Getting Ahead in Singapore: How Neighborhoods,Gender,and Ethnicity Affect Enrollment into Elite Schools

Abstract

Is education the social leveler it promises to be? Nowhere is this question better addressed than in Singapore, the emblematic modern-day meritocracy where education has long been hailed as the most important ticket to elite status. In particular, what accounts for gender and ethnic gaps in enrollment into Singapore’s elite junior colleges—the key sorters in the country’s education system? We consider how the wealth of neighborhoods has combined with the elite status of schools to affect the social mobility of gender and ethnic groups. Analyzing data from 40 years of junior college yearbooks (1971–2010), we find persistent differences in educational opportunity. Women and Malays have historically experienced inequality in Singapore, and their student routes to becoming elites differ markedly. For female students, attending an elite junior college in a wealthy neighborhood is associated with wealthy neighborhoods that have a disproportionate number of elite girls’ secondary schools that feed into the junior colleges. By contrast, for Malays, not attending an elite junior college in a wealthy neighborhood has more to do with wealthy neighborhoods underrepresenting Malays in demographic composition. Elite families thus now include better educated women as well as men, yet Malays still rarely become better educated elites. These results underscore the need to carefully map the complex associations and mechanisms between gender and ethnic categorizations, the status of schools, and the characteristics of neighborhoods.

Keywords

elites schools gender ethnicity inequalities leveling reproducing neighborhoods Singapore

The rules of the reproduction game are the same from the Bronx to Bukit Timah: The key to elite children’s success is living in the right neighborhoods and attending the right schools (Badger and Bui 2018). Elite schools are socializing agents that initiate and instruct students into distinction (Bourdieu 1996; Huang et al. 2018; Reeves et al. 2017; Robinson 2018). This is especially true in Singapore, whose leaders express pride in the country as an all-out meritocracy. Through a scholarship system, Singaporean leaders say they take steps to hire and promote the best (Chua 2017; Quah 1998; Tan 2008).

Yet, not all Singaporeans start out equally in the meritocratic race to elite status: Neighborhoods and schools play a key role, intersecting with individuals’ own socioeconomic status, gender, and ethnicity. Nonacademic factors, alongside academic ones, play an important role in inequality reproduction. As a “social consultant” said to an upwardly married billionaire in the Singapore-focused Crazy Rich Asians trilogy, “Your chief handicap to social success will always be the fact that you did not attend the right kindergarten with any of the right crowd” (Kwan 2015:165).

Our research, based on junior colleges¹ in Singapore (the equivalent of senior high school in the United States and A-levels in the United Kingdom), traces unequal gender and ethnic representation in elite schools to a set of nonacademic factors associated with these schools’ location and surrounding social contexts. We examine how neighborhoods’ socioeconomic characteristics influence student choice and ultimately, patterns of enrollment into elite schools. Our research exploits the unique setting of Singapore’s pre-university system where publicly financed elite and nonelite schools are spatially distributed across wealthy and less wealthy neighborhoods. The Singaporean meritocracy draws its elite students from throughout the island city-state. All Singaporean junior colleges are public schools. They are conceptually distinct from U.S. private elite schools that draw most of their students from a select population of wealthy families. Moreover, as elite schools in many countries concentrate in wealthy neighborhoods—recruiting students who can afford their fees—this precludes meaningful variation for analysis because elite neighborhoods, elite secondary schools, and elite colleges are often associated (Howard and Levine 2004; Lubienski, Gulosino, and Weitzel 2009). Thus, the even spread of elite and nonelite junior colleges places Singapore in a unique position for analysis, allowing the study of enrollment patterns associated with four combinations of school and neighborhood characteristics: (1) elite school located in a wealthy neighborhood, (2) nonelite school located in a wealthy neighborhood, (3) elite school located in a less wealthy neighborhood, and (4) nonelite school located in a less wealthy neighborhood.

We analyzed data in six of Singapore’s elite and nonelite junior colleges (three elite, three nonelite) between 1971 and 2010. We focused on two key correlates of inequality associated with gaining entry into elite junior colleges: whether students come from minority Malay backgrounds (compared to majority Chinese, minority Indian, or other minority backgrounds) and whether students are female or male. We show that women and non-Malays have greater representation in elite junior colleges located in wealthy neighborhoods, and we examine the mechanisms to show why this is the case.

Our article makes four contributions. The first concerns the link between neighborhood and school characteristics (Ainsworth 2002; Johnson 2012). Many studies of neighborhood effects focus on individuals, tracking outcomes such as their educational attainment (Garner and Raudenbush 1991; Oreopoulos 2003; Sampson 2012; Solon, Page, and Duncan 2000). These analyses bypass the school as an institution that occupies the gap between neighborhood and individual characteristics. The few works on the neighborhood-school nexus notwithstanding (e.g., Coleman and Hoffer 1987; Ennett et al. 1997; Jencks and Mayer 1990), “the interrelatedness of the two . . . has not been the subject of any substantial review since 1990” (Johnson 2012:479). Examining the neighborhood-school nexus provides an opportunity to test the nature of compounding inequalities between institutions. Do inequalities in one institution (the neighborhood) reinforce inequalities in another institution (the school)? Inequalities are usually durable (Tilly 1998), and we investigate if they are durable from one context to the next.

Our second contribution concerns the explication of nonacademic factors in school enrollment. Accounts of educational inequality often rely on meritocratic explanations that draw attention to group differences in academic performance (Alon and Gelbgiser 2011; Goldthorpe 2003; Koh 2014). However, elite status goes beyond scholastic considerations. The study of elites is at once a study of a variety of capitals: economic, social, cultural, political, and knowledge capital (Chua 2018; Khan 2012). How do these capitals consolidate elite distinctions? Often, gaining entry to the elite is a process by which one becomes elite through the pathways of elite socialization in the right schools, living in the most well-resourced neighborhoods and families, and embeddedness in the right peer networks (Chua 2018; Cookson and Persell 1985; Gaztambide-Fernandez 2009; Owens 2018; Teo 2018). Therefore, our study considers elements such as gender, spatial proximity, and safety; the flows and path dependencies implicit in the transition between different levels of schooling; peer networks; ethnic culture; and the demographic composition of neighborhoods. All these can be critical nonacademic factors in shaping what are ostensibly purely academic outcomes.

Our third contribution addresses the challenge of specifying mechanisms: “the most important unanswered black box in neighborhood research” (Ainsworth 2002:118; see also Sampson 2012; Small 2004). Our study finds two mechanisms connecting neighborhood and school characteristics. First, female students are better represented in elite junior colleges, particularly junior colleges in wealthy neighborhoods. This is partially explained by the preponderance of elite young women’s secondary schools in wealthy neighborhoods, which feed enrollment into the neighborhoods’ elite junior colleges. Second, lower-status Malays (Singapore’s second largest ethnic group) are the least well-represented ethnic group in these wealthy school neighborhoods, suggesting that the lower share of Malay residents in wealthy neighborhoods dissuades their attendance at the elite junior colleges located there.

By interrogating the dynamics of gender and ethnicity, this article addresses a blind spot within research on elites that “rarely addresses gender and race issues” (Cousin, Khan, and Mears 2018:227). Many studies have scrutinized elites in terms of their class positions as elites, leaving “unmarked (their) gender and racial positions,” yet “elites tell us a lot about race and gender” (Cousin et al. 2018:233).

This study brings us back to a pivotal question within the sociology of education: whether (and to what extent) education is a leveler or reproducer of social inequalities. Research has produced evidence on both sides (Bowles and Gintis 2002; Downey, von Hippel, and Broh 2004; von Hippel, Workman, and Downey 2018), yet clearly the answers stretch beyond a simple dichotomy to encompass unique societal contexts. Much of the work on educational inequalities is centered on the Western context, so expanding this research to an Asian context constitutes our fourth contribution. Singapore is an apt case as it is a developed country heavily imbued with Asian values, norms, and behavior (Chua 2017).

Background: Elite Schools, Group Gaps, and a Life Beyond Grades

An elite education is a coveted prize in Singapore, as it is elsewhere (Bourdieu 1996; Cookson and Persell 1985; Gaztambide-Fernandez 2009; Karabel 2005), paving the way to better and more lucrative jobs (Altonji, Elder, and Taber 2005; Binder, Davis, and Bloom 2016; Evans and Schwab 1995; Gopinathan 2007; Jha and Polidano 2015; Reeves et al. 2017; Sander and Cohen-Zada 2010). Pragmatic Singaporeans have long imbibed the rules of the educational arms race: Students strive hard for educational and occupational success (Gee 2012), and grades have become a national preoccupation (Straits Times 2018). Parents and students know which schools are considered elite, and the private tuition industry to help children gain acceptance is booming (Barr and Skrbiš 2008; Bray 1999; Cheo and Quah 2005; Gee 2012; Tan 2009; Ye and Nylander 2015).

At all educational levels, gender gaps have narrowed considerably in favor of women (Adelman 1991; Alon and Gelbgiser 2011; Breen et al. 2010; Buchmann, DiPrete, and McDaniel 2008; Burgess et al. 2004; Christofides, Hoy, and Yang 2010; DiPrete and Buchmann 2013; Esping-Andersen 2009; Machin and Vignoles 2004; Rosin 2012; Straits Times 2017), but ethnic gaps have narrowed at a slower pace (Eller and DiPrete 2018; Gamoran 2001; Kao and Thompson 2003; Mutalib 2012; Stevens 2007). What accounts for these group gaps? We address this question by examining gender and ethnic gaps in enrollment in Singapore’s elite junior colleges, the key sorters in the country’s education system.

One factor that likely matters is group differences in academic performance. In Singapore’s meritocratic system, entry into the best schools is based on grades obtained in national examinations, such as the O-levels and A-levels. Girls often outscore boys, especially at the primary and secondary levels (Straits Times 2017). However, Malays have not done as well as non-Malays (i.e., Chinese, Indians, and others; Mutalib 2012).

Academic performance has always mattered, but we build on this reality by searching for nonacademic factors, that is, things other than test scores. We propose that the neighborhoods within which elite schools are embedded have a critical role in sustaining—even amplifying—gender and ethnic inequalities. That is, educational inequalities are traceable to a combination of social and spatial factors. Elite status is a multifaceted phenomenon, and schools are a site of elite competition. The imprint of elites extends to school location because “power is power over space” (Pincon and Pincon-Charlot 1998:70). Bukit Timah schools, for example, are well known in Singapore for sitting on vast stretches of prime land, a sizable portion of which is owned privately by schools that have deep historical roots with the earliest Chinese pioneers and philanthropists (Chua 2015a; Goh 2015). The academic aristocracy has produced a class divide demarcating two Singapores—elite and nonelite (Chua 2018), with the difference extending to spatial distinctions (Tan and Tan 2016).

We conceive of enrollment into elite schools as partially influenced by two broad processes: school admissions and student choice. With regard to school admissions, the schools themselves are the actors, determining admissions based on merit and selection (Espenshade and Radford 2009; Golden 2006; Karabel 2005; Lubienski 2007). Under merit, students are allocated places based on their test scores. However, junior colleges can also be selective in whom they recruit through admission policies that privilege certain groups over others, such as applicants whose parents are alumni or students from affiliated schools (Angod 2015; Bian and Ang 1997; Gaztambide-Fernandez 2009). In Singapore, some junior colleges have affiliations with secondary schools, so the gender and ethnic compositions of the associated schools should resemble each other, more so than if no affiliations were present.

With regard to student choice, although schools determine admission, students decide which schools to apply to and if they receive several admittance offers, where to enroll. We associate students’ choices with the role of environmental factors in educational inequality, exploring the idea that students often choose schools based on the kinds of neighborhoods in which schools are embedded (Coleman and Hoffer 1987; Jencks and Mayer 1990). We see this linkage in how housing prices and school quality are often linked (Gibbons and Machin 2003). Our study pays particular attention to neighborhood-institutional linkages (Johnson 2012).

Gender Inequality: The Rise of Female Students

The rise of women in education is a common phenomenon in many Western societies (Breen et al. 2010; Buchmann et al. 2008; Burgess et al. 2004; DiPrete and Buchmann 2013) and increasingly so in East Asia, including Singapore (Marginson 2011; Ministry of Education 2017; Strijbosch 2015). From 1975 to 1999, Singapore experienced an almost six-fold increase in the percentage of university graduates for both men and women (Mukhopadhaya 2001). In 2017, the number of new female university graduates exceeded men—9,574 women and 8,963 men—despite the gender ratio in the population being 50/50 (Department of Statistics 2017). While all junior colleges in Singapore are co-educational, secondary schools are co-educational or single-gender. The best-performing secondary schools—based on Primary School Leaving Examination scores—are typically single-gender, with girls’ schools often ranked near the very top (Straits Times 2017).

Research links women’s achievement to a variety of phenomena: greater self-discipline (Duckworth and Seligman 2006), greater use of cultural capital to succeed in school (Dumais 2002), growing aspirations for lifetime autonomy and good jobs (Goldin 2006), higher motivations for achievement (Martin 2004), and a shift from industrial work using heavy machinery to white-collar work using computers (Rainie and Wellman 2012). Women spend more time in academic pursuits than men (Adler, Kless, and Adler 1992; Legewie and DiPrete 2012; Quadlin 2016). The rise of women in education in the developed world leads us to predict that Singaporean young women will likewise be better represented in elite schooling, in this case, elite junior colleges.

We have several reasons to predict female students will be better represented in elite junior colleges that are located in wealthier neighborhoods. First, parents who send their daughters to elite schools may be especially watchful over them and thus prefer wealthier neighborhoods that they perceive to be safer. Orderly and well-maintained neighborhoods have lower crime rates, and street safety is an important consideration for female residents in particular (Kelling and Coles 1996; Kuo and Sullivan 2001; Pain 2001; Wilson 1987). Singapore is generally a safe society (Yuen 2004), but there are variations in crime—and perceived crime—depending on the neighborhood (Leong and Rahan 2017). Moreover, compared to male teenagers, parents tend to monitor their female teenagers’ activities and travel choices more closely, out of concern for their protection and propriety (Clifton et al. 2011). This makes safety and proximity to school an important consideration in female students’ school choice.

We could not obtain neighborhood-level data on crime or policing dating back to 1970, so we test instead for proximity preferences by examining the association between the percentage of women in the neighborhood and the percentage of female students in the school. A close association suggests residents are choosing nearby schools. Assuming that safety—and therefore proximity to school—is an important consideration in female students’ school choice, we would expect to see a correspondence between the share of female residents and the share of female students within the same neighborhood.

Second, path dependency may foster greater female representation in elite junior colleges within wealthy neighborhoods. If elite girls’ secondary schools are overrepresented in wealthy neighborhoods, elite junior colleges in those wealthy locations will likely also have an overrepresentation of women. The behavioral lock-in implicit in path dependency—that old habits die hard— predicts students will prefer a move to a new school within the same neighborhood (Barnes, Gartland, and Stack 2004), that is, female students who start out in elite secondary schools within wealthy neighborhoods are likely to continue to elite junior colleges within the same wealthy neighborhoods. If there are more elite secondary girls’ schools in wealthy neighborhoods, then we would expect a greater share of women in the elite junior colleges within those wealthy neighborhoods. This is consistent with three characteristics of path dependence: (1) clear temporal order with current outcomes highly sensitive to prior events; (2) outcomes are best explained by the events most immediately preceding them, not by initial conditions further back in time; and (3) persistent inertia—once set in motion, a process stays in motion to produce a predictable outcome (Mahoney 2000).

Third, path dependency is linked to processes of social networking and emulation among parents and peers. The literature suggests that school choices do not always follow from academic considerations. Instead, parents often look at where other wealthy parents are sending their children (Holme 2002). Students’ peer networks matter as well: Where are their friends going? Having begun secondary school in a particular area, students will often follow their friends into postsecondary schools within the same area (Bobonis and Finan 2009). Hence, cohorts of friends travel together, transiting from one level to the next in similar areas. In particular, at the postsecondary level, peer acceptance is associated with the social behavior of female students more than with male students (Wentzel and Caldwell 1997). These three explanatory processes lead to our first hypothesis:

Hypothesis 1: Female students are better represented in elite junior colleges than in nonelite junior colleges, especially elite junior colleges in wealthy neighborhoods.

Ethnic Inequality: Ethnic Solidarity and Underprivilege among Malay Singaporeans

The educational disadvantages of Malays are well documented (Mutalib 2012; Rahim 1998). Between 1970 and 2010, the Malay share of citizens fell from 17 percent to 12 percent. Yet, in 2010, the percentage of Malays with university education was only 5 percent—substantially lower than the 23 percent, 35 percent, and 58 percent of Chinese, Indian, and Other Singaporeans, respectively (Department of Statistics 2011). This is consistent with the expansion of educational opportunities in developed societies that has “disproportionately benefited children from relatively rich families” (Blanden and Machin 2004:230). As of the most recent census (2010), Malays tend to have lower socioeconomic status: Their monthly household income was S$3,844, compared to S$5,100 for majority Chinese, $5,370 for Indians, and S$7,432 for others (SG$1.00 = US$0.74). Malays make up a larger share of low-wealth neighborhoods (22 percent) than high-wealth neighborhoods (9 percent; Department of Statistics 2011; Mutalib 2012).

Wealth inequality affects educational inequality. Wealthy parents are usually better able to deploy their economic capital in a way that fits with the rules of the educational game (Calarco 2014; Holme 2002; Lareau, Evans, and Yee 2016; Schneider, Hastings, and LaBriola 2018, Teo 2018). But culture also has a role: Poorer minorities may reject education, choosing instead to “develop a culture in opposition to schooling” (Downey 2008:107). Therefore, poorer families—many with Malay ethnicity—have extra material and cultural hurdles to surmount.

Given these situations, we predict Malays will be less well represented in elite junior colleges than in nonelite junior colleges. Malay representation should be especially low in elite junior colleges located in wealthy neighborhoods for two reasons. First, the underrepresentation of Malay residents in wealthy neighborhoods could advance the perception that elite schools in these neighborhoods will be populated by students of the dominant ethnic group, making it less likely for Malays to choose them. Minorities often choose schools that are not monopolized by dominant ethnic groups (Ball, Reay, and David 2002). Their preferences for co-ethnics and attending schools near home could affect their lower representation in elite schools, especially if elite schools are in affluent neighborhoods with few Malays (Owens 2018). Despite Singapore’s ethnic integration policy that mandates social mixing in its public housing estates (Sim, Yu, and Han 2003), Malays form tight local communities. Their preference for living near co-ethnics is partly choice and partly circumstance. On one hand, shared cultural practices, such as religion, reinforce social solidarity; on the other hand, their minority status coupled with a shared socioeconomic situation sets them residentially apart from other ethnic groups (Chua 2015b).

Second, Malays could be underrepresented in elite junior colleges in wealthy neighborhoods because those neighborhoods have more elite secondary schools with predominantly non-Malay students (Barr and Skrbiš 2008). Hence, to the extent that students continue their postsecondary education within the same neighborhood where they attend secondary school, we would expect to see a lower share of Malay students in elite junior colleges within wealthy neighborhoods. This is similar to the path dependency arguments presented with gender, in which friends follow each other from one school to the next. These two explanatory processes lead to our second hypothesis:

Hypothesis 2: Malay minorities are less represented in elite junior colleges than in nonelite junior colleges, especially elite junior colleges in wealthy neighborhoods.

Methods

Data

Our study uses data from junior college yearbooks, national censuses, and statistical yearbooks between 1971 and 2010. We chose this timeframe because the first junior college cohort graduated in 1971 and the last Singaporean census was conducted in 2010. We also use school performance indicators compiled by the Singaporean Ministry of Education to distinguish elite from nonelite schools with information obtained from official websites for the seven years available (1995–2002). Differences in entry scores between elite and nonelite junior colleges remained relatively stable over our sample period (see Table 1). Although scholars rarely use yearbooks as a data source, this approach was useful because we could not gain access to individual- or school-level data.

Table 1.

Elite and Nonelite Junior Colleges.

		O-level Entry Scores (Incoming Cohorts)			A-level Points (Outgoing Cohorts)
		High-wealth Neighbor- hood(Central)	Medium-wealth Neighbor- hood(Northeast)	Low-wealth Neighbor- hood(East)	High-wealth Neighbor- hood(Central)	Medium-wealth Neighbor- hood(Northeast)	Low-wealth Neighbor- hood(East)
Elite junior colleges	NJC	8.88 (.26)			67.00(.76)
	AJC		10.73(.39)			62.25(.71)
	TJC			9.56(.30)			66.67(.71)
Nonelite junior colleges	CJC	14.68(.76)			53.00(2.39)
	NYJC		13.51(.47)			57.13(.83)
	TPJC			13.41(.64)			56.67(1.58)

Note: This sample comprises junior college–year observations between 1995 and 2002. Each cell depicts an average by junior college over time. Standard deviations are in parentheses. Neighborhood wealth is defined by the share of private housing. O-level entry scores (also known as L1R5) are derived from the O-level examination results of incoming cohorts, where lower scores denote better performance; averages for elite and nonelite junior colleges are 9.71 and 13.86 points, respectively, and the difference of 4.15 points is statistically significant at the 1 percent level. A-level points reflect the A-level examination results of outgoing cohorts; averages for elite and nonelite junior colleges are 65.36 and 55.64 points, respectively, and the difference of 9.72 points is statistically significant at the 1 percent level.

Because we could not collect data for all 16 junior colleges, we adopted the following sampling design. We first selected the three most populous census neighborhoods—central, northeast, and east (out of five: north, northeast, east, west, and central). Next, in each of the three neighborhoods, we selected a random pair of elite and nonelite junior colleges. The six junior colleges we used are NJC (elite) and CJC (nonelite) in the central area, AJC (elite) and NYJC (nonelite) in the northeastern area, and TJC (elite) and TPJC (nonelite) in the eastern area. Our study is thus based on a population-weighted representative sample. Appendix A summarizes information about all 16 junior colleges (elite and nonelite) from 1971 to 2010, including their spatial patterning.

The six junior college yearbooks we sampled contain the best data we could find on gender and ethnic composition. Our team went through many pages of the yearbooks, recording the number of students in each classroom by gender and ethnicity for the 40-year period. In addition, we collected information about the language medium of each class—whether classes were taught in English, Mandarin, Malay, or Tamil²—as well as their subject specialization—arts, science, or commerce. Our data collection produced a sample of 5,453 classes spanning 40 student cohorts (1971–2010) across six junior colleges. We recorded information on each junior college from its first graduating cohort. For instance, the oldest junior college in our sample (NJC) was founded in 1970, and its first cohort graduated in 1971. In effect, we have a census for six junior colleges amounting to a total of 127,838 students.

The six junior colleges we studied are located in three distinct regions, so we can use national censuses and statistical yearbooks to merge information about neighborhood characteristics with classroom-level information. These neighborhoods differ substantially in terms of residents’ wealth, and they contain elite and nonelite junior colleges. This provides critical variations along two dimensions— neighborhood wealth and elite status of the junior college—allowing us to account for effects of each dimension and estimate interaction effects.

We conducted our analyses at the classroom level because a school’s gender and ethnic composition depends critically on the distribution of its subject specializations. For example, a school that offers more classes in the arts may attract more female students (Alon and Gelbgiser 2011). Likewise, a school that offers more classes in the Malay language may attract more Malays. Hence, any analysis at the school level may be confounded by variations in subject specializations and languages across schools. By conducting analyses at the classroom level, we can directly account for such differences by the use of subject specialization and language fixed effects.

Dependent Variables

Because our focus is the representation of gender and ethnic groups, our two main dependent variables are the classroom-level shares of female and Malay students within a junior college for a given year (hereafter, female share and Malay share). We chose classroom-level outcomes as the dependent variables instead of individual-level outcomes (e.g., test scores) because of data accessibility, not because the study of neighborhoods and schools inherently binds us to using school characteristics.

To ascertain gender and ethnicity, we studied the yearbooks of our six junior colleges, looking at students’ pictures and names. We inferred gender from students’ faces and names. If a photo was unclear, it was usually possible to tell gender from a name. For example, the Chinese characters “Mei” (beautiful), “Min” (intelligent), “Xiu” (elegant), “Xue” (snow), and “Yun” (cloud) are generally female names; the characters “Chang” (flourishing), “Long” (dragon), “Qiang” (strong), and “Yong” (brave) are generally male names. The most direct markers of ethnicity were also indicated via students’ names. For example, signifiers such as “Bin” (for men) and “Binte” (for women) often accompany Malay names, and “S/O” (son of) and “D/O” (daughter of) accompany Tamil Indian names. Interethnic students (e.g., Eurasians) were assigned to the “Other” category, and here too, we could identify them from their names. About 2 percent of students were categorized as “Other” in our sample.

It was critical to get the Malay names right as Malay share is one of our main dependent variables. Our team included a Malay person who could verify the Malay names. If cases were truly ambiguous and we could not place the ethnicity, we considered the cases missing, but these made up less than 1 percent of all students.

Independent Variables

Elite junior colleges

We designated junior colleges as elite if they belong to the upper tail of the academic distribution. Our objective criterion for an elite junior college was that incoming cohorts possessed an average O-level score of 12 or less (lower scores indicate better performance) and its graduating cohorts achieved an average 60 A-level points or more (higher scores indicate better performance). Based on these definitions, NJC, AJC, and TJC are elite, and CJC, NYJC, and TPJC are nonelite. There is a high correlation between entry scores of incoming (based on O-level scores) and outgoing (based on A-level scores) students, implying consistency in the definition of elite (Table 1). Moreover, differences between elite and nonelite junior colleges persisted over time, indicating that elite status is durable (Cheo and Quah 2005; Ye and Nylander 2015).³

Neighborhood wealth

Privately owned housing is substantially more expensive than public housing in Singapore, so the share of private housing represents a good proxy of neighborhood wealth. Public housing in Singapore is not low status, but private housing is still substantially pricier than public housing. In our sample, the average share of private housing within a neighborhood (over 40 years) was 29 percent, with significant variation across neighborhoods and across years.⁴ We use a continuous measure in our regression analyses, but we categorize neighborhood wealth as high, medium, and low for ease of presentation in our cross-tabulations. The wealthiest neighborhood covers the central regions of Tanglin, Novena, and Bukit Timah and has NJC (elite) and CJC (nonelite) within its borders. The moderately wealthy neighborhood comprises the northeastern regions of Ang Mo Kio, Bishan, and Serangoon and contains AJC (elite) and NYJC (nonelite). The least wealthy neighborhood covers the eastern regions of Bedok, Tampines, and Changi and includes TJC (elite) and TPJC (nonelite). In contrast to elite and nonelite junior colleges, which are well distributed across the island, elite and nonelite secondary schools are socioeconomically clustered: 81 percent of secondary schools in the wealthiest central neighborhood are elite compared to 15 percent in the moderately wealthy northeastern neighborhood and 10 percent in the least wealthy eastern neighborhood. A drive through the central Bukit Timah neighborhood reveals an “aristocratic parade” (Bruno and Salle 2018:446) of elite secondary schools standing majestically amid wide manicured green spaces, with riveting architecture and landscaping that reveals the “power of elites over space” (Bruno and Salle 2018:435). Many of these elite Bukit Timah schools are girls’ secondary schools: Elite girls’ secondary schools comprise 49 percent of all secondary schools in the wealthy central region, 5 percent in the moderately wealthy northeastern region, and 0 percent in the least wealthy eastern region.

Control Variables

The first set of control variables comprises classroom characteristics. Given the data obtained from junior college yearbooks (i.e., subject specialization of each class—arts, science, or commerce—and the language medium of each class), we can control for subject specialization and language fixed effects by incorporating their respective dummy variables in our models. For instance, if young women or Malays are more likely to enter arts classes, with subject specialization fixed effects, we can account for differential gender and ethnic representations that may be due to particular junior colleges offering more arts classes in a given year.

Our second set of controls is a suite of fixed effects that exploits the longitudinal richness of our data. These include junior college fixed effects that represent junior college characteristics that do not vary over time and year-specific effects. Junior college fixed effects account for the possibility that some junior colleges—elite or not—prefer to enroll students of a certain gender or ethnic group: For example, if a specific junior college is affiliated with a specific girls’ secondary school, it will likely have greater female representation. These junior college fixed effects could also represent time-invariant students’ preferences for certain schools. For example, Malays may consistently prefer a particular junior college because it offers programs that cater specifically to them. We also incorporate year-specific effects to account for possible trends in female and Malay representation as our period of analysis coincides with substantial socioeconomic development in Singapore.

The third set of controls contains neighborhood characteristics that vary over time. To estimate the interaction effect of neighborhood wealth and elite status of junior colleges, we isolate the effects of neighborhood wealth by including the share of private housing in this set of controls (Table 2). Junior college fixed effects effectively account for the direct effects of (durable) elite status. We also include the share of elite and nonelite women’s’ secondary schools within each neighborhood because this might influence the gender and ethnic mix of incoming cohorts. Finally, because a school’s student population might mirror the gender and ethnic composition of neighborhood residents, our model includes neighborhood shares of female, Chinese, Malay, and Indian residents. Men and members of other ethnic groups are the reference categories. Neighborhood characteristics were only available for census years 1970, 1980, 1990, 2000, and 2010, so we used statistical yearbooks to interpolate characteristics in the intra-census years. Our interpolation method relied on information from two bracketing census years as well as the annual statistical yearbooks (Appendix B provides details).

Table 2.

Descriptive Statistics: Neighborhood Characteristics.

Neighborhood Share of [ . . . ]	High-wealth Neighborhood(Central)	Medium-wealth Neighborhood(Northeast)	Low-wealth Neighborhood(East)	All
Private housing	.482(.197)	.201(.022)	.125(.044)	.282(.200)
Women	.509(.012)	.502(.008)	.497(.008)	.503(.011)
Chinese	.792(.087)	.840(.008)	.693(.012)	.775(.080)
Malay	.095(.076)	.073(.010)	.219(.024)	.128(.080)
Indian	.062(.003)	.073(.009)	.064(.011)	.066(.009)
Elite secondary girls’ schools	.449(.154)	.046(.024)	0(0)	.183(.229)
Nonelite secondary schools	.189(.057)	.814(.095)	.906(.023)	.609(.335)

Note: This sample comprises neighborhood-year observations between 1971 and 2010. Each cell depicts an average by neighborhood wealth cluster over time. Standard deviations are in parentheses. Neighborhood wealth is defined by the share of private housing.

Model

The strength of a panel fixed effects model lies in its ability to control for unobservable and unchanging characteristics on both sides of the equation. These would include the durable academic requirements of elite schools as well as the durable legacies stemming from selective admission practices.

In our fixed effects regression model, the coefficient $β$ indicates whether female or Malay representation in elite junior colleges (relative to nonelite junior colleges) is amplified in wealthy neighborhoods, all else being equal. The dependent variable $Y_{cjnt}$ represents female share (to test Hypothesis 1) or Malay share (to test Hypothesis 2) in classroom $c$ of junior college $j$ , located in neighborhood $n$ , observed in year $t$ .

Y_{cjnt} = β ((elit e_{j} \times private_dwelling_shar e_{nt})) + γ X_{nt} + θ_{c} + δ_{j} + τ_{t} + \propto_{n} + ε_{cjnt} .

In terms of regressors, $elit e_{j}$ refers to the elite junior college dummy, and $private_dwelling_shar e_{nt}$ refers to the share of private housing in the neighborhood. $X_{nt}$ is a vector representing time-varying neighborhood characteristics, and $θ_{c},, δ_{j},, τ_{t},$ and $\propto_{n}$ denote subject specialization and language, junior college, year, and neighborhood fixed effects, respectively. Note that the direct effects of being at an elite junior college are nested within the junior college fixed effects because junior colleges’ elite status is durable. Similarly, the effects of the share of private housing are directly accounted for in the vector $X_{nt}$ . A constant term is omitted due to the set of fixed effects. Figure 1 translates our model into a flowchart that illustrates how a combination of school admissions and student choice factors determines enrollment.

Figure 1.

Flowchart displaying inputs to enrollment.

On the admissions side, junior colleges enroll students based on considerations of merit and selection. Given that elite junior colleges typically admit students with better O-level scores (see Table 1), enrollment patterns may reflect permanent or transitory differences in merit across gender/ethnic groups. In our model, junior college effects account for permanent enrollment differences between gender/ethnic groups arising from their unequal performances. Year effects account for temporal fluctuations in performances. Schools’ selective practices may also affect enrollment—for example, alumni linkages, feeder schools, and subject specializations can factor into school admissions. Our model absorbs all permanent factors relating to junior colleges’ admission practices, and subject specialization effects (specified in the model) do the same for factors related to over-time changes in subject offerings.

On the student choice side, prospective students must also make decisions about which junior college to attend. If students of a particular gender or ethnic group prefer elite junior colleges, then the permanence of such preferences will be absorbed by junior college effects. If—regardless of junior colleges’ elite status—students favor junior colleges in wealthy neighborhoods, then transitory variations in neighborhood wealth (specified in our model) will capture such preferences. Other time-varying factors, such as neighborhood gender/ethnic composition, are also specified and directly accounted for in our model. Finally, student choice may also depend on the siting of junior colleges in neighborhoods.

The shaded area in Figure 1 indicates the interactive effect of elite status of junior colleges and neighborhood wealth on enrollment, and this is the effect ( $β)$ ) we will estimate from Equation 1.

Findings

Female Student Representation in Elite Junior Colleges

There are gender disparities in enrollment between elite and nonelite junior colleges, with neighborhood characteristics amplifying these differences (see Figure 2 and Table 3). The gap in female representation between elite and nonelite junior colleges rises as neighborhood wealth increases: At high levels of neighborhood wealth, the gap exceeds 7 percentage points.

Figure 2.

Female and Malay share and neighborhood wealth.

Table 3.

Descriptive Statistics: Classroom Gender and Ethnic Representation.

	Female Share				Malay Share
	High-wealth Neighborhood(Central)	Medium-wealth Neighborhood(Northeast)	Low-wealth Neighborhood(East)	All	High-wealth Neighborhood(Central)	Medium-wealth Neighborhood(Northeast)	Low-wealth Neighborhood(East)	All
Elite	.581(.173)	.568(.201)	.575(.196)	.575(.191)	.031(.105)	.036(.062)	.050(.071)	.040(.081)
Nonelite	.508(.231)	.579(.206)	.555(.209)	.547(.218)	.026(.046)	.034(.055)	.132(.143)	.061(.102)
All	.543(.208)	.574(.204)	.566(.202)	.561(.205)	.029(.080)	.035(.058)	.085(.115)	.050(.092)

Note: This sample comprises classroom–junior college–year observations between 1971 and 2010. Each cell depicts an average by neighborhood wealth cluster over classrooms, junior colleges, and time. Standard deviations are in parentheses. Neighborhood wealth is defined by the share of private housing.

We examine female representation by estimating Equation 1. For this regression (and all subsequent regressions), we produce robust standard errors clustered by junior college as well as bootstrapped standard errors.⁵ In columns 1, 2, and 3 of Equation 1, we present the coefficient of interest, $β,$ , in three separate regressions in which we add more controls to each subsequent regression (Table 4). Comparing across columns, in each case, the estimate of $β$ is positive and statistically significant, indicating that female students are better represented in elite junior colleges situated in wealthy neighborhoods. Importantly, the size of the effect changes substantially—from column 1 to 2—when we include subject specialization (arts, science, or commerce) and classroom language effects (English, Mandarin, Malay, or Tamil), demonstrating that classroom-level characteristics are potential confounders that ought to be accounted for in our analyses.

Table 4.

Main Results: Female Share.

Dependent Variable: Female Share	(1)	(2)	(3)	(4)	(5)
Elite junior college × Neighborhood share of private housing	.139(.036)[.059]	.177(.036)[.043]	.172(.032)[.042]	.545(.112)[.221]
Elite junior college × Neighborhood share of female population					.003(.001)[.001]
Elite junior college × Neighborhood share of elite secondary girls’ schools					.141(.021)*[.048]
Elite junior college × Neighborhood share of private housing²				−.377(.104)*[.214]^
Junior college and year fixed effects	Y	Y	Y	Y	Y
Subject specialization and language fixed effects	N	Y	Y	Y	Y
Time-varying neighborhood characteristics	N	N	Y	Y	Y
Number of observations	5,453	5,453	5,453	5,453	5,453
R²	.030	.460	.461	.455	.460

Note: Robust standard errors, clustered by junior college, are in parentheses. Bootstrapped standard errors (500 replications) are in brackets. This sample comprises classroom–junior college–year observations between 1971 and 2010. Time-varying neighborhood characteristics include the shares, within each neighborhood, of private housing, elite secondary girls’ schools, nonelite secondary schools, women, Chinese, Malays, and Indians. In column 4, we also include the squared term of neighborhood share of private housing.

p < .10. *p < .05. **p < .01. ***p < .001.

Our preferred specification in column 3 accounts for junior college and year fixed effects: subject specialization and language fixed effects and time-varying neighborhood characteristics. The coefficient of .172 means that moving a pair of elite and nonelite junior colleges from a low-wealth to a high-wealth neighborhood would increase the gap in their female representation by 7.1 percentage points, computed by multiplying the coefficient of .172 by the difference in the average share of private housing between high-wealth and low-wealth neighborhoods (.549 – .134). This is by no means a small effect as average female representation in the classroom over this period is 56.1 percent. We take this result as compelling evidence supporting Hypothesis 1: Female students are better represented in elite junior colleges than in nonelite junior colleges, especially in elite junior colleges in wealthy neighborhoods.

To test for nonlinear effects (Clark 1992), we ran a quadratic model in column 4 by adding the regressors Elite Junior College × Neighborhood Share of Private Housing² and Neighborhood Share of Private Housing² to the model. We find a concave relationship where the effect of elite status on female share diminishes beyond a certain threshold of neighborhood wealth. Nevertheless, this effect remains positive at the average level of neighborhood wealth.

We investigate two possible reasons behind greater female representation in elite junior colleges located in wealthy neighborhoods. Safety is one possibility: Greater female representation in elite schools within wealthy neighborhoods could be due to parents wanting their daughters to attend a nearby elite school (Table 4, column 5). Our results find no support for this safety proposition: The interaction of elite junior college and neighborhood female share yields a coefficient of almost zero (precisely estimated). This might be due to the fact that Singaporean neighborhoods are all predominantly safe places.

Our data fit better with the second possibility of path dependency. The interaction effect of elite junior college and the share of elite secondary girls’ schools in neighborhoods is positive and large (Table 4, column 5). There is a strong correlation between the density of elite secondary girls’ schools in wealthy neighborhoods and female representation in elite junior colleges in these neighborhoods. The presence of elite girls’ secondary schools in wealthy neighborhoods forms a pathway to greater female representation in elite junior colleges in wealthy neighborhoods.

Malay Student Representation in Elite Junior Colleges

In contrast to the situation for female students, Malay students are significantly underrepresented in elite junior colleges compared to nonelite junior colleges. This occurs across virtually all levels of neighborhood wealth (see Figure 2 and Table 3). Moreover, Malay representation is particularly lower in elite junior colleges situated in wealthy neighborhoods. Our tests of Hypothesis 2 find that the coefficient of interest is always significantly negative in Equation 1 even after adding more control variables. The coefficient –.037 in column 3 of Table 5 suggests the gap in Malay enrollment between a pair of elite and nonelite junior colleges decreases by 1.5 percentage points if they relocate from a low-wealth to a high-wealth neighborhood. This is computed by multiplying the coefficient of –.037 by the difference in the average share of private housing between high- and low-wealth neighborhoods (.549 – .134). The effect is considerable given that the average Malay share of junior college participation is only around 5 percent.

Table 5.

Main Results: Malay Share.

Dependent Variable: Malay Share	(1)	(2)	(3)	(4)	(5)
Elite junior college × Neighborhood share of private housing	−.101(.011)**[.059]	−.028(.011)[.015]	−.037(.012)[.016]	−.268(.113)[.137]
Elite junior college × Neighborhood share of Malay population					.103(.040)[.052]
Elite junior college × Neighborhood share of nonelite secondary schools					−.001(.001)[.002]
Elite junior college × Neighborhood share of private housing²				.233(.110)^[.128]^
Junior college and year fixed effects	Y	Y	Y	Y	Y
Subject specialization and language fixed effects	N	Y	Y	Y	Y
Time-varying neighborhood characteristics	N	N	Y	Y	Y
Number of observations	5,453	5,453	5,453	5,453	5,453
R²	.180	.404	.409	.403	.409

p < .10. *p < .05. ***p < .001.

As we found for gender, column 4 of Table 5 shows nonlinear effects. In particular, the effect of Elite Junior College × Neighborhood Share of Private Housing on Malay share is convex, implying that the negative effect weakens beyond a critical point in neighborhood wealth. Although diminishing, the effect remains negative at the average level of neighborhood wealth.

Neighborhood ethnic composition is the key mechanism here. Malays are attending schools in neighborhoods that better reflect their own ethnic group. The results in Table 5, column 5 suggest that Malays are underrepresented in elite junior colleges in wealthy neighborhoods because the ethnic composition of wealthy neighborhoods means those schools are likely in neighborhoods dominated by other ethnic groups. Hence, if one wanted Malays to be better represented in elite junior colleges, a key would be to locate elite junior colleges in neighborhoods with a greater share of Malay residents.

In contrast to gender, we find no support for the path dependency mechanism for Malays (Table 5, column 5). We believe a desire for upward mobility and a wish to follow their peers make the many female students who start out in elite secondary girls’ schools within wealthy areas continue with an elite junior college education within the same wealthy areas. Yet, for the many Malays who start out in nonelite secondary schools in less wealthy areas, a desire for social mobility means they have to leave their peers to enroll in elite junior colleges within wealthy areas. Therefore, a desire for social mobility produces different outcomes for each group. For female students, it produces a path-dependent trajectory connecting secondary and junior college education. For Malays, it produces a less linear trajectory because they have to “beat the odds” to move from a nonelite to an elite milieu.

Robustness Checks

We conducted an over-time robustness test by replacing contemporaneous shares of private housing with shares of private housing in the preceding year. This specification addresses the concern that a neighborhood’s wealth may be endogenously determined. For example, the presence of more female students may make a different contribution to the development of the local community than the presence of more male students, leading to a systematic relationship between the share of female students and the share of private housing. We use a lagged measure of neighborhood wealth because it is less susceptible to an endogenous process. Results suggest our earlier conclusions are robust to such concerns as the estimates continue to be statistically significant and have similar magnitudes (see columns 1 and 3 of Table 6). This suggests the gender and ethnic mix of a school does not significantly influence neighborhood wealth. Columns 2 and 4 of Table 6 also show that our earlier conclusions about mechanisms are robust to lagged measures.

Table 6.

Robustness Tests.

Dependent Variable	Female Share		Malay Share
Dependent Variable	(1)	(2)	(3)	(4)
Elite junior college × Lagged neighborhood share of private housing	.160(.029)[.040]		−.036(.011)[.016]
Elite junior college × Lagged neighborhood share of female population		.003(.001)*[.002]
Elite junior college × Lagged neighborhood share of Malay population				.088(.038)^[.048]^
Elite junior college × Lagged neighborhood share of elite girls’ secondary schools		.016(.001)*[.005]
Elite junior college × Lagged neighborhood share of nonelite sec. schools				−.001(.001)[.002]
Junior college and year fixed effects	Y	Y	Y	Y
Subject specialization and language fixed effects	Y	Y	Y	Y
Time-varying neighborhood characteristics	Y	Y	Y	Y
Number of observations	5,408	5,408	5,408	5,408
R²	.460	.460	.408	.409

Note: Robust standard errors, clustered by junior college, are in parentheses. Bootstrapped standard errors (500 replications) are in brackets. This sample comprises classroom–junior college–year observations between 1971 and 2010. Time-varying lagged neighborhood characteristics include the shares, within each neighborhood, of private housing, elite secondary girls’ schools, nonelite secondary schools, women, Chinese, Malays, and Indians.

p < .10. *p < .05. **p < .01. ***p < .001.

Note that although our suite of fixed effects and time-varying neighborhood characteristics account for many endogenous factors, nonrandom selection can still be problematic insofar as it is not already captured by those control variables. For instance, suppose a certain type of household (e.g., “tiger mom” households that work extraordinarily hard in forcing students to study) sorts into wealthy neighborhoods and their children are more likely to enroll at elite junior colleges. Yet, we cannot observe the share of tiger mom households. If tiger moms are more likely to have daughters (unlikely) or be non-Malay (possibly), then we might be erroneously attributing our results to student choice when instead they are driven by the locational sorting of tiger mom households. Although this appears worrisome at first, if neighborhood gender and ethnic compositions are highly correlated with the share of tiger mom households, selection bias may be kept to a minimum.

Discussion

Is education a social leveler or a reproducer of inequalities (Alexander, Entwisle, and Olson 2007; Bowles and Gintis 2002; Downey et al. 2004)? The debate has oscillated between the two camps, but a more complex question would be how, for whom, why, and in what ways is education a social leveler or reproducer? As Downey and Condron (2016:207) write: “In the half century since the 1966 Coleman Report, scholars have yet to develop a consensus regarding the relationship between schools and inequality.” Rather than one or the other, they suggest our “perspective on schools and inequality [be] guided by the assumption that schools may shape inequalities along different dimensions in different ways.” Our study follows from their suggestion, using Singaporean data to contrast social leveling effects for women with social reproduction effects for Malay minorities.

In addition, our study demonstrates how combinations of neighborhood and school statuses affect educational choice. We show that it is useful to think of schools as part of larger social ecologies (Johnson 2012) in which school locations are critical factors in educational inequality. Elites’ power extends to their spatial dominance occurring alongside their social dominance (Burdick-Will 2018; Hamnett and Butler 2011). In Singapore, education closes the inequality gap for women, on average, but compounds ethnic inequalities.

We found that ethnicity and gender matter in different ways. For female students, attending an elite junior college in a wealthy neighborhood is associated with wealthy neighborhoods having a disproportionate number of elite girls’ secondary schools that feed enrollment into the junior colleges. By contrast, for Malays, not attending an elite junior college in a wealthy neighborhood has more to do with Malays being underrepresented in wealthy neighborhoods.

Our findings also illustrate the horizontal spread of inequalities from one institutional context to the next—from neighborhood to school—in a compounding manner: neighborhood-school combinations of elite characteristics yield specific patterns of inequality. Student enrollments in elite schools nested within wealthy areas are substantially different from those in nonelite schools nested within nonwealthy areas. These differential and specific pathways for women and Malays underscore the need to map the rich and complex associations between gender and ethnicity, status of schools, and characteristics of neighborhoods. The path dependence between secondary school and junior college underscores the need for a comprehensive view of schooling that takes into account the siting of junior colleges vis-à-vis the siting of secondary schools. Elite status follows a compounding principle involving geography, school status, historical legacy, social networks, and path dependence. As a result, in Singapore, the gender gap in educational opportunities has closed more rapidly than the ethnic gap. Both women and Malays have historically experienced inequality in Singapore, but their student routes to becoming elites have differed markedly. The elite status of schools in wealthy neighborhoods has boosted women’s situation, whereas Malays are still less likely to attend schools (especially elite schools) in privileged neighborhoods.

The differences are associated with their demographic positions in Singaporean society. The movement of female students into elite junior colleges from elite secondary schools in the same neighborhoods is an extension of class background reproducing class inequalities (Lee 1998; Leicht 2008; Straits Times 2015). Elite families retain their position, with their daughters joining their sons through elite schooling. In short, education is a social leveler for wealthy female students, bringing them onto the same plane as their wealthy male counterparts. By contrast, Malay students are rarely on the educational route to join the elite (Barr 2006; Downey 2008; Jack 2016).

We have drawn attention to the neighborhood as a serious contender in explanations of educational inequality. The pathway-dependent nature of female students’ move from elite secondary schools to elite junior colleges in wealthy neighborhoods illustrates an “account of temporality,” a process by which elites are formed over time as they travel through “upper-class trajectories” (Toft 2018:341). Malay minorities, on the other hand, remain outside these trajectories, are less connected to influential contacts, and have less information (Chua, Mathews, and Loh 2016; Hamilton, Roksa, and Nielsen 2018; Kaur 2018).

Limitations

One limitation of our study is the “breaking up of the analysis by race and gender” (Leicht 2008:237) given that social identities are intersectional. For example, does female overrepresentation in elite schools located in wealthy neighborhoods exclude female Malays? Or does Malay underrepresentation in those schools include Malay women as well? We must leave such questions to future research. We also wonder about the potentially nonrandom selection of students into neighborhoods. Our suite of fixed effects and time-varying neighborhood characteristics account for many endogenous factors, but nonrandom selection can be problematic when these factors are not captured by control variables.

Data accessibility is another limitation. Data on educational inequalities are not readily available in Singapore. We used only classroom-level data because we were unable to obtain either individual- or school-level data. Hence, we deployed an approach that used student photographs from a range of yearbooks. These yearbooks were not easily available, and coding them was laborious, so we were limited to studying six schools—but these represented a population-weighted random sample.

Conclusions

We have shown how Singapore is a propitious site for studying gender and ethnic issues in educational inequalities given its meritocratic ethos and the spatial arrangement of its schools. We exploited Singapore’s unique pre-university system in which publicly financed elite and nonelite schools are spatially well distributed across a continuum of wealthy and less wealthy neighborhoods. We delved into the longitudinal richness of our data by estimating fixed effects models that account for all time-invariant qualities of junior colleges and time-specific effects that allow us to see more clearly the combined effects of neighborhood wealth and elite status apart from the interference of covariates.

Our approach of fixed effects modeling with panel data allows us to control for institutional practices such as admission by merit (e.g., test scores) and selection (e.g., alumni factors). Both factors are unobserved in this study, so we cannot estimate their unique contributions. However, the strength of a fixed effects panel lies precisely in allowing us to control for time-invariant, institutionally durable factors.

An analogous discussion is occurring in the United States around the question of whether elite universities should selectively admit Asian American students by merit or whether they should discriminate in admissions to foster affirmative action and ethnic diversity (Hsu 2018). Once these institutional factors are accounted for, student choice factors remain to be explained—the focus of our study. To what extent do factors such as the neighborhood influence how students make school choices? The intent of our research has not been to interrogate the value of an elite junior college education set against living in a wealthy neighborhood. Instead, we addressed the ways gender and ethnic inequalities amplify inequalities in elite schooling. Rather than treating wealthy neighborhood–elite schools only as a prized outcome—which it most likely is—our research treats wealthy neighborhood–elite schools as an important source of social stratification.

Our research complements—and implicitly comments on—Western-centric studies of educational inequalities by showing how they operate in an Asian context. In so doing, we extend the conventional bases of elites to encompass gender and ethnicity in Singapore. We explicated the role of spatial inequalities, showing they reinforce inequalities of gender, ethnicity, and school status. We found instances of similarity between Singaporean-Asian and Western educational situations, such as the narrowing of gender inequalities in the West coexisting with stickier ethnic inequalities (Gamoran 2001). Yet, we also found instances of divergence. For example, the even spread of elite and nonelite junior colleges in wealthy and nonwealthy regions presents Singapore as a unique case for studying how social and spatial inequalities intersect in public education. By contrast, elite (predominantly private) schools in the United States and other countries are often located in wealthy regions. Thus, gender and ethnicity are unique systems of inequality in Singapore—each with distinctive characteristics, trajectories, and mechanisms—that can have broader implications for considerations of publicly funded elite education (Reeves et al. 2017).

Footnotes

Appendix A: Junior colleges

Table A1.

Summary of Junior Colleges.

Junior College	Mean O-level Score (L1R5)	Mean A-level Score	Elite	Census Neighbourhood	Share of Private Housing in 2010
ACJC	11.94	59.33	No	Central	.128
AJC	10.81	62.11	Yes	Northeast	.150
CJC	14.70	52.78	No	Central	.950
HCJC	7.69	69.33	Yes	Central	.867
IJC (2005–)	—	—	—	North	.039
JJC	15.13	55.22	No	West	.057
NJC	8.92	67.00	Yes	Central	.867
NYJC	13.47	57.22	No	Northeast	.380
PJC (1999–)	—	—	—	West	.092
RJC	7.47	69.44	Yes	Central (Northeast)	.867 (.244)
SAJC	12.74	59.56	No	Northeast	.088
SRJC	16.32	51.33	No	Northeast	.167
TJC	9.57	66.67	Yes	East	.287
TPJC	13.42	56.56	No	East	.093
VJC	8.83	67.89	Yes	East	.499
YJC	16.64	51.33	No	North	.056

Note: Mean O-level and A-level scores are based on 1995 to 2003 data. No data are available for IJC and PJC. RJC moved from central to northeast in 2004. The elite status of a junior college is based on the objective cutoff of O-level score of 12 or less and A-level score of 60 points or more.

Appendix B: Interpolation of Neighborhood Characteristics in Intra-census Years

We illustrate an example of the interpolation by estimating the Malay population for a particular neighborhood in 1991. First, we identify the Malay population of the neighborhood ( $pop_malay_neig h_{1990}$ and $pop_malay_neig h_{2000}$ ) and the national Malay population ( $pop_mala y_{1990}$ , $pop_mala y_{2000}$ ) from the 1990 and 2000 censuses. We also identify the national Malay population from the 1991 statistical yearbook ( $pop_mala y_{1991}$ ). Weighing census information by proximity to each census so that the more recent census is given greater importance, we derive the first-stage estimate as follows:

\hat{{pop}_{mala y_{neigh}}_{1991}} = \frac{9}{10} \times {\frac{po {p_{malay}}_{1991}}{po {p_{malay}}_{1990}} \times po {p_{mala y_{neigh}}}_{1990}} + \frac{1}{10} \times {\frac{pop_mala y_{1991}}{pop_mala y_{2000}} \times pop_malay_neig h_{2000}} .

This first stage provides us with an estimate that corresponds to the national trend observed in the contemporaneous yearbook. We do the same for each gender-neighborhood-year and ethnicity-neighborhood-year cell.

Second, we adjust the first-stage estimate by taking into account the estimates from other ethnicity cells in the same neighborhood in 1991:

\hat{\hat{pop_malay_neig h_{1991}}} = \frac{\hat{pop_malay_neig h_{1991}}}{\hat{pop_malay_neig h_{1991}} + \dots + \hat{pop_chinese_neig h_{1991}}} \times \hat{pop_neig h_{1991}} .

This second stage ensures that our interpolations are internally consistent; that is, the sum of estimated cells equals the estimated total population for a given neighborhood-year. In this case, $\hat{\hat{pop_malay_neig h_{1991}}}$ is the estimate we take to the regression analysis.

Acknowledgements

This research was supported by the National University of Singapore (grant number: R-111-000-106-133). We are grateful to our associates at the Centre for Family and Population Research (CFPR) for organizing a seminar on the study. We appreciate the feedback colleagues gave us during the 2016 International Sociological Association RC28 (Social Stratification and Mobility) meetings held in Singapore. We thank our research assistants—Woo Wee Meng, Tan Weilie, Mohamad Zulqarnain B Mohamad N, Alaric Choo, Alex Dominic Wong, Elvin Xing—and other associates who provided generous assistance during data collation. We appreciate the extensive comments given by our three anonymous reviewers. The authors had no conflict of interest in doing this research.

Notes

Research Ethics

This study is based on data collated from secondary sources—namely the junior college yearbooks, the census, and the inter-census statistical yearbooks. All the data are publicly available. All data were properly anonymized.

Author Biographies

Vincent Chua is an associate professor in the Department of Sociology at the National University of Singapore and a research associate with the Centre for Family and Population Research (CFPR).

Eik Leong Swee is senior lecturer in the Department of Economics at the University of Melbourne.

Barry Wellman directs the NetLab Network in Toronto and is the Distinguished Visiting Scholar of the Social Media Lab at Ryerson University.

References

Adelman

Clifford

. 1991. Women at Thirtysomething: Paradoxes of Attainment. Washington, DC: US Department of Education, Office of Educational Research and Improvement, Office of Research.

Adler

Patricia

Kless

Steven

Adler

Peter

. 1992. “Socialization to Gender Roles: Popularity among Elementary School Boys and Girls.”Sociology of Education 65(3):169–87.

Ainsworth

James

. 2002. “Why Does It Take a Village? The Mediation of Neighborhood Effects on Educational Achievement.”Social Forces 81(1):117–52.

Alexander

Karl

Entwisle

Doris

Olson

Linda

. 2007. “Lasting Consequences of the Summer Learning Gap.”American Sociological Review 72:167–80.

Alon

Sigal

Gelbgiser

Dafna

. 2011. “The Female Advantage in College Academic Achievements and Horizontal Sex Segregation.”Social Science Research 40(1):107–19.

Altonji

Joseph

Elder

Todd

Taber

Christopher

. 2005. “Selection on Observed and Unobserved Variables: Assessing the Effectiveness of Catholic Schools.”Journal of Political Economy 113(1):151–84.

Angod

Leila

. 2015. “Behind and beyond the Ivy: How Schools Produce Elites through the Bodies of Racial Others.” Doctoral dissertation, Department of Social Justice Education, University of Toronto.

Badger

Emily

Bui

Quoc Trung

. 2018. “Detailed New National Maps Show How Neighborhoods Shape Children for Life.”New York Times, October 1. Retrieved February 20, 2019 (https://www.nytimes.com/2018/10/01/upshot/maps-neighborhoods-shape-child-poverty.html).

Ball

Stephen

Reay

Diane

David

Miriam

. 2002. “Ethnic Choosing: Minority Ethnic Students, Social Class and Higher Education Choice.”Race, Ethnicity and Education 5(4):333–58.

10.

Barnes

William

Gartland

Myles

Stack

Martin

. 2004. “Old Habits Die Hard: Path Dependency and Behavioural Lock-in.”Journal of Economic Issues 38(2):371–77.

11.

Barr

Michael

. 2006. “Racialised Education in Singapore.”Educational Research for Policy and Practice 5(1):15–31.

12.

Barr

Michael

Skrbiš

Zlatko

. 2008. Constructing Singapore: Elitism, Ethnicity and the Nation-building Project. Copenhagen: NIAS Press.

13.

Bian

Yanjie

Ang

Soon

. 1997. “Guanxi Networks and Job Mobility in China and Singapore.”Social Forces 75(3):981–1005.

14.

Binder

Amy

Davis

Daniel

Bloom

Nick

. 2016. “Career Funneling: How Elite Students Learn to Define and Desire ‘Prestigious’ Jobs.”Sociology of Education 89(1):20–39.

15.

Blanden

Machin

Stephen

. 2004. “Educational Inequality and the Expansion of UK Higher Education.”Scottish Journal of Political Economy 51(2):230–49.

16.

Bobonis

Gustavo

Finan

Frederico

. 2009. “Neighborhood Peer Effects in Secondary School Enrollment Decisions.”The Review of Economics and Statistics 91(4):695–716.

17.

Bourdieu

Pierre

. 1996. The State Nobility. Stanford, CA: Stanford University Press.

18.

Bowles

Samuel

Gintis

Herbert

. 2002. “Schooling in Capitalist America Revisited.”Sociology of Education 75(1):1–18.

19.

Bray

Mark

. 1999. The Shadow Education System: Private Tutoring and its Implications for Planners. Paris: UNESCO International Institute for Educational Planning (IIEP).

20.

Breen

Richard

Luijkx

Ruud

Muller

Walter

Pollack

Reinhard

. 2010. “Long-term Trends in Educational Inequality in Europe: Class Inequalities and Gender Differences.”European Sociological Review 26(1):31–48.

21.

Bruno

Isabelle

Salle

Gregory

. 2018. “‘Before Long There Will Be Nothing but Billionaires!’ The Power of Elites over Space on the Saint-Tropez Peninsula.”Socio-Economic Review 16(2):435–58.

22.

Buchmann

Claudia

DiPrete

Thomas

McDaniel

Anne

. 2008. “Gender Inequalities in Education.”Annual Review of Sociology 34:319–37.

23.

Burdick-Will

Julia

. 2018. “School Location, Social Ties, and Perceived Neighborhood Boundaries.”City & Community 17(2):418–37.

24.

Burgess

Simon

McConnell

Brendon

Propper

Carol

Wilson

Deborah

. 2004. “Girls Rock, Boys Roll: An Analysis of the Age 14–16 Gender Gap in English Schools.”Scottish Journal of Political Economy 51(2):209–29.

25.

Calarco

Jessica

. 2014. “Coached for the Classroom: Parents’ Cultural Transmission and Children’s Reproduction of Educational Inequalities.”American Sociological Review 79(5):1015–37.

26.

Cheo

Roland

Quah

Euston

. 2005. “Mothers, Maids and Tutors: An Empirical Evaluation of Their Effect on Children’s Academic Grades in Singapore.”Education Economics 13(3):269–85.

27.

Christofides

Louis

Hoy

Michael

Yang

Ling

. 2010. “Participation in Canadian Universities: The Gender Imbalance (1977–2005).”Economics of Education Review 29(3):400–10.

28.

Chua

Beng Huat

. 2017. Liberalism Disavowed: Communitarianism and State Capitalism in Singapore. Singapore: NUS Press.

29.

Chua

Mui Hoong

. 2018. Singapore, Disrupted: Essays on a Nation at the Crossroads of Change. Singapore: Straits Times Press.

30.

Chua

Vincent

. 2015a. “Do Raffles Boys ‘Rule’ and Anglo-Chinese Boys ‘Own’ Singapore? Analyzing School-to-work Correspondences in an Elite Developmental State.”International Studies in Sociology of Education 25(1):1–19.

31.

Chua

Vincent

. 2015b. “How Do Singaporeans Connect? Ties among Chinese, Malays and Indians.”American Behavioral Scientist 59(9):1115–28.

32.

Chua

Vincent

Mathews

Mathew

Cheng

Loh Yi

. 2016. “Social Capital in Singapore: Gender Differences, Ethnic Hierarchies, and Their Intersection.”Social Networks 47(October):138–50.

33.

Clark

Rebecca L.

1992. Neighborhood Effects on Dropping out of School among Teenage Boys. Washington, DC: The Urban Institute.

34.

Clifton

Kelly J.

Akar

Gulash

Smith

Andrea Livi

Voorhees

Carolyn C

. 2011. “Gender Differences in Adolescent Travel to School: Exploring the Links with Physical Activity and Health.”Women’s Issues in Transportation Conference Proceedings 46(2):203–12.

35.

Coleman

James

Hoffer

Thomas

. 1987. Public and Private High Schools: The Impact of Communities. New York: Basic Books, Inc.

36.

Cookson

Peter

Persell

Caroline

. 1985. Preparing for Power. New York: Basic Books.

37.

Cousin

Bruno

Khan

Shamus

Mears

Ashley

. 2018. “Theoretical and Methodological Pathways for Research on Elites.”Socio-Economic Review 16(2):225–49.

38.

Department of Statistics. 2011. Census of Population 2010: Demographic Characteristics, Education, Language and Religion. Singapore: Ministry of Trade and Industry.

39.

Department of Statistics. 2017. “Graduates from University First Degree Courses by Type of Course and Sex, Annual.” Retrieved February 20, 2019 (http://www.tablebuilder.singstat.gov.sg/publicfacing/createDataTable.action?refId=15207).

40.

DiPrete

Thomas

Buchmann

Claudia

. 2013. The Rise of Women. New York: Russell Sage Foundation.

41.

Downey

Douglas

. 2008. “Black/White Differences in School Performance: The Oppositional Culture Explanation.”Annual Review of Sociology 34:107–26.

42.

Downey

Douglas

Condron

Dennis

. 2016. “Fifty Years since the Coleman Report: Rethinking the Relationship between Schools and Inequality.”Sociology of Education 89(3):207–20.

43.

Downey

Douglas

von Hippel

Paul

Broh

Beckett

. 2004. “Are Schools the Great Equalizer? Cognitive Inequality during the Summer Months and the School Year.”American Sociological Review 69(5):613–35.

44.

Duckworth

Angela

Seligman

Martin

. 2006. “Self- discipline Gives Girls the Edge: Gender in Self-discipline, Grades, and Achievement Test Scores.”Journal of Educational Psychology 98(1):198–208.

45.

Dumais

Susan

. 2002. “Cultural Capital, Gender, and School Success: The Role of Habitus.”Sociology of Education 75(1):44–68.

46.

Eller

Christina

DiPrete

Thomas

. 2018. “The Paradox of Persistence: Explaining the Black-white Gap in Bachelor’s Degree Completion.”American Sociological Review 83(6):1171–214.

47.

Susan

Ennett

Flewelling

Robert

Lindrooth

Richard

Norton

Edward

. 1997. “School and Neighborhood Characteristics Associated with School Rates of Alcohol, Cigarette, and Marijuana Use.”Journal of Health and Social Behavior 38(1):55–71.

48.

Espenshade

Thomas

Radford

Alexandria

. 2009. No Longer Separate, Not Yet Equal. Princeton, NJ: Princeton University Press.

49.

Esping-Andersen

Gosta

. 2009. The Incomplete Revolution. Cambridge: Polity.

50.

Evans

William

Schwab

Robert

. 1995. “Finishing High School and Starting College: Do Catholic Schools Make a Difference?”Quarterly Journal of Economics 110(4):941–74.

51.

Gamoran

Adam

. 2001. “American Schooling and Educational Inequality: A Forecast for the 21st Century.”Sociology of Education 74:135–53.

52.

Garner

Catherine

Raudenbush

Stephen

. 1991. “Neighborhood Effects on Educational Attainment: A Multilevel Analysis.”Sociology of Education 64(4):251–62.

53.

Gaztambide-Fernandez

Ruben

. 2009. “What Is an Elite Boarding School?”Review of Educational Research 79(3):1090–128.

54.

Gee

Christopher

. 2012. The Educational ‘Arms Race’: All for One, Loss for All. Singapore: Institute of Policy Studies.

55.

Gibbons

Steve

Machin

Stephen

. 2003. “Valuing English Primary Schools.”Journal of Urban Economics 53(2):197–219.

56.

Goh

Daniel

. 2015. “Elite Schools, Postcolonial Chineseness and Hegemonic Masculinities in Singapore.”British Journal of the Sociology of Education 36(1):137–55.

57.

Golden

Daniel

. 2006. The Price of Admission: How America’s Ruling Class Buys Its Way into Elite Colleges and Who Gets Left outside the Gates. New York: Crown.

58.

Goldin

Claudia

. 2006. “The Quiet Revolution That Transformed Women’s Employment, Education and Family.”American Economic Review 96(2):1–21.

59.

Goldthorpe

John

. 2003. “The Myth of Education-based Meritocracy.”New Economy 10(4):234–39.

60.

Gopinathan

2007. “Globalisation, the Singapore Developmental State and Education Policy: A Thesis Revisited.”Globalisation, Societies and Education 5(1):53–70.

61.

Hamilton

Laura

Roksa

Josipa

Nielsen

Kelly

. 2018. “Providing a ‘Leg up’: Parental Involvement and Opportunity Hoarding in College.”Sociology of Education 91(2):111–31.

62.

Hamnett

Chris

Butler

Tim

. 2011. “Geography Matters: The Role Distance Plays in Reproducing Educational Inequality in East London.”Transactions of the Institute of British Geographers 36(4):479–500.

63.

Holme

Jennifer

. 2002. “Buying Homes, Buying Schools: School Choice and the Social Construction of School Quality.”Harvard Educational Review 72(2):177–205.

64.

Howard

Adam

Levine

Arthur

. 2004. “Where Are the Poor Students? A Conversation about Social Class and College Attendance.”About Campus 9(4):19–24.

65.

Hsu

Hua

. 2018. “The Rise and Fall of Affirmative Action.”The New Yorker, October15.

66.

Huang

Gejun

Xiaoqian

Chen

Wenhong

Straubhaar

Joseph

. 2018. “Fall-behind Parents? The Influential Factors on Digital Parenting Self-efficacy in Disadvantaged Communities.”American Behavioral Scientist 62(9):1186–206.

67.

Jack

Anthony Abraham

. 2016. “(No) Harm in Asking: Class, Acquired Cultural Capital, and Academic Engagement at an Elite University.”Sociology of Education 89(1):1–19.

68.

Jencks

Christopher

Mayer

Susan

. 1990. “The Social Consequences of Growing up in a Poor Neighborhood,” Pp. 111–86 in Inner-city Poverty in the United States, edited by L. E.

Lynn, Jr

McGeary

M. G. H

. Washington, DC: National Academy Press.

69.

Jha

Nikhil

Polidano

Cain

. 2015. “Long-run Effects of Catholic Schooling on Wages.”The B.E. Journal of Economic Analysis & Policy 15(4):2017–45.

70.

Johnson

Odis

. 2012. “A Systematic Review of Neighborhood and Institutional Relationships Related to Education.”Education and Urban Society 44(4):477–511.

71.

Kao

Grace

Thompson

Jennifer

. 2003. “Racial and Ethnic Stratification in Educational Achievement and Attainment.”Annual Review of Sociology 29:417–42.

72.

Karabel

Jerome

. 2005. The Chosen: The Hidden History of Admission and Exclusion at Harvard, Yale and Princeton. Boston: Houghton Mifflin.

73.

Kaur

Satveer

. 2018. “Reducing Heart Health Disparities: A Culture-centered Approach to Understanding Heart Health Meanings among Low Income Malay Community Members in Singapore.” Doctoral dissertation, National University of Singapore, Department of Communication and New Media, Singapore.

74.

Kelling

George

Coles

Catherine

. 1996. Fixing Broken Windows. New York: Martin Kessler Books.

75.

Khan

Shamus

. 2012. “The Sociology of Elites.”Annual Review of Sociology 38:361–77.

76.

Koh

Aaron

. 2014. “Doing Class Analysis in Singapore’s Elite Education: Unraveling the Smokescreen of ‘Meritocratic Talk.’”Globalization, Societies and Education 12(2):196–210.

77.

Kuo

Frances

Sullivan

William

. 2001. “Environment and Crime in the Inner City: Does Vegetation Reduce Crime?”Environment and Behavior 33(3):343–67.

78.

Kwan

Kevin

. 2015. China Rich Girlfriend. New York: Anchor.

79.

Lareau

Annette

Evans

Shani Adia

Yee

April

. 2016. “The Rules of the Game and the Uncertain Transmission of Advantage: Middle-class Parents’ Search for an Urban Kindergarten.”Sociology of Education 89(4):279–99.

80.

Lee

William

. 1998. “Gender Inequality and Discrimination in Singapore.”Journal of Contemporary Asia 28(4):484–97.

81.

Legewie

Joscha

DiPrete

Thomas

. 2012. “School Context and the Gender Gap in Educational Achievement.”American Sociological Review 77(3):463–85.

82.

Leicht

Kevin

. 2008. “Broken Down by Race and Gender? Sociological Explanations of New Sources of Earnings Inequality.”Annual Review of Sociology 34:237–55.

83.

Leong

Chan-Hoong

Rahan

Ili Kaiyisah

. 2017. In Low-crime Singapore: Understanding the Link between Diversity and Crime at the Neighborhood Level. Singapore: IPS Commons.

84.

Lubienski

Christopher

. 2007. “Marketing Schools: Consumer Goods and Competitive Incentives for Consumer Information.”Education and Urban Society 40(1):118–41.

85.

Lubienski

Christopher

Gulosino

Charisse

Weitzel

Peter

. 2009. “School Choice and Competitive Incentives: Mapping the Distribution of Educational Opportunities across Local Education Markets.”American Journal of Education 115(4):601–47.

86.

Machin

Stephen

Vignoles

Anna

. 2004. “Educational Inequality: The Widening Socio-economic Gap.”Fiscal Studies 25(2):107–28.

87.

Mahoney

James

. 2000. “Path Dependence in Historical Sociology.”Theory and Society 29(4):507–48.

88.

Marginson

Simon

. 2011. “Higher Education in East Asia and Singapore: Rise of the Confucian Model.”Higher Education 61(5):587–611.

89.

Martin

Andrew

. 2004. “School Motivation of Boys and Girls: Differences of Degree, Differences of Kind or Both?”Australian Journal of Psychology 56(3):133–46.

90.

Ministry of Education. 2017. Education Statistics Digest. Singapore: Ministry of Education.

91.

Mukhopadhaya

Pundarik

. 2001. “Changing Labor-force Gender Composition and Male-female Income Diversity in Singapore.”Journal of Asian Economics 12(4):547–68.

92.

Mutalib

Hussin

. 2012. Singapore Malays. Oxford: Routledge.

93.

Oreopoulos

Philip

. 2003. “The Long-run Consequences of Living in a Poor Neighborhood.”The Quarterly Journal of Economics 118(4):1533–75.

94.

Owens

Ann

. 2018. “Income Segregation between School Districts and Inequality in Students’ Achievement.”Sociology of Education 91(1):1–27.

95.

Pain

Rachel

. 2001. “Gender, Race, Age and Fear in the City.”Urban Studies 38(5–6):899–913.

96.

Pincon

Michel

Pincon-Charlot

Monique

. 1998. Grand Fortunes: Dynasties of Wealth in France. New York: Algora.

97.

Quadlin

Natasha

. 2016. “Gender and Time Use in College: Converging or Diverging Pathways?”Gender & Society 30(2):361–85.

98.

Quah

Jon

. 1998. “Singapore’s Model of Development: Is It Transferable?” Pp. 105–25 in Behind East Asia Growth: The Political and Social Foundations of Prosperity, edited by Rowen

. London: Routledge.

99.

Rahim

Lily

. 1998. The Singapore Dilemma: The Political and Educational Marginality of the Malay Community. Kuala Lumpur: Oxford University Press.

100.

Rainie

Lee

Wellman

Barry

. 2012. Networked. Cambridge, MA: MIT Press.

101.

Reeves

Aaron

Friedman

Sam

Rahal

Charles

Flemmen

Magne

. 2017. “The Decline and Persistence of the Old Boy: Private Schools and Elite Recruitment 1897–2016.”American Sociological Review 82(6):1139–66.

102.

Robinson

Laura

. 2018. “The Identity Curation Game: Digital Inequality, Identity Work, and Emotion Management.”Information, Communication & Society 21(5):661–80.

103.

Rosin

Hanna

. 2012. The End of Men and the Rise of Women. New York: Riverhead Books.

104.

Sampson

Robert

. 2012. Great American City. Chicago: University of Chicago Press.

105.

Sander

William

Cohen-Zada

Danny

. 2010. “The Economics of Catholic Schools.” Pp. 354–59 in International Encyclopedia of Education, edited by Peterson

Baker

McGaw

. Oxford: Elsevier.

106.

Schneider

Daniel

Hastings

Orestes P.

LaBriola

Joe

. 2018. “Income Inequality and Class Divides in Parental Investments.”American Sociological Review 83(3):475–507.

107.

Small

Mario

. 2004. Villa Victoria. Chicago: University of Chicago Press.

108.

Sim

Loo Lee

Shi Ming

Han

Sun Sheng

. 2003. “Public Housing and Ethnic Integration in Singapore.”Habitat International 27(2):293–307.

109.

Solon

Gary

Page

Marianne

Duncan

Greg

. 2000. “Correlations between Neighboring Children in Their Subsequent Educational Attainment.”Review of Economics and Statistics 82(3):383–92.

110.

Stevens

Peter

. 2007. “Researching Race/Ethnicity and Educational Inequality in English Secondary Schools: A Critical Review of the Research Literature between 1980 and 2005.”Review of Educational Research 77(2):147–85.

111.

Straits Times. 2015. “Class Divide in the Classroom.”August13.

112.

Straits Times. 2017.”Girls Lead in Primary School but Boys Catch Up: MOE Study.”January22.

113.

Straits Times. 2018. “Don’t Forget to Give Kids Their Childhood.”September24.

114.

Strijbosch

Karlien

. 2015. “Single and the City: State Influences on Intimate Relationships of Young, Single, Well-educated Women in Singapore.”Journal of Marriage and Family 77(5):1108–25.

115.

Tan

Ern Ser

Tan

Min Wei

. 2016. “Two Studies on Class in Singapore: Diversity or Division?” Pp. 121–41 in Managing Diversity in Singapore: Policies and Prospects, edited by Mathews

Chiang

W. F

. Singapore: World Scientific.

116.

Tan

Jason

. 2009. “Private Tutoring in Singapore: Bursting out of the Shadows.”Journal of Youth Studies 12(1):93–103.

117.

Tan

Kenneth Paul

. 2008. “Meritocracy and Elitism in a Global City: Ideological Shifts in Singapore.”International Political Science Review 29(1):7–27.

118.

Teo

You Yenn

. 2018. This Is What Inequality Looks Like. Singapore: Ethos Books.

119.

Tilly

Charles

. 1998. Durable Inequality. Berkeley: University of California Press.

120.

Toft

Maren

. 2018. “Upper-class Trajectories: Capital- specific Pathways to Power.”Socio-Economic Review 16(2):341–64.

121.

von Hippel

Paul

Workman

Joseph

Downey

Douglas

. 2018. “Inequality in Reading and Math Skills Forms Mainly before Kindergarten.”Sociology of Education 91(4):323–57.

122.

Wentzel

Kathryn

Caldwell

Kathryn

. 1997. “Friendships, Peer Acceptance, and Group Membership: Relations to Academic Achievement in Middle School.”Child Development 68(6):1198–209.

123.

Wilson

William

. 1987. The Truly Disadvantaged: The Inner City, the Underclass, and Public Policy. Chicago: University of Chicago Press.

124.

Rebecca

Nylander

Erik

. 2015. “The Transnational Track: State Sponsorship and Singapore’s Oxbridge Elite.”British Journal of Sociology of Education 36(1):11–33.

125.

Yuen

Belinda

. 2004. “Safety and Dwelling in Singapore.”Cities 21(1):19–28.