with prior variable-centered research showing that Highly-Inefficacious profile (Profile 1) than in all other profiles except
teachers who perceive the unavailability of resources report lower self- the Globally-Unconfident Profile (Profile 4). In addition, females were
efficacy (Tschannen-Moran & Hoy, 2007). Given the primacy of class- significantly more likely to be in the Moderate-Globally-and-
room management and pedagogical competencies to teachers’ profes- Instructionally-Confident (Profile 2) and Highly-Efficacious (Profile 3)
sional practice (Ross & Bruce, 2007), it is unsurprising that those tea- profiles than the Globally-Unconfident (Profile 4) and Student-
chers who perceived a greater need for development in these domains Engagement-Efficacious (Profile 5) profiles. These results are consistent
were those in profiles generally characterized by the lowest levels of with early variable-centered studies that showed that females report

200
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

higher global teacher self-efficacy than males (Raudenbush et al., teachers’ capability beliefs may not be effective (Klassen and Chiu,
1992). The higher general confidence beliefs of female teachers have 2010). Instead, principals and other administrators may use informa-
been attributed to gender-role socialization experiences. Based on cul- tion gleaned from the self-efficacy profiles to identify professional de-
turally shared expectations about gender-appropriate professional roles velopment opportunities that are aligned with teachers’ specific self-
(Drudy, 2008), stakeholders may have more favorable expectations for, efficacy configurations. For instance, teachers who show a configura-
and react more favorably to, female teachers’ performance attainments. tion consistent with the Student-Engagement-Inefficacious profile may
As Lent et al. (2002) note, the beliefs of these social agents have im- be targeted for professional development related to strategies to moti-
plications for the development of self-efficacy beliefs as they may lead vate students. What is clear is that tailoring interventions at the level of
to differential opportunities for skill development as well as the inter- latent teacher self-efficacy profiles is likely to be more feasible and
nalization of different performance standards. However, female tea- economically sustainable, especially when offered on a large-scale as a
chers were less likely to be in profiles characterized by high levels of part of statutory professional training, than individual interventions,
self-efficacy for student engagement where other levels of self-efficacy but much more targeted and precise than one-size-fits-all interventions
beliefs were low. Although this may seem at odds with gender-sociali- (Perera & McIlveen, 2017). Information about most-likely latent profile
zation perspectives insofar as cultural-bound expectations for the membership may also be used by individual teachers in their voluntary
communal social roles of females should may lead them to have professional development. Moreover, such professional development is
stronger self-efficacy beliefs for engaging students (Ross et al., 1996), typically required to be implemented over a sustained period of time to
given the primacy of caregiving to the performance of teaching roles be effective (Boyle, Lamprianou, & Boyle, 2005). However, these pre-
across domains, such expectations may have more relevance for self- sent findings remain exploratory, and further work validating the pro-
beliefs concerning teachers’ capability to carry out their roles in gen- files obtained is required, particularly with respect to observable tea-
eral. cher behaviors and student outcomes, before they can be used to tailor
professional development provisions.
4.3. Limitations, implications, and conclusions
Appendix A. Supplementary material
Limitations to this research merit attention. First, though we relied
on social cognitive perspectives to theorize predictors and outcomes of Supplementary data to this article can be found online at https://
teacher self-efficacy profile membership, data were cross-sectional, doi.org/10.1016/j.cedpsych.2019.02.006.
thereby precluding directionality inferences. Research is required to
examine longitudinal relations of these predictors and outcomes. References
Second, we did not account for the reasons for professional develop-
ment participation (mandated vs. voluntary) in examining the relations Aiken, L. H., Clarke, S. P., Sloane, D. M., Sochalski, J. A., Busse, R., Clarke, H., ... Shamian,
of these provisions with profile membership. The reason for participa- J. (2001). Nurses’ reports on hospital care in five countries. Health Affairs, 20(3),
43–53. https://doi.org/10.1377/hlthaff.20.3.43.
tion may affect how teachers engage with professional development Almog, O., & Shechtman, Z. (2007). Teachers' democratic and efficacy beliefs and styles
and, by implication, impact the effects of participation on profile of coping with behavioural problems of pupils with special needs. European Journal of
membership. Third, the study is limited by the use of the TSES-SF, Special Needs Education, 22(2), 115–129.
Arens, A. K., & Morin, A. J. (2017). Improved representation of the self-perception profile
which is designed to reflect an intermediate level of domain-specificity. for children through bifactor exploratory structural equation modeling. American
However, it would seem reasonable that teacher self-efficacy may differ Educational Research Journal, 54(1), 59–87. https://doi.org/10.3102/
as a function of the specific subjects or groups of student taught or the 0002831216666490.
Asparouhov, T. (2005). Sampling weights in latent variable modeling. Structural Equation
specific tasks within a broad domain (Tschannen-Moran et al., 1998; Modeling: A Multidisciplinary Journal, 12(3), 411–434. https://doi.org/10.1207/
Ross et al., 1996), and future research would do well to employ mea- s15328007sem1203_4.
sures of teacher self-efficacy with a greater degree of specificity. An- Asparouhov, T., & Muthén, B. (2009). Exploratory structural equation modeling.
Structural Equation Modeling: A Multidisciplinary Journal, 16(3), 397–438. https://doi.
other possibility is examining profiles of collective teacher efficacy,
org/10.1080/10705510903008204.
reflecting unobserved heterogeneity in the collective perceptions of Bandura, A. (1997). Self-efficacy: The exercise of control. New York, NY: Freeman &
school capacity. A third limitation is our use of self-reported predictors Company.
and outcomes of profile membership as well as self-reported teacher Bauer, D. J., & Curran, P. J. (2003). Distributional assumptions of growth mixture models:
Implications for overextraction of latent trajectory classes. Psychological Methods,
self-efficacy. Although self-reports are the most common and arguably 8(3), 338–363. https://doi.org/10.1037/1082-989X.8.3.338.
the most appropriate method to assess teacher self-efficacy beliefs, job Billingsley, B. S., & Cross, L. H. (1992). Predictors of commitment, job satisfaction, and
satisfaction, and professional development needs, other constructs and intent to stay in teaching: A comparison of general and special educators. The Journal
of Special Education, 25(4), 453–471. https://doi.org/10.1177/
variables, such as classroom disciplinary climate and teacher colla- 002246699202500404.
boration as well as professional development participation and men- Boyle, B., Lamprianou, I., & Boyle, T. (2005). A longitudinal study of teacher change:
toring activities, may be assessed using more objective indices. The What makes professional development effective? Report of the second year of the study,
School Effectiveness and School Improvement, 16(1), 1–27. https://doi.org/10.1080/
reliance on self-reports may have induced biases in ratings of these 09243450500114819.
variables and resulted in the inflation of observed relations due to Brennan, M. D., & Robison, C. (1995). Gender Comparison of Teachers' Sense of Efficacy.
common-method biases. A final limitation concerns the generalizability Washington, DC: ERIC Digests [ED: 384288].
Brophy, J. (1986). Teacher influences on student achievement. American Psychologist,
of the profiles obtained. Although findings support the replicability of 41(10), 10691077. https://doi.org/10.1037/0003-066X.41.10.1069.
the teacher self-efficacy profiles across lower and upper secondary Caprara, G. V., Barbaranelli, C., Borgogni, L., & Steca, P. (2003). Efficacy beliefs as de-
teachers, this evidence should only be considered an initial step in terminants of teachers' job satisfaction. Journal of Educational Psychology, 95(4),
821–832. https://doi.org/10.1037/0022-0663.95.4.821.
determining profile generalizability. Teacher self-efficacy theory would
Caprara, G. V., Barbaranelli, C., Steca, P., & Malone, P. S. (2006). Teachers' self-efficacy
be well served by examining the extent to which the configurations beliefs as determinants of job satisfaction and students' academic achievement: A
generalize across (a) groups defined by different socio-demographic study at the school level. Journal of School Psychology, 44(6), 473–490. https://doi.
characteristics, such as gender and stages of career development org/10.1016/j.jsp.2006.09.001.
Carney, M. B., Brendefur, J. L., Thiede, K., Hughes, G., & Sutton, J. (2016). Statewide
(Klassen and Chiu (2010)), as well as (b) different educational (e.g., mathematics professional development: Teacher knowledge, self-efficacy, and beliefs.
elementary vs. secondary; tertiary vs. secondary) and cultural (e.g., Educational Policy, 30(4), 539–572. https://doi.org/10.1177/0895904814550075.
cross-country) contexts. Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement in-
variance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504.
The results of this study have tentative implications for teacher https://doi.org/10.1080/10705510701301834.
professional development. Support for the self-efficacy profiles suggests Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing
that a “one-size-fits-all” approach to the professional development of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal,

201
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

9(2), 233–255. https://doi.org/10.1207/S15328007SEM0902_5. supporting new teachers: Associations with self-efficacy, reflection, and quality.
Connor, C. M., Piasta, S. B., Fishman, B., Glasney, S., Schatschneider, C., Crowe, E., ... Mentoring & Tutoring: Partnership in Learning, 20(3), 303–323. https://doi.org/10.
Morrison, F. J. (2009). Individualizing student instruction precisely: Effects of child 1080/13611267.2012.701959.
by instruction interactions on first graders’ literacy development. Child Development, Lubke, G., & Neale, M. C. (2006). Distinguishing between latent classes and continuous
80(1), 77–100. https://doi.org/10.1111/j.1467-8624.2008.01247.x. factors: Resolution by maximum likelihood? Multivariate Behavioral Research, 41(4),
Darr, W., & Johns, G. (2008). Work strain, health, and absenteeism: A meta-analysis. 499–532. https://doi.org/10.1207/s15327906mbr4104_4.
Journal of Occupational Health Psychology, 13(4), 293–318. https://doi.org/10.1037/ Lubke, G., & Neale, M. (2008). Distinguishing between latent classes and continuous
a0012639. factors with categorical outcomes: Class invariance of parameters of factor mixture
Davalos, R., & Griffin, G. (1999). The impact of teachers’ individualized practices on models. Multivariate Behavioral Research, 43(4), 592–620. https://doi.org/10.1080/
gifted students in rural, heterogeneous classrooms. Roeper Review, 21(4), 308–314. 00273170802490673.
https://doi.org/10.1080/02783199909553982. Marsh, H. W., Lüdtke, O., Trautwein, U., & Morin, A. J. (2009). Classical latent profile
Dixon, F. A., Yssel, N., McConnell, J. M., & Hardin, T. (2014). Differentiated instruction, analysis of academic self-concept dimensions: Synergy of person-and variable-cen-
professional development, and teacher efficacy. Journal for the Education of the Gifted, tered approaches to theoretical models of self-concept. Structural Equation Modeling,
37(2), 111–127. https://doi.org/10.1177/0162353214529042. 16(2), 191–225. https://doi.org/10.1080/10705510902751010.
Drudy, S. (2008). Gender balance/gender bias: The teaching profession and the impact of McLennan, B., McIlveen, P., & Perera, H. N. (2017). Pre-service teachers' self-efficacy
feminisation. Gender and Education, 20(4), 309–323. https://doi.org/10.1080/ mediates the relationship between career adaptability and career optimism. Teaching
09540250802190156. and Teacher Education, 63, 176–185. https://doi.org/10.1016/j.tate.2016.12.022.
Duffin, L. C., French, B. F., & Patrick, H. (2012). The Teachers' Sense of Efficacy Scale: Millsap, R. E., & Yun-Tein, J. (2004). Assessing factorial invariance in ordered-categorical
Confirming the factor structure with beginning pre-service teachers. Teaching and measures. Multivariate Behavioral Research, 39(3), 479–515. https://doi.org/10.1207/
Teacher Education, 28(6), 827–834. https://doi.org/10.1016/j.tate.2012.03.004. S15327906MBR3903_4.
Durksen, T. L., Klassen, R. M., & Daniels, L. M. (2017). Motivation and collaboration: The Morin, A. J., Arens, A. K., & Marsh, H. W. (2016). A bifactor exploratory structural
keys to a developmental framework for teachers’ professional learning. Teaching and equation modeling framework for the identification of distinct sources of construct-
Teacher Education, 67, 53–66. https://doi.org/10.1016/j.tate.2017.05.011. relevant psychometric multidimensionality. Structural Equation Modeling: A
Dunn, K. E., & Rakes, G. C. (2011). Teaching teachers: An investigation of beliefs in Multidisciplinary Journal, 23(1), 116–139. https://doi.org/10.1080/10705511.2014.
teacher education students. Learning Environments Research, 14(1), 39–58. https:// 961800.
doi.org/10.1007/s10984-011-9083-1. Morin, A. J., Boudrias, J. S., Marsh, H. W., McInerney, D. M., Dagenais-Desmarais, V.,
Eccles, J. S., Midgley, C., Wigfield, A., Buchanan, C. M., Reuman, D., Flanagan, C., & Mac Madore, I., & Litalien, D. (2017). Complementary variable-and person-centered ap-
Iver, D. (1993). Development during adolescence: The impact of stage-environment proaches to the dimensionality of psychometric constructs: Application to psycho-
fit on young adolescents' experiences in schools and in families. American Psychologist, logical wellbeing at work. Journal of Business and Psychology, 32(4), 395–419. https://
48(2), 90–101. https://doi.org/10.1037/0003-066X.48.2.90. doi.org/10.1007/s10869-016-9448-7.
Fives, H., & Buehl, M. M. (2009). Examining the factor structure of the teachers’ sense of Morin, A. J., & Marsh, H. W. (2015). Disentangling shape from level effects in person-
efficacy scale. The Journal of Experimental Education, 78(1), 118–134. https://doi.org/ centered analyses: An illustration based on university teachers’ multidimensional
10.1080/00220970903224461. profiles of effectiveness. Structural Equation Modeling: A Multidisciplinary Journal,
Goddard, Y. L., & Kim, M. (2018). Examining connections between teacher perceptions of 22(1), 39–59. https://doi.org/10.1080/10705511.2014.919825.
collaboration, differentiated instruction, and teacher efficacy. Teachers College Morin, A. J., Meyer, J. P., Creusier, J., & Biétry, F. (2016). Multiple-group analysis of
Record, 120(1), 1–24. similarity in latent profile solutions. Organizational Research Methods, 19(2), 231–254.
Goddard, R. D., Hoy, W. K., & Woolfolk-Hoy, A. (2000). Collective teacher efficacy: Its https://doi.org/10.1177/1094428115621148.
meaning, measure, and impact on student achievement. American Educational Morris-Rothschild, B. K., & Brassard, M. R. (2006). Teachers' conflict management styles:
Research Journal, 37(2), 479–507. https://doi.org/10.3102/00028312037002479. The role of attachment styles and classroom management efficacy. Journal of School
Goldstein, J. (2007). Easy to dance to: Solving the problems of teacher evaluation with Psychology, 44(2), 105–121. https://doi.org/10.1016/j.jsp.2006.01.004.
peer assistance and review. American Journal of Education, 113(3), 479–508. https:// Muthén, B. (2001). Latent variable mixture modeling. In G. A. Marcoulides, & R. E.
doi.org/10.1086/512741. Schumacker (Eds.). New developments and techniques in structural equation modeling
Granziera, H., & Perera, H. N. (2019). Relations among teachers' self-efficacy beliefs, (pp. 1–33). Mahwah, NJ: Lawrence Erlbaum.
engagement, and work satisfaction: A social cognitive view. Contemporary Educational Muthén, L. K., & Muthén, B. O. (1998–2015). Mplus user’s guide, 7th ed., Los Angeles, CA:
Psychology. https://doi.org/10.1016/j.cedpsych.2019.02.003 (Advance online Muthén & Muthén.
publication). Nie, Y., Tan, G. H., Liau, A. K., Lau, S., & Chua, B. L. (2013). The roles of teacher efficacy
Henson, R. K. (2001). The effects of participation in teacher research on teacher efficacy. in instructional innovation: Its predictive relations to constructivist and didactic in-
Teaching and Teacher Education, 17(7), 819–836. https://doi.org/10.1016/S0742- struction. Educational Research for Policy and Practice, 1–11. https://doi.org/10.1007/
051X(01)00033-6. s10671-012-9128-y.
Henson, J. M., Reise, S. P., & Kim, K. H. (2007). Detecting mixtures from structural model Ning, H. K., Lee, D., & Lee, W. O. (2015). Relationships between teacher value orienta-
differences using latent variable mixture modeling: A comparison of relative model fit tions, collegiality, and collaboration in school professional learning communities.
statistics. Structural Equation Modeling: A Multidisciplinary Journal, 14(2), 202–226. Social Psychology of Education, 18(2), 337–354. https://doi.org/10.1007/s11218-015-
https://doi.org/10.1080/10705510709336744. 9294-x.
Hobson, A. J., Ashby, P., Malderez, A., & Tomlinson, P. D. (2009). Mentoring beginning Ning, B., Van Damme, J., Van Den Noortgate, W., Yang, X., & Gielen, S. (2015). The
teachers: What we know and what we don't. Teaching and Teacher Education, 25(1), influence of classroom disciplinary climate of schools on reading achievement: A
207–216. https://doi.org/10.1016/j.tate.2008.09.001. cross-country comparative study. School Effectiveness and School Improvement, 26(4),
Holloway, J. H. (2001). The benefits of mentoring. Educational Leadership, 58(8), 85–86. 586–611. https://doi.org/10.1080/09243453.2015.1025796.
Holzberger, D., Philipp, A., & Kunter, M. (2013). How teachers’ self-efficacy is related to Nylund, K. L., Asparouhov, T., & Muthén, B. O. (2007). Deciding on the number of classes
instructional quality: A longitudinal analysis. Journal of Educational Psychology, in latent class analysis and growth mixture modeling: A Monte Carlo simulation
105(3), 774. https://doi.org/10.1037/a0032198. study. Structural Equation Modeling: A Multidisciplinary Journal, 14(4), 535–569.
Huberman, M. (1989). The professional life cycle of teachers. Teachers College Record, 91, https://doi.org/10.1080/10705510701575396.
31–57. Organisation for Economic Co-operation and Development (2014). TALIS 2013 technical
Klassen, R. M., Bong, M., Usher, E. L., Chong, W. H., Huan, V. S., Wong, I. Y., & Georgiou, report. Paris: Author Available from: https://www.oecd.org/edu/school/TALIS-
T. (2009). Exploring the validity of a teachers’ self-efficacy scale in five countries. technical-report-2013.pdf.
Contemporary Educational Psychology, 34(1), 67–76. https://doi.org/10.1016/j. Perera, H. N. (2016). Construct validity of the social provisions scale: A bifactor ex-
cedpsych.2008.08.001. ploratory structural equation modeling approach. Assessment, 23(6), 720–733.
Klassen, R. M., & Chiu, M. M. (2010). Effects on teachers' self-efficacy and job satisfaction: https://doi.org/10.1177/1073191115589344.
Teacher gender, years of experience, and job stress. Journal of Educational Psychology, Perera, H. N., Granziera, H., & McIlveen, P. (2018). Profiles of teacher personality and
102(3), 741756. https://doi.org/10.1037/a0019237. relations with teacher self-efficacy, work engagement, and job satisfaction.
Klassen, R. M., & Chiu, M. M. (2011). The occupational commitment and intention to quit Personality and Individual Differences, 120, 171–178. https://doi.org/10.1016/j.paid.
of practicing and pre-service teachers: Influence of self-efficacy, job stress, and 2017.08.034.
teaching context. Contemporary Educational Psychology, 36(2), 114–129. https://doi. Perera, H. N., & McIlveen, P. (2018). Vocational interest profiles: Profile replicability and
org/10.1016/j.cedpsych.2011.01.002. relations with the STEM major choice and the Big-Five. Journal of Vocational Behavior.
Klassen, R. M., & Tze, V. M. (2014). Teachers’ self-efficacy, personality, and teaching https://doi.org/10.1016/j.jvb.2017.11.012.
effectiveness: A meta-analysis. Educational Research Review, 12, 59–76. https://doi. Perera, H. N., McIlveen, P., Burton, L. J., & Corser, D. M. (2015). Beyond congruence
org/10.1016/j.edurev.2014.06.001. measures for the evaluation of personality factor structure replicability: An ex-
Lent, R. W., & Brown, S. D. (2006). Integrating person and situation perspectives on work ploratory structural equation modeling approach. Personality and Individual
satisfaction: A social cognitive view. Journal of Vocational Behavior, 69(2), 236–247. Differences, 84, 23–29. https://doi.org/10.1016/j.paid.2015.01.004.
https://doi.org/10.1016/j.jvb.2006.02.006. Perera, H. N., Vosicka, L., Granziera, H., & McIlveen, P. (2018). Towards an integrative
Lent, R. W., Brown, S. D., & Hackett, G. (2002). Social cognitive career theory. In D. perspective on the structure of teacher work engagement. Journal of Vocational
Brown (Ed.). Career choice and development (pp. 225–311). (4th ed.). San Francisco, Behavior, 108, 28–41. https://doi.org/10.1016/j.jvb.2018.05.006.
CA: Jossey-Bass. Perera, H. N., Wiens, P. D., McIlveen, P., Calkins, C., & McLennan, B. (2019). A unifying
Litalien, D., Morin, A. J., & McInerney, D. M. (2017). Achievement goal profiles among perspective on the dimensionality of teacher self-efficacy: Evidence from Australian
adolescent males and females. Developmental Psychology, 53(4), 731–751. https://doi. and US teachers. [submitted for publication].
org/10.1037/dev0000288. Raudenbush, S. W., Rowan, B., & Cheong, Y. F. (1992). Contextual effects on the self-
LoCasale-Crouch, J., Davis, E., Wiens, P., & Pianta, R. (2012). The role of the mentor in perceived efficacy of high school teachers. Sociology of Education, 65(2), 150–167.

202
H.N. Perera, et al. Contemporary Educational Psychology 58 (2019) 186–203

Raykov, T., & Marcoulides, G. A. (2004). Using the delta method for approximate interval 1016/j.tate.2009.11.001.
estimation of parameter functions in SEM. Structural Equation Modeling: A Skaalvik, E. M., & Skaalvik, S. (2014). Teacher self-efficacy and perceived autonomy:
Multidisciplinary Journal, 11(4), 621–637. https://doi.org/10.1207/ Relations with teacher engagement, job satisfaction, and emotional exhaustion.
s15328007sem1104_7. Psychological Reports, 114(1), 68–77. https://doi.org/10.2466/14.02.PR0.114k14w0.
Reyes, M. R., Brackett, M. A., Rivers, S. E., White, M., & Salovey, P. (2012). Classroom Tofighi, D., & Enders, C. K. (2008). Identifying the correct number of classes in growth
emotional climate, student engagement, and academic achievement. Journal of mixture models. In G. R. Hancock, & K. M. Samuelsen (Eds.). Advances in latent
Educational Psychology, 104(3), 700–712. https://doi.org/10.1037/a0027268. variable mixture models (pp. 317–341). Charlotte, NC: Information Age Publishing.
Richter, D., Kunter, M., Lüdtke, O., Klusmann, U., Anders, Y., & Baumert, J. (2013). How Tschannen-Moran, M., & McMaster, P. (2009). Sources of self-efficacy: Four professional
different mentoring approaches affect beginning teachers' development in the first development formats and their relationship to self-efficacy and implementation of a
years of practice. Teaching and Teacher Education, 36, 166–177. https://doi.org/10. new teaching strategy. The Elementary School Journal, 110(2), 228–245. https://doi.
1016/j.tate.2013.07.012. org/10.1086/605771.
Ross, J. A. (1994). The impact of an inservice to promote cooperative learning on the Tschannen-Moran, M., Woolfolk Hoy, A., & Hoy, W. K. (1998). Teacher efficacy: Its
stability of teacher efficacy. Teaching and Teacher Education, 10(4), 381–394. https:// meaning and measure. Review of Educational Research, 68(2), 202–248. https://doi.
doi.org/10.1016/0742-051X(94)90020-5. org/10.3102/00346543068002202.
Ross, J., & Bruce, C. (2007). Professional development effects on teacher efficacy: Results Tschannen-Moran, M., & Woolfolk Hoy, A. (2001). Teacher efficacy: Capturing an elusive
of randomized field trial. The Journal of Educational Research, 101(1), 50–60. https:// construct. Teaching and Teacher Education, 17(7), 783–805. https://doi.org/10.1016/
doi.org/10.3200/JOER.101.1.50-60. S0742-051X(01)00036-1.
Ross, J. A., Cousins, J. B., & Gadalla, T. (1996). Within-teacher predictors of teacher Tschannen-Moran, M., & Woolfolk Hoy, A. (2007). The differential antecedents of self-
efficacy. Teaching and Teacher Education, 12(4), 385–400. https://doi.org/10.1016/ efficacy beliefs of novice and experienced teachers. Teaching and Teacher Education,
0742-051X(95)00046-M. 23(6), 944–956. https://doi.org/10.1016/j.tate.2006.05.003.
Ross, J. A., Hogaboam-Gray, A., & Hannay, L. (2001). Effects of teacher efficacy on Tsouloupas, C. N., Carson, R. L., Matthews, R., Grawitch, M. J., & Barber, L. K. (2010).
computer skills and computer cognitions of Canadian students in grades K-3. The Exploring the association between teachers’ perceived student misbehaviour and
Elementary School Journal, 102(2), 141–156. https://doi.org/10.1086/499697. emotional exhaustion: The importance of teacher efficacy beliefs and emotion reg-
Rots, I., Aelterman, A., Vlerick, P., & Vermeulen, K. (2007). Teacher education, graduates’ ulation. Educational Psychology, 30(2), 173–189. https://doi.org/10.1080/
teaching commitment and entrance into the teaching profession. Teaching and 01443410903494460.
Teacher Education, 23(5), 543–556. https://doi.org/10.1016/j.tate.2007.01.012. Ware, H., & Kitsantas, A. (2007). Teacher and collective efficacy beliefs as predictors of
Rotter, J. B. (1966). Generalized expectancies for internal versus external control of re- professional commitment. The Journal of Educational Research, 100(5), 303–310.
inforcement. Psychological Monographs, 80(1), 1–28. https://doi.org/10.1037/ https://doi.org/10.3200/JOER.100.5.303-310.
h0092976. Watson, G. (2006). Technology professional development: Long-term effects on teacher
Rutkowski, D., Rutkowski, L., Belanger, J., Knowll, S., Weatherby, K., & Prusinski, E. self-efficacy. Journal of Technology and Teacher Education, 14(1), 151–165.
(2013). Teaching and learning international survey 2013: Conceptual framework. Paris: Wolters, C. A., & Daugherty, S. G. (2007). Goal structures and teachers' sense of efficacy:
Organisation for Economic Co-operation and Development Available from http:// Their relation and association to teaching experience and academic level. Journal of
www.oecd.org/edu/school/TALIS%202013%20Conceptual%20Framework.pdf. Educational Psychology, 99(1), 181–193. https://doi.org/10.1037/0022-0663.99.1.
Ryan, A. M., Kuusinen, C. M., & Bedoya-Skoog, A. (2015). Managing peer relations: A 181.
dimension of teacher self-efficacy that varies between elementary and middle school Woolfolk Hoy, A., & Spero, R. B. (2005). Changes in teacher efficacy during the early
teachers and is associated with observed classroom quality. Contemporary Educational years of teaching: A comparison of four measures. Teaching and Teacher Education,
Psychology, 41, 147–156. https://doi.org/10.1016/j.cedpsych.2015.01.002. 21(4), 343–356. https://doi.org/10.1016/j.tate.2005.01.007.
Scherer, R., Jansen, M., Nilsen, T., Areepattamannil, S., & Marsh, H. W. (2016). The Quest Yost, R. (2002). “I think I can”: Mentoring as a means of enhancing teacher efficacy. The
for Comparability: Studying the Invariance of the Teachers’ Sense of Self-Efficacy Clearing House: A Journal of Educational Strategies, Issues, and Ideas, 75(4), 195–197.
(TSES) Measure across Countries. PLoS ONE, 11(3), e0150829. https://doi.org/10. https://doi.org/10.1080/00098650209604930.
1371/journal.pone.0150829. Zee, M., & Koomen, H. M. (2016). Teacher self-efficacy and its effects on classroom
Skaalvik, E. M., & Skaalvik, S. (2007). Dimensions of teacher self-efficacy and relations processes, student academic adjustment, and teacher well-being: A synthesis of 40
with strain factors, perceived collective teacher efficacy, and teacher burnout. Journal years of research. Review of Educational Research, 86(4), 981–1015. https://doi.org/
of Educational Psychology, 99(3), 611–625. https://doi.org/10.1037/0022-0663.99.3. 10.3102/0034654315626801.
611. Zee, M., Koomen, H. M., Jellesma, F. C., Geerlings, J., & de Jong, P. F. (2016). Inter-and
Skaalvik, E. M., & Skaalvik, S. (2010). Teacher self-efficacy and teacher burnout: A study intra-individual differences in teachers' self-efficacy: A multilevel factor exploration.
of relations. Teaching and Teacher Education, 26(4), 1059–1069. https://doi.org/10. Journal of School Psychology, 55, 39–56. https://doi.org/10.1016/j.jsp.2015.12.003.

203

You might also like