POLICY VERSUS EXPLANATION OF OUTCOMES
the possibility that something other than teacher quality also varied acrossclassrooms). If a good teacher makes a big difference, and if we have aneffectivewayofraisingaverageteacherquality,itisworthinvestinginteacher-oriented policies.The need to untangle the confusion between policy effects and analysisof variance seems to be a permanent feature across a variety of social sciencetopics (see, e.g., Manski 20
). In education there is a long history of respectedpolicy advocates and social scientists focusing on the extent to which variationis explained by nonschool factors. Hanushek and Kain (
972, footnote 3) de-scribe this as the “no-school-effect conclusion.” An early example appears inthe Coleman report:It is known that socioeconomic factors bear a strong relation to aca-demic achievement. When these factors are statistically controlled,however, it appears that differences between schools account for onlya small fraction of differences in pupil achievement.
Hanushek and Kain (
36) go on to explain:The closest the [Coleman]
comes to identifying policy instru-ments are its estimates of the
. . .
contribution of individual variablesto explained variance.
. . .
The proportion of explained variance doesnot identify policy instruments and gives little indication of the policyleverageprovidedbydifferentvariables.Parameterestimatesaremuchmore useful in this respect.
. . .
. . .
is simply not avery interesting concept
. . .
to the policy-maker.Morerecentcommentatorsalsoseguefromtheobservationthatmostofthevariability in student outcomes comes from outside school to the conclusionthat in-school factors, notably quality teachers, cannot accomplish the goals of educational reform (note the ﬁnal sentence):[A] central thesis in today’s school reform discussion [is] the idea thatteachers are the most important factor determining student achieve-ment. But this proposition is false. Hanushek has released studiesshowing that teacher quality accounts for about 7.5–
0 percent of stu-dent test score gains. Several other high-quality analyses echo thisﬁnding, and while estimates vary a bit, there is a relative consensus:
. Coleman et al. (
966, pp. 2
22) and cited in Hanushek and Kain (
972, footnote 3). Note that“differences in schools” presumably references differences by race. Ehrenberg and Brewer (
995)show that teacher ability mattered even in the original Coleman data.