Two new national studies raise questions about the how accurate modern teacher evaluations are.
The first study, from the University of Southern California’s Morgan Polikoff and the University of Pennsylvania’s Andrew Porter, finds test-based evaluation scores have little to no link to other teacher quality measures, such as how well instruction matches standards and the content of assessments. Their study included data from Hillsborough County schools in Florida.
The scores are known as a value-added model, and use a statistical formula to predict how well a student should score on standardized tests based on past performance. If the student scores higher or lower than predicted, that difference is attributed to the classroom teacher.
Florida law requires teachers are evaluated based on a combination of how much student test scores improve and in-class observations. Florida is one of a growing number of states which requires school districts to pay teachers based on their evaluations.
“If I had my druthers, I would say we need to slow way down the implementation of these teacher evaluation systems because we just don’t know enough about the quality of these measures,” Polikoff says. “And we have reason to believe a lot of the measures actually aren’t very good quality.”
Some previous research has shown stronger correlations between value-added measures and teacher instruction, while others have shown almost no relationships, he says.
“It’s not clear to me what the reasons are for those differences, but as these systems are rolling out, states need to really study these relationships and think about in the cases where the correlations are really low, what can you do with those data?” Polikoff says.
A second study from the Brookings Institution of classroom observations found the teachers who started the school year with higher achieving students earned better scores, on average, than teachers who started the year with lower achieving students. Most districts don’t adjust their evaluations for this bias.
Brookings recommends teachers are observed and evaluated by someone from a different school and that school districts adjust observation scores for class demographics.
The Brookings study also found that schoolwide value-added scores generally have a positive effect for bad teachers in good schools and negatively affects the ratings of good teachers in bad schools.
Florida and 43 other states are switching to new K-12 math and language arts standards based on Common Core, and evaluations are a concern as the deadline approaches this fall. Supporters say the new standards will be more challenging, and fewer students are expected to meet state goals on new tests tied to the standards.
Teachers, superintendents and some parent groups had asked the State Board of Education and lawmakers to suspend using test scores to determine school grades, teacher pay and other decisions to allow teachers to adjust to the new standards. However, lawmakers decided to suspend the consequences for school grades for the 2014-2015 school year only.
Polikoff and Porter’s study looked at 327 fourth and eighth grade math and English teachers in six school districts, including Hillsborough County. The Brookings study looked at four moderate-sized, anonymous urban districts across the country.