8/3/2019 Paper 2726
1/31
1
Teacher Turnover in New York Citys Public Middle SchoolsA paper prepared for the fall, 2011 APPAM conference
Will Marinell and Aaron Pallas
Several recent studies and a high profile report have underscored the importance oflearning more about the causes and consequences of teacher turnover in New York Citys middleschools. One recent investigation found that rates of turnover were higher among first-year New
York City middle school math teachers than among elementary teachers and that the most
effective middle school math teachers who left their schools after one year tended to leave the
lowest-performing schools.1
Further, a panel of experts recently noted that many of New York
Citys middle schools possess characteristics associated with high rates of turnover, such as large
percentages of underperforming students and high rates of principal turnover.2
Given the role
that students performance in middle school can have on their progress towards post -secondary
work and study,3
it seems critical to learn more about middle school teacher turnover in order tohelp schools develop strategies for recruiting, developing and retaining effective teachers, while
simultaneously limiting turnovers damaging consequences.
To date, however, there has not been a comprehensive resource that addresses central
questions related to New York City middle school teacher turnover and identifies importantavenues for future research. The Research Alliance for New York City Schools aims to fill this
gap through a three-year, mixed-methods study of New York City middle school teacher
turnover. The goals of this project are to describe the rates and patterns of turnover over the past
decade, identify middle school teachers career plans and their impressions of their schools, and
examine the causes and consequences of turnover in middle schools that serve high-need student
populations. This project is a collaboration among researchers at Baruch College (City
University of New York), Teachers College (Columbia), and New York University. The study isorganized into three components, each of which investigates questions related to turnover using
one of the following sources of data: the New York City Department of Educations humanresources administrative records, an original survey of middle school teachers, and case studiesof four middle schools.
This paper presents findings from two of the studys three components. In the firstcomponent of the study, we use human resources data from 2001 to 2010 to identify the
characteristics of New York City middle school teachers, describe the rates and patterns ofturnover over the past decade, and investigate the relationship between turnover and the
characteristics of middle school teachers and middle schools. In doing so, the report aims to
address foundational questions related to turnover and to identify some key questions that futurestudiesincluding the survey and case study components of this larger project should explore
in order to gain a nuanced understanding of middle school teacher turnover. In the secondcomponent of the study, we use data from a survey of 4,214 teachers in grade 6-8 New York City
middle schools to investigate the relationship between middle school teachers impressions oftheir school and their career intentions.
8/3/2019 Paper 2726
2/31
2
Background
Teacher turnover is one of the more widely studied topics in K-12 education, and with
good reason. Research indicates that high rates of teacher turnover can leave schools facinginstructional, financial, and organizational costs, which can be difficult to surmount.4 While
some degree of turnover can be constructive for organizations, a perpetual churning of teachers
through schools requires that administrators devote scarce resources to recruiting and orientingnew teachers. Further, turnover can compromise a schools long-term objectives, such as its
efforts to promote a professional culture or strengthen its instructional core through sequential
professional development.5
Evidence suggests that rates of teacher turnover may be particularly high in urban middleschools, especially those serving disadvantaged students. A recent study found that 60% of
novice math teachers in low-performing middle schools left their schools within two years.6
This
same study revealed a potentially vicious cycle, wherein the least effective middle school
teachers rotate through the schools that serve the largest percentages of underperforming studentsand students from minority backgrounds.7 Outside of New York City, research has shown that
middle schools have relatively high rates of out-of-field teaching, a factor that is associated with
turnover, and that some middle school teachers view their assignments as stepping-stones topositions in elementary or high schools.
Methodology
For the first component of the study the investigation of patterns of turnover over the
past decade this investigation uses a discrete-time survival analysis methodology to estimate
the length of time that teachers remain in their schools. From these estimates, we generatestatistics of the percentage of teachers who left their schools within various lengths of time (e.g.,
after their first year, within three years, etc.). The primary sample for this analysis is the 15,628
teachers who were new to one of New York Citys 196 Grade 6-8 middle schools between 2002and 2009.8The analysis follows these teachers careers from the time they enter their school until
they either depart these schools or are censored by the data set in 2010.
To examine teachers impressions of their schools using our original survey data the
second component of this study we conduct two-level hierarchical logistical regression
analyses to investigate the extent to which teachers personal characteristics and theirimpressions of their school predict two outcomes: 1) teachers consideration of leaving theirschool, and 2) teachers consideration of leaving the teaching profession. In the second stage of
this analysis, we conduct multi-level logistical regression to describe the variation among NewYork City middle schools in the proportion of teachers who report that they are considering
leaving their school and/or teaching. In addition, this component of the study examines the extent
to which the variation across schools is attributable to differences in teachers characteristics,school demographics, and teachers impressions oftheir school as a workplace.
8/3/2019 Paper 2726
3/31
3
Findings: Component 1
Analysis and Text by Will Marinell
How long do middle school teachers remain in their schools?
This question addresses a topic of central importance to school principals, school systemadministrators and organizations invested in preparing teachers for their work and developing
their capabilities on the job: after teachers enter New York City middle schools, how long do
they remain in their schools? On average, middle school teachers who entered their schoolsbetween 2002 and 2009 remained in these schools for roughly three years. More specifically,
27% of middle school teachers left their schools within one year, 55% within three years and
66% within five years. To situate these rates of turnover within the larger context of the NewYork City public school system, we estimated comparable rates of turnover among the
elementary and high school teachers who were new to New York City schools during the same
time period. Exhibit 1 illustrates the cumulative percentage of teachers who left New York Citys
middle, elementary and high schools after various lengths of time. As the exhibit depicts, the
rates of middle school turnover are either comparable or slightly higher than rates of turnover inelementary and high schools. For example, 55% of middle school teachers left their schools
within three years, as compared with 46% of elementary school teachers and 51% of high schoolteachers. Exhibit 1 also demonstrates that, across all of these school levels, the steepest increases
in turnover occurred during teachers first few years in schools, after which turnover ratesgradually leveled off.
8/3/2019 Paper 2726
4/31
4
How have rates of middle school teacher turnover changed over the past decade?
To inform preliminary hypotheses about whether and how turnover may have beeninfluenced by a number of factorssuch as New York Citys movement towards an open-market
hiring systemwe examined whether rates of turnover changed between 2002 and 2007. These
exploratory analyses revealed that rates of middle school teacher turnover declined slightly overthis time period. More specifically, 57% of teachers who entered middle schools during the
2001-2002 school year left those schools within three years. Rates of turnover declined slightly
among teachers who entered their schools over the subsequent five years. Among teachers whoentered middle schools during the 2006-2007 school year, 52% left these schools within three
years. This five percentage-point decrease in turnover rates between 2002 and 2007 represented a
statistically significant negative linear trend; however, the small magnitude of this difference
seemed less notable than the discovery that there was not a year during this time period whenmore than 50% of middle school teachers remained in their schools for longer than three years.
To what extent is turnover characterized by mobility between schools or attrition from the New
York City public school system?
Future studies of middle school teachers career plans and of the factors that influence
whether they remain in their schools, in the New York City public school system, or in teachingshould be grounded in a descriptive analysis of teachers patterns of mobility and attrition. Data
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1 2 3 4 5 6 7 8Cumulative
percentageofteacherswholeftthe
irschools
Years in school
Exhibit 1
The percentage of new-to-school middle, high and elementary school
teachers who left their schools between 2002 and 2009
Middle
High
Elementary
8/3/2019 Paper 2726
5/31
5
from the past decade reveal that both mobility (transferring between New York City schools) and
attrition (leaving the New York City public schools altogether) have contributed to New YorkCity middle school teacher turnover. Exhibit 2 illustrates the patterns of mobility and attrition
among teachers who left their middle schools between 2002 and 2009. As the exhibit reveals,
59% of departing middle school teachers were not employed in the New York City public school
system in the year after their departure (referred to as Leavers). By comparison, 41% ofdeparting middle school teachers transitioned to another New York City public school (referred
to as Movers). Further, as the exhibit indicates, 19% of all of the departing middle school
teachers secured assignments in New York City public schools that did not include the middlegrades (Grades 6-8). Twelve percent of all departing middle school teachers transitioned to
middle schools with traditional Grade 6-8 configurations.
Which teacher characteristics are associated with turnover?
Historically, research on turnover has investigated the relationship between turnover and
the characteristics of teachers and schools in an effort to help practitioners and policymakers
identify which teachers leave schools and which types of schools they leave behind.9
Exhibit 3depicts the percentage of teachers with various background characteristics who left their middle
schools within three years. As this exhibit indicates, while these teacher characteristics are allstatistically significant predictors of turnover, turnover rates are similar among teachers within
many of the larger categories, such as race, gender, and subject area. For instance, roughly 50%
of middle school teachers left their schools within three years, regardless of whether they weremale or female, or from Black, Hispanic, or White racial/ethnic backgrounds. Similarly, the
percentage of math and science middle school teachers who left their schools within three years
did not differ notably from rates of turnover among their colleagues who teach other subjects.
There a few notable exceptions to this general pattern of similarity between teachers
characteristics and rates of turnover. More specifically, turnover rates varied across teachers of
different ages, levels of experience, and degree credentials. For instance, among the teacherswho entered their schools between 2002 and 2009, the teachers with the most experience in New
York City schools were associated with the lowest rates of turnover (44% left their schoolswithin three years). By contrast, 55% of teachers who had worked in New York City schools for
three years or less left their schools within the same period of time. In addition, older and
younger teachers were associated with higher rates of turnover than middle-aged teachers. Morespecifically, 61% of teachers aged 55-or-older and 54% of teachers aged 30-or-younger left their
schools within three years. By comparison, 49% of teachers aged 30-55 left their schools within
the same length of time.
8/3/2019 Paper 2726
6/31
6
NYC middle school teachers who entered their
schools between 2002-09(N= 18,019)
Departed schools during the
period of observation
(N = 10,405)
The Leavers: Left the NYC
public school system
59%
The Stayers: Remained in
schools for the period of
observation
(N= 7,614)
The Movers: Transferred to
other NYC schools
41%
Remained teachers
37%
Transitioned to schools that did not
includegrades 6-819%
Transitioned to schools that included
grades 6-818%
Assumed non-
teaching roles
4%
Grade 6-8 middle
schools
12%
Grade 6-12
schools
3%
Grade K-8schools
3%
High schools
8%
Elementary schools
5%
Schools with other gradeconfigurations and
ungraded schools
6%
Exhibit 2
Patterns of mobility and attrition among teachers who entered NYC middle schools between 2002 and 2009
8/3/2019 Paper 2726
7/31
7
Exhibit 3Estimated percentage of middle school teachers who left their schools within three years
By selected teacher characteristics
Left within 3 years (%)Statistical
significance
GenderFemale 51
***Male 53
Ethnicity
White 53
**Black 52
Hispanic 48
Other 54
Years of experience in NYC schools
3 years or less 55
***Between 3 and 6 years 50Between 6 and 9 years 47
More than 9 years 44
Degree level
B.A. 51
***M.A. or credit equivalent 50
M.A. and 30 additional credits 55
Other 64
Age
30-years-old or younger 54
***Between 30-55 years-old 49
Older than 55 61
Subject area
Math or science teacher 54**
Non-math or science teacher 52SOURCE: Discrete time survival analyses that model the probability of turnover for the various teacher
characteristics, while simultaneously controlling for other teacher characteristics and contextual factors, such as a
schools NYC borough location and its annual change in student enrollment. Please see the full report and
Technical Documentation for detailed information about modeling procedures.
NOTES: Statistical significance key: ~ p < 0.10; * p < 0.05; ** p < 0.01; *** p < 0.001
Which school characteristics are associated with teacher turnover?
Using a variety of publicly-available data, and controlling for various characteristics ofteachers and larger, system-wide contextual factors10, we examined whether turnover rates
differed across various types of middle schools. Exhibit 4 depicts the percentage of middle
school teachers who left their schools within three years across middle schools with differentcharacteristics. As with the previous exhibit, the statistics presented in Exhibit 4 suggest that
rates of turnover are similar across many different types of middle schools, with a few notable
exceptions. For instance, smaller middle schools were associated with higher levels of turnover,
8/3/2019 Paper 2726
8/31
8
on average. Fifty-five percent of the teachers who entered smaller middle schools (i.e., schools
with roughly 700 students) between 2002 and 2009 left these schools within three years. Bycomparison, schools that enrolled approximately twice as many students lost about 48% of
similar teachers within the same time period.
Exhibit 4Estimated percentage of middle school teachers who left their schools with within three years
By selected school characteristics
Left within 3
years (%)
Statistical
significance
School sizea
692 (25th
percentile) 55
***1,122 (50th
percentile) 51
1,383 (75th
percentile) 48
Quality Review score
Under-developed 51
0.3544Under-developed with proficient features 53
Proficient 52
Well-developed 51
Percentage of students in povertyb
59 (25th
percentile) 52
**71 (50th percentile) 51
83 (75th
percentile) 50
Percent Proficient or higher on NY math assessmenta
30 (25th
percentile) 54
***49 (50th
percentile) 5169 (75
thpercentile) 49
Weighted school environment score from School Surveyb
8.1 (25th
percentile) 51
***9.4 (50th percentile) 49
11 (75th
percentile) 47
Teachers race and whether school had a relatively high
proportion of White studentsa
White teacher, high proportion of White students 47
***
White teacher, not a high proportion of White students 55
Black teacher, high proportion of White students 53
Black teacher, not a high proportion of White students 47
Hispanic teacher, high proportion of White students 53
Hispanic teacher, not a high proportion of White students 46
8/3/2019 Paper 2726
9/31
9
Exhibit 4 (continued)
SOURCE: Discrete time survival analyses that model the probability of teacher turnover across schools with
various characteristics, controlling for all of the contextual factors, teacher characteristics, and school
characteristics in the analysis. Please see the full report and Technical Documentation for detailed information
about modeling procedures.
NOTES: Statistical significance key: ~ p < 0.10; * p < 0.05; ** p < 0.01; *** p < 0.001
a: Data from 2002-09
b: Data from 2008
Motivated by the work of Thomas Dee and others11, we examined the relationship
between turnover and the match between teachers and students racial/ethnic backgrounds.
These analyses revealed that middle school teachers remained in their schools longer when theirracial and ethnic characteristics matched those of a substantial proportion of the students in their
schools. In the typical New York City middle school, roughly 16% of students are White and
84% are non-White. Thus, schools where more than 16% of students are White could be
considered schools with a relatively high proportion of White students even though White
students do not represent the predominant racial/ethnic category in the school.As Exhibit 4 depicts, controlling for other factors, White teachers who were working in
schools with a relatively large proportion of White students were associated with lowerrates of
turnover than were White teachers working in schools where White students did not comprise anuncharacteristically large proportion of the student body. Similarly, Black and Hispanic teachers
who were working in schools that had a relatively large proportion of non-White students were
associated with lower rates of turnover than were Black and Hispanic teachers who wereworking in schools that did not have an uncharacteristically large proportion of non-White
students.
While not the focus of this portion of the study, the relationship between turnover and the
contextual factors for which our analyses controlled, such as the borough in which a middleschool was located, yielded patterns of turnover that future studies should investigate further. For
instance, there were wide discrepancies in the percentage of middle school teachers who left
schools across the five boroughs. Rates of turnover were highest in Manhattan, where 66% of the
teachers who entered middle schools between 2002 and 2009 left within three years. By contrast,within the same period of time, 63% of similar teachers left middle schools in the Bronx, 54%
left schools in Brooklyn, 49% left schools in Queens, and 35% left schools in Staten Island.
Findings: Component 2
Analysis and Text by Aaron Pallas (modified by Marinell for APPAM paper)
In the second component of this study, we use our original survey of middle schoolteachers to further examine the relationships between middle school teacher turnover and both
the characteristics of teachers and schools. The survey data provide the study with measures of
schools operational functioning such as, the quality of the administrative leadership and thelevel of influence that teachers have over school policies that are not available from any other
8/3/2019 Paper 2726
10/31
10
source. Our analysis for this component of the study proceeds in four parts. First, we present the
percentage of middle school teachers who reported considering leaving their school and/or theteaching profession. Second, we examine the extent to which a variety of detailed teacher and
school characteristics predict whether a teacher considered leaving her school. In the third
section, we analyze the reasons behind teachers consideration of leaving their school. In the
fourth and final section, we examine whether there is evidence that some middle schools haveunusually high rates of teachers considering leaving their current school and/or leaving teaching.
We describe the characteristics of the schools that have higher-than-average rates of teachers
thinking about leaving, particularly the attributes of the teachers who teach in them and theschools' demographic characteristics and work environments. In our analyses, we explore the
extent to which we can explain this between-school variation with the various teacher and school
characteristics that we use in our previous analyses within this component of the study.
What percentage of middle school teachers are considering leaving their school and/or
teaching?
We begin by describing teachers' responses to two questions about whether theyconsidered leaving their school or the teaching profession during the 2009-2010 school year.
One-third (33%) of the 4,214 responding full-time middle school teachers reported that they hadconsidered leaving their school during the current school year. A smaller, but still substantial,
fraction22%reported that they had considered leaving teaching altogether during the current
school year. Exhibit 5 shows the four different patterns that teachers reported.
Exhibit 5. Patterns of Thinking about Leaving One's School and Teaching
Pattern of Thinking about Leaving Percentage
Considered leaving neither 61
Considered leaving teaching only 7
Considered leaving school only 16
Considered leaving both school and teaching 16
Total 100
Exhibit 5 indicates that 39% of teachers in the study considered leaving their school or
leaving teaching altogether during the 2010 school year. Arguably, the most detached teachersare the 16% who reported that they both considered leaving their school andthat they considered
leaving teaching. An equal number of teachers stated that they had considered leaving their
school, but had not considered leaving teaching. This latter group presumably is still committed
to teaching, but would prefer to teach at another school. But there is a fourth group of teacherswho report that they had considered leaving teaching, but they had not considered leaving their
school. The fraction of such teachers, 7% of the total sample, is not large, but neither can it be
ignored, as these teachers are expressing an attachment to their current school even as they
consider leaving teaching altogether. One possibility, which became clear only after the surveyadministration was complete, is that this latter group of teachers might be considering remaining
in their current schools, but in non-teaching roles such as assistant principal or some other
8/3/2019 Paper 2726
11/31
11
administrative capacity. The wording of the questions in the survey was less precise than
intended, which creates some ambiguity about these teachers future intentions.
One might worry that considering leaving one's school, or leaving teaching, might befleeting thoughtsa bad day spurring a teacher to fantasize about leaving, rather than a more
durable predictor of a teacher's future behavior. There is some evidence that thinking about
leaving is associated with teachers' plans for staying in their school and for remaining in theteaching force. We asked teachers how long they anticipated remaining in their current school:
until the end of the school year; for another year or two; for between 3-5 years; for between 6-10
years; and for more than 10 years. Teachers who reported thinking about leaving their schooland/or leaving teaching anticipated much shorter stints in their current school. Exhibit 6 shows
this association.
Exhibit 6. Association Between Considering Leaving Current School
and Expected Tenure in School
Considered leaving current schoolExpected tenure in current school No Yes
Until the end of this school year 1% 25%
For another year or two 15 45
For between 3-5 years 24 18
For between 6-10 years 20 5
For more than 10 years 39 7
Total 99% 100%
Of the teachers who reported having considered leaving their current school, one-quarter
expected to leave at the end of the current year, and another 45% expected to stay for at most
another year or two. Thus, 70% of the teachers who reported that they had considered leaving
their school expected to be gone within two years. Only 12% of these teachers expect to stay intheir current schools for more than six more years. Conversely, the teachers who had not
considered leaving demonstrate a strong attachment to their current school. Nearly 40% expect to
remain in their current schools for more than 10 years, and 59% expect to remain for another sixyears or more. Only one in six of the teachers who had not considered leaving their current
school anticipated leaving within two more years.
A similar pattern is observed for teachers who reported having considered leaving
teaching altogether. As Exhibit 7 demonstrates, 41% of the teachers who considered leavingteaching expected to stay in teaching for another year or two. The pattern is quite different for
teachers who did not consider leaving teaching; these teachers expect to remain in teaching for asubstantial amount of time. Sixty percent of the teachers who did not consider leaving anticipate
staying in teaching for at least another 10 years, and more than three-quarters (78%) expect tocontinue as teachers for at least another six years. Only 7% of the teachers who did not consider
leaving teaching anticipate leaving the field within two years.
8/3/2019 Paper 2726
12/31
12
Exhibit 7. Association between Considering Leaving Teaching
and Expected Tenure in Teaching
Considered leaving teaching
Expected tenure in teaching No Yes
Until the end of this school year 1% 12%For another year or two 6 29
For between 3-5 years 15 25
For between 6-10 years 18 12
For more than 10 years 60 22
Total 100% 100%
Which teacher and school characteristics predict whether teachers are considering leaving
their school and/or teaching?
In this section, we examine whether a variety of teacher and school characteristics from
the survey predict whether teachers considered leaving their school. We conducted a parallelanalysis with our other outcome of interestconsidering leaving teachingbut do not report the
results, which were similar, for the sake of parsimony. The following five models are estimated:
Model 1 includes a set of teacher characteristics, including experience, certification andcredentials, family status, gender, race/ethnicity, whether the teacher is teaching a new subject or
a new grade level, and commuting time to work.
Model 2 adds to these teacher characteristics four additional social-psychological measures: a
teacher's satisfaction with his/her salary; whether the teacher's peers are encouraging her to stayin the school; the teacher's self-assessment of her effectiveness (i.e., teacher efficacy); and the
teacher's self-report of the strength of her personal ties with students. We might predict that
higher values on these social-psychological dimensions would be associated with a lowerprobability of considering leaving the current school.
Model 3 includes school characteristics, but no teacher characteristics. The school characteristics
included in the model are the school's total enrollment; the percentages of students in poverty,receiving special education services, and English Language learners; the percentage of students
in the school who had been suspended in the prior year; the average median proficiency in math
and English Language Arts for students in the school; the school attendance rate and rate ofstudent stability from the preceding year; and an indicator for the borough in which the school islocated.
Model 4 adds to the Model 3 school characteristics an additional set of school characteristics that
measure various aspects of the school as a workplace. These measures are derived from theaggregated reports of individual teachers in each school and include: teacher collegiality,
principal leadership, parental support, collective responsibility for student learning, school
8/3/2019 Paper 2726
13/31
13
disorder, influence/control over professional matters in the school, the frequency of and
satisfaction with professional development, and the adequacy of the available resources and ofinduction strategies for new teachers. The contrast between Model 3 and Model 4 represents the
unique contribution of the school as a workplace to explaining why teachers consider leaving,
and the extent to which the effects of school structural characteristics in Model 3 are mediated by
the school as a workplace.
Model 5 is the most inclusive model, combining all of the teacher and school characteristics from
the previous models.
Because teachers are nested within schools, we estimate two-level hierarchical logisticregression models with a random school effect and fixed effects for teacher and school
characteristics. This modeling approach adjusts the estimated standard errors for the logistic
regression coefficients for the clustering of teachers with similar characteristics within schools.
As we demonstrate later, there is a significant amount of variation between schools in the
average probability of teachers considering leaving the school (and considering leavingteaching).
To simplify the interpretation of the models, the tables which follow report statistically
significant effects of a variable on the odds of considering leaving. The odds of an event is afunction of the probability of the event, much as we think about odds in games of chance. The
odds of considering leaving ones school is defined as the probability of considering leaving
divided by the probability of not considering leaving. When we refer to a probability of an eventas "50:50," the odds of that event is .50/.50 = 1. Similarly, the phrase "a one in five chance"
equates to an odds of .20/.80 = .25, and a 75% probability represents an odds of .75/.25 = 3.0.
To place these odds in perspective, overall, 22% of the teachers in the study who never
considered leaving teaching reported that they considered leaving their current school. Thisprobability corresponds to an odds of .28. An increase in the odds signifies an increase in theprobability of considering leaving, whereas a decrease in the odds indicates a decrease in the
probability of considering leaving. A 100% increase in the odds of considering leavingthat is,
an odds of .56represents an increase in the relative probability of considering leaving from22% to 36%. A 50% increase in the odds of considering leavingthat is, an odds of .42shifts
the relative probability of considering leaving from 22% to 30%. Conversely, a 50% decrease in
the odds of considering leaving, which corresponds to an odds of .14, represents a decrease in therelative probability of considering leaving from 22% to 12%.
In some cases, distinctions among teachers are represented as contrasts among a set of
categories (e.g., marital status categories), with one category denoted as a "reference category"
against which the other categories are compared. Throughout the analyses, we test whether a setof contrasts collectively is equal to zero, or whether that hypothesis should be rejected in favor ofthe hypothesis that the contrasts collectively differ from zero. Not every contrast in a set needs to
be different for the set collectively to differ from zero, however; for example, the level of some
outcome might differ between divorced teachers and a reference category of teachers who nevermarried, but might not differ between widowed and never-married teachers. The following
exhibits indicate which specific contrasts are significant at p
8/3/2019 Paper 2726
14/31
14
Exhibit 8 displays the extent to which the five models predict the odds of considering
leaving one's school only (compared to not considering leaving the school or leaving teaching).As the exhibit reveals, two features of teachers' work are significant predictors of their
consideration of leaving their school. Teachers who report that they are teaching a subject they
have not taught before are considerably more likely to consider leaving their schools. The 77%
increase in the odds of leaving represents a shift in the predicted probability of leaving from 22%to 33%. Conversely, teachers who report teaching a grade level they have not taught before are
somewhat less likely to consider leaving their school for another school. It is possible that this
pattern reflects the extent to which these changes in teaching assignments are voluntary orinvoluntary. Because middle-school teachers are more likely to be subject-matter specialists than
are elementary school teachers, a shift in subject may represent an unwelcome transition to a new
subject for which a teacher feels less prepared, or which holds less interest. Conversely, a changein grades taught within a middle school may not be nearly as disruptive.
Teachers commuting time also has a powerful effect on the odds of their considering
leaving their school. Not surprisingly, the longer a teacher's commute, the more likely she is to
consider leaving her school. The pattern is particularly pronounced for the longest commutes.
The difference between a commute of 20 minutes or less and a commute of an hour or morerepresents a 147% difference in the odds of considering leaving one's school. An increase of this
magnitude equates to a shift in the probability of considering leaving one's school from 22% to
41%, a very large effect.
Model 2 adds the social-psychological factors of satisfaction with salary, peerencouragement to stay at the current school, teacher self-efficacy and personal ties to students.
For the most part, the patterns displayed in Model 1 are unchanged, suggesting that their
influence on the odds of considering leaving the current school to teach in another school areindependent of these social-psychological factors. When the social-psychological factors are
taken into account, teachers who self-identify as Hispanic or as a race other than Black, Asian or
White are more likely to consider leaving their school than are White or Black teachers.
Three of the four social-psychological factors are associated with the odds of consideringleaving one's current school. Teachers who are more satisfied with their teaching salaries are
more likely to consider leaving, so perhaps the move would not pose a financial risk; a move
within the New York City school system to another teaching position would likely not affect a
teacher's salary and, hence, financial stability. On the other hand, teachers apparently do respondto the influence of others around them and to their own assessments of their teaching
competence. A one standard deviation increase in the strength of the encouragement teachers
receive from those close to them to stay in their current school reduces the odds of consideringleaving the school by 52%. Moreover, those teachers who view themselves as effective are less
likely to consider leaving their current school. But the strength of a teacher's personal ties to
students is not associated with the likelihood of his considering leaving the current school.
8/3/2019 Paper 2726
15/31
15
Exhibit 8. Models Predicting the Odds of Considering Leaving One's School to Teach Elsewhere
(1) (2) (3) (4) (5)
0 to 1 year experience -46% -50% -49%
1 to 3 years experience -21% -17% -20%
4 to 5 years experience 13% 21% 18%
11 to 15 years experience -10% -8% -8%>15 years experience -42% -33% -36%
Black 2% -14%
Asian -15% -19%
Hispanic 69% 48%
Other race 82% 71%
Gender (Male)
Living with domestic partner 12% -2%
Married -32% -33%
Divorced -14% -26%
Separated -76% -69%
Widowed -67% -64%
N of children responsible for -16% -18% -18%
Teaching new subject 77% 63% 58%
Teaching new grade -28% -27%
NYC public high school graduate
NYC Teaching Fellow
Teach For America corps member
Teaching Opportunity Program alum
Other alternative certification
Holds MA degree 39%
Holds MA + 30 credits 5%
Holds doctoral degree 100%
20 to 40 minute commute 29% 15% 11%
40 to 60 minute commute 66% 49% 42%
60 or more minute commute 147% 129% 106%
Satisfaction with salary 17% 14%
Peer encouragement to stay at school -52% -48%
Teacher self-efficacy -22%
Personal ties to students
School total enrollment
School poverty rate
Pct American Indian students 3%Pct Black students -23%
Pct Hispanic students -19%
Pct Asian students -35%
Brooklyn -34% -33%
Queens -40% -50%
Bronx -40% -40%
Staten Island -73% -73%
8/3/2019 Paper 2726
16/31
16
The final point to note before turning to school-level characteristics is the lack of an associationbetween teacher certification status and the probability of considering leaving one's current
school. In both Model 1 and Model 2, there is no evidence that teachers prepared via an alternate
route, such as Teach for America or the New York City Teaching Fellows, are more likely toconsider leaving their current schools for another school than are teachers who entered the field
via a traditional teacher certification program.
Model 3 introduces a set of school characteristics as predictors of the likelihood of ateacher considering leaving her current school. Only two school structural characteristics areassociated with this outcome. First, the teachers in schools located in Manhattan are considerably
more likely to think about moving to another school than the teachers in schools in Staten Island,
the Bronx, and Queens. The odds of teachers in Staten Island schools considering leaving are
77% lower than is observed in Manhattan, which corresponds to a difference in the probability ofconsidering leaving of 10% and 32%, respectively, all else being equal. Second, teachers in
schools with higher rates of student suspensions are more likely to consider leaving their current
school for another school. A one standard deviation increase in the student suspension ratethedifference between 15% and 26%is associated with an 18% increase in the odds of considering
moving to another school. A higher rate of suspensions presumably indicates that the school has
challenges maintaining order and safety, two factors that might be salient to a teacher'sassessment of the suitability of her current school.
Model 4 adds to the equation school-level measures of the school as a workplace. Thedifferences among boroughs persist, which suggests that the reasons that Manhattan teachers are
more likely than teachers in other boroughs to consider leaving their school have little to do with
the school as a workplace. The same logic holds for the fact that teachers in schools with higherrates of student suspensions are more likely to consider moving to another school: it's not that
Exhibit 8 Continued
(1) (2) (3) (4) (5)
Pct Special Education students
Pct ELL students
Pct suspended students 18% 20% 23%
Avg median proficiency
School attendance rate
Student stability rate
Teacher collegiality
Principal leadership
Parental support
Collective responsibility
School disorder
Professional control
Frequency of professional development 15%
Satisfaction with professional development
Adequacy of resources -24%
Adequacy of teacher induction
8/3/2019 Paper 2726
17/31
17
such schools are functioning more poorly, and that teachers are responsive to the organization
and functioning of the school.
In fact, the key story here is that school organization and functioning have relatively littleto do with whether New York City middle school teachers consider leaving their school. This
may be due, in part, to the correlations between these measures then render them insignificant
when added to the model as a group. Of the ten school-level measures of the school as aworkplace, only twothe frequency of professional development activities and the adequacy of
school resourcesare statistically significant predictors of the odds of considering leaving one's
school when the measures were added to the model simultaneously. Somewhat perversely,teachers in schools with more frequent professional development are slightly more likely to
consider leaving their current schools, which may reflect a feeling that professional development
is being imposed on them. (In contrast, satisfaction with professional development is unrelated to
the likelihood of considering moving to another school.) Additionally, teachers who view theresources available to them in the school as adequate are less likely to consider leaving; a one
standard deviation increase in the adequacy of school resources is associated with a 24% decline
in the odds of considering teaching in another school. Other measures of school functioning
such as principal leadership, teacher collegiality, control over school and classroom policies, andschool disorderare not associated with considering leaving the school to teach in another
school. Future modeling approaches will consider the relationship between each of these school
measures and teachers considering of leaving their school by sequentially adding and removingeach measure to the model, one at a time.
One additional pattern that emerges when these school workplace measures are
considered is that the racial/ethnic composition of the school becomes associated with teachers
consideration of leaving the school. Although the racial/ethnic mix of students did not emerge asa statistically significant factor when school structural features alone were considered, there is
evidence that the percentage of Asian students in the school is associated with considering
leaving. A one standard deviation increase in the percentage of Asian students in the school
from, say, 13% to 29%decreases the odds of considering leaving by 35%. This pattern isobserved when school-level measures of academic achievement and disorder are taken into
account, so it does not appear to be attributable to the possibility that schools with high
concentrations of Asian students also tend to be higher-achieving and/or less disorderly.
Finally, Model 5 combines all of the predictors in Model 2 and Model 4 into a singleequation. The results suggest that individual teacher characteristics are more influential in
determining whether New York City middle school teachers considered leaving their school than
are the characteristics of the schools in which they teach. Among the individual teachercharacteristics, experience, teacher race/ethnicity, number of children, teaching a new subject or
a new grade, and time commuting to work all are significant predictors when school-level
characteristics are taken into account at the same time. The same is true for the social-psychological factors of satisfaction with salary and peer encouragement to stay in the school.Moreover, for each of these factors, the magnitude of their effects is virtually the same as in
Model 2, which did not control for school-level factors. This implies that these effects of
individual teacher attributes cannot be explained by how teachers are distributed across middleschools in New York City.
The one instance in which the distribution of teachers across schools might explain the
significance of an individual teacher characteristic is marital status. In Model 2, teachers who
8/3/2019 Paper 2726
18/31
18
were married, separated or widowed were less likely to consider leaving their current schools to
teach in another school than teachers who were never married or who had domestic partners.However, the marital status of teachers differs across schools in the five boroughs, with 38% of
the teachers in Manhattan middle schools reporting that they have never married and do not have
a domestic partner, considerably higher than the percentages in the other boroughs. Taken
separately, teachers in Manhattan schools are more likely to consider leaving for other schoolsthan teachers in schools in the other four boroughs, and teachers who have never married or who
have domestic partners are more likely to consider leaving their current schools. Because of the
association between borough and marital status, the two "cancel each other out" in a statisticalsense.
On the other hand, there is some evidence that some of the effects of schools structural
and workplace characteristics on the odds of considering leaving one's current school might be
due to the distribution of teachers across schools. We have already alluded to the fact thatteachers in Manhattan schools were much more likely to consider leaving their current schools
than teachers in schools in the other four boroughs (Model 4), but the differences among
boroughs are not statistically significant when individual teacher characteristics are taken into
account in Model 5. Part of this may be due to the distribution of teacher marital status acrossboroughs. We also considered the possibility that the differences among boroughs in the
likelihood of considering leaving might vanish due to differences across boroughs in commuting
time. Perhaps a smaller proportion of teachers in Manhattan schools live in Manhattan, due to thecost of living there, and therefore their commute is longer. We have already established that
teachers with longer commutes to their school are much more likely to consider leaving their
school to teach elsewhere than teachers with shorter commutes. Might this explain thedifferences among the boroughs in the odds of considering leaving?
Somewhat surprisingly, commuting time does not account for borough differences in the
probability of considering leaving one's school to teach in another school. To be sure, Staten
Island stands out among the boroughs; 64% of the teachers in Staten Island report a commute of
20 minutes or less, and only 2% of the teachers in Staten Island schools commute more than anhour each way to work. But although Manhattan has the highest proportion of teachers
commuting an hour or more, the percentages are not strikingly different from those in Queens,
the Bronx and Brooklyn. 17% of the teachers in Manhattan middle schools report a commute ofan hour or more, compared to 15% in the Bronx and 13% in Queens and Brooklyn. These
differences, while larger, do not account for the higher probability of considering leaving one's
school observed among teachers in Manhattan schools.
We note one final difference between Model 4, which included both school structuralcharacteristics and school workplace characteristics, and Model 5, which considered individual
teacher characteristics alongside of these school characteristics. Model 4 indicated that teachers
in schools where the resources were judged adequate were less likely to consider leaving theircurrent school to teach in another school. This school-level factor is no longer a statisticallysignificant predictor of considering leaving when individual teacher characteristics are taken into
account. It is possible that the correlation between the adequacy of resources and an influential
teacher-level characteristic might account for this pattern. One candidate is the extent to which ateacher's peers encourage him to remain in his current school. Model 2 indicates that teachers
whose peers think they should stay in their current school are indeed substantially less likely to
consider leaving, and this effect persists once school-level characteristics are taken into account.
8/3/2019 Paper 2726
19/31
19
There is a moderately large association between the amount of peer encouragement and the
adequacy of school resources; a teacher's friends are more likely to encourage her to stay in herschool if the school has more resources than if the school has fewer resources. Thus, the effect of
the adequacy of school resources on a teacher's likelihood of considering leaving her current
middle school may be due to the influence of significant others.
What are the reasons that middle school teachers are considering leaving their schools?
We asked the teachers who had reported considering leaving their school to rate theimportance of 14 different factors in influencing their consideration. Teachers rated each factor
as: not at all important, somewhat important, important, or very important. Exhibit 9 reports the
percentage of teachers indicating that a given factor was very important or important, with thefactors listed in descending order of importance. As the exhibit demonstrates, more than three-
quarters (76%) of all teachers who reported having considered leaving their school indicated that
student discipline problems and/or lack of student motivation was a very important or important
factor. Two-thirds (66%) stated that a lack of support from administrators was very important or
important. More than one-half of all teachers said that wanting to have more influence overschool policies and practices and the quality of their school's facilities were very important or
important considerations (56% and 51%, respectively).
8/3/2019 Paper 2726
20/31
20
Exhibit 9. Importance of Reasons for Considering Leaving Current School for Teachers Considering Leaving
Reason for considering leaving current school
% Not at
All
Important
%
Somewhat
Important
%
Important
% Very
Important
% Very
Important
or
Important
Student discipline problems and/or lack of student motivation 14 11 23 52 76
Lack of support from administrators 17 17 25 41 66
Wanting to have more influence over school policies and practices 20 24 32 24 56The quality of your school's facilities 27 22 31 20 51
Lack of professionalism of teacher colleagues 30 23 27 20 47
Frustration about the amount of testing and test preparation in your school 30 23 25 22 47
Concerns about your job security at this school or in this district due to budget cuts 38 19 19 24 43
Wanting to work at a school closer to your home 37 23 19 21 40
Lack of opportunities for professional development as a teacher 37 24 23 16 39
A change in leadership in the school 48 16 16 19 36
Lack of availability of parking at the school 57 14 12 17 29
Wanting to teach a new subject area or grade 54 19 16 11 27
Wanting to leave teaching altogether 57 16 14 13 27
Personal or family reasons (e.g., spouse's job requires moving, health concerns, etc.) 61 12 14 12 26
8/3/2019 Paper 2726
21/31
21
We sought to understand the individual teacher-level (and, to a lesser extent, school-
level) factors associated with rating a particular factor important. The details of these effects aredisplayed in Exhibit 10. Only the statistically significant effects are reported, with blanks for
those teacher attributes which are not reliably different from zero. (As before, we tested whether
a set of categories collectively differs from zero at the conventional criterion of p
8/3/2019 Paper 2726
22/31
22
Column three reports the relative importance ascribed to the lack of influence a teacherhas over school policies and practices. Three teacher attributes are associated with the
importance of this factor: teacher certification status, commuting time, and New Yorker status.
Teachers from New York view their lack of influence over school policies and practices as more
important to their decision to stay in or leave their current school than do teachers who do nothave the ties to New York that a New York City high school diploma signifies. Those teachers
who entered teaching via the Teaching Opportunity Program (TOP) rate their influence over
school policies and practices as a full point more important than do traditionally-certifiedteachers. It is not obvious what features of the TOP, which is a joint program between CUNY
and the New York City Department of Education, would account for this difference. Nor is it
evident why teachers with a moderately long commute might view influence over school policies
Student
Discipline
Lack of Support
from
Administrators
Wanting
More
Influence
Quality of
School
Facilities
Lack of
Professionalism
Testing/
Test Prep
Job
Security
0 to 1 year experience .762
1 to 3 years experience .445
4 to 5 years experience .149
11 to 15 years experience .254
>15 years experience -.197
Black
Asian
Hispanic
Other ra ce
Gender .162 .225
Living with domestic partner
Married
Divorced
Separated
Widowed
N of children responsible for .130
Teaching new subject
Teaching new grade
NYC public high school graduate .217 .209 .173
NYC Teaching Fellow -.230 .011 -.199 -.346
Teach For America corps member -.457 .091 -.594 -.493
Teaching Opportunity Program alum .150 1.009 .758 -.504
Other alternative certification .150 .178 .083 .086
Holds MA degree
Holds MA + 30 credits
Holds doctoral degree
20 to 40 minute commute -.076 -.027
40 to 60 minute commute -.217 -.229
60 or more minute commute -.340 .012
Table 19a. Predictors of Importance of Reasons for Considering Leaving Current SchoolExhibit 10.
8/3/2019 Paper 2726
23/31
23
and practices as a less important consideration than teachers with a brief commute from home to
school.
The determinants of the importance of the quality of school facilities as a consideration inleaving is reported in column four. New Yorkers view this as a more important factor than do
non-New Yorkers, which might be rooted in their comparison of the qualities of the facilities
they recall from their own experiences as a student in New York City schools and what they seein their schools and classrooms today. Teachers who entered teaching via alternate routes ascribe
a different importance to the quality of a school's facilities in their thoughts about leaving their
school than do traditionally-certified teachers. The quality of school facilities is less important toTeaching Fellows and especially TFA corps members, and more important to TOP teachers, than
this assessment is to the teaching plans of traditionally-certified teachers.
Columns five and six report the determinants of the importance of the professionalism of
a teacher's colleagues and the amount of testing and test preparation at the school, respectively,
as factors influencing considering leaving one's current school. Although these factors are veryimportant to some teachers, and less so to others, the teacher attributes we observe do not
effectively explain why. Female teachers rate the lack of teacher professionalism as slightly more
salient to them than do male teachers, and New Yorkers rate their concerns about testing and testpreparation as a tad more important in their decision-making than do non-New Yorkers.
The final column of Exhibit 10 displays the predictors of the importance of budget-
related job security as a consideration in thinking about leaving one's building. In an era of
declining school budgets and the annual threat of teacher layoffs, it may be that some teachersfeel more vulnerable than others to the potential loss of their positions. Teachers in the earliest
stages of their careers are much more likely to report that this threat is important to them than are
teachers in mid-career. Although teachers in their first year of teaching are less likely than mid-
career teachers to report that they are considering leaving their current schools, those that do
consider leaving are more likely to view job security as an important consideration than are more
experienced teachers, the vast majority of whom have the due process protections of tenure.Teachers with 11 to 15 years of full-time teaching experience also place slightly moreimportance on budgetary threats to their positions than do teachers with six to ten years of
experience.
Independent of career stage, teachers with financial and legal responsibility for children
place more weight on school and district budget issues in considering leaving their schools thando teachers with no children. Each child increases the importance by .13 on the scale of 1 to 4.
This implies that a teacher with responsibilities for three children rates budgetary pressures as
.33 more important in thinking about leaving their school than does a teacher with no children.
Finally, we see additional evidence that teacher certification status influences the mix of
factors that teachers consider when deciding whether to stay in or leave their current school.Teaching Fellows and TFA teachers express less concern about job security due to budget cuts as
a rationale for considering leaving their schools than do teachers who entered teaching through
the traditional route.
8/3/2019 Paper 2726
24/31
24
Do certain New York City middle schools have much higher or lower concentrations ofteachers who are considering leaving their school and/or teaching?
In the final analysis of this component of the study, we shift our focus from the individualto the school as the unit of interest. Although we have considered both the attributes of
individual teachers and of schools, our attention has been directed to understanding why some
teachers consider leaving their schools, and the reasons for leaving that influence theirconsideration. In this section, we examine the variation among schools in the proportion of
teachers who report that they are considering a departure of any kind (i.e., a departure from their
school, from teaching, or from both). Further, we examine whether this variation across schoolsappears to be due to differences among schools in teachers characteristics, school demographics,
or aspects of the school as a workplace.
Across the entire sample of teachers responding to the survey, 39% reported having
considered leaving their current school, leaving teaching, or both. In the average school, the
estimated percentage of teachers considering one of these types of departures from their currentteaching assignment was 43%. We estimated a model which yielded the odds of a teacher
considering any kind of departure for each school. The ICC for this model is .13, indicating that
the school-level average of the predicted percentage of teachers who are considering leavingdoes vary across schools.
The differences among schools are displayed in Exhibit 11, which arrays the estimated
average odds of considering any kind of departure for each of the 125 middle schools in the
study. The lowest average odds, .24, corresponds to an estimated percentage of teachersconsidering some kind of departure of 15%. In contrast, the highest average odds, 3.65,
represents an estimated percentage of 73% of teachers considering some kind of departure. Of
the 125 schools, 24 had an estimated odds of teachers considering a departure that was
significantly below the overall average, and 11 had estimated average odds of teachersconsidering a departure which was significantly above the overall average. As a fraction of the
total number of schools in the study, then, 19% were "stickier" than average, and 9% had moreteachers than average considering departing their school or profession. This is well beyond thefraction of extremely high or extremely low schools that we would expect to observe by chance
alone in a normal distribution of schools, and the differences among schools in the percentages
of teachers considering leaving appear large enough to be of practical importance.
8/3/2019 Paper 2726
25/31
25
Having found that there are differences among schools in the concentration of teachers
considering leaving, we investigated whether we could account for these differences bycontrolling for differences among schools in teachers characteristics, school demographic
characteristics, or various aspects of the school as a workplace. We consider four possibleexplanations for the variation among schools. The first explanation examines the amount of
variation among schools which remains after taking account of individual teachers'
characteristicstheir experience, educational credentials, certification route, race/ethnicity,family status, and so on. This approach reveals the extent to which some middle schools have
teachers with higher predicted rates of considering leaving than others because teachers with
different characteristics are distributed differently across middle schools. The second explanation
adds teacher social-psychological factors to the array of background characteristics. This analysisincorporates teachers' satisfaction with their salaries, the extent to which their peers encourage
them to stay in their current school, their perceived effectiveness in the classroom and theirsocial ties to their students as predictors, and examines the extent to which there continue to be
differences among schools in the average odds of thinking about leaving when these individualteacher attributes are controlled for. This explanation looks at the extent to which teachers with
different individual and social-psychological attributes are distributed differently among middle
schools, and the extent to which that variation alone accounts for variations among schools in theodds of considering leaving.
Exhibit 11.
8/3/2019 Paper 2726
26/31
26
The third and fourth explanations are of a different character, focusing on the attributes of
the schools in which teachers teach rather than on teachers individual characteristics. The thirdexplanation for variation among schools in the average odds of a teacher considering leaving
emphasizes differences among schools in their demographic characteristics, and the extent to
which these differences alone account for the fact that middle schools have different
concentrations of teachers considering leaving. The fourth explanation attends to the variationamong schools in teachers' reports of the school as a workplace, using the same set of school-
level measures discussed in the earlier sections. This approach summarizes the extent to which
differences among schools in how teachers experience the school as a workplace might accountfor differences among schools in the average odds of considering leaving.
The approach here is to examine the variation among schools in the odds of considering
leaving one's school for another school without any attributes of teachers and schools taken into
account (other than a teacher being in one school versus another), and then to see how muchvariation among schools remains when a particular explanation is considered. The shrinkage in
the amount of variation among schools that is observed when teacher or school factors are taken
into account represents the extent to which those factors "explain" the total amount of variation
among schools. This shrinkage can range from 0%, indicating that there is just as much variationamong schools in the odds of considering leaving when a set of factors is taken into accountand hence the factors did not explain any of the variation among schoolsto 100%, indicating
that there is no variation among schools left after a set of factors is consideredand hence thefactors explain all of the variation among schools.
Exhibit 12 reports the results of these analyses. The left panel of the table considers
school-to-school differences in the odds of a teacher considering leaving his or her current
school. As noted earlier, the ICC for this outcome is .15, indicating that most of the variation inconsidering switching schools is among the teachers within a given school, and a relatively small
fraction15%is between schools. The four rows in Exhibit 12 show the relative explanatory
power of the four explanations for why schools differ in their concentrations of teachers
considering moving to other schools. The distribution of teachers' characteristics across schoolsis the least persuasive explanation for school variation in the percentage of teachers considering
switching schools; these factors by themselves account for just 28% of the total variation among
middle schools in the odds of considering leaving ones current school. When the social-psychological characteristics of teacherswhich themselves might be influenced by the
demographic characteristics of the school and the school as a workplaceare also taken into
account, 63% of variation among schools in the odds of considering switching schools which isaccounted for, just under two-thirds of the total variation among schools. (Put differently, there
still are differences among schools in the average odds of considering leaving, but the schools
are bunched much more closely together when teachers' individual attributes and social-
psychological characteristics are taken into account than when they are not considered.)
8/3/2019 Paper 2726
27/31
27
Rows three and four display the extent to which school demographic characteristics and
measures of the school as a workplace, respectively, explain the variation among schools in theodds that a teacher considered leaving his or her current school to teach in another school in the2009-10 school year. Each of these school-level factors by itself accounts for about two-thirds of
the total variation among schools in the average odds of considering switching schools. Overall,
then, we conclude that the sorting of teachers of different backgrounds into different schoolsdoes not provide a powerful explanation of why schools differ from one another in their averageodds that a teacher considered leaving. When teachers' social-psychological attributes also are
taken into account, the combination of individual teacher characteristics and their social-
psychological beliefs are as successful as the structural features of the school or the school as aworkplace in explaining the variation among schools in the propensity to consider leaving one's
current school to teach elsewhere.
The right panel of Exhibit 12 shows the percentage of the variation among schools in the
average odds of considering leaving teaching altogether which is accounted for by the fourdifferent explanations. The ICC for this outcome is .19, indicating that 19% of the variation in
the odds of considering leaving teaching altogether is between schools, and the remaining
variation in the odds of leaving teaching is among the teachers within a given school. The first
row indicates that differences in teachers characteristics are not a powerful explanation for whyschools differ in the average odds of a teacher considering leaving the teaching occupation, as
these characteristics explain only 29% of the total variation among schools in the odds of
thinking about leaving teaching. In fact, the explanatory power of individual teachercharacteristics, as well as the social-psychological factors introduced in row two, is almost
identical to that observed in the left panel of the table.Rows three and four of the right panel of Exhibit 12 report the percentage of the total
variation among schools in the odds of considering leaving teaching altogether which is
accounted for by school-to-school variations in demographic characteristics and in teachers'reports of the school as a workplace, respectively. In both instances, these school characteristics
by themselves explain nearly three-quarters of the variation among schools in the average odds
that a teacher considered leaving the teaching career. We conclude, therefore, that the features ofmiddle schoolsschools demography, as well as how they function as a place for teachers to
ICC
% of Variance inRandom Intercept
Explained ICC
% of Variance inRandom Intercept
Explained
.15 .19
(1) Teacher background variables 28% 29%
(2) Teacher background + social psych factors 63% 64%
(3) School demographics 64% 73%
(4) School as work environment 68% 73%
Table 23. Explanations for Variations among Schools in the Odds that a Teacher Considered Leaving
Thought about Leaving School to
Teach Elsewhere
Thought about Leaving Both
Current School and Leaving
Exhibit 12.
8/3/2019 Paper 2726
28/31
28
workexplain more of the variation among schools in the odds of teachers considerations about
leaving than do the features of individual teachers. This is true even when teachers' social-psychological beliefs, which may be influenced by the features of the school, are taken into
account. Conversely, the features of the school cannot explain why some teachers within a
school are more likely than others to consider leaving; the individual characteristics of teachers
are more important to understanding this kind of variation, and there is a lot more of it than thereis variation from one school to the next.
Discussion and Next Steps
After having completed two of the three components of this study, we have identified
some patterns of middle school teacher turnover that prompt concern. In addition, using originalsurvey data, we have identified some of the teacher and school characteristics that are related to
both previous teachers decisions to stay in or leave their schools and current teachers (2009 -
2010 school year) consideration of whether to leave their school or the teaching profession.
Below, we summarize the notable findings from the study thus far and briefly consider their
potential implications.The main objectives of the first component of our larger study of turnover were to
address foundational questions related to New York City middle school teacher turnover and
establish the context for the subsequent components of our larger study. Despite the studysdesign being more suitable for raising, rather than answering, questions, several key findings
stand out in their own right. First, more than half of the middle school teachers who entered their
schools between 2002 and 2009 left these schools within three years. Among those who left,nearly 60% left the New York City public school system altogether; less than 30% of those who
moved to other schools within the system transferred to Grade 6-8 middle schools. The relatively
small percentage of teachers who secured assignments in other Grade 6-8 schools suggests that
few of the vacancies created by departing teachers were filled by incoming teachers with recent
experience in similar schools.
These rates of turnover are likely to make it challenging for middle school principals, and
for the teachers who remain in their schools, to establish organizational norms and a shared
vision for their schools teaching and learning environment. Turnover of this nature may requireschools administrators to divert resources away from professional development in order to orient
and support teachers who are new to their buildings, new to the New York City schools, or new
to teaching. In addition, turnover may compromise the continuity of the relationships betweenmiddle school teachers and administrators, students, parents, and the staff at organizations that
partner with middle schools. If middle schools are unstable and impersonal, students may find it
even more challenging to manage the transitions into, through, and out of the middle grades a
time period characterized by numerous social and emotional developments.
12
While descriptiveanalyses of this nature do not support causal inferences, we find associations between turnover
and measurable and malleable characteristics of middle schools, such as school size and the
aspects of schools environment measured by the School Survey (e.g., school safety). Thesefindings suggest the possibility that practitioners and policymakers may be able to influence
turnover by influencing these characteristics of middle schools.
The second component of the study uses data from our original survey of over 4,000
middle school teachers to investigate the percentage of teachers who are considering leaving
8/3/2019 Paper 2726
29/31
29
their school or teaching, the reasons behind these considerations, and the extent to which teacher
and school characteristics help us predict: 1) which teachers are considering leaving their schooland/or teaching, and 2) which schools have higher and lower concentrations of teachers who are
considering leaving their school or the teaching the profession. The results from these analyses
reveal that 39% of middle school teachers were considering some type of departure (i.e., from
either their school, teaching, or both) during the 2009-2010 school year. Further, of the teachersconsidering leaving their school, 70% anticipated doing so within two years. Of the teachers
considering leaving the profession, 41% expected to have done so within the same timeframe.
Taken together, the historic observed rates of turnover among new-to-school teachers (fromcomponent 1) and the anticipated rates of turnover among all current teachers (from component
two) suggest that teacher turnover is likely to be an enduring characteristics of New York City
middle schools in the years to come.
It stands to reason that some portion of policymakers and building administrators mayseek to identify interventions and reforms that might alter these rates and patterns of turnover.
Data from our original survey of middle school teachers presents a complicated picture about
how this might work. Give the important role that teachers characteristics play in predicting
whether teachers are considering leaving their school, it seems that a promising approach mightentail developing school- and system-wide interventions that are responsive to the needs of
different types of teachers. For instance, incentives and reforms that reduce teachers commuting
times may be a promising approach for reducing turnover. Similarly, given how much higher the
odds of considering leaving ones school are among teachers who reported teaching a new
subject area, principals might be well advised to make every attempt possible to resolve staffing
challenges in ways that dont require assigning teachers to new subject areas.
The dearth of associations between measures of the school as a workplace and teachersconsideration of whether to leave their school are puzzling and suggest that interventions or
reforms targeted at the malleable aspects of schools may not, in themselves, overcome the role
that teachers individual characteristics play in influencing their decisions about whether to
remain in schools. Contrary to this general trend, however, school disorder (as measures by the
percentage of a schools students who received suspensions) does play an important role in theirconsideration of whether to stay in their school. Similarly, the extent to which teachers have a
supportive network of peers in their school is related to their consideration of leaving theirschool, suggesting another potential avenue for interventions and reforms to explore.
The remaining case study component of this research project will help us investigate
other important questions related to middle school teacher turnover. Dr. Jennifer Goldstein
(Baruch College) is conducting case studies of four middle schoolstwo schools where rates ofturnover have been historically high, and two where rates have been low that serve similar,
high-need student populations. These case studies will help us gain a better understanding of theextent to which teachers and school administrators perceive turnover as a cause or consequence
of their schools operational functioning. Further the case studies will examine schools
strategies for recruiting, developing and retaining effective teachers and for limiting turnoversdamaging consequences.
The Research Alliance also intends to extend the analyses presented here in order to offer
further insight into the potential causes and consequences of the rates and patterns of turnover. In
8/3/2019 Paper 2726
30/31
30
particular, we hope to examine the extent to which patterns of turnover are the result of teachers
voluntary decisions, as opposed to involuntarily transactions initiated by administrators. Utilizing
various sources of data about teachers effectiveness, we also intend to explore why effectivemiddle school teachers leave their schools and whether particular incentives or changes to their
working conditions appear to keep them in their schools. Using the survey data, we will examine
the relationship between measures of the school as a workplace and actual observed rates ofteacher turnover.
To inform our broader research on the middle grades, we plan on gathering more and
better evidence about whether middle school teachers are prepared, and whether their schools are
organized, to support students academic, social and emotional development during this critical
phase of students schooling. This effort will entail conducting rigorous investigations on topicsthat have not been adequately explored, such as whether particular school grade configurations
(e.g. 6-8, K-8, or 6-12 schools) are more suitable for sustaining and stimulating students growthduring the middle grades. Collectively, the findings from the three components of this study of
teacher turnover and from our future investigations will provide policymakers, practitioners, and
researchers with evidence that can inform their efforts to improve middle schools and middle
grades educationtwo features of the New York City public school system that many agree arevital but imperiled.
1Boyd, D., Grossman, P., Lankford, H., Loeb, S., & Wyckoff, J. (2007, September 18). Who leaves? Teacher
attrition and student achievement. Paper presented at the Inaugural Conference of the Research Alliance for New
York City Schools.
2New York City Council. (2007).Report of the New York City Council Middle School Task Force. Retrieved on
December 17, 2010 from:http://council.nyc.gov/html/pr/report/middle_school_task_force_2007.pdf
3Balfanz, R. (2009). Putting middle grades students on the graduation path: A policy and practice brief. Retrieved
on December 17, 2010 from:
http://www2.kapoleims.k12.hi.us/campuslife/depts/electives/dance/Putting%20Middle%20Grades%20Studesnts%20on%20the%20Graduation%20Path.%20%20A%20Policy%20and%20Practice%20Brief.%20%202009.pdf
Balfanz, R., Herzog, L., & Mac Iver, D. J. (2007). Preventing student disengagement and keeping students on the
graduation path in the urban middle-grades schools: Early identification and effective interventions.Educational
Psychologist, 42(4), 223-235.
Roderick, M. (1994). Grade retention and school dropout: Investigating the association.American Educational
Research Journal, 31(4), 729-759.
4Summarized in Johnson, S. M., Berg, J. H., & Donaldson, M. L. (2005). Who stays in teaching and why: A review
of the literature on teacher retention. Cambridge, MA: Harvard Graduate School of Education
Milanowski, A. T., & Odden, A. R. (2007). A new approach to the cost of teacher turnover. School Finance
Redesign Project Working Paper 13. Retrieved on December 21, 2010 from:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.124.2095&rep=rep1&type=pdf
National Commission on Teaching and Americas Future [NCTAF], (2007). The cost of teacher turnover in five
school districts: A pilot study. Retrieved on December 17, 2010 from:
http://www.nctaf.org/resources/demonstration_projects/turnover/documents/CTTFullReportfinal.pdf
5Neild, R. C., Useem, E., Travers, E. F., & Lesnick, J. (2003). Once & for all: Placing a highly qualified teacher in
every Philadelphia classroom. Philadelphia, PA: Research for Action.
http://council.nyc.gov/html/pr/report/middle_school_task_force_2007.pdfhttp://council.nyc.gov/html/pr/report/middle_school_task_force_2007.pdfhttp://council.nyc.gov/html/pr/report/middle_school_task_force_2007.pdfhttp://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.124.2095&rep=rep1&type=pdfhttp://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.124.2095&rep=rep1&type=pdfhttp://www.nctaf.org/resources/demonstration_projects/turnover/documents/CTTFullReportfinal.pdfhttp://www.nctaf.org/resources/demonstration_projects/turnover/documents/CTTFullReportfinal.pdfhttp://www.nctaf.org/resources/demonstration_projects/turnover/documents/CTTFullReportfinal.pdfhttp://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.124.2095&rep=rep1&type=pdfhttp://council.nyc.gov/html/pr/report/middle_school_task_force_2007.pdf8/3/2019 Paper 2726
31/31
6Boyd, D., Grossman, P., Lankford, H., Loeb, S., & Wyckoff, J. (2007, September 18). Who leaves? Teacher
attrition and student achievement. Paper presented at the Inaugural Conference of the Research Alliance for New
York City Schools.
7
Boyd, D., Grossman, P., Lankford, H., Loeb, S., & Wyckoff, J. (2007, September 18). Who leaves? Teacherattrition and student achievement. Paper presented at the Inaugural Conference of the Research Alliance for New
York City Schools.
Goldhaber, D., Gross, B., & Player, D. (2009). Teacher career paths, teacher quality, and persistance in the
classroom: Are schools keeping their best? National Center for Analysis of Longitudinal Data in Education
Research Working Paper, Working Paper 29. August, 2009.
8It should be noted these teachers were not necessarily first-year teachers, nor new to the New York City public
school system.
9Ingersoll, R. M. (2001). Teacher turnover and teacher shortages: An organizational analysis.American Educational
Research Journal, 38(3), 499-534.
10 Our analyses of the relationship between turnover and the characteristics of teachers and schools control for fourfactors that we hypothesized might affect turnover: 1) the school year during which teachers first entered their
NYC middle school; 2) whether a school experienced one or several episodes of principal turnover during the
period of observation; 3) a time-varying measure ofthe annual change in a schools student enrollment; and 4) a
schools NYC borough location. In the full report, we refer to these variables as our baseline covariates. Our
methodology controls for their effects in order to examine the relationships between turnover and the relevant
teacher and school characteristics in our datasets.
11Dee, T. (2005). A Teacher like me: Does race, ethnicity, or gender matter?American Economic Review, 95(2),
158-165.
Consortium on Chicago School Research [CCSR] (2009). Allensworth, E., Ponisciak, S., & Mazzeo, C. The
schools teachers leave: Teacher mobility in Chicago Public Schools. Chicago, IL: Consortium on Chicago School
Research.
12Seidman, E., Aber, J. L., & French, S. E. (2004). The organization of schooling and adolescent development. In
K. Maton, C. Schellenbach, B. Leadbeater, & A. Solarz (Eds.),Investing in children, youth, families, and
communities: Strengths-based researchand policy (pp. 233250). Washington, DC: American Psychological
Association.
Eccles, J., Midgley, C., & Adler, T. F. (1984). Grade-related changes in the school environment: Effects on
achievement motivation. In J. G. Nicholls (Ed.) The development of achievement motivation (pp. 283-331).
Greenwich, CT: JAI Press.
Na