View
12
Download
0
Category
Preview:
Citation preview
1
An Analysis of Teacher Turnover in Michigan
Kolby Gadd Daniel Hubbard
Brian Jacob University of Michigan
February 27, 2016 WORKING PAPER – DO NOT CITE
ABSTRACT: Previous research has established that students learn best from teachers with at least a few years of experience, and that teacher quality is often distributed inequitably in a way that further disadvantages students of color and lower-income students. We use administrative data from Michigan's public school system to determine which factors are associated with teachers leaving their first jobs, and to look at where they go if they choose to change jobs. We find that teachers are more likely to leave when they teach more black students or more low-income students; teachers are less likely to leave if they attended college in Michigan or have master's degrees. Turnover patterns over the business cycle are also fascinating; teachers hired during a recession are less likely to leave, even as the labor market improves during their tenure, indicating that districts may be able to screen candidates effectively when the applicant pool is large.
Special thanks to Sue Dynarski, Katherine Michelmore, Matt Ronfeldt, Kevin Stange, Jeff Smith, and seminar participants at the Michigan Causal Inference in Education Research Seminar (CIERS). Thanks to the Institute of Education Sciences, U.S. Department of Education for providing support through Grant R305E100008 to the University of Michigan. Thanks to our partners at the Michigan Department of Education (MDE) and Michigan's Center for Educational Performance and Information (CEPI). This research result used data structured and maintained by the Michigan Consortium for Education Research (MCER). MCER data are modified for analysis purposes using rules governed by MCER and are not identical to those data collected and maintained by MDE and CEPI. Results, information and opinions are the authors' and are not endorsed by or reflect the views or positions of MDE or CEPI.
2
Regardless of the quality of the other inputs in a student's education, teacher quality
remains one of the most crucial inputs in determining how much a student learns in school. If
teacher quality were distributed randomly throughout the education system, differences in
teacher quality would even out over the course of students' careers, making teacher quality less
of a concern in policymakers' efforts to ensure high-quality education for all. In reality, though,
teacher quality is hardly distributed evenly across schools. Low-income students, African-
American students, Hispanic students, students in urban schools, and students with poor
academic preparation all have their socioeconomic disadvantages compounded by lower-quality
teachers than their white and wealthier peers have (Goldhaber, Lavery, & Theobald 2015;
Clotfelter, Ladd, & Vigdor 2006; Lankford, Loeb, & Wyckoff 2002). This stands to widen the
racial and economic achievement gap, despite all the attention supposedly placed on reducing it.
We argue that teacher turnover is at the heart of this story. If teachers have limited ability
to choose their first jobs, they continue to search for a job that they prefer until they find a better
match. Sometimes this better match involves leaving the public school system altogether. Given
that rookie teachers perform notably worse than teachers with several years of experience
(Murnane & Steele 2007, Clotfelter et al. 2006), if certain schools are systematically losing
experienced teachers, students at these schools will see their academic performance drop
accordingly. Furthermore, if teachers are systematically leaving schools with more
disadvantaged students, this would widen the achievement gap even further. As a majority of
teachers are white and from the suburbs, and teachers tend to prefer to teach closer to home when
possible (Boyd, Lankford, Loeb, & Wyckoff 2005), this result would appear very plausible.
A thorough study of teacher turnover would also be instructive in explaining the behavior
of teacher labor markets; hiring may be drastically different at various points in the business
3
cycle. On the labor supply side, during economic booms, there are more jobs available in the
private sector, and thus teachers' outside option will generally be higher; candidates who still
choose to teach will be either very committed to teaching or "stuck" teaching because their
outside options are particularly bad. During recessions, the outside options will be fewer and less
lucrative, and more workers will pursue teaching positions. On the demand side, during
recessions, there will more likely be more candidates than available jobs, and therefore hiring
schools can be more selective and choose only the best prospective teachers (if they can identify
quality or characteristics associated with it); during booms, schools may not be able to be so
picky and thus might have to hire whatever candidates they can get. If teachers hired during
recessions are more likely to stay in their initial jobs even after the recessions end, this indicates
that schools successfully identify which teachers are the best fits for the open positions.
We use a database of teachers in public and charter K-12 schools in Michigan who start
teaching between 2003-04 and 2008-09 to answer four main questions. First: what factors affect
the probability that a teacher leaves his or her first job within five years of being hired? Second:
what factors affect teachers' choice of a second job, including jobs outside of the Michigan
public education system, new teaching positions in the same school district, and moving across
school districts? Third: how does teacher mobility affect the distribution of teachers across
schools, particularly across socioeconomic levels? Fourth: how do teacher labor markets respond
to events in the business cycle?
Many of our findings will come as no surprise to readers who are well-acquainted with
the existing literature on teacher turnover. Teachers are more likely to leave schools with more
economically disadvantaged students. White teachers are much more likely to leave schools with
larger fractions of African-American students. Black teachers are more likely to leave rural
4
schools, while white teachers are more likely to leave urban schools; given the well-established
"draw of home" results pioneered by Boyd et al. (2005) and Michigan's long history of
residential segregation, it is likely that much of this particular turnover is driven by teachers of
all races looking to move closer to where they grew up, although we do not want to discount
prejudice as a factor in the teachers' decision-making. We do not observe teachers' home town,
although we do observe where they attended college and Michigan-trained teachers tend to be
more likely to stay in their first jobs than teachers who attended college outside of the state.
The business cycle results are fascinating. We find that teachers hired in 2007-08 and
2008-09, when Michigan was in the depths of a recession, are about 5 percentage points more
likely to remain in their first job for five years than teachers hired in 2003-04 through 2005-06,
when Michigan's economy was much healthier, and teachers hired in 2006-07 are about 2.5
percentage points more likely to stay. These recession-era hires are less likely to leave their first
school for any reason, but particularly less likely to leave the Michigan public school teaching
force completely. Although we cannot rule out alternative explanations, this may indicate that
schools are able to screen candidates effectively when there are many applicants for few spots,
and they can successfully find the best fit. Many studies have found that measurable qualities
such as grades, test scores, and college selectivity have little to no power in predicting teacher
performance (see, for instance, Clotfelter et al. 2006), but it appears that schools may be able to
observe some less-quantifiable characteristics that help them to see which teachers are the best
fits for the job.
These results have important implications for practitioners and policy makers. When
hiring teachers, principals and districts will want to take care to find teachers who are less likely
to leave their first job quickly; these teachers include teachers with master's degrees, teachers
5
who attended college nearby, and teachers who are demographically similar to their students.1
However, it seems that when schools have a large number of applicants, they are fairly
successful at locating teachers who are likely to stay for longer periods of time, even as the local
economy improves and they have more outside options. Perhaps the best policy would be simply
to provide incentives that make teaching in public schools more attractive, and then to trust the
schools to pick the best candidates from the larger applicant pool that would result.
Literature Review
The distribution of teachers among schools is a concern inasmuch as it does not support
equitable education opportunities for all students. Goldhaber et al. (2015) find that
disadvantaged students are exposed to lower quality teachers at a higher rate than their
advantaged peers. These authors find that the district and school level sorting rather than
classroom level sorting within schools accounts for most of the variation in the distribution of
teacher quality among students, though this result may not hold in other contexts. In this study,
we examine how teacher placement and mobility affects the distribution of teachers in Michigan
public schools.
Many factors influence where teachers accept their first full-time job. School
characteristics and location seem to have the greatest influence on the distribution of teachers
among students upon entry into teaching. Researchers have found that more highly qualified
teachers tend to accept their first teaching job in schools with higher student achievement, fewer
minority students, and fewer students who are eligible for free and reduced price lunch
(Lankford et al. 2002; Boyd et al. 2005). These studies emphasize the challenge this presents to
1 There is also evidence that less-advantaged students learn better from teachers who are demographically similar and share similar life experiences; see, for instance, Dee (2004).
6
urban schools. As urban schools have higher concentrations of low-achieving, minority, low-
income students than suburban schools, more highly qualified new teachers tend to not accept
jobs in urban schools. Teacher sorting along these school characteristics, however, may be an
artifact of geography. Boyd, et al., argue that teachers prefer to work near where they grew up.
Most new teachers did not grow up in an urban area, so they do not accept jobs in urban areas.
Regardless of the reason teachers choose one school over another, their decision upon entry into
teaching seems to perpetuate inequities in the distribution of teachers among students described
previously (Goldhaber et al.2015).
Factors similar to those that influence the initial placement of teachers also influence
mobility. Factors such as race, sex, academic achievement, and subject area taught as well as
location are associated with teacher attrition (Guarino, Santibanez, & Daley, 2006; Podgursky,
Monroe, & Watson, 2004; Boyd et al. 2005). Moreover, this attrition has heterogeneous effects
on schools. Specifically, schools with high portions of low-income, minority, and low-achieving
students experience higher rates of teacher attrition.
Teacher pay is one mechanism for attrition that has received a significant amount of
attention in literature. Some researchers find that pay can attract experienced teachers to specific
schools or induce them to persist in their job (Stinebrickner, 2001; Hoxby & Leigh 2004;
Podgursky et al. 2004; Figlio, 2002). Other researchers, however, have found that pay has no
effect on teacher mobility. Springer, Swain, and Rodriguez (in press) take advantage of an
incentive program in the state of Tennessee to investigate whether or not a bonus of $5,000 helps
disadvantaged schools retain successful teachers. They find that teachers who were eligible for
the bonus were more no more likely to stay at the school where they taught. Whether or not
these previous studies have found an overall effect of pay on teacher mobility, many have found
7
that the effect of pay is heterogeneous. Most evidence suggests that teachers in tested subject
areas such as math are more responsive to changes in pay than their counterparts in non-tested
subject areas.
Regardless of the true effect of pay on teacher mobility, school characteristics have much
greater bearing than pay on teacher mobility. Hanuschek and Rivkin (2007) analyze teachers’
moves within and outside the school district they originally worked, and find little evidence to
suggest that teachers considered pay when they chose to move to different school either inside or
outside of their original school district. Rather, the demographics of the school had a significant
effect on teachers’ decisions to leave their original school. Indeed, Goldhaber, Gross, and Player
(2011) find that the most effective teachers have lower odds of transfers both inside and outside
the school district and lower odds of leaving teaching than their less successful counterparts, yet
successful teachers leave disadvantaged school at a higher rate than they leave advantaged
schools.
Data
To analyze teacher mobility, we used administrative data from 2003 to 2014 for all
Michigan public schools. Specifically, we used personnel data merged with school-level data for
the location where each employee worked in each year. This dataset made it possible to analyze
the characteristics of schools and teachers associated with mobility in Michigan public schools.
Our analytic sample includes all teachers hired into Michigan public schools between fall
of 2003 and fall of 2008. We restricted this sample to those who were initially hired as full-time
teachers. We defined full-time teachers as those with at least 1 FTE2 in teaching assignments for
2 "FTE" stands for "full time equivalent"; a teacher with an FTE of 1 teaches the same amount as a full-time teacher, while a teacher with an FTE of 0.5 teaches half as many hours as a full-time teacher.
8
the school year. This restriction implies that those who began working in Michigan public
schools as part-time teachers or in any assignment other than teaching are not included in the
analysis. We further restricted our sample to teachers who accepted their first job at a school that
remained open for five years thereafter. By doing so, our analysis of teacher mobility is
restricted to mobility by choice—either on the part of the teacher or the school—rather than by
force.
The main dataset was a panel of the first five years after the date of hire for each teacher
in the analytic sample. To carry out our analysis, we created a balanced data set at the teacher-
year level. We assigned teachers who worked in multiple schools during a single year to the
school where they worked most often. In cases where teachers split their time exactly evenly
between two schools, we broke ties randomly. For teachers who did not work in the Michigan
public school system in a particular year, we created an indicator to note their absence.
For analyses of factors associated with where teachers accept their second full-time
teaching job, we limited our sample to teachers who accepted another job within five years of
their initial date of hire. Of course, some teachers accepted more than two jobs within this time
period. For such teachers, we ignored their jobs subsequent to the second job. Thus, our
analysis pertains to where teachers go when they leave their first full-time job.
In addition to the raw data, we derived a variety of teacher characteristics from the
original data. First, we categorized colleges as either highly selective or not based on Barron’s
measure of college selectivity. We considered a college to be highly selective if the school had a
selectivity rating of 3 or lower on Barron’s 6-point scale. Second, we considered any degree
beyond a bachelor’s as an advanced degree. Third, we determined whether an individual was an
elementary or secondary teacher. Because of variation in grade configuration among schools,
9
building code was unreliable for making this determination. In place of the building code, we
used the lowest and highest grades taught to determine at what level a teacher works. If the
lowest and highest grades taught by an individual fell between 9th and 12th inclusive, we
considered the individual to be a secondary teacher. Otherwise, we considered the individual to
be an elementary teacher.
Portions of our analysis pertain to cohorts of teachers. A cohort consisted of all
individuals hired as full-time teachers during a particular school year. The year variable refers to
the spring of the school year, so the 2007 cohort refers to the group of teachers hired to work
full-time in Michigan public school for the 2006 – 2007 school year.
To analyze teacher mobility outcomes we created indicators to specify teachers’
employment status in Michigan public schools in each year for five years after their initial date
of hire. The primary outcome of interest was whether or not teachers were in the same job at the
same school where they were originally hired. Other outcomes included whether teachers went
on to work in a non-teaching or part-time position, where teachers worked with respect to the
first school and district they worked in, the poverty status of schools where teachers worked in
subsequent jobs, and whether or not teachers worked in Michigan public schools.
Our purpose for studying teacher mobility was motivated in part by the distribution of
teachers among public schools in Michigan. Table 1 shows the distribution of teachers in
Michigan public schools in the 2013 – 2014 school year by poverty status. Throughout our
analyses we use percent of students in the school who are eligible for free or reduced price lunch
(FRL) as a measure of the poverty level in a school. Poverty quartiles in Table 1 are defined by
a teacher weighted measure of school poverty.
10
A few characteristics of this distribution are especially important to notice in order to
motivate our work on teacher mobility. First, teacher race was unevenly distributed across
Michigan public schools in the 2013 – 2014 school year. In particular, black teachers
predominately teach in schools with high FRL. Of the quarter of teachers who teach in the
poorest schools, 19.7% of them are black. In contrast, only 1.4% of the quarter of teachers who
teach in the least poor school are black. Second, beginning teachers were also disproportionately
concentrated among teachers who teach in the poorest schools. Among the quarter of teachers
who work in the poorest schools, 14.2% have 3 or fewer years of experience. Among the quarter
of teachers who work in the least poor schools, however, only 6.7% have 1 – 3 years of
experience. Given such inequalities, we explored the extent to which the initial placement of
teachers and their early-career mobility might shape the distribution of teachers in Michigan
public schools.
Initial placement of teachers between fall of 2003 and fall of 2008, as represented in
Table 2, resembles the distribution of all Michigan public school teachers in the 2013 – 2014
school year. Specifically, disparities in terms of race, degree attainment, and selectivity of
undergraduate institution between teachers who work in schools with high FRL compared to
teachers who work in schools with lower FRL appear in the initial placement of teachers in
Michigan public schools. Given how closely disparities in Table 2 match those in Table 1, the
effect of teacher mobility on Michigan public schools is mostly limited to problems that are a
result of churn. Still, these effects can further disadvantage schools that are already
disadvantaged by the distribution of teachers among schools.
Methodology
11
There are numerous factors that may contribute to teachers' mobility decisions. Some of
these are impossible to observe directly with school administrative data (for instance, a teacher's
spouse may find a job in another state, causing the teacher to move to that state and thus leave
the Michigan public school system), but many factors such as teacher age and school size are
observable. Furthermore, if teachers generally seek to leave schools with difficult working
conditions and struggling students, we can observe this through variables such as the fraction of
students eligible for free or reduced-price lunch or the school's relative performance on state
standardized tests. Combining all these factors, we can get a fuller picture of what drives
teachers to leave their first job, and even what drives them to make particular types of moves.
The teachers who leave their first job to take a new job in a high-poverty school may be very
different from the teachers who leave the teaching profession after their first job.
We express a teacher’s mobility decision, 𝑌𝑌𝑖𝑖𝑖𝑖𝑖𝑖, as follows:
𝑌𝑌𝑖𝑖𝑖𝑖𝑖𝑖 = 𝑓𝑓(𝐷𝐷𝑖𝑖𝑖𝑖 ,𝑃𝑃𝑖𝑖,𝑍𝑍𝑖𝑖𝑖𝑖 ,𝑋𝑋𝑖𝑖,𝐶𝐶𝑖𝑖).
Factors contributing to the mobility decision of teacher i, who accepted a full time
teaching position at school s, after t years since the initial date of hire, are grouped into five
broad categories. The first set of factors, Dst, measures the demographics of the student body.
Some teachers prefer not to teach poorer or minority students; even teachers who have no
prejudices against the students themselves may find schools with many poor students and
students of color to be particularly difficult places to teach, as these schools often have few
resources, low budgets, and low pay for teachers (Hanushek & Rivkin 2007). Additionally,
teachers may find students with unique needs, such as English language learners and special
education students, to be more challenging to teach and may seek a less difficult environment.
12
The second set of factors, Pst, is the academic performance of students at school s, which
we measure as of the time that the teacher is hired. The Michigan data do not match teachers to
individual students in most years of our sample period, so for the purposes of this study, we
measure this at the school-by-year level. Teachers may be frustrated by teaching low-performing
students and may seek to leave low-performing schools. Additionally, low-performing schools
frequently have administrative turmoil and other concerns that may make a teacher want to
leave.
Third, Zst represents characteristics of the school. Teachers in charter or vocational
schools may face different challenges than their counterparts in traditional public schools, or they
may respond differently to different circumstances. Teachers in urban or suburban areas, in
which other schools with job openings are more numerous, may feel more comfortable leaving
their jobs than teachers in rural areas, who might need to move in order to take a new job. The
local unemployment rate may also affect teachers' decision making; if unemployment is high,
teachers may not want to leave teaching and face the open labor market.
The fourth set of factors, Xi, is the teacher's characteristics. Younger teachers may feel
less established in their communities and more comfortable leaving. Teachers in certain fields
that are in high demand, such as special education or high-school math, may have more
opportunities to leave than teachers in other fields. Teachers who attended college in Michigan
may feel more loyalty to their communities, while teachers who attended selective universities or
have master's degrees are likely to have more job opportunities, both in and out of teaching.
Finally, teachers who entered the profession in cohort Ci face different labor markets and
opportunities. Our sample includes teachers who began teaching in 2003-04, when the labor
market in Michigan was robust; the sample also includes teachers who started in 2008-09—at the
13
depth of the U.S. recession and the automotive crisis which crippled Michigan's economy. As the
economy and labor market struggled, teachers had fewer opportunities to find employment both
in and out of the profession. At the same time, struggling districts may have laid off teachers in
the leanest years.
Our primary empirical specification is a linear probability model in which we express the
outcome, an indicator for whether a teacher remains in his or her first job, on a wide variety of
covariates. We measure the outcome during a teacher's fifth year since starting to teach in
Michigan public schools in most specifications, although we also have some specifications in
which we measure it as of his or her second year. Teachers whose first schools closed less than
five years after the teacher was hired are removed from the sample; we aim to study teacher
turnover, not school closing, and the repercussions of each are somewhat different. All school
and student characteristics are measured in the teacher's first school at the time that he or she was
hired, unless explicitly labeled otherwise. We also include missing indicators for any variables
that are ever missing in our data set. Standard errors are clustered at the school level.
The interpretations of most of the covariates are fairly straightforward. The most
interesting coefficients shed light on the labor markets in which these teachers are participating.
For instance, teachers with advanced degrees or bachelor's degrees from selective colleges are
likely to be highly sought-after candidates for other jobs, both in and outside of the Michigan
public school system. If these teachers are more likely to stay in their first job despite these
prevalent outside options, it demonstrates that these teachers are either particularly dedicated to
teaching or they can choose a first job that is a particularly good fit for them. The story behind
the coefficients on the later cohort dummies is similar. Michigan's economy was in a deep slump
in 2007-2009. Because private-sector jobs were few and far between, any available jobs were
14
likely to draw more applicants than usual, and not all applicants for teaching jobs were
necessarily highly committed to teaching. A negative coefficient on the later cohort dummies
would indicate that many people leapt at the first job available with little intent to stay at it, or
even to stay in the teaching sector; they moved on to other jobs as soon as the economy
improved and these jobs became available. Conversely, a positive coefficient on these dummies
would indicate that schools could screen candidates effectively, choosing the best fits from the
large applicant pool and getting teachers who will continue teaching at their schools for years.
We then expand on the linear probability model results by including a multinomial logit
model, which can answer multiple questions at once: which teachers leave, and among the
teachers that leave, which ones go where? This multinomial logit has four possible outcomes,
measured against the base case of staying in one's first job for five full years: teachers may take a
full-time job in another district, they may take a new full-time job in their current district, they
may stay in the public school system but not as a full-time teacher (they may teach part-time or
move into an administrative or coaching position), or they may leave Michigan's public school
system entirely. Each teacher's first move is the one that is recorded in the outcome variable; for
instance, if Ms. Brown moves from Detroit to a full-time teaching job in Lansing in year 2, then
leaves the public school system in year 4, she is coded as taking a full-time job in another
district.
This specification will help us answer several questions about the data. For instance, if
teachers who have large fractions of poor and African-American students are leaving for jobs in
other districts or outside of the public school system, this could lend credence to the "draw of
home" theory from Boyd et al. (2005), as well as to arguments about substandard working
conditions or teacher prejudice; if these teachers are moving within their districts rather than
15
outside of them, however, this may mean simply that these districts are more frequently
reallocating their teachers across their schools. We can also gain further insights about the role of
teachers' outside options. For instance, teachers with a background in science or math are likely
to have lucrative outside options in engineering or health; if these teachers are more likely to
leave the public school system, these outside options are important. On the other hand, during the
sample period, budget cuts put some non-academic subjects such as art and music in peril, and
these teachers have fewer lucrative outside options. If these teachers are more likely to leave the
public school system, this is probably due to the systematic elimination of teaching positions in
art and music.
In order to understand the heterogeneous effects across teacher types or school types, we
run our main specification separately for a variety of different subgroups. This will allow us to
understand how some of the covariates interact with each other, such as whether teachers are still
more likely to leave schools with more black students even if the student population is high-
achieving or well-off. If white teachers are leaving urban schools and schools with more black
students, while black teachers are leaving rural schools but not urban schools and are unaffected
by the size of the school's African-American population, this would be another point in favor of
the "draw of home" theory.
We also run another multinomial logit specification (as well as a separate OLS
regression) in which we separate economically disadvantaged schools from non-disadvantaged
schools. There is an extensive literature about the disparity in teacher quality between poorer and
wealthier schools (Goldhaber et al. 2015, Clotfelter et al. 2006), so by treating movement to an
16
economically disadvantaged school3 as a separate outcome from movement to a wealthier
school, we can determine which types of teachers are willing to leave their first job and take a
job in a poor school. Under the practice referred to by Bridges (1990) as "the dance of the
lemons", low-performing or underqualified teachers move from one poor school to another, often
within the same district, because they cannot easily be fired. If this pattern holds in Michigan,
then teachers who do not have master's degrees or are not highly qualified under the standards of
No Child Left Behind will be more likely to move to poor schools or leave teaching than to move
to non-poor schools or stay put. Similarly, teachers coming from schools with many low-income
or low-performing students will be more likely to move to poor schools or leave teaching. On the
other hand, teachers may also move to seek better working conditions and students who they
deem "easier" to teach, and this would induce teachers in economically disadvantaged schools to
try to leave for wealthier districts.
Finally, we run a handful of interaction specifications to continue to investigate this
point. We interact each of a few measures of teacher qualification (a master's degree, a degree
from a selective undergraduate college, "highly qualified" status, and whether a teacher has all
three of those things at once) with the school disadvantage indicator and include the interaction
term in the otherwise-standard OLS specification. Further extensions also include a triple
interaction term, in which the existing interaction term is interacted with an indicator for whether
a teacher has an assignment in math or science. This will allow us to see whether poor schools
are disproportionately losing their most qualified teachers, and whether skilled math and science
teachers are disproportionately leaving poor schools compared to their skilled peers in other
3 We define a school as "economically disadvantaged" if at least 30% of its students are eligible for free or reduced-price lunch in the first year that we observe that school in the data. This allows for a definition of economic disadvantage that is not time-dependent, which is important given the business cycle fluctuations during the sample period.
17
subjects. Considering the increased policy emphasis on improving disadvantaged students' access
to high-quality education in math and science, these results may be of particular interest to
policymakers.
We attempt to control for as many factors as possible, but some variables remain
unobserved. For instance, teachers may relocate in order to accommodate a spouse's new job, but
we do not observe marital status in the data; as older people are more likely to be married, this
may bias the coefficient on age downward in the main specification. Similarly, we do not
observe distance to home directly; therefore, white teachers who are moving away from urban
black schools may not be reacting specifically to the racial composition of the school or the
urban environment, but instead may be looking to move closer to where they were raised, and
this would bias the coefficients on the race and urbanicity variables downward in the main
specification for these teachers.
We also do not observe whether teacher turnover is driven by the supply side or the
demand side; that is, whether teachers are leaving of their own accord or being pressured to leave
by the school. Some factors may have similar effects on both the supply side and the demand
side; for instance, schools with more poor students are frequently in financial challenges and are
looking to cut or reallocate teachers, but teachers may also leave these schools voluntarily due to
the poor working conditions. However, on other factors, such as teacher qualifications, the
supply and demand sides pull in opposite directions. Schools will want to keep their most
qualified teachers if at all possible, but the most qualified teachers are also the ones who can get
new jobs most easily, so they have more opportunities to leave even when their employers want
them to stay. This will bring the net effect of these factors closer to zero.
Findings
18
We present statistics related to teacher mobility from three perspectives. First, Table 3
shows mobility outcomes for all teachers in our sample as they progress through the first five
years since date of hire into their initial job. After the first year as a full-time teacher, over 20%
of teachers in our sample are no longer full-time teachers at the school where they were
originally hired. Moreover, approximately half of the teachers who left their job left the
Michigan public school system entirely. Each of the outcomes increase or decrease
monotonically until, after five years since date of hire, more than half of the teachers in our
sample no longer worked in their initial job and more than one-fifth of all the teachers hired
between fall of 2003 and fall of 2008 did not work in Michigan public schools in any capacity.
Also of note, teacher mobility into a new school outside the district where teachers originally
work increases more rapidly than mobility into a new school within the district. In year two, the
portion of teachers who accept a new full-time teaching job outside their original district is less
than the portion of teachers who accept a new full-time teaching job within their original district.
In year five, however, the portion of teachers who are full-time teachers outside their original
district is greater than the portion of teachers who work as full-time teachers at a new school
within their original district.
Next, Table 4 shows five-year mobility outcomes by cohort. Cohorts are defined by the
spring of the school year, so teachers who were hired in the fall of 2003 are considered to be part
of the 2004 cohort. Keeping in mind these statistics are unconditional on other observable
factors, the portion of teachers in each mobility outcome five years since date of hire is roughly
constant across all cohorts in our sample. The number of teachers in each cohort, however,
19
decreases drastically in the latter years of our sample. In fact, the smallest cohort of teachers
contains slightly more than half as many teachers as the largest cohort.
Third, Table 5 represents the distribution of mobility outcomes among Michigan public
schools. Teacher mobility outcomes after five years are unequally distributed among teachers
based on FRL in their first school. Teachers who work in the poorest schools have lower
portions of teachers who stay in the same school and same job for five years than teachers who
work in the least poor schools. Slightly more than 70% of the quarter of teachers who work in
schools with the highest portion of students eligible for free or reduced lunch did not work in
their same school and job five years after their date of hire. Among the quarter of teachers who
accepted their first job in the least poor schools, more than half stay in the same school and job
for five years. Teachers who accept their first job in the poorest schools also take a second full-
time teaching job outside their original school district and leave Michigan public schools in
greater proportions than teachers who accept their first job in the least poor schools.
Table 6 presents results from the primary linear probability specification outlined in the
previous section. The outcome variable in columns 1 and 2 is an indicator for remaining in one's
first job for two years; the outcome variable in columns 3 and 4 is an indicator for remaining in
one's first job for five years. In columns 1 and 3, we run separate regressions for each type of
covariate, and the groups are denoted by colors. For instance, one regression has only the
demographic characteristics of the school as independent variables; this regression is highlighted
in red. Another regression has only the students' standardized test quintile indicators as
independent variable; this regression is highlighted in orange. Columns 2 and 4 include all
covariates together. All specifications contain missing indicators (not shown) for any variables
20
that are ever missing, and all standard errors are clustered by school. All teachers whose first
school does not stay open for five years after the teacher is hired are removed from the sample.
As shown in column 2, 78.5% of teachers in the sample stay in their first job for at least
two years. Many of the most substantively significant predictors of staying in one's first job are
related to teachers' credentials. Conditional on all sets of covariates, teachers with master's
degrees are 9.6 percentage points more likely to stay in their first jobs; teachers who are highly
qualified are 6.5 percentage points more likely to stay in their first jobs; and teachers who
attended college in Michigan are 5.1 percentage points more likely to stay in their first jobs.
These results are not terribly surprising; teachers who have fulfilled these requirements have
invested a lot into teaching and are unlikely to want to leave at the first sight of trouble. Teachers
with a connection to the state are less likely to want to leave for other locales, even at a time
when many people were emigrating from Michigan.
Meanwhile, teachers in urban schools and schools with larger fractions of black students
are less likely to stay in their first jobs for two years. Teachers in urban schools are 4.8
percentage points less likely to stay in their first jobs than suburban teachers, and a teacher in a
school in which 30% of students are black is about 2.5 percentage points less likely to stay in his
or her first job than a teacher in a school with no black students. Considering how most urban
schools in Michigan have very high black populations, the combination of the coefficients on the
fraction of black students and the indicator for a school being in a city adds up to a drastic
decrease in the probability that a new teacher in Detroit, Flint, or Saginaw will complete two
years on the job.
Only about 44.7% of teachers complete five full years in their first jobs in the Michigan
public school system. Column 4 shows that many of the same predictors that are associated with
21
staying in one's first job for two years also predict the probability of staying in that job for five
years. Teachers with master's degrees are 15.2 percentage points more likely to stay in their first
jobs for five years; teachers who attended college in Michigan are 5 percentage points more
likely to stay in their first jobs. Perhaps these teachers are most able to pick a first job that they
will like and will be able to keep long-term; teachers who have lived in Michigan are familiar
with the area and may know more about which schools have the most amenable working
conditions, while teachers with master's degrees may have more initial job offers and they can
pick the one that suits them best.
As in the two-year framework, teachers are more likely to leave jobs in urban schools and
schools with large numbers of black students within five years. Teachers in city schools are 3.4
percentage points more likely to leave their first jobs than their suburban counterparts; a teacher
in a school with 30% black students is 4.8 percentage points more likely to leave than a teacher
in a school with no black students. Many of the urban schools with largely black student
populations also have very poor student bodies, and the five-year specifications show that
teachers are also more likely to leave schools with more economically disadvantaged students.
Teachers in schools in which 30% of students are eligible for free or reduced-price lunch are
about 2.6 percentage points more likely to leave within five years than their counterparts in
schools with no students eligible for subsidized meals. Considering how intertwined these factors
are, and how vulnerable these students are even without constant teacher turnover, this is a
worrisome result for policymakers who are concerned about the racial and socioeconomic gaps
in educational outcomes that may be exacerbated by the flow of teachers out of these schools.
Unlike in the two-year specification, we see significant effects of the later cohort
dummies on the probability of staying in one's first job for five years. Teachers who entered the
22
Michigan public school system in 2007-08 or 2008-09 are about 5 percentage points more likely
to stay in their first job for five years than are their colleagues who entered prior to 2006-07.
These teachers entered during a deep and damaging recession, and during that five-year time
span, the job market improved significantly, expanding their set of outside options. Since these
teachers were more likely to stay in their first jobs despite these increased outside options, it
appears that they were particularly well-matched to their first jobs, which implies that schools
and districts that hired teachers during the recession were able to determine which teachers
would be good fits from their candidate pools. Alternatively, there were increases in school
funding that occurred early in the careers of the most recent hires, improving school working
conditions and giving them more incentives to stay in their first jobs.
Table 7 reports odds ratios from a multinomial logit specification with four possible
outcomes measured against the base case of staying in one's first job for five years: getting a new
full-time teaching job in the same district, getting a new full-time teaching job in a different
district, switching to part-time or administrative work in the public school system, or leaving the
Michigan public school system entirely. Teachers with more black students in their schools are
significantly more likely to leave their first school and go anywhere else; teachers with more
economically disadvantaged students are much more likely to take a new job in another district,
but are not significantly more likely to make other types of career moves. The gender variable is
never significant in these results; perhaps surprisingly, women are no more likely than men to
switch from full-time to part-time work in the same school.
Teachers hired in 2007-08 and 2008-09 are less likely to leave their first jobs for any
reason than their colleagues in earlier cohorts, but they are particularly less likely to leave the
public school system or to leave their hiring districts. This supports the hypothesis that districts
23
that hire teachers during recessions are able to determine which teachers are most committed to
teaching and fit best in the district. Similarly, teachers who have master's degrees are much less
likely to leave teaching in Michigan or to switch to part-time work; these teachers have made an
expensive initial investment in teaching and demonstrate their commitment by remaining in the
profession. Conversely, teachers who have degrees from selective colleges are more likely to
leave teaching in Michigan or to switch to part-time work; their credentials are more universal
signals of high ability and possibly of skills that are useful in many sectors, which would give
these teachers greater outside options, whereas a master's degree in education provides skills that
are more specifically tailored to teaching. One might also expect that math and science teachers
would have greater outside options than teachers in general education, English, or social studies;
these teachers are slightly more likely to leave teaching, but the result is only marginally
significant.
In order to see whether at-risk schools are disproportionately losing their most qualified
teachers, we incorporate terms that interact measures of teacher qualification (highly qualified
status, possession of a master's degree, attendance at a selective college, and all of the above
combined) with an indicator for whether a school has an economically disadvantaged student
body. These results are shown in Table 8. In three of the four specifications, the interaction term
is not significant; the exception is the interaction term involving the master's degree. While
teachers with master's degrees are much more likely (18 percentage points) to stay at their first
jobs for five years, this drops to 12.6 percentage points, a 5.4-percentage point decrease, if the
first job is in an economically disadvantaged school. If we believe that a master's degree is a
mark of an effective teacher, then this is a worrisome result for those concerned about equity.
Table 9 adds a triple interaction term, multiplying the existing interaction terms by an indicator
24
for whether the teacher has an assignment teaching math or science. This indicator is never
significant, implying that poor schools are not losing high-quality math and science teachers
more than they are losing high-quality English and social studies teachers.
To focus more on the choice to move into poor schools and to examine the Bridges
(1990) "dance of the lemons" hypothesis, we present odds ratios from another multinomial logit
regression in Table 10, in which three outcomes are compared against the base case of staying in
one's first job for five years: taking a new job in a non-disadvantaged school, taking a new job in
an economically disadvantaged school, and leaving the Michigan public school system. One of
the strongest predictors of taking a new job in an economically disadvantaged school is the
fraction of students in the teacher's first school who are eligible for free or reduced-price lunch;
while this variable has essentially no effect on the probability of taking a new job in a non-
disadvantaged school, it has an extremely large and significant positive relationship with the
probability of taking a new job in a disadvantaged school, indicating that teachers who start in
disadvantaged schools tend to bounce around between different poor schools until they leave
teaching. This is part of the phenomenon that Bridges (1990) calls "the dance of the lemons".
However, it is somewhat surprising that the coefficient on the subsidized lunch variable is
essentially zero for taking a new job in a non-disadvantaged school, as teachers in poorer schools
would seem to want to pursue better working conditions in less-impoverished environments if
possible.
Table 11 continues with the theme of examining who takes jobs in disadvantaged
schools; this table presents linear probability model results from the sample of teachers who
switch schools, examining what factors are associated with taking a job in a poor school instead
of a non-poor school. Once again, the amount of students eligible for free or reduced-price lunch
25
is a key predictor; teachers with 30% of students eligible for subsidized lunch are about 9
percentage points more likely to switch to a poor school than teachers with no students eligible
for subsidized lunch. Worryingly, teachers with advanced degrees and teachers in math and
science are less likely to switch to a poor school, which reinforces the gap in teacher quality and
may leave these schools short-staffed in key subjects.
Finally, in order to evaluate how these factors may interact with each other, we calculate
results separately by a wide variety of subgroups. These heterogeneity specifications, shown in
Table 12, include separate five-year linear probability model results by race, gender, student
achievement, school socioeconomic disadvantage, and teacher qualifications. The most
immediately interesting results come from the split by race. White teachers with 30% black
students are about 4.9 percentage points less likely to stay in their first job than white teachers
with no black students; there is no statistically or substantively significant effect of the size of a
school's black population on whether black teachers stay in their first job or leave it. Taken at
face value, this result is an extremely troubling reflection of white teachers' racial attitudes in
Michigan, and this interpretation should not be discounted. However, the urbanicity variables
add some context. White teachers are 2.9 percentage points less likely to stay in urban schools
than in suburban schools, all else equal; this effect is marginally significant. On the other hand,
black teachers are 53 percentage points less likely to stay in small-town or rural schools than in
suburban schools, while white teachers are 3.5 percentage points more likely to stay in small-
town or rural schools than in suburban schools. Considering the vast racial disparities between
26
rural and urban areas in Michigan4, it is likely that both black and white teachers are interested in
moving closer to home when possible.
One of the more interesting conclusions drawn from Tables 6 and 7 is that teachers hired
in more recent cohorts are more likely to stay in their first schools, even though the labor market
improved during their time in teaching and thus their outside options increased. Table 12 reveals
that this is true both for teachers who started in wealthier schools and teachers who started in
poor schools. For instance, the 2007-08 entry cohort is 5.4 percentage points more likely than the
2003-04 entry cohort to stay in their first jobs in non-disadvantaged schools, while they are 4.6
percentage points more likely to stay at disadvantaged schools; both of these are significant at
the 0.05 level. The result for the 2008-09 cohort is noticeably stronger for poorer schools (6.1
percentage points, significant at 0.05) than for less-poor schools (1.5 percentage points, not
significant at conventional levels). If schools and districts are able to screen candidates for the
best fit during recessions, when the applicant pool is larger, then it is reassuring from a policy
perspective to see that this screening ability is not exclusive to the wealthiest and most-
advantaged hiring districts.
Discussion and Conclusion
There are a variety of potential policy responses to these findings. Because white teachers
are more inclined to leave urban schools with large numbers of black students and black teachers
are not, scholarship programs aimed at African-Americans interested in studying education may
help assure that there are more teachers willing to continue working in these schools instead of
leaving at the first sign of another offer. Alternatively, steeper yearly raises for teachers who
4 As of the 2000 census, Detroit was 81.5% black, Flint was 53.3% black, Lansing was 21.9% black, and Grand Rapids was 20.5% black. The entire state's population was 14.2% black.
27
continue working in these schools may encourage teachers to stay in them for longer periods of
time.
However, if we take the cohort results to mean that schools really can determine teacher
quality (or at least determine whether a teacher is a good fit) when there are many more
applicants than spots, the best thing to do might just be to expand the applicant pool as much as
possible and let schools discern which teachers will be the best fits for their available positions.
Even schools with large populations of low-income students seem to be able to hang onto
teachers hired during the late-2000s recession, when jobs were hard to come by and thus the
applicant pool was large. If the applicant pool can remain large throughout the business cycle,
schools can look for the observable and unobservable characteristics associated with a good fit
and hire teachers who seem most likely to stay.
We measure a large number of characteristics about teachers, but we do not observe
everything. For instance, we do not observe a teacher's home town; our best guess is based on
where he or she attended college. Thus, we do not know exactly how much of the effects of race
and urbanicity are actually due to teachers' desire to move closer to home, and how much are
indeed due to white teachers' preference not to teach black students. Additionally, we do not
have reliable measures of districts' pay structures and working conditions, and these are
significant contributors to teachers' choice to leave for new jobs.
Even so, these results can teach us a great deal, and they contribute to a growing body of
literature about teacher mobility and turnover. White teachers are significantly more likely to
leave their first job if their first school has more black students, black teachers are unaffected.
White teachers are significantly more likely to leave urban schools, while black teachers are
much more likely to leave rural or small-town schools. Poorer schools frequently had trouble
28
hanging on to their teachers, but regardless of socioeconomic status, schools were better able to
keep their teachers who were hired during recessions, and these teachers stayed even as the
economy improved. This is a reassuring result for anyone who sees the teacher turnover problem
as intractable, and it deserves further study.
REFERENCES
Boyd, D., Lankford, H., Loeb, S. & Wyckoff, J. (2005). The draw of home: How teachers'
preferences for proximity disadvantage urban schools. Journal of Policy Analysis and
Management, 24, 113 - 132.
Bridges, E. M. (1990). 9. Evaluation for Tenure and Dismissal. The new handbook of teacher
evaluation: Assessing elementary and secondary school teachers, 147-157.
Clotfelter, C., Ladd, H., & Vigdor, J. (2006). Teacher-student matching and the assessment of
teacher effectiveness. Journal of Human Resources, 41, 778-820.
Dee, T. (2004). The race connection: Are teachers more effective with students who share their
ethnicity? Education Next, Spring 2004, 53-59.
Figlio, D. N. (2002). Can public schools buy better-qualified teachers? Industrial and Labor
Relations Review, 55, 686 - 699.
Goldhaber, D., Gross, B., & Player, D. (2011). Teacher career paths, teacher quality, and
persistence in the classroom: Are public schools keeping their best? Journal of Policy
Analysis and Management, 30, 57 - 87.
29
Goldhaber, D., Lavery, L., & Theobald, R. (2015). Uneven playing field? Assessing the teacher
quality gap between advantaged and disadvantaged students. Educational Researcher, 44,
293 - 307.
Hanushek, E. A., & Rivkin, S. G. (2007). Pay, working conditions, and teacher quality. The
Future of Children, 17, 69 - 86.
Hoxby, C. M. & Leigh, A. (2004). Pulled away or pushed out? Explaining the decline of teacher
aptitude in the United States. American Economic Review, 236 - 240.
Lankford, H., Loeb, S., Wyckoff, J. (2002). Teacher sorting and the plight of urban schools.
Educational Evaluation and Policy Analysis, 24, 37 - 62.
Murnane, J., & Steele, R. (2007). What is the problem? The challenge of providing effective
teachers for all children. The Future of Children, 17, 15-43.
Podgursky, M., Monroe, R., & Watson, D. (2004). The academic quality of public school
teachers: An analysis of entry and exit behavior. Economics of Education Review, 23, 507
- 518.
Springer, M. G., Swain, W. A., & Rodriguez, L. A. (in press). Effective teacher retention
bonuses: Evidence from Tennessee. Educational Evaluation and Policy Analysis,
Stinebrickner, T. R. (2001). A dynamic model of teacher labor supply. Journal of Labor
Economics, 19, 196-230.
Table 1. Distribution of full-time teachers in Michigan public schools in 2014.Table Notes Fewest Poor Students 2nd 3rd Most Poor StudentsTeacher is Black .014 .013 .041 .197Teacher is Hispanic .007 .006 .007 .019Teacher is Asian .007 .004 .005 .009Attended Very Selective College .196 .177 .172 .162Highly Qualified .865 .887 .89 .866Teaches Math .082 .086 .071 .064Teaches Science .081 .078 .061 .048Teaches English .099 .101 .091 .083Teaches Social Studies .073 .073 .056 .046Teaches Special Education .128 .156 .169 .166Elementary Teacher .7 .694 .789 .869Secondary Teacher .3 .306 .211 .131Attended College in Michigan .659 .632 .644 .582Teacher Age 43.33 43.016 43.239 43.115Years Since Hire 14.547 14.355 14.247 13.2441 - 3 Years Experience .067 .067 .081 .1424 - 10 Years Experience .241 .253 .249 .25111 - 20 Years Experience .502 .495 .48 .43220+ Years Experience .19 .184 .19 .174Advanced Degree .687 .623 .563 .552Max % Eligible for Subsidized Lunch 0.282 0.488 0.700 1Number of teachers 18539 18539 18520 18525Number of schools 695 825 1007 916
Notes: Sample includes all full-time teachers in any Michigan public school for the 2013 - 2014 school year. Quartiles of poor students are defined by percent of students in the school who are eligible for free or reduced price lunch. Cutoffs for these quartiles are 28%, 49%, and 70% for the 25th, 50th, and 75th percentiles respectively. Data for attendance at a Michigan college are unavailable after 2012. Teachers who accepted their first full-time teaching job in the 2012 -2013 school year or later are missing for this statistic.
Table 2. Distribution of initial placement of teachers in Michigan public schools from 2004 - 2009
Table Notes Fewest Poor Students 2nd 3rd Most Poor StudentsTeacher is Black .023 .023 .061 .162Teacher is Hispanic .008 .006 .007 .02Teacher is Asian .01 .006 .008 .008Attended Very Selective College .184 .195 .16 .138Highly Qualified .838 .857 .853 .846Teaches Math .094 .09 .089 .073Teaches Science .075 .072 .064 .063Teaches English .11 .11 .085 .077Teaches Social Studies .074 .066 .056 .056Teaches Special Education .18 .191 .181 .148Elementary Teacher .633 .66 .744 .807Secondary Teacher .367 .34 .256 .193Attended College in Michigan .838 .842 .818 .786Teacher Age 30.365 30.121 30.875 32.153Has Advanced Degree .702 .652 .596 .526Number of Teachers 4068 4064 4066 4066Number of Schools 966 1112 1176 793
Notes: Sample includes all full-time teachers in any Michigan public school between the 2003 - 2004 school year and the 2008 - 2009 school year. The sample of teachers is restricted to those who accepted their first job in a school that remained open for at least five years thereafter. Quartiles of poor students are defined by percent of students in the school who are eligible for free or reduced price lunch. Cutoffs for these quartiles are 16%, 34%, and 61% for the 25th, 50th, and 75th percentiles respectively.
Table 3. Mobility outcomes by years since hire date for all teachers hired into Michigan public schools from 2004 - 2009.Table Notes Year 2 Year 3 Year 4 Year 5In a teaching position in the same school, full-time .784 .619 .516 .446
Full-time teacher in a new school outside original district .016 .086 .122 .145
Full-time teacher in a new school within original district .056 .082 .098 .108Employee of Michigan public schools who is not a full-time teacher .035 .052 .062 .073Left Michigan public schools .108 .16 .201 .228
Note: All outcomes are relative to the first year of teaching in Michigan public schools. The sample includes all teachers who accepted a full-time teaching job in any Michigan public school between the 2003 - 2004 school year and the 2008 - 2009 school year. The sample of teachers is restiricted to those who accepted their first job in a school that remained open for at least five years thereafter.
Table 4. Mobility outcomes five years since the date of hire by cohort for all teachers who were hired into Michigan public schools from 2004 - 2009.Table Notes 2004 2005 2006 2007 2008 2009In a teaching position in the same school, full-time .436 .438 .458 .442 .455 .455Full-time teacher in a new school outside original district .151 .153 .127 .147 .147 .142Full-time teacher in a new school within original district .107 .099 .12 .117 .094 .108Employee of Michigan public schools who is not a full-time t .065 .073 .074 .083 .073 .069Left Michigan public schools .24 .236 .219 .21 .229 .224Number of teachers 3435 3538 3268 2514 1987 1820
Note: All outcomes reflect the teachers' status five years after original date of hire relative to their first full-time teaching job. The sample includes all teachers who accepted a full-time teaching job in any Michigan public school between the 2003 - 2004 school year and the 2008 - 2009 school year conditional on the school remaining open for at least five years after initial date of hire.
Table 5. Mobility outcomes, five years after the date of hire, by fraction of poor students for all teachers who were hired into Michigan public schools from 2004 - 2009.
Table NotesFewest Poor
Students 2nd 3rdMost Poor Students
In a teaching position in the same school, full-time .54 .504 .443 .296Full-time teacher in a new school outside district .093 .115 .162 .214Full-time teacher in a new school within original district .113 .118 .114 .089Employee of Michigan public schools who is not a full-time teac .07 .07 .071 .078Left Michigan public schools .183 .194 .21 .323Number of teachers 4065 4059 4063 4062Number of schools 964 1112 1178 792
Note: All outcomes are relative to the first year of teaching. The sample includes all teachers who accepted a full-time teaching job in any Michigan public school between the 2003 - 2004 school year and the 2008 - 2009 school year conditional on the school remaining open for at least five years after the initial date of hire. Poverty quartiles are defined by percent of students eligible for free or reduced price lunch at the school-level. Cutoffs for these quartiles are 16%, 34%, and 61% for the 25th, 50th, and 75th percentiles respectively.
Table 6: OLS Regressions of Teacher Mobility on School and Teacher CharacteristicsDependent variable: indicator for still being in one's first job after a given number of years.Table Notes [1] [2] [3] [4]Frac. Free/Reduced Lunch Eligible -0.054*** -0.025 -0.143*** -0.086***
(0.018) (0.019) (0.021) (0.022)Fraction of Black Students -0.136*** -0.083*** -0.235*** -0.161***
(0.015) (0.019) (0.016) (0.022)Fraction of Hispanic Students -0.022 0.056 0.082 0.144***
(0.051) (0.045) (0.057) (0.052)Frac. Limited English Proficiency Students -0.052* -0.048 -0.176*** -0.147***
(0.03) (0.031) (0.032) (0.028)Frac. Special Education Students -0.016 -0.016 -0.057*** -0.065***
(0.016) (0.017) (0.019) (0.02)Standardized Student Achievement - 2nd Quintile 0.077*** 0.045*** 0.07*** 0.005
(0.016) (0.015) (0.018) (0.017)Standardized Student Achievement - 3rd Quintile 0.099*** 0.025 0.169*** 0.011
(0.015) (0.016) (0.019) (0.02)Standardized Student Achievement - 4th Quintile 0.131*** 0.042** 0.238*** 0.039*
(0.015) (0.017) (0.018) (0.021)Standardized Student Achievement - 5th Quintile 0.127*** 0.035* 0.256*** 0.045**
(0.015) (0.018) (0.018) (0.022)First School is a Charter -0.007 0.03** -0.113*** -0.039***
(0.012) (0.012) (0.015) (0.014)First School is Vocational -0.032 0.079 0.038 0.038
(0.048) (0.052) (0.067) (0.073)First School is Special Education-Only 0.014 0.111*** -0.024 0.104***
(0.021) (0.025) (0.029) (0.036)First School is Alternative -0.046* 0.012 -0.172*** -0.128***
(0.027) (0.025) (0.03) (0.028)First School Located in a City -0.087*** -0.048*** -0.112*** -0.034**
(0.012) (0.012) (0.015) (0.015)First School Located in a Town/Rural Area 0.013 0.007 0.042** 0.035**
(0.013) (0.013) (0.017) (0.017)First School is a Secondary School 0.037*** 0.03*** 0.022** -0.009
(0.008) (0.009) (0.011) (0.012)Average Frac. Change in Enrollment in First School 0.001* 0.001*** -0.001** -0.001
(0.0003) (0.0003) (0.001) (0.001)Local Unemployment Rate -0.005*** 0.002 -0.005*** 0.003
(0.001) (0.002) (0.002) (0.002)Teacher is Highly Qualified 0.036** 0.065*** -0.009 0.022*
(0.014) (0.011) (0.018) (0.012)Teacher Age -0.002*** -0.001*** -0.0004 0.001
(0.0004) (0.0004) (0.0005) (0.0004)Teacher is Female -0.01 -0.003 -0.013 -0.003
(0.008) (0.007) (0.01) (0.009)Teacher is Black -0.067*** -0.002 -0.123*** 0.015
(0.016) (0.016) (0.02) (0.018)Teacher is Hispanic 0.012 0.019 0.02 0.033
(0.035) (0.032) (0.042) (0.034)Teaches Math or Science 0.008 -0.003 -0.007 -0.02*
(0.009) (0.009) (0.012) (0.011)Teaches Special Education 0.011 0.005 -0.059*** -0.093***
(0.012) (0.011) (0.016) (0.012)Teaches English as a Second Language -0.041 -0.011 -0.163*** -0.124***
Same Job in 2 Years Same Job in 5 Years
(0.051) (0.05) (0.048) (0.047)Teaches Art or Music 0.018 0.015 -0.084*** -0.089***
(0.013) (0.013) (0.016) (0.016)Teacher Attended Selective College 0.003 -0.003 -0.002 -0.022**
(0.008) (0.008) (0.011) (0.01)Teacher Attended College in Michigan 0.051*** 0.051*** 0.055*** 0.05***
(0.011) (0.01) (0.013) (0.011)Teacher Has Master's Degree Or Higher 0.103*** 0.096*** 0.172*** 0.152***
(0.007) (0.007) (0.009) (0.008)Entered Teaching in 2004-05 0.017 0.01 0.002 -0.001
(0.011) (0.01) (0.013) (0.012)Entered Teaching in 2005-06 0.006 -0.011 0.022 0.005
(0.012) (0.011) (0.014) (0.012)Entered Teaching in 2006-07 -0.015 -0.017 0.007 0.025*
(0.013) (0.013) (0.015) (0.014)Entered Teaching in 2007-08 0.013 0.02 0.019 0.055***
(0.014) (0.013) (0.019) (0.016)Entered Teaching in 2008-09 -0.014 -0.007 0.019 0.05***
(0.014) (0.015) (0.016) (0.018)Number of Observations 16552 16552 16544 16544Adjusted R² n/a 0.075 n/a 0.103Sample Mean 0.785 0.447
NOTES: Sample includes teachers beginning their first jobs as full-time teachers between fall 2003 and fall 2008. Outcome variable is whether the teacher is in the same job two or five years after starting, conditional on the teacher's first school remaining open for five years. School/student characteristics are measured in the first school at the time of hire unless stated otherwise. Standard errors are clustered at the school level. Differently-colored results in Column [1] and [3] come from separate regressions. Missing indicators are included for all variables that are ever missing.
Table 7: Multinomial Logit of School and Teacher Characteristics on Across-District and Within-District Mobility Outcomes (Relative Risk Ratios Reported)
Dependent variable: indicators for whether a teacher has left his/her first job and taken a new full-time job in another district, a new full-time job in a different school in the same district, a part-time or non-teaching job within the Michigan public school system, or a position outside of teaching, respectively. Base case is staying in one's first job for five years.
Table Notes
Teacher Takes Full-Time Job In Another
DistrictTeacher Takes New Full-Time Job In Same District
Teacher Takes Part-Time or Non-Instructional Job
Teacher Leaves Michigan Public School System
Frac. Free/Reduced Lunch Eligible 2.15*** 1.31 1.25 1.2(0.318) (0.273) (0.224) (0.154)
Fraction of Black Students 1.95*** 2.81*** 1.2 2.34***(0.288) (0.576) (0.207) (0.281)
Fraction of Hispanic Students 0.424** 0.511 0.647 0.506**(0.149) (0.274) (0.254) (0.147)
Frac. Limited English Proficiency Students 1.24 3.27*** 1.54** 2.12***(0.249) (1.01) (0.337) (0.384)
Frac. Special Education Students 1.3* 1.48*** 1.16 1.35**(0.194) (0.208) (0.2) (0.165)
Standardized Student Achievement - 2nd Quintile 0.893 1.58*** 0.972 0.852*(0.104) (0.25) (0.122) (0.077)
Standardized Student Achievement - 3rd Quintile 0.843 1.49** 1.07 0.798**(0.118) (0.259) (0.155) (0.082)
Standardized Student Achievement - 4th Quintile 0.68*** 1.41* 0.998 0.708***(0.097) (0.25) (0.151) (0.079)
Standardized Student Achievement - 5th Quintile 0.515*** 1.55** 0.818 0.75**(0.079) (0.286) (0.133) (0.087)
First School is a Charter 1.78*** 0.077*** 1.67*** 1.75***(0.168) (0.018) (0.162) (0.131)
First School is Vocational 0.645 0.695 0.862 0.872(0.312) (0.405) (0.435) (0.304)
First School is Special Education-Only 0.491*** 0.581** 0.727 0.621**(0.133) (0.134) (0.187) (0.116)
First School is Alternative 1.72** 1.33 2.95*** 1.57**(0.364) (0.3) (0.582) (0.312)
First School Located in a City 1.02 1.54*** 1.1 1.19**(0.101) (0.17) (0.118) (0.092)
First School Located in a Town/Rural Area 0.783** 0.793* 0.912 0.966(0.083) (0.101) (0.122) (0.091)
First School is a Secondary School 1.34*** 0.433*** 1.37*** 1.27***(0.1) (0.047) (0.131) (0.081)
Average Frac. Change in Enrollment in First School 1.42*** 0.887 1.31*** 1.02(0.126) (0.099) (0.118) (0.1)
Local Unemployment Rate 1.01 0.989 0.991 0.974***(0.012) (0.015) (0.013) (0.01)
Teacher is Highly Qualified 1.09 0.895 0.751*** 0.853**(0.092) (0.08) (0.071) (0.061)
Teacher Age 0.975*** 1.01* 1.01* 1(0.003) (0.003) (0.004) (0.003)
Teacher is Female 0.912 0.997 1.02 1.09(0.056) (0.066) (0.08) (0.057)
Teacher is Black 0.868 0.928 1.44*** 0.941(0.11) (0.144) (0.194) (0.107)
Teacher is Hispanic 0.664 1.09 1.17 0.774(0.175) (0.288) (0.317) (0.157)
Teaches Math or Science 1.27*** 0.979 0.856 1.14*(0.09) (0.088) (0.094) (0.076)
Teaches Special Education 2.16*** 1.17* 1.85*** 1.29***(0.176) (0.099) (0.174) (0.1)
Teaches English as a Second Language 2.08** 1.84* 1.22 2.05**(0.749) (0.645) (0.618) (0.622)
Teaches Art or Music 1.8*** 0.945 1.52*** 1.7***(0.19) (0.129) (0.207) (0.159)
Teacher Attended Selective College 1.05 1.03 1.31*** 1.14**(0.073) (0.072) (0.112) (0.072)
Teacher Attended College in Michigan 1.1 1.03 0.759*** 0.567***(0.085) (0.084) (0.068) (0.035)
Teacher Has Master's Degree or Higher 0.981 1.03 0.699*** 0.186***(0.056) (0.066) (0.047) (0.01)
Entered Teaching in 2004-05 0.967 0.916 1.18* 1.04(0.075) (0.082) (0.119) (0.074)
Entered Teaching in 2005-06 0.803** 1.1 1.21* 0.945(0.068) (0.103) (0.128) (0.07)
Entered Teaching in 2006-07 0.889 0.996 1.21 0.76***(0.082) (0.105) (0.143) (0.063)
Entered Teaching in 2007-08 0.821** 0.813* 0.923 0.719***(0.082) (0.092) (0.117) (0.065)
Entered Teaching in 2008-09 0.751** 0.957 0.928 0.726***(0.093) (0.117) (0.131) (0.073)
Number of Observations 16543 16543 16543 16543Pseudo R² 0.126 0.126 0.126 0.126Sample Mean 0.145 0.108 0.073 0.228
NOTES: Sample includes all teachers beginning their first jobs as full-time teachers between fall 2003 and fall 2008. Outcome variable is measured in the teacher's fifth year, conditional on the teacher's first school still being open. School/student characteristics are measured in the first school at the time of hire unless stated otherwise. Odds ratios reported. Standard errors are clustered at the school level. Sample includes one observation per teacher. Missing indicators are included for all variables that are ever missing.
Table 8: OLS Regressions With Interactions of Teacher Quality and School Disadvantage
Dependent variable: indicator for still being in one's first job after five years.
Teacher Is Highly Qualified
Has Master's Degree Or
HigherAttended
Selective College[1], [2], and [3]
Table Notes [1] [2] [3] [4]Frac. Free/Reduced Lunch Eligible -0.074*** -0.054** -0.084*** -0.085***
(0.024) (0.023) (0.023) (0.022)Fraction of Black Students -0.159*** -0.157*** -0.16*** -0.16***
(0.022) (0.022) (0.022) (0.022)Fraction of Hispanic Students 0.146*** 0.15*** 0.146*** 0.145***
(0.052) (0.052) (0.052) (0.052)Frac. Limited English Proficiency Students -0.148*** -0.152*** -0.147*** -0.147***
(0.028) (0.028) (0.028) (0.028)Frac. Special Education Students -0.065*** -0.065*** -0.065*** -0.065***
(0.02) (0.02) (0.02) (0.02)Standardized Student Achievement - 2nd Quintile 0.006 0.009 0.005 0.005
(0.017) (0.017) (0.017) (0.017)Standardized Student Achievement - 3rd Quintile 0.011 0.015 0.011 0.011
(0.02) (0.02) (0.02) (0.02)Standardized Student Achievement - 4th Quintile 0.038* 0.038* 0.039* 0.039*
(0.02) (0.02) (0.021) (0.021)Standardized Student Achievement - 5th Quintile 0.043** 0.042* 0.045** 0.045**
(0.022) (0.021) (0.022) (0.022)First School is a Charter -0.04*** -0.039*** -0.04*** -0.04***
(0.014) (0.014) (0.014) (0.014)First School is Vocational 0.039 0.039 0.039 0.039
(0.073) (0.074) (0.073) (0.073)First School is Special Education-Only 0.106*** 0.113*** 0.105*** 0.104***
(0.036) (0.036) (0.036) (0.036)First School is Alternative -0.127*** -0.124*** -0.128*** -0.128***
(0.028) (0.028) (0.028) (0.028)First School Located in a City -0.035** -0.035** -0.034** -0.034**
(0.015) (0.015) (0.015) (0.015)First School Located in a Town/Rural Area 0.035** 0.037** 0.035** 0.035**
(0.017) (0.017) (0.017) (0.017)First School is a Secondary School -0.01 -0.011 -0.009 -0.009
(0.012) (0.012) (0.012) (0.012)Average Frac. Change in Enrollment in First School -0.001 -0.001 -0.001 -0.001
(0.001) (0.001) (0.001) (0.001)Local Unemployment Rate 0.003 0.003* 0.003 0.003
(0.002) (0.002) (0.002) (0.002)Teacher is Highly Qualified 0.03** 0.023* 0.022* 0.023*
(0.014) (0.012) (0.012) (0.012)Teacher Age 0.001 0.001 0.001 0.001
(0.0004) (0.0004) (0.0004) (0.0004)Teacher is Female -0.002 -0.002 -0.002 -0.002
(0.009) (0.009) (0.009) (0.009)Teacher is Black 0.014 0.015 0.015 0.015
(0.018) (0.017) (0.018) (0.018)Teacher is Hispanic 0.033 0.033 0.033 0.033
(0.035) (0.035) (0.035) (0.034)Teaches Math or Science -0.02* -0.02* -0.02* -0.02*
(0.011) (0.011) (0.011) (0.011)Teaches Special Education -0.093*** -0.094*** -0.093*** -0.093***
(0.012) (0.012) (0.012) (0.012)
Teacher Quality Indicator:
Teaches English as a Second Language -0.126*** -0.126*** -0.124*** -0.124***(0.047) (0.047) (0.047) (0.047)
Teaches Art or Music -0.089*** -0.088*** -0.089*** -0.089***(0.016) (0.016) (0.016) (0.016)
Teacher Attended Selective College -0.023** -0.023** -0.016 -0.019(0.01) (0.01) (0.014) (0.012)
Teacher Attended College in Michigan 0.05*** 0.049*** 0.05*** 0.05***(0.011) (0.011) (0.011) (0.011)
Teacher Has Master's Degree Or Higher 0.151*** 0.18*** 0.151*** 0.153***(0.008) (0.011) (0.008) (0.009)
Teacher Quality × First School is Disadvantaged -0.015 -0.054*** -0.014 -0.017(0.012) (0.013) (0.02) (0.022)
Entered Teaching in 2004-05 -0.001 -0.001 -0.001 -0.001(0.012) (0.012) (0.012) (0.012)
Entered Teaching in 2005-06 0.004 0.004 0.005 0.005(0.012) (0.012) (0.012) (0.012)
Entered Teaching in 2006-07 0.025* 0.025* 0.025* 0.025*(0.014) (0.014) (0.014) (0.014)
Entered Teaching in 2007-08 0.055*** 0.054*** 0.055*** 0.055***(0.016) (0.016) (0.016) (0.016)
Entered Teaching in 2008-09 0.049*** 0.046** 0.05*** 0.05***(0.018) (0.018) (0.018) (0.018)
Number of Observations 16544 16544 16544 16544Adjusted R² 0.103 0.104 0.103 0.103Sample Mean 0.447 0.447 0.421 0.421
NOTES: Sample includes all teachers beginning their first jobs as full-time teachers between fall 2003 and fall 2008. Outcome variable is whether the teacher is in the same job five years after starting, conditional on the teacher's first school remaining open. "Disadvantaged" schools have at least 30% of their students eligible for free or reduced lunch in their first year of data. School/student characteristics are measured in the first school at the time of hire unless stated otherwise. Standard errors are clustered at the school level. The "teacher quality indicator" is the variable in the column heading, which varies by column. Missing indicators are included for all variables that are ever missing.
Table 9: OLS Regressions With Interactions of Teacher Quality, STEM Assignment and School Disadvantage (Secondary Schools Only)
Dependent variable: indicator for still being in one's first job after five years.
Teacher Is Highly Qualified
Has Master's Degree Or
HigherAttended
Selective College[1], [2], and [3]
Table Notes [1] [2] [3] [4]Frac. Free/Reduced Lunch Eligible -0.091** -0.052 -0.094** -0.091**
(0.043) (0.041) (0.041) (0.041)Fraction of Black Students -0.129*** -0.119*** -0.131*** -0.131***
(0.037) (0.037) (0.037) (0.036)Fraction of Hispanic Students 0.031 0.041 0.029 0.032
(0.126) (0.124) (0.126) (0.126)Frac. Limited English Proficiency Students -0.065 -0.069 -0.065 -0.066
(0.057) (0.057) (0.057) (0.057)Frac. Special Education Students -0.015 -0.015 -0.015 -0.014
(0.049) (0.048) (0.049) (0.049)Standardized Student Achievement - 2nd Quintile -0.039 -0.039 -0.04 -0.04
(0.042) (0.041) (0.042) (0.042)Standardized Student Achievement - 3rd Quintile -0.001 0.003 -0.002 -0.002
(0.043) (0.043) (0.043) (0.043)Standardized Student Achievement - 4th Quintile -0.036 -0.034 -0.037 -0.037
(0.045) (0.045) (0.045) (0.045)Standardized Student Achievement - 5th Quintile 0.07 0.072 0.07 0.07
(0.048) (0.048) (0.048) (0.048)First School is a Charter -0.122*** -0.122*** -0.121*** -0.121***
(0.028) (0.028) (0.028) (0.028)First School is Vocational 0.033 0.034 0.031 0.032
(0.085) (0.087) (0.085) (0.085)First School is Special Education-Only 0.011 0.036 0.011 0.01
(0.171) (0.173) (0.17) (0.17)First School is Alternative -0.151*** -0.148*** -0.151*** -0.15***
(0.046) (0.044) (0.045) (0.045)First School Located in a City -0.028 -0.03 -0.028 -0.028
(0.03) (0.03) (0.03) (0.03)First School Located in a Town/Rural Area 0.019 0.019 0.018 0.018
(0.03) (0.03) (0.029) (0.029)Average Frac. Change in Enrollment in First School -0.001 -0.001 -0.001 -0.001
(0.0003) (0.0003) (0.0003) (0.0003)Local Unemployment Rate 0.0002 0.0004 0.0003 0.0004
(0.003) (0.003) (0.003) (0.003)Teacher is Highly Qualified -0.012 -0.01 -0.01 -0.01
(0.026) (0.024) (0.024) (0.024)Teacher Age 0.0004 0.0004 0.0003 0.0003
(0.001) (0.001) (0.001) (0.001)Teacher is Female -0.004 -0.004 -0.004 -0.004
(0.014) (0.014) (0.014) (0.014)Teacher is Black 0.015 0.015 0.017 -0.017
(0.035) (0.035) (0.035) (0.035)Teacher is Hispanic 0.079 0.076 0.08 0.079
(0.059) (0.059) (0.059) (0.06)Teaches Math or Science -0.003 -0.006 -0.011 -0.013
(0.018) (0.016) (0.015) (0.015)Teaches Special Education -0.093** -0.095*** -0.094** -0.094**
(0.036) (0.036) (0.036) (0.036)
Teacher Quality Indicator:
Teaches English as a Second Language -0.09 -0.097 -0.091 -0.091(0.103) (0.102) (0.103) (0.103)
Teaches Art or Music -0.071** -0.07** -0.071** -0.07**(0.029) (0.029) (0.029) (0.029)
Teacher Attended Selective College -0.035* -0.037** -0.038* -0.034*(0.018) (0.018) (0.021) (0.019)
Teacher Attended College in Michigan 0.016 0.013 0.016 0.016(0.022) (0.022) (0.022) (0.022)
Teacher Has Master's Degree or Higher 0.195*** 0.218*** 0.195*** 0.195***(0.016) (0.019) (0.016) (0.016)
Teacher Quality × First School is Disadvantaged 0.007 -0.051* 0.001 -0.026(0.025) (0.028) (0.04) (0.05)
Quality × Disadvantage × Teaches Math/Science -0.02 -0.021 0.017 0.062(0.03) (0.033) (0.06) (0.083)
Entered Teaching in 2004-05 -0.015 -0.016 -0.015 -0.014(0.023) (0.023) (0.023) (0.023)
Entered Teaching in 2005-06 -0.002 -0.004 -0.002 -0.002(0.023) (0.023) (0.023) (0.023)
Entered Teaching in 2006-07 0.044 0.042 0.044 0.044(0.033) (0.033) (0.033) (0.033)
Entered Teaching in 2007-08 0.085** 0.083** 0.085*** 0.085***(0.033) (0.033) (0.033) (0.033)
Entered Teaching in 2008-09 0.072* 0.065* 0.072* 0.071*(0.038) (0.038) (0.038) (0.038)
Number of Observations 4720 4720 4720 4720Adjusted R² 0.116 0.117 0.116 0.116Sample Mean 0.483 0.483 0.483 0.483
NOTES: Sample includes all secondary school teachers beginning their first jobs as full-time teachers between fall 2003 and fall 2008. Outcome variable is whether the teacher is in the same job five years after starting, conditional on the teacher's first school being open. "Disadvantaged" schools have at least 30% of their students eligible for free or reduced lunch in their first year of data. School/student characteristics are measured in the first school at the time of hire unless stated otherwise. Standard errors are clustered at the school level. The "teacher quality indicator" is the variable in the column heading, which varies by column. Missing indicators are included for all variables that are ever missing.
Table 10: Multinomial Logit of School and Teacher Characteristics on Mobility by Student Disadvantage (Relative Risk Ratios Reported)
Dependent variable: indicators for whether a teacher has left his/her first job and taken a new job at a non-disadvantaged school, a disadvantaged school, or a position outside of teaching, respectively. Base case is staying in one's first job.
Table Notes
Teacher Takes New Job in a Non-Disadvantaged
School
Teacher Takes New Job in a Disadvantaged
School
Teacher Leaves Michigan Public School System
Frac. Free/Reduced Lunch Eligible 0.997 3.29*** 1.28*(0.163) (0.593) (0.177)
Fraction of Black Students 2.21*** 2.26*** 2.35***(0.323) (0.366) (0.292)
Fraction of Hispanic Students 0.415** 0.47* 0.418***(0.162) (0.194) (0.14)
Frac. Limited English Proficiency Students 2.26*** 1.7** 2.17***(0.569) (0.353) (0.412)
Frac. Special Education Students 1.36** 1.48*** 1.45***(0.187) (0.21) (0.184)
Standardized Student Achievement - 2nd Quintile 0.953 1.13 0.836*(0.133) (0.131) (0.081)
Standardized Student Achievement - 3rd Quintile 0.994 1.06 0.787**(0.144) (0.152) (0.086)
Standardized Student Achievement - 4th Quintile 0.923 0.869 0.729***(0.134) (0.133) (0.086)
Standardized Student Achievement - 5th Quintile 1.12 0.493*** 0.77**(0.165) (0.087) (0.095)
First School is a Charter 0.798** 0.699*** 1.69***(0.087) (0.071) (0.135)
First School is Vocational 0.541 0.68 0.917(0.226) (0.384) (0.33)
First School is Special Education-Only 0.445*** 0.579** 0.558***(0.103) (0.139) (0.112)
First School is Alternative 1.6*** 1.4 1.34(0.264) (0.314) (0.265)
First School Located in a City 1.09 1.25** 1.26***(0.113) (0.129) (0.101)
First School Located in a Town/Rural Area 0.684*** 0.919 0.972(0.076) (0.113) (0.104)
First School is a Secondary School 0.902 0.791*** 1.22***(0.073) (0.069) (0.086)
Average Frac. Change in Enrollment in First School 1.36*** 1.37*** 1.03(0.117) (0.117) (0.108)
Local Unemployment Rate 0.968** 1.03** 0.967***(0.013) (0.013) (0.01)
Teacher is Highly Qualified 0.892 1.1 0.808***(0.076) (0.101) (0.062)
Teacher Age 0.981*** 0.999 1(0.003) (0.003) (0.003)
Teacher is Female 0.906 1.04 1.14**(0.056) (0.066) (0.064)
Teacher is Black 0.355*** 1.21 0.912
(0.065) (0.153) (0.108)Teacher is Hispanic 0.885 0.873 0.728
(0.231) (0.231) (0.164)Teaches Math or Science 1.21*** 1.06 1.16**
(0.086) (0.085) (0.083)Teaches Special Education 1.55*** 1.65*** 1.27***
(0.123) (0.141) (0.105)Teaches English as a Second Language 2.72*** 1.31 2.22**
(0.859) (0.518) (0.724)Teaches Art or Music 1.68*** 1.1 1.65***
(0.179) (0.14) (0.163)Teacher Attended Selective College 1.04 1.01 1.12*
(0.069) (0.075) (0.075)Teacher Attended College in Michigan 1.08 1.01 0.558***
(0.084) (0.082) (0.037)Teacher Has Master's Degree or Higher 1.25*** 0.802*** 0.161***
(0.075) (0.047) (0.009)Entered Teaching in 2004-05 0.922 0.979 1.05
(0.074) (0.085) (0.078)Entered Teaching in 2005-06 0.891 1 0.96
(0.076) (0.089) (0.076)Entered Teaching in 2006-07 0.929 0.928 0.714***
(0.086) (0.096) (0.064)Entered Teaching in 2007-08 0.722*** 0.904 0.692***
(0.076) (0.096) (0.067)Entered Teaching in 2008-09 1 0.692*** 0.686***
(0.115) (0.091) (0.074)Number of Observations 14860 14860 14860Pseudo R² 0.137 0.137 0.137Sample Mean 0.144 0.137 0.221
NOTES: Sample includes all teachers beginning their first jobs as full-time teachers between fall 2003 and fall 2008. Outcome variable is measured as of the teacher's fifth year, conditional on the teacher's first school still being open. "Disadvantaged" schools have at least 30% of their students eligible for free or reduced lunch in their first year of data. School/student characteristics are measured in the first school at the time of hire unless stated otherwise. Odds ratios reported. Standard errors are clustered at the school level. Sample includes one observation per teacher. Missing indicators are included for all variables that are ever missing.
Table 11: OLS Regressions of Movement into Disadvantaged Schools on School and Teacher Characteristics (Teachers Who Take New Teaching Jobs Only)
Dependent variable: indicator for having taken a new job in a disadvantaged school within a given number of years.Table Notes [1] [2] [3] [4]Frac. Free/Reduced Lunch Eligible 0.68*** 0.468*** 0.466*** 0.298***
(0.074) (0.086) (0.045) (0.047)Fraction of Black Students 0.242*** 0.222*** 0.14*** 0.02
(0.061) (0.068) (0.033) (0.04)Fraction of Hispanic Students 0.124 0.053 0.1 0.097
(0.193) (0.182) (0.131) (0.124)Frac. Limited English Proficiency Students -0.258* -0.13 -0.135* -0.081
(0.149) (0.112) (0.069) (0.068)Frac. Special Education Students 0.038 0.008 0.037 -0.018
(0.063) (0.071) (0.031) (0.034)Standardized Student Achievement - 2nd Quintile -0.066 0.034 -0.032 0.046
(0.052) (0.046) (0.031) (0.03)Standardized Student Achievement - 3rd Quintile -0.268*** -0.005 -0.15*** 0.022
(0.05) (0.055) (0.034) (0.033)Standardized Student Achievement - 4th Quintile -0.341*** 0.011 -0.232*** -0.01
(0.05) (0.061) (0.033) (0.036)Standardized Student Achievement - 5th Quintile -0.614*** -0.17** -0.447*** -0.158***
(0.043) (0.066) (0.031) (0.038)First School is a Charter -0.006 -0.155*** 0.056** -0.027
(0.058) (0.048) (0.026) (0.027)First School is Vocational -0.148 -0.433 0.006 0.032
(0.266) (0.302) (0.117) (0.106)First School is Special Education-Only 0.046 0.03 0.113** 0.11**
(0.091) (0.119) (0.044) (0.05)First School is Alternative 0.049 -0.058 0.014 -0.025
(0.088) (0.096) (0.052) (0.054)First School Located in a City 0.237*** 0.034 0.133*** 0.018
(0.041) (0.046) (0.023) (0.025)First School Located in a Town/Rural Area 0.145** 0.142** 0.096*** 0.089**
(0.07) (0.07) (0.035) (0.035)First School is a Secondary School -0.108*** -0.087** -0.086*** -0.049**
(0.038) (0.043) (0.02) (0.023)Average Frac. Change in Enrollment in First School -0.006 -0.001 0.0004 0.0003
(0.021) (0.021) (0.0003) (0.001)Local Unemployment Rate 0.037*** 0.018 0.023*** 0.014***
(0.006) (0.006) (0.003) (0.004)Teacher is Highly Qualified 0.038 0.028 0.052** 0.031
(0.043) (0.039) (0.025) (0.024)Teacher Age 0.001 -0.002 0.005*** 0.004***
(0.002) (0.002) (0.001) (0.001)Teacher is Female 0.017 0.025 0.039** 0.028
(0.034) (0.029) (0.018) (0.017)Teacher is Black 0.373*** 0.121*** 0.365*** 0.221***
(0.038) (0.042) (0.025) (0.027)
Within 2 Years Within 5 Years
Teacher is Hispanic 0.397*** 0.289** 0.045 -0.052(0.106) (0.117) (0.085) (0.076)
Teaches Math or Science -0.037 -0.016 -0.057** -0.045**(0.043) (0.041) (0.022) (0.02)
Teaches Special Education -0.06 -0.06 0.011 0.007(0.042) (0.037) (0.022) (0.021)
Teaches English as a Second Language -0.021 -0.229* -0.073 -0.138(0.137) (0.134) (0.096) (0.093)
Teaches Art or Music 0.018 -0.015 -0.102*** -0.103***(0.066) (0.056) (0.033) (0.033)
Teacher Attended Selective College -0.071* -0.059* -0.03 -0.013(0.039) (0.034) (0.02) (0.019)
Teacher Attended College in Michigan -0.025 -0.036 -0.007 -0.003(0.042) (0.039) (0.023) (0.021)
Teacher Has Master's Degree or Higher -0.129*** -0.071*** -0.138*** -0.102***(0.03) (0.026) (0.016) (0.016)
Entered Teaching in 2004-05 0.052 0.066 0.02 0.008(0.054) (0.045) (0.025) (0.023)
Entered Teaching in 2005-06 0.029 0.089** 0.006 0.033(0.053) (0.044) (0.026) (0.023)
Entered Teaching in 2006-07 0.013 0.019 0.007 -0.005(0.055) (0.045) (0.028) (0.026)
Entered Teaching in 2007-08 0.093 0.057 0.083*** 0.049*(0.06) (0.051) (0.03) (0.028)
Entered Teaching in 2008-09 0.067 -0.057 0.03 -0.073**(0.06) (0.058) (0.032) (0.032)
Number of Observations 1190 1190 4183 4183Adjusted R² n/a 0.283 n/a 0.158Sample Mean 0.538 0.488
NOTES: Sample includes all teachers beginning their first jobs as full-time teachers between fall 2003 and fall 2008 who take new teaching jobs within the relevant time span (two years in Columns [1] and [2], five years in Columns [3] and [4]). Outcome variable is whether the teacher is in a new job in a disadvantaged school two or five years after starting, conditional on the teacher's first school remaining open for five years. "Disadvantaged" schools have at least 30% of their students eligible for free or reduced lunch in their first year of data. School/student characteristics are measured in the first school at the time of hire unless stated otherwise. Standard errors are clustered at the school level. Differently-colored results in Column [1] and [3] come from separate regressions. Missing indicators are included for all variables that are ever missing.
Table 12: OLS Estimates of the Effect of School and Teacher Characteristics on Teacher Mobility, by SubgroupDependent variable: indicator for still being in one's first job after five years.Table Notes Highly Qualified Not Highly QualifiedFrac. Free/Reduced Lunch Eligible -0.084*** -0.074
(0.025) (0.049)Fraction of Black Students -0.155*** -0.194***
(0.023) (0.053)Fraction of Hispanic Students 0.171*** 0.011
(0.055) (0.117)Frac. Limited English Proficiency Students -0.137*** -0.224***
(0.032) (0.07)Frac. Special Education Students -0.063*** -0.071*
(0.023) (0.04)Standardized Student Achievement - 2nd Quintile -0.005 0.063*
(0.019) (0.033)Standardized Student Achievement - 3rd Quintile 0.008 0.025
(0.022) (0.039)Standardized Student Achievement - 4th Quintile 0.039* 0.04
(0.023) (0.042)Standardized Student Achievement - 5th Quintile 0.045* 0.035
(0.024) (0.045)First School is a Charter -0.035** -0.069**
(0.015) (0.029)First School is Vocational -0.052 0.111
(0.077) (0.073)First School is Special Education-Only 0.076 0.144***
(0.057) (0.042)First School is Alternative -0.151*** -0.026
(0.031) (0.054)First School Located in a City -0.033** -0.024
(0.016) (0.028)
First School Located in a Town/Rural Area 0.045** -0.011(0.019) (0.037)
First School is a Secondary School -0.014 0.048*(0.013) (0.028)
Average Frac. Change in Enrollment in First School -0.001* -0.006(0.001) (0.005)
Local Unemployment Rate 0.001 0.009***(0.002) (0.003)
Teacher is Highly Qualified
Teacher Age 0.001 0.0005(0.001) (0.001)
Teacher is Female -0.002 -0.012(0.01) (0.023)
Teacher is Black 0.016 -0.007(0.021) (0.031)
Teacher is Hispanic 0.034 0.03(0.041) (0.063)
Teaches Math or Science -0.021* 0.01(0.011) (0.045)
Teaches Special Education -0.097*** -0.05**(0.014) (0.023)
Teaches English as a Second Language -0.132** -0.092(0.064) (0.07)
Teaches Art or Music -0.093*** 0.011(0.016) (0.06)
Teacher Attended Selective College -0.02* -0.048*(0.011) (0.029)
Teacher Attended College in Michigan 0.049*** 0.051**(0.012) (0.021)
Teacher Has Master's Degree or Higher 0.159*** 0.106***(0.009) (0.022)
Entered Teaching in 2004-05 0.004 -0.023(0.013) (0.028)
Entered Teaching in 2005-06 0.015 -0.058**(0.014) (0.029)
Entered Teaching in 2006-07 0.03** 0.006(0.016) (0.035)
Entered Teaching in 2007-08 0.059*** 0.022(0.017) (0.033)
Entered Teaching in 2008-09 0.068*** -0.023(0.02) (0.037)
Number of Observations 13839 2654Adjusted R² 0.094 0.143Sample Mean 0.459 0.383
NOTES: Sample includes all teachers beginning their first jobs as full-time teachers between fall 2003 and fall 2008. Outcome variable is an indicator for whether the teacher is in the same job five years after starting, conditional on the teacher's first school remaining open. "Disadvantaged" schools have at least 30% of their students eligible for free or reduced lunch in their first year of data. School/student characteristics are measured in the first school at the time of hire unless stated otherwise. Standard errors are clustered at the school level. Missing indicators are included for all variables that are ever missing.
Master's Degree No Master's Degree Non-Disadvantaged Schools Disadvantaged Schools-0.118*** -0.026 -0.061 -0.058*
(0.032) (0.028) (0.05) (0.03)-0.195*** -0.127*** -0.278*** -0.121***
(0.031) (0.027) (0.042) (0.026)0.189*** 0.094 0.032 0.197***(0.071) (0.065) (0.219) (0.054)
-0.136*** -0.158*** -0.116* -0.147***(0.043) (0.035) (0.063) (0.034)-0.041 -0.088*** -0.019 -0.084***(0.028) (0.028) (0.032) (0.025)-0.003 0.028 -0.0001 0.005(0.024) (0.021) (0.041) (0.018)0.015 0.022 -0.083** 0.038*
(0.028) (0.023) (0.041) (0.022)0.055* 0.037 -0.037 0.061**(0.029) (0.026) (0.039) (0.027)0.047 0.062** -0.027 0.048
(0.029) (0.028) (0.041) (0.037)-0.038* -0.037** -0.044* -0.028(0.02) (0.016) (0.023) (0.018)0.115 0.022 0.071 -0.031(0.09) (0.086) (0.081) (0.106)
0.112** 0.118*** 0.128** 0.11**(0.044) (0.044) (0.06) (0.045)
-0.172*** -0.063* -0.184*** -0.106***(0.037) (0.037) (0.044) (0.036)-0.003 -0.061*** 0.013 -0.049***(0.02) (0.018) (0.023) (0.018)
0.028 0.043* -0.01 0.078***(0.023) (0.023) (0.023) (0.025)0.014 -0.032** -0.008 -0.005
(0.016) (0.015) (0.017) (0.016)-0.007* -0.0004** -0.017** -0.001**(0.004) (0.0002) (0.008) (0.0004)-0.0001 0.006*** 0.008*** 0.0005(0.002) (0.002) (0.003) (0.002)0.042** -0.002 0.024 0.025(0.017) (0.016) (0.018) (0.015)-0.001 0.002*** 0.002** -0.0001(0.001) (0.001) (0.001) (0.001)0.012 -0.017 -0.017 0.013
(0.012) (0.013) (0.013) (0.012)0.011 0.019 0.028 0.017
(0.025) (0.022) (0.041) (0.02)0.06 0.002 0.087 -0.002
(0.047) (0.046) (0.055) (0.043)-0.011 -0.036** -0.013 -0.027*(0.015) (0.016) (0.016) (0.015)
-0.098*** -0.072*** -0.101*** -0.076***(0.016) (0.018) (0.017) (0.017)
-0.158** -0.087 -0.164* -0.098*(0.071) (0.063) (0.09) (0.055)
-0.099*** -0.08*** -0.08*** -0.093***(0.025) (0.02) (0.025) (0.02)-0.017 -0.024 -0.019 -0.027*(0.013) (0.017) (0.014) (0.015)0.032** 0.07*** 0.042** 0.056***(0.015) (0.014) (0.017) (0.014)
0.188*** 0.12***(0.013) (0.011)
-0.003 0.002 0.03* -0.029*(0.015) (0.018) (0.018) (0.016)
-0.013 0.036** 0.014 -0.004(0.016) (0.018) (0.018) (0.017)-0.005 0.07*** 0.029 0.018(0.019) (0.02) (0.021) (0.019)0.016 0.095*** 0.054** 0.046**(0.02) (0.021) (0.024) (0.02)0.013 0.086*** 0.015 0.061**
(0.024) (0.024) (0.027) (0.024)9552 6992 7925 86190.069 0.079 0.08 0.0920.531 0.331 0.525 0.375
White Teachers Black Teachers Male Teachers Female Teachers-0.085*** -0.072 -0.126*** -0.071***
(0.024) (0.063) (0.04) (0.025)-0.163*** 0.007 -0.109*** -0.178***
(0.024) (0.082) (0.036) (0.025)0.139** 0.598*** 0.092 0.171***(0.059) (0.203) (0.095) (0.055)
-0.145*** -0.205 -0.105* -0.162***(0.03) (0.129) (0.06) (0.032)
-0.071*** -0.055 -0.001 -0.085***(0.023) (0.065) (0.036) (0.024)0.007 0.038 0.013 0.003
(0.019) (0.039) (0.028) (0.019)0.009 0.053 0.032 0.004
(0.021) (0.054) (0.035) (0.022)0.037* 0.032 0.053 0.035(0.022) (0.092) (0.035) (0.023)0.045* 0.166** 0.075** 0.038(0.023) (0.084) (0.037) (0.024)
-0.043*** -0.044 -0.041 -0.037**(0.015) (0.039) (0.025) (0.016)0.082 0.114 0.228*** -0.047
(0.075) (0.091) (0.083) (0.09)0.12*** 0.143 0.024 0.129***(0.039) (0.136) (0.06) (0.039)
-0.129*** -0.096 -0.109*** -0.124***(0.029) (0.086) (0.041) (0.034)-0.029* -0.05 -0.034 -0.033**(0.016) (0.04) (0.024) (0.016)
0.035** -0.538*** 0.046 0.03(0.017) (0.147) (0.029) (0.02)-0.011 0.049 -0.006 -0.011(0.012) (0.044) (0.019) (0.014)-0.001 -0.013** -0.022*** -0.001*(0.001) (0.006) (0.006) (0.0005)0.002 0.004 -0.0001 0.003*
(0.002) (0.005) (0.003) (0.002)0.018 0.087*** -0.001 0.033**
(0.013) (0.032) (0.022) (0.014)0.001 -0.001 -0.002** 0.001***
(0.0005) (0.001) (0.001) (0.001)-0.006 0.042(0.009) (0.03)
-0.031 0.025(0.032) (0.019)0.012 0.04
(0.062) (0.039)-0.015 -0.067* -0.011 -0.023*(0.011) (0.038) (0.018) (0.014)
-0.096*** -0.064 -0.074*** -0.094***(0.013) (0.04) (0.027) (0.013)-0.094* -0.051 -0.045 -0.145***(0.056) (0.066) (0.114) (0.052)
-0.085*** -0.091 -0.066** -0.098***(0.017) (0.067) (0.029) (0.018)
-0.023** 0.021 -0.015 -0.027**(0.011) (0.044) (0.019) (0.012)
0.046*** 0.095*** 0.04* 0.053***(0.011) (0.03) (0.021) (0.012)
0.153*** 0.113*** 0.137*** 0.157***(0.009) (0.034) (0.016) (0.01)0.004 -0.056 0.011 -0.006
(0.013) (0.036) (0.022) (0.014)
0.005 -0.0001 0.012 0.0004(0.013) (0.041) (0.024) (0.014)0.029* -0.008 0.067** 0.008(0.015) (0.049) (0.027) (0.016)
0.056*** -0.002 0.062** 0.048***(0.016) (0.054) (0.028) (0.018)
0.059*** 0.002 0.075** 0.042**(0.019) (0.066) (0.033) (0.02)14933 1139 4321 122230.091 0.111 0.117 0.10.462 0.291 0.447 0.446
Bottom Achievement Quintile Achievement Quintiles 2-4 Top Achievement Quintile-0.102** -0.135*** 0.036(0.044) (0.033) (0.096)
-0.197*** -0.125*** -0.217*(0.047) (0.03) (0.113)0.151 0.173*** -0.152
(0.107) (0.061) (0.365)-0.253*** -0.12*** 0.03
(0.056) (0.04) (0.109)-0.051 -0.07** -0.028(0.041) (0.033) (0.055)
-0.038 -0.033* 0.008(0.032) (0.02) (0.035)
0.514*** -0.376***(0.077) (0.034)0.056 -0.052
(0.087) (0.145)-0.054 -0.211*** -0.184*(0.046) (0.051) (0.107)-0.023 -0.038 -0.015(0.029) (0.023) (0.032)
0.03 0.063*** -0.049(0.077) (0.024) (0.051)-0.008 -0.047*** 0.077**(0.03) (0.017) (0.032)
-0.015*** -0.004** -0.015**(0.005) (0.002) (0.006)-0.001 0.002 0.003(0.003) (0.003) (0.006)0.054** 0.019 0.042(0.026) (0.018) (0.033)0.001 0.0002 -0.00004
(0.001) (0.001) (0.001)0.039* -0.014 -0.025(0.021) (0.014) (0.023)0.017 0.002 0.022
(0.027) (0.029) (0.072)0.066 -0.079 0.05
(0.069) (0.055) (0.106)-0.043 -0.008 -0.011(0.029) (0.016) (0.03)-0.06* -0.092*** -0.107***(0.036) (0.017) (0.029)
0.02 -0.192*** -0.388***(0.072) (0.072) (0.116)
-0.134*** -0.081*** -0.047(0.036) (0.022) (0.043)-0.037 -0.024 -0.001(0.033) (0.015) (0.023)0.05** 0.048*** 0.074***(0.023) (0.016) (0.028)
0.094*** 0.143*** 0.176***(0.02) (0.012) (0.022)-0.025 0.009 0.048(0.026) (0.018) (0.034)
-0.037 0.012 0.011(0.033) (0.02) (0.036)-0.007 0.036* 0.015(0.035) (0.02) (0.034)0.032 0.064*** 0.026
(0.035) (0.021) (0.04)0.066 0.068*** 0.032
(0.044) (0.026) (0.044)2468 7596 26690.08 0.076 0.061
0.281 0.442 0.537
Recommended