Bourdieu, networks, and movements: Using the concepts of habitus, field and capital to understand a network analysis of gender differences in undergraduate physics
Steven Martin Turnbull, Frédérique Vanholsbeeck, Kirsten Locke, Dion R J O'Neale
BBourdieu, networks, and movements: Using the concepts ofhabitus, field and capital to understand a network analysis ofgender differences in undergraduate physics
Steven Martin Turnbull ¤ ,* , Fr´ed´erique Vanholsbeeck , Kirsten Locke Dion R. J.O’Neale , Critical Studies in Education, Faculty of Education and Social Work, University ofAuckland, Auckland, New Zealand Te P¯unaha Matatini, Auckland, New Zealand Department of Physics, University of Auckland, Auckland, New Zealand The Dodd-Walls Centre, Auckland, New Zealand ¤ Current Address: The Department of Physics, The University of Auckland, PrivateBag 92019, Auckland, New Zealand* [email protected]
Abstract
Current trends suggest that significant gender disparities exist within Science,Technology, Engineering, and Mathematics (STEM) education at university, with femalestudents being underrepresented in physics, but more equally represented in life sciences(e.g., biology, medicine). To understand these trends, it is important to consider thecontext in which students make decisions about which university courses to enroll in.The current study seeks to investigate gender differences in STEM through a uniqueapproach that combines network analysis of student enrollment data with aninterpretive lens based on the sociological theory of Pierre Bourdieu. We generate anetwork of courses taken by around 9000 undergraduate physics students (from 2009 to2014) to quantify Bourdieu’s concept of field. We explore the properties of this networkto investigate gender differences in transverse movements (between different academicfields) and vertical movements (changes in students’ achievement rankings within afield). Our findings indicate that female students are more likely to make transversemovements into life science fields. We also find that university physics does a poor jobin attracting high achieving students, and especially high achieving female students. Ofthe students who do choose to study physics, low achieving female students are lesslikely to continue than their male counterparts. The results and implications arediscussed in the context of Bourdieu’s theory, and previous research. We argue that inorder to remove constraints on female student’s study choices, the field of physics needsto provide a culture in which all students feel like they belong.
Introduction
Historically, women have been underrepresented in Science, Technology, Engineeringand Mathematics (STEM) disciplines. This is a concerning issue today internationally,and at all stages of higher education [1–3]. More recent studies indicate specific genderdisparities exist within the sub-fields that comprise STEM [4]. Female students tend tobe underrepresented in physics in higher education, and this is evidenced by research
PLOS a r X i v : . [ phy s i c s . e d - ph ] M a r rom the United States [5–8], Europe [2, 3, 9–11], Asia-Pacific regions [12, 13] andAfrica [14]. In contrast, the same research shows that the life science subjects (biologyand medicine) tend to have more of a gender balance. Why do we see gender differencesin the physical and mathematical science subjects, but not the life science subjects?Much research has been dedicated to understanding the extent, causes, and possiblesolutions to this issue [15, 16].The current study investigates the outcomes for male and female physics students atthe University of Auckland (UoA) — the largest university in New Zealand. We adopt aunique approach, by combining quantitative network analysis with a research frameworkbased on Pierre Bourdieu’s sociological theory. Whilst we argue that these twoapproaches can provide a detailed understanding of gender disparities in studentenrolment patterns, there is a lack of research in this area (for examples of how networkanalysis and Bourdieu have been previously used together ,see the work of deNooy [17]), and Bottero and Crossley [18]). We combine these approaches by usingnetwork analysis to provide a representation of Bourdieu’s concept of field, with anemphasis on his ideas of transverse and vertical movements (students moving from onefield to another, and moving upwards and downwards in achievement rankings in afield). In order to avoid misinterpretation of Bourdieu’s theory, which is easily donewhen “bits and pieces” of it are used [19, p.4], we combine our representation of fieldwith Bourdieu’s concepts of habitus and capital. We argue that network analysis canbring to light the complex patterns of students’ subject enrollment, whilst Bourdieu’stheory offers a rich theortical framework to explain these patterns. We place thefindings of our network analysis in a broad socio-cultural context that brings to lightthe complex interactions between society, gender and subject discipline. To avoidconfusion, the following sections will use ‘field’ as a technical term referring to theBourdieu’s definition (which will be explained in more detail in the next section), and‘discipline’ as a non-technical term that describes the different STEM domains.We begin by introducing a simple model of Bourdieu’s theory, using the field ofscience education to illustrate its concepts. We then add to this outline of theory bybuilding our method of network analysis into Bourdieu’s theory. More specifically, wedescribe how network analysis of student enrollment data can provide a representationof field. Exploring the properties of this network structure allows us to understandgender differences in the movements students make within and across fields. Accordingto Bourdieu [20, p.131]:The social space, being structured in two dimensions (overall capital volumeand dominant/dominated capital) allows two types of movement... verticalmovements, upwards or downwards in the same vertical sector, that is in thesame field... and transverse movements, from one field to another, whichmay occur either horizontally or between different levels.In science education, individual’s may move from one field to another (i.e., from physicsto life science), but also upwards and downwards in achievement rankings in the field.We use these concepts of movements to guide our investigation. We seek to understandwhether there are gender differences in the number of students moving from physics toother fields, and also in the changes in achievement rankings of students in physics. Weclose this article with a discussion of our results in the broader context of previousresearch and Bourdieu’s concepts of capital and habitus. The metaphor of the leaky pipeline is often used to describe the attrition of womenfrom physics [9, 21], in that women are more likely to drop out with each transition
PLOS
For Bourdieu, the world is separated into a collection of different fields [20]. A field canbe considered as a system of social locations, where each individual is objectively rankedby the resources (capital) they have relative to others. For example, in the field oftertiary science education, a lecturer ranks higher than a student, whilst a highachieving student ranks higher than a low achieving student. To begin to theunderstand the hierarchical nature of a field, we must first understand the concept of capital . Originally conceived within economics, capital was defined by Adam Smith (in1887) as “That part [of a person’s wealth] that he expects to provide [them] with. . . income. . . ” [24, p.214]. Bourdieu interpreted capital as a legitimate, valuable andexchangeable resource that individuals can use to gain advantage in society [25].Therefore, the rankings are determined by how we define what is valuable andlegitimate in the field. The practices of individuals within the field are judged bycriteria internal to the domain of activity [26]. Individuals with a high volume of valuedcapital will hold power within the field. For example, high achieving students have highvolumes of capital in the field due to their course grades (a signal of success), whilstlecturers and researchers have a greater volume of capital in the form of qualificationsand research experience. In the field of tertiary science education, lecturers andresearchers sit at the top of the hierarchy, and decide what kinds of capital are valued ordevalued (e.g., professors often decide the course content and manner of teaching forundergraduate students at university). We will discuss Bourdieu’s conceptualization ofcapital and the way it can inform gender equity research in the following section. Beforethen, we will outline a brief description of how the field of physics is structured from anobjective point of view in relation to gender.The numbers of male and female students holding qualification in the differentscience disciplines can provide an objective, surface level understanding of the structureof the field. In the United States, only around 20% of students studying physics atbachelors, masters or doctorate level in 2014 were female [5]. This contrasts with biology,where around 50-60% of students studying at bachelors, masters or doctorate level were
PLOS
The objective rankings within a field are defined by the distribution of capital, and thiscan be used to inform gender equity research in science [29]. Different fields havedifferent forms of logic as to what forms of capital are of value. Using a basic example,a science qualification is worth more in the field of science than in other academic fields.Capital is complex and may take many forms, each of which may be valued differentlydepending on the dominant logic of the field. According to Bourdieu [25], capital hasfour forms: economic (e.g., financial resources), cultural (non-financial assets, such asphysical appearance, spoken language, academic achievement), social (e.g., anindividual’s social network), and symbolic (prestige and recognition, such as awards).Individuals who begin their life with more capital, be that through inheritance orimmediate exposure to the dominant culture, will be more able to gain personal andsocial advantages. For example, a student who is born into a family that speaks thedominant language of an educational institution may find it easier to learn, and astudent with greater economic wealth may be more able to afford the costs associatedwith tertiary study (e.g., tuition fees, relocation, travel). The value of capital is notsolely determined by form, but also by factors such as the manner of acquisition, andthe personal characteristics of the owner. Issues emerge when an individual’s capital isdevalued unjustly by the ‘rules’ operating in the field. For example, internationalresearch has shown that female physicists tend to receive fewer opportunities and careerenhancing resources compared to objectively equal male physicists [30]. Previousresearch of tertiary students suggests that female students may be more likely todiscontinue physics education, regardless of performance [31, 32]. Since disparities inenrollment still exist even after controlling for academic achievement, it is likely that
PLOS
Capital, in its various forms, interacts with habitus (Fig 1); a construct defined byBourdieu as a “system of dispositions” [20, p.170] formed in relation to a field. Whilstcapital is what determines one’s position within the field, habitus is what determinesone’s disposition towards it [19]. An individual’s habitus is the internalization of thesocio-cultural and historical context of a field, and it operates “below the level ofconsciousness and language” [20, p.466]. Roy Nash understood habitus as “a system ofschemes of perception and discrimination embodied as dispositions reflecting the entirehistory of the group and acquired through the formative experiences ofchildhood” [42, p.177]. In simple terms, habitus is what we use to determine whetherthe field is something we are interested in, based on evidence present in theenvironment. Whilst habitus is generally formed during childhood within thefamily [43], it is continually reconstructed and transformed as an individual operates insociety. For example, a student who grows up in a family that places high value onscience may share the same disposition [44]. However, an individual may not choose topursue science when faced with evidence that the field is not for them (for example,receiving poor grades, being treated poorly, lack of role models). Based on this internalmatrix of dispositions, an individual’s lifestyle practices are generated. According toBourdieu, the collection of each individual lifestyle produced by habitus then constitutesthe “represented social world” [20, p.170] — the way that things appear to be. As therepresentation of the social world also influences the formation of habitus, the world andhabitus share a reciprocal relationship. This relationship facilitates the culturalreproduction of inequity over time.Habitus can be used as a concept to explain the gender disparities in scienceenrollments. Based on what they see in their represented social world, students will“[refuse] what they are refused (‘that’s not for the likes of us’), [adjust] their expectationsto their chances, [and define] themselves as the established order definesthem.” [20, p.471]. Based on what students see in their environment, they will makedecisions on what they feel is a realistic study choice. Archer and colleagues explain thisidea further: “social axes of ‘race’/ethnicity, social class, and gender all contribute toshaping what an individual perceives to be possible and desirable.” [45, p.885] Themanner by which students perceive the different scientific disciplines, as they arerepresented in society, likely plays an important role in influencing their desire to studythose disciplines.A wealth of research has outlined the various ways that women are subjugated incertain STEM disciplines (especially physics), with the culmination of these factorsbeing referred to as “the smog of bias” [7, p.1] or the “gender filter” [16, p.370]. Nosingle factor can sufficiently explain why women are less likely to pursue physics [7], buta host of factors are likely to interact and impact on the dispositions students hold(habitus). Due to the pervasiveness of these various factors across society, habitus cantake on a collective quality where individuals tend to hold stereotypical views on what isexpected for members of different groups. To provide a simple example, research across34 countries has shown that science tends to be implicitly associated with men morethan with women, and that this level of gender bias predicts gender differences inscience performance [23]. As outlined by Bourdieu, an objective class of individuals canbe considered the “the set of agents who are placed in homogenous conditions ofexistence imposing homogenous conditionings and producing homogenous systems of
PLOS
The current study was motivated by the need to understand any potential genderdifferences in student outcomes in general, and at the University of Auckland (UoA) inparticular. Our study seeks to not only understand the outcomes for physics students atthe UoA, but to employ a unique approach that highlights the complexity of studentenrollments and places them in a wider socio-cultural context. To do so, we employnetwork analysis on student enrollment records to provide a detailed representation ofthe field of physics at the UoA. The network analysis approach builds on the workof [17] and [18] who described the utility of combining network analysis withBourdieu. [18] provide an example of how networks of social relations can provide arepresentation of a field. The current study expands on this area of research byconceptualising academic fields as communities detected in networks of course selection.Furthermore, we draw attention to under-utilised concepts of Bourdieusian theory: theconcepts of transverse movement between fields, and vertical movements within fields.We focus on providing a basic description of the movements that physics students make
PLOS
Bourdieu’s theory encourages us to view student movements across STEM domains inrelation to the structures of the field, the volume of capital a student holds, and themanner by which habitus guides practices in the field. In addition to our objectiverepresentation of the field of physics, we also consider what may motivate thesemovements, based on evidence from previous research.According to Bourdieu, society is structured in a manner that allows individuals toengage in two types of movement: vertical and transverse: “vertical movements,upwards or downwards in the same vertical sector, that is in the same field... andtransverse movements, from one field to another, which may occur either horizontally orbetween different levels” [20, p.131]. Vertical movements upwards require an increase inthe prized capital in the field. In tertiary science education, this may be represented bygrades in science courses over time. Transverse movements entail a shift to a new field,and the conversion of accumulated capital into the capital accepted in the new field. Forexample, a student making a transverse movement from physics to life sciences will haveto assimilate to a different skill set, and even a different culture. Transverse movementscan be used as a strategy to protect a relative vertical position:“transverse movements entail a shift into another field and the reconversionof one type of capital into another or of one subtype into another subtype...and therefore a transformation of the asset structure which protects overallcapital volume and maintains position in the vertical dimension” [20, p.132]When an individual feels that they are slipping in the ranks of the field, they maychoose to make a transverse movement to a new field, where their accumulated capitalholds more translatable value.In the current study, we conceptualize cultural capital in its institutionalized form asmeasured by course grades. The current study, therefore, seeks to understand: • Whether there are gender differences in UoA physics students moving from oneacademic field to another. • Whether there are gender differences in the persistence of UoA students in physics. • Whether there are gender differences in UoA physics students moving upwards ordownwards in academic achievement (as signalled by course grades).Whilst our data do not allow us to conceptualize forms of capital other thaninstitutionalised cultural capital (i.e, course grades), our methodology leaves theopportunity for future research to incorporate other measures of students’ capital. Morespecifically, future research should investigate how other forms of capital are distributedacross fields and relate to the movements that students make.
PLOS aterials and methods
The current study uses administrative student data from the UoA from 2009 to 2014 (N= 8905), including demographic and academic information. For the purposes of thisstudy, the only demographic variable considered in the analysis was gender. Academicvariables include course codes that students were enrolled in, and the year and semesterin which they were enrolled. We did not have information regarding students’ degreeplans or majors. Records of non-physics courses were included as long as a student hadenrolled in at least one physics course during the study period. At the UoA, studentsare required to take two courses outside of their major, with the options being titled asgeneral education courses. We excluded all students who studied a general educationcourse in physics from our analysis. We know that these students are not physicsstudents, and they do not offer a representative sample of students from outside ofphysics.A typical Bachelor of Science physics degree at the UoA takes place over the courseof three years. In their first year, physics students are required to take AdvancingPhysics 1 (AP1) and then Advancing Physics 2 (AP2) before moving onto second yearphysics. Life science students (those majoring in biomedical sciences or medicine) arerequired to take Physics for Life Sciences (PLS) in their first year. PLS is taught by thephysics department. This means that, despite our study population including onlystudents who took a physics course, many of the students present in our data set werelikely majoring in life sciences. Our population therefore allows us to compare theoutcomes for students in the physics and life sciences disciplines. AP1 and PLS coverthe same content, but are presented in a different manner. One significant differencebetween AP1 and PLS is that AP1 assumes a knowledge of calculus, while PLS doesnot. This is an important point to consider, as a mathematics background may be animportant form of science related capital [58], and female students may be more likelyto drop out of physics education after taking calculus [31]. The current study was ableto compare the AP1 and PLS subsets of the general physics population to account for astudent’s first year disciplinary intentions. PLS is still considered an acceptableprerequisite for AP2 in lieu of AP1, although it is rare for students to take this route.
The following variables were used in the analysis: • Grade Point Equivalence (GPE): GPE is an entry level score that provides astandard measure of a student’s prior academic performance at the time ofadmission to university, regardless of the qualification they previously took. It ismeasured on a 0-9 scale, with 9 being the highest performing. It provides anaggregate measure of how well a student did in all of their high school courses [59]. • Grade Point Unit (GPU): GPU is a measure of a student’s university performancein a single course. It is measured on a 0-9 scale, with 0 being equivalent to a fail(D+ or lower), and 9 being equivalent to an A+ grade. GPU was used as ameasure of performance for AP1, AP2 and PLS. • Gender: Due to limitations in the administrative data that were used, gender wasonly recorded as male or female.
PLOS .4 Procedure
Although Bourdieu offers a rich theory to interpret movements within and betweenfields, we are left with the challenge of defining what constitutes a field. Whilst it couldbe argued that every student who takes a physics course at university is a physicsstudent, we believe that this is not sufficient. Students may be enrolled in a subjectdiscipline on paper, but actually be fully engaged in a separate field of study. A goodexample of this is PLS. PLS students may be considered physics students on paper, buttheir main field of study is likely biomedical sciences or medicine. Through networkanalysis, we are able to define academic fields in terms of the patterns of courseselection. We represent course selection patterns as a network, where nodes representuniversity courses and edges represent the enrolments of students within courses. Wethen explore the structure of the network by investigated the communities of coursesthat tend to be taken together by students. Our approach, similar to blockmodellingapproaches [18, 60], allows us to take a complex network and reduce it to its corestructure. It does this by identifying communities of nodes that tend to share moreedges. We can then explore patterns at the level of communities instead of at the levelof nodes. In the current study, we interpret these communities as academic fields.Following this, we are able to investigate gender differences in the transverse thatstudents make across the fields represented in our network. We supplement our networkwith course achievement data to compare vertical movements within and across fields.The following section outlines the series of steps that were used to generate thecourse network and use it to answer our research questions regarding gender differencesin students transverse and vertical movements. Through the analysis of courserelationships, we can take a non-biased approach to defining the fields in which studentsare located.
To begin our network analysis, we structured our data as an adjacency matrix, whereboth rows and columns represent the courses taken by students in our sample, and a cellvalue is the number of students who took both course i and course j within theirundergraduate degree. Whilst we could define edges in relation to the frequency ofstudents who took a pair of courses, this does not truly reveal the underlyingcommunity structures we are interested in. Pairs of courses including one very popularcourse will tend to have higher values regardless of whether the two courses belong tothe same academic field. We take into account course populations by normalising thematrix using a Revealed Comparative Preference (RCP) score. RCP measures thefraction of students from a course j who also took a second course i , relative to theoverall fraction of students taking course i , across all other courses. More specifically: RCP ( i, j ) = x ij /x j x i /x where x ij is the number of students taking both course i and j , x j (or x i ) is the totalnumber of students taking course j (respectively, course i ), and x is the total number ofunique students enrolled in any course. The RCP metric is based on the measureRevealed Comparative Advantage, used in economics [61], and was calculated using theEconGeog package in R [62]. The RCP approach to normalising gives the “revealed”course preferences, controlling for the enrollment numbers of each course (that is, thecourses that tend to be taken together by students in the network more often thanwould be predicted by the course populations alone). RCP values greater than oneindicate that a pair of courses had a ‘preference’ for being taken together, given therelative populations of both courses, whilst RCP values below one indicate no evidence PLOS
Table 1. Compositions of the Communities Detected in the CourseNetwork.
Community Count students Proportion Female Total EnrollmentsAncient History 150 0.48 205Biological Science 5630 0.52 18600Chemical Materials 20 0.35 60Chemistry 1660 0.49 4180Chinese 60 0.27 85Computer Science 4405 0.28 22195Engineering 980 0.36 9315Finance-Marketing 1430 0.29 6735Food Science 640 0.53 1505Geography-Geology 1470 0.38 6095Japanese 85 0.38 180Law 170 0.46 240Liberal Arts 6410 0.38 12750Medical Science 4715 0.53 26790Nursing 70 0.81 300Optometry 550 0.61 1310Pharmacy 645 0.58 2100Physics-Maths 3060 0.25 12125Population Health 200 0.57 510Psychology 1440 0.52 4410Sports Science 245 0.47 575Statistics 1470 0.36 3985Surgery 350 0.43 2800
Counts have been rounded to the nearest 5 to preserve confidentiality. Proportions wereformulated using original values.
To investigate whether there are gender differences in UoA physics students movingfrom one academic field to another during their undergraduate degree (transversemovements), we build on the network outlined in the previous section. We take thesame set of nodes, with the same community structures, but weight edges by the number
PLOS i before course j . From this new directed network, we areable to assess the movements that students make between communities. To answer ourquestions regarding the transverse movements that students make from one field toanother, we aggregate the number of movements from courses within community m tocourses within community n (see Fig 3). For example, the courses in the Physics-Maths community in Fig 2 become a single Physics-Maths node in Fig 3. Outgoing edgesbetween communities are aggregated into a single outgoing edge, with a weightingequivalent to the sum of all outgoing edges weights from nodes in the community. Edgesbetween courses within a community are similarly aggregated, and are represented asself-loops (a link from a node to itself) in Fig 3. To investigate how transversemovements differ by gender, we calculate the odds (with 99% confidence intervals) of afemale student moving from community i to community j over a male student. The newnetwork is represented as a network in Fig 3 and as a heat map in Fig 4. We also seek to investigate how male and female students with differing levels of priorachievement choose to invest their capital. Are there gender differences in the verticalmovements (moving upwards or downwards in the objective rankings in a field) thatstudents make from one stage to the next? Do male and female students with differentlevels of prior achievement choose to invest their capital differently? To understand thenature of students’ vertical movements between within and between fields, weincorporate student achievement data into our previously established network. For eachcourse, we have the student grade point unit score (i.e., their level of achievement). Ourdata set also includes an average high school achievement measure, called Grade PointEquivalence (GPE), for the majority of students in our network. This allows us to lookat the transitions that male and female students make from high school to universitystudy.We are particularly interested in the movements that students make going from highschool to three specific stage one courses: AP1, AP2, and PLS. We also investigate thegender differences in vertical movements that students make from these physics coursesto our detected fields, and between our detected fields. For our detected fields wecalculate a Grade Point Average (GPA) score for each student, in which we take themean of the student’s grade point unit scores for each course they took within thecommunity. For example, the Physics-Maths GPA score will be a student’s meanaverage grade point unit score for all of the courses they took within the Physics-Mathscommunity.As outlined by Bourdieu, an individual’s power in a field is determined by thecomposition and volume of capital they hold relative to other individuals. As our goal isto compare the relative vertical position of students within and between fields, weconvert the achievement scores (GPE for high school, GPU for the key stage onecourses, and GPA for the communities) into percentile ranks. Standardisingachievement in this manner facilitates comparisons across fields. Top achievers in a fieldwill have a percentile rank score of 100, whilst low achievers will have a percentile rankscore closer to 0. We can then compare the change in percentile rank scores for maleand female students across our network. To describe the gender differences in verticalmovements, we use independent 2-group Mann-Whitney U Tests (a non-parametrict-test) to compare differences in percentile rank change between male and femalestudents. We were then able to determine whether there were any significant differencesbetween male and female students gaining in relative performance across fields. We alsoreport the odds (with 99% confidence intervals) of top, middle, and low achievingfemale students enrolling in different fields (where achievement groups are based onpercentile rank split into three equally sized bins). We explore the movements from high
PLOS
Networks showing the revealed communities of courses that students take can be seen inFig 2 and Fig 3. Fig 2 shows the network of courses offered by the University ofAuckland (UoA) between the years 2009 and 2014, with communities indicating coursesthat tend to be taken together within students’ undergraduate degrees (represented bythe different colours). The communities include (in ascending order of aggregatedcourse enrollments): Medical Science, Computer Science, Biological Science, LiberalArts, Physics-Maths, Engineering, Finance-Marketing, Geography-Geology, Psychology,Chemistry, Statistics, Surgery, Pharmacy, Food Science, Optometry, Sports Science,Population Health, Nursing, Law, Ancient History, Japanese, Chinese, and ChemicalMaterials.The use of Revealed Comparative Preference (RCP) in conjunction with thecommunity detection reveals underlying academic fields in which physics studentparticipate, as indicated by the combinations of courses that students enroll in. Physicscourses (including AP1 and AP2, the first prerequisites for a physics major at the UoA)and mathematics courses are located the same field, which we label
Physics-Maths .PLS, a physics course required for students wanting to study medicine, belongs to thefield of
Medical Sciences . We report the the counts of students per community, with thepercentage of female students, in Table 1. Liberal Arts, Biological Science, and MedicalScience were the three largest communities based on number of unique students enrolledin the field. Medical Science, Computer Science, and Biological Science were the largestcommunities in terms of total enrollments (an individual student may be enrolled inmore than one course per field). In terms of the proportion of female students percommunity, Physics-Maths (0.25), Computer Science (0.28), and Chinese (0.27) werethe most male dominated. Nursing (0.81), Optometry (0.61), and Pharmacy (0.58) werethe most female dominated.The network and RCP approach provides a non-biased method of classifying thefields in which students are participating in. The use of RCP shows that disciplinarylabels (i.e., ‘Physics’) are imperfect in classifying the patterns of courses that studentsenrol in. Although PLS is a physics course, it has a higher affinity with the life sciences,and our community detection approach reflects this by locating PLS within the field ofMedical Science. For example, the percentage of female students enrolled in all physicscourses (including PLS) is 40%. Our community detection shows that female studentsonly make up around 25% of the main Physics-Maths community. The differencebetween these percentages is substantial, and raises important implications for the wayin which universities report the number of students studying in different disciplines.
We first wanted to understand whether there are gender differences in UoA physicsstudents moving from one academic field to another. Our results regarding thesetransverse movements (more detail is given in the supplementary material) show thatfemale students were around 1.82 (CI: 1.63-2.02) times more likely to take a course inBiological Science after taking a course in Physics-Maths, and 1.44 (CI: 1.40-1.47) timesmore likely to take a further course in Biological Science after taking a previousBiological Science course. On the other hand, male students were around 1.97 (OR =0.51, CI: 0.48-0.56) times more likely to take a course in Physics-Maths after taking acourse in Biological Science. There were no significant gender differences in students
PLOS
We investigate the impact of student achievement on student enrollment in two ways.We firstly report the number of top, middle, and low achievers who make movementsfrom high school to the key stage one physics courses (AP1, AP2 and PLS), and to thefields revealed in our network. We then assess the vertical movements that studentsmake by analysing gender differences in the change in objective rankings within andbetween fields. We begin by reporting the progression of students from high school touniversity physics.For students ranking in the bottom third of high school students (“low achievers”),17.94% of female students and 35.53% of male students went on to study AP1. Thus,low achieving male students were 2.52 (OR = 0.40, CI :0.30-0.52) times more likely toenter the main physics pathway at the UoA compared to their female counterparts. Forstudents ranking in the middle third of high school students (“middle achievers”),10.69% of female students and 30.58% of male students went on to study AP1. Thus,middle achieving male students were 3.68 (OR = 0.27, CI: 0.20-0.37) times more likelyto enter physics at the UoA compared to their female counter parts.Our findings show that of the students who were ranking in the top third of studentscoming from high school (“top achievers”), very few chose to invest their capital inphysics. Only 8.72% of male students and 5.06% of female students who were topachievers from high school chose to enrol in AP1. These percentages also indicate thatfrom this top achieving group, male students were 1.79 times more likely to go to AP1(OR = 0.56, CI: 0.36-0.87). Thus, not only does it appear that physics is anunattractive option of top achieving high school students, but this is particularly truefor top achieving female students. In contrast, 72.66% of male students, and 88.35% offemale students from this top achieving group enrolled in PLS. Top achieving femalestudents were thus 2.84 (CI: 2.10-3.83) times more likely than their male counter partsto follow this pathway.
PLOS for me ’. For fields such as physics andcomputer science (where we found the most consistent gender disparities in enrollments)students are likely influenced from an early age by the “smog of bias” [7, p.1] thattargets women. Through the combination of a myriad of factors, from the negativegender stereotypes [23], to the ways in which women’s competence is unfairlyquestioned [47–49], students will internalise (via habitus) the perception that physics issomething men do, and where women are unwelcome [68]. Until the ‘smog of bias’ isaddressed, female students will continue to have constrained choice in science.Whilst we could interpret the lower likelihood of a male student studying in the lifesciences as evidence of constrained choice also, we find this an unrealistic interpretation.The sizable representation of male students and researchers in the life sciences presentlyand historically, and the lack of negative factors that impact male students in thisdomain, mean that the life sciences are likely still a realistic study choice for malestudents. To put more simply, male students have more choice on where to invest theircapital, whilst female students are more likely to face obstacles. The rules operating inthe field of physics may require female students to make extra effort to appearcompetent and persevere in the field. As outlined by Ong [47, p.594] in a study ofminority female physics students: “the ways in which women of color organizethemselves to appear competent in the context of physics specify invisible rules aboutthe strict boundaries around local scientific communities.” The idea that women inphysics may have to “relegate social and cultural identities to the margins” [47, p.597] inorder to succeed in physics corresponds to Bourdieu’s idea that individuals lacking in the‘valued’ cultural capital in a field may need to make sacrifices to get ahead [20, p.333].Of the students from AP1 who ranked in the bottom third of achievers (lowachievers), we found that 40.38% of male students, and 28.29% of female studentsprogressed to AP2. Thus, female students from this low achieving group were around1.717 (OR =0.58, CI: 0.35-0.98) times less likely to progress from AP1 to AP2. Therewere no significant gender differences in the middle (OR = 0.84, CI:0.56-1.24) and topachieving (OR = 0.77, CI: 0.49-1.21) AP1 students who went to AP2.
PLOS over-confident . It may be that in our sample, gender differencesin progression from AP1 to AP2 for middle and top achieving students were not presentas the grades received offered evidence that they belong in physics. For the lowachieving students, belonging is not evidenced by their grades. Low achieving malestudents may be buffered by a habitus that, after years of socialization, predisposesthem to physics. Female students, on the other hand, may be less likely to have thisprotective disposition. Whilst further research is needed to substantiate this claim, pastresearch does suggest that students are more likely to make internal attributions offailure for female students in science (i.e., they fail because they are not good at it), andexternal attributions of failure for male students (i.e., unfavourable circumstances) [72].Furthermore research by Ellis, Fosdick and Rasmussan [31] found that female studentsare more likely to discontinue physics after taking an introductory calculus course, withfemale students also being more likely to cite lack of understanding as a reason fordropping out. This may also apply to students in our sample, as AP1 includes contentthat requires knowledge of calculus.We also investigated the rank change for students moving from high school to thekey stage one physics courses, and from those physics courses on to the fields detectedin our network. We found no statistically or practically significant gender differences inthe vertical movements in these pathways, with the exception of students going fromhigh school to AP1. As indicated by Fig 5 and 6, we found that low and middleachieving female high school students were more likely to decrease their rank in the field(i.e., make a vertical movement downwards) in AP1 with this being significant. Onaverage, low achieving female students went down around 6 ranks compared to theirmale counterparts (Difference in Position= − .
71, CI: − .
57 - − . − .
57, CI: − .
86 - − . before university education, particularly for middle andlow achieving students. Many studies point to the late childhood and early teenageyears as a key formative stages [44, 54, 73] for identity within science. Future studies oftertiary education in New Zealand should investigate the role of science identity insubject selection decisions further. The current study offers a detailed account of the movements that students makethrough university physics. Our results show that female students were less likely toprogress from high school to AP1, regardless of prior achievement, while low achievingfemale AP1 students were less likely to progress to AP2. The findings of the currentstudy suggest that more needs to be done to ensure that physics is perceived as a viable
PLOS
PLOS
There are limitations to the current work that future studies should address. Firstly,our data set is limited to UoA physics students, and included only course selection andperformance information, and minimal demographic information. We did not have dataregarding the course selection information of students prior to university, whilst ourmeasure of high school achievement was a general measure and not subject specific.More detailed data would have provided more information regarding students’educational trajectories. With that being said, our results show the utility of workingwith student record data. Our network analysis, whilst simple, also provides a strongframework for working with more complex data; for example, investigating thedistribution of economic, cultural, social capital across the network.Whilst we argue that our network analysis approach enables us to draw manyconclusions from our data, our study would also have benefited from combining ourquantitative analysis with qualitative measures. We have used a quantitative approachto defining the field, and used evidence from other research studies to draw conclusionsfrom our data. Whilst this approach is informative, qualitative approaches can provideeven more context specific details. Bourdieu highlighted the need to break thedichotomy between the aim of understanding the ‘objective reality’ (the overalldistributions of groups and relationships between them) and the aim of understanding“not ‘reality’, but agents’ representations of it” [25, p.482]. Surveys and interviews ofstudents would provide contextual and fine-grained detail that would complement ourquantitative network analysis. Qualitative analysis may also be a more appropriate wayto investigate gender as a non-binary construct.Despite having access to information regarding the ethnicity of students, we decidednot to present this information in the current analysis. This is due to the fact thatpreliminary analysis showed low cell sizes for ethnic groups other than New ZealandEuropean and Asian students in physics, in particular M¯aori and Pacific Island students(these findings are available on request). When possible, future studies should make useof an intersectional research design (one that explores the interaction between gender,ethnicity, social class etc.). This is especially important when using a Bourdieusianframework to interpret results. As suggested by Bourdieu: “The individuals grouped ina class that is constructed in a particular respect... always bring with them secondaryproperties” [20, p.102]. Understanding the intersection of student characteristics wouldallow us to control for the secondary properties that Bourdieu speaks of. The authorsare currently conducting further to understand why there were low cell sizes forminority groups using data from earlier educational stages (i.e., secondary school).
Conclusion
The current study investigated gender differences for undergraduate physics students atthe University of Auckland (UoA) through the use of network analysis on student data,with an interpretive lens based on the work of Pierre Bourdieu. Our network analysisrevealed the different academic fields in which students are situated. We outline theutility of networks in visualising Bourdieu’s concepts of vertical and transversemovements within and across fields. Analysis showed gender differences in transversemovements (moving from one field to another) consistent with gender stereotypes:female students were more likely to enroll in life science fields (Biological Science,
PLOS
References
1. Abraham J, Barker K. Exploring Gender Difference in Motivation, Engagementand Enrolment Behaviour of Senior Secondary Physics Students in New SouthWales. Research in Science Education. 2014;45(1):59–73.doi:10.1007/s11165-014-9413-2.2. Stevanovic B. Girls in Science and Technology in Secondary and Post-SecondaryEducation: The Case of France. British Journal of Sociology of Education.2013;35(4):541–558. doi:10.1080/01425692.2013.791229.3. Smith E. Women into Science and Engineering? Gendered Participation inHigher Education { STEM } Subjects. British Educational Research Journal.2011;37(6):993–1014. doi:10.1080/01411926.2010.515019.4. Mullis I, Martin M, Foy P, Hooper M. TIMSS 2015 International Results inMathematics. TIMSS & PIRLS International Study Center at Boston College;2016.5. National Science Foundation. Women, Minorities, and Persons with Disabilities inScience and Engineering: Field of degree: Women; 2017. .6. Cunningham BC, Hoyer KM, Sparks D. Gender Differences in Science,Technology, Engineering, and Mathematics (STEM) Interest, Credits Earned, andNAEP Performance in the 12th Grade. Stats in Brief. NCES 2015-075. NationalCenter for Education Statistics. 2015;.7. Kost-Smith LE, Pollock SJ, Finkelstein ND. Gender Disparities inSecond-Semester College Physics: The Incremental Effects of a ‘Smog of Bias’.Physical Review Special Topics - Physics Education Research. 2010;6(2):1–17.doi:10.1103/physrevstper.6.020112.
PLOS { STEM } Pathway for Women: What Has Changed?Gifted Child Quarterly. 2012;57(1):39–55. doi:10.1177/0016986212460085.9. Huyer G Sophia; Westholm. Gender indicators in science, engineering andtechnology: an information toolkit. UNESCO; 2007.10. Institute of Physics. It’s Different For Girls: The Influence of Schools. London,United Kingdom: Institute of Physics; 2012.11. Institute of Physics. Closing Doors: Exploring gender and subject choice inschools. London, United Kingdom: Institute of Physics; 2013.12. Education Counts. Secondary subject roll by learning year level & subject name -1 july 2015; 2016. .13. Kennedy JP, Lyons T, Quinn F. The Continuing Decline of Science andMathematics Enrolments in Australian High Schools. Teaching Science.2014;60(2):34–46.14. Semela T. Who is Joining Physics and Why? Factors Influencing the Choice ofPhysics among Ethiopian University Students. International Journal ofEnvironmental and Science Education. 2010;5(3):319–340.15. Brewe E, Sawtelle V. Editorial: Focused Collection: Gender in Physics. PhysicalReview Physics Educucation Research. 2016;12(2):1–4.doi:10.1103/physrevphyseducres.12.020001.16. Blickenstaff JC. Women and Science Careers: Leaky Pipeline or Gender Filter?Gender and Education. 2005;17(4):369–386. doi:10.1080/09540250500145072.17. De Nooy W. Fields and networks: correspondence analysis and social networkanalysis in the framework of field theory. Poetics. 2003;31(5):305–327.18. Bottero W, Crossley N. Worlds, fields and networks: Becker, Bourdieu and thestructures of social relations. Cultural sociology. 2011;5(1):99–119.19. Bourdieu P, Waquant L. The Practice of Reflexive Sociology; 1992. Availablefrom: .20. Bourdieu P. Distinction: A Social Critique of the Judgement of Taste. Routledge;1984.21. Schiebinger L. Has Feminism Changed Science?. vol. 4 of 10. 1st ed. Cambridge,MA: Harvard University Press; 2001.22. Cannady MA, Greenwald E, Harris KN. Problematizing the STEM PipelineMetaphor: Is the STEM Pipeline Metaphor Serving Our Students and the STEMWorkforce? Science Education. 2014;98(3):443–460. doi:10.1002/sce.21108.23. Nosek BA, Smyth FL, Sriram N, Lindner NM, Devos T, Ayala A, et al. NationalDifferences in Gender-Science Stereotypes Predict National Sex Differences inScience and Math Achievement. Proceedings of the National Academy ofSciences. 2009;106(26):10593–10597. doi:10.1073/pnas.0809921106.
PLOS .25. Bourdieu P. The Forms of Capital. In: Richardson J, editor. Handbook of Theoryand Research for the Sociology of Education. Connecticut, US: GreenwoodPublishing Group; 1986. p. 47–58.26. Hilgers M, Mangez E. Introduction to Pierre Bourdieu’s theory of Social Fields.Bourdieu’s theory of social fields: Concepts and applications. 2014; p. 1.27. Education Counts. Secondary Subject Roll by Learning Year Level & SubjectName - 1 July 2015. Wellington, New Zealand: EducationCounts, Ministry ofEducation; 2016. Available from: .28. Bray B, Timewell E. Women in Science: A 2011 Snapshot. Christchurch, NewZealand: Association for Women in the Sciences ( { AWIS } ); 2011. Available from: .29. Kelly A. The Construction of Masculine Science. British Journal of Sociology ofEducation. 1985;6(2):133–154. doi:10.1080/0142569850060201.30. Ivie R, Tesfaye CL, Czujko R, Chu R. The Global Survey of Physicists: ACollaborative Effort Illuminates the Situation of Women in Physics. In: AIPConference Proceedings. { American Institute of Physics } Publishing; 2013. p.53–61.31. Ellis J, Fosdick BK, Rasmussen C. Women 1.5 Times More Likely to LeaveSTEM Pipeline after Calculus Compared to Men: Lack of MathematicalConfidence a Potential Culprit. { PLoS } {
ONE } . 2016;11(7):1–14.doi:10.1371/journal.pone.0157447.32. Ost B. The Role of Peers and Grades in Determining Major Persistence in theSciences. Economics of Education Review. 2010;29(6):923–934.doi:10.1016/j.econedurev.2010.06.011.33. Brown CS, Leaper C. Latina and European American Girls’ Experiences withAcademic Sexism and their Self-Concepts in Mathematics and Science DuringAdolescence. Sex Roles. 2010;63(11):860–870. doi:10.1007/s11199-010-9856-5.34. Hazari Z, Sonnert G, Sadler PM, Shanahan MC. Connecting High School PhysicsExperiences, Outcome Expectations, Physics Identity, and Physics Career Choice:A Gender Study. Journal of Research in Science Teaching. 2010; p. 978—-1003.doi:10.1002/tea.20363.35. Hazari Z, Sadler PM, Sonnert G. The Science Identity of College Students:Exploring the Intersection of Gender, Race, and Ethnicity. Journal of CollegeScience Teaching. 2013;42(5):82–91.36. Hazari Z, Brewe E, Goertzen RM, Hodapp T. The Importance of High SchoolPhysics Teachers for Female Students’ Physics Identity and Persistence. ThePhysics Teacher. 2017;55(2):96–99. doi:10.1119/1.4974122. PLOS
PLOS { STEM } Career Interest in High School: A Gender Study. Science Education.2012;96(3):411–427. doi:10.1002/sce.21007.58. Black L, Hernandez-Martinez P. Re-thinking science capital: The role of ‘capital’and ‘identity’ in mediating students’ engagement with mathematically demandingprogrammes at university. Teaching Mathematics and its Applications.2016;35(3):131–143. doi:10.1093/teamat/hrw016.59. University of Auckland. What is a GPE Grade Point Equivalent?; 2016. https://uoa.custhelp.com/app/answers/detail/a_id/3867/ .60. White HC, Boorman SA, Breiger RL. Social structure from multiple networks. I.Blockmodels of roles and positions. American journal of sociology.1976;81(4):730–780.61. Balassa B. Trade Liberalisation and “Revealed” Comparative Advantage. TheManchester School. 1965;33(2):99–123. doi:10.1111/j.1467-9957.1965.tb00050.x.62. Balland PA. Economic Geography in R: Introduction to the EconGeo package;2017. Available at SSRN: https://ssrn.com/abstract=2962146 or http://dx.doi.org/10.2139/ssrn.2962146 .63. Edler D, Rosvall M. The MapEquation software package; 2017.64. Csardi G, Nepusz T, et al. The igraph software package for complex networkresearch. InterJournal, Complex Systems. 2006;1695(5):1–9.65. Claussen S, Osborne J. Bourdieu’s Notion of Cultural Capital and itsImplications for the Science Curriculum. Science Education. 2013;97(1):58–79.doi:10.1002/sce.21040.66. Archer L, Dawson E, DeWitt J, Seakins A, Wong B. ”Science capital”: Aconceptual, methodological, and empirical argument for extending bourdieusiannotions of capital beyond the arts. Journal of Research in Science Teaching.2015;52(7):922–948. doi:10.1002/tea.21227. PLOS { STEM } Stereotypic Attribution BiasAmong Women in an Unwelcoming Science Setting. Psychology of WomenQuarterly. 2016;40(3):378–397. doi:10.1177/0361684316630965.73. DeWitt J, Archer L, Osborne J. Science-related Aspirations Across thePrimary–Secondary Divide: Evidence from two surveys in England. InternationalJournal of Science Education. 2014;36(0):1609–1629.doi:10.1080/09500693.2013.871659.74. Archer L, DeWitt J, Osborne J, Dillon J, Willis B, Wong B. Science Aspirations,Capital, and Family Habitus: How Families Shape Children’s Engagement andIdentification With Science. American Educational Research Journal.2012;49(5):881–908. doi:10.3102/0002831211433290.75. Cheryan S, Ziegler SA, Montoya AK, Jiang L. Why are some STEM fields moregender balanced than others? Psychological Bulletin. 2017;143(1):1.76. Kelly AM. Social cognitive perspective of gender disparities in undergraduatephysics. Physical Review Physics Education Research. 2016;12(2):020116.77. Master A, Cheryan S, Meltzoff AN. Computing whether she belongs: Stereotypesundermine girls’ interest and sense of belonging in computer science. Journal ofEducational Psychology. 2016;108(3):424.78. Equality Challenge Unit. Athena SWAN Charter; 2018. .79. European Commission. Horizon 2020 - Promoting Gender Equality in Researchand Innovation; 2018. https://ec.europa.eu/programmes/horizon2020/en/h2020-section/promoting-gender-equality-research-and-innovation .80. Archer L, Dawson E, DeWitt J, Seakins A, Wong B. ‘Science Capital’: AConceptual, Methodological, and Empirical Argument for ExtendingBourdieusian Notions of Capital Beyond the Arts. Journal of Research in ScienceTeaching. 2015;52(7):922–948. doi:10.1002/tea.21227.
PLOS
Capital × Habitus Field of Study Practices/Dispositions
GradesCourse enrollment
Fig 1. Simplified Bourdieusian Theoretical Model.
The Bourdieusian framework used in the current study, adapted from the original modeloutlined by Bourdieu [20, p.10] and the work of Archer and colleagues [67, 74, 80]. Astudent’s habitus interacts with their acquired level of capital (in particularscience-related capital) to generate a student’s practices (behaviours, grades etc.) andtheir dispositions towards the field. A student’s habitus, a matrix of internaldispositions [81], is formed in relation to the specific socio-cultural and historicalcontext of a field. A student who is positively predisposed to study in a scientific field,whilst also having access to various forms of science-related capital, will likely achievehigher grades in that field and aspire to study in that field in the future. A student whoencounters bad experiences in the field will likely be dissuaded from future study viatheir habitus (‘this discipline is not for me’).
PLOS
CCTG_101ACCTG_102ACCTG_211ACCTG_221ACCTG_222ACCTG_311ACCTG_312ACCTG_331ANCHIST_100ANCHIST_102ANTHRO_100 ANTHRO_101ANTHRO_102ANTHRO_200BIOMENG_221BIOMENG_241BIOMENG_321 ANTHRO_201BIOINF_301BIOSCI_100 BIOSCI_101BIOSCI_102BIOSCI_103BIOSCI_104 BIOSCI_106BIOSCI_107 BIOSCI_201BIOSCI_202BIOSCI_203BIOSCI_204BIOSCI_205BIOSCI_206BIOSCI_207BIOSCI_208BIOSCI_209BIOSCI_210 BIOSCI_321BIOSCI_328 BIOSCI_322BIOSCI_329 BIOSCI_330BIOSCI_333BIOSCI_335 BIOSCI_340BIOSCI_347BIOSCI_348 BIOSCI_349 BIOSCI_350BIOSCI_351BIOSCI_353BIOSCI_354BIOSCI_356BIOSCI_394ACCTG_371 BUSINESS_101BIOSCI_320BIOSCI_323BIOSCI_337 BIOSCI_358BIOSCI_395BUSINESS_102CHEM_100CHEM_110 CHEM_120CHEM_150 CHEM_210CHEM_220CHEM_230CHEM_240 CHEM_310CHEM_320CHEM_330CHEM_340CHEM_350 CHEM_243CHEM_380 CHEM_390CHEMMAT_121 CHEMMAT_211CHEMMAT_212CHEMMAT_213 CHEMMAT_221CHEMMAT_232 CHEMMAT_312CHEMMAT_313CHINESE_100CIVIL_201CIVIL_210CIVIL_211CIVIL_220 CIVIL_221CIVIL_230 CIVIL_250CIVIL_312CIVIL_313 CIVIL_314CIVIL_322CIVIL_324CIVIL_331CIVIL_332CIVIL_360 ACCTG_321COMLAW_101COMLAW_201COMLAW_203 COMLAW_301ANTHRO_106ASIAN_100 CHEM_392COMPSCI_101COMPSCI_105 COMPSCI_111COMPSCI_210 COMPSCI_215COMPSCI_220COMPSCI_225COMPSCI_230COMPSCI_280COMPSCI_313COMPSCI_314COMPSCI_320 COMPSCI_335COMPSCI_340COMPSCI_345COMPSCI_350COMPSCI_351COMPSCI_367COMPSCI_369 COMPSYS_201 COMPSYS_202COMPSCI_373 ECON_101ECON_111ECON_201ECON_211ECON_351CIVIL_361ELECTENG_101ELECTENG_202COMPSYS_302ELECTENG_204ELECTENG_209ELECTENG_210ELECTENG_303ELECTENG_305ELECTENG_307ELECTENG_309ELECTENG_310CHEMMAT_242ELECTENG_208ELECTENG_311 ENGGEN_115ENGGEN_121ENGGEN_131ENGGEN_140 ENGGEN_150ENGGEN_199ENGGEN_204ENGGEN_299ENGGEN_303 ENGSCI_111ENGSCI_211ENGSCI_233ENGSCI_255 CHEMMAT_315ENGSCI_263ENGSCI_314ENGSCI_331ENGSCI_343ENGSCI_363 ENGSCI_311ENVENG_244ENVSCI_101ENVSCI_201ENVSCI_301ENVSCI_310 ECON_191FINANCE_251FINANCE_261FINANCE_351FINANCE_361FINANCE_362 FOODSCI_201FOODSCI_301FOODSCI_302FOODSCI_303FTVMS_100ENVSCI_311GEOG_101GEOG_102 GEOG_104 GEOG_105GEOG_201GEOG_202GEOG_205GEOG_210GEOG_250 GEOG_261GEOG_305GEOG_302GEOG_307 GEOG_312GEOG_315GEOG_317GEOG_320GEOG_330GEOG_331GEOG_332 GEOG_260GEOG_262GEOG_351GEOLOGY_103 GEOLOGY_104GEOLOGY_201GEOLOGY_202 GEOLOGY_203GEOLOGY_204 GEOLOGY_205GEOLOGY_301GEOLOGY_302GEOLOGY_303GEOLOGY_304GEOLOGY_305GEOLOGY_306GEOLOGY_340 GEOLOGY_361GEOLOGY_372GEOPHYS_330 ECON_221ECON_352FINANCE_383INFOSYS_110INFOSYS_220INFOSYS_222INFOSYS_224INFOSYS_280INFOSYS_320INFOSYS_321INFOSYS_322INFOSYS_330INFOSYS_323INFOSYS_338INFOSYS_339 JAPANESE_130JAPANESE_150JAPANESE_231JAPANESE_232 LAW_131LAW_299MARINE_202 ENGLISH_121ENGWRIT_101 FOODSCI_304FRENCH_101 LINGUIST_100MATHS_101ACADENG_102ANCHIST_103 BUSINESS_309ECON_241ENGSCI_213ENGSCI_313 ENGSCI_391 ENVENG_333ESOL_100ESOL_101 FTVMS_101GEOG_318 GERMAN_101HISTORY_103INFOSYS_341 INNOVENT_202INNOVENT_303JAPANESE_131LINGUIST_103 MATHS_102MATHS_108ESOL_102MATHS_150MATHS_162GEOPHYS_331 MATHS_190MATHS_202 MATHS_208MATHS_250MATHS_253MATHS_255MATHS_260MATHS_269MATHS_270MATHS_315MATHS_320 MATHS_302MATHS_326MATHS_328MATHS_332MATHS_333MATHS_340MATHS_361MATHS_362 MBCHB_203MBCHB_205MBCHB_206MBCHB_210MBCHB_303MBCHB_305MBCHB_306MECHENG_201MECHENG_211MECHENG_222MECHENG_223MECHENG_235MECHENG_234MECHENG_236MECHENG_242MECHENG_270MECHENG_312MECHENG_311 MECHENG_313MECHENG_322MECHENG_325 MECHENG_334MECHENG_340MECHENG_370 HLTHPSYC_122MATHS_153MBCHB_313MEDSCI_142MEDSCI_201 MEDSCI_202MEDSCI_203MEDSCI_204MEDSCI_205MEDSCI_206MEDSCI_301MEDSCI_302 MEDSCI_303MEDSCI_304 MEDSCI_305MEDSCI_306MEDSCI_307MEDSCI_308MEDSCI_309 MEDSCI_310MEDSCI_311MEDSCI_312MEDSCI_313MEDSCI_314MEDSCI_316MGMT_211MGMT_101 MKTG_201 MKTG_202MKTG_301MKTG_303 NURSING_105NURSING_199NURSING_201NURSING_202NURSING_301 OPTOM_110OPTOM_161OPTOM_215OPTOM_165 OPTOM_220OPTOM_262OPTOM_260OPTOM_265 OPTOM_365PHARMACY_101PHARMACY_199PHARMACY_201PHARMACY_202PHARMACY_205PHARMACY_301PHARMACY_303MATHS_363PHIL_100PHIL_101 PHIL_105PHIL_102PHIL_222 HISTORY_102 PHIL_152PHYSICS_102PHYSICS_107PHYSICS_108 PHYSICS_111PHYSICS_120PHYSICS_130PHYSICS_140 MARINE_302 MEDSCI_315MEDSCI_317 NURSING_302OPTOM_270OPTOM_375PHARMACY_304PHYSICS_150 PHYSICS_160PHYSICS_211PHYSICS_213 PHYSICS_220PHYSICS_230PHYSICS_231PHYSICS_240 PHYSICS_250PHYSICS_251PHYSICS_261PHYSICS_315PHYSICS_260PHYSICS_280PHYSICS_325PHYSICS_326PHYSICS_340PHYSICS_350PHYSICS_354PHYSICS_355PHYSICS_356 POPLHLTH_101POPLHLTH_102 POPLHLTH_111POPLHLTH_202POPLHLTH_204POPLHLTH_206POPLHLTH_210EDUC_115PHIL_103PSYCH_108 PSYCH_109PSYCH_201 PSYCH_202PSYCH_203PSYCH_204PSYCH_207 PSYCH_208PSYCH_303PSYCH_305PSYCH_306PSYCH_309PSYCH_310PSYCH_311 PSYCH_313PSYCH_322PHYSICS_390 SCIGEN_101SOCIOL_100SOFTENG_206SOFTENG_211 SOFTENG_250SOFTENG_251SOFTENG_254SOFTENG_306SOFTENG_325SOFTENG_350SOFTENG_351SOFTENG_364 SPORTSCI_101SPORTSCI_103SPORTSCI_201SPORTSCI_202SPORTSCI_203SPORTSCI_301SPORTSCI_204 SPORTSCI_303SPORTSCI_304ACADENG_100ACADENG_101BIOSCI_396 CLASSICS_110EDUC_121GEOG_103GEOG_322GEOG_325 LINGUIST_101MAORI_130PHYSICS_103POLITICS_106 POPLHLTH_211PSYCH_317PSYCH_326 PSYCH_364SCIGEN_201SOCIOL_101SOCIOL_103SOCIOL_105SOCIOL_216 SOCIOL_220SPANISH_104 SPORTSCI_105SPORTSCI_206INTBUS_202MGMT_223 MKTG_306STATS_101STATS_108STATS_125 STATS_150STATS_201STATS_208STATS_210STATS_220STATS_255 STATS_301STATS_302STATS_310STATS_320STATS_325 STATS_326STATS_330STATS_331STATS_340ANTHRO_206BIOMENG_341 CHINESE_101 COMLAW_311COMLAW_314COMPSYS_304 ECON_341ECON_381EUROPEAN_100GEOG_207GEOG_334GEOG_360 INTBUS_201JAPANESE_331 LAW_399MECHENG_352MECHENG_371MGMT_314OPSMGT_255OPSMGT_258OPSMGT_357PHIL_216PHIL_266PHYSICS_391 POLITICS_109 POPLHLTH_203POPLHLTH_212 POPLHLTH_307PSYCH_319SOFTENG_370 SPORTSCI_305STATS_370STATS_380 WINESCI_201
Fig 2. Student Course Network.
A network representing the communities, or fields, of courses formed by studentsco-enrolling in course at the University of Auckland. Each node represents a courseoffered by the university, while links between nodes indicate instances where studentstook those two courses together within their undergraduate degree. Communities wererevealed in a two step process. Firstly, edges were filtered so only those with an RCPvalue over 1 were included. Secondly, the Map Equation software package [63] was usedto highlight the underlying fields. The revealed fields are labeled in Fig 3, and representthe various academic fields that students in our sample were enrolled in.
PLOS ngineeringPhysics-MathsGeography-Geology Finance-MarketingComputer Science Medical ScienceBiological ScienceStatistics PsychologyLiberal Arts Chemistry PharmacyOptometrySports ScienceFood Science Population HealthSurgeryNursingJapanese LawAncient HistoryChemical MaterialsChinese
Fig 3. Course Community Network.
The above directed network represents the network seen in Fig 2, only the links withincommunities (i.e. links between courses belonging to the same community) and betweencommunities have been split by gender and aggregated. Odds ratios comparing thelikelihood of a female student taking a course in community i and community j wereformulated, with the resulting values used as edge weights. The communities werelabeled based on the range of courses that it is comprised of. Edges where femalestudents were more likely to take a course in community i and community j are colouredblue, while edges where male students were more likely to take a course in community i and community j are coloured red. When considering the flow between a pair of nodesconnected by two edges, the direction of flow is outward following the link in a clockwisedirection. The network shows that transverse movements from fields such as computerscience and physics-maths to other domains tend to be female dominated, whilstmovements into these fields are more male dominated. PLOS iological ScienceChemical MaterialsChemistryComputer ScienceEngineeringFinance−MarketingFood ScienceGeography−GeologyMedical ScienceNursingOptometryPharmacyPhysics−MathsPopulation HealthPsychologySports ScienceStatisticsSurgery B i o l og i c a l S c i en c e C he m i c a l M a t e r i a l s C he m i s t r y C o m pu t e r S c i en c e E ng i nee r i ng F i nan c e − M a r k e t i ng F ood S c i en c e G eog r aph y − G eo l og y M ed i c a l S c i en c e O p t o m e t r y P ha r m a cy P h ys i cs − M a t h s P opu l a t i on H ea l t h P syc ho l og y S po r t s S c i en c e S t a t i s t i cs From To −2−101 log(OddsRatio) Likelihood of students taking a course in a community after taking a course in another community
Fig 4. Student Course Heat Map.
The above heat map represents the same underlying data as that which is used in Fig 3.The heat map makes clear the gender differences in the likelihood of students movingfrom one community to another. Orange areas represent instances where femalestudents were more likely to take a course in community i after taking a course incommunity j . Purple areas indicate male students were more likely to take a course incommunity i after taking a course in community j . Areas that are white or emptyindicate no significant relationship. Male students were consistently more likely to takecourses in Computer Science and Physics-Maths after taking courses in each othercommunity. Female students tended to be more likely to take courses in life sciencesubjects (e.g., Biological Science and Psychology). PLOS ighSchoolLowHighSchoolMidHighSchoolTop AP1 LowAP1 MidAP1 TopPLS LowPLS MidPLS Top Gender
FemaleMale
High School to Key Stage One Physics Courses
Fig 5. Student Progression Alluvial.
An alluvial plot showing the progression of students from high school to universityphysics split by achievement bands. PLS (Physics for Life Sciences) and AP1(Advancing Physics 1) represent the two main groups of physics students in our data.As shown in the alluvial plot, PLS is more popular than AP1, especially at theintersection of top achievers and female students.
PLOS igh School − Low High School − Mid High School − Top −
50 0 50 −
50 0 50 −
50 0 50
Change in Percentile Rank D e n s i t y o f r a n k c h a ng es p e r g r oup GenderFemaleMale
From High School going to AP1
Distribution of Rank Change by Gender
Fig 6. Distribution of Rank Change by Gender and High SchoolAchievement Group.
The above density plots show the distribution of rank change going from high school toAdvancing Physics 1 (AP1). Purple represents the distribution of rank changes for malestudents, while orange represents female students. The dotted vertical line show themedian rank change per group. On average, low achieving female students went down 6ranks compared to their male counterparts, while middle achieving female students wentdown 9 ranks relative to their male counterparts. There was no significant genderdifference in rank change for top achievers.