Data and model specification

18 D. Rochat, J.-L. Demeulemeester Economics of Education Review 20 2001 15–26 Table 1 Definition of orientations Orientation 1 Short cycle 2 to 3 years in Economic and Social Sciences Orientation 2 Short cycle 2 to 3 years in Paramedical studies Orientation 3 Short cycle 2 to 3 years in Artistic and Pedagogical studies Orientation 4 Long cycle curricula 4 years or university degree 4 to 7 years in Natural and Medical Sciences Orientation 5 Long cycle curricula 4 years or university degree 5 years in Engineering Orientation 6 Long cycle curricula 4 years or university degree 5 years in Business, Economics and Social Sciences Orientation 7 Long cycle curricula 4 years or university degree 4 to 5 years in Humanities and Psychology the highest probability of success, given his socio-econ- omic background. This methodology has so far only been applied by Cannings et al. 1993 to a quite close topic namely the major choices in undergraduate concentrations.

3. Data and model specification

Our microdata sample consists in 641 freshmen 7 enrolled at Belgian French-speaking higher education institutions either in 1992 or 1993 including univer- sities, long and short cycle non-university higher edu- cation institutions. This data set comes from the PSBH– CREPP 1993–95 survey of the French Community of Belgium. Of those 641 students, 220 are enrolled in short cycle curricula and 421 enrolled in long cycle and uni- versity curricula. We consider the students enrolling in 1992 and 1993 as belonging to a common sample. 8 We classified the students in seven orientations three for the short-cycle curricula and four for the long-cycle curric- ula and university orientations on the basis of the insti- tutional peculiarities of the Belgian system and of the characteristics of the orientations as well as the logical concomitant academic requirements. The list and defi- nition of these orientations are presented in Table 1. Our students sample has the following characteristics: it is made up of 48.36 of men, 45.55 of students aged more than 18 when entering higher education, 9 88.3 of Belgian students. 26.70 of the students benefit either from a tuition fees reduction or of a scholarship, 43.7 come from households with net monthly revenues of 100,000 Belgian francs or more, 55.9 have father hold- 7 i.e. students who began their higher education studies. 8 Their number in some specific orientations are indeed too small to allow for specific estimation on each year separately. Given the fact that exogenous factors do not change a lot from one year to another, we think that this choice does not incur a great cost. 9 31.4 of all students have repeated at least one year while in high school and 34.8 of all students followed other higher education curricula prior their entry in their studies only one quarter of them completed them. ing higher education degree and among them, 52.5 university degree, 51.1 have mother holding higher education degree and among them only 25.9 univer- sity degree, 19.3 of the students have fathers holding “e´lite” occupation see below, 47.4 of students come from households where both parents work, and 23.6 come from separate couples divorce or other atypical situations dead parent, for example. Finally, 18.9 of the students work while studying and 34.5 live on their own. The average success rate is 65.91 in short-cycle curricula and 48 in long cycle curricula and univer- sity orientations. As far as model specification is concerned, we present the set of socio-demographic and ascriptive explanatory variables used in the first two stages of the analysis in Table 2. The assignment procedure of the variables between the multinomial equation explaining choice of orientation and binary probit equation explaining aca- demic success relies upon careful examination of exist- ing literature on the topic see Duru-Bellat and Mingat, 1993, for a survey on the individual and contextual deter- minants of orientation choices and Haveman and Wolfe, 1995 for a survey on students’ attainments as well as compliance to identification requirements. This led us to retain some common factors in the two estimation steps on a priori grounds age, gender, nationality, and vari- ables related to parental education and occupation as well as some specific explanatory variables to each topic investigated see list of variables in Table 2 below. The results of the third step are however quite invariant to minor specification changes made at these two pre- vious steps. In the last step of the estimations conditional logit, we included three more variables besides the estimated probability of success. The first one characterizes the entrance salary by discipline. 10 These data are available for civil servants with short cycle higher education degree and with four years of university education in 10 Berger 1988 asserts that the variable to consider is the expected life-cycle earnings stream rather than the entry-level salary. However, we believe that the latter correctly reflects the future hierarchy of wages. 19 D. Rochat, J.-L. Demeulemeester Economics of Education Review 20 2001 15–26 Table 2 List of explanatory variables and assignment of dummy variables MNL Binary estimations probit Variable Dummy assignment for estimation orientation for academic choice success Gender 1 if male, 0 otherwise X X Age 1 if 19 or more when entering X X Nationality 1 if Belgian, 0 otherwise X X Latin 1 if Latin while in high school X Mathematics 1 if 6 hweek math or more X Single parent family 1 if so, 0 otherwise X X Father’s education 1 if university probit or higher education MNL, 0 otherwise X X Mother’s education 1 if university probit or higher education MNL, 0 otherwise X X 1 if father holds an “e´lite” occupation, i.e. top manager or civil Father with “Elite” occupation X X servant, or professional Both parents work 1 if both parents work, 0 otherwise X X 1 if net monthly household’s income 100,000 Belgian francs, 0 Household’s income X otherwise Repetition during high school 1 if at least one year repeated while in high school, 0 otherwise X Prior studies 1 if prior studies not necessarily completed, 0 otherwise X Number of siblings number of siblings X Change in living arrangements 1 if leaving parental home, 0 otherwise X Scholarship 1 if tuition fees reductions or holding a scholarship, 0 otherwise X Job while studying 1 if working while studying, 0 otherwise X More aged due to prior studies 1 if yes, 0 otherwise X More aged due to repetitions 1 if yes, 0 otherwise X humanities and natural sciences Source: Moniteur Belge. They are also available for employees of the private sectors graduated from university: Business Schools Source: Union des Inge´nieurs Commerciaux Solvay, Engineering Source: FABI, Fe´de´ration Royale d’Associations Belges d’inge´nieurs Civils and various Professions Physicians and Lawyers; Source: MAKLU, Maastricht. 11 We expect that students will be drawn to more remunerative orientations, everything else held 11 The relevance of such data deserves some comments. On the one hand, the knowledge of students about their future earn- ings might be quite partial when entering the university. Betts 1995 showed indeed that students learn about the labor market over time and that their knowledge is basically limited to their field of study. On the other hand, even if students may hold a considerable degree of information about the labour market, the extent of its influence in structuring students preferences is at least open to question. For example, in a survey among US students, only a small proportion 16 in his survey of them consider money as a very important factor in their choice of discipline Freeman, 1989. Following this idea, one could argue that expected economic benefits mostly explain marginal changes among discipline choices from one year to another. Finally, even if students are influenced by labour market infor- mations, they may misuse this flow of informations see Man- ski, 1995. constant. It is true that some authors have pointed out that initial working conditions as initial earnings were less important than the lifetime expected conditions in explaining discipline choices see Berger, 1988, for the importance of the relative present value of the predicted future earnings by subject. However, precise measure of long-term perspectives or career opportunities as earnings stream depend critically on very specific assumptions earnings growth equation as in Willis and Rosen, 1979, 12 assumptions concerning the nature of expectations, etc.... As Oosterbeek and Webbink 1997, we do not include such forward-looking measures. This is mainly because we only have data on wages for groups of disciplines and not for each individuals having fol- lowed an orientation. And it would be quite risky to con- sider estimated lifetime earnings per discipline as good proxies given the very long run nature of educational 12 In their paper, Willis and Rosen 1979 model the choice of whether or not to attend college with a probit model. For those who went to college and for those who did not, separate earnings equations and earnings growth equations are estimated to impute the expected earnings gain from college as an explanatory variable in the college choice equation. They find that a larger expected earnings gain leads to a higher probability to attend college. 20 D. Rochat, J.-L. Demeulemeester Economics of Education Review 20 2001 15–26 investments. Proxying lifetime earnings prospects into the educational choice equation does not mean that our model is at odds with the human capital framework. Indeed age-earning profiles show that starting salaries at least partially reflect future earning differentials see Woodhall, 1987; Demeulemeester, 1995 for evidence on the Belgian labour market. We also introduced a measure of the easiness of inser- tion into the labour market for young graduates by orien- tation. 13 These statistics were taken from Demeulemees- ter and Rochat 1995. This measure might be seen as a complement to our measure of expected future benefits besides wage. Ceteris paribus, we expect that students will prefer orientations whose graduates are perceived to insert more easily on the labour market. Finally, we also introduced the legal minimum length of studies to get the final degree. This will allow us to take into account the risk linked with the increased length of studies, as well as the direct and indirect costs of studies. We expect the latter to be deterrent.

4. Empirical results