Learning Without Loopholes

Congressional leaders have finally acknowledged what most observers have known for months: The No Child Left Behind Act, the signature domestic legislation of the Bush administration, will not be reauthorized in the foreseeable future. And it should not be, at least not until the law’s blatant loopholes are addressed. The truth that no one in power seems willing to admit is that the federal law encourages statistical manipulations that make reports of academic progress suspect and, in some cases, virtually meaningless.

The well-intentioned 2001 legislation declares that all identifiable “subgroups” of students (those from low-income families, members of minority groups, students with disabilities, and others) must make what it calls “adequate yearly progress.” To put pressure on educators, the law says that if just one subgroup fails to make AYP, the entire school is deemed “in need of improvement.”

Ah, but the loopholes! The law lets states decide what constitutes academic proficiency and how large a subgroup needs to be before its scores count. It also allows states to use a statistical trick that can add 50 or more points to a test score.

Because failure under No Child Left Behind has a high price tag in penalties (including loss of jobs and school closings), states would be crazy not to take advantage of whatever loopholes are available to them.

To understand the statistical trick, think of political polls. These report that Candidate X is ahead of Candidate Y, 40 percent to 33 percent, with a margin of error of 4 points. That “plus or minus” band around a score, called a “confidence interval,” makes sense, because political polls are based on a small sample of probable voters, not every registered voter.

Polls generally have a 95 percent confidence interval, meaning that 95 out of 100 times the results would not vary by more than the stated margin. To be 99 percent confident, however, requires a greater margin of error, perhaps 12 points or more, and that would mean that, in the example above, Candidates X and Y are tied.

But schools test every student, not a small sample, and so one might assume that the scores kids get are the scores they get. Not so, because schools apply a confidence interval. With chutzpah reminiscent of Milo Minderbinder in Catch-22, educators rationalize that some students might have done better under different conditions (if they’d eaten a decent breakfast, or hadn’t fought with their parents just before the test, for example). With the blessing of the U.S. Department of Education, states are allowed to calculate these hypothetical scores. And when a subgroup’s actual score falls below the target but is within the confidence interval of hypothetical scores, that subgroup is deemed to have met the NCLB requirements.

What a difference a confidence interval can make! Some 509 schools in Illinois—13.5 percent of those tested—failed to make AYP, but passed anyway because the state employed a 95 percent confidence interval, adding 12 “hypothetical” points.

Thanks to a confidence interval, in only 10 of 161 Kentucky elementary schools did African-American kids fail to make AYP. Without that crutch, however, 83 subgroups would have failed, which would have labeled those 83 schools failures.

Without the escape hatch provided by the confidence interval, the adults in charge would have been publicly embarrassed, but the students would have received additional resources. So, just whose interests were being served?

Parents are kept in the dark. When NCLB results are released, no asterisk tells them their children passed hypothetically but not in reality.

Thirty-six states employ confidence intervals. With the blessing of the federal Education Department, most use the 99 percent level, giving them the largest possible margin of error.

Here’s another glaring loophole: The No Child Left Behind law allows states to decide how large a subgroup will be, and the variation is staggering, ranging from a low of five in Maryland, to 50 in Texas, Virginia, and West Virginia. If a particular elementary school has 35 Hispanic students, but the state subgroup size has been set at 40, that school—as if by magic—does not have a Hispanic subgroup, and how those students perform on the state test doesn’t really matter.

High subgroup numbers let schools off the hook. By definition, it holds fewer schools accountable for educating the most vulnerable students—minority students, low-income students, students with disabilities, and students who have limited proficiency in English. Big achievement gaps are less likely to show up; schools are less likely to tell the public about them or to be pressured to do something to close them.

In effect, the most vulnerable kids—the ones the law was supposed to be helping—disappear (although their scores are mixed in with the results for the entire school). The Associated Press has reported that an estimated 1.9 million minority students have “disappeared.”

Combine a large subgroup size with a confidence interval, as Wisconsin does (40 and 99 percent, respectively), and the results look impressive. In that state, only Milwaukee and Kenosha, of its 440 districts, have not made adequate yearly progress every year.

The law also lets states decide what constitutes passing. So, for example, when Illinois lowered the passing score on its 8th grade math test, the pass rate jumped from 54 percent to 78 percent.

At least seven states lowered the bar in midcourse. Others just started out with a low bar.

Policymakers had hoped that the National Assessment of Educational Progress would shame states into setting a high bar, but that hasn’t worked. On North Carolina’s own 4th grade math test for NCLB, its students had the highest passing rate in the county—92 percent. But when North Carolina students took the NAEP math test, just 40 percent passed. Similar gaps in achievement can be found in most states.

Low standards make schools look good but hurt students. To use a high-jump analogy, if the coach consistently sets the bar at 5 feet, a talented athlete will never know if he or she could have cleared 6 feet, or higher. Lowering the bar also reduces the incentives and the pressure on school districts to do a better job.

Asked if she believed states were taking advantage of the loopholes to avoid being held accountable, U.S. Secretary of Education Margaret Spellings said: “Sometimes they are, and sometimes they aren’t. It’s ‘The Naked City’; there’s lots of different answers to that question. But are some kids being left behind? You bet. Are some people gaming the system, and we ought to be watchful about that and hawkish about it? You bet. And we do our very best to do that.”

But, she added, “I choose to believe that the people in states are working hard to improve education for their kids.”

Perhaps they are, but if the No Child Left Behind law is intended to be a signaling system that sets standards, tests students, and then reveals where students are not learning what they need to know, it is sending—at best—confusing signals. At worst, the data are virtually useless.

Congress is between a rock and a hard place. If it closed the loopholes, someone in Washington would have to police all 15,000 school districts. Does anyone want the U.S. secretary of education to be the nation’s school superintendent?

On the other hand, not closing the loopholes breeds cynicism, hypocrisy, and disrespect for law.

Now a political football, NCLB is unlikely to be reauthorized before the next presidential election. The intervening months are an opportunity for states to work together to set common standards. Presidential candidates could be pressed to pledge adequate resources to allow states to develop common standards and tests. While state participation should be voluntary, once 20 or so states sign on, the rest will follow. And that would be a big step toward a healthy, internationally competitive public education system.