I have long admired the Teach For America (TFA) program. Since 1990 it has recruited more than 10,000 graduates of some of our most selective colleges to teach in our most disadvantaged urban and rural public schools. The way my 19-year-old daughter and her friends talk about TFA, it seems to embody the qualities that inspired new college graduates of my generation to enlist in the Peace Corps.

So I was impressed and pleased when I read the first reports of a new study by the highly respected Mathematica Policy Research Inc. showing that the students of TFA teachers -- who sign up for two years but sometimes stay longer -- outscored their schoolmates in math and matched their performance in reading. Many critics had suggested that the summer of training given to TFA corps members before they started their jobs was not enough, and the Mathematica report seemed to suggest otherwise.

Let me quote from the Mathematica press release, before I get into a disturbing truth at the core of what sounds like a very upbeat story: "The study reveals that after a single school year, students of TFA teachers outscored a randomized control group of non-TFA teachers' students by three percentile points on the math portion of the Iowa Test of Basic Skills -- the equivalent of one extra month of math achievement. On the reading portion of the test, the average gain in test scores of TFA students was nearly identical to that of control students."

The press release quotes Paul Decker, vice president and director of human services research at Mathematica and principal investigator for the report: "TFA teachers not only had more success than other novice teachers but they had more success than teachers with an average of six years of experience in the classroom."

The press release said, "The study is based on a large national sample, including nearly 2,000 elementary school students in 100 classrooms in 17 schools in six geographical high-poverty areas -- Baltimore, Chicago, Houston, Los Angeles (Compton), New Orleans, and the Mississippi Delta. In each school, students in the study were randomly assigned to either TFA teachers or control group teachers in the same grades. Student performance was measured at the beginning and end of the 2002-2003 academic year."

Including several tables, the study is 64 pages long and available at mathematica-mpr.com. I read it and asked Linda Darling-Hammond, professor of education at Stanford, to give me her thoughts about it. Some of those who know Darling-Hammond's oft-expressed antipathy for Teach For America will consider this the equivalent of asking Michael Moore to review Lynne Cheney's latest book, but Darling-Hammond is a scholar, not a polemicist, and her conclusions were pretty much in line with mine, despite my personal support for TFA and my hope that it keeps doing what it has been doing, only better.

The problem is that although the TFA teachers outperformed the non-TFA teachers in the control group, they did not bring their students anywhere near to where they ought to be. As the press release said, their students' math scores increased by three percentile points. What it didn't say, and the full report makes clear, is that the gain was similar to my batting average in the Old Folks League going from .067 to .073. In the fall the TFA students scored on average at the 14th percentile in math. By spring they had risen to the 17th percentile.

This was better than the control group students, who were at the 15th percentile in the fall and still there in the spring, but I don't think anyone would regard this as an impressive gain. In reading, the numbers are similarly depressing: the TFA students went from the 14th to the 15th percentile and the control group students went from the 13th to the 14th percentile.

Darling-Hammond said, "The study documents the failure of the public policy approach these districts and states have adopted for staffing their high-minority schools -- an approach in which slightly trained TFA teachers look better than other poorly trained hires but none of them actually improve student achievement to any great degree."

It gets worse. Darling-Hammond said she cannot reach a firm conclusion until she sees some technical data not included in the report, but she points out that despite the minimal training and experience of the TFA teachers, they turn out to have more preparation for the classroom than the novice teachers in their schools to which the Mathematica study compared them.

The principals of the inner city and rural schools that hire TFA teachers have long said to critics like Darling-Hammond: What do you want us to do? We need teachers willing to work with our very needy children for very little money. At least TFA gives us bright kids who just graduated from Dartmouth and Rice and UCLA and are strongly motivated to help our children. Take a look at the people we would have to hire if we didn't have TFA before you tell us it is a bad idea.

If the Mathematica study is correct -- and I am reluctant to put too much weight on a study that compared only 41 TFA teachers to only 57 control group teachers (18 of them novices) -- then those principals have a point. Teach For America has been working to improve training for its members, with some prompting from states and districts who have told the organization they won't hire their people otherwise. Many first year TFA teachers are spending their afternoons and evenings studying for their teaching certificates. With a median of two years experience, 51 percent of TFA teachers were certified, compared to only 38 percent of the novice control group teachers. (Decker says this difference is erased once you control for different local rules on certification. Darling-Hammond says that adjustment of the data doesn't make sense to her.)

A higher portion of TFA teachers than either novice or veteran teachers in the control group had some student teaching experience. Forty percent of the TFA teachers had a master's degree by the end of their first year, while nobody in the novice control group did.

"The degree to which these districts and the states they are in have been willing to hire just about anyone with very little training to teach kids who are seriously underachieving is remarkable," said Darling-Hammond. Darling-Hammond and the Southeast Center for Teaching Quality in Chapel Hill, N.C., which also analyzed the report, note the many training programs that have given young teachers the tools to make significant progress in the classroom. But it is hard to make this happen for many schools because it takes more money to train and pay their new teachers and more resolve to shake up the education schools, where professors with tenure teach classes that don't prepare students well for inner-city and rural schools.

And it also requires a change in the way the schools attended by low-income students are led and organized. Here the growth of Teach For America has had an unanticipated but wonderfully beneficial effect.

There are scores of small schools throughout the country -- my favorites being the 31 KIPP charter and contract public middle schools -- that have been created by TFA teachers who saw how little they were doing for their students and decided to change that. They established small schools that hired TFA teachers who had gone beyond the first year and figured out how to raise achievement more than three percentile points. They focused on student behavior in ways most low-income schools never do. They spent much time convincing parents of the need to make sure homework was done each night. And their results, year by year, make the little improvements recorded by the Mathematica study look pathetic.

Mathematica did what it was asked to do. It showed that new TFA teachers are no worse than other novice teachers, and maybe a little better. But that is not nearly good enough.

I suspect that Mathematica, Teach For America and everyone else involved in this research project agree that we are not doing what we should to make sure poverty-stricken kids have good teachers. Each year we dither -- giving schools "needs improvement" labels but not much else -- means another group of kids will never get where we want them to go.