85 C. Lo¨fgren, H. Ohlsson Economics of Education Review 18 1999 79–88
resents the possibilities to complete the thesis for each of the students in semester t who did not complete
before. These possibilities are determined both by prefer- ences and ability. Suppose also that Z
it
is a linear func- tion of the explanatory variables x
it
. We have Z
it
5 a 1 bx
it
Assume also that the larger Z
it
, the larger the student’s possibilities to complete in semester t. Say more specifi-
cally that if Z 0 the thesis will be completed. This means that P
it
5 FZ
it
. Let F be a cumulative logistic distribution function. Then
P
it
5 1
1 1 e
− a 1 bx
it
This results in the following likelihood function: L 5 P
n i 5 1
S
1 1 1 e
− a 1 bx
it
D
y
it
S
1 1 1 e
a 1 bx
it
D
1 2 y
it
6 which has been maximized with respect to the para-
meters a and b using the LIMDEP package Greene, 1995. The estimation results are presented in Section 5.
5. Evidence
When describing the data we found that coauthors do much better than single authors. This still holds in the
hazard estimation when simultaneously controlling for the effects of all other variables, as can be seen from
Table 4. A coauthor is more likely to complete. The esti- mations also suggest that coauthors complete because of
ability, not preferences.
5
The coefficient in the estimation for all is almost identical to that for completers. The
theoretical model in Section 2 suggests that, given that adequate data are available, the estimated effects for stu-
dents receiving a specific grade will capture differences in ability or in educational production functions.
For D- Master’s thesis a conclusion also about the preference effect is possible. Students writing D-theses
are less likely to complete than those writing C-theses column 1 in Table 4. However, if the D-students do
complete, and particularly if they do so passing with dis- tinction, they do it in a shorter time than the C-students
who complete column 4 in Table 4. An explanation of this result is that the preference and the ability effects
work in opposite directions. The D-students are at the end of their education. They have a choice of giving up
5
Ability should be broadly interpreted. In the theoretical model differences in ability are defined as differences in the
educational production functions.
their plans for a D-thesis and graduating with a Bach- elor’s degree or completing their D-theses and gradu-
ating with a Master’s degree. This is a question of prefer- ences. But if they do decide to complete, and particularly
if they strive for the highest grade, the reason they do it in a shorter time than the C-students is because of higher
ability higher educational output for given study effort. They already have the training in thesis writing from
their C-thesis experience.
Students in Umeå show a longer completion time than students in Uppsala. A reason for this could be the
stricter rules for thesis work in Uppsala. At this univer- sity students are only allowed to present their theses dur-
ing a two-week period at the end of each semester. There is a long wait for the next occasion if you do not finish
in time. In Umeå students are allowed to present their theses whenever they are finished. It is possible that the
Umeå students in this situation take a little longer time to complete. The way we have interpreted the theoretical
model this will show up as an ability effect — the edu- cational production functions differ between the two uni-
versities.
The hazard rate for thesis completion varies with time. The probability of completing is lower from the 4th sem-
ester to the 7th semester compared to the 1st semester. If we restrict the sample to those completing and those
passing, there are no significant time effects. This sug- gests that the decreasing probability over time has to do
with preferences rather than ability. For the students who pass with distinction the hazard rate is higher in the 2nd
and 3rd semesters compared to the 1st semester. This ability effect says that more effort is needed to receive
the highest grade.
There are also some empirical results with borderline significance worth mentioning: High grades in prior eco-
nomics courses increases the probability of completing. If we restrict the sample to those passing, the estimated
effect is significant while it is still positive but insignifi- cant for the full sample. This suggests that the effect has
more to do with ability than preferences.
We noted in Section 3 that women in Uppsala seem to take a longer time completing than men. The esti-
mation results in Table 4 show a negative effect significant at the 10 level for women when compar-
ing among all students but no such effect within the group of completers. So if women take longer time to
complete than men this seems to be because of prefer- ences and not ability.
6
6
Dynan and Rouse 1997, using data from Harvard Univer- sity, estimate models of the probability of majoring in econom-
ics. They find weak evidence that there remains a difference between men and women after controlling for background vari-
ables. Their interpretation is that this remaining gap is explained by taste differences or differences in other unmeasured charac-
teristics.
86 C. Lo¨fgren, H. Ohlsson Economics of Education Review 18 1999 79–88
Table 4 The probability of completing a thesis, logit estimations
All Completing
Passing Passing with dist.
Background Woman
2 0.46 1.65 0.02 0.06
0.13 0.25 2 0.12 0.18
Age 2 0.07 1.05
2 0.11 1.16 2 0.12 0.67
0.03 0.15 Science program,
2 0.04 0.13 0.10 0.29
0.17 0.31 0.02 0.02
secondary school Grade point average,
0.06 0.18 2 0.33 0.73
0.25 0.32 0.09 0.12
secondary school Study time, economics
2 0.09 0.70 2 0.08 0.41
2 0.04 0.14 2 0.02 0.05
High grades, economics 1.02 1.51
1.65 1.94 4.35 2.36
2 0.70 0.46 Study program
Public administration 0.16 0.41
2 0.08 0.15 2 0.06 0.08
2 0.70 0.47 Business economics
2 0.31 0.66 2 0.74 1.29
2 2.33 2.24 0.44 0.44
Social science 0.08 0.14
2 0.54 0.82 2 1.80 1.18
2 0.89 0.90 International economics
0.46 0.66 0.13 0.15
2 1.05 0.74 2 1.30 0.76
Other programs 2 0.14 0.21
2 0.78 1.04 2 0.40 0.38
2 2.91 1.94 Thesis
Coauthored 1.35 4.56
1.30 3.72 1.44 2.80
1.14 1.64 D-thesis
2 0.78 2.13 0.55 1.16
0.36 0.52 2.42 2.43
Applied econometrics 1.54 2.60
0.80 1.32 0.42 0.46
1.27 1.10 Fixed effects
Spring 1993 2 0.48 1.44
2 0.72 1.87 2 0.84 1.70
0.31 0.33 Umeå
2 1.00 2.50 2 1.57 3.28
2 1.34 1.42 2 2.13 2.68
2nd semester 0.21 0.70
0.66 1.79 2 0.16 0.31
2.20 3.42 3rd semester
2 0.41 1.02 0.48 0.94
2 0.35 0.52 3.90 2.89
4th semester 2 1.55 2.34
2 0.38 0.49 2 0.35 0.43
2 7.20 0.04 5th semester
2 1.12 1.91 0.98 1.13
0.61 0.65 6th semester
2 2.48 2.37 0.04 0.03
0.18 0.14 7th semester
2 1.68 2.17 Constant
1.19 0.54 4.14 1.26
2.59 0.44 2 0.52 0.10
Log likelihood 2 197.50
2 138.40 2 73.69
2 47.57 Restricted log likelihood
2 258.69 2 159.72
2 93.32 2 66.27
x
2
, significance level 0.0000
0.0035 0.0091
0.0071 Number of observations
409 234
136 98
Note: Absolute t-values in parentheses.
Students from the applied econometrics course are more likely to complete their theses than other students.
This appears to be more an effect of preferences than ability since there is no such effect when comparing only
among completers.
Students who first registered for thesis work in the spring semester of 1993, compared to their counterparts
in the fall semester, are likely to take longer completing given that they do complete during the observed seven
semesters. This is shown by the estimation results of models 2 and 3. It is a result opposite to the one we
expected. Our hypotheses, described in the preceding section, was that these students could benefit from hav-
ing the summer vacations “earlier” than the fall students in the row of the seven semesters. By using part of the
summers for thesis work they could complete in a shorter time, all other variables constant, than the fall students.
However, our results indicate that the summers are detri- mental rather than beneficiary for thesis completion. A
plausible explanation is that for students that do not spend the summer working with their thesis the long
pause in thesis work hurts their ability. The break for Christmas vacation is much shorter, only a few weeks.
87 C. Lo¨fgren, H. Ohlsson Economics of Education Review 18 1999 79–88
6. Concluding remarks