Growing to the future?
Believe it or not, I am indeed alive. Turns out that the combination of schoolwork, the newspaper, job hunting and the last month of undergraduate college isn't very conducive to blogging. But, I'm not going anywhere, just perhaps receding in frequency a tad.
The biggest edunews in the past few days is undoubtedly the Federal Department of Education's announcement that they would be letting 10 states do test runs on growth models as a form of accountability. As reported in the Boston Globe:
Really, none of the above. Growth models are a natural evolution from the "status" models epitomized by AYP (though I should point out that AYP has a psuedo-growth model built in via the safe harbor provisions). They are, if applied properly and with appropriate safeguards, in nearly every way superior to status models. Don't get me wrong, they aren't fundamentally different -- most of the negative externalities of high-stakes accountability still exist. But on the quest for the crucial item called accountability, growth models are essentially the next-generation tools.
The key is going to be ensuring that they are applied properly and with appropriate safeguards. At their most basic level, growth models let you see the isolated progress of a student or a class from year to year. This is different than looking at one year's 4th graders vs. the next year's 4th graders in assessing progress of teaching 4th graders -- it's asking the question, how are one year's 4th graders doing in 5th grade, in 6th grade, etc.?
It's not a simple matter of looking at test scores from one year to the next, however; first and foremost, you have to be able to track students as they move through the grades, something only 32 states can do at the moment, and with varying levels of effectiveness. Statistical modeling is also ideally involved to try and tease out the progress -- or lack thereof -- that can actually be attributed to the classroom (though some may argue that the source isn't as important as the result, but I think both are necessary pieces of information).
So I'm pleased that the Feds seem cognizant of these facts and I hope they actually push hard for the logistical prerequisites. A bad growth model is far worse than a medicore status model.
The other, and ultimately more important, concern is what type of growth models will be allowed. There are three main options: Expected-growth, average-growth and growth-to-standard. Expected-growth means that a school/class/student's growth target is equal to what equivalent groups have done in the past. For instance, the Tennessee Value-Added Assessment System (TVAAS), the creme de la creme of existing growth models, currently has it so Tennessee 4th graders have to make the same progress as 4th graders did in a baseline year. Average-growth means that the unit has to make progress equal to the average of its peer group. Growth-to-standard sets a uniform target and a time by which, through whatever amount of growth in each year, everyone must meet the standard.
The growth-to-standard is the only respectable option, even though it's the most difficult to implement and least utilized. Both expected-growth and average-growth take kids who start low and leave them low relative to their higher-performing peers, and neither requires that anyone reaches proficiency. If it's easier, think about a graph of achievement with three students starting at varying levels and a flat proficiency line way above, then think about what expected-growth and average-growth does to the three points. Growth-to-standard, on the other hand, takes all the advantages of growth models -- accuracy, a truer picture, more honed analysis that can be used diagnostically , etc. -- and combines it with the accountability which may be the best part in a largely flawed AYP formula.
So, I'm cautiously encouraged by this move. The Feds need to be vigilant to ensure that growth models are employed appropriately -- with the proper infrastructure and growth-to-standard in place -- but if that happens, they will likely prove far superior to the current system. Now, if we could just salvage those pesky exams...
The biggest edunews in the past few days is undoubtedly the Federal Department of Education's announcement that they would be letting 10 states do test runs on growth models as a form of accountability. As reported in the Boston Globe:
In an experiment that's been months in the making, up to 10 states will be allowed to measure not just how students are performing, but how that performance is changing over time.Having worked on growth models a great deal at the EdTrust over the summer, this news isn't particularly shocking; it was always more of a when than an if. So, what to make of it? The Feds backing down on accountability? Victory for the states? Or just another incarnation of a high-stakes mistake, as the virulently anti-NCLB crowd might have it?Currently, schools are judged based only on how today's students compare to last year's students in math and reading -- such as fourth-graders in 2005 versus fourth-graders in 2004.
Many state leaders don't like the current system of comparing two different years of kids because it doesn't recognize changes in the population or growth by individual students.
[...]The states that win approval for the new flexibility, however, must do more than show growth. They still will have to get all children up to par in reading and math by 2014, as the law requires, and show consistent gains along the way.
The Education Department, eager to show it is not weakening the law, will require states to take many steps before they can qualify for the "growth" option.
States must have data systems to track individual students, close achievement gaps between whites and minorities, and prove they have at least one year of baseline testing.
Really, none of the above. Growth models are a natural evolution from the "status" models epitomized by AYP (though I should point out that AYP has a psuedo-growth model built in via the safe harbor provisions). They are, if applied properly and with appropriate safeguards, in nearly every way superior to status models. Don't get me wrong, they aren't fundamentally different -- most of the negative externalities of high-stakes accountability still exist. But on the quest for the crucial item called accountability, growth models are essentially the next-generation tools.
The key is going to be ensuring that they are applied properly and with appropriate safeguards. At their most basic level, growth models let you see the isolated progress of a student or a class from year to year. This is different than looking at one year's 4th graders vs. the next year's 4th graders in assessing progress of teaching 4th graders -- it's asking the question, how are one year's 4th graders doing in 5th grade, in 6th grade, etc.?
It's not a simple matter of looking at test scores from one year to the next, however; first and foremost, you have to be able to track students as they move through the grades, something only 32 states can do at the moment, and with varying levels of effectiveness. Statistical modeling is also ideally involved to try and tease out the progress -- or lack thereof -- that can actually be attributed to the classroom (though some may argue that the source isn't as important as the result, but I think both are necessary pieces of information).
So I'm pleased that the Feds seem cognizant of these facts and I hope they actually push hard for the logistical prerequisites. A bad growth model is far worse than a medicore status model.
The other, and ultimately more important, concern is what type of growth models will be allowed. There are three main options: Expected-growth, average-growth and growth-to-standard. Expected-growth means that a school/class/student's growth target is equal to what equivalent groups have done in the past. For instance, the Tennessee Value-Added Assessment System (TVAAS), the creme de la creme of existing growth models, currently has it so Tennessee 4th graders have to make the same progress as 4th graders did in a baseline year. Average-growth means that the unit has to make progress equal to the average of its peer group. Growth-to-standard sets a uniform target and a time by which, through whatever amount of growth in each year, everyone must meet the standard.
The growth-to-standard is the only respectable option, even though it's the most difficult to implement and least utilized. Both expected-growth and average-growth take kids who start low and leave them low relative to their higher-performing peers, and neither requires that anyone reaches proficiency. If it's easier, think about a graph of achievement with three students starting at varying levels and a flat proficiency line way above, then think about what expected-growth and average-growth does to the three points. Growth-to-standard, on the other hand, takes all the advantages of growth models -- accuracy, a truer picture, more honed analysis that can be used diagnostically , etc. -- and combines it with the accountability which may be the best part in a largely flawed AYP formula.
So, I'm cautiously encouraged by this move. The Feds need to be vigilant to ensure that growth models are employed appropriately -- with the proper infrastructure and growth-to-standard in place -- but if that happens, they will likely prove far superior to the current system. Now, if we could just salvage those pesky exams...
This entry was posted
on Saturday, November 19, 2005 at 4:15 PM.
|
