The stratification of learning opportunities and achievement in Tasmanian secondary schools.

Social differences in curriculum access and student achievement are a feature of schooling in Australia. These differences are usually documented at the senior secondary level where the curriculum is most differentiated, but the present study found large social, gender and school differences in levels of study at a more junior year-level. Using detailed information from the certificate records of 5491 Year 10 students, this study found that in English and mathematics the level of study students access is strongly related to socioeconomic background, gender and school attended. Differences between schools, measured using multilevel modelling techniques, are related to the type of school, enrolment size, and social composition. The study also found strong effects of curriculum location in the middle years of school on school completion, access to academic courses in senior school, and final year achievement. The results show significant between-school effects suggesting that schools can make a difference.

Introduction

Research on educational inequality has highlighted the role of segregation in the formation of social differences in school achievement and educational outcomes. Although differences can emerge because children who are unequally prepared enter the same classrooms, much of the research reveals that differences emerge because children from different social backgrounds do not receive the same schooling. Students are separated according to the type of school they attend, the area in which they live, the track they enter, and the subjects they take. Organisationally their schooling is divided by type of school or by stream as well as by content and pedagogy. Even if children begin in the same classes, as they ascend school they become separated at key turning points. The further they go, the more differentiated becomes their schooling.

The upshot of this form of stratification is that children from different social backgrounds do not have access to the same learning opportunities. Those who enter private schools are more likely to enrol in academic programs rather than vocational programs. In the senior years, those in academic programs tend more often to take the most profitable subjects, profitable in terms of promoting access to university and to prestigious courses within university. The formal separation of students through the organisation of schools, pupils, and the curriculum produces differences in relationships to school knowledge based on social origin, differences in access to educational resources, differences in the nature and extent of learning, and, ultimately, differences in school success and career choice.

This paper examines social differences in learning opportunities and achievement in the secondary schools of Tasmania. It measures the extent to which the organisation of the curriculum in Tasmanian schools operates as a source of social differentiation. Original data on the levels of study of 5491 Year 10 students in the core subject areas of English and mathematics are used to explore the interactions among social background, gender, school, and access to the curriculum. After controlling for the background characteristics of students, multilevel modelling procedures are used to assess school-level differences in students' chances of being located in particular positions in the curriculum. This analysis considers the question of whether or not some schools promote more equitable patterns of curriculum access. The effects of curriculum differentiation on school achievement and school outcomes are also examined to ascertain whether or not the position students occupy in the curriculum in the middle school years influences their achievement and outcomes in the senior years.

Research on the stratification of learning opportunities and achievement

Much work has been conducted on educational stratification overseas. One strand of this work has reported social differences in the populations of young people studying different types and levels of courses, differences which are maintained after controlling for other factors such as ability and aspirations (e.g. Abraham, 1995; Oakes, 1985). The consequence of these variations in course taking is that children from different backgrounds receive differential treatment and encounter different learning environments. In effect they receive a different schooling by being exposed to qualitatively different types of knowledge, methods of teaching, and resources attached to the different programs or courses in which they are located. Oakes (1985) contended that it leads to differences in rates of learning, academic achievement, entry to post-school education, and employment.

Other work has reported that the extent of stratification varies across schools. Several researchers have shown that the relationship between social background and academic achievement is weaker in some schools than in others, and that some schools are more effective at promoting students' progress than others (e.g. Thomas, Sammons, Mortimore, & Smees, 1997). In exploring the types of schools that are more effective, some work has highlighted differences between private and government schools (e.g. Coleman, Hoffer, & Kilgore, 1982). Other work has shown that course placement and achievement is related to the social composition of schools (e.g. Garet & DeLany, 1988). Recent studies in the growing field of school effectiveness have shown that some schools are more effective than others at promoting the progress of different groups of students and this is not dependent on school type, location or social composition (Thomas et al., 1997). Such work shows that the degree of social differentiation depends, in part, on the schools that students attend. Schools matter.

Less work has been done on curriculum organisation and educational stratification in Australia. One reason is that much of the early work on educational inequality focused on early school leaving as the broad indicator of educational activity. However, with the large changes in school completion during the 1980s, and the growing insensitivity of school completion rates as an indicator of access to schooling, has come an increasing interest in the places students occupy in the curriculum, and on how well they perform in them.

Some of this interest has focused on subject choice and course participation in the senior levels of high school. Ainley, Robinson, Harvey-Beavis, Elsworth, & Fleming (1994) pointed to ways in which social and school factors interact to shape curriculum enrolment patterns. They found social differences in course taking with students from higher socioeconomic backgrounds more often enrolling in the courses that have stronger links with university entry. Teese, McLean, and Polesel (1993) also reported large socioeconomic and gender differences in participation across different subject areas, indicating that access to opportunities in the curriculum in Australia is far from even.

Recent work has also begun to examine the effects of curriculum participation on achievement and outcomes. There is some evidence to support the view that the courses young people participate in are important in shaping educational and occupational futures. Teese et al. reported that the places offered by universities to final year students vary by the field of study, with students from higher socioeconomic backgrounds more often enrolled in the fields of study which receive the highest levels of offers. The authors also reported that final year success in the highest level mathematics, science and humanities subjects, the subjects which play an important role in accessing university places, is strongly related to social origin. Students from higher socioeconomic backgrounds achieved the highest levels of `A' grades whereas children from lower status origins had the highest failure rates.

School differences in processes of differentiation have also attracted research interest. Teese (1989) reported that students in private non-Catholic schools are more likely than others to study academic courses including the physical sciences, the more established foreign languages, and the more traditional humanities. Private schools, particularly the more elite schools, tend to provide a narrower academic curriculum.

Data and methods

The information on curriculum participation used in this paper was extracted from records held by the Tasmanian Secondary School Assessment Board (TASSAB). The TASSAB database contains information on the subject enrolments and subject results for all Tasmanian Certificate of Education (Years 9, 10, 11 and 12) students. It also contains information on the levels of study for each subject, student characteristics (sex, place of residence), and school.

In this paper we focus initially on the English and mathematics courses taken by Tasmanian students in Year 10 in 1993 (N = 5491). At this year level, mathematics and English are compulsory, but both subjects are provided as separate syllabuses based on levels of difficulty and designed for students with different levels of achievement. The top level in both subject areas is intended for students planning to continue the study of the subject into the senior years of secondary school and university. It is the most academically demanding and intellectually rigorous. The other levels represent a decreasing order of difficulty with increased attention given to practical and applied skills.

To examine patterns of social differences in course enrolments, the analysis builds on the tradition of social area research in Australia used to examine socioeconomic and regional differences in education (e.g. Linke, 1988; Teese, 1991). In this work, census derived measures of occupation, income and education are used to compare populations across geographic areas. In the present study, the Census Collection District (approximately 100 to 120 households for Tasmania) was used as the geographical area. Since the home address for each student was available, it was possible to identify the collection district in which he or she lived using a digitised electronic street directory on desktop mapping software (MAPINFO). A socioeconomic index score, calculated at the level of the collection district, was then assigned to the record for each student. The socioeconomic index scores were those derived by the Australian Bureau of Statistics (ABS) and known as the Socioeconomic Indexes For Areas (SEIFA). The ABS provides five indexes. The index of education and occupation was used in the present investigation. It was preferred to the other indexes because recent comparisons using scores at the collection district level found that it correlated more highly with individual social background data (Ainley et al., 1995).

To measure social differences in curriculum participation, the collection districts were divided into deciles based on the scale of socioeconomic status. The Year 10 students were divided into the ten groups according to the social status of the census collection district in which they lived. It was then possible to examine the social distribution of curriculum participation in mathematics and English.

Previous work has shown that it is important to include prior achievement in measuring the independent effects of school, curriculum and student background, otherwise the derived estimates are likely to be biased and over-state the effects of curriculum organisation and school (e.g. Willms, 1987). To control for the effects of prior achievement in the present study, a combined measure of achievement in Year 9 English and mathematics was derived. It was based on the student's recorded five-point moderated grade of assessment in each subject, adjusted for level of study. The separate scores for English and mathematics were added and the resulting scores were then standardised.

To measure the relationships with school, multilevel modelling was used. The technique is a relatively recent development within regression analysis and permits researchers to model student-level outcomes within schools and then to identify and model any between-school differences that exist. Multi-level modelling procedures enable researchers to account more accurately for variance in outcome measures by partitioning within and between school differences more appropriately. The technique is now widely used by investigators to measure variability both within and between schools.

Analyses

Two sets of multilevel analyses were undertaken for the study using hierarchical linear modelling (HLM) software (Bryk & Raudenbush, 1992). The first group of analyses focused on curriculum placement in Year 10. The effects of students' prior achievement, social background and gender on level of study in the subject areas English and mathematics were examined. School effects were also studied, with estimates derived for the effects of school type (government, Catholic independent, or non-Catholic independent), school size, rural or urban location of school, and mean-SES composition of the school. Analyses were conducted separately for each subject. In addition, estimates were provided for an aggregate measure of Year 10 curriculum location. This measure identifies students according to their level of study across both mathematics and English.

The second set of analyses was based on a group of Year 12 academic outcomes. The academic outcome measures included school completion (remained to Year 12 or not), participation in a university qualifying (academic) course or not in Year 12, and academic achievement for those undertaking a university qualifying course. Enrolment in academic subjects was defined as enrolment in subjects for which students obtain a tertiary entry score (TES). Year 12 achievement is measured by tertiary entrance scores which are based on a standardised scale from 0 to 100 derived by aggregating and standardising the results obtained in five subjects.

For the analyses of final year results, the initial procedure was to employ random cross-classified multi-level models. In simple hierarchical models, units at one level are clustered or nested within units at a higher level--for example, students nested within classrooms, classrooms nested within schools. However, in a longitudinal study, students may change schools or, in the case of Tasmania, move from a junior high school to a senior college to complete Years 11 and 12. In this situation, students are not simply clustered within one unit at the school level. Some are assigned to a cross-classification of schools. A procedure for dealing with these more complex structures in multi-level modelling has been applied by Rasbash and Goldstein (1994). The procedure they specified was employed in the present study. It involved building a two-level variance components model with Year 10 school as the Level 2 unit and with a single Level 3 unit spanning the entire data set. Dummy variables, one for each Year 12 school, were created with coefficients random at Level 3 and variances constrained to be equal. The results showed that variation in final year achievement attributable to the Year 12 schools was small and not statistically significant ([[sigma].sup.2.sub.u2], = 3.9, SE = 4.9), but that attributable to the Year 10 schools was larger and significant ([[sigma].sup.2.sub.u2], = 23.3, SE = 8.8). For the purposes of parsimony, and because the changes in sizes of coefficients produced by the cross-classified rather than hierarchical models were negligible, the analyses on final year achievement focused on a simpler two level hierarchical model with Year 10 school at the second level.

Results

Descriptive patterns of curriculum participation

Figure 1 presents the English and mathematics enrolment rates of Year 10 students in Tasmanian schools according to social background and gender. The vertical axis in each figure records the percentages of students located at each level of study. The horizontal axis records the student population divided into the ten socioeconomic groups. At the low end of the scale (1) are the students who live in the collection districts of low social status. At the upper end of the scale (10) are the students living in districts of high social status. The figures for males and females are reported separately.

[FIGURE 1 OMITTED]

The results suggest that, in 1993, social differences in curriculum participation were a feature of the middle years in the secondary schools of Tasmania. Access to the top level of mathematics and English varied according to social area background. Children living in lower socioeconomic status (SES) areas were far less often located in the top levels of study than children in higher SES areas. For example, only 15 per cent of boys in the lowest SES group were taking advanced-level English in 1992 compared with 40 per cent of boys in the highest SES group. The social gradient in participation is steep and generally consistent. Apart from some small fluctuations, rates of participation rise sharply moving from the bottom to the top of the SES scale. This is true of both subjects and for both boys and girls.

The reverse social patterns are apparent when considering participation in the bottom levels of study. Participation at these levels of study is dominated by children living in low SES areas. For example, about 40 per cent of boys from the lowest SES group were located in the least difficult level of mathematics study compared with fewer than 10 per cent of boys from the highest SES group. The levels of enrolment in the bottom levels of study vary across subject area, but the social patterns are consistent.

A further and striking feature of the patterns in Figure 1 is the level of gender differences in levels of study. For English these are substantial. At every point as we ascend the SES scale, females are more often enrolled in the top level. The gender gap at the highest SES level is 22 percentage points and does not fall below 11 points at any other SES level. Girls are also outperforming boys in mathematics, at least in terms of level of study. The average gap between females and males across the SES groups in those accessing the top level of study is three percentage points. The gaps are reversed in terms of participation in the lowest levels of study.

Multilevel analyses of curriculum participation

It is clear from the results that there are social and gender differences in access to levels of study in mathematics and English in Year 10. But do these differences vary across schools? The results from the analyses are presented in Table 1.

The results show that SES and gender exert significant independent effects on the levels of mathematics and English studied in Year 10. Other things being equal, students living in higher SES areas have a significantly higher likelihood of attempting the most difficult levels of study compared with their lower SES peers. They also have a significantly lower likelihood of being relegated to the least difficult levels of study. With reference to gender, other factors remaining equal, being male significantly increases the likelihood of study at the lowest levels in both subjects. These results also apply to the aggregate measure of Year 10 study.

After controlling for the student-level factors, there are significant school differences. Analyses using fully unconditional models predicting level of study show that 13 per cent of the variation in English is due to differences between schools (intraclass correlation, [rho]=0.13) and 12 per cent for mathematics ([rho]=0.12). Some of the factors contributing to this variation are included in the full models presented in Table 1.

School type has an important independent effect on level of study. Other things equal, students attending Catholic schools are more likely to be enrolled in the higher levels of study in mathematics but not in English. The effects for non-Catholic private schools are significant for both subjects. The results suggest that the effect of attending a private non-Catholic school is to improve students' chances of accessing the higher levels of study in mathematics and English. These findings support other research showing independent effects of private schools on educational activity (e.g. Williams & Carpenter, 1990).

Effects of curriculum placement on senior school outcomes

In this section, results are presented from analyses looking at the relationships between levels of study in Year 10 and a range of educational outcomes. Three outcomes are examined: Year 12 completion; enrolment in academic (university-qualifying) subjects in Year 12 or not; and final year achievement measured by university-entry scores.

School completion Table 2 displays the results for school completion. The second column presents results without inclusion of the Year 10 curriculum measure. The results show that, at the student level, sex and social background are significant predictors of school completion. All else equal, females and students from higher SES backgrounds more often remain to Year 12. Prior achievement is also a major influence. Higher achievers in Year 9 were significantly more likely to remain to Year 12, controlling for other background factors. At the school level, there are significant between-school differences in rates of completion. Schools which promote higher than expected rates of completion are those with a higher mean SES composition and private schools.

The third column in Table 2 presents the effects for the model, this time including the aggregate level of Year 10 study in mathematics and English. It shows that curriculum location or level of study in Year 10 is a strong predictor of the likelihood of completing school or dropping out. The coefficient for level of study is positive and significant ([beta]=0.53), indicating that the higher the level of study for mathematics and English, in aggregate, the greater the likelihood of completing school. This is after controlling for prior achievement, social background and gender. Social background is also important, with SES exerting a significant independent effect on survival in school. It suggests that irrespective of level of study in the middle years of school, children from higher SES backgrounds are more likely to complete school. Or, put in another way, children at the same level of performance in Year 10 are more likely to complete school if they are from higher SES origins.

The [beta] coefficient for sex is also statistically significant, though its effect is weaker (significant at the 0.05 levels rather than at the 0.01 level). However, this does not mean that sex does not continue to exert a strong influence. The results for sex need to be considered in conjunction with the effects of curriculum placement in Year 10. Some sense of this can be obtained by looking at the results for the model excluding level of middle school study (column 2). It can be seen that there are significant differences with girls more likely to complete school than boys. When the effects for curriculum placement (level of study) are introduced (column 3), the effect of sex weakens. This is because girls are more likely than boys to be located in the higher levels of English and mathematics, and the levels of study are, themselves, strong predictors of school completion. Level of study in Year 10 relays some of the effects of sex.

In the school-level model, which includes variables on mean social composition, rural or urban location, enrolment size, and sector, the coefficients for school mean SES and private schooling were significant at the 0.01 level. This suggests that students in schools with a high SES composition have a greater likelihood of completing school, even after controlling for differences in selected individual characteristics of students, including individual SES. The results also indicate that, after controlling for other factors, students in Catholic schools have a greater chance of completing school. The effect for private schooling would appear to be modest with Catholic schools exerting an effect significant at the 0.05 level, but the estimate for private non-Catholic schools no longer significant. This is contrary to the strong results on school completion for private schools recorded in other research (e.g. Williams & Carpenter, 1990). The reason is that the effects of private schooling are mediated through the influence of Year 10 level of study.

Academic course in final year Some students enrol in subjects in the final year which qualify them for higher education (academic course) whereas other students enter courses leading to technical and further education or early entry to the labour market (vocational and non-academic courses). Table 3 presents analyses of the chances of enrolling in university-qualifying or academic subjects in Year 12. The results show that Year 10 level of study in English and mathematics strongly influences type of study in the final year (Table 3, column 3). Students who engage in higher levels of study in Year 10 are far more likely to undertake an academic course in the final year. The positive and large coefficient for Year 10 study ([beta] = 0.97, p < 0.01, column 3) suggests that the stratification of learning opportunities in the middle school years substantially shapes the patterns of participation in academic programs in the final year.

Middle-school curriculum location, itself shaped by social origin and prior achievement, mediates the relationship between social background, earlier achievement and final year study. The coefficient for SES in the model with Year 10 level of study ([beta] = 0.04, p > 0.05, column 3) is lower than in the model without Year 10 level of study ([beta] = 0.06, p < 0.01, column 2). Similarly, the coefficient for prior achievement ([beta] = 1.83, p < 0.01, column 3) drops substantially after inclusion of level of study in Year 10, though remains significant ([beta] = 0.87, p < 0.01, column 3). When level of Year 10 study is included in the model, SES no longer exerts a significant independent effect. Prior achievement continues to exert a significant independent effect though it is much weaker. The results suggest that the advantage for children in middle-class areas in accessing academic final-year courses is achieved through higher levels of achievement in Year 10. Level of study in Year 10 shapes the course-taking trajectories of young people as they enter the senior years.

There are school-level influences on final year study. School mean SES is positive and significant indicating that schools with a higher SES composition have more students in the academic track ([beta] = 0.17, p < 0.05, column 3), after controlling for level of middle-school study, prior achievement, social background, and school-level factors. Private schooling also contributes independent effects. Other things being equal, students in Catholic and in non-Catholic private schools more often take university-qualifying subjects than do students in government schools. This supports arguments about the academic focus of private schools (Lamb, 1994) and reflects a smaller number of non-academic course offerings.

Final year achievement Table 4 presents the results for the HLM analyses of final-year aggregate academic achievement--tertiary entrance scores (TES). At the student level, SES has a significant effect on final year results. The SES coefficient ([beta] = 1.34, p < 0.01) suggests that students from the highest SES background had a score advantage of more than ten points on the TES scale over those from the bottom SES group. The advantage remains when level of Year 10 study is included. This indicates that, irrespective of level of Year 10 study, SES continues to contribute a significant and independent influence on final year achievement.

Gender and prior achievement also exert independent significant effects. Girls achieve more highly than boys, all else equal. This is maintained after controlling for level of study in Year 10 indicating that, for males and females at the same level of study and with similar levels of prior achievement, it is females who achieve higher tertiary entrance. Similarly the results suggest that students with higher levels of prior achievement gained higher tertiary entrance scores in Year 12 irrespective of the level of Year 10 study in mathematics and English.

After controlling for prior achievement, gender and SES, the level of study in Year 10 in English and mathematics exerts a strong influence on final year results. The coefficient for level of Year 10 study is ([beta] = 4.82, p < 0.01) and suggests that, for students with similar levels of prior achievement, of the same sex and from the same SES background, being placed in a higher rather than lower level of study in Year 10 leads to a significant increase in final year achievement. The results provide support for the view that placement in the top levels of study in English and mathematics adds substantially to final year achievement, over and above what might be expected given earlier school achievement and student background.

Schools make a difference in this process. Although most of the variation in final year achievement is at the student level, a substantial proportion is between schools. From the HLM estimates in a fully unconditional model, the proportion of variance in student achievement attributable to school-level differences was estimated to be 17.7 per cent ([rho] = 0.177). The intercepts in each of the random effects models were significant, indicating a significant difference between schools in final year achievement, after controlling for the student background factors of prior achievement, sex and SES. There were also differences between schools in the effects associated with Year 10 study. That is, in some schools, level of study in Year 10 was more closely related to final year achievement than was the case in other schools.

The factors that contribute most to the between-school variations in final year achievement are mean SES and non-Catholic private schooling. School mean SES appears to raise final year scores. That is, other things equal, the higher the mean SES of the school, the higher the achievement scores of students. This is also true of private non-Catholic schools. There is an achievement advantage in attending a private non-Catholic school. Thus final year achievement does appear to be stratified between schools.

Conclusion

The results of this research indicate that the organisation of the curriculum in the middle years of secondary school functions as a mechanism of social and gender segregation. Children become separated through assignment to levels of study in key subject areas. Students from higher SES backgrounds are more often located in the top levels of English and mathematics than are students from lower SES origins. They are also more likely to avoid relegation to the bottom classes. The association between social background and curriculum location is strong and consistent. There is also a strong association with gender, though it varies by subject. Girls are outperforming boys by a substantial margin in English, if level of study is a guide. They are more often located in the top level by an average margin of about 13 percentage points. Conversely boys are more likely to be in the bottom levels particularly in English, but also in mathematics. This is a change to the male dominance in mathematics that has been a feature of achievement patterns in the past.

The stratification of learning opportunities through the curriculum mediates the relationship between social background and academic outcomes. Children from lower SES backgrounds, and more often boys than girls, are relegated to the bottom levels of study in the middle years of secondary school. Children in the bottom levels of study are the least likely to complete school, the least likely to enrol in academic courses in the senior years, and the least likely to qualify for university entry. In this context, curricular structure is a foundation for the differentiation of opportunities within schools. The structure of the curriculum serves to transform cultural differences related to social class and gender into academic differences. At the very least, it serves to separate trajectories and experiences in school along social class and gender lines. In effect, structure fortifies culture. It articulates differences based on initial family socialisation and early school experiences into differences in academic trajectories, providing the basis for the social differentiation of opportunity within schools and variation in student achievement.

The organisation and policies of schools are important in this process. Influencing the process of curriculum differentiation is the mechanism of residential segregation. Populations across the different suburbs and regions of Tasmania form concentrations differentiated by levels of income, types of occupations, educational qualifications, quality of housing, lifestyles, and wealth. This process of residential segregation, producing unequal distributions of cultural capital and material wealth, means that the families most likely to enjoy school success (as measured by level of study and academic achievement) tend to be concentrated in particular residential areas. Schools serving these high SES areas achieve higher than expected rates of participation in the advanced levels of study. Students in these schools gain an advantage in terms of accessing the most academically profitable levels of study in the middle years of school, and in terms of final year results. This school effect could be due to the ability of these schools to exploit their large academic core of students with access to substantial, educationally supportive family resources. Neighbourhood resources have a role here. The concentration of families with high levels of cultural capital, parental education, and income in particular geographical areas contributes to a community pooling of resources, and avoidance of socially and culturally mixed environments. The aggregate levels of neighbourhood resources, and the establishment of family and friendship networks in this context, may well account for the neighbourhood or local community effects on achievement reported elsewhere (e.g. Entwistle, Alexander, & Steffel-Olsen, 1994).

Attached to the issue of residential segregation is the role of private schools. There is an uneven distribution of private schools across the different suburbs and regions of Tasmania. The results of the present study suggest that these schools, particularly non-Catholic private schools, provide an advantage in terms of level of middle-school study in English and mathematics, and final year achievement. The findings also suggest that the social distribution of achievement is not more equal in Catholic schools or in non-Catholic private schools than in government schools, for the strength of association does not vary significantly between schools.

Private schools produce gains in achievement and final year scores. Although there are many factors which may account for this--intensive pastoral care and discipline programs, establishment of academic climates through rigorous student management strategies, and a more narrow and focused curriculum (Teese, 1989) --the effect is to reinforce patterns of social stratification. Independent schools cater largely for children from higher SES backgrounds. Their strong gains in achievement allow families to build on the income and educational capital advantages they already possess. In a country with a large private school population, this helps reinforce patterns of privilege and inequality.

Keywords

achievement

curriculum policy

school effectiveness

sex differences

social stratification

socioeconomic influences

Table 1 HLM estimates of level of study in Year 10 English
and mathematics

                                                         Year 10
                                English      Maths      aggregate
Student-level variables
  Socioeconomic status           0.01 **     0.01 *       0.01 *
  Males                         -0.10 **    -0.20 **     -0.09 **
  Year 9 achievement             0.44 **     0.63 **      1.07 **

School-level variables
    Intercept                    2.27 **     3.20 **      5.46 **
  Student composition
    Mean SES                     0.01        0.01         0.03 *
  School type
    Catholic                     0.06        0.10 *       0.11 *
    Non-Catholic private         0.13 **     0.09 *       0.13 *
  School size
    Enrolments                   0.00        0.00         0.00
  Rural or urban location
    Rural school                 0.03        0.03         0.07

* Significant at the .05 level

** Significant at the .01 level
Table 2 HLM estimates of effects of Year 10 curriculum location on
school completion

                               Model without      Model with
                               level of study    level of study

Student-level variables
    SES                            0.05 **           0.04 **
    Males                         -0.23 **          -0.17 *
    Year 9 achievement             1.04 **           0.47 **
    Year 10 level of study                           0.53 **
School-level variables
    Intercept                      0.11 *            0.12 *
  Student composition
    Mean SES                       0.19 **           0.18 **
  School type
    Catholic                       0.65 **           0.48 *
    Non-Catholic private           0.78 **           0.37
  School size
    Enrolments                     0.00              0.00
  Rural or urban location
    Rural school                  -0.02             -0.02

* Significant at the .05 level

** Significant at the .01 level
Table 3 HLM estimates of effects of Year 10 level of study on
enrolment in an academic course in Year 12

                               Model without        Model with
                               level of study     level of study

Student-level variables
    SES                            0.06 **             0.04
    Males                         -0.14 **            -0.24
    Year 9 achievement             1.83 **             0.87 **
    Year 10 level                                      0.97 **
School-level variables
    Intercept                     -0.05 *             -0.08 *
  Student composition
    Mean SES                       0.15 *              0.17 *
  School type
    Catholic                       0.99 **             0.64 **
    Non-Catholic private           0.97 **             0.89 **
  School size
    Enrolments                     0.00                0.00
  Rural or urban location
    Rural school                   0.20                0.20

* Significant at the .05 level

** Significant at the .01 level
Table 4 HLM estimates of effects of Year 10 level of study on final
year achievement

                               Model without        Model with
                               level of study     level of study

Student-level variables
    SES                            1.22 **             1.02 *
    Males                         -3.22 **            -2.96 **
    Year 9 achievement            19.80 **            15.13 **
    Year 10 level of study                             4.82 **
School-level variables
    Intercept                     53.74 **            53.16 **
  Student composition
    Mean SES                       1.34 **             1.31 **
  School type
    Catholic                       2.54                2.24
    Non-Catholic private           4.61 **             4.06 **
  School size
    Enrolments                    -0.02               -0.02
  Rural or urban location
    Rural school                  -1.20               -1.17

* Significant at the .05 level

** Significant at the .01 level

References

Abraham, J. (1995). Divide and school: Gender and class dynamics in comprehensive education. London: Falmer Press.

Ainley, J., Graetz, B., Long, M., & Batten, M. (1995). Socioeconomic status and school education. Melbourne: ACER.

Ainley, J., Robinson, L., Harvey-Beavis, A., Elsworth, G., & Fleming, M. (1994). Subject choice in Years 11 and 12. Canberra: AGPS.

Bryk, A.S. & Raudenbush, S.W. (1992). Hierarchical linear models: Applications and data analysis methods. Newbury Park, CA: Sage.

Coleman, J.S., Hoffer, T., & Kilgore, S.B. (1982) High school achievement: Public, Catholic and private schools compared. New York: Basic Books.

Entwisle, D.R., Alexander, K.L., & Steffel-Olson, L. (1994). The gender gap in math: Possible origins in neighbourhood effects. American Sociological Review, 59(6), 822-838.

Garet, M.S & DeLany, B. (1988). Students, courses and stratification. Sociology of Education, 61, 61-77.

Lamb, S.P. (1994) Private schools and student attitudes: An Australian perspective. Journal of Research and Development in Education, 28(1), 43-54.

Lamb, S.P. (1997). Gender differences in mathematics participation: An Australian perspective. Educational Studies, 23, 105-125.

Linke, R.D. (1988). Regional analysis of socioeconomic trends in educational participation. Canberra: Australian National University.

Oakes, J. (1985). Keeping track: How schools structure inequality. New Haven: Yale University Press.

Rasbash, J. & Goldstein, H. (1994). Efficient analysis of mixed hierarchical and cross-classified random structures using a miltilevel model. Journal of Educational and Behavioural Statistics, 19(4), 337-350.

Teese, R. V. (1989). Australian private schools, specialisation and curriculum conservation. British Journal of Educational Studies, 37(3), 235-252.

Teese, R.V. (1991). A social geography of educational outcomes in Melbourne. In D. Coulby & C. Jones (Ed.), World yearbook of education London: Kogan Page.

Teese, R., McLean, G., & Polesel, J. (1993). Equity outcomes: A report to the Schools Council's taskforce on a broadbanded equity program for schools. Canberra: AGPS.

Thomas, S., Sammons, P., Mortimore, P., & Smees, R. (1997). Differential secondary school effectiveness: Comparing the performance of different pupil groups. British Educational Research Journal, 23(4), 451-468.

Williams, T. & Carpenter, P.G. (1990). Private schooling and public achievement. Australian Journal of Education, 34, 1-24.

Willms, J.D. (1987). Patterns of academic achievement in public and private schools. In E. Haertel, T. James, & H. Levin (Eds.), Comparing public and private schools, Volume 2: School achievement (pp. 113-141). New York: Falmer Press

Stephen Lamb is a Senior Research Fellow in the Department of Education Policy and Management at the University of Melbourne, Parkville, Victoria, 3052. Email: [e-mail]

David Hogan is Professor in the School of Education at the University of Tasmania, GPO Box 252C, Hobart, Tasmania 7001. Email: [e-mail]

Trevor Johnson is a Visiting Research Fellow with the Flinders University Institute of International Education, Bedford Park, South Australia 5042.