In Nevada during the 2011-12 school year, 86 schools were in “restructuring” under the No Child Left Behind Act—the most aggressive sanction under the federal school accountability law.

But after the state got an NCLB waiver, by the 2012-13 school year, 75 of those schools got relief from the toughest interventions. These are schools that hadn’t made adequate yearly progress for six years in a row.

For half of the worst NCLB-era schools in 15 states, waivers proved to be an escape hatch, according to a new paper released today from New America Foundation policy analyst Anne Hyslop, who has delivered some of the most comprehensive research yet on the implications of new NCLB waivers. In the 15 waiver states she surveyed, more than half of schools that were in NCLB restructuring during the 2011-12 school year were not on their state’s list of the bottom 15 percent “priority” or “focus” schools under waivers.

The bottom line of Hyslop’s paper: Far fewer schools are being identified for interventions under waivers, and whether the right schools are being identified is an open and critical question. (See one of her more intriguing charts below.)

While many states did identify similar schools for interventions under waivers as were targeted under the NCLB law, five states in particular took a divergent path: Arizona, Massachusetts, Nevada, Rhode Island, and South Carolina. In these states, at least half of the schools relieved from interventions under waivers were previously in corrective action (the next-to-worst NCLB sanction) or restructuring.

Hyslop’s paper raises questions about the very structure of the waivers, which require states to identify and intervene in the 15 percent of the lowest-performing schools in the states. That was done, in part, to allow states to more carefully target their interventions and alter just which schools got the extra help.

Under NCLB, a limitless number of schools could be identified for interventions that got more punitive as schools continued to miss annual targets.

Under waivers, however, the proportion of schools requiring interventions has essentially been capped—for better or for worse—at 15 percent. This means that the relative ranking of schools compared to others in the same state is what matters—not absolute performance of students.

Hyslop’s research points out that how a state designed its accountability system greatly affected which schools are, and are not, targeted. For example, she found that states using large “super-subgroups” of at-risk students under waivers—rather than small, NCLB-era subgroups broken down by demographics—identified different schools for interventions than did NCLB. This will be of concern to civil rights groups, especially, which have worried that super-subgroups under waivers would mask the low performance of small, individual groups of at-risk students.

What’s most evident is that, under waivers, there are far more questions than answers. And Hyslop’s paper correctly points out the need for much more research.

To that end, the Education Department is working its way through its own analysis of how states are doing in identifying and intervening in schools. It seems doubtful this data will be used to inform waiver renewal decisions, given the scaled-back and revised renewal guidance that recently came out. Still, Education Department officials say they will use this data, which is already in draft form for most of the waiver states, to work with states to help them improve their accountability systems.

Importantly, the department still isn’t sure if it will make its analysis public.

Hyslop told me in an email: “I fear that we’ve entered the Dark Ages of school accountability, and that we’ll wake up in 2016 and have no idea what we’ve learned from this experiment in flexibility. No one has any idea whether relative accountability systems and more targeted school improvement lead to better student outcomes, or whether there are negative effects for schools that just miss the cut and are no longer identified for improvement. And the gap between what we know about waiver implementation, and what we need to know about waiver implementation, isn’t closing fast enough.”