Carole L. Perlman. Using test scores to evaluate teachers creates problems.

LETTER TO THE EDITOR

May 27, 2014
I was director of Chicago Public Schools testing programs for nearly 20 years. I’m all for accountability; I got a great education at CPS and feel strongly that the current students deserve the same. Using test scores to evaluate teachers seems like a great idea on the face of it, but it does not stand up to scrutiny. There are several basic problems, most having to do with the fact that relatively few teachers will have any test data at all and that there aren’t enough students in most classes to yield reliable growth estimates. This has been the case in other school systems such as New York City, where a teacher may have an outstanding rating one year and be in danger of being fired the next — for teaching exactly the same way.

State assessments are generally given in grades 3-8 and two years’ worth of scores are necessary to generate the growth data that’s used for teacher evaluations. That means that teachers of grades K-3 won’t have the requisite pretest and post-test scores — the small kids because they don’t take the state assessments and the third graders because they didn’t take the test the year before. That also leaves out teachers of non-tested subjects (foreign languages, physical education, social studies, the arts), as well as counselors and librarians. At the high school level, the problem is even worse, with only one or two grades being tested in reading and math, and lots of teachers teaching non-tested subjects.

The second and more serious problem is that even when growth estimates can be calculated, they’re wildly imprecise, as they aren’t based on enough students to yield a reliable measure (think of an opinion poll that’s based on only 15-20 people). A typical class will have 25-28 students, many of whom transferred into the school mid-year (CPS kids are a mobile lot). Some will have pretest scores, but many won’t. Others coming in from other states or countries and those who newly left bilingual programs won’t have taken the test the previous year, so if you’re lucky, maybe there exists matching pretest and post-test data for 20. This is a recipe for drawing very wrong conclusions about teacher effectiveness.

Though superficially appealing, using test scores to evaluate teachers will create more problems than it will solve. Excellent teachers will be erroneously labeled as incompetent, while poor teachers may get a pass. Students will not benefit.

— Carole L. Perlman, Chicago