Teacher Evaluation and Performance Measurement
Doug Staiger, Dartmouth College
Not this.
2Weisberg, D., Sexton, S., Mulhern, J. & Keeling, D. (2009) The Widget Effect: Our National Failure to Acknowledge and Act on Differences in Teacher Effectiveness. New York: The New Teacher Project.
Satisfactory (or equivalent) Unsatisfactory (or equivalent)
3
Not this.
4
Transformative Feedback
Recent Work on Teacher Evaluation
Efforts to identify effective teaching using achievement gains– Work with Tom Kane & others in LAUSD, NYC, Charlotte…
www.dartmouth.edu/~dstaiger
Efforts to better identify effective teaching– Measures of Effective Teaching (MET) Project
(Bill & Melinda Gates Foundation)www.metproject.org
– National Center for Teacher Effectiveness (NCTE)(US Department of Education)www.gse.harvard.edu/ncte
5
The Measures of Effective Teaching Project
• Two school years: 2009-10 and 2010-11• Grades 4-8: ELA and Math• High School: ELA I, Algebra I and Biology
Participating Teachers
The MET data is unique … in the variety of indicators tested,
5 instruments for classroom observations (use FFT here)
Student surveys (Tripod Survey)
Value-added on state tests
in its scale,3,000 teachers
22,500 observation scores (7,500 lesson videos x 3 scores)
900 + trained observers
44,500 students completing surveys and supplemental assessments in year 1
3,120 additional observations by principals/peer observers in Hillsborough County, FL
and in the variety of student outcomes studied. Gains on state math and ELA testsGains on supplemental tests (BAM & SAT9 OE)
Student-reported outcomes (effort and enjoyment in class, grit)
7
What is “Effective” Teaching? Can be an inputs based concept
– Observable actions or characteristics
Can be outcomes based concept– Measured by student success
Ultimately, care about impact on student outcomes– Current focus on standardized exams– Interest in other outcomes (college, non-cognitive)
8
Multiple Measures of Teaching Effectiveness
9
10
Measure #1Student Achievement Gains
(“Value Added”)
11
Basics of Value Added Analysis Teacher value added compares actual student achievement at the
end of the year to an expectation for each student
Difference between actual and expected achievement, averaged over all of teacher’s students
Expected achievement is typical achievement for other students who looked similar at start of year– Same prior-year test scores– Same demographics, program participation– Same characteristics of peers in classroom or school
Various flavors, all work similarly– Student growth percentiles– Average change in score or percentile– Based on prior year test or Fall pre-test
There are Large Differences in Teacher Effects on Student Achievement Gains
Most evidence from “value added” analysis, but similar findings from randomized experiments
Huge literature about “teacher effects” on achievement– Large persistent variation across teachers– Difficult to predict at hire– Partially predictable after hire– Improve only in the first few years of teaching– Not related to most determinants of pay
• Certification, degrees, experience beyond first few years
Large Variation in Value Added of LAUSD Teachers is Not Related to Teacher Certification
0.0
3.0
6.0
9.1
2P
ropo
rtion
of C
lass
room
s
-15 -10 -5 0 5 10 15Change in Percentile Rank of Average Student
Traditionally Certified UncertifiedAlternatively Certified
Note: Classroom-level impacts on average student performance, controlling for baseline scores,student demographics and program participation. LAUSD elementary teachers, grade 2 through 5.
by Initial CertificationTeacher Impacts on Math Performance
Variation in Value Added of LAUSD Teachersis Related to Prior Performance
0.0
3.0
6.0
9.1
2P
ropo
rtion
of C
lass
room
s
-15 -10 -5 0 5 10 15Change in Percentile Rank of Average Student
Bottom 3rd Quartile2nd Quartile Top Quartile
Note: Classroom-level impacts on average student performance, controlling for baseline scores,student demographics and program participation. LAUSD elementary teachers, < 4 years experience.
by Ranking After First Two YearsTeacher Impacts on Math Performance in Third Year
Why Not Just Hire Good Teachers? Wise selection is the best means of improving the
school system, and the greatest lack of economy exists wherever teachers have been poorly chosen.
• Frank Pierrepont Graves, NYS Commissioner, 1932
Unfortunately, easier said than done– Decades of work on type of certification, graduate
education, exam scores, GPA, college selectivity, TFA– (Very) small, positive effects on student outcomes
Large Variation in Value Added of NYC Teachers is Not Related to Recruitment Channel
0
24
68
Ker
nel D
ensi
ty E
stim
ate
-.4 -.3 -.2 -.1 0 .1 .2 .3 .4Student Level Standard Deviations
Traditionally Certified Teaching FellowTeach for America Uncertified
Note: Shown are estimates of teachers' impacts on average student performance, controlling for teachers' experience levels and students' baselinescores, demographics and program participation; includes teachers of grades 4-8 hired since the 1999-2000 school year.
Of Course, Teacher Impact on State Test Score is Not All We Care About
Depends on design & content of test
Test scores are proximate measures– But recent evidence suggests they capture long-
run impact on student learning and other outcomes
Test scores are only one dimension of performance– Non-cognitive skills (grit, dependability, …)
Value Added is Controversial “We need to find a way to measure classroom
success and teacher effectiveness. Pretending that student outcomes are not part of the equation is like pretending that professional basketball has nothing to do with the score.” (Arne Duncan 2009)
“There is no way that any of this current data could actually, fairly, honestly or with any integrity be used to isolate the contributions of an individual teacher.” (Randi Weingarten 2008)
18
19
What we learned from MET:Value-added measures
• Identified teachers who caused students to learn more on state tests following random assignment.
• Same teacher’s also caused students to learn more on supplemental assessments and enjoy class more.
• Low year-to-year correlations in value-added (and other performance measures) understate year-to-career correlations.
20
21
-.1-.0
50
.05
.1A
ctua
l Ach
ieve
men
t afte
r Ran
dom
Ass
ignm
ent
-.1 -.05 0 .05 .1Predicted achievement using teacher's past measures of teaching.
Note: Teachers were sorted into 20 groups by their predicted student achievement relative to the randomization group mean. Means are reported for each of the 20. Predictions are adjusted for non-compliance.
of Randomized Classrooms (Math)Figure 1. Actual and Predicted Achievement
Actual = Predicted
22
-.1-.0
50
.05
.1A
ctua
l Ach
ieve
men
t afte
r Ran
dom
Ass
ignm
ent
-.1 -.05 0 .05 .1Predicted achievement using teacher's past measures of teaching.
Note: Teachers were sorted into 20 groups by their predicted student achievement relative to the randomization group mean.
Means are reported for each of the 20. Predictions are adjusted for non-compliance.
of Randomized Classrooms (ELA)Figure 2. Actual and Predicted Achievement
Actual = Predicted
23
Measure #2Classroom Observations
Classroom Observation Using Digital Video
24
Access to Validation Engine:
What you can expect from us:
SEA/LEA chooses a rubric, trains
raters
The MET Project delivers sample
videos
SEA/LEA ratings used to
-Predict value added-Gauge reliability
Helping Districts Test Their Own New Classroom Observations
25
26
Instrument Developer Origin Instructional Focus
Structure Scoring
Framework for Teaching
Charlotte Danielson
Outgrowth ofETS’s PRAXIS III licensing exam
Constructivism
Intellectual Engagement
4 domains; 22 components
MET uses 8 components*
4 Points
Classroom Assessment Scoring System (CLASS)
Robert Pianta, Univ. of Virginia
Tool for research on early childhood development
Teacher-student interactions
3 domains; 12 dimensions
7 Points
Two Cross-Subject Observation Instruments
*not: “flexibility & responsiveness” & “organization of physical space”
27
FFT competencies scored:
CLASSROOM ENVIRONMENTCreating an environment of respect and rapport Establishing a culture of learning Managing classroom procedures Managing Student Behavior
INSTRUCTION
Communicating with Students Using Questioning and Discussion Techniques Engaging Students in Learning Using Assessments in Instruction
28
Instrument Developer Origin Instructional Focus
Structure Scoring
Mathematical Quality of Instruction (MQI)
Heather Hill, Harvard
Outgrowth from written test of math teaching knowledge
Math errors and imprecision
6 overall elements of instruction
3 Points
UTEACH Observation Protocol (UTOP)
Michael Marder, Univ. of Texas-Austin
Teacher prep program for math & science majors
Values different modes, from direct instruction to inquiry-based
4 sections; 22 subsections
5 Points
Math Observation Instruments
29
Instrument Developer Origin Instructional Focus
Structure Scoring
Protocol for Language Arts Teaching Observations (PLATO)
Pam Grossman Stanford
Research on effective middle grade ELA instruction
Modeling, explicit teaching of strategies, guided practice
13 elements
6 elements included in MET study
4 Points
ELA Observation Instrument
30
What we learned from MET:Classroom observations:
• Observation scores were correlated with a teacher’s value-added (.15-.27).
• Different instruments were highly correlated with each other (although subject-specific instruments were distinct from the general-pedagogical instruments).
• Reliability requires certified observers and more than one observer per teacher (because rater judgments differ).
• Principals rate their own teachers higher than other observers do, but their rankings are similar.
• When teachers select their own videos, scores are higher, but ranking remains the same.
31
Four Steps
Four Steps to High-Quality Classroom Observations
Actual scores for 7500 lessons.
Step 1: Define ExpectationsFramework for Teaching (Danielson)
32
Four Steps
Unsa
tisfa
ctor
y Yes/no Questions, posed in rapid succession, teacher asks all questions, same few students participate.
Basic
Some questions ask for student explanations, uneven attempts to engage all students.
Profi
cient
Most questions ask for explanation, discussion develops/teacher steps aside, all students participate.
Adva
nced All questions high quality,
students initiate some questions, students engage other students.
Step 2: Ensure Accuracy of Observers
33
Four Steps
Step 3: Monitor Reliability
34
Four Steps
35
More than 1 observer
One more lesson +.07
One more observer +.16
Step 4: Verify Alignment with Outcomes
36
Four Steps
Teachers with Higher Observation Scores Had Students Who Learned More
37
Measure #3What do students say?
38
Students Distinguish Between TeachersPercent of Students by Classroom Agreeing
39
Students Distinguish Between TeachersPercent of Students by Classroom Agreeing
40
Students Distinguish Between TeachersPercent of Students by Classroom Agreeing
41
Students Distinguish Between TeachersPercent of Students by Classroom Agreeing
42
Students Distinguish Between TeachersPercent of Students by Classroom Agreeing
43
What we learned from MET:Student surveys:
• Surveys are a low-cost way to cover untested grades and subjects.
• Student surveys are related to teacher value-added (.15-.25).
• Student surveys are the most reliable measures we tested.
44
Multiple MeasuresThe “Dynamic Trio”:
Classroom observations, student feedback and student achievement gains.
Dynamic Trio
45
Three Criteria:Predictive power: Which measure could most accurately identify teachers likely to have large gains when working with another group of students?Reliability: Which measures were most stable from section to section or year to year for a given teacher?Potential for Diagnostic Insight: Which have the potential to help a teacher see areas of practice needing improvement? (We’ve not tested this yet.)
Measure Predictive power ReliabilityPotential for
Diagnostic Insight
Value-added
Student survey
Observation
Dynamic Trio
Measures have different strengths …and weaknesses
46
H
ML
M
HM/H
L
MH
Dynamic Trio
Combining Measures Improved Reliabilityas well as Predictive Power
47
Note: For the equally weighted combination, we assigned a weight of .33 to each of the three measures. The criterion weights were chosen to maximize ability to predict a teacher’s value-added with other students. The next MET report will explore different weighting schemes.
Observation alone (FFT)
Student survey alone
VA alone
Combined(Equal Weights)
Combined(Criterion Weights)
.05
.1.1
5.2
.25
Diff
eren
ce in
Mat
h VA
(Top
25%
vs.
Bot
tom
25%
)
0 .1 .2 .3 .4 .5 .6 .7Reliability
Note: Table 16 of the research report. Reliability based on one course section, 2 observations.
The Reliability and Predictive Power of Measures of Teaching:
48
What we learned from MET:Combining measures:
• The teachers identified as more effective caused students to learn more following random assignment.
• Combining value added with student surveys and classroom observations produces two benefits: • Increased reliability• Increased correlation with other outcomes such as value-
added on supplemental assessments and happiness in class• Weighting value-added below .33, though, lowered
correlation with other outcomes and lowered reliability.
49
Can the measures be used for “high stakes”?
High-stakes decisions are being made now, with little or no data. No information is perfect, but better information should lead to
better decisions and fewer mistakes.
Scenario 1: Teacher
You have been teaching biology for 10 years and want to improve your practice. What weaknesses should you focus on and how will you know if you're making progress?
Scenario 2: Principal
A probationary teacher in your school is approaching the end of their 2nd year. If you retain him/her, the teacher automatically earns tenure under the collective bargaining agreement. Should you grant tenure (or recruit a new novice teacher)?
Scenario 3: Superintendent
Your district is considering offering coaching opportunities/higher pay to a subset of your teachers. Should you (i) allocate those slots on the basis of seniority, (ii) ensure that only excellent instructors are coaches? How would you measure effectiveness fairly?
50
No information is perfect.
How do these compare to existing measures?
But better information → better decisions
• Masters Degrees• Years of Experience• Classroom Observations Alone
Compared to What?
Compared to MA Degrees and Years of Experience, the Combined Measure Identifies Larger Differences
51
… on state tests
Compared to What?
…and on low stakes assessments
52
Compared to What?
…as well as on student-reported outcomes.
53
54
The Value of Going Beyond Classroom Observation
+
++
• Observations
• Observations• Student
Perceptions
• Observations• Student
Perceptions• VA on state
tests
55
Compared to Classroom Observations Alone, the Combined Measure Identifies Larger Differences (Math Value Added)
-.2-.1
0.1
.2.3
Ave
rage
mat
h Va
lue
Add
ed, O
ther
Cla
ss
0 20 40 60 80 100Percentile Rank on FFT
Rank using FFT only Rank using FFT and TripodRank using FFT, Tripod, and Value Added
56
Improving TeachingWhat are Districts Doing?
Robust evaluation systems themselves improve teaching outcomes
Source: Eric S. Taylor and John H. Tyler, “Can Teacher Evaluation Improve Teaching?” Education Next, Fall 2012
Teacher Effectiveness Continues to Improve in Better Environments
Source: Matthew A. Kraft and John P. Papay, “Can Professional Environments in Schools Promote Teacher Development? Explaining Heterogeneity in Returns to Teaching Experience,” January 2013 (on NCTE website).
59
The Best Foot Forward Project1. Teachers record their own lessons.
• Record ≥1 lesson every 2 weeks.• Submit 5 lessons over course of the year.• Viewed by principals, content experts.
2. Observers view and discuss videos with teachers.• Observers trained to use video for feedback.• Identify discreet, coachable changes.
3. Teachers can share videos with each other.4. Students provide anonymous feedback.
60
Next Up: Dashboard for Tracking Teacher Evaluations and Benchmarking Performance
1. Distribution of Observation Scores: What are the differences in scores and are the differences between schools, districts, grades and subjects larger than might have occurred by chance?
2. Observations and Value-Added: What are the relationships among the different measures? Do they differ by district, school, grade level, subject? Are they weaker/stronger than we observed in MET?
3. Reliability: How does each measure vary from school to school and year to year?
Useful ResourcesAvailable at: http://www.metproject.org/resources.php
Student surveys: Tripod survey and “Asking Students about Teaching Practitioner Brief”
Roster Validation: Report by Battelle for Kids on ways to allow teachers to verify students in their class: “Identifying The Importance of Accurately Linking Instruction to Students to Determine Teacher Effectiveness”
Software for Certifying Observers using Pre-Scored Videos: Certification engine from Empirical Education
Available at: http://www.gse.harvard.edu/ncte/resources/default.php
Classroom Observation: Links to FFT, CLASS, etc., and webinars with six organizations currently supporting classroom observations
Additional examples of sites with useful resources: TNTP: http://tntp.org/ideas-and-innovations
Pearson: http://educatoreffectiveness.pearsonassessments.com/