Thesis for the Bachelor of Science in Economics
Marco Gregori
January 2016
Student’s name: Marco Gregori Student’s number: 10621210
Programme: BSc Economics and Business Specialization: Economics
Code: BSc ECB
Title of the thesis: Academic dismissal policies and students’ performance Assigned supervisor: Andro Rilovic
This document is written by Marco Gregori who declares to take full respon-sibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the super-vision of completion of the work, not for the contents.
By Marco Gregori⇤
A di↵erence-in-di↵erences estimation is used to investigate the e↵ect of an academic dismissal policy (BSA) on students’ perfor-mance across the first two years of university. Surprisingly, it is found that students who might have targeted the BSA threshold in the first year perform better in the following year. The result seems however due to a substantial proportion of would-be top-performing students who relax, worsening in performance, and thus to an in-accurate design. A dip in performance in the second year is still found across a group of students. Finally, there appears to be some gender di↵erences among the characteristics of this dip.
Academic dismissal policies are a common way to boost students’ performance at university. In the specific case of The Netherlands, students are often required to achieve a minimum number of credits during the first year, known as the Bind-ing Study Advice1 (BSA). This mechanism operates as a selection device as well
as a “stick”, sorting the most talented and eager students and motivating them to study. After the first year, there is no similar requirement. Some studies, such as Arnold (2015) and Sneyers and de Witte (2015), have documented this policy, comparing results before and after the introduction of the BSA, and generally appraising it as an e↵ective selection device. Little is known, however, about its e↵ectiveness on students’ motivation. Indeed, the Education Inspectorate recog-nized in a report (2010) the lack of systematic empirical analysis surrounding the e↵ects of the BSA; the BSA policy may have heterogeneous e↵ects on students’ behaviour that, to date, have not been sufficiently investigated.
The main concern of this thesis is whether the BSA policy is e↵ective only when enacted in the first year or if it induces a permanent change in students’ behaviour, pushing them to improve performance during the full program. Previous literature on cognitive psychology and applied research in higher education in America reports that similar performance devices often yield a short term improvement, without actually inducing a consistent, permanent change in behaviour2.
The main focus is on the e↵ectiveness of the BSA requirement for the program of Economics and Business o↵ered at the University of Amsterdam (UvA). The BSA policy requires students to earn at least 42 credits in the first year and to pass the course “Mathematics”, while no similar requirement is enforced in the successive years. The question is whether this minimum requirement is e↵ective only in
⇤ Thanks to Andro Rilovic for thorough supervision and to Fred Pope for providing the data-set and
for precious suggestions. In particular, subsection V. B. is based upon an original idea of Fred Pope.
1In Dutch, Bindend Studie Advies
2For research in tertiary education the main references are: Lindo, Sanders, and Oreopoulos (2010);
Casey, Cline, Ost, and Qureshi (2015); Fletcher and Tokmouline (2010) 2
the first year, by setting a floor on performance, or if it beneficially influences performance even later. Using a di↵erence-in-di↵erences design, the divergence in performance in the second year of the Bachelor programme is estimated between the groups of students who collected 42 and 60 credits in the first year. The essential idea is that if top performers in the first year are una↵ected by the BSA policy, they will perform equally well in the next year. Students who collected 42 credits, instead, might have targeted the minimum threshold and, since no requirement is enacted in the second year, they might exert less e↵ort.
Contrary to expectations, the initial results indicate that students who collected 42 credits tend to perform better in the second year. However, a more scrupulous investigation reveals that the result is mostly driven by some of the students who collect 60 credits in the first year and perform substantially worse in the next years. The initial identifying assumption in the design seems to be inaccurate. The interpretation is then ambiguous, but, when considering the students who manage to finish the programme in the official time of three year against those who do not manage to, there appears to be some evidence of a dip in performance in the second year, when the BSA requirement is absent. Finally, this dip in performance is more severe for males, which concurs with the previous finding (e.g., Arnold, 2015) that males are more responsive to an academic dismissal policy.
The outline of this thesis is as follows: First, the institutional background of the UvA is introduced, including an overview of the attributes of the BSA policy. A theoretical background is presented in section II, setting forward a theory of motivation. Then, the key variables and descriptive characteristics of the data are presented and the sample selection is discussed in section III. The design for the di↵erence-in-di↵erences estimator is illustrated and motivated in section IV, while the estimation results are presented in section V. The analysis is then extended by changing groups of comparison. Section VI concludes.
I. Institutional Background
The UvA is part of the Dutch system of higher education: its Economics and Business bachelor program comprises 60 credit points per year (following the European Credit Transfer System) over the course of three years. Accordingly, 1 credit corresponds to a workload of 28 hours, with all courses currently o↵ered bearing 6 credits, resulting in a workload of about 168 hours per course (in terms of classes, take-home preparation, exams, and other assessments). The UvA requires a minimum amount of credits to be obtained in the first year by a student to continue his studies, known as the BSA, which has varied over the years. In the academic years 2006-2007 and 2007-2008 the threshold corresponded to 35 credits, in 2009-2010 45 credits and, finally, since 2011 42 credits. This is close to the average across Dutch universities, which is 38 credits (Education Inspectorate, 2010).
students who make insufficient progress with their studies. Initially, the official purpose of the BSA was to identify underachievers as early as possible and to encourage them to pursue another study match, selecting early more suitable students (Education Inspectorate, 2009). Indeed, there appears to be a consistent empirical finding that the majority of students who perform unsatisfactorily in the first year are not successful at university over the whole course of studies (Schneider, 2010; Murtaugh, Burns, & Schuster, 1999). However, the Education Inspectorate recognized in a successive report (2010) that the BSA has also been a useful stick to encourage students to put sufficient e↵ort in their studies, as its introduction corresponded to an increase in initial dropout rate but also to a decrease in time to completion of the bachelor of the continuing students. In the Netherlands, 7 percent of the total higher education budget is allocated based on performance agreements, mainly using dropout and completion rate as criteria (Sneyers & de Witte, 2015): the BSA policy might have been gradually adopted from several universities to enhance their performance in light of these criteria.
Arnold (2015) compares the results within bachelor programs across universi-ties in the Netherlands that did and did not apply the BSA policy: he finds that the average 4-year completion rate is not significantly a↵ected by the introduc-tion of a BSA policy when including the first year drop-outs. However, there is a significant e↵ect on the timing of study success: there is a larger first year dropout, but stronger study progress of first-year survivors, indicating a marked improvement in the selection process. These e↵ects are however highly heteroge-neous: non-significant for programs within the applied sciences, while particularly marked for economics programs, in which the 4-year completion rate rises from 44.1 to 57 percent when excluding the first-year dropouts. Similarly, Sneyers and de Witte (2015) estimate, using a di↵erence-in-di↵erences design, that the BSA policy results in an increase of 7.5 percent in student dropout as well as an in-crease in 3.3 percent in graduation rates, although yielding a dein-crease in students’ satisfaction. Finally, both studies indicate a substantial gender e↵ect, as the BSA policy mostly a↵ects male students’ performance.
II. Theoretical Background
The previous section provided an overview on the e↵ects of introducing the BSA policy across the Netherlands. However, the main concern of this study is on the incentives that the BSA induces on students. Thus, to formulate a testable hypothesis a theory of incentives and motivation is needed. Cognitive evaluation theory, and some prior empirical results, are employed to propose that the BSA motivates students when enacted, positively a↵ecting performance in the first year, but has no impact in the successive year.
The BSA can be categorized as punishment. Due to obvious concerns about students’ welfare, most of the literature in the context of education has focused on rewards as incentives, but some conclusions can be easily extended to pun-ishments. Personnel economics posits that incentives promote e↵ort and
perfor-mance (Gibbons & Roberts, 2013). However, cognitive evaluation theory has long pointed out that incentive schemes in the context of education may backfire, impairing performance in the long-run: Deci (1975) first proposed that rewards, although e↵ective as short-term reinforcers, may crowd out intrinsic motivation, ultimately making individuals less proactive. This sparked a long list of exper-imental research within the context of education, yielding inconclusive answers
3.
The fundamental insights of cognitive evaluation theory have been formalized by Benabou and Tirole (2003), who prove in a two-stage principal agent game how rewards are positive “short-term reinforcers” for the current activity, but “undermine agents’ assessment of the tasks attractiveness”, yielding a negative impact on persistence (p. 497). Although they focus on a positive performance mechanism, their theory yields analogous results with punishments in place of rewards. There have been some applications of their model in the context of academic probation in America. Academic probation is similar to the BSA, but idiosyncratic in that it blends the threat of punishment with encouragement and coaching. In contrast, the BSA sets a clear-cut minimum performance require-ment on students.
In particular, Lindo, Sanders, and Oreopoulos (2010) find that, consistently with Benabou and Tirole’s model, probation initially improves the GPA of the continuing students, but the net impact on the probability of graduating is nega-tive: the probability of dropping out outweigh the initial motivating e↵ect on the continuing students. Their estimates are however highly heterogeneous: notably, being placed in probation after the first year “reduces the probability of gradu-ating within six years by 14.5 percentage points” for students with high school grades above the median (p. 112). Further, the impact of probation is more severe on males, who have a higher probability of dropping, than on females, who tend to be rather unresponsive. This finding is consistent with previous empirical investigations, indicating that women tend to dropout less than men, perhaps due to biological factors or family background (Peltier, Laden, & Matranga, 2000). This strain of literature is also a motivating factor for investigating whether the BSA may have heterogeneous e↵ects on males and females.
Fletcher and Tokmouline (2010) also find, in an enquiry in Texan universities, that probation boosts performance in the subsequent semester, but has no e↵ect in the longer term. In contrast to Lindo et al. (2010), they find di↵erent responses to academic probation based on heterogeneous students characteristic, but they cannot trace any consistent pattern across the various universities. Casey, Cline, Ost, and Qureshi (2015) also dispute the result that probation boosts short-term performance: although they find a similar short-term boost, they document how students engage in “strategic-course taking”: they attempt fewer credits, enrol
3See Cameron and Pierce (1994) for a meta-analysis of experimental results claiming that performance
incentives have negligible extrinsic e↵ects on motivation; see Deci, Koestner, and Ryan (1999) for a contrary opinion.
in easier disciplines, and withdraw more often from attempted classes. They still find substantial heterogeneity in the extent to which students strategically alter their course-taking behaviour.
Finally, it is important to mention, for sake of completeness, that there have been some studies of the selection e↵ect of the BSA policy in some bachelor programs, comparing the first-year results of some cohorts of students immedi-ately before and after its introduction. Paradoxically, both studies by Stegers-Jager, Cohen-Schotanus, Splinter, and Themmen (2011) and Eijsvogels, Goorden, van den Bosch, and Hopman (2015) report that the introduction of the BSA pol-icy did not lead to higher dropout nor higher completion rates at medical school. Examining these results through the analysis of Arnold, however, indicates that academic standards at medical university are already quite high and thus the additional e↵ect of the introduction of the BSA policy is close to nil.
Surprisingly, de Koning, Loyens, Rikers, Smeets, and van der Molen (2014) find in a bachelor program of psychology that after the introduction of the BSA students’ results slightly worsened, although the estimates are imprecise and the number of credits obtained was always, on average, above the minimum require-ments. Students’ reported self-study time and observed learning activities (as reported by tutors) were non-significantly higher in the group a↵ected by the BSA policy. To reconcile the two contrasting results, de Koning et al. suggest that students, although exerting more e↵ort, might target the BSA threshold rather than “acquiring as much knowledge, and hence highest performance, as possible” (p. 845). This finding suggests indeed the possibility that the BSA might yield only short-term motivation, pushing some students to earn exactly the minimum amount of credits required.
III. Data
The data used is from an administrative data-set of the students in the program of Economics and Business of the University of Amsterdam. Observations are at student level and cover all students enrolled at the faculty in the academic years from 2009 to 2014. The data used includes gender, high school GPA4, the BSA
received (maximum if they finished the first year with 60 credits, positive if with less than 60 credits but more than 42), results received per exam, average GPA, and number of credits obtained each year. The average GPA is calculated each year from all recorded attempts, thus including failed as well as passed exams and providing a better proxy of performance; for instance, it includes 3 di↵erent grades for the same subject if the student wrote the exam for the same course in the same year and the first two attempts were unsuccessful.
A number of data is dropped for various reasons, succinctly summarized in Table 1. There has been an administrative change in courses structure (from a
4With the exception of a small group of students coming from vocational school (in Dutch Hoger
semester to a block system) in 2011, rendering the ceteris paribus assumption for the 2009 and 2010 cohorts untenable: briefly, starting 2011 students take two courses each two months and write exams at the end of each two-months block; before 2011 students attended classes during the whole semester and wrote exams at the end of each semester. The 2013 and 2014 cohorts are instead too “young” in the sense that, since the official time to completion is 3 years, they have not finished their studies yet. The remaining cohorts 2011 and 2012 have similar characteristics at the moment of intake and similar academic results in the first and second year; further, the partition across specialization is not extremely di↵erent, as illustrated in Table A1 in the Appendix. Note that for the 2011 cohort the number of students who completed the bachelor in four years is also available. However, the percentage of students who completed the programme in 3 years is similar: 47.1 percent in 2011 and 45.3 percent in 2012. Students in both cohorts tend to perform slightly worse on average in the second year, earning about three/four credits less. Moreover, the standard deviation in the distribution of credits more than doubles, suggesting contrasting di↵erentials in performance.
The students who did not continue their studies because they received a negative BSA are dropped as well, since the focus is on performance across continuing students5. Finally, some students are dropped because they stopped following
the program early in the second year: 12 students discontinued their studies at the begin of the second year, writing less than 3 exams, while 3 students switched to the Econometrics programme.
Table 1—Number of sample removals by reasons
Number of students Reason for removal
3201 Total number of students
2346 1. Exclude cohorts 2009, 2010, 2013, 2014.
855 Cohorts 2011, 2012
245 2. Exclude negative BSA
610 Continuing students
15 3. Did not continue his studies
595 Remaining sample
In the first year, students follow a fixed program out of a mix of Economics and Business courses; in the next year, however, they choose a specialization between Economics, Finance and Organization, Economics and Finance, Business Administration or Accountancy and Control. A new variable is added according
5They have already been omitted from the data-set (under request) students who discontinue their
studies in the first year and students who have received their BSA under exceptional circumstances (e.g. undergoing heavy illnesses or bereavement)
to their specialization chosen6 (between Finance or Economics, Business Studies,
or Accountancy and Control). Students are clustered across specialization to ensure that they approximately take the same courses and write exams of similar difficulty. In the specific case of the specialization of Finance and Organization, Economics, and Economics and Finance it is difficult to distinguish their choice of courses in the second year: students following the Economics specialization or the Economics and Finance specialization follow the same courses, and they have seven (out of ten) courses in common with students following the Finance and Organization specialization. As a result of this, they are clustered in the same group, labelled as Economics or Finance.
IV. Empirical Strategy
This thesis aims at testing the following hypothesis:
Due to di↵erences in responsiveness to the BSA policy, performance of a group of students undergoes a dip in the second year, as they earn less credits and obtain a lower average GPA as compared to the first year
My claim is that the absence of a BSA policy in the second year is the best explanation behind this possible drop, assuming that the right counterfactual is used. Since the design of this study is not a randomized experiment, but based upon a comparison of means, it cannot be proven that this is the causal e↵ect of the BSA policy. Yet, it is a sensible suggestion given that there appears to be no other relevant change (except for courses and programme-related changes, which can partially controlled for) across the two years.
In order to select a group of students that might be responsive to the BSA policy, an identifying assumption is needed. The previous research on academic probation used a regression discontinuity around the cuto↵ for being put into probation, the crucial assumption being that the two groups of students around the cuto↵ were similar in everything except for being put into probation. Thus, the group just above the cuto↵ functioned as counterfactual and the di↵erence in performance with the group just below the cuto↵ across multiple years could be reasonably deduced as the e↵ect of probation7.
Similarly, it appears a priori reasonable to assume in the present study that students who collected 42 credits in the first year have been more responsive to the BSA policy than the students who collected 60 credits, in the sense that, had
6The specialization is not included in the administrative data, but inferred from their course choice,
according to the students’ Bachelor Thesis course, which is unique per each specialization. If they did not write the Bachelor thesis yet, the specialization is inferred from their second year courses choice, which is also di↵erent across specializations. A small number of students may therefore conclude their studies in a specialization di↵erent from that indicated since they might swap in the third year. Apparently one student in the 2011 cohort and one student in the 2012 cohort chose to study Economics in the second year but switched to Business Administration in the third year of their studies.
7This design has been used by Lindo et al. (2010), Casey et al. (2015), and Fletcher and Tokmouline
the threshold not existed or had it been lower, they would have exerted less e↵ort. The students who collected 60 credits are however less, or not at all, responsive to the threshold, being already top performers, and may therefore function as counterfactual. Indeed, the previous research of Stegers-Jager et al. (2011) and Eijsvogels et al. (2015) illustrate that, in medical schools, top performers are una↵ected by the introduction of the BSA. Then, the di↵erence in performance between the two groups across the two years may be the e↵ect of the BSA, as there is no similar minimum requirement in the second year. Cognitive evaluation theory predicts that, as soon as the BSA requirement is lifted performance of the students who were responsive to the BSA policy will drop8.
However, this may not be the case if the identifying assumption does not hold. For instance, the BSA may have a positive feedback, pushing students into exert-ing more e↵ort in the second year because they have discovered to actually enjoy their studies. Alternatively, students who obtained 42 credits in the first year might be highly efficient, targeting a number of credits for minimum e↵ort, and aim for a higher target in the next year. It must be borne in mind that this is not a quasi-experiment, thus the assumption may be ill-judged and the estimates do not necessarily show the causal impact of the BSA. If the right counterfactual is chosen and cognitive evaluation theory holds, the intuitive mechanics of the treatment under an idealized setting are illustrated in Figure 1.
Figure 1. Difference-in-Differences design
For a valid treatment, the two groups need to have similar characteristics and
8The deeper psychological motivation may be due to these students exerting some sort of credits
trend similarly over time. The students within the two groups are all “sufficiently” smart to make it past the threshold. They are clustered within specialization to ensure that they take similar courses9. The essential di↵erence is their respon-siveness to the BSA policy: students who collected 60 credits in the first year are overachievers, already highly motivated, while students who collected 42 credits needed a stick to exert that level of e↵ort. Performance is measured in terms of credits obtained, as 1 credit officially corresponds to about 28 hours of study. The divergence in performance is measured instead using a di↵erence-in-di↵erences es-timator. The following regressions is run, first for all students and then across specialization:
(1) Credits = Y ear2+ 42 Credits + Y ear2⇥ 42 Credits + Intercept + "
Where Credits is the number of credits collected by the individual student at the end of each academic year, Y ear2is a dummy variable equal to 0 if the student
is in the first academic year (2011-2012 for the 2011 cohort and 2012-2013 for the 2012 cohort) and to 1 if student is in the second academic year (2012-2013 for the 2011 cohort and 2013-2014 for the 2012 cohort), 42 Credits is a dummy variable equal to 0 if a student collected 60 credits and to 1 if she collected 42 credits in the first year (using this specification, it follows that its coefficient is always -18, while the Intercept is always 60 credits), Y ear2⇥ 42 Credits is the
di↵erence-in-di↵erences estimator, and " is the error term. The dummy Y ear2 captures
possible di↵erences in courses difficulty and any other possible di↵erence a↵ect-ing all students across the two years while the dummy 42 Credits dista↵ect-inguishes first year performance. The prediction of cognitive evaluation theory is that the coefficient on Y ear2⇥ 42 Credits, indicating the di↵erential in second year
per-formance across the two groups, is significantly negative: students who collected 42 credits will perform significantly worse. As there is no clear reason to assume homoskedastic variance, neither for all students nor across specialization, robust standard errors are used (Stock & Watson, 2007).
Similarly, the same design is used measuring performance in terms of GPA: (2) GP A = Y ear2+ 42 Credits + Y ear2⇥ 42 Credits + Intercept + "
The only requirement is that the coefficient is the same as that of regression (1); an increase in GPA coupled with a drop in number of credits might imply that students are re-taking the easier first year courses, studying less while obtaining higher results.
9Students who collected 42 credits failed three courses in the first year, which they might retake in
the successive year. They select themselves which exams to prepare for among the courses o↵ered, and obviously prefer choosing courses in which they have a better probability to succeed; this weakens the possible objection that a drop in performance may be the consequence of worse first-year preparation. It is indeed theoretically possible that they perform better, by writing exams for courses that they have already followed in the previous year.
A prediction for students who collected 48 or 54 is unclear, since they could be put in between the two groups in terms of motivation. Some regressions are also run where 42 Credits is a dummy variable equal to 0 if a student collected 60 credits and to 1 if she collected 48 or 54 credits respectively, to test whether the estimates have the same sign across groups; this will indicate if there exists a consistent pattern across the various groups of students. For instance, students who collected 42 credits performing worse in the second year as compared to the students in all other groups might indicate that the former group was highly re-sponsive to the BSA policy, in contrast to the other groups, thus upholding the validity of the design. In contrast, strongly heterogeneous performance across groups would suggest that other factors, not captured from the design, are a↵ect-ing performance.
V. Results A. Estimation results
Contrary to expectations, the coefficients of the di↵erence-in-di↵erences esti-mator is positive: the results, shown in Table 2, imply that the students who collected 42 credits in the first year manage to collect a larger number of credits in the next year, while those who collected 60 credits collect a smaller number. The coefficients of Y ear2 show indeed that students who collected 60 credits
tend to perform worse in the second year, collecting about 4 credits less, for an average total of 56 credits. The coefficient on the di↵erence-in-di↵erences estima-tor, Y ear2⇥ 42 Credits is significantly positive and larger than the coefficient of
Y ear2, also when bundling for specialization, with the exception of Accountancy
and Control students, for whom the coefficient is insignificantly negative. The sum of the coefficients of Y ear2 and Y ear2⇥ 42 Credits imply that students who
collected 42 credits in the first year collect on average 1.5 credits more in the second year, for a total of 43.5 credits. There is a significant dip in performance, but not within the group in which I expected to find one.
To understand what is driving the result behind the di↵erence-in-di↵erences estimator, I analyze the specific time trends across specialization, formulating some conjectures on the leading factors behind these results. The pattern is het-erogeneous: students who collected 42 credits in the first year collect on average 3.9 credits more in the second year in the Economics or Finance specialization and 5.3 credits credits more in in the Business specialization. In contrast, those in the Accountancy and Control specialization tend to perform worse, collecting 6.3 credits less in the second year. A preliminary explanation for the di↵erential in performance across specializations might lie in the phenomenon of “strategic course-taking” found by Casey et al. (2015): for instance, Business student are pursuing a specialization where the mathematical curriculum is not equally de-manding as in the first year, when they were bundled with the other students, and thus are able to study faster taking more suitable courses. This explanation
Table 2—2nd Year Performance Differential of students who collected 42 vs. students who collected 60 credits
1. All Students 2. Economics 3. Business 4. Accountancy
or Finance and Control
1. Year2 -4.143⇤⇤⇤ -4.807⇤⇤⇤ -4.645⇤⇤⇤ -2.488
(0.654) (0.646) (1.599) (1.764)
2. 42 credits -18.00⇤⇤⇤ -18.00⇤⇤⇤ -18.00 -18.00⇤⇤⇤
(2.79e-15) (3.97e-15) (0) (1.94e-15)
3. Year2⇥ 5.705⇤⇤⇤ 8.699⇤⇤⇤ 9.939⇤⇤ -3.828
42 credits (2.060) (2.514) (4.348) (4.506)
4. Intercept 60.00⇤⇤⇤ 60.00⇤⇤⇤ 60.00⇤⇤⇤ 60.00⇤⇤⇤
(5.44e-16) (8.60e-16) (4.44e-16) (5.44e-16)
N 734 436 96 202
adj. R2 0.338 0.386 0.365 0.316
Robust standard errors in parentheses
⇤p < 0.10,⇤⇤p < 0.05,⇤⇤⇤p < 0.01
is partially convincing, since the students who collected 60 credits in the first year and who choose the Business specialization tend to perform slightly worse, collecting 4.5 credits less, for an average total of about 55.5 credits. Perhaps, the reason for the slight improvement in performance of the students who collected 42 credits in the first year is that they are retaking first year courses. Nonetheless, the coefficients of the estimates for the Economics or Finance specialization are fairly close to those for the Business specialization: mere strategic course-taking is a wanting explanation when considering that the Economics and Finance spe-cializations are equally demanding in terms of mathematics and statistics in the second year of studies.
All in all, it is difficult to indicate an overarching explanation for the results. The additional results (reported in the appendix) comparing the students who collected 48 or 54 credits against the students who collected 60 credits indicate an inconclusive pattern: for instance, students in the Economics or Finance and Business specializations who collected 48 credits in the second year collect again 48 credits (see Table A2: employing the students who collected 60 credits as ref-erence group the di↵ref-erence-in-di↵ref-erences estimator is again positive ), while those in the Accountancy and Control specialization collect only 38 credits. Students who collected 54 credits in the first year also undergo a dip in performance, col-lecting approximately 50 credits in the second year (see Table A3: the coefficient of the di↵erence-in-di↵erences estimator is insignificant). Performance as
mea-sured in terms of GPA mirrors performance as meamea-sured in terms of credits, as estimated with the second regression (also reported in the appendix, in Tables A4 to A6): there is a small, but significant dip in performance in terms of GPA for students who collected 60 credits in the first year if in the Economics or Finance specialization, while the coefficient is insignificant in the case of Business and Ac-countancy and Control students. In contrast, students who collected 42 credits tend to perform slightly better, although still at a lower level of performance than the students who collected 60 credits.
Overall, the results indicate a dip in performance in the group of the students who collected 60 credits in the first year and an improvement in the group of the students who collected 42 credits. Considering also the groups of students who collected a di↵erent number of credits, there seems to be heterogeneous patterns that this simple design, claiming perhaps a bit naively that students who obtained 42 or 60 credits in the first year follow two oversimplified behavioural rules, is unable to capture. The question is then: is the result imputable to the BSA policy? It is possible that a substantial group of the worse performers in the first year were actually motivated by the BSA to perform better in the second year, but most improvement may stem from 18 credits gained with retaking the easier first year courses. In contrast, some top performers in the first year might have been surprised by more difficult courses in the second year. A di↵erent possibility is that the lower average obtained is actually due to some students obtaining 60 credits in the first year because they were motivated from the BSA in the first year and consequently performing worse when the BSA requirement is absent. The next section extends the investigation, attempting to corroborate this last hypothesis.
B. Tests of the Validity of the DiD design10
The previous sections yielded unclear results. Because the identifying assump-tion in the design is likely wrong, it is not shown yet that there is no dip in performance in the second year. As a matter of fact, the descriptive results (re-ported in Table A1 in the appendix) showed that in the second year the variance in credits obtained more than doubles, as compared to the first year, suggesting that there is a dip in performance across a subset of students. At the end of the previous section, I conjectured that the BSA policy might have motivated a sub-set of students, pushing them to collect up to 60 credits. A dip in performance, starting with the second year, might trigger a delay in their studies, eventually leading not to graduate within the official time. A di↵erent approach is then, instead of taking the ex ante perspective of who were the best performers at the end of the first year, as implemented in the previous section, to take an ex post perspective, analysing who revealed to be the top performers at the end of the three years of bachelor, completing the bachelor on time.
Table 3—Characteristics of the groups of students bundled according to time for comple-tion
Variable Mean Std. Dev. Min. Max.
Did not complete bachelor in 3 years
High School GPA 6.721 0.489 5.79 9
GPA year1 5.940594 0.832 3.89 8.86
Average Credits year1 52.013 6.826 42 72
42 Credits 0.2 0.401 48 Credits 0.253 0.435 54 Credits 0.234 0.424 60 Credits 0.306 0.462 GPA year2 5.808 0.955 3.33 9.35 Credits year2 44.194 15.241 6 78 Male .741 .439 Maximum BSA .313 .464
Completed bachelor in 3 years
High School GPA 6.900 .523 5.88 8.64
GPA year1 6.739 0.880 4.48 9.03
Average Credits year1 57.251 4.947 42 66
42 Credits 0.033 0.178 48 Credits 0.113 0.317 54 Credits 0.138 0.346 60 Credits 0.713 0.453 GPA year2 6.885 0.841 3.75 9.45 Credits year2 59.433 9.463 12 162 Male 0.636 0.482 Maximum BSA 0.716 0.452
Note: The sample of students who completed the bachelor in 3 years consists of 320 students. The high school GPA data is available for 309 students out of 320.
The sample of students who did not complete the bachelor in 3 years consists of 275 students. The high school GPA data is available for 270 students out of 275.
The characteristics of the groups of students who did and did not complete the bachelor programme on time are illustrated in Table 4. At the moment of intake, the average GPA across the two groups is not significantly di↵erent. Nonetheless, only 46.2 percent of all students manage to complete the bachelor in three years. The overwhelming majority of students who finished the bachelor programme on time includes students who collected 60 credits in the first year. However, 30.6 percent of students do not manage to complete the programme in three years collected 60 credits in the first year. They might have relaxed exactly in the second year: if they did so, it is plausible, although not statistically cogent, that they were responsive in the first year to the BSA threshold as well. In contrast, 28.4 percent of the students who complete the bachelor within three years under-perform in the first year, presumably catching up later11. Within this group, only 11.6 percent collected 42 credits in the first year, amounting to 3.3 percent of all successful students. The BSA policy might have had a motivating e↵ect on them, pushing them to pass the threshold in the first year and to put more e↵ort in the subsequent years.
The same methodology illustrated under the empirical strategy is used, chang-ing however the group of comparison: the performance of students who completed the bachelor in three years is employed as a benchmark to measure the divergence in performance of the students who took four or more years (or never completed it). The following regression is run for all students and across specialization:
Credits = Y ear2+ Incomplete Bachelor
+ Y ear2⇥ Incomplete Bachelor + Intercept + "
(3)
The di↵erence in specification is that Incomplete Bachelor is a dummy equal to 1 if the students did not complete the bachelor after three years of studies and 0 if she did. Y ear2⇥ Incomplete Bachelor is then the di↵erence in performance
across the two groups and across the two years. With this specification, the Intercept is the first-year performance of students who completed their bachelor in the official time of three years. The regression is run first across all students, to determine if there is a dip, in general, between the first two years. Then, it is run once again only for students who collected 60 credits in the first year, focusing on the subset of students that were used as counterfactual in the previous design. This approach will shed light on the validity of the counterfactual employed. Nonetheless, it neglects that much could have happened in the third year as well, a↵ecting the performance of students. The results are illustrated in Table 4 and 5 below.
The results of the regression estimation indicate a dip in performance in the second year: the coefficient of the di↵erence-in-di↵erences estimator Y ear2 ⇥ gratefully acknowledged.
11The percentage of students who collect di↵erent numbers of credits does not add up to one in Table
Table 4—2nd year performance differential of students who complete and do not complete bachelor (Bsc) in three years
1. All Students 2. Economics 3. Business 4. Accountancy
or Finance and Control
1. Year2 2.182⇤⇤⇤ 1.840⇤⇤⇤ 2.400 2.720⇤ (0.644) (0.691) (1.520) (1.617) 2. Incomplete Bsc -5.238⇤⇤⇤ -4.905⇤⇤⇤ -6.536⇤⇤⇤ -5.769⇤⇤⇤ (0.484) (0.631) (1.300) (0.884) 3. Year2⇥ -10.00⇤⇤⇤ -9.071⇤⇤⇤ -6.937⇤⇤ -13.51⇤⇤⇤ Incomplete Bsc (1.134) (1.297) (2.897) (2.685) 4. Intercept 57.25⇤⇤⇤ 57.52⇤⇤⇤ 55.56⇤⇤⇤ 57.84⇤⇤⇤ (0.298) (0.402) (0.801) (0.492) N 1190 690 182 318 adj. R2 0.255 0.264 0.232 0.267
Robust standard errors in parentheses
⇤p < 0.10,⇤⇤p < 0.05,⇤⇤⇤p < 0.01
Incomplete Bachelor is negative. The coefficient of Y ear2 imply that students
who complete the bachelor in three years tend to perform slightly better in the second year, collecting about 2 credits more than in the first year. Students who have not completed the bachelor in the official time, however, already have a disadvantaged start of about 5 credits less in the first year, for an average total of 52 credits. Further, the coefficient of the di↵erence-in-di↵erences estimator indicates that, in the second year, they collect 8 credits less as compared to the first year and 15 credits less than the students who manage to complete the bachelor in three years. When considering only the students who collected 60 credits in the first year, the coefficient of the di↵erence-in-di↵erences estimator is similar, as shown in Table 5. The specification however implies that they collect 10.5 credits less than in the first year and 9.5 credits less than the students who collected 60 credits in the first year and manage to complete the bachelor in three years.
With this approach, a negative coefficient for the di↵erence-in-di↵erences esti-mator is found, in line with the initial hypothesis: there is a dip in the second year, that e↵ectively delays graduation. The prior result found seems definitely due to a larger percentage of students who collected 60 credits performing sub-stantially worse, rather than the small percentage of students who collected 42 credits catching up with the top class. To be crystal clear, this does not constitute a randomized control, and “something else” other than the change in BSA policy
Table 5—2nd year performance differential of students who collected 60 credits in the first year divided across completion of the bachelor (Bsc) in three years
1. All Students 2. Economics 3. Business 4. Accountancy
or Finance and Control
1. Year2 -0.980 -1.381⇤⇤⇤ -4.000⇤⇤ 1.286
(0.656) (0.414) (1.688) (1.963)
2. Incomplete Bsc -6.22e-15⇤⇤⇤ 1.54e-15⇤⇤⇤ 4.04e-15⇤⇤ -9.54e-15
(9.93e-16) (5.44e-16) (1.99e-15) (0)
3. Year2⇥ -9.490⇤⇤⇤ -9.119⇤⇤⇤ -5.000 -11.90⇤⇤⇤
Incomplete Bsc (1.394) (1.376) (4.949) (3.643)
4. Intercept 60⇤⇤⇤ 60⇤⇤⇤ 60⇤⇤⇤ 60.00⇤⇤⇤
(3.14e-16) (4.44e-16) (1.11e-16) (6.28e-16)
N 588 362 62 164
adj. R2 0.210 0.354 0.113 0.117
Robust standard errors in parentheses
⇤p < 0.10,⇤⇤p < 0.05,⇤⇤⇤p < 0.01
could have happened in the second year. For instance, some students who collect 60 credits might decide, after a tremendous year, to start o↵ a new bachelor or to set up a business, on which they now focus most of their attention; this cannot be ascertained from the data-set. Still, school regulations do not indicate any relevant change across any of the academic years under consideration. Changes in courses difficulty does not appear to be a determinant factor either: 28.4 per-cent of students who finished within three years underperformed in the first year and the students who collected 60 credits in the first year and graduate on time obtain only 1 credits less, on average, in the second year.
C. A possible gender e↵ect
The previous research has found some evidence of gender di↵erences in respon-siveness to the BSA: Arnold (2015) and Sneyers and de Witte (2015) illustrate that the selectivity of the BSA hit males more severely, while Lindo et al. (2010) show that males undergoing academic probation experience a larger drop in per-formance during the next years of university. If the BSA policy a↵ects di↵erently males and females, then a similar di↵erential in performance across gender should be found. Table 4 illustrates indeed that the larger percentage of students who complete the bachelor in three years are female (63.6 percent of students who complete the bachelor in three years are male, but a larger percentage, 74.1 per-cent, of the students who need four or more years are male). This possibility
is investigated by running a further regression, using the same design illustrated under section B. but bundling students in groups per gender.
Table 6—2nd year performance differential of students who complete and do not complete bachelor in three years across gender
1. All students 2. Male 3. Female
1. Year2 2.182⇤⇤⇤ 2.674⇤⇤⇤ 1.320⇤ (0.644) (0.918) (0.740) 2. Incomplete Bsc -5.238⇤⇤⇤ -5.316⇤⇤⇤ -4.417⇤⇤⇤ (0.484) (0.604) (0.824) 3. Year2 ⇥ -10.00⇤⇤⇤ -11.76⇤⇤⇤ -5.513⇤⇤⇤ Incomplete Bsc (1.134) (1.437) (1.745) 4. Intercept 57.25⇤⇤⇤ 56.71⇤⇤⇤ 58.20⇤⇤⇤ (0.298) (0.414) (0.367) N 1190 824 366 adj. R2 0.255 0.272 0.190
Robust standard errors in parentheses
⇤p < 0.10,⇤⇤p < 0.05,⇤⇤⇤p < 0.01
The results in Table 6 report that males who do not manage to complete the bachelor in three years perform slightly worse in the first year than females who do not manage to complete the bachelor in three years, as indicated from the sum of the coefficients of Intercept and IncompleteBachelor; the same males then tend to perform even worse in the next year, obtaining almost 7 credits less than the same females, as indicated from the di↵erence of the sum of the coefficients of all coefficients in the two separate regression. In contrast, males and females who complete in 3 years perform almost equally in the first year (there is a di↵erence of 1.5 credits) as well as in the second year (there is a di↵erence of only 0.1 credits). Something happens in the second year that impairs more the performance of some male students than that of female students. This finding would concur with the prior literature if the dip in performance is mainly the e↵ect of the absence of a BSA policy. This claim should be taken with a pinch of salt as it has been already emphasized in the previous section that other factors may play a considerable role. Further, the prior literature has been concurring, but not conclusive, about the existence of a gender e↵ect: for instance, Fletcher and Tokmouline (2010) find di↵erent responses to academic probation based on heterogeneous students’ characteristics, but they are not able to trace a consistent gender di↵erence in responsiveness to academic probation across various Texan universities.
VI. Conclusion
This thesis attempted to tackle the topic of the e↵ect of the BSA policy on students’ motivation. Initially, it was found that students performing worse in the first year and presumably more responsive to the BSA policy improve in the next year as compared to the students who performed better in the first year. A deeper analysis of the data revealed that the pattern was mostly due to a larger percentage of students collecting 60 credits in the first year and being unable to maintain an equally high level of achievements in the next year, substantially worsening in performance. It appears that students unable to graduate within the official time of completion of the bachelor substantially worsen in performance in the second year. There appears to be a dip in performance in the second year, regardless of first-year performance: on average, students who do not graduate on time collect in the second year 8 credits less than in the first year, but those who collected 60 credits in the first year collect 10.5 credits less than in the first year. The BSA policy might have a temporary motivating e↵ect on a subset of students, including some students who are highly efficient in the first year. This is in line with the previous findings of Lindo et al. (2010) and Fletcher and Tokmouline (2010) that an academic dismissal policy only has a temporary e↵ect on improving performance.
Still, for an even smaller subset of students the BSA policy might have had a positive feedback, pushing them to work harder even later: indeed, 3.3 percent of the students who complete the bachelor programme on time collected 42 credits in the first year, catching up in the two successive years. This finding suggests that there are highly heterogeneous groups not captured in the design. Finally, males tend to worsen more as compared to females which is, if due to the BSA policy, also in line with the previous findings. Although the results pointed to the fact that the BSA policy has a short term e↵ect, they do not constitute a randomized experiment: there is too much heterogeneity in responsiveness to the BSA policy in the sample and a proper counterfactual was not identified, such that a simple di↵erence-in-di↵erences comparison gives some evidence, but no conclusive verification.
These findings do not necessarily indicate that the BSA a deleterious policy. There is indeed sufficient evidence that the BSA beneficially increases selectivity. The criticism is however that it is possible that some students perform worse in the second year and, as performance is often cumulative (both in terms of prepa-ration and encouragement), perform even worse in the successive years, eventually remaining “trapped” in their programme, unable to graduate and e↵ectively wast-ing three or more years pursuwast-ing a degree that they are unable to complete. This might happen regardless of first-year performance: remarkably, in the group of students who do not manage to graduate in three years the students who collect 60 credits in the first year collect undergo a larger dip in performance than the mean. The e↵ect on graduation of this dip remains hypothetical and cannot be tested, since the data employed covers at most 4 years. It is symptomatic, though,
that in the 2011 cohort, after four years of studies, 24.8 percent of the students who passed the first year have not yet completed their programme. Yet, it is possible that this subset of students might have dropped out anyways in the ab-sence of a BSA policy. Further research should tackle this issue: it is important, for purposes of academic policy, to investigate whether this sort of requirements should be maintained across multiple years to ensure timely graduation.
Concerning the design of this study, there has been a pervasive assumption throughout the thesis: that there is a one-to-one link between the BSA policy and students motivation. The essential idea, as recognized by the Education Inspectorate (2010), is that it is a “stick”, thus predicting motivation. This has actually not been tested: it is possible that it may just set a pace for students’ progress. Before any conclusion can be drawn, further research should tackle this nuance of the BSA.
Finally, there were obvious limitations in the data-set that further research should fill in. First, the data employed was qualitatively limited only to students’ results and within a time frame of four years. However, students’ experiences at university often involves multiple dimensions, beyond mere results, which should be integrated in a thorough analysis: what if the students who collected 60 credits in the first year set up a business in the second year? Then, the results would be misleading. Second, the data concerned only Economics and Business students only from a Dutch university. Past research has already studied Psychology and Medicine students, but there are other faculties left to be investigated. Arnold (2015) has indeed shown that are contrasting di↵erences across bachelor pro-grams. This study has added that there is even substantial heterogeneity, when considering the BSA policy, across students within the same program. All this needs to be tackled in order to ensure a well micro-founded policy.
References
Arnold, I. J. M. (2015). The e↵ectiveness of academic dismissal policies in dutch university education: an empirical investigation. Studies in Higher Educa-tion, 40 (6), 1068-1084.
Benabou, R., & Tirole, J. (2003). Intrinsic and extrinsic motivation. Review of Economic Studies, 70 (3), 489-520.
Education Inspectorate. (2009). Uitval en rendement in het hoger onderwijs. achtergrondrapport bij werken aan een beter rendement. [dropout and suc-cess rates in higher education. background report on working toward a better efficiency].
Education Inspectorate. (2010). Met beide benen op de grond. onderzoek naar de uitvoering- spraktijk van het bindend studieadvies in het hoger onderwijs. [with both feet on the ground. research into the implementation of the binding study advice in higher education].
Cameron, J., & Pierce, W. D. (1994). Reinforcement, reward, and intrinsic motivation: A meta-analysis. Review of Educational research, 64 (3), 363-423.
Casey, M., Cline, J., Ost, B., & Qureshi, J. (2015). Academic probation, student performance and strategic course taking.
Deci, E. L. (1975). Intrinsic motivation. New York: Plenum Press.
Deci, E. L., Koestner, R., & Ryan, R. M. (1999). A meta-analytic review of ex-periments examining the e↵ects of extrinsic rewards on intrinsic motivation. Psychological bulletin, 125 (6), 627-668.
de Koning, B. B., Loyens, S. M. M., Rikers, R. M. J. P., Smeets, G., & van der Molen, H. T. (2014). Impact of binding study advice on study behavior and pre-university education qualification factors in a problem-based psychology bachelor program. Studies in Higher Education, 39 (5), 835-847.
Eijsvogels, T. M. H., Goorden, R., van den Bosch, W., & Hopman, M. T. E. (2015). The binding study advice in medical education: a 2-year experience. Perspectives on Medical Education, 4 , 39-42.
Fletcher, J. M., & Tokmouline, M. (2010). The e↵ects of academic probation on college success: Lending students a hand or kicking them while they are down?
Gibbons, R., & Roberts, J. (2013). The handbook of organizational economics. New Jersey: Princeton University Press.
Lindo, J. M., Sanders, N. J., & Oreopoulos, P. (2010). Ability, gender, and perfor-mance standards: Evidence from academic probation. American Economic Journal: Applied Economics, 2 (2), 95-117.
Murtaugh, P., Burns, L., & Schuster, J. (1999). Predicting the retention of medical students. Research in Higher Education, 40 (3), 355-371.
Peltier, G. L., Laden, R., & Matranga, M. (2000). Student persistence in college: A review of research. Journal of College Student Retention, 1 , 357-376. Schneider, M. (2010). Finishing the first lap: The cost of first year student
attri-tion in americas four year colleges and universities (Tech. Rep.). American Institutes for Research.
Sneyers, E., & de Witte, K. (2015). The e↵ect of an academic dismissal policy on dropout, graduation rates and student satisfaction. evidence from the netherlands. Studies in Higher Education, 1 (1), 1-36.
Stegers-Jager, K. M., Cohen-Schotanus, J., Splinter, T. A. W., & Themmen, A. P. N. (2011). Academic dismissal policy for medical students: E↵ect on study progress and help- seeking behaviour. Medical Education, 45 (10), 987-994.
Stock, J. H., & Watson, M. W. (2007). Introduction to econometrics. Pearson/Addison-Wesley.
Table A1—Characteristics of the students in the 2011 and 2012 cohorts
Variable Mean Std. Dev. Min. Max.
2011 cohort
Male 0.686 0.465
High school GPA 6.847 0.524 5.88 9
Collected 60 Credits 0.529 0.5
Credits year1 54.745 6.584 42 72
GPA year1 6.266 1.039 3.89 9.029
Credits year2 50.765 14.487 6 78
GPA year2 6.288 1.101 3.33 9.35
Accountancy and Control 0.297 0.458
Business Administration 0.186 0.39 Economics or Finance 0.516 0.501 Completed in 3 years .471 Completed in 4 years .281 Completed in 3 or 4 years 0.752 0.433 2012 cohort Male 0.699 0.46
High school GPA 6.759 0.497 5.79 8.33
Collected 60 Credits 0.467 0.5
Credits year1 54.104 6.547 42 66
GPA year1 6.356 0.827 4.5 8.85
Credits year2 51.737 15.459 6 162
GPA year2 6.325 0.996 3.74 9.450
Accountancy and Control 0.235 0.425
Business Administration 0.118 0.323
Economics or Finance 0.647 0.479
Completed in 3 years 0.453 0.499
Note: The sample for the 2011 cohort consists of 306 students who officially continued their studies after the first year. Graduation rate is calculated for 3 and 4 years (official completion time is 3 years). The high school GPA data is available for 298 students out of 306. The sample for the 2012 cohort consists of 289 students who officially continued their studies after the first year. Graduation rate is calculated for 3 years. The high school GPA data is available for 281 students out of 289.
A1. Performance di↵erentials in credits
Table A2—2nd Year Performance Differential of students who collected 48 vs. students who collected 60 credits
1. All students 2. Economics 3.Business 4.Accountancy
or Finance and Control
1. Year2 -4.143⇤⇤⇤ -4.807⇤⇤⇤ -4.645⇤⇤⇤ -2.488
(0.654) (0.646) (1.597) (1.764)
2. 48 Credits -12.00⇤⇤⇤ -12.00⇤⇤⇤ -12.00⇤⇤⇤ -12.00⇤⇤⇤
(3.11e-15) (5.37e-15) (2.51e-15) (5.22e-15)
3. Year2⇥ 1.839 4.372⇤⇤ 4.645 -7.425⇤
48 Credits (1.823) (2.215) (4.097) (4.264)
4. Intercept 60.00⇤⇤⇤ 60.00⇤⇤⇤ 60⇤⇤⇤ 60⇤⇤⇤
(8.88e-16) (1.83e-15) (8.88e-16) (1.17e-15)
N 812 500 102 210
adj. R2 0.235 0.248 0.227 0.255
Robust standard errors in parentheses
Table A3—2nd Year Performance Differential of students who collected 54 vs. students who collected 60 credits
1. All students 2. Economics 3.Business 4.Accountancy
or Finance and Control
1. Year2 -4.143⇤⇤⇤ -4.807⇤⇤⇤ -4.645⇤⇤⇤ -2.488
(0.654) (0.646) (1.596) (1.762)
2. 54 Credits -6.000⇤⇤⇤ -6.000⇤⇤⇤ -6.000⇤⇤⇤ -6.000⇤⇤⇤
(1.40e-15) (3.01e-15) (6.28e-16) (6.59e-15)
3. Year2 ⇥ 0.638 -0.122 6.009⇤⇤ -1.798
54 Credits (1.489) (1.914) (2.620) (3.375)
4. Intercept 60.00⇤⇤⇤ 60⇤⇤⇤ 60⇤⇤⇤ 60.00⇤⇤⇤
(4.44e-16) (6.28e-16) (6.28e-16) (2.24e-15)
N 814 474 106 234
adj. R2 0.122 0.195 0.092 0.075
Robust standard errors in parentheses
A2. Performance di↵erentials in GPA
Table A4—2nd Year Performance Differential in GPA of students who collected 42 vs. students who collected 60 credits
1. All students 2. Economics 3.Business 4.Accountancy
or Finance and Control
1. Year2 -0.145⇤⇤ -0.240⇤⇤⇤ 0.164 -0.0515 (0.0697) (0.0887) (0.191) (0.137) 2. 42 Credits -1.820⇤⇤⇤ -1.867⇤⇤⇤ -1.480⇤⇤⇤ -1.860⇤⇤⇤ (0.0749) (0.110) (0.173) (0.126) 3. Year2 ⇥ 0.555⇤⇤⇤ 0.680⇤⇤⇤ 0.472 0.199 42 Credits (0.122) (0.169) (0.286) (0.210) 4. Intercept 6.982⇤⇤⇤ 7.030⇤⇤⇤ 6.670⇤⇤⇤ 6.994⇤⇤⇤ (0.0412) (0.0516) (0.114) (0.0822) N 734 436 96 202 adj. R2 0.377 0.345 0.433 0.411
Robust standard errors in parentheses
Table A5—2nd Year Performance Differential in GPA of students who collected 48 vs. students who collected 60 credits
1. All students 2. Economics 3.Business 4.Accountancy
or Finance and Control
1. Year2 -0.145⇤⇤ -0.240⇤⇤⇤ 0.164 -0.0515 (0.0696) (0.0886) (0.191) (0.137) 2. 48 Credits -1.415⇤⇤⇤ -1.380⇤⇤⇤ -1.379⇤⇤⇤ -1.438⇤⇤⇤ (0.0630) (0.0789) (0.153) (0.131) 3. Year2 ⇥ 0.260⇤⇤ 0.369⇤⇤ 0.358 -0.226 48 Credits (0.121) (0.157) (0.284) (0.234) 4. Intercept 6.982⇤⇤⇤ 7.030⇤⇤⇤ 6.670⇤⇤⇤ 6.994⇤⇤⇤ (0.0412) (0.0515) (0.114) (0.0822) N 812 500 102 210 adj. R2 0.333 0.304 0.413 0.371
Robust standard errors in parentheses
Table A6—2nd Year Performance Differential in GPA of students who collected 54 vs. students who collected 60 credits
1. All students 2. Economics 3.Business 4.Accountancy
or Finance and Control
1. Year2 -0.145⇤⇤ -0.240⇤⇤⇤ 0.164 -0.0515 (0.0696) (0.0886) (0.191) (0.137) 2. 54 Credits -0.946⇤⇤⇤ -0.950⇤⇤⇤ -0.702⇤⇤⇤ -0.987⇤⇤⇤ (0.0649) (0.0882) (0.147) (0.128) 3. Year2 ⇥ 0.150 0.177 0.243 -0.0865 54 Credits (0.116) (0.158) (0.269) (0.218) 4. Intercept 6.982⇤⇤⇤ 7.030⇤⇤⇤ 6.670⇤⇤⇤ 6.994⇤⇤⇤ (0.0412) (0.0515) (0.114) (0.0821) N 814 474 106 234 adj. R2 0.192 0.177 0.154 0.239
Robust errors in parentheses
