Indicators of teacher and school performance

"Value-Added Measures and the Future of Educational Accountability" is an interesting article by Douglas Harris in the Science magazine of August 12, 2011. The author discusses:

• "point-in-time snapshots of student performance" such as the performance on tests at the end of certain school grades, and
• value added estimates such as those which assess the quality of teachers or schools, and especially those which attribute changes in student performance over a period of time such as a school year to the impact of their teachers or schools.

The data show what I think is quite intuitive, that different measures are somewhat uncorrelated, and that point-in-time estimates confound school/teacher influence with pre-school/home environment influence.

Harris suggests that it one might combine various estimates of teacher quality, such as improvements in test results, measures of qualifications, and evaluations by peers and supervisors to improve the validity of the value added estimate (especially if used in payroll decisions). This also seems reasonable to me.

Bayes would be pleased by the realization, if not by the means of making the changes. One might assume that there is an a priori distribution of teacher quality and that each added piece of information on a particular teacher can be used to find an a posteriori probability distribution of the "true" quality of that teacher. This approach might be used not only with various sources of information in a single year, but adding information from each year that the teacher is in the system.

Still, I am glad that there is an effort to find better ways to evaluate the help teachers and schools are providing their students to learn, and to provide incentives to good teachers and schools as well as protecting students from second rate schools and teachers. Of course it will be hard to do this well, but with work we will get better at it.