A surprisingly low number

Reading an old article on the Atlanta Public School Cheating scandal, Wrong Answer by Rachel Aviv.

There has long been argument about the role of schools as a means of making up for broader familial and social dysfunction as well as arguments about the relative impact of teachers on educational outcomes (as opposed to other influences such as genes, familial home circumstances, etc.). There is a fair amount evidence that high performing teachers have a material impact on high performing student results but we are dealing with averages here.

To explain the improvement in scores, Hall told the investigators that “an effective teacher three years in a row will completely close the gap between a child born in poverty and a child born to a middle-income family.” This theory, in its earliest form, derives from a study by William L. Sanders, a statistician formerly at the University of Tennessee, but the findings, which have contributed to a nationwide effort to rate teachers rigorously, have been overstated to the point of becoming a myth. According to a recent statement by the American Statistical Association, most studies show that teachers account for between one and fourteen per cent of variability in test scores.

I have never seen that figure before, "that teachers account for between one and fourteen per cent of variability in test scores." I wonder about the methodology, replication, etc. But that is a surprisingly low number.