Milwaukee's voucher program prompted sustainable achievement gains for the city's public elementary schools, according to a new study by a Harvard economist.

Researcher Caroline Hoxby followed up on a study of three years ago, in which she concluded that the private school choice program pushed the public schools to improve.

In the new study, she adds test score data from two additional years - the 2000-'01 and 2001-'02 school years - and finds that the gains were sustained, although they did not accelerate. The study was published in the Swedish Economic Policy Review.

"Adding the new years of data allows us to see that the good results have lasted," Hoxby said. "A lot of people thought that this was a blip that was going to go away."

But Martin Carnoy, a professor of education and economics at Stanford University, argued that Hoxby's findings show just that, a blip of one or two years when test scores in some MPS schools improved more sharply than those in a control group of Wisconsin schools outside Milwaukee.

Carnoy added that if Hoxby removed the 1997-'98 school year, with its good test scores, from her analysis, "she would be cooked."

"If one believes her conclusions already, this will fuel their belief," Carnoy said. "But for someone who is a skeptic, this isn't going to change their mind."

Some Milwaukee school administrators have said that the growth of the voucher program - particularly after the 1998 court decision that allowed private schools with a religious affiliation to participate - has prompted them to compete more aggressively for the city's students.

Advocates for vouchers tout this as a key selling point, arguing that vouchers can improve the educational experience of the thousands of children who choose to stay with Milwaukee Public Schools.

Hoxby's argument is that vouchers can spur improvements in public schools by threatening to steal away students and the money that comes with them.

In this study and the earlier one, Hoxby did a complicated review of what she describes as "productivity" at several dozen Wisconsin schools. She defined productivity as the ratio of standardized test scores in various subjects to per-student spending.

Hoxby calculated the productivity from 1996 through 2002 of a set of MPS schools that she predicts would be most affected by choice because they have a large number of students eligible for vouchers; of a set that would be less affected because fewer students are eligible for vouchers; and of a set of Wisconsin schools outside Milwaukee that are in urban areas and have relatively high numbers of poor and black students. The last group is the control set because students outside Milwaukee cannot receive vouchers.

She found that test scores at the most affected schools shot up more than those in the two other groups between 1997 and 1999, when MPS would feel the most competitive pressure because of the state Supreme Court decision allowing the expansion of vouchers.

In all three groups, test scores were relatively level between 1999 and 2002 in most subjects.

"The gains are really remarkable by any American standards," Hoxby said. "Vouchers are not supposed to be a miracle silver bullet but should stimulate schools to make changes that they've needed to make for a long time."

But Emily Van Dunk, research director at the Public Policy Forum, a local research organization, said it is "very unclear when the test score increases she is talking about really happened."

Van Dunk said her own research has not uncovered comparable test score increases. She recently co-wrote a book that concluded that student performance is lower in public schools located near voucher schools, and that MPS schools are insulated from financial harm when they lose students to private ones in the choice program.

"In many ways, her research is creative and sophisticated, but it doesn't reconcile with the real world," she added.

Carnoy argued that not enough is known about the control group outside Milwaukee to ensure that it makes for a good comparison.

"Her position is pretty strong based on her data," he said.

But Hoxby said: "It looks like the achievement is significantly higher and shows no signs of decreasing at all."