Brookings Papers on Education Policy 2004 (2004) 25-39
[Access article in PDF]
[Article by Eric A. Hanushek and Steven G. Rivkin]
Comment by Richard Rothstein
Two important themes in Eric A. Hanushek and Steven G. Rivkin's paper deserve further examination. First, the authors argue that excellent teachers can make a huge difference in raising student achievement. If disadvantaged students could have a string of excellent teachers, these students' achievement would rise to the level of middle-class students, eliminating the test score gap. Second, they say that identifying excellent teachers for hire and retention is extraordinarily difficult. Econometric analysis has found little correlation between students' achievement and their teachers' [End Page 25] characteristics, such as certification, teachers' test scores, their verbal ability, or their education beyond a bachelor's degree. Because using any of these imperfect characteristics as screens for teacher hiring narrows the pool of potential teachers, schools should be permitted to hire any college graduates whom administrators believe are likely to raise student achievement. And schools should retain only those teachers who do raise student achievement.
The Role of Teacher Quality in Closing the Achievement Gap
No dispute can be raised about the important impact of teachers on student achievement. But presenting the impact as this paper does ("a string of good teachers can overcome the deficits of home environment") can be misleading and dangerous from a policy perspective.
Hanushek and Rivkin base their conclusion on analyses they have performed using student test score data from Texas, showing that low-income students can get middle-class test scores if they have five consecutive years of teachers who are "at the 85th percentile of teacher quality"; that is, more effective than 85 percent of all teachers, or a standard deviation above the mean of effectiveness. This is similar to analyses by William L. Sanders, using data from Tennessee, finding that students gain 50 percentile points from having three consecutive years of teachers who are more effective than 80 percent of all teachers. 51 Sanders calls this consequence of good teaching "awesome." His work has been widely cited to support the claim that good teaching can overcome learning impediments stemming from low socioeconomic status.
Neither Hanushek and Rivkin nor Sanders claims to identify the characteristics of such good teachers other than by a circular description—good teachers can raise student achievement, and teachers are defined as good if they raise student achievement. Yet even if good teachers could be defined more usefully (I understand that Sanders is attempting to do this in Tennessee), it might do little to eliminate the socioeconomic test score gap.
Partly it is a matter of logic. Researchers have known at least since the publication of the Coleman Report in 1966 (and subsequent reanalyses of the data) that both schools and families contribute to student achievement. Most researchers conclude that families contribute considerably more than schools but that does not negate the proposition that, in theory, stronger schools could overcome negative family influences. [End Page 26]
Yet Hanushek and Rivkin (along with many others who cite their work) employ their findings to minimize the importance of family and social factors in student achievement. For while they are correct that stronger schools could offset, at least in part, negative family influences, stronger families could also overcome weak schools. Hanushek and Rivkin show that low-income children who have five years of 85th percentile teachers can achieve at middle-class levels. Presumably, low-income children who have 50th percentile teachers but who are the beneficiaries of social policies that raise their family characteristics by a standard deviation would also achieve at middle-class levels.
(Because families have been a more important influence than schools, the achievement gain realized by a standard deviation improvement in teachers could probably also be realized by less than a standard deviation improvement in family characteristics.)
So from a logical viewpoint, improvement in either families or schools (or in a combination of both) could be levers to close the achievement gap. It is curious, therefore, that Hanushek and Rivkin, and Sanders, identify schools alone as the appropriate lever...