Educational Administration Quarterly

文章來源：http://eaq.sagepub.com/
How Principals and Peers
Influence Teaching and
Learning
Jonathan Supovitz,1 Philip Sirinides,1
and Henry May1
Abstract
This paper examines the effects of principal leadership and peer teacher
influence on teachers’ instructional practice and student learning. Using teacher
survey and student achievement data from a mid-sized urban southeastern
school district in the United States in 2006-2007, the study employs
multilevel structural equation modeling to examine the structural relationships
between student learning and theorized dimensions of principal leadership,
teacher peer influence, and change in teachers’ instructional practice. The findings
confirm previous empirical work and provide new contributions to research
on the chain of hypothesized relationships between leadership practice and
student learning. Both principal leadership and teacher peer influence were
significantly associated with teachers’ instructional practices and English language
arts (ELA) student learning. A major contribution of this research is the strong
and significant indirect relationships which mediate education leadership and
student learning. The results indicate the importance of principals work for
student learning because of their indirect influence on teachers’ practices
through the fostering of collaboration and communication around instruction.
Keywords
leadership impacts, distributed leadership, instructional improvement, student
learning, multilevel structural modeling
-----------------------------------------------------------
1University of Pennsylvania, Philadelphia, PA
Corresponding Author:
Jonathan Supovitz, University of Pennsylvania, Graduate School of Education, 3700 Walnut
Street #404, Philadelphia, PA 19104
-----------------------------------------------------------------

The literature on the effects of school leadership on student learning stretches
back for at least 40 years. The accumulation of that literature suggests that
although principals can have a detectable effect on student performance, their
effects are mostly mediated through other aspects of school life that influence
what and how teachers teach in classrooms. More recent explorations of leadership
have incorporated a range of other leadership activities in schools—mostly
leadership enacted by teachers and other “informal” school leaders—that influence
instructional practice.
In this study we combine these two trends and examine the effects of both
principal leadership and peer influence on teachers’ instructional practice and
student learning. Using a data set collected from a school district in the southeastern
United States that allows us to connect teachers’ survey data to students’
learning outcomes, we are able to examine the relationships between both
teacher perceptions of principal practice as well as peer influence on student
learning, as mediated by instructional practice.
We find that both leadership practice and peer influence are related to teacher
instructional practice, which, in our data set, is significantly related to English
language arts (ELA) achievement but not mathematics achievement. Furthermore,
teacher reports of peer influence had an equivalent influence in ELA
and a 2 times greater impact in mathematics on teachers’ practice than do
teacher reports of principal leadership activity. However, principal leadership
also influences instructional practice indirectly by significantly affecting how
teachers report the influence of their peers.
Literature Review
Principal Leadership and Its Effects on Student Achievement
There have been several thorough reviews of the literature on the relationship
between school leadership—mostly defined as the efforts and activities of
school principals—and student outcomes. Hallinger and Heck (1998) synthesized
43 studies conducted between 1980 and 1995 that investigated evidence of
the relationship between principal leadership and student achievement. They
organized the studies into three categories: direct effects of leadership practice
on student outcomes; mediated effects studies, in which principal leadership
was mediated by other people, events, or organizational factors; and reciprocal
effect studies, in which the relationships between leadership efforts and school
and environmental factors were interactive. The authors saw little evidence of
direct effects and few examples of reciprocal effects studies, with most evidence
pointing to indirect effects. They concluded that principals have a measurable,
but indirect, effect on school effectiveness and student achievement.
A second synthesis of the literature on the relationship between school leadership
practices and student outcomes was conducted by Waters, Marzano, and
McNulty (2003), who synthesized 70 research studies relating principal leadership
to student achievement that were conducted from the early 1970s
through the early 2000s. The studies they examined looked at a wide array of
leadership responsibilities, including a focus on school culture, faculty motivation,
instructional support, and emphasis on accountability. They produced
effects sizes for each of the different dimensions of leadership that were examined.
Across these disparate studies, they found an average effect size of .25
and concluded that “there is, in fact, a substantial relationship between leadership
and student achievement” (p. 3).
Witziers, Bosker, and Kruger (2003) conducted a quantitative meta-analysis
of studies that looked at the overall effects of school leadership on student
learning as well as studies that examined the impact of specific principal behaviors
on student outcomes. They found small direct effects across studies of
elementary school principal leadership but no detectable direct impacts of
secondary school principal leadership. They found larger effects, although with
more variability, in studies of more specific leadership behaviors.
A more holistic analysis of a wide range of leadership literature was conducted
by Leithwood, Seashore Louis, Anderson, and Wahlstrom (2004). They
developed a conceptual model of how leadership at different levels of the
education system (state, district, other stakeholders) influenced school leadership,
which interacted with school and student conditions to produce student
outcomes. Through a synthesis of both the quantitative and qualitative studies
of these factors, they concluded that school leadership “is second only to
teaching among school-related factors in its impact on student learning” (p. 5).
One particular empirical study of principal leadership was particularly relevant
to our work, because of both its focus and the methods it employed.
Hallinger, Bickman, and Davis (1996) examined the relationship between
principal leadership and student reading achievement using structural equation
modeling (SEM). SEM allowed them to simultaneously test the independent
effects of multiple antecedent and intervening variables. They found no direct
effects between indicators of principal leadership and student performance.
They then explored the ways that school and classroom variables mediated
the relationship between principal leadership and student achievement. They
found that principal leadership significantly predicted variables of instructional
climate and instructional organization and that those variables were
positively and significantly related to student achievement.
In summary, the accumulated literature on the relationship between principal
practice and student learning indicates two things. First is a confirmation
that principals can have a detectable effect on student learning outcomes. And
second, these effects are more likely to be mediated by other school and classroom
factors than directly by principal actions. This leads to two questions:
What are the key activities of principals that produce changes in classrooms
and students’ performance? And what are the contributions of other school
factors to student improvement?
Key Activities of School Principals
Underneath more global findings of principal support for improved instructional
practice and student learning are a myriad of explorations of what, more precisely,
principals do to produce these outcomes. In their systematic review of
the literature, Waters et al. (2003) provided a list of more than 20 leadership
activities that they found were statistically related to student learning. These
included such diverse activities as setting maintaining order and discipline;
fostering shared belief and cooperative community; securing resources;
involvement in the design and implementation of curriculum, instruction, and
assessment practices; monitoring the effectiveness of school practices; and
recognizing and awarding accomplishments.
To make matters more complex, context is also acknowledged to play
an important role in identifying the essential activities of school leadership. As
Hallinger et al. (1996) observe, “The task of unraveling the effects of administrative
practices on student learning has been complicated by the concurrent effects
that school contexts exert on principals” (p. 528). Several factors, including the
strengths of the principal, the makeup of the school faculty, and the context
facing the school, must be considered when attempting to identify effective leadership
practices. In spite of the challenges of isolating which of the many
emphases of principals best support improvements in teaching and learning, our
analysis of the evidence base points to three factors that seem to be commonly
referenced across the literature. The first factor is the role principals play in
focusing the mission and goals of the organization. The second factor is how
principals encourage an environment of collaboration and trust in the building.
The third factor that has been consistently related to improvements in teaching
and learning is the extent to which principals actively support instructional
improvement.
Setting mission and goals. Many researchers see the key task of principal leadership
to be setting the broad vision and mission of the organization and linking
goals to that mission. Leithwood (1996), for example, argued that setting organizational
direction was one of the core tasks of transformational leadership.
Hallinger and Murphy (1987) contended that instructional leadership focused
first on defining the school mission through a clear vision of what the school was
trying to accomplish. Similarly, Hallinger et al. (1996) identified establishing a
clear school mission as a central activity of instructional leadership. Witziers
et al. (2003) conducted a meta-analysis of seven leadership behaviors and found
“defining and communicating mission” to have the largest effect size of all those
they examined. Goldring and Pasternak (1994) studied principals’ activities and
found that the principals’ roles in framing school goals, establishing a clear mission,
and gaining staff consensus were strong predictors of school outcomes.
Encouraging trust and collaboration. Trust and collaboration point directly to the
cultural heart of the school organization, and many studies identify principals as a
central shaper of their schools’ culture. Bryk and Schneider (2002) used extensive
survey data and case studies in Chicago to examine the connections between what
they called “relational trust” and school outcomes, including student achievement.
They defined relational trust as the social exchanges in schools defined by respect,
personal regard, competence in core role responsibilities, and personal integrity.
Through their analyses, they found that the growth of relational trust in schools
“fuels the multiple strands of the school change process and thereby contributes to
improved student learning” (p. 121). They identified trust levels between the
school’s principal and teachers as a central indicator of trust. Heck, Larson, and
Marcoulides (1990) examined principal supervision and support of teachers. They
found that higher performing elementary and high school principals worked collaboratively
with teachers to coordinate their schools’ instructional programs and
solve instructional problems and supported staff development opportunities.
In their met-analysis, Waters et al. (2003) identified the fostering of shared
beliefs and a sense of community and cooperation to be one of the most significant
leadership predictors of student learning outcomes.
Active support of instruction. A final set of research on effective principal
leadership emphasizes the importance of both creating a learning ethos and
providing more hands-on support for instruction. Leithwood, Jantzi, Silins, and
Dart (1993) investigated how principals developed an instructional emphasis
in schools. Relevant to this review, they found that principals who focused on
developing an instructional vision, setting group goals, holding high expectations,
and providing individual support for teachers positively influenced
school culture and climate. In their review of the literature on leaderships
effects on student achievement, Waters et al. (2003) found leaders’ knowledge
of curriculum, instruction, and assessment to be a significant predictor of student
performance. Supovitz and Poglinco (2001) examined the instructional
leadership practices of urban school principals implementing a comprehensive
school reform model. They found that instructional leaders organized their
schools around an emphasis on instructional improvement supported by a distinct
vision of instructional quality, cultivated a community of instructional
practice in their schools by creating a safe and collaborative environment for
teachers to engage in and deepen their work, and reorganized their own professional
lives, time, and priorities to support instructional improvement.
Emerging Attention to Other Influential Actors in Schools
Concurrent with the recent research on principal leadership, an emerging trend in
the study of leadership looks beyond the principal toward an array of other actors
who either consistently, or situationally, take on a leadership role in schools. These
perspectives come from conceptions of leadership that arise out of authority and
influence. In this view, leadership is not exclusively positional but rather is rooted
in the act of establishing influence over others. Schneier and Goktepe (1983)
define such informal leadership as influence over other group members. Research
from organizational sociology indicates that informal leaders have a strong influence
on group processes, norms, and outcomes (Bass, 1990; Wheelan & Johnston,
1996). Pescosolido (2001) argues that informal leadership that develops within a
group plays a key role in defining the group’s sense of efficacy.
One of the foundational educational theorists on this topic, Peter Gronn (2000),
argues for a reallocation of the tasks and activities that constitute the division of
labor in schools toward a system of “joint performance.” James Spillane and colleagues
have written extensively about a distributed perspective on leadership. In
their view, leadership arises not from formal title or responsibility but rather out of
the interactions among individuals, tasks, and situations (Spillane, 2006; Spillane,
Hallett, & Diamond, 2003; Spillane, Halverson, & Diamond, 2001). Wider conceptions
of leadership have led to recent explorations of the role and influence of
informal leaders and teacher leaders (Mangin & Stoelinga, 2008; York-Barr &
Duke, 2004). Robinson (2008) contrasts views of distributed leadership that
emphasize tasks versus influence and theorizes that emphasizing influence makes
it harder to link distributed leadership to educational outcomes.
Several key factors are emerging within the literature on how teachers influence
their peers in educational settings. These include a collaborative interaction
of faculty around issues of teacher and learning and the development of instructional
advice networks. These are reviewed briefly below.
Active interaction among faculty around teaching and learning. An emerging finding
in the teacher leadership literature is that peers influence each other when they
engage in collaborative discussions about their professional work. In their review
of the literature on teacher leadership, York-Barr and Duke (2004) found that relationship
building and collaboration were the two foremost themes that emerged
when they synthesized the research on teacher leadership activity. LeBlanc
and Shelton (1997) identified collaboration as the primary means by which teachers
affected their peers. Wasley (1991) conducted a series of case studies of teacher
leaders. She found that those with the most influence worked collegially
with other teachers to examine instruction and its effects on student learning.
The Bryk and Schneider (2002) work on relational trust that was discussed earlier
also demonstrated the importance of teacher–teacher trust as a significant factor
in improving school communities and student learning outcomes.
One key strategy by which teachers influence their peers has come to
known as peer coaching (Joyce & Showers, 1995; Showers, 1984). Peer
coaching is a strategy to increase the transfer of professional development by
having teachers do sustained work on what they have learned in professional
development (Showers & Joyce, 1996). Two key elements of peer coaching
are to have teachers observe each other teaching and to examine student work
in relation to assignments. The latter is a central part such educational movements
as understanding by design, in which teachers backward map from
desired results to evidence for results to learning experiences and instruction
(Wiggins & McTighe, 2001).
Strong instructional advice networks. Another way that teachers influence
their peers is via instructional advice networks. Researchers are beginning to
unpack the ways in which teachers provide and seek assistance from each
other through social networks and the influence of these instructional networks
on school improvement efforts and outcomes (Frank, Zhao, & Borman,
2004; Supovitz, 2008). Based on theories of social capital, individual and
collective benefits accrue through dense and interrelated networks among
individuals (Coleman, 1997; Burt, 2000; Lin, 2001). Supovitz (2008) studied
school reform networks and found that most of the instructional support was
provided by teachers who did not hold formal leadership positions. Spillane
(2005) illustrated how leadership practice in primary schools was structured
differently depending on the content area. Weinbaum, Cole, Weiss, and Supovitz
(2008) examined communication networks in high schools implementing
external reforms. They found positive relationships between school communication
patterns and attitudes and behaviors in support of the reforms,
suggesting a relationship between communication and reform practice.
In sum, the literature on how peers influence each other in schools is in a
more nascent stage than the more mature literature base around principal
leadership. However, several important themes are emerging. Foremost is the
ways in which teachers have collaborative opportunities to interact around
issues of teaching and learning. Second are opportunities to observe each
other’s teaching and the resulting conversations. A third, and perhaps related,
trait of instructional interaction among peers is both formal and informal
instructional advice networks.
Conceptual Framework for This Study
Drawing on the different trends in the research on both principal leadership
and how teachers influence their peers in schools, we constructed a conceptual
framework that describes how these two latent factors influence instruction
and student learning. Our conceptual framework is depicted in Figure 1.

Principal leadership is conceptualized as a construct made up of leaders’
emphasis on mission and goals, emphasis on community and trust, and focus
on instruction. We call our second construct “peer influence” to emphasize
the act of teachers’ influencing their colleagues rather than its leadership quality.
In doing so we conceptualized peer influence as a latent factor composed of
instructional conversations, interactions among faculty members around
issues of teaching and learning, and instructional advice networks. These
three overlap, but we viewed them as conceptually distinct. We conceived of
instructional conversations as collegial discussions among peers about instructional
issues. We viewed interactions among faculty members around issues of
teaching and learning as particular acts such as conducting observations, providing
feedback, and reviewing student work together. And we conceptualized
advice networks as the specific seeking of instructional assistance from particular
peers. We also hypothesized principal leadership to unidirectionally affect
the extent to which teachers influenced their peers inside of schools, as represented
by the arrow going from principal leadership to peer influence. Both
principal leadership and peer influence are theorized to influence teachers’
instructional practice, which is conceived to be directly related to student learning
outcomes.

Research Method
Research Questions
Based on this theoretical model, we developed a set of five research questions,
stated below:
1. Is principal leadership associated with teacher change in instruction?
2. Is principal leadership associated with teacher peer influence?
3. What is the relative magnitude of the association of principal leadership
and peer influence with teacher change in instruction?
4. Is there a relationship between teacher change in instruction and
increases in student learning in mathematics and/or ELA?
5. In light of findings from the above questions, what are the indirect
relationships among principal leadership, peer influence, change in
instruction, and student learning?
Sample
The data to address these research questions come from a midsized urban
district in the southeastern United States. They were collected as part of an
ongoing study of educational leadership and principal professional development
discussed elsewhere in this journal edition. Cloverville (a pseudonym)
has 52 schools, 30 elementary schools, 10 middle schools, 8 high schools,
and 4 specialty schools. The district student population is approximately 66%
Black and 27% White, with about 58% of the students on free or reducedprice
lunch. This study utilized two data sources, teacher surveys and student
achievement data.
The teacher surveys provided measures of both leadership practice and peer
influence on teachers. An earlier study of the data from Cloverville had shown
that there were broad differences between principal reports and teacher reports
of principal leadership (Goldring, Huff, Pareja, & Spillane, 2008). Based on
this finding, in combination with the lineage of the literature that indicated that
principals tended to influence student performance indirectly through
influence on teachers, we sought to understand principal leadership from the
perspective of teachers. By doing so, we made the explicit decision to view
principal leadership from how teachers perceived it to be enacted rather than
from the perspective of what principals intended. Therefore, we focused on
teacher perceptions of both principal leadership and peer influence. The teacher
data came from a 2007 administration in which Cloverville teachers completed
a thorough survey regarding their background, the school as a workplace, professional
development, and school change, with an 81% response rate.

Student achievement data for the years 2006 and 2007 were collected and
linked using a district-provided unique student identifier. These records were
then linked to teachers using a teacher identifier for 2007. This study examines
both mathematics and ELA student achievement in Grades 1 to 8. High
school students were not included because testing did not allow us to measure
changes in performance from 2006 to 2007. To ensure that the teacher
link provided by the district was the subject teacher for both mathematics and
ELA, only teachers and students in self-contained classrooms were retained
for the final analysis. Of the 15,053 total number of Grades 2 through 8
students in 2007, 11,397 were used in the analysis. This represents a 24%
reduction in the sample because of an inability to make either a student link
or a teacher link or because the teacher was not in a self-contained classroom.
The final sample included 38 elementary and middle schools and 721 teachers.
Measures
Dependent variables. Two years of student records in Grades 1 through 8
were obtained from district databases, including end-of-year standardized test
scores and administrative data on each student’s race, sex, limited English
proficiency status, and free or reduced-price lunch participation. The 2005-
2006 school year was regarded as the pretest year and the 2006-2007 school
year as the posttest. Concurrent with this time period, the state was transitioning
to an updated version of the state test to maintain alignment with recently
updated state standards. This complicated our analyses in that the transition
to the new test was phased in for some grades earlier than in others. Furthermore,
the two versions of the test were not equated, resulting in very different
scales for the two sets of scores. Of our student sample, 13% were given the
old test in both years, 59% took the new test in both years, and 28% had one of
each. To place the two assessments on the same scale, test scores were standardized
(i.e., converted to z scores) by test version, subject, and grade across
all students in the analyzed sample.
As a result of the within-grade and -subject standardization, the rescaled
test scores reflect performance relative to the average student (for that grade
and subject) in standard deviation units. To better model student learning
during the 2006-2007 school year, we calculated a gain score for each student
by subtracting spring 2006 scores from spring 2007 scores. We chose to use
gain scores as opposed to a covariance approach given that research on Lord’s
paradox (see Holland & Rubin, 1983; Wainer, 1991) suggests that difference
scores produce less biased results than covariance analysis when the dependent
variable does not exhibit natural growth (which is the case with the z
scores used in our analyses).1
Independent variables. In May 2007, Cloverville teachers answered a school
staff questionnaire regarding their background, the school as a workplace,
professional development, and school change. The school staff questionnaire
was administered in the school setting and had an 81% response rate for the
38 elementary and middle schools in the study. For this study, seven areas of
school climate and teaching practice were measured: three relating to principal
leadership, three to peer teacher influence, and one to change in instructional
practice (see the conceptual framework represented in Figure 1).
Scale development was carried out using a combination of previous empirical
work (Camburn, Rowan, & Taylor, 2003) as well as pertinent theory of
each dimension of principal leadership and peer influence. In all, 29 indicator
survey items were used and had 6% missing data, which were imputed using
the Expectation Maximization (EM) algorithm prior to model estimation.
Imputed values were then rounded to the nearest scale value.
The seven scales developed for this study are shown in Table 1. Three scales
of teacher perceptions of principal leadership were derived from survey items:
Mission and Goals, Principal Trust, and Focus on Instruction. These scales were
highly reliable, with Cronbach alphas greater than .90. Three scales representing
peer influence were also developed from the survey: Instructional Conversation,
Interaction Around Teaching and Learning, and Instructional Advice Networks.
These scales were slightly less reliable, with Cronbach alphas around .80.
The scale measuring instructional advice networks was different than the
other scales, which were all developed from closed-ended Likert-type items.
By contrast, the instructional advice network scale was derived from a social
network question that asked teachers who they turned to for assistance in the
appropriate subject area (ELA and mathematics). From these responses, two
numbers were used to produce the Instructional Advice Network scale. The first
was simply the out-degree, or the number of requests for collaboration and assistance
that an individual makes of his or her peers, which we used to represent his
or her instructional resources. The second was the number of peers an individual
sought advice from outside of his or her grade level or content area, which we
took to represent instructional resources beyond his or her immediate network.
The final scale was Teacher-Reported Change in Instruction. This scale was
composed of teachers’ responses to four items that asked them about the
degree to which they had changed various aspects of their teaching. A full
description of the individual items constituting each of these seven scales,
along with their descriptive statistics, is presented in Appendix A.
Because the scales were not developed by exploratory methods alone, a
confirmatory analysis was performed to validate their use. The 29 items were
submitted to an oblique, multiple group, principal components cluster analysis
(Anderberg, 1973; Harman, 1976) to confirm the composition of the theorized
dimensions (see Appendix A). Initial group membership was provided, and
items were then permitted to migrate iteratively to dimensions that better explain
item variance. No item migrated from its hypothesized dimension. Item membership
in respective hypothesized groups was able to explain 73% of the total
variance, as opposed to only 27% when each item was assigned to its best
alternative dimension. All of the factors had a standardized Cronbach’s alpha
coefficient greater than or equal to .70, and all were deemed to be reliable for
this sample in terms of overall internal consistency (Cronbach, 1951) A correlation
matrix for the scales is shown in Appendix B.
Analytic Model
A multilevel structural model with latent variables was specified to investigate
principal and peer influences on change in teacher instruction as it relates to
student learning. All of the seven dimensions listed in Appendix A were specified
by their member items in the measurement model. Our structural model
consists of principal and peer influence as second-order factors that are predictive
of change in teachers’ instructional practice. A structural path was
also included from principal leadership leading to peer influence to allow
principals to have an indirect association with teacher instruction through the
influence of the community of teachers.
Table 1. Scales Used in Study and Reliabilities
Scales of Principal Leadership
Mission and Goals (5 items, a = .90)—Teacher perceptions of the extent to
which their principal has an instructional mission and related goals
Principal Trust (5 items, a = .93)—Teacher perceptions of a trusting relationship
with their principal
Focus on Instruction (5 items, a = .91)—Teacher perceptions of principal
expertise and focus on instruction
Scales of Peer Influence
Instructional Conversation (4 items, a = .81)—Teacher perceptions of
conversations with peers around instructional issues
Interaction Around Teaching and Learning (4 items, a = .76)—Teacher
perceptions of conversations with peers around instructional issues
Instructional Advice Networks (2 items, a = .85)—Social network items that
assessed degree of instructional assistance (density) and extent to which
teachers sought assistance outside of their grade or subject area (spread)
Scale of Change of Instruction
Teacher Change in Instruction (4 items, a = .94)—Teacher self-reported change
in instruction
A multilevel framework was used to allow for the clustering of students
within a teacher’s classroom. This was necessary to calculate an unbiased
estimate and proper confidence interval for the association of change in teacher
instruction with student learning. By specifying two levels, variation in student
learning is partitioned between and within classrooms, such that student covariates
are allowed to explain differences between student learning within a class
and teacher instruction explains the differences between the class averages.
The factor loadings for all 29 observed indicators are given in Appendix C.
Of the 29 factor indicators, 27 were assumed to be normally distributed,
whereas the 2 social network variables were treated as count data by assigning
a Poisson distribution. All disturbance terms among first-order factors as
well as error terms among factor indicators were regarded as uncorrelated.
Each first-order factor was identified by its most reliable indicator; secondorder
factors were each standardized. The model was estimated using Mplus
5.1 (Muthén & Muthén, 2007). The estimation method employed was maximum
likelihood with robust standard errors and, the EM optimization algorithm
was used in conjunction with numerical integration to obtain sample statistics
for model estimation. Convergence was met after 56 and 62 iterations for the
ELA and mathematics models, respectively.
In addition to students and teachers, the clustering of teachers within a school
should be modeled to improve the efficiency of estimates and to correct the confidence
intervals. Because this third level is not available in the analytic software
used, the model estimates may have standard errors that are too large. To conservatively
account for the possible clustering effect, the intraclass correlation and
average cluster size were used to adjust the standard errors by multiplying them
by the square root of the design effect (Higgins & Green, 2006).
Results
The results demonstrate a positive association for both principal and peer influence
with teachers’ change in instructional practice in both ELA and mathematics.
The structural path from principal leadership to peer influence was also shown
to be significant in both subjects. Finally, the direct relationship between teachers’
change in instructional practice and whole-class change in student learning
was demonstrated for ELA but not mathematics. Table 2 presents the structural
path coefficients estimated in the multilevel structural model.
These results allow us to directly address the first four of our research questions.
Our first research question was whether principal leadership was associated
with teacher change in instruction. We found that principal leadership was a
positive and significant predictor of a teachers’ change in instruction for both
ELA and mathematics. This suggests that principals who focus on instruction,
foster community and trust, and clearly communicate school mission and goals
are associated with teachers who report making a greater degree of changes
to their instructional practice.
Our second research question asked if principal leadership was associated
with teacher peer influence. We found that the largest and most significant relationship
in the structural model was the effect of principal leadership on peer
influence. School leadership, characterized in this model by the development of
mission and goals, an environment of collaboration and trust, and a focus on
instructional improvement, appears to foster an environment where teachers
work together and constructively engage with each other around issues of teaching
and learning. Our model also shows that peer influence was a positive and
significant predictor of teachers’ change in instruction for both subjects. Higher
levels of instructional conversation, interaction around teaching and learning,
and advice networks among peer teachers were associated with increases in the
amount of change in instruction that a teacher reports.
Our third research question asked about the relative magnitude of the association
of principal leadership and peer influence with teacher change in
instruction. In both ELA and mathematics, as can be seen by comparing the standardized
coefficients in Table 2, peer influence had a higher direct association
with change in instruction than did principal leadership. In mathematics, the
magnitude of the difference between the effects of peer influence relative to principal
leadership on teachers’ self-reported instructional practice was nearly twice
as large.
The fourth of our research questions asked if there was a relationship between
teacher change in instruction and increases in student learning in mathematics
and/or ELA. Our findings showed that, after controlling for a variety of
student background characteristics, there was a moderately sized and statistically
significantly positive association between teachers’ self-reported change
in instruction and student performance in ELA. There was no significant association
in mathematics.
Our final research question involved further examination of the indirect
relationships among principal leadership, peer influence, and student learning.
Educational leadership influences instructional practice, which changes student
performance. The results of the effect of leadership on instruction are
largely consistent in ELA and mathematics models, although change in instruction
was not a significant predictor of student math learning. Tracing the
relationship of student achievement back to principal leadership and peer influence
allows quantification of the indirect and total effects and corresponding
cluster-corrected standard errors, which are presented in Table 3 for ELA only.
The first four rows of statistics display the indirect effects of principal leadership
and peer influence on change in ELA instruction and student learning.
Principal leadership is significantly related to student learning through change
in instruction. Also displayed are total effects, which are the sum of all associations
between the predictor variable and outcome through direct and indirect
paths. Although the total effect of teacher peer influence on ELA student learning
is .02 (SE = .009), the total principal leadership association is .03 (SE = .00).
Discussion
This research provides both a confirmation of previous work as well as new
contributions to the research on the chain of hypothesized relationships between
leadership practice and student learning. First, consistent with the lineage of
studies of principal leadership on student learning (Hallinger & Heck, 1998;
Leithwood et al., 2004; Waters et al., 2003; Witziers et al., 2003), we found
empirical evidence that principal leadership influences student learning indirectly
through teachers’ instructional practices.
A major contribution of this research is the strong and significant impacts
of teacher peer influence on instructional practice. The second-order latent
factor of peer influence had a statistically significant impact on teachers’
instructional practice in both ELA and mathematics. This provides some of
the first empirical evidence of the influence of teachers’ leadership on the
classroom practices of teachers that produce student learning outcomes. This
supports some of the key contentions in the emerging teacher leadership literature
(Mangin & Stoelinga, 2008; Spillane, 2006).
There was also an important difference between content areas in our findings.
In ELA, the impact was on par with the impact of principal leadership
on instructional practice. In mathematics, the impact of peer influence on
teaching practice was almost twice the magnitude of that of principal leadership.
This suggests that in mathematics, with which principals might tend to
be less comfortable, teacher leaders fill the breach by providing important
support and assistance.
Principals have an indirect association with a teacher’s change in instruction
that is mediated by teacher peer influence. For both ELA and mathematics, the
total effect that principals have on change in instruction, which includes both the
direct and the indirect effects, increases dramatically by a standardized effect
of .08 when including their indirect effect through peer teacher networks.
The indirect effects of principals and peer teachers were also significantly
associated with ELA student learning. Although peer influence has a greater
direct effect on teacher instruction, principal leadership has a greater total
effect on ELA student learning because of the indirect effect through teacher
peer influence. This implies that principals are the most important actor in
student learning in ELA, in part because of their indirect influence on teacher
instruction through collaboration and communication around instruction
between peer teachers. Through fostering a climate of instructional collaboration,
principals have the greatest impact on learning.
The use of multilevel SEM provided several advantages in this investigation.
First, latent variables could be estimated in the same model in which
they were used as independent and dependent variables. Second, this framework
afforded us the ability to specify complex relationships among the
variables of interest, wherein locally dependent variables in parts of the
model were used as predictors of other variables. Teacher instruction and
student learning were specified as the proximal and distal outcomes, respectively,
of principle and peer teacher influences.
Several weaknesses of our study cause us to be conservative and temper our
findings. First, we used teacher self-report data for our key measures. Although
these are relatively easy to collect, they may misrepresent key constructs in our
data. In particular, having teachers self-report on changes in their instructional
practice is less than ideal. In addition, our findings may have differed if we had
used principal reports of their leadership practice rather than teacher reports of
principal leadership practice. One reason for this is that our measures of leadership
practice are likely to be influenced not only by what principals do but also
by teachers’ individual opinions of their principal. On the other hand, it is possible
that the teacher-report measures used here provide a more accurate picture of
principal leadership than principals’ own reports. Furthermore, we would argue
that teachers are affected by principals’ practices in different ways and to varying
degrees. As such, our measures of leadership practice may do a better job of
capturing the variation with which principals influence teachers.
These analyses also point to a series of future analyses that would extend
this work. First, should these analyses be replicated with data from high schools,
we believe the results would look very different. Second, we would like to
see a better measure of instructional practice used. Teacher self-report is potentially
misleading (Cohen, 1990), and more objective measures may produce
different results. Finally, we would also be curious to replicate these results
using principal perceptions of their leadership as opposed to teacher perceptions
of principal leadership and also to explore the factors that explain
variation in teachers’ perceptions of their principal’s practice.
The consuming obsession with accountability in the first decade of the 21st
century has led educators to seek connections between virtually any educational
endeavor and student learning outcomes, regardless of the length of the logic
chain between the two. In this context, school leadership has been scrutinized for
its detectable contributions to student learning. The findings of this study support
many others in the commonsense notion that the main impact of principals is not
directly on students but on teachers who interact with students directly on a daily
basis. Our findings suggest that principal influence is even broader and that principals
work through other leaders in schools to influence what goes on inside of
classrooms. This indirect pathway points to ways that principals’ attention to
such central school improvement concepts as mission and goals, community and
trust, and instructional focus have subtle yet real organizational influence. Principals,
working with and through the range of other school actors who exert
influence on teachers, do affect the instructional practice of teachers that produces
improvements in student learning.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the authorship
and/or publication of this article.
Financial Disclosure/Funding
The authors disclosed receipt of the following financial support for the research and/
or authorship of this article: This research was funded by the Institute of Education
Sciences, U.S. Department of Education (Grant R305E040085).
Note
1. Although there was a clear statistical argument for using difference score analysis
instead of covariance analysis, we also ran the model using a covariance approach
and found similar results, indicating that the transformation of scores did not
influence the findings.