1 New York State Growth Model for Educator Evaluation 2011–12 July 2012 PRESENTATION as of 7/9/12

2 2 Today’s Agenda  Background  The What, Why, and How of Growth Models and Measures  Using Growth Measures for Educator Evaluation  What Data Will Be Available and When?

3 Background

4 4 Evaluating Educator Effectiveness Student growth on state assessments (state- provided) Student learning objectives Growth 20% Student growth or achievement Options selected through collective bargaining Locally Selected Measures 20% Rubrics Sources of evidence: observations, visits, surveys, etc. Other Measures 60%

5 The What, Why, and How of Growth Models and Measures

6 6 By the End of This Section….  You should be able to: –Explain why the state is measuring student growth and not achievement –Describe how the state is measuring growth compared to similar students –Define a student growth percentile and mean growth percentile

7 7 Prior Year Performance for Students in Two Teachers’ Classrooms ─ Proficiency

8 8 Current Year Performance of Same Students ─ Proficiency

9 9 Prior and Current Year Performance for Ms. Smith’s Students Ms. Smith’s Class Prior ScoreCurrent Score Student A Student B Student C Student D Student E600650

ELA Scale Score Student A 450 High SGPs Low SGPs Student A’s Current Year Performance Compared to “Similar” Students If we compare student A’s current score to other students who had the same prior score (450), we can measure her growth relative to other students. We describe her growth as a “student growth percentile” (SGP). Student A’s SGP is the result of a statistical model and in this example is 45, meaning she performed better in the current year than 45% of similar students.

Comparing Performance of “Similar” Students Prior Year Score Current Year Score Given any prior score, we see a range of current year scores, which give us SGPs of 1 to 99.

SGPs for Ms. Smith’s Students Ms. Smith’s Class Prior Score Current Score SGP Student A Student B Student C Student D Student E

Student Growth Percentiles: True or False? 1.A student with an SGP of 50 performed better than 50% of similar students. 2.A student with an SGP of 80 must be proficient. 3.A student with an SGP of 20 grew less than a student with an SGP of The highest SGP that a student can receive is A student with an SGP of 80 grew twice as much as a student with an SGP of 40.

From Student Growth to Teachers and Principals Ms. Smith’s Class SGP Student A45 Student B40 Student C70 Student D60 Student E40 To measure teacher performance, we find the mean growth percentile (MGP) for his or her students. To find an educator’s mean growth percentile, take the average of SGPs in the classroom. In this case: Step 1: =255 Step /5=51 Ms. Smith’s mean growth percentile (MGP) is 51, meaning on average her students performed better than 51% of similar students. A principal’s performance is measured by finding the mean growth percentile for all students in the school.

Which Students Count in a Teacher’s or Principal’s MGP for 2011–12? Student has valid test scores for at least 2011–12 and 2010–11 Student scores do not count for 2011–12 Yes Student meets continuous enrollment standard for 2011–12 No Student growth is attributed to the teacher and the school Yes No Expected for 2012–13: students weighted by duration of instructional linkage

From Student Growth to Teachers and Principals  In order for an educator to receive a growth score, he or she must have a minimum sample size of 16 student scores in ELA or mathematics across all grades taught. Examples: –A teacher has a self-contained classroom with 8 students who take the 4th grade ELA and math assessments; this teacher would then have 16 student scores contributing to his or her growth score. –A teacher has a class with 12 students in varied grades (4th, 5th, 6th) who take the ELA and math assessments for their respective enrolled grade level; this teacher would then have 24 student scores contributing to his or her growth score.  If an educator does not have 16 student scores, he or she will not receive a growth score from the state and will not receive information in the reporting system. –Educators likely to have fewer than 16 scores should use student learning objectives (SLOs).

MGPs and Statistical Confidence 87 Confidence Range Upper Limit Lower Limit MGP NYSED will provide a 95% confidence range, meaning we can be 95% confident that an educator’s “true” MGP lies within that range. Upper and lower limits of MGPs will also be provided. An educator’s confidence range depends on a number of factors, including the number of student scores in their MGP and the variability of student performance in the classroom.

Pause and Reflect: Mean Growth Percentiles  We talked about: –How to find a mean growth percentile (MGP) –How to interpret an MGP –What students are counted in an MGP –How many student scores are needed to provide an MGP –How a measure of statistical confidence (upper and lower limits of a 95% confidence range) will be provided with MGPs and why

Expanding the Definition of “Similar” Students  So far we have been talking about “similar” students as those with the same prior year assessment score  We will now add two additional features to the conversation:  Two additional years of prior assessment scores –Remember—a student MUST have current year and prior year assessment score to be included  Student-level factors – Economic disadvantage (ED) – Students with disabilities (SWDs) – English language learners (ELLs)

Adjustments for Three Student-Level Factors in Measuring Student Growth Student performance Teacher Instruction Other factors (12–13) Economic disadvantage Language proficiency Disability

ELA Scale Score Student A 450 High SGPs Low SGPs Student A’s Current Year Performance Compared to “Similar” Students If we compare student A’s current score to other students who had the same prior score (450), we can measure his or her growth relative to other students. We describe that growth as a student growth percentile (SGP). Student A’s SGP is the result of a statistical model and in this example is 45, meaning student A performed better in the current year than 45% of similar students.

ELA Scale Score Student A 450 High SGPs Low SGPs Expanding the Definition of “Similar” Students to Include Economically Disadvantaged—An Example Now if student A is economically disadvantaged, we compare student A’s current score to other students who had the same prior score (450) AND who are also economically disadvantaged. In this new comparison group, we see that student A now has an SGP of 48.

Further Information on Including Student Characteristics in the Growth Model  The following slides were developed using sample data from 2010–2011. –The “combined” MGPs on the charts have been calculated at the educator level (combining all grades and subjects). –Not all districts provided data linked to teachers for grades 4–8 ELA/Math in 2010–11.

Teacher MGPs after Accounting for Economic Disadvantage Taking student-level characteristics into account helps ensure educators with many students with those characteristics have a fair chance to achieve high or low MGPs. For example, note that for teachers with any percent of economically disadvantaged students, teacher MGPs range from 1 to 99. NOTE: Beta results using available 2010–11 data.

Teacher MGPs after Accounting for SWD NOTE: Beta results using available 2010–2011 data.

Teacher MGPs after Accounting for ELL Percent of ELL Students in Class NOTE: Beta results using available 2010–2011 data.

“Similar” Students: A Summary “Similar” Student Characteristics Unadjusted Mean Growth Percentiles Adjusted Mean Growth Percentiles Up to Three Years of Prior Achievement English Language Learner (ELL) Status Students with Disabilities (SWD) Status Economic Disadvantage Reported to Educators Used for Evaluation

One Last Feature of the Growth Model…. All tests contain measurement error, with greater uncertainty for highest and lowest achieving students The New York growth model accounts for measurement error in computing student growth percentiles.

State Growth Model Summary Regulations allow Prior years of student test results Three student- level variables: SWD, ELL, Econ Disadvantage Measurement error correction Model includes Up to three years, as available All three Measurement error correction Growth model for 2011–12 only for grades 4–8 ELA/Math for teachers and principals

By the End of This Section….  You should be able to: –Explain why the state is measuring student growth and not achievement –Describe how the state is measuring growth compared to similar students –Define a student growth percentile and mean growth percentile

31 Using Growth Measures for Educator Evaluation

By the End of This Section….  You should be able to: –Explain how growth ratings and scores will be obtained, using illustrative data

Growth Ratings and Score Ranges Growth RatingDescriptionGrowth Score Range (2011–12) Highly Effective Well above state average for similar students 18–20 EffectiveResults meet state average for similar students 9–17 DevelopingBelow state average for similar students 3–8 IneffectiveWell below state average for similar students 0–2 The growth scores and ratings are based on an educator’s combined MGP.

Distribution of 2010–11 Teacher-Level MGPs MGP Number of Teachers NOTE: Beta results using available 2010–2011 data. For illustrative purposes only Distribution of Mean Student Growth Percentiles (Teacher Level) Percent of MGPs

MGPs and Statistical Confidence 87 Confidence Range Upper Limit Lower Limit MGP NYSED will provide a 95% confidence range, meaning we can be 95% confident that an educator’s “true” MGP lies within that range. Upper and lower limits of MGPs will also be provided. An educator’s confidence range depends on a number of factors, including the number of student scores included in his or her MGP and the variability of student performance in the classroom.

HEDI Classification Approach for Teachers (using 2010–11 sample data)  Effective requires MGPs within 1 standard deviation of the average MGP of 51. –MGPs between 40 and 61 will earn Effective ratings.  Well Above Average (Highly Effective) requires –MGP of 62 or higher –AND confidence range above 51. (If not, rating is Effective.)  Well Below Average (Ineffective) requires –MGP of 39 or lower –AND confidence range must be less than 51. (If not, rating is Developing.)

Yes No Yes From MGPs to Growth Ratings: Teachers Mean Growth Percentile ≥62 Lower Limit > 51 Highly Effective: Results are well above state average for similar students Mean Growth Percentile ≤39 Upper Limit < 51 Ineffective: Results are well below state average for similar students Developing: Results are below state average for similar students No Effective: Results equal state average for similar students Mean Growth Percentile Confidence RangeGrowth Rating Mean Growth percentile 40–61 Yes Any Yes

HEDI Classification Approach for Principals (using 2010–11 sample data) Same methodology as for Teachers. Slightly different cut scores.  Effective requires MGPs within 1 standard deviation of the average MGP of 50. –MGPs between 43 and 57 will earn Effective ratings.  Well Above Average (Highly Effective) requires: –MGP of 58 or higher –AND confidence range above 50. (If not, rating is Effective.)  Well Below Average (Ineffective) requires –MGP of 42 or lower –AND Confidence Range must be less than 50. (If not, rating is Developing.)

Yes No Yes From MGPs to Growth Ratings: Principals Mean Growth Percentile ≥ 58 Lower Limit > 50 Highly Effective: Results are well above state average for similar students Mean Growth Percentile ≤ 42 Upper Limit < 50 Ineffective: Results are well below state average for similar students Developing: Results are below state average for similar students No Effective: Results equal state average for similar students Mean Growth Percentile Confidence RangeGrowth Rating Mean Growth percentile 43–57 Yes Any Yes

Illustrating Possible Teacher Growth Ratings MGP 1 MGP 99 Well Below Average (39) Average (51) Average (51) Well Above Average (62) Well Above Average (62) MGP

Illustrating Possible Teacher Growth Ratings MGP 1 MGP 99 Well Below Average (39) Well Below Average (39) Average (51) Average (51) Well Above Average (62) Well Above Average (62) MGP

Illustrating Possible Teacher Growth Ratings Ineffective Developing Highly Effective Effective MGP 1 MGP 99 Well Below Average (39) Well Below Average (39) Average (51) Average (51) Well Above Average (62) Well Above Average (62) MGP Developing

Illustrating Possible Teacher Growth Ratings Effective MGP 1 MGP 99 Well Below Average (39) Well Below Average (39) Average (51) Average (51) Well Above Average (62) Well Above Average (62) MGP

Illustrative Results: Teachers (Using 2010–11 sample data) Rating & Points (2011–12 ) Number of Teacher MGPs Percent of Teacher MGPs Highly Effective 18– % Effective 9–17 16,68176% Developing 3– % Ineffective 0– % Points available within each HEDI category will be assigned based on educator MGP

Illustrative Results: Principals (Using 2010–11 sample data) Rating & Points (2011–12) Number of Principal MGPs Percent of Principal MGPs Highly Effective 18– % Effective 9– % Developing 3– % Ineffective 0–2 2417% Points available within each HEDI category will be assigned based on educator MGP

By the End of This Section….  You should be able to: –Explain how growth ratings and scores are obtained

47 What Data Will Be Available and When

Data — What to Expect When Growth scores provided to districts Mid-July Test scores finalized and teacher linkage data final submission Early fall Mid-August Online reporting system available

Data — What to Expect in August Data Elements (for teachers and schools)  Unadjusted mean growth percentiles (Unadjusted MGPs)  Adjusted mean growth percentiles (Adjusted MGPs and upper and lower limits based on confidence range for these adjusted MGPs)  Percent of students above the State median: this will be provided at the teacher and school level, and can be used as a local measure in APPR  Number of student scores included  Growth rating (HEDI)  Growth score (0–20) Breakdowns (by teacher and school)  MGPs by subject, grade, and overall (not HEDI) –Can be used with SLOs as part of the Comparable Measures or Locally Selected Subcomponent  Overall MGPs for subgroups — ELL, SWD, Economic Disadvantage, High- and Low-Achieving –Subgroup scores will not be included on reports if there are fewer than 16 student scores

One Teacher’s Information — August Number of Student Scores Percent of Students Above the State Median Unadjusted MGP Adjusted MGP Growth Rating Growth Score Lower LimitUpper Limit Jane Smith Highly Effective 18 Number of Student Scores Percent of Students Above the State Median Unadjusted MGP Adjusted MGP Lower LimitUpper Limit Jane Smith Math Math Grade ELA ELA Grade Students with disabilities 4* **** English language learners 0 * * * * * Economically disadvantaged 2* **** Low achieving (Level 1) 4* **** High achieving (Level 4) 4* ****

Number of Students Percent of Students Above the Median Unadjusted MGP Adjusted MGP Lower LimitUpper Limit Jane Smith Math Math Grade ELA ELA Grade Students with disabilities 4* **** English language learners 0* ** * * Economically disadvantaged2* **** Low achieving 4* **** High achieving 4* **** Adjusted MGP: 75 Number of Students Percent of Students Above the Median Unadjusted MGP Adjusted MGP Growth Rating Growth Score Lower LimitUpper Limit Jane Smith Highly Effective 18 2 SWD students, 0 ELL students, 1 econ disadvantaged student, 2 high- and 2 low-achieving students Math number of student scores: 28 ELA number of student scores: 28 Total number of student scores: 56 Upper and lower limits of adjusted MGP: 65 and 85 Growth rating of highly effective and growth score of percent of students above the State Median Unadjusted MGP: 70 One Teacher’s Information — August Adjusted MGPs by subject—can be used in an SLO for the Comparable Measures subcomponent No scores reported here since fewer than 16 student scores in a sub-group Unadjusted MGPs by subject

Scavenger Hunt and Quiz 1.What is Ms. Smith’s overall adjusted MGP? 2.What are the upper and lower confidence limits for Ms. Smith’s overall MGP and what do they represent? 3.How many scores are included from Ms. Smith’s class for ELA? 4.What is the adjusted MGP for Ms. Smith’s class in ELA? 5.How do Ms. Smith’s high- achieving students compare to her low-achieving students in terms of growth? 6.What score is Ms. Smith’s growth rating based on?

Definitions  SGP (student growth percentile): the result of a statistical model that calculates each student’s change in achievement between two or more points in time on a state assessment or other comparable measure and compares each student’s performance to that of similarly achieving students  Similar students: students with the same prior test scores, ELL, SWD, and economic disadvantage status  ELLs: English language learners  SWD: students with disabilities  Economic disadvantage: a student who participates in, or whose family participates in, economic assistance programs such as the Free- or Reduced-price Lunch Programs (FRPL), Social Security Insurance (SSI), Food Stamps, Foster Care and others

Definitions  High-achieving, low-achieving: defined by the performance of students based on prior year state assessment scores (i.e., Level 1 = low-achieving, Level 4 = high-achieving)  MGP (mean growth percentile): the average of the student growth percentiles attributed to a given educator  “Unadjusted” MGP: an MGP based on SGPs that have NOT accounted for ELL, SWD, and economic disadvantage status  “Adjusted” MGP: an MGP based on SGPs that HAVE accounted for ELL, SWD, and economic disadvantage status  Growth rating: HEDI rating based on growth  Growth score: growth subcomponent points from 0–20

Definitions  Measurement error: uncertainty in test scores due to sampling of content and other factors  Standard error: a measure of the statistical uncertainty surrounding a score  Standard deviation: a measure that shows the spread of scores around the mean  Upper/lower limit: highest and lowest possible MGP taking statistical confidence into account  Confidence range: range of MGPs within which we have a given level of statistical confidence that the true MGP falls (95% statistical confidence level used for state growth measure)