The unobserved power of context: Can context moderate the effect of expectations on educational achievement?

Abstract

There is growing evidence that shows that social context is becoming less significant in influencing educational achievements and expectations. Additionally, evidence indicates that expectations are high across the board and becoming of declining significance to educational achievement. In light of this, we reexamine and offer an alternative to the renowned linear models connecting background variables to scholarly achievement via pupils' educational expectation. Analysis of GCSE scores, using three consecutive waves of the Longitudinal Study of Young People in England, reveals that the verity of this classical claim is dependent on occupational class, ethnic origin, and school socio-economic status level. These results thus confirm an unrecognized effect of family and school background variables on achievement—not only to engender expectations but also to moderate their influence. This recognition expands and deepens understanding of inequality in educational achievement and social mobility by treating expectations as a resource and analytically distinguishing between the ability of a social group or a certain school to attain or create high expectations and their ability to capitalize on it and translate these into achievements. We find that in many cases, expectations still possess strong potential to transform into achievement, while in others they are high across the board and rather inconsequential.

Keywords

Educational expectations educational achievements class ethnicity social mobility stratification inequality school effects LSYPE

Introduction

Policymakers and researchers have long perceived educational achievement and attainment as a focal key to the understanding of social mobility and inequality of life chances in post-industrial societies (Bourdieu and Passeron, 1990; Coleman, 1988; Goldthorpe, 1996; Jacob and Wilder, 2010). The educational system is considered by many to be a tipping point where, on the one hand, under neo-liberal theories and the rising cultural value of “meritocracy,” it holds opportunity for disadvantaged individuals and social groups to increase capital of all kinds (Deary et al., 2005; Khattab, 2005). On the other hand, it is also considered a chief creator and reproducer of inequality (Johnson et al., 2010).

Given its high importance, scholars have since the late 1960s, tried to discern and model the determinants of educational achievement. Structural and cultural family background variables have shown to greatly influence youths' scholarly achievement (Bozick et al., 2010).¹ Moreover, many models piecing this puzzle, which we will refer to as “classical” or “origin-expectation-destination” models, have incorporated expectations, aspirations, or other measures of motivation of youth as a central chain connecting these socio-demographic background variables to achievement (Breen and Goldthorpe, 1997; Jencks et al., 1983; Marjoribanks, 2005).² Further contributing to this basic model, contemporary scholars vastly expand the variety of background variables to include the school and even neighborhood context as part of the inequality reproducing process (Ainsworth, 2002).

However, in contrast to this discipline of thought, studies utilizing a historical perspective challenge the centrality of expectations as a path through which some groups hold and transfer their advantage. Namely, it is claimed that rising expectations by all social groups have substantially weakened the connection between background variables—such as ethnicity or occupational class—and expectations (Hanson, 1994; Rosenbaum, 2001; Schneider and Stevenson, 1999; Stewart et al., 2007), as well as between expectations and achievement (Jacob and Wilder, 2010; Schneider and Stevenson, 1999). Under this line of thought, we include the developers of the neo-liberal model that have gone as far as suggesting the decoupling of expectations and achievements as part of a wider process of “disembedment” of the individual from his social background (Beck and Beck-Gernsheim, 2002: 30–41; Breen et al., 2009). Other thinkers, such as the “reproduction” hypothesis supporters, suggest that background variables still hold great influence on achievements despite the seemly “meritocratic” expansion of education and expectations. However, since expectations are unrelated to social groups, they are no longer the cardinal vessel connecting background variables to achievements (Blossfeld and Shavit, 1993; Shavit et al., 2007).

When juxtaposing these strands of evidence, a paradox arises: if high expectations are indeed not related to any specific social groups on the one hand or to achievement on the other, how are achievement gaps among social groups still substantial as so many scholars demonstrate? This study aims to address this paradox.

Rather than defaulting to one of these strands of thought as an answer to the paradox, the aim of this study is to offer a third hypothesis—the moderating effect of background—bridging between the two. This hypothesis suggests that both sides of the rift may coexist for even if the relation between background variables and expectation of youth has weakened and expectations are, as a whole, no longer related to achievements; achievement gaps among social groups may persist due to differences among groups in their ability to capitalize on every expectation unit they attain. Hence, the paper expands our understanding of the mechanism connecting background variables to achievement arguing that structural and cultural variables, as a whole affect achievement, not only via the well-known internalization and utilization of norms, expectations and motivation, but also via their ability to alter returns of social actors on already existing traits. Throughout the paper, we examine this hypothesis using four factors: ethnic group; occupational class; school socio-economic status (SES) level; and school minority proportion that to the best of our knowledge, have never been comprehensively scrutinized.

After reviewing the possible hypotheses, we subject them to empirical examination using three-time points of the Longitudinal Study of Young People in England (LSYPE) administered to pupils and their parents. Based on our examination of these data, we conclude that expectation level is still a strong predictor of achievements. Furthermore, the effect of ethnicity, occupational class, school SES, and minority composition are still substantial, in accordance with the “reproduction” hypotheses. Finally, corresponding with the “classical” hypothesis, these effects are partly mediated by expectations. Yet, as predicted by the moderating hypothesis, excluding school minority composition, these background variables also play a role in determining achievements by affecting the relation between expectations and achievements. Once this additional possible effect is modeled, it becomes clear that while expectations do in fact heavily affect achievements within some groups, within others they are almost irrelevant, as predicted by “neo-liberal” supporters. Therefore, we argue that the effect of expectations on achievement is contextual and varies, rendering it strong within some social groups or schools (adhering to classical convention), while within others, it loses its potency as a factor as claimed by neo-liberals.

The disputed role of expectations in the equation

Origin-expectation-destination models all share the notion that expectation, ambition, or motivation of youth to prosper have a stronghold when explaining social mobility and educational inequality.³ Expectations of youth are likely to play a fundamental role as one of the causes of educational achievement—transmitting and reproducing inequality, enabling social mobility or connecting background status to that of the future. They thus suggest a process of internalization of signals and stratified attainment probabilities that transform into (de)motivation or (low) expectation and finally to actual attainment.

Despite fierce disagreements among them regarding the intricate mechanisms of influence and many of the underlying axioms, under this broad contention, we can include a few well-known theories. First, scholars of status attainment (known as the Wisconsin model) (Alexander and Eckland, 1975; Jencks et al., 1983) tie parental SES to educational attianment via a multitude of variables, among them educational expectations which has been dubbed the most influential vessel for intergenerational transmission of social status (Jencks et al., 1972).⁴

Second, the Breen–Goldthorpe model (1997) integrates expectation of youth as the vessel through which primary effects—i.e. determinants of demonstrated initial educational ability or the “potential rage of attainment” (Boudon, 1974; Goldthorpe, 1996) —affects the choice to stay on the education track (Breen and Goldthorpe, 1997: 285–287; Breen and Yaish, 2006).

Third, Bourdieu’s theory (Bourdieu, 1973, 1997; Bourdieu and Passeron, 1990) adds that parents of the dominant class, in calibration with the educational system and other social structures, will tacitly instill in the individual, alongside the socially valued dispositions, orientations, and abilities, a fitting “subjective expectation of objective probability,” which is reinforced by concomitant success thus further elevating and consolidating expectations. On the flip side of the same process, lower class individuals become acutely aware of their relative small chance to succeed which, in turn, leads them to drop out of the educational track (Bourdieu, 1973).

Finally, culture-based explanations, perceiving ethnic or racial gaps within occupational classes (e.g. Fryer and Levitt, 2004; Strand, 2011), also utilize educational expectations to explain both over and under achievements of different minority groups (Kao and Thompson, 2003). Following Coleman’s (1988) line of thought, his successors piece together a possible explanation to such intra-class variance employing expectations. Augmenting the concept of social capital, they claim that a family or community can use their social ties and “norms enforcement” ability to better ingrain educational and social mobility ambitions in the identity of the youth, making social capital a key predictor of educational achievements (Green and White, 2007: 32–59; Goddard et al., 2009; Modood, 2004). These culture-based resources, such as “immigrant optimism” (Kao and Tienda, 1995) or “ethnic capital” (Modood, 2004), have been shown to counter the negative impact of low average SES of some minority groups and therefore, explain their higher expectations and achievements (e.g. Khattab and Modood, 2017; Shah et al., 2010).

Nonetheless, this is not true for all minority groups in all social contexts. Researchers posit that some minority groups are substantially disadvantaged when compared to respective majority groups (Borgna and Contini, 2014; Heath and Brinbaum, 2007; Kristen and Granato, 2007). Interestingly, here too, expectations seem to play an imporant role, partly explaining the disadvantageous achievements of minorities (Karen, 2002; White and Kaufman, 1997).

The research presented thus far has founded, elaborated, and empirically supported an origin-expectation-destination model. Countering these, we posit a school of thought awakened by evident change in the educational systems of post-modern societies. The past few decades have been characterized by consecutive initiatives designed to improve attainment levels of children from all backgrounds and a well-documented expansion in average achievement and qualification levels (Blossfeld and Shavit, 1993; Breen and Jonsson, 2005; Breen et al., 2009; Jacob and Wilder, 2010). Following that came a surge in educational expectations of youth, up to a point where researchers show that the overwhelming majority of youth expect to attain a college degree or higher (Goodman and Gregg, 2010; Jacob and Wilder, 2010; Morgan, 1996; Schneider and Stevenson, 1999: 75). For example, Goyette (2008) demonstrates that while SES is still related to expectations, its effect has decreased substantially in the past decades. Moving to ethnicity, both in the USA and in England, expectations were found to have been equally high for all ethnic groups, after accounting for other social background variables (Modood, 2004).

As for the link connecting expectations to achievements, the rising expectation levels coupled with the still existent variance in actual achievement, logically compels the conclusion that the effect of expectations on achievements should diminish (Jacob and Wilder, 2010).

This growing irrelevance of expectations, alongside other processes, is viewed by the delegates of the “neo-liberal” hypothesis as precursors of the erosion or utter irrelevance of social background in causing achievements (and other outcomes). “Neo-liberal” delegates see this expansion in the educational system, among other factors, as a process the equipped greater numbers of children with “self-reflective” and individualized forms of knowledge or “autonomous reflexivity” (negating communal traditions), “disembed” themselves from their class and social backgrounds (Archer, 2007: 192–265; Beck and Beck-Gernsheim, 2002). Empirical examination does proclaim that, at least in some countries and attainment levels, the effect of class on achievement has mitigated, rendering expectation less important in the process (Blossfeld and Shavit, 1993; Breen et al., 2009; Shavit et al., 2007). In England, evidence is mixed; a steady association between class origin and educational attainment across cohorts and an ever more rigid mobility structure has been found (Breen and Jonsson, 2005). On the other hand, later evidence shows decreasing effect of class on attainment in younger cohorts (Breen et al., 2009).

Reconciling between origin-expectation-destination models—asking to explain existent inequality in achievements via expectations—with the established findings showing that expectations are in the process of becoming unrelated to achievements, we propose the hypothesis of the “moderating” effect of background. In other words, the relationship between expectation and achievement is dependent on different background variables, rendering it strong for some social groups and weak for others. Differential abilities among different groups to capitalize on expectations they attain may counter the process of converging expectation levels and explain persistent achievement gaps.

Although the inadequacy of the linear assumption has already been pointed out (Ganzach, 2000; Marjoribanks, 2002:23–51) and in spite of its great potential explained above, to the best of our knowledge, very few have postulated that the social context can affect returns on expectations. Two noted exceptions, working with USA samples, did follow this path. Hanson (1994) found that lower SES youth were more likely to experience “lost talent” than higher SES youth. With regard to race, Whites were found to have higher odds of “losing talent” than non-Whites, a result the author explains as “selection bias” of her sample (Hanson, 1994: 178).

This study is a good example of the theoretical possibilities obtainable when looking at interaction effects. However, the author uses dichotomous measures for expectations and attainment to create a dichotomous measure of “lost talent,” hence losing a lot of valuable information. Furthermore, this study uses US data that is more than three decades old. Given the large-scale changes of the educational system using newer data seems necessary.

Morgan (2004), using newer US samples but focusing solely on racial gaps, claims that it is probable that expectations do cause achievements, in a way of a “Pygmalion effect,” but no evidence on differing effect of expectations by race can be unequivocally inferred.

To summarize, both studies are confined to the USA. Generating results from England could also capture local processes not seen in these studies and enable cross-national comparison. Moreover, both studies do not use multi-level methods to estimate school-level effects and cross-level interactions. As we elaborate hereinafter, school-level composition variables are important when attempting to explain educational achievements.

School-level factors

While most of the above discussed traits at the student-level (micro), many scholars now see it as essential to add higher level composition variables, claimed by many to have a unique effect on educational achievements (Khattab, 2005; Owens, 2010). For example, school composition variables (e.g. ethnic or SES composition) can influence individual educational outcomes via expectations in two somewhat contradicting ways. On the one hand, there is the “frog pond” or “comparative” effect, whereby the boundaries of the pond are the boundaries of the “reference group” and one’s relative position within this pond determines his/her level of expectations (Davis, 1966; Shavit and Williams, 1985). On the other hand, the “normative” effect suggests that pupils are not affected by their relative position within the pond but rather by the prevailing norm in the pond. In this case, high achieving peers uplift the achievement level of a student by raising the standard and creating expectations and competition (Shavit and Williams, 1985). Of course, the magnitude of the effect of the latter on the former, or vice versa, is dependent on age, the size of the gap between the individual, and his surroundings and cultural environment (Owens, 2010). In both cases, however, it is clear to see that, here too, expectations of youth play an important part connecting school composition variables to achievement.

Keeping in mind all the critiques on the “classical” models presented earlier, it seems worthwhile to also examine potential interaction effects of school composition variables with expectations: a novel attempt to the best of our knowledge. Therefore, we add another aspect to the “moderating” hypothesis—suggesting that the strength of the effect of expectations of the pupil on achievement may depend on the pupil's school SES level or ethnic composition.

In this paper, we set out to examine this third hypothesis focusing on four factors: ethnic affiliation, occupational class, school SES level, and school minority proportion, which, to the best of our knowledge, have never been comprehensively scrutinized. We chose these four variables to enable the examination of the “moderating” hypothesis with variables at both the pupil and the school level, and structural as well as cultural effects on achievement.

Before moving on to present the data and methods, we offer a short summary and formulation of all hypotheses:

The “Classical hypothesis”: the effects of background variables on achievement are strong and partly mediated by expectations.

The “Neo-liberal hypothesis”: background variables are inconsequential in the creation of expectations and achievement. Furthermore, expectations are ineffective in causing achievement.

The “Moderating hypothesis”: background variables may moderate the effect of expectations on achievement.

Data and methods

Survey characteristics

This study utilizes waves 1 to 3 of the LSYPE panel survey and matching administrative data from the National Pupil Database (NPD). The LSYPE was commissioned by the former Department for Education and Skills in England and is managed by the Department for Education. LSYPE data were collected from 2004 (wave 1) and in every consecutive year since, from a sample of 15,770 households including students aged 13–14 attending 658 different schools (for more information see National Center for Social Research, 2010: 5–6). We derived most of the measures of the independent variables in this study from wave 1. However, scholarly expectation of youth, our main independent variable, was measured using wave 2 (youth aged 14–15), while information on achievement, our dependent variable, was measured using administrative data on GCSE exams completed by the end of year 11 (aged 15–16). This procedure of temporal ordering of the measures enabled us to bypass the well-documented question of temporal ordering between expectation and achievement (c.f. Bui, 2007; Mau, 1995)—focusing solely on the unidirectional influence of the former on the latter.

We excluded from the analysis independent schools because they lack all essential school-level information and households with no present parents because they are devoid of any parental background information.

In other cases of missing data, multiple imputation was administered taking into account the multilevel structure of the data (Carpenter et al., 2011). However, this procedure cannot be used for variables that are modeled as interactions. Hence, we also excluded cases with missing values in variables interacted in the model before preforming the imputation.⁵ These exclusions left us with 14,076 pupils within 621 schools that constitute our working.

Variables

Our main dependent variable is the uncapped point score GCSE (KS4) average measured at wave 3, received from NPD. Preliminary tests determine this variable to be close to symmetric (Skewness = −0.3) with a higher peak than a normal distribution (Kurtosis = 3.0). This measure is highly correlated with the number of GCSE qualified at A*-C (p = 0.89) but retains more of the variance among the high achieving pupils and is thus closer to a symmetrical distribution.

Youth scholarly expectations, the main intermediate independent variable, comprised three survey questions, all answered at wave 2 (preceding GCSE scores), and reporting subjective estimated probabilities of the future educational carrier: the first asked whether the youth believed he/she was planning to stay on in full-time education (FTE) after compulsory education (year 11). The second asked regarding the probability, stating four possibilities of ever applying for university from “very likely” to “not likely at all.” The final question detailed the same options regarding probabilities of actually getting into university. Different combinations of these three questions were used to create a five-point scale ranging from “likely to apply and get in” to “likely to drop out of FTE at 16.”⁶

The remainder of the variables were measured at wave 1, thus preceding the measurement of both expectations and GCSE scores. We recoded self-defined ethnic origin of Young Person (YP) into seven categories: “White” (=0), Bangladeshis, Pakistanis, Black Africans, Black Caribbean, Indians, and finally all “mixed” or “other” origin. Parental occupational class is measured using National Statistics Socio-Economic Classification collapsed to three hierarchical categories and a fourth “long-term unemployed” category (Office for National Statistics, Economic Labour and Social Analysis, 2010).

We also measure and control for other household and parental characteristics. Marital status of parents differentiates between students who live with two biological parents; those who live with two parents, one not biological; and those who live with one parent. Number of siblings counts siblings of the YP currently living in the same household as YP.

We measure two aspects of social capital of family. We based the first, measuring family cohesion, on three ordinal variables that summarize the quality and quantity of intra-family relations. The first question refers to quantity of disputes between the student and their parents. The second measures frequencies of quality talks between the student and parents, and the third is the student’s response to the statement “how well does he/she get along with their parents.” We used the first dimension of a categorical principal components analysis (CATPCA) (see: Meulman et al., 2004) to combine these to one continuous.⁷ The second aspect of social capital attempts to capture the families bonds with the community, operationalized using the number of times YP has transferred schools up to wave 1.

Our cultural capital measure is the first dimension of a CATPCA solution (Meulman et al., 2004) including activities such as learning a musical instrument, participating in community work, or reading for pleasure and an additional aspect of satisfaction with school facilities.⁸

We based the scale of parental involvement in school life on four ordinal items measuring different facets of the connection of parents to their children's school life. These include the frequency of the main parent talking with YP about their school reports; frequency of the main parent participation in school activities; attendance of parents at “parents’ evenings”; and a general item on main parent’s perception of involvement in school. These items were converged using CATPCA, creating one continuous dimension to represent all items while losing minimal information (Meulman et al., 2004).⁹

Parental expectation comprises the main parent’s estimations of the probability that the YP will reach higher education using a four-point scale ranging from “very likely” to “not at all likely.” We added to this a fifth category that includes parents who did not think their child would continue in FTE beyond age 16.

We also control for gender (female = 0) and the YP’s previous scholarly achievements measured using KS2 fine-graded point score average.

Our two main school-level variables, measured in wave 1, include first the ethnic composition of the school, measured as the percentage of pupils who define themselves as part of a minority group and second, the school SES level measured using the percentage of pupils eligible for free school meals.

At the school level, we control for achievement level of the school, measured as a simple value-added measure between average KS3 and KS4 scores,¹⁰ school gender composition coded as girls only, boys only or mixed, and school type distinguishing between selective, comprehensive, and special.

Modeling and analysis

Knowing that the data set is structured in hierarchical levels, we were inclined to use multilevel models. Since part of our aim is to analyze second level effects on achievements, which rules out a “fixed effects,” a mixed models approach seems obligatory. This form of analysis is well equipped to deal with intra-class correlation, model between as well as within school variation in achievements and examine the cross-level interactions of interest (Hox, 2010: 4–6). The multilevel regression is introduced as composed from successive steps that progressively incorporate the determinants of achievements (models 1 to 6 in Table 3), enabling us not only to try and account for the variance in achievements but also to infer both direct and indirect paths of influences of independent variables or possible compositional influences.

The following equation represents the general form of the model

where $X_{k i j}$ are k control variables at the individual level, other than ethnicity, class, expectation of pupil, the two nested interaction terms, and the two cross-level interaction terms. $Z_{q j}$ are all control variables at the school level, excluding proportion of minority pupils in the school and a measure of school SES. It should be kept in mind throughout that the variables we mention explicitly are paramount due to our attempt to explore the “moderating hypothesis” posing the possibility of interactions (both cross-level and nested) of these variables and expectation of youth. The number marker placed above the different terms indicates the step number in which the term is added to our main model. It should be noted that preceding models are nested within successive ones, thus enabling the comparison of goodness of fit measures.

Results

We begin this section by presenting the initial effect of ethnicity and class on achievement. Table 1 presents the sample means of GCSE scores by the six ethnic groups of interest and the four occupational classes. Reviewing the bottom row, it is clear that, on average, the higher the class the higher the achievement. Moving to the column designating the ethnic totals, we see the average ethnic gaps, where Indians and Bangladeshis hold the highest average achievement, followed by Whites, Black Africans, Pakistanis, and finally Black Caribbeans. Using a two-way full factorial ANOVA model not including any controls, we find that both class (F = 121.5, p < 0.001) and ethnicity (F = 36.9, p < 0.001) have a strong and significant effect. These results present the achievement gaps between classes and ethnicities that all the aforementioned theories wish to explain. However, they also hint at the kind of non-linear complexities in play—the idea that the process of reproduction of status does not work in the same way for all ethnicities (Kerckhoff et al., 1989; Morgan, 2004). For example, while on average, Indians achieve higher scores than Whites, and these in turn, achieve more than Black Caribbeans, the Whites' percentage increase between working class and professionals is about 1.5 time higher than that of Black Caribbeans and around double the increase within Indians. In other words, it seems that for Whites, class has more of a reproductive power, when compared other minority groups for which class, evidently, matters less.

Table 1.

Estimated GCSE APS means of ethnic groups and occupational classes (N).

Ethnic group	Occupational class				Total
Ethnic group	L.T unemployed	Routine manual	Intermediate	High M.A.P	Total
White	213.5	297.9	372.4	426.4	369.0
White	(264)	(2960)	(2057)	(4109)	(9390)
Black African	292.6	350.8	370.3	413.5	363.6
Black African	(121)	(138)	(78)	(204)	(541)
Black Caribbean	263.6	300.1	344.5	371.0	335.2
Black Caribbean	(45)	(156)	(106)	(226)	(533)
Bangladeshi and Pakistani	345.6	348.5	374.3	422.5	362.0
Bangladeshi and Pakistani	(486)	(548)	(333)	(180)	(1547)
Indian	339.2	397.7	436.8	466.5	425.7
Indian	(58)	(350)	(267)	(286)	(961)
Total	300.0	314.8	378.1	425.4	371.0
Total	(974)	(4152)	(2841)	(5005)	(12,972)

Note: Totals presented are weighted sample averages of every group, thus representing means of the respective group disregarding the second variable in the model. We exclude “Other and mixed” in this table.

Before adding expectation to the analysis and examining the effects of class and ethnicity on the relation between expectations and achievement, we present the distribution of high expectations among ethnicities and classes. Chart 1 shows an interesting relation among the three variables—namely, the lower the class, the larger the variance in expectations among ethnic groups. Going into further details, at the one extreme, Whites seem to exhibit much lower expectations throughout all occupational classes, while at the same time presenting a far larger and clearly hierarchical effect of class on expectations. At the other end, portraying the exact opposite to these characteristics, Black Africans show extremely high expectations across the board and no clear trend connecting class to expectations. The great variety among classes and ethnic groups, especially among whites and the lower classes, perforates the claim that expectations are high across the board, as expected by the “neo-liberal” scholars. However, the position displayed by Black Africans is incomprehensible under the classical, origin-expectation-destination hypothesis, for it means that within this group, class is uncorrelated with youth expectations.

Chart 1.

Proportion of pupils holding highest expectation level by class and ethnicity.

Having showed that expectations vary across classes and ethnicities, we turn to the moderation effect. Chart 2 confirms that, at least prior to controlling for possible confounding effects, ethnic groups and classes do indeed gain different returns on their expectations.¹¹ Scanning the panels clockwise shows the hierarchical effect of class at both expectation levels, for all ethnic groups. However, as also noted earlier, before including expectations in the analysis, this seems to be especially true for Whites who display a larger difference between “routine manual” and “high professionals,” in comparison to all other ethnic groups. In other words, the effect of class is not constant across ethnicities (F = 5.1, p < 0.001). Furthermore, comparing the two left panels (highest and lowest occupational classes) to the two right panels (the two intermediate classes), we note that, over and above the different ethnicities, in the former case, the effect of expectation (the slope of the lines) is more pronounced (F = 2.3, p = 0.03), especially when comparing the lowest class to all other panels.

Chart 2.

GCSE APS by occupational class, ethnicity, and expectation level.

Finally, looking at the slopes of the lines within the panels reveals that Whites and Indians' expectations have the highest impact on achievement, while at the other end of this scale are the Black Caribbeans, for whom expectations make much less of a difference in achievement (F = 4.0, p <0.0001). From this, we tentatively conclude that expectations are of higher value for youth of White or Indian ethnicity, within the high professional class and even more so for those defined as long-time unemployed.

Now that we have closely examined the two level 1 interactions of interest, in the following, we provide the opportunity to examine the gross effects of our two main school-level variables—minority and SES composition—on achievements.

Starting with the unadjusted effects (the first two rows of Table 2), we note that for every 10% increase in the percentage of free meal eligibility, GCSE average decreases by an estimated 5.2 points. In contrast, for every 10% increase in percentage of minority pupils, GCSE average increases by an estimated 1.1 point. The drop in both coefficients of interest seen in model 2 demonstrates that part of the effects found in model 1 can be attributed to school gender and type, although both retain their significant effect even after equalizing these factors among the different SES or minority composition levels. Moving on to model 3, it seems that once the expectation level of the pupil is leveled, the school context influence on pupil's achievements diminishes to a degree where minority composition is no longer consequential, and a 10% increase in school SES level adds only an estimated 3.1 points to the GSCE average. This supplementary drop in coefficients seems to imply that expectation level of pupils mediates part of the effects of both SES and minority composition of school.

Table 2.

HLM regressions on GCSE (KS4) APS by school-level variables and expectation level.

Model	Fixed effects	Estimate	R² level 1	R² level 2
Unadjusted effects	Ln of school SES	−52.2	–	0.45
Unadjusted effects	Ln of ethnic composition	11.6	–	0.45
Adding schoolgender and type	Ln of school SES	−38.6	–	0.61
Adding schoolgender and type	Ln of ethnic composition	6.9	–	0.61
Adding pupil'sexpectation	Ln of school SES	−31.1	0.22	0.69
Adding pupil'sexpectation	Ln of ethnic composition	−2.7^	0.22	0.69
Full model	Constant	366.5	0.23	0.69
	Ln of school SES	−30.7
	Ln of ethnic composition	−3.2
	[Compared to comprehensive]
	Selective school	62.8
	Special school	−185.6
	[Compared to mixed]
	Only boys	0.7^
	Only girls	24.3
	Expectation of pupil	48.4
	School SES × expectation	2.3
	Ethnic composition × expectation	−1.0^

Note: The analysis of all models included 14,076 pupils nested within 621 schools. School SES, ethnic composition, and expectation of the pupil are grand mean centered. ^coefficient was found insignificant at p < 0.05.

Finally, the inclusion of the two cross-level interaction terms of interest in model 4 completes the picture and enables the preliminary examination of the question raised by the moderating hypothesis—can school characteristics like SES or minority composition cause variation in the relation between expectation and achievement?

Reviewing model 4, we find that expectation still has a strong positive estimated effect on achievements. Likewise, when expectation of the pupil is set to average, both school SES and minority composition variables have a significant negative effect on achievements. However, the positive interaction term referring to SES of school and expectations confirms that as school SES amplifies the effect of expectation of pupil is reduced.¹² Contrary to these findings, the negative interaction between the percentage of minority pupils in a school and expectation indicate that a higher percentage of minority pupils in the school decreases the effect of expectations. However, this effect is insignificant.

Although this analysis undoubtedly lacks controls for important variables, it does demonstrate two important points that could not have been exhibited in a more complex model: first, the effect of school context variables is indeed mediated by expectations of the pupil, suggesting that creation of expectations in the pupil is dependent on school SES and minority composition. Second, much in accordance with the “moderating hypothesis,” the efficacy of expectations to create achievements is also dependent on school SES level.

Modeling achievement—Sketching a possible mechanism

Table 3 presents our main multilevel analysis, combining controls for pupil-level variables and school-level variables. We start by presenting achievement differentials among different ethnic groups while controlling for socio-demographic variables like gender, parental education level, and previous achievement. Note that all these variables have an independent significant effect on achievement in the direction theoretically expected, cumulatively accounting for 45% of level 1 variance and 69% of intercept variance. In addition, it seems that a large portion of the gaps among occupational classes are attributable to these variables, since all class-related gaps drop considerably when compared to means presented in Table 1. Moving to ethnic-related gaps, the rather dramatic rise in gaps between Whites on the one hand and Black Africans and Pakistanis on the other, suggest that these groups would have higher achievements if their parents had the education and class distribution of Whites.

Model 2 proceeds by adding our key mediator, as theorized by all classical models, and primary focus—pupil's educational expectations.¹³ Contrary to the stance proposed by staunch neo-liberal advocators, expectations of the pupil still hold a durable and significant effect on later achievement, even after controlling for all other variables in the model. Likewise, the colossal drop in all ethnic-related coefficients reinforces the claimed mediation role prescribed to expectations in the classical hypothesis. However, in current day England, it seems that, as also aforementioned, minority groups are the ones in possession of higher expectations, which helps elevate their achievement (Khattab and Modood, 2017). In sharp contrast to the large differences in ethnic gaps between the models, class-related gaps are hardly affected by the introduction of expectations. However, it is vital to remember, this does not mean that pupils' expectation level does not mediate class effect on achievement, but rather that this is true only after controlling for other variables in the model.¹⁴

Table 3.

HLM regressions on GCSE APS.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Fixed effects
Constant	409.5	401.7	402.9	425.2	423.9	424.5
Other or mixed	29.1	16.6	14.1	13.0	15.1	15.1
Black African	48.4	24.0	29.7	17.7	19.7	19.4
Black Caribbean	8.3^	−3.5^	−0.9	1.3^	5.0^	4.8
Bangladeshi	70.4	54.9	56.4	37.2	36.7	37.4
Pakistani	64.0	44.2	46.2	24.2	25.7	25.0
Indian	78.2	57.9	57.3	39.6	41.7	40.8
Intermediate occupations	−3.0^	0.2^	0.1^	3.7^	3.8^	3.6^
Manual occupations	−24.5	−19.1	−19.2	−10.2	−9.3	−9.7
Long-term unemployed	−39.4	−37.8	−37.4	−19.9	−19.4	−19.3
Male	−29.4	−24.2	−24.0	−22.5	−23.1	−23.3
Higher education below degree	−22.4	−16.5	−17.2	−14.4	−13.9	−14.1
GCSE a level or equivalent	−25.5	−16.4	−17.2	−13.4	−12.7	−13.0
GCSE grades a–c or equivalent	−35.5	−25.7	−26.4	−18.2	−17.0	−17.2
Qualifications at level 1 and below	−44.8	−33.3	−33.9	−21.6	−20.0	−19.7
On qualifications	−49.8	−39.7	−40.1	−28.1	−26.3	−26.5
Key stage 2 achievements^a	23.7	21.0	21.0	19.5	19.3	19.3
Expectations of pupil^a		21.4	19.7	9.3	9.3	10.2
Expectations × other or mixed			3.3^	1.3^	1.2^	−0.4^
Expectations × Black African			−8.9^	−0.5^	0.1^	−2.1^
Expectations × Black Caribbean			−9.6	−7.4^	−7.4^	−9.7
Expectations × Bangladeshi			−11.5	−3.8^	−3.2^	−5.2^
Expectations × Pakistani			−7.5	−1.6^	−1.3^	−2.7^
Expectations × Indian			−0.4^	0.8^	0.6^	−0.2^
Expectations × intermediate occup.			3.4^	2.8^	2.9	2.4
Expectations × manual occupations			2.4^	1.1^	1.2	0.01
Expectations × unemployed			11.4	8.3^	7.9^	5.7^
Living with two parents, one step				−19.0	−18.8	−18.7
Lone parent family				−31.3	−30.6	−30.9
Number of siblings of pupil^a				−2.8	−2.7	−2.6
Fairly high parental expectations				−14.6	−14.7	−14.6
Medium parental expectations				−34.3	−34.3	−34.4
Fairly low parental expectations				−37.1	−37.2	−37.1
Low parental expectations				−66.6	−66.7	−66.2
Parental involvement^a				−9.9	−9.7	−9.7
Cultural capital of pupil^a				−9.2	−8.6	−8.7
Family social capital of pupil^a				−8.6	−8.8	−8.6
Community social capital of pupil^a				−6.5	−6.5	−6.4
SES of school^a					−6.4	−6.2
Ethnic composition^a					−5.7	−5.7
Selective school					11.6	15.1
Special school					−40.4	−40.3
Unisex boys					15.7	14.8
Unisex girls					−15.8	−15.8
School achievement level^a					856.7	852.9
School SES × expectations						2.2
Ethnic composition × expectations						0.1
Variance and goodness of fit
Within school ( $e_{i j})$	11,521	10,911	10,889	10,095	10,091	9875
Constant ( $u_{0 j})$	1929	1843	1841	1713	1330	1322
Expectations slope ( $u_{1 j})$						117
Pseudo R² level 1	0.45	0.48	0.48	0.52	0.52	0.53
Pseudo R² level 2	0.69	0.70	0.70	0.72	0.78	0.78
Total R²/R² expectation slope	0.50	0.53	0.53	0.56	0.58	0.04^b
F test of added variables	857 (7, 25,604)	508 (1, 2954)	2 (9, 12,528)	51 (11, 11,181)	69 (1, 17,158)	3 (2, 3900)

Note: The analysis of all models included 14,076 pupils nested within 621 schools.

^coefficient was found insignificant at p < 0.05.

^aVariable is grand mean centered.

^bMeasure refers to R² of expectation slope.

In model 3, we examine the “moderating hypothesis” by interacting both class and ethnicity with expectation of the pupil. All of the coefficients, except in the case of “other and mixed” are negative, suggest that all of these groups gain less from an increase in their expectation level when compared to Whites. Note that three coefficients (Bangladeshis, Pakistanis, and Caribbean pupils) are also statistically significant. Contrary to this, it seems that lower classes actually gain more from every increase in their expectations when compared to “High Managerial, Administrative, Professional class.” An intuitive way to convey this idea is to chart down predictions for the different social groups, using the fixed portion of the equation only, where the X corresponds to expectation level, Y is predicted GCSE score, and the lines represent different social groups (see top panel of Chart 3).

Chart 3.

Estimated GCSE APS by expectations level, ethnicity, and class.

The chart shows that while high professional Whites gain, on average, 19.7 points for every increase in expectations, unemployed Whites would gain slightly more than a 31-point increase to their GCSE APS for every 1-unit increase in expectations. As for ethnic gaps, Bangladeshis’ gain is estimated at no more than 8.2 and Black African and Caribbean at around 10 points for the same increase in expectation. The overall result is clear—although achievement gaps are still present at the highest expectation level, a pronounced reduction in these gaps between classes and ethnic groups takes place as expectation level increases.

In model 4, we add variables, conceptualized by past researchers as possible causes of achievement, either via the expectation level or directly, such as inner family and community social capital; family composition; cultural capital; parental expectations from the pupils; and their involvement in school life. All variables have a unique significant effect on achievement and help explain a further 4 and 2% of level 1 and intercept variance, respectively. In addition to their direct effect on achievement, they play a significant role as a mechanism through which class, ethnicity, parental educational level, and pupils’ expectations connect to achievement. Furthermore, the estimated differences in the effect of expectation among social groups have decreased below the significant level. However, the effect of expectation for the unemployed is close to double that of high professionals, while the same effect, when looking within Black Caribbeans, is almost canceled out.

Model 5 proceeds by adding all school-level variables. They seem to significantly improve model fit and increase the explained portion of the intercept variance by 6% where, as expected, level 1 variance remains stable. Regarding school-level control variables, it seems sufficient to say that the results are mostly attuned and significant, showing average differences in achievement between schools are indeed partly attributed to them even after controlling for many pupil-level characteristics, as a few notable studies have shown (Owens, 2010; Portes and Hao, 2004). For example, school SES and minority composition have a strong and significant impact, suggesting that lower school SES and higher minority composition are associated with lower average achievement at the school level.

As a final augmentation to the model, we add cross-level interaction terms between school SES and ethnic composition and expectations to reveal possible moderating effects of school context. As the predictions delineated in Chart 4 show, by using this model, we find that the lower the SES of the school the higher the returns on every expectation unit attained by the pupil. Ethnic composition, however, seems to have a negligible impact on the slope of expectations, as can be seen when comparing the two dashed or solid lines to each other.

Chart 4.

Estimated GCSE APS by School SES, ethnic composition and expectation level.

Discussion and concluding remarks: The ability to attain expectations and the ability to utilize them

This study proposes a theoretical mechanism, the “moderating hypothesis,” claiming that educational achievements are also determined by the capacity of structural and cultural background variables (including school level variables) to simultaneously alter the strength of the effect of expectations on achievement. Incorporating school, household, and pupils’ data derived from the first three waves of the LSYPE survey, we have argued that it is vital to decompose the generalized origin-expectation-destination linear claim (Bozick et al., 2010; Ganzach, 2000). We have shown that when permitted by the model, the effect of expectations on achievement varies; it is strong within some social groups or schools (adhering to classical convention), while within others, it loses its potency as claimed by “neo-liberal” proponents (Archer, 2007; Schneider and Stevenson, 1999).

Doing so prompts our understanding of inequality and social mobility by analytically distinguishing between the ability of a social group or school to attain expectations and their ability to transform the gained expectations into achievement. This distinction helps us explain why gaps persist (as claimed by the “reproduction” hypothesis) in an age of “high hope” (as claimed by “neo-liberals”). Furthermore, it may even offer some insight into whether (or in what cases if at all) gaps persist despite high expectations or rather due to low expectations (as originally claim by the classical model).

We began exploring the sustainability and benefits of the moderating hypothesis by presenting gaps in expectation level and achievement among all social groups of interest. Here, we note that variance in expectations and achievement is still very much related to both class and ethnicity. These results repudiate, to some extent, the claim that “expectations are high across the board” and that achievement is “decoupled” from family characteristics. However, the same results also clarify that within some groups, the effect of class or ethnicity on expectations, and indeed on achievements, has become rather negligible, opposing the generalized claim that social background variables uniformly effect expectations or achievements.

Turning next to the triad effects of class, ethnicity, and expectations on achievements, we unearth two further points: first, it seems that expectations still matter, i.e. they have an effect on achievement over and above the different social groups. Second, the impact of expectations on achievement is contingent upon the social group on which we focus. More specifically, this brings to light that it is not only that within some groups ethnicity or class affect expectations more than within others but also that some groups possess a higher capacity to transform expectation into achievement than others—especially when looking within the lowest occupational class or Whites. The power of family context variables, both structural and cultural, to alter the returns on expectation level independently of their effect on expectations or achievement is overlooked under “classical,” “reproduction,” or “neo-liberal” models.

These results gave us a first glance at the capacity of family context variables to moderate the effect of expectations on achievements once a given expectation level has been determined. The next step was the inspection of the hypothesized school context effects, before controlling for pupils’ characteristics. We find that both school-level SES and minority composition have an effect on achievements that is partly mediated by pupil's expectations, affirming the “classical” notion of the centrality of expectations also when analyzing the compound, “overall” mechanism of school-level effects. More specifically, these results lend support to the “normative” effect for it seems that high achieving schools (high minority percentage and high SES schools) create higher expectations, and thus higher achievements (Shavit and Williams, 1985). And yet, closely examining this average effect of expectations by different school types demonstrates that the effect of the pupil's expectation is not constant across schools—rather, it is dependent on school SES level so that within high SES schools’ expectation level of the pupil becomes almost irrelevant, reminiscing a school-level version of the “neo-liberal” notion.

The final endeavor was to paste both levels of analysis together in an attempt to inch towards a model presenting net effects of class, ethnicity, school SES level, minority composition, expectations and, most importantly, the corresponding interactions that constitute the moderating hypothesis.

Focusing first on influences of the “family background” variables, we find that, as would be predicted by the “classical” and “reproduction” hypotheses, both ethnic- and class-related gaps are significant, even after controlling for “pre-existing” characteristics of family and the pupil (previous achievement, gender and parental education, which all also carry a direct impact on achievement). Furthermore, although a large portion of the unconditional gaps among the social groups are explained by many of the variables suggested as factors by theories conglomerated here as “classical models” (community and family social capital (Coleman, 1988), cultural capital (Bourdieu, 1984), parental expectations and involvement (Bozick et al., 2010; Jencks et al., 1983), family structure (Breen and Goldthorpe, 1997), and school-level variables (Owens, 2010; Shavit and Williams, 1985), some of these gaps remain significant throughout. The remaining gaps often remain unexplained also in present-day UK research (Strand, 2011).

Zooming in on the role of expectations, we found that part of the effect of these family context variables does indeed flow via expectations, meaning that they cause variation in expectations, which in turn affects achievement. This lends support to the “classical” hypothesis for it seems that expectations are still of importance, being the cause of these gaps (Alexander and Eckland, 1975; Jencks et al., 1983).

Moving our focus to school-level effects, having controlled for pupil-level variables, in accordance with the “classical” notion, pupil's expectation level mediates a large part of these effects (Bozick et al., 2010; Jencks et al., 1983). Finally, reinforcing research promoting the unique importance of school-level effects (e.g. Goddard et al., 2009), we note that taking into account compositional differences between schools and individual characteristics, helps us explain another large portion of the effects of school SES and minority composition.

Findings stressed so far—pointing to a strong class, ethnic origin, school SES, and minority composition effect, which is mediated by expectations (among other variables) —seem to somewhat rebuttal the “neo-liberal” hypothesis and reinforce the general and renowned “classical” models. However, the main question from this stage onwards becomes whether, and to what extent, are raising expectations beneficial within each specific social group or school prototype.

Here, the analysis revealed that for some groups, such as the lowest class and Whites, it seems that expectations are still a rare and valuable resource in two ways: first, average expectation level within the groups is lower compared to other groups. Second, the relatively strong effect of expectations on achievement within these groups indicates that within them, expectations still have the power to generate achievements. Integrating these two findings, we note that the mechanism causing achievement within these groups is adequately explained by the “classical” model and contradicts the neo-liberal notion (Schneider and Stevenson, 1999). It seems that for some disadvantaged groups (this is true for Whites after controlling for class), expectations hold the potential to raise achievement and generate social mobility if only average expectations were higher. Hence, in these cases, it is not “heightened expectations” that hinder high achievement but rather a persistent low average expectation level that causes the hidden potential of expectations to be largely unrealized. Finally, this hidden potential of expectations also refutes, to some extent, the social “reproduction” theory (Blossfeld and Shavit, 1993) since evidently, when expectations are high, inequality among the social groups (both ethnicities and social classes) is smaller.

This pattern seems to adequately describe not only certain social groups but also certain schools. It was found that the lower the SES of the school, the higher the returns on every expectation unit attained by the pupil. To conclude, when looking at between-school variation in achievement, it seems that expectations hold great potential for mobility, which is, however, largely unrealized due to inequitable distribution of expectation level between the schools.

Shifting focus to other groups, such as high class or Bangladeshi origin, a mirror image of this pattern is illuminated—the high average expectation level means that here they are plentiful, but their small relative effect means they are of less significance. Results within these groups pose a challenge for both the “classical” and the “reproduction” hypotheses since within these groups, inequality is lower possibly due to the smaller effect of class or ethnicity on expectations and of expectations on achievement. These are therefore closer to the “neo-liberal” notion where expectations are high and only slightly impact achievements. Following this logic, the pattern seen in these groups might be an omen of things to come, explained as a general devaluation process where expectations are high across the board and therefore rendered almost inconsequential to achievement (Goyette, 2008; Jacob and Wilder, 2010). An alternative, more circumstantial explanation could be that these advantaged groups receive all the support they need to carry them to the pinnacle of achievements, even if they themselves did not internalize the value of education. Supporting this claim, we find that variables capturing this support system partly explain not only variance in expectations but also the gaps between groups in the effect of expectations. Hence, part of the reason advantaged groups have a more moderate effect of expectations is due to their more advantaged stance in these factors. The same can be said regarding the more advantaged SES schools, within which achievement of the pupil is less dependent on his or her expectation level. Resembling the case of advantaged social groups, this result may be related to higher resources associated with high SES school (e.g. teacher quality and teacher pupil ratio) that causes expectations of the pupil to be almost unneeded.

Two further pattern possibilities remain—the two extreme and opposed cases where, on the one hand, both average expectation level and their effect are low, or, on the other hand, both are high. The former case, evidently best related in the UK to Black Caribbeans, has already been described as the “lost talent” of low classes in the US (Hanson, 1994). Black Caribbeans suffer from two analytically distinguishable penalties—one caused by their inability to attain and hold high expectations (acknowledged under the “classical” premise) and another caused by the low return given to them on expectations in the rare cases where they are attained (acknowledged under the “neo-liberal” premise). Furthermore, it seems that contrary to the previous groups presented, Black Caribbeans' moderate slope, when compared to Whites, is only very mildly explained by family or school context variables that theoretically explain a significant part of the achievement gaps between the groups and cause expectations.

The latter case, exhibited by Indians, seems to bring us back to the “classical” theory in which expectation facilitates achievement contradicting the devaluation of expectations, and yet, their relatively high expectations, throughout the class distribution, supports a “neo-liberal” notion.

The diversity group-specific patterns just presented attest, first and foremost, to the importance of relaxing the classical assumption in which context variables are internalized and manifested in expectations and ultimately as achievements. When attempting to explain scholarly achievement gaps between individuals, groups, or schools, it is essential to analytically distinguish between the ability to attain expectation and the ability to convert them into achievements. This line of thought opens an aperture by which background variables are conceptualized as influential not only in the process of consolidating expectations but also as affecting the ability of the individual to actually utilize the expectations he or she were able to attain.

Footnotes

Acknowledgement

The authors would like to thank the Longitudinal Surveys Team, Department for Education and UK Data Service for providing access to the data. They would also like to thank Laurence Lessard-Phillips and the anonymous reviewer for their helpful and constructive comments. The help of Miriam Shaath in proofreading the paper is greatly appreciated.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Ainsworth

(2002) Why does it take a village? The mediation of neighborhood effects on educational achievement. Social Forces 81(1): 117–152.

Alexander

Eckland

(1975) Contextual effects in the high school attainment process. American Sociological Review 40(3): 402–416.

Archer

(2007) Making Our Way Through the World: Human Reflexivity and Social Mobility. Cambridge: Cambridge University Press.

Beck

Beck-Gernsheim

(2002) Individualization: Institutionalized Individualism and Its Social and Political Consequences. London: SAGE.

Blossfeld

Shavit

(1993) Persisting barriers: Changes in educational opportunities in thirteen countries. In: Shavit

Blossfeld

(eds) Persistent Inequality: Changing Educational Stratification in Thirteen Countries. Boulder: Westview Press, pp.1–24.

Borgna

Contini

(2014) Migrant achievement penalties in Western Europe: Do educational systems matter? European Sociological Review 30(5): 670–683.

Bourdieu

(1973) Cultural reproduction and social reproduction. In: Brown

(ed.) Knowledge, Education, and Cultural Change: Papers in the Sociology of Education. London: Tavistock, pp.71–112.

Bourdieu

(1997) The forms of capital. In: Halsey

Lauder

Brown

et al . (eds) Education: Culture, Economy and Society. Oxford: Oxford University Press, pp.46–58.

Bourdieu

Passeron

J-C

(1990) Reproduction in Education, Society and Culture. 2nd ed. London: SAGE.

10.

Bourdieu P (1984) Distinction: a Social Critique of the Judgement of Taste. Cambridge, Mass.: Harvard University Press.

11.

Boudon R (1974) Education, Opportunity, and Social Inequality: Changing Prospects in Western Society. New York: John Wiley & Sons Canada, Limited.

12.

Bozick

Alexander

Entwisle

et al . (2010) Framing the future: Revisiting the place of educational expectations in status attainment. Social Forces 88(5): 2027–2052.

13.

Breen

Goldthorpe

(1997) Explaining educational differentials: Towards a formal rational action theory. Rationality and Society 9(3): 275–305.

14.

Breen R and Jonsson JO. (2005) Inequality of opportunity in comparative perspective: Recent research on educational attainment and social mobility. Annual Review of Sociology 31: 223–243.

15.

Breen

Luijkx

Müller

et al . (2009) Nonpersistent inequality in educational attainment: Evidence from eight European Countries. American Journal of Sociology 114(5): 1475–1521.

16.

Breen

Yaish

(2006) Testing the Breen-Goldthorpe model of educational decision making. In: Morgan

Grusky

Fields

(eds) Mobility and Inequality: Frontiers of Research in Sociology and Economics. Stanford: Stanford University Press, pp.232–258.

17.

Bui

(2007) Educational expectations and academic achievement among middle and high school students. Education 127(3): 328–331.

18.

Carpenter

Goldstein

Kenward

(2011) REALCOM-IMPUTE software for multilevel multiple imputation with mixed response types. Journal of Statistical Software 45(5): 1–14.

19.

Coleman

(1988) Social Capital in the creation of human capital. American Journal of Sociology 94: S95–S120.

20.

Davis

(1966) The Compus as a frog pond: An application of the theory of relative deprivation to career decisions of college men. American Journal of Sociology 72(1): 17–31.

21.

Deary

Taylor

Hart

et al . (2005) Intergenerational social mobility and mid-life status attainment: Influences of childhood intelligence, childhood social factors, and education. Intelligence 33(5): 455–472.

22.

Fryer RG Jr and Levitt SD (2004) Understanding the black-white test score gap in the first two years of school. Review of Economics and Statistics 86(2): 447–464.

23.

Ganzach

(2000) Parents’ education, cognitive ability, educational expectations and educational attainment: Interactive effects. British Journal of Educational Psychology 70(3): 419–441.

24.

Goddard

Salloum

Berebitsky

(2009) Trust as a mediator of the relationships between poverty, racial composition, and academic achievement: Evidence from Michigan’s public elementary schools. Educational Administration Quarterly 45(2): 292–311.

25.

Goldthorpe

(1996) Class analysis and the reorientation of class theory: The case of persisting differentials in educational attainment. British Journal of Sociology 47(3): 481–505.

26.

Goodman A, Gregg P and Chowdry H (2010) Poorer Children's Educational Attainment: How Important are Attitudes and Behaviour? Joseph Rowntree Foundation York.

27.

Granovetter

(1985) Economic action and social structure: The problem of embeddedness. American Journal of Sociology 91(3): 481–510.

28.

Green AE and White RJ (2007) Attachment to Place Social Networks, Mobility and Prospects of Young People. York: Joseph Rowntree Foundation and Warwick University.

29.

Goyette KA (2008) “College for Some to College for All: Social Background, Occupational Expectations, and Educational Expectations over Time.” Social Science Research 37(2): 461–484.

30.

Hanson

(1994) Lost talent: Unrealized educational aspirations and expectations among U.S. youths. Sociology of Education 67(3): 159–183.

31.

Heath

Brinbaum

(2007) Guest editorial: Explaining ethnic inequalities in educational attainment. Ethnicities 7(3): 291–304.

32.

Hox

(2010) Multilevel Analysis: Techniques and Applications. London: Routledge Academic.

33.

Jacob

Wilder

(2010) Educational Expectations and Attainment. Cambridge: National Bureau of Economic Research.

34.

Jencks C, Smith M, Ackland H, Bane MJ, Cohen D, Gintis H, Heyns B and Michelson S (1972) Inequality: a Reassessment of the Effect of Family and Schooling in America Photocopy. New York: Basic Books.

35.

Jencks

Crouse

Mueser

(1983) The Wisconsin model of status attainment: A national replication with improved measures of ability and aspiration. Sociology of Education 56: 3–19.

36.

Johnson

Brett

Deary

(2010) The pivotal role of education in the association between ability and social class attainment: A look across three generations. Intelligence 38(1): 55–65.

37.

Kao G and Tienda M. (1995) Optimism and Achievement: The Educational Performance of Immigrant Youth. Social science quarterly 76(1): 1–19.

38.

Kao

Thompson

(2003) Racial and ethnic stratification in educational achievement and attainment. Annual Review of Sociology 29(1): 417–442.

39.

Karen

(2002) Changes in access to higher education in the United States: 1980–1992. Sociology of Education 75(3): 191–210.

40.

Kerckhoff

Campbell

Trott

Kraus

. (1989) The transmission of socioeconomic status and prestige in Great Britain and the United States. Sociological Forum 4(2): 155–177.

41.

Khattab

(2005) The effects of high school context and interpersonal factors on students’ educational expectations: A multi-level model. Social Psychology of Education 8(1): 19–40.

42.

Khattab

Modood

(2017) Accounting for British Muslim’s educational attainment: Gender differences and the impact of expectations. British Journal of Sociology of Education 39(2): 242–259.

43.

Kim

(2002) Ethnic differences in academic achievement between Vietnamese and Cambodian children: Cultural and structural explanations. The Sociological Quarterly 43(2): 213–235.

44.

Kristen

Granato

(2007) The Educational attainment of the second generation in Germany: Social origins and ethnic inequality. Ethnicities 7(3): 343–366.

45.

Marjoribanks

(2002) Family and School Capital: Towards a Context Theory of Students' School Outcomes. Norwell, MA: Springer Science & Business Media.

46.

Marjoribanks

(2005) Family environments and children’s outcomes. Educational Psychology 25(6): 647–657.

47.

Mau

(1995) Educational planning and academic achievement of middle school students: A racial and cultural comparison. Journal of Counseling & Development 73(5): 518–526.

48.

Meulman

JJ, V, D

Kooij

Heiser

(2004) Principal components analysis with nonlinear optimal scaling transformations for ordinal and nominal data. In: Kaplan

(ed.) The Sage Handbook of Quantitative Methodology for the Social Sciences. London: SAGE, pp. 49–70.

49.

Mickelson

(1990) The attitude-achievement paradox among black adolescents. Sociology of Education 63(1): 44–61.

50.

Modood

(2004) Capitals ethnic identity and educational qualifications. Cultural Trends 13(2): 87–105.

51.

Morgan SL (1996) “Trends in Black-White Differences in Educational Expectations: 1980–92.” Sociology of Education 69(4): 308.

52.

Morgan

(2004) Methodologist as arbitrator: Five models for Black-white differences in the causal effect of expectations on attainment. Sociological Methods & Research 33(1): 3–53.

53.

National Centre for Social Research (2010) LSYPE User Guide to the Datasets: Wave One to Wave Six. UK: Department for Education.

54.

Office for National Statistics, Economic Labour and Social Analysis (2010) The National Statistics Socio-Economic Classification (NS-SEC Rebased on the SOC2010). Newport: Office for National Statistics.

55.

Owens

(2010) Neighborhoods and schools as competing and reinforcing contexts for educational attainment. Sociology of Education 83(4): 287–311.

56.

Portes

Hao

(2004) The schooling of children of immigrants: Contextual effects on the educational attainment of the second generation. Proceedings of the National Academy of Sciences of the United States of America 101(33): 11920–11927.

57.

Ray

(2006) School Value Added Measures in England: A Paper for the OECD Project on the Development of Value-Added Models in Education Systems. London: Department of Children, School and Family.

58.

Rosenbaum

(2001) Beyond College for All: Career Paths for the Forgotten Half. New York: Russell Sage Foundation.

59.

Schneider

Stevenson

(1999) The Ambitious Generation: America's Teenagers, Motivated but Directionless. New Haven and London: Yale University Press.

60.

Shah B, Dwyer C and Modood T (2010) Explaining Educational Achievement and Career Aspirations among Young British Pakistanis: Mobilizing ‘Ethnic Capital’? Sociology 44(6): 1109–1127.

61.

Shavit

Williams

(1985) Ability grouping and contextual determinants of educational expectations in Israel. American Sociological Review 50(1): 62–73.

62.

Shavit

Yaish

Bar-Haim

(2007) The persistence of persistent inequality. In: Scherer

Pollak

Otte

et al . (eds) From Origin to Destination: Trends and Mechanisms in Social Stratification Research. New York: Campus Verlag, pp.37–57.

63.

Stewart

Simons

(2007) The effect of neighborhood context on the college aspirations of African American adolescents. American Educational Research Journal 44(4): 896–919.

64.

Strand

(2011) The limits of social class in explaining ethnic gaps in educational attainment. British Educational Research Journal 37(2): 197–229.

65.

White

Kaufman

(1997) Language usage, social capital, and school completion among immigrants and native-born ethnic groups. Social Science Quarterly 78: 385–398. ‏

66.

Wrong

(1961) The oversocialized conception of man in modern sociology. American Sociological Review 26(2): 183–193.