On growth models
There's been a lot of skepticism recently about growth models: Russo here and here, WaPo editorial board here. The general qualm is that states are going to use growth models as a cop-out to actually improving student performance, and it's a legitimate concern. But a well-designed growth model can be more than just a quick fix; it can represent the next iteration of assessment systems. The key (and, Mr. Russo, I worked on this issue at EdTrust; I think they are very aware of what I'm about to say) is well-designed. What does that mean? In my opinion, it requires, at the bare minimum, these elements:
1) Growth-to-standard. This is absolutely critical. Growth targets can be set in one of three ways: Average growth (every student expected to improve the same amount which is taken from a base year growth), Expected growth (students expected to improve different amounts depending on how much they improved the previous year) and Growth-to-standard (all students expected to improve enough that they all meet a standard by a certain time). The first two open a Pandora's Box in which low-achieving students are never forced to catch up to their peers. North Carolina just recently changed their growth model formula because it was designed such that low-achieving students were expected to make less growth than higher-achieving ones. Growth-to-standard maintains the same ultimate goal of equity.
2) Vertically aligned curricula. It makes no sense to assess a students growth from 3rd to 4th grade if the 3rd and 4th grade cirricula are wildly disparate, because such an assessment would border on meaningless. There has to be alignment (which, not for nothing, seems educationally sound for a whole bunch of other reasons).
3) Advanced data systems. 18 states currently don't even have unique student identification numbers, and many of those that do have data systems that are primitive at best. Tennessee is the best example of a state-of-the-art data system which can collect enough information for keeping track of students and performing the statistical work needed to make an authentic growth model viable.
At their best, growth models can identify how much each student is progressing in their educational endeavors and identify weaknesses to be addressed. They can allow accountability to be tied not to a binary state, but to a dynamic, richer metric. There are plenty of conceptual concerns, to be sure -- some scholars are dubious about whether an individual teacher's effect can ever be parsed out of student test scores because of the myriad variables. And there's also the whole issue of assessment still being defined solely in terms of often low-level standardized tests. But I think that everyone is a bit hasty to dismiss growth models as nothing more than an attempt by the states and the federal government to abdicate their responsibilities.
Growth models are at the very least promising. Isn't that worth a carefully monitored pilot program?
1) Growth-to-standard. This is absolutely critical. Growth targets can be set in one of three ways: Average growth (every student expected to improve the same amount which is taken from a base year growth), Expected growth (students expected to improve different amounts depending on how much they improved the previous year) and Growth-to-standard (all students expected to improve enough that they all meet a standard by a certain time). The first two open a Pandora's Box in which low-achieving students are never forced to catch up to their peers. North Carolina just recently changed their growth model formula because it was designed such that low-achieving students were expected to make less growth than higher-achieving ones. Growth-to-standard maintains the same ultimate goal of equity.
2) Vertically aligned curricula. It makes no sense to assess a students growth from 3rd to 4th grade if the 3rd and 4th grade cirricula are wildly disparate, because such an assessment would border on meaningless. There has to be alignment (which, not for nothing, seems educationally sound for a whole bunch of other reasons).
3) Advanced data systems. 18 states currently don't even have unique student identification numbers, and many of those that do have data systems that are primitive at best. Tennessee is the best example of a state-of-the-art data system which can collect enough information for keeping track of students and performing the statistical work needed to make an authentic growth model viable.
At their best, growth models can identify how much each student is progressing in their educational endeavors and identify weaknesses to be addressed. They can allow accountability to be tied not to a binary state, but to a dynamic, richer metric. There are plenty of conceptual concerns, to be sure -- some scholars are dubious about whether an individual teacher's effect can ever be parsed out of student test scores because of the myriad variables. And there's also the whole issue of assessment still being defined solely in terms of often low-level standardized tests. But I think that everyone is a bit hasty to dismiss growth models as nothing more than an attempt by the states and the federal government to abdicate their responsibilities.
Growth models are at the very least promising. Isn't that worth a carefully monitored pilot program?
This entry was posted
on Monday, February 27, 2006 at 8:22 AM.
|
