Performance Screens for School Improvement:

The Case of Teacher Tenure Reform in New York City

Susanna Loeb
Stanford University

Luke C. Miller
University of Virginia

James Wyckoff
University of Virginia

October 2014

Abstract
Tenure is intended to protect teachers with demonstrated teaching skills against
arbitrary or capricious dismissal. Critics of typical tenure processes argue that tenure
assessments are superficial and rarely discern whether teachers in fact have the
requisite teaching skills. A recent reform of the tenure process in New York City
provides an unusual opportunity to learn about the role of tenure in teachers’ career
outcomes. We find the reform led to many fewer teachers receiving tenure. Those
not receiving tenure typically had their probationary periods extended to allow them
an opportunity to demonstrate teaching effectiveness. These “extended” teachers
were much more likely to leave their schools and be replaced by a teacher who was
judged to be more effective.

Thanks to Joanna Cannon, Anne-Marie Hoxie and Keely Alexander at the New York
City Department of Education for providing the data employed in this paper and for
answering questions about the NYCDOE tenure policy. We benefited from
comments by Arthur Mckee, Doug Harris and three anonymous reviewers on earlier
drafts of this paper. We appreciate financial support from the National Center for
the Analysis of Longitudinal Data in Education Research (CALDER). CALDER is
supported by IES Grant R305A060018. The views expressed in the paper are solely
those of the authors. Any errors are attributable to the authors.
“Tenure marks a new phase in a teacher’s career and a new commitment by our schools to
those who receive it. Unfortunately, over the years tenure has become an expectation more
than an honor. While we have made progress, we still are not doing enough to set a high
bar for all teachers, recognize excellent teachers, or withhold tenure from all of those who
have not earned it. And a loose tenure system isn’t good for anyone—it hurts students, it
disrespects successful teachers, and it leaves those who are not up to the difficult job to
struggle.” —New York City School’s Chancellor Klein, 2009.

Introduction

Teacher tenure has been controversial since the first tenure provisions were

enacted over a century ago. Proponents typically argue that tenure prevents teacher

dismissal for political purposes or due to capricious decisions by administrators or

politicians. Tenure could guard against dismissal of more experienced, higher paid

teachers during periods of tight budgets when school leaders may be more focused

on reducing costs while meeting class size requirements than they are on student

learning. Tenure does not require schools or districts to retain ineffective teachers

but instead provides a due process mechanism to dismiss tenured teachers for cause.

Critics, however, argue that the costs of due process do, in practice, lead districts to

retain ineffective teachers and as a result tenure not only allows poor teachers to stay

in the classroom but also reduces the incentive for teachers to be as effective as they

could be. They argue that the due process mechanisms for removing teachers with

tenure are so burdensome that they rarely are pursued.

With the availability of large-scale student performance measures linked over

time has come clear evidence that teachers vary substantially in their effectiveness at

2
improving student test performance1 and that these differences can have meaningful

effects on students in both the short run and the long run (Chetty, Friedman, &

Rockoff, 2014; Rivkin, Hanushek, & Kain, 2005; Rockoff, 2004). At least partially as

a result, education reforms in the US recently are focusing on improving the quality

of teaching through human resource policies such as improved evaluation systems

and differentiated pay.2 Given the controversial nature of teacher tenure, it is not

surprising that interest also has increased in changing teacher tenure provisions so

that the due process is less onerous and so that school leaders have greater control

over their workforce. Yet, the evidence on which to base reform decisions is scarce.

We know little about what types of tenure provisions improve the quality of teaching

and what types do not. Similarly, we know little about how long the probationary

period prior to tenure should be, if there is tenure, in order for school systems to

accurately assess teachers’ effectiveness so that they can make well informed

decisions about tenure.

Part of the reason that we have little evidence on the effects of tenure is that

until recently tenure laws have been relatively stable over time and similar, though

not the same, across states. New Jersey passed the nation’s first teacher tenure law in

1909. Over the next several decades other states adopted similar laws: New York in

1917, California in 1921, and Michigan, Pennsylvania, and Wisconsin in 1937. The
1 The research literature on teacher value-added as valid and reliable measures of teacher effectiveness
is not yet settled, as some argue that these measures may be biased (see for example, Rothstein, 2009).
2 Rockoff, Staiger, Kane, & Taylor (2012) examine how increased information can improve teacher

performance and the composition of the teaching workforce. Rothstein (forthcoming) examines how
the supply of teachers matters for improving teacher effectiveness in the context of tenure decisions.

3
state statutes use a variety of synonyms for tenure: continuing contract or service,

permanent status, career status, and post-probationary status. Regardless of the

terminology, these laws have three main components: tenure requirements, reasons

for dismissal, and a process for appeals. The first specifies the length of the

probationary period after which teachers are eligible for tenure. Employers can

dismiss a non-tenured teacher at any time for any reason so long as the decision is

neither arbitrary or capricious nor discriminatory, but tenured teachers can only be

dismissed for the reasons provided in the law. The third component details the

appeals process a dismissed tenured teacher can pursue in an effort to be reinstated.

Of the 48 states in which public elementary and secondary teachers are awarded

tenure, the minimum probationary period exceeds three years in 11 states (National

Council on Teacher Quality, 2012). In most states it is three years, although in a few

states, such as California, teachers typically receive tenure with fewer than three years

of experience.

Tenure laws in the US have been the focus of significant legislative action in

statehouses across the country since the beginning of the 21st Century. In 2000,

Georgia eliminated due process rights for teachers hired after 1 July 2000, but

reinstated these rights three years later. Florida eliminated teacher tenure in 2011.

That same year Idaho enacted a law that would have eliminated teacher tenure had it

not been repealed by voters the following year. Voters in South Dakota turned back

an effort to repeal a 2012 law thereby allowing a law eliminating tenure after 1 July

4
2016 to take effect. North Carolina’s governor signed a bill into law last year that

eliminates teacher tenure by 2018. Though almost all states currently grant tenure,

more than half now require meaningful evaluation during the tenure review process.

As an example, in 2009 only four states used student test performance as a criterion

for tenure; by 2012, 20 states did and 25 states require multiple categories for

teachers in their evaluation, not just satisfactory and unsatisfactory (National Council

on Teacher Quality, 2012).

The debates around teacher tenure have taken on a new intensity as the locus

for those debates has shifted to the courthouse. The 2014 court ruling in Vergara v.

California found elements of California’s tenure provisions to be unconstitutional

setting California on a course to eliminate teacher tenure unless the state’s appeal is

successful. Similar cases are being brought in other states, but it remains unknown

whether strategies to eliminate tenure will ultimately be successful and whether

revisions to tenure policies can achieve the key goals of reformers.

A recent reform by the NYCDOE provides an unusual opportunity to learn

about the role of tenure in teachers’ career outcomes including both strategic

retention on the district side and choice-based retention stemming from teachers’

decisions. While not nearly as provocative as the elimination of tenure, modest

reforms to the tenure process may produce many of the outcomes raised by plaintiffs

in Vergara and other court cases. Knowing more about the role of better

information or longer probationary periods in the composition and effectiveness of

5
the teacher workforce could substantially enlighten the discussion over tenure. In

what follows, we start by describing the New York City reform. We then use data

from NYCDOE and the New York State Education Department (NYSED) to

provide initial evidence on the magnitude of responses to the reform, concluding

with a discussion of the results. We find that the NYCDOE tenure reforms extended

the probationary periods of teachers judged to be less effective and that these

“extended” teachers were substantially more likely than other teachers to leave their

schools. Their likely replacements were typically more effective as judged by their

principals and as measured by value-added. These changes disproportionately

benefited schools with larger percentages of black students.

The Teacher Tenure Process in New York City

The criterion for tenure in New York City is that a teacher possesses

“significant professional skill and a meaningful, positive impact on student learning.”

This criterion is not new. However, prior to academic year (AY) 2009-10 the tenure

process in New York City was similar to that in many other large urban districts. The

receipt of tenure had become an expectation for nearly all teachers and frequently

was based on little evidence of accomplishment. In 2007-08 and 2008-09, well into

the period of accountability reforms, 94 percent of all eligible teachers who stood for

tenure decision were approved for tenure.3

3As is true elsewhere, a portion of probationary teachers leave NYCDOE prior to a tenure decision.
Some of these teachers are being “counseled out” and likely would not have been approved for tenure

6
 Beginning in 2009-10, New York City changed the tenure review process,

infusing more information and increasing the responsibility and accountability of

principals to insure that teachers met challenging performance standards. Tenure

decisions in 2009-10 were informed by sources of information that had been

available previously: classroom observations, evaluations of teacher work products

including lesson plans, and the annual rating sheet that principals completed giving

teachers a Satisfactory, Doubtful, or Unsatisfactory rating. In addition, tenure

decisions in 2009-10 included new student learning measures from the Teacher Data

Reports (which included teacher value-added), in-class assessments aligned with the

New York State standards, and other evidence of student progress (NYCDOE,

2009).

As in previous years, principals sent recommendations to the superintendent

about whether a teacher should be denied tenure, have their probationary period

if they had completed the tenure review process. We have no way of credibly identifying the extent of
this practice or their potential tenure outcomes. However, we get a rough sense of the effect of this
attrition on tenure approval rates by comparing the value-added of teachers who exit prior to a tenure
decision to those who remain. Using the distribution of value-added of teachers with the same level of
teaching experience who persist to a tenure decision, we identify the value-added score of the teacher
at the 6th percentile (the percentile implied by a 94 percent tenure approval rate). We then calculate the
percentage of leaving teachers whose value-added score falls below that score. As might be expected
based on prior research, we find that teachers who exit during their first two years have somewhat
lower value-added than their cohort colleagues who remain (Boyd et al., 2008; Goldhaber et al., 2011),
but no more than 8 percent of leavers fall below the cutoff score, suggesting that 92 percent of the
teachers who exit prior to a tenure decision would have been granted tenure pre-reform. Given that
70 percent of teachers in our sample persist to a tenure decision, the net effect on tenure approval
rates is very small, resulting in an approval rate greater than 93 percent. This result is at best an
approximation. Value-added is a noisy measure, so our estimates should not be viewed as precise.
Principals were probably unaware of teacher value-added and thus counseling out decisions were
made employing other measures unavailable to us. However, research suggests that even in the
absence of value-added data principals identify most of the same low-performing teachers as does a
value-added metric (Jacob and Lefgren, 2008).

7
extended or be granted tenure, but starting in 2009-10 principals had to provide a

rationale for this decision if the evidence available at the district level suggested

either a strong case to approve or deny tenure and this information ran counter to

the principal’s recommendation. The district provided principals with tenure

guidance for teachers for whom there was evidence that performance was

particularly strong or weak. For a teacher whose value-added results had been in the

lowest 50 percent over the past two years (with a 95 percent confidence interval),

who had previously received an Unsatisfactory annual rating, or whose tenure

decision had previously been extended, the principal received guidance from the

district that the teacher should be considered to have “tenure in doubt”. A principal

recommendation to extend or approve tenure for these teachers required a

supporting rationale for the superintendent to consider in his or her review. The

principal received guidance of “tenure likely” for a teacher whose value-added results

had been in the highest 50 percent over the past two years (with a 95 percent

confidence interval). Principals recommending denying tenure or extending the

probationary period for these teachers similarly needed to provide supporting

evidence to the superintendent (NYCDOE, 2009).

The process introduced in 2009-10 remained in place in 2010-11 with some

notable changes (NYCDOE, 2010). New in 2010-11, principals were asked to

evaluate all teachers up for a tenure decision based a four-point effectiveness rating

scale (Highly Effective, Effective, Developing, and Ineffective) as described in the

8
district-developed Effectiveness Framework.4 As in the prior year, the evidence for

these ratings came from measures of the teacher’s impact on student learning such as

value-added measures from the Teacher Data Reports, student work products, and

tests aligned to the New York State standards. Principals also could use evidence

from measures of instructional practice coming from their own classroom

observations, teacher work products, and the annual rating sheet that principals

complete for each teacher.5 In addition to these sources of information, which were

available in the prior year as well, principals in 2010-11 gained information about

professional contributions from surveys of students and parents, from measures of

attendance, from colleague feedback, and from work products related to the

Comprehensive Educational Plan for each school. In contrast to 2009-10, principals

in 2010-11 no longer received “tenure likely” or “tenure in doubt” guidance from the

district but rather were given flags indicating a “low value add” teacher as an “Area

of Concern” and a “high value add” teacher as a “Notable Performance”. Low and

high value-added scores were defined as in the previous year. Other problematic

teacher behaviors flagged as Areas of Concern included: low attendance (defined as

exceeding 20 days in the previous two fiscal years), an Unsatisfactory or Doubtful

4 These effectiveness ratings are distinct from the ratings built into the new statewide teacher
evaluation system which was not implemented until 3 years later in 2013-14. Although they use the
same ratings scale, both the evidence synthesized and the relative weight assigned to the evidence
differs between the two.
5 These sources of evidence were employed in 2009-10 tenure decisions but they were not aggregated

in the effectiveness ratings.

9
rating on a prior Annual Review Sheet, having been previously extended, having

been previously excessed or currently in the Absent Teacher Reserve pool.6

The tenure review process for 2011-12 was very similar to that in 2010-11,

but with two important changes. As before, teachers were evaluated on their impact

on student learning, instructional practice, and professional contributions. Principals

were provided guidance as to the expected (though not required) alignment between

the effectiveness ratings they determined using the Effectiveness Framework and

their tenure recommendations: Highly Effective and Effective ratings were evidence

in favor of granting tenure; a Developing rating, evidence for an extension; and an

Ineffective rating, evidence for denying tenure. Additionally, responsibility for

producing teacher value-added estimates shifted from the district to the NYSED

beginning with 2010-11 and no measures were available for principals to incorporate

them into their 2011-12 tenure decisions (NYCDOE, 2011).

The state-provided value-added estimates did inform principals’ 2012-13

recommendations. Teachers received a growth score (0-20) that corresponded to a

HEDI rating (Highly Effective, Effective, Developing, and Ineffective). No explicit

guidance was provided to principals as to how to incorporate these growth ratings

into their tenure recommendations. They were only told these ratings were a source

of evidence for a teacher’s impact on student learning.

6 The Absent Teacher Reserve pool is composed of teachers who previously taught in NYCDOE but

who currently don’t have a permanent teaching assignments. These teachers continue to be paid and
can be called upon to teach if vacancies arise.

10
Screening Employees to Improve Quality

How might we expect principals and teachers to respond to the New York

City tenure policy changes? Consider a principal who aims to improve the quality of

instruction in her school in the short run. Principals are likely to have this goal for

numerous reasons, including the high rate of principal mobility in urban school

districts (Beteille, Kalogrides, and Loeb, 2012), the substantial accountability pressure

on principals, and the principal training and selection mechanisms in place in NYC.7

To be more specific, we assume that quality of instruction in school s in time t

is a function of the underlying quality of the teachers , of some form of

investment that teachers make to improve their performance and the

performance of their peers, and of other aspects of the school .

, ,

A principal will choose to deny, extend, or grant tenure to teachers in order

to maximize school quality. However, a priori it is not clear how their choice of

tenure for a given teacher will affect who teaches in the school, nor how it will affect

teachers’ investment in their work. These dynamics depend upon how teachers

respond to the decisions both about themselves and about other teachers in the

school. If teachers do not change their choices in response to tenure decisions then

the principal’s decision is straightforward – deny teachers whom they want to leave

7 See Corcoran, Schwartz and Weinstein (2009) as well as descriptions of accountability in NYC at

http://schools.nyc.gov/Accountability/tools/accountability/default.htm

11
the school and extend all other teachers. There would be no reason to recommend

tenure for any teacher since this decision limits the principal’s future actions and

provides no benefits in the present.

However, there are reasons to believe that teachers do care about tenure. As

an example, Brunner and Imazeki (2010) find that districts with longer probationary

periods compensate teachers more, other things equal, indicating that teachers value

shorter probationary periods and are willing to accept somewhat reduced

compensation in return. This finding is consistent with evidence from other

occupations that jobs with greater risk – greater physical risk, or greater risk of

unemployment, or less predictable salaries – have to pay more in order to attract the

same quality workers (Feinberg, 1981). These results lead us to assume that, in

deciding whether or not to stay in a job, teachers value job stability and thus care

about whether or not they have tenure.

Teachers are also likely to value being recognized for strong performance as

well as simply gaining utility from strong performance itself (Kalleberg, 1977).

Evaluation systems that differentiate among employees are likely to increase the

satisfaction of teachers who perform well and may increase their willingness to stay

in the school. For those, on the border line of a positive review, this evaluation may

also increase their interest in investing in their own improvement and the

improvement of their peers. The same differentiation is likely to have negative

12
consequences for the satisfaction of teachers who perform worse on the evaluations,

perhaps decreasing their interest in staying and investing.

Finally, teachers care about the pecuniary and non-pecuniary job attributes

that informed their rationale for becoming a teacher (Flyer and Rosen, 1997). Some

of these characteristics – in particular, the climate of the school such as the attitudes

of teachers and their willingness to support other teachers in the school – may

change as a result of the implementation of tenure reform in each school, thereby

serving as a mediator for the reform’s impact on instructional quality. For some

teachers, the benefits of being a teacher far outweigh those of alternative

occupations, while for others the choice was less clear.

Learning to become an effective teacher is complicated and dependent not

only on the teacher’s ability and motivation but also on prior preparation and the

opportunities, culture, and support they receive in their school. As a result,

identifying teachers as approved, extended, or denied for tenure inevitably means

that each category contains a heterogeneous grouping of teachers who find

themselves in that group for a variety of reasons. Increasing the rigor of

performance screens is intended to reduce this heterogeneity so that teachers

approved for tenure are effective. However, like human resources management in

any profession, this is not an exact science, implying that misidentification occurs

and that some teachers not approved for tenure could have been approved if the

measures were more nuanced or their schools more conducive to their growth.

13
 In shaping the quality of teaching in their schools, principals’ tenure

recommendations are likely informed by their perceptions of how teachers will

respond to tenure outcomes and on the effect of teacher turnover on student

achievement (Ronfeldt, Loeb, & Wyckoff, 2013). Are teachers more likely to

improve or leave their school in response to being extended? Given the dramatic

effect of NYCDOE tenure reform on the percentage of teachers extended, in this

paper we focus on the retention decisions of extended teachers. However, the

reform could well influence other teacher decisions. For example, does the reform

affect whether individuals are attracted to teaching in New York or how much

investment teachers make in their teaching effectiveness? Does it influence the

retention decisions of teachers who are approved for tenure? Conceptually, tenure

reform could influence each of these directly or through changes in the school

learning environment. These decisions are potentially important but in this paper we

focus solely on the effects of tenure reform on teacher retention due to data

limitations.

Consider those teachers who would have received tenure pre-reform but are

extended post-reform. These teachers now receive negative signals about their

teaching with an increased potential of being denied tenure. Each of these increases

their likelihood of leave teaching, particularly at their current school. However, for

some teachers who are extended the appeal of teaching will be high enough to lead

them to choose to stay in teaching. For this subgroup, an increase in investment may

14
increase their probability of future recognition and receipt of tenure. This group may

increase investment as a result of reforms if they assess that this increase will likely

lead to tenure in the future. Understanding the potential effects of tenure reform on

teacher behavior, principals will be more likely to extend teachers who are not

performing well if they believe they can hire more effective teachers. Because school

principals play a central role in the process and because the teacher workforce differs

across schools, we might expect the changes to differ across schools. In keeping with

these potential effects, we address the following three research question in this

paper:

1. Tenure Decisions – How did tenure rates change following reform?

2. Workforce Composition – Of teachers who become eligible for tenure, how

did the composition of those continuing to teach in NYC change following

reform?

3. School Differences – How have schools varied in their tenure decisions and

the subsequent behaviors of their teachers?

Data

In order to assess the effects of NYCDOE tenure reforms, we must

accurately identify teachers eligible for tenure, as well as other teachers potentially

affected by the changes. The Tenure Notification System (TNS) tracked the tenure

review process for all probationary teachers in New York City public schools

between 2007-08 and 2012-13. Each school year, the district made tenure decisions

15
for teachers whose probationary period was scheduled to conclude between

November 1st of the current school year and October 31st of the following school

year. The probationary periods for the 2009-10 cohort, for example, concluded

between November 1, 2009 and October 31, 2010. The TNS provided principals

with a list of teachers at their school eligible for tenure as well as all official guidance

concerning each teacher’s job performance prior to the current year (e.g., prior

Unsatisfactory annual performance ratings, low attendance, value-added

classification, etc.). Principals enter their preliminary and final ratings and

recommendations into the TNS and district superintendents make and record final

tenure decisions in the system.

We assembled additional information on all teachers, not just those in the

TNS, from a variety of sources. NYCDOE provides basic teacher demographic

characteristics, the value-added calculations for 2008-09 and 2009-10, the state’s

value-added calculations for 2011-12, and annual performance ratings used in the

tenure review process. We identify teachers’ pathways into the teaching profession

from state certification records and rosters for the New York City Teaching Fellows

program and Teach for America corps members in the New York City region. State

certification files provide scores on certification exams. From the College Board we

obtain teachers’ SAT scores for those teachers who attended a New York public

school from 1980 to 2008 or a New York private school from 1980 to 2001.

Characteristics of the schools in which teachers teach (e.g., race/ethnicity,

16
free/reduced-price lunch eligibility, AYP status, etc.) come from the annual state-

level School Report Cards database and Institutional Master Files and the federal

Common Core of Data.

Finally, leveraging data from the NYCDOE Teacher Data Initiative, we

observe characteristics of the students taught by specific teachers of grades 4

through 8 mathematics and English language arts (ELA) including demographic and

achievement information. We use these data to estimate our own value-added

measures of teacher effectiveness using the residuals-based approach described in the

technical appendix that controls for individual student, classroom, and school

characteristics.8 Currently, 2010-11 is the final year for which we can calculate these

estimates.

Just over three quarters of the teachers in our post-reform sample are female,

approximately 18 percent are black, and 17 percent are Hispanic. They have average

math and verbal SAT scores of approximately 500 points each. Approximately half

of the teachers entered teaching through traditional teacher preparation programs

that recommended certification, while 22 percent came through the Teaching

Fellows Program, the largest early-entry route serving New York City. These

teachers work at schools where 44 percent of students are Hispanic students and 31

8 We employ our own value-added estimates rather than those estimated by NYCDOE primarily

because NYCDOE estimates are not available for all of the years of our analysis and because we
employ estimates that have been adjusted for measurement error through an empirical Bayes
shrinkage approach.

17
percent are black, with 67 percent eligible for subsidized lunch (see appendix Table

A1).

Recall that principals complete an Annual Rating Sheet for each teacher. Just

2.3 percent of teachers in our sample received an Unsatisfactory rating and one tenth

of one percent of teachers received a Doubtful rating, with the remaining 97.6

percent receiving a Satisfactory rating. On the four-point effectiveness rating scale

assigned by their principals, most teachers received either a Developing (29 percent)

or an Effective (41 percent) rating, while 17 percent received Highly Effective and

two percent received Ineffective ratings. Principals provided no effectiveness rating

for 11 percent of teachers. Eight percent of teachers had what the district classified

as low attendance (more than 20 absences over prior two years), and 12 percent had

low value-added.

Results

Tenure Decisions

As described in Figure 1, 94 percent of teachers were approved for tenure

during AY 2007-08 and 2008-09, the two years prior to the introduction of the

policy. The approval rate dropped to 89 percent in the first year of the policy (2009-

18
10) and averaged 56 percent in the three subsequent years.9 Virtually all of the

decrease in the tenure approval rate resulted in an increase in the percentage of

teachers whose probationary periods were extended, which averaged less than 4

percent prior to the policy, but 41 percent in 2010-11 through 2012-13. The

percentage of teachers denied tenure increased marginally following the introduction

of the new tenure review process from an average of two percent pre-policy to three

percent post-policy.

Principals have played an important role in the determination of tenure

decisions. Principal effectiveness ratings of teachers using the Effectiveness

Framework are highly predictive of tenure outcomes under the new policy. Ninety-

four percent of teachers rated Highly Effective and 83 percent of those rated

Effective were approved for tenure. In contrast, less than two percent of those rated

Developing and less than one percent of those rated Ineffective were approved. The

vast majority (97 percent) of teachers rated Developing were extended, while the vast

majority (81 percent) of those rated Ineffective were denied tenure. Given that

almost all teachers were approved for tenure prior to the reform, many teachers who

9 Given the new policy, it is reasonable to wonder whether teachers voluntarily exited at higher rates
prior to a tenure decision, thus understating the potential effects of the policy. We do not know the
rationale for teacher attrition and thus cannot assess this question directly. However, because we have
value-added for teachers in tested grades and subjects prior to tenure decisions, both before and after
the tenure reform, we can explore whether the reform induced a change in attrition with respect to
value-added. As shown in appendix Table A2, of the two years for which we have post-reform value-
added data and for the first two years of a teacher’s probationary period the only significant pre-
versus-post-reform difference we see is for first-year teachers in 2010. In that case, the value-added of
probationary teachers who exit post-reform is lower than that of probationary teachers leaving pre-
reform, suggesting the pre-versus-post-reform change in the percentage of teachers approved for
tenure would have been even larger if these teachers had remained to stand for tenure.

19
would have been approved prior to the reform received a different outcome under

the new system.

Tenure decisions also correspond with other teacher performance measures.

For teachers in tested grades and subjects, value-added estimates track tenure

decisions. Teachers denied tenure have math value-added estimates that are a full

standard deviation in teacher effectiveness lower than those approved for tenure. On

average, extended teachers are 13 percent of a standard deviation in student

achievement less effective than the average teacher and 38 percent of a standard

deviation less effective than those who are approved. Value-added differences in

ELA are smaller but demonstrate the same pattern. Similarly, extended teachers are

far more likely to have had prior Unsatisfactory or Doubtful annual performance

ratings and to have had low attendance than are teachers approved for tenure (see

appendix Table A3).

Overall, the reforms dramatically reduced the percentage of teachers who

received tenure, but because most teachers who became eligible for tenure were

extended and not dismissed it is unclear a priori whether the reform meaningfully

altered the workforce.

Workforce Composition

Changes in the tenure process can affect the quality of teaching by denying

tenure to less effective teachers. As discussed, denied teachers had lower value-added

20
in both math and ELA than teachers who were extended or approved. However,

even under the new policies, few teachers are dismissed. Larger changes in the

workforce instead may come from changes in voluntary turnover, particularly of

teachers who are extended.

Extended teachers may voluntarily exit from New York City schools,

creating vacancies which can be filled by more effective teachers. We find some

evidence of this phenomenon. Extended teachers were more likely to transfer to

other New York City schools and exit teaching in New York City public schools

altogether in the year following their decision than teachers who were approved for

tenure. Ninety percent of approved teachers return to their schools, while only 75

percent of extended teachers did so.

Being extended appears to meaningfully increase the likelihood of transfers

and exits as extended teachers are more likely to leave even after controlling for

teacher and school characteristics. Table 1 shows regressions with controls for the

final principal effectiveness rating of the teachers and includes school fixed effects,

allowing us to examine the effect of being extended on teacher transfer and exit

behavior for teachers in the same schools with equivalent principal ratings. The

probability of transferring increases by 9 percentage points if the teacher had been

extended rather than approved. This represents a 50 percent increase in the probability

of transferring following a tenure decision. Similarly, extended teachers exit NYC

public schools at a rate that is 4 percentage points higher than approved teachers,

21
holding other factors constant. This represents a 66 percent increase in the probability

of exiting. These results provide suggestive evidence that the new tenure process is

having an effect on the composition of the teaching workforce even without

substantially increasing the percentage of teachers directly denied tenure. However,

we cannot rule out that other factors correlated with a teacher’s Extended status may

account for the increases in transfer and exit rates.

Among extended teachers, those who remain in the same school have

somewhat different measured attributes than those who transfer or exit the system.

Teachers with higher academic qualifications, such as teacher certification exam

scores, are less likely to stay in the same school than to exit. Extended teachers

entering through alternative routes such as the New York City Teaching Fellows

program or Teach for America are less likely to remain in the same school than

teachers entering through college recommended programs. In contrast, the average

value-added estimates of extended teachers who remain in the same school are

higher than those who do not, but the sample sizes are smaller for these measures

and the differences are not statistically significant at traditional levels (see appendix

Table A4).

The voluntary attrition of these less effective teachers would benefit students

only if they are replaced by relatively more effective teachers. We explore this

question by comparing the effectiveness of teachers who were extended and left

22
schools in 2010-11 or 2011-12 with teachers hired at these schools.10 Unfortunately,

teacher effectiveness measures for teachers hired at these schools in 2011-12 and

2012-13 (actual replacements) are unavailable. Rather we employ the effectiveness of

teachers hired at these schools in 2008-09 and 2009-10.11 As might be inferred from

the previous analysis, the extended leavers will be less effective than the average of

teachers hired in their cohorts (extended teachers are less effective than those

approved and extended leavers are less effective than extended teachers who

remain). This analysis differs somewhat in that we include all hires, including within

district transfers, new hires with tenure, and those who exited prior to a tenure

decision. Also our model compares extended leavers to the attributes of new hires

solely in their school, which could make a difference as extended leavers tend to be

concentrated in a more limited group of schools. Finally, we develop a second

replacement group composed of teachers hired from 2006-07 through 2009-10.

It is certainly conceivable that the tenure reforms might change the labor

market for new teachers, and that actual replacements might be less effective than

those hired prior to reforms as they may be concerned that they, too, might be more

10 Teachers who were hired include both those new to teaching and teachers who transferred from
other schools.
11 The vast majority of teachers with tenure decisions in 2010-11 and 2011-12 began their

probationary periods in 2008-09 or 2009-10. We therefore are comparing the extended leavers to
other teachers hired under similar circumstances to themselves. We are making the assumption that
the teachers hired in 2009 and 2010 at the schools where an extended teacher left in 2010-11 or 2011-
12 have measured effectiveness similar to those teachers hired at these schools in 2011-12 and 2012-
13. We have also created a replacement comparison group of teachers by examining teachers who
were hired at these schools from 2006-07 through 2009-10. The proxy replacements include
information on principal effectiveness ratings for all teachers who persist to a tenure decision and
value added data for all teachers who began teaching at these schools during 2009 or 2010.

23
likely to be extended. However, the reverse is also true, especially in schools with

extended teachers, where effective teachers might find the improvement in the

effectiveness of peers an attraction. As a result, the use of proxy replacements should

be viewed as an approximation. For each school with an extended leaver, we

compare the average effectiveness of extended leavers with that of their proxy

replacements, and then average these within school differences across all such

schools. In this way we examine the difference in teacher effectiveness between

extended leavers and proxy replacements in the typical school.

As shown in Table 2, there are substantial differences in the effectiveness of

extended leavers and their proxy replacements. For example, there are 45 percentage

points fewer teachers rated as Highly Effective or Effective among all extended

leavers than their proxy replacements (14 percentage points Highly Effective and 31

percentage points Effective). Estimated value-added in ELA is 20 percent of a

standard deviation higher among the proxy replacements than the extended leavers.12

Although proxy replacement teachers are estimated to outperform extended leavers

in math value-added, this difference is not statistically significant at traditional

significance levels, due primarily to relatively few observations (N=158).

12Employing the sample of teachers entering schools between 2006-07 and 2009-10 as the proxy
replacement comparison group, we estimate the percentage of teachers rated highly effective or
effective is 44 percentage points higher for the proxy replacements than the extended leavers.
Estimated value-added is 13 percent of a standard deviation higher in ELA and 14 percent of standard
deviation higher in math, which are both significant at the 0.06 level.

24
 From a principal’s perspective, these are large effects relative to almost any

other intervention they might contemplate. For example, many principals rightly

privilege experience when hiring teachers as the value-added of a teacher with six

years of experience is estimated to be up to 15 percent of a standard deviation higher

than a novice teacher (Atteberry, Loeb, & Wyckoff, 2013). Extending the

probationary period of teachers with insufficient skills to be approved for tenure and

thereby nudging some teachers to leave the school who are then replaced with a new

teacher has an effect on teacher effectiveness about the same as the gains of hiring a

teacher with six years of experience rather than a novice.

School Differences

While implementation of the policy may have varied across schools, most

schools experienced a substantial change in the percentage of teachers who were

approved for tenure under the new policy. More than 70 percent of schools granted

tenure to fewer than 80 percent of their teachers following the introduction of the

policy as shown in Figure 2. While a cluster of schools approved 100 percent of

eligible teachers, most schools approved far less, with another large cluster of

schools with between 50 and 70 percent approval.

The variation in approval rates seen in Figure 2 corresponds to some school

characteristics, particularly average student attributes, as shown in Table 3. On

average, teachers approved for tenure work in schools in which the percentage of

25
white students is more than 50 percent greater than schools where teachers’

probation is extended. Black students experience the reverse. In schools where

teachers are approved for tenure, black students comprise 27 percent of all students,

but they comprise 35 percent of students in schools where teachers’ probation is

extended. The achievement of students in schools where teachers receive tenure is

15 percent of a standard deviation better in math and 13 percent of a standard

deviation better in ELA than the average achievement in schools where teachers are

extended.

Given the strong link between principal effectiveness ratings and tenure

decisions shown above, it is not surprising that the pattern of differences in school

attributes across principal effectiveness ratings mirror the differences across tenure

outcomes as shown in Table 3. For example, the average Highly Effective teacher

works in schools where the percentage of white students is twice as large as it is for

the average Ineffective teacher. The average Ineffective teacher is located in a school

with 65 percent more black students than their average Highly Effective colleague.

As is also shown in Table 3, the average Ineffective teacher is located in a school

where the ELA performance of students is more than a quarter of standard deviation

lower and more than 30 percent of a standard deviation lower in math than that of

the average Highly Effective teacher. This suggests that replacing Ineffective and

developing teachers with a teacher whose performance is closer to the average would

26
disproportionately improve the quality of teaching in schools with higher percentages

of black students.

To examine the types of schools most affected by the tenure reforms, we

estimated the relationship between school characteristics and tenure decisions in a

multivariate framework controlling for teacher performance measures. When we

estimate a model that includes only the attributes of the students in the school, the

percentage of students who are black is the only measure that is associated with the

likelihood of being extended. This relationship may exist for a variety of reasons,

including the differential effectiveness of teachers initially assigned to schools with

high percentages of black students, as well a lack of mentoring and leadership that

support the growth of teachers in such schools. When teacher attributes are added to

the model, they dominate the determination of whether a teacher is extended. The

estimate for the percent of black students drops substantially in magnitude such that

a 1 standard deviation increase in the percentage of black students (26.4 percentage

points) is estimated to increase the likelihood of a teacher being extended by just

over 1 percent (see appendix Table A5). Thus, the likelihood of teachers being

extended is much greater in schools with larger percentages of black students, but

the differential primarily results from these schools having more teachers judged to

be Ineffective and Developing.

27
Discussion

Teacher tenure has been a hotly debated issue for decades, but there is

surprisingly little research that documents the effects of various tenure policies. This

paper examines an unusual change in the tenure policy in New York City as a step

toward providing evidence to support the design of teacher workforce policies. We

believe this evidence has important implications for the current debate regarding

reforms to the tenure process.

Our analysis documents substantial changes in tenure decisions following the

NYC reforms. While almost all eligible teachers received tenure prior to the change,

after the reforms a large share of teachers instead had their probationary periods

extended to provide more opportunity for them to demonstrate the skills necessary

for effective teaching and for district decision makers to better assess teachers’

performance. Not surprisingly, low-performing and less qualified teachers were more

likely to be extended. Teachers in schools with disproportionate shares of black and

low-performing students also were more likely to be extended. Our analyses provide

some evidence that this differential reflects a uneven distribution of less effective

teachers, which is consistent with recent research (Isenberg, Max, Gleason,

Potamites, Santillano, Hock, & Hansen, 2013; Sass, Hannaway, Xu, Figlio, & Feng,

2012), although we cannot rule out differential application of tenure rules. Finally, we

found evidence that the new tenure policy resulted in additional voluntary attrition of

teachers who were extended, as well as additional involuntary dismissal of the small

28
share of teachers who were denied tenure. Among extended teachers, those with

lower effectiveness, as measured by principals’ ratings, but higher qualifications (e.g.

SAT scores) were more likely to leave, potentially further strengthening the teacher

workforce. Extended teachers who leave their schools are less effective as measured

by principal ratings and value-added estimates than are those likely to replace them.

Because teachers with poor effectiveness ratings are more likely to be in schools with

higher percentages of black students, these schools are most affected by the policy

change and most likely to see attrition of these less effective teachers as a result of

the reforms. These schools on average were able to hire more effective teachers to

fill these vacancies.

New York City’s reforms to the tenure process are still in their early stages

and much remains to be learned Learning to become an effective teacher is

complicated and dependent on many factors, some of which are out of the control

of teachers. Our results suggest large effects but provide only preliminary evidence

because we have not fully ruled out the effects of other factors that may have been at

play in the district simultaneously. Nor do we understand mechanisms by which

some teachers succeed and others choose to leave. In addition our findings should

be viewed as a short run response to the NYCDOE tenure reform. The long run

implications are unclear, but could potentially be less salient if: extended teachers

come to understand they will not be denied tenure, replacement teachers are not

more effective, or the overall applicant pool of new teachers is depressed as a result

29
of tenure reform. Finally, many extended teachers transfer to other NYCDOE

schools. What are the circumstances of these transfers and what is the performance

of these teachers following their transfer? With additional data a causal analysis will

be more feasible and we can address many of these questions. While the direct

effects of the tenure reforms are felt by teachers facing tenure decisions, the labeling

of teachers and increased likelihood of receiving an extension may induce other

teachers in the same school, subject, and/or grade to reassess their positions. These

processes may encourage principals to reassign teachers across grades and subjects or

to reallocate responsibilities in other ways.

Changes in human resource practices including new hiring and evaluation

policies have been hallmarks of many recent reforms. While the tenure process has

been the subject of continual debate, reforms have been slower and less sustained in

this area. In part as a result, research on tenure policies and variety of possible

approaches to probationary periods and screening is sparse. Nearly all districts grant

some form of tenure based at least in theory on teachers demonstrating proficiency.

Yet many districts do only cursory evaluation during the tenure process. As such,

adopting tenure reform similar to that presented here may be comparatively easy

relative to other much discussed human resource policies that require more

controversial policy changes.

30
References
Atteberry, Allison, Loeb, Susanna, & Wyckoff, James (2013). Do first impressions
matter? Improvement in early career effectiveness, CALDER Working Paper No. 90.
Washington, D.C.: CALDER.
Beteille, Tara, Kalogrides, Demetra, & Loeb, Susanna (2012). Stepping stones:
Principal career paths and school outcomes. Social Science Research, 41(4), 904–
919.
Boyd, D., Grossman, P., Lankford, H., Loeb, S., & Wyckoff, J. (2008). Who leaves?
Teacher attrition and student achievement. National Bureau of Economic
Research, Working Paper 14022.
Brown, Steven P., & Leigh, Thomas W. (1996). A new look at psychological climate
and its relationship to job involvement, effort, and performance. Journal of
Applied Psychology, 81(4), 358-368.
Brunner, Eric J., & Imazeki, Jennifer (2012). Probation length and teacher salaries:
Does waiting pay off? Industrial and Labor Relations, 64(1), 164-180.
Chetty, Raj, Friedman, John N., & Rockoff, Jonah E. (2014). Measuring the impacts
of teachers II: Teacher value-added and student outcomes in adulthood.
American Economic Review, 104(9), 2633-2679.
Corcoran, Sean P., Schwartz, Amy Ellen, & Weinstein, Meryle (2009). The New York
City Aspiring Principals Program: A school-level evaluation. New York: IESP.
Feinberg, Robert M. (1981). Earnings-risk as a compensating differential. Southern
Economic Journal, 48(1), 156-163.
Flyer, Fredrick, & Sherwin Rosen (1997). The new economics of teachers and
eEducation. Journal of Labor Economics, 15(1), S104-S139.
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, Volume 30, Issue 1, pages 57–87.
Isenberg, Eric, Max, Jeffrey, Gleason, Philip, Potamites, Liz, Santillano, Robert,
Hock, Heinrich, & Hansen, Michael (2013). Access to effective teaching for
disadvantaged students. Washington, DC: National Center for Education
Evaluation and Regional Assistance, Institute of Education Sciences, U.S.
Department of Education.
Jacob, Brian, Lars Lefgren (2008). “Can Principals Identify Effective Teachers?
Evidence on Subjective Performance Evaluation in Education.” Journal of
Labor Economics 26(1): 101–36.

31
Kalleberg, Arne L. (1977). Work values and job rewards: A theory of job satisfaction.
American Sociological Review, 42(1), 124-143.
Lee, Raymond T., & Ashforth, Blake E. (1996). A meta-analytic examination of the
correlates of the three dimensions of job burnout. Journal of Applied Psychology,
Vol 81(2), 123-133
National Council on Teacher Quality (2012). State of the states 2012: Teacher effectiveness
policies – Area 3: Identifying effective teachers. Washington, D.C.: National Council
on Teacher Quality.
New York City Department of Education (2009). The tenure toolkit, 2009-10. New
York: New York City Department of Education.
New York City Department of Education (2010). The tenure toolkit, 2010-11. New
York: New York City Department of Education.
New York City Department of Education (2011). The tenure toolkit, 2011-12. New
York: New York City Department of Education.
Rivkin, Steven G., Hanushek, Eric A., and Kain., John F. (2005). Teachers, schools,
and academic achievement. Econometrica, 73(2), 417-458.
Rockoff, Jonah E. (2004). The impact of individual teachers on student achievement:
evidence from panel data. American Economic Review Papers and Proceedings,
94(2), 247–252.
Rockoff, Jonah, Staiger, Douglas, Kane, Thomas, and Taylor, Eric (2012)
Information and employee evaluation: Evidence from a randomized
intervention in public schools, American Economic Review, 102(7): 3184-3213.
Ronfeldt, Matthew, Loeb, Susanna, & Wyckoff, James (2013). How teacher turnover
harms student achievement. American Educational Research Journal, 50(1), 4-36.
Rothstein, Jesse (forthcoming). Teacher quality policy when supply matters, American
Economic Review.
Rothstein, Jesse (2009). Student sorting and bias in value-added estimation: Selection
on observables and unobservables. Education Finance and Policy, 4(4), 537-571.
Sass, Tim R., Hannaway, Jane, Xu, Zeyu, Figlio, David N., & Feng, Li (2012). Value
added of teachers in high-poverty schools and lower poverty schools. Journal
of Urban Economics, 72(2), 104-122.

32
100%

80%

60%

40%

20%

0%
2007-08 2008-09 2009-10 2010-11 2011-12 2012-13
Approve Deny Extend

FIGURE 1. Percentage of Teacher Tenure Cases by Tenure Outcome 2007-08 to 2012-13

33
 250
200 150
Frequency
100 50
0

0 .2 .4 .6 .8 1

Proportion Approved

FIGURE 2. Distribution of School Proportion of Tenure Cases Approved 2009-10 through 2012-13

Note. Includes only schools with at least four tenure decisions over the period (81 percent
of all schools).

34
 Table 1
Determinants of Teacher Disposition in Year Following Tenure Decision, 2010-11 and 2011-12
(1) (2) (3) (4) (5) (6)
Variables Transfer Transfer Transfer Exit Exit Exit
Extend 0.145** 0.124** 0.087** 0.057** 0.055** 0.040**
(15.21) (13.04) (6.06) (9.38) (9.07) (4.32)
Student Attributes
Mean Math score -0.024 0.016
(-0.68) (0.68)
Mean ELA score -0.024 -0.019
(-0.64) (-0.82)
Black (%) 0.113* 0.042*
(4.24) (2.46)
Hispanic (%) 0.066~ 0.075**
(2.35) (4.21)
Free lunch (%) -0.099** -0.085**
(-3.12) (-4.23)
Reduced lunch (%) -0.307* -0.187*
(-2.33) (-2.23)
Principal Effectiveness Ratings
Ineffective 0.285* 0.110*
(4.24) (2.54)
Developing 0.071** 0.026*
(3.58) (2.02)
Effective 0.030* 0.007
(2.13) (0.74)
Missing 0.045* 0.013
(2.45) (1.11)
Constant 0.142** 0.135** 0.111** 0.064** 0.037** 0.031**
(4.88) (24.56) (9.60) (3.48) (10.52) (4.22)
School Fixed Effect X X X X

Observations 6,351 8,855 8,855 6,351 8,855 8,855

Note. T-statistics in parentheses. ** p<0.01, * p<0.05, ~p<0.1

35
 Table 2
Mean School Difference in Teacher Effectiveness Measures between Proxy Replacement and Extended
Leavers in Schools with Extended Leavers, 2010-11 and 2011-12
Principal Final Effectiveness Rating (%) Value-Added
Highly
Extended Leaver Status Effective Developing Ineffective ELA Math
Effective
All Extended leavers 14.34*** 30.7*** -36.45*** 1.37* 0.197** 0.119
Extended transfers 11.97*** 30.16*** -34.53*** 1.14 0.127 0.181*
Extended exiters 16.15*** 27.55*** -33.24*** 1.72 0.298* 0.037
Note. Proxy replacement teachers are all teachers hired at the school in 2009 and 2010. Only schools with an
extended leaving teacher in 2011 or 2012 included in all comparisons. Positive values indicate on average within
schools average value for replacement pool exceeds that for the Extended leavers. Comparing extended leavers to
proxy replacements: *** p<0.001, ** p<0.01, * p<0.05.

36
 Table 3
Attributes of the Students in Teacher’s School by Tenure Decision and Principal Effectiveness Rating,
2010-11 and 2011-12
Home
Free Reduced Math ELA
White Hispanic Black Lang
Lunch Lunch Achieve Achieve
(%) (%) (%) Eng
(%) (%) (z-score) (z-score)
(%)
Tenure Decision a
Approve 13.8 44.4 27.4 56.6 72.3 4.4 0.081 0.086
Extend 8.9 44.6 35.1 60.3 77.3 4.1 -0.066 -0.042
Deny 7.1 43.5 39.6 63.3 77.8 4.2 -0.152 -0.093
Principal Effectiveness Rating b

Highly Effective 16.4 42.8 24.1 56.5 69.2 4.8 0.184 0.181
Effective 12.1 45.3 29.9 57.2 74.6 4.2 0.007 0.019
Developing 8.4 45.0 35.3 60.8 78.1 4.1 -0.068 -0.046
Ineffective 7.2 42.4 39.9 62.7 77.7 4.6 -0.161 -0.102
No rating 12.3 42.7 31.0 57.4 71.3 4.1 0.055 0.073
Total 11.7 44.5 30.8 58.3 74.4 4.2 0.015 0.029
Note. a Extended teachers work in schools with different student attributes than approved teachers (p-value less
than 0.01 for all attributes except the percentage of Hispanic students). Teachers denied tenure work in
schools with different attributes than teachers who are extended with respect to the percentage of students
who are black, the percentage whose home language is not English and mean student math scores (p-value less
than 0.05). Differences in other student attributes are not significantly different from zero. b Teachers rated
ineffective work in schools with different student attributes than teachers rated effective or Highly Effective
(p-value less than 0.01 for all attributes except the percentage of Hispanic students and the percentage eligible
for reduced-price lunch). Teachers rated developing work in schools with different student attributes than
teachers rated effective or Highly Effective (p-value less than 0.01 for all attributes except the percentage of
Hispanic students).

37
 Technical Appendix

Estimation of Teacher Value-Added

We estimate teacher-by-year value-added employing a multi-step residual-based

method similar to that employed by the University of Wisconsin’s Value-Added Research
Center (VARC). VARC estimates value-added for several school districts, including until
quite recently New York City.

We initially estimate equation 1, which regresses achievement for student i in

class c at school s taught by teacher j in time t as a function of prior achievement ,
student attributes , and class fixed effects . In this model, the class fixed
effects subsumes both the teacher-by-year fixed effect and any other class or
school-level predictors of student achievement.

(1)

where

Employing these estimates, we calculate the residuals from this regression without
accounting for and then estimate equation 2 which regresses this residual on class and
school characteristics as well as a class random effect to reflect the grouping of
students into classrooms.

(2)

Employing these estimates, we calculate the residuals from this model and calculate
teacher-by-year value-added by averaging across the student-level residuals within a teacher
and year.

̂ (3)

The standard errors of the teacher-by-year value-added estimates are estimated as shown in
equation 4 using the student-level errors ̂ from equation 3 and number
of observations for each teacher-by-year group.

38
 ̂ (4)

We then employ a standard Empirical Bayes shrinkage method to account for the varying
uncertainty associated with each teacher-by-year value-added estimate.

39
 Table A1
Descriptive Statistics for the Analytic Sample, 2009-10 to 2012-13
Variable Obs. Mean Std. Dev.
Tenure Outcome (%)
Approve 19,372 66.6 47.2
Extend 19,372 30.3 46.0
Deny 19,372 3.1 17.2
Teacher Attributes (% unless otherwise noted)
Female 19,335 75.7
Black 17,113 18.5
Hispanic 17,113 16.6
SAT math 9,104 498 102.0
SAT verbal 9,104 501 98.0
Preparation Path (%)
College recommended 19,237 51.8
Teaching Fellow 19,237 21.9
TFA 19,237 2.9
Individual evaluation 19,237 8.5
Temporary license 19,237 5.1
Student Attributes (aggregated to school)
Hispanic (%) 15,124 43.8 25.1
Black (%) 15,124 31.0 26.3
Free lunch (%) 14,057 63.4 27.5
Reduced lunch (%) 14,057 3.9 4.1
Mean ELA score (z-score) 11,010 4.9 44.7
Mean Math score (z-score) 11,022 3.0 49.8
Teacher Performance Measures (%)
U rated 19,372 2.33
D rated 19,372 0.14
Principal Final Effectiveness Ratings
Ineffective 13,080 2.3 15.0
Developing 13,080 29.2 45.5
Effective 13,080 42.6 49.4
Highly Effective 13,080 15.3 36.0
No Rating 13,080 10.6 30.8
Low attendance 13,080 8.0 27.0
VAM ELA 1,498 0.00 1.1
VAM Math 1,538 0.08 1.2
NYC VAM low 2,410 9.3
NYC VAM high 2,408 8.0

40
 Table A2
Average Teacher Value-Added Scores for Teachers who Remain for Tenure Decision and Those
who Leave Prior to the Tenure Decision by Subject, Year of Experience, and Tenure Cohort
2008-2009 2010 2011
Receive tenure Leave prior to Receive tenure Leave prior to Receive tenure Leave prior to
decision tenure decision decision tenure decision decision tenure decision
First year of probationary period
Math -0.241 -0.316 -0.246 -0.512* -0.223 -0.419
ELA -0.191 -0.228 -0.197 -0.443** -0.175 -0.182
Second year of probationary period
Math 0.045 0.001 0.091 -0.046 -0.004 -0.142
ELA -0.062 -0.050 0.021 -0.049 -0.003 0.028
Note. * p < .05, ** p < .01 compared 2008-09 pooled cohort

41
 Table A3
Attributes of Teachers by Tenure Outcomes, 2010-11 through 2012-13a
Tenure Value Added U Rated D Rated Low Attd SAT LAST Preparation Route (%)b
Decision ELA Math (%) (%) (%) Math Verb Exam Coll Rec NYCTF TFA Ind Eval
Approve 0.081 0.248 5.7 22.2 37.1 505 505 257 59.9 49.5 60.2 55.0
Extend -0.138 -0.129 52.1 66.7 56.2 490 494 254 37.8 47.2 38.9 40.7
Deny -0.115 -0.740 42.2 11.1 6.7 469 490 248 2.4 3.2 0.1 4.3
Total -0.009 0.070 100.0 100.0 100.0 498 500 255 100.0 100.0 100.0 100.0
Note. a Means of teachers approved exceed those of teachers extended at a p-value of 0.05 or lower for all attributes. The means of teachers
extended exceed those of teachers denied at a p-value of 0.05 or lower for all variables except ELA value-added and verbal SAT. b The tenure
approval rate is lower for teachers prepared through the NYCTF and IE preparation routes than those from CR programs at p-values of .01 or
lower. There is no statistical difference between CR and TFA.

Table A4
Attributes of 2011 and 2012 Extended Teachers by Disposition in the Following Year
Value Added U-Rated D-Rated Low Attd SAT LAST Preparation Route (%)
Attrition
Status ELA Math (%) (%) (%) Math Verb Exam Coll Rec NYCTF TFA Ind Eval
Same School -0.091~ -0.090 4.0~ 0.2** 10.7 491 495 253** 77.5 70.9** 53.3** 78.8*
Transfer -0.355 -0.421 2.7 0.2 11.2 482 486 253 16.3 15.6 9.0 17.7
Exit -0.332 -0.145 2.9 0.0 9.1 530 539 267 6.2 13.6 37.7 3.5
Note. ** p<0.01, * p<0.05, ~ p<0.10. For Value-Added, U Rated, D Rated, Low Attendance, SAT and LAST Exam, significance levels
denote significant differences between the values of these variables for Extended teachers who remain in same school and those who
either transfer or exit. For Preparation Routes, significance levels denote differences between designated route and College
Recommended.

42
 Table A5
Determinants of Whether Teacher is Extended Relative to being
Approved, 2010-11 and 2011-12
(1) (2)
Extended Extended
(=1) (=1)
Student Attributes
Mean Math score -0.096 -0.073~
(-1.41) (-1.77)
Mean ELA score -0.010 0.021
(-0.14) (0.49)
Black (%) 0.211** 0.048~
(-4.41) (1.80)
Hispanic (%) 0.032 -0.008
(-0.62) (-0.27)
Free lunch (%) 0.012 -0.041
(-0.20) (-1.13)
Reduced lunch (%) -0.066 -0.043
(-0.26) (-0.28)
Teacher Attributes
Low Attendance 0.066**
(3.84)
Unsatisfactory Rated 0.101**
(2.85)
Doubtful Rated -0.125
(-0.75)
Principal Final Rating
Ineffective 0.867**
(25.61)
Developing 0.906**
(95.62)
Effective 0.100**
(8.72)
No rating 0.334**
(12.95)
Constant 0.340** 0.081*
(-6.12) (2.45)
Observations 6,351 6,351
R-squared 0.033 0.613
Note. T-statistics in parentheses. ** p<0.01, * p<0.05, ~ p<0.1

43