330 THE PROB LEMS OF EDUCATION ARE NOT GE NE TICcorrelation “for per for mance self- effi cacy, which was the strongest
correlate (of 50 mea sures).”^16
Similar conclusions have been reached with the United Kingdom’s
A- level grades (at the end of high school). Th ey have always been uncer-
tain predictors of per for mance at university, as admissions tutors have
long realized. Studies in the 1980s found that A- level grades have little
predictive value for per for mance at university, either in medicine or in
nonmedical subjects.^17 A study in the mid-1990s indicated that A- level
grades accounted for only 8 percent of the variation in fi nal degree per-
for mance on average. A study by King’s College London a few years ago
confi rmed that, when picking the candidates with the best potential,
universities might just as well toss a coin. Th e 2012 report by Richardson
and colleagues, mentioned above, found that “in U.K. data, a small cor-
relation was observed between A level points and university GPA (r = 0.25),
again refl ecting previous fi ndings.” An enquiry commissioned by the
U.K. government noted that “many respondents were of the opinion that
A levels do not seem to have a strong relationship to actual per for mance
in higher education— and that those with the best grades can perform
disappointingly (and vice versa).”^18
In the United States, associations between high school and college
grades have been a focus of much debate. One aspect of this has been in-
consistencies between the diff er ent mea sures used as predictors. For ex-
ample, as reported by the College Board in 2008, the correlation between
SAT scores and high school grade point average (HSGPA) is only 0.28. As
the Board says, “Th is fi nding suggests that the SAT and HSGPA may
mea sure diff er ent aspects of academic achievement,” though it does not
suggest what those aspects may be.
Even predictions of high school grades for fi rst- year college grades have
never been very high (0.3–0.6). Correlations with fi nal college grades fall
off to about 0.3–0.4, but this is aft er a number of statistical corrections
that always involve some guesswork. Suggesting they might be “veils of
deception,” one review responded to them as follows: “Th e ETS [Educa-
tion Testing Ser vice] has recently released new estimates of validities of
the GRE [Gradu ate Rec ord Exam] for predicting cumulative gradu ate
GPA. Th ey average in the middle thirties— twice as high as those previ-
ously reported by a number of in de pen dent investigators. It is shown...
This content downloaded from 139.184.14.159 on Tue, 17 Oct 2017 13:58:56 UTChttp://www.ebook3000.com