The content, predictive power, and potential bias in five widely used teacher observation instruments

Many states and districts are adopting commercially available teacher observation instruments for the professional practice component of their evaluation systems. This study was designed to inform decisions about the selection and use of five widely-used teacher observation instruments. The purpose was to explore (1) patterns across instruments in the dimensions of instruction that they measure, (2) relationships between teachers' scores in specific dimensions of instruction and their contributions to student achievement growth (value-added), and (3) whether teachers' observation ratings depend on the types of students they are assigned to teach.

Researchers analyzed the content of the

• Classroom Assessment Scoring System (CLASS), 
• Framework for Teaching (FFT),
• Protocol for Language Arts Teaching Observations (PLATO), 
• Mathematical Quality of Instruction (MQI), 
• and UTeach Observational Protocol (UTOP). 

The content analysis then informed correlation analyses using data from the Gates Foundation's Measures of Effective Teaching (MET) project. Participants were 5,409 4th-9th grade math and English language arts (ELA) teachers from six school districts. Observation ratings were correlated with teachers' value-added scores and with three composition measures: proportions of nonwhite students, low-income students, and low achieving students in the classroom.

Results show that eight of ten dimensions of instruction are captured in all five instruments, but instruments differ in the number and types of elements they assess within each dimension. Observation ratings in all dimensions with quantitative data were significantly but modestly correlated with teachers' value-added scores—with classroom management showing the strongest and most consistent correlations. Finally, among teachers who were randomly assigned to groups of students, observation ratings for some instruments were associated with the proportion of nonwhite and lower achieving students in the classroom, more often in ELA classes than in math classes.

Findings reflect conceptual consistency across the five instruments, but also differences in the coverage and the specific practices they assess within a given dimension. They also suggest that observation scores for classroom management more strongly and consistently predict teacher contributions to student achievement growth than scores in other dimensions. Finally, the results indicate that the types of students assigned to a teacher can affect observation ratings, particularly in ELA classrooms.

When selecting among instruments, states and districts should consider which provide the best coverage of priority dimensions, how much weight to attach to various observation scores in their evaluation of teacher effectiveness, and how they might target resources toward particular classrooms to reduce the likelihood of bias in ratings.