August 7, 2014
EVAAS Value-Added System
MYTH: EVAAS is too complicated to understand.
FACT: Basically, EVAAS compares the change in achievement of a group of students from one year to the next with an expected amount of change based on the students’ prior achievement history. Simply put, “added value” refers to how much the student learned during the school year. While the concept of value-added is simple, the application is more complicated. Any value-added model should take into account variables such as errors in measurement; missing test scores; how tests are scaled; and educators serving different types of students. To use student test data in a reliable way, value-added models must be statistically sophisticated.
MYTH: If I can’t understand how EVAAS scores are calculated, how can I use it to improve my teaching?
FACT: You may not know how to build an iPhone, but you can still use it to do amazing things. While it’s true that EVAAS calculations use complex statistics and multiple years of data, the reports on the EVAAS website (through the ASPIRE portal) do provide usable information.
- Diagnostic reports on students based on their incoming achievement level provide data-driven support to ensure all students reach their potential
- Student-projection reports show teachers and administrators which students are on a positive or negative growth trajectory and need interventions or advanced coursework
- Teacher reports help administrators augment the strengths of effective teachers and implement professional development of struggling teachers
- District-level training helps teachers use EVAAS reports to become more effective
MYTH: HISD's EVAAS system is unreliable "because the error bars [graphed data that indicates error] are so large."
FACT: Standard error is a measure of uncertainty around an estimate. Larger standard errors mean that there is more uncertainty about a teacher’s influence on his or her students' academic progress. Factors that may influence standard error include:
- Range of scores in student performance and number of students in the analysis
- Teachers with fewer students will have larger standard errors
- Multi-year estimates are more stable than single-year estimates and therefore have a smaller standard error
- Fair because they do not depend on the types of students a teacher receives each year
- Valid because various challenges in student testing are taken into account
- Reliable because they are repeatable and consistent; multiple-year estimates protect teachers from misclassification
In general, the range of NCE gains at the teacher level is around -15 to 15 in Houston. The standard errors range from 1 to 4 in HISD. This does not appear to be 30-50% (as purported by opponents to EVAAS) from what the data are showing.
We are able to differentiate a good number of teachers with the measures that we are using.Compared to the district, there are approximately 26% of teachers that either fall into well above or well below categories compared to the district average in any given year. Accounting for standard error (and using multiple years of data) keeps us from misclassifying teachers as well above or well below.
MYTH: A teacher can have a low value-added score one year and a high value-added score the next, which means it is basically an unreliable measure of performance.
FACT: This can be true with single-year estimates, because a teacher may be more effective with one year’s group of students over the next. Less effective teachers may also improve over time.For teachers identified as ineffective or highly effective based on three-year estimates, the results will be more consistent. If a teacher is ineffective for three years, the likelihood is higher that they will remain ineffective. Likewise, a highly effective teacher will continue to be effective.
FACT: Evaluation of EVAAS data (from another state) over a 14-year period provided valuable insight into how teachers may change in effectiveness over time.
- Highly effective teachers are very likely to remain effective
- Teachers identified as highly effective after their first three years of teaching were extremely likely to remain effective three years into the future
- For teachers identified as ineffective based on three-year estimates, approximately 50percent will be ineffective three years later
- Some year-to-year variation among individual teachers is to be expected, depending on the variation in student groups from year to year
MYTH: Teachers identified as "highly effective" based on value-added data do not necessarily produce more learning for their students.
FACT: Students with very effective teachers three years in a row were able to improve their performance on standardized tests by more than 50 percent in comparison to students who had ineffective teachers three years in a row (Sanders and Rivers 1996). A similar study conducted in Dallas ISD found similar results. A teacher's impact on student learning lasts up to four years(Sanders 2005). As teacher effectiveness levels increase, lower-achieving students are the first to benefit (Sanders and Rivers 1996). If a student has an ineffective teacher for two years, this decrease in progress cannot be made up (Rivers 1999).
MYTH: HISD administration is intentionally keeping teachers in the dark about EVAAS and has not provided information and training to explain how scores are calculated.
FACT: HISD began offering EVAAS training through a combination of face-to-face and online courses in August 2008. Since then, more than 8,103 employees – most of them teachers –have completed an EVAAS course.
MYTH: HISD has decided that teachers can be terminated based on their EVAAS scores.
FACT: In February 2010, the HISD Board of Education added EVAAS scores to the list of 34factors that can be taken into account in renewal decisions. Employees are offered individualized support and professional development opportunities throughout the year. Prior to non-renewal of a teacher’s contract, the district will review the teacher’s performance in accordance with the 34 factors. EVAAS scores are never the sole factor for non renewal.
MYTH: Under a new teacher-evaluation system, EVAAS scores are 50 to 60 percent of the teacher-evaluation process.
FACT: In accordance with state law, HISD’s new teacher-evaluation system was developed at the campus level through the Shared Decision Making committees and at the district level through the District Advisory Committee. Taking into account recommendations from these committees, working groups that included teachers and administrators came up with proposals for the numbers used in EVAAS.
- For teachers with value-added data, student performance counts for approximately 50percent of a teacher’s summative rating
- Student performance is made up of value-added data, if available, combined with other growth numbers including comparative growth and student progress on district/campus assessments
- For teachers without value-added data (to ensure consistency and rigor across all grades and subjects), student performance counts for approximately 30 percent of their summative rating
MYTH: EVAAS is a quartile system with 50 percent of teachers receiving a negative score every year, which means half of HISD teachers are deemed "ineffective.”
FACT: EVAAS is not a quartile system. EVAAS measures the growth of students at the classroom level relative to expected growth. Originally, the ASPIRE program used a quartile system, but it has largely moved on from this methodology for a variety of reasons. The ASPIRE program awards core teachers with EVAAS scores based on reaching a set standard (1.0 or higher).
MYTH: EVAAS system developer Dr. William L. Sanders is not an expert in the teacher effectiveness measurement field.
FACT: Dr. Sanders spent over two decades in the teacher-effectiveness measurement field as an academic, a researcher and a provider of value-added analyses and reporting. He was a professor at the University of Tennessee Knoxville and then director of the university’s Value-Added Research and Assessment Center. After retiring from that position, he became senior manager of value-added assessment and research for the SAS Institute in Cary, NC and was a senior research fellow with the University of North Carolina system. Dr. Sanders recently retired.
In 2006, Dr. Sanders testified before the House Education and Workforce Committee about the reauthorization of No Child Left Behind. That year, his value-added research was ranked sixth in the Education Research Center’s list of most influential research on the last decade's national educational policy. In February 2007, Dr. Sanders shared his research in a Senate Health,Education, Labor, and Pensions Committee Round Table discussion on teacher incentives.
Over the last 20 years, Dr. Sanders and his colleagues have developed and refined their methodology.
MYTH: The EVAAS methodology has never been independently validated.
FACT: In the RAND Corporation’s 2003 report, “Evaluating Value-Added Models for TeacherAccountability,” the EVAAS model was studied for its validity in measuring teacher effects. The report concluded that models like EVAAS use an approach that “is likely to be preferable but is computationally demanding.” The full Rand report is available here.
More information about ASPIRE and the EVAAS value-added system, is available here.