In our second paper this month from Dan Goldhaber and Michael Hansen, our star-studded team explores the potential for using valued added models in making personnel decisions such as tenure.

The policy question before districts--as they begin to get increasingly better data--concerns the reliability of using value-added data to judge teacher performance. More specifically, can districts fairly assess a teacher's early-career performance as a reliable measure of how they'll perform in future years?

Drawing upon the same group of teachers as for the teacher licensing and race paper (4th and 5th grades), Goldhaber and Hansen find that teachers who start out much better than their peers tend to stay much better than their peers in future years. Teachers who are much worse continue to hover at the bottom.

It would appear that districts looking to make good decisions about who to keep or not to keep at a four- or even three-year tenure mark can be pretty confident that they are making a fair decision. However any decision absent at least two years worth of data is likely premature and unfair.

Goldhaber and Hansen throw in this useful fact. If districts held to their guns, ensuring that the bottom performing 25 percent of all teachers up for tenure each year didn't earn it (in effect about 13 percent more than currently don't earn tenure), student achievement could be significantly improved. By routinely denying tenure to the bottom 25 percent of eligible teachers, the impact on student achievement would be equivalent to reducing class size across-the-board by 5 students a class...an intervention that comes with a lot lower price tag.