For children of the eighties, the magic eight ball was the go-to source of information about the future, particularly since you could always be sure of getting the answer you wanted if you just asked it enough times. Fortunately, we can use something substantially more reliable when trying to predict whether new teachers will be effective over the long-run: evidence of student learning.

Using value-added data, Allison Atteberry, Susanna Loeb and James Wyckoff demonstrate that novice teachers who do the best in helping their students learn also tend to be the best in years 3, 4 and 5. 

In the researcher's sample of NYC fourth and fifth grade teachers who taught for five years, the performance of a teacher compared with his or her peers does not change that much over time. Teachers who were top performers in years 1 and 2 tend to stay at the top, while teachers at the bottom of the pack in years 1 and 2 were likely to stay there over the next three years. Even though the lowest group of teachers shows the most improvement, they never catch up, on average, to the highest performers.

Some harder numbers:

• For novice teachers who start out in the bottom quintile, only 7 percent will move into the top quintile of performance in years 3-5

• For novice teachers who start out in the top quintile, only 2 percent will move into the bottom quintile of performance in years 3-5

From these findings, it's clear that when it comes to paying for performance, districts would not regret an extra investment made in their top performers based on two years of value-added data (no "Ask again later" messages here). 

Pushing a bit further on these findings, it would be smart to protect young superstars when facing the unfortunate "reductions in force," rather than blindly laying off nontenured teachers. Why not guarantee that teachers who perform at the top of the pack avoid the dreaded pink slip whether they're at the beginning of their careers or not?

Finally, some food for thought: awarding tenure to the most excellent performers (top 10 percent) based on a couple of years of value-added data would be a safe bet and, possibly, a creative retention strategy.