I recently ran across Eric Hanushek’s October 19, 2010 Wall Street Journal op-ed “There is No ‘War on Teachers.’” Here are two key excerpts from what he has to say:
No longer is education reform an issue of liberals vs. conservatives. In Washington, the Obama administration’s Race to the Top program rewarded states for making significant policy changes such as supporting charter schools. In Los Angeles, the Times published the effectiveness rankings—and names—of 6,000 teachers. And nationwide, the documentary “Waiting for 'Superman,’” which strongly criticizes the public education system, continues to succeed at the box office….
My research—which has focused on teacher quality as measured by what students learn with different teachers—indicates that a small proportion of teachers at the bottom is dragging down our schools. The typical teacher is both hard-working and effective. But if we could replace the bottom 5%-10% of teachers with an average teacher—not a superstar—we could dramatically improve student achievement. The U.S. could move from below average in international comparisons to near the top.
The principle that a small fraction of employees usually causes most of the trouble for productivity is well known in manufacturing. (I think of that as one of the ideas behind the “Six Sigma” approach to process improvement.) Eric is applying that principle to education.
I did some illustrative calculations using the assumption of a normal distribution of teacher effectiveness. (If the distribution of teacher effectiveness has fat tails, the results below would be more dramatic.) On the left is the fraction of teachers at the bottom of the distribution who are fired (firing the same percentage of the replacement teachers until none of the teachers that remain are in the bottom of the distribution). On the right is the resulting change in the mean of the distribution of teacher effectiveness in standard deviations of the original distribution of teacher effectiveness.
Mass deleted Improvement in performance
5% +.11 cross-teacher standard deviations
10% +.20 cross-teacher standard deviations
15% +.27 cross-teacher standard deviations
20% +.35 cross-teacher standard deviations
[The left column is the normal distribution function of x, solved for 5%, 10%, 15% and 20%; the right column is the normal density function–giving an incomplete expectation integral–divided by one minus the normal distribution function.]
To help in interpreting the meaning of these improvements measured in standard deviations, I converted the average teacher effectiveness after deleting the bottom of the distribution into the percentile in the original distribution of a teachers of a teacher who has the same effectiveness as the average effectiveness of teachers after the bottom of the distribution has been fired.
Mass deleted Average quality = what percentile of original distribution?
[The right column is the normal distribution function of the performance improvements in the table above.]
Update: @aryal_ga flags some nice graphs of the contribution of teacher quality to long-run student outcomes created by Raj Chetty, John Friedman and Jonah Rockoff: