The intersectionality of postsecondary pathways: the case of high school students with special education needs.

Robson, Karen L. ; Anisef, Paul ; Brown, Robert S. 等

THE EMPHASIS ON postsecondary education (PSE) as a pathway to economic security is not a new idea, but certainly one that is being stressed in the discourse around improving the life chances of individuals. Internationally? numerous scholars have examined how access to PSE is inextricably linked to various fixed characteristics of individuals, such as race, family socioeconomic background, and parental education, and as such, how these predefined characteristics function to perpetuate cycles of advantage and disadvantage. Access to PSE is understood to be a key marker in ensuring national economic competitiveness in a global sense, and also an indicator of equity within a given nation (Finnie and Pavlic 2013). (1)

Increasingly, attention is being focused on another marginalized group of students: students identified with special education needs (SEN). Data from the 2006 Participation and Activity Limitation Survey have revealed an increased number of youth who have been identified as demonstrating "learning limitations." They constitute 2.5 percent of the population aged 15 and older and represent a 40 percent increase in the number of adults identified as having learning disabilities since 2001 (Brennan 2009). Brennan (2009) also reports that being identified with a learning disability not surprisingly--affects education choices. Findings from Human Resources and Social Development Canada (2006) show many discrepancies in educational outcomes for those with and without disabilities. For example, in 2006 around 8 percent of individuals identified with disabilities had bachelor's degrees compared to nearly double that (15 percent) of people without disabilities. These findings are consistent with others from the United States and Canada, which report relatively small numbers of students with SEN transitioning to PSE (Ferguson 2008; Mackenzie 2009; McCloskey et al. 2011; Newman, Davies, and Marder 2003; OECD 2003; Pumfrey 2008; Shaw, Madaus, and Banerjee 2009).

With an augmented focus on increasing the numbers of youth accessing PSE--as illustrated in strategies such as in Ontario where provincial government targets have been set to 70 percent PSE attainment for all adults by 2020 (Ontario Ministry of Finance 2011)--understanding what helps and what hinders access to PSE for students with SEN has become an urgent priority. While a breadth of knowledge exists that focuses on the success of students with SEN, once they reach university and college (for a detailed review, see Boyko and Chaplin 2012), there is a dearth of literature on what factors enhance or hinder access to PSE for such students. This paper addresses this important gap in the literature. In this exploratory study, we ask what factors impact on PSE access for students identified with SEN and how do these compare to students without SEN? We will explore these questions using 2006 data on 17-year-old students in the Toronto District School Board (TDSB).

Special Education Needs, the Life-Course Perspective, and Intersectionality

As a currently constructed parallel system within the realm of public education, the merits, purpose, and function of special education have been hotly debated (Mitchell 2010). The educational profiles of students identified with SEN have been well researched at the school level (K-12) and a growing number of studies have been conducted at the PSE level.

Students who are supported by special education services throughout their tenure in school may experience diminished opportunities to engage in both the school culture, their studies, or with their peers. The proportion of students taught within congregated (segregated) special education classrooms, where there is often a reduced curriculum and lack of access to peers and school participation (Mitchell 2010), varies dramatically across public boards in Ontario (Brown et al. 2013). Although not all students identified with SEN receive segregated support, the very distinction of difference can lead to increased experiences of exclusion.

Given the differential treatment that SEN students may experience, a life-course orientation is employed in this study. Berger and Motte (2007) emphasize the utility of the life-course perspective in capturing the complexities of the journey to PSE and in better understanding the access, persistence, and completion barriers that face many individuals and groups (see also Seabrook and Avison 2012). Specifically, they emphasize two points: (1) factors that determine PSE access (and completion) lie in the individual's life circumstances and are already present in early childhood; and (2) social factors such as socioeconomic status interact with individual characteristics and do so in different ways throughout the life course. This conceptual framework recognizes the importance of three essential ingredients in examining life-course transitions and trajectories--time, social structures (e.g., gender, social class, disability status, ethnic and racial minorities, school-level factors), and personal agency (e.g., educational choices, student engagement). Structure and agency work together in helping shape the life course. When examining PSE transitions through a life-course lens, it is important to view the individual as an agent who constructs a personal pathway within the larger context of social, cultural, and economic forces.

While the life-course orientation underscores the importance of accounting for the individual and combined effects of social structure and personal agency, studies that have been conducted to date rarely examine the impact of multiple social factors (e.g., ethno-racial status, gender, immigrant status, social class) and their intersections on life-course outcomes (McCall 2005). By way of illustration, prior studies have rarely considered how race/ethnicity and gender jointly differentiate PSE pathways; instead they have examined race/ethnicity or gender as if they were separate dimensions of social stratification (Warner and Brown 2011). By contrast, an intersectionality approach systematically examines the interactive influences of race/ethnicity and gender on social mobility across the life course (Mullings and Schulz 2006). That is, an intersectionality approach begins with the premise that forms of oppression (e.g., racism, sexism) overlap, and thus posits that the consequences of race/ethnicity and gender cannot be understood sufficiently by studying these phenomena separately. We use both the life-course approach and intersectionality to conceptualize how the PSE pathways of SEN students are influenced by broad social forces and choices (2) made by individuals.

Determinants of Transition to PSE

The predictors of transition to PSE in Canada have been documented in detail by De Brouker (2005) and Cheung (2007). A significant predictor of youth going on to PSE is the financial status of their family. In fact, young people from high-income families are two to three times more likely to go to university than those from low-income families. In terms of college, however, there are almost no differences in enrollments between high- and low-income students. However, there are other important factors--some that appear to be more influential than family income--that act to shape the PSE trajectories of young people in Canada. Much has been made of single parent status in the past, but both reviews cited above indicate that family structure is largely a proxy for family income, and that once controlled for, youth from single-parent households are no less likely to transition to PSE that youth from two-parent families.

As noted in De Brouker's (2005) review, parental education has been found to be more strongly tied to PSE transition in Canada than family income. In other words, regardless of family income, parents who themselves have PSE are more likely to have children that go on to PSE than are parents who do not have PSE. Finnie et al. (2008, 2010) also found that it was parental education that had an influence on other strong predictors of going on to PSE, such as high school grades, attitude toward education, and educational aspirations. Similarly, parental expectations have also been shown to be an important factor in predicting post high school choices, as youth who perceived that their parents expected them to go on with their schooling were significantly more likely to do so than those who did not have such a perception (Barr-Telford et al. 2003).

Inextricably tied to expectations and grades is program of study (i.e., streaming or academic tracks established in secondary school), which has also been found to be a predictor of PSE transition, with students in applied streams less likely to go on to PSE than students in academic streams. In terms of gender, female students are more likely to pursue PSE than male students (King and Warren 2006) while racialized students--particularly first- and second-generation immigrants--have higher rates of transitioning to PSE, have higher educational aspirations, and are more likely to have parents who have PSE and high educational expectations for their children than nonracialized students (Krahn and Taylor 2005).

SEN and the Predictors of PSE Transition

In this section, we review studies that explore how the factors that are identified as predictors of PSE transitions are associated with the presence of SEN identification. On the surface, barriers to PSE access facing students with disabilities or SEN include limited secondary academic preparation and completion, extensive financial expenses connected to specialized supports, limited employment opportunities, as well as inaccessible infrastructure and instruction styles (Kirby 2009). However, more implicit barriers are embedded within the construction of impairment and disability, which could lead to reduced student engagement, further marginalization from education opportunities, and a reduction of postsecondary access.

Historically, special education classes have long experienced an overrepresentation of racial minorities and male students as well as students from lower-income households, particularly among students identified with higher incidences of exceptionalities or impairment (Artiles et al. 2010). Thus, low family income is associated with both a decreased likelihood of transitioning to PSE (university in particular) and an increased likelihood of being identified with SEN.

As mentioned above, females are more likely than males to go on to PSE. Across North America, gender also appears to be an important factor in the identification of students with SEN. In a number of studies, boys were found to be significantly overrepresented among SEN students (Brown and Parekh 2010; Oswald et al. 2003; Wilkinson 2008). Thus, males are both more likely to be identified with SEN and less likely, regardless of SEN status, to go on to PSE.

In terms of race, data from the United States demonstrate notable overrepresentation of racial minorities within exceptionality categories, particularly within higher incidence categories associated with reduced intellectual capabilities (De Valenzuela et al. 2006; Skiba et al. 2006). Toronto-based research has also uncovered notable incidences of disproportionate representation in special education including an overrepresentation of both white and black students across SEN categories (Brown and Parekh 2010).

In addition to incidences of disproportionate representation of race in special education, similar concerns have been raised regarding the increased proportion of students identified with SEN that are raised in lower-income households (Halfon et al. 2012). As some researchers have suggested, children reared in poverty often experience greater referrals to special education and SEN identifications.

Therefore, students identified with SEN are more likely to be represented among groups that already have a lower likelihood of transitioning to PSE, suggesting that their SEN status may act as an additional "invisible" barrier that is separate, but inextricably linked to being identified as having SEN. In the following section, we test this hypothesis.

DATA AND METHOD

The data employed in this study are derived from a survey (often called the "Student Census") administered from November 6 to 10, 2006, in all TDSB secondary schools and all Grades 7 and 8 in elementary schools. A total of 289 Toronto schools were involved. After all the data were processed and verified, a student census database was created consisting of 34,219 students in Grades 7 and 8 and 71,222 students in Grades 9 to 12. The use of Board enrollment figures of October 31, 2006, revealed a high response rate of 92 percent for Grades 7 and 8 and 81 percent for Grades 9 to 12. The response rate for 17-year-olds, in particular, was 74 percent. Student surveys were administered in classrooms and gathered information on who the students were and how they felt about their school and personal lives. To ensure student confidentiality, teachers were instructed to ask their students to leave their identification numbers, but to black out their names before placing their completed surveys in a return box at the front of the classroom. While the survey was confidential, it was not anonymous. Student identification numbers were linked to other centrally available data sources, such as the TDSB student information system, standardized test results from the Ontario Education Quality and Accountability Office (EQAO), and student report cards. Access to data was granted through the TDSB research department to which one of the authors belongs.

The specific subset of data employed for these analyses is based on 14,048 seventeen-year-old students that were surveyed in the census conducted in Fall 2006. These students were age-appropriate for Grade 12, the age in which most Ontario students start their transition to PSE. Added to these data were other pieces of information: data from the Fall 2006 Student Census completed by Grades 9-to-12 students in the TDSB on a range of socioeconomic, demographic, and attitudinal variables (see Brown 2009); data from the 2001 Canadian Federal Census on household income (matched by postal code to the Census Dissemination Area of around 300 households); (3) and data on postsecondary applications and confirmations as supplied to the TDSB by the Ontario University Applications Centre (OUAC) and Ontario College Applications Centre (OCAS). (4) Because it has been found that students will apply over multiple years (see Sweet et al. 2010), OUAC and OCAS information from the 2007, 2008, and 2009 application cycles were also merged with the data set.

Dependent Variable--College and University Confirmations

This limited cohort study allows us to examine the PSE pathways of 17-year-olds over three years in the TDSB (to the end of 2009). We examined the following three possible pathways: (1) confirmed an offer of admission to an Ontario university; (2) confirmed an offer of admission to an Ontario college; or (3) neither of these two options. (5) Just over half of the sample (50.7 percent) confirmed university acceptance, while 15.2 percent confirmed college acceptance. (6) The term "confirmed acceptance" applies to the situation where a student has applied to a college or university, been accepted, and has accepted the offer. It is more substantial than a simple acceptance at a university or college--it implies the additional intentionality of the student to actually attend.

Independent Variables

Our choice of independent variables reflects both our research questions and the factors that previous literature have identified as critical in understanding the PSE pathways of youth in Canada. Our focus is on how the factors of determining PSE choices may differ according to SEN status. Special education is intended to address the academic needs of students deemed as exceptional or who have been identified by educators as having "special needs that require supports beyond those ordinarily received in the school setting" (Ontario Ministry of Education 2013). Students' special educational needs can be met through accommodations and/or program modifications. In Ontario, there are two main categories of SEN: (1) students who have an Individual Education Plan (IEP) without undergoing a formal identification process, and (2) those who have 1 of 12 exceptionalities, designated through an Identification, Placement, and Review Committee procedure involving diagnostic assessments, usually administered by a school or external psychologist. The operationalization used in this study uses the EQAO definition of SEN that excludes students identified as Gifted: that is, those students with an active exceptionality or an active IEP. (7)

The proportion of students with SEN (excluding gifted) in 2012 to 2013 is estimated as comprising 15 percent of the TDSB population, but the proportion has increased over time. In 2006 to 2007, the proportion was 13 percent, but differed somewhat from grade to grade. Most new exceptionalities and IEPs are assigned in elementary school, and so as new students entered high school they were much less likely to be identified as SEN (Brown 2010). The proportion of students with SEN who were 17 years of age was consequently at 12 percent, below the overall TDSB average. Moreover, the students who completed the Census were also less likely to be students with SEN, further reducing the proportion to 10 percent.

As mentioned above, the financial status of the family has often been touted as a major defining factor of the decisions young people make about PSE. It has also been found, however, that parental education can be even more of an influence on PSE trajectories (Davies, Maldonado, and Zarifa 2014). Therefore, we include both measures in the analysis. In the case of the economic status of the family, we use median neighborhood family income data from the 2001 Canadian census. These figures are by no means perfect, but do give us an idea of the family socioeconomic status based on the general neighborhood income characteristics. Any conclusions about the relationship between family income and PSE in this analysis must be drawn with extreme caution given these data limitations as the TDSB do not hold data on the actual incomes of individual families. In terms of parental education, the students were asked about the highest educational attainment of their parents. The data were recoded into a single variable to capture the highest level of education of either parent with possible response categories being "high school," "college," "university," and "don't know." The variable was dummy-coded so that 1 was equal to university or college (i.e., any PSE).

In terms of student characteristics, high school grades were measured from Grade 11 marks in percentages that were obtained from the administrative database. Attitude toward education was measured with a single Likert-type item stating "I enjoy school" with the response categories of "all of the time," "often," "sometimes," "rarely," and "never" and was reverse-coded so that higher numbers were associated with greater enjoyment of school. While educational aspirations held by the student as well as the student's perception of parental educational aspirations were identified earlier as being important determinants of PSE, the items measuring these variables in the data set were limited to those students that completed "Form B," that is, half the data set. (8) Including these variables would have diminished our sample by half and, more crucially, our subset of SEN students by half. Therefore, the inclusion of these specific variables was not undertaken. However, information on the academic stream of students in Grade 9 was available. A variable measuring academic and applied streams in Grade 9 was included in the analysis and was coded so that 1 was equal to being in the applied stream and 0 was equal to the academic stream.

Sex of student was measured using a variable in which males were coded 1 and females were coded 0. Immigrant generational status of the student was derived from information on students' regions of birth and where their parents were born. First-generation students were those who were born outside of Canada (as were their parents), second-generation students were born in Canada but had one parent born outside Canada, and third-generation students had both parents born in Canada (Anisef et al. 2010). As race has been found to be associated with diagnosis of SEN, we have retained the original coding of the variable in the data, which distinguishes the following groups (self-identified): White, Black, East Asian, South Asian, Southeast Asian, Latin American, Middle Eastern, Mixed, and Other. We also control for school size as measured through a series of categories: [less than or equal to] 100,101 to 200, 201 to 500, 501 to 1,000,1,001 to 1,200,1,201 to 1,400, and >1,400.

Analytic Strategy

The first set of analyses is purely descriptive and also provides comparative information on the mean scores of students with and without SEN on the variables used in the multivariate analyses. These simple mean differences will highlight any discrepancies between students identified with SEN and others that exist in terms of their personal and family characteristics.

The second part of the analysis will employ random intercept models (a type of multilevel model), predicting college and university confirmations. Multilevel models are particularly fitting for examining research questions that inherently involve nested (or clustered) data. Our research questions focus on the effect of SEN status on the university and college confirmations of students within the TDSB, a question that necessitates the examination of "school effects." Students are "nested" within different schools and as such, a data structure like this one violates the ordinary least squares regression assumption of uncorrelated errors. Students within the same schools share many possible unobserved characteristics that influence their postsecondary trajectories. Multilevel modeling techniques allow us to adjust the model for such nested features and allow each individual school to have its own intercept. Such models also allow us to examine variance that is attributable to both individual and school levels.

Because the dependent variable is polytomous in nature, multinomial logistic regression models with random intercepts for school identifiers are used. The models were estimated in Stata 13 using the gllamm procedure. It should be noted that there are 100 different schools in the data set and that students identified with SEN are represented at all but three of these schools. Model-building occurred in five distinct stages as follows: the null (empty model), the bivariate model (with SEN status), the model with the remainder of independent variables of interest, the model that adds school size, and a final model with exploratory interaction terms.

In the final model, we add interaction terms between SEN status and other independent variables. These exploratory interactions are included to examine if combinations of characteristics can differentially impact upon the pathways of students identified with SEN compared to students without SEN. Such a statistical application allows us to operationalize different intersectional ties of students so as to examine how different combinations of fixed characteristics can impact upon their life chances (McCall 2005). As our research questions encompass the exploration of the potential of differential effects of known factors that predict PSE for students with and without SEN, interaction effects are a statistically sound way of examining such hypotheses. A statistically significant interaction will tell us if the effect of a particular independent variable on our PSE outcomes of interest is significantly different, according to SEN status. In other words, does SEN status moderate the path between factors that impact on PSE and PSE outcomes? Because, as the previous literature above has suggested, SEN status, gender, and race appear to be highly linked--particularly for Black males--we explore interaction terms between SEN status and sex and SEN status and self-identifying as Black, as well as three-way interactions between (1) SEN, self-identifying as Black, and being male and (2) SEN, self-identifying as Black, and being in the applied stream. We also explore if neighborhood income, parental PSE, and the applied stream differentially impact upon the PSE pathways of students identified with and without SEN.

Descriptive Results

Table 1 presents the descriptive statistics of the variables used in the analysis by SEN status. With listwise deletion on the variables of interest, our final sample size was 12,518. Of this total, 11,277 were students without SEN, while 1,241 (or about 10 percent) were students with SEN. From the descriptive table it is immediately evident that PSE pathways differ rather dramatically for students identified with and without SEN. Of the total sample, just over half confirmed university, but when we examine subsamples separately, we can see that among students identified with SEN, only 16 percent confirmed university, compared to just over 56 percent of students without SEN. The reverse is true in the case of college confirmations: a sample average of around 16 percent actually masks the fact that among SEN, the college confirmation rate was about a quarter of students identified with SEN, compared to just under 15 percent of students without SEN. And in the case of not confirming either college or university, the sample average is around 32 percent, but in the case of students identified with SEN, it is closer to 60 percent, which is double that of students without SEN.

Other differences between students identified with and without SEN are also evident in Table 1. The parental education is higher for students without SEN--nearly 61 percent of students without SEN have parents with PSE compared to just over half of students identified with SEN. Unsurprisingly, Grade 11 marks are on average 10 percentage points lower for students identified with SEN (60 vs. 70). Students identified with SEN report enjoying school slightly less than those without SEN. Students identified with SEN are also much more likely to be in the applied stream--67 percent compared to just 17 percent of students without SEN. As discussed above, the gender imbalance of students identified with SEN, found repeatedly in the literature, is also replicated here. While males make up about 50 percent of the general sample, they constitute 68 percent of the students identified with SEN. In terms of immigration status, students identified with SEN have characteristics that differ from the overall sample. They are less likely to be first-generation immigrants and more likely to be second or third generation. For example, 32 percent of SEN students are third-generation Canadians, but only 19 percent of the overall sample is third generation.

The literature has also suggested that SEN status is more prevalent among Black students (particularly males). In our sample, students self-identifying as Black comprise around 10 percent of all students, but nearly 19 percent of students identified as having a SEN. Students self-identifying as White also are more highly represented among students identified as SEN--at nearly 45 percent compared to being around 34 percent of the overall sample. Among the three Asian groups considered in Table 1, students self-identifying as East Asian, South Asian, and Southeast Asian are far less represented among students identified as SEN than their respective sample proportions.

In terms of school size, more students identified with SEN were in smaller schools (between 201-500 and 501-1000) than the overall average. Students identified with SEN were less represented in the bigger schools (1,001-1,200 and 1,201-1,400 students), while there was no major difference in the biggest schools in the sample.

PSE Confirmations

Multinomial logistic regressions are somewhat complex to present in an efficient manner. In Table 2, five separation estimations are presented. In each model, the omitted category is "no PSE confirmation" and therefore the columns for each model present the results for one of the remaining categories of the dependent variable--university confirmation and college confirmation. All independent variables are examined as fixed, that is, no random coefficients were included in the model. (9) All results are thus interpreted relative to the omitted category "no PSE confirmation." The fixed effects are presented as the first set of variables and are presented as odds ratios. The Level 2 variance is reported further down the table, as well as various fit statistics. It should be noted that likelihood ratio (LR) [chi square] tests indicated that each model was significantly improved by the inclusion of the additional variables in each step. The reported log-likelihoods also support our position that each set of variables led to model improvement.

The null (empty) model reveals an intraclass correlation (ICC or rho) of 0.142, indicating that about 14 percent of the variance in PSE confirmations can be attributed to differences between schools. When the SEN indicator is added in the second model, the ICC decreases to 13.0 percent, indicating that variance at the school level in PSE confirmations slightly decreases once SEN status is taken into account. The odds ratio for SEN status is statistically significant only in the case of university confirmations, where SEN status reduces the likelihood of university confirmation (relative to no PSE confirmations) by just over 80 percent (compared to students without SEN). Once the independent variables of interest are added in the third model, the ICC drops to just under 7 percent, indicating that much of the variance among schools can be accounted for through characteristics included in the model. SEN status continues to reduce the odds of university confirmation (relative to no PSE confirmations), although once the rest of the variables in the model are taken into account, the reduction in odds is around 27 percent compared to students without SEN. In this model, students identified with SEN have a 20 percent increase in odds in confirming college compared to students without SEN.

In general, the remaining independent variables performed as the literature had suggested: males were less likely to go on to either form of PSE, while both first- and second-generation immigrants were more likely to attend both forms of PSE, relative to third-generation Canadians. Parental education was a significant predictor of university confirmation, but not college confirmation--students whose parents had PSE were about 30 percent more likely to confirm university (compared to no PSE confirmation) than students whose parents did not have PSE. Grades were positively associated with confirming both types of PSE, while higher neighborhood median income reduced the odds of college confirmation and had no effect on university confirmations. Unsurprisingly, students in the applied stream were far less likely to confirm university than students who were in the academic stream (being in applied had no effect on college confirmations), while enjoying school was also associated with confirming university, but had no effect on confirming college. In terms of race, relative to students self-identifying as White (omitted category), self-identifying as Black had no significant association with either college or university confirmations. Students self-identifying as East Asian, Middle Eastern, South Asian, and Southeast Asian were more likely to confirm university than White students, while students self-identifying as South and Southeast Asian were more likely to confirm college than self-identifying White students.

In the fourth model, school size categories were added as exploratory variables of interest, with the largest school size (>1,400) used as the reference category. There is some suggestion that smaller school sizes reduced the odds of confirming PSE, although no clear pattern exists. In particular, schools with 201 to 500 students were considerably less likely to confirm PSE than students in very large schools. The addition of these variables did little to change the impact of the rest of the independent variables of interest.

In the final model, exploratory interaction terms between SEN status and other independent variables were included in the full model. Theoretically, all interactions between SEN status and the independent variables already examined are vehicles by which to test the exploratory hypotheses around the differential effects of PSE determinants on confirmation for students identified with and without SEN, but such a model would be highly complex and overwhelming. Instead, we have chosen to focus on interactions of fixed traits that are intimated in the literature to have differential effects on PSE outcomes.

It is important to note that now the main effects of the composite variables of the interaction cannot be interpreted and compared as they were in previous models because their inclusion in interaction terms changes the way in which main effects are interpreted. The interaction between SEN status and parental PSE achieved statistical significance in the case of college confirmations, suggesting that students identified with SEN whose parents have PSE themselves are more likely to confirm college than students identified with SEN whose parents do not have PSE. The predicted probabilities of confirming college for students identified with SEN whose parents have PSE was .28 compared to .20 for students identified with SEN whose parents do not have PSE. (10) As well, SEN status and median neighborhood income also achieved statistical significance for both college and university confirmations, suggesting that neighborhood characteristics matter in terms of the likelihood of students identified with SEN confirming PSE; those in better off neighborhoods (i.e., wealthier schools) were more likely to confirm a place. These two findings suggest that characteristics beyond the individual student differentially impact upon PSE confirmations according to SEN status.

In terms of three-way interactions, it was particularly striking that self-identifying as Black, being in academic courses, and being male did not impact PSE confirmations differentially for students identified with or without SEN. This result was surprising because self-identified Black males in applied courses are overrepresented among students identified with SEN. (11) However, the three-way combination of self-identifying as Black, being identified as having SEN, and being in applied courses did achieve statistical significance. In order to understand the three-way interaction, the interaction term itself as well as its composite (three) main effects and composite two-way interaction must be taken into account. Such interactions imply that the effect of [x.sub.1] x [x.sub.2] is moderated by [x.sub.3], however the relationship cannot readily be understood from simply looking at Table 2. Graphical illustrations are more useful in such cases. Figure 1 illustrates the predicted probabilities of confirming college, taking the three-way interaction into account. When calculated, (12) the combination of being identified with SEN and being in the applied stream can be seen to differentially impact upon college confirmations for self-identified Black students. In all bars in Figure 1, self-identified Black students have a higher probability of confirming college, except when they are both in the applied stream and have been identified with SEN--in which case their probability of confirming college falls below that of students who do not self-identify as Black.

[FIGURE 1 OMITTED]

DISCUSSION AND CONCLUSIONS

Canada has now reached a universal level of PSE participation, with over 50 percent of Canadians accessing tertiary institutions (Kirby 2009) and Ontario provincial targets set at 70 percent by 2020 (Duncan 2011). Although access to PSE is increasing overall, access to university continues to be largely inequitable, wherein members of historically marginalized groups continue to experience exclusion (Kirby 2009). The current stratification in confirmations reveals the socially constructed valuation embedded within postsecondary programs. The disparity in access indicates a societal preference for university participation over college; however, this preference is not as clear when looking at the postsecondary pathways of students identified with SEN.

The results of our study show that SEN status reduces the likelihood of confirming university, but can increase the likelihood of confirming college application, depending upon what factors are controlled. Our descriptive statistics revealed that even before considering PSE outcomes, various differences in students identified with and without SEN existed. Once these differences were accounted for through the controlling of determinants that the previous literature has deemed to be important factors in all students' PSE outcomes, we were able to see that SEN status did have an important impact on PSE confirmations. We were also able to test if individual and meso-level characteristics impacted students with and without SEN differentially with regard to their likelihood of confirming PSE, thereby activating intersectionality theory. Intersectionality theory understands individuals to occupy multiple identity traits that can--in their combination--influence opportunities and life trajectories (Collins 1990, 1998). We found that indeed these intersections of identity do matter for PSE outcomes.

The PSE outcomes of students identified with SEN are driven by external factors--their parents' PSE outcomes and their neighborhood incomes serve to significantly increase the likelihood of college confirmations for SEN students. The causal mechanism driving the finding around parental PSE is very likely to be the PSE experience of the parents' themselves, valuing the acquisition of higher education, but also focusing their children's sights on the more job-specific training that college programs tend to offer. The finding that the wealth of neighborhoods/schools also differentially influences the trajectories of students identified with SEN suggests that wealthier neighborhoods and schools improve the outcomes for students identified with SEN. The mechanisms through which this may occur is not entirely clear, but it is quite likely that wealthier schools have superior resources to support the learning of students identified with SEN and that the general mentoring environment in such schools may more strongly emphasize the pursuit of PSE, particularly going to university. It should also be noted that neighborhood wealth may also influence the type of SEN identification that a student receives, with students living in wealthier neighborhoods more likely to be classified with learning disabilities compared to mild intellectual disabilities (Parekh, Killoran, and Crawford 2011). PSE trajectories can be quite variant among students identified with a learning disability, but students living in poorer neighborhoods are more likely to be classified as having a mild intellectual disability that severely diminishes the possibility of pursuing any form of PSE.

An inspection of the intersection of self-identifying as Black, being identified with SEN, and being in the applied program revealed that this combination of traits significantly decreased the probability of college confirmation. In the TDSB data, self-identified Black students are overrepresented among students with SEN status and students in the applied program. Such characteristics, particularly being in the applied program, have been highlighted in recent reports by an education advocacy research organization in Ontario (People for Education 2014), demonstrating that socioeconomic status of neighborhood and school is inextricably linked to higher numbers of students in the applied stream in schools, which is also negatively associated with the overall achievement levels of students in those schools. In the TDSB, the streaming process is usually determined through decisions that have been made by Grade 8. The process of identifying SEN is also usually complete by Grade 8. Therefore, structures in place by the end of elementary school are closely connected to postsecondary pathways. It is perhaps no coincidence that this combination of characteristics is strongly associated with such an outcome. In our data, students self-identifying as Black, who also have SEN status and who are in applied courses, comprise a full 25 percent (IV = 204) of all self-identified Black students in our sample. Self-identified Black students in the TDSB are more likely to confirm college (25 percent) over university (23 percent), although "no PSE confirmation" is the largest outcome category (52 percent) for this group. This finding demonstrates that being identified with SEN and being in the applied stream has mixed PSE outcomes for students, with college being the PSE choice for those who do confirm. Having these two characteristics and also being Black depresses the probability of college confirmation. The gap between self-identified and non-Black students in the final bar of Figure 1 differs by 0.06 or 6 percent, which is particularly significant for a group that is already deemed to be "high risk."

Provincial governments across Canada have unreservedly emphasized increased PSE--particularly university--enrollments. Popular political and media discourse has focused on affordability (in response to rising tuitions) and on embracing different forms of technology to deliver courses (e.g., online delivery), but little discussion has emerged around the invisible barriers that prevent various subsections of the population from taking part. SEN status is but one of those barriers that currently reduces university confirmation, but increases college confirmation, provided that various other characteristics of the student and his/her environment are in specific permutations. "Invisible barriers," such as parental and neighborhood characteristics, have been demonstrated to play a significant role, although current discourse emphasizes a meritocratic playing field. Visible barriers, such as differential outcomes for students identified with SEN and students in the applied stream, intersecting with race, clearly demonstrate failings in crucial parts of the education system in Ontario.

References

Anisef, P., R.S. Brown, K. Phythian, R. Sweet and D. Walters. 2010. "Early School Leaving among Immigrants in Toronto Secondary Schools." Canadian Review of Sociology 47:103-28

Artiles, A., E. Kozleski, S. Trent, D. Osher and A. Ortiz. 2010. "Justifying and Explaining Disproportionality, 1968-2008: A Critique of Underlying Views of Culture." Exceptional Children 76(3):279-99

Barr-Telford, L., F. Cartwright, S. Prasil and G. Shimmons. 2003. Access, Persistence and Financing: First Results from the Post-Secondary Education and Participation Survey (PEPS). Ottawa, ON: Statistics Canada.

Berger, J. and A. Motte. 2007. "Mind the Access Gap: Breaking Down Barriers to Post-Secondary Education." Policy Options 28(10):42-46.

Boyko, L. and E.K. Chaplin. 2012. Students with Learning Disabilities: Access to Higher Education in Canada--A Review of the Literature. Retrieved December 12, 2013 (https://tspace. library.utoronto.ca/bitstream/1807/31997/1/Literature%20Review_Boyko_Chaplin%20F IN AL.pdf).

Brennan, S. 2009. Facts on Learning Limitations. Participation and Activity Limitation Survey 2006. Catalogue no. 89628-X-2009014. Ottawa, ON: Statistics Canada. Retrieved December 9, 2013 (http://cansim2.statcan.ca/cgi- win/cnsmcgi.pgm?Lang=E&SP_Action= Result&SP_ID=1963&SP_TYP=62&SP_Sort=-0&SP_Mode=2).

Brown, R., L. Newton, G. Parekh and H. Zaretsky. 2013. Special Education in the TDSB and Ontario: An Overview, 2011-2013. Toronto, ON: Toronto District School Board.

Brown, R.S. 2009. An Examination of TDSB Post-Secondary Patterns: 17 Year Old Students, 2007. Toronto, ON: Toronto District School Board.

Brown, R.S. 2010. The Grade 9 Cohort of Fall 2004. Toronto, ON: Toronto District School Board.

Brown, R.S. and G. Parekh. 2010. Special Education: Structural Overview and Student Demographics. Toronto, ON: Toronto District School Board.

Cheung, S. 2007. Education Decisions of Canadian Youth. Toronto, ON: Higher Education Quality Council of Ontario.

Collins, P.H. 1990. BLACK Feminist Thought: Knowledge, Consciousness, and the Politics of Empowerment. Boston, MA: Unwin Hyman.

Collins, P.H. 1998. "It's All in the Family: Intersections of Gender, Race, and Nation." Hypatia 13(3):62-82.

Davies, S., V. Maldonado and D. Zarifa. 2014. "Effectively Maintaining Inequality in Toronto: Predicting Student Destinations in Ontario Universities." Canadian Review of Sociology 51:22-53.

De Broucker, P. 2005. Getting There and Staying There: Low-Income Students and Post-Secondary Education. Ottawa: Canadian Policy Research Network, Incorporated, Ottawa.

De Valenzuela, J.S., S.R. Copeland, C.H. Qi and M. Park. 2006. "Examining Educational Equity: Revisiting the Disproportionate Representation of Minority Students in Special Education." Exceptional Children 72(4):425-41.

Duncan, D. 2011. Turning the Corner to a Better Tomorrow: 2011 Ontario Budget. Toronto, ON: Ontario Ministry of Finance. Retrieved April 24, 2014 (http://www.fin. gov.on.ca/en/budget/ontariobudgets/2011/papers_all.pdf).

Ferguson, D.L. 2008. "International Trends in Inclusive Education: The Continuing Challenge to Teach Each One and Everyone." European Journal of Special Needs Education 23(2): 109-20.

Finnie, R., M. Frenette, R.E. Mueller and A. Sweetman. 2010. Pursuing Higher Education in Canada: Economic, Social, and Policy Dimensions. Kingston, ON and Montreal, QC: McGill-Queen's University Press.

Finnie, R., R.E. Mueller, A. Sweetman and A. Usher. 2008. Who Goes? Who Stays? What Matters? Accessing and Persisting in Post-Secondary Education in Canada. Kingston, ON and Montreal, QC: McGill-Queen's University Press.

Finnie, R. and D. Pavlic. 2013. Background Characteristics and Patterns of Access to Postsecondary Education in Ontario: Evidence from Longitudinal Tax Data. Toronto, ON, Canada: The Higher Education Quality Council of Ontario. Retrieved April 24, 2014 (http://www.heqco.ca/SiteCollectionDocuments/Final%20LAD%20ENG.pdf).

Halfon, N., A. Houtrow, K. Larson and P.W. Newacheck. 2012. "The Changing Landscape of Disability in Childhood." The Future of Children 22(1):13-42.

Human Resources Development Canada. 2006. Disability in Canada. A 2006 Profile. Retrieved March 13, 2014 (http://publications.gc.ca/collections/collection_2011/rhdcc- hrsdc/HS6411-2010-eng.pdf).

King, A.J.C. and W.K. Warren. 2006. Transition to College: Perspectives of Secondary School Students. Toronto, ON: Colleges Ontario.

Kirby, D. 2009. "Widening Access: Making the Transition from Mass to Universal PostSecondary Education in Canada." Journal of Applied Research on Learning 2:1-17.

Krahn, H. and A. Taylor. 2005. "Resilient Teenagers: Explaining the High Educational Aspirations of Visible-Minority Youth in Canada." Journal of International Migration and Integration 6(3-4):405-34.

Mackenzie, N. 2009. Setting the Direction for Special Education in Alberta: A Review of the Literature. Edmonton, AB: Government of Alberta.

McCall, L. 2005. "The Complexity of Intersectionality." Signs 30(3):1771-800.

McCloskey, L., K. Figura, K. Narraway and B. Vukovic. 2011. Transitions Longitudinal Study: 7th Annual & Final Report to the Ministry of Training, Colleges and Universities. Toronto, ON: Ontario Ministry of Training, Colleges and Universities.

Mitchell, D. 2010. Education that Fits: Review of International Trends in the Education of Students with Special Educational Needs. Wellington, New Zealand: Ministry of Education.

Mullings, L. and A.J. Schulz. 2006. Intersectionality and Health: An Introduction. New York: Jossey-Bass.

Newman, L., E. Davies and C. Marder. 2003. "School Engagement of Youth with Disabilities." In The Achievements of Youth with Disabilities During Secondary School: A Report from the National Longitudinal Transition Study-2 (NLTS2), edited by M. Wagner, C. Marder, J. Blackorby, R. Cameto, L. Newman, P. Levine and E. Davies-Mercier. Washington, DC: U.S. Department of Education.

Ontario Ministry of Education. 2013. An Introduction to Special Education in Ontario. Retrieved April 14, 2014 (http://www.edu.gov.on.ca/eng/general/elemsec/speced/ontario.html).

Ontario Ministry of Finance. 2011. 2011 Ontario Budget: Chapter I: Ontario's Plan for Jobs and Growth Section A: A Better Tomorrow. Retrieved September 12, 2013 (http://www.fin.gov.on.ca/en/budget/ontariobudgets/2011/ch1a.html).

Organisation for Economic Co-operation and Development (OECD). 2003. Disability in Higher Education. Paris: Centre for Educational Research and Innovation.

Oswald, D.P., A.M. Best, M.J. Coutinho and H.A. Nagle. 2003. "Trends in the Special Education Identification Rates of Boys and Girls: A Call for Research and Change." Exceptionality 11(4):223-37.

Parekh, G., I. Killoran and C. Crawford. 2011. "The Toronto Connection: Poverty, Perceived Ability, and Access to Education Equity." Canadian Journal of Education 34(3):249-79.

People for Education. 2014. "Choosing Courses for High School." Retrieved February 28, 2014 (http://www.peopleforeducation.ca/wp-content/uploads/2014/02/choosing-courses- for-high -school-2014.pdf).

Pumfrey, P.D. 2008. "Moving towards Inclusion? The First-Degree Results of Students with and without Disabilities in Higher Education in the UK: 1998-2005." European Journal of Special Needs Education 23(1):31-46.

Seabrook, J.A. and W.R. Avison. 2012. "Socioeconomic Status and Cumulative Disadvantage Processes across the Life Course: Implications for Health Outcomes." Canadian Review of Sociology 49:50-68.

Shaw, S.F., J.W. Madaus and M. Banerjee. 2009. "Enhance Access to Postsecondary Education for Students with Disabilities." Intervention in School and Clinic 44(3): 185-90.

Skiba, R.J., L. Poloni-Staudinger, S. Gallini, A.B. Simmons and R. Feggins- Azziz. 2006. "Disparate Access: The Disproportionality of African American Students with Disabilities Across Educational Environments." Exceptional Children 72(4):411-24.

Sweet, R., P. Anisef, R. Brown, D. Walters and K. Phythian. 2010. Post-High School Pathways of Immigrant Youth. Toronto, ON: Higher Education Quality Council of Ontario.

Tal, B. and E. Enegajor. 2013. "Degrees of Success: Payoffs to Higher Education in Canada." CIBC Report. Retrieved March 14, 2014 (http://research.cibcwm. com/economic_public/download/if_2013-0826.pdf).

Warner, D.F. and T.H. Brown. 2011. "Understanding How Race/Ethnicity and Gender Define Age-Trajectories of Disability: An Intersectionality Approach." Social Science & Medicine 72(8): 1236-48.

Wilkinson, L.A. 2008. "The Gender Gap in Asperger Syndrome: Where are the Girls?" TEACHING Exceptional Children Plus 4(4): 1-9.

Zheng, S.M. 2009. 2006 Student Census: Correlations of School Experience with Student Demographic and Achievement. Etobicoke, ON: TDSB Research Report.

KAREN L. ROBSON AND PAUL ANISEF

York University

ROBERT S. BROWN

Toronto District School Board

GILLIAN PAREKH

York University

We wish to thank Nicholas Dion at the Higher Education Quality Council of Ontario for the French translation of the abstract.

Karen L. Robson, Department of Sociology, 2060 Vari Hall, 4700 Keele Street, York University, Toronto, Ontario M3J1P3. E-mail: [email protected]

(1) However, it should be noted that PSE is not a guarantee of high earnings and is highly contingent upon choice of credential, with holders of liberal arts degrees having less earnings power (Tal and Enenajor 2013).

(2) The word "choice," as it is used within this paper, recognizes the constrained set of choices that may be available to individuals--we do not assume a meritocratic playing field where every outcome is equally accessible to all.

(3.) The 2001 Census was the most recent Canadian census data available when the data set was constructed in 2007. The 2006 Canadian Census had not yet been publically released.

(4.) All students applying to postsecondary in Ontario do so through one of the two institutions. It should be noted that while academic researchers can, upon approval by the TDSB, gain access to student Census data, only TDSB staff can analyze information containing PSE confirmation information.

(5.) This latter category includes those who applied to postsecondary but were not accepted by an Ontario college, those that graduated from high school but did not apply to postsecondary over the three years, and those who dropped out or were still in school at the end of the three years. For the purposes of our analyses, such individuals were similar insofar as they all shared the characteristic of not being accepted to university or college.

(6.) Approximately 9 percent had applied to PSE and had not been accepted, while nearly 10 percent graduated but did not apply to PSE. Just over 15 percent had dropped out or were still in school.

(7.) In the TDSB (2012-2013), over half of students with SEN only had an IEP (57 percent), and around a quarter (23 percent) had been identified with a learning disability exceptionality. The remaining 10 exceptionalities accounted for 20 percent of TDSB students with SEN (Brown et al. 2013).

(8.) Student census questionnaires had common demographic questions, but two different sets of attitudinal questions, with some overlap, referred to "Form A" and "Form B." Students were randomly assigned to one of the two forms (Zheng 2009).

(9.) Our study is largely exploratory. As each random coefficient would add an additional three parameters to our already complex model, we have opted not to allow any coefficients to vary.

(10.) These numbers are derived from manual calculations of predicted probabilities.

(11.) A three-way interaction between Black, gender, and academic stream was also tested with no statistical significance.

(12.) Average probabilities were calculated by calculating predicted values for the college confirmation outcome and then examining the associated probabilities for individuals with specific sets of traits. The average value of the probability of confirming college of these subsets of cases has been reported here.

Table 1
Descriptive Statistics (N = 12,518)

                       Overall             Without SEN      With SEN
                        Mean      Range    (N = 11,277)   (N = 1,241)

PSE pathways
  Confirmed             0.52       0-1         0.56           0.16
    university
  Confirmed college     0.16       0-1         0.15           0.25
  Neither               0.32       0-1         0.29           0.59
Median neighborhood     58.40                 58.38          58.53
  family income
  (000k)
Parental education      0.66       0-1         0.61           0.51
  (1 = PSE)
Grade 11 marks          69.11     0-100       70.11          60.23
Attitude toward         3.42       1-5         3.45           3.25
  school
Applied (1 = yes)       0.22       0-1         0.17           0.67
Male (1 = yes)          0.50       0-1         0.50           0.68
Immigrant generation
  First generation      0.43       0-1         0.45           0.23
  Second generation     0.38       0-1         0.37           0.46
  Third generation      0.19       0-1         0.18           0.32
    or higher
Race
  White                 0.34       0-1         0.33           0.45
  Black                 0.10       0-1         0.09           0.19
  East Asian            0.21       0-1         0.23           0.08
  Latin American        0.02       0-1         0.02           0.02
  Middle Eastern        0.04       0-1         0.05           0.04
  Mixed                 0.06       0-1         0.05           0.06
  Other                 0.01       0-1         0.08           0.03
  South Asian           0.20       0-1         0.20           0.13
  Southeast Asian       0.03       0-1         0.03           0.01
School size
  [less than or         0.00       0-1         0.00           0.00
    equal to] 100
  101-200               0.01       0-1         0.01           0.02
  201-500               0.02       0-1         0.02           0.10
  501-1,000             0.29       0-1         0.27           0.36
  1,001-1,200           0.26       0-1         0.27           0.19
  1,201-1,400           0.27       0-1         0.28          0.160
  1,400                 0.15       0-1         0.15          0.162

PSE, postsecondary education; SEN, special education needs.

Table 2
Hierarchical Multinomial Logistic Regression Predicting Postsecondary
Confirmations (Reference = No Confirmations), N = 12,518

                                 Odds Ratios

                                  Null

                        University      College

Fixed effects
  SEN (1 = yes)
  Male (1 = yes)
  First generation
  Second generation
  Third generation
  Parents have
    postsecondary
    (1 = yes)
  Grade 11 marks
  Median
    neighborhood
    income
  Applied stream
    (1 = yes)
  Attitude toward
    school
  Black
  East Asian
  Latin American
  Middle Eastern
  Mixed
  Other
  South Asian
  Southeast Asian
  White
  School size <100
  101-200
  201-500
  501-1,000
  1,001-1,200
  1,201-1,400
  >1,400
  SEN x male
  SEN x Black
  SEN x parents PSE
  SEN x applied
  SEN x income
  Black x applied
  SEN x Black x
    applied
  SEN x Black x
    male
  Constant                    1.117 ***    0.329 ***
Random effects
  Level 2 variance            0.551
Model fit
  LR [chi square]
  ICC                         0.142
  Log-likelihood        -11,841.97

                                 Odds Ratios

                               +SEN Status

                        University      College

Fixed effects
  SEN (1 = yes)              0.193 ***   1.14
  Male (1 = yes)
  First generation
  Second generation
  Third generation
  Parents have
    postsecondary
    (1 = yes)
  Grade 11 marks
  Median
    neighborhood
    income
  Applied stream
    (1 = yes)
  Attitude toward
    school
  Black
  East Asian
  Latin American
  Middle Eastern
  Mixed
  Other
  South Asian
  Southeast Asian
  White
  School size <100
  101-200
  201-500
  501-1,000
  1,001-1,200
  1,201-1,400
  >1,400
  SEN x male
  SEN x Black
  SEN x parents PSE
  SEN x applied
  SEN x income
  Black x applied
  SEN x Black x
    applied
  SEN x Black x
    male
  Constant                  1.370 ***    0.350 ***
Random effects
  Level 2 variance           0.500
Model fit
  LR [chi square]          246.42
  ICC                        0.130
  Log-likelihood       -11,562.1

                                 Odds Ratios

                                  +IVs

                        University      College

Fixed effects
  SEN (1 = yes)              0.726 **    1.202 *
  Male (1 = yes)             0.888 *     0.832 **
  First generation           1.318 **    1.228 *
  Second generation          1.271 **    1.449 ***
  Third generation           --          --
  Parents have               1.306 ***   0.942
    postsecondary
    (1 = yes)
  Grade 11 marks             1.126 ***   1.027 ***
  Median                          1      0.993 ***
    neighborhood
    income
  Applied stream             0.102 ***   0.887
    (1 = yes)
  Attitude toward             1.083 **   1.004
    school
  Black                       1.17       1.126
  East Asian                  3.443 ***  1.183
  Latin American              0.784      0.987
  Middle Eastern              2.079 ***  1.243
  Mixed                       1.166      0.969
  Other                       1.019      0.786
  South Asian                 3.166 ***  1.698 ***
  Southeast Asian             1.713 **   1.668 **
  White                         --        --
  School size <100
  101-200
  201-500
  501-1,000
  1,001-1,200
  1,201-1,400
  >1,400
  SEN x male
  SEN x Black
  SEN x parents PSE
  SEN x applied
  SEN x income
  Black x applied
  SEN x Black x
    applied
  SEN x Black x
    male
  Constant                    0.001 ***  0.100 ***
Random effects
  Level 2 variance            0.234
Model fit
  LR [chi square]         5,663.69
  ICC                         0.066
  Log-likelihood         -8,730.2

                                 Odds Ratios

                             + School Size

                        University      College

Fixed effects
  SEN (1 = yes)              0.744 **     1.247 *
  Male (1 = yes)             0.872 *      0.819 ***
  First generation           1.302 **     1.228 *
  Second generation          1.274 **     1.464 ***
  Third generation           --           --
  Parents have               1.299 ***    0.939
    postsecondary
    (1 = yes)
  Grade 11 marks             1.125 ***    1.027 ***
  Median                     1.000        0.993 ***
    neighborhood
    income
  Applied stream             0.105 ***    0.907
    (1 = yes)
  Attitude toward            1.086 **     1.009
    school
  Black                      1.184        1.121
  East Asian                 3.419 ***    1.151
  Latin American             0.785        0.983
  Middle Eastern             2.115 ***    1.255
  Mixed                      1.17         0.974
  Other                      1.03         0.804
  South Asian                3.192 ***    1.691 ***
  Southeast Asian            1.763 **     1.690 **
  White                      --           --
  School size <100           1.114        0.325
  101-200                    0.553        0.412 *
  201-500                    0.233 ***    0.365 ***
  501-1,000                  0.887        1.063
  1,001-1,200                1.224        1.261
  1,201-1,400                1.375        1.767 *
  >1,400                     --           --
  SEN x male
  SEN x Black
  SEN x parents PSE
  SEN x applied
  SEN x income
  Black x applied
  SEN x Black x
    applied
  SEN x Black x
    male
  Constant                   0.002 ***    0.094 ***
Random effects
  Level 2 variance           0.149
Model fit
  LR [chi square]          118.02
  ICC                        0.043
  Log-likelihood        -8,670.5

                                 Odds Ratios

                             +Interactions

                        University      College

Fixed effects
  SEN (1 = yes)               0.305 **   0.354 **
  Male (1 = yes)              0.866 *    0.833 **
  First generation            1.292 **   1.229 *
  Second generation           1.274 **   1.481 ***
  Third generation            --         --
  Parents have                1.252 ***  0.862 *
    postsecondary
    (1 = yes)
  Grade 11 marks              1.126 ***  1.027 ***
  Median                      0.999      0.990 ***
    neighborhood
    income
  Applied stream              0.117 ***  0.934
    (1 = yes)
  Attitude toward             1.086 **   1.011
    school
  Black                       1.255      1.204
  East Asian                  3.339 ***  1.128
  Latin American              0.774      0.963
  Middle Eastern              2.093 ***  1.246
  Mixed                       1.167      0.97
  Other                       1.045      0.826
  South Asian                 3.185 ***  1.693 ***
  Southeast Asian             1.723 **   1.645 **
  White                       --         --
  School size <100            0.926      0.275
  101-200                     0.503 *    0.366 **
  201-500                     0.215 ***  0.345 ***
  501-1,000                   0.788      0.962
  1,001-1,200                 1.071      1.121
  1,201-1,400                 1.212      1.583 **
  >1,400                      --         --
  SEN x male                  1.076      0.815
  SEN x Black                 2.406      2.966
  SEN x parents PSE           1.209      1.664 **
  SEN x applied               0.770      1.145
  SEN x income                1.013 **   1.019 ***
  Black x applied             0.626      0.921
  SEN x Black x               1.096      0.276 *
    applied
  SEN x Black x               0.680      0.890
    male
  Constant                    0.002 ***  0.129 ***
Random effects
  Level 2 variance            0.155
Model fit
  LR [chi square]            61.39
  ICC                         0.044
  Log-likelihood         -8,640.4

* p < 0.05, ** p < 0.01, *** p < 0.001

ICC, intraclass correlation; IVs, independent variables; LR,
likelihood ratio; PSE, postsecondary education; SEN, special
education needs.