Issues In Educational Research, Vol 11, 2001   [Contents Vol 11] [IIER Home]

The poor stay poor and the rich stay rich

Bob Peck
Curriculum Council of WA


In 1996, Year 12 students in Western Australia responded to a survey initiated by the University of Melbourne (Teese, 1996) whose aim was to study the educational outcomes of disadvantaged groups. The author's role was to supply data on educational achievement and to tailor the survey form (which will be referred to as the Student Survey) for the Western Australian context. Involvement in the design of the Student Survey provided the opportunity to add some questions on ethnic background and to obtain data on SES indicators. The result was a study on ethnic differences in several educational indicators (Peck, 2000).

While investigating ethnicity a number of other variables had to be statistically controlled. Chief amongst these were gender, socioeconomic status (SES) and rurality. It was noticed that the effects of these variables, in particular SES, was often greater than that of ethnicity. This paper reports a re-working of the original data in which SES, rather than ethnicity, is made the focus of investigation.

In this investigation, rurality and ethnicity were crudely controlled by selecting a subset of the original data. This subset consisted of Anglo-Australians in Year 12 who lived in the Perth metropolitan area. It had been noticed earlier (Peck, 2000, p. 205) that SES variables functioned differently in rural areas compared with the city, and also amongst different ethnic groups. The elimination of ethnicity as a variable limited the scope of the conclusions but allowed much more straightforward statistical models to be used for investigating SES.

A number of investigations (for example: Banks, 1976; Gordon, 1976; Featherman and Carter, 1976; Kerckhoff and Campbell, 1977; Hoeben, 1989; Mortimore, 1988; Brandsma and Knuver, 1989; Blakey and Heath, 1992) found SES to be one of the strongest determinants of educational outcomes. In many studies, the only factor with a greater effect on educational outcomes than SES is prior educational achievement.

Attainments in both the social and economic domain contribute to SES. SES is determined by individual achievements, the most important of which are: educational attainment; employment and occupational status; membership of groups identified as disadvantaged; and income and wealth. In the case of children, SES is conventionally taken to be that of the family.

SES has been estimated by using the postcode of students or schools (Gentilli, 1996), or census statistical collection areas, for example the Ross index (Ross, Farish and Plunkett, 1988) which is currently accepted by most Australian educational systems for identifying the need for school funding to offset social disadvantage. According to Graetz, (1995, p. 42) these "aggregate level methods are efficient and unobtrusive, but concerns have been raised about ( [their] ( accuracy ( in comparison with the individual level attributes they purport to represent".

The most frequently used indicator in the United States is subsidised lunch status (for example, Levine, 1992). In Australia, this measure is not available and it is more usual to form a composite measure of an individual child's SES from information on their parents' education and occupation, and sometimes, additionally, on information about family possessions (for example, Hammond, 1995, p.123).

Data collected by the Student Survey include, for individual students, the highest educational level reached by each of the student's parents and the status (on the ASCO scale) of both parents' occupations. Additional data from the 1996 Australian census (ABS, 1998, private communication) gave household income. The usual approach to deriving a composite SES measure is to use principal components analysis to obtain the optimally weighted combination of these variables (and, if available, other measures of property, wealth, et cetera) for explaining variation in the dependent variable. In the present study, uncombined variables were analysed using multivariate techniques and the SES variables selected for the final models were those that gave what was judged to be the optimal combination of effect size and simplicity.


The subjects of this study were Year 12 students in Western Australia in 1996. The Student Survey was administered to all Year 12 students in a quasi-random selection of schools, across all three major school sectors and in both the country and metropolitan areas. There were 4024 responses to this survey, representing about a fifth of all Year 12 students.

Since it was known that ethnicity and rurality have a significant effect on several educational outcomes, a subgroup (N = 1424) of Anglo-Australian students in the metropolitan area was selected. For this investigation, students were deemed Anglo-Australian if they were born in Australia, state that they were not Indigenous Australians, and had four grandparents whose main language at home was/is English. Additional data were obtained on the whole State population of this age cohort from the 1996 Australian Census of Population and Housing (ABS, 1996) in order to investigate participation in post-compulsory schooling.

The indicators of SES that were available were the occupational category of each student's mother and father (Table 1), and also the highest educational level reached by each student's mother and father. In the case of Census data the overall household income was used as an indicator of SES.

Table 1: Categorisation of parents' occupational level

Student Survey
Number of parents by
occupational category
ProfessionsProfessional /
TechnicianWhite collar /
Business service
Clerk / Sales
Skilled self-employedSkilled trade317942
Skilled salary
Semi-skilled and unskilledUnskilled4115115
Not in paid workNot employedMissing185381

Detailed information from the survey on parental occupations was collapsed (see Table 1) to variables with five levels. These are comparable with the census categories which are based on the ASCO skill levels (ABS, 1997). The categories Not in paid work, Unemployed and Unstated were recoded "Missing" because exploratory work indicated that these categories represent SES anomalously, especially in the case of mothers' occupations.

The highest educational levels reached by each parent were collapsed to a three-point scale, representing the highest educational level completed by students' fathers and mothers respectively. In Table 2, which shows the categories of the new variables, Apprenticeship has been combined with Secondary because this is the form that was taken a generation ago by vocational education, but is now undertaken in post-compulsory schooling. Students who did some secondary education without finishing it are counted in the Primary category since this is the highest level they completed.

Table 2: Categorisation of parents' educational level

Student Survey
Number of parents by
occupational category
Some primaryPrimary1366544
Completed primary
Some secondary
Completed secondarySecondary2342265
Other tertiary
Don't knowMissing dataMissing186160

In the present study, the uncombined variables were initially used for analysis. This allowed the effects of parents' education and occupation to be investigated separately, and the contribution of mothers and fathers to be compared. However, for several of the dependent variables a composite parents' education variable, defined as the number of university educated parents (with values 0, 1 or 2) was used. In calculating this variable students who supplied data on less than two parents were omitted. In many cases this variable was found to explain approximately the same amount of variance as the separate variables for each parent, and since it led to more parsimonious models it was preferred in accordance with the principle of Occam's razor.

Model selection entailed a trial and error process of discovering which combination of significant effects maximised the amount of explained variance in the dependent variable. Gender was initially included since it was not assumed that the gender ratio was the same across all levels of SES. In addition, initial models included interactions of gender with the SES variables.

Since the aim of this study was to investigate the effect of SES in isolation from gender, multivariate methods were used for analysis. When the outcome was dichotomous (as in did/did not enter university) the method chosen was logistic regression (Kleinbaum, 1994). In other cases, where the outcome variable was continuous (as for Tertiary Entrance Score) the General Linear Model procedure in SPSS version 10 (SPSS, 1999) was used to carry out an analysis of variance (Norusis, 1997). The dependent variables, the educational indicators, will be described in more detail in the next section.


Educational participation

Data were obtained from the 1996 Australian Census of Population and Housing (ABS, 1996) on every person aged 16 who was in Western Australia on the night of 6 August 1996. These people comprised the age cohort of students in Year 11 or Year 12 of school. An Anglo-Australian subsample was constructed by eliminating all those in a household where the "head of the family" (Person 1 on the census form) spoke a language other than English at home. It should be noted that the sample and the SES measure for participation are different from those in the rest of this study. Weekly household income was obtained in order to categorise the households of the 16 year-olds by socio-economic status. Three ranges of income were chosen in order to give Low, Medium and High catergories of SES containing approximately equal numbers of persons. Table 3 shows how this variable was categorised. The percentage of post-compulsory students in each category is also shown in this table.

Table 3: Weekly household income and post-compulsory school participation

Weekly Household IncomeSES CategoryNo. of 16-yo Persons% Still at School
$1200 or moreHigh468275.8
$700 to $1199Medium431871.5
Negative to $699Low379966.1
Not availableMissing3345-
Note: The large number in the Not available category arose from households in which one or more persons failed to answer this question.

The data from the census also stated whether each 16-year old person in the State was still at school. Odds ratios for post-compulsory school participation were estimated using logistic regression, with household income and gender as independent variables (Table 4). In this study, odds ratios for SES are presented relative to the category of highest SES.

Table 4: Logistic regression for school participation

Variables CategoryProbOdds Ratio
Household Income (SES)High-1.000
Gender by SESFemale*High-1.000

Both of the main effects were statistically significant. The odds of a person staying at school for post-compulsory education diminish as SES decreases, and males are significantly less likely to finish school than females. A person of low SES has about half the odds of finishing school as one in the group of high SES.

This model also included the interaction between gender and household income. There is a significant interaction between gender and SES at the lowest level of household income, in addition to the main effects (Table 5).

Table 5: Post-compulsory school participation by gender and SES

Gender No of personsPercent of each gender still at school
SES = HighSES = MediumSES = Low
Males 633671.2%66.5%63.4%
Females 609880.6%76.8%68.6%

Educational aspirations

Students were asked about their plans for next year. These ranged from TAFE and university, through other forms of further education, to full-time work with no further training. Over 64 percent of respondents said they expected to go to university, so the numbers selecting other responses were comparatively low. This variable for educational aspirations was recoded to two levels, university and non-university. Because students were asked about aspirations/expectations that would be realised within a year, no distinction was made between aspirations and expectations.

Logistic regression was used to determine the model which best accounted for this outcome. Starting with all the SES variables and gender, and their interactions, the final model contained only gender and an SES variable whose values were equal to the number of university educated parents. This model accounted for 15.7 percent of the variance. The odds ratios for gender and SES are shown in Table 6.

Table 6: Odds ratios for factors associated with aspirations to go to university

Variables CategoryProbOdds ratio
No. of university-
educated parents
Note: Nagelkerke R2 = 0.157

This model shows that males have significantly lower odds of aspiring to go to university than females. Students with no university educated parents had odds of aspiring to university about one seventh of those with two university educated parents.

Occupational aspirations

Students' responses to the question Here is a list of job areas. Which areas would you like to work in when you have finished your studies? were classified according to the ASCO scale. Four levels were obtained (as in Table 1) but since 46 percent of students selected the highest level of aspiration the lowest three levels were combined and this outcome variable was dichotomised into High and Low.

Logistic regression was used to determine the model which best accounted for this outcome (Table 7). Of all the SES variables and gender, and their interactions, the only significant effects were the educational levels of the student's parents. This model accounted for 12.4 percent of the variance.

Table 7: Odds ratios for factors associated with aspirations for high occupational skill

Variables CategoryProbOdds ratio
Father's EducationTertiary-1.000
Mother's EducationTertiary-1.000
Note: Nagelkerke R2 = 0.124

It may be seen that students whose fathers have completed only primary or secondary education can expect to have odds of high occupational aspirations about half that of students whose fathers are university educated. A similar effect is seen in the case of mothers' occupations.

Aspirations for occupational prestige

A second categorisation of occupational aspirations was undertaken on the basis of the perceived prestige of each occupation. For this study, the scale developed by Daniel (1983) was used. Daniel developed the scale with values from 1 to 6.9, with 1 as the most prestigious occupation and 6.9 as the least. Over 1100 occupations are listed alphabetically (Daniel, 1983, pp. 196-206), and classification of the occupations in the data from the Survey consisted of looking up each one to determine a scale value. In this study the values are truncated to the integer below and this had the effect of producing values in the integer range 2 to 5. Since 35.9 percent of this sample aspired to the highest category, the lowest three categories were combined and the variable was dichotomised into High and Low aspirations for prestige.

The SES factors and gender were used to select the logistic regression model which best explained aspiration for high prestige. The only significant factors were father's education and mother's occupation (Table 8).

The results show that if a student's father has not been to university their odds of aspiring to occupations associated with high prestige are less than half of those of students with university educated fathers. The effect of mothers is different in that the mother's occupation has a significant effect, not her educational level. It may be seen that students with mothers in professional/managerial occupations have higher odds of aspiring to high prestige occupations than other students.

The effect of unskilled mothers appears slightly anomalous, since their effect on their children's aspirations for prestige are not significantly different from that of mothers in professional/ managerial occupations. For the unskilled category, is it possible that mother's occupation is not a good indicator of SES.

Table 8: Odds ratios for factors associated with aspirations for high occupational prestige

Variables CategoryProbOdds ratio
Father's EducationTertiary-1.000
Mother's OccupationProfessional/Managerial-1.000
White collar.006.630
Skilled trade.047.404
Note: Nagelkerke R2 = 0.077

Academic self-concept

The academic self-concept scales developed by Peck (2000, chapter 5) were used as the dependent variables. There are two scales, related to English and Mathematics, which can be used to measure how students perceive their efficacy and affectivity towards these two learning areas. The scales were calibrated using the Rasch model, and the scale units are logits. General linear models were used to study the effect of SES and gender.

SES was not significantly associated with mathematics self-concept. All of the available measures of family SES were added to general linear models but the only significant variable was gender (the average mathematics self-concept for males was significantly higher than that for females).

English self-concept was found to be significantly associated with SES. Table 9 summarises the results of the model which explains most of the variance in English self-concept while containing only statistically significant terms.

Table 9: Model for English self-concept

Source of variance dfFProbPartial eta squaredObserved power
Corrected Model361.384<.001.1161.000
Intercept1252.009 <.001.1521.000
Gender1158.205 <.001.1011.000
No. of uni-educated parents214.968<.001.0210.999

Note: R Squared = .116 (Adjusted R Squared = .114)

Of the SES variables available, the one that best explained the variance in English self-concept was found to be the number of parents who had gone to university. No significant interaction between SES and gender was found. It can be seen from the values of the partial correlations in Table 9 that the effect of this measure of SES is about one fifth of the effect of gender. Although the effect of SES was statistically significant, this factor accounts for only 2 percent of the variance in English self-concept.

The estimated marginal means shown in Table 10 show that there is no significant difference in mean English self-concept between students with zero or one university educated parent, but those with two university educated parents have a significantly higher English self-concept.

Table 10: Estimated marginal means of English self-concept by number of parents uni-educated

No. of parents
English self-concept95% Confidence interval
MeanStd. errorLower boundUpper bound
Note: The measurement units for English self-concept are logits.

Subject choice in post-compulsory education

This investigation analysed participation in certain popular patterns of subjects, referred to as "programs of study". The programs of study (Table 11) are specific to this study and have no official standing in the certification of student results in Western Australia. The rationale for defining them is that the education and labour market consequences are likely to reflect the effects of combinations of subjects rather than individual subject enrolments (Lamb and Ball 1999, p. 4).

The variable used to indicate SES is the number of university educated parents for each student. Gender was also included in the logistic regression models, but the interaction between gender and SES was omitted as it was found in every case to be non-significant. Table 11 summarises these models.

It is clear from Table 11 that, relative to students with two university educated parents, those whose parents are not university educated have significantly lower odds of participating in programs of study such as High Maths, Maths & Science and High English, all of which consist of TEE subjects and are manifestly related to university entrance. They also have significantly lower odds of participating in the programs of study Humanities and Social Studies. In addition, they had nearly 6 times the odds of participating in a Non-TES program of study (meaning that they had ruled themselves out of contention for obtaining a Tertiary Entrance Score) than students with two tertiary educated parents.

Table 11: Postcompulsory subject choice Odds ratios for participation by SES
(relative to students with 2 university-educated parents)

Program of study No of university-
educated parents
High Maths: Applicable Maths or Calculus.25.56.104
High English: English or English Literature.16NS.141
Maths & Science: Either Physics or Chemistry, plus
either Applicable Maths or Calculus
Humanities: Three subjects from the Arts, LOTE,
English and Society and Environment
Business: Two subjects from Accounting, Economics,
Law, Work Studies, and the computing area.
Social Studies: Two subjects from Society and Environment.58NS.012
Environment: One subject from {Geography, Geology, Prac
plus one from {Biology, Human Biol and Senior Science}
Non-TES: Three subjects which do not count towards a
Tertiary Entrance Score
Note: Odds ratios are not given unless they are statistically significant.

Year 12 achievement

A measure of overall achievement in Year 12 was estimated for the students in this sample using the Rasch model to analyse their grades. In this analysis, subjects were conceptualised as items and the grades (A, B, C, D and E) within a subject were regarded as a five-point rating scale. Grades are not equally spaced in terms of achievement, and the distances between corresponding pairs of grades in different subjects are not the same. For this reason the unrestricted Rasch Model for analysing polytomous items (Andrich, 1978) was an appropriate analytical model. The hierarchical nature of the grades ensures that they are ordered categories.

A file of grades achieved by each Year 12 student in 1996 was obtained from the Secondary Education Authority. Students who completed two or fewer subjects were deleted, since many of them would not have been typical Year 12 students. Subjects completed by fewer than 100 students were deleted, and the remaining dataset was analysed using the RUMM2010 software (Andrich, Sheridan, and Luo, 1999).

The Rasch Model analysis, using 65 subjects and data from 16,201 students, indicated that the data lay on a highly unidimensional scale, since the person separation index, which is analogous to a traditional Cronbach alpha reliability index, was 0.859. Amongst the output of this analysis was a table of estimates of student locations on the achievement scale. Achievement estimates produced by the Rasch Model analysis take into account the difficulties of the subjects chosen by each student.

This measure of overall achievement differs from the Tertiary Entrance Score in that the grades were derived from non-TES subjects as well as those used for entry to university. Thus, overall estimates of achievement were obtainable for all students whether they were university-bound or not.

The number of university-educated parents was used as an independent variable SES indicator. Model selection involved this variable and gender. The General Linear Model procedure in SPSS version 10 (SPSS, 1999) was used to carry out this analysis of variance. The results are summarised in Table 12.

Table 12 shows that both gender and SES are statistically significant factors in this model of academic achievement. SES accounts for about 11 percent of the variance in this variable. There was no significant interaction between SES and gender.

It can be seen that the mean achievement measures of students increase with the number of university educated parents (Table 13). The means of each category are significantly different from each other.

Table 12: Model for year 12 achievement

Source of variance dfFProbPartial eta squaredObserved power
Corrected Model366.141<.001.1241.000
Intercept12842.213 <.001.6701.000
Gender130.482 <.001.0211.000
No. of uni-educated parents285.408<.001.1091.000

Note: R Squared = .124 (Adjusted R Squared = .122)

Table 13: Estimated marginal means of academic achievement by no. of parents uni-educated

No. of parents
Academic achievement95% Confidence interval
MeanStd. errorLower boundUpper bound
Note: The measurement units for academic achievement are logits.

Tertiary entrance score

For admission to university, the crucial measure of achievement is a ranking (TER) based on the Tertiary Entrance Score. The Tertiary Entrance Score is an aggregate measure of achievement based on scaled marks in four or five subjects. The Tertiary Entrance Score has a maximum possible value of 510.

A general linear model analysis of factors thought to be related to Tertiary Entrance Score revealed that while SES was a significant factor, gender had no significant effect (Table 14). In this model SES was represented by two variables, the highest educational level of the student's mother and father. The father's educational level explains about 5 percent of the variance in Tertiary Entrance Score, while the mother's explains about 1 percent.

Table 14: Model for tertiary entrance score

Source of variance dfFProbPartial eta squaredObserved power
Corrected Model815.950<.001.1261.000
Intercept117109.608 <.001.9511.000
Father's Educational Level226.190 <.001.0561.000
Mother's Educational Level25.455 .003.013.869
Father's Ed * Mother's Ed46.847 <.001.029.992

Note: Power computed using p = .05. R Squared = .126 (Adjusted R Squared = .118)

The significant interaction in this model is further illustrated in Table 15 which shows the estimated marginal means for each level of father's and mother's educational level. The main effect of father's education is manifested in every column of the table, which shows mean Tertiary Entrance Score increasing as the highest educational level increases. This effect is not apparent in the changes across rows of Table 15. In fact, mother's education only has a substantial effect when the father is university educated.

Table 15: Effect of parents' education on TES

Father's education Mother's education
Primary completedSecondary/ ApprenticeTertiary
Primary completed298.0264.8307.1
Secondary/Apprentice290. 4297.4286.4

Admission to university

Data obtained from the four public universities in Perth showed which students were successful in obtaining a university place in the year following the survey. A logistic regression analysis showed that there was no significant interaction between gender and SES (the number of university educated parents), but both of these factors had a significant main effect (Table 16).

As Table 16 shows, the odds of getting in to university for students with no university educated parents are 0.18 of those for students with two university educated parents.

Table 16: Odds ratios for factors associated with admission to university

Variables CategoryProbOdds ratio
No. of university-
educated parents
Note: Nagelkerke R2 = 0.137

Occupational attainment

Although the occupational destinations of students were not available to this study, respondents to the Student Survey provided data on their parents' occupations and educational level.

In order to study the effect of education on occupational attainment, a logistic regression analysis was carried out to relate each parent's education to their occupation. The outcome variable, parent's occupation, was dichotomised to two levels, Professional/managerial and Other.

Table 17: Odds ratios for males attaining a professional/managerial occupation

Variable CategoryProbOdds ratio
Father's EducationTertiary-1.000
Note: Nagelkerke R2 = 0.541

Table 18: Odds ratios for females attaining a professional/managerial occupation

Variable CategoryProbOdds ratio
Mother's EducationTertiary-1.000
Note: Nagelkerke R2 = 0.490

It was found that about half of the variance in occupational level was explained by educational level. For both males (Table 17) and females (Table 18) the odds ratios of attaining a professional/managerial occupation are extremely low for individuals who do not complete tertiary education relative to those of people who do so.

Discussion and conclusions

A simplistic description of an idealistic role for education would be to allow individual students to develop to their own potential. By this means, a capable student from a disadvantaged group in society would be able, through education, to become socially mobile. As a corollary, incompetent people from socially privileged backgrounds would not rise automatically to positions requiring high levels of expertise. Thus, by maximising the skill available to the workforce the cost of education would return a benefit to society in general.

However, despite the plausibility of this line of reasoning, there is little evidence to suggest that this is what occurs. On the contrary, many writers have discussed the role of the education system in perpetuating social inequality (Fitzgerald, 1976; Meade, 1978; Branson and Miller, 1979). For example, the Fitzgerald Report concluded, after reviewing the evidence on disadvantage in Australia

The structural inequalities in our society are nowhere more evident than in our school systems. Far from being a way out for poor people, schools act as a sorting, streaming mechanism helping to maintain the existing distribution of status and power. (Fitzgerald, 1976, p. 227)
Society's resistance to the educational measures intended to increase accessibility is addressed by a number of theories of social and cultural reproduction. Chief amongst these are the ideas of Bourdieu and Passeron (1977), Bowles and Gintis (1976), Birrell, T. (1978, p. 107), Taylor (1982) and Bernstein (1971).

For the most part, the effects of privilege on educational outcomes are only observed in their crudest forms-a good word put in, the right contacts, a well resourced home, extra tuition, information on the educational system and job outlets. However, this is only because the most extensive and substantial effects of privilege and power are invisible. The dominant culture is in a position to select, structure, evaluate and deliver 'appropriate' knowledge through its education system.

According to Bourdieu and Passeron (1977), schools do not transmit culture but they play an essential part in reproducing the dominant culture. This thesis was elaborated by Taylor (1982) who argued that schools are reality-defining institutions where there is a dialectical relationship between identity and social structure, mediated through school culture.

As institutions, schools have the power to structure the objective reality of those people who are part of them. Thus the reality internalized by students tends to be a particular definition of reality imposed on them through the structuring of school experience... This process of imposition of reality is hidden beneath an ostensibly neutral system which favours those with power in society. (Taylor, 1982, p. 152)
The power to impose or define reality is hegemonic. Because it is not coercive power, it appears to be neutral and is invisible to most people, even those who are disadvantaged by the existing social structure. Hegemonic power is involved in selecting what knowledge is of value and defines the agenda and the limits of any debate by presenting certain concepts and relationships as normal (Watkins, 1980, p. 458). Thus, a student who makes a "free" choice (to enrol for a particular course, for example) may genuinely believe it to be uninfluenced by societal factors. The neutrality of the curriculum can not be assumed, even when it is well intentioned.

In this study, the measured educational outcomes and indicators were the culmination of twelve years of schooling in Western Australia. The widespread and significant effect of variables representing SES provides evidence that education is part of a positive feedback cycle for the reproduction of inequality. This does not imply that education is necessary for social class to reproduce itself, simply that it is currently involved in the reproduction of class.

It was found that low SES, as compared with high SES, was associated with

Parents with high educational attainments tend to find occupations with high social status and their children, who have high aspirations and a high English self-concept, stay in the education system longer and tend to opt for the subjects from which a Tertiary Entrance Score can be formulated. They tend to achieve at a higher level at school and their higher Tertiary Entrance Scores give them access to university courses that lead to high status occupations. (The causative relations, if any, between these indicators were not part of this study.)

The effect of SES on Tertiary Entrance Scores was such that students from high SES backgrounds could expect to gain access to many university courses which are not accessible to students from less privileged backgrounds. University admission is based on a student's Tertiary Entrance Rank (TER) which is a percentile rank based on the 17 year-old age cohort in the general population.

Students with two university educated parents had an average score of 342 (TER approximately 90), while those with no university educated parents had an average of 298 (TER approximately 79). Many university courses have entry cut-offs at TERs between 79 and 90 (TISC, 2001, pp. 25-34) and hence would tend to discriminate between these two groups of students. Since many university courses are aligned with occupations, this score at the end of schooling is a major determinant of future occupational attainment.

Current plans to implement the Curriculum Framework (Curriculum Council, 1998; 2000) and the Post-compulsory Review constitute what is arguably the most radical reform of Western Australian schooling in history. Social inequity has been identified as one of the targets to be addressed by this change. In view of the extensive theorisation of society's resistance to change, this is a very ambitious undertaking. The history of educational measures to redress social inequity is a record of numerous failures; if our new policies are to succeed, they will have to do so against this trend. During and after the Curriculum Framework's implementation, inclusivity and access will be monitored with interest. This work provides a baseline against which to measure future indicators of access.


Andrich, D. (1978). A rating formulation for ordered response categories. Psychometrika, 43(4), pp. 561-573.

Andrich, D., Sheridan, B. & Luo, G. (1999). RUMM2010. 17 Dondonaea Place, Duncraig, Western Australia, Perth: Rumm Laboratory.

Australian Bureau of Statistics (1996). 1996 Census of population and housing. Canberra ACT: Commonwealth of Australia.

Australian Bureau of Statistics. (1997). Australian Standard Classification of Occupations (2nd Edition). Canberra, ACT: Commonwealth of Australia.

Banks, O. (1976). The sociology of education. New York: Schocken.

Bernstein, B. (1971). On the classification and framing of educational knowledge. In M.F.D. Young (Ed), Knowledge and Control. New Directions for the Sociology of Education. London: Collier-Macmillan.

Birrell, T. (1978). Migration and the dilemmas of multiculturalism. In R. Birrell and C. Hay (Eds), The Immigration Issue in Australia. Bundoora: Department of Sociology, La Trobe University.

Blakey, Louise S. & Heath, Anthony F. (1992). Differences between comprehensive schools: Some preliminary findings. In P. Cuttance and D. Reynolds (Eds), School effectiveness: Research, policy and practice, pp. 96-121. London: Cassell. ISBN 0-304-32295-4.

Bourdieu, P. (1974). The School as a Conservative Force. In J. Eggleston (Ed), Contemparary Research in the Sociology of Education. London: Methuen.

Bourdieu, P. & Passeron, J. C. (1977). Reproduction in Education, Society and Culture. London: Sage.

Bowles, S. & Gintis, H. (1976). Schooling and Capitalist America: Educational Reform and Contradictions of Economic Life. London: Routledge and Kegan Paul.

Brandsma, H. P. & Knuver, J. W. M. (1989). Organisational differences between Dutch primary schools and their effect on pupil achievement. In B. Creemers, T. Peters and D. Reynolds (Eds), School effectiveness and school improvement: Proceedings of the first international congress London 1988. Groningen: RION; Cardiff: University of Wales.

Branson, J. & Miller, D. B. (1979). Class, Sex and Education in Capitalist Society. Malvern, Victoria: Sorrett Publishing.

Curriculum Council (1998). The Curriculum Framework. Perth, Western Australia: Curriculum Council.

Curriculum Council (2000). Post-compulsory Review Position Paper. Perth, Western Australia: Curriculum Council.

Featherman, D. L. & Carter, T. M. (1976). Discontinuities in schooling and the socioeconomic life cycle. In W. H. Sewell, R. M. Hauser and D. L. Featherman (Eds), Schooling and Achievement in American Society. New York: Academic Press.

Fitzgerald, R. T. (Commissioner) (1976). Poverty and Education in Australia: Commission of Inquiry into Poverty (Fifth main report). Canberra, ACT: AGPS.

Gordon, E.W. (1976). Group differences versus individual development ineducational design. In S. Messick, (Ed), Individuality in Learning. San Francisco: Jossey Bass.

Graetz, Brian (1995). Socioeconomic Status and School Education. Canberra, ACT: AGPS. ISBN 0 644 35991 9.

Hammond, John T. (1995). Teacher beliefs and educational attainment: A three-level model of the relation of student, classroom and school variables to public examination performance. PhD Thesis, Murdoch University.

Hoeben, W. Th. J. G. (1989). Educational innovation or school effectiveness: a dilemma? In B. Creemers, T. Peters and D. Reynolds (Eds), School effectiveness and school improvement: Proceedings of the second international congress, Rotterdam 1989. Amsterdam/Lisse: Swets and Zeitlinger.

Kerckhoff, A. C. & Campbell, R. T. (1977). Black-white differences in the educational attainment process. Sociology of Education, 50, pp. 15-27.

Kleinbaum, David G. (1994). Logistic Regression. New York: Springer-Verlag. ISBN 0-387-94142-8.

Lamb, Stephen & Ball, Katrina (1999). Curriculum and Careers: the education and labour market consequences of Year 12 subject choice. Longitudinal Surveys of Australian Youth, Report No. 12. Camberwell, Victoria: ACER.

Levine, Daniel U. (1992). An interpretive review of US research and practice dealing with unusually effective schools. In P. Cuttance and D. Reynolds (Eds), School effectiveness: Research, policy and practice, pp. 96-121, London: Cassell. ISBN 0-304-32295-4.

Meade, Phil (1978). Youth, Education and Life Chances: An Investigation Into the Processes of Education and Social Selection of Youth in Society. Sydney, NSW: School of Education, University of NSW.

Mortimore, P. (1988). School matters: The junior years. Wells: Open Books.

Norusis, M. J. (1997). SPSS Professional Statistics 7.5. Chicago, Ill: SPSS Inc.

Peck, Bob (2000). Ethnicity: A study of participation, aspirations and achievement in education. Doctor of Education Thesis, Murdoch University.

Ross, K., Farish, S. and Plunkett, M. (1988). Indicators of socioeconomic disadvantage for Australian schools. Geelong, Victoria: Deakin Institute for Studies in Education.

SPSS (1999). SPSS Version 10.0 (Software). Chicago: SPSS Inc.

Taylor, Sandra. (1982). Schooling and Social Reproduction. The Australian Journal of Education, 26(2), 144-154.

Teese, R. (1996). The Educational Outcomes Questionnaire: Survey of Western Australian Year 12 Students. University of Melbourne.

Tertiary Institutions Service Centre (2000). Tertiary Institutions Service Centre 2000 Guide. Perth, WA: TISC.

Watkins, Peter E. (1980). Work Experience, Correspondence and Hegemony. Paper presented at the AARE Annual Conference, November 1980, Sydney. ISSN 0159-2130.

Author: Bob Peck is Senior Consultant Measurement and Research at the Curriculum Council of Western Australia and Adjunct Research Fellow at Edith Cowan University, Perth. His research interests are applications of the Rasch model and educational assessment in the paradigm of an outcomes-focussed standards framework. Email:

Please cite as: Peck, B. (2001). The poor stay poor and the rich stay rich. Issues In Educational Research, 11(2), 45-64.

[Contents Vol 11] [IIER Home]
© 2001 Issues In Educational Research. This URL:
HTML: Clare McBeath [] and Roger Atkinson []
Created 16 May 2002. Last revision: 4 Sep 2013.