Schools Move Toward Following Students’ Yearly Progress on Tests

From July 6th's NYTIMES:

The Cohoes city school district, outside Albany, is considering a gifted program for elementary students and adding college-level courses after discovering that its top students improved less on standardized tests in the past two years than everyone else in the district.

In Ardsley, N.Y., a Westchester County suburb, administrators intend to place more special education students in regular classes after seeing their standardized test scores rise in the last year.

And as the New York City Department of Education begins grading each public school A to F for the first time this fall, more than half the evaluation will be based on how individual students progress on standardized tests.

All three changes resulted from an increasingly popular way of analyzing test scores, called a “growth model” because it tracks the progress of students as they move from grade to grade rather than comparing, say, this year’s fourth graders with last year’s, the traditional approach.

Concerned that the traditional way amounted to an apples-to-oranges comparison, schools in more than two dozen states have turned to growth models. Now a movement is mounting to amend the federal No Child Left Behind Act, which is up for reauthorization this year, to allow such alternative assessments of student progress.

Many urban educators contend that growth models are a fairer measure because they recognize that poor and minority students often start out behind, and thus have more to learn to reach state standards. At the same time, many school officials in affluent suburbs favor growth models because they evaluate students at all levels rather than focusing on lifting those at the bottom, thereby helping to justify instruction costs to parents and school boards at a time of shrinking budgets.

Adding growth models as a way to satisfy federal requirements to demonstrate “adequate yearly progress” could make it easier for some schools to avoid penalties because they would receive credit for students who improve performance but still fall below proficiency levels. It could also increase pressure on high-performing schools that sail above state standards to prove that their students are continuing to advance.

Federal education officials agreed in 2005 to a pilot program allowing up to 10 states to experiment with growth models, but emphasized that they remained responsible for ensuring that all students would reach reading and math standards by 2014, and show consistent gains along the way. Seven states — North Carolina, Tennessee, Arkansas, Delaware, Ohio, Florida and Iowa — have joined the pilot so far, federal officials said, and on Tuesday, the Education Department green-lighted Alaska and Arizona to use growth models to analyze data from the 2006-7 school year.

“A growth model is a way for states that are already raising achievement and following the bright-line principles of the law to strengthen accountability,” Margaret Spellings, the secretary of education, said in a statement. “We are open to new ideas, but when it comes to accountability, we are not taking our eye off the ball.”

In New York, education officials are developing a statewide growth model that will be in place by the 2008-9 school year, to be used as an additional way to measure student learning. Fifteen New York school districts, mainly in the Albany and Catskill regions, have experimented with growth models on their own through a voluntary program started by two regional support educational agencies in 2005. The districts typically pay these agencies from $1,000 to $6,000 to train administrators and staff, and an additional $2.50 a year for each student for the data analysis, which is partly reimbursed through state aid.

“There is absolutely a need for this kind of data,” said Timothy G. Kremer, executive director of the New York State School Boards Association, which represents about 700 districts. “It keeps the focus on student achievement, and not on whether you’re going to pave the parking lot or who’s going to get hired as next year’s coach.”

But as growth models become more widespread, some teachers and parents have complained that they are hard to understand and place too much focus on test scores. Teachers’ unions, even while supporting the concept, have protested the use of growth models for performance reviews and merit pay.

“It’s detrimental for education,” said Aimee Bolender, president of the Alliance-AFT, which represents 9,000 teachers and other staff members in the Dallas schools. “It is pulling apart teams of teachers and it doesn’t look at why test scores are low. From the very beginning, we viewed it as a slippery slope that did not do anything valuable to improve the educational environment in the schools.”

Ms. Bolender’s union is fighting a decision by the Dallas school district to remove about 30 teachers from five middle and high schools this summer after not enough of their students passed the state tests, and too many failed to show adequate progress on growth models. Ms. Bolender said that many teachers question the reliability of the growth model data, calling it “voodoo math” because “you have to be a Ph.D. in statistics to even comprehend it.”

Even some supporters of growth models have expressed concerns that they could shift attention and resources away from the neediest students. Kati Haycock, director of the Education Trust, an advocacy group for disadvantaged children, said she was worried about growth models’ focusing too much attention on students at the top. “It risks so broadening the federal government’s involvement that its historical role will be dissipated,” she said.

While growth models have existed for at least two decades, they were not widely used by school systems until recently because few states had the extensive testing data required for the analysis. But under the No Child Left Behind Act, which requires annual testing for students in Grades 3 through 8, states have developed larger databases ripe for the growth-model approach, which many experts see as a more thorough picture.

“When you look at achievement, every single wealthy suburb has high test scores,” noted Theodore Hershberg, a professor of public policy and history at the University of Pennsylvania. “That’s a terrible way to measure the performance of a school or an individual teacher because what you’re really looking at is family background or family income.”

In the high-performing Ardsley schools, where more than 87 percent of the students passed state reading tests this spring, district officials have long mined scores on their own, compiling a thick data book for review and coining the saying: “In God we trust, everybody else bring data.”

But this year, they employed a more sophisticated growth model, which showed, for instance, that seventh-grade special education students had benefited from learning in regular classes. So this fall the district will expand the mainstreaming to the elementary and high schools. “This gives us the ability to measure whether a program has any teeth or is all fluff,” said Richard Maurer, the superintendent.

Cohoes school officials have spent more than $1 million on programs for their most struggling students in the past five years, and wanted to find out how much they had progressed. They learned that the lowest-level students were doing fine, while their high achievers were starting to fall behind.

Charles S. Dedrick, superintendent of the 2,200-student district, said that parents had complained that their children were scoring too low on the Advanced Placement exams to receive college credit, but he thought there was just a problem with the A.P. coursework. Now, after examining over time the state test scores of students in advanced classes, he sees a more systemic problem. So the district has made top-level students a priority, too, and is considering starting a gifted program, expanding A.P. and college-level courses, and adding an International Baccalaureate program to keep them challenged.

“The fact is we serve all students, and not just the lower-end students,” said Mr. Dedrick, who travels across the state to speak about growth models to school superintendents. “If you’re just concentrating on one group of kids, it’s not fair because both sets of parents pay taxes.”