August 2008

Using Early-Warning Data to Improve Graduation

Rates: Closing Cracks in the Education System
By Lyndsay Pinkus
With a national high school graduation rate, hovering around 70 percent,1 far too many of the nation’s students
are falling through the cracks of the education system and leaving high school without the skills necessary for
success in college, work, and life. Anecdotal evidence and national statistics show that educational outcomes for
poor and minority children are generally worse than that of their peers. Yet recent research is consistently reveal-
ing academic factors—known as “early-warning data,” “risk factors,” or “on-track measures”—that more
accurately predict whether or not a student is likely to drop out than socioeconomic factors do.2 This knowledge
puts power in the hands of educators to eradicate the “soft bigotry of low expectations” and to strategically focus
on the academic factors of success they can positively impact.

Simply identifying at-risk students does nothing to mitigate their risk factors and help them graduate. However,
the predictive power of early-warning data is being harnessed by schools, districts, states, and support organiza-
tions around the country and being used to guide interventions and embed prevention strategies throughout
schools. These approaches, still in their early stages, are helping educa-
tors prevent students from falling off the track to graduation and to The power of early-warning indicators
target interventions and support to students who need them most. Early- lies in the willingness and capacity of
school leaders and educators to trans-
warning data also can be used to better understand—and target re- form insightful data into strategic
sources to—low-performing schools where concentrated numbers of decisionmaking that leads to improved
students require significantly improved schools in order to succeed. student outcomes.

By preventing students from falling through the cracks and ensuring that they receive the appropriate level of at-
tention, instruction, engagement, and support needed to succeed in their classes, educators can give every student
the chance to graduate from high school prepared for college, the modern workforce, and life. This brief explores
the predictive power of early-warning data, offers examples3 of current efforts to use such data to guide secondary
school interventions across the country, and discusses the policies that can support these efforts.

Getting to the Roots of the Problem: Early-Warning Data

A glance at high school graduation statistics reveals scribe who is dropping out of high school, but it does
the following: High school dropouts are more likely not explain why they are dropping out.
to be poor, minority, and male. They are more likely
to have had their education interrupted due to chal- Over the past five decades, research has shown that
lenges such as mobility, homelessness, pregnancy, “[dropping out of] high school represents a conflu-
incarceration, or abuse. This information may de- ence of individual, social, family, cultural, socio-
economic, and institutional factors … Dropping out
is best understood as the ultimate end to a gradual failing grade in math or English or attendance
process that often begins as early as school entry rates below 80 percent in the eighth grade.7
and through which students lose interest and
significance for what schooling offers them.”4 • Seventy-five percent of the dropouts from the
Boston Public Schools’ Class of 2004 fit into one
The decision to drop out is rarely the result of a sin- of four distinct categories: 1) students with mul-
gle life event; in fact, many students exhibit tiple ninth-grade course failures; 2) students with
academic warning signs years before they leave high one or more eighth-grade risk factors (attendance
school. Using longitudinal data—information about below 80 percent, two or more years over-age, or
individual students’ progress over time—highly res- failing multiple core courses); 3) late-entrance
pected researchers from Johns Hopkins University, English language learners; or 4) special education
the Consortium on Chicago School Research students taught in “substantially separate” class-
(CCSR), and the Parthenon Group, among others, rooms.8
have examined the academic history of dropouts to
identify their shared academic characteristics. Stu- • Chicago Public Schools’ eventual graduates and
dents who dropped out usually had received a failing dropouts were accurately identified 80 percent of
grade in core courses (especially in math or Eng- the time using an “on-track” indicator based on
lish), earned a low grade point average (GPA), or the number of credits earned and the number of
scored low on achievement tests. They were often failures in core courses by the end of the ninth
retained in grade because they had not earned grade.9 (See graph below.)
enough credits to be promoted; as a result, many
were older than the other students in their class. Fur-
thermore, as demonstrated by low attendance rates
and disciplinary problems, these students were fre-
quently not engaged in their education or aware of
its importance to future opportunities.

The early-warning signs of dropping out

When analyzed in combination, these academic cha-

racteristics can provide strong indications
of which students are at risk of becoming dropouts.
Examples from recent research include the follow-
ing:

• As early as fourth grade, future dropouts from a

Fall River, Massachusetts, study received lower
grades than future graduates did. The “early dro-
pouts”—those who would eventually drop out of
high school between seventh and ninth grade—
Graduation rates of Chicago Public School freshmen identi-
generally earned a C- academic GPA and ranked fied by on-track indicator.
in the twenty-fifth percentile of their fourth-grade Source: E. M. Allensworth and J. Q. Easton, What Matters for Staying on
class.5 Track and Graduating in Chicago Public High Schools (Consortium on Chi-
cago School Research at the University of Chicago, 2007).
• Students in Philadelphia who in sixth grade failed
either a math or English course, had an atten-
dance rate of under 80 percent, or had a final The benefits of early-warning data
“unsatisfactory” behavior mark in at least one
class had at least a 75 percent chance of dropping The ability to pinpoint and the opportunity to
out of high school.6 address these academic issues is good news for
• Three out of four students who ultimately educators, reformers, and advocates interested in
dropped out of Philadelphia schools had either a improving student outcomes. First, it is no longer

2
necessary to rely only on the broader socioeconomic and actionable in a strategic way throughout a
markers of demographics and economic status to school. Some examples of this include:
identify students at risk of dropping out. Instead,
specific academic early-warning data and well- • shining a light on the exact indicators that are the
designed monitoring systems can be used both to “tipping point” for student success;
identify at-risk students and to gauge their level and • serving as a mechanism for pulling together
cause of risk. Second, while educators cannot disparate data and sharing it with classroom
change the out-of-school factors that may contribute teachers who wouldn’t have access to it otherwise;
to a student’s decision to drop out, by focusing on • helping to prioritize among students along the
improving students’ academic performance they can spectrum of weaker performance;
reduce how much those nonacademic factors inter- • acting as an organizing tool for structuring
fere with students’ eventual educational success. interventions; and
Lastly, patterns in early indicators can be examined • identifying goals for the interventions.
at the school and district levels to identify systemic
weaknesses in schools that are actually increasing
the likelihood that students will drop out. A recent analysis of Los Angeles’ students’ outcomes
found that “academic experiences explained six times
It may seem intuitive that students with the academic more of the difference in graduation rates among
risk factors described in this brief are more likely to students…than demographic characteristics.”
drop out, and that educators would use this informa- Source: D. Silver, M. Saunders, and E. Zarate, What Factors Predict
tion to target students for intervention. Even so, High School Graduation in the Los Angeles Unified School District
(California Dropout Research Project at the University of California
formalizing the use of early-warning data can serve Linguistic Minority Research Institute, 2008).
as a way to make that information understandable

Developing Early-Warning Systems

The indicators isolated in the studies described remain unidentified would be more likely to receive
above are examples of “high-yield” indicators: col- the help they need.10 Educators, reformers, and poli-
lectively, they identify a significant portion of future cymakers alike may choose to embrace early-
dropouts and identify students who—absent inter- warning data systems as tools for the more effective
vention—have very low odds of graduating. and efficient use of scarce resources to improve
student outcomes.
Early-warning data offers a way to address a prob-
lem that plagues many dropout prevention efforts: Some indicators are usually strong predictors
efficiently targeting the right students. By using
early-warning data to predict future dropouts, educa- For decades, researchers have studied dozens of
tors can reduce the inaccurate targeting—to students factors that have an impact on whether a student is
who are unlikely to benefit from them—of these more or less at risk of dropping out.11 However,
kinds of interventions. Specifically, students who several indicators continually rise to the top of the
might be considered at risk of dropping out due to predictive list; these indicators were paramount in
demographic factors, but who are actually successful the various early-warning studies cited above and
in school, would not receive unnecessary dropout critical to early-warning systems development in
prevention interventions, while students with high- cities across the country. Researchers from the
yield academic risk factors who would otherwise Everyone Graduates Center at Johns Hopkins
University are currently working with more than ten
A large-scale evaluation of federally funded dropout school districts to develop early-warning data
efforts found that “dropout prevention programs often systems. “In the high-poverty districts we are work-
serve students who would not have dropped out, and ing in, we have not found one district where grades,
do not serve students who would have dropped out.” credits, and attendance are not highly predictive,”
Source: P. Gleason and M. Dynarski, “Do We Know Whom to Serve? noted Robert Balfanz, co-director of the center.12
Issues in Using Risk Factors to Identify Dropouts,” Journal of Educa-
tion for Students Placed at Risk 7, no. 1 (2002): 25–41.

3
• Course success missed less than a week of school per semester
graduated, regardless of their eighth-grade test
Research demonstrates that course success may be scores.17 Dropouts themselves acknowledge the im-
the best predictor of eventual graduation. CCSR’s portance of attendance; in a recent survey of those
study of the Chicago schools found that students who chose to leave school without attaining a dip-
with a B average or better in their freshman year loma, missing too many days of school and having
have more than a 95 percent chance of graduating. trouble catching up was the second most reported
On the other hand, those freshmen with less than a C reason for dropping out.18
average are more likely to drop out than to graduate.
In fact, once students’ freshman-year grades are While research on student engagement and atten-
known, other background information, including test dance demonstrates the complexity of these issues
scores, does not improve the predictability of gradu- and their solutions, successful high schools and high
ation. Course grades also are the best predictors of school reform models often take a strategic approach
improvements on test scores and college gradua- that includes both preventive and responsive strate-
tion.13 gies. One study that surveyed strategies for
increasing attendance described them as falling into
This supports the commonsense notion that lack of four overlapping categories: sound and reasonable
success in academic courses serves as a barrier to attendance policies with consequences for missing
graduating from high school. As students fall farther school; early interventions; targeted interventions for
behind in their course work, they lack the number of students with chronic attendance problems; and
credits to be promoted to the next grade, are required strategies to increase engagement and personaliza-
to repeat classes and grades, and become older than tion with students and families that can affect
their classroom peers, all of which makes them more attendance rates.19
likely to drop out. Fortunately, there are many class-
room strategies educators can use to increase • Academic skills
students’ course success, including improving and
personalizing instruction, providing extra learning Literacy is the gateway skill that students must have
time, striking a balance between relevance and rigor, mastered if they are to be successful in any course;
and providing support for students who are strug- low literacy levels translate into poor grades, grade
gling with skills or content. repetition, and eventual disengagement from school,
all of which tend to precede a student’s decision to
• Attendance drop out.20

Nowhere is the saying “Showing up is half the bat- In an average high-poverty urban school—where
tle” more true than in education. Especially for dropping out is sometimes more common than gra-
students who are struggling academically or lack duating—approximately half of incoming ninth-
academic support at home, the primary way to build grade students read at a sixth- or seventh-grade lev-
the skills and knowledge necessary for success is el.21 This means that they cannot navigate or
through participation in classes. Student attendance comprehend textbooks designed for ninth graders
is more difficult to control, of course, as students get and will quickly become lost in their courses. “Kids
older and have increased autonomy over their sche- would rather be bad than be embarrassed. When stu-
dules and transportation; absenteeism due to reasons dents are able to acquire the skills needed to
other than illness and cutting individual classes in- succeed, their grades improve, attendance improves,
creases with each grade, starting in the sixth.14 Not and behavior improves,” notes former high school
surprisingly, absenteeism, cutting classes, and truan- principal Mel Riddile.22 (See box on page 5 regard-
cy all have been found to be highly correlated with ing the use of literacy data at his high school in
dropping out.15 In a large longitudinal survey, stu- Virginia.) Successful high schools and high school
dents who had high absenteeism, cut classes at least reform models across the country are identifying
once a week, or were tardy ten or more times in a struggling readers and including adolescent literacy
single month were more than six times as likely as support as part of their game plan.
their peers to drop out.16 Recent research in Chicago
found that nearly 90 percent of freshmen who

4
 researchers found that less than 80 percent atten-
Using Literacy to Measure Progress
dance was the tipping point in eighth grade, while
At T. C. Williams High School in Alexandria, Virginia, less than 70 percent attendance was the tipping point
educators explicitly set college and work readiness
at graduation as the goal for all students. As part of a
in ninth grade.23
comprehensive school improvement approach that
necessitates strong leadership and staff buy-in, the To ensure that the indicators being used have the
school places a high value on literacy as a predictor of highest yield possible, educators and administrators
graduation. Former principal Mel Riddile recently noted, using early-warning systems must be prepared to
“Our literacy rate is the same as our graduation rate, and
that is no coincidence.” Using the Lexile scale for read- adjust their indicators and triggers as new or better
ing, which matches reader ability to text difficulty, they information becomes available. For example, when
identified a Lexile score of 1,300 as the proxy for twelfth- Chicago Public Schools adopted CCSR’s freshman
grade college and work readiness. Working with the on-track metric as part of a new early-warning inter-
middle schools, a Lexile goal was set for each grade
from sixth grade forward. Students are assessed regu-
vention program, district officials added easy-to-
larly to see if they are on or off track to achieving the access indicators (including multiple disciplinary
literacy skills needed for graduation. Educators then infractions and eligibility for, but nonparticipation
implement strategies to address the three factors that in, summer transitional programs) at the suggestion
schools can control in supporting at-risk students: time, of district staff and program educators.24 Similarly,
setting, and method. Struggling readers are enrolled in
an additional intensive reading class with a reduced at Abbeville (Louisiana) High School, former prin-
student-to-teacher ratio and are taught by teachers who cipal Ralph Thibodeaux constantly reevaluated the
have received training in literacy instruction methods. triggers in his early-warning data system to make
Prior to these interventions, only 10 percent of students sure he was accurately identifying at-risk students.25
who did not pass the state reading test on the first try
passed on the second try. In just the first year of imple-
mentation, the second-try pass rate increased to 55 Leaders creating early-warning systems must also be
percent—and it continues to rise. aware of their capacity to intervene with the identi-
Source: M. Riddile, personal communication, March 5, 2008. fied students. Staff burden was one of Thibodeaux’s
constant concerns as he adjusted the early-warning
triggers to identify only a manageable number of
Approaches to building early-warning data students. Bobby Franklin, a Louisiana Department
systems of Education official, echoes this point, cautioning
against the use of early-warning indicators and trig-
While some indicators almost always have value in gers that place too many students in the at-risk
predicting dropouts, the yield of each indicator also category. Not only will this reduce the sense of
depends on its “trigger”—that is, how it is applied. urgency for educators with regard to the at-risk
For example, at what number, at what percent, or in students, but, as he says, “if the list is too long, the
which grade is the indicator a stimulus for action? educators won’t be able (or willing) to carry the
Research on attendance rates in Philadelphia pro- burden.”26 As a result, limited resources may
vides one example. In identifying factors that gave inevitably influence decisionmaking around the
students at least a 75 percent chance of dropping out, use of early-warning data.

Two Birds With One Stone: Improving Test Scores and Graduation Rates
A recent analysis by CCSR found that the very factors that keep students on track to graduation⎯course success,
attendance, and behavior⎯also drive success on test scores. Therefore, by focusing on the quality of course work and
instruction and student behaviors, educators can increase both test scores and graduation rates. In other words, improving
both test scores and graduation rates are manageable and compatible goals for improving low-performing high schools.
Source: E. M. Allensworth, M. Correa, and S. Ponisciak, From High School to the Future: ACT Preparation—Too Much, Too Late (Consortium on Chicago
School Research at the University of Chicago, 2008).

5
Because both the pattern of high-yield risk factors New Visions for Public Schools’
and the capacity for intervention vary, schools and On-Track Metric
districts need to determine the data that is most pre-
New Visions for Public Schools, an education reform
dictive and actionable for their particular system.27
organization in New York City (NYC) that supports more
There are several approaches to recognizing early- than sixty-three of the city’s public schools, has developed
warning indicators and developing the data systems an on-track metric for use in its high schools. The metric is
to track them and identify students that exhibit them. based partially on CCSR research and findings from NYC’s
While this work can be completed using sophisti- Department of Education’s Office of Multiple Pathways.
New Visions uses New York State’s graduation require-
cated systems that have tracked student data for ments (a combination of course credits and pass grades in
years, it can also be done with simpler research and a series of end-of-course state Regents exams) to map out
tools available at the school level. the progress students need to make each year to graduate
with a regular diploma in four years. In addition, the metric
includes indicators of college readiness that go beyond the
• One approach (used in Chicago, Boston, and state’s high school requirements. Using this metric, each
Philadelphia) is to analyze historical data on past student’s status is described through the use of color-
dropouts to identify common academic risk fac- coded categories: on track to college readiness (blue); on
tors.28 Even without expensive longitudinal track to graduation (green); almost on track to graduation
studies, simpler analyses can be conducted using (yellow); or off track to graduation (red). New Visions ana-
lyzes the data for each school and provides administrators
past transcript data and school records.29 with a schoolwide report on student performance based on
the metric. It also creates snapshots that can show an indi-
• Even in the absence of historical information, vidual student’s progress toward graduation and college
data currently available at the school level can readiness, and encourages students to create a plan to get
(or stay) on track. New Visions’ schools also employ a
be useful. For example, using graduation re- comprehensive data-driven model of instructional and
quirements and working backward to identify operational improvement and school leadership develop-
interim benchmarks of success, New Visions for ment known as the Scaffolded Apprenticeship Model
Public Schools in New York City developed a (SAM). Each school convenes one or more teams com-
prised of the principal or assistant principal, teachers, and
metric to measure whether students are on track school counselors who are trained to analyze data and to
to graduation (see box to the right). develop and assess tailored strategies to improve instruc-
tion for a group of targeted, low-achieving students. This
• There is also an increasing interest in developing strategy helps ensure that data use becomes an intrinsic
component of the school’s daily operations.
and using indicators that measure a student’s
preparedness for college; as a result, many ef- The color-coded tools are easy to understand; as a result,
forts are under way to define and track progress schools are using the information in a variety of ways.
toward that goal. Examples include efforts by School leaders and staff are employing schoolwide and
several states (including Florida, Massachusetts, individual student data to plan interventions and decide
what professional development their teachers need.
and Texas); assessment developers (including Student snapshots are shared with students and parents,
ACT, the College Board, and the National making it easier to understand the complicated graduation
Assessment Governing Boarda); organizations requirements and the steps a student must take in order to
(such as New Visions and CCSR); and even get on track. A number of high schools served by New
Visions use the on-track metric as a tool for creating indi-
individual schools (for instance, T. C. Williams
vidual graduation plans for each student. New Visions’
High School in Virginia; see box on page 5). new college-readiness metric, based on college entrance
requirements, is helping students and families understand
the difference between high school graduation and college
readiness. New Visions is now developing a high school-
readiness metric for use in its middle schools.

While it is too soon to quantify the effects of these efforts,

the New Visions experience thus far illustrates how easy-
a to-understand information can lead a variety of stakehold-
The National Assessment Governing Board (NAGB)
ers—educators, students, and families—to make data-
administers the National Assessment of Education driven decisions that are likely to positively impact student
Progress (NAEP) assessment, also known as the Nation’s success and outcomes.
Report Card. NAGB is in the process of redesigning Source: R. Genn and B. Gunton, personal communication, 2008.
the twelfth-grade NAEP to measure postsecondary
preparedness.

6
 Putting the Data to Work: Addressing Students’ Academic Challenges
Early-warning indicators provide a helpful way of educators to transform actionable data into strategic
looking at the academically diverse populations decisionmaking that leads to improved student out-
served by America’s secondary schools, including: comes. But early-warning data and the identification
of at-risk students are not silver bullets; rather, they
• Students who struggled in middle school and
are factors to be included in the complex mix of
entered high school already off track or at risk.
elements that together can result in comprehensive
Many incoming ninth graders already lack the
secondary school improvement.
necessary academic skills, behavioral habits, and
course-taking history that are necessary for suc-
As school leaders strive to implement comprehen-
cess. These students need close monitoring and
sive intervention and improvement strategies that
support that starts in the middle grades and
address the needs of all the students in the school,
continues through high school. Particular atten-
the emerging use of early-warning data has led to the
tion must be paid to these students as they move
application of tiered intervention models (described
through the notoriously difficult transition from
in the following pages) to complement those
middle to high school, where any existing chal-
strategies. Such models include applying broadly
lenges are often compounded.
supported preventive efforts populationwide;
• Students who struggle during ninth grade. continuously identifying at-risk individuals; and
Success in ninth grade is critical to graduation. providing successive levels of intervention for
Some students struggle immediately, failing those who need them. This approach can help all
multiple classes or having high absentee rates in secondary schools meet the differing needs of their
the first term of their ninth-grade year. These academically diverse student populations and should
students need immediate interventions; if these be considered part of the data-driven decisionmaking
actions are not successful by the end of the that is a critical component of secondary school de-
freshman year, it may indicate that the student sign and reform efforts; the use of early-warning
needs more intense intervention or an alternative data is one way to do this.
pathway that allows for a more individualized
approach.30 Other students may fail one or two ►Preventive strategies to ease transitions and fo-
classes in their first semester or year of high cus on progress toward graduation: Preventive
school, or fail no classes but achieve less than a strategies can be used schoolwide to reduce student
C average. These students are likely to be risk factors and the number of dropouts by providing
positively influenced by modest intervention students with consistent attention to their progress,
efforts, such as summer school or peer-support motivation to succeed, sense of control and accoun-
groups. tability for their success, and support at critical
junctures. Such strategies—including personalized
• Students who start high school on track but drop graduation planning processes, various ninth-grade
out in later years. For many of these later dro- transition strategies, and strong behavior and atten-
pouts, the high school experience itself is what dance policies (described above)—can create a
contributes most to their lack of success. The stable learning culture. Research on ninth-grade
relatively small number of students who start transition strategies as summarized recently by the
high school on track but whose grades or atten- Council of the Great City Schools “suggests that
dance begin to drop in later years usually strategies that offer personalized attention, supple-
experience some crisis while they are in high mental academic instruction, and increased exposure
school and benefit from a short-term interven- to core courses, can improve the ninth grade transi-
tion to get them back on track. tion and students’ progress through high school,
particularly for high needs students.”31 If prevention
Using tiered interventions to address student strategies are successful, fewer students will demon-
challenges strate risk factors and more will progress toward
graduation.32
The power of early-warning indicators lies in the
willingness and capacity of school leaders and

7
►Group strategies to address particular risk services such as child care, housing, mental health,
factors: Second-tier interventions are intended to and so on), and blended academic and career
address the group of individuals (commonly 15 to 20 programs.35
percent) who are not responsive to or protected by
the first-tier interventions. Group strategies can be Success with Multiple Pathways
directed toward students who share a particular risk Following an analysis of their student outcomes
factor or struggle with similar challenges. Imple- that demonstrated that 93 percent of dropouts
mentation should ensure that group strategies were over-age and under-credited, the New York
continue to match the individual students’ academic City Department of Education created multiple
pathways designed specifically for such students.
challenges. Targeted interventions can include daily The graduation rate for this population increased
attendance check-ins, behavior checklists brought to from 19 percent in the traditional schools to 56
each class, or extra-help courses. These strategies percent at small, transfer schools and 39 percent in
can be particularly useful for students who are not evening programs.
yet completely disengaged but show clear signs of Sources: American Youth Policy Forum, Forum Brief:
Development and Implementation of Multiple Pathways to
risk. Graduation in New York City—Friday, October 26, 2007,
http://www.aypf.org/forumbriefs/2007/fb102607.htm (accessed
June 4, 2008); J. Lynch, “Using Data to Guide the Development
►Individual strategies to reach the most chal- of Multiple Pathways,” paper presented at the Symposium on
Federal Policy and Persistently Failing Schools, May 14, 2008.
lenged students: In the third tier, the small
percentage (5 to 10 percent) of individuals whose
needs are not addressed by the group strategies are Examples in action
supported with individually delivered, intensive in-
terventions. Individual strategies can be used to Below are examples of how a school reform model,
provide intensive attention for students who need a school district, and a high school have used early-
one-on-one support to succeed.33 Such support may warning data to implement tiered interventions:
include individual counseling or tutoring, behavior
contracts, or the involvement of social workers or • The researchers behind the Talent Development
psychologists to reach these students.34 secondary school reform models, working with
the Philadelphia Education Fund and the School
►Alternative strategies to serve students not likely District of Philadelphia, have used their findings
to succeed in their current setting: A small num- from early-warning studies to design and pilot a
ber of students face unique challenges that are not prevention and intervention program in
likely to be solved in the traditional educational Philadelphia’s middle schools. As mentioned
settings they have struggled in for years. For exam- earlier, they use four easily and commonly
ple, one subset of this group of students is known as measured factors—poor attendance, poor beha-
“over-age and under-credited.” Academically, these vior marks, a failing grade in math, or a failing
students are two or more years off track from gradu- grade in English—that are highly effective in
ation and likely need more individualized instruction identifying the majority of sixth graders who fall
than can be offered in most high schools. They are off the path to graduation. The prevention and
not comfortable socially in classes with much intervention plan addresses each factor in three
younger students, become increasingly frustrated by stages (whole school, targeted, and intensive) to
their lack of academic success, and feel further attend to the needs of all students within the
isolated in remedial pullout sessions. Many of these school, as illustrated in the figure to the right.36
students also have additional financial or family Whole-school strategies include student adviso-
responsibilities outside school that make attending a ries, consistent responses beginning with the
traditional high school difficult. Fortunately, districts student’s first unexcused absence, and research-
across the country are successfully using data to based instructional programs. Targeted interven-
identify over-age and under-credited students. tions can include daily attendance check-ins,
Common alternative settings include comprehensive behavior checklists brought to each class, or ex-
alternative schools, programs targeted to meet the tra-help courses. Intensive support can include
needs of populations with alternative learning styles in-depth behavioral assessments, one-on-one tu-
or schedules, integrated-services schools (providing toring, and bringing in appropriate social

8
 services. An early-warning indicators section • Using data from a districtwide system, Louisi-
has been added to the school district’s data sys- ana’s Abbeville High School has developed an
tem, enabling classroom teachers in the pilot early-warning system and a five-level interven-
schools to view continually updated indicator tion system that includes a comprehensive
profiles of the students in their classes. Teachers program for all ninth-grade students, known as
sharing common sets of students meet regularly the “freshman academy.” At-risk students are
to monitor student progress, intervene early, and identified every three weeks, using discipline,
track the success of interventions.37 attendance, overall GPA, and decline in GPA as
warning indicators, from data system-generated
reports. Educators then provide these students
with a variety of progressively intense interven-
tions, including mandatory tutoring, staff and
student mentors, and increased levels of com-
munication with families. In its first year of full
implementation, the combined use of early-
warning systems and the freshman academy de-
creased the number of ninth graders who were
not promoted to tenth grade by 50 percent.38 At
the end of the 2008 school year, a full 90 percent
of Abbeville’s ninth graders were promoted to
tenth grade.39

• Chicago Public Schools is piloting an early-

warning and intervention system in six of its
high schools. Using students’ historical data that
correlates with the CCSR on-track metric (based
on attendance, grades, age, and discipline
record), the central office analyzes the student
This figure illustrates how schools participating
data and provides each school with a list of at-
in the Talent Development middle schools pilot
organize interventions around the four risk risk students. The school then uses this informa-
factors. tion and other school-based data (extracurricular
Source: R. Neild and L. Herzog, “The Secondary Conti- activities, parent engagement information, and
nuum—Keeping 6th–12th Grade Students on Track to academic and discipline records) to develop, for
Graduation,” paper presented to the 4th Annual Middle
Grades Matter Conference, May 25, 2007, each student, an intervention strategy that ad-
www.philaedfund.org/mgm/MGM%202007%20Neild%20
and%20Herzog.ppt (accessed on March 3, 2008).
dresses four issues: parent engagement,
connection to the school community, social and
emotional support, and academic support.40

Using Early-Warning Indicators at the School and District Levels

While early-warning data can certainly be used to can help to identify schools that face high concentra-
identify and intervene at the student level, there is tions of incoming, off-track students; have large
also application for this data at the school and dis- clusters of students with certain academic risk fac-
trict levels. tors; have a history of contributing to risk factors; or
are unable to mitigate risk factors. This kind of data
First, early-warning data can be used to better under- can help educators pinpoint and address problems at
stand the nature of the academic problems in low- the school and student levels systemically. For
performing secondary schools. Examining the pat- example, a school that sees low student attendance
terns of early-warning indicators can unearth during morning classes can develop schoolwide
systemic weaknesses and enable schools and dis- strategies to get students to school on time. Or a
tricts to address them head on. Early-warning data feeder middle school whose students consistently

9
struggle through the first semester of ninth grade improving graduation rates. There is emerging con-
might benefit from partnering with the high school sensus that it may take several years for whole-
to develop a transition program. This data is useful school reform strategies to manifest themselves in
for those charged with improving school outcomes the consistent, large increases in graduation rates and
and for the policymakers interested in targeting re- test scores41 that are often used to make accounta-
sources to the schools that face the most significant bility judgments and allocate resources. In the
academic challenges. meantime, schoolwide early-warning data can pro-
vide a way for school leaders to demonstrate interim
Second, early-warning data can be used to assess the improvements that research indicates will lead to
effectiveness of strategies in a timely manner. End- increased graduation rates.
of-the-year test scores or years-away graduation out-
comes are an inefficient feedback loop for assessing Lastly, easy-to-understand early-warning data can be
the impact of educators’ strategies and interventions. a powerful tool for garnering support from key
Early-warning data that is available to school leaders stakeholders for needed interventions. Across the
and educators on a more frequent basis—such as country (see box below), educators and advocates
attendance, behavior, and course grades—can pro- are using this data to engage disenchanted students
vide a quick read on these efforts and guide in shaping their own futures; to involve sometimes
adjustments and targeting of resources. elusive high school parents in their children’s aca-
demic progress; to empower educators with
Early-warning data also provides a way to demon- actionable data; and to solicit resources from local
strate whether an entire school is on track to partners.

Data Can Empower Stakeholders to Embrace Interventions

Early-warning data is being used in a variety of ways to engage critical stakeholders in actions to improve student out-
comes.
A New York City high school principal employing New Visions’ on-track metric (described in the box on page 6) has post-
ed each student’s individual on-track status on a large board in his office. This has generated significant excitement from the
students, who are now actively engaged in their on-track status and sometimes update their own status on the board.
At T. C. Williams High School in Alexandria, Virginia, students’ reading scores are sent home to parents each quarter.
With timely and easy-to-understand student performance data, parents are more willing to accept intervention decisions,
such as requiring a student to give up an elective course or participate in afterschool sessions in order to receive more lite-
racy instruction. (See box on page 5.)
In summer 2008, Connected by 25, a coalition in Portland, Oregon, of thirty-five education, business, and community or-
ganizations, will launch the Ninth Grade Counts program, which will target 1,500 of Portland’s most at-risk students (as
identified through the use of early-warning data). Using a concentrated marketing campaign, Connected by 25 has success-
fully communicated the importance of helping these students—future Portland citizens and workers—graduate from high
school ready for work and college. As a result, the local partners are committing time, funding, and staffing resources to
improve outcomes for the city’s most at-risk students.
Sources: R. Genn, M. Riddile, and T. Somerville, personal communication, 2008.

Policies to Support the Implementation of Early-Warning Systems

There is no question that the real test for using early- tise, and technological tools—to analyze student da-
warning data exists at the school level, where teach- ta, select indicators and triggers, identify at-risk
ers and other school staff interact directly with students, communicate this information to necessary
students. However, most schools and educators stakeholders, and train and support school staff to
already have too much to do. There might be consi- maximize the power of these systems.
derable enthusiasm for putting an early-warning data Partners and policymakers at all levels can—and
and intervention program in place, but enthusiasm must—play valuable roles in the development and
alone is not enough. Many schools simply do not use of early-warning data systems to improve
have the internal capacity—including time, exper- student outcomes.

10
 information to school leaders. By aggregating data
1. Building the infrastructure to predict future beyond the school, district and state officials can
dropouts identify high-need schools and districts or even
Implicit in the application of early-warning data feeder patterns of low-performing middle and
systems is the existence of and access to necessary elementary schools.
data. Current efforts across the country at the state
and district levels have demonstrated that data sys- In states with high-quality statewide data systems
tems can be expensive to build and maintain over that serve as a central repository for district infor-
time. By investing in systems that respond to both mation, state education agencies are also well
short- and long-term needs, policymakers can help positioned to evaluate and compare intervention
ensure that educators have the most up-to-date in- programs across the state to ensure that student
formation and tools to make informed decisions. outcomes are improving, and to identify district-
or statewide patterns.
As states and districts scurry to respond to the call
for data-driven decisionmaking and develop data 2. Building capacity to implement early-
systems to meet their needs, there is increasing warning and intervention systems
concern about the lack of alignment between those Even when technical and data issues are ad-
systems. This is often because state systems and dressed, educators may need help in using early-
district systems collect and house very different warning data to improve student achievement and
kinds of information. In the case of early-warning outcomes. As one Boston educator noted, “Having
systems, end-of-year statewide test scores⎯the outside people come into a district and help look at
student performance data most commonly col- data is crucial.”42 External support organizations
lected at the state level⎯will not suffice. States that have expertise in data analysis can serve as
interested in analyzing or providing early-warning valuable partners to schools, districts, and states
data need to include real-time early-warning data managing and using data to identify at-risk stu-
usually only stored at the local level⎯such as dents. In fact, in nearly every example in this brief
course attendance, failures and grades, and test of a school or district early-warning data system,
scores⎯in the statewide system. This is not the the school or district has benefited from this type
case in most states; in fact, of all the early-warning of support from a nonprofit organization, school
data systems identified in the preparation of this reform organization, institution of higher educa-
brief, only Louisiana’s (see box on page 12) de- tion, or private-sector partner.
monstrates an alignment between the state and
district systems that allows for the sharing of ear- With or without the support of external partners,
ly-warning data. finding time to analyze data and collaborate with
colleagues can be a challenge for educators. Ac-
Developing an early-warning system can require cording to nonprofit technical assistance provider
schools to overcome a range of technical Learning Points Associates, “Schools that are
issues, such as purchasing or building data systems committed to using data to guide their work allo-
and working with vendors. There are cate time for teachers to meet, discuss, reflect
also numerous considerations related to the data it- upon data, and make informed instructional deci-
self, including accessing necessary information sions. Schools identify the need for this time, then
from multiple systems and conducting the analyses find it through a combination of creative schedul-
to identify indicators and set ing ... and priority setting.”43 School and district
triggers. policies, such as those involving school structure
and school improvement planning, must recognize
By handling some of these issues at the district or this need.
state levels, officials can reduce burdens on
schools and maximize resources. This support can Educators are more likely to use data when it is
include purchasing new data systems, streamlining communicated clearly and in a way that allows
existing ones, or embedding technology that au- them to translate it directly into action. States, dis-
tomatically analyzes the data and reports the tricts, and partner organizations can provide a

11
 valuable service by creating easy-to-use tools for funding, have supported pilot early-warning and
understanding the data, such as New Visions for intervention programs in Chicago, Philadelphia,
Public Schools’ color-coded metric, Philadelphia Louisiana, and Ohio.
Education Funds’ early-warning indicator, and the
CCSR on-track measure. The Chicago, Louisiana, Political leaders’ bully pulpits can facilitate coop-
and Philadelphia pilot intervention systems go eration between agencies or school districts and
even further by automatically providing school strengthen communication among or build joint ef-
leaders with lists of students who have been identi- forts between districts. For example, observers of
fied as being at risk. Louisiana’s implementation of a statewide effort
(described in the box below) have noted that the
State and district staff and partner organizations passion for the project on the part of the governor,
that specialize in school improvement also can the superintendent, and a state board of education
help identify appropriate interventions to use with member were critical to quickly implementing
given groups of students. The intervention guides such a large and technically and politically chal-
offered by the Philadelphia and Chicago pilot pro- lenging project.
grams connect the data to intervention types and
categories, but leave decisionmaking to individual The involvement of political leaders also increases
educators. This frees up teachers and other school the visibility of the data and the intervention ef-
staff to maximize their own expertise—putting in- forts, leading to increased interest by the public
terventions and instructional approaches into and an expectation that there will be accountability
action. for results. For example, on-track data is now in-
cluded as part of the school report card that is
3. Bringing efforts to scale mailed home from Chicago high schools to par-
Political leadership at the district or state level can ents. Recent legislation passed by the Indiana
also be a powerful driver for initiating and expand- legislature requires high schools to report the
ing early-warning and intervention efforts. The number of ninth graders who fail to earn enough
commitment of new, dedicated funding or estab- credits for promotion to the tenth grade on their
lishment of appropriate policies can leverage school’s report card, and to implement a compre-
action at the local level or replicate existing ef- hensive intervention strategy to improve student
forts. New initiatives, accompanied by new transitions and graduation rates.

Louisiana’s Dropout Early Warning Systems (DEWS)

As attention increasingly focused on the state of high schools nationally, Louisiana’s low graduation rate—61 percent,
according to independent estimates—could not be ignored. State leaders have recently taken a number of steps to
improve student outcomes; in 2007, the Louisiana Department of Education (DOE) decided to develop and implement
the Louisiana Dropout Early Warning Systems (DEWS). Just one year later, with the help of state grants, half of the
state’s districts (44) and special and charter schools (11) have purchased the early-warning system and begun to
implement a complementary intervention system.
By managing the data and technical issues at the state level, the DOE was able to leverage partnerships and resources,
while reducing the burden and duplicative efforts that would have resulted from each school or district doing this work
individually. To develop the data system, the DOE contracted with a regional educational software vendor that already
had an “early at-risk monitoring” module. The vendor and researchers at the state university analyzed historical data to
identify trigger points for four early-warning indicators: student attendance, student course achievement, student
behavior, and student age. School and district leaders automatically receive information about at-risk students twice a
month.
The DOE’s program has also leveraged use of the data to drive interventions. In exchange for the grant, data, and tech-
nology, participating schools are required to generate, analyze, and validate the accuracy of the DEWS data, develop an
intervention program for ninth graders, and focus on students who have recently fallen off track.
Sources: A.Harris, B.Franklin, and W. Loftin, personal communication, 2008.

12
 Federal Policy Implications
As the emerging practices in schools, districts, and in and collaborative time for the appropriate use of
states across the country take hold, federal policymak- data; and
ers should pay close attention to the results and lessons • invest in additional research on early-warning data
learned from their experiences. To support the devel- and evaluation of accompanying interventions and
opment and effective use of early-warning data to practices.
improve student outcomes, federal policy should:
By supporting the innovative use of academic data to
• make an increased investment in statewide data
make decisions about education policy and practice,
systems that also considers the need for high-
federal policy can improve America’s secondary
quality local data systems by leveraging the align-
schools and ensure that every student graduates pre-
ment with such systems;
pared for college, work, and life in the twenty-first
• encourage the use of early-warning and other data
century.
in the school improvement process as part of the
method for differentiating among low-performing
Existing Federal Legislation
schools, designing school improvement plans, and
evaluating school progress; Several pieces of federal legislation have been intro-
duced that have implications for the development and
• use early-warning and other data to prioritize fund- implementation of early-warning indicator systems in
ing for school improvement to those middle and the nation’s schools:
high schools with populations exhibiting risk fac- • Graduation Promise Act (S. 1185 and H.R. 2928)
tors;
• GRADUATES Act (S. 1920 and H.R. 3763)
• build capacity for schools and districts to partner
with external organizations that assist with data • Striving Readers Act of 2007 (S. 958 and H.R. 2289)
analysis and intervention support as part of the • Statewide Data Systems bill (S. 2014) and METRICS
school improvement process; Act (H.R. 3253)
• increase the ability and willingness of educators to • Success in the Middle Act (H.R. 3406 and S. 2227)
use early-warning and other data to inform practice For more information visit www.all4ed.org.
by directing resources to professional development

Conclusion
As the nation has come to recognize the academic, civil systems to develop tiered or differentiated intervention
rights, and economic imperatives of improving high systems has provided some guidance to educators on
school graduation rates, educators, advocates, and poli- what exactly “data-driven decisionmaking” can look
cymakers are increasingly interested in systemic, cost- like in a school or district.
efficient, successful approaches to secondary school
reform. Simultaneously, there are calls for the use of This emerging research and practice should be sup-
data-driven decisionmaking to individualize instruction ported by educators, researchers, advocates, and
and ensure that all students graduate ready for college policymakers across the country. In addition, it must be
and work. widely adapted, replicated, and further studied to ensure
that more students manifesting early-warning signs re-
Early-warning data is one means to advance that objec- ceive immediate assistance to avoid falling through the
tive. No longer should socioeconomic factors and life cracks of the education system and graduating from
experiences outside the schools be considered the sole high school with a meaningful diploma and an assur-
barometer of the likelihood of a student graduating ance that they are prepared for the challenges and
from high school. And the use of early-warning data opportunities of college, work, and life.

Lyndsay Pinkus is a senior policy associate at the Alliance for Excellent Education.

This brief was made possible by a grant from Carnegie Corporation of New York. The statements made and views expressed
are solely the responsibility of the author.

13
Endnotes
1
Editorial Projects in Education, Diplomas Count: The Graduation Project 2008 (Editorial Projects in Education, 2008).
2
E. M. Allensworth and J. Q. Easton, What Matters for Staying On-Track and Graduating in the Chicago Public High Schools (Consortium on Chicago School
Research at the University of Chicago, 2007); C. D. Jerald, Identifying Potential Dropouts: Key Lessons for Building an Early-Warning Data System: A Dual
Agenda of High Standards and High Graduation Rates, white paper prepared for Staying the Course: High Standards and Improved Graduation Rates, a joint
project of Achieve and Jobs for the Future, funded by Carnegie Corporation of New York (2006).
3
The research and best practice highlighted in this report are included as examples and not intended to be exhaustive.
4
M. Janosz, I. Archambault, J. Morizot, and L. S. Pagani, “School Engagement Trajectories and Their Differential Predictive Relations to Dropout,” Journal of
Social Issues 64, no. 1 (2008): 21–40, http://www.blackwell-synergy.com/doi/pdf/10.1111/j.1540-4560.2008.00546.x (accessed on March 3, 2008).
5
M. Roderick, The Path to Dropping Out (Westport, CT: Auburn House, 1993).
6
R. C. Neild, R. Balfanz, and L. Herzog, “An Early-Warning System,” Educational Leadership 65, no. 2 (2007): 28–33.
7
R. C. Neild and R. Balfanz, Unfulfilled Promise: The Dimensions and Characteristics of Philadelphia’s Dropout Crisis, 2000–2005 (Baltimore: Center for
Social Organization of Schools, Johns Hopkins University, 2006).
8
Parthenon Group, Boston Public Schools: Strategic Planning to Serve Off-Track Youth (Data Review and Strategic Implications, 2007).
9
Allensworth and Easton, What Matters for Staying on Track.
10
Jerald, Identifying Potential Dropouts.
11
S. Battin-Pearson, R. D. Abbott, K. G. Hill, R. F. Catalano, J. D. Hawkins, and M. D. Newcomb, “Predictors of Early High School Dropout: A Test of Five
Theories,” Journal of Educational Psychology 92, no. 3 (2000): 568–82; C. Hammond, D. Linton, J. Smink, and S. Drew, Dropout Risk Factors and Exemplary
Programs (Clemson, SC: National Dropout Prevention Center, Communities in Schools, Inc., 2007); L. Kennelly and M. Monrad, Approaches to Dropout
Prevention: Heeding Early Warning Signs with Appropriate Interventions (Washington, DC: National High School Center, 2007); G. Montes and C. Lehmann,
Who Will Drop Out from School? Key Predictors from the Literature (Rochester, NY: Children’s Institute, 2004); National Center for Education Statistics,
National Education Longitudinal Study of 1988: Characteristics of At-Risk Students in NELS:88 (Washington, DC: U.S. Department of Education, 1992).
12
R. Balfanz, personal communication, June 2008.
13
Allensworth and Easton, What Matters for Staying on Track.
14
J. Railsback, Increasing Student Attendance: Strategies from Research and Practice (Portland, OR: Northwest Regional Educational Laboratory, 2004).
15
Hammond et al., Dropout Risk Factors; National Center for Education Statistics, National Education Longitudinal Study of 1988; Railsback, Increasing
Student Attendance.
16
National Center for Education Statistics, National Education Longitudinal Study of 1988.
17
Consortium on Chicago School Research, Freshman Year: The Make-It or Break-It Year (Chicago Consortium on School Research, 2007).
18
J. M. Bridgeland et al., The Silent Epidemic (Washington, DC: Civic Enterprises, 2006).
19
Railsback, Increasing Student Attendance.
20
G. Biancarosa and C. E. Snow, Reading Next: A Vision for Action and Research in Middle and High School Literacy, report to Carnegie Corporation of New
York (Washington, DC: Alliance for Excellent Education, 2004).
21
R. Balfanz, J. M. McPartland, and A. Shaw, Re-conceptualizing Extra Help for High School Students in a High Standards Era (Baltimore: Center for Social
Organization of Schools, Johns Hopkins University, 2002).
22
M. Riddile, personal communication, June 11, 2008.
23
R. C. Neild and L. Herzog, “The Secondary Continuum—Keeping 6th–12th Grade Students on Track to Graduation,” paper presented to the 4th Annual
Middle Grades Matter Conference, May 25, 2007, www.philaedfund.org/mgm/MGM%202007%20Neild%20and%20Herzog.ppt (accessed on March 3, 2008).
24
P. Ponder, personal communication, March 25, 2008.
25
R.Thibodeaux, personal communication, February 29, 2008.
26
B. Franklin, personal communication, March 6, 2008.
27
Jerald, Identifying Potential Dropouts; R. Balfanz, L. Herzog, and D. J. Mac Iver, “Preventing Student Disengagement and Keeping Students on the Gradua-
tion Path in Urban Middle-Grades Schools: Early Identification and Effective Interventions,” Educational Psychologist 42, no. 4 (2007): 223–35.
28
Jerald, Identifying Potential Dropouts.
29
Balfanz, Herzog, and Mac Iver, “Preventing Student Disengagement.”
30
E. Allensworth, personal communication, June 17, 2008.
31
A. Horwitz and J. Snipes, Supporting Successful Transitions to High School (Washington, DC: Council of the Great City Schools, 2008).
32
E. Allensworth, personal communication, June 17, 2008.
33
Jerald, Identifying Potential Dropouts.
34
Balfanz, Herzog, and Mac Iver, “Preventing Student Disengagement.”
35
J. D. Hoye and C. Sturgis, The Alternative Pathways Project: A Framework for Dropout Reduction and Recovery (Chicago: Alternative Pathways Project,
2005).
36
Balfanz, Herzog, and Mac Iver, “Preventing Student Disengagement.”
37
R. Balfanz, personal communication, June 19, 2008.
38
R. Thibodeaux, personal communication, February 29, 2008.
39
R. Thibodeaux, personal communication, June 11, 2008.
40
P. Ponder, personal communication, February 29, 2008.
41
Symposium on whole school reform, Alliance for Excellent Education, May 22, 2008.
42
L. M. Myatt, remarks at “Turning Around Low-Performing High Schools: Lessons for Federal Policy from Research and Practice,” Alliance for Excellent
Education briefing, August 16, 2007.
43
Learning Points Associates, Guide to Using Data in School Improvement Efforts: A Compilation of Knowledge from Data Retreats and Data Use at Learning
Points Associates (Learning Points Associates, 2004).

14