For years, education advocates, policymakers and scholars have been trying to put an end to the underprepared novice teacher. The hope has been to find a training model that is just right, pairing theoretical knowledge and practical skills necessary for the messy reality of the classroom.

Now some think they’ve found an approach that works: teacher residencies.

Writing in the New York Times, three staff members of Bank Street College argued for this idea, comparing it to how doctors are trained.

“Aspiring teachers need well-designed and well-supported preparation,” wrote Shael Polakow-Suransky, Josh Thomases, and Karen DeMoss. “Year-long co-teaching residencies, where candidates work alongside an accomplished teacher while studying child development and teaching methods, offer a promising path.”

Indeed, there is consistent research showing that teachers trained through residencies are more likely to stay in the profession, potentially reducing churn in schools and costs of finding and training new teachers.

“When it’s done well, it’s kind of a solution to the teacher shortage problem that has plagued urban districts,” said Linda Darling-Hammond, president of the Learning Policy Institute, an education think tank that recently put out a report praising the residency model.

But preliminary new research focusing on Denver’s residency program showed that teachers trained through the program were less effective at improving student achievement in math than other novice teachers in Denver.

This echoes the findings of a study on the Boston Teacher Residency, a prominent example of the approach. In that case, residency-trained teachers also were less effective in math in their first years in the classroom — though they improved fairly quickly.

Together, the positive impacts on teacher retention and the more tepid effectiveness results might still suggest that residencies are worthwhile. But some see the enthusiasm getting ahead of the evidence, particularly in light of the steep price tag of such models.

“I am amazed by how much enthusiasm this idea seems to be generating, despite the fact that we don’t have much evidence to support it,” said Marty West, a Harvard professor who studied the Boston program.

A teacher residency has several key components, according to proponents.

Darling-Hammond’s group identifies several characteristics: a full year of student teaching under an experienced, effective mentor; a partnership between a school district and university so that practice and theory are closely linked; continuing mentorship after candidates become full-fledged teachers; and payment of student teachers during the residency year in return for a three- to four-year teaching commitment.

The final aspect is part of what makes the program appealing as well as costly.

“As I think about the common elements of residency program, there’s a lot that seems very promising — if also, potentially, very expensive,” West said.

Under a traditional university training model, students pay tuition; under the residency model they get paid, albeit modestly. The Boston Teacher Residency, for instance, is free for those who teach in Boston for three years, and offers candidates a $12,600 stipend as well as health insurance for their residency year. (In that program, teachers do have to pay tuition to UMass Boston to receive a master’s degree as part of the program.)

The upside is that those who go through residencies seem to remain teachers in their school districts for longer. In Denver, for instance, residents were 16 percentage points more likely than other novice teachers to return to the district. A national study of 12 teacher residency programs also showed higher retention rates.

This, Darling-Hammond hypothesizes, is explained by the quality of residence programs.

“I think that amount of student teaching and the mentor teacher being a true expert probably has a lot to do with the retention rate being strong,” she said. “You’re getting everything a beginning teacher should get.”

Although research on what makes teacher training effective has generally not come to clear conclusions, there is evidence for the idea that giving teachers practice in a real classroom is important.

But when it comes to the initial effectiveness of residency-trained teachers — at least as measured by the impact on students’ standardized English and math test scores — the evidence is mixed, and in some cases even negative.

West and colleagues found that teachers who go through the Boston Teacher Residency program were initially less effective at improving student achievement in math and no better in English, compared to other beginning teachers.

To West, these findings were counterintuitive.

“I was excited about the opportunity to evaluate the Boston Teacher Residency because I was optimistic,” he said. “I was surprised by our finding that residents were less effective than other new hires, at least initially.”

Darling-Hammond points out — and West agrees — that the teaching corps is likely to be particularly strong in Boston, where there is a robust higher education sector, so that it might be especially difficult for one program to prove particularly effective.

The Denver study, though, produced similar results: negative impacts on former residents’ students in math and essentially no effect in reading.

But there were bright spots in both evaluations. The teachers in Boston improved swiftly over time to the point that those teaching for five years were more effective than other experienced teachers. Combined with the lower turnover rates, the study estimates that the program had a modestly beneficial effect on student achievement over the long run.

And in Denver, the researchers also examined teachers’ classroom ratings, assigned by trained observers. There, former residents came out ahead of other teachers.

Other research on residency programs is thin but paints a more positive picture. A report on the New Visions Hunter College teacher residency in New York City showed that their teachers outperformed other novices in five high school exam areas, but underperformed in three others. A recent state analysis of 40 teachers trained through the Memphis Teacher Residency found they had above-average impacts on student test scores.

West says he is still optimistic about the residency model. The key question, he says, is whether the costs of the program outweigh the benefits — but no such comprehensive analysis has been done.

Darling-Hammond notes that some programs have tried to save costs by, for instance, using residents as substitutes one day a week or having them take the place of teacher aides. She also emphasizes the impact, financial and otherwise, of residencies on reducing teacher attrition.

“If you think about the costs of replacing teachers … this ends up being a cost-effective investment,” she said.