Thank you for sending your comment. I agree that it is not yet clear how to measure teacher’s effectiveness, since we do not yet have a definition of “effectiveness” that covers the whole spectrum of what makes teachers effective. I think Dr. Pianta’s research is very valuable to get us closer to this point.

The problem I have with value added analysis is that it uses percentile ranks, and changes in percentile rank to determine teacher effectiveness (based on standardized test scores only), and to compare teachers to one another. A percentile rank is a measure of how someone performed relative to everybody else; on a standardized test, you would look at how one student performed relative to all other students in their grade level in the entire school district. So if someone’s score is at the 64th percentile, this means that their score is better than the scores of 64 percent of students at their grade level.

Percentile ranks are not absolute scores, because the percentile rank that is tied to a particular score depends on the scores of everybody else. So if Johnny scores a 70 on his spelling test, he might be in the 50th percentile in school district A, and at the 85th percentile in school district B. Another feature of percentile ranks is that they cluster in the middle and spread out at the extremes (think of the famous “bell curve”). Most students score in the average range, so if Johnny’s score of 70 is at the 50th percentile in district A, and at the end of the year he scores an 80, his percentile rank may have gone up from 50 to 64. However, if this same score of 70 was at the 85th percentile in district B, and his score went up to 80 at the end of the year, his percentile rank may have only gone up to 86. The same absolute gain in points (10 points) leads to a 14 point gain in percentile rank in school district A, but only 1 point gain in percentile rank in school district B.

If you have teachers with students whose scores are mostly in the average range on standardized tests, you will likely see much more fluctuation in percentile ranks. On the other hand, when a teacher has a class of students who score really high, you will probably not see as much fluctuation in percentile ranks, simply because there are generally fewer students who score really high, because most students score around average. Therefore, if you have a group of really high scoring students, it will be much more difficult to actually make gains in percentile rank. In fact, it is much more likely to see drops in percentile rank at the high scores because of a statistical phenomena called “regression to the mean”. One could then erroneously conclude that teachers who happen to be teaching high achieving students are not very effective, because you wouldn’t see as many gains in percentile points.

This is why I have a problem using percentile ranks to base very important and far reaching decisions on about who is effective or ineffective, and who should keep or lose their job or receive a bonus. Because most people don’t understand how percentile ranks work, many people erroneously regard percentile ranks as absolute scores.

I hope this helps to clarify why I don’t think value added analysis is a fair approach.

Thank you again for your comment, and I hope you will keep reading my blog!

Iektje

Thank you for the link. That looks really interesting, I will be sure to look into that.

Iektje