is not robust to the inclusion of the number of hours of study. Dolton et al.
(2003), applying stochastic frontier techniques to a large sample of Spanish
students, find tha both formal study and self study are significant determi-
nants of exam scores but that the former may be up to four times more
important than the latter. However, they also find that self study time may
be insignificant if ability bias is corrected for.
All of these studies, with the exception of Marburger (2001) and Rodgers
(2001), are based on cross-sectional data sets. As a consequence, as observed
by Romer (1993), the possibility that the estimated relationship between at-
tendance and exam performance reflects the impact of omitted factors rather
than a true effect cannot be ruled out. In the following we thus report re-
sults obtained using panel data on Introductory Microeconomics students to
estimate the net effect of attendance on academic performance.
3 Data
The data for this study were collected by conducting a survey of 766 students
attending the Introductory Microeconomic course at the University of Milan
in the academic years 2001 to 2004.6 The course, taught over twelve weeks in
the spring semester, is taken by all students in the first-year of study of the
Economics degree. The exam is based on four mid-term tests, administered
every three weeks, covering equal fractions of the course and carrying the
same weight for the final grade. Questionnaires were distributed to students
with each of the four test papers, and compiled before starting the tests.
This produced four independently pooled panels (one for every year), each
with a cross-section of about two hundred students observed over four tests,
resulting in a potential balanced panel of 3064 observations (N = 766 times
T = 4). The number of observations for the actual (unbalanced) panel is
2913 (T = 3.8), due to inconplete questionnaires and a number of students
dropping out before the end of the course.7
Summary statistics for the main variables are reported in table 1. Aca-
demic performance is measured by students’ test score (SCO). The actual
6 There are 7 parallel sections of the Introductory Microeconomic course offered to about
1,500 first year students. All sections have the same content (syllabus and textbook), even
though the number and format of intermediate tests may differ.
7Note that only the three main variables of interest (test score, lecture attendance and
class attendance) are time varying, whereas all other variables are time invariant.