Project topics and materials
Get Complete Project Material File(s) Now! »
Competitive Schools and the Gender Gap in the Choice of Field of Study
Abstract
In most developed countries, students have to choose a major field of study during high school. This is an important decision as it largely determines subsequent educational and occupational choices. Using French data, this paper reveals that enrollment at a more selective high school, with higher-achieving peers, has no impact on boys, but a strong impact on girls’ choices: they turn away from scientific fields and settle for less competitive ones. Our results are not consistent with two commonly-advanced explanations for gender differences in field of study, namely disparities in prior academic preparation and in sensitivity to rank in class.
In most developed countries, male and female students still choose very different major fields of study during high school or during college. In French high schools for instance, male students are about 40% more likely than female students to specialize in science1. These gender differences have attracted considerable attention as they likely explain a significant part of labor market differentials across gender groups. The choice of science as a major field of study is typically associated with the best prospective outcomes, but female students are still dramatically underrepresented in this field.
A long-standing literature has explored the causes of the gender gap in the choice of field of study, with a specific emphasis on gender differences in ability, expectations or preferences. Several influential studies have also emphasized the role of teaching practices and teachers’ stereotypes2. In this paper we analyze the role of another potential determinant of students’ choices, namely the school environment in which they make their decisions. Specifically we look at whether the choice of field of study of girls and boys depends on the academic level of the schoolmates with whom they have to compete. In more selective schools, with higher-achieving peers, students may be induced to form new expectations about their chances of success in the different fields, which may eventually affect their choices.
It has long been recognized that enrollment at a more selective school, with higher-achieving peers, may affect students’ subsequent performance and graduation probability, even though empirical evidence are mixed (see Pop-Eleches and Urquiola, 2013 or Abdulkadiro˘glu, Angrist and Pathak, 2014 for example). Much less is known, however, on whether enrollment in selective schools affects students’ choice of field of study. One basic reason for this lack of evidence is that the choice of major field of study often occurs at the same time as the high school choice, so that it is logically very difficult to define the impact of the latter on the former. Another basic difficulty is that the choice of enrolling in a more selective school and the choice of field of study likely depend on the same explanatory factors, such as students’ willingness to compete. In such a context, where two decisions potentially share the same causes, it is typically very hard to evaluate the influence that they exert on each other, even when they do not take place at the same time. Students enrolled in selective schools tend to choose more demanding fields of study, but it does not follow that their choice is influenced by their school environment.
In this paper, we build on the features of the Parisian high school system to overcome these issues. One first feature of this system is that middle school students are assigned to high schools through a centralized process that gives priority to students with the best average grades in middle school. About half of high schools receive more applications than they have places to offer and enrollment at each of these schools is restricted to students whose middle school average grade is above a specific cut-off level computed by the system. The second basic feature of the system is that students do not have to choose their major field of study at the start of high school, but only after one year of familiarization (that is, grade 10). In this context, it is possible to isolate the impact of enrollment at an oversubscribed high school on subsequent choices of field of study by comparing students whose middle school achievements were either just above or just below the specific cut-off level of this more selective high school.
This Regression Discontinuity (RD) analysis first confirms that eligibility for enrollment in a more selective school is associated with a very significant increase in the ability level of high school peers. It also reveals that this increase in peer ability is even more significant in science than in humanities, so that enrollment in a more selective school is first and foremost associated with an increase in competition in science. Importantly these first stages effects on peer ability are very similar for boys and girls, consistent with the assumption that, in our set-up, there is no gender gap in the willingness to attend higher ranked schools.
By contrast, eligibility for admission into a more selective school, with higher-achieving peers, has very different effects on the choice of major field of study made by boys and girls one year later, at the end of grade 10. Specifically, it has no effect on boys, but induces a significant decrease in the probability that girls choose science and a symmetrical increase in the probability that they choose humanities. Eventually, enrollment at a more selective school is not followed by any significant change in students’ overall graduation probability, but by a significant decrease in the share of girls who graduate in science.
Generally speaking, our results are consistent with experimental findings showing that female students are more likely to turn away from competitive settings than their male counterparts (Croson and Gneezy, 2009; Niederle and Vesterlund, 2011). They are also consistent with Rask and Tiefenthaler (2008) or Goldin (2015), who suggest that female students tend to be more responsive to a decline in performances than males. Our paper contributes to showing the decisive impact of these gender differences on the choices made by girls and boys during high school. Because science is the field of study where competition increases the most in sought-after schools, gender differences in attitudes towards grades and competition appear to induce many female students to turn away from science. Our findings are also reminiscent of the literature on college major choice in the US and on the role played by students’ expectations in this choice (Stinebrickner and Stinebrickner, 2014; Wiswall and Zafar, 2015). In a related contribution, Arcidiacono, Aucejo and Hotz (2016) show that the probability that a minority student graduates in science may be much lower in more selective Californian universities than in less selective ones. Finally, our paper contributes to the literature on the effect of going to a more selective school, with higher-achieving peers (Abdulkadiro˘glu et al., 2017; Clark and Del Bono, 2016; Cullen, Jacob and Levitt, 2006; Jackson, 2013). Several recent papers build on a similar regression discontinuity design to provide evidence on the effect of selective schools on students’ performance in various institutional contexts (Abdulkadiro˘glu, Angrist and Pathak, 2014; Dobbie and Fryer Jr, 2014; Pop-Eleches and Urquiola, 2013). This literature finds mixed evidence on the impact of elite schools on student academic performance3. Because French students have to choose a major field of study at the end of their first year of high school, we are able to look not only at the impact on academic performance, but also on the choice of field of study. The effect of gaining admission to a more selective school appears to be much stronger on field of study than on academic performance. There is a small recent literature which documents similar findings about tracking within schools (He, 2016; Dougherty et al., 2017).
The paper is organized as follows. Section 2 describes the institutional context. Section 3 describes our administrative data sources while Section 4 provides basic graphical evidence on the impact of being eligible for admission into a selective school on students subsequent choices of field of study or graduation probabilities. Section 5 develops our Regression Discontinuity analysis. Section 6 explores the candidate mechanisms that may explain that more competitive school contexts induce female students to turn away from science.
Institutional Context
In this section, we provide information on how middle school students are assigned to high schools in Paris as well as on the exams that they have to take and the choices that they have to make during their high school years. In the following sections, the main research question will be whether the high school to which a middle school student is assigned affects her subsequent choices and performance on exams.
The Assignment of Middle School Students to High Schools
In France, middle school runs from grade 6 to grade 9. Students complete grade 9 the year they turn 15. The curriculum is the same in all middle schools and there is no streaming by ability. At the end of grade 9, students enter into high school, which runs from grade 10 to grade 12. This paper focuses on students who completed 9th grade in public middle schools in the education region of Paris, in either 2009 or 2010.
France is divided into thirty education regions and the education region of Paris represents about 3% of French middle school students. In this region, there are about 100 public middle schools and about 50 public high schools where middle school students can pursue general education courses. Middle school students are assigned to public high schools through a centralized process called Affelnet, which is completely gender-blind and which is described in detail in Hiller and Tercieux (2014) or in Fack, Grenet and He (2015). Students are first asked to list up to six choices of public high schools in descending order of preference4. Paris is divided into four geographical districts (West, East, North, and South), and there is a very strong incentive to apply to high schools in one’s district of residence since the system gives priority to home-district over out-of-district applications. Also, within each district, a priority is given to low-income students, namely the 20% students eligible to means-tested financial assistance. For the other students, the system ranks their applications according to the average of their 9th grade marks across all subjects and assigns them to as many seats as possible using a deferred acceptance algorithm (Roth, 1982) and a multi round process5. There are twelve subjects (Mathematics, Physics, Biology, Technology, French, History/Geography, Sport, two foreign languages, Art, Music and Discipline) and the marks used to compute the average score used by the system are first standardized at the region level. Standardization amounts to weighting each mark by the inverse of its standard deviation. These weights being revealed only ex post (that is, after all students have submitted their choices), the weighted average marks used by the system are ex ante impossible to precisely predict or manipulate.
In substance, the algorithm first assigns the students with the best 9th grade average marks to their preferred schools until one school starts being oversubscribed. This top school is then dropped from the application lists of the remaining (not yet assigned) students. These students are then re-ranked and the process is reiterated until another school starts being oversubscribed, and so on. At the end of this first round, there are no seats left in a fraction of schools (that is, the oversubscribed ones), whereas the other fraction is still undersubscribed. Similarly, a fraction of students are assigned to one of the schools of their list whereas the other fraction are still unassigned (that is, they applied for oversubscribed schools only). To further improve the assignment rate, each unassigned student is then asked to form new choices, namely to apply to at least one of the undersubscribed schools, and the process is reiterated. At the end of this second round, some students are still unassigned, and the education administration helps them find a seat in one of the remaining undersubscribed schools in an informal way. Undersubscribed schools are typically those that end up admitting a significant proportion of out-of-district students.
The key feature of this assignment process is that it is possible to define a minimum admission score for a large fraction of high schools, namely the oversubscribed ones. As discussed below, in years 2009 and 2010, about half of the public high schools of the region of Paris appear to be oversubscribed, with very significant discontinuities in the rate of enrollment of students at specific cut-off points of the distribution of 9th grade scores. This feature will make it possible to build on a regression discontinuity analysis to evaluate the effect of being admitted to these schools on subsequent educational outcomes. Specifically, we will focus on students whose 9th grade scores fall either just above or just below the cut-off point of an oversubscribed school and we will compare the outcomes of those just above with those just below. The vast majority of these students continue general education in high school, and the question will be whether enrollment at a higher ranked school affects either their major field of study in high school or the level of their academic performance6.
Generally speaking, our research strategy relies on the assumption that individual 9th grade scores and schools’ minimum admission scores cannot be manipulated and predicted. As discussed above, there is little scope for manipulation of individual scores. With respect to schools’ minimum admission scores, they depend on many factors that are no easier to predict than individual scores, such as the exact number of low-income students and the exact distribution of their choices across schools in each district. In this set up, there is again little scope for manipulation or prediction7. As discussed below, we find no evidence of discontinuities in the density of individual scores – or in students’ pre-assignment characteristics – at the cut-off.
Major Field of Study and High School Exit Exams
At the end of their first year of high school (grade 10), French students can either pursue general education or enter a technical or a vocational education program. Furthermore, those who pursue general education have to choose a major field of study. There are three possible fields: science (field « S »), economics and social sciences (field « E/S ») or languages and literature (field « L »). The number of students per school and field of study is not set once and for all. It can vary significantly from one year to another so as to meet the choices of students8. At the end of 10th grade, students are asked the field of study that they prefer and, eventually, the vast majority is allowed to pursue the track that they prefer. This is a key choice: each field of study corresponds to a very specific curriculum, specific high school examinations, and specific opportunities after high school.
For example, for those who choose to specialize in science, the scientific subjects represent 50% to 60% of compulsory courses in grade 11 and 12. By contrast, for those who choose to specialize in languages and literature, scientific subjects represent less than 5% of compulsory courses. With respect to post-secondary education, it is virtually impossible to enter an engineering school or a medical school after non-scientific studies in high school9.
Generally speaking, science is the most prestigious field of study. Students in the scientific track performed on average 60% of a standard deviation higher in middle school than students in the social sciences track and 75% of a standard deviation higher than those in the languages and literature track. In our working sample, about 31% of students choose the scientific track, 13% specialize in literature and languages, 22% specialize in social sciences whereas about 34% opt for a more technical or vocational program, or drop out from education.
The first year of high school (grade 10) is dedicated to exploring the different subjects and to choosing a major field of study. After this exploration year, students have very little leeway to change their major field of study. In Paris, only about 2.5% of students change their major field of study after 10th grade. The two last years of high school (11th and 12th grade) are dedicated to the preparation of the national high school exit exam, the baccalauréat, which is a prerequisite for entry into post-secondary education. Students have to take one exam per subject, and they obtain their diploma if their weighted average mark across subjects is 10/20 or more, where subjects taken and weights depend strongly on the major field of study. For instance, the weights of exams in scientific subjects represent about 50% of the total for students who choose these subjects as major field of study, whereas the weights of these subjects is only about 20% for those who choose social sciences and 5% for those who choose languages and literature. Most exams are taken at the end of 12th grade, except for exams in French (oral and written) which are taken at the end of 11th grade. Students whose weighted average mark across subjects is 12/20 or more graduate with honors. Graduation with honors is granted to about half of the students each year in each field.
Selective Undergraduate Programs
High school graduation is a prerequisite for admission into post-secondary education programs. About half of these programs are selective, and selection depends on the grades obtained in the two last years of high school as well as on students’ ranks within their class. The Classes Préparatoires aux Grandes Écoles (CPGE) are among the most prestigious such selective programs. These two-year programs prepare students to take the entry exams of the most prestigious graduate programs (so called Grandes Écoles). The last important question addressed in this paper is whether enrollment at a more selective high school at the end of grade 9 affects the subsequent probability of gaining access to CPGE programs at the end of grade 12.
There are different types of CPGE programs: some are specialized in science (they prepare for entry exams at engineering graduate schools), some are specialized in economics and business (they prepare for entry exams at business schools) and some are specialized in lit-erature and languages (they prepare for entry exams at top graduate programs in this field). Each year, in Paris, about 20% of students from high school general education programs gain access to a CPGE. The vast majority graduate from high school with honors. When a student applies to a CPGE program, her high school has to provide the average marks (as assessed by teachers) that the student obtained in each subject for each quarter during 11th and 12th grade, as well as the corresponding rank within their class. Hence students from more selec-tive high schools may benefit from the prestige of their schools, but may suffer from being less well ranked within their class.
Data and Methods
Data
In our empirical analysis, we use administrative data providing detailed information on students who finished middle school (9th grade) in either 2009 or 2010 in the education region of Paris. For each student, we know the high school to which she was assigned after 9th grade, the field of study chosen at the end of 10th grade as well as the field of study in which she graduated at the end of 12th grade. For each student, we also know whether she repeated a grade during high school, whether she dropped out from education before graduation and whether she got admitted into a CPGE program after high school.
With respect to students’ academic performance, we know the average marks given by teachers in each subject during 9th grade as well as results at the national exams taken in Mathematics and French at the end of 9th grade (externally set and marked). We also know students’ results at the national examination (baccalauréat, externally set and marked) taken at the end of high school (12th grade). As discussed previously, the score used to assign middle school students to high schools corresponds to an average of the average marks given by 9th grade teachers.
To construct this dataset, we used schools’ registration records as well as administrative records with exhaustive information on results at the end of middle school and at the end of high school national exams, for each year between 2009 and 2014. We were able to match these different data sources using students’ ID number.
Finally, we augmented our individual-level database with information coming from school-level administrative datasets, namely information on high school size as well as on the proportion of female teachers and on the distribution of teachers’ age.
Cut-off Scores
In this section, we consider the 52 public high schools which enter the centralized assignment system in 2009 or 2010. For each cohort and each district, we focus on the 9th grade students whose applications went through the standard assignment process, that is, they are not low-income and come from a public middle school of the district10. Our purpose is to identify the public high schools which received more applications than they could accommodate, and to estimate the lowest 9th grade score that students had to earn to gain admission into these schools.
Our data do not provide information on students’ rank order lists, so that it is not possible to identify which schools are oversubscribed (and their minimum admission scores) directly from the data. To overcome this issue, we built on a method developed by Hansen (2000). Among all possible minimum admission scores, the method first identifies for each school those which coincide with a significant discontinuity in enrollment rates. If any such thresholds exist, the method amounts to choosing for each school the specific threshold which corresponds to the most significant discontinuity. More details on how we implemented this procedure are provided in Appendix 1.B.
Table of contents :
1 Competitive Schools and the Gender Gap in the Choice of Field of Study
1 Introduction
2 Institutional Context
2.1 The Assignment ofMiddle School Students to High Schools
2.2 Major Field of Study and High School Exit Exams
2.3 Selective Undergraduate Programs
3 Data andMethods
3.1 Data
3.2 Cut-off Scores
4 Graphical Evidence
4.1 First-Stage : Effect of Eligibility on Enrollment
4.2 Field of Study at the End of 10th Grade
5 Regression Results
5.1 First Stage Effects on School Environment
5.2 Major Field of Study and Performance on Exams
5.3 Robustness and Falsification Tests
6 Mechanisms
6.1 Attitude towards Competition
6.2 Rank Consideration
6.3 Comparative Advantages and Capacity Constraints
6.4 Teachers’ Characteristics
7 Conclusion
Appendices
1.A Additional Tables and Figures
1.B Cut-off Scores
2 Les redoublants nuisent-ils aux autres étudiants ? Le cas des classes préparatoires scientifiques
1 Introduction
2 Contexte
2.1 Les concours
2.2 Les classes préparatoires
2.3 Classes étoilées
3 Données utilisées
3.1 Les données du concoursMines-Ponts
3.2 Les données du ministère
3.3 Identification des étudiants qui redoublent après avoir été classés
3.4 Échantillon de travail
4 Les redoublants : profils et performances
4.1 Les effets directs du redoublement sur les redoublants
4.2 Le niveau académique des redoublants classés et non classés
5 L’influence des redoublants sur les non-redoublants
5.1 Effets des redoublements sur la taille des classes l’année suivante
5.2 Effets des 5/2 sur les performances des 3/2
5.3 Redoublants classés vs. non classés
6 Conclusion
Annexe
3 From Employment to Engagement? Stable Jobs, Temporary Jobs, and Cohabiting Relationships
1 Introduction
2 Stable Jobs, Temporary Jobs, and Cohabiting Relationships
2.1 Data
2.2 Event Study
2.3 Timing of Events Analysis
2.4 Stable Jobs, Temporary Jobs, Cohabiting Relationships and Fertility
3 Gender, Employment and Cohabiting Relationships before the 1950s
3.1 Data
3.2 Event Study
3.3 Timing of Events Analysis
4 Discussion
5 Conclusion
Appendix
Conclusion générale
References
