Hiring new teachers is expensive--especially when high rates of teacher turnover create a revolving door in classrooms. A new study by Columbia University economist Jonah Rockoff looks at effects of New York City's now defunct $40 million mentor program (designed with UC Santa Cruz?s New Teacher Center) that hoped in part to curb some of the city?s startling turnover rates for new teachers.

Rockoff finds that mentors who share similar educational backgrounds and subject matter experience as their mentees don?t seem to have any impact, good or bad, on teacher retention or student performance--despite the fact that this type of matching is often stressed by state law and supporters of mentoring programs.

So what characteristics do seem to matter? Strong evidence shows that having a mentor who previously worked in the same school has an important impact on whether a teacher returns to that school the following year.

Rockoff explains, "If I?m a new teacher in P.S. 101 and have a mentor who's been teaching there for 15 years, that could be really helpful because that mentor can help me figure out how to deal with Vice Principal Jones, who's really tough, or where to get the cheapest school supplies or even what bus to take to get there on time. A mentor who's been teaching 20 years in another part of town may not understand the students or administrators at that school."