“Most researchers have no interest in who anybody is. If I ever released confidential private information publicly, I’d be in deep trouble.” “Schools’ first responsibility is to protect the information on students. ... We’re not going to do a new regulation just on this.”
“More and more people are starting to look into questions of the federal privacy-protection law and how it applies” to those newer databases, said Eric A. Hanushek, a Stanford University scholar who has made extensive use of longitudinal testing data in Texas and other states. “And some of these interpretations have the possibility of shutting down some of the best research that’s been done over the last decade.”
The conflicts are arising, in part, out of two movements that have swept the U.S. education landscape in recent years: the push to hold schools accountable for students’ educational progress, and the emphasis on schools’ use of only programs and practices deemed to be backed by scientifically based research.
Those efforts, both of which are embedded in the 4-year-old federal No Child Left Behind Act, are raising the demand for more and better data on student progress.
Most states have responded by creating data-collection systems that assign unique “student identifiers” so youngsters can be tracked anonymously over the years as they move from school to school, or district to district.
According to a count last year by the Denver-based Education Commission of the States, at least 31 states now have databases that can link test scores, the length of time students have been enrolled in given schools, and graduation records over time.
But the longer that researchers follow students, the greater the possibility that the students’ identities will be revealed.
“With kids moving from school to school, if you start talking specifically about, let’s say, a low-income African-American female in special education, well, she could end up being one of two students in the school that fit that kind of a description,” said Chrys J. Dougherty, the research director for the National Center on Educational Accountability, a nonprofit research group based in Austin, Texas.
1974 Law
To prevent such identification from occurring, officials in some states are tightening the lids on their data sets by invoking the Family Educational Rights and Privacy Act, or FERPA.
The Family Educational Rights and Privacy Act of 1974, or FERPA, is a federal law designed to protect the privacy of students’ education records. It applies to all educational institutions that receive federal funding. While the law allows for some exceptions, it generally prohibits schools and state and local education agencies from releasing student records, or information from those records that might be “personally identifiable,” without parents’ written consent
SOURCE: U.S. Department of Education
The 1974 federal law generally prohibits schools and education agencies from releasing information from students’ educational records without their parents’ written consent. The law makes an exception for studies conducted by the schools or districts themselves or by researchers working under contract to them, but that exception is generally thought to exclude researchers like Mr. Dougherty, who often work independently or with private funding.
Moreover, researchers say, states are not likely to want to underwrite or endorse studies that criticize their practices.
“This is a law designed for a different set of circumstances and a different time,” argued Mr. Hanushek, who is a senior fellow at the Hoover Institution, a think tank based at Stanford.
According to Mr. Hanushek, he is now spending half his time working to resolve conflicts related to the federal law. He contended that part of the problem may be that some states are “hiding behind” the law to better control the information they release.
“The reality is that most researchers have absolutely no interest in who anybody is,” Mr. Hanushek said, referring to the students whose data are used in studies. “If I ever released confidential private information publicly, I’d be in deep trouble as a researcher.”
Researchers and data managers also note that they have a repertoire of techniques for masking identities when “cell sizes” in studies become small enough that students could be identified.
For their part, federal officials point out that the law is intended to guide schools, not to help researchers. States that fail to follow the law ultimately stand to lose federal funding if they fail to protect students’ privacy.
“Schools’ first responsibility is to protect the information on students,” said LeRoy S. Rooker, the director of the U.S. Department of Education’s family-compliance office, which handles FERPA issues.
“That’s not necessarily a rare instance,” he said, “when you have a particular student in a particular subgroup who’s failing or all the students in the subgroup are failing, and you can identify the school, and you can identify the class, and you can identify the subgroup.”
Thomas Hutton, a staff lawyer for the National School Boards Association, based in Alexandria, Va., said that student privacy could become an even bigger issue if states move to using “growth models” to measure student progress under the No Child Left Behind law. Such models involve tracking the performance of the same students over time, rather than comparing, for instance, one year’s 4th graders against the previous year’s 4th graders.
“This is one of those instances where school districts are trying to figure out, ‘How do we comply with these competing demands?’ ” Mr. Hutton said.
Clamping Down
To guide officials struggling to interpret the privacy law, Mr. Rooker’s office points to a letter it issued in late 2004 to Tennessee school officials. In it, federal officials said that states are free to release data on students not considered “personally identifiable”—data, for instance, that may already have been “scrubbed” of names, birth dates, or Social Security numbers.
But it also reminded Tennessee officials that information contained in data cells small enough to make students’ identities traceable is considered personally identifiable under the law.
“It does clarify the issue, but it does it in a way that makes it very difficult for researchers to ever have longitudinal data,” Mr. Dougherty said of the letter to Tennessee.
The letter said that the department intended to write future regulations to define “non-personally-identifiable information, thus allowing disclosure, without parental consent, but with proper privacy safeguards.” But department officials said last week that no plans were yet in the works to change the existing regulations.
“We’re not going to do a new regulation just on this,” Mr. Rooker said. “It’s just something we’d like to see in the next round of regulations that are promulgated.”
In the meantime, state and local officials vary in how they interpret the law. Researchers at Duke University and in the University of North Carolina system, working under an agreement with the North Carolina education department, can obtain scrubbed student data from a third-party group, the North Carolina Education Research Data Center.
“If the study will improve the lives of students and teachers, that research is allowed,” said Elizabeth J. Glennie, the director of the center, located on Duke’s campus in Durham, N.C.
In Ohio, scholars and state school officials have a similar arrangement with an outside data-management group. State officials there say they also set a minimum cell size—in other words, the number of people associated with a given category of information—at 10 subjects. A given cell in an education study, for instance, might contain all the male 8th graders who failed the math portion of a state exam.
Experts say a number of other states also have rules on cell size, but the numbers vary widely, from no less than five subjects allowed per cell to hundreds.
Colorado, citing FERPA restrictions, has created a “data store,” through which researchers can obtain only prepackaged analyses of student data, according to Margaret E. Raymond, a researcher who has been trying to use that state’s data.
“My big itch with them is they’ve pretty much commandeered the ability to do independent research,” said Ms. Raymond, who directs the Hoover Institution’s Center for Research on Education Outcomes.
Colorado education officials said Ms. Raymond’s characterization was only half-right. While the system contains preset analyses to answer 80 percent to 90 percent of the information requests the department gets, they said, it also contains data “cubes.” Such a cube is essentially a multidimensional matrix on a given topic that allows for more-independent analyses.
But the state, citing the federal law, does not give schools or researchers access to students’ educational histories in previous schools or districts.
“We’re doing all we can with the FERPA restrictions in place,” said Daniel Domagala, the supervisor of Colorado’s data warehouse.
Mr. Hanushek, meanwhile, said he has run into conflicts with Texas school officials over a study in which he hopes to use student-level scores from state exams to track students’ learning growth as they move in and out of charter schools over the course of their school careers.
“I think it’s a simple matter that, if we want our schools to improve, we have to make it possible to do some of the fundamental research on what affects achievement,” he said. “If we stop that, we’re stuck.”
