Publications

Culture, Tourism and the Centre for Education Statistics

Ability in mathematics and science at age 15 and program choice in university: differences by gender

Data and Methods

Warning View the most recent version.

Archived Content

Information identified as archived is provided for reference, research or recordkeeping purposes. It is not subject to the Government of Canada Web Standards and has not been altered or updated since it was archived. Please "contact us" to request a format other than those available.

Archived

This page has been archived on the Web.

Data – Youth in Transition Survey (YITS), Cohort A

This report uses the linked Youth in Transition Survey (YITS)-Programme for International Student Assessment (PISA) data, which contains data from the Canadian component of the PISA survey from 2000 when youth were aged 15, and longitudinal YITS data to age 25 (Cycle 6). Using these data allows for the linking of characteristics during adolescence with educational outcomes in young adulthood.1

Data from Cycle 1 to Cycle 6 (age 25) were used to take into account as many youth as possible with first PSE programs. Other work (see Finnie and Childs 2010 and OECD 2010), which measured program choice only up to age 21 (or Cycle 4), benefitted from a larger sample size; however, the disadvantage is a potential loss of information. For example, Finnie and Childs (2010) found that by age 21, 25% of youth had not yet started PSE. In the current sample extending to age 25, that figure drops to below 20% (see Table 1A).  Moreover, given the current interest in male and female differences, measuring to only age 21 may lead to the unnecessary exclusion of males who eventually go on to PSE beyond age 21. Finnie and Childs (2010) found that 31% of males and 19% of females had not started PSE by age 21:  a substantial gap, which decreases to about 24% for males and 12% for females by age 25 in the current sample (analysis not shown). Thus, by including up to Cycle 6, this paper is likely including more males who go into university than in some alternate work.  

Outcome: First University Program

The first university program is measured from Cycle 2 (age 17) through to Cycle 6, when the youth reached age 25. In this study, only university-bound youth are included because of the inherent difficulty in comparing programs across institution type. The non-university-bound youth are interesting and very important, but are left to future work.2

Using YITS’s program rosters from Cycles 2 through 6, a respondent’s first university program was determined using Classification of Instructional Program (CIP) codes for the first main field of study or specialization.  While a person’s first program type may not be their final program upon graduation (if they graduate), first program type does indicate a person’s initial interests out of high school. Switching of programs does occur for some youth during their time in university; however, some past work has found that the majority of youths’ first programs are those they remain in throughout five years of university. Finnie and Childs (2010) found, for example, that 38% of students that were not in science, technology, engineering and mathematics (non-STEM) fields either switched programs or left university by their fifth year in university, while the proportion who left one of these STEM programs was much less at 24%.

In this study, the program-type measure re-categorized the 13 primary groupings in CIP (Classification of Instructional Programs) 2000 into five categories. The five new categories, informed from some past literature (see Montmarquette, Cannings, and Mahseredjian 2002) are as follows: (1) Social Sciences (includes arts, education, humanities, social sciences, and law), (2) Business/Management/Public Administration, (3) Science/Mathematics/Computer Science, Engineering and Agriculture, (4) Health, Parks, Recreation and Fitness, and (5) Other.  For sake of parsimony, the five category titles are shortened to: (1) Social Sciences, (2) Business, (3) STEM, (4) Health, and (5) Other. (See Appendix Table 1 for a description of the re-categorization).3

Table 1A shows the distribution of the programs for all youth including those who never went to university by age 25 (including those who went to some other type of PSE) and for just the university-bound population across both the mathematics and science samples.4

Table 1A First program type

Social science programs are the most common first programs in university. Next most common are the STEM programs, followed by Business, Health, and Other categories. Notably, around 20% of young adults had not started a first PSE program by December 2009 when they were aged 25.

Among young women, the most common first university programs are in the Social Sciences: 50% of females choose these types of programs as their first at university (Table 1B). This is much higher than the proportion of males going into the Social Sciences, at only 32%. For young men, the most common first university programs are in STEM fields, at 44%. In contrast, only 20% of females’ first university programs are in STEM fields. Meanwhile, the proportion of both men and women who enter Business programs is basically identical at 14%. Women in turn appear to be significantly more likely than men to have a first university program in the field of Health.5

Table 1B First program type for university bound students, by gender

Mathematics and Science PISA Scores at Age 15

In the 2000 survey when the youth were aged 15, reading skill was the main focus and so only a subsample of randomly chosen respondents responded to a much smaller number of mathematics and science items than the entire sample that was assessed for reading literacy (OECD 2010; Statistics Canada 2004). In total, 32 mathematics questions and 35 science questions were included in the PISA 2000 assessments. Nonetheless, the mathematics and science scales in PISA were developed to measure mathematical and science literacy of 15 year olds, and through the pairing with the longitudinal aspect of YITS the relationship between mathematical and science literacy at age 15 can be examined with numerous outcomes into young adulthood in Canada.

Mathematical literacy is used in the current context to “indicate the ability to put mathematical knowledge and skills to functional use rather than just mastering them within a school curriculum”.  Meanwhile, science literacy is defined as “the capacity to use scientific knowledge, to identify questions, and to draw evidence-based conclusions in order to understand and help make decisions about the natural world and the changes made to it through human activity” (Bussière, Cartwright, Crocker, Ma, Oderkirk and Zhang 2001: 86). 

This study uses average levels of mathematics and science test scores from Cycle 1 when the youth were age 15. Proficiency levels in mathematics and science were also created and used to form a measure tapping into high levels of mathematical and science ability: youth defined as having ‘high’ mathematical or science ability are in the 4th proficiency level or higher.6  In general, youth at higher mathematical and science proficiency levels are better able to integrate and use mathematics and science concepts. Their reflection and insight into concepts as well as mathematical and science problem solving skills are stronger than youth at lower levels.

Control Variables

Several variables are included to take into account of factors that may affect the decision to enter a particular program in university. These controls therefore are less affiliated with ‘access’ to PSE, instead, access is taken as a given because the sample under consideration has started a program at university prior to age 25.7

Control variables are grouped into three main categories:

  1.  Factors related to student achievement and educational interests in high school (measured at the age of 15 unless otherwise specified) — includes marks in either mathematics or science, an index of self-rated mathematical ability8, reading ability measured via the PISA reading tests, grade level when the PISA tests were administered, frequency of science lab usage, and an index measuring sense of control/mastery.
  2. Demographic Characteristics include gender, province of residence at age 15, whether the respondent lived in a rural area at age 15, and whether the respondent and the parents were born in Canada.
  3. Parental/family influence –Several familial and/or parental factors are included: parental education, the degree to which parents know their child’s teacher, and two factors that measure the parent–child relationship, namely, the frequency that a parent and teen talk about the teen’s future career and education, as well as the frequency of discussions of current political and social issues between parents and teens.9

Four of the covariates at the student level are of particular interest: marks in mathematics or science, self-rated mathematical ability, the frequency of science-lab usage, and reading ability. Marks, which are closely linked to the objective PISA measures, measure the degree of potential difficulty youth have with either mathematics or science (at least the difficulty as measured through mathematics/science course-taking), while self-rated mathematical ability and frequency of science-lab usage potentially measure one’s interest and degree of confidence in the subject. However, marks may not be accurately measuring one’s ability but instead may be confounded by a myriad of factors including self-discipline, one’s ability to better navigate the education system in high school, and teacher perception (Cornwell et. al 2011).  Meanwhile, self-rated aptitude in a subject can be confounded by gender, because girls may not rate their aptitude in a particular subject as high as it is in reality, whereas boys often have the tendency to exaggerate their abilities (Cech, Rubineau, Silbey, and Seron 2011; Correll 2001). 

Thus, achievement test scores, such as the PISA scores, may be the preferred option to measure aptitude, because it may offer a truer measure of one’s’ ability in mathematics and science, and  may not be as affected by issues such as self-discipline, aptitude inflation/deflation, or differential treatment in the school system based on gender. Nonetheless, each of these factors is important to include because they could potentially alter the relationship between objective test scores and choice of university program, and could also operate differently for young men and young women.  Finally, reading ability is included to help measure students’ overall academic ability beyond math and science aptitude, and it is an important indicator predicting program choice in its own right (see see Finnie and Childs 2010; and OECD 2010).

Method

Three research questions are asked in this report.  First, which programs are students with high mathematical and high science abilities choosing when they first enter university?  Second, does gender matter? That is, do males and females of equal mathematical and science ability choose similar programs in university? Third, is the relationship between gender, mathematical/science ability and program choice affected by other factors? The analysis will include descriptive and multivariate methods using multinomial logistic regression.

The analysis will proceed as follows: First, average levels of mathematics/science PISA scores are compared across first university program for the total university-bound sample, and also separately by gender (Table 2). Second, the intersection of gender and levels of mathematical/science (high vs. low) ability is compared with first university program (Table 3). Third, the relationship between mathematical/science ability and gender, predicting first university program choice, is assessed through a series of multivariate multinomial regression models. The first model, the bivariate model, includes only the mathematical/science ability by gender measure. The second model includes all controls, except for mathematics/science marks and self-assessed mathematical ability, while the third model (the full model) adds in these latter two measures (Table 4 and 5).10 Multinomial logistical regression is useful for the current analysis, since it allows for multiple categories in the dependent variable without any real hierarchy, such as multiple program types (other work examining program choice has used this same technique — see Finnie and Childs 2010; Leppel et. al 2001; OECD 2010; Zarifa 2012).11

The analysis is restricted to the university-destined youth only; the sample under consideration does not include those whose first PSE program is in a non-university setting and also those who do not go on to a PSE program before the age of 25.12 The choice was made to consider only university-bound youth because of comparability challenges between programs at the university and non-university levels. For example, engineering programs are offered at both colleges and universities but can be quite different, with the former being oriented more toward practical job skills. Furthermore, in multinomial models, the comparison between university program choice (STEM for example) and not going on to PSE is much less interesting or relevant than the comparison between STEM and another type of university program.13 In all analyses, the appropriate survey weights are used, as well as the corresponding bootstrap weights. Stata’s (Version 11) specialized survey procedures are used for all analyses.

Notes

  1. See especially OECD (2010) for a description of these linked PISA-YITS data as well as their usefulness.
  2. The non-university bound population made up approximately 35% of the total sample, while the university bound made up 46% and the group who had not attended any PSE by December 2005 made up 19%.
  3. For a list of the constituent CIP series and subseries associated with the CIP primary groupings, refer to Statistics Canada (2005), pages 7–9.
  4. Estimates are very similar (two percentage points or less separate them) across the two subsamples so, unless otherwise specified, results are discussed relative to the mathematics subsample.
  5. These trends using YITS Cohort A are also in line with statistics from the Post-Secondary Student Information System (PSIS) except that since 2004, in PSIS, the most common university program for men is the Social Sciences, followed by STEM. Comparisons between PSIS and those data presented in Table 1B are difficult though, because the current study uses the ‘first’ university program whereas PSIS provides data on university programs reported in the past year.  Also, the current report uses a cohort of youth aged 15, in 2000, that entered university for the first time prior to their 25th birthday in 2010, while the estimates from PSIS are based on 18- to 24- year-olds in each year, from 1992 to 2009. Nonetheless, the comparison between YITS-Cohort A and PSIS is interesting, because using PSIS allows for a broader perspective of university program trends in Canada over time.  The PSIS results are available upon request.
  6. Mathematics and science proficiency levels were not officially derived in 2000 because, in that year, only proficiency levels for reading were derived since it was the main focus. Instead, in this paper, proficiency-level cut points for mathematics are measured from 2003 (see Bussière, Cartwright, Knighton, and Rogers 2004), while proficiency-level cut points for science are from 2006 (see Bussière et al. 2007). These two years (2003 and 2006) were the closest available to the 2000 mathematics and science assessments.
  7. Approximately 46% of the sample went into university, while 35% went into some other type of PSE, and 19% had not started any PSE by age 25.
  8. Self-rated mathematical ability was from Cycle 2 (age 17). It was not measured in Cycle 1.
  9. With regard to parental factors, parental occupation is also a potential factor in choice of major in university (see Leppel, Williams, and Waldauer 2001). However, using these same data, OECD (2010) found no significant link between parental occupation and university program choice. As a result, the decision was made to not include it in this analysis. Parental education was retained; however, because it has been found to figure more prominently in university program choice (OECD 2010; Zarifa 2012).
  10. The analysis was done this way to isolate the effects of mathematics/science marks and self-assessed mathematical ability. These two are both correlated, more than other covariates, with mathematics/science PISA scores and also in some subsequent analyses were shown to be robust predictors of university program choice.
  11. Only results from models ran with the interaction between mathematics/science score and gender are presented. Models were estimated with main effects for mathematics/science scores and gender, and are available upon request. This supplemental research showed that high mathematics/science scores led to a greater chance of entering STEM fields; however, this effect was removed after the inclusion of gender and other important controls. The main effect for gender indicated that females are much less likely to enter STEM fields than others.
  12. In some additional analyses (not shown), which included the non-university-bound as well as those youth who had not started any PSE prior to age 25, the basic relationship between mathematical/science ability and gender and university program choice did not substantially change. The results are available upon request.
  13. Other work also deals with these different postsecondary populations separately. See Boudarbat and Montmarquette (2009) and Boudarbat (2008).
Date modified: