cators (column 5) leaves the coefficient virtually unchanged (0.073) relative
to the specification that includes only ability controls. These results, con-
sistent with the findings in Romer (1993), suggest that ability is positively
related to both attendance and performance, so that in estimating the effect
of attendance on performance it is crucial to take into account the effect of
unobserved ability. Controlling for effort and motivation, instead, does note
seem to have a major impact on the estimated coefficient for attendance.19
Focusing on the complete specification (column 5), all the ability indi-
cators have a positive and significant coefficient. One additional percentage
point of HSG or GPA corresponds to 0.20 and 0.22 percent improvements in
test score, respectively. The point estimates for exam speed and calculus are
also quantitatively large: one additional exam per annum is associated to a
1.19 per cent higher test score, and students who have passed calculus have
a test score 3.42 percentage points higher than the others.20 The indicators
of motivation and effort also have the expected sign: one additional hour of
study per week produces a 0.14 percentage point increase in performance, al-
though the coefficient is only marginally significant. Subject evaluation has a
positive coefficient (0.07) significant at the 10 per cent level, whereas teacher
evaluation has a significant and larger coefficient: one additional percentage
point in teacher evaluation corresponds to a 0.16 percentage point increase
in test score.
Looking at the other controls in column 5, an additional year of registra-
tion has a significant negative impact on test score of 2.6 percentage points:
the older a student, in terms of academic career, the worse his/her perfor-
mance. Speaking a foreign language and living away from home both have
very large negative and statistically significant effects on test score (-5.03
and -4.45, respectively). The coefficient on the female dummy, on the other
hand, is negative but not statistically significant, indicating that gender does
not have a significant effect on performance, consistently with the results in
Williams et al. (1992) and Durden and Ellis (1995).21.
The results in table 2 indicate that controlling for ability, effort and mo-
19 This could be interpreted as indicating either that effort and motivation are not cor-
related with attendance, or that student and teacher evaluation and hours of study are
not good proxies for motivation and effort.
20This finding is consistent with the results in Brasfield et al. (1992) and Durden and
Ellis (1995).
21 Other studies, however, report significant gender-related differences in performance
(see e.g. Sigfried (1979), Lumsden and Scott (1987)).
12