scholarships for school leaders: impacts of the north ... · scholarships for school leaders:...
TRANSCRIPT
Scholarships for School Leaders: Impacts of the North Carolina
Principal Fellows Program
Kevin C. Bastian, Research Associate and Director of the Teacher Quality Research Initiative,
Education Policy Initiative at Carolina and University of North Carolina at Chapel Hill
Sarah C. Fuller, Research Assistant Professor
Education Policy Initiative at Carolina and University of North Carolina at Chapel Hill
Acknowledgements
We wish to thank the University of North Carolina General Administration for its on-going
financial support through the Teacher Quality Research Initiative and the Deans and department
heads from the colleges, schools, and departments of education at the 15 UNC system
institutions engaged in teacher and school leader education for their valuable feedback and
collaboration.
Manuscript presented at the 41st annual Association for Education Finance and Policy
conference in Denver, Colorado
Scholarships for School Leaders: Impacts of the North Carolina
Principal Fellows Program
Abstract
In the present study we assess the North Carolina Principal Fellows program, a statewide
scholarship loan initiative designed to attract academically-competitive individuals into
university-based school leadership preparation and to enhance the effectiveness of the school
leadership workforce. Descriptively, we find that Principal Fellows score significantly higher on
their principal licensure exams and are much more likely to hold school administrator positions.
While estimating principal effectiveness is methodologically challenging, evidence suggests that
Principal Fellows have positive impacts on working conditions, teacher retention, and student
absences. Impacts on student achievement are mixed and inconclusive. Continued research is
needed to identify effective recruitment and training practices so that states and districts can
strengthen their principal pipelines and staff schools with higher quality leaders.
Introduction
Nascent research shows that principals have significant effects on educational outcomes
of interest. Through both direct and indirect means principals impact the achievement gains,
attendance, and graduation rates of students (Branch, Hanushek, & Rivkin, 2012; Clark,
Martorell, & Rockoff, 2009; Coelli & Green, 2012; Grissom, Kalogrides, & Loeb, 2015); the
quality and retention of a school’s teachers (Boyd, Grossman, Ing, Lankford, Loeb, & Wyckoff,
2011; Grissom & Loeb, 2011; Grissom, Loeb, & Master, 2013; Horng, Klasik, & Loeb, 2010;
Ladd, 2011); and the working and learning environments of schools (Leithwood & Jantzi, 2005).
After teachers, principals are the most important school-based resource influencing student
outcomes (Branch, Hanushek, & Rivkin, 2013).
Given the importance of principals to students and schools, there are several approaches
that states and districts may take to improve principal quality. Akin to teacher policy, one
approach is for states and districts to develop principal evaluation rubrics which hold principals
accountable for school achievement and key domains of principal action and provide principals
with formative feedback to drive improvements in school leadership practice. For example,
districts such as Chicago and states such as North Carolina have codified principal evaluation
systems with standards focused on key elements of principal practice—e.g., instructional
leadership, school vision and culture—and student achievement growth (Chicago Public Schools,
2013; NCDPI, 2013).
Another approach seeks to alter the composition of the school leader workforce through
policies that recruit highly-competitive individuals into the principal pipeline and training
practices that prepare principal candidates to be effective school leaders. Here, most research
and policy attention has been directed towards high-profile alternative routes into the
principalship—e.g., New Leaders for New Schools or New York City’s Aspiring Principal
Program—and their impacts on academic outcomes (Clark, Martorell, & Rockoff, 2009;
Corcoran, Schwartz, & Weinstein, 2012; Martorell, Heaton, Gates, & Hamilton, 2010). While
such programs may serve as sources of innovation for traditional school leader preparation
programs, these alternative routes typically prepare only a small number of graduates for work in
high-need schools.
In the present study, we consider another innovation to alter the composition of the
school leader workforce—a merit-based state scholarship loan program to attract academically-
competitive individuals into university-based school leader preparation programs. Many states
have scholarship loan programs to recruit competitive candidates into their teacher workforce
(Hirsch, Koppich, & Knapp, 2001) and research shows that such programs produce graduates
who are more effective and likely to remain in teaching (Henry, Bastian, & Smith, 2012). To our
knowledge, North Carolina is the only state with a comparable merit-based scholarship loan
program for school leadership. Rigorously evaluating the extent to which the outlay of state
funds towards scholarship loans results in academically-competitive applicants and a stable
supply of effective and persistent school leaders is a key policy concern for North Carolina and
for other states exploring ways to improve the quality of their school leader workforce.
To consider the impact of school leader scholarship loans, we examine the North
Carolina Principal Fellows program. In existence since 1993, the Principal Fellows program
provides merit-based scholarship loans to academically-competitive individuals to attend a
participating university, earn a Master’s of School Administration (MSA) degree (one year of
full-time coursework, a one year school internship, and additional enrichment opportunities), and
serve as a school administrator in a North Carolina public school (NCPS). Since its inception,
over 1,200 Principal Fellows have completed the program (NC Principal Fellows, 2015). Given
that the program aims to attract an academically-competitive and stable supply of school leaders
and that principals influence a range of school and student outcomes, we assess Principal
Fellows across multiple measures—two constructs of school working conditions, teacher
retention, student achievement, student absences, and principal retention. Our research questions
are as follows:
(1) What are the characteristics of Principal Fellows and the schools that they lead?
(2) Do Principal Fellows impact school and teacher outcomes?
(3) Do Principal Fellows impact student academic outcomes?
(4) Do Principal Fellows persist in school principal positions?
Overall, we find that Principal Fellows have higher principal licensure exam scores and
are much more likely to enter school administrator positions—assistant principals and
principals—than other in-state public university MSA graduates. Additionally, nearly 90 percent
of Principal Fellows fulfill the school administrator service requirements specified by their
scholarship loan. Principal Fellows are an academically-competitive and reliable source of
school leaders. From an equity perspective, however, Principal Fellows lead schools with fewer
racial/ethnic minority and high-poverty students. Regarding Principal Fellow impacts on school,
teacher, and student outcomes, we acknowledge that estimating principal performance is
methodologically challenging since principal effects are often indirect, take time to develop, and
are difficult to separate from the effect of the school. Despite these challenges several results
emerge across model specifications: (1) positive effects of Principal Fellows versus an all other
principal category on two school working conditions constructs and teacher retention; (2)
positive effects of Principal Fellows versus other in-state public MSA graduates on student
absences; and (3) mixed and inconclusive impacts of Principal Fellows on student achievement.
In the sections that follow, we first provide more background on the Principal Fellows
program and the impact of principals and principal preparation on outcomes of interest. Next,
we describe our research sample and our empirical methods to estimate principal effects.
Finally, we address each of our research questions, providing additional details on sample,
measures, and methods, before concluding with implications for policy and practice.
The North Carolina Principal Fellows Program
In the early 1990s, North Carolina initiated two broad reforms to principal preparation at
in-state public universities. First, the state abolished all existing school leadership programs and
created the Masters of School Administration (MSA) degree. These MSA programs were to be
more selective in admissions, better resourced, staffed by highly-qualified faculty, acquainted
with the challenges of contemporary school environments, connected to K-12 schools through
internship programs, and aligned with national school leadership standards (Handa, Thompson,
Marcus, & Smith, 2010). Originally, in-state public institutions competed for the right to offer
MSA degrees; now, 13 of the 15 in-state public institutions have MSA programs.1
Second, as part of this restructuring of school leadership preparation, the North Carolina
General Assembly created the Principal Fellows program in 1993. The goal of the Principal
Fellows program is to prepare outstanding candidates for full-time school administrator
positions—assistant principals and principals—in the state’s public schools. Towards this end,
the Principal Fellows program provides merit-based scholarship loans to academically-
competitive individuals with teaching or other relevant experiences who want to enter school
1 There is a 16
th in-state public university, the North Carolina School of the Arts, that does not offer any teacher or
school leader preparation.
administration in North Carolina. To be awarded a scholarship loan, candidates must have a
strong academic record (at least a 3.2 GPA in the last 60 hours of study); be admitted into a
participating MSA program; demonstrate effective leadership, communication, and management
skills; and complete an interview with the Principal Fellows commission. Each March and April
the Principal Fellows commission meets and selects candidates for the award (NC Principal
Fellows, 2015).
Principal Fellows use the scholarship loans to attend one of 11 participating in-state
public universities and earn a MSA degree.2 The Principal Fellows program offers one year of
full-time academic coursework and a one year full-time school internship in a North Carolina
public school (NCPS). During their full-time coursework, Principal Fellows receive a $30,000
scholarship loan; during their school internship, Principal Fellows receive $4,100 for tuition and
fees, a scholarship loan equivalent to 60 percent of a first-year assistant principal’s salary, and an
internship stipend equivalent to 40 percent of a first-year assistant principal’s salary. Taken
together, this is approximately $75,000 in financial awards for each Principal Fellow (NC
Principal Fellows, 2015). After completion of the MSA, Principal Fellows promise to seek and
obtain employment as an assistant principal or principal in a NCPS for four years (within a six-
year period) or repay their scholarship loan at an interest rate of 10 percent.
Participating in-state public universities have a program coordinator who mentors
Principal Fellows and organizes a variety of additional enrichment activities designed to better
prepare Principal Fellows for school leadership positions. Examples of such enrichment
activities include visits to local schools, meetings with school district leaders, attending board of
2 The 11 participating in-state public universities are as follows: Appalachian State University, East Carolina
University, Fayetteville State University, North Carolina Agricultural and Technical State University, North
Carolina Central University, North Carolina State University, UNC Chapel Hill, UNC Charlotte, UNC Greensboro,
UNC Wilmington, and Western Carolina University.
education and district-level professional meetings, and pre-internship rotations with school
principals. Each year, Principal Fellows attend five day-long enrichment seminars—10 seminars
over their two-year MSA period—designed to provide leadership experiences not available
through their university program and allow them to form professional and collegial relationships
with their peers. Seminar topics include: instructional leadership, teacher evaluation, working
with students in poverty, managing financial resources, and working effectively with students,
staff, parents, and the district. During their year-long school internship, Principal Fellows work
under the direction and supervision of the sitting principal and complete a variety of
administrative responsibilities.
The Impact of School Leadership and School Leader Preparation
Research examining the impacts of principals or principal characteristics on student and
school outcomes is conceptually and methodologically challenging for three primary reasons.
First, principals do not directly impact many outcomes of interest. For instance, the effects of
principals on student learning are indirect and happen through principals’ ability to hire and
retain teachers, establish a positive school climate, and oversee instruction (Branch, Hanushek, &
Rivkin, 2012; Grissom, Kalogrides, & Loeb, 2015). Second, principals’ effects may not be
immediate, but rather, may develop over time as principals shape their teaching staff, promote a
shared vision and culture, and build relationships (Coelli & Green, 2012; Grissom, Kalogrides, &
Loeb, 2015; Miller, 2013). Finally, principals are not randomly assigned to schools. Methods to
separate the effects of principals from the confounding effects of schools may not produce
generalizable results, since such methods (principal and school fixed effects) require principal
transitions and often limit comparisons to a small network of principals working in the same
schools (Chiang, Lipscomb, & Gill, 2012; Grissom, Kalogrides, & Loeb, 2015).
Despite these challenges, nascent research indicates that principals significantly influence
a range of student and school outcomes. Regarding student-level outcomes, Branch and
colleagues show that a one standard deviation increase in effectiveness for principals switching
schools is associated with a 0.10 standard deviation increase in student achievement (Branch,
Hanushek, & Rivkin, 2012). Using principal and school fixed effects, Grissom and colleagues
return estimates of 0.060 and 0.034 standard deviations in mathematics and reading, respectively
(Grissom, Kalogrides, & Loeb, 2015). Additionally, principals one standard deviation higher in
the effectiveness distribution increase high school graduation rates by 2.6 percentage points
(Coelli & Green, 2012), while those in the top quartile of effectiveness are associated with
significantly higher student attendance rates (Branch, Hanushek, & Rivkin, 2012). Beyond
student outcomes, evidence indicates that principals’ organizational management—hiring,
retaining, and assigning teachers—and elements of their instructional leadership—coaching and
evaluating teachers—are critical components of their success (Grissom & Loeb, 2011; Grissom,
Loeb, & Master, 2013; Horng, Klasik, & Loeb, 2010). Finally, principals significantly influence
the working conditions of their schools and the retention decisions of their teachers (Boyd,
Grossman, Ing, Lankford, Loeb, & Wyckoff, 2011; Ladd, 2011).
To date, only a few studies have investigated the relationships between principal
preparation and principal effectiveness. Work by Vanderhaar and colleagues shows that
principals prepared in university training programs perform comparably to those prepared in
school district programs (Vanderhaar, Munoz, & Rodosky, 2006). Two studies of the New York
City Aspiring Principals Program find that schools led by program graduates are initially lower
performing but narrow gaps over time (Clark, Martorell, & Rockoff, 2009; Corcoran, Schwartz,
& Weinstein, 2012). First-year principals trained by New Leaders for New Schools are
associated with slightly reduced student achievement; however, more experienced New Leaders
are associated with positive student achievement outcomes (Martorell, Heaton, Gates, &
Hamilton, 2010). Finally, work in North Carolina indicates that graduates of the newly
constituted MSA programs at in-state public universities performed similarly to principals with
other forms of training (Handa, Thompson, Marcus, & Smith, 2010). We add to this body of
research by examining whether a statewide scholarship loan program to recruit and train
academically-competitive applicants produces higher performing school leaders.
Research Sample
To address our descriptive and empirical research questions, we use data provided by the
North Carolina Department of Public Instruction (NCDPI) and the University of North Carolina
General Administration (UNCGA) to link students and teachers to schools, principals to their
type of preparation, and principals to the schools that they lead. The UNCGA provided a list of
graduates of in-state public university MSA programs, with separate identifiers for the subset of
these graduates who were Principal Fellows. For our analyses the study sample consists of the
most comprehensive set of available data—full-time principals in the 2005-06 through 2011-12
academic years. To define a full-time principal we specified the following rules: (1) an
individual had to begin work as a principal at a school in one of the fiscal year’s first three pay
periods (July, August, or September) and (2) an individual had to remain as a full-time principal
at that school for at least eight pay periods (months). We compare the outcomes for Principal
Fellows with those of other MSA graduates (non-Principal Fellows) prepared at in-state public
universities and with all other types of principal preparation. This all other category includes
those earning a MSA at an in-state private university, those earning principal preparation degrees
at in-state institutions prior to the establishment of MSA programs, and those earning principal
preparation degrees out-of-state. Together, these groups allow us to assess how Principal
Fellows fair versus other principals with similar training in in-state public university MSA
programs and versus those with more diverse forms of preparation.
Empirical Methods
In this study we use a modeling approach comparable to that of Grissom and colleagues
(2015) to estimate principal effects on two constructs of school working conditions—school
vision and culture and instructional leadership—teacher retention, student achievement, and
student absences. Specifically, we estimate the impact of principals with two types of models.
Model one is a principal-by-school fixed effects approach that generates an estimate of
school effectiveness for each principal-school combination in the sample. A benefit of this
approach is that effectiveness comparisons are made across all of the principals in the analysis.
Furthermore, we use the Stata felsdvregdm fixed effects procedure, which means that there is not
a left-out principal-by-school reference group and effectiveness estimates are centered across all
of the unique principal-school combinations in the analysis (Mihaly, McCaffrey, Lockwood, &
Sass, 2010). The key concern with this approach, however, is that the model attributes the entire
school effectiveness estimate to the principal, which may incorrectly debit or credit the principal
with results that are outside of her control (Grissom, Kalogrides, & Loeb, 2015). For instance,
schools may vary in how well prepared their students are when they enter the school. The basic
estimation equation for this principal-by-school fixed effects model is as follows. Here, outcome
Y for each principal-school combination is a function of model covariates (which we discuss
more fully below), X, and year fixed effects, 𝜏𝑡, and the measure of principal effectiveness is the
principal-by-school fixed effect 𝛿𝑝𝑠.
𝑌 = 𝛽𝑋 + 𝜏𝑡 + 𝛿𝑝𝑠 + 휀𝑝𝑠 (1)
Model two is a principal and school fixed effects approach that generates a within school
estimate of principal effectiveness. This approach requires principal transitions—that two or
more principals lead the school during the study period—and then compares the effectiveness of
the school during the tenure of principal 1 to the effectiveness of the school during the tenure of
principal 2. The main benefit of this approach is that it attempts to separate the effect of the
principal from the effect of the school. The key concern with this model, however, is the
generalizability of the results, since schools without principal transitions do not contribute to
effectiveness estimates and comparisons are limited to a small number of principals. Given the
sorting of principals into schools (Loeb, Kalogrides, & Horng, 2010), these limited comparisons
may unfairly measure principal performance since a principal’s effectiveness estimate is relative
to the effectiveness of other principals who led the same school (Grissom, Kalogrides, & Loeb,
2015). Additionally, this approach can attribute effects to the wrong principal, since it may take
time for principal effects to develop (Coelli & Green, 2012). As with model one, we use the
Stata felsdvregdm procedure to estimate these principal and school fixed effects models.
Estimates from these models are centered within the network of principals who have led the
same school during the study period.3 The basic estimation equation for this principal and school
fixed effects model is as follows. Here, outcome Y for each principal is a function of model
covariates, X, year fixed effects, 𝜏𝑡, and a school fixed effect, 𝜑𝑠, and the measure of principal
effectiveness is a principal fixed effect, 𝛿𝑝.
𝑌 = 𝛽𝑋 + 𝜏𝑡 + 𝜑𝑠 + 𝛿𝑝 + 휀𝑝 (2)
3 One limitation of this principal and school fixed effects model with the felsdvregdm procedure is that the model
will not run when principals are observed in more than one school in the analysis data. Therefore, for principals
who led more than one school during our study period, we followed Grissom, Kalogrides, and Loeb (2015) by (1)
identifying the school in which the principal had served the longest and (2) keeping observations from only that
school for principal and school fixed effects analyses.
For each of our outcomes of interest model 1generates an estimate of effectiveness for
each principal-school combination (the principal-by-school fixed effect, 𝛿𝑝𝑠) and model 2
generates an estimate of effectiveness for principals within a network of schools (the principal
fixed effect, 𝛿𝑝). Post-estimation, we take these effectiveness estimates and use t-tests to
determine whether there are significant differences in (1) the mean effect of Principal Fellows
versus the mean effect of other in-state public MSA graduates and (2) the mean effect of
Principal Fellows versus the mean effect of all other principals. We prefer this t-test approach,
rather than second-stage regressions with Principal Fellows as the reference category, because
readers can see whether the mean effect for our principal preparation categories is above or
below zero.
Given that principals have a direct impact on school working conditions, we begin our
empirical analyses by assessing the relationships between the principal preparation categories
and two working conditions constructs—vision and culture and instructional leadership. Next,
we consider another outcome upon which principals have a more direct influence—teacher
retention. Finally, we examine two student outcomes upon which principals have indirect
effects—student achievement and student absences. By ordering outcomes this way we present
the most proximate principal effects first and then progress to outcomes that are more
conceptually challenging to estimate and interpret since principals impact them through indirect
mechanisms.
Across all outcome measures, we control for year fixed effects and a select set of school
characteristics—school size, total per-pupil expenditures, average teacher salary supplements,
percentage of racial and ethnic minority students, and percentage of students qualifying for
subsidized school meals. Since Principal Fellows average fewer years of principal experience
and tenure than all other principals and research shows that these characteristics can impact
principal performance (Clark, Martorell, & Rockoff, 2009; Coelli & Green, 2012), we also
control for a linear and quadratic form of principal experience and tenure at a school in all of our
outcome models.4 In the results sections below, we further detail covariates used in specific
analyses.
What are the characteristics of Principal Fellows and the schools that they lead?
To assess the individual characteristics of Principal Fellows, we use MSA completer data
provided by the UNCGA and certified salary files provided by the NCDPI to compare the values
for Principal Fellows graduates from 1996 through 2012 to the other in-state public university
MSA graduates over the same time period. Table 1 shows that there have been approximately
1,200 Principal Fellows graduates and 4,500 other in-state public MSA graduates over this 17
year period. Overall, a significantly higher percentage of Principal Fellows are female, while a
significantly lower percentage are racial or ethnic minorities. Principal Fellows score
significantly higher on their principal licensure exams and are on a slightly faster track than other
in-state public MSA graduates. On average, Principal Fellows teach for fewer years in NCPS,
complete their MSA at an earlier age, and become principals at an earlier age. Finally, a
significantly higher percentage of Principal Fellows assume school leadership positions in the
state’s public schools—96 percent of Principal Fellows, versus 61 percent of other in-state public
MSA graduates, have been or currently are an assistant principal; likewise, 46 percent of
Principal Fellows, versus 27 percent of other in-state public MSA graduates, have been or
currently are a principal. These characteristics indicate that Principal Fellows are an
academically-competitive and reliable source of school leaders.
4 Average principal experience is 3.62 years for Principal Fellows, 2.82 years for other in-state public MSA
graduates, and 7.54 years for all other principals. Average principal tenure is 2.24 years for Principal Fellows, 1.82
years for other in-state public MSA graduates, and 3.64 years for all other principals.
[Insert Table 1 about here]
To complement these individual characteristics, we use school-level descriptive and
expenditure data provided by the NCDPI to display school characteristics for first-time principals
and for all principals during our study period. This allows us to assess the types of schools that
Principal Fellows are initially hired to lead and the types of schools they lead overall. Given that
principals can impact several of the characteristics in Table 2, the values for first-time principals
are from the year before a principal assumes leadership—the school characteristics that first-time
principals inherit.
Overall, the left panel of Table 2 shows that Principal Fellows who are first-time
principals inherit schools with (1) fewer minority students than other in-state public MSA
graduates; (2) more students passing their End-of-Grade (EOG) and End-of-Course (EOC)
exams (performance composite); (3) lower per-pupil expenditures than all other principals; and
(4) a mix of teacher credentials—more early-career and NBC teachers and fewer advanced
degree teachers. Principal Fellows who are first-time principals are also much more likely to
lead an elementary school and less likely to lead a high school. Regarding all principals, the
right panel of Table 2 displays characteristics comparable to those for first-time principals.
Principal Fellows lead schools with (1) fewer subsidized school lunch and minority students; (2)
more students passing their EOG and EOC exams; (3) lower per-pupil expenditures;5 and (4)
more NBC teachers. Again, Principal Fellows are more likely to lead elementary schools and
less likely to lead high schools. Given this concentration of Principal Fellows in elementary
schools and the potential for our outcomes of interest to differ by school level, we estimate
separate models at the elementary, middle, and high school levels.
5 These spending differences between groups may be attributable to the fact that (1) in North Carolina, high-poverty
and low-performing schools spend more, per-pupil and (2) Principal Fellows are less likely to work in high schools,
where per-pupil expenditures are the highest.
[Insert Table 2 about here]
Do Principal Fellows impact school and teacher outcomes?
Vision and Culture and Instructional Leadership
Background: Research shows that principals may impact student outcomes and teacher
retention through the school vision and culture that they help establish and through their actions
as instructional leaders (Boyd, Grossman, Ing, Lankford, Loeb, & Wyckoff, 2011; Grissom,
Loeb, & Master, 2013; Robinson, Lloyd, & Rowe, 2008). Therefore, to assess the extent to
which Principal Fellows fulfill these key roles, we used teachers’ responses to the North Carolina
Teacher Working Conditions (TWC) survey to create constructs for school vision and culture
and instructional leadership. North Carolina administers the TWC survey biennially—in the
2005-06, 2007-08, 2009-10, and 2011-12 school years during our study period—and items focus
on time management, facilities and resources, community support and involvement, student
conduct, teacher leadership, school leadership, professional development, and instructional
practices and supports.6
To create our outcome constructs, we examined each of the four TWC surveys during our
study period and identified sets of items that were included in each TWC administration and that
conceptually mapped onto the constructs of school vision and culture or instructional leadership.7
Within schools and years we averaged teachers’ responses to each of these items, creating a
single school-by-survey value for each survey item of interest, and then estimated Cronbach’s
alpha values to assess the internal consistency of these items as measures of vision and culture
and instructional leadership. Table 3 shows the TWC items we included in each outcome
6 Teacher response rates to the North Carolina TWC survey were approximately 69, 89, 93, and 91 percent,
respectively, in the four survey iterations included in our analyses. 7 To view the items from the most recent survey included in our analyses (2011-12 school-year) please see the
following: http://2012.ncteachingconditions.org/sites/default/files/attachments/NC12_survey_main.pdf
construct. Overall, the Cronbach’s alpha values for each construct were above 0.90—indicative
of high levels of internal consistency. Additional confirmatory factor analysis indicated that
each survey item significantly loaded onto the latent constructs of vision and culture or
instructional leadership.
[Insert Table 3 about here]
For each school-survey combination we summed the items for each construct and then
standardized each of these constructs within school level (elementary, middle, and high) and
survey-year. With the school-level constructs for vision and culture or instructional leadership as
outcome variables, we estimated separate models at each school level controlling for year fixed
effects, average teacher experience at the school, school covariates, and a linear and quadratic
form of principal experience and tenure. Again, we estimated these models with a principal-by
school fixed effect and a principal and school fixed effect.
Results: Focusing first on school vision and culture, the top left panel of Table 4 shows
that across school levels, the measure of vision and culture is significantly higher in schools led
by Principal Fellows versus schools led by all other principals. These differences are
approximately 20, 32, and 65 percent of a standard deviation in elementary, middle, and high
school, respectively. Comparing within networks of principals who have led the same schools,
the bottom left panel of Table 4 indicates that our measure of vision and culture remains
significantly higher in elementary and middle schools led by Principal Fellows versus all other
principals. Conversely, measures of vision and culture are significantly higher in middle and
high schools led by other in-state public MSA graduates.
[Insert Table 4 about here]
Regarding our instructional leadership construct, the top right panel of Table 4 indicates
that across elementary and middle schools, the measure of instructional leadership is
significantly higher in schools led by Principal Fellows versus schools led by all other principals.
These differences are approximately 20 and 40 percent of a standard deviation, respectively.
Comparing within networks of principals who have led the same schools, these significant
differences in instructional leadership persist between Principal Fellows and all other principals.
Teacher Retention
Background: Research shows that principal quality significantly impacts teachers’
retention decisions (Boyd, Grossman, Ing, Lankford, Loeb, & Wyckoff, 2011) and that teacher
attrition can have harmful effects on school performance (Ronfeldt, Loeb, & Wyckoff, 2013).
Given this, we use data from the 2005-06 through 2011-12 school years to assess the extent to
which Principal Fellows are able to retain teachers in the schools that they lead.
In these analyses the outcome variable is a dichotomous indicator for an individual
teacher’s retention decision, where ‘1’ means returning to teach in the same school in the
following year and ‘0’ means leaving teaching at the school. We estimate separate linear
probability models for elementary schools, middle schools, and high schools. In all of these
teacher retention models, we control for year fixed effects, a set of teacher characteristics—
gender, minority status, age, experience, graduate degree status and NBC status—school
characteristics, and a linear and quadratic form of principal experience and principal tenure at the
school. We estimate the impact of principals on teacher retention with a principal-by-school
fixed effect and a principal and school fixed effect.
Results: For the principal-by-school estimates, the top panel of Table 5 indicates that
across school levels, teacher retention is higher in schools led by Principal Fellows versus
schools led by all other principals. Adjusting for covariates, these differences in teacher
retention are approximately three percentage points in elementary and middle schools and eight
percentage points in high schools. Conversely, teacher retention is lower in elementary schools
led by Principal Fellows versus other in-state public MSA graduates. When using principal and
school fixed effects to generate within school estimates, many of the teacher retention results are
no longer statistically significant. Comparing within networks of principals who have led the
same elementary schools, however, individual teacher retention is significantly higher in schools
led by Principal Fellows versus schools led by all other principals. Given the concentration of
Principal Fellows in elementary schools, this result has additional practical significance.
[Insert Table 5 about here]
Do Principal Fellows impact student academic outcomes?
Student Achievement
Background: We use student test score and demographic data from the 2005-06 through
2011-12 academic years to assess the impact of Principal Fellows on adjusted-average student
achievement. For these analyses we focus on exams in five grade-level/subject-area
combinations: End-of-Grade (EOG) exams in elementary grades mathematics, elementary
grades reading, middle grades mathematics, and middle grades reading and End-of-Course
(EOC) exams in the five high school courses that were taken by nearly all NCPS students during
the study period—algebra I, English I, biology, US history, and civics and economics. The
outcome variable for these analyses is a student’s standardized EOG (within subject, grade, and
year) or EOC (within subject and year) exam score.
In addition to our common set of control variables, these models control for a rich set of
student characteristics: prior scores in both mathematics and reading,8 peer ability, days absent,
mobility (within-year, between-year and structural), underage for grade, overage for grade,
giftedness, disability, subsidized school meals, gender, race/ethnicity, and limited English
proficiency.9 In these analyses we exclude classroom and teacher characteristics since these are
a primary way in which principals may indirectly impact student achievement. We estimate
these student achievement models with either a principal-by-school fixed effect or a principal
and school fixed effect.
Results: For the principal-by-school effectiveness estimates, the top panel of Table 6
shows that adjusted-average student achievement is significantly higher in schools led by
Principal Fellows than in schools led by other in-state public MSA graduates in four
comparisons—elementary mathematics, elementary reading, middle grades mathematics, and
middle grades reading. These differences are all approximately three percent of a standard
deviation in student achievement. Principal-by-school estimates are also significantly higher for
Principal Fellows versus all other principals for high school EOC exams; conversely, all other
principals have significantly higher principal-by-school effectiveness estimates in elementary
grades reading.
[Insert Table 6 about here]
When using principal and school fixed effects to generate within-school estimates, results
differ substantially. The bottom panel of Table 6 indicates that adjusted-average student
8 For the high school End-of-Course exams we control for 8
th grade mathematics and reading test scores. We also
include subject-area indicator variables in reference to algebra I. 9 Unlike teachers, who do not influence a student’s prior test score, principals can influence a student’s current and
prior year test scores. This presents an additional challenge to estimating principal effectiveness. Following
Grissom and colleagues (2015), in these analyses we include students’ prior test scores in value-added models. We
acknowledge this as a potential limitation, but note that coefficients still capture whether achievement is greater than
expected given model covariates.
achievement is significantly lower for Principal Fellows versus all other principals in elementary
mathematics, elementary reading, and middle grades mathematics and significantly lower than
other in-state public MSA graduates in high school EOC exams. Since these fixed effects
results—principal-by school and principal and school—are based on both real differences in
student achievement and measurement error, we shrank the fixed effects estimates using the
empirical Bayes method (Grissom, Kalogrides, & Loeb, 2015; McCaffrey, Lockwood, Koretz,
Louis, & Hamilton, 2004). As shown in Appendix Table 1, the shrunken fixed effects estimates
and the relationships between principal categories are comparable with and without adjustment
for measurement error.
The differences between the principal-by-school effectiveness estimates and the within-
school effectiveness estimates may be attributable to better isolating the effect of principals from
the effect of schools and/or changes in the estimation sample. Given that descriptive
characteristics show Principal Fellows leading schools that are higher-performing and that have
fewer minority and high-poverty students, the principal-by-school estimates may be a result of
these differences in school environments. However, changes in estimation sample may also be
an issue, since the principal and school fixed effects (1) exclude schools without principal
transitions during the study period and (2) make comparisons within limited sets of principals
who have led the same schools during the study period.
To investigate how changes in estimation sample may have influenced the student
achievement results, we took two additional steps. First, for each of our value-added models, we
identified the average number of principals contributing to effectiveness estimates when using a
principal and school fixed effect. In elementary, middle, and high school the averages were 2.32,
2.30, and 2.29, respectively. Essentially, the modal category was comparing the effectiveness of
two principals. Second, we re-ran our principal-by-school effectiveness models focused only on
the sample of principals and schools in our within-school effectiveness approach. As shown in
Appendix Table 2, adjusted-average student achievement in elementary grades reading remains
higher in schools led by all other principals versus schools led by Principal Fellows—this result
is consistent across modeling approaches. Other results differ between Table 6 and Appendix
Table 2, suggesting that the estimation sample and methods meaningfully influence principal
effectiveness estimates.
Student Absences
Background: Research shows that more experienced principals and principals in the top
quartile of value-added effectiveness are associated with significantly higher student attendance
rates (Branch, Hanushek, & Rivkin, 2012; Clark, Martorell, & Rockoff, 2009). Given this, we
use data from the 2005-06 through 2011-12 school years to assess principals’ impact on number
of days absent for students in elementary, middle, and high schools. In all student absence
models, we control for year fixed effects, school covariates, a linear and quadratic form of
principal experience and tenure, and a set of student demographic characteristics—race/ethnicity,
gender, subsidized school meals, limited English proficiency, giftedness, disability, underage for
grade, and overage for grade. As with our other outcome measures we used two modeling
approaches—a principal-by-school fixed effect and a principal and school fixed.
[Insert Table 7 about here]
Results: The top panel of Table 7 reports student absence estimates from the principal-
by-school fixed effects models. Here, results indicate that the level of student absences is lower
in high schools served by Principal Fellows compared to high schools led by other in-state public
MSA graduates and all other principals. In middle schools, the number of student absences at
schools led by Principal Fellows is lower than in schools led by other in-state public MSA
graduates but higher than in schools led by all other principals. Comparing within networks of
principals who have led the same schools, the bottom panel of Table 7 reports student absence
results in middle schools that are comparable to the principal-by-school fixed effects models.
Results at the elementary school level show that student absences are significantly lower in
schools led by Principal Fellows versus schools led by all other principals.
Do Principal Fellows Persist in School Principal Positions?
Beyond attracting competitive candidates and producing effective school leaders, it is
important to know whether the outlay of state funds to the Principal Fellows program reduces
principal attrition. This is particularly important since inexperienced principals (less than five
years as a school principal) comprise more than 50 percent of the principal workforce in North
Carolina (Bastian & Henry, 2015) and research shows that transitions to first-time principals
have adverse effects on student achievement, student attendance, and teacher persistence
(Beteille, Kalogrides, & Loeb, 2012; Burkhauser, Gates, Hamilton, & Ikemoto, 2012; Dhuey &
Smith, 2013; Miller, 2013). Therefore, we examine school leader persistence by (1) reporting
the percentage of Principal Fellows who fulfill their program service requirements and (2)
estimating models predicting that principals return in the following school year.
As a condition of their scholarship loan, in the six year period post MSA graduation,
Principal Fellows must work as a full-time school administrator—assistant principal or
principal—in a NCPS for at least four years or repay their scholarship loan with interest. To
determine the percentage of Principal Fellows fulfilling this requirement, we used salary data
provided by the NCDPI to track the employment status of the 1996 through 2006 Principal
Fellows graduating cohorts over the six year period post MSA graduation. In total, 87 percent of
the 867 Principal Fellows in these graduating cohorts fulfilled their service requirement. When
we expanded our analysis to include the 2007 and 2008 graduating cohorts and limited the post-
graduation period to five or four years, respectively, 86 and 85 percent of eligible Principal
Fellows fulfilled their service requirement.
To assess whether Principal Fellows, relative to other in-state public MSA graduates and
all other principals, return to a school principal position in the following school year, we used
data from the 2005-06 through 2011-12 academic years to estimate linear probability models for
principal persistence. Since principal attrition from a school can be influenced by factors beyond
a principal’s control—e.g., districts re-assigning an effective principal to a low-performing
school—we specified models predicting principal persistence in the same school or in any
NCPS. We estimated these models in elementary, middle, and high schools and controlled for
year fixed effects, a linear and quadratic form of principal experience and tenure, school
characteristics, and basic principal demographics—gender, race/ethnicity, and age.
[Insert Table 8 about here]
For principal persistence in the same school, the top panel of Table 8 indicates that in
high schools other in-state public MSA graduates are significantly more likely to return as
principals in the following school year. This result is consistent with other positive results—
school vision and culture and student achievement—for other in-state public MSA graduates at
the high school level. For principal persistence in any NCPS, the bottom panel of Table 8
reports significantly higher probabilities of retention for other in-state public MSA graduates in
high schools; conversely, Principal Fellows are more likely to return to a principal position than
all other middle school principals. This result for Principal Fellows is consistent with other
positive results—school vision and culture, instructional leadership, teacher retention—for
Principal Fellows versus all other principals in middle schools.
Discussion
In this study we assessed the impacts of the Principal Fellows program, a statewide,
merit-based scholarship loan designed to attract academically-competitive individuals into
university-based school leader preparation and school administrator positions. Given program
goals and the range of school and student outcomes that principals influence, we examined the
characteristics of program participants and the schools that they lead; the impacts of Principal
Fellows on school, teacher, and student outcomes; and principal retention. We compared
Principal Fellows to two groups—other principals who receive MSAs from in-state public
universities and all other principals. These groups complement each other. Other in-state public
university MSAs received similar training to Principal Fellows; all other principals are not as
similar in training or career progression, but make up the bulk of the principal workforce in
North Carolina and are more likely to have served the same schools before or after a Principal
Fellow.
The stated goal of the Principal Fellows program is “to prepare the most outstanding
candidates for full-time school administration in North Carolina” (NC Principal Fellows, 2015).
Compared to other in-state public university MSA graduates, we find that Principal Fellows have
stronger academic credentials, as measured by licensure test scores, and are significantly more
likely to work as an assistant principal or principal in NCPS after completing their degree.
Furthermore, nearly 90 percent of Principal Fellows fulfill the school administrator service
requirements specified by the scholarship loan. In policy terms, this evidence suggests that a
scholarship loan program for prospective school leaders can serve as a mechanism to provide a
supply of highly qualified principals and assistant principals.
This study faces several limitations in attempting to estimate impacts of the Principal
Fellows on school, teacher, and student outcomes. First, the effect of principals on many
outcomes of interest, especially those at the student level, is indirect. Second, principal effects
may take time to develop. This can make it difficult to identify the influence of a principal,
particularly in models comparing principals who have led the same school, since the influence of
the first principal is likely to continue to be observed into the tenure of the next principal.
Finally, principals are not randomly assigned to schools, and as such, it is difficult to separate the
influence of school characteristics from the effect of the principal. This limitation is particularly
salient in our analyses because there are some significant differences between the schools led by
Principals Fellows and those led by principals with other forms of preparation.
Given that principals have a more direct impact on school working conditions and teacher
retention, we began our analyses by assessing relationships between principal preparation and
two working conditions constructs and teacher retention. Here, Principal Fellows appear to have
an advantage relative to all other principals. Principal Fellows lead elementary and middle
schools rated higher on school vision and culture and instructional leadership; they also lead
schools with higher levels of teacher retention, particularly in elementary schools where
Principal Fellows are concentrated. This suggests that Principal Fellows compare favorably to
the largest group of principals currently working in NCPS. In comparison to other in-state public
university MSA graduates, Principal Fellows lead schools rated lower on a school vision and
culture construct and have similar levels of teacher retention.
Results for student outcomes—which principals impact indirectly—are less conclusive.
Regarding student test scores, estimates show that adjusted-average student achievement in
elementary and middle schools led by Principal Fellows is generally higher than in schools led
by other in-state public university MSA graduates but lower than in schools led by all other
principals. For high schools this trend is reversed—adjusted-average student achievement in
schools led by Principal Fellows is higher than in schools led by all other principals but lower
than in schools led by other in-state public university MSAs. Importantly, these differences
between groups are not consistently statistically significant across the principal-by-school and
principal and school fixed effects models. For absences, students in middle schools led by
Principal Fellows have fewer absences than their peers attending middle schools led by other in-
state public university MSA graduates but more absences than peers attending middle schools
led by all other principals. The remaining significant results suggest fewer absences in schools
led by Principal Fellows but are not consistently significant across models. Overall, these
student-level results are mixed—perhaps reflecting the challenge of estimating principal effects
on student outcomes—and suggest that any differences in student outcomes between principal
preparation groups may be relatively small.
Moving forward, states and school districts need more evidence about the effectiveness
of the principal recruitment and preparation programs that supply their school leader workforce.
More broadly, further research is needed about specific principal recruitment and training
practices—such as competitive scholarship loans—that are effective. With these data states and
school districts can strengthen their principal pipelines and staff schools with higher quality
school leaders.
References
Bastian, K.C. & Henry, G.T. (2015). The apprentice: Pathways to the principalship and student
achievement. Educational Administration Quarterly, 51(4), 600-639.
Beteille, T., Kalogrides, D., & Loeb, S. (2012). Stepping stones: Principal career paths and
school outcomes. Social Science Research, 41, 904-919.
Boyd, D., Grossman, P., Ing, M., Lankford, H., & Loeb, S. (2011). The influence of school
administrators on teacher retention decisions. American Educational Research Journal,
48(2), 303-333.
Branch, G., Hanushek, E., & Rivkin, S. (2012). Estimating the effect of leaders on public sector
productivity: The case of school principals. NBER Working Paper 17803.
Branch, G., Hanushek, E., & Rivkin, S. (2013). School leaders matter. Education Next, 13(1),
1-8.
Burkhauser, S., Gates, S., Hamilton, L., & Ikemoto, G. (2012). First-year principals in urban
school districts: How actions and working conditions relate to outcomes. Santa Monica,
CA: RAND Education.
Chiang, H., Lipscomb, S., & Gill, B. (2012). Is school value-added indicative of principal
quality. Mathematica Policy Research. Available from: http://mathematica-
mpr.com/publications/pdfs/education/value-added_principal_quality.pdf
Chicago Public Schools. (2013). Principal quality and effectiveness: Board Briefing. Available
from: http://www.cpsboe.org/content/documents/principal_evaluations_012313.pdf
Clark, D., Martorell, P., & Rockoff, J. (2009). School principals and school performance. Calder
Institute Working Paper 38.
Coelli, M., & Green, D. (2012). Leadership effects: School principals and student outcomes.
Economics of Education Review, 31(1), 92-109.
Corcoran, S., Schwartz, A.E., Weinstein, M. (2012). Training your own: The impact of New
York City’s Aspiring Principals Program on student achievement. Educational Evaluation
and Policy Analysis, 34(2), 232-253.
Dhuey, E., & Smith, J. (2013). How school principals influence student learning. Retrieved from
http://www4.ncsu.edu/~rghammon/workshop/S13_Dhuey.pdf
Grissom, J., Kalogrides, D., & Loeb, S. (2015). Using student test scores to measure principal
performance. Educational Evaluation and Policy Analysis, 37(1), 3-28.
Grissom, J., & Loeb, S. (2011). Triangulating principal effectiveness: How perspectives of
parents, teachers, and assistant principals identify the central importance of managerial skills.
American Educational Research Journal, 48(5), 1091-1123.
Grissom, J., Loeb, S., & Master, B. (2013). Effective instructional time use for school leaders:
Longitudinal evidence from observations of principals. Educational Researcher, 42(8), 433-
444.
Handa, S., Thompson, C.L., Marcus, J.V., & Smith, A.A. (2010). The effects of UNC Master of
School Administration programs on NC public schools, 2004-2008. Carolina Institute for
Public Policy. Available from: http://publicpolicy.unc.edu/research/publications-
presentations-and-reports.
Henry, G.T., Bastian, K.C., & Smith, A.A. (2012). Scholarships to recruit the “best and
brightest” into teaching: Who is recruited, where do they teach, how effective are they, and
how long do they stay? Educational Researcher, 41(3), 83-92.
Hirsh, E., Koppich, J.E., & Knapp, M.S. (2001). Revisiting what states are doing to improve the
quality of teaching: An update on patterns and trends. Center for the Study of Teaching and
Policy. Available from: http://depts.washington.edu/ctpmail/PDFs/States-HKK-02-2001.pdf
Horng, E., Klasik, D., & Loeb, S. (2010). Principal’s time use and school effectiveness.
American Journal of Education, 116(4), 491-523.
Ladd, H. (2011). Teachers’ perceptions of their working conditions: How predictive of planned
and actual teacher movement? Educational Evaluation and Policy Analysis, 33(2), 235-261.
Leithwood, K. & Jantzi, D. (2005). A review of transformational school leadership research,
1996-2005. Leadership and Policy in Schools, 4(3), 177-199.
Loeb, S., Kalogrides, D., & Horng, E.L. (2010). Principal preferences and the uneven
distribution of principals across schools. Educational Evaluation and Policy Analysis, 32(2),
205-229.
Martorell, F., Heaton, P., Gates, S.M., & Hamilton, L.S. (2010). Preliminary findings from the
New Leaders for New Schools evaluation. Rand Education. Available from:
https://192.5.14.43/content/dam/rand/pubs/working_papers/2010/RAND_WR739.pdf
McCaffrey, D.F., Lockwood, J.R., Koretz, D., Louis, T.A., & Hamilton, L. (2013). Models for
value-added modeling of teacher effects. Journal of Educational and Behavioral Statistics,
29(1), 67-101.
Mihaly, K., McCaffrey, D., Lockwood, J.R., & Sass, T. (2010). Centering and reference groups
for estimates of fixed effects: Modifications to felsdvreg. The Stata Journal, 10(1), 82-103.
Miller, A. (2013). Principal turnover and student achievement. Economics of Education Review,
36, 60-72.
North Carolina Department of Public Instruction. (2013). North Carolina standards for school
executives. Available from: http://www.dpi.state.nc.us/docs/district-
humanresources/evaluation/standardsadmin.pdf
North Carolina Principal Fellows. (2015). http://www.ncpfp.org/
Robinson, V., Lloyd, C., & Rowe, K. (2008). The impact of leadership on student outcomes: An
analysis of the differential effects of leadership types. Educational Administration Quarterly,
44(5), 635-674.
Ronfeldt, M., Loeb, S., & Wyckoff, J. (2013). How teacher turnover harms student achievement.
American Educational Research Journal, 50(1), 4-36.
Vanderhaar, J., Munoz, M. & Rodosky, R. (2006). Leadership as accountability for learning:
The effects of school poverty, teacher experience, previous achievement, and principal
preparation programs on student achievement. Journal of Personnel Evaluation in Education,
19(1-2), 17-33.
Table 1: Characteristics of Principal Fellows and Other In-State Public MSA Graduates
Characteristics Principal Fellows Other In-State Public MSA
Total Number of Graduates
(1996-2012) 1228 4530
Percentage Female 71.12** 64.62
Percentage Racial/Ethnic
Minority 26.01
** 31.99
Age at Completion of MSA 36.07** 38.17
Std. Principal Licensure
Exam Scores 0.313
** 0.043
Taught in NCPS 0.963** 0.906
Years Teaching in
NCPS Pre-MSA 6.58
** 8.98
Assistant Principal in NCPS 0.963** 0.611
Principal in NCPS 0.461** 0.269
Age at First Principalship 38.45** 40.58
Note: This table displays descriptive characteristics for Principal Fellows and all other in-state public MSA
graduates from 1996 through 2012. +, *, and ** indicate statistically significant differences between Principal
Fellows and other in-state public MSA graduates at the 0.10, 0.05, and 0.01 levels, respectively.
Table 2: School Characteristics for First-Time Principals and All Principals
First-Time Principals All Principals
School
Characteristics Principal Fellows
Other In-State
Public MSA All Other Principal Fellows
Other In-State
Public MSA All Other
Subsidized School
Lunch Percentage 56.26 58.85 57.32 54.54 59.37
** 55.88
*
Minority Student
Percentage 46.56 49.93
+ 45.50 46.18 49.32** 46.42
Performance
Composite 73.93 70.75
** 71.35
* 72.64 70.36**
70.97**
Total Per-Pupil
Expenditures 8932.98 9219.30 10305.53
* 8779.73 9623.95**
9433.55**
Novice Teacher
Percentage 24.19 22.75
+ 22.78
+ 21.76 20.96** 21.36
NBC Teacher
Percentage 9.68 8.74
+ 10.23 12.14 11.41**
11.63*
Advanced Degree
Percentage 26.61 25.95 28.84
** 28.51 26.95**
29.08*
School Level
Elementary 71.29 61.59**
57.62** 66.51 56.68
** 57.09
**
Middle 18.39 20.21 20.28 19.00 19.42 17.10*
High 9.68 16.77**
19.77** 13.77 22.48
** 23.57
**
K-12 0.64 1.43 2.33* 0.72 1.42
** 2.24
**
Unique Principal
Count 322 891 828 520 1175 2565
Principal-by-Year
Count 322 891 828 2368 4727 9903
Note: The left panel of this table displays school-level characteristics for first-time principals during our study period. These characteristics are from the year
before the principal assumes leadership—the characteristics the first-time principal inherits. The right panel of this table displays school-level characteristics for
all principal-year combinations during our study period. +, *, and ** indicate statistically significant differences between Principal Fellows and the other
principal preparation categories at the 0.10, 0.05, and 0.01 levels, respectively.
Table 3: TWC Survey Items Included in Outcome Constructs
School Vision and Culture
Teachers and staff work in a school environment that is safe
The school leadership consistently enforces rules for student conduct
Teachers are centrally involved in decision making about educational issues
In this school we take steps to solve problems
There is an atmosphere of trust and mutual respect within the school
The school leadership consistently supports teachers
The faculty and staff have a shared vision
Instructional Leadership
Teacher performance evaluations are handled in an appropriate manner
The procedures for teacher performance evaluations are consistent
Teachers are held to high professional standards for delivering instruction
Teachers receive feedback that can help them improve teaching
Professional development provides teachers with the knowledge and skills most needed to teach
effectively
Table 4: Results for School Vision and Culture and Instructional Leadership
Approach 1: Principal-School Estimates
School Vision and Culture Instructional Leadership
Elementary
Schools
Middle
Schools
High
Schools
Elementary
Schools
Middle
Schools
High
Schools
NCPF 0.097
(1.083)
0.117
(1.043)
0.404
(0.938)
0.075
(1.107)
0.188
(0.988)
0.111
(1.256) Principal-School
Count 464 136 107 464 136 107
Other In-State
Public MSA
0.175
(1.117) 0.325
+
(1.085)
0.372
(1.211)
0.177
(1.128)
0.320
(1.072)
0.022
(1.411) Principal-School
Count 838 289 342 838 289 342
All Other -0.118
**
(1.266) -0.208
**
(1.101) -0.247
**
(1.436) -0.113
**
(1.292) -0.223
**
(1.126)
-0.028
(1.735) Principal-School
Count 1625 529 691 1625 529 691
Approach 2: Within School Estimates
NCPF 0.239
(1.059)
0.069
(0.864)
0.008
(0.585)
0.305
(1.296)
0.138
(0.691)
-0.095
(0.794)
Principal Count 240 61 48 240 61 48
Other In-State
Public MSA 0.360
(1.113) 0.313
+
(0.822) 0.182
+
(0.664)
0.468
(1.314)
0.197
(0.726)
-0.185
(0.840)
Principal Count 491 165 168 491 165 168
All Other -0.248
**
(1.229) -0.192
*
(0.931)
-0.083
(0.660) -0.321
**
(1.555) -0.140
**
(0.733)
0.096
(0.924)
Principal Count 944 291 372 944 291 372 Note: For the school vision and culture and instructional leadership constructs, the top panel of this table displays mean principal-by school fixed effects
estimates for each principal preparation category. The bottom panel of this table displays mean principal fixed effects estimates for each principal preparation
category. Standard deviations of the mean fixed effect estimates are in parentheses. +, *, and ** indicate a statistically significant mean difference between
Principal Fellows and other in-state public MSA graduates/all other principals at the 0.10, 0.05, and 0.01 levels, respectively.
Table 5: Teacher Retention Results
Approach 1: Principal-School Estimates
Principal Category Elementary Schools Middle Schools High Schools
NCPF 0.014
(0.124)
0.013
(0.097)
0.052
(0.193)
Principal-School Count 478 142 118
Other In-State Public MSA 0.032
*
(0.127)
0.024
(0.122)
0.045
(0.185)
Principal-School Count 872 303 365
All Other -0.021
**
(0.156) -0.016
**
(0.148) -0.031
**
(0.239)
Principal-School Count 1677 553 744
Approach 2: Within School Estimates
NCPF 0.021
(0.093)
0.005
(0.054)
0.004
(0.101)
Principal Count 252 65 54
Other In-State Public MSA 0.029
(0.089)
0.003
(0.062)
0.021
(0.106)
Principal Count 518 176 177
All Other -0.021
**
(0.107)
-0.003
(0.060)
-0.010
(0.115)
Principal Count 977 306 412 Note: The top panel of this table displays mean principal-by-school fixed effects estimates for each principal
preparation category. The bottom panel of this table displays mean principal fixed effects estimates for each
principal preparation category. Standard deviations of the mean fixed effect estimates are in parentheses. +, *, and
** indicate a statistically significant mean difference between Principal Fellows and other in-state public MSA
graduates/all other principals at the 0.10, 0.05, and 0.01 levels, respectively.
Table 6: Student Achievement Results
Approach 1: Principal-School Estimates
Principal Category Elementary Math Elementary Reading Middle Math Middle Reading High School EOC
NCPF 0.002
(0.142)
-0.018
(0.136)
0.020
(0.153)
0.013
(0.103)
0.045
(0.216)
Principal-School Count 469 469 223 223 115
Other In-State
Public MSA -0.025
**
(0.200) -0.046
**
(0.157) -0.012
*
(0.187) -0.016
**
(0.136)
0.039
(0.245)
Principal-School Count 862 862 491 491 366
All Other 0.012
(0.164) 0.029
**
(0.170)
0.002
(0.198)
0.006
(0.152) -0.026
**
(0.256)
Principal-School Count 1666 1666 885 885 758
Approach 2: Within School Estimates
NCPF -0.008
(0.080)
-0.027
(0.108)
-0.009
(0.083)
0.001
(0.064)
-0.004
(0.107)
Principal Count 249 249 116 116 54
Other In-State
Public MSA
-0.013
(0.087)
-0.037
(0.108)
-0.016
(0.086)
-0.006
(0.062) 0.050
**
(0.128)
Principal Count 513 513 281 281 175
All Other 0.009
**
(0.092) 0.026
**
(0.129) 0.011
*
(0.097)
0.003
(0.081)
-0.020
(0.142)
Principal Count 975 975 498 498 422 Note: The top panel of this table displays mean principal-by-school fixed effects estimates for each principal preparation category. The bottom panel of this
table displays mean principal fixed effects estimates for each principal preparation category. Standard deviations of the mean fixed effect estimates are in
parentheses. +, *, and ** indicate a statistically significant mean difference between Principal Fellows and other in-state public MSA graduates/all other
principals at the 0.10, 0.05, and 0.01 levels, respectively.
Table 7: Student Absences Results
Approach 1: Principal-School Estimates
Principal Category Elementary Schools Middle Schools High Schools
NCPF -0.075
(0.047)
0.405
(0.123)
-0.977
(0.306)
Principal-School Count 468 138 109
Other In-State Public MSA -0.005
(0.039) 0.780
*
(0.104) -0.217
*
(0.212)
Principal-School Count 850 298 345
All Other 0.010
(0.030) -0.548
**
(0.087) 0.190
**
(0.161)
Principal-School Count 1648 542 701
Approach 2: Within School Estimates
NCPF -0.055
(0.455)
0.143
(1.014)
-0.031
(1.619)
Principal Count 250 65 53
Other In-State Public MSA -0.078
(0.540) 0.480
*
(1.097)
-0.045
(1.641)
Principal Count 513 175 177
All Other 0.055
**
(0.584) -0.306
**
(1.488)
0.023
(1.692)
Principal Count 966 306 407 Note: The top panel of this table displays mean principal-by-school fixed effects estimates for each principal
preparation category. The bottom panel of this table displays mean principal fixed effects estimates for each
principal preparation category. Standard deviations of the mean fixed effect estimates are in parentheses. +, *, and
** indicate a statistically significant mean difference between Principal Fellows and other in-state public MSA
graduates/all other principals at the 0.10, 0.05, and 0.01 levels, respectively.
Table 8: Principal Persistence
Principal Persistence in the Same School
Elementary Schools Middle Schools High Schools
Other In-State Public MSA -0.002
(0.014)
0.000
(0.027) 0.078
*
(0.032)
All Other 0.010
(0.014)
-0.027
(0.024)
0.042
(0.031)
Cases 7644 2366 2799
Principal Persistence in NCPS
Elementary Schools Middle Schools High Schools
Other In-State Public MSA 0.014
(0.011)
-0.010
(0.021) 0.053
+
(0.029)
All Other 0.007
(0.011) -0.044
*
(0.021)
0.034
(0.028)
Cases 7644 2366 2799 Note: Cells report coefficients from linear probability models in reference to Principal Fellows with standard errors
in parentheses. +, *, and ** indicate statistically significant differences at the 0.10, 0.05, and 0.01 levels,
respectively.
Appendix Table 1: Student Achievement Results with Empirical Bayes Shrinkage
Approach 1: Principal-School Estimates
Principal Category Elementary Math Elementary Reading Middle Math Middle Reading High School EOC
NCPF 0.003
(0.132)
-0.014
(0.120)
0.024
(0.134)
0.016
(0.078)
0.044
(0.194)
Principal-School Count 469 469 223 223 115
Other In-State Public
MSA -0.016
*
(0.130) -0.035
**
(0.111) -0.004
*
(0.161) -0.009
**
(0.102)
0.043
(0.213)
Principal-School Count 862 862 491 491 366
All Other 0.013
(0.139) 0.024
**
(0.134)
0.009
(0.167)
0.009
(0.109) -0.015
**
(0.228)
Principal-School Count 1666 1666 885 885 758
Approach 2: Within School Estimates
NCPF -0.006
(0.068)
-0.023
(0.094)
-0.005
(0.064)
0.003
(0.049)
-0.001
(0.093)
Principal Count 249 249 116 116 54
Other In-State Public
MSA
-0.010
(0.071)
-0.031
(0.092)
-0.015
(0.071) -0.007
*
(0.039) 0.039
**
(0.102)
Principal Count 513 513 281 281 175
All Other 0.006
*
(0.072) 0.021
**
(0.105) 0.009
*
(0.077)
0.003
(0.042)
-0.015
(0.114)
Principal Count 975 975 498 498 422 Note: The top panel of this table displays mean principal-by-school fixed effects estimates for each principal preparation category. The bottom panel of this
table displays mean principal fixed effects estimates for each principal preparation category. Standard deviations of the mean fixed effect estimates are in
parentheses. +, *, and ** indicate a statistically significant mean difference between Principal Fellows and other in-state public MSA graduates/all other
principals at the 0.10, 0.05, and 0.01 levels, respectively.
Appendix Table 2: Student Achievement Results Using Restricted Sample
Approach 1: Principal-School Estimates using Within School Effectiveness Sample
Principal Category Elementary Math Elementary Reading Middle Math Middle Reading High School EOC
NCPF 0.007
(0.132)
-0.012
(0.119)
0.003
(0.172)
0.022
(0.125)
-0.104
(0.242)
Principal-School Count 249 249 116 116 54
Other In-State Public
MSA
0.010
(0.146)
-0.024
(0.131)
-0.021
(0.172)
0.003
(0.145)
-0.071
(0.233)
Principal-School Count 513 513 281 281 175
All Other -0.007
(0.160) 0.016
**
(0.141)
0.011
(0.226) -0.007
*
(0.155) 0.043
**
(0.404)
Principal-School Count 975 975 498 498 422 Note: This table displays mean principal-by-school fixed effects estimates for each principal preparation category. Standard deviations of the mean fixed effect
estimates are in parentheses. +, *, and ** indicate a statistically significant mean difference between Principal Fellows and other in-state public MSA
graduates/all other principals at the 0.10, 0.05, and 0.01 levels, respectively.